article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
the classification of small bodies in the inner solar system as either asteroids or comets has historically been attempted by different scientists using different techniques and employing different criteria .
observational astronomers classify small bodies having transient , unbound atmospheres ( usually made visible by the scattering of sunlight from entrained micron - sized dust particles ) as `` comets '' .
bodies having instead a constant geometric cross - section are called `` asteroids '' . to planetary scientists ,
comets and asteroids are distinguished by their ice content or perhaps by their formation location .
comets are icy ( because they formed beyond the `` snow - line '' ) while asteroids are not ( supposedly because they formed at higher mean temperatures inside it ) .
lastly , to dynamicists , comets and asteroids are broadly distinguished by a dynamical parameter , most usually the tisserand parameter measured with respect to jupiter ( kresak 1982 , kosai 1992 ) .
it is defined by @xmath1^{1/2}\cos(i ) \label{tisserand}\ ] ] where @xmath2 , @xmath3 and @xmath4 are the semimajor axis , eccentricity and inclination of the orbit while @xmath5 = 5.2 au is the semimajor axis of the orbit of jupiter .
this parameter , which is conserved in the circular , restricted 3-body problem , provides a measure of the close - approach speed to jupiter .
jupiter itself has @xmath6 = 3 .
main belt asteroids have @xmath7 and @xmath8 3 while dynamical comets have @xmath9 . the three systems of classification ( observational , compositional and dynamical ) are independent but imperfect .
for example , whether a coma or tail is detected on a given object depends strongly on the parameters of the observing system used .
a puny telescope may not reveal a coma that is easily rendered visible by a more powerful one .
there is neither an agreed quantitative ice fraction with which to divide comets from asteroids nor , more importantly , any reliable way to measure the ice fraction in a small body .
the formation locations and dynamical histories of small bodies are rendered uncertain by hard - to - model non - gravitational forces from both electromagnetic radiation ( the yarkovsky effect ) and mass loss itself ( the rocket effect of whipple 1950 ) and also by the chaotic nature of solar system dynamics .
the tisserand criterion is an imperfect classifier because it is based on an idealized representation of the solar system ( e.g. jupiter s orbit is not a circle , the gravity of other planets is not entirely negligible , and so on ) .
therefore , the utility of equation ( [ tisserand ] ) as a dynamical discriminant is limited for objects with @xmath6 very close to 3 . as an example , the quasi - hilda comets at @xmath10 4.0 au have @xmath11 2.9 - 3.04 but are clearly interacting with jupiter through the 3:2 mean - motion resonance ( toth 2006 ) .
some recognized jupiter family comets ( e.g. 2p / encke with @xmath6 = 3.03 ) also fall in this category ( fernandez et al .
2002 , levison et al .
2006 ) . given these and other imperfections
it is remarkable that , for a majority of objects , the observational , compositional and dynamical definitions of asteroids and comets lie in close agreement . for the most part , objects classified as asteroids ( comets ) based on their orbits , have the physical properties expected of asteroids ( comets ) as far as can be observed .
exceptions have given rise to a somewhat confusing and evolving system of nomenclature , used to describe small solar system bodies by a combination of their orbital properties and physical appearances .
to clarify this we show , in figure ( [ classification ] ) , a two - parameter classification based on morphology , on the one hand , and the tisserand parameter on the other .
traditional comets lose mass and have @xmath12 3 .
those with 2 @xmath13 3 are called jupiter family comets and are thought to originate in the kuiper belt .
comets with @xmath12 2 are long - period or halley family comets , with a source in the oort cloud .
inactive counterparts to the jupiter family comets are called , variously , extinct , dead or dormant comets ( hartmann et al .
they are presumably former comets in which past heating by the sun has removed all near surface ice , although buried ice might remain and these objects could , in principle , reactivate .
inactive counterparts to the long - period and halley family comets are called damocloids ( jewitt 2005 ) .
again , these are likely objects in which near - surface ice has been lost . in this paper , we focus attention on the sub - set of the small - bodies which are dynamically asteroid - like ( @xmath14 3 ) but which lose mass , like comets .
these were called `` main - belt comets '' by hsieh and jewitt ( 2006 ) but here we use the term `` active asteroids '' , since some of the examples to be considered , while dynamically asteroid - like and showing comet - like properties , are not in the main - belt .
numerical integrations show that these are not recently captured comets from the kuiper belt ( fernandez et al .
2002 , levison et al .
scientific interest in these objects lies in the possibility that primordial water ice could have survived in asteroids despite early heating from embedded radioactive nuclei ( grimm and mcsween 1989 ) and heating by the sun .
even greater interest is added by the possibility that the outer asteroid belt may have supplied part of the volatile inventory of the earth ( morbidelli et al .
2000 ) . additionally , active asteroids are a source of dust for the zodiacal cloud , while unseen counterpart bodies may supply dust to the debris disks of other stars ( e.g. shannon and wu 2011 ) .
after briefly summarizing the current observational evidence concerning these active asteroids , we discuss the surprisingly varied mechanisms through which a body is capable of losing mass .
a recent and complementary discussion focused on observational properties has been offered by bertini ( 2011 ) .
to - date , eleven active ( or mass - shedding ) asteroids have been reported .
the nine spatially resolved examples are shown for comparison in figure ( [ image_compo ] ) , while their positions in the semimajor axis vs. eccentricity plane are plotted in figure ( [ mbc_ae_plot ] ) .
the comet - like morphologies and their orbital separation from the domain of the comets are obvious in these two figures .
all eleven objects are located inside the 2:1 mean - motion resonance with jupiter at 3.3 au and all but three lie below the mars - crossing line ( marked @xmath15 in the figure ) .
the orbital properties of these objects are summarized in table ( [ orbital ] ) , the physical properties in table ( [ physical ] ) and we review them briefly here , in order of decreasing @xmath6 . * ( 3200 ) phaethon : @xmath6 = 4.508 * is dynamically associated with the geminid meteor stream ( e.g. williams and wu 1993 ) and with several other small asteroids including 2005 ud ( ohtsuka et al .
2006 , jewitt and hsieh 2006 , konoshita et al .
2007 ) and 1999 yc ( ohtsuka et al .
2008 , kasuga and jewitt 2008 ) .
all these objects may be related to a precursor body that disintegrated @xmath010@xmath16 yrs ago ( ohtsuka et al .
evidence for modern - day mass loss comes from the observed brightening of phaethon by a factor of two within a few hours of perihelion ( @xmath17 = 0.14 au ) in 2009 ( jewitt and li 2010 ) .
this rapid brightening can not be due to scattering from the @xmath05 km diameter nucleus alone and , instead , indicates transient mass loss with the ejection of @xmath010@xmath18 kg of particles each @xmath19 mm in radius .
with @xmath19 = 1 mm , this is only 10@xmath20 of the geminid stream mass , but raises the possibility that the decay of phaethon is a continuing process .
the phenomenon has not been observed to repeat , but comparable observations when near perihelion are difficult to secure because of the small angular separation from the sun .
* p/2010 a2 : t@xmath21 = 3.582 * the object showed a distinctive morphology in which a leading , point - like nucleus about 120 m in diameter ( table [ physical ] ) , is followed by an extended tail ( or trail ) of dust in which are embedded ribbon - like structures ( jewitt et al .
the position angle of the tail and its variation with time are consistent with the action of radiation pressure on mm to cm sized dust particles , following impulsive ejection at very low speeds ( @xmath00.2 m s@xmath22 ) in 2009 february - march ( jewitt et al .
2010 , snodgrass et al .
( note that an opposite conclusion was reached by moreno et al .
2010 but based on limited data ) .
p/2010 a2 went unnoticed for its first @xmath09 months largely because of its angular proximity to the sun ( jewitt et al .
2011a ) . during this time
, a large quantity of fast - moving particles are presumed to have left the vicinity of the main nucleus .
the mass of particles remaining in the tail at discovery is estimated to be in the range ( 6 to 60)@xmath2310@xmath24 kg ( jewitt et al . 2010 , moreno et al .
2010 , snodgrass et al .
2010 ) . *
( 2201 ) oljato : @xmath6 = 3.299
* magnetometers on the pioneer venus spacecraft revealed multiple , symmetric disturbances in the solar wind magnetic field , clumped non - randomly in time ( russell et al .
1984 ) . about 25% of these events are associated with planet - crossing asteroid ( 2201 ) oljato , whose orbit lies interior to venus when near perihelion .
russell et al .
suggested that the magnetic disturbances result from decelleration of the solar wind , perhaps caused by mass loading from ionized gases released by an unknown process from debris distributed along oljato s orbit
. a mass loading rate of only @xmath05 kg s@xmath22 is reportedly needed . however , a spectroscopic search for gas produced by oljato itself proved negative ( chamberlin et al . 1996 ) , with upper limits to the cn production rate near 10@xmath25 s@xmath22 . with a standard
h@xmath26o / cn mixing ratio of 360 , the corresponding limit to the mass production rate in water is @xmath271.5 kg s@xmath22 .
whatever the cause of the repetitive magnetic disturbances , they are not products of an inert asteroid and imply mass loss from oljato .
a dynamical simulation indicates that ( 2201 ) oljato has negligible chance of being a captured jupiter family comet ( bottke et al .
2002 ) .
* p/2008 r1 : t@xmath21 = 3.216 * this object was observed over a @xmath045 day interval in 2008 , when near @xmath17 = 2 au , as having the appearance of an active comet with a typically flared tail ( jewitt et al .
the object intrinsically faded by a factor of about 2 over the above interval of observations . an upper limit to the nucleus radius of 0.7 km
was set ( red geometric albedo 0.05 assumed , see table [ physical ] ) , later reduced to @xmath28 0.2 km by subsequent observations ( h. hsieh , private communication , 2011 ) .
spectral observations limited the production of the cn radical to q@xmath29 1.4@xmath2310@xmath25 s@xmath22 , again corresponding to a water production rate @xmath271.5 kg s@xmath22 assuming h@xmath26o / cn = 360 .
p/2008 r1 is located near the 8:3 mean - motion resonance with jupiter and is also affected by the @xmath30 secular resonance .
the dynamical lifetime in this orbit is short ( 20 to 30 myr ) compared to the age of the solar system , suggesting that p/2008 r1 was scattered into its present location from elsewhere in the asteroid belt . *
( 596 ) scheila : t@xmath21 = 3.208 * ( 596 ) scheila , a 113 km diameter object with red geometric albedo @xmath00.04 ( table [ physical ] ) , developed a coma in late 2010 .
over the course of a month , this coma expanded with a characteristic speed @xmath060 m s@xmath22 and faded in response to the action of solar radiation pressure ( bodewits et al .
2011 , jewitt et al .
2011 , moreno et al .
2011 ) , apparently without any continued replenishment of particles from the nucleus .
the gas production from the nucleus was reportedly limited to q@xmath3110@xmath32 s@xmath22 ( moreno et al .
2011 ) to q@xmath3110@xmath33 s@xmath22 ( howell and lovell 2011 ) , corresponding to water production rates @xmath270.3 kg s@xmath22 to 3 kg s@xmath22 ( the meaning of these limits is unclear given the non - steady nature of the mass loss event from scheila ) .
the mass of dust in micron - sized grains was 4@xmath2310@xmath24 kg ( jewitt et al .
2011 ) while more model - dependent attempts to account for mass in larger particles gave 6@xmath2310@xmath34 kg ( bodewits et al .
2011 ) to 2@xmath2310@xmath35 kg ( moreno et al .
no ice was observed in the coma ( yang and hsieh 2011 ) . *
300163 ( 2006 vw139 ) : t@xmath21 = 3.203 * discovered in 2006 and first observed to be active on ut 2011 aug .
30 ( hsieh et al .
2011d ) , little is yet known about this object . with reported absolute magnitude @xmath36 = 16.6 and an assumed albedo @xmath37 = 0.04 , appropriate to its outer belt location ,
the estimated diameter is @xmath03 km .
two thin tails , one near the projected orbit and another roughly antisolar , show that dust leaves the nucleus very slowly and point to mass loss over a protracted period .
there is no reported evidence for gas or for repetitive mass loss , but new observations of this object are still be acquired at the time of writing . * 133p / elst - pizarro : t@xmath21 = 3.184 * the appearance is typically that of a point - like nucleus with a thin tail ( or `` trail '' ) of 10 @xmath38 m sized dust following in the projected orbit .
order of magnitude mass loss rates in gas , inferred from spectroscopy ( licandro et al .
2011 ) and in dust , inferred from surface photometry of the tail ( hsieh et al .
2004 ) reach @xmath00.02 kg s@xmath22 .
the thin tail indicates that particles are ejected from the nucleus at speeds @xmath01.5 m s@xmath22 , comparable to the escape speed from the [email protected] km diameter nucleus . * 176p : t@xmath21 = 3.167 * a cometary appearance was detected for 176p / linear over a single , month - long interval in 2005 ( hsieh et al .
during this time , the object showed a fan - shaped tail and was about 30% brighter than the bare nucleus , leading to an implied dust mass @xmath010@xmath40 kg .
the properties of the dust can be approximately matched by models in which the characteristic particle size is 10 @xmath38 m , the ejection speed @xmath05 m s@xmath22 and the dust production rate @xmath0 0.07 kg s@xmath22 , all similar to values inferred in 133p .
the [email protected] km diameter nucleus rotates with a period near 22.2 hr ( table [ physical ] ) .
* 238p / read (= p/2005 u1 ) : t@xmath21 = 3.152 * like p/2008 r1 , the nucleus of 238p / read is tiny , with a diameter @xmath00.8 km ( hsieh et al .
it was observed to be in an active state in both 2005 and 2010 , but not in between , with a coma dust mass of order 10@xmath40 kg and a production rate estimated ( from published photometry ) near @xmath00.1 kg s@xmath22 . also like p/2008 r1 ,
238p / read is dynamically short - lived , with a survival time of order 20 myr ( haghighipour 2009 ) .
* p/2010 r2 ( la sagra ) : t@xmath21 = 3.098 * the object was observed to be active from 2010 september to 2011 january , at @xmath17 = 2.6 to 2.7 au .
moreno et al .
( 2011b ) inferred dust production at the peak rate of @xmath04 kg s@xmath22 , with centimeter - sized particles ejected at about 0.1 to 0.2 m s@xmath22 . a limit to the outgassing rate @xmath413@xmath2310@xmath25 s@xmath22 ( corresponding to 3 kg s@xmath22 in water )
was placed spectroscopically ( hsieh et al .
hsieh et al .
( 2011c ) show ( neglecting possible non - gravitational forces due to outgassing ) that the orbit of p/2010 r2 is stable on timescales @xmath0100 myr and argue that this object was likely formed in - situ . *
107p/(1949 w1 ) wilson - harrington : t@xmath21 = 3.083 * fernandez et al . ( 1997 ) analyzed two photographic plates taken on 1949 november 19 , when at @xmath17 = 1.148 au .
the object is trailed in both but , in the blue plate , shows a prominent diffuse tail about 2@xmath42 in length .
the red plate shows only a hint of this tail . the color ( b - r
= -1 ) is inconsistent with scattering from dust , but suggests instead resonance fluorescent scattering from an ion tail .
the position angle of the tail , being about 15@xmath43 from radial to the sun , is also more consistent with the expected direction of a plasma tail than with a dust tail .
curiously , 107p was re - observed on november 22 and 25 but then showed no trace of a tail ( cunningham 1950 ) , and no comet - like activity has been reported since ( lowry and weissman 2003 , ishiguro et al .
2011 ) . based on a statistical dynamical model , bottke et al .
( 2002 ) concluded that there is a 4% chance that 107p is a captured jupiter family comet .
we consider a variety of processes capable of launching dust from a small body . in each case , the number of unknown but relevant physical parameters prevents any exact treatment , but it remains instructive to consider the range of action of the mechanisms in the context of the asteroids . since whipple ( 1950 ) , sublimation has been explored in great detail as the driver of mass loss from the classical comets . it need not be re - described in detail here .
although simple in concept , detailed studies of comets show that sublimation is a remarkably complex process when factors such as the porosity of the surface , nucleus rotation , the conduction of heat into the interior and the development of a refractory mantle are considered ( guilbert - lepoutre et al .
one simplification possible for the present objects is the assumption that asteroids contain no amorphous ice , since temperatures in the asteroid belt are too high for it to escape crystallization .
accordingly , we address only the highly idealized case of sublimation from an exposed crystalline ice surface in thermal equilibrium with sunlight .
the sublimation mass flux per unit area , @xmath44 ( kg m@xmath45 s@xmath22 ) from a patch of surface whose normal is inclined to the sun - direction by angle @xmath46 is determined by solution of the energy balance equation @xmath47 here @xmath48 = 1360 w m@xmath45 is the solar constant , @xmath49 is the heliocentric distance in au , @xmath50 is the bond albedo , @xmath51 0.9 is the emissivity of the surface , @xmath52 = 5.67@xmath2310@xmath53 w m@xmath45 k@xmath20 is the stefan - boltzmann constant , @xmath54 is the equilibrium temperature and @xmath55 is the latent heat of sublimation .
the temperature dependence of @xmath44 may be estimated from thermodynamics and the clausius - clapeyron equation , or from direct measurement ( washburn 1926 ) .
the conduction term , @xmath56 , requires knowledge of the thermophysical parameters and the spin vector of the nucleus and is small .
we neglect it here .
the angle term , @xmath57 , depends on the nucleus rotation vector , changes with the position on the nucleus and varies with time as the nucleus rotates .
therefore , the total mass loss rate must be obtained by integrating equation ( [ subl ] ) over the surface and with respect to time . to avoid sinking in complexity
, we here focus on two limiting cases for the surface temperature and resulting sublimation rate .
the maximum temperature , @xmath58 , is reached at the subsolar point on a non - rotating body , where @xmath59 = 1 .
the minimum effective temperature , @xmath60 , occurs when the surface is isothermal , corresponding to the largest possible radiating area . on a sphere ,
the average value of the angle term is @xmath61 = 1/4 .
accordingly , we estimate the minimum and maximum temperatures and specific sublimation rates from equation ( [ subl ] ) for @xmath61 = 1 and 1/4 .
we considered both low ( @xmath50 = 0.05 ) and high ( @xmath50 = 0.50 ) ice albedos and calculated sublimation mass fluxes from equation ( [ subl ] ) .
clean ( high albedo ) ice sublimates too slowly to drive activity at asteroid belt distances and so we present only the dirty ( low albedo ) ice solutions for the sublimation rate as a function of distance , @xmath17 , in figure ( [ sublimation ] ) .
the curves are convergent towards small @xmath17 because the sublimation term in equation ( [ subl ] ) is then dominant , but separate dramatically at larger @xmath17 as the sublimation term drops exponentially , leaving all the energy for radiation . across the 2.0 @xmath62 3.3 au strip corresponding to the asteroid belt ,
sublimation rates from @xmath010@xmath20 kg m@xmath45 s@xmath22 ( sublimation at @xmath58 ) to @xmath6310@xmath53 kg m@xmath45 s@xmath22 ( sublimation at @xmath60 ) are found .
mass loss rates near 1 kg s@xmath22 ( table [ physical ] ) require sublimating areas from 10@xmath64 m@xmath65 ( @xmath58 ) to 10@xmath34 m@xmath65 ( @xmath60 ) .
since the latter exceeds the entire surface area of ( for example ) 133p , we can exclude the isothermal ( @xmath60 ) model as the activity driver . the main conclusion to be drawn from figure ( [ sublimation ] )
is that exposed water ice can sublimate across the full range of heliocentric distances occupied by the main belt provided it is dirty .
two other quantities are directly calculable once @xmath44 is known .
first , the rate of recession of the sublimating surface is given by @xmath66 and indicated on the right - hand axis of figure ( [ sublimation ] ) .
with ice density @xmath67 = 1000 kg m@xmath68 , we find 350 @xmath38 m yr@xmath22 @xmath69 3 m yr@xmath22 over the 2 to 4 au distance range swept by the majority of active asteroids ( see right hand axis of the figure ) .
second , the size of the largest ejectable particle is estimated by balancing the gas drag force with the gravity , assuming spherical particles and nucleus and neglecting all other forces .
we take the densities of the ejected grains and of the bulk nucleus , in the absence of contrary evidence , to be equal .
the resulting maximum particle radius , computed from @xmath70 is plotted in figure ( [ critical ] ) as a function of heliocentric distance for low albedo , subsolar sublimation . in equation ( [ drag ] ) , @xmath71 is the dimensionless drag coefficient , here taken to be unity , and @xmath72 is the gas speed at the nucleus surface , taken from biver et al .
the equation offers , at best , a crude estimate of the critical radius ; the highly anisotropic outflow of gas from a real nucleus can strongly affect dust acceleration and dynamics relative to the simple isotropic case used here ( e.g. crifo et al .
still , it is obvious from figure ( [ critical ] ) that sublimation gas drag can easily remove small solid particles across the full range of distances swept by the main - belt asteroids .
grains larger than @xmath73 remain bound to the asteroid and contribute to the development of a refractory mantle , eventually leading to a stifling of the sublimation . while exposed ice sublimates quickly , ice protected from direct illumination by a covering or mantle of refractory debris can survive nearly indefinitely .
a meter - thick layer of regolith can suppress sublimation sufficiently to retain ice for gyr timescales ( schorghofer 2008 ) .
then , to initiate sublimation , this protective mantle must be removed , which could occur naturally through impact .
collisions between asteroids occur at characteristic speeds of several km s@xmath22 ( bottke et al .
1994 ) and are therefore highly erosive .
impact yields , defined as the ratio of the ejecta mass , @xmath74 , to the projectile mass , @xmath75 , are @xmath761 .
material ejected by impacts may explain observed activity in some asteroids . in this section
we aim to obtain a relation between the impactor properties and the resulting brightening caused by ejected material . in an impact , the bulk of the ejecta travel at the lowest speeds . for equal target and projectile densities ,
the mass of ejecta , @xmath74 , traveling faster than a given speed , @xmath77 , can be roughly expressed by a power law @xmath78 in which @xmath75 and @xmath79 are the impactor mass and speed and @xmath80 0.01 is a constant ( housen and holsapple 2011 ) .
the index @xmath81 depends slightly on the properties of the target but is reasonably well approximated for a range of materials as @xmath81 = -1.5 . only ejecta traveling with @xmath82
can escape from an asteroid to produce an increase in the scattering cross - section , while the rest must fall back to coat the surface around the impact site . with the escape velocity given by @xmath83
it is clear that equation ( [ housen ] ) defines , for a given target density @xmath67 and impact speed @xmath79 , a relation between the impact yield and the radius , @xmath84 , of the impacted asteroid .
a relation between @xmath74 and the scattering cross - section of the ejecta and hence the change in brightness caused by impact , can be established .
the size distribution of the ejecta from nasa s deep impact mission to comet 9p / tempel 1 has been modeled as a power law ( kadono et al .
their figure 4 gives a differential power law size index @xmath85 = 3.7 over the 1 @xmath86 100 @xmath38 m size range , while other investigators have found slightly steeper distributions ( lisse et al .
2006 ) . separately ,
the size distribution of the ejecta from p/2010 a2 has been modeled as a power law with @xmath85 = [email protected] for @xmath87 mm ( jewitt et al .
a wider range of power laws ( from @xmath85 = 3 to as steep as @xmath85 = 6 ) has been reported in small - scale , hypervelocity impact experiments ( takasawa et al . 2011 ) .
for the sake of the present discussion , we adopt @xmath85 = 3.5 and find that the relation between the scattering cross - section , @xmath88 , and the mass of particles having radii in the range @xmath89 is @xmath90 where @xmath91 .
for example , with @xmath92 = 0.1 @xmath38 m , @xmath93 = 0.1 m , @xmath94 = 10@xmath20 m , or 0.1 mm .
we combine equations ( [ housen ] ) , ( [ escape ] ) and ( [ mvsc ] ) to compute the ratio of the cross - section of the ejecta to the cross - section of the target asteroid @xmath95 where we have taken the projectile to be a sphere of the same density as the target and with a radius , @xmath96 . substituting @xmath50 = 0.01 , @xmath94 = 0.1 mm , @xmath79 = 5 km s@xmath22 , @xmath67 = 2000 kg m@xmath68 and @xmath81 = -3/2
, we obtain @xmath97 a 1 m projectile impacting a 1 km target asteroid will produce ejecta with a cross - section 30 times the geometric cross - section of the asteroid
. lines of constant brightening , computed from @xmath98 \label{deltamag}\ ] ] are shown in figure ( [ dmag ] ) , as a function of asteroid and projectile radius .
the curves are plotted only where the ejecta mass computed by equation ( [ housen ] ) is less than the target mass .
the figure shows that , for example , the impact of a 1 m projectile into a 1 km asteroid would cause a brightening by @xmath99 3.5 magnitudes , while impact of the same projectile into a 10 km asteroid would cause a brightening of only a few @xmath230.01 magnitude , and would almost certainly escape detection .
the published parameters of ( 596 ) scheila and p/2010 a2 , interpreted as impact ejecta , are also plotted on figure ( [ dmag ] ) .
the location of scheila overlaps the @xmath100 = 1 magnitude brightening line , consistent with observations .
p/2010 a2 was discovered nearly a year after disruption .
the magnitude of its initial brightening is thus unknown , but figure ( [ dmag ] ) suggests that it was a dramatic @xmath101 15 .
the curves in figure ( [ dmag ] ) are uncertain , both because of the simplicity of the treatment and because of the many uncertainties in the impactor and dust grain parameters . for example
, if @xmath79 were larger by a factor of two , equation ( [ dm ] ) shows that the ejecta mass and cross - section would be larger by a factor of @xmath02.8 and the ejecta would be brighter by @xmath01 magnitude . nevertheless , equation ( [ dm ] ) and figure ( [ dmag ] )
provide a useful first estimate of the observable effect of a given impact in the main belt .
the figure leads naturally to the question of the impact rate in the main - belt , which we discuss in section ( [ discussion ] ) .
while some ejecta can be launched from the impact site at speeds comparable to the impact speed ( i.e. @xmath102 5 km s@xmath22 ) , most material travels much more slowly , following equation ( [ housen ] ) .
the slowest escaping material leaves the target at a speed comparable to the gravitational escape speed , thus setting a relation between the size of the target and the duration of the brightening caused by impact .
for example , consider two target asteroids of size 100 m and 100 km , observed from 1 au distance using a typical charge - coupled device camera with a field of view 5@xmath42 ( corresponding to @xmath010@xmath40 km half - width ) .
ejecta from the 100 km body will travel at @xmath0100 m s@xmath22 and cross the field of view in @xmath010@xmath103 s ( about 10 days ) .
ejecta from the 100 m body will travel at characteristic speed @xmath104 0.1 m s@xmath22 and take so long to cross the field ( @xmath010@xmath105 s ) that radiation pressure and keplerian shear will dominate the distribution of the dust .
the persistence of impact - produced coma is thus a measure of the size of the target body .
the equatorial , centripetal acceleration on a strengthless rotating object of mass density @xmath67 ( kg m@xmath68 ) equals the gravitational acceleration at the critical rotational period , @xmath106 , given by @xmath107^{1/2}. \label{rotation}\ ] ] here , @xmath108 is a dimensionless constant dependent on the shape of the object .
a sphere has @xmath108 = 1 .
a prolate body with axes @xmath109 in rotation about its shortest dimension has @xmath110 .
expressed in hours , equation ( [ rotation ] ) may be written @xmath111 \sim 3.3 k \left[\frac{1000}{\rho}\right]^{1/2}. \label{rotation2}\ ] ] for example , a sphere with @xmath67 = 2000 kg m@xmath68 has @xmath106 = 2.3 hr while an elongated body with the same density and @xmath112 = 2 has @xmath106 = 4.7 hr , by equation ( [ rotation2 ] ) .
although equation ( [ rotation2 ] ) shows that @xmath106 is independent of the object size , numerous measurements of asteroids show that only objects smaller than @xmath00.1 - 0.3 km rotate with periods @xmath632 hr ( pravec et al .
the probable reason is that sub - kilometer asteroids have significant tensile strength , and so can resist rotational disruption , while larger bodies are structurally weak and rotationally unstable when @xmath113 .
a weak ( in tensile strength ) aggregate or `` rubble pile '' structure could be produced by impact fracturing or , in the case of cometary nuclei , by the gentle settling together of icy planetesimals formed separately .
given the evidence from the distribution of asteroid rotation rates , it is reasonable to conjecture that rotational instability might be a cause of observable mass - loss in the main - belt .
a number of torques are capable , in principle , of driving an asteroid into rotational instability .
for example , collisions between asteroids lead to a random walk in angular momentum that can end in an unstable state .
anisotropic radiation of photons can lead to a net torque ( the so - called yorp effect ) that can do the same ( marzari et al .
2011 , jacobson and scheeres 2011 ) .
an object losing mass through sublimation can , depending on the rate and angular dependence of the mass loss , experience a torque orders of magnitude larger than torques due to either collisions or yorp , quickly being driven towards rotational break - up ( c.f .
drahus et al .
2011 ) . while , in principle , rotational instability can affect any asteroid regardless of its size , in practice the spin - up times vary inversely with radius and the objects most likely to experience rotational bursting are the smallest ( exactly as observed in the near - earth population ) .
could some active asteroids be rotationally disrupted objects ?
the action of electrostatic forces in moving dust particles across a planetary surface is best established on the moon .
images from the surveyor lunar lander spacecraft showed an unexpected `` horizon glow '' , caused by forward scattering from dust particles @xmath010 @xmath38 m in size located @xmath01 m above the surface ( rennilson and criswell 1974 ) .
later , the lunar ejecta and meteorites ( leam ) dust detection experiment of apollo 17 recorded the impact of low velocity ( @xmath27100 m s@xmath22 ) dust particles , and showed particularly intense fluxes of such particles near local sunrise and sunset ( berg et al . 1976 ) .
the measured dust particles fluxes are @xmath07 orders of magnitude too large to be associated with churning of the surface by micrometeorite bombardment . instead
, electrostatic levitation was suggested as the source of these particles . on
the day - side , positive charging of the lunar surface occurs by ejection of electrons owing to the photoelectric effect from uv and x - ray solar photons . in the absence of a discharging current
, the potential can rise to the energies of the most energetic photons ( 10@xmath16 v , or more , c.f .
de and criswell 1977 ) but , in practice , the measured dayside potential is about + 10 v ( colwell et al .
2007 ) . on the unilluminated side , or in shadows , the surface becomes negatively charged as a result of the greater flux of solar wind electrons ( the solar wind is electrically neutral , but the low - mass electrons travel much faster than solar wind protons , leading to a net flux of electrons onto the surface ) .
the typical energy of solar wind electrons is @xmath010 ev leading to a negative potential of order 10 v. near the terminator , the same processes lead to positive and negative charging of the surface , but on much smaller spatial scales corresponding to the scale of the local topography , leading to locally high electric field gradients . near a shadow edge , the photoelectrons produced in sunlight can travel to and stick in unilluminated regions , causing local electric field gradients estimated at @xmath114 10 to 100 v m@xmath22 ( colwell et al .
the positive , sunlit surface attracts a cloud of electrons , effectively neutralizing the gradient on length scales @xmath1151 m. the electrostatic processes that move dust particles on the moon presumably operate also on the asteroids .
the charging time on the moon is @xmath010@xmath65 to 10@xmath16 s ( de and criswell 1977 ) ; photoelectron charging currents will be 9 times weaker at 3 au and the charging times will be 9 times longer but the potentials attained , for a given dielectric constant , will remain the same .
the principal difference in the asteroid belt is that , whereas levitated lunar dust is retained by the gravity of the moon , dust ejection speeds on small bodies can exceed the escape velocity , @xmath116 .
therefore , electrostatic processes are potentially capable of leading to mass loss from asteroids .
the charge on a spherical grain of radius @xmath2 is related to the potential on the grain , @xmath117 , by @xmath85 = @xmath118 , where @xmath119 = 8.854@xmath2310@xmath120 f m@xmath22 is the permittivity of free space .
the force on a charged particle exposed to an electric field @xmath121 ( v m@xmath22 ) is just @xmath122 . as our criterion for dust ejection
, we demand @xmath123 , where @xmath77 is the grain speed achieved by accelerating across the shielding distance @xmath124 and @xmath125 is the gravitational escape speed at the surface . assuming that the grain and the asteroid are spherical , of radius @xmath2 and @xmath84 , respectively , and that both have density @xmath67
, this criterion gives the critical grain size for electrostatic ejection as @xmath126 substituting the lunar values , @xmath117 = 10 volts , @xmath124 = 1 m , @xmath121 = 10 to 100 v m@xmath22 , and using @xmath67 = 2000 kg m@xmath68 as the canonical asteroid density , equation ( [ electro ] ) gives @xmath127 = 1.5 to 5 @xmath38 m for a @xmath84 = 1 km asteroid ( c.f .
figure [ critical ] ) .
these sizes are somewhat smaller than the @xmath010 @xmath38 m particle sized particles inferred from observations of some active asteroids ( e.g. 133p ; hsieh et al .
2004 ) but , given the many uncertainties in both the model and in the interpretation of observations , perhaps the differences are acceptable .
in contrast , equation ( [ electro ] ) shows that on the moon ( @xmath84 = 1600 km ) , only nanometer - sized grains can be ejected , meaning that the process is irrelevant there . even for asteroid ( 596 ) scheila ( @xmath84 = 56 km and @xmath127 = 0.04 to 0.1 @xmath38 m )
, any electrostatically ejected particles would be smaller than a wavelength and inefficient optical scatterers .
figure ( [ critical ] ) shows that @xmath128 , for a given asteroid radius .
we conclude that electrostatic ejection of particles large enough to scatter optical photons is a plausible mass - loss mechanism only for smaller asteroids .
there are major uncertainties concerning electrostatic effects on the moon , especially regarding the tendency for small particles to stick to each other and to surfaces , through the action of van der waals and other contact forces .
these uncertainties are magnified in importance on the asteroids because , unless the forces of cohesion can be overcome , electrostatic levitation and ejection will be impossible ( hartzell and scheeres 2011 ) .
unfortunately , ignorance of grain cohesion limits our ability to know whether electrostatic ejection is important on the asteroids , although flat , pond - like structures on eros have been interpreted this way ( hughes et al .
sonnett et al .
( 2011 ) report that @xmath05% of main - belt asteroids show ultra - faint comae that can not be detected individually but which are collectively significant .
if these comae are real , electrostatic ejection of sub - micron grains might offer a plausible source mechanism that is approximately independent of distance from the sun .
thermal fracture can occur when the stress associated with a change in the temperature of a material exceeds the tensile strength of the material .
the thermal stress is @xmath129 , where @xmath81 ( k@xmath22 ) is the thermal expansivity , @xmath130 ( n m@xmath45 ) is the young s modulus and @xmath131 ( k ) is the responsible temperature change .
the expansivities of common rocks are @xmath132 ( lauriello 1974 , richter and simmons 1974 ) .
young s moduli @xmath133 n m@xmath45 are typical for rock ( pariseau 2006 ) . if converted into kinetic energy with efficiency @xmath134 , thermal strain energy can generate fractured material with speeds ( jewitt and li 2010 ) @xmath135 substituting @xmath134 = 1 gives an upper bound to the ejection speeds of fracture fragments as @xmath136 20 m s@xmath22 , for fracture at @xmath131 = 1000 k. this is sufficient to launch particles above the gravitational escape speed from asteroids up to radius @xmath137 20 km . thus
, thermal fracture is a potential source of small particles and an agent capable of ejecting these particles from asteroids , but only those approaching the sun very closely .
some carbonaceous chondrite classes ( probable fragments of outer - belt asteroids ) contain 10% - 20% water by weight , bound into hydrated minerals ( jarosewich 1990 ) .
if it could be liberated , this bound water might drive observable mass loss by entraining small grains in the gas flow produced as the water escapes into space .
laboratory experiments show that thermal dehydration of hydrous phyllosilicates of planetary relevance ( e.g. serpentine , brucite , muscovite , talc ) is characterized by activation energies in the range 325 - 400 kj mol@xmath22 ( bose and ganguly 1994 ) .
thermal dehydration is accompanied by a net volume change which can crack the dehydrating material and provide a source of small particles . however , the above activation energies correspond to temperatures ( @xmath01000 k ) , much higher than normally found in the asteroid belt . for this reason ,
thermal dehydration is unlikely to play a role in a majority of the active asteroids since they have perihelia @xmath1382 au and temperatures too low to trigger dehydration .
dehydration can be caused by shock pressure waves in impacts .
for example , shock dehydration of brucite ( mg(oh)@xmath26 ) begins near 12 gpa and is complete by 56 gpa ( 50% of the trapped water is lost at pressures 17 gpa ) ( duffy et al .
pressures this high are confined to a region comparable in size to the projectile diameter , so that the volume of shock - dehydrated target material is small compared to the volume of the resulting impact crater .
nevertheless , water released by impact could sustain mass loss from an asteroid for a timescale longer than that on which the impact ejecta dissipates .
for example , a 10 m radius projectile would dehydrate @xmath010@xmath24 kg of target . with a water fraction of 10%
, some 10@xmath103 kg of water could potentially leak from the impact site , enough to sustain a dusty flow at 0.1 kg s@xmath22 for @xmath0100 days .
while shock dehydration must be considered , at best , a process of secondary importance , it does raise the prospect that gaseous products detected following an impact might , in the future , be misinterpreted as evidence for sublimating ice .
dust particles can be swept from the surface of an asteroid through the action of radiation pressure . to see this
, we consider a grain located just above the surface and compare the gravitational acceleration towards the nucleus with the acceleration in the anti - sunward direction caused by radiation pressure , the latter expressed as @xmath139 , where @xmath140 is the gravitational acceleration to the sun .
a dust particle can be swept away by radiation pressure if @xmath141 in which we have assumed that the nucleus is at heliocentric distance @xmath17 and is spherical , of density @xmath67 , radius @xmath84 and non - rotating .
also , @xmath142 is the gravitational constant and @xmath143 is the mass of the sun . the dimensionless quantity @xmath144 depends upon the grain shape , composition and porosity but is principally a function of particle size , @xmath2 .
as a useful first approximation , we take @xmath145 , where @xmath146 is the grain radius expressed in microns ( bohren and huffman 1983 ) .
then , the condition for a grain to be lost to radiation pressure becomes @xmath147 substituting @xmath67 = 2000 kg m@xmath68 and expressing the nucleus radius in km and the heliocentric distance in au , we find the critical radius for grain loss to be @xmath148 equation ( [ amu ] ) ( also plotted in figure [ critical ] ) shows that , for example , a 1 km radius object at @xmath17 = 3 au would lose dust grains smaller than about @xmath10 1 @xmath38 m . optically active particles ( @xmath149 0.1 @xmath38 m ) can be swept away throughout the asteroid belt provided their parent bodies are smaller than about 10 km .
larger particles can be lost if the nucleus is rotating and has an aspherical shape .
the above considerations are simplistic in that they take no account of the direction of the radiation pressure acceleration relative to the nucleus .
grains released on the day - side , for instance , will be pushed back into the nucleus by radiation pressure from above . on the other hand , grains released near
the terminator will feel a net force in a direction determined by the vector sum of the local gravitational acceleration and radiation pressure acceleration .
these grains have the potential to escape . the principal limitation to efficacy of radiation pressure sweeping
is , as with electrostatic launch , set by contact forces that hold small particles to the asteroid surface .
if these forces can be overcome , then radiation pressure sweeping can play a role across the asteroid belt for 10 km sized asteroids and smaller .
the key observational property of these objects is that the activity is recurrent ( hsieh et al .
2004 , hsieh et al . 2010 ) , qualitatively consistent with sublimation in the same way that mass loss from the classical comets is modulated by the varying insolation around the orbit . other mechanisms in section ( [ mechanisms ] ) are either inconsistent with recurrent mass loss ( impact , rotational instability ) , or can be reconciled with it only in the most contrived way . with equilibrium sublimation rates
@xmath150(3 to 5)@xmath2310@xmath151 kg m@xmath45 s@xmath22 at @xmath17 = 2.6 to 3.0 au ( figure [ sublimation ] ) , the most stringent production rates @xmath1500.01 kg s@xmath22 ( table [ physical ] ) imply mass loss from areas of exposed ice @xmath150200 to 300 m@xmath152 , corresponding to circles of radius @xmath010 m ( hsieh et al .
2004 ) .
the eccentricities of 133p and 238p are small ( 0.165 and 0.253 , respectively ) and , over the range of heliocentric distances swept , the expected sublimation mass flux varies by less than an order of magnitude ( upper curve in figure [ sublimation ] ) .
this is a problem because a ten - fold reduction in activity at aphelion relative to perihelion would still produce detectable dust , whereas none is observed . instead
, mass loss might be confined to a window near perihelion because of a shadowing effect of local topography on the nucleus , perhaps caused by recession of the sublimating surface , forming a pit ( equation [ recession ] ) .
the surface recession rate at @xmath17 = 2.6 to 3.0 au is ( 3 to 5)@xmath2310@xmath53 m s@xmath22 , so that a sublimation pit meters deep should develop in only a few orbits .
however , that the ice must still be close to the physical surface is shown by the appearance of coma near and soon after perihelion .
ice more than a few centimeters deep would be thermally decoupled from the deposition of heat on the surface . the evidence for ice sublimation , although self - consistent , remains indirect .
detection of gaseous products of sublimation would provide definitive evidence for the role of sublimation in active asteroids .
just as in the classical comets , practical concerns render the daughter species ( molecular fragments ) optically more visible than the parent molecules sublimated from the ice .
the most prominent band in the optical spectra of comets is due to the trace species cn , itself a product of dissociation of the parent molecule hcn . as noted in section 2 ,
several attempts to detect cn 3889 in active asteroids have been reported , typically resulting in limits to the total gas production rate 0.1 - 1 kg s@xmath22 , about two to three orders of magnitude smaller than found in bright comets near 1 au . to set this on a quantitative basis , we show in figure ( [ cn_plot ] ) the expected cn band flux as a function of @xmath17 for several outgassing rates , as marked , computed as described in jewitt et al .
the fluxes in the figure were evaluated for opposition ( i.e. @xmath153 = @xmath17 - 1 ) and assume the haser model and a spectroscopic slit 1@xmath1545@xmath155 .
the swings effect ( in which the resonance fluorescence cross - section is affected by the heliocentric velocity through the doppler effect ) amounts to less than a factor of two , and has been neglected for simplicity .
plotted on the figure are published limits to the cn band flux in active asteroids .
state of the art measurements , indicated by the horizontal grey band in the figure , reach cn fluxes @xmath0 ( 2 to 3)@xmath2310@xmath156 erg cm@xmath45 s@xmath22 , corresponding to @xmath158 10@xmath25 s@xmath22 at @xmath17 = 2 au but , by @xmath17 = 3 au , the limits barely reach @xmath158 10@xmath159 s@xmath22 , showing how difficult is the spectroscopic detection of gas at larger distances . the steep distance dependence is made worse by the rapidly declining sublimation rates from 2 au to 3 au ( figure [ sublimation ] ) . for the most favorable ( high temperature ) case in that figure ( albedo 0.05 ) the sublimation rate drops by a factor of four from 2 au to 3 au . for these reasons
, spectroscopic detection of gas at main - belt distances is challenging even in many classical comets , where sublimation is the undoubted driver of the observed activity .
in so far as spectroscopic emission detections of gas in active asteroids are concerned , absence of evidence is not evidence of absence .
the brightening of scheila was sudden , the scattering cross - section declined smoothly with time , and the morphology evolved in a way consistent with the expansion of an impulsively ejected coma under the action of radiation pressure ( bodewits et al .
2011 , jewitt et al .
2011 , moreno et al .
2011 , hsieh et al .
2011 , ishiguro et al .
all these signatures are consistent with impact production but are difficult or impossible to reconcile with the other mechanisms of section ( [ mechanisms ] ) .
scheila is also alone amongst the active asteroids , so far , in being very large ( 113 km diameter ) .
it presents a substantial cross - section for impact , but the resulting strong gravity precludes the action of electrostatic ejection , while the slow rotation eliminates any possibility of rotational instability .
sublimation of water ice can eject small grains from scheila , but the short coma fading time is difficult to understand given the sublimation rates in figure ( [ sublimation ] ) .
the outburst brightness was @xmath01 mag brighter than the bare asteroid ( larson 2010 , jewitt et al .
2011 ) . figure ( [ dmag ] ) is consistent with the conclusion that the projectile was between 35 m and 60 m in diameter ( jewitt et al .
2011 , ishiguro et al .
bodewits et al .
2011 ) .
impact may account for the dust tail of p/2010 a2 , but the interpretation in this case is non - unique .
this body is comparatively tiny ( only 120 m in diameter vs. 113 km for scheila ) , with a tail mass suggesting collision with a projectile only a few meters in size ( jewitt et al .
2010 , snodgrass et al .
a key observational difference is that , whereas ( 596 ) scheila faded on a timescale of one month , the tail of p/2010 a2 remained prominent @xmath015 months after formation , despite the addition of no new dust ( jewitt et al .
2010 , 2011 ) .
this difference in fading timescales arises because the slowest particles to escape from a body leave the target with a speed comparable to the gravitational escape speed . for p/2010 a2
this is @xmath00.1 m s@xmath22 , allowing large , slow grains to persist for @xmath1601 yr after impact while , for scheila , particles ejected at less than the 60 or 70 m s@xmath22 escape speed simply fell back on the surface ( jewitt et al .
2011 , ishiguro et al .
2011 ) . furthermore , most grains traveling fast enough to be ejected were also small enough to be deflected by solar radiation pressure , providing another mechanism to quickly clear the coma .
there are no direct measurements of the total , early - time brightening , only a limit @xmath161 19 ( jewitt et al . 2011 ) .
however , the model suggests that if the p/2010 a2 were hit by a two meter - sized projectile , the brightening would be @xmath99 15 ( figure [ dmag ] ) , which is consistent with the empirical constraint .
extreme brightening caused by impact has been suggested as an explanation of `` guest stars '' recorded in ancient chinese records ( reach 1992 ) .
rotational instability offers another explanation for mass loss from p/2010 a2 ( jewitt et al .
the small diameter of the primary body ( 120 m ) corresponds to a short spin - up timescale under the action of yorp . in fact , for sub - kilometer main - belt objects the yorp spin - up timescale is less than the timescale for collisional disruption ( marzari et al .
2011 , jacobson and scheeres 2011 ) .
while the rotation period of the primary body in p/2010 a2 is unknown , it would not be surprising to find that it had exceeded the centripetal limit .
unfortunately , there are few published predictions for the detailed appearance of a rotationally disrupted body ( richardson et al .
2005 ) .
object ( 3200 ) phaethon is distinguished from the other active asteroids by its small perihelion , @xmath85 = 0.14 au , allowing it to reach extreme sub - solar temperatures ( @xmath01000 k ) .
these are sufficient to trigger thermal fracture and to dehydrate water - bearing minerals , if they are present ( jewitt and li 2010 ) .
it has been suggested that phaethon was recently scattered inwards from a pallas - like orbit ( @xmath2 = 2.771 au , @xmath3 = 0.281 , @xmath4 = 33@xmath43 ; de leon et al .
2010 ) where a large proportion of the asteroids are volatile - rich b- and c - types .
if phaethon originated in the vicinity ( or as part ) of pallas then finding hydrated minerals would not be surprising , since pallas itself contains hydrated minerals .
the small asteroids associated dynamically with phaethon ( 1999 yc and 2005 ud ) should have similar compositions , but have not been observed close to perihelion and have shown no evidence for mass loss .
in addition to the creation of dust and fragments by thermal fracture and dehydration shrinkage , we conjecture that the escape of particles produced by these thermal mechanisms is facilitated by radiation pressure sweeping .
this is especially important in ( 3200 ) phaethon near perihelion relative to other asteroids . according to equation ( [ amu ] ) , particles up to 300 @xmath38 m in diameter
can be swept from its surface when at @xmath85 = 0.14 au .
particles produced by thermal fracture or dehydration cracking , or lifted from the surface by electrostatic repulsion , would be rapidly swept away from it by radiation pressure ( jewitt and li 2010 ) .
the mechanisms behind the mass loss in other active asteroids remain less clear .
activity in 176p can be interpreted in terms of a sublimation model that is able to fit the observed coma morphology ( hsieh et al .
models involving impulsive ejection ( e.g. by impact ) reportedly provide less good fits to the imaging data but , given the many unknown parameters in the models ( e.g. the relation between dust particle size , scattering efficiency and speed ) the identification of sublimation in 176p can not be regarded as definitive .
the same must be said for observations of p/2008 r1 , where sublimation can fit the data ( jewitt et al .
2009 ) but where an impact origin may also apply .
we also regard the nature of activity in p/2008 r1 as unknown .
the coma isophotes of p / la sagra are consistent with dust emission over an extended period ( moreno et al .
2011 ) , as is integrated light photometry ( after correction for the unmeasured phase function of the comet , c.f .
hsieh et al . 2011 ) . again ,
isophote models involve the assumption of numerous , unknown parameters and provide solutions which are non - unique .
this is especially true for objects viewed at small angles to their orbital planes , as are most of the active asteroids .
clear evidence of the weakness of models is the fact that moreno et al .
( 2010 ) , using limited data , were able to fit the isophotes of p/2010 a2 with a model in which particles were released over an eight month period , inconsistent with the finding of an impulsive origin based on more and better imaging data ( jewitt et al .
2010 , snodgrass et al .
2010 ) . with this as a cautionary reminder
, we regard the evidence concerning activity in 176p , p/2008 r1 , p / la sagra and 300163 as insufficient to diagnose the cause .
evidence concerning 107p and ( 2201 ) oljato is even more fragmentary , and the cause of their activity also unknown .
the results of this and the previous section are summarized for convenience in table ( [ summary ] ) and shown schematically in figure ( [ summary_figure ] ) .
the figure shows the plane of object radius , @xmath84 , vs. heliocentric distance , @xmath17 , with the regions of action of the different physical processes considered in section ( [ mechanisms ] ) marked as color fields . the eleven reported
active asteroids are plotted at the heliocentric distances at which activity was observed .
the figure shows that , for example , ( 596 ) scheila lies in a region of the @xmath84 vs. @xmath17 plane in which only impact and sublimation are capable of ejecting particles .
( as described above , detailed studies of the morphology and time - dependence of the coma of scheila show that impact is the true explanation ) .
conversely , a majority of the active asteroids lie in regions of the @xmath84 vs. @xmath17 plane where many processes are potentially important . for example , in 238p , sublimation , electrostatic ejection , rotational instability , radiation pressure and impact process are all potentially active . only through detailed physical investigation
is it possible to discriminate between these possibilities ( in favor of sublimation , in the case of 238p , based principally on the repetition of the observed mass - loss ) . for many active asteroids ,
the physical observations needed to discriminate amongst mechanisms do not exist .
collisions are implicated in active asteroids both directly , as in the case of ( 596 ) scheila and , perhaps , p/2010 a2 , and indirectly as a trigger for activity ( for example , to expose buried ice ) , as in 133p and 238p . here ,
we briefly examine the expected rate of collision between asteroids . the typical collision probability per unit area in the asteroid belt
is @xmath162 3@xmath2310@xmath163 km@xmath45 yr@xmath22 , with variations by a factor of several reflecting a collisional environment that varies with location in the belt ( bottke et al .
the interval between impacts onto an asteroid of radius @xmath84 is @xmath164 where @xmath165 is the number of impactors larger than @xmath96 .
estimates of the size distribution of the asteroids are many and varied , with significant uncertainties resulting from the unmeasured albedos of most asteroids , as well as from severe observational bias effects ( jedicke et al .
the uncertainties are particularly acute for sub - kilometer asteroids because such objects are faint and remain largely unobserved . for radii
@xmath166 1 km , the best - fitting differential power law index is about -2 , albeit with significant , size - dependent deviations from this value . for radii
@xmath167 1 km , perhaps the best constraints on the distribution come from the impact crater size distribution on asteroid gaspra .
there , craters from 0.4 km to 1.5 km in diameter ( caused by projectiles perhaps 10 to 20 times smaller ) are distributed as a power - law with a differential size index [email protected] ( belton et al .
1992 ; note that chapman et al .
( 1996 ) report that _ fresh _ craters on gaspra follow an even steeper distribution , with differential power law index [email protected] ) .
we assume that the total main - belt population is @xmath01.4@xmath2310@xmath168 asteroids with diameters @xmath1601 km . combining these results and integrating over the size distribution
we take the number of projectiles with radius @xmath169 as @xmath170 where @xmath96 is expressed in meters .
equations ( [ tauc ] ) and ( [ n ] ) together give @xmath171 \label{tauc2}\ ] ] with @xmath84 in km .
lines of constant collision time from equation ( [ tauc2 ] ) are plotted in figure ( [ tauc_plot ] ) .
impacted asteroid ( 596 ) scheila is marked on figure ( [ tauc_plot ] ) with an error bar indicating different estimates of the projectile radius from jewitt et al .
( 2011 ; @xmath10 17 m ) and ishiguro et al .
( 2011 ; @xmath10 40 m ) , respectively .
the corresponding collision times are 3@xmath23 10@xmath172 @xmath1733@xmath23 10@xmath174 yr . given that there are @xmath0250 known asteroids as large as or larger than ( 596 ) scheila , the mean interval between similar events is @xmath0(10 - 100 ) yr , statistically consistent with the detection of scheila within the first decade of efficient , nearly real - time sky monitoring
. disrupted asteroid p/2010 a2 ( interpreted as a @xmath060 m radius object impacted by a projectile of characteristic size 2 to 4 meters ; jewitt et al .
2010 ) is also marked in figure ( [ tauc_plot ] ) , indicating a collision time @xmath175 7 to 50@xmath2310@xmath103 yr . with @xmath176 4@xmath2310@xmath34 ( equation [ n ] ) , the expected rate of similar events is @xmath08 to 60 yr@xmath22 .
jewitt et al .
( 2011 ) estimated the detection probability of p/2010 a2 clones as @xmath1506% , so that of 8 to 60 similar events per year we would currently detect only @xmath00.5 to 4 .
this is still higher than the actual rate of detection , counted as one object in perhaps a decade of efficient , nearly real - time sky monitoring by linear and other survey telescopes .
the high rate could simply indicate the extreme uncertainty in using equation ( [ n ] ) to estimate the number of meter - sized projectiles .
for example , the equation gives @xmath177 = 3@xmath2310@xmath178 , while published estimates of this number range from @xmath010@xmath179 to @xmath010@xmath180 ( davis et al .
2002 ) , indicating a large uncertainty in @xmath181 at small projectile sizes . or it could indicate that activity in p/2010 a2 is caused by another process , possibly yorp spin - up , as remarked in section ( [ p2010a2 ] ) .
further progress in understanding the impact rate in the asteroid belt hinges strongly upon better measurements of the sub - kilometer size distribution .
future wide - area sky surveys may help elucidate the mechanisms operating to cause asteroidal mass loss .
for example , we expect the spatial distribution of collisionally produced or triggered objects to be correlated with the regions of the asteroid belt in which the collision probability per unit time is highest . unfortunately , the published surveys for active asteroids so far are either biased ( hsieh 2009 ) or , if unbiased , detected no objects and so provide insufficient information ( gilbert and wiegert 2009 , 2010 , sonnett et al .
all but one of the known active asteroids were discovered serendipitously by a variety of surveys and methods , most too poorly quantified in terms of reported areal coverage and limiting magnitude to be used to interpret the spatial distribution .
this dismal situation deserves to be soon corrected by well - characterized , all - sky surveys having sensitivity sufficient to discover substantial numbers of active asteroids .
the number of mechanisms capable of producing mass loss from asteroids rivals the number of asteroids showing evidence for mass loss .
no single mechanism can account for the varied examples of activity observed , but preferred explanations can be suggested for particular objects . 1 . sublimation of crystalline ice is effective to the outer edge of the asteroid belt and in asteroids up to a few 100 km in size .
observational evidence for the sublimation of water ice is strongest in the two repeatedly active objects 133p / elst - pizarro and 238p / read . 2 .
activity in ( 596 ) scheila is unambiguously caused by the impact of a decameter - sized projectile , as indicated by the photometric and morphological evolution of this body .
rotational instability is likewise broadly operable , but it is the sub - kilometer asteroids that are most likely to be driven towards instability by yorp effects .
the 120 m diameter p/2010 a2 , if it is not also an impact relic , could be a rotationally disrupted body .
4 . thermal fracture and dehydration cracking are expected to supply dust particles only on asteroids that are very close to the sun ( @xmath1821 au ) and smaller than about 20 km .
radiation pressure sweeping can remove grains from 10 km asteroids up to 3 au , but is more effective at smaller distances .
all these processes may operate in small perihelion object ( 3200 ) phaethon .
ejection of optically active grains ( i.e. @xmath160 0.1 @xmath38 m ) by the electrostatic mechanism is possible for asteroids up to about 10 - 20 km in radius with an efficiency that is independent heliocentric distance .
no clear examples of this process have been identified .
the causes of mass loss from the other active asteroids can not be reliably determined given the limited available data ( see table [ summary ] ) .
davis , d. r. , durda , d. d. , marzari , f. , campo bagatin , a. , & gil - hutton , r. 2002 , asteroids iii , w. f. bottke jr . , a. cellino , p. paolicchi , and r. p. binzel ( eds ) , university of arizona press , tucson , p. 545
- 558 \(3200 ) phaethon & 4.508 & 1.271 & 0.890 & 22.17 & 0.140 & 2.402 + p/2010 a2 & 3.582 & 2.291 & 0.124 & 5.26 & 2.007 & 2.575 + ( 2201 ) oljato & 3.299 & 2.172 & 0.713 & 2.52 & 0.623 &
3.721 + p/2008 r1 ( garradd ) & 3.216 & 2.726 & 0.342 & 15.90 & 1.794 & 3.658 + ( 596 ) scheila & 3.208 & 2.928 & 0.165 & 14.66 & 2.445 & 3.411 + 300163 ( 2006 vw139 ) & 3.203 & 3.052 & 0.201 & 3.24 & 2.438 & 3.665 + 133p / elst - pizarro & 3.184 & 3.157 & 0.165 & 1.39 & 2.636 & 3.678 + 176p / linear ( 118401 ) & 3.167 & 3.196 & 0.192 & 0.24 & 2.582 & 3.810 + 238p / read & 3.152 & 3.165 & 0.253 & 1.27 & 2.364 & 3.966 + p/2010 r2 ( la sagra ) & 3.098 & 3.099 & 0.154 & 21.39 & 2.622 & 3.576 + 107p
/ wilson - harrington & 3.083 & 2.638 & 0.624 & 2.79 & 0.993 &
4.284 + \(3200 ) phaethon@xmath183 & 5 - 7 & 0.08 - 0.17 & 3.6 & [email protected] &
n / a + p/2010 a2@xmath65 & 0.12 & 0.1 & ? & ? & n / a + ( 2201 ) oljato@xmath172 & 1.8 & [email protected] & ? & ? & 5 ? ( gas ) + p/2008 r1 ( garradd)@xmath64 & @xmath630.4 & 0.04 & ? & [email protected] & @xmath271.5 ( gas ) , 0.01 + ( 596 ) scheila@xmath184 & 113@xmath392 & [email protected] & 15.848 & [email protected] & @xmath273 ( gas ) + 300163 ( 2006
vw139)@xmath103 & 3 & 0.04 & ? & ? & ?
+ 133p / elst - pizarro@xmath185 & [email protected] & [email protected] & [email protected] & [email protected] & @xmath630.04 ( gas ) , 0.01 , 0.7 - 1.6 + 176p / linear ( 118401)@xmath186 & [email protected] & [email protected] & [email protected] & [email protected] & 0.1 + 238p / read@xmath105 & 0.8 & 0.05 & ? & [email protected] & 0.2 + p/2010 r2 ( la sagra)@xmath35 & 1.4 & 0.04 & ? & ? & 4 + 107p / wilson - harrington@xmath178 & [email protected] & [email protected] & 7.15 & ? & @xmath27150 ( gas ) + \(3200 ) phaethon & @xmath23 & ? & ? & ? & @xmath187 + p/2010 a2 & @xmath23 & @xmath187 & @xmath23 & @xmath187 & @xmath23 + ( 2201 ) oljato & ? & ? & ? & ? & @xmath23 + p/2008 r1 ( garradd ) & ? & ? & ? & ? & @xmath23 + ( 596 ) scheila & @xmath23 & @xmath187 & @xmath23 & @xmath23 & @xmath23 + 300163 ( 2006 vw139 ) & ? & ? & ? & ? & @xmath23 + 133p / elst - pizarro & @xmath187 & @xmath23 & ? & ?
@xmath23 + 176p / linear ( 118401 ) & ? & ? & ? & @xmath23 & @xmath23 + 238p / read & @xmath187 & @xmath23 & @xmath23 & ? &
@xmath23 + p/2010 r2 ( la sagra ) & ? & ? & ? & ? & @xmath23 + 107p / wilson - harrington & ? & ? & ? & @xmath23 & @xmath23 +
|
some asteroids eject dust , unexpectedly producing transient , comet - like comae and tails .
first ascribed to the sublimation of near - surface water ice , mass losing asteroids ( also called `` main - belt comets '' ) can in fact be driven by a surprising diversity of mechanisms . in this paper
, we consider eleven dynamical asteroids losing mass , in nine of which the ejected material is spatially resolved .
we address mechanisms for producing mass loss including rotational instability , impact ejection , electrostatic repulsion , radiation pressure sweeping , dehydration stresses and thermal fracture , in addition to the sublimation of ice . in two objects ( 133p and 238p ) the repetitive nature of the observed activity leaves ice sublimation as the only reasonable explanation while , in a third ( ( 596 ) scheila ) ,
a recent impact is the cause .
another impact may account for activity in p/2010 a2 but this tiny object can also be explained as having shed mass after reaching rotational instability . mass loss from ( 3200 ) phaethon is probably due to cracking or dehydration at extreme ( @xmath01000 k ) perihelion temperatures , perhaps aided by radiation pressure sweeping . for the other bodies ,
the mass loss mechanisms remain unidentified , pending the acquisition of more and better data . while the active asteroid sample size remains small
, the evidence for an astonishing diversity of mass loss processes in these bodies is clear .
|
super - massive black holes ( smbhs ) of millions to billions of solar masses are believed to exist in the centre of most galaxies .
the galactic center black hole candidate , sgr a * , is the closest and best studied smbh , making it the perfect source to test our understanding of galactic nuclei systems in general . this compact object was first observed as a radio source by @xcite . since then
, observations have constrained important parameters of sgr a * , such as its mass and distance , estimated at @xmath2 and @xmath3 , respectively @xcite .
the accretion rate has also been constrained by polarisation measurements , using faraday rotation arguments @xcite and is estimated to be in the range @xmath4 .
other key parameters , such as the spin , inclination , and magnetic field configuration , are still under investigation .
multi - wavelength observations of sgr a * have been performed from the radio to the gamma ray ( see reviews by @xcite , and references therein ) .
more recently , important progress has been achieved in the infrared ( ir ; e.g. , * ? ? ?
* ) and sub - millimeter ( sub - mm ) domains .
all observations agree that sgr a * is a very under - luminous and weakly accreting black hole ; indeed its accretion rate is lower than has been observed in any other accreting system . in the near future
, the next milestone will be observing the first black hole shadow from sgr a * @xcite with the proposed `` event horizon telescope '' @xcite thanks to the capabilities of very long base interferometry at sub - mm wavelengths .
such a detection would be the first direct evidence for a black hole event horizon , and may also constrain the spin of sgr a*. in order to make accurate predictions for testing with the event horizon telescope , however , we need to have reliable models for the plasma conditions and geometry in the accretion ( in / out)flow .
general relativistic magnetohydrodynamic ( mhd ) simulations offer significant promise for this class of study , as they can provide both geometrical and spectral predictions .
sgr a * has already been modeled in several numerical studies ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
all of these models consist of two separate codes : a grmhd code describing the dynamics , and then a subsequent code to calculate the radiative emission based on the output of the first one .
the under - luminous and under - accreting state of sgr a * ( @xmath5 and @xmath6 erg / s in the 0.5 to 10 kev band , where @xmath7 is the eddington luminosity ; @xcite ) , is the common argument given to justify ignoring the cooling losses and simplify the description of sgr a*. even though this approach seems reasonable , especially for the peculiar case of sgr a * , we propose to quantify to what extent it really applies .
this question is especially important if one wants to extend these studies to more typical nearby low - luminosity active galactic nuclei ( llagn ) .
an important example is the nuclear black hole in m87 , which is the other major target for mm - vlbi and is significantly more luminous than sgr a * ( @xmath8 ) . to test whether or not ignoring the radiative cooling losses is a reasonable approximation , we need to take into account the radiative losses self - consistently in the dynamics , which is now possible with the _ cosmos++ _ astrophysical fluid dynamics code @xcite . by allowing the gas to cool , energy can be liberated from the accretion flow , potentially changing the dynamics of the system .
the basic result that we show in this paper is that cooling is indeed negligible at the lowest end of the possible accretion rate range of sgr a * , but plays an increasingly important role in the dynamics and the resulting spectra for higher accretion rates ( relevant for most nearby llagn and even at the high end of the range of sgr a * ) .
this paper is organized as follows : in section 2 we describe the new version of _
cosmos++ _ used to perform this study . in section 3 we present the initial set - up of the simulations . in section 4
we show the importance of the accretion rate parameter and assess the effect of including radiative cooling . in section 5
we perform a parameter survey to investigate the influence of the initial magnetic field configuration and the spin of the smbh ( we also discuss the effect of a retrograde spin on the dynamics ) on the resulting accretion disc structure and outflow . in section 6
we end with our conclusions and outlook .
the spectra generated from these simulations are analyzed in detail in a companion paper ( drappeau et al .
, in preparation , hereafter referred to in the text as d12 ) .
within _ cosmos++ _ we solve the following set of conservative , general relativistic mhd equations , including radiative cooling , @xmath9 where @xmath10 is the generalized fluid density , @xmath11 is the generalized boost factor , @xmath12 is the transport velocity , @xmath13 is the fluid 4-velocity , @xmath14 is the 4-metric , @xmath15 is the 4-metric determinant , @xmath16 is the total energy density , @xmath17 is the specific enthalpy , @xmath18 is the specific internal energy , @xmath19 is the fluid pressure , @xmath20 is the magnetic pressure , @xmath21 is the stress - energy tensor , @xmath22 is the cooling function , and @xmath23 is the covariant momentum density . with indices , @xmath24 indicates the geometric connection coefficients of the metric . without indices ,
@xmath24 is the adiabatic index . for this work we use an ideal gas equation of state ( eos ) , @xmath25 with @xmath26 .
there are multiple representations of the magnetic field in our equations : @xmath27 is the magnetic field measured by an observer comoving with the fluid , which can be defined in terms of the dual of the faraday tensor @xmath28 , and @xmath29 is the boosted magnetic field 3-vector .
the un - boosted magnetic field 3-vector @xmath30 is related to the comoving field by @xmath31 the magnetic pressure is @xmath32 .
note that , unlike in previous versions of _ cosmos++ _ , we have absorbed the factor of @xmath33 into the definition of the magnetic fields , so - called lorentz - heaviside units ( @xmath34 ) .
we note that the mhd equations as written are all in the form of conservation equations @xmath35 where @xmath36 , @xmath37 , and @xmath38 represent the conserved quantities , fluxes , and source terms , respectively .
these are solved using a new high resolution shock capturing ( hrsc ) scheme recently added to the _ cosmos++ _ computational astrophysics code @xcite .
the new hrsc scheme is modeled after the original non - oscillatory central difference ( nocd ) scheme of _ cosmos _ @xcite .
it also has many of the same elements as the publicly available _ harm _ code @xcite , although with staggered magnetic fields ( instead of harm s zone - centered fields ) and the inclusion of a self - consistent , _ physical _ cooling model .
more details of the new scheme are provided in a separate paper @xcite .
briefly , as in other , similar conservative codes , the fluxes , @xmath37 , are determined using the harten - lax - van leer ( hll ) approximate riemann solver @xcite @xmath39 a slope - limited parabolic extrapolation gives @xmath40 and @xmath41 , the primitive variables at the right- and left - hand side of each zone interface .
from @xmath40 and @xmath41 , we calculate the right- and left - hand conserved quantities ( @xmath42 and @xmath43 ) , the fluxes @xmath44 and @xmath45 , and the maximum right- and left - going waves speeds , @xmath46 and @xmath47 .
the bounding wave speeds are then @xmath48 and @xmath49 . for stability , the equations are integrated using a staggered leapfrog method ( 2nd order in time ) .
first a half - time step , @xmath50 , is taken to project the conserved variables @xmath51 forward to @xmath52 . from these ,
a new set of primitives @xmath53 can be computed .
these intermediate primitives are then used in calculating the fluxes and source terms needed for the full time step , @xmath54 . to find the new primitive variables , @xmath55 , after each update cycle , we use either the @xmath56 or @xmath57 numerical methods of @xcite . along with the evolution equations , we must also satisfy the divergence - free constraint @xmath58 . to accomplish this , _
cosmos++ _ now has the option to use a staggered magnetic field with a constrained - transport ( ct ) update scheme akin to the newtonian version described in @xcite . in the present work
, we assume the whole system is optically thin , i.e. radiation escapes freely from the system . however , for the calculation of the radiation we consider the appropriate optical depth of the gas at a given location and time , which depends on the state .
this approximation takes into account the optical depth without performing radiative transfer , and is valid as long as the ( assumed thermal ) peak of the radiating particle distribution corresponds to energies greater than the self - absorption frequency , which is almost always the case for the regions under study .
the radiative cooling term , @xmath22 , in equations ( [ eqn : en ] ) and ( [ eqn : mom ] ) , is the cooling function introduced to _
cosmos++ _ in @xcite and is based on @xcite , which includes treatments of bremsstrahlung , synchrotron , and the inverse - compton enhancement of each of these .
since sgr a * is optically thin , we do not need to worry about multi - scattering events in the treatment of the inverse compton emission .
the cooling function is calculated locally in each zone from the fluid density , electron temperature , magnetic field strength , and scale height [ @xmath59 .
the local electron temperature is recovered by assuming a constant ion - to - electron temperature ratio , @xmath60 , ( fixed for each simulation ) and calculating the ion temperature from the fluid variables @xmath61 , where @xmath62 is the mean molecular weight ( appropriate for solar abundances ) , @xmath63 is the mass of hydrogen , and @xmath64 is boltzman s constant .
the temperature scale height is defined as @xmath65 .
our modifications to the original @xcite cooling function are described in @xcite and are also described in more detail in our companion paper ( d12 ) . in taking the ion and electron temperatures to be held in a fixed ratio
, we are assuming there is some efficient coupling process at work between the ions and electrons .
the same was done in @xcite and @xcite , except here we relax the assumption somewhat so that the coupling does not have to be exact ( @xmath60 can be greater than 1 ) .
this procedure is not entirely self - consistent since the expression for ion - electron collisions in bremsstrahlung cooling [ equation ( 17 ) of @xcite ] assumes the ions and electrons have the same temperature .
therefore , whenever @xmath66 , we are actually underestimating the amount of bremsstrahlung cooling .
however , because bremsstrahlung is such a minor contributor to the cooling , this omission is not significant in the context of our current work . obviously , assuming a fixed ratio of @xmath60 is not likely to be physically accurate .
however , it is the current standard in most simulations ( e.g. * ? ? ?
* ; * ? ? ?
* ) and recent comparisons with more sophisticated treatments has so far not shown large discrepancies ( dexter , quataert , priv .
comm . ) . because the cooling time of the gas , @xmath67 , can sometimes be shorter than the mhd timestep , @xmath68 , where @xmath69 and @xmath70 are the characteristic zone length and velocity , respectively
, we allow the cooling routine to operate on its own shorter timestep ( subcycle ) , if necessary .
to prevent runaway loops inside the cooling function , we limit the cooling subcycle to 4 steps , with a maximum change to the specific internal energy each subcycle of @xmath71 . even with these restrictions ,
the cooling has the potential to decrease the internal energy ( and hence temperature ) by up to nearly 95% each mhd cycle .
our method for generating spectra is described in detail in our companion paper ( d12 ) .
however , we want to mention one important point related to how we present spectra in this paper . mhd simulations of accretion discs , such as the ones used in our work , generically show significant variability due to the stochastic nature of mri - generated turbulence and magnetic reconnection .
together , these effects can sometimes lead to large flares , similar to those observed in the solar corona .
since our goal is to depict `` representative '' spectra , we have to make a choice about what to consider `` typical '' behavior .
one option would be to produce many simulations using the same initial setup , but with different random seeds , and then select only those simulations that give desirable results .
this was the procedure in @xcite .
we have instead chosen a different approach , in that we perform only one simulation for each set of parameters , and then present the median spectrum for each simulation as the representative one . by choosing the median value of all spectra ,
we minimize the impact of `` extreme '' episodes , especially a few very brief synchrotron / inverse compton flares that appear to be attributable to numerical limitations of the simulations , most particularly the forced axisymmetry . for error bars ,
we give the `` 1-sigma '' variation about the median , i.e. the limits within which 68% of the spectra fall .
for example , if we have 50 individual spectra ( corresponding to 50 different simulation times ) , as is typically the case in this work , then for each spectral energy bin , we drop the 8 highest and 8 lowest data points . in fact dropping just the 4 highest and lowest data points already gives similar results , but we use the 1-sigma values throughout .
in this paper we only include spectra computed over the inner @xmath72 of the simulation domain .
we have confirmed , via comparison with spectra computed using the entire simulation domain , that most of the emission comes from this inner region .
the emission from the jets in our simulations is completely negligible because of their extremely low density .
this extreme level of magnetic domination is likely partially a byproduct of adhering to ideal mhd , where matter can not effectively load the jets .
we hope to explore the jet contribution to the spectrum more in future works .
we start each simulation with a torus of gas around a compact object situated at the origin . the mass of the central compact object is set to the mass of sgr a * ( @xmath73 ) , and the initial density profile inside the torus
is chosen to produce a desired mass accretion rate , somewhere in the range @xmath74 and @xmath75 ( depending on the simulation ) at the inner grid boundary .
the free parameters that describe the torus are the black hole spin @xmath76 ( where @xmath77 is the angular momentum of the black hole and @xmath78 ) , the inner radius of the torus ( @xmath79 ) , the radius of the pressure maximum of the torus ( @xmath80 ) , and the power - law exponent ( @xmath81 ) used in defining the specific angular momentum distribution , @xmath82 we then follow the procedure in @xcite to solve for the initial internal energy distribution of the torus @xmath83 .
this sets its initial temperature profile , @xmath84 . for the purpose of initialization
, we assume an isentropic equation of state @xmath85 , so that now the density is given by @xmath86^{1/(\gamma-1)}$ ] . we can then use the parameter @xmath87 to set the density ( and mass ) normalization of the initial torus .
the torus is seeded with weak poloidal magnetic field loops to drive the magnetorotational instability ( mri ) @xcite .
the non - zero spatial components are @xmath88 and @xmath89 , where @xmath90 & \mathrm{for } & \rho\ge\rho_{\rm cut}~ , \\ 0 & \mathrm{for } & \rho<\rho_{\rm cut}~ , \end{array } \right .
\label{eq : torusb}\ ] ] @xmath91 is the number of field loop centers , and @xmath92 . in this work
we consider configurations with @xmath93 and 4 , as illustrated in figure [ initial_magneticfield ] , in order to investigate the effect of changing @xmath91 .
the parameter @xmath94 is used to keep the field a suitable distance inside the surface of the torus initially , where @xmath95 is the initial density maximum within the torus . using the constant @xmath96 ,
the field is normalized such that initially @xmath97 throughout the torus , where @xmath20 is the magnetic pressure .
this normalization is to ensure that the initial magnetic field is weak .
this way of initializing the field is slightly different than what other groups have done , e.g. @xcite , who used a volume integrated @xmath98 , or @xcite , who used @xmath99 , to set the field strength ( see * ? ? ? * for a discussion of the different methods ) .
we also tested one simulation with @xmath100 .
the higher @xmath101 case ends up with a density ( over @xmath102 ) that is about a factor of 2 higher , a temperature that is 30 - 50% colder , a scale - height that is about 40% thinner , and a magnetic pressure that is about 50% smaller .
the higher @xmath101 case contributes @xmath10370% more flux in the infrared waveband , and @xmath10386% more in the x - ray band , compared to the weaker case ( see d12 ) .
the fact that @xmath104 affects our final result is not surprising since our simulations do not reach a saturated ( magnetically arrested or magnetically choked ) state @xcite .
our resulting field strength is dictated by how much magnetic field is made available to the black hole through our initial conditions . in the background region
not specified by the torus solution , we set up a low density non - magnetic gas .
numerical floors are placed on density and energy density with the following forms : @xmath105 and @xmath106 .
these floors are never applied within the disc , nor in most of the background region .
they are only seldom applied very close to the outer boundary ( the few last external cells ) , and more frequently along the vertical axis in the funnel region .
the most commonly applied floor is that on the ratio @xmath107 .
whenever this quantity drops below 0.01 ( almost always within the jets ) , both @xmath108 and @xmath109 are rescaled by a factor appropriate to maintain this ratio .
all of the simulations are performed in 2.5 spatial dimensions ( all three spatial components of vector quantities are evolved in a single azimuthal slice , although symmetry is assumed in the azimuthal direction ) using a spherical polar coordinate grid .
the grid used in the majority of the simulations consists of 256 radial zones and 256 zones in polar angle . in the radial direction ,
we use a logarithmic coordinate of the form @xmath110 , where @xmath111 is the radius of the black hole horizon ; therefore , @xmath112 .
the inner and outer radial boundaries are set at @xmath113 and @xmath114 , respectively .
note that , because we use the kerr - schild form of the kerr metric , we are able to place the inner radial boundary some number of zones ( usually 4 ) inside the black hole horizon . in principle , this choice should keep the inner boundary causally disconnected from the flow , since mhd signals can not physically radiate out from within the event horizon .
this situation is preferable to having the inner boundary of the grid outside the event horizon , which can lead to artificial behavior within the flow .
the spatial resolution near the black hole horizon is @xmath115 ; near the initial pressure maximum of the torus , it is @xmath116 .
we use outflow boundary conditions at both the inner and outer boundaries ( the mhd primitive variables are copied from interior zones into ghost zones , except the radial velocity component , which is zeroed out if it would lead to inflow onto the grid ) .
since we consider a fairly small radial range , there is some concern that our jets ( discussed in section 5.1 ) might be affected by reflections of waves off the outer boundary of the grid . to test this
, we performed one simulation , b4s9t3m9ce , that used an extended radial grid with @xmath117 .
we found that most properties of this simulation were very similar to the corresponding results done on the smaller ( default ) grid .
the deviations were consistent with those expected from using a different random perturbation seed , as was the case here since the new simulation used a different processor distribution which affects the random seeding . in the angular direction , we include the full range @xmath118 , with reflecting boundary conditions applied at the poles ( meaning that the mhd primitive variables are copied from interior zones into ghost zones , with the sign reversed on the @xmath119 component of all vector quantities , plus the staggered magnetic field component zeroed out at the pole ) .
we use a concentrated latitude coordinate @xmath120 of the form @xmath121 with @xmath122 , which gives us a better resolution near the midplane ( @xmath123 ) .
figure [ initialconfiguration ] shows the initial state of the simulation together with the grid resolution .
our choice of resolution is sufficient to ensure that the fastest growing poloidal field mri modes are well resolved [
@xcite ] for at least the first few orbits of the simulation .
we also confirm that @xmath125 , as expected @xcite .
we also performed select simulations at one - half and at double our default resolution to directly test the numerical convergence of our results ( simulations b4s9l and b4s9h in table 1 ) .
we find very little variation between our default resolution ( such as in simulation b4s9 ) and its higher resolution counterpart ( b4s9h ) , suggesting our results are well converged .
since our ultimate goal is to compare spectra from our simulations with actual data , our true criterion for demonstrating reasonable convergence is to compare spectra generated from simulations at different resolutions .
figure [ resolutiontest ] demonstrates that simulation b4s9 , with a resolution of @xmath126 , gives a spectrum that agrees with simulation b4s9h , with a resolution of @xmath127 , to within our error bars . on the other hand , simulation b4s9l , with a resolution of @xmath128 ,
produces a markedly different spectrum that diverges from the other two .
recently , @xcite confirmed that once a certain resolution is achieved , further increases in resolution have little effect on the spectra .
because the simulations are performed assuming axial symmetry , the anti - dynamo theorem stating that an axisymmetric magnetic field can not be maintained via dynamo action @xcite , prevents the mri from being self - sustaining . after an initial phase of vigorous growth of the mri channel modes , the turbulence gradually dies out as the simulations progress
this 2d approximation is a main caveat of the simulations we have run , and it is therefore important for us to carefully choose the time interval that we wish to analyse .
we want it to be after the initial phase of mri growth , but before the mass accretion has begun to decay too dramatically ( longer simulation times are not necessarily beneficial in 2d for this reason ) . since we have a target mass accretion rate in mind for each simulation , we can use that as a guide for selecting our time interval .
figure [ mdot ] shows the variation of the accretion rate as a function of time for three simulations that include cooling ( b4s9t3m9c , b4s9t3m8c , and b4s9t3m7c ) .
we see that it takes roughly 1.5 orbits to have the accretion reach its peak value , and that the accretion dies out ( returns to the background rate ) after about orbit 5 , where we are referring to the circular orbital period at @xmath129 , i.e. @xmath130 . for these three simulations , the target accretion rates were @xmath74 , @xmath131 , and @xmath132 , respectively .
figure [ mdot ] then suggests that an appropriate interval would be between 2.5 and 3.5 orbits .
we therefore use this as the standard time interval for analysis . in the rest of this paper , @xmath133 .
we confirmed that our region of interest ( @xmath134 ) has reached inflow equilibrium over this time interval , based , for instance , on the mass flux being constant as a function of radius over this region .
therefore , we can also be confident that our selected time interval is late enough not to be affected by transient solutions .
as we have stated , our primary goal in this paper is to assess the importance of including radiative cooling losses self - consistently in numerical simulations of llagn like sgr a*. we do this by comparing two types of simulations : those that include radiative cooling losses self - consistently in the dynamics ( indicated by a `` c '' in the simulation name or cooling `` on '' in table 1 ) and those that neglect them ( no `` c '' , cooling `` off '' ) .
we performed a set of 25 different simulations to assess the importance of radiative cooling losses , as well as study the influence of parameters such as the initial magnetic field configuration , the spin , the ion to electron temperature ratio , and the mass accretion rate .
table 1 summarizes all of the simulations performed .
the name given to each simulation is derived from its parameter choices in the order they appear in the table .
we also include the average and root mean square ( rms ) fluctuations of the mass accretion rates actually measured in our simulations .
this value is to be compared with our target values of @xmath135 . in simulations without cooling ,
this value is arbitrary since the initial density ( or mass ) of the disc is a free parameter ( other than the requirement that the total mass of the disc @xmath136 be negligible compared to @xmath137 since we ignore the self - gravity of the disc ) .
its choice does not affect the subsequent evolution of the simulation , so can be freely rescaled after the fact .
this freedom does not extend to the simulations that include cooling , since the cooling function , @xmath22 , depends directly on @xmath108 .
therefore , to test different mass accretion rates , it is sufficient in simulations without cooling to do a single simulation and then simply rescale it to each of the target accretion rates , whereas in the simulations with cooling , a new simulation must be conducted for each new mass accretion rate .
this physical scaling is an important distinction between our work and that of previous authors each of our simulations with cooling has a real , _ physical _ mass accretion rate associated with it ; it is no longer a parameter that can be freely fit to the data . when comparing simulations with cooling against those without , we always assume the same initial accretion rate . [ cols="<,^,^,^,^,^,^,^,^,^",options="header " , ] in our models we start with a relatively weak ( @xmath138 ) magnetic field contained entirely within the disc , initially in the form of poloidal loops .
we tested two different configurations : a single set of poloidal loops centered on the pressure maximum of the torus and following contours of pressure / density ( an example is simulation b1s9t3m9c ) and four sets of poloidal loops spaced radially , with alternating field directions in each successive loop ( an example is simulation b4s9t3m9c ) .
the 1-loop configuration yields a significantly more energetic outflow ( mostly associated with the `` jets '' ) more than an order of magnitude stronger than for the more complex , though perhaps more realistic , configuration consisting of 4 initial loops .
we have confirmed that , as expected , the overall power of both components scales roughly linearly with the mass accretion rate , meaning the efficiencies of the outflows are not strongly effected by the cooling . aside from the outflow
, the initial magnetic field configuration also has a strong influence on the accretion rate as seen by comparing the measured @xmath135 in table 1 for simulations b4s9t3m9c and b1s9t3m9c .
the 1-loop simulation has a substantially higher average @xmath135 , as well as much larger statistical fluctuations .
this difference is likely a consequence of the axisymmetric nature of these simulations . in our simulations ,
accretion is driven by the mri .
for the 1-loop case , the dominant channel solution is able to disrupt nearly the entire disc , driving large amounts of material onto the black hole in multiple short , intense episodes . in the 4-loop cases ,
the channel solutions are restricted in their radial range and can not disrupt the disc as effectively ; accretion proceeds more smoothly , and at an overall lower normalization .
[ the channel solution @xcite is a particularly violent form of the poloidal mri and is characterized by prominent radially extended features . ]
these effects are illustrated in figure [ rho_1b4b ] . ) of the system for simulations b1s9t3m9c and b4s9t3m9c at the same time step ( t= 3 orbits ) .
the two simulations started with the same initial torus configuration as shown in figure [ initialconfiguration ] .
the only initial difference between the two models is that the left panel started with a single magnetic loop configuration ( as shown in the bottom panel of figure [ initial_magneticfield ] ) , while the right panel started with a 4 magnetic loop configuration ( as shown in the top panel of figure [ initial_magneticfield ] ) .
we can see that in the case of the single initial magnetic loop , the channel solution extends through the entire disc , resulting in a higher mass accretion rate at the black hole horizon and a more extended disk . _ _ ] figure [ sed_b1b4_rt_median ] compares the spectra for the two different initial magnetic field configurations , the `` 1-loop '' configuration ( in black ) , and the `` 4-loop '' configuration ( in red ) .
it is clear that the magnetic field configuration has a strong influence on the resulting spectra and is therefore one of the big uncertainties affecting all such models of the accretion flow .
the lack of _ a priori _ knowledge of the field configuration introduces an effective error in all simulation conclusions that is comparable to the difference introduced by including cooling .
thus , while we can treat the cooling losses self - consistently , it is difficult to overcome the variations that different choices in initial magnetic field configurations will introduce . .
_ _ ] via its formation and accretion history , sgr a * is likely a rotating smbh , but as with all black holes it is currently difficult to reliably determine its spin .
although the observed spectrum should depend on spin , it also depends on the mass accretion rate , which is somewhat constrained , as well as the geometry of the accretion flow .
nevertheless , spectral fitting of sgr a * has already placed some meaningful constraints on the value of @xmath139 @xcite , although there can be additional complicating factors , such as disk tilt @xcite .
other methods of inferring the isco , and then the spin , such as observing relativistically - broadened line profiles or modeling the thermal emission from the disc , require the presence of a geometrically thin , optically thick disc , which is not observed at the low accretion rates of sgr a*. the ultimate determination of spin will likely require future observations from mm - vlbi @xcite .
we investigated the spin dependence of our results by performing simulations at six different values of @xmath139 : -0.9 , 0 , 0.5 , 0.75 , 0.9 , and 0.98 , where the minus sign indicates retrograde rotation ( disc and black hole rotating in opposite senses ) .
we will discuss the effects of spin on our spectral fits in companion paper d12 .
concerning the disc evolution , there does not seem to be any clear trend in @xmath135 , see table 1 . if anything there appears to be a peak in the distributions , with the highest mass accretion rates occurring at intermediate spins , @xmath140 .
the outflow power on the other hand , shown in table 2 , reveals a clear trend with spin , as expected based on theoretical grounds ( e.g. * ? ? ?
* ; * ? ? ?
* ) and from previous numerical work @xcite .
this trend is seen in both the fluid and magnetic components of the outflow and the correlations are shown in figure [ jetkineticenergy_spin ] . the effect of spin on jet power is a question that has been debated recently in the literature , specifically using claimed spin observations from black hole x - ray binaries .
so far the observational results are not definitive ( see , e.g. , the opposing conclusions found in @xcite and @xcite ) .
our results , while focusing on smbhs , would be predicted to scale to smaller mass black hole systems via the fundamental plane of bh accretion relation @xcite .
semi - analytical models for the compact jets observed in weakly accreting systems predict a dependance of radio luminosity on total jet power of @xmath141 ( e.g. , * ? ? ?
thus if we consider the total power as the sum of the fluid and magnetic components in figure [ jetmagneticenergy_spin ] , we would also predict a correlation between observed radiative power and spin . in all our simulations we started with the same disk thickness
( @xmath142constante @xmath143 ) in order to start with a thick disk , appropriate for the galactic center .
then the shape of the disk is modified during the simulation and is especially sensitive to the cooling as shown in figure [ rho_cool ] .
we also investigated thinner disk simulation , the thinner disk emmit a little bit less that the thick disk but they could be compatible within one sigma .
for the first time we have been able to assess the importance of the radiative losses in numerical simulations of llagn , specifically sgr a * [ see also @xcite ] .
we show that radiative losses can affect the dynamics of the system and that their importance increases with accretion rate .
we set a rough limit of @xmath144 , above which radiative cooling losses should be included self - consistently in numerical simulations .
otherwise , many important derived dynamical quantities , such as density , magnetic field magnitude , and temperature , may be off by an order of magnitude or more , especially when the accretion rate reaches @xmath145 , which correspond to @xmath146 for sgr a*. since several recent works suggest that accretion physics is similar across the mass scale , this accretion rate in eddington units should likely be an important limit for all black holes .
thus , we predict that the inclusion of self - consistent radiative cooling above @xmath147 should be important for llagn in general . overall , this study allows us to have a more consistent model of accretion , even for the well - studied and under - luminous source , sgr a*. not only do we have a more realistic model by including the physics of radiative losses , but we are also able to show the influence of the accretion rate on the resulting spectra .
the spin of the central black hole and the initial magnetic field configuration of the torus also have important consequences on the dynamics of the system and the resulting spectra . by including the cooling losses in our study
, we can discuss the influence of these free parameters with more accuracy .
for instance , initial magnetic field configurations consisting of a single set of poloidal loops result in significantly more powerful outflows than the four - loop cases , and we find that the jet power increases with the spin of the central black hole . as mentioned in section [ sec : time_interval ] , there is concern as to whether or not our simulations have run long enough to reach meaningful equilibrium states .
this concern is especially pertinent for our case using 2.5d simulations , as these can never truly reach a steady state .
the reason is that , after a period of initial vigorous growth , the mri in 2.5d simulations steadily decays because the dynamo action that normally sustains it requires access to non - axisymmetric modes that are obviously inaccessible . we , therefore , chose a time interval for analysis when the mass accretion rate closely approximated our target value for most simulations .
clearly simulations in 3d with a longer duration ( to ensure a proper equilibrium is reached ) would be beneficial , for comparison .
one other concern with only performing 2.5d , axisymmetric simulations is that it precludes the possibility of exploring the effects of misalignment between the angular momentum of the gas and the black hole . in reality , most supermassive black holes are unlikely to be accreting from matter sharing the same orbital plane as the black hole spin .
@xcite recently showed that accounting for such a `` tilted '' accretion flow can dramatically alter the best - fit characteristics of sgr a * and produce important new features in the spectra .
as a final note , one can justify neglecting full radiative transfer in simulations of sgr a * and likely most other weakly accreting llagn because the inner regions are generally thought to be optically thin . however , to treat the outer regions of the accretion flow , or higher luminosity sources , a more thorough treatment of radiative transfer will need to be implemented into the simulations .
similarly , to approach important questions such as the mass loading and particle acceleration in the jets , likely resistive ( non - ideal ) mhd will need to be considered .
we see this work as an important step towards these next technological `` horizons '' .
we acknowledge support from the european community s seventh framework programme ( fp7/2007 - 2013 ) under grant agreement number itn 215212 black hole universe .
sm and sdi gratefully acknowledge support from a netherlands organization for scientific research ( nwo ) vidi fellowship .
this work was also partially supported by the national science foundation under grants ast 0807385 and phy11 - 25915 and through teragrid resources provided by the texas advanced computing center ( tacc ) .
we thank sara computing and networking services ( www.sara.nl ) for their support in using the lisa computational cluster .
pcf acknowledges support of a high - performance computing grant from oak ridge associated universities / oak ridge national laboratory .
s. , agol e. , backer d. , baganoff f. , bower g. c. , broderick a. , fabian a. , fish v. , gammie c. , ho p. , honman m. , krichbaum t. , loeb a. , marrone d. , reid m. , rogers a. , et al .
2009 , in astro2010 : the astronomy and astrophysics decadal survey vol .
2010 of astronomy , imaging an event horizon : submm - vlbi of a super massive black hole .
s. s. , weintroub j. , rogers a. e. e. , plambeck r. , freund r. , tilanus r. p. j. , friberg p. , ziurys l. m. , moran j. m. , corey b. , young k. h. , smythe d. l. , titus m. , marrone d. p. , cappallo r. j. , bock d. c .- j .
2008 , , 455 , 78 r. , ott t. , genzel r. , hofmann r. , lehnert m. , eckart a. , mouawad n. , alexander t. , reid m. j. , lenzen r. , hartung m. , lacombe f. , rouan d. , gendron e. , rousset g. , lagrange a .- m .
, brandner w. , et al .
2002 , , 419 , 694
|
we present general relativistic magnetohydrodynamic ( grmhd ) numerical simulations of the accretion flow around the supermassive black hole in the galactic centre , sagittarius a * ( sgr a * ) .
the simulations include for the first time radiative cooling processes ( synchrotron , bremsstrahlung , and inverse compton ) self - consistently in the dynamics , allowing us to test the common simplification of ignoring all cooling losses in the modeling of sgr a*. we confirm that for sgr a * , neglecting the cooling losses is a reasonable approximation if the galactic centre is accreting below @xmath0 i.e. @xmath1 . but above this limit , we show that radiative losses should be taken into account as significant differences appear in the dynamics and the resulting spectra when comparing simulations with and without cooling .
this limit implies that most nearby low - luminosity active galactic nuclei are in the regime where cooling should be taken into account .
we further make a parameter study of axisymmetric gas accretion around the supermassive black hole at the galactic centre .
this approach allows us to investigate the physics of gas accretion in general , while confronting our results with the well studied and observed source , sgr a * , as a test case .
we confirm that the nature of the accretion flow and outflow is strongly dependent on the initial geometry of the magnetic field .
for example , we find it difficult , even with very high spins , to generate powerful outflows from discs threaded with multiple , separate poloidal field loops .
[ firstpage ] accretion discs black hole physics mhd
radiation mechanisms : thermal relativistic processes methods : numerical galaxy : center galaxy : nucleus galaxies : jets .
|
semidetached binary systems are interacting binaries , in which one of the stars fills its critical surface , resulting in mass exchange between the components of the system . in general , the critical surface has a complicated shape [ 1 ] , and special mathematical models are required to describe matter flows in the system ( a review of such models is given in [ 2 ] ) .
however , the large amount of observational data confirming the small eccentricity of the orbits and the high degree of synchronization in these binary systems makes it possible to simplify the models used . in the standard formulation
, the components of semidetached binary systems are assumed to have circular orbits and to rotate in synchrony with the orbital motion . in this case , the critical surface is identified with the inner surface ( roche surface ) in the restricted three - body problem , and it is assumed that the mass exchange between the system components occurs through the vicinity of the inner lagrange point @xmath0 , where the pressure gradient is not balanced by the force of gravity . the gas dynamics of mass transfer through the inner lagrange point @xmath0 have been studied by many authors .
a thorough analysis of the flow of matter in the vicinity of @xmath0 was performed by lubow and shu [ 3 ] , who estimated the basic characteristics of the flow in a semi - analytical approximation using a perturbation method .
the small parameter @xmath1 used in this analysis was the ratio of the sound speed @xmath2 at the star s surface to the characteristic orbital velocity @xmath3 ( @xmath4 is the angular velocity of rotation of the system components and @xmath5 is the distance between the component centers ) . in another approach based on bernoulli integral analysis , the parameters of the stream of matter were refined , and a relationship obtained between the mass transfer rate and the degree of overfilling of the roche lobe by the donor star [ 4,5 ] . to adequately describe the mass exchange process
, it is necessary , in addition to determining the stream parameters , to consider the further behavior of the stream lines after the motion of matter from @xmath0 .
it is the mass transfer process that forms the overall pattern of the flow and , consequently , defines its basic observational manifestations ; therefore , much attention has been given to the study of this problem .
warner and peters [ 6 ] , lubow and shu [ 3 ] , and flannery [ 7 ] pioneered the investigation of particles ejected from @xmath0 and moving in the gravitation field of the binary system .
the results obtained in these and a number of subsequent studies have gained wide recognition , but their application to the analyses of specific systems led to the discovery of substantial discrepancies with the observational data .
these discrepancies are associated with the use of a simplified ballistic approach in the gas motion analysis , which did not take into account gas dynamics effects . to consider the influence on gas motions of a circumstellar envelope during and after its formation , and , consequently , to accurately describe matter flows , it is necessary to solve a complete system of gas dynamics equations ; this is possible only using quite sophisticated mathematical models .
the use of numerical methods to study the gas dynamics of mass transfer in semidetached binary system was for a long time limited by the development of computer techniques , so that two - dimensional ( 2d ) models were used for the flow analysis . in spite of the simplification of the 2d approach
, it enabled a correct treatment of some particular features of the matter flow , and made it possible to obtain a number of interesting results ( see , e.g. , [ 812 ] ) .
in recent years , numerical investigations of mass transfer gas dynamics using more realistic three - dimensional ( 3d ) models have become possible [ 1320 ] .
such studies have considered , in particular , accretion disk formation in semidetached binary systems [ 13 , 15 ] , and also the region of interaction between the stream flowing from @xmath0 and the disk [ 14 , 20 ] .
unfortunately , these studies were conducted in a limited formulation , without considering the process of establishing the flow on long time scales ; this prevented investigation of the real morphology of the systems and , correspondingly , evaluation of the impact of a common envelope on the flow pattern .
some progress in the investigation of the overall flow pattern was achieved in the studies of molteni , belvedere , and lanzafame [ 1619 ] , who conducted 3d numerical simulations over rather long time intervals .
a number of interesting results were obtained in these papers , however , the use of the smoothed particle hydrodynamics ( sph ) method to solve the system of gas dynamics equations made it impossible to study the influence of a common envelope on the flow pattern .
the restrictions imposed by the sph method do not allow investigation of flows with large density gradients , and , therefore , the action of the rarefied envelope gas on the mass transfer gas dynamics was not taken into account entirely correctly . here , we present the results of 3d numerical simulations of the flow patterns in semidetached binary systems .
the total variation diminishing ( tvd ) method we use to solve the system of gas dynamics equations does not have the disadvantages of the sph method .
this made it possible to investigate gas flow morphology in the system and the effect of the common envelope of the system , despite the presence of substantial density gradients .
we studied the gas dynamics of matter flows in semidetached binary systems over long time intervals , which enabled us to consider the basic features of the flow pattern in a steady - state ( stationary ) regime .
semidetached binary systems exhibiting extremely interesting observational manifestations of their interaction include cataclysmic binaries , low mass x - ray binaries ( lmxb ) , and supersoft x - ray sources .
the available observations of lmxb x - ray light curves provide striking evidence for the presence of complicated flow patterns , and suggest some characteristic features of the gas flow structure .
in particular , in a number of x - ray `` dipping '' sources , of which the most widely known is the x1822371 system , a substantial decrease of the emission is observed at phase 0.8 of the orbital period , associated with a thickening in the disk in the place where it is perturbed by the stream of gas flowing from @xmath0 [ 2022 ] . in our studies of matter flow patterns
, we considered semidetached binary systems with parameters typical for low mass x - ray binaries , where a main sequence dwarf fills its roche lobe and transfers mass to a neutron star . for more specific calculations , we chose a system with parameters close to those of x1822371 [ 20 ] .
we assumed that the primary star filling its roche lobe has a mass @xmath6 equal to @xmath7 and a surface gas temperature @xmath8 ; the mass of the companion ( a compact object with radius @xmath9 ) has @xmath10 ; the orbital period of the system is @xmath11 ; and the distance between the centers of the two stars is @xmath12 . to adequately describe the matter flow structure in the binary system we have chosen
, we must take into account the effect of radiation processes upon the gas dynamics .
including non - adiabatic processes in the numerical model substantially increases the calculation time . for this reason , based on the available computational resources and the need to let the calculations run over long time intervals ( to obtain a steady - state regime for the flow ) , we restricted our consideration to an ideal gas model with adiabatic index @xmath13 and equation of state @xmath14 ( where @xmath15 is pressure
, @xmath16 is density , and @xmath17 is specific internal energy ) .
the ideal gas assumption in the model formally neglects radiative losses ; however , in the calculations presented , the adiabatic index was assumed to be close to unity ( namely , @xmath13 = 1.01 ) , which approaches the isothermic case [ 23 ] .
this , in turn , allows us to treat the @xmath18 model used as a system with energy losses . a similar technique for taking into account radiative losses is well known and has been applied in a number of studies ( see , for example , [ 16 , 24 ] )
this procedure clearly prevents accurate allowance for radiative cooling of gas in the system , but , nonetheless , brings the model used closer to a realistic case .
we used a 3d system of gas dynamics equations , closed by the ideal gas equation of state with adiabatic index @xmath19 , to describe the gas flow .
we used an integral form of this system , since these equations allow , together with continuous solutions , nonisoentropic solutions satisfying the rankin
hugoniot conditions at discontinuities ( shock waves and contact discontinuities ) .
the system has the form @xmath20 @xmath21 @xmath22 the specific external force * f * includes the coriolis force , and also the attractive force of the two gravitational centers and the centrifugal force ( both described by the roche potential ) .
it has the form where the roche potential can be written @xmath24 here , @xmath25 is the velocity , @xmath26 , @xmath27 , @xmath28 and @xmath29 are the centers of the system components , and @xmath30 is the center of mass of the system .
we adopted the following boundary conditions .
\(1 ) the outflowing star fills its roche lobe . over the entire surface of this component ,
the gas velocity is normal to the surface .
the gas speed is assumed to equal the local sound speed @xmath31 .
\(2 ) the conditions of free matter outflow are set both on the accretor and at the outer boundary .
\(3 ) the density @xmath16 in the outflowing component s roche lobe was taken to be @xmath32 .
note that the boundary value of the density in the outflowing component does not affect the solution , since the system of equations scales with @xmath16 and @xmath15 .
we used an arbitrary value for @xmath32 in the calculations .
however , when a specific system with a known mass loss rate is considered , the real density values can be found simply by adjusting the calculated density values according to the scale implied by the ratio of the real and model densities at the surface of the outflowing component .
we took the gas in the calculation region to initially have low density and be at rest .
this gas is later displaced by the gas flowing outward from the primary component . a fundamental problem in the numerical realization of gas dynamics models is the choice of a method for solution of the corresponding system of equations .
currently , finite difference methods recommend themselves most highly . of the wide variety of finite difference schemes ,
the most accurate are so - called godunov methods [ 25 ] .
these are based on approximating the initial gas dynamics equations in divergence form ( corresponding to the conservation laws ) , which makes it possible to obtain correct values at discontinuities in the case of discontinuous solutions .
the important features of finite difference methods , which , in many respects , determine the resulting adequacy of the solutions , are the approximation viscosity and the monotonicity of the scheme . increasing the approximation order decreases the numerical viscosity .
it is possible to decrease the numerical viscosity in a scheme with a given order using characteristic properties of the initial system of gas dynamics equations , which is hyperbolic .
this approach leads to so - called total variation diminishing ( tvd ) schemes , which have the property of monotonicity ( corresponding to the absence of nonphysical oscillations of the solutions near discontinuities ) [ 26 ] . increasing the approximation order and introducing special flow restrictions that make it possible to preserve the monotonicity lead to high - order tvd methods with minimal approximation viscosity [ 27 ] . in spite of their relative complexity ,
such schemes are currently the most promising for the numerical realization of gas dynamics models .
we used a high - order tvd scheme to numerically solve the system of equations .
special studies showed that this method makes it possible to adequately reflect flow structures that include shock waves and tangential discontinuities , without leading to artificial oscillations or smearing of features in the flow pattern .
in addition , the method allows us to study flows with large density gradients , which is especially important when treating the influence of common envelope gas on the flow structure .
we solved the gas dynamics equations in a cartesian coordinate system specified as follows : - origin at the center of the outflowing star ; - @xmath33 axis directed along the line connecting the star centers , from the outflowing component to the accretor ; - @xmath34 axis directed along the rotation axis ; - @xmath35 axis defined so that the resulting system is right - handed .
the calculation region is the parallelepiped @xmath36 ; by virtue of the symmetry of the problem , we carried out the calculations only for the upper half - space .
we used a non - uniform ( denser in the accretor zone ) difference grid with @xmath37 nodes for the numerical calculations .
the system of equations was solved using a method for transition to a steady state , from arbitrarily chosen initial conditions to the entry of the solution to a stationary regime .
note that the characteristic gas dynamics timescale for the establishment of the flow for the system under consideration ( the ratio of the system s characteristic size @xmath5 to the propagation velocity of perturbations @xmath2 ) is of the order of three orbital periods .
thus , to ensure that the resulting solution was stationary , we carried out the calculations over a substantially longer time interval more than ten orbital periods .
we checked the degree of stationarity of the solution using both local and integrated characteristics of the flow .
we will consider some characteristic features of the flow structure in semidetached binary systems obtained using the 3d gas dynamics model described in section 2 . the overall flow pattern , which illustrates the morphology of gas flows in the system , is shown in fig . 1 , which presents three - dimensional images of density contours at the levels of 0.01 ( fig .
1a ) and 0.05 ( fig .
1b ) of the gas density at the point @xmath0 .
to better show details of the calculated flow structure , fig .
2 shows cross sections of the density contours in fig .
1a made by the @xmath38 and @xmath39 planes passing through the accretor . the flow pattern in fig . 1 and fig .
2 is time steady , and corresponds to the steady flow regime achieved after a time interval exceeding ten orbital periods from the starting time of the calculations .
these results reveal the following features of the calculated flow structure .
\1 ) the matter of the stream can be conditionally divided into three parts : the first forms a quasi - elliptical accretion disk ; the second bends around the accretor and , remaining inside the system , acts upon the overall flow pattern ; and the third leaves the system through the vicinity of lagrange point @xmath40 .
\2 ) the interaction between the stream and the disk is not a shock interaction .
\3 ) the dimensions of the stream of matter flowing from the vicinity of @xmath0 change as the stream propagates toward the accretor : its height decreases and its width in the orbital plane increases .
\4 ) the height of the accretion disk that forms in the system is much less than the height of the stream ; this is confirmed by the data presented in fig . 3 , which shows density contours up to the value @xmath41 and velocity vector in the @xmath39 plane , which forms a transverse cut through the disk when it passes through the accretor .
we can analyze the structure of gas flows in more detail and estimate the disk s linear size by considering the flow pattern in the equatorial plane .
figure 4 presents the density field and velocity vector field in this plane inside an area with dimensions from 2 to @xmath42 along the @xmath33 axis and from -3 to @xmath43 along the @xmath35 axis .
four stream lines , marked with the letters a , b , c , d and indicating the directions of flows in the system , are also shown .
analysis of the results presented in fig .
4 confirms the above conclusion that some of the stream material leaves the system through the vicinity of the lagrange point @xmath40 without getting close to the accretor ( stream line @xmath5 ) .
some material falls directly onto the disk ( stream line d ) and further , losing its angular momentum under the action of viscosity , takes part in the accretion process .
the quantitative estimates obtained indicate that about 75of gas injected into the system is accreted . of the most interest for the purposes of this paper
is the material that acts upon the flow pattern while remaining inside the system ( stream lines b and c in fig . 4 ) .
we will call this material the `` common envelope '' of the system .
note that a significant part of the common envelope gas ( see stream line b ) bends around the donor star and interacts with the matter outflowing from its surface .
the influence of this part of the common envelope on the flow structure leads to substantial changes in the mass exchange regime in the system .
we will provide a detailed consideration of this effect in a future paper .
another part of the common envelope ( see stream line c ) passes completely around the accretor and undergoes a shock interaction with the edge of the stream facing into the orbital motion .
this type of interaction results in substantial changes in the overall flow pattern , in particular , in the absence of a `` hot spot '' in the disk and the formation of an extended shock wave along the edge of the stream .
we present a detailed description of the impact of this part of the common envelope on the gas flow morphology in the system below . to determine the linear dimensions of the disk , we must find the limiting ( `` last '' ) stream line along which matter flows directly onto the disk . the preceding stream line ( between lines c and d ) , which directs matter around
the accretor and then back to interact with the stream , still passes outside the disk .
the matter along this line belongs to the common envelope and not to the disk , although later ( after interaction with the stream ) , it can also be accreted . the limiting stream line in fig .
4 is d , and we can easily estimate the dimensions of the calculated quasi - elliptical disk by examining this line : @xmath44 .
it follows from fig . 3 that the disk thickness increases with distance from the accretor , varying from @xmath45 to @xmath46 .
note that to determine the disk thickness more accurately , we must increase the number of grid nodes , since the minimum calculated thickness of the disk was limited by the size of a difference cell .
the variations of the parameters for the gas flowing along the stream lines presented in fig . 4 show that the flow is smooth along all lines belonging to the disk , up to the limiting line d. the absence of discontinuities indicates a shockless interaction between the stream and disk material , which , in turn , implies the absence of a hot spot in the disk .
the results presented in figs . 1 , 2 and 4
provide an explanation for this shockless morphology for the stream disk system .
examination of the flow pattern makes it clear that the stream deflected by the action of the common envelope gas ( stream line c in fig .
4 ) approaches the disk along a tangent , and , therefore , does not cause any shock perturbation at the disk surface . at the same time
, the interaction of the stream with the common envelope leads to the formation of an extended shock wave along the edge of the stream facing into the orbital motion .
we can estimate the shock parameters , as well as the total amount of energy released in the shock , using the calculation results .
one of the characteristic features of the shock is its variable intensity .
this is illustrated in fig . 5 , which presents the distribution of the specific rate of energy release along the shock in the equatorial plane @xmath47 , normalized to unity .
the vertical dashed lines in fig .
5 show the edges of the stream : the left line shows the point where the stream originates ( the point @xmath0 ) , and the right line the place where the stream ends , i.e. , the place where the stream comes into contact with the disk .
we can easily seen in fig . 5 that the bulk of the energy in the system is released in the compact shock region , adjacent to the accretion disk .
this is extremely important for interpretations of observational data , because the compactness of the energy release region can explain the phase relations of features in binary light curves in essentially the same way as has been done previously assuming the presence of a hot spot , in the framework of a fundamentally different model for the matter flow in the system . to verify that the shock forming along the edge of the stream is an adequate substitution for the hypothetical hot spot in the disk , let us consider the total energy release in both of these formations . in accordance with standard models , we assume the hot spot to be generated at a distance @xmath48 from the accretor . in this case , the energy release resulting from the stream and disk interaction is ( see , e.g. , [ 28 ] ) : @xmath49 where @xmath50 is the mass loss rate of the donor star . for the adopted values of the density @xmath32 and temperature ( sound speed @xmath2 ) at the star s surface ,
the value of @xmath50 in standard models is determined by the relation @xmath51 where @xmath52 is the cross - sectional area of the stream in the vicinity of @xmath0 [ 3 ] .
using these formulas and determining @xmath48 from the calculation results given above , assuming the spot is generated at the point where the stream contacts the disk , we estimated @xmath53 for the gas density and temperature at the surface of the donor star adopted in our model .
comparison of the energy release rate in the shock @xmath54 with our @xmath53 estimate shows that @xmath54 is twice @xmath53 .
these approximately equal energy release rates and the quite limited size of the region where the bulk of the energy in the shock is released ( @xmath55 60% of the energy is released in the one - third of the shock that is adjacent to the disk see fig .
5 ) , suggests that the observational manifestations of this shock will be roughly equivalent to those of a hot spot in the disk .
finally , we note that , in spite of the expected similarity of the observational effects of the hypothetical hot spot and our calculated shock wave , the calculated matter flow morphology in semidetached binary systems is fundamentally different from the standard picture , and this , in turn , calls for a reexamination of a number of commonly accepted concepts .
for instance , the presence of a common envelope leads to significant changes in the mass transfer rate in the system . in the case
considered , the common envelope causes matter to flow from a much ( order of magnitude ) larger part of the surface of the donor star than is supposed in standard models ( this point will be considered more thoroughly in a future paper ) .
this means that for approximately equal energy release rates in the different ( `` shock wave '' and `` hot spot '' ) models , the mass exchange rates will be substantially different .
correct concepts about the flow morphology are especially important when interpreting observations , since comparisons between model calculations and observational data can provide information about the system .
we have presented the results of 3d numerical simulations of the matter flow pattern in a semidetached binary system similar to the low mass x - ray binary x1822371 .
our results provide evidence for the substantial influence of the rarefied common envelope gas on the structure of gas streams in the system . the common envelope gas interacts with the stream outflowing from the vicinity of @xmath0 and deflects it , leading to a shockless ( tangential ) interaction between the stream and the outer edge of the accretion disk , and , consequently , to the absence of a hot spot in the disk . at the same time , the interaction of the common envelope gas with the stream results in the formation of an extended shock wave of varying intensity along the edge of the stream .
a preliminary analysis suggests that the observational manifestations of this shock and the standard hypothetical hot spot are similar , and , therefore , this shock can be considered equivalent to a hot spot in the interpretation of observations .
note that these results were obtained for a steady flow regime . in a non - stationary regime , when the flow morphology is determined by external factors and is not self - consistent , regions of shock interaction between the disk and the gas stream are also possible . for example , if the disk is already formed before the donor star fills its roche lobe , a hot spot can be created where the matter stream makes contact with the outer edge of the disk after the mass exchange through the vicinity of @xmath0 begins . since a self - consistent solution without a hot spot is expected after a steady - state flow regime is achieved , it is important to determine the life time for the hot spot .
it is natural to adopt for its characteristic life time the time required for the amount of matter injected into the system by the stream to become comparable to the accretion disk mass , since after complete mass exchange , the solution will become self - consistent . with mass exchange and accretion disk parameters typical for the system under consideration [ 20 ]
, we expect that the stationary flow regime will already be established after a time of the order of 100 rotation periods .
this suggests that the probability of observing a hot spot is extremely low , and all the observational manifestations that have been associated with hypothetical hot spots are , in fact , consequences of energy release in the shock on the stream edge .
in summary , we conclude that correctly taking into account the common envelope in numerical simulations of the gas dynamics of mass transfer in semidetached binary systems has revealed new features of the gas flow structures , which significantly change our concepts of the morphology of flows in these systems .
|
results of numerical simulations of matter flows in a semidetached binary system similar to the low - mass x - ray binary x1822371 are presented .
three - dimensional modeling of the mass transfer gas dynamics makes it possible to investigate gas streams in the system and to study the influence of a common envelope .
the presence of the common envelope leads to the absence of shock interaction between the stream of matter flowing from the inner lagrange point and the gas in the accretion disk .
the stream is deflected by the gas in the common envelope and approaches the disk tangentially , so that it does not cause any shock perturbation ( `` hot spot '' ) on the disk . at the same time
, the interaction of the stream with the common envelope leads to the formation of an extended shock wave along the edge of the stream . the observational manifestation of this shock is estimated to be equivalent to that of a hot spot in the disk .
the calculated accretion disk parameters are presented .
|
a topological phase of matter with a symmetry @xmath0 is an equivalence class of gapped quantum lattice systems with a symmetry @xmath0 .
one can study either ground states or hamiltonians . for classification purposes , it is the same@xcite . in terms of ground states ,
the equivalence relations are of two kinds : tensoring with a product state , eg .
the ground state of a trivial paramagnet ( this adds new degrees of freedom ) , and local unitary transformations of the ground state commuting with @xmath0 .
topological phases of matter can be `` stacked together '' , by taking tensor product of hilbert spaces , hamiltonians , and ground states , and taking the @xmath0 symmetry of the stack to be the diagonal one .
this operation makes the set of topological phases with a symmetry @xmath0 into a commutative unital semigroup , a set with an associative and commutative binary operation and a neutral element , but not necessarily with an inverse for every element .
a short - range - entangled ( sre ) topological phase with symmetry @xmath0 is a topological phase with symmetry @xmath0 which has an inverse .
sre topological phases in @xmath1 spatial dimensions with symmetry @xmath0 form an abelian group .
sre topological phases are interesting in part because they are more manageable than general topological phases but still retain many interesting topological properties .
fermionic sre topological phases ( fsre phases ) are particularly rich .
free fsre phases , ie .
equivalence classes of quadratic hamiltonians of hopping fermions , have been classified in all spatial dimensions @xcite . in the interacting case
, there is a fairly complete picture of fsre phases in dimensions @xmath2 and @xmath3 @xcite ( the abelian group structure on the set of 1d fsres was recently studied in @xcite ) .
gu and wen also constructed a large class of fsre phases in all dimensions using the `` supercohomology '' approach @xcite .
but it is clear by now that this construction does not produce all possible fsre phases .
it was conjectured in @xcite that fsre phases can be classified using spin - cobordism of the classifying space of @xmath0 .
this conjecture is supported by a recent mathematical result that relates ( spin ) cobordisms with unitary invertible ( spin ) tqft @xcite .
the drawback of this approach is that the relation between tqfts and topological phases of matter is not well understood .
in particular , given a spin - cobordism class it is not clear in general how to construct a lattice fermionic system which belongs to the corresponding fsre phase .
neither is it clear which physical properties distinguish systems corresponding to different cobordism classes . in dimensions @xmath2 and @xmath3
these problems have been solved , at least in the case when @xmath0 acts unitarily .
for example , 1d fsre phases are classified by a triple @xcite : @xmath4 to each such triple one can assign a concrete integrable lattice hamiltonian as well as a spin - cobordism class of @xmath5 @xcite .
the physical significance of each member of the triple is understood ( they describe properties of the edge modes of the system ) .
similar results for 2d fsre systems have been obtained in @xcite .
the main goal of this paper is to extend some of these results to dimension @xmath6 .
our approach is based on the idea of bosonization / fermionization .
it is a well - known result that every lattice fermionic system in one spatial dimension corresponds to a lattice bosonic system with a global @xmath7 symmetry .
this is usually explained using the jordan - wigner transformation . in @xcite
it was argued that one can obtain fermionic systems in @xmath1 spatial dimensions starting from bosonic systems with a global @xmath8-form @xmath7 symmetry generated by a fermionic quasiparticle .
more precisely , the @xmath7 symmetry must have a particular t hooft anomaly which is trivialized when the spin structure is introduced .
the fermionic system can be recovered by gauging the @xmath7 symmetry , i.e. by coupling the bosonic system to a dynamical @xmath1-form gauge field valued in @xmath7 as well as a simple fermionic system . in 1d and 2d
every fsre phase arises in this way from a suitable bosonic system , and it is natural to conjecture that this is also true in higher dimensions .
in fact , we will argue that in 3d a new phenomenon occurs which makes the bosonization approach a bit more involved .
namely , the fermion parity operator @xmath9 can get contributions from both particle and string states , and the string contribution can not be written in a local way .
microscopically , these strings carry a 1d fsre phase ( the kitaev chain @xcite ) , which may have a fermionic ground state depending on how it is embedded into space .
we call these objects kitaev strings . from the mathematical viewpoint
, this means that the bosonic shadow has both 2-form and 1-form global @xmath7 symmetries , with a nontrivial `` interaction '' between them , and both need to be gauged in order to get an fsre phases .
we propose a generalization of the gu - wen supercohomology which accounts for this new phenomenon .
we also write down a concrete 3d lattice bosonic model which , when coupled to a background @xmath0 gauge field , gives the bosonic shadow of a general 3d fsre phase .
this theory is interesting in its own right as a very simple non - abelian 3 + 1d topological order , analogous to the ising anyons in 2 + 1d .
the organization of the paper is as follows . in sections 2 and 3 we recall topological bosonization in one and two spatial dimensions and how it is used to classify fsre phases .
we also interpret the classification in terms of properties of domain walls and their junctions in a broken symmetry phase . in section 4
we describe our proposal for 3d bosonization and propose a classification of 3d fsre phases . in section 5
we write down a 3d bosonic model which can serve as a bosonic shadow for 3d fsre phases . in section 6
we briefly discuss a new class of 3d phases which seem to be neither bosonic nor fermionic , although they contain `` fermionic strings '' . in section 7
we summarize our results and discuss possible higher - dimensional generalizations
. we will be interested in models where the fermion number is conserved modulo @xmath3 .
accordingly , the eigenvalues of the fermion number operator @xmath10 are defined only as elements of @xmath11 .
the eigenvalues of the fermion parity operator @xmath9 are @xmath12 .
we will freely use simplicial cochains and operations on them , including steenrod squares .
some properties of steenrod squares and stiefel - whitney classes are recalled in appendix [ steenrodreview ] .
a. k. would like to thank greg brumfiel , john morgan and anibal medina for communicating to him some of their unpublished results .
r. t. would like to thank dominic williamson , dave aasen , and ethan lake for many enlightening discussions .
this paper was supported in part by the u.s .
department of energy , office of science , office of high energy physics , under award number de - sc0011632 .
the work of a. k. was also supported by the simons investigator award .
r. t. is supported by an nsf grfp grant .
a. k. and r. t. are grateful to kitp , santa barbara , for hospitality during the initial stages of this project .
it is well - known that 1d fermionic systems can be mapped to bosonic systems with @xmath7 symmetry by means of the jordan - wigner transformation@xcite .
this tranformation is not an equivalence , as it does not preserve certain physical properties .
for example , it maps the majorana chain @xcite ( a discretization of the massive majorana fermion ) to the quantum ising chain . depending on the values of the parameters ,
the latter model can have a doubly - degenerate ground state . on the other hand ,
the ground state of the majorana chain is always unique .
it is best to think about the jw transformation as gauging the fermion parity " .
this becomes more obvious when one considers the bosonization transformation on a circle@xcite .
while the massive fermion on a circle requires a spin structure , the corresponding bosonic system does not . on the other hand , since it has a @xmath7 symmetry , it can be coupled to a @xmath7 gauge field . to obtain the bosonic hilbert space from the fermionic hilbert space
one has to sum over spin structures " .
conversely , the fermionic hilbert space can be obtained from the bosonic one by `` summing over @xmath7 gauge fields . ''
the scare quotes indicate that certain topological terms are important in these sums .
one can describe the connection between bosonic and fermionic hilbert spaces on a circle in complete generality .
the bosonic hilbert space has a @xmath7-untwisted sector and a @xmath7-twisted sector , which we denote @xmath13 and @xmath14 .
each of these can be further decomposed into eigenspaces of the @xmath7 global symmetry : @xmath15 on the other hand , the fermionic hilbert space has an ns sector and a r sector , which we denote @xmath16 and @xmath17 , and each of them decomposes into eigenspaces of the fermion parity @xmath18 : @xmath19 these decompositions are related as follows : @xmath20 in particular , the @xmath21 component of the total fermionic hilbert space , @xmath22 , is the untwisted sector of the bosonic theory @xmath23 , while the @xmath24 component @xmath25 , is the twisted sector of the bosonic theory @xmath26 .
these relations can be interpreted as follows : to get the fermionic hilbert space from the bosonic one , one gauges the @xmath7 symmetry and identifies the holonomy of the @xmath7 gauge field @xmath27 as the fermion parity @xmath18 . for each value of @xmath28
, one needs to project to a particular value of the @xmath7-charge to select either the ns or r sector states : more precisely , if we label the spin structures by @xmath29 so that @xmath30 corresponds to the ns sector and @xmath31 corresponds to the r sector , then the generator of @xmath7 acts in the sector with the holonomy @xmath27 with the weight @xmath32 . note that the weight is a ( exp-)linear function both of the spin structure and the @xmath7 gauge field on a circle . as an example , consider the majorana chain @xcite and the quantum ising chain .
the quantum ising chain has a gapped phase with an unbroken @xmath7 ( paramagnet ) and a gapped phase with a spontaneously broken @xmath7 ( ferromagnet ) .
consider the limit of an infinite energy gap .
then in the unbroken phase the system has a unique ground state both for the trivial and the nontrivial @xmath7 gauge field .
on the other hand , in the broken phase , the system has two ground states with a trivial @xmath7 gauge field ( a @xmath33-even one and a @xmath7-odd one ) , and no ground states when the @xmath7 gauge field is turned on because in the limit of infinite energy gap the energy of the domain wall between the two vacua is infinite .
the majorana chain also has two phases , depending on the sign of the parameter which corresponds to the fermion mass in the continuum limit . for both signs of the mass
, there is a unique ground state for either choice of the spin structure on a circle .
the difference is that for a positive mass the ramond - sector ground state has @xmath21 , while for a negative mass it has @xmath24 .
the ground state in the ns sector has @xmath21 in both cases .
the jw transformation maps the positive - mass majorana chain to the ising chain with a spontaneously broken @xmath7 , while the negative - mass majorana chain is mapped to the ising chain with an unbroken @xmath7 .
note that the majorana chain ( for either sign of the mass ) is an fsre , but the quantum ising chain in a phase with a spontaneously broken @xmath7 is not a bosonic sre phase .
thus bosonization and fermionization do not map sre phases to sre phases .
this also applies in higher dimensions , as we will see . when considering 1d systems on a circle , it is easy to mistake a spin structure for a @xmath7 gauge field .
the distinction between them becomes clearer when we consider systems on a curved space - time with a nontrivial topology .
it will be useful to write down a relation between the partition functions of the fermionic theory and its bosonic `` shadow '' on a general riemann surface @xmath34 .
the fermionic partition function depends on a spin structure on @xmath34 , while the bosonic partition function depends on a @xmath7 gauge field @xmath35
( i.e. an element of @xmath36 ) .
a nice way to think about a spin structure on @xmath34 is as follows @xcite : every spin structure @xmath37 gives rise to a quadratic function @xmath38 , such that @xmath39 conversely , every such quadratic function corresponds to a spin structure on @xmath34 .
one says that @xmath40 is a quadratic refinement of the bilinear form @xmath41 .
note that the set of spin structures is not an abelian group ( there is no natural way to define a group operation on the set of quadratic refinements of a fixed bilinear form ) . on the other hand ,
the set of equivalence classes of @xmath7 gauge fields is an abelian group .
we note for future use that the latter group naturally acts on the set of spin structures : for all @xmath42 we let @xmath43 given this relation between spin structures and quadratic refinements , the relation between the partition functions can be written as a nonlinear discrete fourier transform : @xmath44 in our example of the ising / majorana correspondence , @xmath45 is a delta function @xmath46 setting @xmath47 in the ferromagnetic phase ( because of the infinite energy of the domain wall ) and the constant 1 in the paramagnetic phase .
so the former coincides with the constant fermionic partition function while the latter coincides with the arf invariant , which is the partition function of the kitaev chain @xcite , agreeing with what we expect from the microscopic jw transformation .
fsres in 1d ( with arbitrary interactions ) have been classified in @xcite using bosonization and matrix product states .
see also @xcite , where the same results were obtained using fermionic mps .
the result is that the set of fsre phases with a unitary symmetry @xmath0 is classified times fermion parity , rather than an extension of @xmath0 by fermion parity .
the generalization to nontrivial extensions is straightforward .
] by triples @xmath48 of group cohomology classes ( [ onedfsre ] ) .
all of these parameters can be be interpreted in terms of properties of the edge zero modes .
the parameter @xmath49 is the number modulo two of majorana zero modes at each edge of the system .
for example , the negative - mass majorana chain @xcite has @xmath50 and a single majorana zero mode at every edge .
the parameter @xmath51 tells us whether a particular element @xmath52 commutes ( @xmath53 ) or anti - commutes ( @xmath54 ) with the fermion parity @xmath9 when acting on the edge zero modes .
the parameter @xmath55 controls the projective nature of the action of @xmath0 on the edge zero modes .
if any one of these parameters is non - vanishing , the system must have nontrivial edge zero modes , and therefore the ground state on an interval is degenerate in the large - volume limit .
it is instructive , although somewhat nontrivial , to interpret these parameters without appealing to the edge zero modes @xcite .
it is helpful to introduce a nontrivial spacetime geometry and a fixed background @xmath0 gauge field .
let us imagine that the ir limit of the system is described by a unitary continuum 2d quantum field theory , then we can wick - rotate it and place it on an arbitrary riemann surface @xmath56 , perhaps with a nonempty boundary @xmath57 .
for a fermionic system , this requires choosing a spin structure on @xmath56 , which also induces a spin structure on each boundary circle in @xmath57 .
there are two spin structures on a circle : periodic ( ramond ) and anti - periodic ( neveu - schwarz ) .
they are also known as non - bounding and bounding spin structures , respectively , since the ns spin structure on a circle can be obtained by restricting the unique spin structure on a disk , while the ramond spin structure can not be so obtained .
since we are dealing with an fsre phase , the ground state on a circle is non - degenerate for either choice of the spin structure , and one can show that in the ns sector it is always parity - even @xcite .
the parameter @xmath58 tells us whether the ground state in the ramond sector is bosonic ( @xmath59 ) or fermionic ( @xmath50 ) .
one can also couple the system to a flat @xmath0 gauge field and consider the ground states on a circle with a holonomy @xmath52 ( and an arbitrary spin structure ) . for any @xmath52
there is a unique ground state ( again by the sre assumption ) .
the parameter @xmath60 tells us whether it is bosonic a fermionic for the ns spin structure ( for the ramond spin structure , the fermion parity of the ground state is shifted by @xmath58 ) . when the symmetry @xmath0 is broken , turning on a holonomy @xmath61 around the circle leads to a particle - like domain wall ; the paremeter @xmath60 tells us whether it is bosonic or fermionic .
finally , the parameter @xmath62 describes the `` s - matrix '' of the domain walls obtained when the symmetry @xmath0 is spontaneously broken . to be more precise ,
let us assume that @xmath63 .
then all domain walls are bosonic , and since the theory is trivial away from the domain walls , one should be able to compute the partition function by summing over possible domain - wall worldlines .
the parameter @xmath64 is a phase attached to a junction of domain walls labeled by @xmath65 and @xmath66 .
together , these parameters define a 2-dimensional spin cobordism class of @xmath5 via the atiyah - hirzebruch spectral sequence .
recently , it has been shown that a 2 + 1d lattice fermionic system can be obtained from a 2 + 1d bosonic system ( its bosonic `` shadow '' ) with an anomalous @xmath7 1-form symmetry .
let us remind what this means @xcite .
a parameter of a global 1-form @xmath7 symmetry is a @xmath7 gauge field , i.e. a 1-cocycle ( cech or simplicial ) with values in @xmath7 , defined up to a @xmath7 gauge transformation ( i.e. up to adding an exact 1-cocycle ) .
this symmetry is assumed to preserve the action , but can not be gauged .
that is , one can not promote the parameter @xmath67 to a general @xmath7-valued 1-cochain even at the expense of introducing a 2-form gauge field @xmath68 ( i.e. a 2-cocycle with values in @xmath7 which transforms as @xmath69 ) while maintaining gauge invariance .
the anomaly of a bosonic shadow has a very specific form : the partition function @xmath70 on a closed oriented 3-manifold @xmath71 transforms under @xmath69 by a factor @xmath72 it was shown in @xcite that one can obtain the fermionic partition function by performing a nonlinear discrete fourier transform : @xmath73\in h^2(y,{{\mathbb z}}_2 ) } z_b(b ) ( -1)^{q_{\zeta}(b)}\ ] ] here we use an observation @xcite that to every spin structure @xmath74 on a triangulated closed oriented 3-manifold @xmath71 one can associate a quadratic function @xmath75 which under @xmath69 transforms as @xmath76 the construction and properties of the function @xmath77 are discussed in appendix [ qeta ] . thanks to ( [ quadratictransf3d ] ) , the summand in ( [ fourier3d ] ) is a well - defined function on @xmath78 . unlike in 2d
, the definition of the quadratic function @xmath77 depends on additional choices : a branching structure on the triangulation .
the bilinear form on @xmath79 corresponding to the quadratic function @xmath77 is independent of @xmath74 but depends on these extra choices : @xmath80 where @xmath81 is a certain bilinear operation @xmath82 introduced by steenrod @xcite ( see appendix [ steenrodreview ] ) .
one can show that spin structures on @xmath71 are in one - to - one correspondence with quadratic refinements of this bilinear form which transform according to ( [ quadratictransf3d ] ) , see appendix [ qeta ] and@xcite .
the equation ( [ fourier3d ] ) says that the fermionic theory is obtained from the bosonic one by gauging the @xmath7 1-form symmetry .
the factor @xmath83 is needed to cancel the gauge anomaly .
it is instructive to see how the gauging works on the hamiltonian level .
consider a space - time of the form @xmath84 , where @xmath34 is a closed riemann surface .
there are two sectors in the gauged bosonic theory distinguished by the flux of the 2-form gauge field @xmath68 through @xmath34 .
the untwisted sector @xmath85 is identified with the @xmath21 sector of the fermionic hilbert space , while the twisted sector @xmath86 is identified with the @xmath24 sector of the fermionic hilbert space .
the gauge 1-form @xmath7 symmetry acts in each sector by unitary operators @xmath87 , @xmath88 . by fixing a gauge
, we can assume that @xmath67 is closed , so that each sector is acted upon by @xmath89 .
this action is projective because of the t hooft anomaly .
the corresponding 2-cocycle is computed following a standard procedure , see appendix [ descent ] and @xcite .
we get @xmath90 in particular @xmath91 . as in the 1d case ,
the sector corresponding to a particular spin structure @xmath37 on @xmath34 is obtained by decomposing the hilbert space into eigenspaces of @xmath87 , namely @xmath92 this is consistent with ( [ projcocycle3d ] ) thanks to ( [ quadratic2d ] ) . )
can be trivialized by defining @xmath93 and requiring physical states to be invariant under @xmath94 for all @xmath95 . ]
let us now recall the classification of 2d fsre phases proposed in @xcite .
they are labeled by triples @xmath96 which satisfy the equations @xmath97 the first two of these are the gu - wen equations which describe supercohomology phases .
the bosonic shadow of all these fsre phases can be taken to be the toric code equivariantized with respect to @xmath0 @xcite .
in particular , the homomorphism @xmath98 tells us which elements of @xmath0 exchange the @xmath99 and @xmath100 excitations of the toric code .
the toric code has an action @xmath101 and a global 1-form @xmath7 symmetry which acts by @xmath102 @xmath103 , @xmath104 .
one can check that this 1-form symmetry has the right t hooft anomaly .
one can interpret the data @xmath105 in physical terms .
as stated above , a nonzero @xmath106 means that the element @xmath61 acts as particle - vortex symmetry of the toric code .
this implies that an insertion of a flux @xmath61 of the background gauge field carries a majorana zero mode ( or more precisely , an odd number of majorana zero modes ) . in the symmetry - broken phase ,
this insertion becomes an endpoint of a domain wall , and thus the corresponding domain wall carries a negative - mass majorana chain . in what follows we will call a 1d defect with this property a kitaev string . let us denote by @xmath107 the domain wall corresponding to the group element @xmath61 .
note that since fusing @xmath107 and @xmath108 produces @xmath109 , and the number of majorana zero modes must be preserved modulo @xmath3 , we must have @xmath110 i.e. @xmath58 is a homomorphism .
the parameter @xmath111 is most easily interpreted if @xmath59 .
then the endpoint of each domain wall carries no fermionic zero modes , and one might as well assume that the endpoint has fermion parity zero .
but when considering networks of domain walls , we might need to assign fermion parity @xmath112 to each triple junction , where @xmath113 and @xmath114 meet .
requiring that the fermion number of the network does not change under pachner moves , one gets a constraint saying that @xmath115 is a 2-cocycle ( with values in @xmath7 ) .
equivalently , one may consider the surface of a tetrahedron , and regard each edge as a domain wall . since this network can be consistently continued into the interior of the tetrahedron ,
the fermion parity of the network must vanish .
this again gives the condition @xmath116 .
note also that since every domain wall has two ends , we can shift the fermion parity of the endpoint of @xmath107 by @xmath117 without changing the net fermion parity of the network .
this shifts @xmath115 by a coboundary : @xmath118 thus only the cohomology class of @xmath119 has a physical meaning .
when @xmath106 is non - vanishing , the situation is not very different .
the key point is that at the junction of three domain walls we have an even number of majorana zero modes , thanks to the condition ( [ sigmaishomomorphism ] ) .
they act irreducibly on a fermionic fock space , and one can imagine turning on a local interaction at the junction that lifts the degeneracy and makes one of these states the ground state .
the fermion parity of this ground state is @xmath115 .
the same arguments as above show that @xmath115 is a 2-cocycle defined up to a coboundary .
the parameter @xmath120 has the same meaning as in the bosonic case , i.e. it describes the amplitude assigned to a point - like junction of four domain wall worldsheets in space - time . to derive a constraint on it
, one needs to consider a 3-sphere triangulated into a union of four tetrahedra , pass to the dual cell complex and insert a domain wall along every 2-face of this cell complex . on the one hand ,
the amplitude muct be trivial , because such a configuration of domain walls can be created out of a trivial one .
on the other hand , one can evaluate it taking into account the fermionic statistics of the triple domain wall junctions @xcite .
the resulting constraint is the gu - wen equation @xmath121 when the parameter @xmath58 is nontrivial , some domain walls are kitaev strings and consequently carry fermion number when wrapping cycles with ramond spin structure .
note that a homologically trivial kitaev string automatically carries zero fermion number , because the spin structure induced on it by the spin structure in the ambient space is of the ns type .
therefore the contribution of the kitaev strings to the fermion number is nonlocal and depends on the homology class of the string network . to determine its form , note first that the homology class of the kitaev strings is the poicare - dual of @xmath122 , where @xmath123 is the @xmath0 gauge field on @xmath34 . assuming that the fermion number depends only on the homology class of the string
, we may assume that the kitaev string wraps a closed curve @xmath124 on @xmath34 whose homology class is dual to @xmath125 .
then the spin structure induced on @xmath124 is ramond precisely if @xmath126 .
therefore we can identify @xmath127 with the contribution of kitaev strings to the fermion number @xmath10 .
note that it is nonlocal , as expected , and conserved .
this explains why we could ignore it when identifying the fermion number with a local expresson @xmath128 : in 2d fsre phases , the particle and strings contrubutions to the fermion number are separately conserved . as in the 1 + 1d case ,
these triples define spin cobordism classes of @xmath5 via the atiyah - hirzebruch spectral sequence . in this section
we discuss ground states of a simple lattice model which is a bosonic shadow of the trivial 2d fsre phase following @xcite .
this is a warm - up for a similar discussion of 3d fsre phases in later sections .
we need a bosonic tqft which has a 1-form @xmath7 symmetry with the correct anomaly .
a bosonic tqft can be constructed from a spherical fusion category @xmath129 .
its objects can be thought of as boundary line defects for a particular boundary condition .
bulk line defects are described by objects in a modular tensor category @xmath130 , the drinfeld center of @xmath129 .
a generator of a 1-form @xmath7 symmetry is a bulk line defect and thus corresponds to an object @xmath131 with a fusion rule @xmath132 .
such an object has topological spin @xmath133 which satisfies @xmath134 .
it measures the anomaly of the 1-form @xmath7 symmetry .
since we want the anomaly to be of order @xmath3 , the topological spin must be @xmath135 , i.e. @xmath136 must be a fermion .
the simplest 2 + 1d tqft with these properties is the @xmath7 gauge theory , also known as the toric code .
the corresponding category @xmath129 is the category of @xmath7 graded vector spaces and has two irreducible objects : @xmath2 and @xmath10 , with the fusion rule @xmath137 .
one can think of the boundary line defect @xmath10 as the result of fusing @xmath136 with the boundary .
the toric code has two more irreducible line defects , @xmath99 and @xmath100 , such that @xmath138 , and @xmath139 .
the objects @xmath99 and @xmath100 are bosons ( @xmath140 ) and thus correspond to non - anomalous @xmath7 symmetries , but since they are muutually nonlocal , their bound state @xmath136 is a fermion .
let the spatial slice be a closed oriented 2d manifold @xmath34 with a chosen triangulation .
the toric code has @xmath141 linearly independent ground states on @xmath34 .
the string - net construction describes these ground states as particular linear combinations of states @xmath142 , where @xmath143 .
a 1-cocycle on a triangulated surface can be thought of more geometrically as a 1-cycle on a dual cell complex , ie .
a bunch of closed curves , a string net " .
the string - net hamiltonian is a commuting projector hamiltonian whose ground states have the property that the coefficient @xmath144 of the state @xmath145 is invariant under local rearrangements of the string - net which do not change its homology class , or dually , the cohomology class @xmath146\in h^1(m,{{\mathbb z}}_2)$ ] .
thus a general ground state is @xmath147)\vert\gamma\rangle.\ ] ] there are several natural bases in the space of ground states associated to various 1-form @xmath7 symmetries of the toric code .
the most obvious basis @xmath148\rangle=\sum_{[\gamma]=[\beta]}\vert\gamma\rangle,\quad [ \beta]\in h^1(m,{{\mathbb z}}_2)\ ] ] can be characterized by the property that @xmath149\rangle$ ] is a simultaneous eigenvector of the 1-form symmetry transformations @xmath150 the 1-form symmetry which acts by @xmath151 has simultaneous eigenvectors of the form @xmath152\rangle=\sum_\gamma ( -1)^{\int_m \beta\cup\gamma } \vert\gamma\rangle.\ ] ] these two 1-form symmetries are non - anomalous and correspond to @xmath99 and @xmath100 bulk line defects .
the `` diagonal '' 1-form symmetry which acts by @xmath153 corresponds to the bulk line defect @xmath136 .
its simultaneous eigenvectors are labeled by spin structures @xmath37 : @xmath154 the anomalous 1-form symmetry ( [ anomalousoneform ] ) acts on these states as follows : @xmath155 upon gauging the 1-form @xmath7 symmetry , the state @xmath156 gives rise to the unique ground state of the fermionic tqft on the spin manifold @xmath157 . in order to get a nontrivial fsre with symmetry @xmath0 , one has to couple the toric code to a background @xmath0 gauge field . as explained in @xcite , this leads to the most general 2d fsre with parameters @xmath105 .
it was mentioned in @xcite that one should be able to construct fermionic phases in 3d from bosonic phases with an anomalous global 2-form @xmath7 symmetry .
the anomaly is again quite special : it trivializes when a spin structure is specified .
the most concise way to describe the anomaly is to write down a 5d topological action for a 3-form @xmath7 gauge field @xmath158 whose variation is a boundary term cancelling the variation of the partition function of the anomalous theory on @xmath159 . in the present case ,
this anomaly action is @xmath160 where @xmath161 is the steenrod square .
this action is invariant under @xmath162 , @xmath163 when @xmath18 is closed .
when @xmath18 has a nonempty boundary @xmath164 , the action varies as follows : @xmath165 note that the variation vanishes when @xmath166 and @xmath164 is a spin 4-manifold .
this means that the variation of @xmath167 can be interpreted as an t hooft anomaly for a 3 + 1d bosonic phase which has a global 2-form @xmath7 symmetry on a spin 4-manifold . as usual
, the anomaly implies that the global 2-form @xmath7 symmetry acts projectively on the hilbert space of the bosonic theory associated to a compact 3-manifold @xmath71 .
the 2-cocycle on @xmath79 corresponding to this projective action is computed in appendix [ descent ] and turns out to be @xmath168 this is a symmetric bilinear form on @xmath79 , and we know from the previous section that its quadratic refinements correspond to spin structures on @xmath71 .
thus once we fixed a spin structure @xmath74 on @xmath71 , we can impose a gauss law constraint selecting the states in the fermionic hilbert space for @xmath74 : @xmath169 we also identify the fermion parity operator @xmath9 with @xmath170 .
note that the 2-cocycle ( [ twococycle3d ] ) is not invariant under @xmath171 , and neither is @xmath172 .
so the anomaly is more severe than in the 2d case . to obtain the supercohomology phases of gu and wen
, we take the bosonic shadow to be the simplest crane - yetter - kauffman - walker - wang model @xcite : @xmath173 where @xmath174 are subject to gauge symmetries @xmath175 the global 2-form @xmath7 symmetry acts by shifting @xmath176 , @xmath177 .
this transformation shifts the action by @xmath178,\ ] ] where @xmath179 is the 2nd stiefel - whitney class . if @xmath164 is a closed spin 4-manifold , then @xmath180 and the action is invariant for arbitrary @xmath181 .
alternatively , if @xmath164 is not assumed to be spin , the action is invariant only if we impose a constraint @xmath182=0 $ ] . to gauge this 2-form symmetry ,
we introduce a 3-form gauge field , i.e. a 3-cocycle @xmath183 .
we modify the action to @xmath184 the variation of @xmath185 under a gauge transformation is independent of @xmath186 and given by ( [ 5dvariation ] ) .
thus the theory has the correct t hooft anomaly to be a bosonic shadow of a fermionic theory .
we can promote this theory to a @xmath0-equivariant model by replacing @xmath187 , where @xmath188 .
this does not change the anomaly of the 2-form @xmath7 symmetry , but introduces an anomaly for @xmath0 . to simply notation ,
let us denote @xmath189 ; then @xmath190 the last term is exact and thus does not lead to anomaly ( it can be absorbed into a contact term @xmath191 in the action ) .
the first term gives the usual anomaly for the 2-form symmetry , while the second term leads to an anomaly for @xmath0 .
this anomaly can be canceled if and only if there exists a 4-cochain @xmath62 on @xmath0 with values in @xmath192 such that @xmath193 then we can cancel the anomaly by adding the term @xmath194 to the 4d action .
the equation ( [ guwen3d ] ) is the gu - wen equation for 3d supercohomology phases . before gauging the 2-form @xmath7 symmetry , the model ( [ ww ] ) has loop observables and surface observables .
the surface observable localized on a 2d submanifold @xmath195 is @xmath196 .
it is invariant under the gauge symmetry ( [ wwgaugesymmetry ] ) .
it is also charged under the global 2-form @xmath7 symmetry : @xmath197 the loop observable localized on a 1d submanifold @xmath198 is @xmath199 .
it is invariant under the gauge symmetry ( [ wwgaugesymmetry ] ) .
when @xmath200 for some 2-chain @xmath201 , this loop observable generates the 2-form gauge symmetry with a parameter @xmath202 which is poincar dual to @xmath201 . after gauging the 2-form @xmath7 symmetry ,
@xmath203 is not longer an observable , because it is not gauge - invariant .
the loop observable @xmath204 vanishes if @xmath124 is homologically nontrivial ( this follows from the fact that @xmath204 is charged under the global 1-form symmetry @xmath205 for @xmath206 ) , while for homologically trivial @xmath124 is a generator of a 2-form gauge transformation and therefore is @xmath2 when inserted into any correlator .
the conclusion is that gauging the 2-form @xmath7 symmetry leads to a theory without any nontrivial observables except the partition function , which depends on the spin structure as well as the @xmath0 gauge field @xmath123 .
this suggests that the gauged theory is a fermionic spt .
singularity , where four zippers meet . with the x axis along the blue - grey junction and the y axis along the green - grey junction
, the planes @xmath207 cut through this picture to give a movie of the f move as we vary @xmath208 through zero.,title="fig:"][a3 ] let us discuss the physical significance of the 3-cochain @xmath209 . in a fixed gauge of @xmath123 at a particular instant of time ,
we see a network of @xmath0 domain walls .
we will denote by @xmath107 the codimension-1 domain wall labeled by @xmath52 .
the spt ground state should be invariant under the reconnection of the domain wall network .
there are several kinds of defects in different dimensions corresponding to different degrees of group cocycles .
for example , there is a string - like zipper " @xmath210 where the domain walls @xmath107 , @xmath108 , and @xmath211 meet .
there is also a particle - like fusion junction @xmath212 where four zippers
meet we call the f - junction or @xmath213 singularity .
this is because if we choose a foliation of space by planes transverse to the f - junction , as we scan across we see a movie of the f move " or associator where one would apply the f symbol in tensor category theory .
these are particle - like objects , and @xmath214 can be thought of as a way of assigning fermion parity to f - junctions : some are fermionic , some are bosonic .
we will see this interpretation is natural from the atiyah - hirzebruch spectral sequence later .
it is exactly analogous to the 2 + 1d situation where the analogous 2-cocycle @xmath115 defines the fermion parity of the triple junction of domain walls .
the fact that @xmath119 is a 3-cocycle follows from the conservation of fermion number ( mod 2 ) , if we assume that the fermion number is the sum of fermion numbers of the f - junctions . indeed ,
consider a 4-simplex @xmath215 whose boundary @xmath216 consists of five 3-simplices and is homeomorphic to a 3-sphere .
the dual of this triangulation of @xmath216 contains 10 zippers ( dual to 2-simplices of @xmath216 ) meeting at four f - junctions ( dual to 3-simplices of @xmath216 ) .
if the dual of every 1-simplex of @xmath216 is a domain wall labeled with an element of @xmath0 , then the f - junctions are labeled by three elements of @xmath0 and have fermion parity determined by @xmath119 . on the other hand ,
since @xmath216 is a boundary of a 4-simplex , the net fermion number of this configuration of domain walls must vanish mod 2 .
this is equivalent to the condition @xmath116 .
alternatively , we can require the fermion parity of a network of domain walls to remain unchanged under 3d pachner moves .
this leads to the same condition on @xmath119 .
since every zipper has two ends , we can flip the fermion parity of each end without changing the fermion number of the whole network . but this changes the 3-cocycle @xmath119 . if the fermion parity of the endpoint of @xmath210 is shifted by @xmath217
, it is easy to see that the fermion numbers of the f - junctions change according to @xmath218 the class @xmath219\in h^3(bg,{{\mathbb z}}_2)$ ] is unchanged .
the ground - states of the model ( [ ww ] ) can be constructed by categorifying the string - net approach @xcite .
roughly speaking , instead of a spherical fusion category , one needs to take a spherical semi - simple monoidal 2-category .
unfortunately , there is no generally accepted definition of this object , and consequently there is no completely general method of constructing 4d tqfts .
but there is a well - understood special case , the cykww model @xcite , and the model ( [ ww ] ) belongs to this class .
the input of the cykww construction is a braided fusion category @xmath129 whose objects represent boundary defect lines for a particular boundary condition . in the present case ,
the bulk tqft has a 2-form @xmath7-symmetry , so we expect that there is an invertible line defect on the boundary which we denote @xmath136 and which satisfies @xmath132 .
it is a fermion and therefore must have topological spin @xmath135 .
this encodes the fact that the 2-form symmetry has a nontrivial anomaly .
if we assume that there are no other irreducible objects in the braided fusion category , then @xmath129 is equivalent to the category of super - vector spaces
. let @xmath71 be a closed oriented 3-manifold with a triangulation .
the cykww construction describes the ground states of the model ( [ ww ] ) as linear combinations of states @xmath220 , where @xmath221 . in the poincare - dual picture , @xmath68 is a network of @xmath136 line defects .
a 1-form gauge transformation @xmath222 corresponds to a local rearrangement of the string network .
a general state has the form @xmath223 the string - net hamiltonian is a commuting projector hamiltonian whose ground states are distinguished by the way their components transform under a re - arrangement of the string network .
namely , under @xmath222 , @xmath224 , one must have @xmath225 the explanation for this rule is the following .
the category of supervector spaces occurs as a subcategory of the category of bulk line defects for the toric code .
specifically , the line defect @xmath136 can be identified with the @xmath136 of the toric code .
each configuration of @xmath136 lines in @xmath71 can be viewed as a network of @xmath136 lines in the toric code , or equivalently as the toric code coupled to a 2-form @xmath7 gauge field @xmath68 .
this gauge field is associated to the anomalous 1-form @xmath7 symmetry of the toric code whose generator is @xmath136 .
rearranging the @xmath136 lines is equivalent to 1-form gauge transformations @xmath222 .
the rules of the 3d string - net construction tell us that the coefficients @xmath226 transform in the same way as the partition function of the toric code , i.e. ( [ anomaly3d ] ) .
this gives ( [ ctransf ] ) .
the transformation rule ( [ ctransf ] ) makes it clear that the number of linearly independent ground states is given by @xmath227 . a natural basis in the space of ground states is labeled by spin structures on @xmath71 .
namely , given a spin structure @xmath74 , we let @xmath228 after we gauge the 2-form @xmath7 symmetry , @xmath229 gives rise to a unique ground state of the 3d fsre on the spin manifold @xmath230 .
the supercohomology phases do not exhaust all possible 3d fsre phases .
there are several ways to see this .
for example , one may ask if a zipper @xmath210 ( a junction of three domain walls @xmath107 , @xmath231 and @xmath211 ) may carry the nontrivial 1d fsre , i.e. the kitaev string .
the endpoint of such a zipper would have an odd number of majorana zero modes .
such a phase would be characterized by a new parameter @xmath232 which tells us whether @xmath210 carries the kitaev string or not .
this parameter must be a 2-cocycle .
indeed , consistency requires an even number of majorana zero modes at each @xmath213 singularity , which is equivalent to the 2-cocycle condition on @xmath58 .
there is an ambiguity in the definition of @xmath233 .
the zipper is a place where three domain walls meet .
we can attach to the boundary of the domain wall @xmath107 a closed kitaev string ; this does not affect any observables , like degeneracies and fermion parities ( because the boundary of every domain wall is closed and can be contracted to a point ) , but it shifts the 2-cocycle @xmath233 by a coboundary .
which is easily seen to be a homomorphism from @xmath0 to @xmath33 . since we assumed that @xmath0 is finite , this parameter vanishes .
but this is an interesting possibility if @xmath0 is infinite and should lead to a new class of fermionic spt phases . ]
when @xmath58 is nonvanishing , the constraint on the 3-cochain @xmath119 is modified . to see how this comes about ,
let us try to guess the contibution @xmath234 of kitaev strings to the fermion number .
it is clear that such a contribution can be present whenever @xmath235\in h^2(y,{{\mathbb z}}_2)$ ] is nonzero , because this means that there are kitaev strings wrapping noncontractible loops on @xmath71 .
in fact , @xmath235 $ ] is the poincare - dual of the homology class of the kitaev strings .
@xmath234 must depend linearly on the spin structure on @xmath71 .
indeed , shifting the spin structure @xmath74 by a 1-cocycle @xmath35 should shift by one the fermion number of a kitaev string wrapping a curve @xmath124 if and only if @xmath236 ( since this is when the spin structure induced on @xmath124 by the ambient spin structure @xmath74 is flipped by the shift @xmath237 ) .
hence we expect @xmath238 the quadratic function @xmath239 depends on @xmath74 as expected , provided we identify @xmath240 .
we therefore propose that @xmath241 an important property of @xmath242 is that it is not invariant under replacing the 2-cocycle @xmath125 with a cohomologous one . in the language of kitaev strings
, this means that @xmath234 changes when kitaev strings are reconnected .
since the net fermion number must be conserved , we propose that whenever the 2-cocycle @xmath58 is shifted by @xmath243 , the 3-cocycle @xmath119 determining the @xmath244 contribution to the fermion number @xmath10 shifts : @xmath245 then this compensates for the transformation rules of @xmath246 so @xmath247 is gauge - invariant .
the constraint @xmath116 is not invariant under the shift ( [ rhoshift ] ) , which is not surprising , since the fermion number is now carried both by the junctions @xmath212 and the zippers @xmath210 . but
a modified constraint @xmath248 is invariant under the shift ( [ rhoshift ] ) accompanied by @xmath249 .
we can reach the same conclusion assuming the relation between 3d fsre phases and spin - cobordism .
the atiyah - hirzebruch spectral sequence converging to the spin - cobordism of @xmath5 indicates that there should be three parameters : @xmath250 , @xmath251 and @xmath252 .
these parameters must satisfy constraints which at linearized level are simply @xmath253 , but have corrections which are encoded in the differentials of the spectral sequence .
the spectral sequence immediately implies that the constraint on @xmath58 is not modified by the differentials , i.e. @xmath254 , in agreement with the physical argument above , but that other constraints are modified .
the 1st differential in the spectral sequence is known to be the steenrod square @xmath255 , suggesting that the equation for @xmath119 is modified to ( [ rhosigmaeq ] ) .
the equation @xmath256 is modified at leading order as well , to the gu - wen equation @xmath257 , but it must receive higher - order modifications as well , in order to be consistent with ( [ rhosigmaeq ] ) .
it is shown in appendix [ eanomalies ] that there is an essentially unique modification of the gu - wen equation consistent with ( [ rhosigmaeq ] ) .
we thus propose that 3d fsre phases are classified by solutions of the equations @xmath258 where dots denote terms which depend only on @xmath58 which are required to make the r.h.s . of the first equation closed mod integers .
we give an explicit formula for @xmath259 in the appendix .
there are also several nontrivial identifications on the set of solutions .
the abelian group structure is also highly nontrivial , since the equations appear nonlinear @xcite .
suffice it to say that the space of solutions has an obvious subgroup corresponding to solutions of the form @xmath260 , where @xmath261 .
this subgroup consists of bosonic sre phases . taking a quotient by this subgroup leads to a more manageable object ( the group of 3d fsre phases modulo bosonic sre phases ) which consists of equivalence classes of pairs @xmath262 satisfying @xmath263 @xmath254 .
the equivalence relation arises from a gauge symmetry @xmath264 the abelian group structure on these equivalence classes is easily guessed : @xmath265 indeed , since @xmath266 the r.h.s .
( [ grouplaw3dfsre ] ) satisfies the equation ( [ rhosigmaeq ] ) provided the l.h.s .
it is straightforward to verify the group axioms .
let us interpret the above proposal for the classification of 3d fsre phases in terms of their bosonic shadows .
it was argued in @xcite that one can construct a 3d fermionic system from a 3d bosonic system with a global @xmath7 2-form symmetry provided this 2-form symmetry has a suitable t hooft anomaly
. a natural generalization of this construction is to combine it with gauging some other symmetries of the bosonic system . now , suppose the symmetry of the bosonic system is not simply a product of the 2-form @xmath7 symmetry and other symmetries , but a more general structure .
specifically , since the general 3d fsre phases are supposed to contain both a condensate of fermionic particles and a condensate of kitaev strings , we are led to consider bosonic shadows with both a 2-form @xmath7 symmetry and a 1-form @xmath7 symmetry .
particles will be associated with generators of the 2-form symmetry , while strings will be associated with generators of the 1-form symmetry . in general , when a field theory has 0-form , 1-form and 2-form symmetries , the whole symmetry structure is described by a 3-group .
a general 3-group is quite a complicated object , but it simplifies when we ignore 0-form symmetries . in that case , the 3-group is characterized by its 1-form symmetry group @xmath267 , its 2-form symmetry group @xmath268 , and a postnikov class taking values in @xmath269 . in the present case @xmath270 , and @xmath271 , so there is only one nontrivial possibility for the postnikov class .
if the postnikov class vanishes , the 3-group is simply a product of 1-form and 2-form symmetries .
if it is nontrivial , the 2-form gauge field @xmath68 is still closed , while the 3-form gauge field @xmath272 satisfies @xmath273 note the similarity with eq .
( [ rhosigmaeq ] ) .
the modified bianchi identity ( [ threegroup ] ) gives rise to a modified group law for global symmetry transformations . to derive the group law
, we assume that 2-form symmetry transformations leave @xmath68 invariant and shift @xmath272 : @xmath274 while 1-form symmetry transformations shift @xmath68 : @xmath275 then ( [ threegroup ] ) requires @xmath272 to transform as follows under 1-form gauge symmetry : @xmath276 now consider the effect of two 1-form symmetry transformations with parameters @xmath67 and @xmath277 on the configuration @xmath278 .
we get @xmath279 the first equation shows that this is equivalent to a 3-group symmetry transformation with a 1-form symmetry transformation @xmath280 and an undetermined 2-form symmetry transformation with a parameter @xmath281 .
the second equation then implies that @xmath282 specializing to closed @xmath181 and @xmath67 , we conclude that the group law for global 3-group symmetry transformations is @xmath283 consider now coupling the bosonic theory to a @xmath0 gauge field @xmath123 by letting @xmath284 and @xmath240 for some @xmath251 and @xmath252 .
since @xmath68 must be closed , @xmath58 must be a 2-cocycle .
since @xmath272 satisfies ( [ threegroup ] ) , we must subject @xmath119 to ( [ rhosigmaeq ] ) . to get the 3d fsre
, we gauge the 3-group symmetry , while keeping @xmath123 fixed .
to ensure gauge - invariance with respect to @xmath0 gauge transformations , we need to impose further constraints on the data @xmath119 and @xmath58 , like the first equation in ( [ genguwen3d ] ) .
thus our proposal for 3d bosonization can be formulated as follows : every fermionic theory has a bosonic shadow with a global 3-group symmetry as above ( we will denote this 3-group @xmath285 ) and an t hooft anomaly @xmath286 ( see appendix f for the definition of the latter ) . in particular , we propose that every 3d fsre can be constructed in this way .
this construction is more general than that proposed in @xcite . to see this , note that the 2-form @xmath7 symmetry is a proper subgroup of the 3-group symmetry , so we are free to gauge it first and get a 3d fermionic phase as in @xcite .
but this fermionic phase is not an fsre yet : it has nontrivial observables charged under the global 1-form @xmath7 symmetry ( this symmetry is what remains of the 3-group symmetry after we gauge the 2-form symmetry ) . to get an fsre we must also gauge this 1-form symmetry .
the order of the steps in this two - step procedure can not be reversed , since the 1-form symmetry is not a subgroup of the 3-group symmetry , and can not be gauged without gauging the whole 3-group .
the description of the 3-group gauging as a two - step process makes it intuitively clear that the resulting phase is a fermionic phase , since the spin structure is introduced already at the first step .
but it is not clear that a further geometric structure is not needed at the second step .
the question boils down to computing possible anomalies for a 1-form @xmath7 symmetry in a fermionic theory , taking into account that the 2-form gauge field @xmath68 satisfies the constraint @xmath287=0 $ ] .
it is shown in the appendix that no anomaly is possible , and thus the 3-group symmetry with the above anomaly can always be gauged on a spin 4-manifold .
the goal of this section is to construct a 3 + 1d tqft which is a 3 + 1d analogue of the ising tqft in 2 + 1d , and has global 3-group symmetry @xmath285 .
physically we imagine a gapped superconductor with fermionic charges and vortex lines which terminate at majorana zero modes on the boundary .
since the ising category describes the behavior of majorana zero modes , it will also describe the behavior of these charges and vortex strings .
this tqft is nonabelian , so we will need an algebraic approach to construct it .
this approach is a 4d analog of the turaev - viro construction and takes a monoidal 2-category as an input @xcite . from the physical viewpoint ,
this monoidal 2-category describes boundary defects for a particular topological boundary condition .
since the 4d tqft has both 2-form and 1-form @xmath7 symmetries , it contains a codimension-3 defect ( the generator of the 2-form @xmath7 symmetry ) and a codimension-2 defect ( the generator of the 1-form @xmath7 symmetry ) .
let us denote by @xmath136 the fusion of the codimension-3 defect with the boundary .
it is a line defect on the boundary , or equivalently a line defect on the `` transparent '' surface defect .
algebraically , the ` transparent' defect is the identity object @xmath285 of the monoidal 2-category , and thus @xmath288 .
the fusion of the codimension-2 defect with the boundary gives us another object which we denote @xmath289 .
we have the fusion algebra @xmath290 we postulate that there are no further indecomposable defects in the 4d tqft , and thus every object in the monoidal 2-category is a direct sum of several copies of @xmath285 and @xmath289 ( this is a natural assumption since gauging both 2-form and 1-form symmetries should lead to a theory with no nontrivial observables ) .
next we need to describe morphism categories .
@xmath291 is a braided fusion category , and by assumption it is generated by @xmath136 and the identity object @xmath2 .
the @xmath7 fusion rule @xmath132 means that @xmath291 is equivalent to the category of @xmath7-graded vector spaces as a fusion category .
there are two braided structures on it : one corresponds to the usual tensor product , and the other one to the supertensor product .
they correspond to two possible anomalies for the 3-form symmetry : the trivial one and the one with the anomaly action @xmath292 .
we need the latter option , so that @xmath291 is equivalent to the category of supervector space as a braided fusion category . assuming that @xmath289 is its own dual
, we can also compute @xmath293 : @xmath294 finally , we need to describe the categories @xmath295 and @xmath296 .
we postulate that both @xmath296 and @xmath295 are non - empty and each of them has a single irreducible object which we denote @xmath58 .
this means that the surface defect @xmath289 can terminate on the boundary .
nevertheless , @xmath289 is not equivalent to @xmath285 , because @xmath58 is not invertible .
we postulate the simplest non - invertible fusion rule : @xmath297 we also necessarily have @xmath298 because @xmath136 is invertible . to complete the construction of the monoidal 2-category we need to specify all the associator morphisms and the pentagonator 2-morphisms @xcite .
this is facilitated by the fact that the monoidal 2-category we are constructing has a very special form : its data are equivalent to those of a braided @xmath7-crossed braided category @xcite .
this is a @xmath7-graded fusion category @xmath299 with a compatible @xmath7-action and additional data which generalizes braiding and reduces to it when the @xmath7-action is trivial . in our case , @xmath300 , and @xmath301 , and
the @xmath7 action is trivial .
thus @xmath129 is an ising braided fusion category ( and therefore is a braided @xmath7-crossed category ) .
all possible braided fusion structures on an ising category are known , and it turns out there are eight inequivalent ones , naturally labeled by a complex number @xmath302 such that @xmath303 . to any braided @xmath7-crossed category one can associate a 4d tqft using a generalization of the cykww construction @xcite .
we will call the 4d tqft obtained by taking the braided ising category as an input for this construction a 2-ising model .
we will show that the 3-group symmetry generated by @xmath289 and @xmath136 is isomorphic to @xmath285 and has the correct anomaly to be a bosonic shadow .
we propose that the bosonic shadows of all 3d fsres can be obtained by taking the 2-ising model and coupling it to a background @xmath0 gauge field while keeping the anomalies intact .
below we provide some evidence for this . in this section
we describe the state sum for the 2-ising model .
the state sum is a sum over colorings of a triangulation of @xmath164 with a fixed branching structure .
a coloring is an assignment of ( simple ) objects to edges , ( simple ) morphisms to triangles , and ( simple ) 2-morhisms to tetrahedra .
the weight of each coloring is a product over the 15j symbol in each 4-simplex .
the partition function on @xmath164 is a sum of weights over all colorings .
let us spell out what this means for 2-ising .
* edges are labeled either @xmath304 or @xmath305 .
* at a triangle , we assign a morphism in the fusion space ( a category ) @xmath306 where we have labeled the triangle according to the chosen branching structure . in particular , we have either two @xmath289 s , in which case the only morphism in the fusion space is @xmath58 ; or no @xmath289 s , in which case the morphism may be either 1 or @xmath136 .
* at a tetrahedron , it is useful to imagine the dual picture , shown in fig 1 , where six sheets are meeting , with a 1 , @xmath136 , or @xmath58 on each of four fusion junctions which meet at a point in the center . at this point
, we need to have something gluing together the fusion junctions .
the rules for this is precisely the same as in the usual ising category .
that is , we can forget the sheets and just think of this as a junction of 1 , @xmath136 , and @xmath58 lines .
choosing a resolution of the 4-valent vertex into two 3-valent vertices defines a basis for this fusion space .
* the coloring around a 4-simplex is a collection of @xmath289-sheets and @xmath136 and @xmath58 lines around its boundary , a 3-sphere .
the branching structure defines a framing of this 3-sphere and we can use the rules of the ising category @xcite to evaluate it to a number .
this defines the `` 15j '' symbol of @xcite and is the weight of the coloring in the state sum .
see fig 2 .
+ [ 4simplex ] + by line objects . representing triangles ,
these edges are labeled by triples of vertices , and there are 5 choose 3 of those .
there are some extra edges where we have resolved 4-way intersections to make the graph trivalent .
these are labeled by tetrads of vertices and coincide with the 5 choose 4 tetrahedra of the dual 4-simplex .
the state sum gives us a labeling of these edges by 1 , @xmath136 , and @xmath58 , and the rules of the ising braided fusion category of @xcite gives us a way to evaluate this picture to a number .
this defines the 15j symbol.,title="fig:",width=302][flat4simplex ] it is useful to encode the state sum as a sum over cochains .
we define the following cochains in spacetime @xmath164 : * @xmath307 is poincar dual to the @xmath289 worldvolume .
* it follows from the fusion rules that @xmath308 is poincar dual to the @xmath58 worldsheet .
* @xmath309 is poincar dual to the @xmath136 worldsheet . *
it follows from the rules of the ising category that @xmath310 this last point deserves some elaboration .
we can imagine each configuration in the state sum on @xmath164 as a movie of fluctuating @xmath58 and @xmath136 lines and @xmath289 surfaces which evolve according to the local moves of the usual ising category , except for the @xmath289 surfaces making the @xmath58 worldsheet always a boundary and inducing the into - the - page framings for the evaluation of the ising @xmath311 and @xmath10 matrices . on the boundary of a 4-ball in @xmath164
we see a snapshot of the action . in this snapshot
, we may have two @xmath58 lines in a hopf - link formation with @xmath289 surfaces defining the into - the - page framing as shown in fig .
3 . according to the rules of the ising category ( see e.g. @xcite ) , this configuration can only be filled into the 4-ball if those @xmath58 lines have a @xmath136 connecting them .
for this rule to be insured by the local dynamics of the 2-ising hamiltonian , the term which creates small discs of @xmath289 surface must create @xmath136 lines along the intersections of @xmath289 surfaces .
there are also terms which create small loops of @xmath136 line , but these can not move the endpoints of the @xmath136 lines , which will be where the @xmath289 surface intersects the @xmath58 line . from this
follows the equation ( [ hopfconstraint ] ) .
[ hopf ] anyons ( black circles ) in a hopf link formation .
the @xmath58 s are the boundary of the @xmath289 surface ( orange discs ) . where the @xmath58 s intersect the orange disc ( red stars ) , we have a @xmath136 anyon ( wavy black curve ) being born.,width=302 ]
the equation ( [ hopfconstraint ] ) implies that the 2-ising model carries an action of the symmetry 3-group @xmath285 . to see this , note that the global 2-form symmetry acts by @xmath312 , @xmath313 while leaving @xmath314 unchanged , where @xmath71 is a spatial 3-manifold .
the 1-form @xmath7 symmetry shifts @xmath315 , @xmath316 , and to be consistent with ( [ hopfconstraint ] ) one must also transform @xmath317 : @xmath318 .
a general symmetry transformation is parameterized by a pair @xmath319 and acts as follows : @xmath320 performing two consecutive transformation we get the group law eq .
( [ egrouplaw ] ) : @xmath321 thus the 2-ising model is acted upon by the 3-group @xmath285 . as explained in section 4.5 , with the proper anomaly such a symmetry is `` fermionic '' in that we can gauge it by introducing a spin structure .
the relation ( [ hopfconstraint ] ) has other interesting consequences . to evaluate the right hand side on a tetrahedron
, we need to order the vertices 0 , 1 , 2 , 3 and compute @xmath322 this quantity depends on the choice of ordering , which we take to be defined by a branching structure on @xmath164 .
one way to understand this is that @xmath323 computes the mod 2 self - linking number of the @xmath58 curves with respect to the framing defined by the @xmath289 surfaces .
that is , it equals the mod 2 linking number of the @xmath58 curve and the curve obtained from @xmath58 by displacing it a small distance into the piece of @xmath289 surface that bounds it .
the integral of @xmath324 is counting crossings between these two curves , and of course where the crossings are depends on the local framing of space .
the @xmath136 lines are line defects in the 4d theory whose endpoints represent the fundamental fermion .
the equation ( [ hopfconstraint ] ) then says that the total fermion number of the state is the self - linking of the @xmath58 loops framed by @xmath289 .
the configurations which appear in the 2-ising state sum all have even net fermion number , but the number of points where the @xmath136 lines are attached depends crucially on the framing . for example , we can create a hopf link of @xmath58 loops without any @xmath136 lines by having an @xmath289 surface which is a twice - twisted ribbon .
4.-matrix because the framing of @xmath58 induced by the @xmath289 surface does not extend to any framing of @xmath325 .
all of the ising category numbers are computed in 2-ising by choosing the @xmath289 surfaces so that they induce into - the - page framings and then computing the wavefunction overlap with the empty picture . ]
this is a new ingredient for topological order in 3 + 1d .
quasiparticles can only be the boundary of string operators in one way , but 2-ising illustrates how the statistics of a quasistring depends on how it is framed by its bounding surface operator .
[ hopf2 ] anyons ( black circles ) in a hopf link formation given by the boundary of a twice - twisted ribbon of @xmath289 surface ( orange skeleton ) . with this configuration of the @xmath289 surface ,
the self - linking of each component is even , so there is no need for a @xmath136 line connecting them .
this contrasts with the into - the - page framed hopf link we drew above , where the framing induced by the @xmath289 surface has odd self - linking in each component , so the two components are fermionic and there must be a @xmath136 line connecting them.,width=302 ] when we gauge the @xmath285 symmetry , the @xmath289 surfaces will proliferate and the @xmath58 loops will lose their framings .
their density will be measured by a @xmath7 2-form @xmath68 which is poincar dual to the worldsheet ( its integral over a surface counts the number of @xmath58 strings piercing it ) . in order for the fermion number @xmath326 of the loops to be well - defined ,
one needs some geometric input that stands in for the framing of the @xmath58 loops . from what we have discussed so far
, the function has to satisfy @xmath327 which is a special case of eq .
( [ rhoshift ] ) replacing @xmath67 by @xmath328 .
it is possible to achieve this by framing all of space since this frames all curves so that their mod 2 self - linking numbers are well - defined , but this is not very physical and too restrictive for our goals . as discussed in section 4.4 , we can define such an @xmath234 given a spin structure on spacetime .
this is very physical , since we wish to describe fermionic systems by gauging the @xmath285 symmetry .
we conclude that the fermion number of the @xmath58 loops will depends on the spin structure @xmath74 .
let us write the string fermion number @xmath329 .
we can guess the dependence of @xmath329 on the spin structure @xmath74 by thinking about the spin structure induced by a framing . as discussed in @xcite
, one can think of a spin structure in 3d as a mod 2 invariant of any framed curve which increments by 1 when the framing of the curve is twisted .
it also flips by @xmath330 when the spin structure @xmath74 is shifted by a @xmath7 1-cocycle @xmath35 to @xmath331 .
thus , changing the spin structure is equivalent to twisting certain framings .
this also changes the mod 2 self - linking by the same amount , so we find that the fermion number is linear in the spin structure : @xmath332 as discussed in appendix [ qeta ] , this requirement essentially fixes @xmath333 to be the function @xmath239 . gauging the @xmath285 symmetry
also frees the fundamental fermion , the endpoint of the @xmath136 lines , turning the 3-coboundary @xmath334 into a 3-cochain @xmath272 .
it follows from eq .
( [ hopfconstraint ] ) that @xmath335 this reflects the non - trivial postnikov class of the 3-group @xmath285 .
now we want to enlarge 2-ising to a theory with a global @xmath0 symmetry .
we do this by extending @xmath0 , considered as a monoidal 2-category with trivial morphisms , by 2-ising . the data for this will consist of a group 2-cochain @xmath336 , a group 3-cochain @xmath337 , and a group 4-cochain @xmath338 , satisfying some conditions we presently derive .
physically , the presence of a global symmetry @xmath0 means that for every @xmath52 there is a codimension-1 invertible defect .
their fusion obeys the group law of @xmath0 .
if the symmetry is unbroken on the boundary , each such defect gives rise to an invertible surface defect on the boundary which we denote @xmath339 .
fusing each of them with @xmath289 ( the generator of the 1-form symmetry ) , we get another surface defect @xmath340 .
obviously , @xmath341 .
although the fusion of bulk domain walls obeys the group law of @xmath0 , the fusion of @xmath339 and @xmath340 is governed by the group law of an extension of @xmath0 by @xmath7 .
this happens because the termination of a bulk defect is not canonically defined , so for a given @xmath61 one can always swap @xmath339 and @xmath340 . if @xmath342 is a 2-cocycle describing this extension , then the fusion rule is @xmath343 associativity of the fusion algebra is equivalent to @xmath344 geometrically , this means the following : a zipper @xmath345 is part of the boundary of @xmath289 surfaces if and only if @xmath346 .
thus , in the gauge where @xmath58-lines are absent , we must have @xmath347 where @xmath348 represents the configurations of @xmath0 labels on objects on edges in the state sum . in the bulk , where an @xmath289 surface can not terminate
, we postulate that ( [ gsigmaconstraint ] ) holds without any restrictions .
next we interpret the 3-cochain @xmath349 .
whenever this 3-cochain is nonzero , the @xmath213 singularities where four zippers meet are sources of @xmath136 lines .
thus the 2-cochain @xmath317 representing @xmath136-lines must satisfy @xmath350 the second term is required to ensure that the constraint is invariant under the action of the @xmath285 symmetry ( [ epsilonaction ] ) .
now let us integrate ( [ rhoconstraint ] ) over @xmath71 , assuming that @xmath351 .
taking into account that all cochains in ( [ rhoconstraint ] ) are restrictions of cochains on @xmath164 and using the stockes theorem and ( [ gsigmaconstraint ] ) , we get @xmath352 since @xmath123 and @xmath164 are arbitrary , we must have @xmath353 with the constraints in eqs . [ nu2closed ] and
[ nupostnikov ] , the pair @xmath354 describes precisely a map @xmath355 , where @xmath356 is the classifying space for the 3-group @xmath285 .
thus a @xmath0-crossed 2-ising model is an equivariantization of the 2-ising model .
one more constraint on @xmath119 and @xmath58 should follow from the topological invariance of the @xmath0-crossed 2-ising model . in principle , it can be obtained by evaluating the partition function on a boundary of a 5-ball and requiring it to be @xmath2 for an arbitrary gauge field @xmath123 on the 5-ball and arbitrary @xmath314 and @xmath317 satisfying all the constraints . for a trivial gauge field
@xmath123 , this is ensured by the properties of the 2-ising 15j symbol . instead of performing this rather formidable computation for general @xmath123 , we can take a short - cut and require the symmetry @xmath0 to be non - anomalous . since we embedded @xmath0 into @xmath285 by letting @xmath284 and @xmath240
, this means that the cocycle @xmath357 must be exact . here
the choice of the sign in @xmath259 depends on the braiding structure of the 2-ising theory we use .
thus there must exist a 4-cochain @xmath358 such that @xmath359 we propose that the general 3d fsre with symmetry @xmath0 can be obtained by gauging the 3-group symmetry @xmath285 of the @xmath0-crossed 2-ising model . when @xmath59 , the sector of the @xmath0-crossed 2-ising containing @xmath289 surfaces and @xmath58-lines decouples , and we can restrict our attention to the networks with @xmath360 .
the remaining constraints simplify to @xmath361 thus for a fixed network of @xmath0 domain walls , we sum over all networks of @xmath136 lines satisfying the following condition : each @xmath213 singularity @xmath212 with @xmath362 is a source for a @xmath136-line , and @xmath136-lines can not end anywhere else . with non - zero @xmath119 but @xmath363 .
the tetrahedra where @xmath364 have a non - conservation of @xmath136 lines , indicated by red curves coming out of the resolved 4-way junctions ( @xmath213 singularities ) dual to the tetrahedra .
the @xmath136 lines go and join `` the condensate '' , represented by a red ball which may absorb any number of @xmath136 lines . in evaluating the diagram according to the rules of the ising category ,
we get contributions from crossings .
the red with black give a sign contribution of @xmath135 to power @xmath365 .
the black with black crossing gives a contribution of @xmath135 to power @xmath366.,title="fig:",width=302][flat4simplex ] let us study the 15j symbol of this 2-category .
this is the quantity we will multiply 4-simplex - by-4-simplex along a triangulation of @xmath164 to obtain the partition function of the @xmath0-crossed 2-ising theory . as we see from fig .
5 , the 15j symbol is the exponential of @xmath367 this is analogous to eq ( 3.12 ) in @xcite .
evaluating the partition function on the boundary of a 5-ball should give @xmath2 , which is equivalent to the condition @xmath368 .
since @xmath369 , this is equivalent to the gu - wen equation @xmath370 now we can write the partition function in a fixed @xmath0 background @xmath348 as ( up to positive multiplicative factors ) @xmath371 consider now coupling this theory to a background 3-form gauge field @xmath183 . this is achieved by replacing the constraint ( [ eps2rhoconstraint ] ) with @xmath372 this ensures symmetry under the 2-form @xmath7 gauge symmetry @xmath312 , @xmath162 , where @xmath373 .
thus we must merely replace @xmath374 with @xmath375 .
the partition function is thus @xmath376 now consider the effect of the gauge transformation @xmath162 . making a change of variables @xmath312
, we find after some work : @xmath377 observe the appearance of the first descendant of @xmath378 ( see appendix [ descent ] ) evaluated at the value of the 3-form gauge field @xmath379 .
this is almost the expected transformation law for the partition function , except that we expect @xmath272 , not @xmath379 .
this is easily fixed by multiplying the partition function @xmath380 by an additional factor @xmath381 this is a non - minimal contact - term coupling between @xmath272 and @xmath123 ensuring that the model has the proper anomaly for the 2-form symmetry to be a bosonic shadow of a fermionic phase .
in fact , it also shows that the full @xmath285 symmetry of this theory has the @xmath382 symmetry , since by our results in appendix [ eanomalies ] , this is determined once one knows the anomaly for the @xmath272 part only , though we can not decide whether to take the @xmath383 or @xmath384 extension .
we leave the explicit construction of state sums for more general 3d fsres to future work .
as discussed in appendix [ eanomalies ] , the two possible anomalies for a bosonic shadow of a fermionic theory are @xmath382 which differ by @xmath385 in this section we would like to investigate the physics of this term alone .
that is , we consider a bosonic theory with a 1-form @xmath7 symmetry and an anomaly given by ( [ bsqbanomaly ] ) . since the 3-form gauge field @xmath272
does not enter the anomaly , it is irrelevant whether the 2-form @xmath7 symmetry is present or not . if it is present , one can gauge it without introducing the spin structure ( since the 2-form @xmath7 symmetry is nonanomalous now ) and reduce to the case when it is absent . since the anomaly ( [ sqtilde ] ) is trivialized by the spin structure , and ( [ bsqbanomaly ] )
is twice ( [ sqtilde ] ) , this means that the latter anomaly is also trivialized by the spin structure . to see this more directly ,
we use the following identities in @xmath386 , where @xmath18 is any closed oriented 5-manifold ( see appendix [ oneformanomalies ] ) : @xmath387\cup sq^1 b.\ ] ] for a closed spin 5-manifold @xmath18 , @xmath388=0 $ ] , so the anomaly is trivial .
but there are no fermionic particles , because the 2-form symmetry , even if present , is not anomalous , and gauging it merely leads to a condensation of bosons ( the worldlines of the @xmath272-field ) .
it seems that we have a violation of the @xmath389 relation unless one considers also string statistics on the right hand side .
in fact , it appears possible to fermionize the theory with something less restrictive than a spin structure : a @xmath390-structure @xcite . just like a spin structure can be thought of as a trivialization of @xmath391 , a @xmath390-structure on an oriented @xmath392-manifold @xmath393 is a 2-cocycle @xmath394 such that @xmath395 , defined up to exact 2-cocycles .
clearly , any two @xmath390 structures differ by an element of @xmath396 , so the set of @xmath390 structures can be identified with @xmath396 , but not canonically . to see the relevance of @xmath390-structures
, we note that @xmath397 , hence @xmath398\cup sq^1 b = [ w_3]\cup b.\ ] ] hence the anomaly is trivial on a closed orientable 5-manifold @xmath18 satisfying @xmath399=0 $ ] . on a 5-manifold with a boundary @xmath164
, we need a trivialization @xmath400 of @xmath401 to define a counterterm @xmath402 which cancels the anomaly .
a model which depends on a @xmath390-structure but does not have fermions evades the contradiction with the spin - statistics relation .
but it does not correspond to a normal bosonic phase either . in the remainder of this section
we make a few remarks about such unusual phases .
first , although not every closed oriented 4-manifold is spin ( a counter - example being @xmath403 ) , every closed oriented 4-manifold admits a @xmath390 structure .
this can be easily shown using @xmath404 and properties of steenrod squares .
second , gauging a 1-form @xmath7-symmetry means proliferating strings .
their worldsheets are poincar - dual to @xmath221 .
the anomalous nature of the 1-form symmetry means that these strings need a @xmath390-structure for their definition .
such strings were discussed recently in a somewhat different context by one of us @xcite and were dubbed fermionic strings . their normal bundle is framed , and the wavefunction is multiplied by @xmath135 when the framing is twisted by one unit . thus phases requiring @xmath390 structure may be called fermionic string phases . a simple way to construct a fermionic string phase is to start with a bosonic model with a 2-form @xmath7 symmetry and anomaly ( [ guwenanomaly ] ) and set @xmath405 .
this means that we are embedding the 1-form @xmath7 symmetry into the 2-form @xmath7 symmetry group and then gauge the 1-form symmetry .
the resulting theory clearly has no @xmath7-grading on its hilbert space , because @xmath406 vanishes for any oriented 3-manifold @xmath71 .
this means that in general fermionic string phases do not have a conserved @xmath7-valued charge analogous to @xmath9 .
it also illustrates that ordinary fermionic phases do not really come in two types corresponding to the @xmath407 in @xmath259 because we can flip the sign by a redefinition of the symmetry operators corresponding to the shift @xmath408 .
a further insight is obtained by noticing that while the homology 2-cycle dual to @xmath68 represents the string worldsheet @xmath56 , the homology 1-cycle dual to @xmath409 can be thought of as the 1-cycle on @xmath56 which is dual ( in the 2d sense ) to the 1st stiefel - whitney class of the normal bundle of @xmath56 . note that while the 4-manifold @xmath164 is assumed to be oriented , @xmath56 need not be orientable . since we may think of @xmath272 as @xmath409
, we conclude that fermionic string phases have fermion worldlines confined to fermionic string worldsheets .
finally , let us give a couple of examples of bosonic shadows of fermionic string phases .
first , as remarked above we can take a shadow of any standard fermionic phase ( with a 2-form @xmath7 symmetry only ) and embed the 1-form @xmath7 symmetry into the 2-form @xmath7 symmetry .
for example , we can take the model ( [ ww ] ) and consider a global 1-form @xmath7 symmetry which acts as follows : @xmath410 it is easy to see that the action is invariant for any closed oriented @xmath164 and any @xmath67 .
another way to obtain a shadow of a fermionic string phase is to start with a model with both a 1-form and a 2-form @xmath7 symmetries and a mixed anomaly @xmath411 and then set @xmath412 .
that is , we embed the 1-form @xmath7 symmetry into a product of 1-form and 2-form @xmath7 symmetries in a nonstandard way . as a simple example , consider the @xmath7 gauge theory in 3 + 1d with an action @xmath413 one can get the desired anomaly ( [ bsqbanomaly ] ) by considering the following action of a global 1-form @xmath7 symmetry : @xmath414 gauging this symmetry means proliferating the strings and the particles of the @xmath7 gauge theory , but with particles confined to the string worldsheets in a particular way .
since the particles are not local with respect to the strings , one is forced to choose a @xmath390-structure on @xmath164 to make the result well - defined .
we have argued that every 3d fermionic model has a bosonic shadow which has a certain 3-group symmetry @xmath285 with an anomaly .
further , we argued that 3d fsres with a finite unitary symmetry @xmath0 are classified by triples @xmath105 satisfying certain rather complicated equations generalizing the gu - wen supercohomology .
we proposed that bosonic shadows of all such models are @xmath0-equivariant versions of a certain 4d tqft which we called the 2-ising model .
if @xmath415 vanishes , we can replace this 4d tqft with the simplest crane - yetter - walker - wang model and recover the supercohomology phases . gauging the anomalous 3-group symmetry @xmath285
is achieved by proliferating fermionic particles and kitaev strings .
it would be interesting to construct explicitly the resulting lattice model .
our proposed classification of 3d fsre phases can be made concrete once we pick a particular symmetry group @xmath0 .
let us give a few examples .
if @xmath416 with @xmath392 odd , both @xmath119 and @xmath58 vanish , and 3d fsre phases are classified by the same data as bosonic fsre phases . if @xmath416 with @xmath392 even , we only get gu - wen supercohomology phases , because the parameter @xmath58 vanishes . indeed ,
while @xmath417 if @xmath392 is even , the generator of this @xmath7 does not square to zero ( in fact , it generates a polyominal ring inside @xmath418 @xcite ) .
hence the equation @xmath419 has solutions only if @xmath420=0 $ ] . similarly ,
if @xmath0 is a product of several copies of @xmath7 , we only get supercohomology phases , since the cohomology ring of @xmath0 with @xmath7 coefficients is a polynomial ring @xcite , and any element of @xmath421 which squares to zero must be trivial .
the simplest example where there are phases which are not supercohomology phases is @xmath422 .
@xmath423 is generated by two elements @xmath424 of degree @xmath2 and an element @xmath425 of degree @xmath3 .
the only relation is @xmath426 .
thus @xmath421 has a unique nontrivial nilpotent element @xmath427 .
if we set @xmath428 , we get 3d fsre phases where the zipper @xmath429 , where @xmath430 and @xmath66 is a kitaev string , while all other zippers are `` trivial '' .
overall , for @xmath422 we get four supercohomology phases ( including the trivial phase ) , and four non - supercohomology phases .
we also found a new class of phases which are neither bosonic , nor fermionic , in that they have `` fermionic strings '' but no fermionic particles .
their partition function depends on a @xmath390-structure on the 4-manifold .
it would be very interesting to explore the physics of these new phases .
our discussion was not as systematic as that of @xcite because we lack an algebraic description of completely general unitary 4d tqft .
in particular , while we outlined the structure of various monoidal 2-categories relevant to us , we did not describe all the data which enter into a definition of these objects .
we hope to return to this issue in the future .
our results suggest that bosonization in higher dimensions will get progressively more complicated as the dimension increases .
this complexity reflects the topological complexity of the spin - bordism spectrum .
for example , in four spatial dimensions presumably one would have to deal with a symmetry 4-group which involves 3-form , 2-form and 1-form symmetries .
bosonic shadows of gu - wen supercohomology phases would be quite special since they only possess 3-form symmetries .
we review here some definitions and results from @xcite and @xcite .
in this paper we mostly work with simplicial cochains of a triangulated manifold @xmath164 with values in @xmath7 .
we assume a local order on vertices of the triangulation .
there is a coboundary operation @xmath431 satisfying @xmath432 . as usual , a cochain annihilated by @xmath433
is called a cocycle , and the space of @xmath434-cocycles is denoted @xmath435 .
the cohomology class of a cocycle @xmath436 is denoted @xmath437 $ ] .
there is a well - known product operation @xmath438 it is bilinear and associative and satisfies the leibniz rule @xmath439 the cup product is not ( super)commutative on the level of chains , rather one has @xmath440 where the new product @xmath81 has degree @xmath135 : @xmath441 note that the cup product is commutative on the level of cohomology classes , i.e. @xmath437\cup [ b]=[b]\cup[a]$ ] .
the @xmath81 product is not commutative either , rather one has @xmath442 where yet another product @xmath443 appears , etc .
one defines an operation @xmath444 , @xmath445 , by the following formula on the cochain level : @xmath446 despite appearances , this operation is linear on the level of cohomology classes , i.e. @xmath447=sq^1[a]+sq^1[b].$ ] we note that @xmath448 is a differential , i.e. @xmath449\cup [ b])=sq^1 [ a]\cup [ b]+ [ a]\cup sq^1 [ b].\ ] ] note also that if @xmath437\in h^p(x,{{\mathbb z}}_2)$ ] , then @xmath450=[a]\cup [ a]$ ] . in particular , for any @xmath437\in h^1(x,{{\mathbb z}}_2)$ ] one has @xmath437\cup[a]=sq^1[a]$ ] . on an @xmath392-manifold @xmath164
we have stiefel - whitney classes @xmath451 , @xmath452 .
the class @xmath453 is an obstruction to orientability .
if @xmath453 vanishes , then the class @xmath391 is an obstruction to having a spin structure .
these classes satisfy a number of relations . in particular , @xmath397 for all @xmath392 .
there also relations which depend on @xmath392 .
for example , for @xmath454 we have @xmath455 , so any orientable 2-manifold admits a spin structure . on a closed @xmath392-manifold @xmath164 we also have the wu formula : @xmath456=v_{n - p}\cup [ a],\quad [ a]\in h^p(x,{{\mathbb z}}_2),\ ] ] where @xmath457 is a certain polynomial in stiefel - whitney classes independent of @xmath164 .
it is known as the wu class .
the lowest wu classes are @xmath458 , @xmath459 , @xmath460 .
the wu formula has many useful consequences .
for example , it implies that on any orientable @xmath392-manifold one has @xmath461=0 $ ] for any @xmath462 . in particular , on a riemann surface @xmath164 the square of every element of @xmath463 vanishes .
another consequence is that on a spin 4-manifold @xmath164 the square of every element in @xmath464 vanishes .
the t hooft anomalies reveal themselves in how the symmetry algebra is realized projectively by unitary operators on the hilbert space . for an anomaly @xmath465 coming from group cohomology
, one can compute the so - called descendants to find the class @xmath466 of the projective action .
conversely , a system with a projective symmetry in class @xmath466 has the anomaly @xmath465 . in this section
, we discuss the descent procedure for anomalies of 1-form and 2-form @xmath7 symmetries relevant for fermionization .
unfortunately the calculation of descendants of the @xmath285 symmetry anomaly is beyond the scope of this paper . as a warm - up
, we consider the 1-form @xmath7 symmetry in 2 + 1d with the anomaly @xmath467 this cohomology class also defines an effective action for a bosonic spt in 3 + 1d protected by the 1-form @xmath7 symmetry .
we can use it to compute the spt ground state on a closed oriented 3-manifold @xmath164 .
we consider the path integral on the cone with base @xmath164 , denoted @xmath468 .
@xmath468 can be made from the cylinder @xmath469 $ ] by collapsing @xmath470 to a point . from this description
one sees that a 2-form gauge field on the cone @xmath471 is the same thing as a homotopy from the trivial gauge field on @xmath472 to some other gauge field on @xmath473 .
this is of course the same thing as a gauge transformation on @xmath164 and is parametrized by a @xmath7 1-cochain @xmath474 .
computing the sum over all these @xmath67 where we remember the ( ungauged ) boundary condition on @xmath164 we obtain the state @xmath475 we can rewrite this state as @xmath476 this expression makes it clear that @xmath477 is short - range entangled , since it is produced from a product state @xmath478 by time @xmath479 evolution of a hamiltonian defined by @xmath480 this function is called the first descendant of @xmath481 .
it also determines the variation of the partition function of the 2 + 1d theory under the 1-form gauge symmetry transformation .
now we consider a global symmetry transformation @xmath482 , where @xmath181 is a @xmath7 1-cocycle , @xmath483 .
we find a variation in the exponent @xmath484 which defines the second descendant @xmath485 this means that for closed @xmath164 the spt ground state is invariant under global 1-form symmetry , but when @xmath486 is nonempty , it is invariant only up to a boundary term : @xmath487 it also tells us whether the global 1-form symmetry acts projectively on the hilbert space of the 2 + 1d theory .
that is , whether transforming by @xmath488 is any different than transforming by @xmath489 followed by @xmath490 .
this is measured by @xmath491 so indeed we do have a projective symmetry action measured by a bilinear form on the symmetry 2-group @xmath492 .
as we have seen above , such cocycles are trivialized by quadratic refinements of the form , which in this dimension we may obtain from a spin structure .
next we consider 2-form @xmath7 symmetry in 3 + 1d with an anomaly @xmath493 where @xmath494 $ ] is the generator of @xmath495 .
it can also be regarded as an effective action of a 4 + 1d spt with a 2-form @xmath7 symmetry . as before , on a 4-manifold @xmath164 we obtain an spt ground state @xmath496 where @xmath497 is a @xmath7-valued 2-cochain parametrizing a gauge transformation of @xmath272 .
we compute @xmath498 so we can rewrite @xmath499 using the first descendant @xmath500 now we compute the transformation of this state under a global symmetry @xmath501 for some @xmath7-valued 2-cocycle @xmath502 .
we find @xmath503 this variation involves the boundary variation we expected from the previous calculation but also a new ingredient : a prefactor @xmath504 this factor is not a boundary term for a general @xmath164 and @xmath181 . for example , when @xmath505 and @xmath181 represents the unique nonzero degree-2 class , this prefactor is @xmath135 .
such a symmetry transformation multiplies the ground state by @xmath135 .
thus we are dealing with an spt phase only if we restrict to those @xmath181 for which @xmath182=[\beta\cup\beta]=0 $ ] .
alternatively , we can restrict @xmath164 to be a spin 4-manifold . after this
is done , we can extract the second descendant which measures to what extent the action of the global 2-form symmetry on the hilbert space of the 3 + 1d theory is projective : @xmath506 an interesting feature about this term is that it is not invariant under 2-gauge transformations @xmath507 , where @xmath474 .
indeed , it has a variation @xmath508 which looks like it could be exact but unfortunately is not .
this means that the symmetry action on the boundary is only defined for cocycles @xmath509 and not cohomology classes .
likewise , the bilinear form @xmath510 is only well - defined on @xmath511 .
we summarize here the definition and properties of the function @xmath512 here @xmath71 is a closed oriented 3-manifold equipped with a triangulation and a branching structure , and @xmath74 is a spin structure on @xmath71 . given an oriented 3-manifold @xmath71 , there exists an oriented 4-manifold @xmath164 with boundary @xmath71 such that any @xmath513 extends to a 2-cocycle @xmath514 on @xmath164 . in other words ,
the restriction map @xmath515 is surjective .
given such @xmath164 , we may consider the relative stiefel - whitney class @xmath516 which measures the obstruction to extending @xmath74 to a spin structure on @xmath164 .
recall also that the cup product makes @xmath517 into a module over the algebra @xmath518 , and that the fundamental homology class @xmath519 $ ] takes values in @xmath520 .
thus it makes sense to consider the expression @xmath521\cup w_2(x , y,{\zeta})\cap [ x].\ ] ] we write it somewhat schematically as @xmath522 to indicate that when the spin structure @xmath74 is shifted by @xmath316 , the quantity ( [ w2bevaluation ] ) shifts by @xmath523 in other words , the set of spin structures on @xmath71 is an affine space over the vector space @xmath524 , and the quantity ( [ w2bevaluation ] ) is an affine linear function on it .
the quantity ( [ w2bevaluation ] ) depends on the cohomology class @xmath525\in h^2(x,{{\mathbb z}}_2)$ ] , but not on the concrete representative .
but it also depends on the choice of @xmath164 as well as the choice of the extension of @xmath68 from @xmath71 to @xmath164 .
given two such choices , @xmath526 and @xmath527 , the difference between the corresponding expressions is @xmath528 where the 4-manifold @xmath215 obtained by gluing @xmath164 and @xmath529 along @xmath71 is closed , and @xmath530 restricts to @xmath514 on @xmath164 and @xmath531 on @xmath529 . but
this is the same as @xmath532 this implies that the expression @xmath533 does not depend either of the choice of extension of @xmath68 from @xmath71 to @xmath164 , nor on the choice of @xmath164 . on the other hand ,
since @xmath514 is an absolute 2-cocycle , it does depend on the choice of @xmath68 within its cohomology class in @xmath78 .
it is easy to see that @xmath534 it is also easy to see that @xmath535 now suppose the spin structure @xmath74 extends to @xmath164 .
then @xmath536 vanishes , and therefore we get @xmath537 this property is used in section 4 to argue the conservation of fermion number .
conversely , for a fixed triangulation and branching structure , the function @xmath77 completely determines the equivalence class of @xmath74 .
indeed , given any two spin structures @xmath74 and @xmath538 , the difference @xmath539 is linear and depends only on the cohomology class of @xmath68 and thus must have the form @xmath540 for some @xmath316 . on the other hand ,
this difference is equal to @xmath541 . by poincar duality
, @xmath542 for all @xmath68 implies that @xmath543 is exact , which means that @xmath544 .
in this section we classify possible anomalies for a 1-form @xmath7 symmetry in 3 + 1d , both for bosonic and fermionic theories , assuming the space - time symmetry is orientable ( i.e. ignoring time - reversal symmetries , if any ) . in the bosonic case , we need to compute the oriented cobordism group @xmath545 it is the pontryagin - dual of the oriented bordism group @xmath546 .
physically , these classify possible 5d topological actions built out of a 2-form gauge field @xmath547 , where @xmath18 is a closed oriented 5-manifold .
all such topological terms will be integrals of densities made out of @xmath68 and certain characteristic classes of the tangent bundle of @xmath18 , namely the stiefel - whitney classes and the pontryagin classes .
actually , since pontryagin classes modulo 2 can be expressed through stiefel - whitney classes , it is sufficient to consider the latter .
an obvious approach is to construct elements in @xmath548 and then embed them into @xmath386 using the embedding @xmath549 , keeping in mind that distinct elements of @xmath548 can become identical elements of @xmath386 .
let us write down candidate independent terms in @xmath548 .
orientability implies @xmath550 and it follows @xmath551 for any @xmath552 @xcite , so we have four candidates : @xmath553 the 2nd and the 3rd are actually the same thanks to the wu formula @xcite .
further , since @xmath448 satisfies the leibniz rule w. r. to the cup product , and @xmath397 , the 4th one is the same as the 3rd one .
thus we are left with only two independent elements of @xmath548 .
now we must map these classes to @xmath386 .
in fact , we will find they map to the same ( nonzero ) element . to see this
, one needs to use the long exact sequence @xmath554 where we have introduced the short - hand @xmath555 .
the 1st map is multiplication by @xmath3 , and the 2nd map is the bockstein homomorphism associated to the short exact sequence @xmath556 .
we are interested in the image of the bockstein homomorphism .
the group @xmath557 is isomorphic to @xmath558 and is generated by @xmath559 times the pontryagin square of @xmath560\in h^2(k({{\mathbb z}}_2,2),{{\mathbb z}}_2)$ ] , for which a representative may be written @xmath561 where @xmath562 is an integral lift of @xmath68 . using the defining property of @xmath81 , we find that the bockstein of ( [ pontryaginsq ] ) is @xmath563 which is a cocycle formula for @xmath564 in @xmath565 . by exactness
, it follows that this difference maps to zero in @xmath566 . to verify that this exhausts all possible topological actions
, one can use the atiyah - hirzebruch spectral sequence computing oriented bordism groups starting from the homology groups @xmath567 .
the integral homology groups of @xmath555 are known , and the nonzero ones up to degree @xmath568 are @xmath569 the first term defines a purely gravitational anomaly @xmath570 which splits off from the anomaly group .
then the spectral sequence implies that the map @xmath571 is an isomorphism .
hence the group of gauge anomalies @xmath572 , and the nontrivial anomaly action can be written as @xmath573 in the fermionic case , we need to compute @xmath574 . note that since @xmath575 for a closed orientable spin manifold @xmath18 , the bosonic action ( [ bosonic5d ] ) becomes trivial for such @xmath18 . that is , the image of the map @xmath576 is trivial .
however , there can also be elements of @xmath574 which do not come from @xmath577 .
these topological terms use the spin structure in a key way and wo nt be just integrals of characteristic classes . looking at the atiyah - hirzebruch spectral sequence , we see that the only such element arises from the @xmath578 term @xmath579 .
this cohomology group is @xmath7 and is generated by @xmath580=[sq^2b]$ ] .
the corresponding spin - topological action is evaluated as follows : we take the homology class in @xmath581 which is poincar - dual to @xmath580 $ ] and pick a closed 1d submanifold @xmath124 which realizes it .
then we restrict the spin structure of @xmath18 to @xmath124 and evaluate the corresponding holonomy .
thus @xmath582 .
note that if @xmath68 satisfies the constraint @xmath580=0 $ ] , then the corresponding 5d spin - topological action is zero .
thus the fermionic anomaly is necessarily trivial for such @xmath68 .
in this section we wish to discuss possible t hooft anomalies for bosonic and fermion systems with 2-form @xmath7 symmetry .
the calculations are much the same as the previous section , except where now the classifying space @xmath583 , an eilenberg - maclane space with only nonzero homotopy group @xmath584 . for bosonic systems
, the possible anomalies are @xmath585 the first two are actually equal thanks to the wu formula , while the third does not involve the gauge field @xmath586 and so describes a purely gravitational anomaly .
so as before , we find that ( 3-)group cohomology describes all the gauge anomalies .
when we consider these terms on a closed spin 5-manifold @xmath18 , they all vanish because @xmath587 . for new anomalies
we look to @xmath588 and @xmath589 in the atiyah - hirzebruch spectral sequence .
the first does not survive the @xmath590 differential since @xmath591 .
the second , however , defines a topological term which measures the holonomy of the spin structure along the curve poincar dual to @xmath592 .
we have discussed the appearance of the 3-group symmetry @xmath285 , whose elements are pairs @xmath593 .
the group law is not the product , but has a twist @xmath594 in this section we discuss possible anomalies for such symmetr in bosonic and fermionic systems . like the @xmath7 1-form symmetry considered in the previous appendix
, @xmath285 has a classifying space denoted @xmath356 with @xmath595 , @xmath584 and the postnikov class @xmath255 , reflecting the twisted group law .
this means that one can think of a map
@xmath596 as a pair @xmath597 satisfying @xmath598 from this one sees that there is a map @xmath599 by forgetting @xmath272 .
this map is a fibration with fiber @xmath600 .
this fibration is very useful for computing the cohomology of @xmath356 .
for instance , to compute @xmath601 , a first approximation to the bosonic anomaly group @xmath602 , we use the serre spectral sequence which starts with @xmath603 .
the three possible terms are @xmath604 \in e_2^{5,0 } \quad \frac{1}{2 } [ b \cup c ] \in e_2^{3,2 } \quad \frac{1}{2 } [ b\cup \frac{\delta \hat b}{2}]\in e_2^{0,5},\ ] ] where @xmath605 is an integral lift of @xmath68 .
the differential in this spectral sequence comes from eq .
( [ apostnikov ] ) .
for example , the differential of the second term above is @xmath606 , which is a non - zero class in @xmath607 , so we throw it out .
the differentials of the third term are all zero , but it may be a differential of something else .
however , there is no candidate in the right degree , so the third term survives to give a nontrivial class in @xmath608 .
the first term is a little complicated , but after some work one finds that it also survives , and that the corresponding element of @xmath608 looks as follows : @xmath609 here @xmath610 is a mod-2 cochain defined by the equation @xmath611 .
an explicit simplicial expression for it is @xcite @xmath612 the expression ( [ sqtilde ] ) restricts to @xmath613 when one sets @xmath614 , i.e. it is a extension of @xmath615\in h^5(l , u(1))$ ] to the total space @xmath356 .
note that the extension is not unique , and that the two possible extensions differ by @xmath616 $ ] .
note also that @xmath617 = \frac12\left [ b\cup \frac{\delta \hat b}{2}\right]=\frac12 [ bsq^1b].\ ] ] this means that the cohomology group @xmath608 is isomorphic to @xmath558 , and that it is generated by the cohomology class of eq .
, for either choice of the sign .
the ambiguity in the sign is simply the ambiguity in choosing the generator of @xmath558 .
the anomalies for 3 + 1d bosonic systems with @xmath285 symmetry are actually classified by the cobordism group @xmath602 .
there is another useful spectral sequence for computing this , the atiyah - hirzebruch - serre spectral sequence , which goes from @xmath618 to @xmath619 .
we have shown in the above section that the map @xmath620 is an isomorphism and it is not hard to show that @xmath621 is also an isomorphism for all @xmath622 .
thus , except for the purely gravitational anomaly @xmath623 , all the @xmath285 anomalies are classified by @xmath624 . note that the shift @xmath408 exchanges @xmath625 and @xmath626 , so the difference between the anomalies is not really physical , amounting to a redefinition of the symmetry operators .
further , @xmath627 = [ \frac{1}{2}b \frac{\delta \hat b}{2}]$ ] .
now we want to consider what happens with the map @xmath628 . for this
we can use naturality of the atiyah - hirzebruch - serre spectral sequence .
we know that our nonzero classes in @xmath602 come from @xmath629 and @xmath630 . because the first is proportional to @xmath391 and the second to @xmath390 , these map to zero on closed spin 5-manifolds , which all have @xmath587 and @xmath631 .
this implies that the map @xmath628 sends all the gauge anomalies to zero ( and @xmath623 too , for that matter ) .
s. ryu , a. schnyder , a. furusaki , a. ludwig , `` topological insulators and superconductors : ten - fold way and dimensional hierarchy , '' new . j. phys .
* 12 * , 065010 ( 2010 ) [ arxiv:0912.2157[cond - mat.mes - hall ] ] .
a. kitaev , `` homotopy - theoretic approach to spt phases in action : z16 classification of three - dimensional superconductors '' , talk given at ipam program `` symmetry and topology in quantum matter '' in january , 2015 f. d. m. haldane , `` luttinger liquid theory of one - dimensional quantum fluids . i. properties of the luttinger model and their extension to the general 1d interacting spinless fermi gas , '' journal of physics c : solid state physics , volume 14 , number 19 ( 1981 ) z - c .
gu and x - g .
wen , `` symmetry protected topological orders for interacting fermions - fermionic topological nonlinear @xmath58 models and a special group supercohomology theory , '' phys .
b * 90 * , 115141 ( 2014 ) [ arxiv:1201.2648[cond - mat.str - el ] ] .
|
we discuss bosonization and fermionic short - range - entangled ( fsre ) phases of matter in one , two , and three spatial dimensions , emphasizing the physical meaning of the cohomological parameters which label such phases and the connection with higher - form symmetries .
we propose a classification scheme for fermionic spt phases in three spatial dimensions with an arbitrary finite point symmetry @xmath0 .
it generalizes the supercohomology of gu and wen .
we argue that the most general such phase can be obtained from a bosonic `` shadow '' by condensing both fermionic particles and strings .
|
99 email : ( smb ) [email protected] ( dm ) [email protected] , _ proc .
natl . acad .
usa _ , * 94 * , 11935 ( 1997 ) ; t. strunz _
_ , _ proc .
usa _ ; m. rief , h. clausen - schaumann , and h.e .
gaub , _ nature struct .
biol . _ * 6 * , 346 ( 1999 ) . ,
_ , * 79 * , 4489 ( 1997 ) , _ phys .
e _ , * 58 * , 2386 ( 1998 ) .
s. m. bhattacharjee cond - mat/9912297 , _ j. phys .
a _ * 33 * , l423 ( 2000 ) ; 9003(e ) ( 2000 ) ; cond - mat/0010132 .
d. k. lubensky , d. r. nelson , _ phys .
lett . _ * 85 * , 1572 ( 2000 ) , cond - mat/0107423 ; k. l. sebastian , _ phys .
e _ , * 62 * , 1128,(2000 ) , h. zhou , cond - mat/0007015 .
d. marenduzzo _ et .
_ , _ phys .
lett . _ * 88 * 028102 ( 2002 ) , d. marenduzzo , a. trovato , a. maritan , _ phys .
e _ * 64 * , 031901 ( 2001 ) .
u. gerland , r. bundschuh , and t. hwa , _ biophys .
j. _ * 81 * , 1324 ( 2001 ) .
d. poland , h. a. scheraga , _ j. chem .
phys . _ * 45 * , 1464 ( 1966 ) ; m. peyrard , a. r. bishop , _ phys .
_ , * 62 * , 2755 , ( 1989 ) .
s. mukherji , phys .
e * 50 * , r2407(1994 ) .
r. m. neumann , _ phys .
a _ , * 31 * , r3516 ( 1985 ) , r. a. guyer , j. a. y. johnson , _ phys .
a _ , * 32 * , 3661 ( 1985 ) , j. t. titantah _ et al . _
e _ , * 60 * , 7010 ( 1999 ) . m. peyrard , _ europhys
_ , * 44 * , 271 ( 1998 ) .
r. e. thompson , e. d. siggia , _ europhys .
_ , * 31 * , 335 ( 1995 ) .
|
the possibility of detecting mutations in a dna from force measurements ( as a first step towards sequence analysis ) is discussed theoretically based on exact calculations .
the force signal is associated with the domain wall separating the zipped from the unzipped regions .
we propose a comparison method ( `` differential force microscope '' ) to detect mutations .
two lattice models are treated as specific examples .
87.14gg , 87.80fe , 05.20gg the possibility of a force induced unzipping transition @xcite has opened up new ways of exploring the properties of biomolecules .
since the critical force for a double stranded dna depends on its sequence , an inverse problem can be posed : `` can the dna sequence be detected from the force required to unzip it ? '' a still simpler question would be : `` can the mutations in a dna be detected by force measurements ? '' .
a positive answer to either of these questions would lead to the possibility of testing ( and detecting ) mutations and of sequence determination in a non - destructive way . based on a few simple models used earlier for the dna denaturation and unzipping transitions@xcite ,
we show here the signature of mutations on the @xmath0-vs-@xmath1 curve ( force versus relative distance of the end points of the two strands ) .
the inverse problem is then to get the position of the mutations from such an experimentally realizable curve .
our emphasis in this paper is on the one base pair mutation problem also known as point mutation .
this is not just of academic interest .
the replication of dna is a high - fidelity process , thanks to the inbuilt proof - reading and repair mechanisms , so that mistakes are very rare though even one could play havoc .
our proposal ( solution of the inverse problem ) , based on exact calculations , is to obtain the force difference ( `` differential force '' ) between identically stretched native and mutant dnas .
this could be done by using _
e.g. _ two atomic force microscope tips : we name this apparatus a `` differential force microscope '' .
the position of the mutation can be obtained from a calibration curve involving the extremum position ( or its value ) of the differential force curve . in our models
, we can find the nature of the mutation from the sign of the differential force .
let us model a double stranded ( ds ) dna by two n - monomer polymers interacting at the same contour length ( or monomer index ) @xmath2 of the strands through a contact attractive potential @xmath3 , which might depend on the contour length .
the interaction energy is @xmath4 , where @xmath5 denotes the relative distance of the @xmath6th base pair and @xmath7 denotes the kronecker delta .
other features like the self- and mutual - avoidance , base stacking energy , helicity , etc are ignored in this study ( but see below ) in order to focus on the base pairing energy .
such a model exhibits a denaturation transition at a model - dependent temperature @xmath8@xcite from a low @xmath9 double stranded configuration to a high @xmath9 phase of two unpaired single strands . a few definitions : starting with a sequence @xmath10 or bases of a dna ( to be called the native dna )
, we define a mutant dna as one with almost the same base sequence as the native molecule _ except _ a few base pairs . a homogeneous dna or homo - dna is a dna with identical base pairs ( i.e. @xmath11 for all @xmath2 ) while a heterogeneous dna is one with heterogeneity in the sequence ( @xmath2-dependent @xmath12 ) .
notice that the model we have defined does not consider base pair stacking interaction and therefore does not distinguish _
e.g. _ between an at and a ta ( or cg and gc ) base pair .
consequently , the mutations we are referring to are only those involving at ( or ta ) @xmath13 cg ( or gc ) .
two specific examples are considered here because of their exact tractability ( analytical and numerical ) : ( mi ) a two dimensional @xmath14 directed dna model ( with the strands directed along @xmath15 ) with base - pair interaction and mutual avoidance ( forbidden crossing of the strands ) , and ( mii ) two gaussian polymers with the base - pair interaction in @xmath16-dimensions . in both cases , only the relative chain , involving the separation * r * of the bases at the same contour length , need be considered .
these models in addition to the denaturation transition also show , for @xmath17 , an unzipping transition in presence of a force at the free end ( @xmath18)@xcite . in this paper , we consider the conjugate fixed distance ensemble and henceforth restrict ourselves to @xmath17 .
consequently , we put @xmath19 unless explicitly shown , where @xmath20 is the boltzmann constant . for any ds dna
having their first monomers ( @xmath21 ) joined and their last monomers ( @xmath18 ) at a relative lattice distance @xmath22 , the force @xmath23 required to maintain this relative distance @xmath22 is @xmath24 where @xmath25 represents the free energy of the system in the fixed-@xmath22 ensemble@xmath26}$ ] . by definition ,
@xmath27 = @xmath28 which gives the free energy of binding or the work required to unzip completely dna .
the two ensembles , fixed - force and fixed - stretch , are expected to give identical results in the @xmath29 limit , though for finite @xmath30 inequivalence might be expected @xcite ( this is indeed the picture valid for the homo - dna ) , but a more serious situation arises if , _
e.g. _ for the scalar case , @xmath31 in the fixed - stretch ensemble because , in the fixed - force ensemble , @xmath32 @xmath33 , where @xmath34 denotes thermal averaging .
this is the case for a heterogeneous dna as shown in fig .
1 .
the regions of `` wrong '' sign ( reminiscent of preshocks in burgers turbulence@xcite or of `` slip '' in ref.@xcite ) go away only after quenched averaging but do survive in the thermodynamic limit for each individual realization . _
the absence of self - averaging in the system is encouraging , because it implies that individual features , typical of a single realization of the sequence , are maintained in the characteristic curves .
_ we draw attention to the overlap of the @xmath0 vs @xmath35 curves over a range of @xmath35 in fig . 1 for different lengths with identical sequence near the open end .
this overlap is a sign that the modulation is a characteristic of the sequence , and we have found this to be true quite generally .
let us now consider the one mutation problem where the @xmath36-th pairing energy of the native dna has been changed from @xmath37 to @xmath38 .
for this one site change in @xmath39 , the partition function @xmath40 is given by [ mutation_one ] z_n|k(*r*,n)=z_n(*r*,n)+c_k z_n(*r*,n , k ) . where @xmath41 and @xmath42 is the native dna partition function with last monomers at a relative distance @xmath22 , while @xmath43 is with the additional constraint of the @xmath44th pair being zipped .
notice that eq .
[ mutation_one ] applies for both heterogeneous and homogeneous cases ( and also for self avoiding strands ) .
we define a zipping probability @xmath45 @xmath46 , which is the conditional probability that site @xmath36 is zipped when the free ends are at a distance @xmath22 .
the force difference ( to be called the differential force ) to keep the free ends of both the mutant and the native dna at the same distance @xmath22 can be written as @xmath47 [ mut3 ] here @xmath48 is a generalized force , the one necessary to keep the free ends of the native dna at a relative distance @xmath22 when the @xmath44th monomers are zipped .
except @xmath49 , all other quantities in eq .
[ mut3 ] refer to the native dna .
this fact allows the inverse problem to be tackled as we show explicitly for a few cases . since for dna we have generally two possible choices for @xmath50 , ( @xmath51 and @xmath52 )
, the sign of @xmath49 determines the sign of @xmath53 in eq .
[ mut3 ] .
thus , in our simple model , the nature of the mutation can be identified from the sign of the differential force curve .
eqs .
[ mutation_one],[mut3 ] can be generalized to more than one mutations and to models with other local energy parameters , though they become algebraically more involved .
these more complicated cases will be discussed elsewhere .
we consider the simplest case here .
if @xmath54 is the direction of stretching , the quantity @xmath55 has a kink - like behavior ( as in fig .
[ fig:2 ] ) . in the fixed stretch ensemble
, the chain separates into an unzipped and a zipped region separated by a domain wall , which to a very good approximation can be fit by a @xmath56 function : @xmath57 } /{w_{\rm d}(k ) } , \end{aligned}\ ] ] where @xmath58 and @xmath59 are the position and the width of the wall ( kink ) , and @xmath60 is a constant .
eq .
[ eq:3 ] suggests that @xmath61 , where @xmath62 is a scaling function .
it is possible to extend the above analysis to the general case where more than one mutation is present .
for example , the partition function for the case with two mutations at positions @xmath63 and @xmath64 , in an obvious notation , is @xmath65 the last term of the above equation , representing correlation of the mutations , gives an additional contribution to the differential force over and above the individual contributions of the mutations .
this additional contribution is negligible if the two mutations are far away or , more quantitatively , not in the same domain wall .
we now use these general results for cases ( mi),(mii ) . in the two dimensional model ( mi ) ,
the partition function for two directed chains having their last monomers at a relative lattice distance @xmath35 ( along @xmath66 and in unit of the elementary square diagonal ) , and their first monomers joined , can be written in terms of n monomer - to - monomer transfer matrices @xmath67 ( @xmath68 ) with matrix elements @xmath69 @xmath70 @xmath71 @xmath72 being the position vectors ( with the constraint @xmath73 ) . for a homo - dna with contact energy @xmath50 ,
the largest eigenvalue of @xmath74 determines the free energy and the thermodynamic properties in the limit @xmath75 . for @xmath76 , the melting temperature , @xmath77 with @xmath78 .
indeed , in the fixed - stretch ensemble , any finite ( @xmath79 ) stretch puts the chain on the phase coexistence curve . exploiting the equivalence of the ensembles valid for for homo - dna
, we find @xmath80 , where [ characteristic ] ( y)&=&2 ^-1(\{y , y_0 } ) , ( y_0^ ) , is a piecewise continuous non - analytic function . for finite @xmath30 , there is no singularity but the approximation of eq .
[ characteristic ] still works quite well . at @xmath81
the force curve increases sharply ( see also fig.1 , curves ( a ) and ( c ) ) .
we now come to the explicit results for the one - mutation case where one @xmath50 is replaced by @xmath82 . for a homo - dna , by starting from eq .
[ characteristic ] , using eqs .
[ mut3],[eq:3 ] , one can find analytical approximations to the shapes of the previously introduced @xmath83 and @xmath53 in terms of piecewise continuous functions .
we find _ e.g. _
@xmath83 = @xmath84 , where ( if @xmath85 ) : @xmath86 [ eq:2 ] and @xmath87}$ ] .
eq .
[ mut3 ] now simplifies because @xmath88 .
the characteristics of the differential force curve @xmath53 vs @xmath35 , such as the extremum value , its position and the width , @xmath89 , @xmath90 and @xmath91 , can be determined from eqs .
[ eq:2 ] as [ results ] f_max(k)~ [ w_f(k)]^-1 ~|k^-1/2 , x_f(k ) = |kx_f(|k ) with @xmath92 and @xmath93 , where @xmath94 is defined in eq .
[ characteristic ] .
the area of the peak , which yields the difference ( with respect to the native case ) in the work necessary to unzip completely the molecule , is constant as expected .
the scaling form introduced after eq .
[ eq:3 ] suggests that the differential force is significant only in the domain wall region and the width of the domain wall @xmath95 as we do see in the numerical results . for model ( mii ) , with one - dimensional gaussian polymers ( @xmath96 )
, @xmath83 has been calculated exactly by a transfer matrix method and is shown in fig.2a . the validity of eq .
[ results ] ( @xmath97 ) is apparent from the data - collapse of the various @xmath53 curves in fig.2b where @xmath98 is plotted against @xmath99 .
the peak force difference , @xmath89 as a function of the mutant position @xmath36 is in accord with the @xmath100 law of eq .
[ results ] .
the results for model ( mi ) are shown in fig .
[ fig:3 ] . for @xmath101 ( @xmath102 as above )
the situation is similar as , _
e.g. _ , the data - collapse of fig .
[ fig:2 ] and eq .
[ results ] remain valid . coming to the case of one mutation on a heterogeneous dna consisting of two energies @xmath51 and @xmath103 chosen with equal probability
, the shapes of the zipping probability curves are found to be similar to the homo - dna case , in fact indistinguishable on the plot of fig .
[ fig:2 ] , with @xmath58 and @xmath59 sequence dependent .
this indicates the validity of the domain wall interpretation even for a heterogeneous dna .
the mutation involves a change of the energy at site @xmath36 ( i.e. @xmath104 ) .
the signals @xmath53 for various mutations are shown in fig .
[ fig:4 ] .
as already mentioned , in our models the sign of @xmath53 tells us the nature of the mutation .
these individual curves can again be collapsed on to a single one as for homo - dna , though the nature of collapse is not as good , mainly because the area under the curve is no longer strictly a constant .
this reflects the importance of local sequences around the mutation point .
figure 3 ( curves b , c ) shows the @xmath36-dependence of @xmath90 . unlike for homo - dna , @xmath89 does not seem to have the simple form of curve ( a ) in fig .
3 .
although the linearity is maintained for @xmath90 as for the homogeneous case , there are regions of nonmonotonicity at small scales which hamper the inversion . fig .
3 gives a basis for a calibration curve .
this could be @xmath90 or @xmath105 vs k ( or both ) for a homogeneous dna , though for heterogeneous dna , we find the @xmath106-vs-@xmath36 curve to be more reliable .
given the value of @xmath90 , one can look up in fig.3 for the corresponding @xmath36 .
the accuracy of the method relies on the ability to resolve close - by mutation points , _
i.e. _ mutation points in the same domain wall .
there are differences in the full profiles of the @xmath53 curves for mutations at , say , @xmath36 , @xmath107 , @xmath108 , though translating that information back to the identification of the position is yet to be achieved .
a better resolution is in any case obtained by changing the point at which the strands are pulled .
we now argue on how our calculation can compare with a potential experiment . in @xcite , the typical force arising in the unzipping
is between @xmath109 and @xmath110 pn , while the resolution is set below @xmath111 pn , so in percentage it is @xmath112 .
dynamical effects ( important in @xcite ) are almost negligible already at the lowest stretching velocity used in @xcite ( @xmath113 nm / s ) and are less important as the velocity is lowered .
our values for the typical force at @xmath114 are at the border of this present day resolution ( see fig .
3 , where it appears that the resolution is around @xmath115 for @xmath116 , _
i.e. _ on the middle of the chains ) . in principle
they can be improved above it by lowering @xmath9 , though it is not clear to what extent this will work because the experimental temperature range is rather limited . in ref.@xcite ,
the authors suggest that there will be experimental difficulties which would hamper the acquisition of the base by base sequence of dna by means of force measurements ( but would however allow to get information over groups of @xmath117 bases ) .
this difficulty , though absent in our exact analysis of the models , might also set a lower limit on the error on the position of the mutation .
summarizing , we prove that mutations are detectable in the theoretical models .
numbers coming from our models suggest that this measurement might be a benchmark for present day real technology .
we finally propose an algorithm for sequencing dna from the unzipping force in our models .
this is defined so that the energy of the @xmath118th base pair is @xmath52 if the average force at stretch @xmath119 is above the force signal of a homo - dna with an attractive energy @xmath120 ( @xmath9 is low enough that @xmath121 ) .
our algorithm differs from the discussion in @xcite because @xmath9 and extra constraints ( see below ) play crucial roles .
if we define the `` score '' of the algorithm as the fraction of base pairs correctly predicted , from our data we observe that for any finite - size sample the score is 100% for @xmath122 , with @xmath123 for @xmath124 . however , @xmath125 monomers near the open end can be sequenced at @xmath126 , no matter how big the total @xmath30 is .
once this is done , we restart this time keeping the corresponding bases at position @xmath125 from the open end at a distance @xmath35 with the constraint that the monomers at @xmath30 are at a distance @xmath127 , to prevent rejoining in the already unzipped @xmath125 monomers . in this way we would sequence another @xmath125 monomers and so on .
we have verified this for models ( mi ) and ( mii ) . in conclusion , we studied the @xmath128 vs. @xmath22 characteristic curve in the fixed stretch ensemble for simple models of dna focusing on the base pairing energy only .
we have seen that for a homo - dna , the force difference between a native and the corresponding one mutation case when pulled to the same distance contains enough signature to locate the position of the mutation .
this could be the basis of a differential force microscope to detect mutations . for a heterogeneous dna
, the mutation point can not be localized always as accurately as for homo - dna .
accuracy could be achieved by taking cognizance of the full features of the @xmath53 curve .
we have shown that the differential force curve can be understood as due to the domain wall separating the zipped and the unzipped phases as the strands are pulled apart . moreover , we found ( fig .
1 ) that the modulations in the force curve are connected to the local sequence .
this holds promise of extension of our proposal to cases beyond point mutation .
_ acknowledgements : _ smb thanks ictp for hospitality .
we thank a. maritan and f. seno for useful discussions .
|
numerous studies have attempted to understand how gas is converted into stars .
observational data indicate a power - law relation between these two quantities : @xmath6 where @xmath7 and @xmath8 are the star formation rate ( sfr ) and gas surface densities in units of m@xmath9 yr@xmath10 kpc@xmath11 and m@xmath9 pc@xmath11 .
such relations are often called the kennicutt - schmidt ( k - s ) laws ( schmidt 1959 ; kennicutt 1998 ) .
only recently have observations provided high - resolution , multiwavelength images of nearby galaxies for spatially - resolved , sensitive measurements of @xmath8 and @xmath7 , allowing exploration of the k - s relation on sub - kpc scales .
for instance , bigiel et al .
( 2008 , 2011 ) used the co(j=2 - 1 ) line as an h@xmath12 gas tracer and far - ultraviolet ( fuv ) plus 24@xmath13 emission as a sfr tracer to study the k - s relation at sub - kpc resolution in nearby disc galaxies observed as part of the hera co - line extragalactic survey ( heracles ; leroy et al .
they proposed a linear k - s relation ( and thus a constant molecular gas depletion time ( t@xmath1 ) ) , independent of local conditions such as orbital timescale , mid - plane gas pressure and disc stability parameter @xmath14 ( leroy et al .
2008 ) . some subsequent works , however , reported non - linear relations between @xmath7 and @xmath15 using different fitting methods ( shetty et al .
2013 ) or new data ( e.g. , momose et al .
2013 ; pan et al .
2014 ) .
@xcite investigated the relation between integrated molecular gas depletion time and global galaxy parameters statistically using galaxies observed as part of the cold gass project , which obtained iram @xmath1622@xmath17 single beam co(1 - 0 ) line measurements for @xmath16360 galaxies with stellar masses @xmath18 in the redshift range 0.02 0.05 @xcite .
sfr was estimated by fitting stellar population models to optical and ultraviolet ( uv ) broad - band photometry .
an estimate of the molecular gas depletion time of the full galaxy was made by applying aperture corrections to the co line luminosities .
molecular gas depletion time was found to correlate well with various galaxy parameters such as stellar mass , stellar surface density , concentration of the light ( i.e. , bulge - to - disc ratio ) , near - uv ( nuv)@xmath19r color and specific star formation rate ( ssfr ; defined as the ratio of sfr to stellar mass ) .
we subsequently improved on this work by estimating sfr with the combination of galex fuv and wise 22@xmath13 data within fixed apertures matched to the 22@xmath17 beam size of the gas observations @xcite .
dependences of the depletion time on galaxy structural parameters such as stellar surface density and concentration index were then found to be much weaker , or even absent .
we went on to demonstrate that the _ primary dependence _ of t@xmath1 is on ssfr .
all other remaining correlations , such as the t@xmath1stellar mass relation , were demonstrated to be secondary , i.e. , induced by the fact that ssfr in turn correlates with a variety of other global galaxy parameters .
we note that a comprehensive study based on 500 star forming galaxies at redshifts from z = 0 to 3 also supports our findings ( genzel et al .
2014 ) .
we also compared the results obtained from the cold gass galaxies to estimates of @xmath20 on 1 kpc scales using high - resolution co and sfr maps from the heracles survey .
the global t@xmath1ssfr relation derived from the cold gass sample and that derived from 1 kpc grids placed on heracles galaxies agree remarkably well .
this suggests that the local molecular gas depletion time is dependent on the local fraction of young - to - old stars and that galaxies with high current - to - past - averaged star formation activity , will consume their molecular gas reservoir sooner .
another aspect that has not been fully understood is the role of galaxy structures such as bulges , arms , bars , or rings in determining the molecular gas depletion time .
we note that simply using all available grids from the whole galaxy or integrated values of gas mass and sfr to study molecular gas depletion times might smooth out variations caused by different structures .
many studies have investigated whether the star formation efficiency ( sfe ) varies in differnt galactic environments .
for instance , meidt et al . (
2013 ) studied the inner region of the spiral galaxy m51 with a high resolution co ( 1 - 0 ) map ( @xmath1640 pc ) and proposed that streaming motions , induced by gravitational instabilities due to bars and arms , cause variations in gas depletion time .
momose et al .
( 2010 ) observed a barred spiral galaxy , ngc 4303 , with co ( 1 - 0 ) observations and found that its arms have twice higher sfes than its bar , although strong co emission is both seen in the bar and the disc region along the arms .
fujimoto et al . ( 2014 ) proposed that these results could be understood in a model where star formation is regulated by collisions between molecular clouds .
however , other studies have come to the opposite conclusion about the viability of the cloud collision model ( foyle et al .
2010 ; eden et al .
other observational studies using co(1 - 0 ) line luminosities as the gas tracer and h@xmath21 or 24@xmath13 luminosities as the sfr tracer have demonstrated that the sfe in spiral arms is enhanced compared to regions outside the arms ( e.g. , lord & young 1990 ; knapen et al .
1996 ; rebolledo et al . 2012 ; hirota et al . 2014 ) .
we note that most of these studies explored the influence of structures on sfe based on samples of one or only a few galaxies , and they did not consider variations in sfe with respect to the t@xmath4ssfr relation . in this paper
, we try to understand whether galaxy structures induce significant variation in molecular gas depletion time by systematically separating the grids into those in bulge , arm , bar , and ring regions , and by studying how the local depletion time from these different regions varies with respect to the established t@xmath4ssfr relation .
our paper is organized as follows .
we describe the multiwavelength data sets used in this work in section 2 , and our methods for identifying galaxy structures and deriving t@xmath1 and ssfr on local scales in section 3 .
our results are presented in section 4 , where we look at the t@xmath1ssfr relations for the grids in the bulge , arm , bar , and ring regions respectively .
finally , we compare our results with previous studies and discuss our findings in section 5 .
in this section we introduce the data sets we use in this work .
the heracles sample provides spatially - resolved co maps with resolution @xmath22 1kpc , appropriate for pixel - by - pixel studies . because of this good resolution , we are able to identify the pixels lying in galaxy structures such as bulges or spiral arms .
atlas@xmath0 galaxies are mostly early - type galaxies and are included because this data supplements the number of data points we are able to extract from the bulge - dominated regions of galaxies .
the cold gass project targets more distant sources and therefore provides only integrated values of the molecular gas mass .
nevertheless , the large sample size gives us the opportunity to carry out a statistical analysis of global relations between gas depletion time and other galaxy properties . to study molecular gas depletion time on sub - kpc scales , we use public data from the hera co - line extragalactic survey ( heracles ; @xcite ) .
heracles has released co ( @xmath23 ) maps for 48 nearby galaxies , achieving a spatial resolution @xmath1613 arcsec and an average h@xmath24 surface density detection limit of @xmath163 m@xmath9pc@xmath11 .
we analyze 21 massive galaxies with log ( m@xmath25/m@xmath9 ) @xmath26 10 from the catalog in @xcite so that we avoid those low - mass galaxies with less reliable @xmath27 mass measurements .
all our selected galaxies are located within a distance of @xmath1620 mpc .
a variety of ancillary data is available for the heracles sample , which allow us to derive the sfr and stellar mass surface densities .
the relevant data includes fuv images from galex all - sky imaging survey ( ais ) and nearby galaxy survey ( ngs ; @xcite ) , 24 @xmath13 data from the spitzer infrared nearby galaxies survey ( sings ; @xcite ) , hi maps from the hi nearby galaxies survey ( things ; @xcite ) , and optical images from sloan digital sky survey ( sdss ) .
fuv images from galex satellite have effective wavelength 1528 and angular resolution @xmath164.3 arcsec fwhm .
sings provides mips 24@xmath28 m images with an angular resolution @xmath166 arcsec and 3 @xmath29 sensitivity @xmath160.21 mjy sr@xmath10 .
the maps ( natural - weighting ) from things have an angular resolution of @xmath1611 arcsec and are sensitive to @xmath30 @xmath26 0.5 m@xmath9 pc@xmath11 . combining these public data sets , we estimate gas , sfr , and stellar mass surface densities as a function of position in the galaxy by dividing the galaxy into a set of square cells with 1 kpc@xmath2 size . in this way , we are able to separate the grids within bulges , arms , bars , rings from those in the rest of the galaxy .
galaxy parameters such as distance , inclination angle , position angle are taken directly from @xcite . to increase the amount of data in bulge regions , we make use of a subset of the galaxies observed as part of the atlas@xmath0 project @xcite , with publicly - available resolved interferometric co(1 - 0 ) maps @xcite .
the full atlas@xmath0 sample consists of a volume - limited sample of 260 nearby early - type ( elliptical e and lenticular s ) galaxies ( etgs ) with a cut - off absolute magnitude of m@xmath31 = @xmath1921.5 .
the atlas@xmath0 galaxies were observed in co(1 - 0 ) and co(2 - 1 ) by the iram 30 m telescope @xcite , and those with detected co integrated fluxes larger than 19 jy kms@xmath10 were imaged using the combined array for research in millimeter astronomy ( carma ) @xcite .
this data set includes 30 etgs .
together with 10 co - detected galaxies that have interferometric co(1 - 0 ) data from the literature , there are a total 40 etgs with resolved co maps in the atlas@xmath0 sample .
the reader is referred to @xcite for further details about the sample , the observations and the analyses .
we derive sfr using the combination of fuv emission from galex data and mid - infrared emission at 22 from wise data .
the fuv maps are retrieved from galex data release 7 products .
when several atlas images from different observing programs are available , we always use the one with the longest exposure time . the exposure time for @xmath162/3 of the sample is longer than 1500 seconds ; the rest of galaxies have data from the ais program with typical exposure time @xmath16100 200 seconds and limiting magnitude @xmath1620.5 ab magnitude .
there are 3 galaxies without fuv observations from galex .
wise provides 22 images of the whole sky , with angular resolution 12 arcsec and 5@xmath29 point - source sensitivity @xmath16 6mjy .
all galaxies have wise observations .
the stellar mass is derived from sdss 5-band images using a sed - fitting method ; the sdss images are downloaded from sdss data release 9 . in figure 1
we compare the basic properties of the galaxies in the heracles and atlas@xmath0 samples . generally , their stellar masses are comparable .
the heracles galaxies are much closer than the atlas@xmath0 galaxies , so the angular disc sizes ( r@xmath32 ) of the former sample are larger than the latter sample .
the distributions of the concentration index , r@xmath33/r@xmath34 , which is tightly related to the bulge - to - disc ratio of the galaxy ( @xcite ; @xcite ) , reflect the fact that the atlas@xmath0 galaxies are mostly etgs and the heracles sample consists of mostly disc galaxies .
( r@xmath32 ) , concentration index ( r@xmath33/r@xmath34 ) for heracles galaxies ( black line ) and atlas@xmath0 galaxies ( red line ) . ]
the sample is drawn from the cold gass survey catalogue @xcite , which contains co ( @xmath35 ) line measurements from the iram 30 m telescope for @xmath16360 nearby galaxies with stellar masses in the range @xmath36 and redshifts in the range @xmath37 .
we only select the galaxies with detected co fluxes in our studies .
the reader is referred to @xcite for a detailed description of the sample selection and the observations . in the same fashion as was done for the atlas@xmath0 sample , we derive sfr by summing galex fuv and wise 22@xmath13 luminosities and we estimate stellar mass by fitting stellar population synthesis models to sdss 5-band images .
we retrieve the fuv maps from galex data release 7 products ; the maps with the longest exposure time are always preferred .
most of our fuv images are from the medium imaging survey ( mis ) with typical exposure time @xmath161500 seconds ; 81 out of 366 galaxies only have data from the ais program with typical exposure time @xmath16100 200 seconds .
the limiting magnitude for mis is 23.5 ab magnitude and for ais , it is 20.5 ab magnitude .
all galaxies have both wise and sdss imaging data . sample .
all maps are convolved to 13 arcsec resolution .
the blue ellipses and red circles indicate the aperture derived from co(1 - 0 ) map and the half - light radius r@xmath38 in k - band taken from cappellari et al .
the green bar shows the physical scale of 1kpc in each map . ]
molecular gas depletion time is defined as @xmath39/@xmath40 or m@xmath41/sfr , where @xmath42 and @xmath40 are molecular gas and sfr surface densities , and m@xmath43 is molecular gas mass .
specific sfr ( ssfr ) is defined as sfr / m@xmath44 or @xmath40/@xmath45 , where @xmath45 is stellar mass surface density .
we use the first definition for the spatially - resolved heracles sample ; the second definition is used for the atlas@xmath0 and cold gass samples to derive an integrated depletion time and ssfr . for the heracles sample ,
the spatial resolution of the co maps is @xmath1613 , which is @xmath221kpc .
the galex fuv and spitzer 24@xmath13 maps have better resolution , so we degrade them to the same resolution as the co maps . the sizes of structures we want to study , such as bulges , arms , bars and rings , are well - resolved with 1kpc smoothing .
we therefore use 1kpc @xmath46 1kpc grid cells as our sampling elements .
atlas@xmath0 includes a subset of @xmath1640 etgs with resolved co maps and supplements data for bulge regions .
given that the size of the co - emitting regions is only a few kpcs or even smaller and that the gas surface density sensitivity limit varies from one galaxy to another for the atlas@xmath0 project , we do not derive the depletion time within 1kpc@xmath47 grids , but within the co - emitting region .
we first smooth co maps to the same 13 resolution as the wise image .
we run sextractor on the co moment0 maps and use the kron aperture as the size of co - emitting region .
the co flux within the kron aperture is converted to the molecular gas mass by applying a galactic co - to - h@xmath24 conversion factor .
we note that we measure the sfr and the stellar mass using the same aperture as the co emission .
an example of our multi - band images with the aperture derived from the co map is shown in figure 2 .
the co emission for the cold gass galaxies is observed within the iram 22@xmath17 gaussian beam .
once again , sfr must be measured within the same region where molecular gas is observed ; thus , we derive sfr for the cold gass galaxies by placing a 22 gaussian beam on the fuv and 22@xmath13 maps .
as pointed out in @xcite , the depletion timescales that we derive for the cold gass galaxies agree better with the heracles results than those in saintonge et al ( 2011b ) . in this section
, we fit srsic models to sdss g - band images and identify bars / rings / spiral arms from the residual images .
after the main features of galaxies , i.e. the bulge and the disc , are removed , the residual maps can reveal fine structures such as rings or arms more clearly .
we first mask stars and background galaxies in the sdss images using the colour - based masks described in @xcite .
we inspect the images again and mask stars by hand when necessary .
the masked regions are marked as bad pixels which will not be used for fitting .
the maps of @xcite are used to correct for galactic extinction .
we next apply a two - dimensional image decomposition program called galfit @xcite to the sdss g - band images to decompose galaxies into bulges and disks .
to fit the luminosity profile of the galaxies , the code assumes a two - component model , srsic bulge plus srsic disk profile , where the srsic @xmath48 values for the bulge and disk are allowed to vary from 1.5 to 4 and from 0.8 to 1.2 , respectively .
the grids within the effective radius of the bulge model are designated as the bulge grids .
one example of the fitting result is shown in figure 3 .
the surface brightness profile of the galaxy is clearly separated into two components .
the effective radii of the bulge and disc are 0.8 and 8.0 kpc .
we identify seven galaxies with large arms , three barred galaxies , and one galaxy with a ring from the residual maps .
we visually select grids within the arm , bar and ring regions .
examples are shown in figure 4 .
note that we exclude those grids with @xmath15 below the co detection sensitivity limit or @xmath49 m@xmath9 yr@xmath10 kpc@xmath11 , because the sfr calibration does not work for these low values ( see section 3.4.1 ) . as can be seen
, we are able to sample between @xmath50 grid cells for each galaxy , with the exception of some of the barred galaxies and the ring galaxy .
the reason why there are only a few grids on these galaxies is that their @xmath15 values are generally low and thus many grids are discarded . for the resolved heracles maps , molecular gas depletion time
is defined as @xmath52 .
we calculate the molecular gas depletion time in 1-kpc @xmath46 1-kpc grid cells .
the fluxes from the reduced heracles co maps are converted to @xmath53 in each 1-kpc grid cell adopting the galactic co - to - h@xmath12 conversion factor , 4.35 m@xmath9 pc@xmath11 ( k km s@xmath10)@xmath10 .
note that we discard the bins with @xmath53 @xmath54 3 m@xmath9 pc@xmath11 , which is the sensitivity limit for the heracles co maps .
for the cold gass galaxies , the observed co fluxes in the catalog of @xcite are converted to the molecular gas mass . for atlas@xmath0 , the co flux within the kron aperture is converted to the molecular gas mass . in both cases ,
the galactic co - to - h@xmath12 conversion factor is used .
we estimate sfr by combining the fuv and ir data .
fuv traces the emission from unobscured massive stars formed over the past @xmath16100myr .
uv photons absorbed by the surrounding dust cause emission at ir wavelengths .
the linear combination of the fuv and ir emission can recover both the unobscured stellar emission and the dust - reprocessed emission .
we use the fuv and 24 images to estimate @xmath7 .
the same method applied to the optical images is used to mask stars and background galaxies .
we re - fill the masked region with the local background .
we correct the fuv images for galactic extinction using the maps of @xcite .
all the images are convolved to a resolution of gaussian 13 using the kernels released in @xcite .
next we apply the method suggested in section 8.2 of @xcite to remove the cirrus contribution originating from evolved stellar populations .
a first - order correction is done using the total gas surface density , i.e. the sum of @xmath15 and @xmath30 , where @xmath30 is obtained from things data . after we remove this cirrus emission from the 24 @xmath13 emission , we then adopt the updated calibration coefficient in @xcite to calculate the sfrs from the linear combination of the fuv and 24 @xmath13 luminosities . those bins with
@xmath55 @xmath54 10@xmath56 m@xmath9 yr@xmath10 kpc@xmath11 are excluded in the following analysis , because the sfr calibration becomes poor in the very low sfr regime @xcite . sample .
only galaxies with 22 flux s / n@xmath57 3 are shown .
the black solid line is the best - fit linear regression to the co non - detected galaxies .
the blue plus symbols represent galaxies excluded during the fitting process ( see text ) . ]
we use the fuv and 22 images to estimate sfrs within the aperture derived from the co maps .
both the galex fuv and wise 22 images are convolved to 13 resolution using the kernels from @xcite .
the steps to process fuv and wise 22 images including star / background galaxy masking , sky removal , galactic foreground extinction correction and wise color correction , are the same as described in @xcite .
as the atlas@xmath0 galaxies are etgs , the contribution from old stellar populations to the ir emission is significant ( e.g.,kennicutt 1998 ) .
@xcite estimate that the average fractional contribution from evolved stars to the 22 luminosities of etgs is @xmath1625% .
these authors found a tight correlation between k - band and 22 luminosities for a subset of galaxies without co detections and used this relation to correct the total 22 emission for the contribution from old stars .
we retrieve 2mass k - band atlas images from the nasa / ipac infrared science archive for both co - rich and co non - detected galaxies from the atlas@xmath0 survey . to remove stars or background galaxies , we run sextractor and mask detected sources , with the exception of the central galaxies .
we estimate the sky background as the median value of the pixels with values @xmath54 3@xmath29 above the median value of the whole image .
the k - band images are then convolved to the same 13 resolution as the wise 22 images .
we derive k - band and 22 luminosities for the whole galaxy using the kron aperture in sextractor .
the k - band absolute magnitude is plotted against 22 luminosity for co - rich and co non - detected etgs in figure 5 . as can be seen
, the k - band absolute magnitude correlates well with the 22 luminosity for co non - detected galaxies ( black points ) , while there is no clear relation for the co - rich galaxies ( magenta points ) .
we perform a linear fit to the co non - detected galaxies and plot the result with a black solid line in figure 5 .
the best - fit is given by @xmath58 note that there are 5 outliers ( blue plus signs in figure 5 ) with scatter larger than 0.5 dex .
they include one galaxy with apparent spiral arms , one with a star - forming ring , one with a high hi mass fraction , and two with either a star or small galaxy nearby .
we exclude these galaxies from our fits .
we measure the k - band magnitudes of co - rich etgs within the aperture derived from co maps , estimate the 22 luminosity from old stars using equation ( 1 ) , and subtract this from the total 22 emission .
we then scale the wise 22 fluxes so that they have the same normalization as the mips 24 fluxes @xcite .
we convert the fuv and 22 luminosities to sfrs using the same calibration given in @xcite . as in the previous section ,
we use galex fuv and wise 22@xmath13 images to estimate sfr for cold gass galaxies within the central 22 region .
the procedure is the same as mentioned above . for consistency ,
we use the same calibration as we apply to heracles and atlas@xmath0 sample to convert fuv plus 22@xmath13 into sfr .
stellar mass is derived using the sed - fitting method in @xcite by which the stellar population synthesis models of @xcite are fit to sdss 5-band magnitudes .
the parameter , ssfr , is the sfr divided by the stellar mass .
we plot the depletion time and the ratio of sfr@xmath59 to sfr@xmath60 ( ir / uv ) as a function of radius for bulge and disc grids in figure 6 .
whereas the depletion time of the bulge grids drops towards the center of the galaxy , the depletion time of the disc grids remains almost constant as a function of radius .
in contrast , the ir / uv ratios of the disc and bulge grids decline from the center towards the outskirts this is not surprising because the ir / uv ratio is a robust indicator of dust attenuation , and the outer parts of galaxies are expected to have lower metallicities and dust content .
we note that the fact that different radial trends are found for depletion time and for ir / uv ratio make it seem unlikely that variations in co - to - h@xmath12 conversion factor are responsible for the drop in the depletion time in the central bulge .
we note that this result is consistent with results presented in @xcite ( see section 5 for further discussion ) . as in the previous section
, we use galex fuv and wise 22@xmath13 images to estimate sfr for cold gass galaxies within the central 22 region .
the procedure is the same as mentioned above . for consistency
, we use the same calibration as we apply to heracles and atlas@xmath0 sample to convert fuv plus 22@xmath13 into sfr .
stellar mass is derived using the sed - fitting method in @xcite by which the stellar population synthesis models of @xcite are fit to sdss 5-band magnitudes .
the parameter , ssfr , is the sfr divided by the stellar mass .
we plot the depletion time and the ratio of sfr@xmath59 to sfr@xmath60 ( ir / uv ) as a function of radius for bulge and disc grids in figure 6 . whereas the depletion time of the bulge grids drops towards the center of the galaxy ,
the depletion time of the disc grids remains almost constant as a function of radius .
in contrast , the ir / uv ratios of the disc and bulge grids decline from the center towards the outskirts this is not surprising because the ir / uv ratio is a robust indicator of dust attenuation , and the outer parts of galaxies are expected to have lower metallicities and dust content .
we note that the fact that different radial trends are found for depletion time and for ir / uv ratio make it seem unlikely that variations in co - to - h@xmath12 conversion factor are responsible for the drop in the depletion time in the central bulge .
we note that this result is consistent with results presented in @xcite ssfr relation for the bulge and disc grids of the heracles sample , as well as the atlas@xmath0 galaxies .
grey and blue solid points show grids in the disc and bulge regions of the heracles galaxies .
red solid points show the atlas@xmath0 sample .
ols bisector fits to the disc , bulge and atlas@xmath0 data points are shown as grey , blue , and red dashed lines .
the blue dotted line shows the ols bisector fit to the bulge grids , excluding those those with r @xmath54 0.1@xmath61 .
the green dashed line is the ols bisector fit to the heracles bulge data points plus the results from the atlas@xmath0 sample . ] in figure 7 , we investigate the t@xmath4ssfr relation for bulges ( blue solid points ) and discs ( grey solid points ) regions .
we also plot the atlas@xmath0 bulge - dominated sample for comparison ( red solid points ) .
we perform ols - bisector fits to the data and plot the fitting results in figure 7 as well .
there are two obvious results : first , the scatter in depletion time is larger in bulges than in discs ( 0.33 dex in @xmath62 , compared to 0.22 for discs ) . excluding bulge grids with r @xmath54 0.1 r@xmath63 , which may be influenced by agn , makes no difference to this conclusion .
second , in the bulge , t@xmath1 is systematically shifted to smaller values at a given ssfr compared to the disc grids .
we note that there are some bulge grids whose depletion times lie above the grey line derived for disc grids in figure 7 .
we check the location of these grids and find that they are generally in the outer region of the bulge . in figure 8 , we plot the t@xmath4ssfr relation for galaxies with grand - design arms , bars and rings .
magenta points denote the grids located within the region of the targeted structure , such as the grand - design arm , bar or ring ; blue points denote other grids in the same galaxies .
grids from other galaxies , i.e. those galaxies that do not contain the designated structure , are shown in grey .
the ssfrs in the interstellar medium ( ism ) of galaxies with grand - design arms are higher than in galaxies without arms , and the ssfrs within the arms themselves are the highest .
this is expected , because gas is compressed by spiral density waves and star formation rate scales with gas surface density .
interestingly , t@xmath1 is shifted to higher values at a given ssfr in galaxies with spiral arms .
we apply ols - bisector fits to the spiral arm data and the results are shown as dashed lines in figure 8 . indeed , the grids in the regions with ( blue dashed line ) or without ( magenta dashed line ) arms are systematically shifted to longer t@xmath1 at fixed ssfr compared to the grey dashed line , which is for galaxies without spiral structure . in barred galaxies ,
we find the opposite trend .
the ssfrs in the interstellar medium of galaxies with bars ( magenta points in the middle panel ) are much lower than in galaxies without bars .
bars are long - lived structures .
although they can act to channel gas towards the center of a galaxy and fuel starbursts , this is a transient phenomenon and the majority of present - day barred galaxies are relatively quiescent .
the ols - bisector fit to the grids in barred galaxies indicates that t@xmath1 is shortened at fixed ssfr compared to grids from galaxies without bars .
this is exactly the opposite to what was found for galaxies with spiral arms .
there is only one galaxy with a ring in heracles sample , so we only have a few grids available for analysis . based on the limited data , we conclude that ringed galaxies follow the same trends as barred galaxies .
so far , we have analyzed molecular gas depletion time variations in terms of location within or outside specific structures of the galaxy , such as bulges , bars , arms or rings .
we now ask whether all of these variations can be linked to a physical parameter that varies continuously within galaxies .
for example , daddi et al . ( 2010 ) and genzel et al . (
2010 ) proposed that the combination of high gas density and short dynamical timescale could explain the short global molecular gas consumption timescales in strongly starbursting galaxies .
the parameters that we examine in this section are , 1 ) ir / uv ratio as a proxy for dust content , 2 ) h@xmath12 surface density , 3 ) stellar surface density . * _ ir / uv ratio _ * in figure 9 , we plot the distributions of ir / uv ratios for inner bulge , outer bulge and disk grids in the top left panel . in the top right panel , we plot distributions for spiral arm , inter - arm and control grids .
the bottom two panels are the same as the top right panel , except that results are shown for bars and rings .
as we have seen previously , ir / uv ratios are highest in the inner regions of bulges , from which we infer that these are the dustiest regions of nearby galaxies .
the high dust content in the centers of bulges means that it is unlikely that we have under - estimated the h@xmath12 content on the galaxy by using the galactic co - to - h@xmath12 conversion factor .
the short depletion times observed in central bulges are thus not likely to be an artifact of wrongly estimated molecular gas masses .
the least dusty regions of the ism are found within galaxies with grand - design spiral arms . by a similar argument
, we infer that that higher molecular gas depletion times observed in these regions are not likely to be an artifact of conversion factor variations either .
grids within bars in barred galaxies have similar ir / uv ratio as the outer parts of galactic bulges , while those outside bars have ir / uv ratios similar to the general disc population .
grid within rings appear less dusty than grids within bars , but we caution that our sample consistent of only one ringed system . in summary , we deem it unlikely that variations in co - to - h@xmath12 conversion factor are responsible for the different t@xmath4ssfr relations among grid cells lying within different structures . *
_ h@xmath12 surface density _ * figure 10 is the same as figure 9 , except that distributions of h@xmath12 surface density are shown .
if we exclude the single ringed galaxy , the only significant shift in h@xmath12 surface density occurs within bulges , which have molecular gas densities up to a factor of 30 higher than discs , particularly in their inner regions .
the enhancement in gas density in the grids with spiral arms is only about 50% , and grids that lie outside spiral arms have h@xmath12 densities that are identical to grids with galaxies without arms .
gas density is enhanced more strongly within bars , but still not as strongly as in bulges . in conclusion , although high gas densities may be a promising avenue for explaining the short depletion time in bulges and starburst systems , it does not appear to vary in a systematic way in other environments such as spiral arms and bars , so as to explain other observed trends .
we will demonstrate this in more detail later in this section .
* _ stellar surface density _ * in figure 11 , we plot the distributions of stellar surface density for the grids within different structures . in this case , there is a very clear separation between grid cells in bulges , bars and rings , which peak at log @xmath64 m@xmath9 pc@xmath11 while grids in the spiral - arm galaxies peak at log @xmath65 m@xmath9 pc@xmath11 .
grids in the inner regions of bulges , where @xmath20 is shortest , are shifted to even higher @xmath45 than the outer bulges .
spiral arm and inter - arm grids also peak at lower @xmath45 values than grids from galaxies without spiral arms . in summary , variations in the @xmath45 distributions between different galaxy structures mirror the variations in the @xmath20-ssfr relations described in the previous section extremely well .
we now fit a linear relation to the t@xmath4ssfr relation for all grid data points and plot the residuals against @xmath45 , @xmath39 and a third quantity , the interstellar pressure , in figure 12 . following @xcite , the pressure , p@xmath66 ,
is expressed as @xmath67 where @xmath68 and @xmath69 are velocity dispersions of gas and stars , and @xmath70 and @xmath45 are the gas and stellar surface densities .
data from atlas@xmath0 and cold gass are also added in this figure .
figure 12 shows that @xmath45 is the only quantity that clearly correlates with @xmath71 .
grids in bulges , bars , and rings and the atlas@xmath0 sample clearly separate from the spiral arm , inter - arm and disc grids in this plot .
we can also understand the reason why we did not observe a correlation between @xmath71 and @xmath45 for the cold gass sample in @xcite : the dynamic range in @xmath45 of the cold gass sample is too small .
the integrated stellar surface densities of the cold gass sample span only 1 dex on the x - axis .
motivated by these results , we fit a plane to the two - dimensional relation between the depletion time , specific star formation rate and the stellar surface density . following the method in bernardi et al .
( 2003 ) , we find the best - fit linear correlation of the form @xmath72 we plot all the data , including the grids from heracles , the atlas@xmath0 sample and the cold gass sample in figure 13 , and we show the t@xmath4ssfr relation in the next panel for comparison .
the 1:1 relation is shown by a black solid line .
the values of the coefficients , a , b , and c from the best - fit result are @xmath190.36 , @xmath190.14 , and 5.87 respectively .
the pearson linear correlation coefficient r between the @xmath73 and @xmath74 axes is 0.68 , and the scatter in t@xmath1 about the plane is 0.18 dex . the reduction in scatter , compared to the one - dimensional relation for t@xmath1 versus ssfr , which has a scatter of 0.19 , is only 5% .
this is because our sample is heavily dominated by the number of grids from the disc - dominated regions of galaxies .
the main effect of the two - parameter fit is to bring the outlying points , which are mainly from bulge and bar grid cells , into better agreement with the rest of the sample .
finally , we note that for 1 kpc@xmath2 grid cells , we can rewrite equation ( 3 ) above as @xmath75 this means that the surface density of evolved stars in a grid cell contributes @xmath161.4 times more weight to our prediction of depletion time than the surface density of newly formed stars ! we will speculate on possible reasons for this in the next section .
using the resolved maps of the nearby heracles galaxies , we have calculated the molecular gas depletion time in 1-kpc @xmath46 1-kpc grid cells and studied the t@xmath4ssfr relation for grids lying within different galaxy structures such as bulges , arms , bars , and rings .
our main results can be summarized as follows : \(i ) bulge regions have shorter gas depletion times than the disk regions at a given value of the ssfr .
this effect is strongest in the central regions of the bulge .
\(ii ) the t@xmath4ssfr relation for the grid cells in spiral - arm galaxies shifts to longer t@xmath1 at fixed ssfr compared to grid cells from galaxies without spiral arms . note that this shift applies to _ all grid cells _ from spiral arm galaxies . within a spiral arm ,
the ssfr values are higher than in the inter - arm region , so the average depletion times will be shorter .
( iii)grid cells located within bars have reduced depletion time at fixed ssfr .
grids located in barred galaxies outside of the barred region occupy a narrow range of low ssfr values .
\(iv ) we identify the stellar surface density @xmath76 as the parameter that can best predict the shifts between the relations found for bulges , bars and spiral arms .
our results on star formation efficiencies in spiral arms are largely consistent with those presented in figure 7 of foyle et al .
( 2010 ) and figure 13 of rebolledo et al .
( 2013 ) even though our methodology for identifying spiral arms differs from these papers .
several previous papers have claimed that star formation efficiencies in bars are low ( e.g. , reynaud & downes 1998 ; momose et al .
2010 ; sorai et al .
we note that in our sample , the molecular gas depletion time in barred regions spans a range of more than a factor of 10 . only in bars with high gas densities
do we find grid cells with short depletion times and high sfes .
this is consistent with the work of sheth et al .
( 2005 ) , who find that barred spiral galaxies generally have higher molecular gas concentrations in their central regions .
however , barred galaxies exist that have no molecular gas detected in the nuclear region and very little within the bar corotation radius .
bars are long - lives , and we likely observe them at different evolutionary stages , which explains why the sfe has such a large spread .
our result that bulge grids have reduced depletion time compared to disc grids is consistent with results presented in leroy et al .
these authors studied radial trends in molecular gas depletion time as a function of radius and observed that the average @xmath20 was smaller in the central @xmath161kpc regions .
the reduced depletion time was even more pronounced after they applied a co - to - h@xmath12 conversion factor that depended on dust - to - gas ratio .
once again , these authors did not study trends of depletion time at fixed specific star formation rate .
we note that davis et al .
( 2014 ) found that star formation efficiencies ( sfes ) of the etgs in atlas@xmath0 survey were _ reduced _ compared to spiral galaxies .
the main reason for this apparently discrepant result is the definition of star formation efficient adopted by these authors : sfe was computed as the ratio of the total ( atomic+molecular ) gas mass divided by the total star formation rate of the galaxy .
the analysis in this paper only considers the star formation efficiency of the molecular gas component of the galaxy .
we note that _ no previous study has considered variations in sfe at fixed specific star formation rate _ , which is the main focus of this paper why should the star formation efficiency of the molecular gas be so sensitive to the local surface density of evolved stars ?
helfer & blitz ( 1993 ) studied the properties of the dense molecular gas in the bulges of a sample of 19 normal spiral galaxies .
they observed the 3 mm emission from the molecules hcn and cs , which traces much denser gas than co. they found that the cs to co ratio in the external bulges was consistent with that measured for the bulge of the milky way and a factor of 2 or more larger than in the galactic disc . in subsequent work ,
helfer & blitz ( 1997 ) carried out interferometric observations of emission from the hcn molecule in ngc 6946 , ngc 1068 , and the milky way and found that the ratio of hcn to co at the galactic centres was 5 - 10 times higher than in the discs of these galaxies .
helfer & blitz ( 1993 ) put forward two possible explanations for their findings : 1 ) because the stellar potential is larger in the bulge than in the disc , the gas is subject to much higher pressures , 2 ) the giant molecular clouds ( gmcs ) have different mass and/or structure in the bulge than in the disc .
launhardt , zylka & mezger ( 2002 ) emphasized that most of the ism in the central bulge of the milky way is likely in the form of very dense , compact clouds .
these conclusions arise from the lack of a significant diffuse component in high resolution submm maps of the galactic centre .
they speculate that the physical reason for the observed clumpiness of the ism is the tidal stability limit clouds have to be dense enough to be stabilized against tidal forces by their own gravitation .
the results in this paper suggest that there may be a _
continuum _ of molecular cloud properties set by the local stellar surface density .
if so , this would have very interesting consequences for our understanding of galaxy evolution .
gas - to - evolved star ratios are much higher in the early universe than they are today ( tacconi et al 2010 ) and star formation may likewise be proceeding in a quite different manner inside gmcs .
we thank linda tacconi , reinhard genzel , jing wang , richard dsouza , sambit roychowdhury for helpful discussions . this research has made use of the nasa / ipac infrared science archive , which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration .
funding for the sdss and sdss - ii has been provided by the alfred p. sloan foundation , the participating institutions , the national science foundation , the us department of energy , the national aeronautics and space administration , the japanese monbukagakusho , the max planck society , and the higher education funding council for england .
the sdss is managed by the astrophysical research consortium for the participating institutions .
the participating institutions are the american museum of natural history , astrophysical institute potsdam , university of basel , cambridge university , case western reserve university , university of chicago , drexel university , fermi national accelerator laboratory , the institute for advanced study , the japan participation group , johns hopkins university , the joint institute for nuclear astrophysics , the kavli institute for particle astrophysics and cosmology , the korean scientist group , the chinese academy of sciences ( lamost ) , the leibniz institute for astrophysics , los alamos national laboratory , the max - planck - institute for astronomy ( mpia ) , the max - planck - institute for astrophysics ( mpa ) , new mexico state university , ohio state university , university of pittsburgh , university of portsmouth , princeton university , the us naval observatory and the university of washington . 99 aniano g. , draine b. t. , gordon k. d. , sandstrom k. , 2011 , pasp , 123 , 1218 alatalo k. et al . , 2013 , mnras , 432 , 1796 bernardi m. et al . , 2003 , aj , 125 , 1866 bigiel f. , leroy a. k. , walter f. , brinks e. , de blok w. j. g. , madore b. , thornley m. d. , 2008 , aj , 136 , 2846 bigiel f. et al . , 2011 ,
apj , 730 , l13 bruzual g. , charlot s. , 2003 , mnras , 539 , 718 calzetti d. , 2013 , star formation rate indicators .
cambridge univ . press , cambridge ,
cappellari m. et al . , 2011 ,
mnras , 413 , 813 daddi e. et al . , 2010 ,
apjl , 714 , 118 davis t. a. et al . , 2014 , mnras , 444,3427 dib s. , helou g. , moore t. j. t. , urquhart j. s. , dariush a. , 2012 , apj , 758 , 125 eden d. j. , moore t. j. t. , morgan l. k. , thompson m. a. , urquhart j. s. , 2013 , mnras , 431 , 1587 foyle k. , rix h .- w . , walter f. , leroy a. k. , 2010 , apj , 725 , 534 fujimoto y. , tasker e. j. , habe a. , 2014 , mnras , 445 , l65 gadotti d. a. et al . , 2009 , mnras , 393 , 1531 genzel r. et al . , 2010 , mnras , 407 , 2091 genzel r. et al . , 2015 , apj , 800 , 20 gil de paz a. et al . , 2007 , apjs , 173 , 185 hao c
.- n . , kenicutt r. c. jr . , johnson b. d. , calzetti d. , dale d. a. , moustakas j. , 2011 , apj , 741 , 124 helfer t. t. , blitz l. , 1993 , apj , 419 , 86 helfer t. t. , blitz l. , 1997 , apj , 478 , 162 hirota a. et al . , 2014 , pasj , 66 , 46 huang m .-
l . , kauffmann g. , 2014 , mnras , 443,1329 kennicutt r. c. jr . , 1998 , ara&a , 36 , 189 kennicutt r. c. jr .
et al . , 2003 , pasp , 115 , 928 knapen j. h. , beckman j. e. , cepa j. , nakai n. , 1996 , a&a , 308 , 27 launhardt r. , zylka r. , mezger p. g. , 2002 , a&a , 384 , 112 leroy a. k. et al . , 2008 ,
aj , 136 , 2782 leroy a. k. et al . , 2012 ,
aj , 144,3 leroy a. k. et al . , 2013 ,
aj , 146,19 lord s. d. , young j. s. , 1990 , apj , 356 , 135 muoz - mateos et al . , 2009 , apj , 703,1569 meidt s. e. et al . , 2013 , apj , 779 , 45 momose r. , okumura s. k. , koda j. , sawada t. , 2010 , apj , 721 , 383 momose r. et al .
, 2013 , apj , 772 , l13 pan h .- a . , kuno n. , hirota a. , 2014 , pasj , 66 , 27 peng c. y. , ho l. c. , impey c. d. , rix h .- w .
et al . , 2002 , aj , 124,266 rebolledo d. , wong t. , leroy a. , koda j. , meyer d. j. , 2012 , apj , 757 , 155 reynaud d. , downes d. , 1998 , a&a , 337 , 671 saintonge a. et al . , 2011a , mnras , 415 , 32 saintonge a. et al .
, 2011b , mnras , 415 , 61 saintonge a. et al .
, 2012 , apj , 758 , 73 schlegel d. j. , finkbeiner d. p. , davis m. , 1998 , apj , 500,525 schmidt m. , 1959 , apj , 129 , 243 sheth k. , vogel s. n. , regan m. w. , thornley m. d. , teuben p. j. , 2005 , apj , 632 , 217 shetty r. , kelly b. c. , bigiel f. , 2013 , mnras , 430 , 288 sorai k. et al . , 2012 , pasj , 64 , 51 tacconi l. j. et al . , 2010 , nature , 463 , 781 walter f. , brinks e. , de blok w. j. g. , bigiel f. , kennicutt r. c. jr . , thornley m. d. , leroy a. k. , 2008 , aj , 136 , 2563 wang j. et al .
, 2011 , mnras , 412 , 1081 weinmann s. m. , kauffmann g. , van den bosch f. c. , pasquali a. , mcintosh d. h. , mo h. , yang x. , guo y. , 2009 , mnras , 394 , 1213 young l. m. et al . , 2011 , mnras , 414 , 910
|
we use a data set of nearby galaxies drawn from the heracles , atlas@xmath0 , and cold gass surveys to study variations in molecular gas depletion time ( t@xmath1 ) in galaxy structures such as bulges , grand - design spiral arms , bars and rings .
molecular gas is traced by co line emission and star formation rate ( sfr ) is derived using the combination of far - ultraviolet and mid - infrared ( mir ) data .
the contribution of old stars to mir emission for the atlas@xmath0 sample is corrected using 2mass k - band images .
we apply a two - dimensional image decomposition algorithm to decompose galaxies into bulges and discs .
spiral arms , bars and rings are identified in the residual maps , and molecular gas depletion times are derived on a square grid of 1 kpc@xmath2 size . in previous work , we showed that t@xmath1 correlates strongly with specific star formation rate ( ssfr ) .
we now find that at a given ssfr , the bulge has shorter t@xmath1 than the disc .
the shift to shorter depletion times is most pronounced in the inner bulge ( @xmath3 ) .
grids from galaxies with bars and rings are similar to those from galactic bulges in that they have reduced t@xmath1 at a given ssfr .
in contrast , the t@xmath1 versus ssfr relation in the discs of galaxies with spiral arms is displaced to longer t@xmath1 at fixed ssfr .
we then show that the differences in the t@xmath4ssfr relation for bulges , discs , arms , bars and rings can be linked to variations in _ stellar _
, rather than gas surface density between different structures .
our best current predictor for t@xmath1 , both globally and for 1 kpc grids , is given by @xmath5 .
[ firstpage ] galaxy formation
|
in non-@xmath1-wave cooper pairing states , quasiparticle scatterings from system surfaces and nonmagnetic impurities can lead to suppression of the order parameter .
one typical example of such anisotropic pairing states would be the spin - triplet @xmath2-wave one which is realized in superfluid phases of liquid @xmath0he .
in contrast to superconductors whose properties are closely connected to the underlying crystal symmetries , bulk superfluid @xmath0he is a highly clean homogeneous system , so that the cooper pair can potentially take all the possible symmetries within the spin - triplet @xmath2-wave sector . in reality , however , only two superfluid phases are realized in the bulk @xmath3he , b and a phases which are respectively identified with the balian - werthamer ( bw ) and the anderson - brinkman - morel ( abm ) pairing states @xcite .
the former is a full gapped state appearing in the wide range of the temperature - pressure phase diagram , and the latter with point nodes at both poles of the fermi sphere is stabilized by the strong - coupling effect only in the high - temperature and high - pressure region .
the most interesting question is whether or not pairing states other than the abm and bw ones may appear in superfluid @xmath0he confined in thin slabs , narrow cylinders , and porous media such as aerogels . since orbital degrees of freedom of the superfluid order parameter
are strongly restricted due to the scatterings from such foreign objects , the pure bulk states can be unstable in these geometries . in the bw state in a film geometry , surface - normal components of the order parameter are locally suppressed near specular system surfaces . as the film thickness becomes small ,
these components get smaller even inside the film and the bw state is eventually deformed into the planar pairing state the nodal structure of which is the same as that of the abm state @xcite . the relative stability between the planar - distorted bw pairing state and the abm one depends on the strength of the strong - coupling effect which varies as a function of pressure @xcite . in experiments ,
it has been observed that the stability region of the a phase is extended in thin slabs @xcite . on the other hand , in aerogels , interconnected silica strands work as nonmagnetic impurities for superfluid @xmath0he , giving rise to intriguing phenomena .
there , pinning effects of the @xmath4-vector in the abm state have been extensively studied @xcite , and the polar pairing state with a line node along the equator of the fermi sphere was predicted to appear in uniaxially stretched media @xcite .
recently , the existence of this polar phase has been experimentally confirmed in nematically - ordered aerogels @xcite .
also in narrow cylinders , the occurrence of the polar state is theoretically predicted @xcite , but so far , any signature of this state has not been observed in experiments @xcite .
the key ingredient common to the above cases is a global anisotropy .
it lowers the effective dimensionality of the system favoring the anisotropic pairing states . in such quasi - low dimensional systems
, one may naturally expect that the superfluid state is spatially uniform along scattering - free directions , namely , the in - plane and cylinder - axis directions for the slabs and cylinders , respectively . near the continuous transition from the bw state into the planer or polar states , however , the surface scattering yields periodic spatial structures along these scattering - free directions @xcite . from recent theoretical studies of the surface scattering effect on unconventional superconductors and superfluids , it is becoming clear that the surface - induced gap distortion triggers instability of spatially modulated cooper pairing states in quasi - low dimensional systems @xcite
they are analog of the fulde - ferrell - larkin - ovchinnikov ( fflo ) state @xcite which was originally studied in the context of spin - singlet superconductors in a strong zeeman field , and are characterized by a nonzero center of mass momentum of the cooper pair @xmath5 .
the occurrence of the fflo - like @xmath6 state in superfluid @xmath0he was first pointed out by vorontsov and sauls for a film with two specular surfaces @xcite . near the bw - planar transition in the film ,
the bw state exhibits a one - dimensional modulation characterized by @xmath5 within the film plane .
this modulated bw state is called `` stripe order '' . after their prediction
, it is shown that the stripe order may be possible also in cylindrical geometry @xcite . in narrow cylinders , however ,
the stripe order can be stable only when the system surface is specular and diffusive along the cylinder axis and rim , respectively .
this indicates that the stability of the stripe order is sensitive to whether the surface scattering is specular or diffusive . for the slab geometry ,
it is not clear whether the stripe order can survive for any scattering conditions or not ( quite recently , the strong - coupling effect and surface - roughness effect have been theoretically studied @xcite , and we will discuss these issues later on ) .
in addition , it is also an interesting question how the midgap states in the @xmath0he b phase are modified by the periodic stripe structure . in this paper
, we will consider thin films of superfluid @xmath0he and investigate ( 1 ) surface - roughness effects on the stripe order and ( 2 ) quasiparticle excitations in this spatially modulated phase . throughout this paper ,
we restrict ourselves to liquid @xmath0he at 0 bar which is considered to be in the weak - coupling limit @xcite .
it will be shown that the stripe order can exist even in a film with one specular and one diffusive surfaces and that the stripe structure induces a new gapped bound state distinctly different from the conventional surface andreev bound state .
the remainder of this paper is organized as follows : in sec.ii , we introduce the ginzburg - landau ( gl ) free - energy functional and examine the surface scattering effects on the bw pairing state .
the mechanism of the stripe order and its stability against surface roughness are discussed based on the gl theory . in sec.iii
, we investigate low - energy quasiparticle excitations in the stripe order by solving the eilenberger equation to obtain the local density of states ( ldos ) .
there , for simplicity , we have used the spatial profile of the order parameter obtained in sec.ii , i.e. , in the gl theory .
finally , we summarize our results in sec.iv .
in this section , we will discuss the stability of the stripe order in the bw pairing state confined in a quasi - two dimensional system with thickness @xmath7 .
the system geometry is shown in fig.[fig:4](a ) .
the two - dimensional plane is extending in the @xmath8-@xmath9 plane , and quasiparticles are scattered on the system surfaces located at @xmath10 .
hereafter , three patterns of surface conditions will be considered : ( a ) both surfaces are specular , ( b ) the upper surface is specular while the lower one is diffusive , and ( c ) both surfaces are diffusive .
the case ( a ) corresponds to the ideal superfluid @xmath0he film which can be realized by coating slab surfaces with @xmath11he @xcite .
the situations ( b ) and ( c ) describe superfluid @xmath0he on a rough substrate and in a slab without the @xmath11he coat , respectively .
we will first visit the simplest case ( a ) , which has been already discussed in @xcite , to clarify the mechanism of the stripe order near the bw - planer transition , and next examine the surface roughness effect . throughout this section
, we ignore the abm pairing state because it has the same condensation energy as that of the planar pairing state in the weak coupling limit .
the gap function of the spin - triplet @xmath2-wave cooper pairing state is generally given by @xmath12 with pauli matrices @xmath13 @xmath14 and complex variables @xmath15 which play a role of the order parameter of this system .
properties of liquid @xmath0he near the superfluid transition temperature @xmath16 are well described by the gl theory @xcite .
as our main focus is on pairing states of a superfluid film with its thickness @xmath7 less than the dipole length @xmath17 m , we will neglect the dipole interaction .
the corresponding functional is obtained as an expansion in the order parameter @xmath15 @xmath18 in the weak coupling limit , the coefficients are given by @xmath19 , @xmath20 , @xmath21 , and @xmath22 with @xmath23 as density of states per spin on the fermi surface and @xmath24 as the superfluid coherence length at @xmath25 .
note that this definition of @xmath26 is based on ref.@xcite and is different from that in ref.@xcite by the factor of @xmath27 .
the effect of the system surface can be incorporated by the boundary condition on the order parameter @xmath28 . for the specular surface scattering , the surface - normal component of the quasiparticle momentum @xmath29 changes its sign by the mirror reflection at @xmath30 , i.e. , @xmath31 , where @xmath32 denotes a unit vector normal to the surface . because the gap function should satisfy the equation @xmath33 at the surfaces ,
the surface - normal component of the order parameter must vanish at the system surface while parallel components should be unchanged .
thus , the boundary condition on @xmath28 at the specular surface reads @xmath34 when the surface is sufficiently rough such that quasiparticles are randomly scattered independent of the incident direction , the boundary condition eq.([eq : boundary_sp ] ) is replaced with the diffusive one , i.e. , @xmath35 for any @xmath36 and @xmath37 .
now , we turn to the instability of the stripe order with @xmath6 in the bw state . in the bw state in the slab geometry , the order parameter is usually assumed to be uniform in the two - dimensional plane ( @xmath8-@xmath9 plane ) , and its basic form is given by @xmath38 .
because of the boundary condition eq.([eq : boundary_sp ] ) , @xmath39 vanishes at @xmath10 .
it has been known that due to the surface - induced gap suppression , the bw state is gradually deformed into the planar pairing state @xmath40 as film thickness @xmath7 becomes small @xcite .
below , we will see that near the bw - planar transition @xmath41 , the bw state lowers the energy by introducing the spatial modulation @xmath5 in the @xmath8-@xmath9 plane and extends its stability region to higher temperatures . for brevity , we take into account spatial variations in the order parameter only along the @xmath8 direction , assuming that the modulation , namely , periodic stripe structure is introduced in the @xmath8 direction ( see fig.[fig:4](a ) ) . as @xmath8 components of the order parameter
are generally relevant to spatial variations in the @xmath8 direction , we start from the bw state of the form @xmath42 by expanding @xmath43 in a fourier series with respect to @xmath8 and picking up relevant leading order terms , we have the following form of the order parameter which can be continuously deformed into @xmath44 : @xmath45 where @xmath46 is the wave vector characterizing the spatial modulation of the order parameter in the @xmath8 direction , and @xmath47 has been dropped because its contribution is known to be negligibly small @xcite . as we will see below , not only the off - diagonal component itself @xmath48 but also the phase difference of @xmath49 between sine waves in @xmath48 and @xmath50 is important for the occurrence of the stripe order . in a film with thickness @xmath7 and specular surfaces .
@xmath41 for the bw state with a one - dimensional spatial modulation in the film plane ( solid curve ) is higher than that for the uniform bw state ( black dashed one ) .
inset shows the evolution of the characteristic wave number @xmath51 along the @xmath52 curve .
( b ) temperature dependences of @xmath53 ( green ) , @xmath54 ( blue ) , @xmath55 ( red ) , and @xmath51 ( black ) at @xmath56 .
( c ) the spatial profile of the order parameter along the @xmath8-axis in the striped bw state with @xmath57 .
[ fig:1 ] ] the mechanism of the stripe order in the cylindrical geometry has been already addressed in ref.@xcite .
the origin of the stripe order in film superfulid @xmath0he can be understood in a similar way .
we first consider the case with specular surface scattering and take the following trial state satisfying eq.([eq : boundary_sp ] ) , @xmath58 translational symmetry breaking in the bw state is signaled by finite values of @xmath51 .
inserting the expression eq.([eq : op_trial ] ) into eq.([eq : gl ] ) and integrating over @xmath59 and @xmath8 , we obtain @xmath60 , \nonumber\\ f_g & = & \delta_2 ^ 2 \big(\frac{3}{4 } \frac{\pi^2}{d^2}+ \frac{1}{4 } q^2 \big ) + \delta_3 ^ 2 \big ( \frac{1}{4}\frac{\pi^2}{d^2 } + \frac{3}{4}q^2
\big ) \nonumber\\ & - & \delta_2 \delta_3 \ , q \ , \frac{\pi}{d } \nonumber\\ & = & c_q \big ( q-\frac { \pi \delta_2 \delta_3}{2 c_q \ , d }
\big)^2 + \frac{\pi^2 \delta_2 ^ 2}{4 d^2}\big ( 3 + \frac{\delta_3 ^ 2}{\delta_2 ^ 2 } -\frac{\delta_3 ^ 2}{c_q}\big ) \end{aligned}\ ] ] with @xmath61 .
the coefficients are calculated as @xmath62 , @xmath63 , @xmath64 , @xmath65 , @xmath66 , @xmath67 , and @xmath68 .
it should be emphasized here that the gradient term linear in @xmath51 shows up . by minimizing @xmath69 with respect to @xmath51
, we obtain @xmath70 which is nonzero as long as @xmath54 and @xmath55 are nonzero .
the gradient energy for the optimal @xmath51 is lower than that for @xmath71 by @xmath72 .
thus , the system tends to introduce the modulation @xmath51 to lower @xmath69 leading to the stripe order .
this situation is in sharp contrast to conventional fflo states for which a @xmath51-linear gradient term does not exist and nonzero @xmath51 appears just because the coefficient of the @xmath73 term becomes negative at very low temperatures for strong zeeman fields @xcite .
the present system is , however , rather similar to non - centrosymmetric superconductors ( ncs ) with rashba spin - orbit coupling @xcite in a magnetic field where the so - called helical phase with a field - induced phase modulation @xmath74 is believed to be realized @xcite . in rashba - type ncs
, broken inversion symmetry allows a @xmath51-linear gradient term coupled with the field , and as a result , the phase modulation exists even at the superconducting transition temperature in the magnetic field . from the analogy to ncs , it is inferred that the modulated bw state would emerge from high temperatures near @xmath16 .
figure[fig:1](a ) shows the @xmath52 curves for the bw states of the form eq.([eq : op_trial ] ) with ( solid curve ) and without ( dashed one ) a one - dimensional modulation in the two - dimensional plane . the @xmath41 transition temperature is higher for @xmath75 than that for @xmath71 , which implies that the superfluid state with broken translational symmetry is stabilized as the lowest energy bw state .
it is striking that the stripe order with @xmath75 appears from relatively high temperatures near @xmath16 , while the conventional fflo state is realized only in the low temperature region . in the superfluid @xmath0he film ,
the @xmath51-linear term coupled with @xmath55 ( amplitude of @xmath48 ) yields the @xmath75 pairing state , and thus , the internal degrees of freedom of the order parameter play an essential role for the striped superfluid phase .
we also note that the @xmath51-linear term exists only when we have the phase difference of @xmath49 between sine waves in @xmath48 and @xmath50 . in fig.[fig:1](b )
, one can see that @xmath54 and @xmath55 grow up with the @xmath76 dependence suggestive of the second order bw - planar transition .
the inset of fig.[fig:1](a ) shows the evolution of the modulation @xmath51 along the @xmath41 transition curve .
@xmath51 develops with decreasing film thickness .
the result obtained here is valid only when the trial state eq.([eq : op_trial ] ) well describes the exact spatial profile of the order parameter . in particular , for rough surfaces , @xmath59-dependence of @xmath28 is not trivial . in the next subsection
, we will determine the stability region of the stripe order by numerically solving gl equations .
he b phase confined in film geometries with thickness @xmath7 .
the t - d@xmath77 phase diagram for two specular ( one specular and one diffusive ) surfaces is shown in ( a ) [ ( b ) ] .
the spatial profiles of the order parameter components along @xmath59-axis at the point @xmath56 and @xmath78 , which is denoted by the blue colored symbol * in ( a ) [ ( b ) ] , are shown in ( c ) [ ( d ) ] .
green , blue , and red curves in ( c ) and ( d ) denote the spatial variations of @xmath79 , @xmath80 , and @xmath81 , respectively . in ( a)-(d ) , the solid and dashed curves are obtained by numerically solving gl equations , whereas the black dotted ones in ( a ) and ( c ) are the results obtained by the variational method with the use of the trial functions eq.([eq : op_trial ] ) .
[ fig:2 ] ] here , we consider the three patterns of surface conditions ( a ) specular - specular , ( b ) specular - diffusive , and ( c ) diffusive - diffusive . to obtain the @xmath82 phase diagram in each case
, we will solve gl equations @xmath83 numerically under corresponding boundary conditions for various values of @xmath51 at fixed points of @xmath84 and @xmath7 , and determine the lowest energy state .
for the specular surface , the boundary condition eq.([eq : boundary_sp ] ) is imposed , while for the diffussive one , @xmath35 is used .
figures [ fig:2](a ) and ( b ) show the numerically obtained @xmath82 phase diagrams of the superfluid @xmath3he b - phase film with surfaces of type ( a ) and ( b ) , respectively . in the region sandwiched by red and orange curves ,
the striped bw state with the modulation @xmath85 is more stable than the uniform one with @xmath86 .
the upper boundary is determined from the single mode instability with the use of eq.([eq : bw_op_q ] ) , and the lower boundary denotes the transition from the uniform bw state into the inhomogeneous one consisting of two domains with opposite signs of @xmath50 . in obtaining the lower boundary ,
we have used the general form of @xmath28 , eq.([eq : bw_op ] ) . comparing figs.[fig:2](a ) and
( b ) , one can see that the stability region of the stripe order shrinks as the surface scattering becomes more diffusive .
figure[fig:2](c ) shows @xmath59-dependences of @xmath87 near the upper boundary of the stripe - order stability region in a film with two specular surfaces , where solid and dotted curves denote the results obtained by numerically solving gl equations and by the variational method with the use of the trial state eq.([eq : op_trial ] ) , respectively .
one can see that the trial state eq.([eq : op_trial ] ) well describes the spatial profiles of @xmath88 .
correspondingly , the stripe - planar transition curve obtained by solving gl equations is in quantitative agreement with the one calculated with the variational method based on eq.([eq : op_trial ] ) . on the other hand , in the case ( b ) , as is shown in fig.[fig:2](d ) , @xmath89 are squashed to zero at the lower diffusive surface , and as a result , @xmath89 exhibit large spatial variations near the diffusive surface .
although such large spatial variations cost much gradient energy , the stripe order is still stable as the @xmath51-linear gradient term can lower the _ net _ gradient energy like in the case with two specular surfaces . in the case ( c ) where both surfaces are diffusive , @xmath89 must be zero at the both upper and lower surfaces , which results in a large spatial variation in @xmath90 in the striped bw state with @xmath75 . by using the spatial profiles of @xmath89 compatible with the diffusive condition
@xmath91 we can evaluate the gradient energy as @xmath92 comparing eqs.([eq : gl_trial ] ) and ( [ eq : gl_trial_df ] ) , one can see that @xmath51-relevant terms are almost unchanged , while the associated energy cost for @xmath90 , which corresponds to @xmath93 in eq.([eq : gl_trial_df ] ) , is much enhanced . because the large spatial variation in @xmath90 lowers the net energy gain , the stripe order becomes less stable . in the variational method using eq.([eq : op_trial_df ] ) , @xmath52 for the @xmath75 state is higher than the one for the @xmath86 state only by less than @xmath94 of @xmath52 for @xmath95 .
such a result is also obtained from exact numerical solutions of the gl equations .
thus , the uniform and striped bw states are almost degenerate near @xmath52 .
this suggests that the stability region of the stripe phase is restricted in the vicinity of the @xmath52 curve .
in fact , numerical calculations for determining the lower phase boundary show that the uniform bw phase extends up to near @xmath52 and the striped bw phase is possible only in a very narrow region below @xmath52 .
it is , however , difficult to accurately determine the lower phase boundary because near @xmath52 the free energy difference between the two states is very small compared with numerical errors .
the phase diagram in the case ( c ) is therefore not shown in fig.[fig:2 ] .
the above our results are obtained in the weak coupling limit without the strong - coupling effect which favors the abm pairing state in the bulk . in general , when the strong - coupling corrections are incorporated , the stability region of the stripe order may be modified .
recent theoretical work has shown that even in the presence of the strong - coupling contributions , the stripe order can survive for moderately diffusive surfaces with its stability region being suppressed at high temperatures , while it exists only in the vicinity of @xmath25 for maximally diffusive surface scatterings @xcite .
although in ref.@xcite , it is assumed that the strong - coupling contributions in a slab can be evaluated from the corresponding bulk values , they might be quite different from those in the bulk like in the case of superfluid @xmath0he in globally anisotropic aerogels @xcite , but this issue is beyond the scope of this work .
in this section , we will investigate quasiparticle excitations in the stripe order . as our focus is on how the midgap state is affected by the stripe structure in the pair potential @xmath12
, we will consider the typical case with two specular surfaces .
for the specular scattering , we have seen in the previous section that the simplified form of the order parameter eq.([eq : op_trial ] ) well describes superfluid properties near the stripe - planar transition , so that we could expect that eq.([eq : op_trial ] ) should work also for examining quasiparticle excitations in the stripe order .
below , we will solve the eilenberger equation with the use of eqs.([eq : bw_op_q ] ) and ( [ eq : op_trial ] ) to obtain the angle - resolved ldos which gives detailed informations on the low - energy excitations .
the quasiclassical spinful eilenberger equation has been extensively used for studies of superfluid @xmath0he . in general
, it should be solved self - consistently in combination with the superfluid gap equation . as we have mentioned above
, however , in this study we will use the fixed form of the gap function eq.([eq : op_trial ] ) instead of solving the gap equation .
this procedure should be valid for qualitative discussion on the low - energy excitations near @xmath16 because in this high temperature regime , the gap equation is reduced to the gl equation whose solution is well approximated by eq.([eq : op_trial ] ) . it has been known that the eilenberger equation can be easily solved by introducing so - called riccati amplitudes @xmath96 and @xmath97 which are @xmath98 matrices obeying the riccati equations @xmath99 with matsubara frequency @xmath100 @xcite .
the angle - resolved ldos @xmath101 can be expressed in terms of the riccati amplitude @xmath96 and @xmath97 as @xmath102\big| _ { i\varepsilon_n\rightarrow e+i\eta } \big\},\ ] ] where @xmath103 is a positive infinitesimal constant .
the amplitudes @xmath104 and @xmath105 can be obtained by solving eq.([eq : riccati ] ) with the replacement @xmath106 for the known pair potential . in this paper , the positive constant is chosen to be @xmath107 .
although the numerical integration of the differential equation ( [ eq : riccati ] ) requires initial values of @xmath96 and @xmath97 , we do not know the initial values anywhere in the film .
thus , we start from initial guess for @xmath96 and @xmath97 and numerically integrate along a classical trajectory until convergence is reached @xcite .
the concrete procedure is as follows .
let @xmath108 ( @xmath109 ) be a point which the classical trajectory starting from @xmath110 along @xmath29 ( -@xmath29 ) reaches after @xmath111 times reflections at the surfaces .
the riccati amplitude @xmath96 ( @xmath97 ) at a position @xmath110 for momentum @xmath29 are obtained by numerically integrating eq.([eq : riccati ] ) forward ( backward ) along the classical trajectory with an arbitrary initial value at @xmath109 ( @xmath108 ) .
note that @xmath111 must be sufficiently large such that the total length of the trajectory is much longer than the superfluid coherence length @xmath26 and the obtained riccati amplitudes do not depend on initial values . in this work ,
we take @xmath112 and parametrize the momentum direction as @xmath113 ( see fig.[fig:4](a ) ) . and @xmath114 [ the black colored symbol * in fig.[fig:2 ] ( a ) ] on the surface of the superfluid film with two specular surfaces .
( a ) , ( b ) , and ( c ) are results at the positions a ( @xmath115 ) , b ( @xmath116 ) , and c ( @xmath117 ) in fig.[fig:4 ] ( a ) , respectively .
left ( right ) panels show ldos s for the quasiparticle trajectory across ( along ) the stripe , namely , @xmath118 ( @xmath119 ) .
[ fig:3 ] ] figure [ fig:3 ] shows the angle - resolved ldos @xmath101 on the surface of the film with the thickness @xmath56 at three points along the modulation ( a , b , and c in fig.[fig:4](a ) ) , where @xmath120 .
we first consider the quasiparticle trajectory perpendicular to the modulation ( @xmath119 ) , or equivalently , parallel to the striped domain structure . here , at the domain center ( the point a in fig.[fig:4](a ) ) , @xmath50 is nonzero except just at the surfaces , while at the domain boundary ( the point c in fig.[fig:4](a ) ) , @xmath50 vanishes everywhere within the domain - wall plane . for @xmath119
, the quasiparticle excitations should be essentially the same as those for the uniformly deformed bw state without the modulation .
as one can see in the right panel of fig.[fig:3](a ) , @xmath101 at the domain center shows two branches in its low - energy part , which is qualitatively consistent with the result obtained by self - consistently solving the eilenberger and the gap equations @xcite .
this suggests that our theoretical approach using the fixed form of the pair potential is a good approximation for the qualitative discussion on the quasiparticle excitations .
the upper branch originates from the surface andreev bound state gapped by the overlap of the wave functions at the two surfaces @xcite , whereas the lower one corresponds to the gapless excitations due to the surface - induced suppression of @xmath50 .
as the domain boundary is approached , the energy gap between the andreev bound state and the bulk continuum becomes smaller because the surface - normal component of the order parameter @xmath50 gradually decreases .
as one can see in the right panel of fig.[fig:3](c ) , the gap is completely closed at the domain boundary where @xmath50 vanishes . ) , c ( @xmath117 ) , and b ( @xmath116 ) on the film surface represent the domain center , the domain boundary , and the middle point between them , respectively .
( b ) the amplitude of the superfluid gap @xmath121 at @xmath56 and @xmath114 along the quasiparticle trajectory which passes through the position a with the azimuthal angle @xmath122 ( @xmath123 ) and the polar angle @xmath119 ( parallel to the stripe ) . the origin @xmath124 corresponds to the position a. ( c)-(f ) the superfluid gap @xmath121 at @xmath56 and @xmath114 along the trajectory running across the stripe with the polar angle @xmath118 .
the origins @xmath124 in left [ ( c ) and ( e ) ] and right [ ( d ) and ( f ) ] panels correspond to the positions a and c , respectively . the azimuthal direction of the trajectory for ( c ) and ( d ) [ ( e ) and ( f ) ] is characterized by @xmath123 ( @xmath125 ) .
[ fig:4 ] ] in contrast to the conventional behavior for @xmath119 , an additional midgap state shows up for the quasiparticle trajectory running across the stripe ( @xmath118 ) . in the left panels of fig.[fig:3 ]
, one can find an additional branch between the andreev bound state and the bulk continuum , and the end point of this branch approaches @xmath126 as one goes from the domain center toward the domain boundary .
to understand the origin of this new bound state , we will examine the spatial profile of the pair potential along the trajectory @xcite .
the spatial variation in the amplitude of the pair potential @xmath127/2}$ ] along the quasiparticle trajectory is shown in fig.[fig:4 ] , where @xmath124 corresponds to a point on the surface ( a or c in fig.[fig:4](a ) ) .
for the trajectory slightly tilted from the surface normal , @xmath121 is very small at the surfaces [ figs.[fig:4](b)-(d ) ] , while for a highly tilted one , @xmath121 is relatively large at the surfaces [ figs.[fig:4](e ) and ( f ) ] . in the case of @xmath119 ( parallel to the stripe ) , @xmath121 exhibits a simple periodic behavior associated with multiple surface scatterings , which can be seen in fig.[fig:4](b ) .
by contrast , in the case of @xmath118 ( perpendicular to the stripe ) , an additional longer - period oscillation occurs , as is shown in figs.[fig:4](c)-(f ) .
this large oscillation is associated with the one - dimensional modulation of the stripe order . in figs.[fig:4](c ) and ( d ) , one can see that near @xmath124 , the large oscillation works as a confinement potential for the quasiparticles at the domain boundary ( fig.[fig:4](d ) ) , while it works as a energy barrier at the domain center ( fig.[fig:4](c ) ) .
that s the reason why the bound state exists at the domain boundary ( see the left panel of fig.[fig:3](c ) ) , while it does not near @xmath126 at the domain center ( see the left panel of fig.[fig:3](a ) ) . comparing figs.[fig:4](c - d ) with ( e - f ) , one finds that as the trajectory is tilted from the surface normal , namely , @xmath128 increases , the amplitude of the large oscillation becomes small .
this indicates that such a reduction of the energy barrier should lead to the overlap of the wave functions localized at each valley in @xmath121 , resulting in the midgap state appearing from a finite value of @xmath128 at the domain center . for the specular - surface film with the thickness @xmath56 .
( a ) and ( b ) are results at the domain center a ( @xmath115 ) and the domain boundary c ( @xmath117 ) , respectively .
red solid ( green dashed ) curves denote ldos s at the surface ( center ) of the film . in the inset of ( b ) , the bound state energy gap @xmath129 is plotted as a function of @xmath51 ; from right to left , data points correspond to @xmath130 at @xmath114 , and @xmath131 at @xmath132 .
[ fig:5 ] ] figures [ fig:5](a ) and ( b ) show the angle - averaged ldos s at the domain center and boundary for the thickness @xmath56 , respectively .
the overall feature of the ldos is almost unchanged along the @xmath59 axis for a fixed @xmath8 , while the low - energy behavior depends on @xmath8 .
as one can see in fig.[fig:5](a ) , the ldos at the domain center with nonzero @xmath50 exhibits a linear behavior near @xmath133 due to the contribution of the surface states extending from the upper and lower surfaces , which is consistent with the previous result for thin films @xcite . on the other hand , one can see in fig.[fig:5](b ) that the spectrum at the domain boundary has a kink near @xmath134 .
this kink originates from the bound state associated with the modulation of the stripe order .
actually , the kink position coincides with the bound state energy at @xmath126 @xmath135 ( see the left panel of fig.[fig:3](c ) ) .
now , the question is what determines the bound state energy gap @xmath129 .
since the bound state is formed due to the confinement potential shown in fig.[fig:4](d ) , it is natural to expect that @xmath129 is related to the shape of the potential which is characterized by the period of the stripe structure @xmath51 . in the inset of fig.[fig:5](b ) , @xmath129 is plotted as a function of @xmath51 for various film thicknesses .
@xmath129 is well scaled by @xmath51 , suggesting that the bound state energy gap @xmath129 is closely connected to the period of the stripe .
in this paper , we have examined the surface - roughness effect on the stability of the stripe order in thin films of the superfluid @xmath0he b phase based on the ginzburg - landau theory in the weak coupling limit , and also investigated the quasiparticle excitations in this striped superfluid phase by solving the eilenberger equation .
although the occurrence of the stripe order was originally pointed out for a film with two specular surfaces , it is found that the stripe order survives even in a film with one specular and one diffusive surfaces which should correspond to superfluid @xmath0he on a substrate .
our numerical results on the angle - resolved local density of states ( ldos ) show that a new bound state distinct from the surface andreev bound state appears for classical trajectories running across the stripe .
this unconventional bound state originates from the one dimensional modulation of the order parameter , namely , the stripe structure , and is reflected as a kink in the angle - averaged ldos .
the author is grateful to t. mizushima , k. machida , and y. tsutsumi for useful discussions .
this work is supported by a grant - in - aid for scientific research ( grant no . 25800194 ) .
60 d. volhardt and p. wolfle , _ the superfluid phases of helium 3 _ ( taylor and fransis , london , 1990 ) .
y. nagato , m. yamamoto , and k. nagai , j. low temp . phys . * 110 * , 1135 ( 1998 ) . a. b. vorontsov and j. a. sauls , phys . rev .
b * 68 * , 064508 ( 2003 ) .
s. murakawa , y. tamura , y. wada , m. wasai , m. saitoh , y. aoki , r. nomura , y. okuda , y. nagato , m. yamamoto , s. higashitani , and k. nagai , phys .
lett . * 103 * , 155301 ( 2009 ) .
y.h . li and t. l. ho , phys .
b * 38 * , 2362 ( 1988 ) . l. v. levitin , r. g. bennett , a. casey , b. cowan , j. saunders , d. drung , th .
schurig , and j. m. parpia , science * 340 * , 841 ( 2013 ) .
l. v. levitin , r. g. bennett , e. v. surovtsev , j. m. parpia , b. cowan , a. j. casey , and j. saunders , phys .
* 111 * , 235304 ( 2013 ) g.e .
volovik , jetp lett . * 63 * , 301 ( 1996 ) .
k. aoyama and r. ikeda , phys .
b * 72 * , 012515 ( 2005 ) ; r. ikeda and k. aoyama , phys . rev .
b * 79 * , 064527 ( 2009 ) .
k. aoyama and r. ikeda , phys .
b * 73 * , 060504 ( r ) ( 2006 ) .
g. e. volovik , j. low temp . phys .
* 150 * , 453 - 463 , ( 2008 ) .
t. kunimatsu , t. sato , k. izumina , a. matsubara , y. sasaki , m. kubota , o. ishikawa , t. mizusaki , and yu .
m. bunkov , jetp lett .
* 86 * , 216 ( 2007 ) .
j. elbs , yu . m. bunkov , e. collin , h. godfrin , and g. e. volovik , phys .
* 100 * , 215304 ( 2008 ) .
dmitriev , a.a .
senin , a.a .
soldatov , and a.n .
yudin , phys .
lett . * 115 * , 165304 ( 2015 ) .
askhadullin , v. v. dmitriev , d. a. krasnikhin , p. n. martynov , a. a. osipov , a. a. senin , and a. n. yudin , jetp letters * 95 * , 326 ( 2012 ) a. l. fetter and s. ullah , j. low temp . phys . * 70 * , 515 ( 1988 ) . j. j. wiman and j. a. sauls , phys .
b * 92 * , 144515 ( 2015 ) . m. t. manninenm and j. p. pekola , j. low temp phys . *
52 * , 497 ( 1983 ) .
v. kotsubo , k. d. hahn , and j. m. parpia , phys .
58 * , 804 ( 1987 ) .
j. p. pekola , j. c. davis , z. yu - qun , r. n. r. spohr , p. b. price , and r. e. packard , j. low temp . phys . * 67 * , 47 ( 1987 ) .
j. saunders , d. s. betts , d. f. brewer , s. j. swithenby , and w. s. truscott , phys .
. lett . * 40 * , 1278 ( 1978 ) .
a. yamaguchi , t. hayakawa , h. nema , and h. ishimoto , j. phys .
: conference series * 150 * , 032123 ( 2009 ) . a. b. vorontsov and j. a. sauls , phys .
* 98 * , 045301 ( 2007 ) .
a. b. vorontsov , phys .
. lett . * 102 * , 177001 ( 2009 ) . m. hachiya , k. aoyama , and r. ikeda , phys .
b * 88 * , 064519 ( 2013 ) .
k. aoyama , phys .
b * 89 * , 140502(r ) ( 2014 ) .
n. miyawaki and s. higashitani , phys .
b * 91 * , 094511 ( 2015 ) .
s. higashitani and n. miyawaki , j. phys .
jpn . * 84 * , 033708 ( 2015 ) . m. hkansson , t. lfwander , and m. fogelstrm , nat . phys . * 11 * , 755 - 760 ( 2015 ) . a. vorontsov , arxiv:1602.06917 . j. j. wiman and j. a. sauls , arxiv:1605.01047 .
p. fulde and r. a. ferrell , phys . rev . * 135 * , a550 ( 1964 ) .
a. i. larkin and yu .
n. ovchinnikov , sov .
jetp * 20 * , 762 ( 1965 ) .
m. r. freeman , r. s. germain , e. v. thuneberg , and r. c. richardson , phys .
lett . * 60 * , 596 ( 1988 ) .
h. adachi and r. ikeda , phys .
b * 68 * , 184510 ( 2003 ) .
e. i. rashba , sov .
phys . solid state * 2 * , 1109 ( 1960 ) .
, edited by e. bauer and m. sigrist , springer 2012 .
o. v. dimitrova and m. v. feigelman , jetp lett . * 78 * , 637 ( 2003 ) .
k. v. samokhin , phys .
b * 70 * , 104521 ( 2004 ) .
r. p. kaur , d. f. agterberg , and m. sigrist , phys .
* 94 * , 137002 ( 2005 ) . k. aoyama and m. sigrist , phys .
lett . * 109 * 237007 ( 2012 ) .
r. ikeda , phys .
b * 91 * , 174515 ( 2015 ) .
y. nagai , k. tanaka , and n. hayashi , phys .
b * 86 * , 094526 ( 2012 ) .
t. mizushima , phys .
b * 86 * , 094518 ( 2012 ) .
y. tsutsumi , m. ichioka , and k. machida , phys .
b * 83 * , 094510 ( 2011 ) .
m. ichioka , t. mizushima , and k. machida , phys .
b * 82 * , 094516 ( 2010 ) .
|
surface scattering effects in thin films of the superfluid @xmath0he b phase have been theoretically investigated , with an emphasis on the stability of the stripe order with spontaneous broken translational symmetry in the film plane and quasiparticle excitations in this spatially inhomogeneous phase .
based on the ginzburg - landau theory in the weak coupling limit , we have shown that the stripe order , which was originally discussed for a film with two specular surfaces , can be stable in a film with one specular and one diffusive surfaces which should correspond to superfluid @xmath0he on a substrate .
it is also found by numerically solving the eilenberger equation that due to the stripe structure , a midgap state distinct from the surface andreev bound state emerges and its signature is reflected in the local density of states .
|
star formation ( sf ) is a fundamental process in the evolution of galaxies . but this process is far from being well understood .
the sf is usually characterized by the initial mass function ( imf ) and the total sf rate ( sfr ) , which depends on many factors such as the density of the interstellar gas , its morphology , its metallicity , _ etc .
_ according to larson ( 1987 ) , four major factors drive star formation in galaxies : large scale gravitational instabilities , cloud compression by density waves , compression in a rotating galactic disk due to shear forces , and random cloud collisions . in galaxies with previous stellar generations ,
additional sf triggers exist , such as shock waves from stellar winds and supernova explosions . in dense environments , such as clusters of galaxies and compact groups , tidal interactions and collisions with other galaxies
probably play some role in triggering star formation ( _ e.g. , _ the blue compact dwarf and the i m galaxies apparently formed at the ends of the tidal tails of arp 105 : duc and mirabel 1994 ) .
while `` global''phenomena , such as the first two sf triggers of larson ( 1987 ) , play a large part in grand design spirals , random collisions of interstellar clouds may provide the best explanation for dwarf galaxies with bursts of sf .
therefore , understanding sf in dwarf galaxies should be simpler than in other types of galaxies .
understanding this type of objects has also cosmological implications , as some star forming dwarf galaxies may be genuinely young .
one way to derive the star formation history of galaxies is with a combination of credible spectral synthesis models and observations in as many spectral bands as possible , which should provide constraints on the imf .
this approach is usually followed for a sample of galaxies of similar type , in which the star formation parameters would be basically similar .
surface photometry in broad - bands , from which the spectral energy distribution ( sed ) is derived , together with a tracer of massive stars , such as h@xmath0 or uv measurements , makes it possible to obtain the total ongoing sfr .
the h@xmath0 radiation is directly coupled to the radiation at wavelengths @xmath1912 ( _ e.g. _ , kennicutt 1983 ) , indicating the ongoing star formation .
uv radiation longward of the lyman break , mainly emission detected by space experiments such as iue and hst , which is longward of lyman @xmath0 , is produced not only by the most massive stars , but also by cooler stars that do not ionize hydrogen significantly .
almoznino ( 1995 ) studied a sample of bcds in the virgo cluster selected from the binggeli , sandage , and tamman ( 1985 , hereafter bst ) catalog , with hi measurements from hoffman ( 1987 , 1989a ) . among the objects in his sample he found one exceptional object , vcc 144 , which could be a truly young galaxy . here
we describe our observations of this object , combine them with existing information , compare them with evolutionary population synthesis models , and present our interpretation .
the virgo cluster contains a very unusual object ( hi 1225 + 01 ) , discovered serendipitously by giovanelli & haynes ( 1989 ) as an extragalactic hi cloud within the cluster .
subsequently , an optical counterpart was identified ( _ e.g. , _
djorgovski 1990 ) , was studied by salzer ( 1991 ) , and was recently mapped in hi with the vla ( chengalur 1995 ) .
salzer showed it to be a system which witnessed a burst of star formation @xmath21 gyr ago , with signs of a recent ( @xmath210 myr ) second burst .
chengalur ( 1995 ) found rotation in the hi distribution and identified the star - forming clump as the only region where the hi surface density exceeds 4.6 10@xmath3 atoms @xmath4 .
the system is probably the result of a tidal interaction between two hi components .
a general review of properties of dwarf galaxies was given by thuan ( 1992 ) and a review of spectroscopic properties of star - forming dwarfs was presented by vlchez ( 1995 ) .
lately , a number of papers reported extended hi envelopes , or hi companions , of dwarf galaxies ( _ e.g. , _
taylor 1995 ; van zee 1995 ; hoffman 1996 ; szomoru 1996a , 1996b ) .
as will be seen below , we detected an extended hi envelope connected with vcc 144 .
it is possible that this object represents a more intensely star - forming case than hi 1225 + 01 and that dwarf galaxies with extended hi distributions have other common characteristics .
therefore , we will compare our measurements and derived properties for this virgo bcd with those of the optical counterparts of the dwarf galaxies with extended hi and with hi 1225 + 01 .
below we first review published observations of vcc 144 and describe our own observations together with additional data collected from the literature or data banks .
we adopt here a uniform distance of 18 mpc to the virgo cluster , and by inference to all its members , including vcc 144 .
a smaller cluster distance , @xmath5 mpc , is suggested by the [ oiii ] distances to planetary nebulae ( jacoby 1990 ) , but visvanathan & griersmith ( 1979 ) give 17.9 mpc , in agreement with sandage & tammann ( 1974 ) .
we note other values for the distance to the virgo cluster , in particular 16.8 mpc ( tully 1988 ) and the estimate of the distance to m100 from observations of cepheids ( 17.1 mpc : freedman 1995 ) , while considering the depth of the cluster to be some 3 - 4 mpc .
note that vcc 144 could be located in the w cloud ( hoffman 1989b ) .
if this is the case , its distance could be 1.7 to 2 times greater than adopted here .
bst identified vcc 144 [ optical peak at @xmath0(j2000)=12@xmath6 15@xmath7 [email protected] , @xmath9(j2000)=+05@xmath1045 39``.2 ] , a bcd galaxy , as a member of the virgo cluster , despite it being just outside the borders of the w cloud .
note that the coordinates given in bst are off the peak of the optical image by @[email protected] and @xmath12=4''.7
. de vaucouleurs ( 1991 ) quote for this galaxy an optical heliocentric velocity v@xmath13=1960@xmath1452 km s@xmath15 .
gallagher & hunter ( 1986 ) reported on single - aperture photoelectric photometry of this object , along with 64 other similar objects .
they measured within a 19 " aperture [email protected] , ( b v)[email protected] , and ( u b)[email protected] .
gallagher & hunter ( 1989 , gh89 ) surveyed spectroscopically bcds in the virgo cluster . for vcc 144
they measured an h@xmath16 equivalent width ew[h@xmath16]=35 . the low resolution ( @xmath215 )
spectrum with the irs on the kpno 0.9 m telescope , shown as their fig .
2 , exhibits a blue - sloping continuum with strong and narrow emission lines .
these are ( in order of strength ) [ oiii ] @xmath185007 , [ oii ] @xmath183727 , [ oiii ] @xmath184959 , h@xmath16 , h@xmath19 , h@xmath9 ( very weak ) , and a blend longward of [ oii ] , that can be mostly hei @xmath183888 or an incomplete sky subtraction .
the region between h@xmath16 and h@xmath19 is free of the heii @xmath184686 and niii @xmath184641 high excitation lines .
gh89 correct upward the h@xmath16 line by 10% to account for underlying absorption ( which is not evident in their spectrum ) , and obtain a line flux of 1.14 10@xmath20 erg @xmath4 s@xmath15 , with ratios of other lines to h@xmath16 : [ oii ] @xmath183727=3.16 , [ oiii ] @xmath184959=1.03 , and [ oiii ] @xmath185007=3.68 . to these
we add , by measuring from their fig . 2 : [ oiii ] @[email protected] , [email protected] , [email protected] , he i @[email protected] ( if real ) .
vcc 144 was observed with the ccd camera of the wise observatory ( wiseobs ) on april 11 , 1991 .
the galaxy was imaged through standard b , v , r and i filters . in total , we collected three b images each of 10 minute exposure , three 200 sec v - band images , and four 120 sec exposures for r and for i with the rca - ccd .
this is a thinned , back - illuminated chip , with 512@xmath25320 pixels .
each pixel subtends 0``.87 at the f/7 focus of the 1 m wiseobs reflector .
a reference star close enough to appear on the galaxy s ccd frame was imaged together with the galaxy .
this star was calibrated against landolt standards ( landolt 1973 , 1992 ) with photoelectric photometry at the wiseobs using a 30 '' round aperture .
the galaxy is shown as the combined b - band frame , underlying the contours in fig .
[ fig4 ] .
the galaxy was also imaged through narrow - band filters , to derive the h@xmath0 emission .
two filters were used : one containing the h@xmath0 line at the redshift of the object and the other sampling the continuum at @xmath26 ) .
three 20 min exposures were obtained through each of the two filters .
in addition , the hz44 spectrophotometric standard star was measured several times each night through different air masses with the same filters , in order to derive the atmospheric extinction and the absolute photometric calibration for the h@xmath0 images .
ea observed vcc 144 with iue in 19 jan 93 for a full vilspa observing shift with @xmath26 hours of integration .
the spectrum was obtained through the iue large aperture with the short wavelength primary camera in low dispersion setup .
vcc 144 is a very compact object , and the radiation comes almost entirely from its central 10 " , thus iue collected essentially all of its uv flux .
the iue spectrum ( swp 46761 ) was very weak after its extraction .
we measured two regions 100 wide and averaged the flux densities within each to obtain two uv photometric indices at 1350 and at 1850 .
the monochromatic magnitudes in these two bands are [ 1350][email protected] and [ 1850][email protected] and the respective flux densities are [email protected] 10@xmath27 and [email protected] 10@xmath27 erg s@xmath15 @xmath4 @xmath15 .
vcc 144 is not listed in the iras point source catalog .
the iras data consist of flux densities in four infra - red ( ir ) bands centered at 12@xmath28 , 25@xmath28 , 60@xmath28 and 100@xmath28 . iras scanned most of the celestial sphere several times , and it is possible to co - add the various iras scans for a given location to obtain a deeper detection or upper limit at this location in the sky .
the iras faint source catalog lists a source very close to the position of vcc 144 with the following flux densities : [ 12][email protected] jy , [ 25][email protected] jy , [ 60][email protected] jy , and [ 100][email protected] jy . we adopt this as the fir emission from the galaxy .
as mentioned above , our sample is based on the hi measurements of hoffman _
( 1987 ) and hoffman _ et al . _
( 1989 ) with the arecibo radio telescope .
they list for vcc 144 a flux integral of 2305@xmath1419 mjy km s@xmath15 and a 135 km s@xmath15 hi line width at 20% of the peak .
the profile is triangular - shaped , asymmetric , and shows no sign of a two - horn profile expected for a flat rotation curve .
while a solid - body rotation curve can give rise to a triangular profile ( skillman 1996 ) , it is unusual to find evidence of significant rotation in dwarf galaxies which exhibit triangular central beam profiles as narrow as this one ( hoffman 1996 ) .
the widths at 50% and 80% of the peak intensity are 82 and 44 km s@xmath15 .
the heliocentric velocity from the hi measurement , adopted by hoffman ( 1989 ) as the mid - point of the hi profile edges at 50% peak intensity , is 2014 km s@xmath15 . within the measuring accuracy ,
the optical and hi redshifts coincide .
vcc 144 was observed in hi and radio continuum at the very large array ( vla ) .
observations in the l band continuum and hi line were conducted for 15 and 82 minutes on - source , respectively , on 12 and 16 april 1988 in the c array ; c band continuum observations were made for 52 minutes on - source on 10 september 1988 in the d array .
the same pointing center ( 1950 ) , @xmath29 , was used throughout .
the spectral line observations used 32 channels with channel spacing of 10.45 km @xmath30 centered on heliocentric velocity 2014 km @xmath30 .
continuum observations were conducted with 50 mhz bandpasses at 6 and 20 cm .
calibration was performed using sources from the vla calibrator list in the usual way , with map - making and deconvolution performed using the standard routine imagr within the nrao aips software package . for the hi data , the continuum was subtracted in the @xmath31 domain , and zero - spacing fluxes
were interpolated for each channel from the arecibo spectrum .
the rms noise achieved after cleaning was 0.21 mjy / beam for the l band continuum data , 0.053 mjy / beam for the c band data , and 2.0 mjy / beam in each channel for the hi line data .
we measured for vcc 144 the following integrated photometric properties : [email protected] , ( b v)[email protected] , ( v r)[email protected] , and ( r i)[email protected] .
the photometry was performed through a polygonal aperture , which traced the `` limits '' of the galaxy as estimated visually on the deepest ( v - band ) image .
this corresponds to a surface brightness of @xmath225.5 mag / square arcsec .
this rather shallow limit is the result of the low exposure collected for vcc 144 , which is one of the brightest galaxies in our sample .
although the v magnitude we measure reproduces that by gh86 , there is a possible discrepancy in the b band , as the color ( b v ) is marginally redder ( 2@xmath32 ) than what they measured . the difference could be related to a radial color gradient in the galaxy , as gh86 measured only the innermost 19 `` while we include an additional @xmath210 '' external region .
however , a test of the image profiles in the b , v , and r bands ( fig .
[ fig1 ] , below ) shows that they are @xmath2identical , thus vcc 144 is _ not _ redder in its outer regions . with the measured photometry , and an adopted distance of 18 mpc to the virgo cluster
, vcc 144 appears to be a low luminosity dwarf galaxy ( [email protected] ) .
if in cloud w , as mentioned above and @xmath34 further away , [email protected] mag , it would still be a dwarf galaxy , although on the brighter side .
the h@xmath0 frames were combined and the final image was shifted to match the broad - band images , and the galaxy was measured inside the same polygonal area used for its broad - band image .
the method of measurement was basically the same as for the broad - band frames , with the exception that the sky background level was computed several times using different boxes around the objects ( see almoznino 1993 ) .
the mean of these results was adopted as the object counts , while the error was taken as the largest of the errors of each measurement .
the standard deviation of the results was smaller than this .
the integrated h@xmath0 flux was measured to be [email protected] 10@xmath20 erg s@xmath15 @xmath4 .
the average line equivalent width , obtained by dividing the net line image by the off - line continuum image , is 159@xmath1422 , with a peak value of 390 .
the intense h@xmath0 emission implies that @xmath213% of the r band photons originate from this line and from [ nii ] emission .
we have not corrected the magnitudes and colors for line emission , either for h@xmath0+[nii ] in r , for [ oiii ] in v , or for [ oii ] in u. the contribution of line emission is strongest in the r band ; the other lines have smaller equivalent width and are located near the edges of bands .
we smoothed all final images , combined from different images through the same filter , with a gaussian of full width at half - maximum of 4 pixels=3``.5 , which corresponds to the typical seeing in our images .
we then obtained an intensity profile for each image , by fitting ellipses to the isophotes .
the intensity profiles for the b , v , r , and i bands , as well as that for the net - h@xmath0 image , were scaled to the same peak intensity ( of the b image ) , and the widths of the profiles were compared .
we found that all five images had virtually identical profiles , to within 5% of the peak intensity at @xmath212 pixels@xmath2411 '' galactocentric distance .
the intensity profiles are plotted in fig .
the images are all resolved , with very similar eccentricities of @xmath20.6 in all colors .
the surface brightness is high with an average value of @xmath35=20.5 mag/@xmath36 " .
note that because of the smoothing , there are only very few `` independent '' points , thus fitting the light distribution to any of the usual profiles seems pointless .
the high surface brightness , the strong h@xmath0 emission , and the compact and blue appearance of the galaxy point to an interpretation favoring a strong burst of star formation taking place at present in vcc 144 .
we did not find any faint extensions or disturbances in our images ; thus the triggering of star formation in vcc 144 can not be attributed to a recent interaction with another visible galaxy . both vla continuum images ( figs .
[ fig2 ] and [ fig3 ] ) show a single gaussian feature centered on the optical image of the galaxy . at 20 cm ,
the feature has a size @xmath37 after deconvolution with the @xmath38 beam .
the peak intensity of a fitted gaussian is @xmath39 mjy / beam and the integrated intensity over the feature is @xmath40 mjy . at 6 cm
the galaxy is still only marginally resolved ( the other features on the map are all well outside the optical and hi boundaries , and are almost certainly unrelated to vcc 144 ) , with a size about @xmath41 after deconvolution with the @xmath42 beam , a gaussian peak at @xmath43 mjy / beam and an integrated intensity of [email protected] mjy .
this gives a global spectral index of [email protected] if the integrated intensities are used , or [email protected] if the peak intensities are used instead .
the 21 cm line data were processed at full resolution ( 16 `` ) and at 45 '' , to bring out extended features at lower column densities .
the maps at full and at reduced resolution are presented below and are discussed in detail . the total hi map at 16 " resolution was computed by integrating over the emission - bearing channels with no blanking applied .
the resulting contour map , overlaid on a grey scale version of the b - band image , is shown in fig . [ fig4 ] and reveals one concentration of about @xmath44 coincident in position with the optical image , with hints of more diffuse emission extending perhaps one beam width to the southwest , and about 1.5 arcmin to the northeast . to confirm that the extension is real
, we convolved the map with a gaussian to smooth it to a resolution of 45 arcsec .
the result is shown in fig .
[ fig5 ] , and the ne extension stands out clearly .
in fact , the two figures show that the hi distribution extends on both sides of the minor axis of the galaxy .
integrating over the entire hi distribution gives a total mass of @xmath45 , about 50% larger than the flux estimated from the single - beam spectrum obtained with the arecibo telescope , which partially resolves the hi cloud .
our images and inspection of the palomar sky survey copies show no optical counterpart within the ne extension to a surface brightness limit of @xmath224 mag / square arcsec .
the hi linear extent is about 11.5 kpc .
all masses are appropriate to a distance of 18 mpc as assumed throughout , and would be larger by a factor of @xmath46 if vcc 144 were at the w cloud distance . to obtain isovelocity contours at 16 " resolution it was necessary to blank the cube in two stages : first everything outside the lowest contour on the smoothed total hi map ( fig .
[ fig5 ] ) , presumably unrelated to vcc 144 , was blanked .
then only features in each channel that exceeded 2 mjy / beam were retained for computation of the first moment map , shown in fig .
[ fig6 ] . a 45 " resolution
first moment map was also produced , retaining only features in each channel of the spatially smoothed cube that exceeded 5 mjy / beam .
everything outside the lowest contour on the smoothed total hi map was then blanked from the smoothed first moment map , which is shown in fig .
[ fig7 ] . inspection of the isovelocity contours in figs .
[ fig6 ] and [ fig7 ] , and the individual channel maps shown in figs .
[ fig8 ] and [ fig9 ] as mosaics , leads us to the following interpretation : the dominant clump , centered on the optical image , has a mean velocity around 2010 km @xmath30 with a roughly gaussian velocity profile of width about 80 km @xmath30 .
the ne extension lies generally at higher velocity , but has a more complex velocity profile with prominent narrow features at about 2070 , 2040 and 1980 km @xmath30 .
this suggests some bulk motion of the diffuse gas with respect to the main concentration , but we can not say conclusively from these observations whether that motion is more properly described as infall , ejection , or rotational motion . other considerations , to be discussed below , lead us to favor a gas blow - out .
we do not have adequate signal nor spatial resolution to determine whether the main concentration rotates about its own axis , or has purely turbulent internal motions . in any case ,
turbulent motions appear to be at least as significant as ordered motions in the dynamics of the clump and of the system as a whole .
we estimate here a representative total dynamical mass for the system , and masses of its different components ( stars , gas , and dust ) .
dwarf irregular galaxies often show little rotation and their dynamics are dominated by turbulent motions ( _ e.g. , _ sargent 1983 ) .
it is possible to estimate an indicative gravitational mass m@xmath47 for the galaxy using the half - width at half - intensity of the single beam 21 cm profile @xmath48 : @xmath49 for a typical hi maximal extent from the center of the hi distribution of r=4 kpc the gravitational mass is m@xmath47=1.56 10@xmath50 m@xmath13 .
the total b magnitude measured by us , with the adopted distance to the galaxy , yields a total blue luminosity of @xmath23.6 10@xmath51 l@xmath13 , making the indicative mass - to - blue light ratio m@xmath47/[email protected] .
this is somewhat higher than m / l values of blue star - forming galaxies .
note though that if we are witnessing a collision between two clouds , and if the relative velocities of the clouds are approximately perpendicular to our line of sight , there is no requirement for the gas to be bound to the system . in this case
we might be seeing a tidal disruption taking place , or the blow - out of gas from a system experiencing strong star formation .
the indicative dynamical mass of the dominant clump , assuming a profile half - width of 40 km s@xmath15 and a radius of 1.6 kpc , is @xmath25.7 10@xmath51 m@xmath13 . for the densest part of the ne extension ( visible as a distinct clump in the 16 " resolution
total hi map ) , taking `` profile half - width '' to be half the difference between the velocities of the two narrow peaks and a similar radius , we get 3.6 10@xmath51 m@xmath13 .
we calculated the hi content of vcc 144 from the hi flux density integrated over the entire map , following the prescription in de vaucouleurs _
the flux integral value from hoffman ( 1989a ) yields a total hi mass of [email protected] 10@xmath51 @xmath53 m@xmath13 , and the total integrated flux from the vla map yields 2.7 10@xmath51 @xmath53 m@xmath13 with similar uncertainty .
the vla map of the dominant clump , which coincides with the optical image , contains 1.2 10@xmath51 m@xmath13 of hi .
the hi mass is thus comparable with the stellar mass , as derived from the blue luminosity for a m / l@xmath541 assumption , but is only @xmath220% of the indicative dynamical mass . the location of vcc 144 in a fir color - color diagram , such as that by helou ( 1986 ) , coincides with that of starburst , bcd , and extragalactic hii region galaxies , and the exact location of the point indicates that most fir emission is produced by the current star formation , _
i.e. , _ there is no significant `` cirrus '' cloud contribution .
the total dust content can be estimated from the relations of thronson ( 1988 ) , bothun ( 1989 ) , or thuan & sauvage ( 1992 ) .
the harmonic mean of the three values yields m@xmath55=9.3 10@xmath56 m@xmath13 as representative total dust mass and the dust - to - gas mass ratio is @xmath23.4 10@xmath57 if the dust is distributed within the entire hi entity or @xmath27.8 10@xmath57 if it is relegated to the optical galaxy and its hi cloud .
some 10@xmath58 massive stars must have existed prior to the present star formation burst to have produced this amount of dust , or the dust could be pre - stellar in this galaxy .
it is useful to consider the entire sed of the galaxy , displayed in fig .
[ fig10 ] .
the points originate mainly from our wiseobs broad - band photometry and from measurements with the iue . in order to derive the total u magnitude we assumed that the ( u b ) color measured by gh86 for the inner 19 " is representative for the entire galaxy .
the sed shows a steep decline with increasing wavelength , both in the uv and in the optical .
since our derived sed is very sparse because of having only broad - band colors , it can not be tested against detailed spectra , but can be compared , in general , to various typical seds of different stellar populations . a comparison with the experimentally - derived spectral templates of kinney ( 1996 ) in fig .
[ fig10 ] shows a good fit to the `` negligible extinction '' [ e(b v)@xmath210.10 ] template of a starburst ( their fig .
thus , from the sed alone , we could conclude that vcc 144 _ is _ an unextinguished , star - bursting galaxy .
the `` recent '' star formation rate may be estimated from the blue luminosity ( gallagher 1984 , ght ) as : @xmath59 averaged over the past 0.46 gyrs .
the assumption in ght is that the sfr is constant over the lifetime of the galaxy and the imf is a salpeter law with an upper mass cutoff of 100 m@xmath13 .
a recent starburst would have correspondingly more blue and luminous stars , lowering the sfr resulting from the same l@xmath60 .
the galaxy shows the most intense h@xmath0 emission among objects in our sample .
the line has very high equivalent width , peaking at 390 .
the entire @xmath21 kpc optical galaxy is a huge hii region .
the line radiation comes from the entire galaxy and , as mentioned above , no optical extension to the galaxy is visible beyond the line emitting region even in the broad - band images .
this intense h@xmath0 emission indicates a high sfr at present
. the lower limit on the [ oiii ] line ratio from the gallagher & hunter ( 1989 ) spectrum indicates an electron temperature @xmath61k .
the present star formation rate is estimated from the h@xmath0 , again using a relation from ght : @xmath62 this requires correcting the h@xmath0 emission for extinction . adopting our h@xmath0 flux and comparing it with the h@xmath16 from gh89
, we find h@xmath0/h@xmath16=6.18 , apparently indicating a significant amount of extinction for case b recombination theory .
the comparison is valid , although our h@xmath0 measurement is _ global _ while that of gh89 was taken through a 22 `` aperture , because the h@xmath0 image of vcc 144 s is @xmath63 pixels@xmath2422 '' wide , but is elongated .
gh89 estimate e(b
v)@xmath540.0 from h@xmath16/h@xmath19 .
the h@xmath0/h@xmath16 inconsistency , produced by two different observations , may result if gh89 missed part of the total h@xmath16 flux in their round aperture .
we conclude that the extinction for vcc 144 is probably `` negligible '' , as indicated by the reasonable fit to the kinney ( 1996 ) sed .
this yields sfr@xmath64=0.29 m@xmath13/yr , assuming no extinction of the h@xmath0 line .
the discrepancy between sfr@xmath60 and sfr@xmath64 found above may result from an incompatibility of the model .
a similar result was found for an irregular galaxy by thronson ( 1988 ) . as will be shown below
, our preferred interpretation of vcc 144 is of a recent and first star formation burst .
therefore , the ght method of deriving sfr@xmath60 will not be correct .
a comparison with the spectral energy distribution of different young starbursts may be more adequate .
the lack of heii and typical wr emission signatures indicates that high mass stars ( m@xmath6540m@xmath13 ) are missing from the present population mix , while ght assumed an upper mass cutoff of 100 m@xmath13 ; only if the sf burst happened instantaneously a few 10@xmath66 yrs ago , the most massive stars and their wr remnants could be absent now , which would explain the discrepancy between sfr@xmath60 and sfr@xmath64 .
another hint about the present population of massive stars , derived from the radio continuum mapping , will be discussed below .
the far - ir flux , from lonsdale ( 1985 ) , is : @xmath67 \ , 10^{-14}\ ] ] for vcc 144 , at the nominal distance of 18 mpc , this implies a fir luminosity of 7.6 10@xmath68 l@xmath13 , thus @xmath69=4.7 and the galaxy is not a strong fir source .
the sfr in vcc 144 can also be estimated from the iras data , using the method of thronson and telesco ( 1986 ) .
this assumes that essentially the entire luminosity of the massive ob stars is absorbed by the dust and is reemitted in the infrared .
this yields the following relation : @xmath70 with the fir luminosity @xmath71 given in solar units and approximated by : @xmath72 where @xmath73 is the distance to the object in mpc and @xmath74 and @xmath75 are the two iras flux densities in jy .
the sfr resulting from this method is similar to that derived from the h@xmath0 data ( @xmath20.3 m@xmath13 yr@xmath15 ) , increasing our confidence in the `` negligible extinction '' case for this galaxy and , coincidentally , in the assumption that most photons from ob stars heat the dust .
it is worthwhile to consider a distance - free parameter , such as the star formation per unit galactic area .
this can be derived using the size of the galaxy from its ccd image , or profile , as exhibited in fig .
1 . in this case
, the ongoing sfr per unit area is given by : @xmath76 where the sfr / area is in m@xmath13/yr/@xmath77 and @xmath78 is the observed h@xmath0 flux per square arcsecond .
the high sfr / area value , 2.28@xmath79 , originates from the strong luminosity and compactness of the object .
not surprisingly , with a high specific sf but small size vcc 144 undergoes strong star formation per unit galactic area , but shows only modest total sfr . for comparison
, the milky way has a total sfr of @xmath80 and sfr / area of @xmath81 .
large , active , star forming galaxies have sfr / area similar to that of vcc 144 ( _ e.g. , _ pogge and eskridge 1987 ) .
thus , once the star formation activity is triggered , its intensity , as manifested by the sfr / area , is similar in small and large galaxies .
this interpretation implies that mechanisms believed to be responsible for star formation in large galaxies but not in dwarf galaxies , such as compression due to density waves or from rotational shear forces , are as efficient as other mechanisms , such as random collision between clouds or gravitational instabilities caused by other effects .
otherwise , there would be significantly more star formation per area in large star - forming galaxies than in late - type dwarfs .
it is worth mentioning that here the _ efficiency _ of the star formation process , caused by the various mechanisms , is tested , rather than how wide - spread these mechanisms are .
therefore , the conclusion at this stage is that once star formation is induced , its efficiency is approximately the same , regardless of the mechanism which induced it .
the radio synthesis observations indicate that the continuum source is connected with the presence of stars , as the emission is localized in the cloud which coincides with the optical counterpart .
the spectral index is consistent with thermal emission processes .
the thermal radio continuum can be calculated from the physical conditions in the hii region : @xmath82 for the observations at 20 cm , with t@xmath83=1.3 10@xmath58k ( upper limit , from the [ oiii ] lines ) , with the number of ionizing photons calculated from h@xmath0 ( see below ) and at 18 mpc , s(1420 mhz)=0.9 mjy . at 5 ghz
the thermal radio continuum is 0.8 mjy and makes up @xmath230% of the detected flux .
the fraction of non - thermal 20 cm contribution is @xmath21.5 mjy and at 6 cm it is @xmath21.0 mjy .
the resultant non - thermal spectral index is @xmath20.3 . with an average flux from a snr at 18 mpc of @xmath21 mjy
there should be at present at most a couple of radio snrs in vcc 144 .
this is consistent with the present number of early - type stars ; with @xmath22000 o7 stars present ( see below ) , with a lifetime to supernova explosion of @xmath210@xmath68 years and a radio snr lifetime of 10@xmath58 yrs , a couple of snrs should be detectable at present in radio continuum .
above we derived a number of sfrs , which can be combined with the observationally - derived masses to yield two indicative time scales .
in particular , it is possible to use the `` star formation timescale '' , defined by hodge ( 1993 ) as : @xmath84 where @xmath85 is the total mass in stars , estimated with the assumption of @xmath86 , and @xmath87 is the present sfr . for vcc 144
we obtain @xmath88=1.2 gyrs , similar to hodge s value for the most extreme starburst galaxy in his list .
this is an indication on the production of the present - day luminosity with solar - type stars ; for an imf rich in young stars
m / l is lower .
mateo ( 1992 ) found one magellanic cluster with m / [email protected] for an age of 8 10@xmath68 yrs .
for three magellanic clusters , elson and freeman ( 1985 ) found m / l between 0.11 and 0.56 ; if this is the case for vcc 144 , the mass in stars would be lower and @xmath88 could be @xmath89 yrs or even less .
the roberts time scale measures the time to exhaust the galactic hi reservoir given the present sfr : @xmath90 we obtain for vcc 144 @xmath91=0.9 gyrs if the entire hi reservoir is considered , or 0.4 gyrs for the hi coincident with the optical image .
the estimate of the roberts time scale assumes no recycling of material from the stars ; otherwise , the time to exhaust the hi will be longer for a constant sfr . for a star formation proceeding in bursts followed by long periods of quiescence
, the galaxy could survive for a hubble time , provided the star formation duty cycle is @xmath542% .
this constraint can be relaxed if the total hi would include the diffuse extended cloud of hydrogen .
we compared our results with population synthesis models of leitherer & heckman ( 1995 ) and found reasonable matches for single star burst models 5 - 7 myrs old .
we emphasize that the matches are not perfect , but indicative .
we tested also the h@xmath0 data through the number of ionizing photons , following osterbrock ( 1989 ) for a simple case b recombination theory and with the h@xmath0 luminosity in erg / s : @xmath92 the color index ( h@xmath0v ) for the models , for this purpose , is obtained as @xmath93=129.8 - 2.5 \ , log(n_c)-m_v\ ] ] we compared the observed color index [ h@xmath0v]=2.5 log [ f(h@xmath0)]v=15.56 , ( b v ) , and m@xmath94 with the model predictions .
the indices fit a @xmath26 myr old single star burst model from leitherer & heckman ( 1995 ) but are not compatible with continuous star formation .
these are much too blue to fit the observations . from the value of n@xmath95 found for vcc 144 , using l(h@xmath0)=2.7 10@xmath96 erg s@xmath15 and adopting a lyman continuum photon flux of 10@xmath97 s@xmath15 from a typical o star , it appears that some 2000 late - o or b stars are needed in vcc 144 at present to ionize the hydrogen . for a salpeter imf
, this implies that the present star burst produced some 10@xmath66 m@xmath13 in stars of all masses .
the observational and derived parameters related to vcc 144 are summarized in the table below . for comparison
, we included a few `` typical '' values from the compilation of vlchez ( 1995 , v ) for star - forming dwarfs , from huchra ( 1977 , h ) for markarian galaxies , and from thuan ( 1992 , t ) for late - type dwarfs .
the metallicity was derived from the [ oii]+[oiii ] lines as measured from gh89 data and it compares well with the typical metallicity found by vlchez ( 1995 ) for his sample of dwarf galaxies .
we conclude that our observations , combined with data from the literature , indicate that a single , short and recent burst of star formation can account for the observed properties of this object .
ccc observed properties & vcc 144 & sf dg + v & 14.83 + b v & 0.46 & 0 .
4 ( v ) + u b & 0.51 & 0.36 ( h ) + v r & 0.22 & 0.081.17 ( h ) + r i & 0.23 + uv v & 1.33 + iras ? & weak + d@xmath98d@xmath99 & @xmath224``@xmath2512 '' + velocity [ km s@xmath15 ] : optical & 1960@xmath1452 + velocity [ km s@xmath15 ] : radio & 2014 + total h@xmath0 flux [ erg s@xmath15 @xmath4 ] & [email protected] 10@xmath20 + ( [ oii]+[oiii])/h@xmath16 & 7.78 + hi flux integral [ jy km s@xmath15 ] & 2.31 + + derived properties & vcc 144 & sf dg + d [ kpc ] & @xmath21 & @xmath1020.520.6 ( v ) + l@xmath60 [ l@xmath13 ] & 3.6 10@xmath51 + l(h@xmath104 ) [ erg s@xmath15 ] & 2.7 10@xmath96 & 10@xmath10510@xmath106 ( v ) + n(o7 ) & 2000 + m(hi ) [ m@xmath13 ] & 1.8 10@xmath51 & 10@xmath51 ( t ) + m(hi)/l@xmath60 [ m@xmath13/l@xmath13 ] & 2.0 & 2 , 4 ( t ) + m(dust ) [ m@xmath13 ] & 9.3 10@xmath56 + 12+log(o / h ) & @xmath1028.2 or 7.7 & 7.78.9 ( v ) + m@xmath47 [ m@xmath13 ] & 1.6 10@xmath107 + sfr
[ m@xmath13 yr@xmath15 ] & 0.3 + log(sfr / area ) [ m@xmath13 yr@xmath15 pc@xmath108 & 6.64 + log @xmath91 [ yrs ] & 8.78 + we mentioned in the introduction the unique object hi 1225 + 01 , a faint blue dwarf galaxy with large amounts of hydrogen and a companion hi cloud , which are probably interacting ( chengalur 1995 ) .
hi 1225 + 01 is only @xmath21.6 mpc away from vcc 144 ( in projected distance ) and both are located in the same general region of the virgo cluster .
other similar objects are some of the hii galaxies with hi companions mapped by taylor ( 1995 , 1996 ) , and some of the bcds found by van zee ( 1995 ) to have extended hi envelopes .
even the arch - type young galaxy i zw 18 shows an extended hi envelope ( skillman 1996 ) .
however , none are identical with vcc 144 . in many aspects ,
the optical counterparts of hi 1225 + 01 and i zw 18 are similar to vcc 144 .
the objects are blue and can be understood as recent starbursts .
salzer ( 1991 ) identified signs of an older starburst in hi 1225 + 01 , which enriched it with nitrogen and now produces at most @xmath230% of the b light .
chengalur ( 1995 ) concluded that the present burst of activity in hi 1225 + 01 was produced by a recent tidal interaction with the second hi lobe .
thus , its star forming activity could be explained , despite the object being relatively isolated .
vcc 144 appears to be similarly isolated from other optical galaxies at the southern outskirts of the virgo cluster , with ngc 4197 at @xmath254 kpc projected distance as nearest neighbor , but shows a much more intense star forming activity .
the bcds studied by van zee ( 1995 ) are sometimes extended in hi ( 16/41@xmath5440% of the cases ) .
an extended ( @xmath109 size of the optical galaxy ) , but symmetrical , envelope was found for the nearby irregular galaxy leo i ( young & lo 1996 ) . in most cases of diffuse hi near bcds
the hi extension is @xmath2symmetrical around the optical object .
however , in vcc 144 the diffuse hi appears in two extensions off the optical axis . as explained above , the iso - density contours of hi shown in figs .
[ fig4 ] and [ fig5 ] appear as two lobes on the extension of the galaxy minor axis .
the extensions , mainly the large one to the ne , represent large amounts of kinetic energy , just from considering the mass and velocity difference with respect to the hi on the optical counterpart ( @xmath25 10@xmath110 erg ) .
at least an equal amount of mechanical energy was probably injected as turbulent motion in the blown - out gas .
leitherer & heckman ( 1995 ) calculated a total mechanical energy injected by supernovae and stellar winds of @xmath21.6 10@xmath111 erg ( scaled to the present m@xmath94 and for an instantaneous starburst 6 10@xmath66 yrs old ) .
this indicates that a scenario of ism blow - out from vcc 144 , as a result of a recent starburst event , is a viable proposition .
1 . optical observations of vcc144 show a blue compact , elliptical , bright dwarf galaxy .
the strong h@xmath0 emission coincides and overlaps the broad band optical image of the galaxy .
hi mapping with the vla shows a main concentration coincident with the optical galaxy and an asymmetric , diffuse extension which has no optical counterpart , and which is extended along the minor axis of the galaxy .
rather large mass is required for the structure to be gravitationally confined .
comparisons with evolutionary synthesis models indicate that a most probable explanation is of a first burst of star formation in the last 10@xmath68 years .
the object does , however , contain dust .
4 . similarities with other dwarf galaxies with extended hi envelopes are noted .
we argue that a blow - out scenario is more appropriate for vcc 144 to explain the shape of the hi envelope , than either accretion , or a collision with an hi companion .
the conclusion is that galaxy formation , in the form of dwarf galaxies , takes place at present in the southern outskirts of the virgo cluster .
observations at the wise observatory are partly supported by a grant from the israel science foundation .
uv studies at the wise observatory are supported by special grants from the ministry of science and arts , through the israel space agency , to develop tauvex , a uv space imaging experiment , and by the austrian friends of tel aviv university .
glh was supported in part by us national science foundation grant ast-9316213 to lafayette college .
nb acknowledges the hospitality of prab gondhalekar and of the iras postmission analysis group at ral , as well as iras faint source catalog searches by rob assendorp .
we are grateful to the referee , daniel kunth , for constructive remarks which improved the quality of this paper .
freedman , w.l .
, madore , b.f .
, mould , j.r . , hill , r. , ferrarese , l. , kennicutt , r.c .
, saha , a. , stetson , p.b . ,
graham , j.a . ,
ford , h. , hoessel , j.g .
, huchra , j. , hughes , s.m . & illingworth , g.d .
1994 nature 371 , 757 .
|
we describe results of a multi - spectral study of a blue compact dwarf galaxy in the virgo cluster .
the object was observed with broad - band and h@xmath0 imaging , ultraviolet observations , and radio synthesis .
our data were combined with previously published optical observations , with hi single - beam observation and with far - infrared data , and were compared to results of evolutionary synthesis programs .
the radio synthesis observations revealed a compact concentration of hi coincident with the optical galaxy , embedded in a diffuse , asymmetric hi cloud which has no trace of optical emission . while the overall velocity dispersion and size of the hi structure suggests that the total mass in stars and gas is not sufficient to gravitationally bind the system as a whole , the hi clump coincident with the optical galaxy requires little or no dark matter to be self - gravitating .
the diffuse cloud has more complex velocity structure and is extended in a direction approximately perpendicular to the optical major axis .
the optical - uv data can be explained by a single population of stars formed in a recent burst , indicating that this is a genuine young galaxy .
the efficiency of star formation is similar to that in large disk galaxies .
the ir emission indicates the presence of dust ; this must have been formed very recently , or was already present within the original hi cloud from which the galaxy was formed .
the round and smooth isophotes , the correspondence of the optical and hi redshifts , and the lack of any suitable nearby galaxy , indicate that the starburst was probably not triggered by an external interaction with a visible galaxy .
the distribution of hi fits better a blow - out scenario than an accretion or collision with a companion .
we point out features in common with other actively star - forming dwarf galaxies and conclude that , at least in the southern outskirts of the virgo cluster , intensive star formation , perhaps for the first time in some objects , takes place at present .
|
in recent years several attempts at controlling turbulence through a number of wall - based forcing methods have been reported @xcite , often aimed at frictional drag reduction ( dr ) .
a large number of works , exploiting both numerical and experimental approaches , has been devoted to this goal .
in this paper we focus on spanwise wall - based forcing , i.e. on a class of forcing methods designed to modify favorably the turbulent flow by introducing an external action directed in the spanwise direction .
early work addressing the modification of wall turbulence by creating a cross flow can be traced back to bradshaw & pontikos @xcite .
spanwise forcing of turbulent flows has been reviewed by karniadakis & choi @xcite . in 1992 ,
jung , mangiavacchi & akhavan @xcite introduced the spanwise - oscillating wall technique , where the wall of a fully turbulent channel flow is subject to an alternate harmonic motion in the spanwise direction .
if @xmath0 indicates the spanwise velocity component at the wall , the law that defines this forcing method is : @xmath1 where @xmath2 is time , and @xmath3 and @xmath4 are the oscillation amplitude and period , respectively .
once @xmath3 and @xmath4 are set within the optimum range from the viewpoint of dr performance , the authors observed , by means of direct numerical simulations ( dns ) , that a strong suppression of turbulence occurs in the wall region , accompanied by a significant reduction of the mean friction . the analysis has been carried on in successive studies : we recall quadrio & sibilla @xcite for the pipe flow , and j .- i .
choi , xu & sung @xcite and quadrio & ricco @xcite , as dns - based analyses which contributed new datasets and detailed descriptions of the flow .
laboratory experiments , due to laadhari _
_ @xcite and k. s. choi @xcite , complemented the numerical works and addressed the issue of dependency of dr on the value of the reynolds number .
alltogether these works have contributed to demonstrating that the natural friction drag of the turbulent flow can be reduced ( at least at moderate values of @xmath5 ) up to 45% for @xmath6 ( quantities with the @xmath7 superscript are made dimensionless with viscous wall units ) .
an optimal value of the oscillation period exists , namely @xmath8 , that yields the maximum dr at all amplitudes .
the net energy saving , that substracts from the reduced flow - driving pumping power the power expense required to move the wall against viscous resistance , has been addressed first by baron and quadrio @xcite , and today it is recognized @xcite that it can reach up to 10% . crediting a commonly accepted qualitative explanation , the transverse oscillating boundary layer induced by the wall motion explains the drag reduction , since it produces a phase displacement between the wall - layer turbulence structures , capable of weakening the viscous wall cycle . when phase - averaged , this time - alternating layer in the turbulent regime has been found to coincide with the oscillating laminar stokes layer , for which an analytic solution is known as a classic solution of the boundary layer equations @xcite .
this has been recently exploited to determine a parameter @xcite that is capable of scaling linearly with dr and thus makes it possible to predict dr capabilities .
advantages and drawbacks of the oscillating - wall technique are obvious .
it presents energetic performance that could make it worth of practical implementations , and thanks to its open - loop character it does not need distributed sensors or actuators , that would be still unpractical with the technology available today . on the other hand , by its very nature
this technique requires moving parts , and thus does not lend itself to be implemented as a passive device , which is on the other hand the most appealing possibility application - wise . in this paper ,
we aim at extending the oscillating - wall technique , to take one further step towards the long - term objective of a successful practical implementation ; in particular we want to translate the time - dependent forcing law expressed by eq.([eq : time ] ) into a stationary formula .
this goal can be achieved by exploiting the convective nature of wall - bounded flows .
though at the wall the mean velocity profile annihilates , it is well known that the convection velocity of a turbulent wall flow ( or , more precisely , the convection velocity of turbulent fluctuations ) resembles the mean velocity profile only in the bulk of the flow .
kim & hussain @xcite have shown some years ago that near the wall , say below @xmath9 , the convection velocity becomes essentially independent upon the wall distance , and remains constant at the value @xmath10 .
it is this near - wall value of the convection velocity that will enable us to translate the temporal forcing from eq.([eq : time ] ) into a spatially - oscillating forcing , yielding the following forcing law that is expected to modify the turbulent flow in a similar manner to the oscillating wall : @xmath11 here @xmath12 denotes the streamwise coordinate , and @xmath13 is the forcing wavelength .
the resulting distribution of spanwise wall velocity is sketched in fig.[fig : sketch ] , where the coordinate system employed in this paper is also indicated . though the control law ( [ eq : space ] ) has never been considered in the literature , one paper where a similar space - time extension has been discussed in the past
is the one by berger and coworkers @xcite . in a parametric dns - based study
a spanwise - oriented lorentz volume force was simulated to obtain the following two forcing configurations : @xmath14 @xmath15 where @xmath16 is the penetration depth of the forcing , and @xmath17 its intensity . for certain values of the parameters ,
the streamwise - dependent law ( [ eq : berger - space ] ) produced a dr comparable to that of the time - dependent law ( [ eq : berger - time ] ) , with the added benefit of an improved energetic balance .
however , the difference between a body force and a wall - based forcing is substantial , as discussed for example by zhao et al .
@xcite , and no definite conclusion can thus be drawn _ a priori _ from the study by berger et al .
@xcite with respect to the oscillating velocity , which is strictly a wall - based forcing .
indeed , it will be shown later that some of the results described in ref .
@xcite ( for example the existence of an optimal wavelength that depends on the forcing intensity ) do not apply at all to the type of forcing considered here .
schoppa & hussain @xcite showed that a significant amount of drag reduction can be achieved by introducing in a channel flow a spanwise velocity gradient , generated in the dns by a large - scale , streamwise - aligned and @xmath12-independent rolls .
such velocity gradient weakens the near - wall cycle by suppressing the transient growth of streaks that would otherwise be stable according to normal - mode analysis .
this mechanism is then addressed in a subsequent paper @xcite , where the role of streamwise - varying spanwise perturbations is highlighted in the context of the turbulence regeneration cycle .
the dominant wavelength of the streak waviness is found to be very similar to the streamwise length of the coherent structures educed from conditional analysis of turbulent flow fields @xcite , namely about 300400 viscous units .
aim of the present paper is to investigate the dr and energetic performance of the steady control law ( [ eq : space ] ) .
thanks to an accurate data set , purposely obtained by several dns simulations , the @xmath18 parameter space will be explored in detail
. the value of the reynolds number will be fixed at @xmath19 ( based on the friction velocity of the reference flow and half the channel gap ) ; any dependence on this flow parameter is not discussed here .
the effects of the forcing ( [ eq : space ] ) on a laminar channel flow will be preliminarly studied .
we will see that a laminar solution , although approximated , can be obtained in analytical form , and this solution will then be used to help understanding the dr properties of the forcing ( [ eq : space ] ) when used on a turbulent flow . in the end
, we will be able to connect this forcing to a physical application that does not necessarily involve a moving wall .
the paper is organized as follows .
sec.[sec : laminar ] contains a theoretical discussion of the associated laminar flow , that is useful for predicting some of the energetic characteristics of the present technique ( further analytical details are deferred to an appendix ) . in sec.[sec
: num ] the numerical simulations of the turbulent case and their discretization parameters are described .
results about dr and energetic performances are reported and discussed in sec.[sec : results ] , together with a visual and statistical description of the forced flow , compared to the reference unperturbed one .
some conclusive remarks , including examples of how the present forcing can be implemented in practice , are given in the concluding sec.[sec : conclusions ] .
we consider first the laminar flow in a plane channel subject to either the temporal boundary forcing ( [ eq : time ] ) or the spatial boundary forcing ( [ eq : space ] ) .
for the temporal case , it is easy to show that the incompressible momentum equation for the spanwise component @xmath20 decouples , so that the streamwise flow is described by the classic poiseuille parabolic solution , and the entire flow consists in this parabolic profile plus a spanwise alternating motion , that is identical to the oscillating transversal boundary layer that develops in a still fluid bounded by a wall subject to harmonic oscillation ( i.e. the so - called stokes second problem , that possesses a classic analytical solution @xcite ) . this oscillating boundary layer will be called temporal stokes layer ( tsl ) in the following .
we address now the laminar flow subject to the space - varying boundary forcing ( [ eq : space ] ) , to define and discuss the spatial equivalent of the tsl , that will be called spatial stokes layer ( ssl ) .
we leave some analytical details to appendix [ sec : analytical ] , but the main ideas are given here as follows .
field for the laminar tsl and ssl .
top : temporal forcing @xmath21 .
bottom : spatial forcing @xmath22 .
contours at @xmath23 , dashed lines indicate negative values . ] in both the tsl and ssl , the field of the spanwise velocity component @xmath20 is a function of two independent variables , namely @xmath24 for tsl and @xmath25 for the ssl .
we start by observing in fig.[fig : wcont ] how closely these @xmath20 fields for tsl and ssl resemble each other .
the abscissa in fig.[fig : wcont ] is @xmath26 for the tsl , whereas it is changed to @xmath27 for the ssl .
both fields are computed with the dns code , although the analytical expression of the former is available @xcite : @xmath28 , \ ] ] where @xmath29 is a ( real ) normalization constant , @xmath30 is the angular frequency of the oscillation and @xmath31 is the thickness of the tsl , defined as : @xmath32 figure [ fig : wcont ] , where the parameters @xmath4 and @xmath13 are in the range of interest for turbulent dr purposes , illustrates the wall - normal structure of the tsl and the ssl . within this parameter range ,
both stokes layers are thin compared to the channel half - width @xmath33 .
the convection velocity being approximately constant over such a small wall - normal extension explains why the two contours in figure [ fig : wcont ] look very similar .
an analytical solution for the ssl can be arrived at under the small thickness approximation . after some analytical efforts , described in the appendix ,
the @xmath34 field of the laminar ssl can be shown to obey an expression , similar to eq.([eq : solution - tsl ] ) , that contains an airy function instead of an exponential function : @xmath35 , \ ] ] where @xmath36 is a ( real ) normalization constant , @xmath37 is the forcing wavenumber , @xmath38 is the airy function of the first kind @xcite , and @xmath39 is the thickness of the ssl , defined as : @xmath40 in this expression , @xmath41 represents the gradient of the streamwise mean velocity profile evaluated at the wall .
the @xmath34 laminar field computed by dns and plotted in fig.[fig : wcont ] ( bottom ) is virtually indistinguishable from the same field as computed from the analytical solution ( [ eq : solution - ssl ] ) . comparing the two expressions ( [ eq : solution - tsl ] ) and ( [ eq : solution - ssl ] )
, a qualitative difference between the ssl and the classical tsl can be observed : the former is not decoupled from the longitudinal mean velocity profile @xmath42 , but it depends on @xmath41 through the thickness @xmath39 given by ( [ eq : deltax ] ) .
this remains without consequences in the laminar case , where the streamwise flow is not affected by the wall forcing , whereas it will become important in the turbulent case .
fields for temporal and spatial forcing . left : analytical solution ( lines ) of the tsl and turbulent @xmath43 field ( symbols ) computed by dns .
right : analytical solution ( lines ) of the ssl and turbulent @xmath44 field ( symbols ) computed by dns .
profiles are taken at different ( temporal and spatial ) phases during the oscillation .
the laminar solutions are linear in @xmath3 , whereas the turbulent profiles are for @xmath45 .
turbulent temporal case is at @xmath19 and @xmath46 .
turbulent spatial case is at @xmath19 and @xmath47 . to emphasize collapse with proper scaling ,
the ssl analytical solution is computed for a different wavelength , namely @xmath48 , yielding a different @xmath39 , so that the role of @xmath49 as similarity variable is highlighted . ]
it is well documented @xcite for the time - oscillating wall that the phase - averaged @xmath50 profile is identical to the laminar solution expressed by eq.([eq : solution - tsl ] ) , except for the initial transient where the oscillation is started from rest .
( the operator @xmath51 indicates averaging along the homogeneous directions @xmath12 and @xmath52 . )
this is well illustrated by fig.[fig : w - lamturb ] ( left ) , where the wall - normal distribution of @xmath20 in the tsl after eq.([eq : solution - tsl ] ) is plotted at various phases during the cycle , and compared with the turbulent @xmath43 field .
the agreement , which has been related by ricco & quadrio @xcite to the vanishing @xmath53 component of the reynolds stresses tensor , is , as expected , excellent .
the same result is shown in fig.[fig : w - lamturb ] ( right ) to hold true for the ssl case .
now , of course , the analytical solution ( [ eq : solution - ssl ] ) has to be compared with the turbulent @xmath44 field .
the role of @xmath54 as similarity variable is well known for the tsl . to emphasize that @xmath49 plays the same role for the ssl ,
the comparison between laminar and turbulent profiles is carried out for two cases with different @xmath55 and thus , according to ( [ eq : deltax ] ) , with different @xmath39 .
the agreement between the two profiles is excellent .
this confirms the analogy between tsl and ssl in terms of the connection between the laminar solution and the phase - averaged turbulent solution .
this analogy can be used ( see [ sec : results ] ) to increase our predictive capabilities when the spatial forcing ( [ eq : space ] ) is applied to a turbulent flow .
we turn now to the turbulent case , that will be dealt with by using the direct numerical simulation ( dns ) technique . the computer code used to solve the incompressible navier
stokes equations is a parallel dns solver , based on mixed discretization ( fourier expansion in the homogeneous directions , and compact fourth - order accurate explicit compact finite difference schemes in the wall - normal direction ) , recently developed by luchini and quadrio @xcite .
the turbulent flow in an indefinite plane channel is considered , with gap half - width given by @xmath33 .
one reference simulation without forcing has been carried out at @xmath19 , where @xmath56 is the friction reynolds number , defined based on @xmath33 and the friction velocity @xmath57 , that will be used as reference velocity throughout the paper unless otherwise noted .
the size of the computational domain is @xmath58 , @xmath59 and @xmath60 .
the total averaging time is rather large , to allow for well converged time - averaged results , and amounts to approximately @xmath61 viscous time units . the employed number of modes / points is given by @xmath62 , @xmath63 and @xmath64 , that yields a standard spatial resolution . for the reference simulation a value of @xmath65 ( defined based on the bulk velocity )
is obtained , in agreement with the correlation @xmath66 given by pope @xcite .
some 40 simulations with eq.([eq : space ] ) used as boundary condition at the channel walls have been then carried out for different values of @xmath13 and @xmath3 .
owing to the periodic boundary conditions employed in the homogeneous directions , an integer number of forcing wavelength @xmath13 must be contained in the domain length @xmath67 .
the systematic variation of @xmath13 has thus required , mostly for the largest values of @xmath13 , slight adjustments of @xmath67 .
compared to the reference value of @xmath68 , 3 cases had an actual value of @xmath69 , and one was at @xmath70 . since such changes
are of limited entity , the number of streamwise fourier modes has been left unchanged , with the implied little change in streamwise spatial resolution for these few cases . properly accounting for the initial transient , where the wall friction decreases from the unperturbed value to the reduced value given by the forcing ,
is enforced according to the procedure described by quadrio & ricco @xcite , who reported an overall error in determining the friction coefficient below 1% .
we have also verified that the results are not affected by the chosen value of @xmath67 .
one simulation with @xmath71 and large dr has been repeated by doubling @xmath67 at @xmath72 ( and of course doubling the number of streamwise fourier modes ) .
the value of @xmath73 measured in the simulation with the longer domain differs from the one computed by the simulation with standard domain length by much less than the error , mentioned above , related to the finite averaging time .
the simulations were run on a computing system available in dedicated mode at the universit di salerno , taking advantage of its large computing throughput to run several cases at a time .
the system possesses 64 computing nodes , each of which is equipped with two dual - core amd opteron cpus .
the single computational case took about 10 days of wall - clock time when run in parallel by using 8 nodes .
up to 8 cases can be run simultaneously , so that the wall - clock time for the entire study was about 7 weeks .
the effectiveness of the steady spatial forcing ( [ eq : space ] ) in reducing the frictional drag is assessed by examining the value of the skin - friction coefficient obtained in several different simulations , in which the amplitude @xmath3 and the wavelength @xmath13 of the wall forcing are systematically varied .
the measured coefficients are then compared to the value of the reference ( unforced ) flow .
the savings in driving power will then be compared to the energetic cost of the wall velocity distribution , in order to asses the net power saving made possible by the spatial forcing .
our simulations are performed at a fixed flow rate , so that a decrease of the frictional drag translates into a proportional decrease of the mean streamwise pressure gradient and of the power required to drive the flow .
the power to drive the flow against viscous resistance is defined as : @xmath74 where @xmath75 and @xmath76 are the space - averaged value of the streamwise component of the wall shear stress , evaluated at the lower and upper wall respectively , and @xmath77 is the time interval over which the time averaged is carried out , after discarding initial transients .
the percentage saved power @xmath78 is expressed as percentage of the power @xmath79 required to drive the flow in the reference case , and is defined as follows : @xmath80 the quantity @xmath78 exactly corresponds to the percentage of friction drag reduction , and is expected to be a function of @xmath3 and @xmath13 .
as a function of the forcing wavelength .
empty symbols are temporal cases from quadrio & ricco @xcite : for them @xmath81 ( see text ) . ] in fig.[fig : psav - lambda ] @xmath78 is plotted first as a function of @xmath82 , for different values of the amplitude @xmath83 .
available data @xcite obtained with temporal forcing are also plotted , with the oscillation period translated into an oscillation wavelength through the near - wall value of the convection velocity , namely @xmath84 .
the capital information drawn from this plot is the validation of our working hypothesis : the present steady forcing indeed parallels the unsteady oscillating - wall technique when oscillation period and forcing wavelength are related through @xmath85 .
in analogy to the oscillating wall , that yields the maximum dr at a well - defined oscillation period @xmath86 , namely @xmath87 , the steady forcing yields the maximum dr at a well - defined wavelength @xmath82 , namely @xmath88 .
it is striking how well the prediction @xmath89 is confirmed by our simulations .
dr is observed over a very wide range of wavelenghts , @xmath90 , analogously to the oscillating wall .
the effects of the ssl on the turbulent flow are , however , only qualitatively similar to those of the tsl .
quantitative differences do exist , and in particular at a given wavelength the spatial forcing is observed to attain larger values of dr when compared to the temporal forcing at the same amplitude and equivalent period . as a function of forcing amplitude . as in fig.[fig : psav - lambda ] , empty symbols are temporal cases from quadrio & ricco @xcite : for comparison recall that @xmath91 ( see text ) . ]
fig.[fig : psav - a ] shows @xmath78 as a function of the amplitude @xmath83 , for different values of @xmath82 .
the plot contains data from spatial as well as from temporal forcing cases .
the dependence of dr on the forcing amplitude is qualitatively very similar for tsl and ssl : dr monotonically grows with @xmath3 , but the increase saturates to an apparently asymptotic behavior .
again , the space - dependent forcing appears to be more effective in reducing turbulent friction : for a given amplitude the spatial forcing yields a @xmath92 larger dr than the temporal forcing .
a maximum dr of about @xmath93 is observed at @xmath47 with a forcing amplitude of @xmath94 .
in addition to @xmath95 , the forced channel flow has an additional power input @xmath96 , that is required to enforce the wall motion against the spanwise shear stress .
@xmath96 is defined as : @xmath97 where @xmath98 is the space - averaged value of the spanwise component of the wall shear stress , and @xmath0 is the spanwise velocity of the walls . as a function of the forcing wavelength . as in fig.[fig : psav - lambda ] , empty symbols are temporal cases from quadrio & ricco @xcite . the line - connected asterisks represent @xmath99 obtained from the laminar solution with @xmath45 , from formula ( [ eq : solution - tsl ] ) for the tsl and from formula ( [ eq : solution - ssl ] ) for the ssl .
[ fig : preq - lambda ] ] the required percentage power @xmath99 is defined in terms of the friction power @xmath79 of the reference flow as @xmath100 .
@xmath99 is presented in fig.[fig : preq - lambda ] as a function of the forcing wavelength . again
, spatial as well as temporal forcing data are included .
of course @xmath99 assumes negative values , i.e. work has to be done against the fluid viscosity . comparing the two forcing methods , one can easily appreciate how the spatial forcing presents an energetic cost that is smaller than the cost of temporal forcing , by approximately a factor of 2 .
the line - connected points in fig.[fig : preq - lambda ] represent the values @xmath99 that can be computed from laminar theory .
tsl and ssl are considered at @xmath45 , for which formulas ( [ eq : solution - tsl ] ) and ( [ eq : solution - ssl ] ) respectively are analytically integrated .
the good agreement was expected , since it was already observed ( see fig.[fig : w - lamturb ] ) that laminar and turbulent mean profiles of spanwise velocity are coincident . as a function of the forcing wavelength .
[ fig : pnet - lambda ] ] lastly , a net percentage power saving @xmath101 is easily defined by comparing @xmath99 and @xmath78 , as follows : @xmath102 the net power saving @xmath101 is plotted in fig.[fig : pnet - lambda ] . since it has already been observed how ssl yields higher dr at lower energetic cost , of course here a significant difference in net gain is expected when comparing ssl and tsl .
@xmath101 shows indeed an interesting maximum value of about @xmath103 at @xmath104 , while remaining positive over a wide range of wavelengths .
moreover , positive @xmath101 are found for rather large amplitudes : at @xmath45 a net gain of about 5% can still be observed , whereas the tsl at the same amplitude presents a net loss of about 30% .
it is worth noting that the search for the maximum of @xmath101 can not be considered exhaustive , and thus the presently observed maximum value of 23% at @xmath104 should be regarded as a starting point for a refined search . .
flow is from left to right .
top : reference flow ; bottom : spatial forcing with @xmath47 and @xmath45 . ] observing the main statistics of the flow and even a few instantaneous snapshots may help understanding how ssl affects turbulence .
( additional statistical quantities are presented elsewhere @xcite ) . in fig.[fig : ucont ] isosurfaces are visualized for the @xmath105 quantity introduced by jeong & hussain @xcite and since then often used to identify turbulent vortical structures .
the level is set at @xmath106 .
the top plot is for the reference flow , and the bottom plot is for the same flow subject to the effects of the spatial forcing , with @xmath107 and @xmath108 .
this is one of the cases with the highest drag reduction , i.e. 45% .
the ssl clearly appears to modify the near - wall turbulence and its structures . .
] similar considerations can be drawn from single realizations of the flow as well from its statistical description .
fig.[fig : uav ] reports the mean velocity profile in the law - of - the - wall form .
the modification to the profile for the tsl and ssl cases are analogous when compared to the reference flow , but the effects are larger for the ssl .
the drag reduction manifests itself through the thickening of the viscous sublayer , that results in the upward shift of the logarithmic portion of the velocity profile , as previously documented for other dr techniques , for example riblets @xcite .
another relevant statistical quantity that is modified by the action of the ssl is the turbulence intensity .
fig.[fig : urms ] presents the r.m.s .
value of the fluctuations of the streamwise velocity component . to examine the effect of the lower reynolds number implied by the strong drag reduction in a constant - flow - rate simulation
, this plot is still scaled using inner variables , but both the friction velocity of the reference flow and the actual friction velocity of the drag - modified flow are used as velocity scale .
the strong reduction of fluctuations is deemphasized with the latter form of scaling , and what remains is the structural effect of the ssl after subtracting the effect due to a smaller @xmath56 .
the curves however remain significantly different ; the main effect is a decrease of the fluctuation intensity , together with a displacement of the position of the maximum intensity further form the wall .
similar effects have been already documented for other dr methods .
in particular the same observation has been put forward @xcite for the oscillating - wall .
it may be useful to remind the reader that effective dr methods exist , for example the active opposition control @xcite , where @xmath109 is unaffected when properly scaled ; on the other hand , as discussed for example by jimnez @xcite , experimental evidence exists that wall roughness may reduce the near - wall peak of turbulence intensities while increasing drag .
this paper has studied a new form of boundary forcing for wall - bounded turbulent flows , that consists in imposing at the wall a steady distribution of spanwise velocity , modulated in the streamwise direction . in this study only sinusoidal modulation has been considered .
main motivation was to find a steady counterpart to the oscillating - wall technique .
the link between the two kinds of forcing is the convection velocity of the turbulence fluctuations , that takes a well - defined non - zero value @xmath85 at the wall and is thus capable of transforming a time scale into a length scale and viceversa .
thanks to a number of direct numerical simulations , the behavior of this new forcing in the parameter space has been determined , and dr up to @xmath93 has been observed for @xmath94 and @xmath47 . for all amplitudes , the forcing wavelength that yields the maximum dr has been found to correspond to the optimal period of the oscillating wall converted in length through @xmath85 , thus confirming the validity of the analogy between temporal and spatial forcing . this analogy has been extended further by studying the laminar case : this was known to be relevant to the oscillating - wall technique , since in the turbulent case the spanwise profile after space - time averaging is identical to the laminar solution .
the laminar solution , that can be written in terms of airy functions , has been determined for the spatial case too , and it has been additionally verified that the turbulent spanwise flow when phase - averaged is identical to the laminar solution .
this property can be leveraged to predict the power required for the control .
we hope that a predictive quantity , similar to the parameter discussed by quadrio & ricco @xcite and capable of describing the dr effects of the forcing , could be envisaged on the basis of this analytical solution .
together with qualitative analogies , there are quantitative differences between temporal and spatial forcing .
the spatial forcing is more efficient in terms of dr , from the point of view of both absolute dr and net power saving . in particular , a net power saving as high as @xmath103 has been computed at @xmath104 , with one unit of forcing power translating into 3 units of saved power .
this is more than 3 times the largest net saving documented for the oscillating wall , and also significantly larger than the benefit obtained with passive devices like riblets , that are reportedly capable of a saving up to 8 - 10% in laboratory conditions @xcite .
the present form of spatial forcing is certainly realizable in principle , and an experimental setup is indeed under construction that will help shedding light on effects like the dependence of dr on @xmath5 . at the same time
, we do not consider the present forcing , though steady , directly suitable as yet for practical applications .
however , the successful design of a steady control law is one important step towards the realization of a passive drag - reducing device . in this framework ,
the sensitivity of the turbulence near - wall cycle to a well - defined streamwise lengthscale is a fundamental result , that paves the way to the search for an efficient drag - reducing type of roughness . in this context , we have recently applied for a patent @xcite concerning a design method of roughness distributions yielding improved benefits over conventional riblets .
such distributions may contain riblets with sinusoidally ondulated crests with wavelength @xmath110 . a later paper by peet _
_ @xcite has confirmed by numerical simulations the potential success of similar sinusoidal riblets , reporting a 50% improvement over conventional riblets . since in @xcite
the wavelength was in the optimum range , but only a single amplitude was tested , we are confident that even larger benefits can be attained by such a purely passive technique .
cv has been supported by italian ministry of university and research through the grant prin 2005 on _ large scale structures in wall turbulence_. we acknowledge interesting discussions with dr p.ricco .
part of this work has been presented by cv in june 2007 at the xi european turbulence conference , porto ( p ) .
in this appendix an analytical , approximate solution of the navier stokes equations for the laminar flow between indefinite plane walls is described , where the non - homogeneous boundary condition @xmath111 is imposed to the spanwise velocity component .
the boundary forcing creates a layer of alternate spanwise motion , which develops close to the wall and resembles the temporal stokes layer .
we have referred to it in this paper as the spatial stokes layer ( ssl ) , in comparison to the conventional temporal stokes layer ( tsl ) .
an analytical solution will be derived now for the velocity profile of the ssl , under the assumption that the wall - normal length scale characteristic of the layer is small in comparison to the channel width .
the solution is steady , and thanks to the spanwise invariance of the differential system , including its boundary conditions , all the @xmath52 derivatives can be dropped from the momentum equations , which then read : this highlights that eqs.([eq : n - s4 ] ) and ( [ eq : n - s5 ] ) decouple from eq.([eq : n - s6 ] ) to form an independent two - dimensional problem , unaffected by the inhomogeneous boundary condition ( [ eq : bc - z ] ) .
its solution is thus the classical laminar poiseuille solution , that gives a parabolic longitudinal velocity profile , and predicts a wall - normal velocity @xmath115 everywhere .
thus eq.([eq : n - s6 ] ) can be further simplified as : @xmath116 at this point a boundary - layer approximation is introduced .
we suppose the ssl to be confined in a thin region close to the wall , and to vanish at a distance definitely smaller than @xmath33 .
if @xmath39 indicates the characteristic thickness of the ssl , we are requiring that @xmath119 . even before giving a precise definition of @xmath39 , we have seen in fig.[fig : wcont ] that this requirement is satisfied when the flow parameters are set within the range of interest ( which is where the spatial forcing achieves a substantial dr in the turbulent regime ) .
if @xmath13 is comparable or larger than @xmath33 , the boundary - layer approximation implies also that @xmath122 in eq.([eq : prob1 ] ) is negligible compared to @xmath123 .
we are thus left with the well - posed problem : @xmath124 with boundary conditions : @xmath125 to be solved in the domain @xmath126 .
infinite solutions of an airy equation exist for @xmath138 spanning the whole complex plane , when derivatives are considered in the sense of analytic functions .
these solutions are linear combinations of the two special functions @xmath139 and @xmath140 @xmath141 which are called airy functions of the first and second kind respectively .
the general solution ( [ eq : combination ] ) has an alternate representation @xcite , which turns out to be useful in our case : @xmath142 where @xmath143 .
it can be shown @xcite that among @xmath139 , @xmath140 , @xmath144 and @xmath145 , the only base function satisfying conditions ( [ eq : bc3 ] ) ( up to a normalization factor ) is @xmath145 . as a consequence ,
the solution that satisfies the required boundary conditions is : @xmath146 m. quadrio , c. viotti , and p. luchini .
skin - friction drag reduction via steady streamwise oscillations of spanwise velocity . in j.m.l.m .
palma and silva lopes a. , editors , _ advances in turbulence xi _ , advances in turbulence xi , pages 659661 .
springer , berlin , 2007 .
|
steady forcing at the wall of a channel flow is studied via dns to assess its ability of yielding reductions of turbulent friction drag . the wall forcing consists of a stationary distribution of spanwise velocity that alternates in the streamwise direction . the idea behind the forcing builds upon the existing technique of the spanwise wall oscillation , and exploits the convective nature of the flow to achieve an unsteady interaction with turbulence .
the analysis takes advantage of the equivalent laminar flow , that is solved analytically to show that the energetic cost of the forcing is unaffected by turbulence . in a turbulent flow
, the alternate forcing is found to behave similarly to the oscillating wall ; in particular an optimal wavelength is found that yields a maximal reduction of turbulent drag . the energetic performance is significantly improved , with more than 50% of maximum friction saving at large intensities of the forcing , and a net energetic saving of 23% for smaller intensities .
such a steady , wall - based forcing may pave the way to passively interacting with the turbulent flow to achieve drag reduction through a suitable distribution of roughness , designed to excite a selected streamwise wavelength .
|
this paper deals with moving solitons of the discrete nonlinear schrdinger ( dnls ) equation .
the earliest applications of the cubic dnls equation were to the self - trapping of electrons in lattices ( the polaron problem ) and energy transfer in biological chains ( davydov solitons ) see the reviews @xcite for references .
relatedly , the equation arises in the description of small - amplitude breathers in frenkel - kontorova chains with weak coupling @xcite . in optics
the equation describes light - pulse propagation in nonlinear waveguide arrays in the tight - binding limit @xcite .
most recently the dnls equation has been used to model bose - einstein condensates in optically induced lattices @xcite .
the question of the existence of moving solitons in the dnls equation has been the subject of debate for some time @xcite .
recently , gmez - gardees , flora , falo , peyrard and bishop @xcite have demonstrated that the stationary motion of pulses in the cubic one - site dnls ( the ` standard ' dnls ) is only possible over an oscillatory background consisting of a superposition of plane waves .
this result was obtained by numerical continuation of the moving ablowitz - ladik breather with two commensurate time scales . in our present paper ,
we study the travelling discrete solitons analytically and independently of any reference models .
consistently with the conclusions of @xcite , we will show that solitons can not freely move in the cubic dnls equation ; they emit radiation , decelerate and eventually become pinned by the lattice .
we shall show , however , that this radiation is exponentially small in the soliton s amplitude , so that broad , small - amplitude pulses are highly mobile and are for all practical purposes indistinguishable from freely moving solitons . in the context of optical waveguide arrays
important not only in themselves but also as a first step to understanding more complicated optical systems such as photonic crystals
interest among experimentalists @xcite has recently shifted away from media with pure - kerr nonlinearity ( which gives rise to a cubic term in the dnls equation ) and towards photorefractive media , which exhibit a saturable nonlinearity @xcite . in practice
such arrays may be optically induced in a photorefractive material @xcite or fabricated permanently
see @xcite , for example .
the study of solitons in _ continuous _ optical systems with saturable nonlinearitites has a long history ; interesting phenomena here include bistability @xcite , fusion @xcite and radiation effects @xcite which do not arise in the cubic equation . as for the _ discrete _ case ,
the work of khare , rasmussen , samuelsen and saxena @xcite suggests that the saturable one - site dnls may be _ exceptional _ in the sense of ref.@xcite ; that is , despite not being a translation invariant system , it supports translationally invariant stationary solitons .
this property is usually seen as a prerequisite for undamped motion in discrete equations ( see e.g. @xcite ) and indeed , the numerical experiments of vicencio and johansson @xcite have revealed that soliton mobility is enhanced in the saturable dnls equation .
the dnls equation with a saturable nonlinearity is the second object of our analysis here ; our conclusions will turn out to be in agreement with the numerical observations of ref.@xcite .
we will show that for nonlinearities which saturate at a low enough intensity , solitons can _ slide_that is , move without radiative deceleration at certain isolated velocities . these ` sliding ' solitons are examples of embedded solitons .
the usual saturable dnls equation is the discrete form of the vinetskii - kukhtarev model @xcite : @xmath0 in order to encompass both cubic and saturable nonlinearities in a single model , we shall instead consider the equation @xmath1 obtained from eq . by making the transformation @xmath2 and letting @xmath3 . in the form above
, @xmath4 represents the saturation threshold of the medium @xcite , which tends to infinity as one approaches the pure kerr ( cubic ) case of @xmath5 .
the higher the value of @xmath6 , the lower is the intensity at which the nonlinearity saturates .
this paper is structured in the following way . in section [ section2 ] , we construct a small - amplitude , broad travelling pulse as an asymptotic series in powers of @xmath7 , its amplitude . the velocity and frequency of this soliton are obtained as explicit functions of @xmath7 and its carrier - wave wavenumber . then in section [ section3 ] , the main section of this paper , we derive an expression for the soliton s radiation tails and measure their amplitude using the method of asymptotics beyond all orders . in section [ section4 ] ,
we investigate the influence of this exponentially weak radiation on the soliton s amplitude and speed .
finally , in section [ section5 ] , we summarise our work and make comparisons with some earlier results .
we begin by seeking solutions of the form @xmath8 where @xmath9 and @xmath7 is a parameter . by analogy with the soliton of the continuous nls , we expect the discrete soliton to be uniquely characterised by two parameters e.g .
@xmath7 and @xmath10while the other two ( @xmath11 and @xmath12 ) are expected to be expressible through @xmath7 and @xmath10 .
substituting the ansatz , into gives a differential advance - delay equation @xmath13 this can be written as an ordinary differential equation of an infinite order : @xmath14 where @xmath15 .
from now on we assume that @xmath7 is small .
our aim in this section is to find an approximate solution to eq.and hence eq.with @xmath16 .
( that is , we are looking for small , broad pulses which modulate a periodic carrier wave . ) to this end , we expand @xmath17 , @xmath11 and @xmath12 as power series in @xmath7 : [ expansions ] @xmath18 ( we are not expanding @xmath10 as we consider it , along with @xmath7 , as one of the two independent parameters characterising our solution . ) substituting these expansions into eq . gives a hierarchy of equations to be satisfied at each power of @xmath7 by choosing @xmath19 and @xmath20 properly .
in nonlinear oscillations , this perturbation procedure is known as lindstedt s method @xcite . at the order @xmath21
we obtain @xmath22 while the order @xmath23 gives @xmath24 these two relations correspond to the dispersion of linear waves . at the power @xmath25
we obtain the following nonlinear equation for @xmath26 : @xmath27 this is the stationary form of the nls equation , which has the homoclinic solution @xmath28 with @xmath29 returning to the original variable @xmath17 , we note that the amplitude @xmath30 can always be absorbed into @xmath7 , the parameter in eq.([eps_wid ] ) and in front of @xmath26 in ( [ exp_psi ] ) .
that is , there is no loss of generality in setting @xmath31 and letting @xmath7 describe the amplitude ( and inverse width ) of the pulse instead .
this allows us to set @xmath32 note that the coefficients in eq .
are periodic functions of the parameter @xmath10 with period @xmath33 ; therefore it is sufficient to consider @xmath10 in the interval @xmath34 .
also , is invariant with respect to the transformation @xmath35 , @xmath36 , @xmath37 ; hence it is sufficient to consider positive @xmath10 only .
finally , our perturbative solution does not exist if @xmath38 is negative .
thus , from now on we shall assume that @xmath39 . at the order @xmath40 , where @xmath41 , we arrive at the following equations for the real and imaginary parts of @xmath42 : [ genlinear ] @xmath43 where
@xmath44 and @xmath45}\left(\frac{2\cos k}{(2j+2)!}\psi_{n-2j}^{(2j+2)}-\omega_{2j+2}\psi_{n-2j}\right ) + i\sum_{j=1}^{[\frac{n+1}{2}]}\left(\frac{2\sin k}{(2j+1)!}\psi_{n-2j+1}^{(2j+1)}-v_{2j}\psi'_{n-2j+1}\right ) \\ + \sum_{m=1}^{n-1}2\psi_0(\psi_m{{\psi}^{\ast}}_{n - m}-\psi_{n - m}{{\psi}^{\ast}}_m ) + \sum_{m=1}^{n-1}\sum_{\ell=0}^{n-1-m}2\psi_{n - m-\ell}\psi_m{{\psi}^{\ast}}_{\ell}\\ + \mu\sum_{m=0}^{n-1}\sum_{\ell=0}^{n-1-m}\bigg [ \sum_{j=1}^{[\frac{n - m-\ell}{2}]}\left(\frac{2\cos k}{(2j)!}\psi^{(2j)}_{n - m-\ell-2j } -\omega_{2j}\psi_{n - m-\ell-2j}\right)\\ + i\sum_{j=1}^{[\frac{n - m-\ell-1}{2}]}\left(\frac{2\sin k}{(2j+1)!}\psi^{(2j+1)}_{n - m-\ell-2j-1 } -v_{2j}\psi'_{n - m-\ell-2j-1}\right ) \bigg]\psi_m{{\psi}^{\ast}}_{\ell}. \label{all_p}\end{gathered}\ ] ] the linear nonhomogeneous ordinary differential equations must be solved subject to a boundedness condition .
the bounded homogeneous solutions of eqs and ( @xmath46 and @xmath47 , respectively ) correspond to the translation- and u(1)-invariances of eq .. including these zero modes in the full solution of eqs would amount just to the translation of @xmath48 by a constant distance in @xmath49 and its multiplication by a constant phase factor .
these deformations are trivial , and hence we can safely discard the homogeneous solutions at each order of @xmath7 . as @xmath26 s real part is even and its imaginary part is odd ( zero ) , @xmath50 s real and imaginary parts are also even and odd , respectively .
the same holds , by induction , to all orders of the perturbation theory .
indeed , assume that @xmath51 have even real parts and odd imaginary parts .
then it is not difficult to verify that the function @xmath52 is even and @xmath53 is odd . since the operators @xmath54 and @xmath55 are parity - preserving , and since we have excluded the corresponding homogeneous solutions , this means that @xmath42 has an even real part and an odd imaginary part . finally , the homogeneous solution of being even and that of being odd , the corresponding solvability conditions are satisfied at any order . note that since the solvability conditions do not impose any constraints on @xmath20 and @xmath19 , the coefficients @xmath20 with @xmath56 and @xmath19 with @xmath57 can be chosen completely arbitrarily . solving eqs successively ,
we can obtain the discrete soliton to any desired accuracy . here
, we restrict ourselves to corrections up to the cubic power in @xmath7 .
the order @xmath25 is the lowest order at which the saturation parameter @xmath6 appears in the solution . on the other hand ,
it is high enough to exemplify and motivate our choice of the coefficients in and . letting @xmath58 in eq .
, we have @xmath59 the corresponding solution of eq .
is @xmath60.\end{aligned}\ ] ] here the term proportional to @xmath61 decays to zero as @xmath62 ; however , it becomes greater than @xmath63 for sufficiently large @xmath64 , leading to nonuniformity of the expansion . in order to obtain a uniform expansion ,
the term in question should be eliminated .
being free to choose the coefficients @xmath20 with @xmath56 , we use this freedom to set @xmath65 this leaves us with @xmath66 after ` distilling ' in a similar way the correction @xmath67 , where we fix @xmath68 to eliminate a term proportional to @xmath69 , we obtain @xmath70 + { \cal o}(\epsilon^3)\big\},\end{aligned}\ ] ] where @xmath71\cos k , \label{omegadef } \\ v & = 2\left[1+\frac{1}{3!}\epsilon^2+{\cal o}(\epsilon^4)\right]\sin k. \label{cdef}\end{aligned}\ ] ] in the previous subsection we have shown that fixing suitably the coefficients @xmath19 and @xmath20 can lead to a uniform expansion of @xmath17 to order @xmath25 . our approach was based on solving for @xmath42 explicitly and then setting the coefficients in front of the ` secular ' terms @xmath69 to zero . here , we show that the secular terms can be eliminated to all orders and without appealing to explicit solutions .
assume that the secular terms have been suppressed in all nonhomogeneous solutions @xmath72 with @xmath73 up to @xmath74 ; that is , let @xmath75 ( since we know that the real part of @xmath72 is even and the imaginary part odd , it is sufficient to consider the asymptotic behaviour at one infinity only . ) the constant @xmath76 may happen to be zero for some @xmath73 in which case the decay of @xmath72 will be faster than @xmath77 .
our objective is to choose @xmath19 and @xmath20 in such a way that @xmath78 will also satisfy . to this end
, we consider the function .
all terms which are trilinear in @xmath79 and derivatives of these functions decay as @xmath80 or faster ; these terms in @xmath81 can not give rise to the secular terms proportional to @xmath69 in @xmath42 .
on the other hand , the terms making up the first line in tend to @xmath82 } \left [ \frac{2\cos k}{(2j+2 ) ! } -\omega_{2j+2}\right ] c_{n-2j } \\ & \quad + i e^x \sum_{j=1}^{[\frac{n+1}{2}]}\left [ \frac{2\sin k}{(2j+1 ) ! } -v_{2j}\right ] c_{n-2j+1}\end{aligned}\ ] ] as @xmath83 . these are ` asymptotically resonant ' terms in the sense that their asymptotics are proportional to the asymptotics of the homogeneous solutions of and .
it is these terms on the right - hand sides of and that give rise to the secular terms in the solution @xmath42 .
the resonant terms will be suppressed if we let @xmath84 after the resonant terms have been eliminated , the bounded solution to will have the asymptotic behaviour @xmath85 as @xmath83 . by induction , this result extends to all @xmath86 . substituting , together with , into eqs and , and summing up the series , we obtain @xmath87 the frequency and velocity of the discrete soliton parameterised in terms of @xmath10 and @xmath7 .
equations coincide with the expressions @xcite of the ablowitz - ladik soliton s velocity and frequency in terms of its amplitude and wavenumber .
the difference between the two sets of answers is in that eqs pertain to small - amplitude solitons only , whereas ablowitz and ladik s formulas are valid for arbitrarily large amplitudes .
note , also , that the velocity and frequency do not depend on the saturation parameter @xmath6 .
this is in contrast to the stationary ( @xmath88 ) soliton of eqs and obtained by khare , rasmussen , samuelsen and saxena @xcite .
the soliton of ref.@xcite has its amplitude and frequency uniquely determined by @xmath6 .
as @xmath83 , the series reduces to @xmath89 . the convergence of the series @xmath90 for all @xmath49 would imply , in particular , the convergence of the series @xmath91
therefore , if the series converged , the solution @xmath48 would be decaying to zero as @xmath83 . however , although we have shown that the series @xmath90 is asymptotic to all orders , it does not have to be convergent .
for instance , it is easy to see that the series can not converge for @xmath92 and any @xmath7 ; since the advance - delay equation is translation invariant , this would imply that we have constructed a family of stationary solitons with an arbitrary position relative to the lattice .
this , in turn , would contradict the well established fact that the ` standard ' cubic dnls solitons can only be centered on a site or midway between two adjacent sites @xcite ( that is , that the ` standard ' discretisation of the cubic nls is not exceptional @xcite ) .
thus , we expect that the perturbative solution satisfies @xmath93 as @xmath62 only for some special choices of @xmath12 , @xmath7 and @xmath6
. can eqs and have a bounded solution despite the divergence of the corresponding series @xmath94 ? to gain some insight into this matter , we linearise eq . about @xmath95 and
find nondecaying solutions of the form @xmath96 where @xmath97 is a root of the dispersion relation @xmath98 it is easy to check that there is at least one such harmonic solution [ i.e. eq
. has at least one root ] if @xmath99 .
these harmonic waves can form a radiation background over which the soliton propagates ( as suggested by the numerics of @xcite and the analysis of a similar problem for the @xmath100 kinks in @xcite ) .
being nonanalytic in @xmath7 , such backgrounds can not be captured by any order of the perturbation expansion .
as we will show below , not only the wavenumbers but also the amplitudes of the harmonic waves are nonanalytic in @xmath7 .
this phenomenon was first encountered in the context of the breather of the continuous @xmath100 model , where eleonskii , kulagin , novozhilova , and silin @xcite suggested that the radiation from the breather could be exponentially weak .
segur and kruskal @xcite then developed the method of ` asymptotics beyond all orders ' to demonstrate that , in the limit @xmath101 , such radiation does exist .
we will use segur and kruskal s method to measure the magnitude of the radiation background of the travelling discrete soliton .
qualitatively , the fact that the radiation is not excited at any order of the perturbation expansion is explained by the fact that the soliton exists on the long length scale @xmath49 , whereas the radiation has the shorter scale @xmath102 . to all orders , the two are uncoupled . using the relations , we can rewrite as @xmath103 the left hand side is plotted in fig.[nicedisprelfig ] .
for @xmath104 it has minima at multiples of @xmath105 where the curve is tangent to the horizontal axis . for nonzero @xmath7 ,
the minima of the curve are lifted off the @xmath97 axis slightly .
the minima with larger values of @xmath97 have smaller elevations above the horizontal axis ; i.e. , the minima come closer and closer to the @xmath97 axis as @xmath97 grows .
we see from the figure that for @xmath106 , where @xmath107 , there is only one radiation mode .
note also that the left - hand side of eq .
is negative for negative @xmath97 ; since we have assumed that @xmath108 , this implies that eq . can not have negative roots . the left hand side of eq .
for two values of @xmath7 .
the root(s ) @xmath109 of the dispersion relation are located where this graph is intersected by a horizontal straight line with ordinate equal to @xmath110 . ]
although we originally constructed the expansion as an asymptotic approximation to a solution which is stationary in the frame of reference moving with the velocity @xmath12 , it can also represent an approximation to a time - dependent solution @xmath111 . here
@xmath111 is related to @xmath112 , the discrete variable in eq .
, by the substitution : @xmath113 the coefficients @xmath42 in the asymptotic expansion of @xmath111 will coincide with the coefficients in the expansion of the stationary solution @xmath114 if the time derivatives @xmath115 lie beyond all orders of @xmath7 and hence the time evolution of the free parameters @xmath10 and @xmath7 occurs on a time scale longer than any power of @xmath116 .
physically , one such solution represents a travelling soliton slowing down and attenuating as the cherenkov radiation left in its wake carries momentum and energy away from its core . substituting into eq . , gives @xmath117 where @xmath118 .
we consider two solutions of this equation which both have the same asymptotic expansion , denoted @xmath119 and @xmath120 , such that @xmath121 as @xmath122 and @xmath123 as @xmath83 .
since the difference @xmath124 is small ( lies beyond all orders of @xmath7 ) , and since the solution @xmath125 can be regarded as a perturbation of @xmath126 , @xmath127 obeys the linearisation of eq . about @xmath126 to a good approximation .
that is , @xmath128 since @xmath129 , we can solve eq . to leading order in @xmath7 by ignoring the last two terms in it ; the resulting solutions are exponentials of the form @xmath130 .
we make a preemptive simplification by setting @xmath131 . [
that @xmath132 has to be set equal to zero follows from matching these exponentials to the far - field asymptotes of the stationary ` inner ' solution ; see section [ iiid ] below .
physically , @xmath131 implies that the travelling pulse will only excite the radiation with its own ( zero ) frequency in the comoving frame . ] the leading - order solution of eq .
is therefore @xmath133 where @xmath109 ( @xmath134 ) are the roots , numbered in order from smallest to largest , of the dispersion relation .
( recall that since we have taken @xmath10 in the interval @xmath135 $ ] , all the roots @xmath109 are positive . ) for @xmath136 , there is only one root , @xmath137 .
higher - order corrections to the solution can be found by substituting the ansatz @xmath138e^{i { \cal q}_n x/\epsilon } \\
+ \sum_n a_n^ * [ \epsilon^2 g_2^{(n)}(x ) + \epsilon^3 g_3^{(n)}(x)+\cdots]e^{-i{\cal q}_n x/\epsilon}\end{gathered}\ ] ] into eq . , expanding the advance / delay terms @xmath139 and @xmath140 in taylor series in @xmath7 , and making use of the asymptotic expansion , for @xmath126 .
for instance , the first few corrections are found to be @xmath141 since @xmath142 as @xmath83 , it follows that @xmath143 as @xmath83 , and hence , once we know the amplitudes @xmath144 , we know the asymptotic behaviour of @xmath125 .
we shall now employ the method of asymptotics beyond all orders to evaluate these amplitudes .
segur and kruskal s method allows one to measure the amplitude of the exponentially small radiation by continuing the solution analytically into the complex plane .
the leading - order term of @xmath17 , @xmath145 , has singularities at @xmath146 , @xmath147 . in the vicinity of these points ,
the radiation becomes significant ; the qualitative explanation for this is that the @xmath148 function forms a sharp spike with a short length scale near the singularity point , and hence there is a strong coupling to the radiation modes , unlike on the real axis .
the radiation , which is exponentially small on the real axis , becomes large enough to be measured near the singularities .
we define a new complex variable @xmath149 , such that @xmath150 is small in absolute value when @xmath49 is near the lowest singularity in the upper half - plane : @xmath151 the variables @xmath149 and @xmath49 are usually referred to as the ` inner ' and ` outer ' variables , respectively the transformation to @xmath149 effectively ` zooms in ' on the singularity at @xmath152 .
we also define @xmath153 and @xmath154 . continuing eq . to the complex plane i.e .
substituting @xmath155 for @xmath48 and @xmath156 for @xmath157we obtain , in the limit @xmath158 , [ inner - equations ] @xmath159 here we have used the fact that @xmath160 and @xmath161 as @xmath101 .
equations are our ` inner equations ' ; they are valid in the ` inner region ' @xmath162 , @xmath163 .
( the solution can not be continued up from the real @xmath49 axis past the singularity at @xmath164 . ) solving the system order by order , we can find solutions in the form of a series in powers of @xmath165 .
alternatively , we can make the change of variables in the asymptotic expansion and send @xmath101 .
this gives , for the first few terms , [ pertexpy ] @xmath166\frac{i}{y^3 } + \mathcal{o}(y^{-4})\right\ } , \\ \hat{w } & = \sqrt{\cos k}\left\ { -\frac{i}{y } - \tan k\frac{1}{2y^2 } + \left[\frac{1}{3}(1-\mu\cos k)+\frac{7}{12}\tan^2k\right]\frac{i}{y^3 } + \mathcal{o}(y^{-4})\right\}.\end{aligned}\ ] ] we are using hats over @xmath167 and @xmath168 to distinguish the series solution from other solutions of eq .
that will appear in the next section .
the asymptotic series may or may not converge .
we note a symmetry @xmath169 of the power - series solution . in order to obtain an expression for the terms which lie beyond all orders of @xmath165
, we substitute @xmath170 into eqs .
since @xmath171 and @xmath172 solve the equations to all orders in @xmath165 , then provided @xmath173 and @xmath174 are small , they will solve the linearisation of eqs about @xmath175 for large @xmath176 .
formal solutions to the linearised system can be constructed as series in powers of @xmath165 . because @xmath171 and @xmath172 are both @xmath177 , the leading - order expressions for @xmath173 and @xmath174 as @xmath178 are obtained by substituting zero for @xmath171 and @xmath172 in the linearised equations .
this gives @xmath179 where @xmath180 ( @xmath181 ) are the roots of @xmath182 note that the roots @xmath180 with @xmath183 are given by the @xmath184 limits of the roots of the dispersion equation : @xmath185 .
in addition , there is a root @xmath186 which does not have a @xmath187 counterpart . the full solutions ( i.e. solutions including corrections to all orders in @xmath165 ) will result if we use the full inverse - power series for @xmath188 and @xmath189 ; these solutions should have the form [ expexpansion ] @xmath190 \exp(-iq_ny).\end{aligned}\ ] ] note that we have excluded the terms proportional to @xmath191 from this ansatz ( i.e. set the amplitudes @xmath192 to zero ) as they would become exponentially large on the real @xmath49 axis .
[ one can readily verify this by making the change of variables in eq .. ] the coefficients @xmath193 , @xmath194 , and @xmath195 , @xmath196 , are found recursively when the ansatz is substituted into the linearised equations and like powers of @xmath165 collected . in particular , the first few coefficients are @xmath197\cos k}{[\sin(k+q_n ) -\sin k]^2}\end{aligned}\ ] ] and @xmath198 where @xmath199 .
having restricted ourselves to considering the linearised equations for @xmath173 and @xmath174 , we have only taken into account the simple harmonics in and .
writing @xmath200 and @xmath201 , where @xmath202 is an auxiliary small parameter ( not to be confused with our ` principal ' small parameter @xmath7 ) ; substituting @xmath203 and @xmath204 in eqs , and solving order - by - order the resulting hierarchy of nonhomogeneous linear equations , we can recover all nonlinear corrections to @xmath173 and @xmath174 .
the @xmath205-corrections will be proportional to @xmath206 ; higher - order corrections will introduce harmonics with higher combination wavenumbers . later in this section
it will become clear that @xmath202 is actually of the order @xmath207 [ see eq .
below ] ; hence the amplitudes of the combination harmonics will be exponentially smaller than that of @xmath208 .
now we return to the object that is of ultimate interest to us in this work that is , to the function @xmath127 of section [ radi_soli ] representing the radiation of the moving soliton .
we need to match @xmath127 to the corresponding object in the inner region . to this end
, we recall that @xmath209 , where @xmath125 and @xmath126 are two solutions of the outer equation which share the same asymptotic expansion to all orders . in the limit @xmath184
, the corresponding functions @xmath210 solve eqs and share the same inverse - power expansions .
we express this fact by writing @xmath211 therefore , the difference @xmath212 ( which results from the analytic continuation of the function @xmath127 ) can be identified with @xmath173 and @xmath213 with @xmath174 .
continuing eq . and its complex conjugate gives , as @xmath184 , [ cont12 ] @xmath214 \exp(iq_ny ) \nonumber\\ & - \sum_n \lim_{\epsilon \to 0 } a_n^*(\epsilon ) \exp \left(\frac{\pi \mathcal{q}_n}{2 \epsilon } \right ) \left[\frac{2 \cos k}{\omega + vq_n- 2 \cos(k - q_n ) } \frac{1}{y^2}+{{\cal o}\left({\frac{1}{y^3}}\right)}\right ] \exp(-iq_ny ) \label{cont1}\end{aligned}\ ] ] and @xmath215 \exp(-iq_ny ) \nonumber\\ & - \sum_n \lim_{\epsilon \to 0 } a_n(\epsilon ) \exp \left(-\frac{\pi \mathcal{q}_n}{2 \epsilon } \right ) \left[\frac{2 \cos k}{\omega + vq_n- 2 \cos(k - q_n ) } \frac{1}{y^2}+{{\cal o}\left({\frac{1}{y^3}}\right)}\right ] \exp(iq_ny ) .
\label{cont2}\end{aligned}\ ] ] [ here we have used eq .. ] matching to and to yields then @xmath216 and @xmath217 for @xmath218 .
we note that eq
. follows from eq . ,
while the latter equation can be written , symbolically , as @xmath219 our subsequent efforts will focus on the evaluation of the constants @xmath220 . for @xmath10 greater than @xmath221 ( approximately @xmath222 ) ,
there is only one radiation mode and therefore only one pre - exponential factor , @xmath223 . for smaller @xmath10
we note that the amplitude of the @xmath224th radiation mode , @xmath144 , is a factor of @xmath225 smaller than @xmath226 , the amplitude of the first mode .
referring to fig.[nicedisprelfig ] , it is clear that for @xmath227 , the difference @xmath228 will be no smaller than @xmath229 . as for the second mode , it becomes as significant as the first one only when @xmath230 in which case @xmath231 .
but in our asymptotic expansion of section [ section2 ] we assumed , implicitly , that @xmath10 is of order @xmath232 and so the case of @xmath233 is beyond the scope of our current analysis .
therefore , for our purposes all the radiation modes with @xmath56 ( when they exist ) will have negligible amplitudes compared to that of the first mode , provided @xmath223 is nonzero and @xmath7 is small .
for this reason , we shall only attempt to evaluate @xmath223 in this paper .
pomeau , ramani , and grammaticos @xcite have shown that the radiation can be measured using the technique of borel summation rather than by solving differential equations numerically , as in segur and kruskal s original approach .
the method has been refined by ( among others ) grimshaw and joshi @xcite and tovbis , tsuchiya , jaff and pelinovsky @xcite , who have applied it to difference equations .
most recently it has been applied to differential - difference equations in the context of moving kinks in @xmath100 models @xcite .
this is the approach that we will be pursuing here . expressing @xmath155 and @xmath156
as laplace transforms [ laplace ] @xmath234 and @xmath235 where @xmath236 is a contour extending from the origin to infinity in the complex @xmath237 plane , the inner equations are cast in the form of integral equations [ integral - equations ] @xmath238\ast u\ast w + u\ast u\ast w = 0 , \label{fpu } \\ f(-p)w + \mu [ f(-p)w]\ast w\ast u + w\ast w\ast u = 0 , \label{gpu}\end{gathered}\ ] ] where @xmath239 the asterisk @xmath240 denotes the convolution integral , @xmath241 where the integration is performed from the origin to the point @xmath237 on the complex plane , along the contour @xmath236 . in deriving eqs ,
we have made use of the convolution theorem for the laplace transform of the form , where the integration is over a contour in the complex plane rather than a positive real axis .
the theorem states that @xmath242 e^{-py } dp .
\label{conv_thm}\ ] ] the proof of this theorem is provided in appendix [ a ] .
we choose the contour so that @xmath243 as @xmath244 along @xmath236 . in this case
we have @xmath245 as @xmath246 for all @xmath149 along any line @xmath247 with @xmath163 .
therefore , the integrals in converge for all @xmath149 along this line and any bounded @xmath248 and @xmath249 .
the function @xmath248 will have singularities at the points where @xmath250 vanishes while the sum of the double - convolution terms in does not .
similarly , @xmath249 will have a singularity wherever @xmath251 vanishes [ while the sum of the double - convolution terms in does not ] .
therefore , @xmath252 and @xmath253 may have singularities at the points where @xmath254 with the top and bottom signs referring to @xmath248 and @xmath249 , respectively .
the imaginary roots of eq . with the top sign are at @xmath255 and those of eq . with the bottom sign at @xmath256 , where @xmath180 are the real roots of .
( we remind the reader that all roots @xmath180 are positive . )
the point @xmath257 is not a singularity as both double - convolution terms in each line of vanish here .
there is always at least one pure imaginary root of eq .
( and _ only _ one if @xmath258 , where @xmath107 ) .
in addition , there are infinitely many complex roots .
the complex singularities of @xmath252 [ complex roots of the top - sign equation in ] are at the intersections of the curve given by @xmath259 with the family of curves described by @xmath260 and at the intersections of the curve @xmath261 with the family of curves described by @xmath262 here @xmath263 and @xmath264 are the real and imaginary part of @xmath237 : @xmath265 .
the curve looks like a parabola opened upwards , with the vertex at @xmath266 , and the curve like a parabola opened downwards , with the vertex at @xmath267 , @xmath268 . the curves and , on the other hand ,
look like parabolas for small @xmath263 but then flatten out and approach horizontal straight lines as @xmath269 .
[ in compiling the list of these ` flat ' curves in and , we have taken into account that the curve with @xmath270 does not have any intersections with the parabola , and the curve with @xmath271 does not have any intersections with the parabola . ]
as @xmath10 is reduced , the vertex of the parabola moves down along the @xmath264 axis ; the intersections of this parabola with the ` flat ' curves approach , pairwise , the @xmath264 axis . after colliding on the @xmath264 axis ,
pairs of complex roots move away from each other along it .
[ the parabola does not move as @xmath10 is reduced , but only steepens , which results in the singularities in the upper half - plane approaching the imaginary axis but not reaching it until @xmath272 . ] in a similar way , the complex singularities of @xmath249 move onto the imaginary axis as @xmath10 is decreased . in section [ blt ] below we will use the fact that the distance from any complex singularity to the origin is larger than @xmath33 ; this follows from the observation that the closest points of the curves and to the origin are their intersections with the @xmath264 axis .
these are further away than @xmath33 from the origin .
in addition to singularities at @xmath273 , the function @xmath248 will have singularities at points @xmath274 , @xmath134 .
these are induced by the cubic terms in eq . ; for instance , the singularity at @xmath275 arises from the convolution of the term proportional to @xmath237 in @xmath276 with the function @xmath253 which has a singularity at @xmath275 . [ that @xmath248 has singularities at @xmath274 can also be seen directly from eq .. ] similarly , the function @xmath249 will have singularities at points @xmath277 , @xmath134 . by virtue of the nonlinear terms
there will also be singularities at the ` combination points ' @xmath278 , @xmath279 , etc . the formal inverse - power series , which we can write as @xmath280 result from the laplace transformation of power series for @xmath248 and @xmath249 : @xmath281 the series converge in the disk of radius @xmath282 , centred at the origin , and hence can be integrated term by term only over the portion of the contour @xmath236 which lies within that disk
. however , by watson s lemma , the remaining part of the contour makes an exponentially small contribution to the integral and the resulting series are asymptotic as @xmath283 . the functions @xmath155 and @xmath156 defined by eqs give the borel sums of the series @xmath284 and @xmath285 .
consider now some horizontal line in the inner region ; that is , let @xmath286 be fixed and @xmath287 vary from @xmath288 to @xmath289 . if the integration contour @xmath236 is chosen to lie in the first quadrant of the complex @xmath237 plane , the functions @xmath155 and @xmath156 generated by eqs will tend to zero as @xmath290 along this line . similarly , if it is chosen to lie in the second quadrant , they will tend to zero as @xmath291 .
suppose there were no singularities between two such contours : then the one could be continously deformed to the other without any singularity crossings ; i.e. they would generate the same solution which , therefore , would decay to zero at both infinities .
[ that is , the oscillatory tails in would have zero amplitudes , @xmath292 . ] in general , however , @xmath248 and @xmath249 have singularities both on and away from the imaginary axis . in order to minimise the number of singularities to be crossed in the deformation of one contour to the other
, we choose the contours to lie above all singularities with nonzero real part .
( that this is possible , is shown in appendix [ b ] . )
note also that the imaginary part of the singularity grows faster than its real part and hence @xmath293 should tend to @xmath294 as @xmath244 along @xmath236 ; this was precisely our choice for the direction of the contours @xmath236 in the beginning of section [ blt ] .
the integration contours @xmath295 and @xmath296 used to generate the solutions @xmath297 and @xmath298 respectively via eqs .
the dots are singularities of @xmath249 .
shown is the situation where the linear dispersion relation has only one real root , @xmath282 . ]
let @xmath295 and @xmath296 be two contours chosen in this way , with @xmath295 lying in the first quadrant and @xmath296 in the second quadrant .
the solutions @xmath299 generated by eqs with the contour @xmath295 will tend to zero as @xmath300 ( with @xmath301 fixed ) .
hence they can be identified with solutions @xmath302 obtained by the continuation of the outer solution @xmath125 which has the same asymptotic behaviour .
similarly , the solutions generated by eqs with the contour @xmath296 coincide with solutions @xmath303 , @xmath304 like @xmath303 , @xmath304 , the solutions generated by eqs tend to zero as @xmath305 ( with fixed @xmath306 ) .
consider , first , solutions @xmath307 and @xmath304 . since the contours @xmath295 and @xmath296 are separated by singularities of @xmath249 on the positive imaginary axis , they can not be continuously deformed to each other without singularity crossings and so the solution @xmath307 does not coincide with @xmath304 , unless the residue at the singularity happens to be zero .
if we deform @xmath295 to @xmath308 and @xmath296 to @xmath309 as shown in fig.[integration - contours ] , without crossing any singularities , then the only difference between the two contours is that @xmath308 encircles the singularities , whereas @xmath309 does not .
therefore , the difference @xmath213 can be deduced exclusively from the leading - order behaviour of @xmath249 near its singularities
. there can be two contributions to this difference : the first arises from integrating around the poles and is a sum of residues , while the second arises if the singularity is a branch point . to find the singularity structure of the function @xmath249
, we equate @xmath310 to the expansion .
the first term in , @xmath311 , arises from the integration of a term @xmath312 in @xmath249 .
for such a term , the difference of the two integrals in reduces to an integral around a circle centered on the point @xmath274 : @xmath313 the term @xmath314 in , with @xmath315 , arises from the integration of a term @xmath316 in @xmath249 .
this time , @xmath274 is a branch point . after going around this point along the circular part of @xmath308 , the logarithm increases by @xmath317 and the difference between the integrals in is given by @xmath318 where @xmath319 is the part of @xmath308 extending from @xmath274 to infinity .
this equals exactly @xmath320 .
thus , in order to generate the full series we must have @xmath321 + w_{\rm reg}(p),\end{gathered}\ ] ] where @xmath322 denotes the part of @xmath253 which is regular at @xmath274 , @xmath134 . by the same process ,
matching @xmath212 to @xmath173 in yields @xmath323 the solution to eqs is nonunique ; for instance , if @xmath324 is a solution , then so is @xmath325 with any complex @xmath326 and @xmath327 .
also , if @xmath324 is a solution , @xmath328 is another one .
we will impose the constraint @xmath329 this constraint is obviously compatible with eqs .
it is not difficult to see that the reduction singles out a unique solution of eqs .
the motivation for imposing the constraint comes from the symmetry @xmath169 of the power - series solution of eq .. using this symmetry in eq . , we get @xmath330 and then eq . implies . in view of ,
the singularities of @xmath248 in the upper half - plane are singularities of @xmath249 in the lower half - plane , which fall within @xmath331 , and vice versa .
thus we have , finally , @xmath332 + \tilde{w}_{\text{reg}}(p),\end{gathered}\ ] ] where @xmath333 is regular at @xmath334 .
we also mention an equivalent representation for which turns out to be computationally advantageous : @xmath335 \\ + \check{w}_{\text{reg}}(p).\end{gathered}\ ] ] here @xmath336 is an integral map : @xmath337 the notation @xmath338 should be understood as @xmath339 we have also introduced @xmath340 and @xmath341 for economy of notation .
the only difference between eqs and is that the double - sum term on the right - hand side of includes some terms which are regular at @xmath342 , whereas in eq .
, all regular terms are contained in @xmath333 .
the residues @xmath220 at the poles of @xmath249 are known as the stokes constants .
the leading - order stokes constant @xmath223 can be related to the behaviour of the coefficients in the power - series expansion of @xmath249 . indeed
, the coefficients in the power series satisfy @xmath343 this is obtained by expanding the singular part of the expression in powers of @xmath237 .
( coefficients of the regular part become negligible in the limit @xmath344 compared to those of the singular part . ) note that we have ignored singularities with nonzero real part and singularities on the imaginary axis other than at @xmath345 .
the reason is that all these singularities are further away from the origin than the points @xmath346 ( in particular all complex singularities are separated from the origin by a distance greater than @xmath33 ) , and their contribution to @xmath347 becomes vanishingly small as @xmath348 .
we have also neglected singularities at the ` combination points ' because of their exponentially small residues .
the coefficients @xmath347 can be calculated numerically ; once they are known , it follows from that @xmath349 ( where we have recalled that @xmath350 and @xmath351 ) .
we now turn to the numerical calculation of the coefficients @xmath347 . to make our forthcoming numerical procedure more robust , we normalise the coefficients in the power series by writing @xmath352 substituting the expansions with eq . as well as the constraint into either of equations and equating coefficients of like powers of @xmath237 , yields the following recurrence relation for the numbers @xmath353 ( @xmath86 ) : @xmath354 } \\
\times\left(\sum_{j=0}^m ( -1)^{m - j}\delta_{m - j } \delta_j\frac{j!(m - j)!}{m!}\right)\frac{m!(n - m)!}{n!}. \label{recu}\end{gathered}\ ] ] here @xmath355 solving eq . with @xmath356
gives @xmath357 .
thereafter it can be solved for each member of the sequence @xmath358 in terms of the preceding ones , and thus each @xmath353 can be calculated in turn .
[ since all the coefficients in the recurrence relation are real , the sequence @xmath359 turns out to be a sequence of real numbers . ]
once the sequence @xmath358 has been generated , expression can be used to calculate the stokes constant @xmath223 : @xmath360 for sufficiently large @xmath361 .
unfortunately , the convergence of the sequence @xmath359 is slow and thus the above procedure is computationally expensive .
the convergence can be accelerated by expanding eq . in powers of small @xmath362 : @xmath363,\ ] ] whence @xmath364 . \label{k1}\end{gathered}\ ] ] according to eq . , eq .
gives @xmath223 with a relatively large error of order @xmath362 . on the other hand ,
a two - term approximation @xmath365 is correct to @xmath366 .
more precisely , the relative error associated with the answer is given by @xmath367 in our calculations , we set @xmath368 .
since @xmath369 , @xmath370 and @xmath371 are known constants [ given by and ] , eq .
tells us what @xmath361 we should take
i.e. how many members of the sequence @xmath359 we should calculate in order to achieve the set accuracy . figure [ series ] illustrates the convergence of the approximate values of @xmath223 calculated using and as @xmath361 is increased .
note the drastic acceleration of convergence in the latter case .
convergence of the sequence on the right - hand side of ( dashed line ) and the ` accelerated ' sequence defined by the right - hand side of ( solid line ) .
shown are the @xmath361th approximations to the stokes constant @xmath223 [ the @xmath361th members of the sequences and ] divided by @xmath372 to get a real value . in this plot ,
@xmath5 and @xmath373 . ]
the stokes constant @xmath223 for various values of @xmath6 . note the logarithmic scale on the vertical axis .
the downward spikes extend all the way to @xmath288 ; hence each spike corresponds to a zero crossing .
each panel shows only a portion of the full range @xmath39 ; there are no additional zero crossings in the part which is not shown . ]
the stokes constant @xmath223 for various values of @xmath6 . note the logarithmic scale on the vertical axis .
the downward spikes extend all the way to @xmath288 ; hence each spike corresponds to a zero crossing .
each panel shows only a portion of the full range @xmath39 ; there are no additional zero crossings in the part which is not shown . ]
figure [ stokes](a ) shows the calculated stokes constant as a function of @xmath10 for various values of the saturation parameter @xmath6 .
first of all , @xmath374 does not have any zeros in the case of the cubic nonlinearity ( @xmath375 ) .
this means that solitons of the cubic one - site discrete nls equation [ eq . with @xmath375 ] can not propagate without losing energy to radiation . for @xmath377
the stokes constant does have a zero , but at a value of @xmath10 smaller than @xmath378 ( where @xmath379 ) .
since higher radiation modes do exist in this range of @xmath10 , there will still be radiation from the soliton unless the ` higher ' stokes constants @xmath380 , @xmath381 , , happen to be zero at the same value of @xmath10 .
finally , for @xmath382 the zero is seen to have moved just above @xmath378 and for @xmath383 it has an even higher value .
there are no @xmath384 radiations for these @xmath10 ; hence the zeros of @xmath374 define the carrier wavenumbers at which the soliton ` slides ' i.e. travels without emitting any radiation .
equation then gives the corresponding sliding velocities , for each @xmath7 .
figure [ stokes](b ) shows the stokes constant @xmath374 for higher values of the parameter @xmath6 . for @xmath385
a second zero of the stokes constant has appeared while for @xmath386 , the function @xmath374 has three zeros .
as @xmath6 is increased , the existing zeros move to larger values of @xmath10 while new ones emerge at the origin of the @xmath10 axis . for not very large @xmath64 ,
the solution @xmath125 is close to the localised pulse found by means of the perturbation expansion in section [ section2 ] .
as @xmath122 , it tends to zero , by definition , while the @xmath83 asymptotic behaviour is found from @xmath387 . here
the solution @xmath126 decays to zero as @xmath388 and hence @xmath125 approaches the oscillatory waveform @xmath127 given by eqs , , and : @xmath389e^{i\mathcal{q}_n x/\epsilon}\end{gathered}\ ] ] as @xmath390 .
equation describes a radiation background over which the soliton is superimposed .
as we have explained , we can ignore all but the first term in the sum . to determine whether the radiation is emitted by the soliton or being fed into it from outside sources , we consider a harmonic solution @xmath391 of the linearised eq . ; the corresponding dispersion relation is @xmath392 the radiation background @xmath127 consists of harmonics with @xmath131 and @xmath393 where @xmath394 are roots of eq .. the group velocities of these harmonic waves are given by @xmath395 the @xmath394 s are zeros of the function @xmath396 and the group velocities are the slopes of this function at its zeros ; thus the group velocities @xmath397 , @xmath398 , , have alternating signs
. the first one , which is the only one that concerns us in this work , must be negative .
indeed , the value @xmath399 is positive , and hence the slope of the function @xmath396 as it crosses the @xmath400 axis at @xmath401 is negative .
therefore , the first radiation mode , extending to @xmath288 , carries energy _
away _ from the soliton .
the even - numbered radiation modes ( where present ) in our asymptotic solution have positive group velocities and hence describe the flux of energy fed into the system at the left infinity .
a more interesting situation is obviously the one with no incoming radiation ; the corresponding solution is obtained by subtracting off the required multiple of the solution of the linearised equation , e.g. @xmath402 .
one would then have a pulse leaving the odd modes in its wake and sending even modes ahead of it . if the first stokes constant @xmath374 has a zero at some @xmath403 while @xmath404 is the only radiation mode available ( as happens in our saturable model with @xmath6 greater than approximately @xmath405 ) , then according to eq .
, the radiation from the soliton with the carrier - wave wavenumber @xmath406 is suppressed completely .
to find the radiation - induced evolution of the travelling soliton , we use conserved quantities of the advance - delay equation associated with eq .. in the reference frame moving at the soliton velocity @xmath12 this equation reads @xmath407 the discrete variable @xmath112 in eq .
is related to the value of the continuous variable @xmath408 at the point @xmath409 : @xmath410 . for future use
, we also mention the relation between @xmath408 and the corresponding solution of eq .
: @xmath411 we first consider the number of particles integral : @xmath412 multiplying eq . by @xmath413 , subtracting the complex conjugate and integrating yields the rate of change of the integral @xmath414 : @xmath415 in eq . ,
@xmath416 and @xmath417 stands for the complex conjugate of the immediately preceding term .
the integration limits @xmath418 and @xmath419 are assumed to be large ( @xmath420 ) but finite ; for example one can take @xmath421 .
consider the soliton moving with a positive velocity and leaving radiation in its wake .
this configuration is described by the solution @xmath125 of eq . ; the corresponding solution of eq . has the asymptotic behaviour @xmath422 as @xmath423 . substituting the leading - order expression for the soliton s radiation tail into eq .
yields the asymptotic behaviour at the other infinity : @xmath424e^{i(k+{\cal q}_n)x+i(\omega+kv)t } \label{asympie}\ ] ] as @xmath425 .
since , as we have explained above , the @xmath404 radiation is dominant , it is sufficient to keep only the @xmath58 term in . substituting into and evaluating the integral over the region @xmath426 in the soliton s wake , we obtain @xmath427 $ ] is the group velocity of the first radiation mode , @xmath397 , which , as we have established , is negative ; hence @xmath428 .
we now turn to the momentum integral , @xmath429 multiplying by @xmath430 , adding its complex conjugate and integrating , gives the rate equation @xmath431_{a}^b.\end{gathered}\ ] ] evaluating the right - hand side of similarly to the way we obtained and substituting eq . for @xmath11 , produces latexmath:[\[\label{pdot } \frac{dp}{dt } = using the leading - order term of the perturbative solution and eq .
, we can express @xmath414 and @xmath433 via @xmath7 and @xmath10 : @xmath434 in calculating @xmath414 and @xmath433 we had to integrate from @xmath435 to @xmath436 . since the integrands decay exponentially , and since @xmath418 and @xmath419 were assumed to be much larger than @xmath116 in absolute value , it was legitimate to replace these limits with @xmath288 and @xmath289 , respectively .
the error introduced in this way is exponentially small in @xmath7 .
taking time derivatives of @xmath414 and @xmath433 above and discarding @xmath21 corrections to @xmath437 terms , we deduce that @xmath438 finally , substituting for @xmath439 and @xmath440 from eqs and , we arrive at the dynamical system [ paramev ] @xmath441 where the group velocity @xmath442 with @xmath443 , and @xmath444 is the smallest root of eq .. the phase portrait of the system in the case ( a ) where the stokes constant @xmath374 does not have zeros , and ( b ) where @xmath374 has one zero . in ( b ) , the dashed line is the line of nonisolated fixed points @xmath403 .
note that in ( b ) , the phase portrait has been replotted on the @xmath445 plane ; hence the dashed line gives the value of the sliding velocity for each value of @xmath7 .
in ( a ) , @xmath375 ; in ( b ) , @xmath446 . ] the phase portrait of the system in the case ( a ) where the stokes constant @xmath374 does not have zeros , and ( b ) where @xmath374 has one zero . in ( b ) , the dashed line is the line of nonisolated fixed points @xmath403 .
note that in ( b ) , the phase portrait has been replotted on the @xmath445 plane ; hence the dashed line gives the value of the sliding velocity for each value of @xmath7 . in ( a ) , @xmath375 ; in ( b ) , @xmath446 . ] the vector field is defined for @xmath447 , where @xmath448 is the value of @xmath10 for which the roots @xmath404 and @xmath449 merge in fig.[nicedisprelfig ] ( i.e. the smallest value of @xmath10 for which the roots @xmath404 and @xmath449 still exist ) . when @xmath450 , the group velocity @xmath397 becomes equal to zero ; therefore , the equation @xmath451 defines a line of ( nonisolated ) fixed points of the system .
assume , first , that the saturation parameter @xmath6 is such that @xmath374 does not vanish for any @xmath10 .
for @xmath452 , the factor @xmath397 in is negative ; since @xmath404 and @xmath38 are positive for all @xmath453 , the derivative @xmath454 satisfies @xmath455 for all times .
hence , all trajectories are flowing towards the line @xmath451 from above .
it is worth noting that the derivative @xmath456 is also negative , and that the @xmath10 axis is also a line of nonisolated fixed points .
however , no trajectories will end there as follows from the equation @xmath457,\ ] ] which is a consequence of and .
representative trajectories are plotted in fig.[phaseplane](a ) .
since @xmath458 is very small for small @xmath7 , the line @xmath451 is practically indistinguishable from the horizontal axis .
therefore , we can assert that , to a good accuracy , @xmath459 as @xmath460 .
this means that the soliton stops moving and stops decaying at the same time . for @xmath10
smaller than @xmath448 , the vector field is undefined and we can not use it to find out what happens to the soliton after @xmath10 has reached @xmath448 .
the reason for this is that the soliton stops radiating at the wavenumber @xmath404 as @xmath10 drops below @xmath448 .
in fact our analysis becomes invalid as soon as @xmath10 becomes @xmath461i.e .
even before @xmath10 reaches @xmath448because we can no longer disregard the @xmath462 radiation here . at the qualitative level it is obvious , however , that the parameter @xmath10 should continue to decay all the way to zero , in a cascade way .
first , the @xmath462 radiation will become as intense as the @xmath58 mode when @xmath10 approaches @xmath463 .
subsequently i.e . for @xmath10 smaller than @xmath463the @xmath464 harmonic will replace the radiation with the wavenumbers @xmath404 and @xmath449 as a dominant mode .
the @xmath465 mode will become equally intense near @xmath466 ; as @xmath10 drops below @xmath467 , both @xmath468 and @xmath469 will cede to @xmath470 , and so on . if @xmath6 is such that the stokes constant @xmath374 vanishes at one or more values of @xmath10 , the system has one or more lines of nonisolated fixed points , @xmath471 .
the corresponding values of @xmath12 , @xmath472 , define the _ sliding _ velocities of the soliton
i.e. velocities at which the soliton moves without radiative friction .
one such velocity is shown by the dashed line in fig.[phaseplane](b ) .
the fixed points @xmath473 are semistable ; for @xmath10 above @xmath474 , the flow is towards the line @xmath471 but when @xmath10 is below @xmath474 , the flow is directed away from this straight line [ see fig.[phaseplane](b ) ] .
since both @xmath456 and @xmath454 are negative , the soliton s velocity @xmath475 will generally be decreasing until it hits the nearest underlying sliding velocity @xmath476 and locks on to it .
the ensuing sliding motion will be unstable ; a small perturbation will be sufficient to take the soliton out of the sliding regime after which it will resume its radiative deceleration .
however , since @xmath454 is proportional to the _ square _ of the stokes constant and not to @xmath374 itself , small perturbations @xmath477 will be growing linearly , not exponentially , in @xmath478 . as a result
, the soliton may spend a fairly long time sliding at the velocity @xmath476 .
it is therefore not unreasonable to classify this sliding motion as _
we conclude that the soliton becomes pinned ( i.e. @xmath479 and so @xmath480 ) before it has decayed fully ( i.e. before the amplitude @xmath7 has decreased to zero ) .
the exponential dependence of @xmath481 and @xmath482 on @xmath483 implies that tall , narrow pulses will be pinned very quickly while short , broad ones will travel for a very long time before they have slowed appreciably .
next , if @xmath6 is such that there is one or more sliding velocity available in the system , and if the soliton is initially moving faster than some of these , its deceleration will be interrupted by long periods of undamped motion at the corresponding sliding velocities . the soliton s decay rate as a function of its velocity , for fixed @xmath484 .
( here @xmath375 . ) in the inset , the same curve is replotted using a logarithmic scale on the vertical axis . ]
it is worth reemphasising here that if the amplitude @xmath7 is small , then even if the soliton is _ not _ sliding , its deceleration will be so slow that it will spend an exponentially long time travelling with virtually unchanged amplitude and speed . the deceleration rate @xmath485 is shown in fig.[decayrate ] , as a function of the soliton s velocity @xmath12 , for fixed amplitude @xmath7 .
note that the decay rate drops , exponentially , as the velocity is decreased ; this drop is due to the exponential factor @xmath486 in eq .. as the velocity @xmath12 ( and hence the wavenumber @xmath10 ) decreases , the root @xmath487 grows towards the limit value of approximately @xmath33 ( see fig.[nicedisprelfig ] ) .
this variation in @xmath404 is amplified by the division by small @xmath7 and exponentiation in @xmath486 .
in this paper , we have constructed the moving discrete soliton of the saturable nls equation [ and hence ] as an asymptotic expansion in powers of its amplitude .
the saturable nonlinearity includes the cubic nls as a particular case ( for which @xmath375 ) .
our perturbation procedure is a variant of the lindstedt
poincar technique where corrections to the parameters of the solution are calculated along with the calculation of the solution itself .
although the resulting asymptotic series for the soliton is generally not convergent , the associated expansions for its frequency and velocity sum up to exact explicit expressions . from the divergence of the asymptotic series it follows , in particular , that the travelling discrete soliton does not decay to zero at least at one of the two infinities . instead , the soliton approaches , as @xmath488 or @xmath83 , an oscillatory resonant background where the amplitudes @xmath144 of its constituent harmonic waves lie beyond all orders of @xmath7
for the soliton moving with a positive velocity and approaching zero as @xmath488 , the oscillatory background at the left infinity represents the cherenkov radiation left in the soliton s wake . to evaluate the amplitudes of the harmonic waves arising as @xmath83
, we have continued the radiating soliton into the complex plane , where it exhibits singularities .
we then matched the asymptotic expansion of the background near the lowest singularity on the imaginary @xmath49 axis to the far - field asymptotic expansion of the background solution of the ` inner ' equation
i.e. of the advance - delay equation ` zoomed in ' on this singularity .
the asymptotic expansions here are in inverse powers of the zoomed variable , @xmath149 .
the amplitudes of the radiation waves were found to be exponentially small in @xmath7 , with the pre - exponential factors ( the so - called stokes constants ) being dependent only on the soliton s carrier - wave wavenumber . representing solutions to the inner equation as borel sums of their asymptotic expansions ,
the stokes constants can be related to the expansion coefficients ; we have calculated these coefficients numerically , using algebraic recurrence relations .
the upshot of the calculation of the leading stokes constant @xmath223 is that in the case of the cubic nonlinearity ( i.e. for @xmath375 ) , @xmath374 does not vanish for any @xmath10 .
this means that the cubic discrete soliton can not ` slide '
i.e. can not move without radiative friction .
the saturable solitons , on the other hand , _ can _ slide provided the saturation parameter @xmath6 is large enough .
this is because , for a sufficiently large @xmath6 , the stokes constant @xmath374 is found to have one or more zeros @xmath489 . since the soliton with wavenumber @xmath490 can have no higher - order resonances , those zeros which satisfy @xmath491
do define the wavenumbers at which sliding motion occurs .
for each value of the soliton s amplitude @xmath7 , the formula then gives the sliding velocities @xmath492 .
the calculation of asymptotics beyond all orders is useful not only for determining the sliding velocities . knowing the radiation amplitudes has allowed us to derive a two - dimensional dynamical system for the soliton s parameters .
trajectories of this dynamical system describe the evolution of the soliton travelling at a generic speed .
the evolution turns out to be simple : if @xmath6 is such that the stokes constant @xmath374 does not vanish anywhere in the region @xmath493 , the soliton decelerates , although very slowly .
eventually it becomes pinned to the lattice with a decreased but finite amplitude .
however , if @xmath6 is such that there are sliding velocities in the system , and if the soliton starts its motion with the velocity higher than some of these then , although it will finally become pinned to the lattice , its deceleration will be interrupted by long periods of metastable sliding motion at these isolated velocities .
it is interesting to tie up the above results with our previous work on exceptional discretisations of the @xmath100 model @xcite .
in that case we discovered that for some exceptional models ( in which the stationary soliton possesses an effective translational symmetry ) , sliding velocities , at which the radiation disappears , do exist .
the fact that all these models involve a complicated nonlocal discretisation of the nonlinearity leads one to wonder whether the nonlinearity _ has _ to be discretised nonlocally in order for sliding solitons to exist .
this question is answered negatively by our present work which gives a counterexample of a simple , local and physically motivated nonlinearity supporting sliding solitons .
a remaining open issue is whether exceptionality of the system is a prerequisite for the existence of sliding solitary waves .
indeed , we have yet to encounter a nonexceptional discrete system permitting sliding motion .
we conclude this section by placing our results in the context of earlier studies of discrete solitons . moving solitons in the cubic dnls equation
have previously been studied by ablowitz , musslimani , and biondini @xcite ( among others ) , using a numerical technique based on discrete fourier transforms as well as perturbation expansions for small velocities .
as we have done , these authors suggest that sliding ( radiationless ) solitons may not exist .
they also observe , with the aid of numerical simulations , that strongly localised pulses are pinned quickly to the lattice , while broader ones are more mobile results which are in agreement with ours .
duncan , eilbeck , feddersen and wattis @xcite have studied the bifurcation of periodic travelling waves from constant solutions in the cubic dnls equation , using numerical path - following techniques .
they find that the paths terminate when the soliton s amplitude reaches a certain limit value , and in the light of our results it seems likely that this is the point at which the radiation becomes large enough to have an effect on the numerics .
solitary waves which decay to constant values at the spatial infinities , despite the fact that the generic asymptotic behaviour in the underlying model is oscillatory , are commonly referred to as embedded solitons .
an example is given by the sliding solitons reported in this paper as well as the sliding kinks of @xcite ; both types of solitary waves propagate without exciting the resonant background oscillations .
embedded solitons have been studied for some time in continuous systems @xcite , but their history in lattice equations is younger .
recently , malomed , gonzlez - prez - sandi , fujioka , espinosa - cern and rodrguez have considered certain lattice equations with next - to - nearest - neighbour couplings and shown ( by means of explicit solutions ) that both stationary @xcite and moving @xcite embedded solitons exist .
stationary embedded solitons in discrete waveguide arrays have also been analysed by yagasaki , champneys and malomed @xcite . finally , while preparing the revised version of this paper , we learnt that melvin , champneys , kevrekidis , and cuevas have obtained results very similar to ours . using a combination of intuitive arguments and numerical computations , they have found sliding solitons for certain values of the parameters of a saturable dnls model @xcite .
the authors would like to thank dmitry pelinovsky and alexander tovbis for helpful discussions relating to this work .
o.o . was supported by the national research foundation of south africa and the university of cape town .
i.b . was supported by the nrf under grant 2053723 and by the urc of uct .
this appendix deals with the convolution property of the modified laplace transform of the form , where the integration is along an infinite contour in the complex plane rather than the positive real axis . in the context of asymptotics beyond all orders ,
this transform was pioneered by grimshaw and joshi @xcite . since no proof of the convolution result is available in literature , and
since it requires a nontrivial property ( concavity ) of the integration contour , we produce such a proof here .
we wish to show that @xmath494 dp,\end{aligned}\ ] ] where @xmath495 stands for an integral along the curve @xmath236 from the origin to the point @xmath237 on that curve .
we assume that the curve @xmath236 extends from the origin to infinity on the complex plane ; lies in its first quadrant ( @xmath496 , @xmath497 ) ; is described as a graph of a single - valued function @xmath498 ( i.e. never turns back on itself ) , and is concave - up everywhere : @xmath499 we also assume that the function @xmath500 is analytic in the region between the contour @xmath236 and the imaginary axis .
deformation of the integration contour @xmath501 to @xmath502 . ]
we begin by writing the left - hand side of as @xmath503 dp,\end{aligned}\ ] ] where @xmath504 .
the curve @xmath501 is the path traced out by @xmath505 as @xmath506 traces out the curve @xmath236 , for a given @xmath237 ( which also lies on @xmath236 ) this is depicted in fig .
[ conv - paths ] .
the path @xmath501 is the same as the path @xmath236 but translated from the origin to the point @xmath237 . for each given @xmath237
we deform the path @xmath501 so that it now lies on @xmath236 , still starting at the point @xmath237 .
we call this deformed path @xmath502 . the point @xmath507 , which lay on the path @xmath236 before the deformation , will now move inside the region bounded by @xmath236 ( i.e. the region to the left of @xmath236 ) , since all points on @xmath502 lie inside the region bounded by @xmath501 .
this follows from the fact that the path @xmath501 is concave - up .
the point @xmath508 will , however , stay to the right of the imaginary axis , since after the deformation all points @xmath505 on @xmath502 lie to the right of the point @xmath237 .
this follows from the fact that the curve @xmath236 never turns back on itself .
since @xmath509 is analytic for all @xmath506 between the imaginary axis and @xmath236 , the value of the integral will not be affected by the deformation .
[ in the particular case of the functions @xmath248 and @xmath249 considered in section [ blt ] , we have deliberately chosen the contours of integration so that the analyticity condition is satisfied . ] after the deformation , both integrals in follow the path @xmath236 , with the inner one starting at the point @xmath237 . by parameterising the path
@xmath236 it is now straightforward to change the order of integration , and end up with the desired result .
the argument holds , with the obvious modifications , for paths in the second quadrant .
the roots of eq . with the top and bottom signs give singularities of @xmath248 and @xmath249 , respectively .
our aim here is to show that the integration contours @xmath296 and @xmath295 in section [ blt ] can be chosen to lie above all the complex singularities
i.e. , that singularities of @xmath248 and @xmath249 do not accumulate to the imaginary axis . the roots of eq .
are zeros of the functions @xmath510 , where @xmath511 the imaginary zeros of @xmath512 are at @xmath513 , where @xmath180 are the real roots of eq .. we let @xmath414 denote the number of these imaginary zeros ; for @xmath514 , @xmath414 is finite .
we recall that all @xmath180 are positive ; hence all @xmath414 imaginary zeros of @xmath512 are on the negative imaginary axis , and all @xmath414 zeros of @xmath515 on the positive imaginary axis .
let @xmath263 and @xmath264 be the real and imaginary part of @xmath237 : @xmath516 .
we let @xmath517 denote the rectangular region in the complex-@xmath237 plane bounded by the vertical lines @xmath518 and horizontal lines @xmath519 and @xmath520 , where @xmath202 is small and @xmath521 is large enough for the region to contain all @xmath414 zeros of @xmath515 on the positive imaginary axis . by the argument principle , the total number of ( complex ) zeros of the function @xmath515 in the region @xmath517 is given by @xmath522 times the variation of its argument ( i.e. , its phase ) along the boundary of @xmath517 . in a similar way , one may count the number of zeros of the function @xmath512 in @xmath517 .
we have @xmath523 points on the vertical line @xmath524 satisfy @xmath525 consider , first , the case of the bottom sign in eq .. the expression between the bars in crosses through zero @xmath414 times as the line @xmath524 is traced out .
each time zero is crossed , the logarithmic derivative in jumps from @xmath288 to @xmath526 , and the argument of @xmath515 increases by @xmath229 as we move from one crossing to the next one .
the net increase of the argument as the line @xmath524 is traversed from its bottom to the top , is @xmath527 . in the case of the top sign in
, on the other hand , the expression in @xmath528 never crosses through zero and hence the total increment of the argument of @xmath512 is zero .
as we move along the vertical line @xmath529 from top to bottom , the argument of @xmath515 increases by another @xmath527 while the argument of @xmath512 does not acquire any increment .
since there are no zero crossings on the horizontal segments , the net change of the argument of @xmath515 along the boundary of @xmath517 is @xmath530 while the total increment of @xmath531 is zero .
therefore @xmath515 has only @xmath414 zeros in the region @xmath517 ( and they all lie on the imaginary axis ) while the function @xmath512 has no zeros in @xmath517 .
this implies that complex singularities of @xmath248 and @xmath249 can not accumulate to the positive imaginary axis .
h. feddersen , in _ nonlinear coherent structures in physics and biology .
proceedings of the 7th interdisciplinary workshop held at dijon , france , 4 - 6 june 1991 . _
m. remoissenet and m. peyrard , lecture notes in physics vol .
393 ( springer , berlin , 1991 ) , pp .
159167 .
j. c. eilbeck and m. johansson , in _ proceedings of the third conference on localization and energy transfer in nonlinear systems , san lorenzo de el escorial , madrid , spain , june 1721 , 2002 _ ( world scientific , singapore , 2003 ) , pp .
|
using the method of asymptotics beyond all orders , we evaluate the amplitude of radiation from a moving small - amplitude soliton in the discrete nonlinear schrdinger equation .
when the nonlinearity is of the cubic type , this amplitude is shown to be nonzero for all velocities and therefore small - amplitude solitons moving without emitting radiation do not exist . in the case of a _
saturable _ nonlinearity , on the other hand , the radiation is found to be completely suppressed when the soliton moves at one of certain isolated ` sliding velocities ' .
we show that a discrete soliton moving at a general speed will experience radiative deceleration until it either stops and remains pinned to the lattice , or in the saturable case locks , metastably , onto one of the sliding velocities .
when the soliton s amplitude is small , however , this deceleration is extremely slow ; hence , despite losing energy to radiation , the discrete soliton may spend an exponentially long time travelling with virtually unchanged amplitude and speed .
|
extensive air showers ( eas ) resulting from primary cosmic - ray particles of energies above 1 eev ( @xmath0 ev ) produce an ionization trail which is comparable to that of micro - meteors , which have been detected for many decades using radar methods ( lovell 1948 ; greenhow , 1952 ; hanbury brown & lovell 1962 ) .
in fact , blackett & lovell ( 1940 ) proposed that earlier detections of sporadic radio reflections at altitudes of 5 - 20 km from the troposphere and mesosphere ( cf .
colwell & friend 1936 ; watson watt et al .
1936a , 1936b , 1937 ; appleton & piddington 1938 ) might be attributed to cosmic ray eas , although at the time both the energies and fluxes of such events were only crudely known .
later confirmation of the meteoric origin of higher altitude ionization trails may have contributed to a lack of follow - up on this prescient suggestion for the source of these low - altitude sporadics .
the idea was re - examined in the early 1960 s ( suga 1962 ) and a proposed experiment was described by matano et al . ( 1968 ) ; however , no results from this test have appeared in the literature to date .
eas ionization trails are now commonly detected by their air fluorescence emission at near - uv wavelengths ( cf .
baltrusaitas et al .
future large eas detector arrays such as the auger project ( gurard et al .
1998 ) and the proposed space mission owl / airwatch ( scarsi et al . 1999 ; krizmanic et al .
1999 ) have made fluorescence detection of eas a centerpiece of their approach , since it can provide information such as the position of the shower maximum and the total shower energy which can be difficult to pin down with particle detectors alone .
one of the most compelling reasons to extend the sensitivity of eas detectors in this energy regime is the possibility that neutrinos of energy @xmath1 ev may be an important component of the primary particles ( capelle et al .
1999 ) , contributing to the fluxes at these energies either directly or indirectly .
neutrinos and other highly penetrating particles will induce air showers that develop much deeper in the atmosphere than hadron - initiated showers
. a unique signature of their presence would be the detection of highly inclined ( near - horizontal ) showers for which the point of maximum development is many radiation lengths deeper than is possible for a hadronic or electromagnetic shower .
thus any technique which ( like the air fluorescence approach ) has a particular sensitivity to showers of this type is of interest in addressing this problem . to our knowledge
no one has reported any studies the ionization trail of eas using radar echo techniques , although as we will show here , the signals should be clearly detectable using standard radar methods .
one reason for this may be the transient nature of the expected echoes ; short duration transients are typically rejected by radar systems designed to detect or track targets where the reflectivity does not rapidly decay .
[ [ ionization - columns - meteors - vs .- eas . ] ] ionization columns : meteors vs. eas .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + although there are many significant differences between the ionization caused by meteors and that of eas , we can use the basic formalism developed for meteor study as a starting - point for analysis of eas ionization .
meteor ionization trails are commonly parameterized in terms of their ionization line density @xmath2 ( electrons m@xmath3 ) , a measure of the total ionization content divided by the length of the meteor track .
typical radar - detected meteors occur at heights of 80120 km , and have line densities of @xmath4 to @xmath5 m@xmath3 . at the lowest detectable line densities , the incident meteor has a mass of @xmath6 g , with a radius less than 100 @xmath7 m . at typical velocities of @xmath8 km s@xmath3 , the implied kinetic energy of these meteor grains is 0.05 joules or more , much of which goes into ionization of the air along it path .
a cosmic ray proton of energy @xmath0 ev also has a kinetic energy of order 0.1 j , and much of this energy also ultimately ends up in the form of ionization and excitation of atoms of the air along the path of the shower of charged particles that results from the proton s collision with nuclear hadrons .
the primary differences between the meteor track and cosmic - ray - induced eas are in the way the ionization column forms , and in the resulting ionization density profile . for the meteor , ablation of material from its surface yields atoms with kinetic energies of @xmath9 ev , which ionize air molecules by direct collision with a mean free path of several cm .
this produces an ionization column with an approximately uniform distribution of radial density , and an initial radius of 2 - 10 m. the density then evolves with time due to diffusion , convective processes , bulk motions of the air , and the lorentz forces of the ambient electric and geomagnetic fields . electron attachment and recombination
eventually complete the process of dissipating the ionization column . at radar frequencies in the hf to lower vhf range ( 10 - 100 mhz ) , echoes from typical meteors may be detectable for several seconds after the meteor is gone .
the ionization in an eas , in contrast , is not produced by a single body , but rather by the collective effects of the disk highly energetic particles ( mostly electrons and positrons ) that make up the body of the shower . because the lateral distribution of these particles spreads out as the shower progresses ,
the ionization column is formed with a different initial distribution than that of a meteor , reflecting the evolution of the cross - sectional charged particle density .
in addition , since the shower propagates essentially at the speed of light , it appears almost instantaneously compared to even the fastest meteors at @xmath10 km s@xmath3 .
an important measure of the transverse charged particle distribution in an eas is the moliere radius @xmath11 within which of order 90% of the charged particles can be found . for air at sea level , @xmath12 m , but it is important to note that within @xmath11 the radial distribution is a power law , and most showers retain a tight core of particles of diameter several m or less which may look much like the column initially produced by a meteor .
in fact , of order 10% of the total number of electrons in an eas are likely to be found within a radius of @xmath13 , comprising much less than 1% of the area of the moliere disk . [ [ lightning - ionization - columns . ] ] lightning ionization columns .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + lightning discharges also produce intensely ionized regions which have also been the subject of many radar studies since the early 1950 s ( ligda 1950 , 1956 ; miles 1952 ) .
ionization charge densities produced by lightning are many orders of magnitude higher than those produced by either eas or small meteors , and the plasma channels produced by lightning are also of much smaller diameter than either meteor or eas ionization columns .
however they do occasionally occur in clustered networks with transverse scales comparable to those of eas and meteors , and at altitudes that overlap the altitudes of eas maxima .
thus they can provide examples of radar targets with features that are in some ways relevant to our discussion . in the following section we present the concept of the radar cross section and outline specific cases relevant to eas ionization columns .
section 3 develops a semi - analytical method for determining the ionization densities for a given air shower energy and altitude , and discusses some issues related to the evolution of the electron density .
section 4 then makes estimates of the radar echo power based on the formalism developed in the previous two sections .
section 5 concludes the paper with a discussion of some applications to specific experimental conditions .
for radar detection of the columns that result from meteors , eas events , or lightning , there are two regimes to consider , depending on the plasma frequency @xmath14 of the ionized region : @xmath15 where @xmath16 is the electron density in @xmath17 .
these two regimes are known in as the under- and over - dense regimes , respectively , and traditionally ( in the meteor radar literature ) are divided at the line density of @xmath18 m@xmath3 . however , because of the significant differences in the radial distribution of the electrons in eas ionization compared to that of meteors , we distinguish them only on the basis of the ratio of radar frequency to plasma frequency , a distinction that is also common to radar lightning measurements .
thus we will consider the over - dense portion of an ionization column to be that region where the electron density is high enough that the plasma frequency exceeds the radar frequency , and the radar can not penetrate it and is reflected .
underdense columns are those which have electron densities such that the local plasma frequency is below the frequency of the incoming radar , which can therefore penetrate the ionized region .
radar targets are most commonly described in terms of their effective radar cross section ( rcs ) , a measure of the equivalent physical area of an ideal scattering surface . here
we will develop concepts useful to understanding the rcs characteristic to ionization columns , to prepare for later estimation of the expected radar return power from eas - induced ionization . in the following subsections we introduce general formulas for the rcs for both over- and under - dense cases .
an additional distinction is also useful in each case : the long wavelength , or _
regime involves targets with characteristic dimensions smaller than the radar wavelength used ; and the short wavelength , or _
regime involves cases where the characteristic target size is much larger than the radar wavelength .
when the electron density @xmath19 is high enough to produce a surface where the plasma frequency exceeds the frequency of the incident radiation at some critical radius @xmath20 from the track , the resulting index of refraction becomes imaginary , and total external reflection of the radiation occurs . under these conditions the surface at @xmath20
can be treated to first order as a metal cylinder , and the rcs is accordingly greatly enhanced . when @xmath21 , the cylindrical region of reflecting plasma can be approximated as a thin wire . as we will show later
when we develop an estimate of the ionization density for eas , this is the case that will be most relevant to low - frequency radar echoes from eas . in this case
the maximum of the rcs , which occurs for normal incidence , can be approximated as ( crispin & maffett 1965 ) : @xmath22 \right ) ^2 } \ ] ] where @xmath23 the angle of linear polarization with respect to the axis of the wire .
the behavior of @xmath24 at angles other than normal incidence is given by : @xmath25 \right ) ^2 \right ] } \ ] ] where @xmath26 is the angle as measured from the axis of the wire ( or eas in our case ) .
this approximation is valid in the range @xmath27 , but underestimates the cross section at angles closer to the axis of the wire , due to what are known as _ traveling wave _ effects , which produce large cross sections near the axis , in some cases comparable to that at normal incidence ( medgyeshi - mitschang & putnam 1985 ) , and significantly broader in angular range .
however , the treatment of such effects is very complex and beyond our scope at present . when @xmath28 , the reflection from the critical surface becomes almost entirely specular , and for normal incidence we have ( kraus 1988 ) @xmath29 where @xmath30 denotes the maximum value of the rcs , and @xmath31 is the length of the cylindrical reflecting surface .
radar cross sections for metal cylinders have been studied in detail ( cf .
medgyeshi - mitschang & putnam 1985 ) and are complicated functions of angle , with resonances and nulls that depend strongly on the length and other factors . as we shall establish below , eas ionization columns will generally not fall into the optical regime for the overdense case , except at very high eas energies and very low radar frequencies .
for this reason we do not treat these effects further here . for radar frequencies well above the plasma frequency @xmath14 , the radiation penetrates the entire ionization column , and the electrons within the ionization column scatter independently according to the thomson cross section @xmath32 the total effective radar backscatter cross section @xmath24 will then depend on the individual phase factors of each of the scattering electrons .
for @xmath33 , and angles that are nearly perpendicular to the track , all of the electrons scatter coherently over a longitudinal region of the track @xmath34 where @xmath35 is the perpendicular distance to the track and @xmath36 the radar wavelength ( hanbury - brown & lovell 1962 ) .
this region of the track is known as the first fresnel zone of the track , and the total radar backscatter cross section then becomes @xmath37 where @xmath38 is the number of electrons within a single fresnel zone along the track .
the radar cross section depends of the square of the electron density because of the assumption of full coherence . except for radar frequencies
@xmath39 mhz , most radar ranging of eas ionization columns is not in the rayleigh regime .
however , the concept of the fresnel zone , which establishes a characteristic length scale over the longitudinal extent of the track , will still be useful in what follows . for tracks where the radar frequency @xmath40 , but @xmath41 , the assumption of coherent scattering is no longer satisfied , and the individual phase factors of the electrons must be included . for this case the effective rcs can be written ( cf .
wehner 1987 ; also see appendix ) : @xmath42 where @xmath43 is the wave vector of the incident field ( @xmath44 ) , @xmath45 is the vector distance to the volume element at which the scattering takes place .
here we are neglecting refractive effects of the ionization column on the incident radiation , and assuming that the incident and backscattered waves satisfy @xmath46 . note
that the argument of the exponential here includes an extra factor of two to represent the two - way phase .
this is due to the fact that as the scattered radiation returns on the same path as the incident radiation , it picks up an additional phase factor equal to that of the incoming radiation .
the term @xmath47 represents the differential contribution to the scattered electric field of a volume element of electrons which scatter coherently .
if we now write @xmath48 then equation [ phasor - eq ] becomes @xmath49 equation [ fft - eq ] thus reduces the problem of estimating the effective cross section in the underdense case to that of calculating the fourier transform ( and the resulting power spectrum ) of the electron number density distribution . in the appendix
we describe the geometry more explicitly and show how this relation arises .
a broader implication of this result can be stated as follows : _ a measurement of the complex amplitude of the radar echo from an eas in the underdense case is proportional to a measurement of one fourier component of the shower ionization profile .
_ this will be true even in the case of a bistatic radar ( receiver location not coincident with transmitter ) although there will be an additional coefficient to account for the bistatic cross section and its angular behavior .
in the previous section we introduced the mechanisms for radar reflection from cylindrical plasma columns in two regimes of ionization density . here we estimate the spatial distribution of the expected ionization from a high energy cosmic - ray air shower , which will then allow us to assess their effective radar cross section .
there are many years of development of the theory of eas production .
the most accurate treatments of the evolution of the electron density in the shower require numerical simulations , but there are a number of parameterizations available that yield results accurate enough for our needs . here
we use the analytical model due originally to kamata & nishimura ( 1958 ) , and greisen ( 1965 ) , often referred to as the nkg approximation . in the nkg model the longitudinal ( along - track ) development of the shower at a depth @xmath50 is parameterized by its _ age _
@xmath51 : @xmath52 where @xmath53 mev for electrons in air , and @xmath54 gm @xmath55 is the electron radiation length in air .
the evolution of the total number of charged particles ( virtually all electrons and positrons ) with depth is then approximated by @xmath56 \over \sqrt{\ln ( e / e_{crit } ) } } ~.\ ] ] although this approach does not account for any of the large fluctuations that are possible in high energy air showers , it describes the average behavior reasonably well .
the transverse charged particle densities are described in a similar fashion , also parameterized by the age of the shower and the moliere radius @xmath11 ( bourdeau 1980 ) : @xmath57 where @xmath58 and @xmath59 is the gamma function , and @xmath60 .
the moliere radius for air is given by @xmath61 where @xmath62 is the density of the air at the altitude under consideration , and @xmath63 is the sea level density .
the calculated density @xmath64 is in units of charged particles per unit area , passing through a plane transverse to the shower axis at the given depth .
since the shower thickness is typically of order several meters or less over its length , @xmath64 can be approximately equated with the number density in a slab transverse to the shower . to convert these particle number densities to the resulting ionization density
, we use the average value for electron energy loss given by @xmath65 mev gm@xmath3 @xmath55 , divided by the mean energy per ion - pair for air @xmath66 ev ( segre 1977 ) which accounts for the inefficiency of ion - pair production .
this procedure is similar to that used by the fly s eye ( baltrusaitas et al .
1985 ) in estimating shower parameters based on fluorescence yield .
we note here that the nkg approach and subsequent modifications to it do not accurately account for the electron distribution at large radii from the core , and will in general underestimate the electron density beyond the moliere radius .
however , since the ionization density at large radii is low compared to the core , there is very little coherent contribution from these electrons , in spite of their large integrated number .
our results remain basically unchanged by this effect .
[ ne - fig ] shows curves of calculated ionization line densities for several eas of different primary energies over the range of @xmath67 ev .
the showers are assumed to be propagating horizontally at an altitude of @xmath68 km , and the shower parameters are corrected for the lower air density at this altitude .
the bottom axis shows the along - track distance corresponding to the depth shown along the top axis . as noted above
, the line densities at these energies correspond to typical line densities of radio meteors , which are detected at heights of 80 - 120 km
. however , the lateral distributions will be quite different as noted above .
we note also that the showers are physically very long , with the region of shower maximum extending over many km of physical distance for most showers at these energies . in fig .
[ cumne - fig ] we show the effective line density as a function of the radius within which the electron density is integrated , for the inner 40 m radius for the same set of showers shown in fig . 1 .
here it is evident that the effective line densities of the inner @xmath69 m radius core of the showers are about a factor of 10 lower than those which include all the electrons produced in the shower .
however , this area corresponds to only 0.25% of the area of the moliere disk ( @xmath70 m at 10 km ) , it is thus evident that the cores of the showers are highly concentrated in ionization density with respect to the bulk of the shower disk area . in fig .
[ rhoe - fig ] , we show the lateral ionization density distributions near shower maximum for the same four eas presented in fig .
[ ne - fig ] ( solid lines ) .
included also are curves of the effective plasma frequency corresponding to the density at each radial distance ( dashed lines ) .
given the uncertainty in the accuracy of the analytical model for eas development at very small core radii ( cf .
@xmath71 cm ) , the implied highest radar frequencies that will undergo total reflection are in the range of 10 - 50 mhz , with a strong dependence on the primary energy .
it is evident from figures [ cumne - fig ] and [ rhoe - fig ] that the cores of the showers are likely to be most important in determining the radar response , since the electron density is much higher near the core . in fig .
[ rhoevsalt - fig ] , we plot the core electron density again near shower maximum for a @xmath72 ev horizontal shower , at a variety of different altitudes to display the effects of the different atmospheric density . to first order the shape of the core is preserved with altitude , but the overall electron density decreases at higher altitudes because the production rate of ionization depends on the density of the air . before estimating the radar echo power , we first address the important topic of the expected lifetime of the free electrons in the ionization column .
if eas detection by radar is to be proven practical , it is first necessary to first establish that the ionization column can be treated as quasi - static over the timescale it might take to interrogate the shower with a radar pulse , either as a standalone system , or if a separate air shower detector provided the trigger . as an example of the issue with respect to triggered operation , consider a large air shower array with radius 30 km .
for typical electronics used in such systems the trigger forms within several tens of @xmath7s .
round trip transit time to an eas within @xmath73 km requires another @xmath74s .
thus in this case we require eas electrons to remain mostly free for of order 250 @xmath7s or longer , if an external trigger is to be used to allow interrogation of a detected shower .
in general , if the free electron lifetime is @xmath75 , and the trigger formation and pulse initiation time is @xmath76 , then the maximum range for a triggered system is @xmath77 if a radar detection system is operated in a standalone mode , the free electron lifetime will determine how often the detection volume must be pulsed in order to have a reasonable efficiency at detecting the eas . in the limit where the free electron lifetime becomes so short that it is comparable to the pulse duration ,
then pulsed radar detection will become impractical .
other methods which use continuous - wave ( cw ) techniques may still apply in this case ; these will be discussed briefly in a later section .
there are a number of effects that cause the electron density to evolve with time : 1 .
diffusion of the electrons through the ambient air , lowering their density and thus their net radar cross section ; 2 .
electronic recombination with ions ; 3 .
attachment of electrons to neutral molecules , or dissociative attachment to atoms ; 4 .
collisional detachment of previously attached electrons , which frees electrons that have been attached to atoms ; although the conditions of the atmosphere at the 80 - 120 km heights that meteor echoes are observed at are quite different from those at eas heights , it is useful to preface our analysis with a discussion of the effects that determine the time scale for meteor echoes .
the duration of meteor radar echoes has been modelled and studied experimentally in detail for many decades ( cf .
hanbury brown & lovell 1957 ; kaiser 1968 ; kaiser et al .
1969 ; jones & jones 1990 ; jones 1991 ) .
the power of a meteor radar echo is found in the underdense regime to decay exponentially with a time constant @xmath78 where @xmath79 is the ambipolar ( or ion neutral ) diffusion coefficient . at the altitudes that meteors
are observed with radar , @xmath80 m@xmath81 s@xmath3 , and the typical decay times for underdense trails are thus several tens of ms for frequencies in the vhf regime .
to estimate diffusion effects at the lower altitudes of eas ionization columns , we note that @xmath82 where @xmath83 is the kinetic temperature and @xmath84 the collision frequency ( cf .
buonsanto et al .
thus at 10 km altitude , the diffusion coefficient is much smaller , @xmath85 cm@xmath81 s@xmath3 , due mainly to the much higher collision frequency at lower altitudes . the implied time constant , assuming diffusion progresses in the same manner for eas as for meteor ionization columns , is @xmath87 s for @xmath88 m. s - band radar echoes from lightning at altitudes of @xmath69 km measured by williams et al .
( 1984 ) were frequently found to persist for several hundred milliseconds . for a wavelength of 11 cm ,
the diffusion time constant at this altitude is 240 ms , which suggests that diffusion may be the dominant effect in the evolution of lightning radar cross section , at least at microwave frequencies .
however , lightning ionization densities are estimated to be initially @xmath89 @xmath17 , with plasma frequencies @xmath90 or more times higher than for eas columns .
thus other effects relevant to the conditions of this much denser plasma may be important .
since diffusion is less effective at dissipating the ionization column of an eas than it is for a meteor , at least at vhf frequencies , we expect that attachment and recombination may play a more significant role .
the recombination rate for atmospheric electrons is described by ( cf .
thomas 1971 ) : @xmath91 where @xmath92 are the initial and evolving electron density , respectively , @xmath93 is the electron - ion recombination coefficient ( in units of @xmath94 s@xmath3 ) , @xmath95 and is the electron attachment rate ( s@xmath3 ) .
here we have ignored terms in thomas ( 1971 ) which relate to the presence of negative ions in the ionosphere since we are considering eas within the troposphere where such ions are not prevalent .
the electronic recombination coefficient for hydrogenic atoms can be approximated by ( seaton 1959 ) @xmath96 @xmath97 where @xmath98 and @xmath99 is the electron kinetic temperature , with @xmath100 k typical for ionization in this case .
for air in the troposphere we thus expect @xmath101 to @xmath102 @xmath94 s@xmath3 , and the implied recombination time for the highest electron densities we have considered is several minutes or more .
thus it is highly probable that the electrons will attach before recombining .
@xmath94 s@xmath3 are found in the d region of the ionosphere ( thomas 1971 ) .
such a high value is likely to be due to the large variety of ion species present , which increases the total recombination cross section per electron . ]
molecular oxygen is electronegative , forming ions with free electrons primarily through dissociative attachment , via the process @xmath103 where the resulting oxygen ion and atom carry off the excess kinetic energy of the electron , and the electron affinity of o@xmath104 is of order 0.5 ev .
molecular nitrogen , in contrast , does not form attachments with electrons but can be more easily collisionally excited , and may play a role in mediating detachment processes in air .
several studies of electron attachment in air using electron swarm techniques ( see gallagher et al .
1983 for a compilation ) have indicated a much lower attachment rate than that which is expected from using the attachment coefficient for pure molecular oxygen reduced by the molar ratio in air ( moruzzi & price 1974 & references therein ) .
measurements of the attachment coefficient are very sensitive to contamination by carbon dioxide , and co@xmath104 was not removed from the samples in many early measurements . although co@xmath104 can be present in significant concentrations in the near sea level atmosphere , its average concentration in the atmosphere as a whole is @xmath105% .
thus electron attachment rates for dry , high - altitude air must be determined with co@xmath104 largely removed from typical air samples .
moruzzi & price ( 1974 ) , after carefully removing all co@xmath104 from their samples , confirmed earlier measurements in being unable to detect any attachment in pure dry air .
this result was attributed to a competing process : rapid collisional detachment from n@xmath104 interactions , yielding an effective attachment rate that was less than 10% of the ionization rate and thus too small too measure .
their limit on the effective attachment coefficient @xmath106 for pure air is @xmath107 where @xmath108 is the number density of the attaching molecular species .
the attachment rate @xmath95 is related to the effective attachment coefficient through @xmath109 where @xmath110 is the electron drift velocity . for a typical electron drift velocity of @xmath111 cm@xmath3 s@xmath3 , the limit on @xmath113 gives @xmath114 s@xmath3 , implying that attachment dominates completely over recombination .
figure [ attach - fig ] shows the behavior of the electron density for three different altitudes for a @xmath115 ev shower core region .
here we have plotted the electron density time evolution for both the case where the attachment rate is equal to the limit of moruzzi & price ( 1974 ) , and one - tenth of the limit ( dashed lines ) .
it is evident that the expected electron lifetimes are at least 10 - 20 @xmath7s , increasing slowly at the higher altitudes .
[ attach - fig ] although this estimate of the electron attachment lifetime is important from the point - of - view of evaluating whether there is sufficient time for triggered - interrogation of an eas with radar , it is also important to establish an upper limit on the electron lifetime , both for purposes of evaluating the possibility of multiple - pulse interrogation of a shower , and for considerations of radar clutter from echoes of multiple showers for a radar system with very large range .
the lightning radar measurements noted above ( williams et al .
1984 & references therein ) often found radar echo persistence of several hundred ms using repetitive radar pulses .
however , the plasmas produced by lightning are extremely overdense and hot relative to those we are considering , and thermal detachment may play a significant role . in addition , the large amount of water vapor present in typical lightning radar measurements complicates any comparison with eas in dry air .
thus comparison of radar echoes from lightning with those from eas ionization must be done with caution .
an rough upper limit for the free electron lifetime can be obtained by considering the equilibrium level of free electron content of the mesosphere in the region of maximum cosmic ray ionization , peaking at about 12 - 14 km altitude .
here the dominant ionization is caused by low energy ( several hundred gev and above ) cosmic ray air showers , which produce of order 30 ion pairs per @xmath94 per second ( rossi 1966 ; gregory & clay 1982 ) , roughly 15 times the sea level rate of ion production .
the dc conductivity of the atmosphere at 12 km is @xmath116 ( dolezalek 1982 ) .
the cosmic ray ionization at this altitude induces a tenuous plasma , with dc conductivity @xmath117 where @xmath118 is the dielectric permittivity constant , and the electron - neutral molecule collision frequency @xmath119 can be written as @xmath120 where @xmath108 is the molecule number density ( @xmath17 ) at the altitude considered .
the electron temperature @xmath99 is uncertain but likely to be in the range of 300 - 2000 k. at 12 km we thus expect @xmath121 s@xmath3 for @xmath122 k. inverting equation [ dc - sigma - eq ] for the plasma frequency , which in turn depends on the equilibrium electron number density ( equation 1 ) , we find @xmath123 for the values determined above we have @xmath124 @xmath17 which implies a mean lifetime of less than 20 ms
. the actual electron lifetime will be less than this since the detailed dc conductivity also includes contributions from ionic terms which we have neglected here .
there will also be an altitude dependence on this upper limit since it depends on the collision frequency . in summary
, we find that the available time for eas interrogation by radar is bounded at 10 km altitude approximately by @xmath125 thus , for the shortest values of @xmath75 , triggered interrogation of air showers appears to be impractical except for ranges of a few km or less ; for the largest possible values , ranges of hundreds to even a thousand km or more are not excluded for higher altitude showers . in the latter case ,
no existing system can yet trigger on showers at these great distances , but planned space - based systems such as owl / airwatch will have such capability .
the @xmath126 ms upper limit to the electron lifetime shows also that radar clutter from echoes of multiple showers should be completely negligible .
for example , even assuming an energy threshold as low as @xmath5 ev could be attained out to 10 km , the shower rate at this energy over this area is still only several events per second .
radar systems also generally allow range - gated triggers , and this can facilitate rejecting any nearby clutter from low - energy showers within a few km of the transmitter .
having established that the duration of the eas radar targets is likely to be sufficient for their detection , and introduced the radar cross sections expected , we now turn to estimation of the return power of the echo .
radar return power @xmath127 is described in terms of a model where the radiation is emitted from an antenna with peak transmitted power @xmath128 and directivity gain @xmath129 , where @xmath130 is the solid angle of the main beam of the antenna .
the radiation is assumed to then scatter from objects in the antenna beam and be re - radiated isotropically in the frame of the scatterer , producing a @xmath131 dependence in the returned power as a function of range @xmath35 .
deviations from isotropic scattering are thus absorbed into the effective radar backscatter cross - section @xmath24 , which can be larger or smaller than the physical cross - section of the object .
in addition any real transmitting and receiving system will have less than unity efficiency , which we designate here as @xmath113 .
the radar equation under these conditions is ( cf .
skolnik 1990 ) @xmath132 here we are assuming that the transmitting and receiving antennas are identical , and we are neglecting for the moment any polarization effects or losses in the medium . where @xmath133 is the largest projected dimension of the scattering target . in our case
, this is not generally satisfied , since the length of the ionization columns can be tens of km .
however , we have already accounted for these fresnel zone effects by limiting our analysis to the first fresnel zone as noted above . ]
given equation [ rad - eq ] , the problem of determining the detectability of eas - initiated ionization columns reduces to that of determining the effective rcs @xmath24 for a given choice of operating radar frequency , and the noise power of the specific radar in use .
the noise power is given by @xmath134 , where @xmath135 is the system noise temperature , @xmath136 is boltzmann s constant , and @xmath137 the effective receiving bandwidth , assumed here to be matched to the transmitting bandwidth .
[ [ pulse - compression - radar . ] ] pulse compression radar .
+ + + + + + + + + + + + + + + + + + + + + + + + almost all modern pulsed radar systems now use what is known as _ pulse compression _ , a method which allows the receiver bandwidth , and thus the noise power , to be minimized ( cf .
wehner 1987 ) .
pulse compression is typically implemented by effectively dispersing a band - limited pulse with an initial bandwidth @xmath138 through a filter , transforming it into a frequency chirp which spans the original bandwidth , but now has a duration @xmath139 .
the receiver then uses an inverse filter to de - disperse the received pulse .
the effective bandwidth is thus @xmath140 , but the range resolution @xmath141 of the full bandwidth is recovered .
[ [ cw - radar . ] ] cw radar .
+ + + + + + + + + continuous wave radar systems which transmit at 100% duty cycle , can derive range information by using frequency- or phase - encoded waveforms to uniquely tag the detected echoes to a particular time segment of the transmitted waveform .
processing of the received signal typically involves mixing with a reference signal , which transforms different ranges to different frequencies , followed by detection in a filterbank .
the effective bandwidth is then determined by the bandwidth of the frequency channels used . for simplicity we will assume a pulsed radar system here and in what follows . combining the noise power equation above with equation [ rad - eq ] , the signal - to - noise ratio ( snr ) of the received power is @xmath142 evaluating equation [ sn0-eq ] for a nominal choice of parameters gives the snr per received radar pulse per square meter of rcs : @xmath143 @xmath144 we note that the reference values chosen here represent a modest radar system ; in particular the peak power of 1 kw is easily attained by current standards , and the directivity @xmath145 ( beam size @xmath146 sr ) represents a relatively low - gain antenna .
the system temperature @xmath147 k is realistic for 100 mhz ( @xmath88 m ) however , since the brightness temperature of the sky is quite high at these frequencies .
radar systems also routinely use repetitive pulsing to increase snr , which then grows roughly as @xmath148 where @xmath149 is the number of pulses that are averaged . in the case of an eas ,
the number of pulse repetitions is limited by the free electron lifetime , with large uncertainty .
for the shortest possible electron lifetimes , it is unlikely that repetitive pulsing can provide much improvement in snr , but for lifetimes of order 1 ms , repetitive pulsing could provide substantial improvements in snr .
we now consider in detail the expected rcs for eas ionization columns in the various density regimes that were introduced above . at radar frequencies below the plasma frequency of the ionized core of the eas ,
the rcs is comparable to that of a metal cylinder , as noted above , with a radius of @xmath20 and a length equal to that of the fresnel zone .
based on the analytical shower model used above and the results in fig .
[ cumne - fig ] , the critical radius at a given wavelength and eas cascade energy @xmath150 can be empirically approximated by @xmath151 it is evident from this result that for frequencies above 10 mhz , and for all but the highest eas energies considered here , @xmath21 and we can use the thin wire approximation discussed above . in fig .
[ thinwire - fig ] we show results for radar cross sections of the first fresnel zone for a shower of @xmath152 ev , for three different frequencies , for a range of angles centered on normal incidence , and a polarization angle of 1 rad with respect to the shower axis .
the cross section at normal incidence is greatly enhanced due to the specular reflection at that angle .
we can empirically express the maximum cross section for this case as @xmath153 @xmath154 it is clear from this analysis is due to the fact that the size of the first fresnel zone grows with distance .
in fact , although we have used the length of the fresnel zone @xmath155 as the characteristic length of the `` thin wire '' portion of the shower , the actual length of this region is likely to be much longer than @xmath155 . ]
that overdense showers will show very large cross sections when the reflection is nearly specular .
however , as seen in fig . 6 , the cross sections decrease rapidly for non - specular reflections .
this behavior will be seen to contrast sharply with the case of underdense scattering , to be treated below .
any consideration of the detectability of a particular shower must account for both contributions to the cross section . in the analysis above we have also ignored the additional contribution due to the partially coherent scattering of the much larger volume of underdense plasma outside of @xmath20 .
this region could have refractive effects on the incoming radiation . in an appendix
we treat this issue in some detail and show that it produces a negligible effect for the showers under consideration here . when either the frequency is high enough or the ionization density low enough that there is effectively no region of the column where the critical density obtains , the radar return is due to thomson scattering of the free electrons in the column , modulated by the phase factor associated with the physical extent of the ionized region as described in equation [ fft - eq ] above .
we have made numerical estimates of the effective cross section for a restricted set of cases here , with the methods described in more detail in an appendix .
here we summarize the results . for the case of radar observations at normal incidence to the shower ,
the near constancy of the ionization density along the track allows us to estimate the frequency dependence of the phase factor via a fast fourier transform ( fft ) of the two dimensional ionization distribution in a plane transverse to the track .
we have taken advantage of this feature to make numerical estimates of the frequency - dependent behavior of the phase factors at normal incidence .
note that the requirement of normal incidence for the radar rays interrogating the shower is not as restrictive as it might seem .
for example , consider a horizontal shower whose maximum is at a range of @xmath156 from the radar transmitter / receiver system , and whose axis is inclined an angle @xmath26 with respect to the ray that joins the radar source and the shower maximum . under these conditions , the radar system can still interrogate the shower along a ray that intersects the shower at normal incidence at a distance @xmath157 from the shower maximum .
thus a horizontal shower at 10 km altitude , with 15 km range to its maximum , and inclined at @xmath158 will have a ray that intersects it at normal incidence at a longitudinal distance along the shower of 7.5 km from shower maximum , at a range of 13 km , where the ionization line density is still @xmath159% of its maximum value .
thus the corresponding acceptance solid angle over which these results apply is not negligible , if the beam of the radar system is not too restrictive in solid angle .
[ phasefac - fig ] shows our estimates of the frequency dependence of the phase factor for two horizontal showers at altitudes of 5 and 10 km , for @xmath160 ev .
the phase factor depends strongly on both frequency and shower altitude .
the frequency dependence is due to loss of coherence as noted above , and the altitude dependence arises from the change in moliere radius to smaller values at lower altitudes .
we have also investigated the behavior at other energies in this regime and found these results to be insensitive to energy over the range @xmath67 ev .
for the two cases shown we have included fitted power laws @xmath161 @xmath162 here @xmath163 is the radar frequency in units of 10 mhz .
these relations are individually accurate to about 20% over the range from 10 mhz to 0.3 ghz . given the power loss associated with the phase factor , the total cross section is then estimated by summing the cross sections of the individual volume elements over one fresnel zone of the track , by analogy to the overdense case , and multiplying by the overall phase factor for the radar frequency and shower altitude . evaluating this modified cross section for
its energy and frequency dependence , we derive an empirical expression for the case of a horizontal eas at altitude of 10 km @xmath164 @xmath165 and similarly for an altitude of 5 km : @xmath166 @xmath165 these relations should be valid out to ranges of several tens of km . at the highest energies and lowest frequencies the underdense case may not obtain , but these estimates will then provide a lower limit on the rcs . for shower angles away from normal incidence ,
it is necessary to consider all of the individual scattering contributions of the electrons over the entire volume of the shower , and to treat the pulse echo behavior explicitly .
we discuss this in detail in an appendix .
we have numerically estimated the integrated rcs for several wavelengths at a distance of 10 km to display some of the characteristics of the angular dependence .
the results of this are shown in figures [ thetfac10-fig ] and [ thetfac5-fig ] .
we have plotted the total rcs as a function of angle over about 1 radian around normal incidence , for a horizontal shower at heights 10 and 5 km and energy @xmath72 ev , and several different radar wavelengths .
the dependence on angle is quite different from the overdense case where specular reflection is important ; here , the cross section has an approximately gaussian profile with angle , with a fwhm of order @xmath167 .
the fairly steep wavelength dependence is also independent of angle over this range .
thus the empirical relations derived in the previous section can be used with appropriate factors for the angle of incidence shown here .
we caution here that the angular dependence of the rcs at distances that are much greater than the length of the shower is probably not accurately reflected by these results .
we have not yet derived a general analytical treatment of the behavior for large distances , but we note that the methods described in the appendix allow for numerical estimates to be made at any distance .
we note that , due to self - imposed limitations of scope , a number of effects have not been treated in our analysis .
we list several of these here . [
[ geomagnetic - effects . ] ] geomagnetic effects .
+ + + + + + + + + + + + + + + + + + + + we have neglected geomagnetic charge separation in the shower development .
although this is an important effect in determining the detailed shower structure , it depends greatly on the shower direction with respect to the magnetic field . its effect on the rcs
would also depend on the angle of observation and the treatment of this complex interaction is beyond our scope at present .
[ [ landau - pomeranchuk - migdal - lpm - effect . ] ] landau - pomeranchuk - migdal ( lpm ) effect .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we have neglected the lpm effect in the showers considered here .
this effect , which retards the initial growth of the shower , is important at the energies discussed and should be accounted for in a more detailed treatment .
however , it affects primarily shower which interact initially through the electromagnetic rather than the hadronic channel , and thus there are still many eas at these energies for which the approximation used is still accurate .
[ [ secondary - fresnel - zone - echoes . ] ] secondary fresnel zone echoes .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for simplicity we have only treated the echo from the primary fresnel zone here . in fact , for the high snr case , many distinct echoes may be detected from successive fresnel zones , and these will certainly add significantly to the understanding of any detected shower . on the other hand , for underdense showers that are not observed near normal incidence , the time spread of the return will also tend to decrease the instantaneous snr .
such effects are important in understanding the response of a radar system to the wide variety of possible eas directions and relative angles and will need to be explored further in any real application .
[ [ near - vertical - eas . ] ] near vertical eas .
+ + + + + + + + + + + + + + + + + + for convenience we have analyzed a horizontal air shower , but such showers are of general interest , since such highly - inclined showers imply very deep initial interactions and are thus a possible signature of neutrino primaries . for more typical showers with small zenith angles ,
the moliere radius changes significantly during the shower development and complicated the analysis considerably .
generalization of our analysis to eas at all angles is best done with a full numerical simulation .
preliminary results of such a simulation ( d. bergman 2000 , pers . comm . )
do not show significant differences in the ionization profiles , and still show a tight core in the radial ionization density .
[ [ pulse - smearing - by - the - shower . ] ] pulse smearing by the shower .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + note that our results for the underdense case above apply in a general way only to non - pulsed radar , since the superposition implied by the fourier transforms assumes that each fourier component extends spatially out to dimensions that are much larger than the shower column . for pulses
whose compressed time scale is physically shorter than the time - projected thickness of the ionization column of a shower , we must explicitly account for the behavior of the individual pulse echoes from different portions of the shower .
however , the effective rcs corresponding to the integrated power of the echo is still described by our analysis to first order .
a general consideration of the complications arising from the pulsed nature of the radar are beyond our scope at present but must be included in any detailed estimate of the pulsed radar echo behavior . in the appendix we describe a method developed to deal with pulsed - radar reflections , and we utilize this method in some later results . [ [ plasma - resonance - effects . ] ] plasma resonance effects . + + + + + + + + + + + + + + + + + + + + + + + + + we have treated the underdense case in a simplified way which neglects the possibility of plasma resonance effects .
such effects are in fact seen in meteor echoes ( poulter & baggaley 1977 ) and can play an important role in enhancing the rcs , and in producing more complex time structure in the radar return .
in the previous sections we have presented evidence that eas ionization columns are well within the range of vhf radar detection for cascade energies above @xmath168 ev .
now we turn to a discussion of the applicability of this approach to present efforts at detection and characterization of such eas . to establish a basis for comparison , we first outline the capabilities of fluorescence detector systems , to which eas radar detection , if possible , would be most similar .
[ [ fluorescence - detector - capabilities . ] ] fluorescence detector capabilities .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the most capable present fluorescence detector system is the high resolution fly s eye system in utah ( abu - zayyad et al .
1999 ) operating for several years now in `` monocular '' mode , and now beginning to come on line with stereo - mode operation , using a pair of telescope clusters separated by 12.6 km .
hires uses about 1 photomultiplier per square degree of focal - plane sky coverage , with 2 - 3 m aperture low - resolution optical telescopes used as the light collectors .
hires can detect events out to 20 - 25 km , although at this distance the effects of aerosols increase the uncertainty of the measurements . to achieve reasonable precision in the measurement of 6 shower parameters (
three geometric and three associated with the energy and shower development ) requires typically a minimum of about 20 photomultiplier pixel hits in monocular mode , giving at this level an energy resolution of @xmath169% and precision of about 30 gm @xmath55 in the position of shower maximum ( p. sokolsky 2000 , pers .
significant improvement in the precision of these measurements is expected once the stereo mode comes into full operation ( abu - zayid et al . 1999 ) .
observation efficiency for any fluorescence detector is limited to about 10% since the observations can only be made on clear , moonless nights .
the total acceptance aperture for hires is expected to be about 6000 km@xmath81 sr in the region of @xmath72 ev shower energy . here
we consider what type of radar system would be required to produce eas measurements that can be compared to those expected from hires .
because of the issues of interference rejection , a standalone eas radar system would almost certainly have to be constructed with multiple stations configured to both transmit and receive , with pulse encoding such that each station would be sensitive to pulses produced by the other stations .
bistatic radar cross sections of eas will vary only modestly from the monostatic rcs values treated here .
global synchronization of the stations to several tens of ns or less is routine with the global positioning system , particularly with the disabling of selective availability ( sa ) of the high precision gps codes . to optimize the signal - to - noise ratio of the detected echoes , we wish to operate at the lowest frequency that is practical from considerations of background noise , including ionospheric noise from daytime operation .
although lower frequency operation may be possible , we choose here to consider only frequencies above about 30 mhz . in the hf - vhf regime the effective system noise temperature for a remote site ( distant from urban noise sources ) can be written as ( skolnik 1986 ) : @xmath170 which yields @xmath171k at 30 mhz . assuming that each station receives an echo from its own as well as all others in an array of @xmath172 transmit / receive stations , there are @xmath173 complex measurements for each combined pulse of the @xmath172 stations . for six eas parameters to be fitted ,
the minimum number of stations required is @xmath174 ( giving 6 total geometric ranges and echo amplitudes ) , assuming that the measurements are independent . given that there will be some covariance of the measurements , and requiring that the fit be well - constrained , at least 4 or 5 stations will be needed , giving 12 - 20 ranges and amplitudes .
additional closure - delay ( or triangle ) measurements number @xmath175 , and will provide an extra 4 - 10 independent constraints for the @xmath176 cases , respectively .
however , these latter constraints apply only to the geometry , which is typically well - constrained in any case , and not the echo amplitude .
addition of dual - polarization capability , which is not uncommon in radar systems , would double the number of measurements .
the measured data can be characterized as follows : [ [ range - data - shower - angle . ] ] range data & shower angle .
+ + + + + + + + + + + + + + + + + + + + + + + + + + the time delay to the leading edge of the echo at each station will provide the range from the transmitter to the receiver .
the ranges can be measured typically to a precision of roughly the compressed pulse width divided by the snr . for a snr=10 , and a compressed pulse width of 250 ns
, the range resolution is below 10 m. a complete determination of the angular resolution possible using radar is beyond our scope , but simple arguments can show that a set of 4 or 5 stations should provide reasonable estimates of the track direction . a schematic view of the geometry associated with the range measurements for a pair of stations is shown in fig .
[ angres - fig ] . here
the stations 1 and 2 measure leading - edge ranges @xmath156 and @xmath177 to separate portions of the shower at along - track positions @xmath178 with respect to shower maximum . the stations are separated by a baseline @xmath179 which is not generally parallel to the shower track .
let a vector @xmath180 bisect the baseline and the shower segment from @xmath181 to @xmath182 .
the shower direction can then be characterized by angles @xmath2 , which measures the rotation of the shower track around the axis @xmath180 with respect to the projected baseline , and @xmath183 which measures the angle between the shower and @xmath180 .
[ angres - fig ] it is evident from fig .
[ angres - fig ] that changes in @xmath183 will under most conditions produce first order changes in the range values , eg .
@xmath184 where @xmath185 is the mean range error and @xmath186 the resulting angular resolution in @xmath183 .
since for tracks that are primarily transverse to the baseline @xmath179 , we have @xmath187 , then baselines of order 5 km or more with range errors of @xmath69 m will give angular resolution in @xmath183 of order several mrad , or about @xmath188 .
changes in @xmath2 for a pair of stations under typical expected detection geometry do not give first order effects in range , and are thus more problematic . because a change in @xmath2 corresponds to a kind of `` torsion '' mode of the range contraints ( using a rigid - rod analogy ) ,
the error in @xmath2 for a two station solution is likely to generally behave as @xmath189 which applies most directly to the case where @xmath190 .
for example , for @xmath191 km , @xmath192 km , and @xmath193 m , we have @xmath194 , much lower precision than can be achieved in air fluorescence detectors .
we note however that the addition of more radar stations , with the inclusion of bistatic ranges and multiple polarization measurements should rapidly improve the precision as more constraints are added , but further estimates are beyond our scope at present . [
[ echo - amplitude - data . ] ] echo amplitude data .
+ + + + + + + + + + + + + + + + + + + + the echo amplitude is more difficult to interpret . for underdense showers ,
the echo amplitude is directly proportional to the number of electrons present , and thus is similar to fluorescence detectors in terms of its ability to estimate the number of ions within a certain region of the track .
however , since the radar echo may be spread out over a band of different ranges , determination of the contributing portion of the shower for each portion of the returned echo at each receiving station is a clearly more complicated problem than that faced by fluorescence detectors , which rely on imaging precision to solve this problem .
however , since the geometric parameters of the shower track are easily separable via the range data , the resulting amplitude of the leading edge of the return can be attributed unambiguously to the portion of the ionization column that is nearest to the receiving station . by thus reconstructing a profile of these prompt received amplitudes at their relative positions along the geometric track , the profile of the shower development can be extracted . for overdense showers
, the radar echos may show anomalously high strength compared to expectations based on an underdense echo . in this case ,
interpretation of the data may rely on distinguishing these two cases on an event - by - event basis .
estimates of the effective conductivity of the plasma can then still yield the electron content of the shower . in either of these cases , the amplitude data ,
although not as straightforward to interpret at that of fluorescence measurements , still provides the type of calorimetry of an eas that is necessary to determine the shower energy .
this is a feature that particle counter arrays can not match in detail , and it sets apart the fluorescence detection and potential radar detection techniques from other methods .
perhaps the most interesting feature of radar as a potential eas detection approach is the possibility of much higher duty cycle observations over detection volumes comparable to those of fluorescence detectors . in principle
a radar system is not limited to night operation , and suffers very little attenuation from clouds or aerosols .
although operation in the presence of lightning within the detection volume might produce some spurious events , avoiding operation in periods of lightning would cost at most a few percent of the livetime .
the largest uncertainty in efficiency is probably due to the free electron lifetime , since if this is very short , it may be difficult to ensure adequate pulse repetition frequency to detect all showers . and although radar propagation is unaffected by aerosols and clouds , the free electron lifetime is not immune to variations in these components of the atmosphere , and in practice they may be important limitations to the efficiency .
in addition , the presence of the sun above the horizon will increase the system temperature somewhat , especially near solar maximum , and this will act to increase the detection threshold for daytime operation . in figure
[ array - fig ] we show a plan view of the geometry of a radar array using both monostatic and bistatic detection of an eas , here shown with 4 stations at 3 - 5 km spacing , providing 10 ranges and 10 pulse amplitudes . here
we describe the parameters required of each transmit / receive station to satisfy the needs of the array described .
[ array1 ] if we assume that we require an energy threshold of @xmath115 ev out to a range of 20 km , the implied rcs for the underdense case is @xmath195 m@xmath196 at this energy .
the fresnel zone length at this range is about 250 m for a shower at 10 km altitude , corresponding to about 10 g @xmath55 along the track . using a commercial vhf radar system , it is straightforward to achieve 60 kw peak power and pulse repetition rates of 10 - 50 khz for @xmath197s pulses .
the sky brightness temperature will lead to a system temperature of @xmath171 k at 30 mhz as noted above .
since we wish to be sensitive to as large a volume as possible , we use an antenna with a broad beam ( thus corresponding to a low gain ) .
a suitable system is a vertical monopole or discone system ( balanis 1997 ) which provides good @xmath198 azimuthal coverage from @xmath199 to about @xmath200 elevation , giving an overall gain of @xmath201 with respect to an isotropic antenna .
we also assume here a somewhat pessimistic overall transmit / receive efficiency of @xmath202 , and a single pulse echo ( no pulse averaging ) to account for possible rapid decay of the free electron population . using these assumptions we show the snr budget , in dbm ( decibels referenced to 1 milliwatt ) in table [ auger - tab ] .
this budget corresponds to a single monostatic or bistatic pulse measurement in the array .
the resulting snr of 6.4 provides a range resolution of 34 m at a 30 mhz operating frequency , using a chirp bandwidth of 10% ( 3 mhz in this case ) .
lcc parameter & value & @xmath203dbm + + & @xmath204 & + & @xmath205 & + + peak transmit power & 60 kw & 77.8 + pulse duration & @xmath206s & ... + chirp bandwidth & 3 mhz & ... + number of repetitions & 1 & 0.0 + antenna gain & 3 & 9.54 + wavelength & 10 m & 20.0 + @xmath24 at @xmath207 ev & 3.8 m@xmath81 & 5.8 + range to eas & 20 km & -172.0 + xmit / rcv efficiency & 0.05 & -13.0 + @xmath208 & @xmath209 & -33.0 + + * received power * & & -104.9 dbm + + & @xmath210 & + + boltmann s constant & @xmath211 & -198.6 + & @xmath212 & + system temperature & 3650 k & 35.6 + effective bandwidth & 100 khz & 50.0 + + * noise power * & & -113.0 dbm + + * snr * & 6.4 & 8.0 db + + we can also estimate the rate of shower detection as a function of energy . to do this
we estimate that the acceptance solid angle of the system is @xmath146 sr , based on the rcs within @xmath213 of the peak values .
this is significantly lower than that of a fluorescence detector , and reflects the rapid decrease in the rcs as the angle of incidence moves away from the normal to the shower axis .
we also consider only the rcs for showers at 10 km altitude , although showers at lower altitudes will have significantly higher cross sections in general .
the integral number of events seen above a given energy @xmath214 is given by @xmath215 where @xmath216 is the observing efficiency , @xmath83 the observation duration , @xmath217 is the effective solid angle , and @xmath218 is the integral cosmic ray spectrum at these energies . here
the maximum range for detection is determined by inverting equation [ sn - eq ] above , which yields : @xmath219 @xmath220 @xmath221 this relation is approximately valid to energies of order @xmath72 ev , where the detection range is @xmath222 km . beyond this range ,
the minimum beam elevation is such that it exceeds the initial assumption of 10 km shower altitude , and the detection efficiency as a function of shower altitude must be more explicity addressed however , this exercise highlights the large potential detection area for a radar system [ easrate ] in figure [ easrate - fig ] , we plot the expected event rate per year for 80% observing efficiency , which is achievable with a fully operational system . here
the assumed minimum @xmath223 . also plotted with respect to the right axis is the effective aperture , in km@xmath81 sr , for the same assumptions .
it is evident from this analysis that eas detection by radar certainly merits further investigation as a high - duty cycle alternative to existing eas detection systems .
although it is not a competitive technique at lower energies , it appears that this approach may have significant benefits for detection of eas at the highest energies .
although it may not be able to compete with fluorescence detection in its ultimate accuracy in the estimation of shower parameters , a system which utilizes a trigger provided by an existing particle detector array such as the auger observatory could potentially gain valuable complementary information on daytime events where fluorescence information can not be obtained .
we have demonstrated that , using standard models for the average behavior of extensive air shower development , the resulting ionization appears straightforward to detect using radar techniques in the vhf frequency range ( 30100 mhz ) , for primary energies greater than about @xmath0 ev .
we estimate that a relatively modest ground - based system could provide information on eas that is comparable to that available from other techniques , and may do so with a much higher duty cycle and potentially a larger effective aperture .
the approach requires no new technology , making use of standard radar equipment and detection techniques , and may thus also provide a low - cost alternative to existing methods .
we thank an anonymous referee who provided a very valuable critique .
we thank douglas bergman , george resch , pierre sokolsky , paul sommers , david saltzberg , and trevor weekes for useful discussion and comments .
this research has been performed at the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration .
antoni , t. , et al .
, 2000 , astroparticle phys .
, in press , ( also arxiv : astro - ph/0004233 ) .
appleton , e. v. , and piddington , j. h. , 1937 , proc .
a , 164 , 467 .
balanis , c. a. , 1997 , _ antenna theory _ , 2nd edition , ( new york : wiley & sons ) .
baltrusaitis , r. m. , cady , r. , cassiday , g. l. , cooper , r. , elbert , j. w. , gerhardy , p. r. , ko , s. , loh , e. c. , salamon , m. , steck , d. , & sokolsky , p. , 1985 ,
nim a240 , 410 .
blackett , p. m. s. , and lovell , a. c. b. , 1940 , proc .
royal soc .
( london ) ser .
a , 177 , 183 .
bourdeau , m. f. , et al . , 1980 ,
phys . 6 , 901 .
buonsanto , m. , sipler , d. p. , davenport , g. b. , & holt , j. m. , 1997 , j. geophys .
102 ( 17 ) , 267 .
capelle , k.s .
, cronin , j.w . , parente , g. , & zas , e. , 1998 , astropart . phys . 8 , 321 .
colwell , r. c. , and friend , a. w. , 1936 , phys .
50 , 632 .
dawson , b.r .
et al . , 1997 ,
25th international cosmic ray conf . , durban , og 10.6.16 .
dolezalek , h. , 1982 , in hand .
phys . , ed .
weast & astle , f-172 .
gallagher , j. w. , beaty , e. c. , dutton , j. , and pitchford , l. c. , 1983 , j. phys .
chem . ref .
data , 12(1 ) , 109 .
greenhow , j. s. , 1952 , proc .
b , 65 , 169 .
gregory , a. , and clay , r. w. , 1982 , in hand .
phys . , ed .
weast & astle , f-175 .
greisen , k. , 1965 , in prog .
cosmic ray physics vol .
iii , j.g .
wilson ed . ,
( north holland : amsterdam ) 1 .
gurard , c. k. , et al .
1998 , proc .
workshop front .
astroph . & part .
physics , vulcano , sicily .
hanbury brown , r. , & lovell , a. c. b. , 1962 , _ the exploration of space by radio _
, 2nd ed . , ( new york : wiley ) .
hightower , c. h. , brown , t. , & soon , n. y. , 1993 , proc . of the ieee nat .
radar conf . ,
jones , j. & jones , w. 1991 , planet .
space sci .
19(9 ) , 1289 .
jones , w. , 1991 , planet .
space sci .
1991 , 19(9 ) , 1283 .
kaiser , t. r. , & closs , t. , 1952 , phil . mag .
43 , 1 .
kamata , k. , & nishimura , j. , 1958 , suppl .
phys . 6 , 93 .
kraus , j. d. , 1988 , _
antennas _ , 2nd ed . , ( new york : mcgraw - hill ) .
krizmanic , j. et al . , 1999 ,
cosmic ray conf . , eds . kieda , dingus & salamon , ( univ .
utah : salt lake city ) vol 2 , 388 .
lidga , m. g. h. , 1950 , bull .
soc . , 31 , 279 . lidga , m. g. h. , 1956 , j. atm .
phys . 9 , 329 .
lovell , a.c.b . , & clegg , j. a. , 1948 , proc .
60 , 491 .
matano , t. , nagano , m. , suga , k. , and tanahashi , g. , 1968 , can . journ .
46 , s255 . miles . v. g. , 1952 , nature 170 .
poulter , e. m. , & baggaley , w. j. , 1977 , journ .
39 , 757 .
rosner , j. l. , and wilkerson , j. f. , 1997 , appendix to the proposal for the auger air shower array ; hep - ex/9702008 .
rossi , b. b. , 1964 , _ cosmic rays _
, new york : mcgraw - hill ) .
sahr , j. d. , and lind , f. d. , 1997 , radio sci .
32(6 ) , 2345 .
scarsi , l. et al . , 1999 ,
cosmic ray conf .
, eds . kieda , dingus & salamon , ( univ .
utah : salt lake city ) vol 2 , 384 .
seaton , m. , 1959 , 119 , 81 .
segre , e. , 1977 , _ nuclei and particles _
, 2nd ed . , ( reading , ma : benjamin / cummings ) .
skolnik , m. , 1990 , _ radar handbook _ ,
( new york : mcgraw - hill ) , 2nd ed .
stalio , r. et al . 1999 ,
cosmic ray conf . , eds . kieda , dingus & salamon , ( univ .
utah : salt lake city ) vol 2 , 403 .
sovers , o. j. , fanselow , j. l. , and jacobs , c. s. , 1998 rev .
70(4 ) , 1393 .
suga k. , 1962 , proc .
fifth interamerican seminar cosmic rays , 2 , xlix .
thomas , l. 1972 , journ .
33 , 157 .
watson watt , r. a. , bainbridge - bell , l. h. , wilkins , a. f. , and bowen , e. g. , 1936 , nature 137 , 866 .
watson watt , r. a. , wilkins , a. f. , and bowen , e. g. , 1936 , proc .
a , 161 , 181 .
wehner , d. r. , 1987 , _ high resolution radar _ , ( norwood , ma : artech house ) .
williams , e. r. , geotis , s. g. , and bhattacharya , a. , b. , 1989 , j. atm .
sciences , 46(9 ) , 1173 .
zhou , q. h. , and kelley , m. c. , 1997 , journ . atm .
& sol .- terr
59(7 ) , 739 .
to estimate the rcs for the underdense regime of an eas ionization column , we need to evaluate the individual contributions of the scattering electrons and the phase of the resulting scattered fields .
as noted in section 2 above , under certain conditions this is equivalent to determining the fourier transform of the electron distribution .
here we describe the basis for this conclusion and provide details of the methods used for numerical evaluation of the rcs under these circumstances . the geometry for this evaluation is shown in fig .
[ geom - fig ] .
here we let @xmath224 represent the vector distance from the origin @xmath225 ( at the radar transmitter / receiver ) to the position of shower maximum .
we choose the @xmath226 plane to be coincident with the plane containing the axis of the shower and @xmath224 . within the ionization column itself ,
we use cylindrical coordinates @xmath227 to calculate the proper ionization density at position @xmath228 with respect to the position of shower maximum ; @xmath229 is the vector separation of @xmath228 from shower maximum .
the vector joining the origin to @xmath228 is denoted as @xmath45 .
the angle between the shower axis and @xmath224 is denoted @xmath26 .
given these definitions , we find @xmath230 which expresses the distance to an arbitrary point in the ionization column as a function of the distance to shower maximum , the local cylindrical coordinates of the point , and the angle of obliquity @xmath26 of the shower
. the electric field that is incident at @xmath228 can be written as @xmath231 where @xmath232 is the magnitude of the transmitted field , @xmath233 is the transmitter gain factor ( as defined in the main text ) , @xmath44 , and @xmath45 is defined in fig .
[ geom - fig ] . assuming that @xmath43 and @xmath45 are parallel ( that is , neglecting the effective dielectric behavior of the underdense plasma in the column ) ,
the received field from a small volume of electrons @xmath234 which scatter coherently is then given by @xmath235 where @xmath236 is the effective gain of the electron scattering process ( @xmath237 for isotropic scattering ) , corresponding to electric dipole emission . however , this response is complicated in any real plasma by a number of effects , and we have assumed @xmath237 for all cases considered here . ] and @xmath238 are the wavevector and radial vector of the scattered wave .
in this case we have @xmath239 and @xmath240 , and thus @xmath241 which displays the expected radial dependence of the field ( equivalent to a @xmath242 dependence in received power ) , and also the additional factor of two in the spatial phase which is due to the two - way trip of the radiation .
if we assume for the moment that the time variation of the field is steady state , we can drop the time dependence , and integrate over the volume of the shower to yield the total received field strength , assuming that the transmitted beam is larger than the dimensions of the shower .
if we also assume that the distance from the shower is large enough that @xmath243 , then we have @xmath244 which shows ( taking @xmath48 ) that the received field is proportional to the fourier transform of the electron density distribution . under conditions where the pulse is incident on the ionization column at normal incidence , and
if we assume that the ionization column is essentially constant along its longitudinal length , the received field becomes proportional to the two dimensional fourier transform of the cross sectional electron density , times the length of the fresnel zone @xmath155 ( defined as the length over which the scattered fields are in phase ) : @xmath245 we have numerically evaluated this using fast fourier transform techniques , since the area @xmath246 over which this is done is of order @xmath247 and the number of @xmath248 cells in the grid is @xmath249 , where we have assumed @xmath250 m , giving a @xmath251 grid .
the result of the fft can be normalized to provide a frequency dependent phase factor , which we have plotted in fig .
[ phasefac - fig ] above . in the previous analysis
we have assumed a steady state incident field ( which corresponds to carrier wave or cw radar ) .
however , in our analysis of section 4 , we assumed that the radar is pulsed , and in fact pulsed radar is preferable to achieve the kind of time resolution required in this application .
for the overdense case , the analysis presented in section 4 is adequate , since we are considering reflections from a localized source at the critical radius of the ionization column .
also , for normal incidence , the reflections from the first fresnel zone arrive closely in phase and can be treated without explicitly accounting for the pulse echo behavior .
when we allow for oblique shower angles , however , the concept of the fresnel zone loses its usefulness , and the received echo comes from the entire volume of the shower and is thus significantly spread out in time .
we require a method of treating the behavior of a discrete pulse under these conditions .
the approach we have followed is intuitively simple : we first produce a time - projected profile of the electron density ; that is , we calculate the round - trip time associated with the global phase of each of the contributing differential volume elements of the scattering electrons @xmath252 here @xmath253 is the round - trip time for the @xmath254th volume element at location @xmath255 .
if we consider all of the volume elements for which @xmath256 within a specified interval @xmath257 , then we define a shell at distance @xmath258 with a thickness of 1 radian of phase at the radar angular frequency @xmath259 .
this is depicted in fig .
[ tprogeom - fig ] . summing all of the contributing elements in this shell
gives the time
projected square root of the total cross section at that phase : @xmath260 once @xmath261 is determined for a given radar wavelength , shower angle and energy , and distance , then the pulse response of the ionization column is determined by convolving the electric field of the pulse profile with @xmath261 , and squaring the result . if the pulse is defined to have unit amplitude , then the resulting convolution can be expressed as an effective radar cross section for use in equation [ sn - eq ] . for example ,
if the pulse is assumed to be a @xmath262-function in time , the resulting convolution @xmath263 is @xmath264 and the received power is then proportional to @xmath265 . for other pulse profiles we numerically evaluate the convolution explicitly , and take the maximum of the resulting profile to be the effective radar cross section . in fig .
[ timepro - fig ] we show an example of the square of @xmath261 , giving the time
projected cross section calculated according to the manner described here . in each case
the time bins correspond to a radian of angular phase according to the wavelength indicated , and the value for each bin is estimated according to equation [ y - eq ] .
the profile in this case shows a characteristic rapid rise on its leading edge , reflecting the rise in cross sectional ionization density of the nearest portion of the shower .
the more slowly falling trailing edge is due to the contributions from portions of the shower further away .
we note that @xmath261 defined in this way is analogous to a green s function for the time response of the system .
it can be used in this manner to evaluate arbitrary radar pulse shapes , for a given eas energy and direction . for extremely broad - band radar systems
, the effective rcs values will begin to approach those of @xmath261 . for more narrow - band radar pulses
, the convolution of the field oscillations with @xmath261 will act to partially cancel and the effective rcs will decrease .
the refractive index of a plasma is given by spitzer ( 1962 ) in the quasilongitudinal approximation by @xmath266^{1/2}\ ] ] where we have assumed that the frequency of the radiation is well above the electron gyrofrequency in the earth s magnetic field ( @xmath267 khz ) .
in this case the phase delay of the radiation after propagating a distance @xmath268 through the plasma is given as ( cf .
sovers et al .
1998 ) : @xmath269 where the variable @xmath270 here represents the column density of the electrons integrated along the ray path : @xmath271 where @xmath272 is the classical electron radius and @xmath273 is the speed of light . the negative phase delay , or phase advance given above is characteristic of propagation in a plasma .
more appropriate for consideration here is the group delay , obtained by differentiating @xmath274 with respect to frequency @xmath275 which constitutes a net additive delay that increases at lower frequencies .
numerically , this delay has the value @xmath276 where @xmath277 is the electron column density in e@xmath278 @xmath55 integrating the electron densities in the halo of a @xmath72 ev shower along a radial path from outside the shower down to a radius of 1 m and then back out again ( to approximate the radar ray path ) yields @xmath279 e@xmath278 @xmath55 , and the corresponding delay at 30 mhz is 0.6 ns , or about @xmath280 of rf phase . a group delay of this order will have a negligible effect on the received power compared to the geometric phase factors .
at lower frequencies , such at 10 mhz , the group delay is an order of magnitude larger , but the induced rf phase lag is still only @xmath281 .
thus it appears that the refractive effects of the radial plasma halo can be ignored to first order for the eas ionization columns considered here .
similar analysis of ray - path deviation along chords in the outer halo ( due to the gradient in the plasma density ) shows that the refraction is not significant until the chords approach within several m of the eas core for the highest energy showers .
thus although it is important to include such effects in any final calibration of parameters estimated by radar measurements , they do not appear to bear significantly on the detectability of the showers .
|
we revisit and extend the analysis supporting a 60 year - old suggestion that cosmic rays air showers resulting from primary particles with energies above @xmath0 ev should be straightforward to detect with radar ranging techniques , where the radar echoes are produced by scattering from the column of ionized air produced by the shower .
the idea has remained curiously untested since it was proposed , but if our analysis is correct , such techniques could provide a significant alternative approach to air shower detection in a standalone array with high duty cycle , and might provide highly complementary measurements of air showers detected in existing and planned ground arrays such as the fly s eye or the auger project .
the method should be particularly sensitive to showers that are transverse to and relatively distant from the detector , and is thus effective in characterizing penetrating horizontal showers such as those that might be induced by ultra - high energy neutrino primaries .
|
the atlas detector at cern is one of two large general purpose detectors at the large hadron collider ( lhc ) @xcite , a 7 tev on 7 tev proton accelerator .
it was designed to have excellent tracking , calorimetry and muon spectroscopy over the entire energy range of the lhc allowing discovery of any new physics in that range .
particular discovery potentials include details of the standard model and beyond , such as discovery of the higgs boson , supersymmetry ( susy ) and extra dimensions .
more complete discussion of the physics discovery potential of atlas is given in reference @xcite .
the atlas detector , under construction for well over a decade , was ready to record collision events on september 10 , 2008 when the lhc started up .
single beams were circulated around the accelerator and resulting events were recorded in the detector . as is well known , about one week later , an incident in a superconducting splice between two dipoles resulted in a shutdown of the machine before any beam collisions occurred
. this shutdown is still continuing and atlas has used this time to fix minor problems with the detector and commission the subsystems with cosmic ray events . the lhc accelerator is now scheduled to start again in november , 2009 .
this paper will describe the current state of each subsystem and present results of these cosmic ray events showing that atlas now fully ready for collision data that is expected later this year .
the atlas detector is large in more than one way : the outer dimensions ( a cylinder 44 m long and 25 m in diameter ) and the collaboration : over 2500 scientists from nearly 200 institutes in 37 countries . an overview schematic of the detector is shown in fig .
[ atlas_detector ] where the major subsystems are labelled .
the outermost part of atlas is the muon spectrometer and its associated superconducting toroidal magnets which give atlas its name : a toroidal lhc apparatus .
the key design in the muon spectrometer is minimal material provided by having air - core toroidal magnets .
next , moving inward is the calorimeter system consisting of three different technologies .
the hadronic calorimeter is a scintillating tile / iron structure covering @xmath0 . in the endcap region
@xmath1 hadronic calorimetry is provided by a copper / liquid argon structure providing four longitudinal samples .
forward hadronic calorimetry is provided by a tungsten / liquid argon structure covering @xmath2 with three longitudinal samples .
electromagnetic calorimetry is provided by a lead / liquid argon detector @xmath3 with three longitudinal samples together with a presampler detector in the @xmath4 region .
inside of the calorimeter is the inner detector consisting of a transition radiation tracker ( trt ) , semiconductor tracker , and a pixel detector shown in fig .
[ inner_detector ] .
the magnetic field for the inner detector is provided by a solenoid magnet providing a 2 tesla field .
the trt provides @xmath5 separation over the energy range @xmath6 gev .
the pixel readout has 80 million channels , another one of the superlatives in atlas .
complete details of the atlas detector are given in reference @xcite
during the brief single beam running in september , 2008 , atlas recorded many events from beam halo ( interactions with gas in the beam pipe ) and so called beam - splash events produced when the beam was run into a collimator approximately 150 m upstream of the detector .
the beam splash events provided rather spectacular event display pictures , such as shown in fig .
[ splash_event ] , but also allowed time synchronization between the subsystems and the lhc machine .
since the end of the beam in sept . 2008 , atlas has been accumulating cosmic ray events in all subsystems of the detector .
[ cosmic_runs ] shows the running periods and the state of the magnets during each run . over 200 million events
have been recorded .
the following sections will review the status of each atlas subsystem and show results from these cosmic ray runs .
the atlas inner detector has three major detector components : the pixel detector , the silicon detector and the transition radiation tracker as shown above in fig .
[ inner_detector ] .
the pixel detector currently has approximately @xmath7 live channels , a hit efficiency of @xmath8 and a noise occupancy of @xmath9 the semiconductor tracker has over @xmath10 of its 6 million channels operational .
the noise occupancy is @xmath11 in the barrel region and @xmath12 in the endcap .
the transition radiation tracker ( trt ) has @xmath13 of its 350k channels operational .
the transition radiation aspects of the trt have been proven in testbeam studies where the full turn on of transition radiation is seen and in the cosmic ray runs where the beginning of the turn on is seen .
this is shown in fig .
[ trt_muons ] where the probability of a signal above the high threshold is shown as a function of the lorentz gamma factor of the cosmic muon .
transition radiation photons produced by particles with high gamma - factor cause a larger energy deposition in the straw tubes of the 1000 trt .
the atlas liquid argon ( lar ) calorimeter systems , described above and shown in more detail in fig .
[ calorimeter ] , has @xmath14 of channels operational .
the hadronic tile calorimeter has @xmath15 dead channels , which will be fixed at the next shutdown .
the electronic calibration systems for all the calorimeter systems are fully operational .
an example of studies with cosmic rays is shown in fig .
[ lar_response ] , where the response of the lar barrel electromagnetic calorimeter is shown as a function of pseudorapidity .
here the response is the most probable value ( mpv ) determined by two different algorithms .
the larmuid algorithm is a variable size algorithm - only cells above a given threshold are added to the cluster .
the 3x3 cluster algorithm is fixed in size .
also shown are the true cluster response from monte carlo simulation and the cell depth .
the data match the simulation and follow the cell depth as expected .
the uniformity of response agrees with the simulation at the @xmath16 level .
point between 0.3 and 0.4,width=302 ]
the muon spectrometer in atlas has two different technologies for the precision measurement and two for the trigger as shown in fig .
[ muon_detector ] . the monitored drift tube ( mdt ) system together with the cathode strip chambers ( csc ) in the forward region provide a spatial resolution of @xmath17 .
the mdt tubes are `` monitored '' by a 12232 channel optical alignment system which provides position accuracy of @xmath18 .
the mdt system has @xmath19 dead channels .
the csc system has @xmath20 dead channels .
the muon trigger is provided by resistive plate chambers ( rpc ) in the barrel and thin gap chambers ( tgc ) in the endcap .
these provide a spatial resolution of @xmath21 mm and a time resolution sufficient to allow bunch crossing identification with the @xmath22 ns bunch crossing time of the lhc .
the rpc system has @xmath23 channels operational with an additional @xmath24 recoverable during a shutdown .
the tgc system is @xmath25 operational with less than @xmath26 noisy channels .
overall , the muon system provides a stand alone ( that is , not combined with inner detector measurements ) resolution of @xmath27 up to 1 tev .
a typical cosmic ray muon seen in the spectrometer is shown in fig .
[ muon_event ] . as can been seen by the track bending ,
the magnetic field was on for this event .
effects of the muon alignment system can be seen in fig . [ muon_residuals ] where the sagitta measured in the bend plane of the magnet is shown from tracks that go through one particular chamber in the muon barrel system .
the top plot in that figure shows the sagitta with just the nominal geometry for the chambers that the track passes through . the middle plot shows the sagitta after corrections given by the optical alignment system .
the bottom plot shows the final distribution obtained after applying corrections from track - based alignment .
for this analysis , 3.5 million cosmic ray events which passed the high - level trigger track reconstruction in the barrel inner detector were used . after filtering for electromagnetic cluster candidates with transverse energy above @xmath283 gev and with a loose track match in @xmath29 ( pointing downwards ) , about 11 000 candidates remain .
these remaining events are required to satisfy medium electron cuts ( lateral shower shapes in first and second layers of em calorimeter and track- cluster matching in @xmath30 for tracks with at least 25 trt hits ) and are then split into two categories : a ) a sample consisting of 1229 muon bremsstrahlung candidates , with only one track reconstructed in the barrel inner detector b ) a sample consisting of 85 ionisation electron candidates , with at least two tracks reconstructed in the barrel inner detector in fig .
[ electrons_1 ] , the red boxes correspond to candidates satisfying additional cuts defined for standard tight high@xmath31 electron identification in atlas at @xmath32 .
these cuts are @xmath33 and @xmath34-dependent and the ones applied to most of the events are illustrated by the dashed red lines in figures [ electrons_1 ] and [ electrons_2 ] : @xmath35 and high to low - threshold trt hit ratio @xmath36 ( indicating the detection of transition radiation produced only by relativistic particles ) .
most of the events in the muon bremsstrahlung sample have small e / p and few high - threshold trt hits ( only muons above 100 gev momentum are expected to produce transition radiation ) .
the events with low e / p and high trt ratio contain a large fraction of muons with combined momentum measured in the muon spectrometer and i d above 100 gev .
only 19 of the 1229 events satisfy the signal criteria . in contrast ,
in the electron ionisation sample , a large fraction of events , 36 out of a total of 85 , satisfy the signal criteria .
these events are interpreted as high- energy @xmath37-rays produced in the inner detector volume by the incoming cosmic muons . in fig .
[ electrons_3 ] , the red curve is the projection of a two dimensional binned maximum likelihood fit to the two dimensional plot in fig .
[ electrons_2 ] , excluding the signal candidates .
the shape of the projected distributions is obtained from the muon bremsstrahlung sample , but the parameters are fitted using the ionisation sample .
this fit is used as one of the methods to estimate the background contamination to the signal .
a clear excess of events with a large ratio of high to low - threshold trt hits is observed , indicating the presence of an electron signal .
, after applying the cut indicated on the ratio of high to low - threshold trt hits .
the red curve shows the background , estimated as explained in fig .
[ electrons_3 ] and projected on the e / p distribution .
the background under the signal is estimated to be ( @xmath38 ) events , using this first method described above for fig .
[ electrons_3].,width=302 ] figure [ electrons_4 ] , shows the distribution of energy to momentum ratio for the electrons candidates . a clear accumulation of signal events around e / p = 1 is observed , as expected for electrons . a second independent method to estimate the background uses the measured ratio of negatively to positively charged muons coupled to the fact that the electron ionisation signal should contain no positrons : - the @xmath39 ratio obtained from the muon bremsstrahlung sample is @xmath40 - out of the 36 signal candidates , four have a positive charge , leading to an expectation of ( @xmath41 ) background events in the signal sample , in good agreement with the estimate from the first method
the final sample consists of the 32 candidates with measured negative charge .
due to space limitations , some systems of atlas have been omitted from this paper .
notably the trigger system which has also been commissioned during this period and is also ready for fist collisions .
in addition , the atlas distributed computing system , with its world - wide computing infrastructure , a crucial component of the analysis and data processing system , has been undergoing continuous commissioning . we are greatly indebted to all cern s departments and to the lhc project for their immense efforts not only in building the lhc , but also for their direct contributions to the construction and installation of the atlas detector and its infrastructure .
we acknowledge equally warmly all our technical colleagues in the collaborating institutions without whom the atlas detector could not have been built .
furthermore we are grateful to all the funding agencies which supported generously the construction and the commissioning of the atlas detector and also provided the computing infrastructure .
we acknowledge the support of anpcyt , argentina ; yerevan physics institute , armenia ; arc and dest , australia ; bundesministerium fr wissenschaft und forschung , austria ; national academy of sciences of azerbaijan ; state committee on science & technologies of the republic of belarus ; cnpq and finep , brazil ; nserc , nrc , and cfi , canada ; cern ; nsfc , china ; ministry of education , youth and sports of the czech republic , ministry of industry and trade of the czech republic , and committee for collaboration of the czech republic with cern ; danish natural science research council ; european commission , through the artemis research training network ; in2p3-cnrs and dapnia - cea , france ; georgian academy of sciences ; bmbf , desy , dfg and mpg , germany ; ministry of education and religion , through the epeaek program pythagoras ii and gsrt , greece ; isf , minerva , gif , dip , and benoziyo center , israel ; infn , italy ; mext , japan ; cnrst , morocco ; fom and nwo , netherlands ; the research council of norway ; ministry of science and higher education , poland ; grices and fct , portugal ; ministry of education and research , romania ; ministry of education and science of the russian federation , russian federal agency of science and innovations , and russian federal agency of atomic energy ; jinr ; ministry of science , serbia ; department of international science and technology cooperation , ministry of education of the slovak republic ; slovenian research agency , ministry of higher education , science and technology , slovenia ; ministerio de educacin y ciencia , spain ; the swedish research council , the knut and alice wallenberg foundation , sweden ; state secretariat for education and science , swiss national science foundation , and cantons of bern and geneva , switzerland ; national science council , taiwan ; taek , turkey ; the science and technology facilities council and the leverhulme trust , united kingdom ; doe and nsf , united states of america .
expected performance of the atlas experiment - detector , trigger and physics , the atlas collaboration , ( submitted on 28 dec 2008 ( v1 ) , last revised 14 aug .
2009 ( v4 ) ) http://arxiv.org/abs/0901.0512v3 [ arxiv:0901.0512v3 ] http://arxiv.org/abs/0901.0512v3
|
the atlas detector at the large hadron collider , cern has been under construction for more than a decade .
it is now largely complete and functional .
this paper will describe the state of the major subsystems of atlas .
results from the brief single beam running period in 2008 will be shown .
in addition , results from a long period of cosmic ray running will be shown .
these results show that atlas is prepared to make major new physics discoveries as soon as we get colliding beams in late 2009 .
|
the macro detector was located in the gran sasso underground laboratory @xcite .
thanks to its large area , fine tracking granularity and up - down symmetry , it was a proper tool for the study of upward - travelling muons and neutrino interactions in the apparatus .
the different kinds of neutrino events detected by macro are shown in fig .
[ topo]a : ( 1 ) upward - throughgoing muons , ( 2 ) semicontained upgoing muons , ( 3 ) upgoing stopping and ( 4 ) semicontained downgoing tracks . the sample ( 1 )
is due to more energetic neutrinos ( @xmath1 ) producing muons also at long distances from the detector .
the other samples are due to neutrinos of lower energy ( @xmath2 ) .
the samples ( 3 ) and ( 4 ) are indistinguishable and therefore they are studied together ( 3 + 4 ) . in fig.s
[ topo]b and [ low ] the angular measurements are shown for samples ( 1 ) , ( 2 ) and ( 3 + 4 ) , respectively .
all distributions are so far from the expectation assuming no oscillation . assuming @xmath3 oscillation the measurements become compatible with expectation @xcite .
the best fit parameters result @xmath4 and @xmath5 .
the measurement of the multiple coulomb scattering @xcite permits to estimate the energy of muons in sample ( 1 ) .
neutrino energies are inferred by means of monte carlo methods . in fig .
[ mcs]a the data / expectation ratio as a function of estimated @xmath6 is shown .
the last point is due to sample ( 2 ) . in fig .
[ mcs]b the 90% c.l . allowed regions assuming @xmath3 oscillation are shown .
the smaller area is estimated by normalization and angular distribution of sample ( 1 ) .
the medium area is due to @xmath7-energy estimate for the same sample .
the larger area is deduced by ( 2 ) and ( 3 + 4 ) low energy samples .
a sample of 1356 upward - travelling muons ( @xmath8 ) has been studied to look for astrophysical signals @xcite .
the data do not show @xmath9-excess from selected pointlike sources .
neutrino - induced muon - flux limits have been established at the level of @xmath10 .
recently the microquasar gx339 - 4 has been proposed as a source of tev neutrinos @xcite .
we note that macro detected 7 events in the direction of gx339 - 4 , while 2.3 atmospheric neutrinos were expected @xcite .
another analysis searched for high - energy upward - travelling muons due to astrophysical high - energy diffuse neutrinos .
no statistically significant signal has been found and an upper limit has been set on diffuse neutrino flux assuming @xmath11 as power law spectrum @xcite .
99 s. ahlen _ et al _ , nucl .
methods a * 324 * , 337 ( 1993 ) ; m. ambrosio _ et al _ , nucl .
methods a * 486 * , 663 ( 2002 ) m. ambrosio _ et al _ , phys .
b * 434 * , 451 ( 1998 ) ; m. ambrosio _ et al _ , phys .
b * 478 * , 5 ( 2000 ) ; m. ambrosio _ et al _ , phys .
b * 517 * , 59 ( 2001 ) m. ambrosio _ et al _ , physics/0203018 , accepted by nucl .
methods a m. ambrosio _ et al _ , astrophys .
j. * 546 * , 1038 ( 2001 ) c. distefano _ et al _ , astro - ph/0202200 m. ambrosio _ et al _ , astro - ph/0203181 , accepted by astroparticle physics
|
macro experiment operated in the gran sasso underground laboratory .
neutrino events collected by this detector are used in order to study the atmospheric neutrino flux .
different measurements in different energy samples are in full agreement and show evidence of neutrino oscillation phenomenon with maximal mixing and @xmath0 . also the search for neutrino astrophysical sources
is reported .
= 14.5pt
|
applications of gravitational lens theory was recently expanded when first paczynsky ( 1986 ) predicted the possibility of detecting dark matter in the galactic halo by searching for microlensing effects on millions of stars in the lmc , and , by now , the four collaborations : the macho group ( for the latest results see the world wide web at : wwwmacho.mcmaster.ca/ or wwwmacho.anu.edu.au/ ) , the eros group ( aubourg et al .
1993 ) , the ogle group ( see www.astro.princeton.edu/ ogle/ or www.astrouw.edu.pl ) , and the duo group ( alard 1995 ) have detected more than 100 microlensing candidates . to distinguish between variable stars and microlensing a number of criteria was used , one of which was that the light curve be symmetric , i.e. only microlensing by a single object was considered . however , as the observations have shown , binary lenses do indeed exist ; notable examples being the duo 2 ( alard et al . 1995 ) , ogle 7 ( udalski et al .
1994 ) , and the lmc binary event ( bennett et al . 1996 ) candidates . a priori , binary lensing is expected in about 10% of lensing events ( mao & paczyski 1991 ) , and mao & di stefano ( 1995 ) developed a method that determines the lensing parameters from a given light curve .
this method was successfully applied to duo 2 ( alard et al .
1995 ) and ogle 7 ( udalski et al .
1994 ) showing that their light curves are fit well by binary lens models ; thereby , raising the question of how many of the objects previously believed to be variable stars could , in fact , be microlensing candidates .
in addition , dominik & hirshfeld ( 1995 ) showed that the light curve of the macho lmc # 1 candidate could be explained better by a binary lens model than a single lens model .
a recent review of the galactic microlensing searches can be found in paczyski ( 1996 ) . in an investigation similar to this one
, della valle & livio ( 1995 ) found that the distribution of the amplitudes relative to the durations of ordinary dwarf nova eruptions is significantly different from those exhibited by the main body of the microlensing candidates from the macho and ogle collaborations .
they concentrated on single lens models which limits the investigation to cataclysmic variables with symmetric light curves .
they concluded that cataclysmic variables with symmetric light curves are not a significant source of contamination , but speculated that the most serious source of potential contamination may be due to old novae that exhibit dwarf nova outbursts . by adopting the current estimates for the space density of such old novae
, they found that a significant fraction of the microlensing candidates could be cataclysmic variables . in this paper ,
the possibility of confusing microlens candidates and cataclysmic variables is investigated further , and , since della valle & livio ( 1995 ) estimated that only 20 - 40% of cataclysmic variables have symmetric light curves , only binary lens models , which can produce asymmetric light curves , are considered .
because of the well sampled light curve and its frequency - amplitude relation , ss cygni is used as a typical cataclysmic variable , and the features of ss cygni is compared to light curves caused by binary microlenses .
since the theory of gravitational microlensing has been presented in detail by numerous authors ( e.g. schneider et al .
1992 or paczyski 1996 ) , only the most basic ideas and notation are included in this section .
assume that a point source is located at an angular diameter distance , @xmath0 , and a binary lens at distance @xmath1 from the observer .
let a light ray from the source pass at a distance @xmath2 from the @xmath3 lens with mass @xmath4 .
define the dimensionless mass of the @xmath3 lens as : @xmath5 and define @xmath6 which makes all angles measured in units of the einstein ring radius , @xmath7 , which is given by : @xmath8 for a lens with unit total mass .
@xmath9 is the angular diameter distance between lens and source .
@xmath7 is a few au for lensing in the lmc or the galactic halo .
the lens equations for a binary lens then becomes : @xmath10 where @xmath11 , @xmath12 , and @xmath13 are the angles of the source , the image of the source , and @xmath3 lens , respectively , as seen by the observer , and @xmath14 is the impact parameter in the observer plane .
the theoretical light curves can then be computed from the lens equations ( [ lens1 ] ) and ( [ lens2 ] ) using the method described in witt ( 1990 ) which simplifies the lens equations into a fifth - order polynomial , which in turn can be solved numerically by laguerre s method ( see mao & di stefano 1995 ) .
as can be seen from the previous section , the light curve of a background point source created by a binary gravitational lens can be characterized by six variables : the ( four ) positions of the two lenses in the observer plane ( ( @xmath15,@xmath16 , ) and ( @xmath17,@xmath18 ) all measured in units of @xmath7 ) , the mass ratio of the lenses ( @xmath19 ) , and the time parameter @xmath20 measured in days .
it is not necessary to solve for the source position , since changing ( @xmath21 ) is equivalent to move the lenses .
therefore , define @xmath22 as 0.0 and let @xmath23 vary between [ -2.5,2.5 ] .
in addition , it is necessary to know the baseline , i.e. the magnitude of the unlensed source , and which fraction of the observed light is caused by nearby unlensed sources . these two parameters can be estimated by other means than including them as free parameters , and they are , therefore , treated as input parameters .
it is not possible to analytically solve for the six parameters for a given light curve , and it is , therefore , necessary to fit the observed and modeled light curves numerically .
this can be done by minimizing @xmath24 given by @xmath25 where @xmath26 is the number of observations , @xmath27 and @xmath28 are the theoretical and observed _ magnifications _ , and @xmath29 is the measurement error . for large numbers of observations ,
@xmath24 is a normal distribution with mean @xmath30 and standard deviation @xmath31 ( see e.g. press et al . , 1992 ) .
due to the nature of microlensing light curves , which can have several sharp peaks appearing and disappearing with only small changes in the parameters , @xmath24 has many local minima . using a simulated light curve ( with @xmath32 for all @xmath33 ) ,
[ param ] shows how @xmath24 changes as a function of one parameter , the coordinate of one of the lenses .
as can be seen from the figure , the curve shows numerous local minima which makes it impossible to use standard numerical methods for minimizing a function ( for instance the downhill - simplex method or simulated annealing , which is even designed to disregard local minima , but fails in this case , see press et al .
1992 ) . looking closely at fig .
[ param ] notice that : as the function gets closer to the global minimum , the local minima get smaller .
it would seems that a `` simulated tunnel effect '' could jump from one local minima to another , since most of them are closely spaced . using these observations ,
a numerical method was developed that first uses the downhill - simplex method to find a ( local ) minima , and then simulates a tunnel effect by randomly searching a six dimensional cube for nearby ( local ) minima that are smaller than the present minima . to overcome features like the double peak at @xmath34 and 0.79 in fig .
[ param ] , the cube is gradually expanded as long as no smaller minima is found .
when a new ( and smaller ) minima is found , the whole process repeats itself .
the procedure is terminated when the cube reaches a specified maximum size . when the cube is expanded sufficiently slow and
far , this method reliably finds the global minimum using simulated light curves .
however , it turns out to be time consuming if the initial guess of the six parameters is far from the global minimum . to increase the efficiency of the program , the initial guesses are determined by defining a grid in the 6 dimensional space , finding the ( local ) minimum for each initial grid point , and
then use the parameters of the best minimum as initial guesses ( which , depending on how fine the grid is , may well be the global minimum ) .
( a)(b ) as a test of the program , it was applied to the observations of duo 2 ( alard et al .
1995 ) in the b band , which has the most complex light curve of the microlensing candidates so far , and , therefore , should be the most difficult to fit .
assume a 30% blend from an unlensed source and adopt a baseline of @xmath35 mag from the unlensed part of the light curve .
the first attempts at fitting the observations showed that the quoted measurements errors were overestimated , and the errors were instead obtained from the scatter in the unlensed part of the curve . in this case , the best fit has a @xmath24 of 78 for 80 data points ( 74 degrees of freedom ) . using the same notation as above ,
the best fit parameters are : @xmath36 from which the time of closest approach , @xmath37 , can be calculated : @xmath38 j.d .
- 2440000 or with the notation from alard et al.(1995 ) : @xmath39 j.d . - 2449536.64861 . for direct comparison , the parameters used by alard et al .
( 1995 ) are introduced : the distance between the two lenses , @xmath40 ( in units of @xmath7 ) , the closest approach to the center of mass , @xmath41 ( in units of @xmath7 ) , and the angle between the binary and the trajectory , @xmath42 , this model yields : @xmath43 the observations and the best fit model are shown in fig .
this result is remarkably close to the solution found by alard et al.(1995 ) .
the best fit parameters only deviate a few percent from each other which is surprising since there s neither no guarantee that the global minimum is found nor is the solution unique .
in addition , alard et al .
( 1995 ) fitted both the r and b band data , while only the b band data was considered here and included a finite source size .
it should also be noted that alard et al .
( 1995 ) used a different minimizing method , namely levenberg - marquardt s method .
the downhill - simplex method was used here because it is more robust ( but slower ) . notice that the method , described in this paper , does not rely on any human decisions once the fitting parameters ( maximum size of cube etc . )
are determined ; the same program can be applied to any light curve without changes or decisions on the best guesses of the parameters . in particular ,
alard et al .
( 1995 ) had to estimate @xmath37 , and @xmath44 as an intermediate step . by fitting all six parameters simultaneously ,
this was avoided here .
assume an unusual ( and unlikely ) scenario : let the cataclysmic variable exhibit one event during the observing run , let that run be shorter than normal for microlensing searches , say one month , and assume that subsequent runs happens to be in between bursts .
this will give a worst case scenario for contamination of binary lens candidates .
therefore , only cataclysmic variables with an outburst frequency larger than one month are considered in the following .
the relation between amplitude and frequency is given by the kukarkin - parengo relation : @xmath45 + ( 1.90 \ , [ \pm 0.22 ] ) \
, { \rm log}(t ) \label{kuk}\ ] ] where @xmath46 is the amplitude of the burst in magnitudes and @xmath47 is the elapsed time between bursts in days ( see e.g. warner 1995 ) . a one month frequency then corresponds to an amplitude of about 3.5 magnitudes which again corresponds to a magnification of a factor of 25 .
notice that , according to eq .
( [ kuk ] ) , a low amplitude variable would exhibit several events during the same observing run , and would , therefore , immediately be classified as a variable star . based on the above ,
a good example of a cataclysmic variable is then ss cygni , a dwarf nova of u geminorum type , with an average outburst frequency of 50 days and an amplitude corresponding to a magnification of a factor of 30 .
the american association of variable star observers combined observations from numerous people ( see mattei et al .
1991 , which can be found at www.aavso.org/monographs.html , for details and acknowledgments ) thereby creating a well sampled light curve of ss cygni over a long period of time . a small part of this light curve is shown in fig . [ sscyg ]
as we have seen in the previous sections , only high magnification events ( defined here conservatively as more than a factor of 10 ) are of interest . according to schneider et al .
( 1992 ) , the probability that the magnification of a point source is larger than a certain value , @xmath48 , is proportional to @xmath49 independent of lens model . since only single peak light curves have any interest , @xmath50 250,000 light curves and their peak magnification using binary lens models were calculated .
the resulting histogram is shown in fig .
as can be seen from the figure , of the total 250,000 light curves only @xmath50 3% are high magnification events with one peak .
the possibility of mistaking cataclysmic variables as binary lens candidates is , therefore , exceedingly small ; out of the total of @xmath50100 microlensing candidates , one would expect @xmath5010% to show binary lens signatures ( mao & paczyski 1991 ) , and only 3% of these would exhibit light curves similar to cataclysmic variables .
( a)(b ) however , if the light curve is well sampled , it is actually impossible to confuse the two phenomena .
[ shape ] shows the shape of one event in ss cygni and a light curve of a high magnification event on the same scale .
as can be seen , the two light curves are incompatible with each other , especially in the wings . to quantify this statement , the probability of a magnification interval @xmath51 , was calculated from all events in mattei et al .
( 1991 ) and all models fitting the more than a factor ten , one peak criteria . to avoid numerical artifacts from the large scatter in the ss cygni observations ,
@xmath51 was set equal to a magnification of a factor of 2.5 ( corresponding to 1.0 magnitudes ) .
the resulting histogram is shown in fig .
[ hist ] , which clearly shows that the shapes of the light curves are incompatible with each other . as a final test , one event from ss cygni s light curve
was fitted with a binary lens model .
ss cygni s light curve was smoothed by averaging over every four points thereby getting estimates of the measurement errors and reducing computation time by a factor of four .
not surprisingly , the `` best '' fit has a @xmath24 of 9,550 with 50 degrees of freedom , which clearly shows that the event in ss cygni can not be modeled by binary lenses . as can be seen from fig .
[ cygfit ] , the modeled light curve not only has more than one peak , but show no resemblance to the observations whatsoever .
it is again concluded that cataclysmic variables can not be confused with binary microlens candidates .
it has been demonstrated that the developed numerical method reliably finds the global minimum in all the simulated data .
the program also determines a best fit model to the binary lens candidate duo 2 which is compatible with the models published in alard et al .
( 1995 ) .
ss cygni was chosen as an example of a typical cataclysmic variable star , and it was demonstrated that not only is it highly unlikely that the observed events can be explained by binary microlens models , but , due to the difference between the shapes of the observed events and the models , the two phenomena are incompatible with each other when the light curve is well sampled .
the attempt to fit one of the observed events with a binary lens model was , as expected , a complete failure .
i would like to thank b. paczyski for many helpful ideas , discussions , and comments on this paper .
i would also like to thank s. mao for letting me use the duo 2 observations , for helpful discussions on the numerical method , and for supplying me with a program that generates the light curves from eqs .
( 3 ) and ( 4 ) . in this research ,
i have used and acknowledge with thanks data from the aavso international database , based on observations submitted to the aavso by variable star observers worldwide .
this work was supported by nsf grant ast 93 - 13620 .
alard , c. : 1995 , in `` proc .
iau symp . 173 , '' eds . kochanek , c. s. , hewitt , j. n. , in press ( kluwer academic publishers ) , 215 alard , c. , mao , s. , guibert , j. : 1995 , aa 300 l17 aubourg , e. , bareyre , p. , brhin , s. , gros , m. , lachlze - rey , m. laurent , b. , lesquoy , e. , magneville , c. , milsztajn , a. , moscoso , l. , queinnec , f. , rich , j. , spiro , m. , vigroux , l. , zylberajch , s. , ansari , r. , cavalier , f. moniez , m. , beaulieu , j .p . ,
ferlet , r. , grison , p. , vidal - madjar , a. , maurice , e. , prvt , l. , gry , c. , guibert , j. , moreau , o. , tajahmady , f. : 1993 , nature 365 , 623 bennett , d. p. , alcock , c. , allsman , r. a. , alves , d. , axelrod , t.s
. , becker , a. , cook , k. h. , freeman , k. c. , griest , k. , guern , j. , lehner , m. j. , marshall , s. l. , minniti , d. , peterson , b. a. , pratt , m. r. , quinn , p. j. , rhie , s. h. , rodgers , a. w. , stubbs , c. w. , sutherland , w. , welch , d. : 1996 , preprint , astro - ph/9606012 della valle , m. , livio , m. : 1996 , apj 457 , l77 dominik , m. , hirshfeld , a. c. : 1995 , preprint mao , s. , di stefano , r. : 1995 , apj 440 , 22 mao , s. , paczyski , b. : 1991 , apj 374 , l37 mattei , j. a. , waagen , e. o. , foster , e. g. : 1991 , aavso monograph 1 , supplement 1 , ss cygni light curves 1985 - 1990 , aavso , cambridge , ma paczyski , b. : 1986 , apj 304 , 1 paczyski , b. : 1996 , preprint , astro - ph/9604011 press , w. h. , flannery , b. p. , teukolsky , s. a. , vetterling , w. t. : 1992 , `` numerical recipes '' ( cambridge university press ) schneider , p. , ehlers , j. , falco , e. e. : 1992 , `` gravitational lenses '' ( springer - verlag , berlin ) udalski , a. , szymaski , m. , mao , s. , di stefano , r. , kaluzny , j. , kubiak , m. , mateo , m. , krzemiski , w. : 1994 , apj 436 , l103 warner , b. : 1995 , `` cataclysmic variable stars '' ( cambridge university press ) witt , h. j. : 1990 , aa 236 , 311
|
a common remark at conferences and meetings has been : `` one can fit _ any _ light curve with binary lens models because of the large number of parameters ! ''
it is especially tempting to claim that some of the possible binary events in the galactic microlensing searches could be caused by cataclysmic variables since both phenomena exhibits asymmetric light curves .
this paper presents a statistical analysis comparing the light curves of cataclysmic variable stars to the light curves generated from binary lens models .
it is shown that the light curves of high amplitude cataclysmic variables are one example of light curves that can _ not _ be explained with binary lens models .
it is , therefore , not only unlikely , but also impossible that cataclysmic variables can contaminate the binary microlens candidates . 2 ^ 2
|
let @xmath6 be the algebra of all @xmath7 complex matrices .
the schur product @xmath8 of two matrices @xmath9 and @xmath10 in @xmath6 is defined by @xmath11 and given @xmath12 , the schur map @xmath13 on @xmath6 is defined by @xmath14 for all @xmath15 .
although a schur map certainly appears unlikely to be a homomorphism with respect to ordinary matrix multiplication , one can verify that , for example:@xmath16 is both a schur map and a @xmath17-homomorphism . in this article
, we develop several characterizations of multiplicative schur maps . in the infinite - dimensional case
, there is a natural notion of a schur multiplier of a schatten ideal of @xmath18 if the hilbert space is taken to be finite - dimensional , a schur multiplier is a schur map . in our discussion , @xmath19 denotes the @xmath7 matrix with all entries equal to @xmath20 , and @xmath21 is the usual identity matrix .
we use the notation @xmath22 @xmath23 for the standard basis of @xmath6 or their infinite
dimensional analogues , as in section @xmath24 .
if a has no zero entries then by @xmath25}$ ] we mean the matrix whose entries are reciprocals of the entries of @xmath26 .
by @xmath27 we mean the set of all eigenvalues of the matrix @xmath28 the notation @xmath29 stands for the conjugate transpose of @xmath26 . by a positive matrix ,
we always mean a hermitian matrix with nonnegative spectrum . throughout this paper ,
we generally exclude the zero schur map from the set of schur maps in order to avoid some trivial cases in the statements of our results . in section @xmath30
, we characterize multiplicative schur maps over @xmath6 and it turns out that in the case of positive schur maps , the entries of the corresponding matrix are nothing but the extreme points of the closed unit disk , thus , are necessarily scalars of modulus 1 . in section @xmath31 , we enumerate all positive schur maps over @xmath5 . in section @xmath24 , we try to generalize our results to the infinite dimensional setting .
finally , in section @xmath32 we explore the relation of schur maps to the extreme points of certain sets .
a map @xmath33 from @xmath6 to @xmath6 is called _ positive _ if it maps the positive matrices in @xmath6 to positive matrices in @xmath6 . [ main ]
let @xmath34 be a nonzero schur map
. then the following are equivalent : 1 .
@xmath13 is multiplicative , 2 .
@xmath13 is spectrum preserving , 3 .
@xmath35 where @xmath36 has multiplicity @xmath37 and the diagonal entries of @xmath26 are equal to @xmath20 , 4 .
@xmath26 is of rank one with diagonal entries equal to @xmath20 , and 5 .
@xmath38 = @xmath39 @xmath40 and @xmath41 for all @xmath42 .
@xmath43 if @xmath13 is multiplicative we claim that it is bijective and that the inverse is a schur map . in other words
, @xmath26 has a schur inverse or equivalently @xmath44 for all @xmath45 . to prove the claim , note that if there exist @xmath46 and @xmath47 such that @xmath48 then @xmath49 which implies that the kernel of the multiplicative map @xmath50 is nonempty . since @xmath6 is simple
, this would imply that @xmath50 is in fact the zero map , which by hypothesis is not the case .
thus , the kernel is @xmath51 , and @xmath13 is an automorphism , but any automorphism over @xmath6 is inner . in other words , there exists an invertible matrix @xmath52 such that @xmath53 for all @xmath15 .
thus , @xmath13 is given by a similarity transformation , implying that the spectrum of @xmath54 is equal to the spectrum of @xmath55 @xmath56 : since @xmath13 is spectrum preserving , we can conclude that @xmath57 the spectrum of @xmath58 hence of @xmath59 is @xmath60 the kernel of @xmath13 is @xmath51 because @xmath50 is a linear and spectrum - preserving map ( see ( * ? ? ?
2.1.i ) . ) in finite dimensions , injectivity implies surjectivity , so that @xmath50 is hence surjective .
being spectrum - preserving , @xmath13 preserves the trace , but the equation @xmath61 means that @xmath62 for all choices of sequence @xmath63 and this can only occur if @xmath41 for all @xmath64 .
hence , @xmath65 and since the spectrum of @xmath26 is @xmath66 we deduce that the algebraic multiplicities of @xmath36 and @xmath67 in the spectrum are respectively @xmath37 and @xmath20 .
@xmath68 since rank is the number of nonzero eigenvalues , with multiplicity , we have @xmath69 .
@xmath70 since @xmath69 , we conclude that the @xmath71 column of @xmath26 is a scalar multiple of the @xmath72 column , that is , @xmath73 but for @xmath74 we have @xmath75
. therefore , @xmath38 = @xmath39 @xmath40 and @xmath41 for all @xmath42 .
@xmath76 for arbitrary matrices @xmath77 and @xmath52 , it follows that @xmath78 @xmath79 @xmath80 for all @xmath81 .
matrices of positive elements satisfying the condition @xmath38
= @xmath39 @xmath40 that appears in condition @xmath82 of theorem [ main ] are in fact the so - called consistent matrices that are used in one branch of mathematical economics ( see @xcite . )
the requirement of having unit elements on the diagonal is then equivalent to the condition that the matrix is what is called a reciprocal matrix .
one could say that theorem [ main ] characterizes multiplicative schur maps in terms of a generalization of the consistent reciprocal matrices .
an unital schur map @xmath83 is a homomorphism with respect to the usual matrix product if and only if the coefficients of @xmath26 are of the form @xmath1 where @xmath84 is a vector of nonzero complex numbers .
moreover , if @xmath50 is a homomorphism , it is constructed as a similarity transformation by a diagonal matrix.[cor : diag ] if @xmath50 is a homomorphism , then by the above theorem @xmath38 = @xmath39 @xmath40 where @xmath41 and all the @xmath38 are nonzero , so that @xmath85 we can set @xmath86 conversely , if @xmath1 then it follows that for any matrix @xmath77 : @xmath87 where the operation on the right is ordinary matrix multiplication .
but , then @xmath50 is a homomorphism with respect to matrix multiplication .
[ norm ] if @xmath34 is multiplicative then the following hold : 1 .
@xmath88}}$ ] and @xmath26 is diagonalizable .
2 . in general , @xmath89 and if @xmath90 then @xmath91 .
3 . if @xmath92 then @xmath13 is numerical range preserving .
@xmath93 by theorem , @xmath94 which implies @xmath95 for all @xmath45 .
since the polynomial @xmath96 has distinct roots and annihilates @xmath26 , it follows that the minimal polynomial of @xmath26 has distinct roots , and thus @xmath26 is diagonalizable .
@xmath97 since @xmath50 is multiplicative , theorem [ main ] shows that the spectrum of @xmath26 is @xmath98 so the spectral radius @xmath99 of @xmath26 is @xmath100 therefore @xmath101 if @xmath102 , then @xmath103 .
@xmath104 as in the proof of corollary [ cor : diag ] , we have that @xmath105 where the diagonal entries of the diagonal matrix @xmath106 come from the first row of @xmath28 but by part @xmath46 and the fact that @xmath107 we conclude that @xmath108 .
this property carries over to the entries of @xmath106 so that @xmath109 but then @xmath110 , making @xmath106 a unitary .
we thus have @xmath111 as claimed .
by @xcite it follows that @xmath13 is numerical range preserving .
[ mainp ] let @xmath112 be an unital schur map
. then the following are equivalent : 1 .
@xmath13 is multiplicative and @xmath4-preserving , 2 .
@xmath13 is a completely positive isomorphism , 3 .
@xmath26 is of rank one and normal with diagonal entries equal to @xmath20 , 4 .
@xmath26 is of rank one with unimodular entries and diagonal entries equal to @xmath20 , and 5 .
@xmath26 is self - adjoint with @xmath35 where @xmath36 has multiplicity @xmath37 and @xmath113 , 6 .
@xmath26 and @xmath25}$ ] are positive matrices with diagonal entries equal to @xmath20 .
@xmath43 since @xmath13 is @xmath4-preserving , @xmath26 is self - adjoint and by theorem it has positive spectrum .
therefore , @xmath26 is positive , and by ( * ? ? ?
* theorem 3.7 ) the map @xmath50 is completely positive . on the other hand , since by theorem the map @xmath13 has no kernel it is , by finite dimensionality , an invertible operator .
positive , unital , multiplicative , and invertible maps of matrix algebras have inverses that are positive , unital , and multiplicative .
clearly @xmath114}}$ ] is an inverse for @xmath50 . arguing as previously , @xmath114}}$ ] is completely positive . therefore , @xmath13 is a completely positive isomorphism .
@xmath115 if @xmath97 holds then @xmath50 is a @xmath17-isomorphism of matrix algebras .
the map @xmath50 thus maps a projection of rank @xmath116 to a projection of rank @xmath116 . applying @xmath50 to the orthogonal projection of rank 1 given by @xmath117 we find that @xmath118 is a projection of rank 1 , and is positive .
thus @xmath26 is positive , hence normal , and of rank 1 as a matrix .
the matrix @xmath26 has diagonal elements @xmath20 since @xmath50 is unital .
@xmath68 since @xmath26 is normal , it has a decomposition of the form @xmath119 where @xmath120 is a unitary matrix and @xmath121 where @xmath122 are the eigenvalues of @xmath26 with the decreasing order .
since @xmath26 has rank one with diagonal entries equal to @xmath20 we can conclude by theorem that @xmath123 . therefore , @xmath124 but @xmath125 unless @xmath126 where @xmath127 .
hence , @xmath128 which implies @xmath129 for all @xmath130 .
thus , @xmath131 for all @xmath45 .
@xmath70 since @xmath26 is of rank one with diagonal entries equal to @xmath20 , theorem implies that @xmath35 where @xmath36 has multiplicity @xmath37 . if @xmath26 has unimodular entries then @xmath132 for all @xmath45 . by repeated use of theorem
we can conclude that @xmath133 for all @xmath45 .
hence , @xmath134 for all @xmath45 .
thus , @xmath26 is normal with positive spectrum and the spectral theorem for normal matrices implies that @xmath26 is positive and it is known that for positive matrices @xmath135
@xmath136 it is clear that @xmath26 is positive .
therefore , by corollary we conclude that @xmath137}}$ ] which implies @xmath138})=\spec(a)=\{0,n\}$ ] .
thus , @xmath25}$ ] is also positive . since @xmath26 is positive , @xmath139 . on the other hand , @xmath140 .
therefore , the diagonal entries of @xmath26 are equal to @xmath20 .
@xmath141 if @xmath26 and @xmath25}$ ] are positive with diagonal entries equal to @xmath20 then by ( * ? ? ?
* theorem 3.7 ) we can conclude that @xmath13 and @xmath114}}$ ] are completely positive and unital . denoting these maps by @xmath142 and @xmath143 respectively , by the schwarz inequality ( * ? ? ?
* corollary 2.8 ) we have @xmath144 . by using the schwarz inequality
once more for @xmath143 we have @xmath145 . hence , @xmath146
therefore , @xmath147 .
let us define @xmath148 by @xmath149 if we identify @xmath150 with 2-by-2 matrices of operators ( block matrices ) we can write @xmath151 now , if we apply the schwarz inequality to the positive map @xmath152 and the element @xmath153 we get @xmath154 thus we conclude that the following matrix is operator positive : @xmath155 but we have already shown using the schwartz inequality that the entries on the diagonal of the above matrix happen to be zero , so then we have that @xmath156 is operator positive .
let @xmath157 and note that an operator matrix of the form @xmath158 where @xmath159 is a non - negative scalar , is operator positive if and only if @xmath160 .
this can be deduced from , for example , ( * ? ? ?
* lemma 3.1 ) .
but , then it follows that @xmath161 is zero .
[ idempotent]let @xmath162 be a multiplicative schur map .
the following are equivalent : 1 .
the map is @xmath17-preserving , 2 .
the matrix @xmath26 has norm @xmath163 3 .
the matrix @xmath164 is an orthogonal projection , and 4 .
the operator norm of @xmath50 is 1 .
@xmath43 if @xmath50 is @xmath17-preserving then @xmath26 is self - adjoint and by corollary , we have @xmath165 @xmath115 recall that by applying @xmath50 to the equation @xmath166 we find that @xmath167 is a not necessarily orthogonal projection . by ( * ? ? ?
* theorem 5.4 ) , a projection @xmath168 is an orthogonal projection if and only if it has operator norm @xmath169 thus , the norm condition @xmath91 holds if and only if @xmath170 and hence @xmath59 is positive .
@xmath171 if @xmath118 is a projection then @xmath26 is self - adjoint and consequently @xmath50 is @xmath17-preserving . for the equivalence of @xmath46 and
@xmath172 we compute @xmath173 with respect to the norm induced by the operator norm on the range and domain .
since @xmath50 is multiplicative , we note that @xmath174 by corollary 2.2 , and we have @xmath175 this bound is in fact achieved if we take @xmath77 equal to a suitable matrix unit .
thus , @xmath176 and this is equal to 1 if and only if the @xmath177 all have modulus 1 .
but by theorem 2.4 this occurs if and only if the map @xmath50 is @xmath4-preserving .
in this section , @xmath178 denotes a field of characteristic zero .
although matrices can very rarely be diagonalized over a general field , we do still have that @xmath179 is simple and that all automorphisms are inner ( as a consequence of the noether - skolem theorem ) .
moreover , it is still true that the matrix units @xmath180 satisfy @xmath181 and from these facts it follows that : let @xmath182 be a field .
let @xmath183 be a schur map
. then the following are equivalent : 1 .
@xmath50 is a nonzero homomorphism , 2 .
@xmath184 for some invertible diagonal matrix @xmath185 and 3 .
@xmath38 = @xmath39 @xmath40 and @xmath41 for all @xmath42 .
we define generalized toeplitz matrices by @xmath186 .
we define @xmath187 to be the set of all @xmath188 such that @xmath189 is multiplicative .
if we equip @xmath187 with the schur product then @xmath187 turns into an abelian group and @xmath190 is a subgroup , consisting of toeplitz matrices .
it is clear that @xmath187 is an abelian group with respect to schur product because @xmath191 and if @xmath192 then @xmath25}\in l^{n}_{\mathbb{f}}$ ] because the inverse of a homomorphism is also a homomorphism . since @xmath193 , @xmath190 consist of toeplitz matrices and the relations
@xmath194 imply that it is a subgroup of @xmath187 .
we remark that for @xmath195 , we have @xmath196 . in general , returning to the case of the real or the complex field , the problem of determining all the coefficients of the matrix of a multiplicative schur map given partial information on the coefficients leads to a gauss - jordan problem in abstract linear algebra . transforming the equations of theorem by @xmath197 we obtain the following linear system @xmath198 over @xmath199
if we want the schur map to be multiplicative and @xmath17-preserving , then , by theorem , the equations for the transformed coefficients become : @xmath200 over @xmath201 in general we are free to specify @xmath37 of the unknowns , and in the case where we specify one row or column of the matrix @xmath202 $ ] , it is straightforward to solve for the remaining unknowns
. let @xmath203 be the circle group , then it is easy to check that the map @xmath33 from @xmath204 to @xmath205 which sends the vector @xmath206 to the first row of the matrix @xmath26 ( note that the first entry of the matrix is always @xmath20 ) is in fact an isomorphism between topological groups . from this discussion , denoting by @xmath205 the positive matrices within @xmath207 : the lie group @xmath207 has complex dimension @xmath208 and the compact lie group @xmath205 is isomorphic to the @xmath209torus .
if we are interested in schur maps that are multiplicative , @xmath17-preserving , and have only real entries in their matrix of coefficients , then all the entries must , by theorem , be in @xmath210 .
then , for example , for @xmath211 where @xmath212 is the subgroup of positive matrices within @xmath213 the only possibilities are : @xmath214 these are all of the 3-by-3 positive and real schur multiplicative matrices . for the 2-by-2 complex case
the only possibility is : @xmath215 the above matrix has operator norm @xmath216 which is consistent with proposition . for the 3-by-3 complex case
, we have a nice expression for the general form of a matrix for a multiplicative schur map , as a schur factorization : @xmath217 where @xmath218 denotes the schur product .
the discrete group @xmath219 has cardinality @xmath220 . if @xmath221 then by theorem we have either @xmath222 or @xmath223 for all @xmath45 . on the other hand , since @xmath26 is of rank one and @xmath224=@xmath20 , @xmath26 can be completely determined by exactly @xmath37 of its first row s entries . in other words
, we have @xmath37 entries and two choices for each entry .
each choice gives us a distinct schur map , and therefore , @xmath225 .
if we consider only schur matrices with positive real entries , then the only such schur map that is multiplicative and @xmath4-preserving is the identity map @xmath226 .
in this section , @xmath227 denotes the separable infinite dimensional hilbert space and @xmath228 denotes the algebra of compact operators over @xmath227 . if we fix , once and for all , an orthonormal basis for our hilbert space , we can then identify bounded operators on @xmath229 denoted by @xmath230 with their ( infinite ) matrix representation with respect to that basis
. we may then consider schur multipliers with respect to various classes of linear operators in @xmath18 the schatten
von neumann classes , @xmath231 , are algebraic ideals within @xmath232 and are defined by @xmath233 where @xmath234 is the canonical trace on @xmath18 the schatten @xmath168-norm , which makes @xmath231 a banach space within @xmath228 , is defined by @xmath235 in particular , the schur multipliers of @xmath236 are precisely the matrices whose coefficients are uniformly bounded .
since the schatten classes are ideals , we may take products of elements of a schatten class @xmath231 and obtain elements of @xmath231 .
( for @xmath237 more is true : a product of elements in @xmath238 is also in @xmath239 ) a schur multiplier @xmath240 with no kernel satisfies @xmath241 if and only if @xmath242 where @xmath243 is a bounded sequence that is bounded away from zero . if @xmath242 then given two operators @xmath244 that are of finite rank with respect to the chosen basis , we choose a matrix subalgebra @xmath6 within @xmath228 that is a corner of our space of infinite matrices , and contains both @xmath245 and @xmath246 we then have that @xmath241 for @xmath245 and @xmath247 of finite rank .
but since @xmath248 is a bounded operator , we then have by approximation that the equation @xmath241 holds for all @xmath245 and @xmath247 in @xmath238 . for the converse ,
if @xmath248 is multiplicative , then for every finite - dimensional matrix corner @xmath6 the map @xmath50 restricts to @xmath249 the restricted maps are homomorphisms with no kernel , and thus by finite - dimensionality of @xmath6 they are isomorphisms . in particular , then , since the unit of an algebra is unique , the restricted maps @xmath50 must take the unit of @xmath6 to the unit of @xmath250 our results for the finite - dimensional case then imply that @xmath1 with @xmath86 taking larger and larger corners , we conclude that for all @xmath46 and @xmath47 we have @xmath1 with @xmath86 if either @xmath84 were not a bounded sequence , or @xmath251 were not a bounded sequence , then we could find an unbounded sequence of coefficients of @xmath59 and @xmath50 would not be a schur multiplier of @xmath252 we next consider the case of schur multipliers of the compact operators .
let @xmath26 be an infinite matrix in @xmath253 and @xmath228 be the algebra of compact operators on @xmath254 represented by formal infinite matrices .
then @xmath13 is a bounded operator from @xmath228 to @xmath228 .
it is sufficient to consider the case of elementary compact operators @xmath255 .
since @xmath256 , we have @xmath257 where @xmath258 .
hence , @xmath13 is continuous and since it maps finite rank operators to finite rank operators , it is a map from @xmath228 to @xmath228 .
let @xmath259 be a schur map where a is in @xmath253
. then @xmath13 is multiplicative if and only if the columns of @xmath26 are scalar multiples of the first column , and moreover there are 1 s on the diagonal
. if @xmath13 is multiplicative then @xmath260 for all @xmath45 because if there exist a @xmath261 and @xmath67 in @xmath262 such that @xmath263=@xmath36 then @xmath264=@xmath36 which implies @xmath265 but this contradicts the fact that @xmath228 is simple .
since @xmath13 is multiplicative we have @xmath266=@xmath267= @xmath268 which implies @xmath38=@xmath269 . if we let @xmath270 we have @xmath271 and therefore @xmath41 . for @xmath272 ,
the above equations imply that @xmath273 if we let @xmath274 then @xmath26= [ @xmath275 , @xmath276 , @xmath277 , @xmath278 , ... ] . in other words , by abuse of terminology , we have @xmath69 .
conversely if @xmath69 in this sense then we have @xmath279 since a is unit diagonal , if we let @xmath74 then we have @xmath280 and consequently we recover the equations @xmath281 . hence , @xmath282 . therefore ,
@xmath283 .
let @xmath259 be @xmath4-preserving where a is in @xmath253 .
if @xmath13 is multiplicative then @xmath284=@xmath20 and @xmath285=@xmath20 . by the above theorem
, @xmath26 satisfies the equations @xmath20=@xmath285=@xmath286 .
therefore , @xmath26 has unimodular entries .
let @xmath26 be an infinite matrix in @xmath287 .
if @xmath26 has 1 s on the diagonal and linearly dependent columns then a is not bounded in the operator norm . choosing an arbitrary @xmath288 , consider the @xmath67-by-@xmath67 top right corner of @xmath26 and denote it by @xmath289 . by theorem
, we can find an eigenvector @xmath290 such that @xmath291 .
construct the vector @xmath292 such that @xmath293 for all @xmath294 , with the other entries set to zero .
we can multiply @xmath295 by a scalar such that the norm of @xmath295 is 1 .
but , @xmath296 thus , we have a sequence of elements of the unit ball upon which @xmath26 is unbounded , or in other words , @xmath26 is not bounded in the operator norm .
a linear completely positive map between von neumann algebras is called normal if it is continuous with respect to the @xmath297-weak topologies on domain and range .
a markov map is a linear normal trace preserving unital completely positive map from @xmath6 to @xmath6 and we use the notation @xmath298 for the set of all markov maps . a positive matrix with unit diagonal entries
is called a correlation matrix and the set of all correlation matrices is denoted with @xmath299 .
if @xmath34 is @xmath4-preserving and multiplicative then by theorem , the entries of @xmath26 are the extreme points of the closed unit disk . in addition
, @xmath26 is a rank one correlation matrix and consequently by ( * ? ? ?
* theorem 1 ) it is an extreme point of @xmath299 .
however , this can also be proved by the above proposition .
note that by ( * ? ? ?
* theorem 3.5 ) , @xmath13 is an extreme point of the set of all positive unital schur maps . in other words
, it is an extreme markov map and the above proposition implies that @xmath26 is an extreme correlation matrix .
let @xmath300 be a schur map with no kernel , where @xmath26 is the matrix of an operator in the unit ball @xmath301 of @xmath232 .
if @xmath13 belongs to an extremal ray of the set of schur maps , then @xmath26 is either a scalar multiple of an isometry or a scalar multiple of a co - isometry . by considering the unit ball
, we may work with extreme points rather than extremal rays .
suppose @xmath13 is extreme and @xmath302 where @xmath303 and @xmath304 are matrices of operators in the unit ball @xmath301 . by the properties of schur maps , @xmath305 and @xmath306 are in the unit ball of the space of all schur maps from @xmath232 to @xmath307 ) and @xmath308 .
since @xmath13 is extreme , @xmath309 . therefore , @xmath310 is the zero map .
we claim that @xmath311 because if @xmath312 then it has a nonzero entry such as @xmath313 and by applying @xmath310 to @xmath314 we have @xmath315 which is a contradiction .
therefore @xmath26 is an extreme point of the unit ball of @xmath232 , and for @xmath232 these are the isometries and co - isometries .
|
we give necessary and sufficient conditions for a schur map to be a homomorphism , with some generalizations to the infinite - dimensional case . in the finite - dimensional case , we find that a schur multiplier @xmath0\colon m_{n } ( \mathbb{c } ) \rightarrow m_{n } ( \mathbb{c } ) $ ] distributes over matrix multiplication if and only if @xmath1 for some @xmath2 where @xmath3 is the set of nonzero complex numbers .
in addition , it is shown that it is possible to enumerate all @xmath4-preserving multiplicative schur maps on @xmath5 .
we also study the relation of schur map to the extreme points of certain sets .
|
the southern planetary nebula ( pn ) suwt 2 ( pn g311.0 + 02.4 ) is a particularly exotic object .
it appears as an elliptical ring - like nebula with much fainter bipolar lobes extending perpendicularly to the ring , and with what appears to be an obvious , bright central star .
the inside of the ring is apparently empty , but brighter than the nebula s immediate surroundings .
an overall view of this ring - shaped structure and its surrounding environment can be seen in the h@xmath2 image available from the supercosmos h@xmath2 sky survey ( shs ; * ? ? ?
@xcite classified suwt 2 as of intermediate excitation class ( ec ; @xmath3@xmath4 ; * ? ? ?
* ) based on the strength of the heii @xmath54686 and [ oii ] @xmath53728 doublet lines .
the line ratio of [ n ii ] @xmath56584 and h@xmath2 illustrated by @xcite showed a nitrogen - rich nebula that most likely originated from post - main - sequence mass - loss of an intermediate - mass progenitor star . over a decade ago , @xcite discovered that the apparent central star of suwt 2 ( nsv 19992 ) is a detached double - lined eclipsing binary consisting of two early a - type stars of nearly identical type .
furthermore , @xcite suggested that this is potentially a triple system consisting of the two a - type stars and a hot , unseen pn central star .
however , to date , optical and uv studies have failed to find any signature of the nebula s true ionizing source ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
hence the putative hot ( pre-)white dwarf would have to be in a wider orbit around the close eclipsing pair .
@xcite recently derived a period of 4.91 d from time series photometry and spectroscopy of the eclipsing pair , and concluded that the centre - of - mass velocity of the central binary varies with time , based on different systemic velocities measured over the period from 1995 to 2001 .
this suggests the presence of an unseen third orbiting body , which they concluded is a white dwarf of @xmath6 , and is the source of ionizing radiation for the pn shell .
there is also a very bright b1ib star , sao 241302 ( hd 121228 ) , located 73 arcsec northeast of the nebula .
@xcite speculated that this star is the ionizing source for suwt 2 .
however , the relative strength of heii@xmath54686 in our spectra ( see later ) shows that the ionizing star must be very hot , @xmath7 @xmath8 100,000k , so the b1 star is definitively ruled out as the ionizing source .
narrow - band h@xmath2+[nii ] and [ oiii]5007 images of suwt 2 obtained by @xcite show that the angular dimensions of the bright elliptical ring are about @xmath9arcsec@xmath10@xmath11arcsec at the 10% of maximum surface brightness isophote @xcite , and are used throughout this paper .
@xcite used the mosaic2 camera on the cerro tololo inter - american observatory ( ctio ) 4-m telescope to obtain a more detailed h@xmath2+[nii ] image , which hints that the ring is possibly the inner edge of a swept - up disc .
the [ nii ] image also shows the bright ring structure and much fainter bipolar lobes extending perpendicular to the ring plane .
we can see similar structure in the images taken by bond and exter in 1995 with the ctio 1.5 m telescope using an h@xmath2+[nii ] filter .
fig.[suwt2:fig1 ] shows both narrow - band [ nii ] 6584 and h@xmath2 images taken in 1995 with the eso 3.6 m new technology telescope at the la silla paranal observatory using the eso multi - mode instrument ( emmi ) .
the long - slit emission - line spectra also obtained with the emmi ( programme i d 074.d-0373 ) in 2005 revealed much more detail of the nebular morphology .
the first spatio - kinematical model using the emmi long - slit data by @xcite suggested the existence of a bright torus with a systemic heliocentric radial velocity of @xmath12 kms@xmath13 encircling the waist of an extended bipolar nebular shell . in this paper , we aim to uncover the properties of the hidden hot ionizing source in suwt 2 .
we aim to do this by applying a self - consistent three - dimensional photoionization model using the mocassin 3d code by @xcite . in section [ suwt2:sec : observations ] , we describe our optical integral field observations as well as the data reduction process and the corrections for interstellar extinction . in section
[ suwt2:sec : kinematic ] , we describe the kinematics . in section [ suwt2:sec : tempdens ] , we present our derived electron temperature and density , together with our empirical ionic abundances in section [ suwt2:sec : abundances ] . in section [ suwt2:sec : photoionization ] , we present derived ionizing source properties and distance from our self - consistent photoionization models , followed by a conclusion in section [ suwt2:sec : conclusions ] .
integral - field spectra of suwt 2 were obtained during two observing runs in 2009 may and 2012 august with the wide field spectrograph ( wifes ; * ? ? ?
wifes is an image - slicing integral field unit ( ifu ) developed and built for the anu 2.3-m telescope at the siding spring observatory , feeding a double - beam spectrograph .
wifes samples 0.5 arcsec along each of twenty five @xmath14arcsec @xmath10 @xmath15arcsec slits , which provides a field - of - view of @xmath16arcsec@xmath10@xmath14arcsec and a spatial resolution element of @xmath17arcsec@xmath10@xmath18arcsec ( or @xmath19 for y - binning=2 )
. the spectrograph uses volume phase holographic gratings to provide a spectral resolution of @xmath20 ( 100 kms@xmath13 full width at half - maximum , fwhm ) , and @xmath21 ( 45 kms@xmath13 fwhm ) for the red and blue arms , respectively .
each grating has a different wavelength coverage .
it can operate two data accumulation modes : classical and nod - and - shuffle ( n&s ) .
the n&s accumulates both object and nearby sky - background data in either equal exposures or unequal exposures .
the complete performance of the wifes has been fully described by @xcite . and [ nii]6584 taken with the european southern observatory ( eso ) 3.6-m telescope ( programme i d 055.d-0550 ) .
the rectangles correspond to the wifes fields of view used for our study : 1 ( red ) and 2 ( blue ) ; see table [ suwt2:tab : observations ] for more details .
bottom panels : spatial distribution maps of flux intensity and continuum of @xmath22o iii@xmath23 $ ] @xmath55007 for field 2 and locations of apertures ( @xmath24arcsec@xmath10@xmath25arcsec ) used to integrate fluxes , namely ` a ' the ring and ` b ' the interior structure .
the white contour lines show the distribution of the above narrow - band h@xmath2 emission in arbitrary unit .
north is up and east is towards the left - hand side . , title="fig:",width=316 ] + and [ nii]6584 taken with the european southern observatory ( eso ) 3.6-m telescope ( programme i d 055.d-0550 ) .
the rectangles correspond to the wifes fields of view used for our study : 1 ( red ) and 2 ( blue ) ; see table [ suwt2:tab : observations ] for more details .
bottom panels : spatial distribution maps of flux intensity and continuum of @xmath22o iii@xmath23 $ ] @xmath55007 for field 2 and locations of apertures ( @xmath24arcsec@xmath10@xmath25arcsec ) used to integrate fluxes , namely ` a ' the ring and ` b ' the interior structure .
the white contour lines show the distribution of the above narrow - band h@xmath2 emission in arbitrary unit .
north is up and east is towards the left - hand side . , title="fig:",width=163 ] and [ nii]6584 taken with the european southern observatory ( eso ) 3.6-m telescope ( programme i d 055.d-0550 ) .
the rectangles correspond to the wifes fields of view used for our study : 1 ( red ) and 2 ( blue ) ; see table [ suwt2:tab : observations ] for more details .
bottom panels : spatial distribution maps of flux intensity and continuum of @xmath22o iii@xmath23 $ ] @xmath55007 for field 2 and locations of apertures ( @xmath24arcsec@xmath10@xmath25arcsec ) used to integrate fluxes , namely ` a ' the ring and ` b ' the interior structure .
the white contour lines show the distribution of the above narrow - band h@xmath2 emission in arbitrary unit .
north is up and east is towards the left - hand side .
, title="fig:",width=163 ] our observations were carried out with the b3000/r3000 grating combination and the rt560 dichroic using n&s mode in 2012 august ; and the b7000/r7000 grating combination and the rt560 dichroic using the classical mode in 2009 may .
this covers @xmath2633005900 in the blue channel and @xmath2655009300 in the red channel .
as summarized in table [ suwt2:tab : observations ] , we took two different wifes exposures from different positions of suwt 2 ; see fig .
[ suwt2:fig1 ] ( top ) .
the sky field was collected about 1 arcmin away from the object . to reduce and calibrate the data ,
it is necessary to take the usual bias frames , dome flat - field frames , twilight sky flats , ` wire ' frames and arc calibration lamp frames .
although wire , arc , bias and dome flat - field frames were collected during the afternoon prior to observing , arc and bias frames were also taken through the night .
twilight sky flats were taken in the evening .
for flux calibration , we also observed some spectrophotometric standard stars .
.journal of suwt 2 observations at the anu 2.3-m telescope . [ cols="<,^,^,^",options="header " , ] + @xmath27 shock excitation largely enhances the s@xmath28/h@xmath28 ionic abundance ratio .
table [ suwt2:tab : modelresults ] compares the flux intensities calculated by our models with those from the observations .
the fluxes are given relative to h@xmath29 , on a scale where h@xmath30 .
most predicted line fluxes from each model are in fairly good agreement with the observed values and the two models produce very similar fluxes for most observed species .
there are still some discrepancies in the few lines , e.g. @xmath22o ii@xmath23 $ ] @xmath263726,3729 and @xmath22s ii@xmath23 $ ] @xmath266716,6731 .
the discrepancies in @xmath22o ii@xmath23 $ ] @xmath263726,3729 can be explained by either recombination contributions or intermediate phase caused by a complex density distribution ( see e.g. discussion in * ? ? ?
@xmath22s ii@xmath23 $ ] @xmath266716,6731 was affected by shock - ionization and its true flux intensity is much lower without the shock fronts .
meanwhile , @xmath22ar iii@xmath23 $ ] 7751 was enhanced by the telluric line .
the recombination line h@xmath31 @xmath54102 and he ii @xmath55412 were also blended with the o ii recombination lines .
there are also some recombination contributions in the @xmath22o ii@xmath23 $ ] @xmath267320,7330 doublet .
furthermore , the discrepancies in the faint auroral line [ n ii ] @xmath55755 and [ o iii ] @xmath54363 can be explained by the recombination excitation contribution ( see section 3.3 in * ? ? ?
table [ suwt2:tab : temperatures ] represents mean electron temperatures weighted by ionic abundances for models 1 and 2 , as well as the ring region and the inside region of the pn .
we also see each ionic temperature corresponding to the temperature - sensitive line ratio of a specified ion .
the definition for the mean temperatures was given in @xcite ; and in detail by @xcite .
our model results for @xmath32o iii@xmath23 $ ] compare well with the value obtained from the empirical analysis in
[ suwt2:sec : tempdens ] .
[ suwt2:fig16 ] ( top left ) shows @xmath33 obtained for model 2 ( adopted best - fitting model ) constructed in @xmath34 cubic grids , and with the ionizing source being placed in the corner .
it replicates the situation where the inner region has much higher @xmath33 in comparison to the ring @xmath33 as previously found by plasma diagnostics in [ suwt2:sec : tempdens ] .
in particular the mean values of @xmath32o iii@xmath23 $ ] for the ring ( torus of the actual nebula ) and the inside ( spheroid ) regions are around @xmath35 and @xmath36k in all two models , respectively .
they can be compared to the values of table [ suwt2:tab : tenediagnostics ] that is @xmath32o iii@xmath23=12\,300$]k ( ring ) and @xmath37k ( interior ) . although the average temperature of @xmath32n ii@xmath23\simeq11\,700$]k over the entire nebula is higher than that given in table [ suwt2:tab : tenediagnostics ] , the average temperature of @xmath32o iii@xmath23\simeq13,000$]k is in decent agreement with that found by our plasma diagnostics .
it can be seen in table [ suwt2:tab : tenediagnostics ] that the temperatures for the two main regions of the nebula are very different , although we assumed a homogeneous elemental abundance distribution for the entire nebula relative to hydrogen .
the temperature variations in the model can also be seen in fig .
[ suwt2:fig16 ] .
the gas density structure and the location of the ionizing source play a major role in heating the central regions , while the outer regions remain cooler as expected .
overall , the average electron temperature of the entire nebula increases by increasing the helium abundance and decreasing the oxygen , carbon and nitrogen abundances , which are efficient coolants .
we did not include any dust grains in our simulation , although we note that a large dust - to - gas ratio may play a role in the heating of the nebula via photoelectric emissions from the surface of grains .
results for the fractional ionic abundances in the ring ( torus ) and inner ( oblate spheroid ) regions of our two models are shown in table [ suwt2:tab : ionfraction ] and fig .
[ suwt2:fig16 ] .
it is clear from the figure and table that the ionization structures from the models vary through the nebula due to the complex density and radiation field distribution in the gas .
as shown in table [ suwt2:tab : ionfraction ] , he@xmath38/he is much higher in the inner regions , while he@xmath28/he is larger in the outer regions , as expected .
similarly , we find that the higher ionization stages of each element are larger in the inner regions . from table
[ suwt2:tab : ionfraction ] we see that hydrogen and helium are both fully ionized and neutrals are less than 8% by number in these best - fitting models .
therefore , our assumption of @xmath39 is correct in our empirical method .
table [ suwt2:tab : ionratio ] lists the nebular average ionic abundance ratios calculated from the photoionization models .
the values that our models predict for the helium ionic ratio are fairly comparable with those from the empirical methods given in [ suwt2:sec : abundances ] , though there are a number of significant differences in other ions .
the o@xmath40/h@xmath40 ionic abundance ratio is about 33 per cent lower , while o@xmath41/h@xmath40 is about 60% lower in model 2 than the empirical observational value .
the empirical value of s@xmath28 differs by a factor of 8 compared to our result in model 2 , explained by the shock - excitation effects on the @xmath22s ii@xmath23 $ ] @xmath266716,6731 doublet .
additionally , the ne@xmath38/h@xmath40 ionic abundance ratio was underestimated by roughly 67% in model 2 compared to observed results , explained by the properties of the ionizing source .
the ar@xmath42/h@xmath40 ionic abundance ratio in model 2 is 56% lower than the empirical results .
other ionic fractions do not show major discrepancies ; differences remain below 35% .
we note that the n@xmath28/n ratio is roughly equal to the o@xmath28/o ratio , similar to what is generally assumed in the @xmath43(n ) method .
however , the ne@xmath38/ne ratio is nearly a factor of 2 larger than the o@xmath38/o ratio , in contrast to the general assumption for @xmath43(ne ) ( see equation [ suwt2:eq_ne_cel1 ] ) .
it has already been noted by @xcite that an alternative ionization correction method is necessary for correcting the unseen ionization stages for the neon abundance . in fig .
[ suwt2:fig15 ] we compared the values of the effective temperature @xmath44 and luminosity @xmath45 obtained from our two models listed in table [ suwt2:tab : modelparameters ] to evolutionary tracks of hydrogen - burning and helium - burning models calculated by @xcite .
we compared the post - agb age of these different models with the dynamical age of the ring found in
[ suwt2:sec : kinematic ] .
the kinematic analysis indicates that the nebula was ejected about 2340026300 yr ago .
the post - agb age of the hydrogen - burning model ( left - hand panel in fig .
[ suwt2:fig15 ] ) is considerably shorter than the nebula s age , suggesting that the helium - burning model ( vltp ; right - hand panel in fig .
[ suwt2:fig15 ] ) may be favoured to explain the age .
the physical parameters of the two a - type stars also yield a further constraint .
the stellar evolutionary tracks of the rotating models for solar metallicity calculated by @xcite imply that the a - type stars , both with masses close to @xmath46 and @xmath47k , have ages of @xmath48 myr .
we see that they are in the evolutionary phase of the `` blue hook '' ; a very short - lived phase just before the hertzsprung gap .
interestingly , the initial mass of @xmath49 found for the ionizing source has the same age .
as previously suggested by @xcite , the pn progenitor with an initial mass slightly greater than @xmath46 can be coeval with the a - type stars , and it recently left the agb phase .
but , they adopted the system age of about 520 myr according to the y@xmath50 evolutionary tracks @xcite .
the effective temperature and stellar luminosity obtained for both models correspond to the progenitor mass of @xmath49 .
however , the strong nitrogen enrichment seen in the nebula is inconsistent with this initial mass , so another mixing process rather than the hot - bottom burning ( hbb ) occurs at substantially lower initial masses than the stellar evolutionary theory suggests for agb - phase @xcite .
the stellar models developed by @xcite indicate that hbb occurs in intermediate - mass agb stars with the initial mass of @xmath51 for the metallicity of @xmath52 ; and @xmath53 for @xmath54@xmath55 .
however , they found that a low - metallicity agb star ( @xmath56 ) with the progenitor mass of @xmath49 can also experience hbb .
our determination of the argon abundance in suwt 2 ( see table[suwt2:tab : modelparameters ] ) indicates that it does not belong to the low - metallicity stellar population ; thus , another non - canonical mixing process made the abundance pattern of this pn .
the stellar evolution also depends on the chemical composition of the progenitor , namely the helium content ( @xmath57 ) and the metallicity ( @xmath58 ) , as well as the efficiency of convection ( see e.g. * ? ? ?
more helium increases the h - burning efficiency , and more metallicity makes the stellar structure fainter and cooler .
any change in the outer layer convection affects the effective temperature .
there are other non - canonical physical processes such as rotation , magnetic field and mass - loss during roche lobe overflow ( rlof ) in a binary system , which significantly affect stellar evolution .
@xcite calculated a grid of stellar evolutionary tracks with rotation , and found that n / h at the surface in rotating models is higher than non - rotating models in the stellar evolutionary tracks until the end of the central hydrogen- and helium - burning phases prior to the agb stage .
the modules for experiments in stellar astrophysics ( mesa ) code developed by @xcite indicates that an increase in the rotation rate ( or angular momentum ) enhances the mass - loss rate .
the rotationally induced and magnetically induced mixing processes certainly influence the evolution of intermediate - mass stars , which need further studies by mesa .
the mass - loss in a binary or even triple system is much more complicated than a single rotating star , and many non - canonical physical parameters are involved ( see e.g. binstar code by * ? ? ?
* ; * ? ? ?
@xcite used the cambridge stellar evolution ( stars ) code developed by @xcite to study numerically evolution of population i binaries , and produced a helium - rich outer layer .
similarly , @xcite developed a helium white dwarf from a low mass progenitor in a close binary system .
a helium enrichment in the our layer can considerably influence other elements through the helium - burning mixing process .
in this paper we have analysed new optical integral - field spectroscopy of the pn suwt 2 to study detailed ionized gas properties , and to infer the properties of the unobserved hot ionizing source located in the centre of the nebula .
the spatially resolved emission - line maps in the light of @xmath22n ii@xmath23 $ ] @xmath56584 have described the kinematic structure of the ring .
the previous kinematic model @xcite allowed us to estimate the nebula s age and large - scale kinematics in the galaxy .
an empirical analysis of the emission line spectrum led to our initial determination of the ionization structure of the nebula .
the plasma diagnostics revealed as expected that the inner region is hotter and more excited than the outer regions of the nebula , and is less dense .
the ionic abundances of he , n , o , ne , s and ar were derived based on the empirical methods and adopted mean electron temperatures estimated from the observed @xmath22o iii@xmath23 $ ] emission lines and electron densities from the observed @xmath22s ii@xmath23 $ ] emission lines .
we constructed photoionization models for the ring and interior of suwt 2 .
this model consisted of a higher density torus ( the ring ) surrounding a low - density oblate spheroid ( the interior disc ) .
we assumed a homogeneous abundance distribution consisting of eight abundant elements .
the initial aim was to find a model that could reproduce the flux intensities , thermal balance structure and ionization structure as derived from by the observations .
we incorporated nlte model atmospheres to model the ionizing flux of the central star . using a hydrogen - rich model atmosphere , we first fitted all the observed line fluxes , but the time - scale of the evolutionary track was not consistent with the nebula s age .
subsequently , we decided to use hydrogen - deficient stellar atmospheres implying a vltp ( born - again scenario ) , and longer time - scales were likely to be in better agreement with the dynamical age of the nebula .
although the results obtained by the two models of suwt 2 are in broad agreement with the observations , each model has slightly different chemical abundances and very different stellar parameters .
we found a fairly good fit to a hydrogen - deficient central star with a mass of @xmath59 with an initial ( model ) mass of @xmath60 .
the evolutionary track of @xcite implies that this central star has a post - agb age of about 25000 yr .
interestingly , our kinematic analysis ( based on @xmath61 from * ? ? ?
* ) implies a nebular true age of about 2340026300 yr .
table [ suwt2:tab : modelparameters ] lists two best - fitting photoionization models obtained for suwt 2 .
the hydrogen - rich model atmosphere ( model 1 ) has a normal evolutionary path and yields a progenitor mass of @xmath49 , a dynamical age of 7,500 yr and nebular @xmath62 ( by number ) .
the pg 1159 model atmosphere ( model 2 ) is the most probable solution , which can be explained by a vltp phase or born - again scenario : vltp @xmath63 [ wcl ] @xmath63 [ wce ] @xmath63 [ wc]-pg1159 @xmath63pg1159 @xcite . the pg1159 model yields @xmath64 and a stellar temperature of @xmath65 kk corresponding to the progenitor mass of @xmath49 and much longer evolutionary time - scale . the vltp can be characterized as the helium - burning model ,
but this can not purely explain the fast stellar winds ( @xmath66kms@xmath67 ) of typical [ wce ] stars .
it is possible that an external mechanism such as the tidal force of a companion and mass transfer to an accretion disc , or the strong stellar magnetic field of a companion can trigger ( late ) thermal pulses during post - agb evolution .
the abundance pattern of suwt 2 is representative of a nitrogen - rich pn , which is normally considered to be the product of a relatively massive progenitor star @xcite .
recent work suggests that hbb , which enhances the helium and nitrogen , and decreases oxygen and carbon , occurs only for initial masses of @xmath685@xmath69 ( @xmath52 ; * ? ? ?
* ; * ? ? ?
* ) ; hence , the nitrogen enrichment seen in the nebula appears to result from an additional mixing process active in stars down to a mass of 3@xmath69 .
additional physical processes such as rotation increase the mass - loss rate @xcite and nitrogen abundance at the stellar surface ( end of the core h- and he - burning phases ; * ? ? ?
the mass - loss via rlof in a binary ( or triple ) system can produce a helium - rich outer layer @xcite , which significantly affects other elements at the surface .
ad warmly acknowledges the award of an international macquarie university research excellence scholarship ( imqres ) .
qap acknowledges support from macquarie university and the australian astronomical observatory .
we wish to thank nick wright and michael barlow for interesting discussions .
we are grateful to david j. frew for the initial help and discussion .
we thank anna v. kovacevic for carrying out the may 2009 2.3 m observing run , and lizette guzman - ramirez for assisting her with it .
we also thank travis stenborg for assisting ad with the 2012 august 2.3 m observing run .
ad thanks milorad stupar for his assistance in the reduction process .
we would like to thank the staff at the anu siding spring observatory for their support , especially donna burton .
this work was supported by the nci national facility at the anu .
we would also like to thank an anonymous referee for helpful suggestions that greatly improved the paper .
h. e. , 2000 , in astronomical society of the pacific conference series , vol .
199 , asymmetrical planetary nebulae ii : from origins to microstructures , j. h. kastner , n. soker , & s. rappaport , ed .
, p. 115
h. e. , obrien m. s. , sion e. m. , mullan d. j. , exter k. , pollacco d. l. , webbink r. f. , 2002 , in astronomical society of the pacific conference series , vol .
279 , exotic stars as challenges to evolution , c. a. tout & w. van hamme , ed . , p. 239
|
the planetary nebula suwt 2 ( pn g311.0 + 02.4 ) , is an unusual object with a prominent , inclined central emission ellipse and faint bipolar extensions .
it has two a - type stars in a proven binary system at the centre .
however , the radiation from these two central stars is too soft to ionize the surrounding material leading to a so far fruitless search for the responsible ionizing source . such a source is clearly required and has already been inferred to exist via an observed temporal variation of the centre - of - mass velocity of the a - type stars .
moreover , the ejected nebula is nitrogen - rich which raises question about the mass - loss process from a likely intermediate - mass progenitor .
we use optical integral - field spectroscopy to study the emission lines of the inner nebula ring .
this has enabled us to perform an empirical analysis of the optical collisionally excited lines , together with a fully three - dimensional photoionization modelling .
our empirical results are used to constrain the photoionization models , which determine the evolutionary stage of the responsible ionizing source and its likely progenitor .
the time - scale for the evolutionary track of a hydrogen - rich model atmosphere is inconsistent with the dynamical age obtained for the ring .
this suggests that the central star has undergone a very late thermal pulse .
we conclude that the ionizing star could be hydrogen - deficient and compatible with what is known as a pg 1159-type star .
the evolutionary tracks for the very late thermal pulse models imply a central star mass of @xmath0 , which originated from a @xmath1 progenitor .
the evolutionary time - scales suggest that the central star left the asymptotic giant branch about 25,000 years ago , which is consistent with the nebula s age .
[ firstpage ] ism : abundances planetary nebulae : individual : pn suwt 2
|
the study of quantum effects in nonintertial frames of reference has been thoroughly discussed in the literature .
the pioneer articles of fulling and unruh @xcite showing the non - equivalence of the quantization of scalar fields in rindler and minkowski coordinates and the thermal character of the radiation were the origin of a large body of articles devoted to analyze quantum measurement processes in uniformly accelerated frames and possible interpretations thereof .
the advantage of considering rindler coordinates are many .
they can be associated with a uniformly accelerated observer .
they also possess a global timelike killing vector and the ( massive and massless ) klein - gordon as well as the dirac equations are separable in the rindler coordinates .
the rindler coordinates can be extended to cover the whole space time and thermal effects can be related , via the equivalence principle , with the hawking effect @xcite .
the study of quantum effects in non uniformly accelerated frames of reference presents , at first glance , different technical problems . among them
we can mention that the system of coordinates associated with non - rindler kinematics do not possess in general a timelike killing vector , and therefore a standard interpretation of positive and negative frequency solutions is absent .
the complete separation of variables of the klein - gordon and the dirac equations is possible only in a restricted set of coordinates , and those coordinates allowing separability sometimes present coordinate singularities ; therefore the quantization scheme fails . among the non - static coordinate systems where the klein - gordon and dirac equations separate we have@xcite @xmath0 the line element associated with the coordinate transformation ( [ 1 ] )
is @xmath1 the separability of he klein - gordon equation in ( [ 1 ] ) has been discussed by kalnins @xcite , one the authors has accomplished a complete separability of the dirac@xcite equation in ( [ 1 ] ) .
the kinematics associated with the coordinates ( [ 1 ] ) have been exhaustively analyzed by costa @xcite .
the proper time along @xmath2 is @xmath3 considering @xmath4 as the evolution parameter , we have that an observer co - moving to the system ( [ 1 ] ) has a four - velocity given by @xmath5 and experiences an acceleration whose components are @xmath6 from the absolute value of the acceleration , we readily obtain that @xmath7 takes the form @xmath8 from ( [ 3 ] ) we obtain that the accelerated frame becomes inertial as @xmath9 , and on the other hand it evolves toward an uniformly accelerated frame as @xmath10 . quantum effects in the non - inertial frame ( [ 1 ] ) have been discussed by costa @xcite and by percoco and villalba @xcite for dirac particles .
the nonexistence of a global timelike killing vector for the line element ( [ 2 ] ) precludes making a straightforward identification of the positive and negative frequency solutions of the scalar and dirac wave equations . in order to circumvent this difficulty
, we identify positive and negative frequency modes comparing the asymptotic solutions of the wave equations with those obtained for the relativistic hamilton - jacobi equation recently , bautista @xcite , has discussed , in a rindler accelerated frame , vacuum effect associated with a spin 1/2 particle with anomalous magnetic moment in a constant magnetic field directed along the acceleration .
the author obtains a planckian distribution of created particles that depends on the magnetic field via the non - minimal anomalous coupling .
the author also discusses deviation of the energy density from a thermal distribution due to the magnetic field . as a preliminary step towards a deeper understanding of quantum processes in non inertial frames of reference , in the present paper we analyze vacuum effects associated with scalar and spin @xmath11 particles in the coordinates ( [ 1 ] ) , when a constant magnetic field is also present .
the article is structured as follows : in section ii , we solve the relativistic hamilton - jacobi equation with a constant magnetic field in the accelerated frame ( [ 1 ] ) . in sec .
iii , we solve the klein - gordon equation and compute the rate of scalar particles created . in sec .
iv , we solve the dirac equation with anomalous magnetic moment and compute the density of particles related by the magnetic field in the non inertial frame ( [ 1 ] ) . finally , we discuss the results obtained in this article in sec . v.
the relativistic hamilton - jacobi equation coupled to an electromagnetic field can be written as@xcite @xmath12 where here and elsewhere we adopt the units where @xmath13 , and @xmath14 the vector potential associated with a constant magnetic field @xmath15 directed along the acceleration ( [ acel ] ) has the form @xmath16 it is not difficult to verify that ( [ a ] ) satisfies the conditions @xmath17 and @xmath18 . substituting the line element ( [ 2 ] ) into ( [ hj ] )
we obtain @xmath19 + \left ( \omega _ { o}z-\partial _ { y}s\right ) ^{2}+\left ( \partial _ { z}s\right ) ^{2}+m^{2}=0 \end{array } \label{k24}\ ] ] where @xmath20 . the solution of eq .
( [ k24 ] ) has the form @xmath21 which in the asymptotic limit as @xmath22 and @xmath9 reduces to @xmath23 the wave function @xmath24 gives the quasi - classical asymptotes of the solutions of the klein - gordon and dirac equations . in the remote past , as @xmath25 @xmath26 takes the form @xmath27 where in ( [ k27 ] ) the upper and lower signs correspond respectively to positive and negative frequency modes .
analogously , we have that as @xmath22 and @xmath28 ( [ k25 ] ) reduces to @xmath29 and consequently , @xmath30 where , in the present case , the upper sign corresponds to positive frequency modes and the lower sign to negative modes . the results ( [ k27 ] ) and ( [ k29 ] ) give the quasi - classical asymptotic behaviors of the relativistic wave equations in the accelerated coordinates ( [ 1 ] ) .
in this section we proceed to solve the klein - gordon equation , coupled to a constant magnetic field , in the accelerated coordinates with the line element given by ( [ 2 ] ) . the covariant generalization of the klein gordon equation is@xcite @xmath31 where @xmath32 is the covariant derivative , and @xmath33 is the vector potential given by ( [ a ] ) . substituting eq .
( [ 2 ] ) into ( [ kg ] ) we readily obtain @xmath34 \psi = 0 .
\label{k6}\ ] ] since eq .
( [ k6 ] ) commutes with the operator @xmath35 we can look for a solution of the form @xmath36 which reduces eq .
( [ k6 ] ) to @xmath37 eq .
( [ k7 ] ) can be separated in the form @xmath38 .
the resulting equations are @xmath39 \eta = 0 \label{k8}\ ] ] @xmath40 f(z ) \label{k9}\ ] ] where @xmath41 is a separation constant
. variables @xmath42 and @xmath4 can be separated in eq .
( [ k8 ] ) after making the substitution : @xmath43 .
the resulting equations for @xmath42 and @xmath4 are @xmath44 @xmath45 where @xmath46 is a constant of separation .
( [ k11 ] ) takes a more familiar form in terms of bessel functions@xcite after introducing the variable @xmath47 @xmath48 whose solution are the modified bessel functions@xcite @xmath49 and @xmath50@xmath51 where @xmath52 , @xmath53 are arbitrary constants and @xmath54 .
the solution of eq .
( [ k12 ] ) can be obtained in the same manner . introducing the change of variables @xmath55 in ( [ k12 ] ) we get the bessel equation @xmath56 whose
solutions can be expressed in terms of the hankel functions @xcite @xmath57 and @xmath58 @xmath59 where @xmath60 and @xmath61 are arbitrary constants . in order to solve eq.([k9 ] ) , we introduce the new variable @xmath62 .
( [ k9 ] ) takes the form @xmath63 f(x)=0 \label{k14}\ ] ] where @xmath64 .
( [ k14 ] ) is the parabolic cylinder equation @xcite .
therefore the solution of eq .
( [ k9 ] ) is @xmath65 let us analyze the asymptotic behavior of the solutions of eq .
( [ kgc ] ) as @xmath22 and @xmath66 we obtain that @xmath67 = \\ = e^{-\frac{\lambda } { \omega } e^{\omega x}}\left [ a_{-\infty } ^{\prime } e^{i% \frac{\lambda } { \omega } e^{-\omega t}}+b_{-\infty } ^{\prime } e^{-i\frac{% \lambda } { \omega } e^{-\omega t}}\right ] \end{array } \label{k30}\ ] ] comparing ( [ k30 ] ) with ( [ k27 ] ) we identify the first right hand side term as a positive frequency mode , and the second term as a negative frequency . on the other hand
when @xmath22 and @xmath28 we obtain @xmath68 = \\ = e^{-\frac{\lambda } { \omega } e^{\omega x}}\left [ a_{\infty } ^{\prime } \left ( e^{-\omega t}\right ) ^{i\nu } + b_{\infty } ^{\prime } \left ( e^{-\omega t}\right ) ^{-i\nu } \right ] = e^{-\frac{\lambda } { \omega } e^{\omega x}}\left ( a_{\infty } ^{\prime } e^{-i\sqrt{k}t}+b_{\infty } ^{\prime } e^{i\sqrt{k}% t}\right ) \end{array } \label{k31}\ ] ] also , comparing ( [ k31 ] ) with ( [ k29 ] ) we can identify the first and second right hand side terms in ( [ k31 ] ) as positive and negative frequency modes respectively .
now , we are going to express an inertial positive frequency mode ( @xmath9 ) @xmath69 in terms of the accelerated modes in the asymptotic future ( @xmath70@xmath71 where @xmath72 and @xmath73 are normalization constants according to the standard inner product@xcite @xmath74 @xmath75 for the klein gordon equation .
the recurrence relation between @xmath76 and @xmath77 permits one to express @xmath78 in terms of @xmath79 as follows @xmath80 .
\label{relat}\ ] ] since the inertial and accelerated modes are related via the bogoliubov coefficients@xcite @xmath81 and @xmath82 @xmath83 we have that eq .
( [ relat ] ) gives immediately the values of @xmath84 and @xmath85@xmath86 since @xmath87 and @xmath88 the density of created particles@xcite has the form @xmath89 which can be identified as a planck distribution with a temperature @xmath90 the temperature measured by the accelerated observer can be obtained using the relation @xcite : @xmath91 in the asymptotic limit as @xmath92 we have @xmath93 and the temperature @xmath4 takes the value @xmath94 then we obtain that the temperature is proportional to the asymptotic value of the acceleration .
in this section we solve the dirac equation with anomalous magnetic moment in the accelerated coordinates ( [ 1 ] ) when a constant magnetic field is present .
the covariant generalization of the dirac - pauli equation in curvilinear coordinates is@xcite @xmath95 where @xmath96 are the curvilinear dirac matrices satisfying the anticommutation relations @xmath97 , @xmath98 are the spinor connections and @xmath99 is the electromagnetic tensor .
the curvilinear @xmath96 matrices are related to the constant minkowski @xmath100 matrices with @xmath101 via the tetrad @xmath102@xmath103 in order to write the curvilinear dirac matrices in eq ( [ dirac1 ] ) we have to choose a tetrad @xmath104 here we are going to work in the diagonal tetrad gauge , where @xmath105 takes the form @xmath106 it is easy to verify that @xmath107 . in this tetrad gauge
, the curvilinear dirac matrices can be expressed in terms of the constant @xmath100 as follows @xmath108 in the diagonal tetrad gauge ( [ diag ] ) the spinor connections , defined by the relation@xcite : @xmath109 with @xmath110 , take the form @xmath111 substituting ( [ gammas ] ) , ( [ rel7 ] ) and ( [ a ] ) into ( [ dirac1 ] ) we obtain @xmath112 + \overline{\gamma } % ^{2}\left ( \frac{\partial \psi } { \partial y}-i\frac{eb_{x}}{\sqrt{2}}z\psi \right ) + \overline{\gamma } ^{3}\frac{\partial \psi } { \partial z}+ \\
+ \frac{\overline{\gamma } ^{4}}{\sqrt{e^{2\omega x}+e^{-2\omega t}}}\left [ \frac{\partial \psi } { \partial t}-\frac{\omega e^{-2\omega t}}{2\left ( e^{2\omega x}+e^{-2\omega t}\right ) } \psi \right ] + m\psi + i\mu b_{x}% \overline{\gamma } ^{2}\overline{\gamma } ^{3}\psi = 0 . \end{array}\ ] ] introducing the spinor @xmath113 @xmath114 we eliminate the contribution terms due to the spinor connections .
the dirac equation takes the form @xmath115 where @xmath116 . in order to solve eq .
( [ dirac2 ] ) we proceed to separate variables using the algebraic method of separation developed by shishkin and villalba@xcite the idea behind the method is to reduce eq .
( [ dirac2 ] ) to a sum of two commuting first order differential operators @xmath117 _ { -}=0 \label{k}\ ] ] where in the present case , @xmath118 and @xmath119 since @xmath120 and @xmath121 satisfy ( [ k ] ) they satisfy the eigenvalue equations @xmath122 now , we proceed to solve equation @xmath123
@xmath124 applying the transformation @xmath125 defined by : @xmath126 we reduce eq .
( [ c8 ] ) to the form @xmath127 \phi + \\ + \overline{\gamma } ^{4}\overline{\gamma } ^{2}\overline{\gamma } ^{3}\left [ \partial _ { t}\xi ( x , t)+i\partial _ { x}\theta ( x , t)\right ] \phi = -i\lambda \sqrt{e^{2\omega x}+e^{-2\omega t}}e^{2i\overline{\gamma } ^{1}\overline{% \gamma } ^{4}\theta ( x , t)}\phi \end{array } \label{c10}\ ] ] in order to separate variables in ( [ c10 ] ) we demand that the terms inside the brackets vanish , i.e. @xmath128 the solution of eq .
( [ c11 ] ) is @xmath129 which determines the spinor transformation s eq .
( [ c9 ] ) .
( [ c10 ] ) reduces to @xmath130 let @xmath131 and @xmath132 be the commuting operators @xmath133 = 0.\ ] ] eq .
( [ c13 ] ) can be expressed in terms of @xmath131 and @xmath132 as follows : @xmath134 in order to solve eq .
( [ c16 ] ) we introduce the auxiliary spinor @xmath135 substituting ( [ c17 ] ) into ( [ c16 ] ) we have that @xmath136 this allows separation of variables as follows : @xmath137 @xmath138 where @xmath46 is a constant of separation .
when we choose the following representation for the dirac matrices@xcite @xmath139 : @xmath140 then the spinor @xmath141 has the structure : @xmath142 substituting ( [ c23 ] ) into eq .
( [ c20 ] ) we obtain @xmath143 analogously , eq . ( [ c21 ] ) reduces to @xmath144 in order to solve the system of equations ( [ c24 ] ) , it suffices to solve the second order equation @xmath145 where @xmath146
analogously , we have that solving the system of equations ( [ c25 ] ) is equivalent to solve the whittaker@xcite differential equation @xmath147 where we have introduced the change of variable @xmath148 the solution of eqs .
( [ c27 ] ) and ( [ c29 ] ) can be expressed in terms of a combination of whittaker functions @xmath149 the solutions of eqns .
( [ c27 ] ) and ( [ c29 ] ) are @xmath150 where @xmath151 the spinor @xmath152 can be computed with the help of the inverse transformation ( [ c17 ] ) : @xmath153 the system of equations ( [ c24 ] ) can be written as a coupled system of ordinary differential equations @xmath154 analogously , we have that the eq .
( [ c25 ] ) is equivalent to @xmath155 substituting ( [ c35 ] ) and ( [ c36 ] ) into ( [ c34 ] ) we arrive at @xmath156 where @xmath157 and @xmath158 using the recurrence relations for the whittaker functions we find that @xmath159 has the form @xmath160 where @xmath161 recalling that @xmath125 where @xmath162 is given by ( [ c9 ] ) we have that @xmath163 takes the form @xmath164 now we proceed to solve the equations ( [ c7 ] ) governing the dependence of the spinor solution of the dirac equation on the coordinates @xmath165 and @xmath166 @xmath167 since @xmath168 _ { -}=0 $ ] , we have that @xmath169 can be written as @xmath170 substituting ( [ c54 ] ) into ( [ c53 ] ) we obtain @xmath171 where @xmath20 . introducing the spinor @xmath172 with @xmath173
we find that eq ( [ c55 ] ) reduces to @xmath174 substituting into ( [ c59 ] ) the dirac matrices in the representation ( [ c22 ] ) and considering a spinor with the structure @xmath175 we reduce our problem to that of solving the following system of partial differential equations : @xmath176 \theta _ { 4}=\left ( \mu + \lambda + \mu _ { o}\right ) \theta _ { 1},\ \left [ \partial _ { z}+\left ( -k_{y}+\omega _ { o}z\right ) \right ] \theta _ { 1}=\left ( \mu -\lambda -\mu _
{ o}\right ) \theta _ { 4 } , \label{uno}\ ] ] @xmath176 \theta _ { 2}=\left ( \mu + \lambda + \mu _ { o}\right ) \theta _ { 3},\ \left [ \partial _ { z}+\left ( -k_{y}+\omega _ { o}z\right ) \right ] \theta _ { 3}=\left ( \mu -\lambda -\mu _ { o}\right ) \theta _ { 2}. \label{dos}\ ] ] looking at ( [ uno ] ) and ( [ dos ] ) we see that @xmath177 and @xmath178 and therefore it is only necessary solve one of the systems of coupled equations . making the change of variable @xmath179 we have @xmath180 \theta _ { 1}=0 , \label{c65}\ ] ] @xmath181 \theta _ { 4}=0 , \label{c66}\ ] ] where @xmath182
. equations ( [ c65 ] ) and ( [ c66 ] ) are parabolic cylinder equations @xcite and their solutions are @xmath183 with the help of the recurrence relations for the parabolic cylinder equation and the equations ( [ uno]),([dos ] ) , we can find the relation between the coefficients @xmath184 : @xmath185 and @xmath186 consequently @xmath187 the spinor @xmath188 is obtained using the matrix transformation @xmath189 ( [ c58 ] ) : @xmath190 from ( [ c58 ] ) and ( [ c54 ] ) we obtain @xmath191 combining ( [ c52 ] ) with ( [ c75 ] ) we find that the spinor @xmath192 is @xmath193 now , we proceed to analyze the asymptotic limits as @xmath194 and @xmath92 .
we will confine our attention to the solutions of the spinor ( [ c78 ] ) . in the asymptotes
we obtain a time dependent term multiplied by a factor depending on space variables . here
, we proceed like we did in sec .
iii for the scalar case .
the relation between @xmath195 and @xmath196@xcite will be helpful @xmath197 taking into account expression ( [ c79 ] ) we find that the solutions of eq .
( [ c36 ] ) are related as follows @xmath198 looking at the quasi - classical behavior given by eq .
( [ k29 ] ) , we identify the positive and negative accelerated modes as @xmath199 where @xmath200 is a normalization constant .
also an inertial @xmath201 positive frequency mode @xmath202 is given by @xmath203 where @xmath204 is a normalization constant .
looking at eq .
( [ rec ] ) and recalling that the inertial modes can be expressed in terms of the accelerated positive and negative modes via the bogoliubov coefficients @xmath205 we get that @xmath206 and taking into account that @xmath207 , we find that the density of particles created is@xcite @xmath208 a result that can be identified as a fermi - dirac distribution of particles associated with a temperature @xmath209 and , consequently , the temperature detected in the accelerated frame is@xcite @xmath210 that in the asymptotic limit as @xmath92 takes the form @xmath211 showing that the temperature is proportional to the asymptotic value of the acceleration .
in the present paper we have separated variables and solved the klein - gordon and dirac equations in the curvilinear coordinates ( [ 1 ] ) when a constant magnetic field ( [ a ] ) is present . the algebraic method of separation @xcite has been applied to reduce dirac - pauli equation to a system of coupled ordinary differential equations . using the obtained exact solutions we calculated the density of scalar and spin @xmath11 particles detected by an accelerated observer associated with system of coordinates ( [ 1 ] ) when a constant magnetic field in the direction of acceleration is present .
the identification of positive and negative frequency modes was carried out comparing the relativistic solutions with the quasi - classical hamilton - jacobi solutions ( [ k27 ] ) and ( [ k29 ] ) . the results obtained in sec iii and iv indicate that the magnetic field does not modify the thermal character of spectrum . the temperature associated with the thermal bath
is not modified by @xmath53 .
this result has a classical counterpart : since @xmath53 is colinear to the motion , it does not accelerate the particle and no radiation is caused by it . the presence of a non minimal coupling in eq .
( [ dirac1 ] ) does not affect the density ( [ c87.2 ] ) .
the proportionality between temperature and accelerations remains valid even if uniform accelerations are reached asymptotically .
the role of the anomalous magnetic moment in the energy spectrum density will be discussed in a forthcoming publication .
|
in the present article we solve the dirac - pauli and klein gordon equations in an asymptotically uniformly accelerated frame when a constant magnetic field is present .
we compute , via the bogoliubov coefficients , the density of scalar and spin 1/2 particles created .
we discuss the rle played by the magnetic field and the thermal character of the spectrum .
|
the amount of texts readily available in the world is growing at an astonishing rate ; classifying these texts through machine learning techniques , promptly and without much human intervention , has thus become an important problem in data mining .
much research , in the field of supervised learning , has been done to find accurate algorithms to classify documents in a dataset to their appropriate categories , e.g. see @xcite or @xcite for a detailed survey of text categorization .
the most widely used model for text categorization is the vector space model ( vsm ) @xcite . under this model
, a data dictionary @xmath0 consisting of unique words across the documents in the dataset is constructed .
the documents are represented by real - valued vectors in the space @xmath1 with dimension equaling to the size of the dictionary . given @xmath2 , the @xmath3-th coordinate of a vector is the relative frequency of the word @xmath3 in a given document .
when some of the documents actual class labels are known and used for training , many well - known classifiers in supervised machine learning , such as svm @xcite , @xmath4-nn @xcite , and random forest @xcite , can then be applied to categorize documents .
text categorization decidedly comes across as a problem of detecting similarities between a given text and a collection of texts of a particular type . although distance - based learning rules for text categorization , such as the @xmath4-nearest neighbour classifier , e.g. @xcite , are not new , they are currently based on the entire feature space , while any dimension reduction steps are done independently beforehand @xcite .
we aim to fill this gap by suggesting a novel supervised learning algorithm for text categorization , called the domain - specific classifier .
it discovers specific words for each category , or domain , of documents in training and classifies based on similarity searches in the space of word frequency distributions supported on the respective domain - specific words . for each class label , @xmath5
, our algorithm extracts class , or domain , specific words from labeled training documents , that is , words that appear in the class @xmath6 more frequently than in all the other document classes combined ( modulo a given threshold ) .
now a given unlabeled document is assigned a label @xmath6 if the normalized frequency of domain - specific words for @xmath6 in the document is higher than for any other label . to see that this classifier is indeed similarity search based , let @xmath7 be a binary vector whose @xmath3-th coordinate is @xmath8 if and only if @xmath3 is domain - specific to @xmath6 , and 0 otherwise . normalize @xmath9 according to the @xmath10 distance , and let a document be represented by a vector @xmath11 .
then the label assigned to @xmath12 is that of the closest neighbour to @xmath12 among @xmath13 with regard to the simplest similarity measure , the inner product on @xmath14 . in other words ,
we seek to maximize the value of @xmath15 over @xmath5 .
notice that the well - known cosine similarity measure , cf .
e.g. @xcite , corresponds to the special case @xmath16 .
this algorithm was first used in the 3rd cybersecurity data mining competition ( cdmc 2012 ) to notable success , as the team of authors placed second in the text categorization challenge , and first in classification accuracy @xcite .
in addition , the classification performance of the algorithm was validated on a sub - collection of the popular reuters 21578 dataset @xcite , consisting of single - category documents from the top eight most frequent categories , with the standard `` modapt '' training / testing split . in terms of accuracy ,
our classifier performs slightly better than svm with a linear kernel , and is significantly faster .
this paper is organized as follows .
section [ backg ] surveys common feature selection and extraction methods and classifiers considered in the text categorization literature .
section [ domainclass ] explains the new domain - specific classifier in detail and casts it as a similarity search problem .
section [ discuss ] discusses results from the cdmc 2012 data mining competition and experiments from the competition and on the reuters 21578 dataset .
finally , section [ conclude ] concludes the paper with some discussion and directions for future work .
in this section , we describe the vsm model and provide a brief survey on widely known methods for text categorization . from this section
onwards , following notation similar to @xcite , we let @xmath17 denote the dataset of documents , with size @xmath18 , and @xmath19 the data dictionary of all unique words from documents in @xmath20 , with size @xmath21 . given a document @xmath22 and a word @xmath23
, @xmath24 denotes the number of words in @xmath22 and @xmath25 indicates that the word @xmath3 is found in @xmath22 .
the vector space model ( vsm ) @xcite is the most common model for document representation in text categorization . according to @xcite
, there is usually a standard preprocessing step for the documents in @xmath20 , where all alphabets are converted to lowercase and all stop words , such as articles and prepositions , are removed .
sometimes , a stemming algorithm , such as the widely used porter stemmer @xcite , is applied to remove suffices of words ( e.g. the word `` connection '' @xmath26 `` connect '' ) . in vsm , the data dictionary , consisting of all unique words that appear in at least one document in @xmath20 ,
is first constructed .
sometimes , @xmath27-grams , which are phrases of words , are also included in the dictionary ; however , the benefit of these additional phrases is still up for debate @xcite .
given the data dictionary , each document can be represented as a vector in the real - valued vector space with dimension equaling the size of the dictionary .
two common methods for associating a document to a vector are explained below .
the simplest method assigns to a document @xmath22 the vector consisting of the relative term frequencies for @xmath22 , see e.g. @xcite .
the second , known as the @xmath28-@xmath29 method , assigns @xmath22 to the vector consisting of the products of term and inverse document frequencies @xcite . mathematically speaking ,
a document @xmath22 is mapped to a real - valued vector of length @xmath30 : @xmath31 , for @xmath32 or @xmath33 where @xmath34 denotes the number of times the word @xmath35 appears in @xmath22 .
other representations include binary and entropy weightings and the normalized @xmath28-@xmath29 method @xcite .
once the documents are represented as vectors , the dataset can be interpreted as a data matrix @xmath36 of size @xmath37 .
however , a main challenge for text categorization is that the size of the data dictionary is usually immense so the data matrix is extremely high dimensional .
dimension reduction techniques must often be applied before classification to reduce complexity @xcite . due to the potentially large size of the data dictionary , feature selection and extraction methods
are often applied to reduce the dimension of the data matrix .
feature selection methods assign to each feature , a word in the data dictionary , a statistical score based on some measure of importance .
only the highest scored features , past some defined threshold , are kept and a lower dimensional data matrix is created from only these features . some known feature selection methods in text categorization include calculating the document frequency , e.g. @xcite , mutual information @xcite , and @xmath38 statistics , e.g. @xcite .
see e.g. @xcite or @xcite for a thorough study of text feature selection methods .
feature extraction methods transform the original list of features to a smaller list of new features , based on some form of feature dependency .
common well - known feature extraction techniques , as surveyed in @xcite , are latent semantic indexing ( lsi ) @xcite , linear discriminant analysis ( lda ) @xcite , partial least squares ( pls ) @xcite , and random projections @xcite .
well - known classifiers that have been applied to text categorization include the @xmath4-nearest neighbour classifier @xcite , support vector machines @xcite , the naive bayes classifier @xcite , and decision trees @xcite .
we ask the reader to refer to indicated references , or to survey articles , such as @xcite , @xcite , and @xcite .
the paper @xcite provides comparable empirical results on some of these classifiers .
the standard approach in literature for text categorization is that one , or more , feature selection or extraction technique is first applied to a data matrix , since the original data matrix is often extremely high - dimensional . a learning algorithm , independent of the dimension reduction process ,
is then used for classification @xcite .
the novel approach in this paper is that we consider a new classifier based only on extracted class specific words , which naturally reduces time complexity and the dimension of the dataset . in other words ,
the domain - specific classifier both performs dimension reduction and classifies , in consecutive and dependent steps .
our algorithm consists of two distinct stages : extraction of domain - specific words from training samples and classification of documents based on the closest labeled point determined by these domain - specific words .
fix an alphabet @xmath39 and denote @xmath40 as the set of all possible words " formed from @xmath39 .
a document @xmath22 is then simply an ordered sequence of words " , @xmath41 , and the data dictionary @xmath42 is a subset of @xmath40 .
given a set of labeled documents , we can denote it as @xmath43 , where @xmath44 is a document and @xmath45 is its label , out of a possible @xmath4 different labels .
in addition , we can partition @xmath46 into subsets of documents according to their labels : @xmath47 where @xmath48 is the set of documents of label @xmath6 . then , for a particular label @xmath6 and a word @xmath23 in the data dictionary , we denote @xmath49 as the average proportion of times the word @xmath3 appears in documents with label @xmath6 : @xmath50 domain - specific words are those words which appear , on average , proportionally more often in one label type of documents in @xmath46 than other types .
let @xmath51 .
a word @xmath52 in the data dictionary is _ domain _ ( or _ class _ ) _ @xmath6 specific _ if @xmath53 this definition of domain - specific words depends on the parameter @xmath54 and hence , so does the domain - specific classifier .
as @xmath54 increases from @xmath55 , the number of domain - specific words for each class label decreases ; as a result , @xmath54 can be thought of as a threshold parameter , and an optimal choice for @xmath54 is determined through cross - validation using training data .
let now @xmath22 be an unclassified document .
we associate to it a vector @xmath56 ( a relative frequency distribution of words ) as in eq .
( [ eq : frequency ] ) , that is , for every @xmath57 , @xmath58 let @xmath6 be a label . denote @xmath59 the set of domain - specific words to @xmath6 . define the total relative frequency of domain @xmath6 specific words found in @xmath22 : @xmath60 = \sum_{t\in cs_j } w(t)=\frac{1}{\vert d\vert}\sum_{t\in cs_j } c(t , d).\ ] ] the classifier assigns to @xmath22 the label @xmath6 for which the following ratio is the highest : @xmath61}{\vert cs_i\vert^{1/p}}.\ ] ] here , @xmath62 $ ] is a parameter , which normalizes a certain measure with regard to the @xmath10 distance , cf .
below in section [ spacemeasure ] .
a ( positive ) measure on a finite set @xmath0 is simply an assignment @xmath63 to every @xmath57 of a non - negative number @xmath64 ; a probability measure also satisfies @xmath65 .
denote @xmath66 the set of all positive measures on @xmath42 . fix a parameter
@xmath62 $ ] .
the following is a positive measure on @xmath42 : @xmath67 if @xmath68 , we obtain a probability measure uniformly supported on the set of domain @xmath6 specific words . in general , values of @xmath69 $ ]
correspond to different normalizations of the uniform measure supported on these words , according to the @xmath10 distance .
( the case when @xmath70 , that is , the @xmath71 distance , corresponds to non - normalized uniform measure . ) among the similarity measures on @xmath72 , we single out the standard inner product @xmath73 notice that for every @xmath74 and each @xmath6 , @xmath75}{\vert cs_j\vert^{1/p}},\ ] ] and for this reason , the classification algorithm ( [ eq : classifier ] ) can be rewritten as follows : @xmath76 our classifier is based on finding the closest point @xmath77 to the input point @xmath12 in the sense of the simplest similarity measure , the inner product .
the similarity workload is a triple @xmath78 , consisting of the domain @xmath79 , the similarity measure @xmath80 equal to the standard inner product ( a rather common choice in the problems of statistical learning , cf .
@xcite ) , and the dataset @xmath81 of normalized uniform measures corresponding to the text categories and domain - specific words extracted at the preprocessing stage
. note that the well - known cosine similarity measure arises in the special case when the normalizing parameter is @xmath16 ; hence , it is not necessary to consider this measure separately .
our experiments have shown that different datasets require different normalizing parameters for @xmath9 , and that the optimal normalization depends on the sizes of the document categories ; section [ conclude ] includes a discussion on this topic .
this section details the experiments and results obtained for the domain - specific classifier , in the 2012 cybersecurity data mining competition and on the reuters 21578 dataset .
all of the programming for this section were done with standard packages in r @xcite and with the specialized packages @xmath82 @xcite and @xmath83 @xcite , on a desktop running windows 7 enterprise , with a intel i5 3.10 ghz processor and 4 gb of ram . the 3rd cybersecurity data mining competition ( cdmc 2012 ) @xcite , associated with the 19th international conference on neural information processing ( iconip 2012 ) in doha , qatar from november 12 - 15 , 2012 @xcite , included three supervised classification tasks : electronic news ( e - news ) text categorization , intrusion detection , and handwriting recognition .
the domain - specific classifier was first developed by the team of authors for the e - news text categorization challenge , which required classifying news documents to five topics : business , entertainment , sports , technology , and travel .
the documents were collected from eight online news sources .
the words in these documents were obfuscated , and all punctuations and stop words removed . here is a sample scrambled text document paragraph from the competition : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ hujr xj gjxzmuxe fajjaek uo jwxea ursek uyjmx xji k seew eowrjjeer zarwzdek uak wdjkmzxzme kxr ua erreaxzur bmerxzja rzaze zjowuazer xjkuj omrx uzzjowrzrx ojde izxx weied wejwre uxe ojrx rmzzerrwmr rxudxmwr omrx _
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in total , 1063 e - news documents for training , each labeled as one of the @xmath84 topics , were given for the goal of classifying 456 documents .
.information on the dataset size for e - news classification task .
[ cols="^,<,^",options="header " , ] the domain - specific classifier performed slightly better than svm with the linear kernel , and better than random forest in terms of accuracy . with respect to the f - measure ,
our classifier performed better than svm for categories with large sizes , and better than random forest in 6 of the 8 categories , while svm had a higher f - measure on two of the smaller categories , undoubtably due to svm s class weight adjustment .
computationally , our classifier ran considerably faster than svm and random forest .
in this paper , we have introduced a novel text categorization algorithm , the domain - specific classifier , based on similarity searches in the space of measures on the data dictionary .
the classifier finds domain - specific words for each document category , which appear in this category relatively more often than in the rest of the categories combined , and associates to it a normalized uniform measure supported on the domain - specific words . for an unlabeled document ,
the classifier assigns to it the category whose associated measure is most similar to the document s vector of relative word frequencies , with respect to the inner product .
the cosine similarity measure arises as a special case corresponding to the @xmath85 normalization .
our classifier involves a similarity search problem in a suitably interpreted domain .
we believe that this is the right viewpoint with the aim of further improvements .
it is worthwhile noting that our algorithm is unrelated to previously used distance - based algorithms ( e.g. the @xmath4-nn classifier @xcite ) .
the dataset in the similarity workload is completely different , and as a result , unlike most algorithms in text categorization , this classifier does not require any separate dimension reduction step beforehand .
the process of selecting domain - specific words in our algorithm is actually an implicit feature selection method which is class - dependent , something we have not seen before from a classifier in text categorization . for each class , instead of a centroid
, we are choosing an outlier , a uniform measure supported on the domain - specific words , which is representative of this class and not of any other class . not only does each uniform measure lead to a reduction in the dimension of the feature space ( as most words are not domain - specific ) for similarity calculations , it does so dependent of the class labels , since domain - specific words are chosen relative to all classes .
this algorithm was first developed for the 2012 cybersecurity data mining competition and brought the team of authors 2nd place in the text categorization challenge , and 1st place in accuracy .
this is evidence that our algorithm outperformed many existing text categorization algorithms , as surveyed in section [ backg ] .
in addition , our algorithm was evaluated on a sub - collection of the reuters 21578 dataset against two state - of - the - art classifiers , and shown to have a slightly higher classification accuracy than svm , with a higher f - measure for the larger categories , and overall performed better than random forest .
computationally , our classifier ran significantly faster than either , especially in the training stage .
the normalizing parameter @xmath86 plays a significant role : it is to account for class imbalance .
when there are categories with very few documents , @xmath70 should be used to avoid over - emphasizing the smaller categories ; and small values of @xmath86 should be used when the categories have roughly the same number of documents . for future work , we hope to test the domain - specific classifier on biological sequence databases .
other definitions of domain - specific words can be investigated , for instance the one proposed in @xcite .
we would like to experiment with assigning non - uniform measures on the domain - specific words , for instance , by putting weights based on their relative occurrences or on @xmath54 .
finally , we would like to extend the process of selecting domain - specific words to a general classification context , by defining class - specific features relative to the classes and performing classification on only these class - dependent features .
bingham , e. , mannila , h. : random projection in dimensionality reduction : applications to image and text data . in : kdd
01 , san francisco , usa 2001 .
proceedings of 7th acm sigkdd int .
conf . knowledge discovery and data mining , pg .
245 - 250 .
dimitriadou , e. , hornik , k. , leisch , f. , meyer , d. , weingessel , a. : e1071 : misc functions of the department of statistics ( e1071 ) , tu wien .
r package version 1.6 , available at http://cran.r-project.org/package=e1071 ( 2011 ) .
joachims , t. : text categorization with support vector machines : learning with many relevant features . in : ecml-98 , chemnitz , germany 1998 .
proceedings of 10th european conference on machine learning , pg .
137 - 142 .
keim , d.a . ,
oelke , d. , rohrdantz , c. : analyzing document collections via context - aware term extraction . in : nldb 09 saarbrcken , germany .
proceedings of 14th inter . conf .
on appl . of natural language to information systems , pg .
154 - 168 .
r development core team : r : a language and environment for statistical computer .
r foundation for statistical computing , vienna , austria .
isbn 3 - 900051 - 07 - 0 , available at http://www.r-project.org ( 2008 ) .
schtze , h. , hull , d.a . , pedersen , j.o . : a comparison of classifiers and document representations for the routing problem . in : sigir 95 seattle , usa .
proceedings of sigir 1995 , 18th acm international conference on research and development in information retrieval , pg . 229 - 237 .
yang , y. , liu , x. : a re - examination of text categorization methods . in : sigir 99 , berkeley , usa 1999 .
proceedings of the 22nd annual international acm sigir conference on research and development in information retrieval , pg .
42 - 49 .
yang , y. , pedersen , j.o . : a comparative study on feature selection in text categorization . in : icml 97 ,
nashville , usa 1997 .
proceedings of icml-97 , 14th international conference on machine learning , pg .
412 - 420 .
|
we present a supervised learning algorithm for text categorization which has brought the team of authors the 2nd place in the text categorization division of the 2012 cybersecurity data mining competition ( cdmc2012 ) and a 3rd prize overall .
the algorithm is quite different from existing approaches in that it is based on similarity search in the metric space of measure distributions on the dictionary . at the preprocessing stage ,
given a labeled learning sample of texts , we associate to every class label ( document category ) a point in the space of question . unlike it is usual in clustering , this point is not a centroid of the category but rather an outlier , a uniform measure distribution on a selection of domain - specific words . at the execution stage ,
an unlabeled text is assigned a text category as defined by the closest labeled neighbour to the point representing the frequency distribution of the words in the text .
the algorithm is both effective and efficient , as further confirmed by experiments on the reuters 21578 dataset .
|
the study of _ universal crossing probabilities _ in the standard random percolation problem at criticality has been the subject of much interest over the last decade ( see ref .
@xcite for a review ) . at the percolation threshold , @xmath5 , on a rectangular lattice with length @xmath6 and width @xmath7 , the probability to have at least one cluster crossing the system in one of the two directions ( or in both ) is a _ scale - invariant _ quantity in the _ continuum limit _ where the lattice spacing @xmath8 . actually , the crossing probability is a function @xmath9 of the aspect ratio , @xmath10 , and depends also on the boundary conditions .
it remains universal in the sense that it does not depend on the lattice type or whether one is considering site , bond or site - bond percolation .
following the pioneering numerical work of langlands @xcite , cardy @xcite obtained an analytical expression for the crossing probability between opposite sides of a rectangle .
he combined the correspondence between percolation and the limit @xmath11 of the @xmath12-state potts model @xcite with boundary conformal field theory techniques to calculate the crossing probability between two segments on the boundary of a half - plane at criticality .
a conformal mapping could then be used to transform the half - plane result into the rectangular geometry .
cardy s and another crossing probability formula have been recently proven rigorously @xcite .
monte carlo @xcite and exact results @xcite have been obtained for the probability to find @xmath13 incipient spanning clusters connecting two disjoint segments at the boundary of a finite system .
more recently , crossing probabilities on same - spin ising clusters in two dimensions have been also considered @xcite .
let us notice that the existence of a _ scale - invariant _ crossing probability is not a trivial result . in the case of ordinary percolation , for example , it is linked to the vanishing , in the limit @xmath11 , of the scaling dimension @xmath14 of a boundary condition changing operator of the potts model @xcite . in this letter
, we investigate the behaviour of crossing probabilities in anisotropic critical systems .
we first give an exact treatment of weakly anisotropic standard percolation .
we show how a length rescaling technique , which has been used in other contexts @xcite , allows to generalize cardy s exact results in this situation
. then we turn to crossing probabilities in strongly anisotropic systems . to our knowledge ,
such problems have not been considered before , at least with the appropriate anisotropic scaling analysis .
we start with the random walk with absorbing boundary conditions in the continuum limit which can be considered as a linearized mean - field approximation @xcite for directed percolation @xcite and illustrates on an exactly solvable example the behaviour of the crossing probability in strongly anisotropic systems .
finally we present the results of monte carlo simulations for the crossing probability in directed percolation , a strongly anisotropic problem which has so far defied all attempts at an exact solution , even in two dimensions .
we consider the random bond percolation on a square lattice with anisotropic bond occupation probabilities .
the system is rectangular and the bond probability is @xmath15 ( @xmath16 ) in the direction parallel to the side with length @xmath6 ( @xmath7 ) .
the correlation lengths , @xmath17 , @xmath18 , diverge along the critical line where @xmath19 with the same exponent @xmath20 for both directions but different amplitudes @xmath21 and @xmath22 .
the isotropy can be restored @xcite by rescaling the lattice parameter anisotropically so that @xmath23 .
the aspect ratio @xmath24 is then changed into an effective one @xmath25 in the second expression the effective aspect ratio involves the lengths measured in units of the corresponding correlation lengths .
the critical amplitude ratio @xmath26 , which is non - universal , is known for the @xmath12-state potts model on the square lattice @xcite . in the percolation limit , @xmath11 ,
one obtains : @xmath27 which solves the problem for a weak anisotropy .
when the anisotropy axis is not parallel to one edge of the rectangle , the rescaling leads to an isotropic system with the shape of a parallelogram .
standard conformal methods do not apply in strongly anisotropic systems .
the simplest example of such a system is the random walk in @xmath0 dimensions , which may be considered as a directed walk in space - time . in the continuum limit
, the probability density @xmath28 satisfies the diffusion equation @xmath29 we assume a uniform initial probability density @xmath30 and absorbing boundary conditions at @xmath31 and @xmath32 .
the solution is then obtained as the following eigenvalue expansion : @xmath33 \exp\left[-\frac{(2p+1)^2\pi^2dt}{l^2}\right]\ , .
\label{e - p}\ ] ] on the rectangle @xmath34 , the crossing probability in the time direction is given by : @xmath35\ , .
\label{e - cross - rw}\ ] ] thus the crossing probability shown in fig .
[ f.1 ] is a universal function of the scaled variable @xmath36 , product of the aspect ratio @xmath37 , where @xmath38 is the dynamical exponent for the random walk , by the non - universal inverse diffusion constant @xmath39 . in strongly anisotropic systems ,
such a behaviour generally follows from anisotropic scaling @xcite . under the anisotropic scale transformation , @xmath40 , @xmath41
, the crossing probability behaves as : @xmath42 for a dimensionless crossing probability , @xmath43 , one obtains the scale - invariant behaviour mentioned above with @xmath44 . the probability density for a walker starting at @xmath19 with uniform probability on the segment @xmath45 $ ] to die at time @xmath46 is given by @xmath47 = \frac{4d}{l^2}\,\vartheta_2\left [ \exp\left(-\frac{4\pi^2dt}{l^2}\right)\right]\ , , \label{e - surv - rw}\ ] ] where @xmath48 is a jacobi theta function .
since the linearized mean - field approximation for directed percolation @xcite is also governed by the diffusion equation ( [ e - dif ] ) , the mean - field crossing probability for directed percolation is given by ( [ e - cross - rw ] ) when the aspect ratio is sufficiently small .
the critical crossing probability in the time direction , @xmath49 , has been studied via monte carlo simulations for bond and site directed percolation on the square lattice , with the time axis along the diagonal direction .
space and time coordinates take integer and half - integer values alternatively on successive spatial rows and @xmath50 .
there are two directed bonds leaving a site at @xmath51 and terminating on the sites at @xmath52 .
sites may be wet or dry and a directed percolation cluster is a collection of connected wet sites starting from some source at @xmath19 @xcite . in the bond problem ,
bonds are open with probability @xmath53 .
a site is wet at time @xmath54 when it is connected through an open bond to a site which was wet at time @xmath46 . in the site problem
, sites are occupied with probability @xmath53 and all the bonds are open .
a site is wet at time @xmath54 when it is occupied and connected to a site which was wet at time @xmath46 .
a _ scale - invariant crossing probability _ is obtained when all the sites are wet in the initial state at @xmath19 .
the critical crossing probability was calculated for @xmath55 samples with sizes @xmath56 to @xmath57 and different values of the aspect ratio , @xmath37 , ranging from @xmath58 to @xmath59 .
the percolation thresholds @xmath60 , @xmath61 and the dynamical exponent @xmath62 were taken from ref .
either periodic boundary conditions ( pbc ) or free boundary conditions ( fbc ) were used at @xmath63 and @xmath32 .
the samples were grouped into 20 packets in order to evaluate the errors on @xmath49 .
the raw data @xmath49 versus @xmath64 are shown in fig .
2a for the largest sizes .
the statistical errors are actually smaller than the size of the symbols .
the scale invariance is manifest and finite size corrections to the leading behaviour are quite weak .
there is a shift in @xmath64 between the curves for site and bond percolation which is the same for pbc and fbc . as shown in fig .
2b , an universal curve is obtained for each type of boundary condition after the curves corresponding to the site problem have been shifted by the same amount . in this strongly anisotropic problem ,
the correlation lengths , @xmath65 in the space direction and @xmath66 in the time direction , diverge as @xmath67 thus , generalizing the behaviour for the weakly anisotropic system in eq .
( [ e - reff ] ) , for each type of boundary conditions , @xmath68 or @xmath69 , one expects the crossing probability @xmath49 to be an universal function @xmath70 of the scale - invariant ratio @xmath71 in agreement with anisotropic scaling in eq .
( [ e - scaling ] ) . as above for weakly anisotropic percolation and for the random - walk ,
the constant @xmath4 is non - universal and takes different values , @xmath72 and @xmath73 , for the site and bond directed percolation on the diagonal square lattice . for a given type of boundary conditions ,
the crossing probability @xmath49 is the same for couples of values of the aspect ratio @xmath74 and @xmath75 such that the arguments of the universal function @xmath70 are the same for the site and bond problems .
this happens when @xmath76 .
thus , with a logarithmic scale , the two curves have a relative shift @xmath77 this shift was estimated for the different sizes for 6 points in the ascending part of the the curves .
since the points for the site and bond problems correspond to slightly different ordinates , a direct estimation of the shift was not possible .
instead we used a polynomial interpolation for the curve corresponding to the bond problem and the value of the abscissa associated with the common value of the ordinate was determined through the tangent method .
the results for the ratio of non - universal amplitudes as a function of the inverse size are shown in fig .
[ f.3 ] for pbc and fbc .
they converge from above and from below to the common asymptotic value @xmath78 corresponding to the shift @xmath79 which was used to obtain the data collapse in fig .
the same universal crossing probability is obtained for any finite homogeneous density of wet sites in the initial state at @xmath19 although the finite - size corrections to the leading scale - invariant behaviour is stronger at large values of @xmath24 ( small @xmath46 ) when the initial density decreases . with a single wet site at @xmath80 in the initial state , i.e. , for an asymptotically vanishing initial density , the crossing probability remains a function of @xmath81 but the scale - invariance is lost , @xmath49 decaying as @xmath82 .
the probability @xmath83 to find @xmath13 clusters crossing the system in the time direction at @xmath5 has been also studied .
details will be given elsewhere .
this work can be extended by considering the crossing probability on same - spin clusters in strongly anisotropic spin systems , like the three - dimensional uniaxial annni model at the lifshitz point , for which presumably exact results for the form of the correlation functions , following from a generalization of local scale invariance @xcite , have been obtained recently @xcite .
|
scale - invariant universal crossing probabilities are studied for critical anisotropic systems in two dimensions . for weakly anisotropic standard percolation in a rectangular - shaped system , cardy s exact formula
is generalized using a length - rescaling procedure . for strongly anisotropic systems in @xmath0 dimensions ,
exact results are obtained for the random walk with absorbing boundary conditions , which can be considered as a linearized mean - field approximation for directed percolation .
the bond and site directed percolation problem is itself studied numerically via monte carlo simulations on the diagonal square lattice with either free or periodic boundary conditions .
a scale - invariant critical crossing probability is still obtained , which is a universal function of the effective aspect ratio @xmath1 where @xmath2 , @xmath3 is the dynamical exponent and @xmath4 is a non - universal amplitude .
|
the poplmark challenge is a set of common programming language problems meant to test the utility of modern proof assistants and techniques for mechanized metatheory . in response to this challenge
, significant strides have been made in making it easier to mechanize the metatheory of programming languages , especially regarding variable binding @xcite .
however , little progress has been made in the direction of modularity : it is still difficult to separately develop the definitions and meta - theory of language fragments and then link the fragments together to obtain the definitions and meta - theory for a language composed of such fragments .
dependent types have formed the foundation of a broad and rich range of type systems that allow values and types to be freely mixed .
programmers can express propositions as types viewed as sets , and proofs as objects viewed as inhabitants of those sets .
this style of theorem proving suggests the use of familiar engineering abstractions as general solutions to questions about theorem proving . rather than relying on semi - automated proof search such as coq s ltac we propose a method of proof composition using simple abstractions whereby components are defined piecewise and `` tied '' together at the end using a wrapper datatype acting as a tagged union .
the method of language definition used is iterative .
components are defined separately from one another and are composable along with their proofs .
thus we would like for separate language designers to be able to reuse one anothers work without the need for sophisticated proof search algorithms or with effort spent copying and pasting terms .
the language we present is one of simple expressons using agda as the implementation language and proof assistant .
we begin by defining a series of language syntaxes for sums , options , and arrays .
we chose to include arrays because they not only can result in runtime errors requiring the inclusion of the @xmath0 type but like addition , they use the natural numbers , forcing consideration of how value types can be shared across otherwise isolated components .
we continue by defining evaluation semantics and typing rules . the language is defined piecewise , each component is built in isolation alongside a proof of type preservation .
we conclude with a presentation of how these components can be composed and a proof of type preservation for the combined language can be immediately derived from the componentwise proofs .
the motivation for our technique is drawn from a solution to the expression problem where languages are defined as the disjoint sum of smaller languages by removing explicit recursion .
we show that this idea can be recast from types and terms , to proofs .
when modeling a problem with a functional flavor often the natural solution emerges as several recursive cases handled by some helper functions .
the expression problem states that this type of solution presents us with a choice : we may ordain our data structure forever unchanging , making it easy to add new functions without changing the program ; or we may leave our data structure open , making it difficult to extend the original program with new functions . while many solutions to the expression problem have been proposed over the years ,
here we make use of the method described by malcom @xcite which generalizes recursion operators such as fold , from lists to polynomial types .
the problem we encounter arises as a result of algebraic data types being _ closed _ : once the type has been declared , no new constructors for the type may be added without amending the original declaration and the solution presented lies at the heart of our work .
the idea is simply to remove immediate recursion and split a monolithic datatype into components to be later collected under the umbrella of a tagged union . throughout this paper
we will work with a simple evaluator over natural numbers and basic arithmetic operators ; in agda we might first consider + + this definition has the advantage of being direct and simple , however a problem lies within the explicit recursion ; notice that when later extending expressions with arrays and option types we can make no reuse of expr+ due to the closed nature of algebraic data types . to extend expr+
we must define a whole new data type , as in the following definition of monolithicexpr .
+ + + + + + + 0.06em + + + + + ( ) + ( ) suppose instead we begin with polymorphic definitions such as the following .
( ) + + ( ) + + + 0.06em + ( ) + + + ( ) + we then introduce recursion as follows , combining components as a disjoint sum , written @xmath1 in agda .
+ + + + + more generally , this type of data can be captured using a `` categorical approach '' where recursion is introduced as the fixed point of a functor : ( ) + ( ) + ( ) + + + it is easy to see that this new type is equivalent to monolithicexpr up to isomorphism @xmath2 \ ; \textpipe \ ; -!!- \ ; \textpipe \ ; -[-]:=- \\\end{aligned}\ ] ] the functor @xmath3 , passed into @xmath4 above , serves as the key abstraction allowing us to represent expressions as least fixed points .
functors are a special mapping defined over both types and functions satisfying the so called _ functor laws _ ; a functor f 1
. assigns to each type @xmath5 , a type @xmath6 2 .
assigns to each function @xmath7 , a function @xmath8 such that 1 .
identity is preserved : @xmath9 , and 2 . when @xmath10 is defined : @xmath11 .
one familiar example is the @xmath12 functor mapping each type @xmath5 to @xmath13 and each function @xmath7 to @xmath14 which applies @xmath15 to each element of a list .
here we define the least fixed point over a restricted class of functors called the _
polynomial functors_. polynomial functors are a subset roughly equivalent to the more familiar algebraic polynomials , @xmath16 where addition is disjoint sum and multiplication is cartesian product . in agda
, ulf norell@xcite expresses this class as a datatype @xmath17 along with an interpretation as a set @xmath18 $ ] + + + + + + + [ 1.5mu 0.06em 1.5mu ] + [ 1.5mu 1.5mu ] + [ 1.5mu 1.5mu ] + [ 1.5mu 1.5mu ] + [ 1.5mu 1.5mu ] with least fixed point ( ) + ( ) then to reexpress @xmath19 as a polynomial functor we use sum @xmath20 to define cases within a type , and product @xmath21 to represent arguments of a particular case + + + + + + + + + unfolding @xmath22 yields the same value calculated above as we should hope ! + + ( ) + ( ) + ( ) + ( ) + + ( ) + ( ) ( ) + ( ) ( ) ( ) ( ) ( ) what do values in @xmath22 look like ?
written directly they appear nonsensical , consider @xmath23 + ( ( ( + ( ( ( ( ) ) ) ) + , ( ( ( ( ) ) ) ) ) ) ) notice here the role that the injections and @xmath24 functions play .
traditionally we would provide a unique name for each branch in an algebraic datatype , however here we only have two names @xmath25 and @xmath26 so we instead rely on nesting to create unique prefixes .
once we have tagged a value we must give it a well known type so that parent expressions can expect a common child type , this is the role of @xmath24 .
although cumbersome we can hide much of this complexity provided the right abstractions + + + + + ( ( ( , ) ) )
we are now ready to define a simple language and its operational semantics .
the language is small including just sums , an option type , and an array with assignment and lookup . in agda
, the unit type is written @xmath27 and has only one member : @xmath28 .
@xmath27 is used to represent constructors that take no arguments such as @xmath29 , the empty list .
+ + + + + + + + + what do each of these definitions mean ? the maybe type has two constructors : some , which wraps a single expression ; and none taking no arguments .
we define more descriptive constructors for tagging these two types of values + ( ( ( ( ) ) ) ) + + giving a convenient constructor for @xmath30 is similarly straightforward + + + ( ( ( , ) ) ) and to define arrays we have assignment taking an array , an index , and a value to assign at that index ; nil , the empty array ; and lookup which accepts an array and an index 0.06em + ( ( ( ( , , ) ) ) ) + + ( ( ( ) ) ) + + ( ( ( , ) ) ) so far the definition of our syntax has used fairly standard techniques but we have failed to give any sort of meaning to these expressions . we first define a monolithic static and dynamic semantics for this language , then show how to modularize their definition later in this section .
figure [ fig : value typing rules ] defines a simple set of typing rules using metavariables @xmath31 to range over experssions and @xmath32 to range over values ; figure [ fig : evaluation semantics ] gives a small step operational semantics .
c c c & + + & + [ fig : language definition ] while agda is expressive enough to implement these rules , directly and indeed they are nearly a direct reflection of that implementation , recall that our goal is to create several independant languages each carrying their own semantics .
we begin by defining monolithic semantics for @xmath19 and proceed to determine points of failure and to dissect the definition into independant constituents . to simplify things we define our notion of @xmath33 as a closed adt + + + and here is the definition of the monolithic type system and evaluation relation in agda .
+ \{1.5mu 1.5mu } ( ) + \{1.5mu 1.5mu } + + ( ) + + \{1.5mu 1.5mu } + + + ( ) + \{1.5mu 1.5mu } + + + + ( ) + + + \{1.5mu 1.5mu } + + + \{1.5mu 1.5mu}\{1.5mu 1.5mu } + + + \{1.5mu 1.5mu } + ( ) + \{1.5mu 1.5mu } + + + \{1.5mu 1.5mu } + , the function @xmath34 is the lookup function that evaluates to @xmath35 when @xmath36 has been defined and @xmath37 otherwise . notice that we currently do not restrict the values of @xmath32 enough in the @xmath38 rule ; our typing rules require that n be a value while in agda we have only required it be an expression .
some notion of value is needed and a common solution is to add a tag @xmath39 to the @xmath40 type and pattern match ; here @xmath39 is called @xmath41 $ ] and in a dependantly typed context we might then define a predicate over @xmath39 .
however because the sum type has only one type of value , a number , it is simpler to use @xmath42 directly .
this method for defining semantics is common with the advantage of being direct and concise , but similar to our first implementation of @xmath43 and @xmath44 above : there is no simple mechanism for code reuse .
the answer is again to delay recursion . in order to modularize the evaluation rules we define a separate step relation for each functor making up our @xmath19 type .
first note that @xmath45 does nt make use of _ how _ the step from @xmath46 to @xmath47 occurs so we can factor this top - level relation \{1.5mu 1.5mu } + + \{1.5mu 1.5mu } + + \{1.5mu 1.5mu}\{1.5mu 1.5mu } + + + \{1.5mu 1.5mu } + ( ) while this is better there is still an undesirable reference to the datatype @xmath19 .
applying the same factorization here to the underlying functor requires parametrization by two extra coercion functions , these are the @xmath45 and @xmath42 functions defined previously . the new names @xmath48 and @xmath49 used here are meant to imply that a subtype is being `` lifted '' into its supertype \{1.5mu 1.5mu}\{1.5mu 1.5mu } + \{1.5mu ( ) 1.5mu}\{1.5mu 1.5mu } + + \{1.5mu 1.5mu } + ( , ) ( , ) + \{1.5mu 1.5mu}\{1.5mu 1.5mu } + + ( , ) ( , ) + \{1.5mu 1.5mu } + ( , ) ( ) unfortunately this definition falls short too . when we lift terms into the expression type @xmath50 , agda `` forgets '' the constituents @xmath46 and @xmath47in turn we lose the ability to reason about these distinct components of the sums @xmath51 and @xmath52 .
this later becomes a problem when , for example , attempting to abstract the welltyping relation .
an intelligent human can peel away @xmath48 and see that the terms @xmath46 and @xmath47 in @xmath53 and @xmath54 are the same because @xmath48 is _
injective_. however agda is unconvinced , and rightfully so , for it does not require a particularly great deal of ingenuity to find a counterexample , consider taking @xmath55 so that @xmath56 + ( , ) the problem is that our abstraction is too general .
what we require is a proof that @xmath57(\mu e)$ ] and @xmath58 are subtypes of the top - level expression datatype @xmath50 .
the solution to the problem is drawn from the notion of a categorical subobject .
we proceed by delaying application of injections and view the objects as injectable , existential terms .
the importance of this approach is two - fold : firstly this allows us to take inverses of lift functions while we are secondly able to retain the perspective of operating on a single type @xmath50 .
a subobject of a type @xmath59 is a left invertible function with codomain @xmath59 , @xmath60 .
being restricted to polynomial functors , we know that all our subobjects @xmath61 will be some composition of @xmath24 , @xmath25 and @xmath26 so a proof that @xmath62 is a subtype of @xmath50 is merely a description of which direction to move at each point in a disjoin sum + + \{1.5mu 1.5mu } + + \{1.5mu 1.5mu } + + \{1.5mu 1.5mu } + now we can define containment on a functor s interpretation as a set + + \{1.5mu 1.5mu } + ( ) ( ) with conversion functions defined as \{1.5mu 1.5mu } ( ) + + ( ) + ( ) + \{1.5mu 1.5mu } ( ) + ( ) recall the two goals we had in mind .
we first wished to take the inverse of a lift function to gain access to its arguments , in the case of @xmath30 these were @xmath46 and @xmath47 . by representing an injection as a delayed application of a subobject because the constructor s arguments are stored as a part of the coercion
finding left inverses will become a trivial case of pattern matching . to delay function application allowing agda to effectively peel away the @xmath63 functions we define a @xmath64 datatype from type @xmath5 to @xmath65 representing the _ intention _ of coercing an object @xmath66 while treating it at the type - level as @xmath65 .
a lazy coercion is then an injection @xmath67 along with an object in @xmath5 + \{1.5mu 1.5mu } ( ) + \{1.5mu 1.5mu } + ( ) our second goal was to operate on objects of a single type .
why is this the case ? recall that the type of our step relation is indexed by two expressions : @xmath68 .
we should expect the same of the final abstraction over step relations because it can not easily name the underlying type of its indexing expressions .
instead we have packaged the indices as _
existentials _ which are viewed as the type @xmath65 .
we seem to be close to a modular step relation @xmath54 , defining at each point another level of abstraction to delay immediate application . to modularize datatypes , recursion is delayed and types are viewed as polynomial functors , then to modularize step relations , evaluation is parametrized and expression upcasts are delayed by viewing them as an intention .
attempting again to define a step relation for addition we find very little has changed \{1.5mu 1.5mu } + \{1.5mu 1.5mu } + \{1.5mu ( ) 1.5mu } + \{1.5mu 1.5mu } + ( ) ( ) + + \{1.5mu 1.5mu } + + ( , ) ( , ) + \{1.5mu 1.5mu } + + ( , ) ( , ) + \{1.5mu 1.5mu } + ( , ) + ( ) it appears we ve littered an otherwise simple definition with @xmath69 but we ve replaced our arbitrary arrows with objects having constructors we can match on . using the above techniques we can modularize the welltyping relation over sums for free \{1.5mu 1.5mu } + \{1.5mu 1.5mu } + \{1.5mu ( ) 1.5mu } + ( ) + \{1.5mu 1.5mu } + + ( ( , ) ) the above definitions nearly wrote themselves .
the simplicity comes from the fact we are just abstracting as many terms as possible , keeping in mind we can fill them in naturally later because the abstraction is so general there are few options available .
we proceed by defining the step and welltypedness relations on arrays that can be combined with the relations on sums .
the definitions for evaluation and welltypedness should look similar to those for sums @xmath54 .
0.06em \{1.5mu 1.5mu } + \{1.5mu 1.5mu } + \{1.5mu ( ) 1.5mu } + \{1.5mu 1.5mu } + \{1.5mu ( ) 1.5mu } + ( ) ( ) + + \{1.5mu 1.5mu } + ( ) ( ) + \{1.5mu ( ) 1.5mu}\{1.5mu 1.5mu } + ( ) + , to define the typing relation we again follow the format of @xmath70 above and we are done .
\{1.5mu 1.5mu } + \{1.5mu 1.5mu } + \{1.5mu ( ) ( ) 1.5mu } + \{1.5mu 1.5mu } + ( ) + ( ) + \{1.5mu 1.5mu } + + ( ( ) ) + \{1.5mu 1.5mu } + + ( ( ) )
the type preservation lemma states that if a term is welltyped and can step , then the type of the term is preserved after evaluation @xmath71 prior to considering how type preservation might look for each of the previously defined components we should review what type preservation looks like for the @xmath44 language .
the proof is standard , proceeding by structural induction on the shape of the welltyping tree .
\{1.5mu 1.5mu}\{1.5mu 1.5mu } + + + + ( ) ( ) + ( ) + ( ) ( ) + ( ) + + ( \{1.5mu 1.5mu}\{1.5mu 1.5mu } ) ( ) + ( ) + ( ) ( ) + ( ) + + ( \{1.5mu 1.5mu}\{1.5mu 1.5mu } ) ( ) + , there are three items worth noting here : the first is the use of the function @xmath72 which we have assumed produces a pair with first component an expression and second component a proof that the expression is a welltyped option ; the second is that recursion acts as our induction hypothesis ; and finally that agda is smart enough to notice there is only a single possible welltyping constructor for each step constructor in agda all functions are total .
we should expect the modular type preservation lemmas to look similar because there is little global knowledge involved .
the induction hypothesis and values aside , each case is `` contained within its own world '' in the sense that each evaluation rule relies only on the fact that subterms are welltyped but ignoring the _ reason _ they are welltyped . to show type preservation for sums we might start with \{1.5mu 1.5mu}\{1.5mu 1.5mu } + \{1.5mu ( ) 1.5mu }
+ + + + + ( \{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu } ) ( ) however recall that @xmath54 requires the top - level step relation and proof that @xmath73 contains both sums and naturals .
there is a second mistake in writing preservation this way we would like to show that @xmath74 is welltyped in the expression language , not just necessarily in the modular sum language , this reflects our desire to expose as little about each component as possible . a second formulation might then begin as follows but we again fail .
\{1.5mu 1.5mu } + \{1.5mu 1.5mu } + \{1.5mu 1.5mu } + \{1.5mu ( ) 1.5mu } + \{1.5mu 1.5mu } + \{1.5mu 1.5mu } + \{1.5mu ( ) 1.5mu } + \{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu } + \{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu } + ( ) + ( ) ( ) + ( ) + ( ) ( ) + ( ) + ( ) + ( ) it seems we re only missing two pieces : we need to be able to lift welltyped sums and naturals into @xmath75 ; and we need some way of expressing the induction hypothesis which states that because @xmath46 is welltyped and stepped , @xmath76 is welltyped too .
the induction hypothesis is slightly stranger than was the case in our @xmath44 s because we know @xmath46 and @xmath76 are welltyped despite the fact that they are any expressions , not necessarily just sums .
this motivates our solution which takes the induction hypothesis as an explicit assumption .
\{1.5mu 1.5mu } + \{1.5mu 1.5mu } + \{1.5mu 1.5mu } + \{1.5mu ( ) 1.5mu } + \{1.5mu 1.5mu } + \{1.5mu 1.5mu } + \{1.5mu ( ) 1.5mu } + ( ( ) ( ) ) + ( \{1.5mu 1.5mu}\{1.5mu 1.5mu } + \{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu } ( ) + ( ) ) + ( \{1.5mu 1.5mu}\{1.5mu 1.5mu } ) + \{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu } + \{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu } + ( ) + + ( ) ( ) + ( ( ) ) + + ( ) ( ) + ( ( ) ) + + ( \{1.5mu 1.5mu}\{1.5mu 1.5mu } ) ( ) + ( ) we are pleased with how similar this is to the original , monolithic formulation .
notice again that the solution was to factor out assumptions about the outside world similar to the previous abstractions .
proving type preservation for arrays is similarly natural : \{1.5mu 1.5mu } + \{1.5mu 1.5mu } + \{1.5mu 1.5mu } + \{1.5mu ( ) ( ) 1.5mu } + \{1.5mu 1.5mu } + \{1.5mu ( ) ( ) 1.5mu } + \{1.5mu 1.5mu } + \{1.5mu ( ) 1.5mu } + ( ( ( ) ) ( ) ) + ( \{1.5mu 1.5mu}\{1.5mu 1.5mu } + \{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu } ( ) + ( ) ) + ( \{1.5mu 1.5mu}\{1.5mu 1.5mu } ) + 0.06em \{1.5mu
1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu } + \{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu } + ( ) + + ( ) ( ) + ( ( ) ) + + ( \{1.5mu 1.5mu}\{1.5mu 1.5mu } ) ( ) + , it would seem we re nearly done and the final pieces should be entirely guided by the selected abstractions .
the @xmath77 functions each have a unique solution : + ( ( ( ) ) ) + + ( ( ( ) ) ) + + ( ( ( ) ) ) + + ( ) but how should we define welltypedness for @xmath40 ? again
the notion of what it means to be welltyped has already been defined and we simply need to `` tie the knot '' as @xmath78 did above + ( ) ( ) + ( ) + ( ) + \{1.5mu 1.5mu}\{1.5mu 1.5mu } + \{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu } ( ) + ( ) + \{1.5mu 1.5mu}\{1.5mu 1.5mu } + \{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu } + ( ) + ( ) to define a step relation on @xmath40 , @xmath53 we provide a similar wrapping for each language component + \{1.5mu 1.5mu}\{1.5mu 1.5mu } + \{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu } + ( ) + + \{1.5mu 1.5mu}\{1.5mu 1.5mu } + 0.06em + \{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu}\{1.5mu 1.5mu } + ( ) + the only piece remaining is to prove type preservation .
we begin in the same way we have for each of the previous proofs using the step relation s constructors as a guide .
the type signature should not have changed \{1.5mu 1.5mu}\{1.5mu 1.5mu } + and there are two cases @xmath79 and @xmath80 $ ] ; moreover we should expect to merely apply preservation- * to each case , supplying the necessary lift functions and the induction hypothesis .
this is indeed the case : ( ) ( ) + + + ( ) ( ) + + having shown type preservation it is interesting to see the similarity between how terms are shown to be welltyped and to evaluate and how the terms are expressed in @xmath81 .
recall that each term in @xmath40 is wrapped by a tag given by @xmath25 and @xmath26and the constructor @xmath24 plays the role of recursion . to reiterate
consider the convenience functions , + ( ( ( ) ) ) + + ( ( ( ) ) ) + 0.06em + ( ) + + ( ) + + ( , ) we may then ask : why is the term + ( ) ( ) welltyped ? the answer given by @xmath82 is + ( ) + + ( ) + + ( ( ) + ( ) ( ) ) + ( ) + ( + ( ) ( ) ) the @xmath83 functions play the same role in @xmath84 as @xmath24 does in @xmath40 ; however rather than using the generalized approach of a series of disjoint sums we bundle the tag and recursion into a single constructor for each language component .
evaluation displays a similar symmetry ( ) ( ) + ( ) + ( ( ) ) what does the proof that @xmath85= { \textrm{nat}\;}1)\,!e\,{\textrm{nat}\;}1 $ ] is welltyped look like ?
we can compute it by invoking which evaluates to + ( + ( + ( ( ) ( ) ( ) ) ) + ( ) )
independent and concurrently with our work , delaware , et al .
@xcite developed a solution to moduler meta - theory in coq .
both their approach and ours relies on the principle of representing data types as functors ; however they have chosen to express inductive types using church encodings and recursive evaluation using mendler algebras , which requires some extra sophistication . here
we express types as data members of the family of polynomial functors and apply recursive evaluation directly .
their approach presented is further along and has shown the important level of robustness required by most languages while there are more unanswered questions regarding the method presented here .
we should ask if we have accomplished the goal that we set out with .
the language @xmath40 was given componentwise and the boiler - plate necessary to wrap each welltyping and step relation is minimal .
the proof of type preservation was almost immediate , requiring only an invocation of previously defined proofs for each component .
moreover there is no copy and paste necessary and the repetitive components should be automatically producable given a sophisticated macro system where terms can be inspected by name
set equality is non - deterministic rather than value . using agda as a proof language ,
although convenient , leaves the question of consistency open .
we regard this as a minor problem and hope that our implementation would port to coq .
a more pertinent problem is the definition of @xmath86 for @xmath40agda is unable to prove termination and we plan to address this soon .
the language presented is quite simple , unable to express even euclid s algorithm , and the method of polynomial functor s used to express @xmath19 precludes the possibility of first class function types which are critical for functional programming .
various solutions to this problem have been proposed @xcite and the area of recursion schemes is rich @xcite .
a real world language calls for much heavier sophistication , but the ideas presented here are new and their reach is open to question and requires further exploration . erik meijer and graham hutton . _ bananas in space : extending fold and unfold to exponential types . _ in proceedings of the seventh international conference on functional programming languages and computer architecture ( fpca 95 ) .
acm , new york , ny , usa , 324 - 333 .
|
while methods of code abstraction and reuse are widespread and well researched , methods of proof abstraction and reuse are still emerging .
we consider the use of dependent types for this purpose , introducing a completely mechanical approach to proof composition .
we show that common techniques for abstracting algorithms over data structures naturally translate to abstractions over proofs .
we first introduce a language composed of a series of smaller language components tied together by standard techniques from malcom @xcite .
we proceed by giving proofs of type preservation for each language component and show that the basic ideas used in composing the syntactic data structures can be applied to their semantics as well .
|
usual approach in the area of quantum games meyerdavid , ewl , multiplayer , du , cheon , flitney , shimamura , iqbal , ichikawa consists of analyzing a quantum system manoeuvred by participating agents , recognized as players , who possess necessary means for their actions on parts of the system .
the quantum system evolves to its final state and players payoffs , or utilities , mathematically expressed as expectation values of self - adjoint payoff operators , are generated from quantum measurement @xcite .
thus the usual constructions of quantum games involve the concepts of quantum state vectors , entangled states , quantum measurement , expectation values , trace operation , and that of density operators etc .
this may seem normal because as being part of the research field of quantum computation @xcite quantum games are expected to exploit relevant tools from quantum mechanics .
however , in our experience , this noticeable reliance of the models of quantum games on the tools of quantum mechanics also succeeds to keep many readers away from this inter - disciplinary area of research .
ideally , they would like to see genuine quantum games constructed from elementary probabilistic concepts , as it is the case with many examples in game theory @xcite .
we find this situation as an opportunity to present a probabilistic approach in which quantum games are constructed without referring to the tools of quantum mechanics .
while looking for the possibility of such an approach , it is encouraging to find that the most unusual character of quantum mechanics can be expressed in terms of probabilities @xcite only .
for example , bell inequalities @xcite can be written in terms of constraints on joint probabilities relevant to pairs of certain random variables .
as probabilities are central to usual analyses in game theory , it seems natural to use the peculiar probabilities , responsible for the violation of bell inequalities , to construct quantum games .
we , therefore , suggest to construct quantum games from the probabilities arising in the einstein - podolsky - rosen ( epr ) experiments @xcite performed to test the violation of bell inequalities .
the most unusual character of the epr probabilities , that they may not be factorizable , motivates us , in this paper , to find how non - factorizability can be used to construct quantum games . in other words , we search for the role of non - factorizable probabilities in game - theoretic solution concepts when epr experiments provide the sets of non - factorizable probabilities .
this explicitly probabilistic approach towards quantum games is expected to be of interest to the readers from such areas as economics @xcite and mathematical biology @xcite , where game theory finds extensive applications and the tools of quantum mechanics are found to be rather alien . secondly , because of its exclusively probabilistic content , this approach promises to provide a unified perspective for both the classical and the quantum games .
the rest of the paper is organized as follows .
section [ fvbi ] describes the role of factorizability in deriving bell inequality .
section [ tpgfps ] discusses playing two - player games using factorizable probabilities and presents two- and four - coin setups to play the well known games of prisoner s dilemma ( pd ) , stag hunt ( sh ) , and chicken @xcite .
section [ tpgueprs ] describes playing two - player games with epr experiments .
section [ tpgunfps ] develops a framework in which factorizable probabilities lead to the classical game whereas non - factorizable probabilities result in the quantum game .
section discussion discusses the results and presents a view for further work .
factorizability is known to be an interesting property of coupled systems with separated parts , saying @xcite that for such systems the probability for a simultaneous pair of outcomes can be expressed as the product of the probability for each outcome separately .
mathematically , this is expressed by writing joint probabilities as arithmetic product of their respective marginals @xcite .
it turns out to be the most important technical ingredient in the proof of bell inequality .
in fact , j. s. bell used factorizability in his mathematical formulation of locality @xcite . others @xcite have recognized it , for example , by such terms as conditional stochastic dependence .
we describe factorizability with reference to the standard epr setup aspect et al , peres used to derive bell inequality .
this setup consists of two spatially - separated participants , known as alice and bob , who share two - particle systems emitted by the same source .
we specify the state of a coupled system by @xmath0 .
we denote alice s parameter by @xmath1 that can be set either at @xmath2 or at @xmath3 and denote bob s parameter by @xmath4 that can be set either at @xmath5 or at @xmath6 . in a run ,
alice sets her apparatus either at @xmath2 or at @xmath3 and , in either case , on receiving her particle she makes a measurement , the outcome of which is @xmath7 that is either @xmath8 or @xmath9 . in the same run , bob sets his apparatus either at @xmath10 or at @xmath6 and , in either case , on receiving his particle he makes a measurement , the outcome of which is @xmath11 that can be either @xmath8 or @xmath9 .
alice and bob record the outcomes of their measurements for many runs as they receive two - particle systems emitted from the same source .
we denote the probability that alice obtains the outcome @xmath12 or @xmath13 by @xmath14 and , similarly , we denote the probability that bob obtains the outcome @xmath15 or @xmath9 by @xmath16 .
also , we denote the probability that alice and bob obtain the outcomes @xmath7 and @xmath11 , respectively , by @xmath17 .
these outcomes result from their choices of the parameters @xmath1 and @xmath4 , i.e. which one of the four pairs @xmath18 @xmath19 @xmath20 @xmath21 is realized in a run . according to quantum theory ,
the outcomes @xmath7 and @xmath11 both are completely random and alice and bob can only find the probabilities to obtain @xmath8 or @xmath9 at the outcomes of their measurements . as it is described in the section [ tpgueprs ] , alice s and bob s parameters @xmath1 and @xmath4 decide these probabilities . in many runs , alice can choose between @xmath2 or @xmath3 with some probability .
similarly , in many runs , bob can choose between @xmath22 or @xmath6 with some probability .
assume that the source emits a total of @xmath23 two - particle systems .
we denote by @xmath24 the number of times alice gets the outcome @xmath7 when she may set her parameter @xmath1 either at @xmath2 or at @xmath3 .
similarly , we denote by @xmath25 the number of times bob gets the outcome @xmath11 when he may set his parameter @xmath4 either at @xmath26 or at @xmath6 . and
, we denote by @xmath27 the number of times when alice gets the outcome @xmath7 and bob gets the outcome @xmath11 , wherever they may set their parameters @xmath1 and @xmath4 , respectively . when @xmath23 is large , the ensemble probabilities are defined as @xmath28 for many runs we consider an ensemble of states emitted from the source that may not be the same . to allow mixture of states
we denote the normalized probability density , characterizing the ensemble of emissions , by @xmath29 .
the ensemble probabilities are @xmath30 where @xmath31 is the space of states @xmath0 .
now , factorizability states that @xmath32 that is , for each @xmath0 the joint probabilities are arithmetic product of their respective marginals .
factorizability being the most important technical ingredient in the proof of bell inequality can be seen as follows .
following inequalities must hold when probabilities are sensible quantities @xmath33 we now refer to a theorem stating that if six numbers @xmath34 , @xmath35 , @xmath36 , @xmath37 , @xmath38 , @xmath39 are given such that @xmath40 then the function @xmath41 is constrained by the inequalities @xmath42 proof of this theorem can , for example , be found in ref .
@xcite . inequalities ( [ factorizability ( a ) ] ) together with the inequalities ( [ factorizability ( c ) ] ) give @xmath43 for each @xmath0 .
the central role of factorizability follows as when the definition ( [ factorizability ( mathematical ) ] ) of factorizability holds , the multiplication of ( [ factorizability ( d ) ] ) by @xmath44 and the subsequent integration over @xmath0 gives @xmath45 which is the bell ( or clauser - horne ) inequality @xcite .
note that the definition ( [ factorizability ( mathematical ) ] ) is crucial in order to obtain the inequality ( [ chsh inequality ] ) .
to bring non - factorizability into the realm of two - player games , we consider a symmetric two - player , two - strategy , non - cooperative game @xcite represented by the matrices @xmath46 where all @xmath47 are real numbers .
players can go for one of the two available strategies : @xmath48 , @xmath49 for alice and @xmath50 , @xmath51 for bob .
as factorizability is central to obtain bell inequality , in this paper we construct quantum games from non - factorizable probabilities that exploit epr setup .
this rests on fine s view @xcite that the violation of bell inequality in epr experiments shows that quantum theory violates factorizability .
this view allows us to construct quantum games for which factorizability always corresponds to the classical game .
we recognize key features of an epr setup being that these relate to a probabilistic system divided into two parts such that a ) each observer has access to one part of the system b ) each observer can select between two available choices c ) observers can not communicate between themselves d ) observers can make independent selections between the available choices e ) probabilities relevant to each part of the system are normalized and that f ) probabilities are sensible quantities .
it is worth mentioning here that the experimental testing of bell inequality involves four correlation experiments that correspond to combining @xmath2 with @xmath5 , @xmath2 with @xmath6 , @xmath3 with @xmath26 , and @xmath3 with @xmath6 , respectively .
these experiments are mutually exclusive in the sense that for any given experiment alice has to select between @xmath2 and @xmath3 and bob has to select between @xmath5 and @xmath6 .
that is , alice ( bob ) can not go for @xmath2 @xmath52 and @xmath3 @xmath53 simultaneously because the corresponding observables are incompatible , and can not be measured simultaneously . whereas , in the above derivation of the bell inequality it is assumed that @xmath54 all have definite values which can be measured simultaneously in pairs
. the above mentioned features are remindful of coins which , if distributed between players , are found to have all the above mentioned properties . for coins factorizability
has a straightforward meaning in that the associated probabilities remain factorizable .
hence , we develop an analysis of two - player games with non - factorizable probabilities by first translating playing of three well known games in terms of the games played when players share coins .
it turns out that a version of this translation provides the right comparison with the probabilities involved in the epr experiments and opens the way to the next step i.e. to introduce non - factorizable probabilities into the playing of two - player games .
we now consider pairs of coins and use it to play a two - player game ( [ a and b matrices ] ) .
for example , this game can be played when each player receives a coin , head up , and ` to flip ' or to ` not to flip ' is a player s strategy .
both coins are then passed to a referee who rewards the players after observing the state of both coins .
assume @xmath2 ( to flip ) and @xmath3 ( not to flip ) are alice s strategies and @xmath5 ( to flip ) and @xmath6 ( not to flip ) are bob s strategies .
that is , with reference to the matrices ( [ a and b matrices ] ) , we make the association @xmath55 , @xmath56 , and @xmath57 , @xmath58 . in two - coin setup ,
we assume that the strategies @xmath2 and @xmath5 represent alice s and bob s actions ` to flip ' the coin , respectively ; and , similarly , @xmath3 and @xmath6 represent alice s and bob s actions ` not to flip ' the coin , respectively . in repeated runs of the game players
can play mixed strategies .
alice s mixed strategy @xmath59 @xmath60 $ ] is the probability to choose @xmath2 over @xmath3 and similarly bob s mixed strategy @xmath61 @xmath60 $ ] is the probability to choose @xmath5 over @xmath6 .
players payoffs are written as @xmath62 where @xmath63 is for transpose and the subscripts @xmath64 and @xmath65 refer to alice and bob , respectively .
the first and the second entries in a parentheses are alice s and bob s payoffs , respectively .
assume @xmath66 is a nash equilibrium ( ne ) @xcite : @xmath67 in rest of this paper we will use newhen we refer to either a nash equilibrium or to nash equilibria , assuming that the right meaning can be judged from the context .
we identify this arrangement to play a two - player game using two coins as the _ two - coin setup_. the game ( [ a and b matrices ] ) can also be played using four coins instead of two .
it is arranged by assigning two coins to each player before the game is played . in a run
each player has to choose one coin .
two coins out of four are , therefore , chosen by the players in each turn .
these coins are then passed to a referee who tosses them together and observes the outcome .
it is assumed that the players do not need to share fair coins .
we recall that in two - coin setup @xmath2 and @xmath5 are alice s and bob s strategies , respectively , that represent players actions ` to flip ' the coin that a player receives in a turn .
instead of flipping or not flipping , in four - coin setup a player s strategy is to choose one out of the two coins that are made available to each player in a turn .
the four - coin setup is relevant as , in a run , choosing a coin out of the two corresponds to choosing one of the two directions in which measurement is performed in standard epr experiment , the outcome of which is @xmath8 or @xmath9 . in repeated game ,
a player s strategy is defined by the selection she / he makes over several runs of the game .
for example , a player plays a pure strategy when she / he goes for the same coin over all the runs and plays a mixed strategy when she / he finds a probability to choose one coin out of the two over many runs .
referee rewards the players according to their strategies , the underlying statistics of four coins obtained from the outcomes of many tosses each one of which follows every time the two players choose two out of the total four coins , and the matrices ( [ a and b matrices ] ) representing the game being played .
we identify the arrangement using four coins to play a two - player game as the _ four - coin setup_. note that in four - coin setup the players rewards depend on the outcomes of repeated tosses even for pure strategies .
a large number of runs are , therefore , necessary whether a player plays a pure - strategy or a mixed - strategy .
four - coin setup provides an inherently probabilistic character to playing a two - player game and facilitates a probabilistic analysis when we seek to play the game ( [ a and b matrices ] ) using epr experiments .
as the four - coin setup uses a different definition of a strategy relative to the two - coin setup , we call @xmath2 and @xmath3 being alice s coins and @xmath26 and @xmath6 being bob s coins . when selecting a coin
is a player s strategy and we want to play the game given by the matrices ( [ a and b matrices ] ) , it is reasonable to make the association @xmath68 , @xmath56 , and @xmath69 , @xmath57 .
we represent the head of a coin by @xmath8 and its tail by @xmath9 and adapt this convention in the rest of this paper . for coins , alice s outcome of @xmath70 or @xmath9 (
whether she goes for @xmath2 or @xmath3 ) is independent from bob s outcome of @xmath15 or @xmath9 ( whether he goes for @xmath26 or @xmath6 ) and relevant joint probabilities are factorizable . referring to the definition ( [ factorizability ( mathematical ) ] ) of factorizability and noticing that probabilities associated to coins are factorizable , we use the same notation that is introduced in section fvbi to consider , for example , the probability @xmath71 that can be factorized as @xmath72 .
we define probabilities @xmath73 $ ] by @xmath74 and @xmath75 saying that @xmath76 is the probability of getting head for alice s first coin @xmath2 and @xmath77 is the probability of getting head for bob s first coin @xmath5 .
factorizability then allows us to write @xmath78 and @xmath79 where @xmath80 and @xmath81 i.e. @xmath82 and @xmath83 are the probabilities of getting head for alice s and bob s second coin , respectively . in four - coin setup
we find it useful to have the following table : @xmath84 from which we define payoff relations for the players : @xmath85 where @xmath86 for example , @xmath87 is alice s payoff when , in repeated runs of coin tossing , she always goes for her first coin , i.e. @xmath88 , while bob goes for his second coin , i.e. @xmath89 .
as it is the case with two - coin setup , alice s mixed strategy in four - coin setup is the probability with which she chooses her pure strategy @xmath2 over her other pure strategy @xmath3 during repeated runs of the experiment .
similarly , bob s mixed strategy is the probability with which he chooses his pure strategy @xmath5 over his other pure strategy @xmath90 during repeated runs of the experiment .
assume that alice plays @xmath2 with probability @xmath59 and bob plays @xmath5 with probability @xmath61 , their mixed - strategy payoff relations are @xmath91 the ne can then be found from ( [ ne factorable 1 ] ) , which is written as @xmath92 in the following , before we make a transition to playing our game using epr experiments , we consider playing three well known games using both the two- and the four - coin setups . we analyze the games of pd , sh , and chicken in two- and four - coin setups and afterwards make a transition to the epr setup .
pd is known to be a representative of the problems of social cooperation @xcite and has been one of the earliest @xcite and favorite topics for quantum games .
hence it is worthwhile to analyze this game in the setup using non - factorizable probabilities .
our second game is sh that , like pd , describes conflict between safety and social cooperation .
our third game is chicken , also known as the hawk - dove game @xcite , which is considered an influential model of conflict for two players in game theory .
pd is a noncooperative game @xcite that is widely known to economists , social and political scientists , and in recent years to quantum physicists .
it is one of the earliest games to be investigated in the quantum regime @xcite .
its name comes from the following situation : two criminals are arrested after having committed a crime together .
each suspect is placed in a separate cell and may choose between two strategies : _ to confess _ @xmath93 and _ not to confess _
@xmath94 , where @xmath95 and @xmath96 stand for cooperation and defection . if neither suspect confesses , i.e. @xmath97 , they go free , which is represented by @xmath98 units of payoff for each suspect .
when one prisoner confesses @xmath93 and the other does not @xmath94 , the prisoner who confesses gets @xmath99 units of payoff , which represents freedom as well as financial reward i.e. @xmath100 , while the prisoner who did not confess gets @xmath101 , represented by his ending up in the prison .
when both prisoners confess , i.e. @xmath102 , both are given a reduced term represented by @xmath23 units of payoff , where @xmath103 , but it is not so good as going free i.e. @xmath104 . referring to the matrices ( [ a and b matrices ] ) we make the association @xmath105 @xmath95 and @xmath106 @xmath96 and require that @xmath107 .
we define @xmath108 @xmath109 @xmath110 which requires @xmath111 for this game . in two - coin setup , the inequalities
( [ ne factorable 1 ] ) give @xmath112 and @xmath113 comes out as a unique ne at which players payoffs are @xmath114 . in the four - coin setup ,
the pd game as defined above is played as follows . using the mixed - strategy payoff relation ( [ coins payoff relations ] ) ,
the pair of pure strategies @xmath21 is represented by @xmath115 .
if we require this strategy pair to be a ne then we also need to know about the constraints this requirement imposes on @xmath116 .
when @xmath113 the ne inequalities ( [ ne factorable 2 ] ) for pd are reduced to @xmath117 now , for the ne inequalities ( [ ne generalized pd ] ) to hold , it is required that @xmath118 and @xmath119 both when @xmath120 and when @xmath121 .
this , of course , is possible if @xmath122 which must hold if the strategy pair @xmath113 is to be a ne in pd .
( [ s=0=s ] ) should be true along with that the probabilities @xmath123 are factorizable . as we find it , this result provides the basis on which the forthcoming argument for the quantum version of this game rests .
notice that , from ( [ bilinear payoffs ] ) , we obtain @xmath124 when eq .
( [ s=0=s ] ) holds .
the constraint ( [ s=0=s ] ) appears when the strategy pair @xmath125 is assumed to be the ne .
one can assume other strategy pair , for example @xmath126 , to be a ne for which , instead of the requirement ( [ s=0=s ] ) , we obtain @xmath127 however , it is found that this freedom does not affect the forthcoming argument for a quantum game . along with pd
, the game of sh provides another interesting context to study problems of social cooperation .
it describes the situation when two hunters can either jointly hunt a stag ( an adult deer that makes a large meal ) or individually hunt a rabbit ( which is tasty but makes substantially small meal ) .
hunting a stag is quite challenging and hunters need to cooperate with each other , especially , it is quite unlikely that a hunter hunts a stag alone . it is found that , in contrast to pd that has a single pure ne , the game of sh has three ne , two of which are pure and one is mixed .
the two pure ne correspond to the situations when both hunters hunt the stag as a team and when each hunts rabbit by himself .
sh differs from pd in that mutual cooperation gives maximum reward to the hunters . when compared to pd , sh is considered a better model for the problems of ( social ) cooperation .
referring to the matrices ( [ a and b matrices ] ) the game of sh is defined by @xmath128 in two - coin setup the ne inequalities for this game remain the same as the inequalities ( [ pd ne 2 coins ] ) except that now we have @xmath129 instead of the condition @xmath111 that hold for pd . here
@xmath130 are defined in the section [ pd(c ) ] .
this leads to three ne : @xmath131 and the corresponding payoffs at these equilibria , obtained from eqs .
( 2-coin payoff relations ) , are @xmath132 where we define @xmath133 .
now consider playing this game within the four - coin setup in which the ne inequalities ( [ ne factorable 2 ] ) reduce to @xmath134(x^{\star } -x)\geq 0 , \\
( r^{\prime } -s^{\prime } ) [ x^{\star } ( r - s)\triangle _ { 3}+(s\triangle _ { 3}-\triangle _ { 2})](y^{\star } -y)\geq 0.% \end{array } \label{ne stag hunt}\ ] ] from the inequalities ( [ ne stag hunt ] ) the ne @xmath135 results when @xmath136 and , similarly , the ne @xmath137 results when @xmath138 also , the inequalities ( [ ne stag hunt ] ) hold when @xmath139 , @xmath140 and for @xmath141 to be a ne we require @xmath142 these constraints on @xmath143 hold along with the probabilities @xmath144 being factorizable .
the game of chicken is about two drivers who drive towards each other from opposite directions .
one driver must turn aside , or both may die in a crash .
if one driver turns aside but the other does not , s / he will be called a chicken . while each driver prefers not to yield to the opponent , the outcome where neither driver yields is the worst possible one for both . in this anti - coordination game
it is mutually beneficial for parties to play different strategies .
sometimes , chicken is also known as the hawk - dove game , that originates from the parallel development of the basic principles of this game in two different research areas : economics and mathematical biology .
economists , and the political scientists too , refer to @xcite this game as chicken , while mathematical biologists refer to @xcite it as the hawk - dove game .
the game of chicken differs from pd in that in chicken the mutual defection ( the crash when both players drive straight ) is the most feared outcome . while in pd cooperation while the other player defects is the worst outcome .
a version of the chicken game is obtained from the matrices ( [ a and b matrices ] ) when @xmath145 playing this game in two - coin setup the inequalities ( [ ne factorable 1 ] ) are reduced to @xmath146 and three ne emerge : @xmath147 the corresponding payoffs at these equilibria , obtained from eqs .
( 2-coin payoff relations ) , are @xmath148 now we play this game using the four - coin setup .
the ne inequalities come out to be the same as the ones given in ( [ ne stag hunt ] ) except that now we have @xmath149 and @xmath150 .
then for @xmath151 we require @xmath152 similarly , for @xmath153 we require @xmath154 at @xmath155 the inequalities ( [ ne stag hunt ] ) reduce to @xmath156 which puts constraint on @xmath157 given as @xmath158 a special case is the one when @xmath159 and the strategy pair @xmath160 becomes a ne which imposes certain constraints on @xmath143 .
for this ne the inequalities ( [ ne factorable 2 ] ) , for the game defined by ( [ a and b matrices ] , chicken game ) , are reduced to @xmath161
so , we require @xmath162 if the strategy pair @xmath163 is to be a ne in this game . along with this , the probabilities @xmath123 are to be factorizable .
section [ tpgfps ] describes playing a two - player game with four coins such that choosing a coin is a strategy while players payoffs are given by their strategies , the matrix of the game , and the underlying statistics of the coins .
this facilitates transition to playing the _ same _ game using epr experiments . in epr setup , alice and bob are spatially separated and are unable to communicate with each other . in an individual run ,
both receive one half of a pair of particles originating from a common source . in the same run of the experiment both
choose one from two given ( pure ) strategies .
these strategies are the two directions in space along which spin or polarization measurements can be made . keeping the notation for the coins
, we denote these directions to be @xmath2 , @xmath3 for alice and @xmath5 , @xmath6 for bob .
each measurement generates @xmath8 or @xmath9 as the outcome , like it is the case with coins after their toss in the four - coin setup .
experimental results are recorded for a large number of individual runs of the experiment and payoffs are awarded depending on the directions the players go for over many runs ( defining their strategies ) , the matrix of the game they play , and the statistics of the measurement outcomes . for epr experiments
, we retain cereceda notation @xcite for the associated probabilities : @xmath164 in this notation , for example , we write @xmath165 for the probability @xmath166 and @xmath167 for the probability @xmath168 .
one can then construct the following table of probabilities @xmath169 this table allows to transparently see how the probabilities @xmath170 are linked to the probabilities @xmath123 , where we recall that @xmath1 can be set at @xmath2 or at @xmath3 and , similarly , @xmath4 can be set at @xmath3 or at @xmath6 . in cereceda
s notation the epr probabilities @xmath171 are normalized as they satisfy the following relations @xmath172 notice that the factorizable probabilities ( [ table of factorizeable probabilities ] ) are also normalized and ( [ normalization ] ) holds for them .
payoff relations ( [ bilinear payoffs ] ) are originally constructed when the game given by the matrices ( [ a and b matrices ] ) is played with four coins and their mathematical form convinces one to use the following recipe @xcite to reward players when the same game is played using epr probabilities ( [ epr probabilities ] ) : @xmath173 here @xmath174 , for example , is alice s payoff when she plays @xmath2 and bob plays @xmath6 . like it is the case with four coins , the payoff relations for bob are obtained from ( [ epr payoffs ] ) by the transformation @xmath175 in eqs .
( [ epr payoffs ] ) .
when @xmath171 are factorizable in terms of @xmath176 , a comparison of ( [ epr payoffs ] ) with ( [ bilinear payoffs ] ) requires @xmath177 that is , the factorizability of @xmath171 in terms @xmath178 makes the game played by epr probabilities equivalent to the one played by using coins
. however , the epr probabilities @xmath171 , appearing in ( [ bilinear payoffs ] ) , may not be factorizable in terms of @xmath143 , whereas for both the payoff relations ( [ bilinear payoffs ] , [ epr payoffs ] ) the normalization ( [ normalization ] ) continues to hold .
as it is the case with the coin game , alice s mixed strategy is defined to be the probability to choose between @xmath2 and @xmath3 and we can use , once again , the payoff relations ( [ coins payoff relations ] ) which , however , now correspond to the possible situation when @xmath171 may not be factorizable .
so that , the relations ( [ bilinear payoffs ] ) can be replaced with the relations ( [ epr payoffs ] ) in alice s mixed - strategy payoff relation in ( [ coins payoff relations ] ) . the same applies to bob s payoff relations .
note that when @xmath171 are factorizable , using ( [ pis written bilinearly ] ) allows the probabilities @xmath179 to be expressed in terms of @xmath171 : @xmath180 which are useful relations for the forthcoming argument for a quantum game .
along with the normalization ( [ normalization ] ) , the epr probabilities @xmath181 @xmath182 also satisfy certain other constraints imposed by the requirements of causality .
cereceda @xcite writes these constraints as @xmath183 which is referred to as the _ causal communication constraint _ cereceda .
notice that the constraints ( [ constraints on ps ] ) , of course , also hold when @xmath171 are factorizable and are written in terms of @xmath116 as in ( [ pis written bilinearly ] ) . essentially , these constraints state that , on measurement , alice s probability of obtaining particular outcome ( @xmath8 or @xmath9 ) , when she goes for @xmath2 or @xmath3 , is independent of how bob sets up his apparatus ( i.e. along @xmath5 or along @xmath6 ) .
the same applies to bob i.e. on measurement his probability of obtaining a particular outcome ( @xmath8 or @xmath9 ) , when he goes for @xmath5 or @xmath6 , is independent of how alice sets up her apparatus ( i.e. along @xmath2 or along @xmath184 ) .
other authors may like to call the constraints ( [ constraints on ps ] ) with some different name , for example , winsberg and fine winsbergfine have described them as the _ locality constraint_. notice that because of normalization ( [ normalization ] ) half of the eqs .
( [ constraints on ps ] ) are redundant that makes eight among sixteen probabilities @xmath171 to be independent .
a convenient solution cereceda of the system ( [ normalization ] , [ constraints on ps ] ) , for which the set of variables : @xmath185 is expressed in terms of the remaining set of variables : @xmath186 is given as follows @xmath187 the relationships ( [ dependent probabilities ] ) between joint probabilities arise because both the normalization condition ( normalization ) and the causal communication constraint ( [ constraints on ps ] ) are fulfilled . from eqs .
( [ dependent probabilities ] ) one can obtain other constraints considering that the sum of any combination of probabilities from the set @xmath188 must be non - negative . in the following are some results to be used later in this paper . in ( [ dependent probabilities ] )
the sum @xmath189 is non - negative and it requires that @xmath190 in ( [ dependent probabilities ] ) the sum @xmath191 is non - negative and it requires that @xmath192 similarly , the sum @xmath193 is non - negative and it requires that @xmath194 both of the inequalities ( [ p1 and p8 ] , [ p1 and p12 ] ) are found useful in developing a quantum version of pd .
the inequality ( [ p8 and p13 ] ) is found useful in developing quantum version of sh . using ( [ epr payoffs ] ) in ( [ coins payoff relations ] ) , with the assumption that @xmath66 is a ne , one obtains @xmath195\geq 0,% \end{array } \label{nex1}\ ] ] and @xmath196\geq 0,% \end{array } \label{ney1}\ ] ] where @xmath197 now use ( [ dependent probabilities ] ) to write ( [ nex1 ] ) and ( [ ney1 ] ) in terms of the probabilities appearing in the set @xmath198 given in ( second set of probabilities ) to obtain @xmath199\geq 0,% \end{array } \label{nex}\ ] ] @xmath200\geq 0.% \end{array } \label{ney}\ ] ] notice that the probabilities associated to the epr experiments can be factorized only for certain directions of measurements even for singlet states .
for these directions the game played using epr experiments can thus be interpreted within the four - coin setup .
essentially , we obtain quantum game from the classical as follows .
referring to the four - coin setup developed in the section [ tpgfps ] , the factorizability of associated probabilities in terms of @xmath201 allows us to translate the requirement that the resulting game has a classical interpretation into certain constraints on @xmath201 .
we find that from factorizability the relations ( discrete prob ) follow and from these relations the constraints on @xmath202 can be re - expressed in terms of @xmath171 @xmath203 .
we now obtain a quantum version of the game by retaining these constraints and afterwards allowing @xmath171 to become non - factorizable . in this procedure retaining the constraints ensures that classical outcome results when probabilities become factorizable . in the following we consider the impact of non - factorizable probabilities on the ne in pd , sh , and the chicken game .
recall that section [ tpgfps ] states the result that when pd is played with four coins we require the condition ( [ s=0=s ] ) to hold if the strategy pair @xmath204 is to exist as a ne . along with this
the probabilities @xmath171 are to be factorizable .
this motivates to construct a quantum version of pd when probabilities @xmath205 are not factorizable while the constraint ( [ s=0=s ] ) remains valid .
the condition ( [ s=0=s ] ) ensures that with factorizable probabilities the game can be interpreted classically . notice
that when the probabilities @xmath171 are factorizable , i.e. they can be written as in ( [ pis written bilinearly ] ) , the constraint ( [ s=0=s ] ) can hold when numerical values are assigned to certain probabilities among @xmath171 : @xmath206 where , because of the normalization ( [ normalization ] ) , @xmath207 requires that @xmath208 @xmath209 and @xmath210 .
this can also be noticed more directly from ( [ discrete prob ] ) .
this assignment of values to certain probabilities reduces eqs .
( [ normalization ] ) and eqs .
( constraints on ps ) to @xmath211 substituting from ( [ probability value assignment ] ) into ( [ nex ] , ney ) gives @xmath212\geq 0;% \end{array } \label{ne(a)1}\ ] ] @xmath213\geq 0.% \end{array } \label{ne(b)1}\ ] ] note that from ( [ qpd cond2 ] ) we obtain @xmath214 , that for factorizable probabilities , becomes equal to @xmath76 when we refer to eqs .
( [ discrete prob ] ) .
similarly , from ( [ qpd cond2 ] ) we obtain @xmath215 .
substituting these in ( [ ne(a)1 ] ) along with that @xmath216 gives @xmath217 . in similar way we find from ( [ qpd cond2 ] ) that @xmath218 which for factorizable probabilities becomes equal to @xmath219 when we use eqs .
( [ discrete prob ] ) .
likewise , from ( [ qpd cond2 ] ) we obtain @xmath220 .
substituting these in ( [ ne(b)1 ] ) , along with that @xmath216 , gives @xmath221 .
this can be described as follows : when probabilities @xmath222 and @xmath223 are factorizable and the values assigned to them in ( [ probability value assignment ] ) hold , the inequalities ( [ ne(a)1 ] , [ ne(b)1 ] ) ensure that the strategy pair @xmath204 becomes a ne .
now we ask about the fate of the ne strategy pair @xmath21 when in ( [ probability value assignment ] ) the values assigned to certain probabilities , resulting from the requirement ( [ s=0=s ] ) , hold while @xmath181 do not remain factorizable in terms of @xmath157 .
allow the probabilities @xmath222 and @xmath223 not to be factorizable and use ( [ probability value assignment ] ) in ( [ dependent probabilities ] ) to get @xmath224 and the inequalities ( [ ne(a)1 ] , [ ne(b)1 ] ) take the form : @xmath225 p_{1}\geq 0,% \end{array } \label{qpd q ne 1}\ ] ] @xmath226 p_{1}\geq 0,% \end{array } \label{qpd q ne 2}\ ] ] where @xmath227 and @xmath228 are defined in the section ( pd(c ) ) .
note that ( [ p1 and p8 ] ) and ( [ p1 and p12 ] ) give @xmath229 and @xmath230 so that @xmath231 which results , once again , in the strategy pair @xmath232 being a ne , which is the classical outcome of the game .
notice that this ne emerges for non - factorizable epr probabilities along with our requirement that factorizable probabilities must lead to the classical game .
this result for pd appears to diverge away from the reported results in quantum games @xcite .
we believe that part of the reason resides with how payoff relations and players strategies are defined in the present framework , which exploits epr setup for playing a quantum game .
section [ sh ] describes playing sh in the four - coin setup for which three ne emerge . for each of these three ne there
correspond constraints on @xmath202 for factorizable probabilities . in the following we first translate these constraints in terms of the epr probabilities @xmath205 and
afterwards allow @xmath171 to assume non - factorizable values when the constraints on @xmath143 , expressed in terms of @xmath171 , continue to hold . in the following
we follow this procedure for each individual ne that arises when sh is played in the four - coin setup .
[ [ xstarystar_100 ] ] @xmath233 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + refer to eqs .
( [ sh 2 coin three ne ] ) in section [ sh ] and consider the ne @xmath234 for which the constraint on probabilities are ( [ s=0=s ] ) as it is the case with pd .
analysis for quantum pd from section [ qpd ] , therefore , remains valid and we can directly use the inequalities ( [ qpd q ne 1 ] , [ qpd q ne 2 ] , [ qpd q ne 3 ] ) to obtain @xmath235 where @xmath236 .
this gives rise to three equilibria : @xmath237 where the superscript @xmath238 refers to ` quantum ' .
from the relations ( [ qsh ( x*,y*)=(0,0)a ] , [ qsh ( x*,y*)=(0,0)b ] ) and the inequalities ( [ p1 and p8 ] , [ p1 and p12 ] ) it turns out that @xmath239 emerges without any further constraints apart from the ones given by re - expressed form of ( [ s=0=s ] ) i.e. ( [ probability value assignment ] ) ; @xmath240 emerges when non - factorizable probabilities are such that , apart from ( probability value assignment ) to hold , both @xmath241 and @xmath242 have values in the interval @xmath243 $ ] ; and @xmath244 emerges when , apart from ( [ probability value assignment ] ) being true , both @xmath245 and @xmath246 are non negative .
[ [ xstarystar_2triangle_2triangle_3triangle_2triangle_3 ] ] @xmath247 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + refer to section [ sh ] and use ( [ discrete prob ] ) and ( constraints on ps ) , along with the normalization ( [ normalization ] ) , to express the constraints ( [ sh probs constraints ] ) as @xmath248 the normalization ( [ normalization ] ) , then , assigns zero value to the remaining @xmath249 probabilities .
now substitute the constraints ( [ qsh probs constraints ] ) in eqs .
( [ nex ] , [ ney ] ) to obtain the ne inequalities that will correspond to the non - factorizable probabilities : @xmath250\geq 0 , \\
\pi _ { b}(x^{\star } , y^{\star } ) -\pi _ { b}(x^{\star } , y)=(y^{\star } -y)[x^{\star } \triangle _ { 3}-\triangle _ { 2}]\geq 0.% \end{array}%\ ] ] this results in identical to the classical situation and the three ne : @xmath251 emerge when ( [ qsh probs constraints ] ) hold .
[ [ xstarystar_311 ] ] @xmath252 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for this ne in ( [ sh 2 coin three ne ] ) the constraint on probabilities are ( [ sh pure 2 ] ) i.e. @xmath253 . for factorizable probabilities
, this constraint can be rewritten using normalization ( [ normalization ] ) along with ( [ discrete prob ] , [ constraints on ps ] ) as @xmath254 from which using the normalization ( [ normalization ] ) it then follows that @xmath255 @xmath256 @xmath257 .
the constraints ( [ sh ne3 constraints ] ) reduce the nash inequalities ( [ nex ] , [ ney ] ) to @xmath258\geq 0;% \end{array } \label{sh ne3 ( a)}\ ] ] @xmath259\geq 0.% \end{array } \label{sh ne3 ( b)}\ ] ] using the constraints ( [ sh ne3 constraints ] ) in the @xmath260-th equation in ( [ dependent probabilities ] ) results in @xmath261 which then simplifies the nash inequalities ( [ sh ne3 ( a ) ] , [ sh ne3 ( b ) ] ) to @xmath262 and gives rise to three ne that are described below .
@xmath263 for this ne we have @xmath264 and the inequalities ( [ sh ne r=0=r ] ) become @xmath265 where from ( [ p1 and p12 ] , [ p8 and p13 ] ) we have @xmath266 and @xmath267 .
that is , @xmath268 will be a ne when @xmath269 and @xmath270 are such that @xmath271 hold true along with the constraints ( [ sh ne3 constraints ] ) .
@xmath272 from the inequalities ( [ sh ne r=0=r ] ) the strategy pair @xmath273 where the strategy pair @xmath274 @xmath275 can exist as a ne when @xmath269 and @xmath270 are such that @xmath276 $ ] .
together with this the constraints ( [ sh ne3 constraints ] ) are to hold .
@xmath277 for the possible ne @xmath278 the inequalities ( [ sh ne r=0=r ] ) become @xmath279 which are to hold along with that the constraints ( [ sh ne3 constraints ] ) being true . using @xmath280 with ( [ p1 and p12 ] , [ p8 and p13 ] )
we find that the inequalities ( [ qsh r=0=r ( 0,0)c ] ) will always hold and the only requirement for the strategy pair @xmath281 to be a ne is that the constraints ( [ sh ne3 constraints ] ) hold .
so that the list of possible ne , that can arise in the quantum sh , consists of five members i.e. @xmath282 , @xmath283 , @xmath284 @xmath285 , @xmath286 @xmath287 , and @xmath288 @xmath289 $ ] . which one , or more , from this list are going to arise depends on the set of non - factorizable probabilities .
for example , we notice that there exist @xcite two sets of non - factorizable probabilities that maximally violate the quantum prediction of the clauser - holt - shimony - horne ( chsh ) sum of correlations .
the first set is @xmath290 and the second set is @xmath291 where @xmath198 and @xmath292 are defined in ( [ first set of probabilities ] ) and in ( [ second set of probabilities ] ) , respectively .
the probabilities in these sets are non - factorizable because for both sets a solution of the eqs .
( [ pis written bilinearly ] ) will involve one or more of the probabilities @xmath157 being negative or greater than one .
now for sh the requirement that factorizable probabilities are to lead to classical game gives rise to three sets of constraints on epr probabilities given by ( [ probability value assignment ] ) , ( [ qsh probs constraints ] ) , and ( [ sh ne3 constraints ] ) .
these sets of constraints correspond to the ne @xmath293 @xmath294 and @xmath295 respectively .
unfortunately , the probabilities from either of the two sets ( [ first set ] , [ second set ] ) , which maximally violate the quantum prediction of the chsh sum of correlations , do not satisfy these constraints .
stated otherwise , the probabilities from the sets ( [ first set ] , [ second set ] ) are in conflict with the requirement that factorizable probabilities must lead to the classical game of sh .
however , other sets of non - factorizable probabilities can be found that are consistent with this requirement and , depending on the elements of a set , one or more out of five possible ne may emerge .
this situation can be described by saying that in sh non - factorizability can lead to new ne but , unfortunately , either of the sets ( [ first set ] , [ second set ] ) can not be used for this purpose .
refer to section [ cg ] and use eqs .
( [ chicken game constraints on r , s , r,s ] ) express the constraints on @xmath143 in this setup .
these constraints are imposed for the strategy pair @xmath296 is to be a ne .
( discrete prob ) and the normalization ( [ normalization ] ) to translate these constraints in terms of the epr probabilities @xmath171 : @xmath297 addition and subtraction of ( [ qcg constraints on epr probs ( b ) ] ) to and from ( [ qcg constraints on epr probs ( a ) ] ) gives @xmath298 and eqs .
( [ dependent probabilities ] ) then allow us to re - express eqs .
( [ q chicken 1,2 ] ) in terms of the probabilities in set @xmath198 , defined in ( [ second set of probabilities ] ) , to obtain @xmath299 two probabilities can be eliminated from the inequalities ( [ nex ] , ney ) using eqs .
( [ q chicken a , b ] ) .
we select ( arbitrarily ) these to be @xmath223 and @xmath300 and express them in terms of other probabilities in the set @xmath198 i.e. @xmath301 notice that for the chicken game , defined in ( [ chicken game ] ) , the definition of @xmath302 in section [ pd(c ) ] gives @xmath303 now eliminate @xmath223 and @xmath300 from the inequalities ( [ nex ] , ney ) using eqs .
( [ quantum chicken 1,2 ] ) and substitute from eqs .
( qcg deltas ) . the inequalities ( [ nex ] , [ ney ] ) then read @xmath304;% \end{array } \label{qchickengamene1}\ ] ] @xmath305.% \end{array } \label{qchickengamene2}\ ] ] inequalities ( [ qchickengamene1 ] , [ qchickengamene2 ] ) ensure that for factorizable probabilities the classical ne @xmath306 comes out as the outcome of the game .
what is the fate of this equilibrium when probabilities are not factorizable ? to address this question we consider a special case when @xmath307 , for which @xmath308 is the classical mixed - strategy outcome of the game . to obtain this outcome within the four - coin setup the constraints on @xmath116
are given in eqs .
( [ chicken constraint ] ) .
the inequalities ( [ qchickengamene1 ] , [ qchickengamene2 ] ) reduce to @xmath309 @xmath310 notice that the inequalities ( [ sh ne r=0=r ] ) do not allow either of the strategy pairs @xmath311 and @xmath312 to be ne . like it has been the case with quantum sh ,
we now ask which of these nine possible ne will emerge when probabilities become non - factorizable .
to answer this we refer to the set ( [ first set ] ) of probabilities and assign the value @xmath313 to each of the probabilities @xmath314 @xmath315 @xmath316 @xmath317 @xmath318 @xmath319 . using eqs .
( [ quantum chicken 1,2 ] ) the assumption that @xmath307 then also assigns the same value , i.e. @xmath313 , to both @xmath223 and @xmath320 .
the inequalities ( [ qcgc ] , [ qcgd ] ) are @xmath321 with the result that the strategy pairs @xmath322 emerge as the new equilibria ) we recall that when @xmath323 there are three equilibria i.e. @xmath324 and @xmath325 in the classical chicken game , at which they get rewarded by @xmath326 and @xmath327 respectively . here
the first and the second entry refers to alice s and bob s reward , respectively . ] .
so that , in this case , the set ( [ first set ] ) of non - factorizable probabilities indeed leads to the new equilibria of the game . using ( [ coins payoff relations ] ) one finds that at the equilibrium @xmath283 both players get @xmath328 while at @xmath282 both players get @xmath329 as their payoffs .
note that , when re - expressed in terms of the epr probabilities @xmath171 using eqs .
( [ discrete prob ] ) , the constraints ( [ chicken constraint ] ) can be written as @xmath330 which , of course , continue to hold for the set ( [ first set ] ) when the probabilities @xmath171 are allowed to be non - factorizable . similarly , referring to the second set ( [ second set ] ) ,
we assign the value @xmath331 to the probabilities @xmath314 @xmath315 @xmath316 @xmath332 @xmath318 @xmath319 that appear in ( [ qcgc ] ) and ( [ qcgd ] ) . the eqs .
( [ quantum chicken 1,2 ] ) , with the assumption that @xmath307 , then assign the value of @xmath331 both to @xmath223 and @xmath300 and the inequalities ( [ qcgc ] , [ qcgd ] ) read @xmath333 which are the same as the ones given in ( [ qcgn(a , b ) ] ) , apart from extra negative signs .
this results in three strategy pairs @xmath334 to come out as the equilibria . once again , using ( coins payoff relations ) one finds that at all of these three equilibria both players get equally rewarded by the amount @xmath329 . hence , while referring to ( [ 3 equilibria in chicken ] ) , we find that in this special case when @xmath307 the set ( [ second set ] ) of probabilities leads to new equilibria of the game .
notice that , like it is the case with the set ( [ first set ] ) of probabilities , the constraints ( [ qcg constraints ] ) continue to hold also for the set ( [ second set ] ) when @xmath171 are allowed to be non - factorizable .
in typical quantization procedure @xcite of a two - player game , two quantum bits ( qubits ) are in a quantum correlated ( entangled ) state that are given to the players alice and bob .
players strategies consist of performing unitary actions on their respective qubits .
classical game remains a subset of the quantum game in that both players can play quantum strategies that correspond to the strategies available classically .
as more choices are allowed to the players that now also include superpositions of their classical moves , it gives ground to the argument that enk and pike @xcite have put forward . the setup proposed in this
paper uses epr experiments to play a two - player game and a quantum game is associated to a classical game such that it becomes hard to construct an enk and pike - type argument as both the payoff relations and the players sets of strategies remain identical @xcite in the classical and the associated quantum game . in the present setup
it is non - factorizability
responsible for the violation of bell inequality in epr experiments that gives rise to the new solutions in quantum game .
when players play a game using a physical system for which joint probabilities are factorizable the classical game results always .
in other words , the role of non - factorizable probabilities is sought in the game - theoretic solution concept of a ne , when the physical realization for these probabilities is provided by the epr experiments .
this analysis introduces a new viewpoint in the area of quantum games in which non - factorizability gets translated into the language of game theory .
the argument put forward in this paper can be described as follows .
firstly , players payoffs are re - expressed in the form @xmath335 where @xmath171 are the sixteen joint probabilities ; @xmath336 are players strategies , and @xmath337 are players payoff matrices defined in ( [ a and b matrices ] ) .
secondly , nash inequalities are used to impose constraints on @xmath171 that ensure that with factorizable @xmath171 the game has classical outcome and the resulting payoffs can be interpreted in terms of classical mixed - strategy game .
it is achieved by playing the game in the four - coin setup and using nash inequalities to obtain constraints on the coin probabilities @xmath143 which reproduce the outcome of the classical mixed - strategy game . using ( discrete prob ) , which results from factorizability , these constraints on @xmath202 are then translated in terms of constraints on @xmath181 .
thirdly , while referring to the epr setup , @xmath171 are allowed to be non - factorizable when the constraints on @xmath171 continue to hold .
fourthly , and lastly , it is observed how non - factorizability leads to the emergence of new solutions of the game .
note that for a game different sets of constraints are defined depending on which ne is to be the solution of the game .
for example , for three ne in chicken we require three different sets of constraints on @xmath201 .
considering one of these three sets at a time we repeat the four steps stated above .
the same procedure is then repeated for other sets of constraints corresponding to other ne . that is , in this setup not all solutions of a game are re - expressed in terms of a single set of constraints on @xmath143 . instead
, a separate set of constraints is found for each ne .
it seems that the four - coin setup is the minimal arrangement that allows one to introduce , in a smooth way , the epr probabilities into a game - like setting .
we suggest that with increasing the number of coins , shared by each of the two players , all the ne of a game can be translated to a single constraint on the underlying coin probabilities which are subsequently translated in terms of @xmath181 .
this will then allow to see the role of non - factorizability on solution of a game from a single set of constraints .
however , this will be obtained at a price : firstly , more coins will be involved resulting in more mathematical complexity ; secondly , for more coins player s strategy will need to be redefined such that it permits to incorporate epr probabilities . note
that the usual approach uses entangled states to construct quantum games and this paper uses non - factorizability to the same end .
mathematically , non - factorizability comes out to be a stronger condition than the condition that translates entanglement into constraints on joint probabilities .
that is , a non - factorizable set of probabilities always corresponds to some entangled state but an entangled state can produce a factorizable set of probabilities .
for example , in case of singlet state the outcomes of two measurements violate bell inequality only along certain directions , and not along other directions .
in other words , in a quantum game exploiting entangled states , the joint probabilities may still be factorizable but for a quantum game , resulting from non - factorizable probabilities , bell inequality is bound to be violated .
non - factorizability being a stronger condition may well be suggested as the reason why it can not be helpful to escape from the classical outcome in pd .
the proposed setup demonstrates how non - factorizability can change outcome of a game .
we suggest to extend @xcite this setup to analyze multi - player quantum games @xcite where players share physical systems for which joint probabilities can not be factorized .
one of us ( ai ) is supported by the japan society for promotion of science ( jsps ) .
ai gratefully acknowledges hospitality and support from the international relations center and the research institute at the kochi university of technology .
this work is supported , in part , by the grant - in - aid for scientific research provided by japan ministry of education , culture , sports , science , and technology under the contract numbers 18540384 and 18.06330 .
m. a. nielsen and i. l. chuang , _ quantum computation and quantum information _ , cambridge university press ( 2000 ) .
see also , c. p. williams and s. h. clearwater , _ explorations in quantum computing _ , springer - verlag inc , new york ( 1998 ) .
|
a probabilistic framework is developed that gives a unifying perspective on both the classical and the quantum games .
we suggest exploiting peculiar probabilities involved in einstein - podolsky - rosen ( epr ) experiments to construct quantum games . in our framework
a game attains classical interpretation when joint probabilities are factorizable and a quantum game corresponds when these probabilities can not be factorized .
we analyze how non - factorizability changes nash equilibria in two - player games while considering the games of prisoner s dilemma , stag hunt , and chicken . in this framework
we find that for the game of prisoner s dilemma even non - factorizable epr joint probabilities can not be helpful to escape from the classical outcome of the game .
for a particular version of the chicken game , however , we find that the two non - factorizable sets of joint probabilities , that maximally violates the clauser - holt - shimony - horne ( chsh ) sum of correlations , indeed result in new nash equilibria .
|
two ubiquitous aspects of large - scale data analysis are that the data often have heavy - tailed properties and that diffusion - based or spectral - based methods are often used to identify and extract structure of interest . in the absence of strong assumptions on the data , popular distribution - independent methods such as those based on the vc dimension fail to provide nontrivial results for even simple learning problems such as binary classification in these two settings . at root ,
the reason is that in both of these situations the data are formally very high dimensional and that ( without additional regularity assumptions on the data ) there may be a small number of `` very outlying '' data points . in this paper , we develop distribution - dependent learning methods that can be used to provide dimension - independent sample complexity bounds for the maximum margin version of the binary classification problem in these two popular settings . in both cases , we are able to obtain nearly optimal linear classification hyperplanes since the distribution - dependent tools we employ are able to control the aggregate effect of the `` outlying '' data points . in particular , our results will hold even though the data may be infinite - dimensional and unbounded .
spectral - based kernels have received a great deal of attention recently in machine learning for data classification , regression , and exploratory data analysis via dimensionality reduction @xcite .
consider , for example , laplacian eigenmaps @xcite and the related diffusion maps @xcite . given a graph @xmath0 ( where this graph could be constructed from the data represented as feature vectors ,
as is common in machine learning , or it could simply be a natural representation of a large social or information network , as is more common in other areas of data analysis ) , let @xmath1 be the eigenfunctions of the normalized laplacian of @xmath2 and let @xmath3 be the corresponding eigenvalues .
then , the diffusion map is the following feature map @xmath4 and laplacian eigenmaps is the special case when @xmath5 . in this case , the support of the data distribution is unbounded as the size of the graph increases ; the vc dimension of hyperplane classifiers is @xmath6 ; and thus existing results do not give dimension - independent sample complexity bounds for classification by empirical risk minimization ( erm ) .
moreover , it is possible ( and indeed quite common in certain applications ) that on some vertices @xmath7 the eigenfunctions fluctuate wildly even on special classes of graphs , such as random graphs @xmath8 , a non - trivial uniform upper bound stronger than @xmath9 on @xmath10 over all vertices @xmath7 does not appear to be known .
even for maximum margin or so - called `` gap - tolerant '' classifiers , defined precisely in section [ sxn : background ] and which are easier to learn than ordinary linear hyperplane classifiers , the existing bounds of vapnik are not independent of the number @xmath11 of nodes . on the vc dimension of gap - tolerant classifiers applied to the diffusion map feature space corresponding to a graph with @xmath11 nodes .
( recall that by lemma [ lem : vap1 ] below , the vc dimension of the space of gap - tolerant classifiers corresponding to a margin @xmath12 , applied to a ball of radius @xmath13 is @xmath14 . )
of course , although this bound is quadratic in the number of nodes , vc theory for ordinary linear classifiers gives an @xmath9 bound . ]
a similar problem arises in the seemingly very - different situation that the data exhibit heavy - tailed or power - law behavior .
heavy - tailed distributions are probability distributions with tails that are not exponentially bounded @xcite .
such distributions can arise via several mechanisms , and they are ubiquitous in applications @xcite .
for example , graphs in which the degree sequence decays according to a power law have received a great deal of attention recently .
relatedly , such diverse phenomenon as the distribution of packet transmission rates over the internet , the frequency of word use in common text , the populations of cities , the intensities of earthquakes , and the sizes of power outages all have heavy - tailed behavior .
although it is common to normalize or preprocess the data to remove the extreme variability in order to apply common data analysis and machine learning algorithms , such extreme variability is a fundamental property of the data in many of these application domains .
there are a number of ways to formalize the notion of heavy - tailed behavior for the classification problems we will consider , and in this paper we will consider the case where the _ magnitude _ of the entries decays according to a power law .
( note , though , that in appendix [ sxn : learnht ] , we will , for completeness , consider the case in which the probability that an entry is nonzero decays in a heavy - tailed manner . ) that is , if @xmath15 represents the feature map , then @xmath16 for some absolute constant @xmath17 , with @xmath18 . as in the case with spectral kernels , in this heavy - tailed situation
, the support of the data distribution is unbounded as the size of the graph increases , and the vc dimension of hyperplane classifiers is @xmath6 .
moreover , although there are a small number of `` most important '' features , they do not `` capture ''
most of the `` information '' of the data .
thus , when calculating the sample complexity for a classification task for data in which the feature vector has heavy - tailed properties , bounds that do not take into account the distribution are likely to be very weak . in this paper , we develop distribution - dependent bounds for problems in these two settings . clearly , these results are of interest since vc - based arguments fail to provide nontrivial bounds in these two settings , in spite of ubiquity of data with heavy - tailed properties and the widespread use of spectral - based kernels in many applications . more generally , however , these results are of interest since the distribution - dependent bounds underlying them provide insight into how better to deal with heterogeneous data with more realistic noise properties .
our first main result provides bounds on classifying data whose magnitude decays in a heavy - tailed manner . in particular , in the following theorem we show that if the weight of the @xmath19 coordinate of random data point is less than @xmath20 for some @xmath21 , then the number of samples needed before a maximum - margin classifier is approximately optimal with high probability is independent of the number of features .
[ heavy - tailed data ] [ thm : learn_heavytail ] let the data be heavy - tailed in that the feature vector : @xmath22 satisfy @xmath23 for some absolute constant @xmath17 , with @xmath18 .
let @xmath24 denote the riemann zeta function .
then , for any @xmath25 , if a maximum margin classifier has a margin @xmath26 , with probability more than @xmath27 , its risk is less than @xmath28 where @xmath29 hides multiplicative polylogarithmic factors .
this result follows from a bound on the annealed entropy of gap - tolerant classifiers in a hilbert space that is of independent interest .
in addition , it makes important use of the fact that although individual elements of the heavy - tailed feature vector may be large , the vector has bounded moments .
our second main result provides bounds on classifying data with spectral kernels .
in particular , in the following theorem we give dimension - independent upper bounds on the sample complexity of learning a nearly - optimal maximum margin classifier in the feature space of the diffusion maps .
[ spectral kernels ] [ thm : learn_spectral ] let the following diffusion map be given : @xmath30 where @xmath31 are normalized eigenfunctions ( whose @xmath32 ) norm is @xmath33 , @xmath34 being the uniform distribution ) , @xmath35 are the eigenvalues of the corresponding markov chain and @xmath36 .
then , for any @xmath25 , if a maximum margin classifier has a margin @xmath26 , with probability more than @xmath27 , its risk is less than @xmath37 where @xmath29 hides multiplicative polylogarithmic factors . as with the proof of our main heavy - tailed learning result , the proof of our main spectral learning result makes essential use of an upper bound on the annealed entropy of gap - tolerant classifiers . in applying it ,
we make important use of the fact that although individual elements of the feature vector may fluctuate wildly , the norm of the diffusion map feature vector is bounded . as a side remark ,
note that we are not viewing the feature map in theorem [ thm : learn_spectral ] as necessarily being either a random variable or requiring knowledge of some marginal distribution as might be the case if one is generating points in some space according to some distribution ; then constructing a graph based on nearest neighbors ; and then doing diffusions to construct a feature map .
instead , we are thinking of a data graph in which the data are adversarially presented , _
e.g. _ , a given social network is presented , and diffusions and/or a feature map is then constructed .
these two theorems provides a dimension - independent ( _ i.e. _ , independent of the size @xmath11 of the graph and the dimension of the feature space ) upper bound on the number of samples needed to learn a maximum margin classifier , under the assumption that a heavy - tailed feature map or the diffusion map kernel of some scale is used as the feature map .
as mentioned , both proofs ( described below in sections [ sxn : gaphs_htapp ] and [ sxn : gaphs_skapp ] ) proceed by providing a dimension - independent upper bound on the annealed entropy of gap - tolerant classifiers in the relevant feature space , and then appealing to theorem [ thm : vapnik ] ( in section [ sxn : background ] ) relating the annealed entropy to the generalization error . for this bound on the annealed entropy of these gap - tolerant classifiers ,
we crucially use the fact that @xmath38 is bounded , even if @xmath39 is unbounded as @xmath40 .
that is , although bounds on the individual entries of the feature map do not appear to be known , we crucially use that there exist nontrivial bounds on the magnitude of the feature vectors .
since this bound is of more general interest , we describe it separately .
the distribution - dependent ideas that underlie our two main results ( in theorems [ thm : learn_heavytail ] and [ thm : learn_spectral ] ) can also be used to bound the sample complexity of a classification task more generally under the assumption that the expected value of a norm of the data is bounded , _
i.e. _ , when the magnitude of the feature vector of the data in some norm has a finite moment . in more detail
: * let @xmath41 be a probability measure on a hilbert space @xmath42 , and let @xmath43 . in theorem [ thm : ddb_hs ] ( in section [ sxn : gaphs_annent ] ) , we prove that if @xmath44 , then the annealed entropy of gap - tolerant classifiers ( defined in section [ sxn : background ] ) in @xmath42 can be upper bounded in terms of a function of @xmath45 , @xmath12 , and ( the number of samples ) @xmath25 , independent of the ( possibly infinite ) dimension of @xmath42 . it should be emphasized that the assumption that the expectation of some moment of the norm of the feature vector is bounded is a _ much _ weaker condition than the more common assumption that the largest element is bounded , and thus this result is likely of more general interest in dealing with heterogeneous and noisy data .
for example , similar ideas have been applied recently to the problem of bounding the sample complexity of learning smooth cuts on a low - dimensional manifold @xcite . to establish this result
, we use a result ( see lemma [ lem : vap1 ] in section [ sxn : gaphs_vcdim ] . ) that the vc dimension of gap - tolerant classifiers in a hilbert space when the margin is @xmath12 over a bounded domain such as a ball of radius @xmath13 is bounded above by @xmath46 .
such bounds on the vc dimension of gap - tolerant classifiers have been stated previously by vapnik @xcite . however , in the course of his proof bounding the vc dimension of a gap - tolerant classifier whose margin is @xmath12 over a ball of radius @xmath13 ( see @xcite , page 353 . ) , vapnik states , without further justification , that due to symmetry the set of points in a ball that is extremal in the sense of being the hardest to shatter with gap - tolerant classifiers is the regular simplex .
attention has been drawn to this fact by burges ( see @xcite , footnote 20 . ) , who mentions that a rigorous proof of this fact seems to be absent . here
, we provide a new proof of the upper bound on the vc dimension of such classifiers without making this assumption .
( see lemma [ lem : vap1 ] in section [ sxn : gaphs_vcdim ] and its proof . ) hush and scovel @xcite provide an alternate proof of vapnik s claim ; it is somewhat different than ours , and they do not extend their proof to banach spaces .
the idea underlying our new proof of this result generalizes to the case when the data need not have compact support and where the margin may be measured with respect to more general norms .
in particular , we show that the vc dimension of gap - tolerant classifiers with margin @xmath12 in a ball of radius @xmath13 in a banach space of rademacher type @xmath47 $ ] and type constant @xmath48 is bounded above by @xmath49 , and that there exists a banach space of type @xmath50 ( in fact @xmath51 ) for which the vc dimension is bounded below by @xmath52 .
( see lemmas [ lem : banach_ub ] and [ lem : banach_lb ] in section [ sxn : gapbs_vcdim ] . )
using this result , we can also prove bounds for the annealed entropy of gap - tolerant classifiers in a banach space .
( see theorem [ thm : ddb_bs ] in section [ sxn : gapbs_annent ] . ) in addition to being of interest from a theoretical perspective , this result is of potential interest in cases where modeling the relationship between data elements as a dot product in a hilbert space is too restrictive , and thus this may be of interest , _
e.g. _ , when the data are extremely sparse and heavy - tailed .
gap - tolerant classifiers see section [ sxn : background ] for more details are useful , at least theoretically , as a means of implementing structural risk minimization ( see , _ e.g. _ , appendix a.2 of @xcite ) . with gap - tolerant classifiers ,
the margin @xmath53 is fixed before hand , and does not depend on the data .
e.g. _ , @xcite . with maximum margin classifiers , on the other hand ,
the margin is a function of the data . in spite of this difference ,
the issues that arise in the analysis of these two classifiers are similar .
for example , through the fat - shattering dimension , bounds can be obtained for the maximum margin classifier , as shown by shawe - taylor _ et al . _
@xcite . here
, we briefly sketch how this is achieved .
let @xmath54 be a set of real valued functions .
we say that a set of points @xmath55 is @xmath56shattered by @xmath54 if there are real numbers @xmath57 such that for all binary vectors @xmath58 and each @xmath59=\{1,\ldots , s\}$ ] , there is a function @xmath60 satisfying , f_(x_i ) = \ { ll > t_i + , & + < t_i - , & . for each @xmath61 , the _ fat shattering dimension
_ @xmath62 of the set @xmath54 is defined to be the size of the largest @xmath56shattered set if this is finite ; otherwise it is declared to be infinity .
note that , in this definition , @xmath63 can be different for different @xmath64 , which is not the case in gap - tolerant classifiers .
however , one can incorporate this shift into the feature space by a simple construction .
we start with the following definition of a banach space of type @xmath50 with type constant @xmath48 .
[ banach space , type , and type constant ] [ def : rad ] a _ banach space _ is a complete normed vector space . a banach space @xmath65 is said to have ( rademacher ) type @xmath50 if there exists @xmath66 such that for all @xmath11 and @xmath67 _ t^p _
i=1^n x_i_^p . the smallest @xmath48 for which the above holds with @xmath50 equal to the type ,
is called the type constant of @xmath65 . given a banach space @xmath68 of type @xmath50 and type constant @xmath48 , let @xmath69 consist of all tuples @xmath70 for @xmath71 and @xmath72 , with the norm @xmath73 noting that if @xmath68 is a banach space of type @xmath50 and type constant @xmath48 ( see sections [ sxn : gapbs_prelim ] and [ sxn : gapbs_vcdim ] ) , one can easily check that @xmath69 is a banach space of type @xmath50 and type constant @xmath74 . in our distribution - specific setting , we can not control the fat - shattering dimension , but we can control the logarithm of the expected value of @xmath75 for any constant @xmath76 by applying theorem [ thm : ddb_bs ] to @xmath69 . as seen from lemma 3.7 and corollary 3.8 of the journal version of @xcite ,
this is all that is required for obtaining generalization error bounds for maximum margin classification . in the present context
, the logarithm of the expected value of the exponential of the fat shattering dimension of linear @xmath33-lipschitz functions on a random data set of size @xmath25 taken i.i.d from @xmath41 on @xmath68 is bounded by the annealed entropy of gap - tolerant classifiers on @xmath69 with respect to the push - forward @xmath77 of the measure @xmath41 under the inclusion @xmath78 .
this allows us to state the following theorem , which is an analogue of theorem 4.17 of the journal version of @xcite , adapted using theorem [ thm : ddb_bs ] of this paper .
[ thm : margin_bs ] let @xmath79 .
suppose inputs are drawn independently according to a distribution @xmath41 be a probability measure on a banach space @xmath68 of type @xmath50 and type constant @xmath48 , and @xmath80 .
if we succeed in correctly classifying @xmath25 such inputs by a maximum margin hyperplane of margin @xmath53 , then with confidence @xmath27 the generalization error will be bounded from above by @xmath81 where @xmath29 hides multiplicative polylogarithmic factors involving @xmath82 and @xmath53 .
specializing this theorem to a hilbert space , we have the following theorem as a corollary .
[ thm : margin_hs ] let @xmath79 .
suppose inputs are drawn independently according to a distribution @xmath41 be a probability measure on a hilbert space @xmath42 , and @xmath83 .
if we succeed in correctly classifying @xmath25 such inputs by a maximum margin hyperplane with margin @xmath53 , then with confidence @xmath27 the generalization error will be bounded from above by @xmath84 where @xmath29 hides multiplicative polylogarithmic factors involving @xmath85 and @xmath53 .
note that theorem [ thm : margin_hs ] is an analogue of theorem 4.17 of the journal version of @xcite , but adapted using theorem [ thm : ddb_hs ] of this paper .
in particular , note that this theorem does not assume that the distribution is contained in a ball of some radium @xmath13 , but instead it assumes only that some moment of the distribution is bounded . in the next section ,
section [ sxn : background ] , we review some technical preliminaries that we will use in our subsequent analysis . then , in section [ sxn : gaphs ] , we state and prove our main result for gap - tolerant learning in a hilbert space , and we show how this result can be used to prove our two main theorems in maximum margin learning . then , in section [ sxn : gapbs ] , we state and prove an extension of our gap - tolerant learning result to the case when the gap is measured with respect to more general banach space norms ; and then , in sections [ sxn : discussion ] and [ sxn : conclusion ] we provide a brief discussion and conclusion . finally , for completeness , in appendix [ sxn : learnht ] , we will provide a bound for exact ( as opposed to maximum margin ) learning in the case in which the probability that an entry is nonzero ( as opposed to the value of that entry ) decays in a heavy - tailed manner .
in this paper , we consider the supervised learning problem of binary classification , _ i.e. _ , we consider an input space @xmath86 ( _ e.g. _ , a euclidean space or a hilbert space ) and an output space @xmath87 , where @xmath88 , and where the data consist of pairs @xmath89 that are random variables distributed according to an unknown distribution .
we shall assume that for any @xmath90 , there is at most one pair @xmath91 that is observed .
we observe @xmath25 i.i.d .
pairs @xmath92 sampled according to this unknown distribution , and the goal is to construct a classification function @xmath93 which predicts @xmath87 from @xmath86 with low probability of error . whereas an ordinary linear hyperplane classifier consists of an oriented hyperplane , and points are labeled @xmath94 , depending on which side of the hyperplane they lie , a
_ gap - tolerant classifier _ consists of an oriented hyperplane and a margin of thickness @xmath12 in some norm .
any point outside the margin is labeled @xmath95 , depending on which side of the hyperplane it falls on , and all points within the margin are declared `` correct , '' without receiving a @xmath95 label .
this latter setting has been considered in @xcite ( as a way of implementing structural risk minimization
apply empirical risk minimization to a succession of problems , and choose where the gap @xmath12 that gives the minimum risk bound ) .
the _ risk _ @xmath96 of a linear hyperplane classifier @xmath97 is the probability that @xmath98 misclassifies a random data point @xmath99 drawn from @xmath41 ; more formally , @xmath100 $ ] . given a set of @xmath25 labeled data points @xmath101 , the _ empirical risk _
@xmath102 of a linear hyperplane classifier @xmath97 is the frequency of misclassification on the empirical data ; more formally , @xmath103 $ ] , where @xmath104 $ ] denotes the indicator of the respective event .
the risk and empirical risk for gap - tolerant classifiers are defined in the same manner .
note , in particular , that data points labeled as `` correct '' do not contribute to the risk for a gap - tolerant classifier , _
i.e. _ , data points that are on the `` wrong '' side of the hyperplane but that are within the @xmath12 margin are not considered as incorrect and do not contribute to the risk . in classification , the ultimate goal is to find a classifier that minimizes the true risk , _
i.e. _ , @xmath105 , but since the true risk @xmath96 of a classifier @xmath98 is unknown , an empirical surrogate is often used . in particular ,
_ empirical risk minimization ( erm ) _ is the procedure of choosing a classifier @xmath98 from a set of classifiers @xmath106 by minimizing the empirical risk @xmath107 .
the consistency and rate of convergence of erm see @xcite for precise definitions can be related to uniform bounds on the difference between the empirical risk and the true risk over all @xmath108 .
there is a large body of literature on sufficient conditions for this kind of uniform convergence .
for instance , the vc dimension is commonly - used toward this end .
note that , when considering gap - tolerant classifiers , there is an additional caveat , as one obtains uniform bounds only over those gap - tolerant classifiers that do not contain any data points in the margin the appendix a.2 of @xcite addresses this issue . in this paper ,
our main emphasis is on the annealed entropy : [ annealed entropy ] [ def : an_ent ] let @xmath41 be a probability measure supported on a vector space @xmath42 .
given a set @xmath106 of decision rules and a set of points @xmath109 , let @xmath110 be the number of ways of labeling @xmath111 into positive and negative samples such that there exists a gap - tolerant classifier that predicts _ incorrectly _ the label of each @xmath112 .
given the above notation , h_ann^(k ) : = _ p^k ( z_1 , , z_k ) is the annealed entropy of the classifier @xmath106 with respect to @xmath41 .
note that although we have defined the annealed entropy for general decision rules , below we will consider the case that @xmath106 consists of linear decision rules . as the following theorem states
, the annealed entropy of a classifier can be used to get an upper bound on the generalization error .
this follows from theorem @xmath113 in @xcite and a remark on page 198 of @xcite .
note that the class @xmath114 is itself random , and consequently , @xmath115 is the supremum over a random class .
[ thm : vapnik ]
let @xmath114 be the family of all gap - tolerant classifiers such that no data point lies inside the margin .
then , @xmath116 < 8 \exp{\left(\left({h_{ann}^\lambda(\ell)}\right ) - \frac{\epsilon^2\ell}{32}\right)}\ ] ] holds true , for any number of samples @xmath25 and for any error parameter @xmath117 .
the key property of the function class that leads to uniform bounds is the sublinearity of the logarithm of the expected value of the `` growth function , '' which measures the number of distinct ways in which a data set of a particular size can be split by the function class .
a finite vc bound guarantees this in a distribution - free setting .
the annealed entropy is a distribution - specific measure , _
i.e. _ , the same family of classifiers can have different annealed entropies when measured with respect to different distributions . for a more detailed exposition of uniform bounds in the context of gap - tolerant classifiers ,
we refer the reader to ( @xcite , appendix a.2 ) .
note also that normed vector spaces ( such as hilbert spaces and banach spaces ) are relevant to learning theory for the following reason .
data are often accompanied with an underlying metric which carries information about how likely it is that two data points have the same label .
this makes concrete the intuition that points with the same class label are clustered together .
many algorithms can not be implemented over an arbitrary metric space , but require a linear structure .
if the original metric space does not have such a structure , as is the case when classifying for example , biological data or decision trees , it is customary to construct a feature space representation , which embeds data into a vector space
. we will be interested in the commonly - used hilbert spaces , in which distances in the feature space are measure with respect to the @xmath118 distance ( as well as more general banach spaces , in section [ sxn : gapbs ] ) . finally , note that our results where the margin is measured in @xmath118 can be transferred to a setting with kernels . given a kernel @xmath119 , it is well known that linear classification using a kernel @xmath119 is equivalent to mapping @xmath120 onto the functional @xmath121 and then finding a separating halfspace in the reproducing kernel hilbert space ( rkhs ) which is the hilbert space generated by the functionals of the form @xmath121 . since the span of any finite set of points in a hilbert space can be isometrically embedded in @xmath118 , our results hold in the setting of kernel - based learning as well , when one first uses the feature map @xmath122 and works in the rkhs .
in this section , we state and prove theorem [ thm : ddb_hs ] , our main result regarding an upper bound for the annealed entropy of gap - tolerant classifiers in @xmath118 .
this result is of independent interest , and it was used in a crucial way in the proof of theorems [ thm : learn_heavytail ] and [ thm : learn_spectral ] .
we start in section [ sxn : gaphs_annent ] with the statement and proof of theorem [ thm : ddb_hs ] , and then in section [ sxn : gaphs_vcdim ] we bound the vc dimension of gap - tolerant classifiers over a ball of radius @xmath13 .
then , in section [ sxn : gaphs_htapp ] , we apply these results to prove our main theorem on learning with heavy - tailed data , and finally in section [ sxn : gaphs_skapp ] , we apply these results to prove our main theorem on learning with spectral kernels .
the following theorem is our main result regarding an upper bound for the annealed entropy of gap - tolerant classifiers .
the result holds for gap - tolerant classification in a hilbert space , _ i.e. _ , when the distances in the feature space are measured with respect to the @xmath118 norm .
analogous results hold when distances are measured more generally , as we will describe in section [ sxn : gapbs ] .
[ annealed entropy ; upper bound ; hilbert space ] [ thm : ddb_hs ] let @xmath41 be a probability measure on a hilbert space @xmath42 , and let @xmath43 .
if @xmath44 , then then the annealed entropy of gap - tolerant classifiers in @xmath42 , where the gap is @xmath12 , is @xmath123 let @xmath25 independent , identically distributed ( i.i.d ) samples @xmath124 be chosen from @xmath125 .
we partition them into two classes : @xmath126 and @xmath127 our objective is to bound from above the annealed entropy @xmath128 $ ] . by lemma [ lem : submul ] , @xmath129 is sub - multiplicative .
therefore , @xmath130 taking an expectation over @xmath25 i.i.d samples from @xmath125 , @xmath131 \leq { \mathbb{e}}[n^{\lambda}(x_1 , \dots , x_{\ell - k})n^{\lambda}(y_1 , \dots , y_k)].\ ] ] now applying lemma [ lem : vap1 ] , we see that @xmath131 \leq { \mathbb{e}}[n^{\lambda}(x_1 , \dots , x_{\ell -k})(k+1)^{r^2/\delta^2 + 1 } ] .\ ] ] moving @xmath132 outside this expression , @xmath131 \leq { \mathbb{e}}[n^{\lambda}(x_1 , \dots , x_{\ell -k})](k+1)^{r^2/\delta^2 + 1 } .\ ] ] note that @xmath133 is always bounded above by @xmath134 and that the random variables @xmath135 $ ] are i.i.d .
let @xmath136 $ ] , and note that @xmath137 is the sum of @xmath25 independent bernoulli variables .
moreover , by markov s inequality , @xmath138 \,\ , \leq \,\ , \frac{{\mathbb{e}}[\|z_i\|^2]}{r^2 } , \ ] ] and therefore @xmath139 .
in addition , @xmath140 \leq { \mathbb{e}}[2^{\ell - k}].\ ] ] let @xmath141 $ ] denote an indicator variable . @xmath142 $ ] can be written as @xmath143 } ] = ( 1+\rho)^\ell \leq e^{\rho \ell}.\ ] ] putting everything together , we see that [ eqn : jensen1 ] [ n^(z_1 , , z _ ) ] ( ( ) ^2 + ( + 1 ) ( + 1 ) ) .
if we substitute @xmath144 , it follows that @xmath145 \\ & \leq & \left(\ell^{\frac{1}{2}}\left(\frac{r}{\delta}\right ) + 1\right ) ( 1+\ln(\ell+1 ) ) .\end{aligned}\ ] ] for ease of reference , we note the following easily established fact about @xmath146 .
this lemma is used in the proof of theorem [ thm : ddb_hs ] above and theorem [ thm : ddb_bs ] below .
[ lem : submul ] let @xmath147 be a partition of the data @xmath148 into two parts . then
, @xmath146 is submultiplicative in the following sense : @xmath149 this holds because any partition of @xmath150 into two parts by an element @xmath151 induces such a partition for the sets @xmath152 and @xmath153 , and for any pair of partitions of @xmath152 and @xmath154 , there is at most one partition of @xmath148 that induces them . as an intermediate step in the proof of theorem [ thm : ddb_hs ] , we needed a bound on the vc dimension of a gap - tolerant classifier within a ball of fixed radius .
lemma [ lem : vap1 ] below provides such a bound and is due to vapnik @xcite .
note , though , that in the course of his proof ( see @xcite , page 353 . ) , vapnik states , without further justification , that due to symmetry the set of points that is extremal in the sense of being the hardest to shatter with gap - tolerant classifiers is the regular simplex .
attention has also been drawn to this fact by burges ( @xcite , footnote 20 ) , who mentions that a rigorous proof of this fact seems to be absent .
vapnik s claim has since been proved by hush and scovel @xcite . here
, we provide a new proof of lemma [ lem : vap1 ] . it is simpler than previous proofs , and in section [ sxn :
gapbs ] we will see that it generalizes to cases when the margin is measured with norms other than @xmath118 . [ vc dimension ; upper bound ; hilbert space ] [ lem : vap1 ] in a hilbert - space , the vc dimension of a gap - tolerant classifier whose margin is @xmath12 over a ball of radius @xmath13 can by bounded above by @xmath155 .
suppose the vc dimension is @xmath11 .
then there exists a set of @xmath11 points @xmath156 in @xmath157 that can be completely shattered using gap - tolerant classifiers .
we will consider two cases , first that @xmath11 is even , and then that @xmath11 is odd .
first , assume that @xmath11 is even , i.e. , that @xmath158 for some positive integer @xmath159 .
we apply the probabilistic method to obtain a upper bound on @xmath11 .
note that for every set @xmath160 $ ] , the set @xmath161 can be separated from @xmath162 using a gap - tolerant classifier .
therefore the distance between the centroids ( respective centers of mass ) of these two sets is greater or equal to @xmath163 . in particular , for each @xmath164 having @xmath165 elements , @xmath166 suppose now that @xmath164 is chosen uniformly at random from the @xmath167 sets of size @xmath159 .
then , @xmath168 \\ & = & k^{-2}\left\{\frac{2k+1}{2k}\sum_{i=1}^n \|x_i\|^2 - \frac{\|\sum_1^n x_i\|^2}{2k } \right\ } \\ & \leq & \frac{4(n+1)}{n^2 } r^2 .\end{aligned}\ ] ] therefore , @xmath169 and so @xmath170 next , assume that @xmath11 is odd .
we perform a similar calculation for @xmath171 .
as before , we average over all sets @xmath164 of cardinality @xmath159 the squared distance between the centroid of @xmath172 and the centroid ( center of mass ) of @xmath173 . proceeding as before , @xmath174 \\ & = & \frac{\sum_{i=1}^n \|x_i\|^2 ( 1 + \frac{1}{2n } ) -
\frac{1}{2n}\|\sum_{1 \leq i \leq n } x_i\|^2}{k(k+1 ) } \\ & \leq & \frac{\sum_{i=1}^n \|x_i\|^2 ( 1 + \frac{1}{2n})}{k(k+1 ) } \\ & = & \frac{4k+3}{2k(2k+1)(k+1)}\{(2k+1)r^2\ } \\ & < & \frac{4r^2}{n-1 } .\end{aligned}\ ] ] therefore , @xmath175 . for a random data sample @xmath120 , ^2 & & _ i=1^n ( c i^-)^2 + & & c^2 ( 2 ) , where @xmath176 is the riemann zeta function .
the theorem then follows from theorem [ thm : margin_hs ] . a diffusion map for the graph @xmath177 is the feature map that associates with a vertex @xmath120 , the feature vector @xmath178 , when the eigenfunctions corresponding to the top @xmath179 eigenvalues are chosen .
let @xmath34 be the uniform distribution on @xmath180 and @xmath181 .
we note that if the @xmath182 are normalized eigenfunctions , _
i.e. _ , @xmath183 ^2 & = & 1 .
the above inequality holds because the eigenvalues have magnitudes that are less or equal to @xmath33 : @xmath184 the theorem then follows from theorem [ thm : margin_hs ] .
in this section , we state and prove theorem [ thm : ddb_bs ] , our main result regarding an upper bound for the annealed entropy of gap - tolerant classifiers in a banach space .
we start in section [ sxn : gapbs_prelim ] with some technical preliminaries ; then in section [ sxn : gapbs_vcdim ] we bound the vc dimension of gap - tolerant classifiers in banach spaces over a ball of radius @xmath13 ; and finally in section [ sxn : gapbs_annent ] we prove theorem [ thm : ddb_bs ] . we include this result for completeness since it is of theoretical interest ; since it follows using similar methods to the analogous results for hilbert spaces that we presented in section [ sxn : gaphs ] ; and since this result is of potential practical interest in cases where modeling the relationship between data elements as a dot product in a hilbert space is too restrictive , _ e.g. _ , when the data are extremely sparse and heavy - tailed . for recent work in
machine learning on banach spaces , see @xcite .
recall the definition of a banach space from definition [ def : rad ] above .
we next state the following form of the chernoff bound , which we will use in the proof of lemma [ lem : banach_ub ] below .
[ chernoff bound ] [ lem : chernoff ] let @xmath185 be discrete independent random variables such that @xmath186=0 $ ] for all @xmath64 and @xmath187 for all @xmath64 .
let @xmath188 for all @xmath64 and @xmath189 be the variance of @xmath90
. then @xmath190 \leq 2e^{-\lambda^2/4}\ ] ] for any @xmath191 .
the idea underlying our new proof of lemma [ lem : vap1 ] ( of section [ sxn : gaphs_vcdim ] , and that provides an upper bound on the vc dimension of a gap - tolerant classifier in hilbert spaces ) generalizes to the case when the the gap is measured in more general banach spaces
. we state the following lemma for a banach space of type @xmath50 with type constant @xmath48 .
e.g. _ , that @xmath51 for @xmath192 is a banach space of type @xmath193 and type constant @xmath33 .
[ vc dimension ; upper bound ; banach space ] [ lem : banach_ub ] in a banach space of type @xmath50 and type constant @xmath48 , the vc dimension of a gap - tolerant classifier whose margin is @xmath12 over a ball of radius @xmath13 can by bounded above by @xmath194 since a general banach space does not possess an inner product , the proof of lemma [ lem : vap1 ] needs to be modified here . to circumvent this difficulty
, we use inequality ( [ ineq : rad ] ) determining the rademacher type of @xmath65 .
this , while permitting greater generality , provides weaker bounds than previously obtained in the euclidean case .
note that if @xmath195 , then by repeated application of the triangle inequality , @xmath196 this shows that if we start with @xmath197 having norm @xmath198 , @xmath199 for all @xmath64 .
the property of being shattered by gap - tolerant classifiers is translation invariant .
then , for @xmath200 $ ] , it can be verified that @xmath201 the rademacher inequality states that [ ineq : rad]_t^p _
i=1^n x_i^p.using the version of chernoff s bound in lemma [ lem : chernoff ] [ twelve ] [ |_i=1^n _ i| ] 1- 2 e^-^2/4 . we shall denote the above event by @xmath202 .
now , let @xmath197 be @xmath11 points in @xmath65 with a norm less or equal to @xmath13 .
let @xmath203 as before .
@xmath204\\ & \geq & { \mathbb{e}}_\epsilon [ \|\epsilon_i ( x_i-\mu)\|^p| e_{\lambda}]\,\,{\mathbb{p}}[e_{\lambda}]\\ & \geq & { \mathbb{e}}_\epsilon [ ( n - \lambda^2)^p(2\delta)^p ( 1 - 2 e^{-\lambda^2/4})]\,\,\end{aligned}\ ] ] the last inequality follows from ( [ eleven ] ) and ( [ twelve ] ) .
we infer from the preceding sequence of inequalities that @xmath205 the above is true for any @xmath206 , by the conditions in the chernoff bound stated in lemma [ lem : chernoff ] .
if @xmath207 , choosing @xmath208 equal to @xmath113 gives us @xmath209 therefore , it is always true that @xmath210 finally , for completeness , we next state a lower bound for vc dimension of gap - tolerant classifiers when the margin is measured in a norm that is associated with a banach space of type @xmath47 $ ] . since we are interested only in a lower bound , we consider the special case of @xmath211 .
note that this argument does not immediately generalize to banach spaces of higher type because for @xmath212 , @xmath51 has type @xmath213 .
[ vc dimension ; lower bound ; banach space ] [ lem : banach_lb ] for each @xmath47 $ ] , there exists a banach space of type @xmath50 such that the vc dimension of gap - tolerant classifiers with gap @xmath12 over a ball of radius @xmath13 is greater or equal to @xmath214 further , this bound is achieved when the space is @xmath51 .
we shall show that the first @xmath11 unit norm basis vectors in the canonical basis can be shattered using gap - tolerant classifiers , where @xmath215 .
therefore in this case , the vc dimension is @xmath216 .
let @xmath217 be the @xmath218 basis vector . in order to prove that the set @xmath219 is shattered , due to symmetry under permutations , it suffices to prove that for each @xmath159 , @xmath220 can be separated from @xmath221 using a gap - tolerant classifier .
points in @xmath51 are infinite sequences @xmath222 of finite @xmath51 norm .
consider the hyperplane @xmath223 defined by @xmath224 .
clearly , it separates the sets in question .
we may assume @xmath217 to be @xmath225 , replacing if necessary , @xmath159 by @xmath226 .
let @xmath227 clearly , all coordinates @xmath228 of @xmath120 are @xmath229 . in order to get a lower bound on the @xmath51 distance
, we use the power - mean inequality : if @xmath192 , and @xmath230 , @xmath231 this implies that @xmath232 for @xmath233 , the type of @xmath51 is @xmath213 @xcite . since @xmath234 is a decreasing function of @xmath50 in this regime , we do not recover any useful bounds .
the following theorem is our main result regarding an upper bound for the annealed entropy of gap - tolerant classifiers in banach spaces .
note that the @xmath118 bound provided by this theorem is slightly weaker than that provided by theorem [ thm : ddb_hs ] .
note also that it may seem counter - intuitive that in the case of @xmath118 ( _ i.e. _ , when we set @xmath235 ) , the dependence of @xmath12 is @xmath236 , which is weaker than in the vc bound , where it is @xmath237 .
the explanation is that the bound on annealed entropy here depends on the number of samples @xmath25 , while the vc dimension does not .
therefore , the weaker dependence on @xmath12 is compensated for by a term that in fact tends to @xmath238 as the number of samples @xmath239 .
[ annealed entropy ; upper bound ; banach space ] [ thm : ddb_bs ] let @xmath41 be a probability measure on a banach space @xmath68 of type @xmath50 and type constant @xmath48 .
let @xmath240 , and let @xmath241 .
if @xmath242 , then the annealed entropy of gap - tolerant classifiers in @xmath68 , where the gap is @xmath12 , is @xmath243 the proof of this theorem parallels that of theorem [ thm : ddb_hs ] , except that here we use lemma [ lem : banach_ub ] instead of lemma [ lem : vap1 ] .
we include the full proof for completeness .
let @xmath25 independent , identically distributed ( i.i.d ) samples @xmath124 be chosen from @xmath125 .
we partition them into two classes : @xmath126 and @xmath127 our objective is to bound from above the annealed entropy @xmath128 $ ] . by lemma [ lem : submul ] , @xmath129 is sub - multiplicative .
therefore , @xmath130 taking an expectation over @xmath25 i.i.d samples from @xmath125 , @xmath131 \leq { \mathbb{e}}[n^{\lambda}(x_1 , \dots , x_{\ell - k})n^{\lambda}(y_1 , \dots , y_k)].\ ] ] now applying lemma [ lem : banach_ub ] , we see that @xmath131 \leq { \mathbb{e}}[n^{\lambda}(x_1 , \dots , x_{\ell -k})(k+1)^{(3tr/\delta)^{\frac{p}{p-1}}+ 64 } ] .\ ] ] moving @xmath244 outside this expression , @xmath131 \leq { \mathbb{e}}[n^{\lambda}(x_1 , \dots , x_{\ell -k})](k+1)^{(3tr/\delta)^{\frac{p}{p-1}}+ 64 } .\ ] ] note that @xmath133 is always bounded above by @xmath134 and that the random variables @xmath135 $ ] are i.i.d .
let @xmath136 $ ] , and note that @xmath137 is the sum of @xmath25 independent bernoulli variables .
moreover , by markov s inequality , @xmath138 \,\ , \leq \,\ , \frac{{\mathbb{e}}[\|z_i\|^\gamma]}{r^\gamma } , \ ] ] and therefore @xmath245 .
in addition , @xmath140 \leq { \mathbb{e}}[2^{\ell - k}].\ ] ] let @xmath141 $ ] denote an indicator variable .
@xmath142 $ ] can be written as @xmath143 } ] = ( 1+\rho)^\ell \leq e^{\rho \ell}.\ ] ] putting everything together , we see that [ eqn : jensen2 ] [ n^(z_1 , , z _ ) ] ( ( ) ^ + ( k+1)(64 + ) ^ ) . by setting @xmath246 and adjusting @xmath13 so that @xmath247 we see that @xmath248 thus , it follows that @xmath249 \\ & \leq & \left(\eta^{-\eta}(1-\eta)^{-1+\eta } \left(\frac{\ell}{\ln(\ell+1 ) } \left(\frac{3tr}{\delta}\right)^\gamma\right)^\eta + 64\right ) \ln(\ell+1 ) .\end{aligned}\ ] ]
in recent years , there has been a considerable amount of somewhat - related technical work in a variety of settings in machine learning .
thus , in this section we will briefly describe some of the more technical components of our results in light of the existing related literature .
* techniques based on the use of rademacher inequalities allow one to obtain bounds without any assumption on the input distribution as long as the feature maps are uniformly bounded .
e.g. _ , @xcite .
viewed from this perspective , our results are interesting because the uniform boundedness assumption is not satisfied in either of the two settings we consider , although those settings are ubiquitous in applications . in the case of heavy - tailed data ,
the uniform boundedness assumption is not satisfied due to the slow decay of the tail and the large variability of the associated features . in the case of spectral learning ,
uniform boundedness assumption is not satisfied since for arbitrary graphs one can have localization and thus large variability in the entries of the eigenvectors defining the feature maps . in both case , existing techniques based on rademacher inequalities or vc dimensions fail to give interesting results , but we show that dimension - independent bounds can be achieved by bounding the annealed entropy . * a great deal of work has focused on using diffusion - based and spectral - based methods for nonlinear dimensionality reduction and the learning a nonlinear manifold from which the data are assumed to be drawn @xcite .
these results are very different from the type of learning bounds we consider here .
for instance , most of those learning results involve convergence to an hypothesized manifold laplacian and not of learning process itself , which is what we consider here .
* work by bousquet and elisseeff @xcite has focused on establishing generalization bounds based on stability .
it is important to note that their results assume a given algorithm and show how the generalization error changes when the data are changed , so they get generalization results for a given algorithm .
our results make no such assumptions about working with a given algorithm .
* gurvits @xcite has used rademacher complexities to prove upper bounds for the sample complexity of learning bounded linear functionals on @xmath51 balls .
the results in that paper can be used to derive an upper bound on the vc dimension of gap - tolerant classifiers with margin @xmath12 in a ball of radius @xmath13 in a banach space of rademacher type @xmath47 $ ] .
constants were not computed in that paper , therefore our results do not follow . moreover , our paper contains results on distribution specific bounds which were not considered there .
finally , our paper considers the application of these tools to the practically - important settings of spectral kernels and heavy - tailed data that were not considered there .
we have considered two simple machine learning problems motivated by recent work in large - scale data analysis , and we have shown that although traditional distribution - independent methods based on the vc - dimension fail to yield nontrivial sampling complexity bounds , we can use distribution - dependent methods to obtain dimension - independent learning bounds . in both cases , we take advantage of the fact that , although there may be individual data points that are `` outlying , '' in aggregate their effect is not too large . due to the increased popularity of vector space - based methods ( as opposed to more purely combinatorial methods ) in machine learning in recent years , coupled with the continued generation of noisy and poorly - structured data ,
the tools we have introduced are likely promising more generally for understanding the effect of noise and noisy data on popular machine learning tasks .
in this appendix section , we state and prove a second result for dimension - independent learning from data in which the feature map exhibits a heavy - tailed decay .
the heavy - tailed model we consider here is different than that considered in theorem [ thm : learn_heavytail ] , and thus we are able to prove bounds for exact ( as opposed to maximum margin ) learning . nevertheless , the techniques are similar , and thus we include this result in this paper for completeness .
consider the following toy model for classifying web pages using keywords .
one approach to this problem could be to associate with each web page the indicator vector corresponding to all keywords that it contains .
the dimension of this feature space is the number of possible keywords , which is typically very large , and empirical evidence indicates that the frequency of words decays in a heavy - tailed manner .
thus the vc dimension of the feature space is very large , and in a distribution - free setting it is not possible to classify data in such a feature space unless the number of samples is of the order of the vc dimension .
more generally , one might be interested in a bipartite graph , _
e.g. _ , an `` advertiser - keyword '' or `` author - to - paper '' graph , in which the nodes are the stated entities and the edges represent some sort of `` interaction '' between the entities , in which case similar issues arise . here , we show that if the probability that the @xmath19 keyword in the above toy example is present is heavy - tailed as a function of @xmath64 , then the sample complexity of the binary classification problem is dimension - independent .
more precisely , the following theorem provides a dimension - independent ( _ i.e. _ , independent of the size @xmath11 of the graph and the dimension of the feature space ) upper bound on the number of samples needed to learn by erm , with a given accuracy and confidence , a linear hyperplane that classifies heavy - tailed data into positive and negative labels , under the assumption that the probability of the @xmath19 coordinate of a random data point being non - zero is less than @xmath20 for some @xmath21 .
the proof of this result proceeds by providing providing a dimension - independent upper bound on the annealed entropy of the class of linear classifiers in @xmath250 , and then appealing to theorem [ thm : vapnik ] relating the annealed entropy to the generalization error .
* remark : * note that although the generalization bound provided by the following theorem seems to be pessimistic in @xmath97 , the dependence on @xmath97 is tight , at least as @xmath97 tends to @xmath33 . clearly , when @xmath251 , the expected number of @xmath33 s in a random sample becomes asymptotically equal to @xmath252 , where @xmath11 is the dimension , in which case , we do not expect a sample complexity that is dimension - independent .
[ bounds for heavy - tailed data ] [ thm : learnht_exact ] let @xmath41 be a probability distribution in @xmath250 .
ci^{-\alpha}$ ] for some absolute constant @xmath17 , with @xmath18 .
then , the annealed entropy of ordinary linear hyperplane classifiers is h_ann^ ( ) & & ( ^+1)consequently , the minimum number of random samples @xmath254 needed to learn , by erm , a classifier whose risk differs from the minimum risk @xmath96 by @xmath255 with probability @xmath256 is less than or equal to 2 ( ( + ) ) ^ ( ( ( + ) ) ^ ) .
let the event that a sample @xmath257 has a non - zero coordinate @xmath258 for some @xmath259 be denoted @xmath260 .
the probability of this event can be bounded as follows . if @xmath261 and @xmath262 , then [ e_i ] & = & [ k > ^1/ , z_ik 0 ] + & & c_i = k+1^i^- + & & .
we partition the @xmath112 into two classes : @xmath263 and @xmath264 @xmath146 is sub - multiplicative by lemma [ lem : submul ] .
taking an expectation over @xmath25 i.i.d samples from @xmath41 , [ ( z_1 , , z _ ) ] & & [ ( x_1 , , x_-m)(y_1 , , y_m)]the dimension of the span of @xmath265 is at most @xmath159 , and by a result from vc theory ( @xcite , page 159 ) we have @xmath266 then , @xmath267 \leq { \mathbb{e}}[\nl(x_1 , \dots , x_{\ell -m})e m^k].\ ] ] moving @xmath268 outside this expression , @xmath267 \leq { \mathbb{e}}[\nl(x_1 , \dots , x_{\ell -k})]e m^k .\ ] ] note that @xmath269 is always bounded above by @xmath134 and that the events @xmath270 are independent identically distributed .
let @xmath271 $ ] , and note that @xmath137 is the sum of @xmath25 independent @xmath50-bernoulli variables .
in addition , @xmath272 \leq { \mathbb{e}}[2^{\ell - k}],\ ] ] and @xmath142 $ ] can be written as _
i=1^(1 + [ e_i ] ) & = & ( 1+p)^ + & & e^p + & = & e^().putting everything together , we see that @xmath267 \leq e(\ell)^k
e^{\frac{c \ell\ , k^{-a + 1}}{\a - 1 } } .\ ] ] since @xmath273 , we see that h_ann^ ( ) & = & [ ( z_1 , , z _ ) ] + & & ( ^+1 ) ( ) . in order to obtain sample complexity bounds , we need to apply theorem [ thm : vapnik ] and substitute the above expression for annealed entropy . for the probability that the error of erm exceeds @xmath274 to be less than @xmath275 ( where @xmath98 is the optimal classifier ) ,
it is sufficient that @xmath25 satisfy @xmath276 for this to be true , it is enough that @xmath277 a calculation shows that @xmath278 is a value of @xmath25 that satisfies the previous expression . l. gurvits .
a note on a scale - sensitive dimension of linear bounded functionals in banach spaces .
in _ proceedings of the 8th international conference on algorithmic learning theory _ , pages 352363 , 1997 . v. koltchinskii and d. panchenko .
rademacher processes and bounding the risk of function learning . in e.
gine , d. mason , and j. wellner , editors , _ high dimensional probability ii _ , pages 443459 .
birkhauser , 2000 .
j. leskovec , k.j .
lang , a. dasgupta , and m.w .
statistical properties of community structure in large social and information networks . in _
www 08 : proceedings of the 17th international conference on world wide web _ ,
pages 695704 , 2008 .
j. leskovec , k.j .
lang , and m.w .
mahoney . empirical comparison of algorithms for network community detection . in _ www 10 : proceedings of the 19th international conference on world wide web _ , pages 631640 , 2010 .
l. k. saul , k. q. weinberger , j. h. ham , f. sha , and d. d. lee .
spectral methods for dimensionality reduction . in o.
chapelle , b. schoelkopf , and a. zien , editors , _ semisupervised learning _ , pages 293308 .
mit press , 2006 .
|
two ubiquitous aspects of large - scale data analysis are that the data often have heavy - tailed properties and that diffusion - based or spectral - based methods are often used to identify and extract structure of interest .
perhaps surprisingly , popular distribution - independent methods such as those based on the vc dimension fail to provide nontrivial results for even simple learning problems such as binary classification in these two settings . in this paper
, we develop distribution - dependent learning methods that can be used to provide dimension - independent sample complexity bounds for the binary classification problem in these two popular settings .
in particular , we provide bounds on the sample complexity of maximum margin classifiers when the magnitude of the entries in the feature vector decays according to a power law and also when learning is performed with the so - called diffusion maps kernel .
both of these results rely on bounding the annealed entropy of gap - tolerant classifiers in a hilbert space .
we provide such a bound , and we demonstrate that our proof technique generalizes to the case when the margin is measured with respect to more general banach space norms .
the latter result is of potential interest in cases where modeling the relationship between data elements as a dot product in a hilbert space is too restrictive .
|
after the discovery of the higgs boson in july 2012 @xcite , a solution to the hierarchy problem the question why the electroweak scale is so much lower than the planck scale is more urgently needed than ever . among the numerous possibilities to solve the hierarchy problem ,
the most popular approach is low - scale supersymmetry .
a promising alternative is given by models with a warped extra dimension @xcite , in which the standard model ( sm ) is embedded in a slice of anti - de sitter space , while the higgs field is localized on the `` infra - red ( ir ) brane '' , one of the two four - dimensional hyper - manifolds that bound the extra dimension .
these so - called randall - sundrum ( rs ) models can provide a natural explanation for this vast hierarchy , since the fundamental ultra - violet ( uv ) cutoff is the warped planck scale , which near the ir brane takes values in the tev range . moreover
, by allowing the fermion fields to propagate into the bulk , warped extra - dimension models can also provide an explanation for the hierarchies observed in the flavor sector @xcite and the smallness of flavor - changing neutral currents @xcite .
unfortunately , none of the kaluza - klein ( kk ) particles predicted in extra - dimensional extensions of the sm have been observed yet , and electroweak precision measurements indicate that these particles are probably too massive for a direct detection at the lhc .
thus , indirect searches like the precision measurements of the higgs - boson couplings to sm particles , which are accessible via studies of both the higgs production cross section and its decay rates , become more and more attractive . concerning warped extra dimensions , especially the loop - mediated higgs couplings to gluons and photons
could give hints about the existence of additional kk particles . while the gluon - fusion process has been discussed extensively in several works @xcite ,
the present paper focuses on the higgs decay into two photons , which was investigated in @xcite .
the first analysis of the effects of the kk tower of the @xmath2 boson on the @xmath1 amplitude was performed in @xcite .
the first complete calculation of the @xmath7 decay rate , in which both the yukawa couplings to the @xmath8-even and @xmath8-odd fermions were included , was performed in @xcite .
it was found in this paper that the higgs decay rate into two photons is enhanced relative to the sm due to the effect of the kk fermions , which turned out to give the dominating correction . at about the same time , an independent analysis of the @xmath7 decay rate came to the opposite conclusion @xcite .
an explanation for these deviating results was first given in @xcite , followed by a five - dimensional ( 5d ) analysis in @xcite , which found that the two results belong to two different scenarios of the rs model .. up to different factors for the color multiplicity and electric charges , an analogous discussion holds for the quark and lepton kk - tower contributions to @xmath7 .
] while these papers focused on the contribution stemming from the fermionic kk states , in the present work we provide a detailed analysis of the bosonic loop contributions to the @xmath7 amplitude , which in unitary gauge stem from the @xmath2 bosons and their kk excitations .
the advantages of this approach are that we are able to derive an exact result , which includes the full dependence on the higgs - boson mass and holds to all orders in @xmath3 , where @xmath9 is the higgs vacuum expectation value ( vev ) , and @xmath10 sets the mass scale for the low - lying kk excitations of the sm particles .
it is also straightforward to extend our formulas to the case where the higgs boson lives in the bulk of the extra dimension .
our approach also allows us to carefully study the effects of the fifth components of the gauge fields , whose profiles are discontinuous on the ir brane , similar to the @xmath8-odd fermion profiles which indeed require a careful treatment .
our paper is structured as follows : in section [ sec : pre ] we define the set - up and comment on the necessity to distinguish between the so - called brane - higgs and narrow bulk - higgs scenarios when calculating the fermionic loop contributions to the @xmath11 and @xmath1 amplitudes . in section [ sec : analysis ] we discuss the general structure of the @xmath7 amplitude and summarize the results for the fermionic contributions from charged quarks and leptons propagating in the loop .
we then focus on the bosonic loop contributions , calculate them in the kk - decomposed theory and show that the result for the contributions of each individual kk mode is gauge invariant .
in the next step we resum the kk towers and derive an exact formula for the @xmath7 amplitude in terms of an overlap integral over the higgs - boson profile and the transverse part of the 5d gauge - boson propagator , including the exact dependence on the higgs - boson mass . to the best of our knowledge , such a formula has not been obtained before .
we derive an explicit , closed expression for the 5d propagator and show that the overlap integral is insensitive to the precise details of shape of the higgs - boson profile , once this profile is localized very close to the ir brane . by expanding our results in powers of @xmath3 , we can identify the contributions from the @xmath2 bosons ( with modified couplings to the higgs boson ) and their kk towers , confirming the results of @xcite . in section [ sec : custodial ]
we generalize our findings to an extended version of the rs model with a custodial symmetry protecting electroweak precision observables @xcite .
again , we obtain an exact formula for the @xmath7 amplitude and , for the first time , for the 5d gauge - boson propagator in the custodial rs model .
when expanded to order @xmath3 , our findings for the contributions of the @xmath2 boson and its kk excitations are consistent with the findings of @xcite .
phenomenological implications of our results in the context of the latest lhc data are discussed in section [ sec : pheno ] , where we study the higgs decay into two photons in two different versions of the minimal and the custodially protected rs model .
we illustrate the magnitude of the effects as a function of the mass of the lightest kk gluon state and the scale of the 5d yukawa couplings , and derive the regions in parameter space that are already excluded by recent lhc measurements .
our main results are summarized in the conclusions .
we focus on rs models where the electroweak symmetry - breaking sector is localized on or near the ir brane .
the extra dimension is chosen to be an @xmath12 orbifold parametrized by a coordinate @xmath13 $ ] , with two branes localized on the orbifold fixed - points at @xmath14 ( uv brane ) and @xmath15 ( ir brane ) .
the rs metric reads @xcite @xmath16 where @xmath17 with @xmath18 is referred to as the warp factor , and the size @xmath19 and curvature @xmath20 of the extra dimension are assumed to be of planck size , @xmath21 .
the quantity @xmath22 measures the size of the extra dimension and is chosen to be @xmath23 in order to explain the hierarchy between the planck and the tev scales . with the help of the curvature @xmath20 and the warp factor evaluated at the ir brane , @xmath24 , one defines the kk mass scale as @xmath25 . on the right - hand side of we have introduced the dimensionless coordinate @xmath26 defined by @xmath27 $ ] , by the rescaling @xmath28 . ] which will be used throughout this work .
our paper deals with the minimal rs model based on the sm gauge group , as well as with an extended rs model with a larger bulk gauge group , which after electroweak symmetry breaking includes an @xmath29 custodial symmetry protecting the @xmath30 parameter and suppressing larger corrections to the @xmath31 couplings @xcite . in both versions of the rs model , all fermions and gauge bosons are allowed to propagate into the bulk , resulting in infinite towers of heavy kk copies of the sm particles .
in contrast , the scalar sector is assumed to reside on or near the ir brane so as to provide a solution to the hierarchy problem .
we recall that rs models are effective field theories valid up to a position - dependent uv cutoff @xcite @xmath32 at this scale , gravity becomes strong and the model needs to be uv - completed into a theory of quantum gravity . the fact that rs models are defined with an inherent uv cutoff can be used to distinguish between different scenarios for the localization of the higgs sector .
we shall consider a scalar field @xmath33 localized on or very near the ir brane .
its profile along the extra dimension is described by a normalized distribution @xmath34 with width @xmath35 . for @xmath36
this profile becomes a @xmath37-function , corresponding to a strictly brane - localized scalar field .
following @xcite , we introduce the characteristic width of the higgs field along the extra dimension as @xmath38 .
if the inverse width is larger than the inherent uv cutoff near the ir brane , @xmath39 , then the higgs profile can not be resolved by the high - momentum modes contained in the effective lagrangian of the rs model , and hence for all practical purposes such a scalar field can be regarded as a _ brane - localized higgs field_. on the other hand , if the inverse higgs width is smaller than the cutoff scale , then the modes of the effective theory can resolve its profile , and we speak of a _ bulk higgs field_. while calculations in generic bulk - higgs models are rather complicated , it has been shown in @xcite that there is the possibility of obtaining analytical results for the special case of a _ narrow bulk - higgs field _ , whose inverse width is such that @xmath40 .
as has been explained in @xcite , the fermionic loop contributions to the @xmath11 and @xmath1 amplitudes are sensitive to the details of the localization mechanism , and indeed the results obtained in a brane - localized higgs scenario @xcite differ significantly from those derived for a narrow bulk - higgs field @xcite .
we thus need to distinguish between the two types of localization mechanisms in our phenomenological analysis .
as we will show , however , the bosonic contributions to the @xmath1 amplitude are insensitive to the precise localization of the scalar sector and approach an unambiguous result in the limit where @xmath35 .
our goal is to calculate the @xmath7 decay amplitude entirely in terms of the 5d propagators for both gauge bosons and fermions .
while the contributions from quarks and charged leptons can be easily deduced from the corresponding results for the @xmath11 amplitude , a detailed consideration of the gauge - boson contribution has not yet been performed in 5d language .
our approach in the present work is the following : in a first step , we calculate the bosonic contributions to the @xmath7 amplitude in the kk - decomposed , 4d effective theory and show that at each kk level the sum of all diagrams is gauge - invariant .
the only contributing diagrams in unitary gauge are those with vector bosons propagating in the loop .
we can then rewrite the amplitude , summed over kk states , as an expression involving the 5d gauge - boson propagator in the mixed momentum - position representation @xcite .
we show that in the limit of a very narrow higgs profile the amplitude approaches an unambiguous value , which is insensitive to the details of the higgs localization mechanism . at the end of this section
, we employ our exact results to derive expressions for the contributions of the zero modes ( the standard @xmath2 bosons ) and their infinite towers of kk excitations to the @xmath1 amplitude .
we begin with the calculation in the minimal rs model with the sm gauge group in the bulk , broken to @xmath41 on the ir brane , where the higgs field develops a vev .
details for the implementation of the higgs , gauge - boson , and gauge - fixing sectors in the context of this model and using our notations have been given in @xcite , while appendix [ app : fr ] includes a summary of the relevant feynman rules needed for our analysis .
here , it suffices to mention that we decompose the 5d @xmath2-boson field into 4d mass eigenstates @xmath42 where @xmath43 are the kk modes of the @xmath2 bosons with masses @xmath44 .
the scalar particles @xmath45 are `` unphysical '' in the sense that they provide the longitudinal degrees of freedom of the @xmath2 bosons ( @xmath46 ) and the massive @xmath2-boson kk modes ( @xmath47 ) , and thus they can be gauged away .
indeed , the scalar fields @xmath48 mix with the charged goldstone bosons arising from the higgs sector . assuming for the time being that the scalar sector is localized on the ir brane , we parameterize the higgs doublet after electroweak symmetry breaking in the usual form @xmath49 throughout this paper , @xmath9 denotes the higgs vev in the rs model , which differs from the sm value @xmath50 by a small amount @xcite .
we determine @xmath9 from the shift to the fermi constant derived in the rs model by considering ( at tree level ) the effect of the exchange of the infinite tower of kk gauge bosons on the rate for muon decay , using the definition @xmath51 .
this yields @xcite @xmath52 where @xmath53 in the minimal rs model .
when we generalize our analysis to the case of an extended rs model with a custodial symmetry in section [ sec : custodial ] , this relation will still hold , but the parameter @xmath54 will then take on a non - trivial value .
the decomposition of the scalar fields @xmath55 into the mass eigenstates @xmath45 reads @xcite @xmath56 where @xmath57 is the leading contribution to the @xmath2-boson mass in an expansion in powers of @xmath3 .
the relation between the two parameters can be written as @xcite @xmath58 , \ ] ] where once again @xmath53 in the minimal rs model .
since the profile of the zero mode is flat up to corrections of order @xmath3 , it follows that @xmath59 is close to 1 , and hence the fields @xmath55 coincide with @xmath60 to leading order . mixing effects arise at order @xmath3 and higher
. note also that one can adjust the gauge - fixing lagrangian so as to cancel any mixings between @xmath61 and the scalar fields @xmath48 and @xmath55 @xcite .
the one - loop feynman diagrams contributing to the @xmath1 decay amplitude are shown in figure [ fig : alldiagrams ] for a general @xmath0 gauge . in the subsequent section
we will demonstrate that the full amplitude is gauge invariant . in the unitary gauge only the diagrams @xmath62@xmath63 contribute .
in order to present our results , we find it convenient to parametrize the @xmath7 amplitude , including the contributions from sm particles , by means of two wilson coefficients @xmath64 and @xmath65 define via @xmath66 where @xmath67 with @xmath68 .
each wilson coefficient can be written as a sum of three terms , @xmath69 where in a general gauge @xmath70 includes the bosonic contributions from gauge bosons , scalar bosons , and ghosts .
the calculation of these bosonic contributions is the main subject of this work .
the fermionic loop contributions due to virtual quarks and leptons shown in diagram @xmath71 can be readily deduced from expressions derived in @xcite .
they will be summarized in section [ subsec : fermions ] .
( origin ) at ( 0,0 ) ; ( 1,1 ) ( -1,0 ) ; ( 1,-1 ) ( -1,0 ) ; ( 1,-1 ) ( 1,1 ) ; ( -1,0 ) ( -3,0 ) ; ( 3,1.3 ) ( 1,1 ) ; ( 3,-1.3 ) ( 1,-1 ) ; ( -1,0 ) node [ vertex ] ; ( 1,1 ) node [ vertex ] ; ( 1,-1 ) node [ vertex ] ; ( -2,2 ) node @xmath71 ; ( -1,0.6 ) node @xmath26 ; ( 1,1.6 ) node @xmath72 ; ( 1,-1.6 ) node @xmath73 ; ( 2.7,0.6 ) node @xmath74 ; ( 2.7,-0.6 ) node @xmath75 ; ( 1,1 ) ( -1,0 ) ; ( 1,-1 ) ( -1,0 ) ; ( 1,-1 ) ( 1,1 ) ; ( -1,0 ) ( -3,0 ) ; ( 3,1.3 ) ( 1,1 ) ; ( 3,-1.3 ) ( 1,-1 ) ; ( -1,0 ) node [ vertex ] ; ( 1,1 ) node [ vertex ] ; ( 1,-1 ) node [ vertex ] ; ( -2,2 ) node @xmath76 ; ( -1,0 ) ( -3,0 ) ; ( 3,1.3 ) ( 1,0 ) ; ( 3,-1.3 ) ( 1,0 ) ; ( -1,0 ) node [ vertex ] ; ( 1,0 ) node [ vertex ] ; ( -2,2 ) node @xmath77 ; ( 1,0 ) arc ( 0:360:1 ) ; ( 1,1 ) ( -1,0 ) ; ( 1,-1 ) ( -1,0 ) ; ( 1,-1 ) ( 1,1 ) ; ( -1,0 ) ( -3,0 ) ; ( 3,1.3 ) ( 1,1 ) ; ( 3,-1.3 ) ( 1,-1 ) ; ( -1,0 ) node [ vertex ] ; ( 1,1 ) node [ vertex ] ; ( 1,-1 ) node [ vertex ] ; ( -2,2 ) node @xmath78 ; ( 0,1 ) ( -3,1 ) ; ( 3,1.3 ) ( 0,1 ) ; ( 3,-1.3 ) ( 0,-1 ) ; ( 0,1 ) node [ vertex ] ; ( 0,-1 ) node [ vertex ] ; ( 0,-1 ) arc ( -90:90:1 ) ; ( 0,-1 ) arc ( 270:90:1 ) ; ( -2,2 ) node @xmath79 ; ( 1,1 ) ( -1,0 ) ; ( 1,-1 ) ( -1,0 ) ; ( 1,-1 )
( 1,1 ) ; ( -1,0 ) ( -3,0 ) ; ( 3,1.3 )
( 1,1 ) ; ( 3,-1.3 )
( 1,-1 ) ; ( -1,0 ) node [ vertex ] ; ( 1,1 ) node [ vertex ] ; ( 1,-1 ) node [ vertex ] ; ( -2,2 ) node @xmath80 ; ( 1,1 ) ( -1,0 ) ; ( 1,-1 )
( -1,0 ) ; ( 1,-1 ) ( 1,1 ) ; ( -1,0 )
( -3,0 ) ; ( 3,1.3 )
( 1,1 ) ; ( 3,-1.3 )
( 1,-1 ) ; ( -1,0 ) node [ vertex ] ; ( 1,1 ) node [ vertex ] ; ( 1,-1 ) node [ vertex ] ; ( -2,2 ) node @xmath81 ; ( 1,1 ) ( -1,0 ) ; ( 1,-1 ) ( -1,0 ) ; ( 1,-1 ) ( 1,1 ) ; ( -1,0 ) ( -3,0 ) ; ( 3,1.3 )
( 1,1 ) ; ( 3,-1.3 )
( 1,-1 ) ; ( -1,0 ) node [ vertex ] ; ( 1,1 ) node [ vertex ] ; ( 1,-1 ) node [ vertex ] ; ( -2,2 ) node @xmath82 ; ( 1,1 ) ( -1,0 ) ; ( 1,-1 ) ( -1,0 ) ; ( 1,-1 ) ( 1,1 ) ; ( -1,0 ) ( -3,0 ) ; ( 3,1.3 )
( 1,1 ) ; ( 3,-1.3 )
( 1,-1 ) ; ( -1,0 ) node [ vertex ] ; ( 1,1 ) node [ vertex ] ;
( 1,-1 ) node [ vertex ] ; ( -2,2 ) node @xmath83 ; ( -1,0 ) ( -3,0 ) ; ( 3,1.3 ) ( 1,0 ) ; ( 3,-1.3 ) ( 1,0 ) ; ( -1,0 ) node [ vertex ] ; ( 1,0 ) node [ vertex ] ; ( 1,0 ) arc ( 0:360:1 ) ; ( -2,2 ) node @xmath84 ; ( -1,0 )
( -3,0 ) ; ( 3,1.3 )
( 0,1 ) ; ( 3,-1.3 ) ( 0,-1 ) ; ( 0,1 ) node [ vertex ] ; ( 0,-1 ) node [ vertex ] ; ( 1,0 ) arc ( 0:360:1 ) ; ( -2,2 ) node @xmath85 ; in our analysis we will also discuss the case of a very narrow higgs boson localized near the ir brane , where the higgs profile @xmath34 has a characteristic width @xmath86 subject to the condition @xmath87 . in principle
, such a scenario gives rise to a tower of _ physical _ scalar particles @xmath88 , which in some sense are the kk excitations of the charged components of the higgs doublet . as discussed in detail in @xcite ,
these fields are defined in terms of a gauge - invariant superposition of @xmath48 and @xmath55 .
it has been shown in the same reference that the effect of these heavy scalar particles on the @xmath1 amplitude is @xmath89 where @xmath90 , and the function @xmath91 \ , , \quad \mbox{with } \quad f(\tau ) = \arctan^2\frac{1}{\sqrt{\tau-1 } } \,,\ ] ] approaches 1 for @xmath92 . in the limit of a very narrow higgs profile
the couplings @xmath93 scale like @xmath94 , while the masses of the heavy scalar particles scale like @xmath95 .
it follows that @xmath96 , and hence this contribution decouples in the limit @xmath36 , as expected .
we will therefore not consider the corresponding feynman diagrams in our analysis .
the one - loop contributions to the @xmath1 amplitude due to the exchange of virtual quarks and leptons can be derived in a straightforward way from analogous results for the quark contributions to the @xmath11 amplitude , which were studied in @xcite . here
we will use expressions derived in our previous work @xcite , where a variety of rs models were considered using the same 5d approach employed in the present work .
all that is necessary is to include appropriate factor of color and electric charges .
the exact result can be written in the form @xmath97 , \\
c_{5\gamma}^q & = 2 n_c \sum_{f = u , d}\,q_q^2\,\int_0 ^ 1\!dx \int_0^{1-x}\!dy \left[\ , t_-^q(-xy m_h^2 ) - t_-^q(\lambda_{\rm tev}^2 ) \right ] , \end{aligned}\ ] ] where @xmath98 and @xmath99 are the electric charges of the quarks , and @xmath100 is the number of colors .
the functions @xmath101 are defined in terms of linear combinations of overlap integrals of the higgs - boson profile with the chirality - odd components of the 5d fermion propagator ( see eq . ( 15 ) in @xcite for more details ) .
an analogous expression , with @xmath102 replaced by 1 and @xmath103 replaced by @xmath104 holds for the charged - lepton contribution .
these exact results can be simplified by neglecting some terms of order @xmath105 and chirally - suppressed @xmath106 terms , which is an excellent approximation numerically .
this leads to the explicit expressions @xcite @xmath107 n_c\,q_u^2\,a_q(\tau_t ) + n_c\,q_d^2\,a_q(\tau_b ) + \sum_{q = u , d } n_c\,q_q^2\,\,{\rm re}\,{{\text}{tr}}\,g({\boldsymbol{x}}_q ) \
, , \nonumber\\ c_{5\gamma}^q & \approx - \frac{v^2}{3{m_{\text{kk}}}^2}\,{\rm i m } \left [ \frac{(\bm{y}_u \bm{y}_u^\dagger \bm{y}_u ) _ { 33}}{(\bm{y}_u)_{33 } } \right ] n_c\,q_u^2\,b_q(\tau_t ) +
\sum_{q = u , d } n_c\,q_q^2\,\,{\rm im}\,{{\text}{tr}}\,g({\boldsymbol{x}}_q ) \
, , \end{aligned}\ ] ] and @xmath108 where the contributions from the sm fermions and the kk excitations can now readily be identified .
the loop functions are given by ( with @xmath109 ) @xmath110 \ , , \qquad b_q(\tau ) = \tau f(\tau ) \,.\ ] ] they both approach 1 for @xmath92 .
for values @xmath111 the function @xmath112 in ( [ scalarloop ] ) must be analytically continued , with @xmath113 .
the quantities @xmath114 are defined in terms of the dimensionless 5d yukawa matrices of the rs model , whose entries are assumed to be random complex numbers of order 1 .
the observed hierarchies in the spectrum of fermion masses and mixing angles are generated when these anarchic yukawa matrices are combined with the values of the fermion profiles near the ir brane , which are exponentially small for all light fermions @xcite .
note that with the hermitian matrices @xmath115 the traces over matrix - valued functions @xmath116 are real , so that @xmath117 and the only contribution to the coefficient @xmath118 arises from the top - quark contribution given by the first term on the right - hand side of ( [ eqn : cga ] ) . the precise form of the function @xmath119 depends on the details of the localization of the scalar sector on or near the ir brane .
for the two scenarios with a brane - localized higgs and a narrow bulk higgs , as defined in section [ sec : pre ] , one finds @xcite @xmath120 so that the effect of the kk tower is approximately equal but of opposite sign in the two cases .
the difference is due to a `` resonance effect '' in the narrow bulk - higgs scenario , where very heavy kk modes with masses of order the inverse higgs width @xmath121 give an unsuppressed contribution to the loop amplitude @xcite . at a technical level
, the difference arises from the subtraction term at large euclidean momentum in ( [ ourbeauty ] ) , which is relevant for the function @xmath122 only . for a brane - localized higgs ,
this function approaches a plateau at large momenta , such that @xmath123 . for a narrow bulk higgs , on the other hand
, the function @xmath124 vanishes like @xmath125 in the region of large euclidean momenta @xmath126 , and hence @xmath127 can be set to zero . in @xcite
we have also considered a variant of the brane - higgs scenario with two different yukawa matrices @xmath128 and @xmath129 for the @xmath8-even and @xmath8-odd fermion fields . in this so - called type - ii brane - higgs model
the matrices @xmath130 are no longer hermitian , but to leading order @xmath131 is still a hermitian matrix .
the type - ii brane - higgs scenario is thus rather similar to the original brane - higgs model with identical yukawa matrices @xmath132 .
numerically , we find that the main difference is a slightly larger spread of the distribution of scatter points when one scans over the parameter space of the model . in our phenomenological analysis in section [ sec : pheno ] , we will therefore restrict ourselves to a study of the two cases shown in ( [ eqn : gfunc ] ) . in the sm
, a recent paper @xcite has thoroughly discussed the @xmath133 independence of the @xmath1 amplitude in dimensional regularization and has shown that the calculation can be performed consistently in the unitary gauge @xmath134 . in the case of the rs model , it is convenient to first work in the kk - decomposed theory , where 4d feynman propagators have the same structure as in the sm .
the feynman rules required to evaluate the one - loop diagrams shown in figure [ fig : alldiagrams ] are summarized in appendix [ app : fr ] . from these rules
, it follows that : * all vertices involving one or two external photons but no higgs boson are diagonal in kk number after one integrates over the extra - dimensional coordinate of the vertex with measure @xmath135 .
the feynman rules for these vertices have the same form as in the sm after one identifies the 4d electromagnetic coupling as @xmath136 @xcite . for the mass - dependent
vertex connecting a photon to @xmath137 , one must replace @xmath138 .
* as a result , all one - loop diagrams contributing to the @xmath1 amplitude involve a single kk particle in the loop .
hence , only kk - diagonal higgs couplings are required in the calculation . *
all kk - diagonal higgs couplings have the same structure as in the sm but come with an overall prefactor @xmath139 ^ 2 = \frac{2\tilde m_w^2}{v}\,2\pi \left [ \chi_n^w(1 ) \right]^2 , \ ] ] which replaces the corresponding factor @xmath140 in the sm .
in addition , for each scalar boson @xmath45 a factor @xmath141 appears , which replaces @xmath142 in the corresponding sm feynman rule for vertices involving the goldstone bosons @xmath55 .
it follows from these observations that , diagram by diagram and in a general @xmath0 gauge , the bosonic loop contributions obtained in the rs model resemble those of the sm up to trivial substitutions , such that @xmath143 ^ 2 \left [ \frac{v_{\rm sm}}{m_w^2}\,\mathcal{a}_{{\text}{sm}}^w(h\to{\gamma}{\gamma } ) \right]_{m_w\to m_n^w } .\ ] ] the statement made in the first bullet point above requires some comments . for vertices involving a photon and a pair of vector bosons , fermions or ghosts , the statement that the vertex is diagonal in kk number is a direct consequence of the flatness of the photon profile , enforced by @xmath41 gauge invariance , and the orthogonality of the relevant vector - boson and fermion profiles
however , the situation is different for vertices involving the scalar bosons @xmath45 , which receive contributions from the 5d fields @xmath48 and @xmath55 , see ( [ eqn : kkdecomm ] ) and ( [ varphidecomp ] ) . in this case
the vertex becomes diagonal only after one adds up these two contributions .
consider , as an example , the vertex ( origin ) at ( 0,0 ) ; ( 15,0 )
( 17 , -2)node[right]@xmath144 ; ( 15,0)(17,2)node[right]@xmath145 ; ( 13,2 ) node[left ] @xmath146(15,0 ) ; ( 13,-2 ) node[left ] @xmath147(15,0 ) ; needed for diagram @xmath84 in figure [ fig : alldiagrams ] . after integrating over the coordinate of this vertex
, we obtain the feynman rule @xmath148
\left [ \partial_t\chi_n^w(t ) \right ] + \frac{\tilde m_w^2}{m_m^w\,m_n^w}\,2\pi\,\chi_m^w(1)\,\chi_n^w(1 ) \right ] , \ ] ] where the first contribution originates from the @xmath149 term contained in the yang - mills action for the @xmath2-boson fields using the kk decomposition ( [ eqn : kkdecomm ] ) , while the second contribution arises from the @xmath150 term contained in the kinetic term for the higgs doublet using the kk decomposition ( [ varphidecomp ] ) .
we now integrate by parts in the first term use the equations of motion @xcite @xmath151 for the gauge - boson profiles along with the boundary conditions @xmath152 corresponding to a neumann boundary condition on the uv brane and a mixed boundary condition on the ir brane .
the notation @xmath153 means that the ir brane is approached from the left ( @xmath154 ) .
such a prescription is necessary because the derivative of the profile function is discontinuous on the ir brane . in this way
, we obtain the feynman rule @xmath155 , \ ] ] where the boundary term cancels the contribution arising from the @xmath150 term .
using finally the orthonormality relation @xcite @xmath156 for the gauge - boson profiles , we obtain the sm expression @xmath157 for the vertex .
let us now return to the general result ( [ eqn : super ] ) and explore its consequences .
obviously , this relation implies that for each single kk mode the @xmath1 amplitude in the rs model is gauge invariant provided the amplitude is gauge invariant in the sm .
since , as we will demonstrate below , the sum over kk modes is convergent , it follows that gauge invariance is maintained also in the 5d theory .
we recall that to show gauge invariance in the sm one divides the @xmath2-boson propagator in @xmath0 gauge into two parts , @xmath158 = \frac{i}{p^2-m_w^2 } \left ( \frac{p^\mu p^\nu}{m_w^2 } - \eta^{\mu\nu } \right ) - \frac{i}{p^2-\xi m_w^2}\,\frac{p^\mu p^\nu}{m_w^2 } \,,\ ] ] where the first part coincides with the propagator in unitary gauge and the second part has the same structure as the scalar - boson and ghost propagators .
it has been shown in @xcite that , after adding up all diagrams , many intricate cancellations occur , and at the end only the diagrams @xmath76 and @xmath77 in figure [ fig : alldiagrams ] with the @xmath2-boson propagators in unitary gauge , as well as the fermion loop contributions shown in diagram @xmath71 , remain .
we have repeated this analysis and checked these cancellations by explicit calculation , thereby confirming that it is justified to use unitary gauge also in the rs model .
we now repeat the calculation of the bosonic loop contributions to the @xmath1 amplitude using a 5d approach . based on the findings of the previous section we adopt unitary gauge and consider only the contributions of diagrams @xmath76 and @xmath77 in figure [ fig : alldiagrams ] .
we employ the mixed momentum - position representation of the 5d gauge - boson propagator @xmath160 @xcite , in which the extra - dimensional coordinate is kept in position space .
this is very convenient from a technical point of view , but it is also physically well motivated , as by ads / cft correspondence the position along the extra dimension defines the natural mass scale of the model @xcite . it is well known that even in the sm the loop - momentum integral must be regularized dimensionally in order to preserve gauge invariance .
we will thus introduce a dimensional regulator @xmath161 on the loop integral in intermediate steps .
this regulator can be removed at the end of the calculation .
we also regularize the higgs profile by replacing the @xmath37-function profile of the brane - localized higgs field by a smooth function @xmath34 of width @xmath35 .
such a regularization is important in the calculation of the fermionic loop contributions to the @xmath11 and @xmath1 amplitudes .
however , we will find that in the calculation of the bosonic loop contributions the limit @xmath36 can be taken without encountering any ambiguities .
diagrams @xmath76 and @xmath77 give rise to the amplitude @xmath162 \ , , \end{aligned}\ ] ] where @xmath163 arises from the product of two triple gauge - boson vertices , with @xmath164 . our goal is to rewrite this result as a feynman parameter integral over a _ single _
5d gauge - boson propagator , which should be possible since in the kk - decomposed theory only a single kk mode propagates in the loops . in order to simplify the answer
, we decompose the 5d propagator as @xmath165 and use the kk decomposition @xmath166 in an intermediate step .
the use of the kk representation is merely a mathematical trick , similar to the use of feynman parameters in conventional loop calculations .
it is justified because all expressions are finite and the kk sum converges . at the end of the calculation
we obtain an expression without any reference to kk modes . due to the mode - diagonality of the vertices involving a photon
, we can perform the integration over @xmath72 and @xmath73 using the orthonormality relation ( [ orthonorm ] ) for the gauge - boson profiles . working out the dirac algebra and making use of passarino - veltman reductions
, we can reduce the answer to a simple feynman parameter integral .
after the contributions from the various diagrams have been combined , the dimensional regulator @xmath167 can be set to 0 .
we find @xmath168 ^ 2 \left [ \frac{1}{\left(m_n^w\right)^2 } + 6\int_0 ^ 1\!dx\int_0^{1-x}\!dy\ , \frac{1 - 2xy}{\left(m_n^w\right)^2-xy m_h^2 - i0 } \right ] , \ ] ] and @xmath169 .
it is now a simple exercise to recast the answer in terms of the 5d propagator function @xmath170 defined in ( [ eqn:5dpropkk ] ) .
we obtain @xmath171 \,,\ ] ] where @xmath172 denotes the overlap integral of the higgs profile with the transverse part of the 5d @xmath2-boson propagator evaluated at @xmath173 , @xmath174 for the case of a brane - localized scalar sector , an explicit expression for the function @xmath175 will be derived in the following section . the more general case of a narrow bulk higgs
will be briefly considered at the end of section [ sec : analzeromrs ] and will be described in more detail elsewhere .
such an analysis demonstrates that the above integral exhibits a smooth behavior in the limit of small @xmath86 , so that the last identity holds and the regulator on the higgs profile can be taken to zero without encountering any ambiguities .
relation is one of the main results of this work .
it shows the exact result for the wilson coefficient @xmath64 in dependence of overlap integrals of the higgs profile and the 5d gauge - boson propagator . with the help of the findings in @xcite
, it can be shown that this relation also holds for an arbitrary bulk - higgs profile @xmath176 , provided one uses the corresponding 5d gauge - boson propagator in the bulk - higgs model .
then the regularized @xmath37-function in ( [ twdef ] ) must be replaced by @xmath177 where @xmath178 is the profile of the higgs vev , and we use the conventions of @xcite .
note , however , that in this case it is necessary to also include the contribution ( [ c1phi ] ) due to the physical scalar excitations of the bulk higgs field .
the last equation in ( [ bulkreplace ] ) holds ( up to very small corrections ) in the particular bulk - higgs model analyzed in @xcite .
the parameter @xmath179 is related to the 5d mass parameter of the bulk scalar field . in the region where @xmath180 , the function on the right - hand side indeed approaches a regularized @xmath37-distribution , with a characteristic width given by @xmath181 .
note that relation ( [ eq23 ] ) results after integrating a feynman loop integrand of the type @xmath182 ^ 3 $ ] over @xmath183 ( after the wick rotation ) , which corresponds to the integral over the second derivative @xmath184 .
in order for this integral to exist , we need to require that both @xmath185 and @xmath186 vanish for very large euclidean momenta .
we will show in the following section that this is indeed the case .
the calculation of the propagator function @xmath175 in ( [ eqn:5dpropkk ] ) in the rs model with a brane - localized higgs field is a straightforward exercise and has been performed , for instance , in @xcite .
this function is the solution to the differential equation ( with @xmath187 ) @xmath188 subject to the boundary conditions @xcite @xmath189 note the close similarity with the corresponding equations for the gauge - boson profiles @xmath190 in ( [ chindeq ] ) and ( [ chinbcs ] ) . integrating the differential equation ( [ deg ] ) over an infinitesimal interval around @xmath173 , we derive the jump condition @xmath191 the propagator itself is continuous at @xmath173 . in the region of time - like momenta ( @xmath192 ) ,
the general solution can be written in the form @xmath193 d_{10}(t_<,{\epsilon } ) } { \hat p d_{00}(1,{\epsilon } ) - b_1 d_{10}(1,{\epsilon } ) } \,,\ ] ] where @xmath194 for space - like momenta , we find instead ( with @xmath195 and @xmath196 ) @xmath197 d_{10}(t_<,{\epsilon } ) } { { { \hat p_e}}d_{00}(1,{\epsilon } ) - b_1 d_{10}(1,{\epsilon } ) } \,,\ ] ] with @xmath198 for our result ( [ twdef ] ) we need the propagator in the time - like region , evaluated at @xmath199 .
using the general solution in , we obtain ( with @xmath200 ) @xmath201^{-1 } \equiv \frac{1}{2\pi\tilde m_w^2}\,\hat t_w(-p^2 ) \,,\ ] ] which is exact to all orders in @xmath3 . and @xmath202 .
] it follows from this expression that @xmath203 .
we have thus succeeded in deriving a closed analytic expression for the wilson coefficient @xmath204 in , valid for the minimal rs model with a higgs sector localized on the ir brane .
note that we have kept the quantity @xmath57 , which is the leading - order contribution to the mass of the physical @xmath2 boson , in the prefactor above , since it will cancel against a corresponding factor in the definition of the wilson coefficient .
indeed , our final result for this coefficient takes the form @xmath205 \,.\ ] ] before proceeding , we briefly study the behavior of the propagator function in the region of large space - like momenta . for large euclidean momenta @xmath206
, this function approaches an inverse power - law behavior given by @xmath207 it follows that both @xmath185 and @xmath186 vanish for large euclidean momenta @xmath208 , and hence the conditions required for the validity of our relation ( [ eqn : c1w ] ) are indeed satisfied .
our exact expression for the overlap integral @xmath172 in ( [ eqn : iw2 ] ) contains the contribution of the zero mode the standard @xmath2 boson with its modified coupling to the higgs field as well as the infinite tower of kk excitations .
it is instructive to isolate the contribution from the zero mode and the kk tower explicitly . to this end
, we expand the exact formula in powers of @xmath3 , using that we need this function for values @xmath209 much smaller than the kk scale @xmath210 .
we find @xmath211 + \frac{m_w^2}{2{m_{\text{kk}}}^2 } \left ( \frac{l}{c_\vartheta^2 } - 1 + \frac{1}{2l } \right ) + { \cal o}\bigg ( \frac{v^4}{{m_{\text{kk}}}^4 } \bigg ) \,,\ ] ] where @xmath53 in the minimal rs model . in section [ sec : propcust ] we will show that the same result holds in the custodial rs model , where however the parameter @xmath212 takes a different value . in the above result we have replaced the parameter @xmath57 by the physical @xmath2-boson mass @xmath213 using relation ( [ wmassrela ] ) , which was derived in @xcite by solving the eigenvalue equation for the @xmath2-boson profiles and extracting the lowest eigenvalue . in appendix [ sec : massren ]
, we present an alternative approach , where the above relation is derived with the help of our expressions for the 5d gauge - boson propagator . based on the formulas above
, we can perform the integration over the feynman parameters in and find the wilson coefficient @xmath214 + { \cal o}\bigg ( \frac{v^4}{{m_{\text{kk}}}^4 } \bigg ) \ , , \qquad c_{5\gamma}^w = 0 \,,\ ] ] where @xmath215 , and the function @xmath216\ ] ] with @xmath112 from ( [ scalarloop ] ) approaches 1 for @xmath92 .
the first contribution to @xmath64 arises from the standard @xmath2 boson , whose coupling to the higgs boson is modified , compared with the sm , by a factor @xmath217 times @xmath218 .
the last factor is accounted for by using the higgs vev in the rs model in the definition of the effective operators in ( [ eqn : effamp ] ) .
the term @xmath219 in ( [ eqn : c1wmin ] ) is due to the kk excitations .
explicitly , we obtain @xmath220 note that at this order @xmath221 , such that the rs corrections to @xmath64 in ( [ eqn : c1wmin ] ) would cancel in the limit @xmath222 .
this simple relation is however not preserved in higher orders .
our result for @xmath64 agrees with a corresponding expression derived in @xcite .
notice also that the value of @xmath217 is consistent with relation ( [ prefac ] ) , which gives @xmath223 ^ 2 $ ] .
we close this section by returning briefly to the case of a narrow bulk - higgs model , in which the scalar sector is localized not on but near the ir brane . as a concrete model
, we adopt the scenario discussed in @xcite featuring a bulk scalar field , which acquires a vev due to a mexican - hat potential localized on the ir brane .
as discussed earlier , relation ( [ twdef ] ) still holds in this model provided one makes the replacement ( [ bulkreplace ] ) and calculated the gauge - boson propagator in the background of a bulk - higgs field .
it is the solution to the differential equation @xmath224 with neumann boundary conditions @xmath225 for @xmath226 .
we will present the details of such an analysis elsewhere . here
it suffices to note that in the limit where @xmath227 , one obtains @xmath228 instead of ( [ eqn : c1wmin ] ) .
this demonstrates that the result for the bosonic loop contributions to the @xmath1 amplitude interpolates smoothly from the narrow bulk - higgs scenario into a scenario with a brane - localized scalar sector .
we will now present the generalization of the above results to the rs model with custodial protection , which has been proposed to mitigate the large corrections to electroweak precision observables , so that the lightest kk particles are in reach for the direct detection at the lhc @xcite .
we will consider an rs model based on the enlarged bulk symmetry @xmath229 , whose @xmath29 subgroups are broken on the ir brane via the symmetry - breaking pattern @xmath230 .
this symmetry breaking is accomplished by means of the higgs field transforming as a bi - doublet under the two @xmath29 gauge groups . in component notation , it is given by @xmath231 where @xmath232 are real scalar fields , @xmath233 , and @xmath9 denotes the higgs vev in the custodial rs model .
the resulting @xmath234 supplies the custodial symmetry and protects the @xmath30 parameter @xcite .
the @xmath235 symmetry prevents the left - handed @xmath236 coupling @xcite and its flavor - changing counterparts @xcite from receiving too large corrections . on the uv brane
, the symmetry breaking @xmath237 generates the sm gauge group .
this is achieved by an interplay between uv and ir boundary conditions .
thorough discussions of this model containing many technical details can be found in @xcite , and we will adopt the notations of the first reference throughout this analysis . as a consequence of the discrete @xmath235 symmetry , the left - handed bottom quark needs to be embedded in an @xmath238 bi - doublet with isospin quantum numbers @xmath239 .
this assignment fixes the quantum numbers of the remaining quark fields uniquely .
in particular , the right - handed down - type quarks have to be embedded in an @xmath240 triplet in order to obtain a @xmath241-invariant yukawa coupling .
one arrives at the following multiplet structure for the quark fields with even @xmath8 parity : @xmath242 the field @xmath243 transforms as @xmath244 under @xmath238 , while @xmath245 transforms as @xmath246 .
the fields with odd @xmath8 parity have the opposite chirality .
their profiles are related to those of the @xmath8-even fields by the field equations .
the inner and outer subscripts on the various fields denote their @xmath41 and @xmath241 charges , respectively , which are connected through the relations @xmath247 and @xmath248 .
the superscripts on the fields specify the type of boundary conditions they obey on the uv boundary .
fields with superscript @xmath249 obey the usual mixed boundary conditions allowing for a light zero mode , meaning that we impose a dirichlet boundary condition on the profile functions of the corresponding @xmath8-odd fields .
these zero modes correspond to the sm quarks .
fields with superscripts @xmath250 correspond to heavy , exotic fermions with no counterparts in the sm .
for these states , the dirichlet boundary condition is imposed on the @xmath8-even fields , so as to avoid the presence of a zero mode .
the uv boundary conditions for the fields of opposite @xmath8 parity are of mixed type and follow from the field equations .
note that we chose the same @xmath238 quantum numbers for all three quark generations , which is necessary to consistently incorporate quark mixing in the anarchic approach to flavor in warped extra dimensions .
altogether , there are fifteen different quark states in the up sector and nine in the down sector ( for three generations ) .
the boundary conditions give rise to three light modes in each sector , which are identified with the sm quarks .
these are accompanied by kk towers consisting of groups of fifteen and nine modes of similar masses in the up and down sectors , respectively .
in addition , there is a kk tower of exotic fermion states with electric charge @xmath251 , which exhibits nine excitations in each kk level .
the fermionic loop contributions to the @xmath1 amplitude in the custodial rs model can be parameterized in terms of the same @xmath252 yukawa matrices appearing in the minimal model , however with different coefficients that reflect the embeddings of the various fermion species under the enlarged bulk gauge group .
the generalizations of relations ( [ eqn : cga ] ) for the quark contributions have been worked out in @xcite .
they are @xmath253 n_c\,q_u^2\,a_q(\tau_t ) + n_c\,q_d^2\,a_q(\tau_b ) \\ & \quad\mbox{}+ n_c\,q_u^2\,\,{\rm re}\,{{\text}{tr}}\,g(\sqrt{2}{\boldsymbol{x}}_u ) + n_c \left ( q_u^2 + q_d^2 + q_\lambda^2 \right ) { \rm re}\,{{\text}{tr}}\,g(\sqrt{2}{\boldsymbol{x}}_d ) \ , , \\
c_{5\gamma}^q & \approx - \frac{2v^2}{3{m_{\text{kk}}}^2}\,{\rm i m } \left [ \frac{(\bm{y}_u \bm{y}_u^\dagger \bm{y}_u ) _ { 33}}{(\bm{y}_u)_{33 } } \right ] n_c\,q_u^2\,b_q(\tau_t ) \\ &
\quad\mbox{}+ n_c\,q_u^2\,\,{\rm im}\,{{\text}{tr}}\,g(\sqrt{2}{\boldsymbol{x}}_u ) + n_c \left ( q_u^2 + q_d^2 + q_\lambda^2 \right ) { \rm im}\,{{\text}{tr}}\,g(\sqrt{2}{\boldsymbol{x}}_d ) \ , .
\end{aligned}\ ] ] for various rs models with a brane - localized scalar sector or a narrow bulk - higgs sector , the explicit forms of the function @xmath119 have been given in and .
recall that the taylor expansion of these functions starts with @xmath254 , and thus the factors of @xmath255 arising in the quark contributions in the custodial model approximately double the contribution arising in the minimal model .
combined with the large electric charge of the @xmath256-type quarks , one finds that due to the higher multiplicity of kk quark states the contribution in the custodial rs model is much larger than in the minimal model @xcite , by approximately a factor 68/5 .
the result for the loop contributions to the @xmath1 amplitude involving charged leptons depends on the way in which the lepton fields are embedded into the extended gauge symmetry of the custodial rs model . as a first possibility
, we consider a model in which the lepton multiplets are chosen in analogy to the quark multiplets in ( [ quarkmultiplets ] ) .
this choice was adopted in @xcite . in component notation ,
the corresponding fields are @xmath257 there are fifteen different lepton states in the neutrino sector and nine in the charged - lepton sector .
the boundary conditions give rise to three light modes in each sector , which are identified with the sm neutrinos and charged leptons .
these are accompanied by kk towers consisting of groups of fifteen and nine modes in the two sectors , respectively .
in addition , there is a kk tower of exotic lepton states with electric charge @xmath258 , which exhibits nine excitations in each kk level .
the gauge - invariant yukawa interactions for these fields are constructed in complete analogy with the quark yukawa interactions @xcite .
they can be expressed in terms of two dimensionless @xmath252 yukawa matrices @xmath259 and @xmath260 , which we assume to have an anarchic structure . when dressed with the fermion profiles on the ir brane , these matrices give masses to the sm leptons .
the resulting contributions to the wilson coefficients have the same structure as in ( [ eqn : cqcust ] ) , except that there are no zero - mode contributions ( they are proportional to @xmath261 and thus can be neglected ) and that we must replace @xmath262 , @xmath263 , @xmath264 , and @xmath265 , @xmath266 , @xmath267 .
we thus obtain @xmath268 with @xmath269 as defined in ( [ xfdef ] ) .
it follows that the leptonic contribution in the custodial rs model is approximately 4 times larger than in the minimal model . as a second possibility
, we consider a model with a more minimal embedding of the leptons into the extended gauge group . the simplest assignment is to put the left - handed neutrino and electron into an @xmath270 doublet ( as in the sm ) and the right - handed electron along with a new , exotic neutral particle @xmath271 into an @xmath240 doublet . the lepton fields with even @xmath8 parity
are then chosen as @xmath272 and they transform as @xmath273 and @xmath274 , respectively .
the choice of the boundary conditions is such that the zero modes correspond to the light leptons of the sm , without a right - handed neutrino .
the gauge - invariant yukawa interaction that can be built using these fields is @xmath275 where @xmath276 .
upon electroweak symmetry breaking this generates a mass term for the zero modes of the charged leptons .
the sm neutrinos remain massless at this order .
their masses can be explained by means of higher - dimensional operators .
the only additional lepton field is the right - handed neutrino , which is charged under @xmath240 but electrically neutral , so that it does not affect the @xmath1 decay amplitude .
the lepton contribution is therefore the same as in the minimal version of the rs model , namely @xmath277 as in ( [ leptonterm ] ) . in order to derive the feynman rules and the 5d gauge - boson propagator it is inevitable to understand the bosonic sector of the custodial rs model , whose 5d action reads @xmath278
since it is of no significance for our discussion , we refrain from presenting the gauge - fixing term @xmath279 , whose explicit form can be found in @xcite .
the gauge - kinetic terms read @xmath280 where @xmath281 denotes the 5d metric with determinant @xmath282 , and where a sum over the gauge - group indices @xmath283 is implied .
we choose the 4-components of the gauge fields to be even under the @xmath8 parity , while the fifth components are chosen to be odd , in order to derive at a low - energy spectrum that is compatible with observation .
as in the previous section , it does not make any difference if we consider a narrow - bulk or a brane - localized higgs sector , and we thus focus on the scenario in which the higgs lagrangian is localized on the ir brane .
the higgs transforms as a bi - doublet under @xmath238 and is neutral with respect to @xmath241 , see ( [ bidoublet ] ) . in order to show how the symmetry breaking @xmath230 is accomplished
, we use the covariant derivative @xmath284 where @xmath285 and @xmath286 are the 5d gauge couplings associated with @xmath287 , and @xmath288 are the corresponding generators . in order to evaluate the kinetic term for the scalar bi - doublet , it is convenient to rotate the gauge bosons @xmath289 and @xmath290 into a new basis of fields @xmath291 and @xmath292 , such that @xcite @xmath293 where @xmath294 the @xmath235 symmetry , which is imposed to protect the left - handed @xmath31 couplings from receiving large corrections @xcite , enforces that @xmath295 , and hence @xmath296 . in our discussion in this section
we will however keep the value of @xmath297 as a free parameter .
the higgs vev @xmath298 then generates a mass term @xmath299 for the fields @xmath291 , while the fields @xmath292 remain massless .
we can also read off the coupling to the higgs boson , once we replace @xmath300 by @xmath301 .
note that only the fields @xmath291 couple to the higgs boson @xmath302 .
this will become important for the derivation of the propagator .
the further symmetry breaking via boundary conditions is not relevant for our discussion , and we again refer to @xcite for details . notice that relation ( [ basistrafo ] ) represents the connection between the uv basis fields ( right ) , which obey dirichlet boundary conditions on the uv brane , and the ir basis fields ( left ) , which obey dirichlet boundary conditions on the ir brane .
we now focus on the charged sector and define the gauge - boson fields @xmath303 whose kk decomposition can be written in a form analogous to ( [ eqn : kkdecomm ] ) , such that @xcite @xmath304 the orthonormality relation for the gauge - boson profiles reads @xmath305 the profiles @xmath306 are @xmath8-even functions on the orbifold .
their upper ( lower ) components are `` untwisted '' ( `` twisted '' ) functions . untwisted even functions
obey neumann boundary conditions on the uv brane , allowing for light zero modes .
twisted even functions obey dirichlet boundary conditions on the uv brane and are thus not smooth at this orbifold fixed point .
the upper ( lower ) components of the rotated profiles @xmath307 obey mixed ( neumann ) boundary conditions on the ir brane , such that @xmath308 which generalizes ( [ chinbcs ] ) . here
@xmath309 is a projector on the upper component , and from now on we use the abbreviations @xmath310 and @xmath311 . as in the minimal rs model ,
the parameter @xmath57 is the leading contribution to the @xmath2-boson mass in an expansion in powers of @xmath3 , see ( [ wmassrela ] ) .
it is now straightforward to deduce the feynman rules in the custodial model from the ones in the minimal model compiled in appendix [ app : fr ] .
using , we can convince ourselves that the @xmath312 couplings to the photon are not changed at all .
this statement is independent of the basis , since the rotation matrix @xmath313 drops out in the orthonormalization condition .
in contrast , as mentioned earlier the higgs only couples to the ir basis fields @xmath314 with a strength proportional to @xmath315 .
this can be taken into account with the help of the projection operator @xmath316 rotated into the ir basis and accompanied by a factor @xmath317 .
it follows that , compared with the sm , all kk - diagonal higgs couplings in the custodial rs model come with a prefactor @xmath318 which replaces the corresponding factor ( [ prefac ] ) in the minimal model .
here we have introduced @xmath319 in analogy with ( [ level42 ] ) , we now define the propagator function @xmath320 in terms of gauge - boson profiles in the uv basis . an explicit expression for this function will be derived in the next section . in analogy with expression ( [ eqn : super ] ) valid in minimal rs model , we find that the @xmath1 amplitude in the custodial rs model can be written as @xmath321_{m_w\to m_n^w } .\ ] ] it follows that expression ( [ eqn : c1w ] ) for the wilson coefficient @xmath322 derived in section [ sec:3.3 ] remains valid , provided we replace the quantity @xmath172 defined in ( [ twdef ] ) with @xmath323 .\ ] ] we now derive the exact expression for the 5d gauge - boson propagator in the rs model with custodial symmetry , which to the best of our knowledge has not been done before .
the differential equation for the propagator function @xmath324 is the same as in the minimal model , see ( [ deg ] ) .
however , the boundary conditions are modified to @xcite @xmath325 the first equation follows from the boundary conditions for the uv fields @xmath326 and @xmath327 , where we have defined @xmath328 .
the second equation is a direct consequence of ( [ irbcs ] ) .
we find that , in the region of time - like momenta ( @xmath329 ) , the general solution for the propagator function reads @xmath330 d_{01}(1,{\epsilon } ) - b_1 \frac{4s_\vartheta^2}{\pi^2\hat p^2{\epsilon } } } \\ & \times \bigg\ { \left [ \big [ \hat p d_{10}(t_>,1 ) - b_1 d_{11}(t_>,1 ) \big ] d_{01}(1,{\epsilon } ) - b_1 \frac{2s_\vartheta^2}{\pi\hat p}\,d_{11}(t_>,{\epsilon } ) \right ] d_{10}(t_<,{\epsilon})\,\bm{p}_+ \\ & \quad\mbox{}+ \left [ \big [ \hat p d_{00}(1,{\epsilon } ) - b_1 d_{10}(1,{\epsilon } ) \big ] d_{10}(t_>,1 ) + b_1 \frac{2s_\vartheta^2}{\pi\hat p}\,d_{10}(t_>,{\epsilon } ) \right ] d_{11}(t_<,{\epsilon})\,\bm{p}_- \\ & \quad\mbox{}- b_1 \frac{2s_\vartheta c_\vartheta}{\pi\hat p } \big [ d_{10}(t,{\epsilon})\,d_{11}(t',{\epsilon})\,\bm{p}_{12 } + d_{11}(t,{\epsilon})\,d_{10}(t',{\epsilon})\,\bm{p}_{21 } \big ] \bigg\ } \ , , \end{aligned}\ ] ] where the functions @xmath331 have been defined in , and we have introduced the @xmath332 matrices @xmath333 and @xmath334 , which have an entry 1 at the corresponding position indicated by the subscripts and entries 0 otherwise . note that up to irrelevant @xmath335 terms we can replace @xmath336 for @xmath337 .
this gives rise to a simpler expression , in which the spurious @xmath338 term in the denominator is removed . in the limit @xmath339
, we can identify the coefficient of @xmath316 in ( [ eqn : buv ] ) with the result ( [ eqn : bbranemin ] ) obtained in the minimal rs model .
moreover , for the special case @xmath340 our result reduces to equation ( 54 ) in @xcite .
our general results above valid for arbitrary momentum are however new .
it is now straightforward to calculate the quantity @xmath172 in ( [ twdefcusto ] ) , which we need for the calculation of the wilson coefficient @xmath64 in ( [ eqn : c1w ] ) . expanding this answer in powers of @xmath3 and for @xmath341
, we recover expression ( [ eqn : timin ] ) . with respect to the minimal rs model , the only modification concerns the coefficient of the leading @xmath342-enhanced correction terms , which is enhanced by @xmath317 .
this affects both the contributions from the @xmath2 boson and the kk tower . in the custodial rs model with @xmath235 symmetry , this enhancement factor is equal to 2 .
note that with @xmath343 the expressions in ( [ eqn : cwmrs ] ) are compatible with corresponding results obtained in @xcite . in this reference the wilson coefficient @xmath64 belonged to the operator @xmath344 instead of the one in , and hence @xmath345}=\kappa_w/\kappa_v^2 $ ] .
we now present a numerical study of the higgs decay into two photons in both the minimal and the custodial rs model , which can be directly compared to experimental data . as in our recent work on higgs production @xcite , we distinguish the two cases of a brane - localized and a narrow bulk - higgs scenario .
we consider the ratio of the measured @xmath346 cross section normalized to its sm value , @xmath347 \big ( |\kappa_{\gamma}|^2 + |\kappa_{\gamma5}|^2 \big ) } { \kappa_v^2\,\kappa_h } \,,\ ] ] where we have included the two main higgs production channels via gluon fusion ( gf ) and vector - boson fusion ( vbf ) , with probabilities of @xmath348 and @xmath349 at the lhc with @xmath350tev @xcite .
other higgs production channels , such as the associated production with a @xmath351 pair or a vector boson , can be neglected to a very good approximation .
the quantities @xmath352 and @xmath353 ( with @xmath354 ) parametrize the values of the relevant wilson coefficients normalized to their sm values , @xmath355 explicit expressions for the wilson coefficients @xmath64 and @xmath65 in the rs model have been derived in sections [ sec : analysis ] and [ sec : custodial ] .
the corresponding sm value is @xmath356-\frac{21}{4 } a_w(\tau_w)$ ] .
the rs effects on the gluon - fusion production process were studied in @xcite .
the values of the wilson coefficients @xmath357 and @xmath358 can be obtained from @xmath64 and @xmath65 by replacing @xmath359 , @xmath360 , and @xmath264 . in the sm we have @xmath361 .
concerning the vbf production process , we note that using @xmath217 as a correction factor in ( [ eqn : rgaga ] ) is only approximate but sufficient for our purposes @xcite .
the parameter @xmath362 in ( [ eqn : rgaga ] ) parameterizes the shift of the higgs vev in the rs model relative to the sm and has been given in ( [ kappav ] ) .
finally , we take into account the rs corrections to the sm higgs width @xmath363mev ( for @xmath364gev ) @xcite by means of the parameter @xmath365 where the corrections to the decays @xmath366 have a numerically insignificant effect and therefore can be neglected ( the combined branching ratio for these channels is 12% in the sm ) .
neglecting some small chirally - suppressed terms , the correction to the higgs coupling to a @xmath367 pair can be well approximated by @xcite @xmath368 it is an important goal of future lhc and ilc analyses to determine as many of the effective higgs couplings @xmath352 as possible from a global fit to the data .
a detailed discussion of the individual effective higgs couplings to fermions and gauge bosons in the context of rs models will be presented in a future work @xcite . at present , however , the experimental groups have not yet presented a detailed , model - independent analysis of higgs couplings @xcite , and we will thus focus on the ratio @xmath369 in the present work . note also that , in contrast to the observable @xmath370 , the quantities @xmath352 and @xmath353 are not directly observable . the gluon - fusion rate is proportional to the sum of the absolute squares of @xmath371 and @xmath372 , and no observable sensitive to a different combination of these parameters is experimentally accessible . in the case of @xmath1 decay
, it is in principle possible to access the cp - violating coefficient @xmath373 by studying the distribution of the two electron - positron pairs in events in which both photons undergo nuclear conversions @xcite , however this will be very challenging experimentally .
figure [ fig : rggmrs ] shows our predictions for @xmath370 obtained in the minimal rs model with a brane - localized higgs sector ( left plot ) and a narrow bulk - higgs state ( right plot ) .
the new - physics effects arising in these scenarios scale with @xmath374 .
we find it useful to convert the mass parameter @xmath375 to the physical mass @xmath376 of the lightest kk gluon ( or kk photon ) state , which is independent of the details of the localization of the scalar sector and of the choice of the electroweak gauge group in the bulk @xcite .
our numerical results also depend on the rs volume @xmath377 and the dimensionless 5d yukawa matrices @xmath378 , @xmath379 and @xmath380 . typical values for @xmath342 fall in the range @xmath381 , corresponding to @xmath382tev , and we take @xmath383 for concreteness .
we work with anarchic yukawa matrices , whose individual entries are taken to be random complex numbers subject to the condition that @xmath384 . as a further constraint
, we impose that these matrices correctly reproduce the wolfenstein parameters @xmath385 and @xmath86 of the unitarity triangle ( see @xcite for explicit formulae ) and that , with appropriately chosen bulk mass parameters @xmath386 , one can reproduce the correct values for the masses of the sm quarks .
the dominant corrections to the @xmath11 and @xmath1 amplitudes arise from fermionic loop contributions and scale with @xmath387 @xcite
. the value of @xmath388 should be naturally of @xmath389 , and requiring that one - loop corrections to the yukawa couplings remain perturbative one can derive an upper bound @xmath390 @xcite ( see also @xcite for a detailed discussion ) .
the green , red , and blue scatter points in the figure correspond to rs model points obtained using three different values of @xmath388 .
the latest experimental values for @xmath370 reported by the atlas and cms collaborations are @xmath391 ( at @xmath392gev ) @xcite and @xmath393 ( at @xmath394gev ) @xcite which we naively average to obtain @xmath395 .
the @xmath396 error band corresponding to this result is shown by the blue band in the two plots .
model points falling outside these bands are excluded at 68% confidence level ( cl ) .
it is interesting to observe that for relatively large values for @xmath388 the data already disfavor kk gluon masses in the low tev range .
the tensions between our theoretical predictions and the experimental data are stronger for the brane - higgs model due to the mild tendency of an enhanced cross section seen in the data , which is in conflict with the suppression of the predicted cross section in this case .
we emphasize , however , that using the individual values for @xmath369 reported by atlas and cms one would obtain different conclusions .
[ ] @xmath397}$ ] [ b]@xmath370 [ ] [ cols= " < " , ] one can also read the exclusion plots in figures [ fig : rggexcmrs ] and [ fig : rggexccrs ] in a different way . under the optimistic assumption that the first kk gluon resonance is in reach for direct production at the lhc , these plots allow one to impose bounds on @xmath388 .
for instance , in the minimal rs model with a hypothetical kk gluon mass @xmath398tev , our results imply an upper bound of @xmath399 at 95% cl in the brane - higgs model , and @xmath400 at 68% cl in the narrow bulk - higgs scenario . in the custodial rs model , those bounds
are tightened to @xmath401 for a brane higgs and @xmath402 for a narrow bulk higgs , both at 95% cl .
even though the constraints are rather strong in the case of the custodial rs model , they do not quite compete with those stemming from the decays @xmath403 @xcite .
this is due to the fact that the rs corrections to the decay into two photons partially compensate the huge effect in the gluon - fusion production process .
this compensation does not occur in the decays into two weak gauge bosons , whose couplings to the higgs are only slightly affected by new - physics effects .
the discovery of a higgs boson at the lhc in the summer of 2012 @xcite has marked the beginning of a new era in elementary particle physics .
the couplings of this new particle are found to be non - universal and indeed very close to those predicted for the elementary scalar boson of the sm .
an explanation for the hierarchy problem is thus more pressing than ever .
measuring precisely the higgs couplings to various sm particles provides an important tool to discover and distinguish between new - physics models that can address the hierarchy problem .
especially interesting are loop - induced processes like higgs production via gluon fusion @xmath404 and the radiative decay @xmath405 . in this paper , we have focused on the higgs decay into two photons in different scenarios of rs models , where the higgs sector is localized on or near the ir brane , while the remaining fields propagate in the bulk of the warped extra dimension .
while the contribution from diagrams with virtual fermions ( quarks and charged leptons ) in the loop has been extensively discussed in recent works , mostly in the context of gluon fusion @xcite , here we have further concentrated our analysis on the diagrams with bosonic fields propagating in the loops , generalizing the findings of @xcite .
we have shown that the relevant diagrams , calculated in the general @xmath0 gauge , add up to a gauge - invariant result .
working in unitary gauge , we have derived an exact expression for the @xmath1 decay amplitude in terms of an integral over the 5d @xmath2-boson propagator with the higgs - boson profile along the extra dimension , given in .
this expression can be used to calculate the bosonic contributions to the amplitude as long as one is able to derive an analytic expression for the 5d @xmath2-boson propagator .
we have shown that the 5d loop diagrams with bosonic fields are insensitive to the precise details of the localization of the scalar sector on or near the ir brane .
this finding is in contrast to the case of fermionic loops , for which one finds different results for the cases of a strictly localized higgs sector and a scenario in which the higgs is a narrow bulk field @xcite . in our approach
we retain the exact dependence of the amplitude on the higgs - boson mass . taking a 5d perspective on the calculation
, we have not distinguished between the sm modes and the kk particles .
however , in our final results we have been able to identify the contributions from the @xmath2 boson and the infinite tower of its kk excitations .
finally , we have generalized our findings to the rs model with custodial protection of electroweak precision observables , for which we have discussed two different embeddings of the lepton sector in the extended gauge group of the model . in the phenomenological part of this paper
, we have analyzed the new - physics effects on the cross section times branching ratio for the process @xmath346 , including effects on the production cross sections via gluon fusion or vector - boson fusion , the @xmath7 decay rate , and the total higgs width .
we have focused on the ratio @xmath369 representing the cross section times branching ratio in the rs model normalized to its sm value , and distinguished between the brane - localized and the narrow bulk - higgs scenarios of both the minimal and the custodial rs models .
we have also derived approximate formulas for @xmath369 , which allow for a reasonable parametrization of our results in terms of the kk mass scale and the maximal value @xmath388 imposed on the magnitude of the individual entries of the anarchic 5d yukawa matrices .
we have pointed out the fact that the rs corrections to higgs production via gluon fusion act in the opposite direction as those to the decay into two photons , where generally the new - physics effects on the gluon - fusion rate provide the dominant source of corrections to the sm predictions . for not too small values of @xmath388 and not too large values of @xmath375 , new - physics effects in rs models can lead to significant deviations of @xmath369 from 1 . while in all brane - higgs scenarios considered here @xmath406 in most regions of parameter space ( in good agreement with the cms data ) , rs scenarios with the higgs implemented as a narrow bulk field allow for the possibility that @xmath407 ( as suggested by the atlas data ) . more precise measurements of @xmath369 and of other higgs production and decay rates would be required to differentiate between different incarnations of rs models .
importantly , significant deviations of the higgs couplings from their sm values could arise even if the kk mass scale is so large that kk modes are out of reach for direct production at the lhc .
comparing our results with the latest atlas and cms data @xcite , we have derived exclusion regions in the @xmath408@xmath388 parameter space of the various models , shown in figures [ fig : rggexcmrs ] and [ fig : rggexccrs ] . in the minimal model , we can exclude kk gluon masses lighter than @xmath409 at 95% cl for the brane - localized scenario , and @xmath410 at 68% cl for the narrow bulk - higgs scenario . in case of the custodially protected rs model and for a brane - localized higgs
, we can exclude kk masses below @xmath411 at 95% cl . for a narrow bulk - higgs field
, we find the excluded region @xmath412 at 95% cl . together with our previous work @xcite , the present paper allows for a full treatment of the loop - mediated higgs couplings and can be supplemented by the remaining tree - level couplings to arrive at a comprehensive description of higgs physics in rs models , where the scalar sector lives on or near the ir brane . with increasing experimental precision on the extracted higgs couplings , it will be exciting to compare our results for various rs models with the data .
even if no kk excitations will be discovered at the lhc , it is conceivable that future precision measurements of higgs couplings at the lhc and ilc could provide a first hint on the existence of a warped extra dimension .
we are grateful to paul archer for useful discussions .
this research is supported by the advanced grant eft4lhc of the european research council ( erc ) , the cluster of excellence _ precision physics , fundamental interactions and structure of matter _ ( prisma exc 1098 ) , grant 05h12ume of the german federal ministry for education and research ( bmbf ) , the rhineland - palatinate research center _ elementary forces and mathematical foundations _ , and the dfg graduate school grk 1581 _ symmetry breaking in fundamental interactions_. with the exception of c.h .
, the authors gratefully acknowledge kitp santa barbara for hospitality and support during the early stages of this work .
kitp is supported by the national science foundation under grant no .
nsf phy11 - 25915 .
at ( -2,3 ) @xmath71 ; ( 0,0)node[left ] @xmath302 node[above](3.5 , 0 ) ; ( 3.5,0)node[above left](5.5,2)node[right]@xmath146 ; ( 3.5,0)(5.5,-2)node[right]@xmath413 ; ( 4.6,2 ) node[above left]@xmath414(3.5,1 ) ; ( 1,-0.5 )
node[below]@xmath415(2.5,-0.5 ) ; at ( -2,3 ) @xmath76 ; ( 0,0)node[left ] @xmath302 ( 3.5 , 0 ) ; ( 3.5,0)(5.5,2)node[right]@xmath416 ; ( 3.5,0)(5.5,-2)node[right]@xmath413 ; at ( -2,3 ) @xmath77 ; ( 2,0 )
( 4 , -2)node[right]@xmath417 ; ( 2,0)(4,2)node[right]@xmath413 ; ( 0,2)node[left ] @xmath302(2,0 ) ; ( 0,-2)node[left ] @xmath146(2,0 ) ; at ( -2,3 ) @xmath78 ; ( 0,0)node[left ] @xmath302 ( 3.5 , 0 ) ; ( 3.5,0)(5.5,2)node[right]@xmath418 ; ( 3.5,0)(5.5,-2)node[right]@xmath419 ; at ( -2,3 ) @xmath79 ; ( 5,0 ) ( 7 , -2)node[right]@xmath417 ; ( 5,0)(7,2)node[right]@xmath420 ; ( 3,2)node[left ] @xmath421(5,0 ) ; ( 3,-2)node[left ] @xmath422(5,0 ) ; at ( -2,3 ) @xmath80 ; ( 2,0 ) ( 5.5 , 0)node[right]@xmath417 ; ( 0,2)node[left ] @xmath423node[above right](2,0 ) ; ( 0,-2)node[left ] @xmath424node[below right](2,0 ) ; ( 1,2 ) node[above right]@xmath425(2,1 ) ; ( 1,-2 ) node[below right]@xmath426(2,-1 ) ; at ( -2,3 ) @xmath81 ; ( -2,0) ( -5.5 , 0)node[left ] @xmath417 ; ( 0,2)node [ right ] @xmath427(-2,0 ) ; ( 0,-2)node [ right ] @xmath428(-2,-0 ) ; at ( -2,3 ) @xmath82 ; ( 2,0 ) ( 4 , -2)node[right]@xmath417 ; ( 2,0)(4,2)node[right]@xmath420 ; ( 0,2)node[left ] @xmath423(2,0 ) ; ( 0,-2)node[left ] @xmath424(2,0 ) ; at ( -2,3 ) @xmath83 ; ( 2,0 ) ( 5.5 , 0)node[right]@xmath417 ; ( 0,2)node[left]@xmath429node[above right](2,0 ) ; ( 0,-2)node[left ] @xmath430(2,0 ) ; ( 2,1 ) node[above right]@xmath431(1,2 ) ; at ( -2,3 ) @xmath84 ; ( 0,0)node[left]@xmath302 ( 3.5 , 0 ) ; ( 3.5,0)(5.5,2)node[right]@xmath432 ; ( 3.5,0)(5.5,-2)node[right ] @xmath433 ; at ( -2,3 ) @xmath85 ; ( 4.5,0.5)node[above left]@xmath434 ( 3,0.5 ) ; ( 1,2 ) node[above right]@xmath425(2,1 ) ; ( 1,-2 ) node[below right]@xmath426(2,-1 ) ; ( 2,0) ( 5.5 , 0)node[right ] @xmath417 ; ( 0,2)node[left]@xmath435(2,0 ) ; ( 0,-2)node[left]@xmath436(2,-0 ) ; here we list the feynman rules needed for the calculation of the one - loop gauge - boson , scalar , and ghost contributions to the @xmath1 decay amplitude in the kk - decomposed version of the minimal rs model , using the notations of @xcite .
we work in a general @xmath0 gauge and use mass eigenstates of all sm particles and their kk excitations .
the feynman rules for the vertices shown above are : @xmath437 where @xmath9 is the higgs vev in the rs model , the parameter @xmath57 has been defined in ( [ varphidecomp ] ) , and the tensor structure @xmath438 of the 3-boson vertex has been given in the text after ( [ eqn:5damp ] ) .
for the analysis of the zero - mode contributions in section [ sec : analzeromrs ] we need to expand the @xmath2-boson propagator at @xmath199 in powers of @xmath3 . in this context
, we also need the relation between the physical @xmath2-boson mass and the model parameter @xmath57 in ( [ varphidecomp ] ) beyond the leading order .
a corresponding formula was derived in @xcite by solving the eigenvalue equation for the @xmath2-boson profiles and extracting the lowest eigenvalue .
we now present an alternative approach based on our exact expression for the 5d gauge - boson propagator , which allows us to derive explicit expressions for the @xmath2-boson mass and profile to any order in @xmath3 .
starting from the exact expression , we perform an expansion in powers of @xmath3 while keeping @xmath440 and @xmath441 fixed and of order @xmath300 .
this yields @xmath442 + \sigma(p^2 ) + i0 } \,,\ ] ] where @xmath443 \right\ } , \end{aligned}\ ] ] which are valid up to terms of order @xmath105 .
the zero of the denominator of the propagator in ( [ eqn : propexp ] ) defines the physical mass @xmath213 of the ground state , and the residue of the pole determines the corresponding product of profile functions @xmath444 . indeed , for @xmath445 we find @xmath446 where @xmath447 , \\ z_2(t ,
t ' ) & = 1 - \pi(t , t';m_w^2 ) - \frac{{\partial}\sigma(p^2)}{{\partial}p^2 } \big|_{p^2 = m_w^2 } \\ & = 1 + \frac{\tilde m_w^2}{4{m_{\text{kk}}}^2 } \left [ 2 - \frac{2}{l } + t^2 \left ( 1 - 2l - 2\ln t \right ) + t'^2 \left ( 1 - 2l - 2\ln t ' \right ) \right ] + \dots \ , . \end{aligned}\ ] ] these results are valid up to @xmath448 corrections .
the relation in the first line is equivalent to ( [ wmassrela ] ) . rewriting the second relation in the form @xmath449 ,
we can extract @xmath450 + \dots \right\ } , \ ] ] which gives the first non - trivial correction to the profile of the @xmath2 boson @xcite .
g. aad _ et al . _ [ atlas collaboration ] , phys .
b * 716 * , 1 ( 2012 ) [ arxiv:1207.7214 [ hep - ex ] ] .
s. chatrchyan _ et al . _
[ cms collaboration ] , phys .
b * 716 * , 30 ( 2012 ) [ arxiv:1207.7235 [ hep - ex ] ] .
l. randall and r. sundrum , phys .
lett . * 83 * , 3370 ( 1999 ) [ hep - ph/9905221 ] .
y. grossman and m. neubert , phys .
b * 474 * , 361 ( 2000 ) [ hep - ph/9912408 ] .
t. gherghetta and a. pomarol , nucl .
b * 586 * , 141 ( 2000 ) [ hep - ph/0003129 ] .
s. j. huber and q. shafi , phys .
b * 498 * , 256 ( 2001 ) [ hep - ph/0010195 ] .
k. agashe , g. perez and a. soni , phys .
lett . * 93 * , 201804 ( 2004 ) [ hep - ph/0406101 ]
. k. agashe , g. perez and a. soni , phys .
d * 71 * , 016002 ( 2005 ) [ hep - ph/0408134 ] . c. csaki , a. falkowski and a. weiler , jhep * 0809 * , 008 ( 2008 ) [ arxiv:0804.1954 [ hep - ph ] ] . s. casagrande , f. goertz , u. haisch , m. neubert and t. pfoh , jhep * 0810 * , 094 ( 2008 ) [ arxiv:0807.4937 [ hep - ph ] ] .
m. blanke , a. j. buras , b. duling , s. gori and a. weiler , jhep * 0903 * , 001 ( 2009 ) [ arxiv:0809.1073 [ hep - ph ] ] .
m. blanke , a. j. buras , b. duling , k. gemmler and s. gori , jhep * 0903 * , 108 ( 2009 ) [ arxiv:0812.3803 [ hep - ph ] ] . m. bauer , s. casagrande , u. haisch and m. neubert , jhep * 1009 * , 017 ( 2010 ) [ arxiv:0912.1625 [ hep - ph ] ] . a. djouadi and g. moreau , phys .
b * 660 * , 67 ( 2008 ) [ arxiv:0707.3800 [ hep - ph ] ] .
a. falkowski , phys .
d * 77 * , 055018 ( 2008 ) [ arxiv:0711.0828 [ hep - ph ] ] .
g. cacciapaglia , a. deandrea and j. llodra - perez , jhep * 0906 * , 054 ( 2009 ) [ arxiv:0901.0927 [ hep - ph ] ] .
g. bhattacharyya and t. s. ray , phys .
b * 675 * , 222 ( 2009 ) [ arxiv:0902.1893 [ hep - ph ] ] . c. bouchart and g. moreau , phys .
d * 80 * , 095022 ( 2009 ) [ arxiv:0909.4812 [ hep - ph ] ] .
s. casagrande , f. goertz , u. haisch , m. neubert and t. pfoh , jhep * 1009 * , 014 ( 2010 ) [ arxiv:1005.4315 [ hep - ph ] ] .
a. azatov , m. toharia and l. zhu , phys . rev .
d * 82 * , 056004 ( 2010 ) [ arxiv:1006.5939 [ hep - ph ] ] .
a. azatov and j. galloway , phys
. rev .
d * 85 * , 055013 ( 2012 ) [ arxiv:1110.5646 [ hep - ph ] ] .
f. goertz , u. haisch and m. neubert , phys .
b * 713 * , 23 ( 2012 ) [ arxiv:1112.5099 [ hep - ph ] ] .
m. carena , s. casagrande , f. goertz , u. haisch and m. neubert , jhep * 1208 * , 156 ( 2012 ) [ arxiv:1204.0008 [ hep - ph ] ] .
r. malm , m. neubert , k. novotny and c. schmell , jhep * 1401 * , 173 ( 2014 ) [ arxiv:1303.5702 [ hep - ph ] ] .
k. agashe , a. delgado , m. j. may and r. sundrum , jhep * 0308 * , 050 ( 2003 ) [ hep - ph/0308036 ] . c. csaki , c. grojean , l. pilo and j. terning , phys .
* 92 * , 101802 ( 2004 ) [ hep - ph/0308038 ] .
k. agashe , r. contino , l. da rold and a. pomarol , phys .
b * 641 * , 62 ( 2006 ) [ hep - ph/0605341 ] .
l. randall and m. d. schwartz , jhep * 0111 * , 003 ( 2001 ) [ hep - th/0108114 ] .
a. pomarol , phys .
lett . * 85 * , 4004 ( 2000 ) [ hep - ph/0005293 ] .
choi , h. d. kim and i .- w .
kim , jhep * 0211 * , 033 ( 2002 ) [ hep - ph/0202257 ] .
w. d. goldberger and i. z. rothstein , phys .
lett . * 89 * , 131601 ( 2002 ) [ hep - th/0204160 ] . c. delaunay , j. f. kamenik , g. perez and l. randall , jhep * 1301 * , 027 ( 2013 ) [ arxiv:1207.0474 [ hep - ph ] ] . m. frank , n. pourtolami and m. toharia , phys .
d * 87 * , 096003 ( 2013 ) [ arxiv:1301.7692 [ hep - ph ] ] .
m. puchwein and z. kunszt , annals phys .
* 311 * , 288 ( 2004 ) [ hep - th/0309069 ] .
r. contino and a. pomarol , jhep * 0411 * , 058 ( 2004 ) [ hep - th/0406257 ] .
m. s. carena , a. delgado , e. ponton , t. m. p. tait and c. e. m. wagner , phys .
d * 71 * , 015010 ( 2005 ) [ hep - ph/0410344 ] . c. csaki , y. grossman , p. tanedo and y. tsai , phys .
d * 83 * , 073002 ( 2011 ) [ arxiv:1004.2037 [ hep - ph ] ] .
p. archer , m. carena , a. carmona and m. neubert , in preparation .
w. j. marciano , c. zhang and s. willenbrock , phys .
d * 85 * , 013002 ( 2012 ) [ arxiv:1109.5304 [ hep - ph ] ] .
h. davoudiasl , j. l. hewett and t. g. rizzo , phys .
b * 473 * , 43 ( 2000 ) [ hep - ph/9911262 ] .
a. pomarol , phys .
b * 486 * , 153 ( 2000 ) [ hep - ph/9911294 ] . for a review , see : t. gherghetta , arxiv:1008.2570 [ hep - ph ] . g. cacciapaglia , c. csaki , g. marandella and j. terning , jhep * 0702 * , 036 ( 2007 ) [ hep - ph/0611358 ] .
p. r. archer , jhep * 1209 * , 095 ( 2012 ) [ arxiv:1204.4730 [ hep - ph ] ] .
c. csaki , j. erlich and j. terning , phys .
d * 66 * , 064021 ( 2002 ) [ hep - ph/0203034 ] .
f. goertz , arxiv:1112.6387 [ hep - ph ] .
m. s. carena , e. ponton , j. santiago and c. e. m. wagner , nucl .
b * 759 * , 202 ( 2006 ) [ hep - ph/0607106 ] .
g. cacciapaglia , c. csaki , g. marandella and j. terning , phys .
d * 75 * , 015003 ( 2007 ) [ hep - ph/0607146 ] .
r. contino , l. da rold and a. pomarol , phys .
d * 75 * , 055014 ( 2007 ) [ hep - ph/0612048 ] .
m. s. carena , e. ponton , j. santiago and c. e. m. wagner , phys .
d * 76 * , 035006 ( 2007 ) [ hep - ph/0701055 ] .
m. e. albrecht , m. blanke , a. j. buras , b. duling and k. gemmler , jhep * 0909 * , 064 ( 2009 ) [ arxiv:0903.2415 [ hep - ph ] ] . g. burdman and l. da rold , jhep * 0811 * , 025 ( 2008 ) [ arxiv:0809.4009 [ hep - ph ] ]
. s. heinemeyer _ et al . _
[ lhc higgs cross section working group collaboration ] , arxiv:1307.1347 [ hep - ph ] .
r. malm , m. neubert and c. schmell , in preparation .
a. denner , s. heinemeyer , i. puljak , d. rebuzzi and m. spira , eur .
j. c * 71 * , 1753 ( 2011 ) [ arxiv:1107.5909 [ hep - ph ] ] .
f. bishara , y. grossman , r. harnik , d. j. robinson , j. shu and j. zupan , arxiv:1312.2955 [ hep - ph ] .
g. aad _ et al . _ [ atlas collaboration ] , phys .
b * * ( 2013 ) [ arxiv:1307.1427 [ hep - ex ] ] . m. h. chan _ et al .
_ [ cms collaboration ] , cms - pas - hig-13 - 005 ( april 17 , 2013 ) .
t. plehn and m. rauch , europhys .
lett . * 100 * , 11002 ( 2012 ) [ arxiv:1207.6108 [ hep - ph ] ] .
m. e. peskin and t. takeuchi , phys .
d * 46 * , 381 ( 1992 ) .
m. s. carena , a. delgado , e. ponton , t. m. p. tait and c. e. m. wagner , phys .
d * 68 * , 035010 ( 2003 ) [ hep - ph/0305188 ] .
a. djouadi , phys .
rept . * 457 * , 1 ( 2008 ) [ hep - ph/0503172 ] .
|
a detailed five - dimensional calculation of the higgs - boson decay into two photons is performed in both the minimal and the custodially protected randall - sundrum ( rs ) model , where the standard model ( sm ) fields propagate in the bulk and the scalar sector lives on or near the ir brane .
it is explicitly shown that the @xmath0 gauge invariance of the sum of diagrams involving bosonic fields in the sm also applies to the case of these rs scenarios . an exact expression for the @xmath1 amplitude in terms of the five - dimensional ( 5d ) gauge - boson and fermion propagators
is presented , which includes the full dependence on the higgs - boson mass .
closed expressions for the 5d @xmath2-boson propagators in the minimal and the custodial rs model are derived , which are valid to all orders in @xmath3 .
in contrast to the fermion case , the result for the bosonic contributions to the @xmath1 amplitude is insensitive to the details of the localization of the higgs profile on or near the ir brane .
the various rs predictions for the rate of the @xmath4 process are compared with the latest lhc data , and exclusion regions for the rs model parameters are derived .
mitp/13 - 87 + december 20 , 2013 + arxiv:1312.5731 * higgs decay into two photons in a warped + extra dimension * juliane hahn@xmath5 , clara hrner@xmath5 , raoul malm@xmath5 , matthias neubert@xmath6 , + kristiane novotny@xmath5 and christoph schmell@xmath5 +
|
now we proof by the method of mathematical induction , that in the `` non - crossing '' approximation the electron green s function , averaged over impurity configurations , has the form of eq .
( [ gg ] ) : @xmath167with @xmath168 being independent of @xmath43 and @xmath14 . in the born approximation with one impurity the electron self energy @xmath169
does not depend on @xmath43 and @xmath14 , but depends on the difference @xmath170 : @xmath171and its real part is an odd function of @xmath172 , re@xmath173re@xmath174 , while its imaginary part is an even function of @xmath172 .
hence , in the born approximation @xmath175 .
assume that our condition is satisfied when each irreducable self - energy part contains no more than @xmath176 impurities .
then the green s function in coinciding points depends only on energy , @xmath177 : @xmath178when we include one more impurity to the irreducable electron self - energy and average over its position , it remains independent of @xmath179 and depends only on @xmath172 : @xmath180 .\]]the green s function @xmath181 also keeps the form of eq .
( [ gg ] ) . to show this ,
let us find the function @xmath182 \left [ 1-ug_{j}\left ( \varepsilon , r^{\prime } , r^{\prime } \right ) \right ] } + .. \\ & = & g_{0}\left ( \varepsilon , r_{1},r_{2}\right ) + \frac{u}{1-ug_{j}\left ( \varepsilon \right ) } \times \\ & & \times \sum_{n , k_{y},k_{z}}\frac{\psi _ { n , k_{y},k_{z}}^{0}(r_{1})}{\varepsilon -\epsilon _
( n , k_{z}\right ) -\sigma _ { 0,n}\left ( \varepsilon \right ) } \\ & & \times \sum_{n^{\prime } , k_{y}^{\prime } , k_{z}^{\prime } } \frac{\psi _ { n^{\prime } , k_{y}^{\prime } , k_{z}^{\prime } } ^{0\ast } ( { \boldsymbol{r}}_{2})}{\varepsilon -\epsilon _ { 3d}\left ( n^{\prime } , k_{z}^{\prime } \right ) -\sigma _ { 0,n^{\prime } } \left ( \varepsilon \right ) } \\ & & \times \int d^{3}r\psi _ { n , k_{y},k_{z}}^{0\ast } ( r)\psi _ { n^{\prime } , k_{y}^{\prime } , k_{z}^{\prime } } ^{0}({\boldsymbol{r}})+ .. \end{aligned}\]]the integration over @xmath183 in the last line gives @xmath184 , and the function @xmath185 becomes @xmath186 ^{2}}+ ..
\notag \\ & = & \sum_{n , k_{y},k_{z}}\frac{\psi _ { n , k_{y},k_{z}}^{0}(r_{1})\psi _ { n , k_{y},k_{z}}^{0\ast } ( { \boldsymbol{r}}_{2})}{\varepsilon -\epsilon _ { 3d}\left ( n , k_{z}\right ) -\sigma _ { 0,n}\left ( \varepsilon \right ) } \notag \\ & & \times \sum_{i=0}^{\infty } \left ( \frac{un_{i}/\left [ 1-ug_{j}\left ( \varepsilon \right ) \right ] } { \varepsilon -\epsilon _ { 3d}\left ( n , k_{z}\right ) -\sigma _ { 0,n}\left ( \varepsilon \right ) } \right ) ^{i } \notag \\ & = & \sum_{n , k_{y},k_{z}}\frac{\psi _ { n , k_{y},k_{z}}^{0}(r_{1})\psi _ { n , k_{y},k_{z}}^{0\ast } ( { \boldsymbol{r}}_{2})}{\varepsilon -\epsilon _ { 3d}\left ( n , k_{z}\right ) -\sigma _ { j+1,n}\left ( \varepsilon \right ) } , \label{gj1}\end{aligned}\]]where the new self - energy part @xmath187\]]depends only on energy @xmath36 and on ll number @xmath12 . the green s function in eq .
( [ gj1 ] ) is again of the form of eq .
( [ gg ] ) .
the difference of impurity scattering to the same and to other lls may give additional dependence of @xmath188 on ll number @xmath12 , which requires further study .
if @xmath189 depends only on @xmath190 , then @xmath191 entering above formulas also depends only on @xmath172 , and @xmath192 is also a function of @xmath172 only .
if @xmath193 depends only on @xmath36 but not on @xmath12 , and if the impurity scattering to the same and to other lls has the same matrix element , then @xmath191 also depends only on @xmath36 , and so does @xmath194 .
( 120,55 ) ( 0 , 5 ) ( 30,5)(0,14)3(0,1)10 ( 30 , 46 ) ( 28 , 41)x ( 30 , 5 ) ( 36 , 5 ) ( 54,5)(7,14)3(1,2)5 ( 92,5)(-7,14)3(-1,2)5 ( 73 , 46 ) ( 70 , 41 ) ( 54 , 5 ) ( 92 , 5 ) ( 54 , 6)(1,0)38 ( 54 , 4)(1,0)38 ( 96 , 5 ) ( 114,5)(7,14)3(1,2)5 ( 133,5)(0,14)3(0,1)10 ( 152,5)(-7,14)3(-1,2)5 ( 133 , 46 ) ( 130 , 41 ) ( 114 , 5 ) ( 152 , 5 ) ( 133 , 5 ) ( 114 , 6)(1,0)38 ( 114 , 4)(1,0)38 ( 156 , 5 ) some nonzero longitudinal magnetoresistance may appear in the @xmath195-approximation only in some special cases of complicated fs and particular directions of magnetic field , when there are singular electron trajectories.@xcite this notation @xmath197 denotes the area electron concentration on one ll .
it coincides with @xmath197 in ref .
@xcite and differs from the same notation in ref .
@xcite by the factor @xmath15 , because in ref .
@xcite @xmath197 denoted the volume ll degeneracy . in the 3d case the diagrams with the intersections of impurity lines are small by the parameter @xmath198 , where @xmath62 and @xmath199 are the volume impurity and electron concentrations . in 2d case in magnetic field
these crossing diagrams are not small by this parameter .
however , the calculations of the dos in refs .
@xcite show that these diagrams give a small correction to the result of `` non - crossing '' approximation , they describe the exponentially small tails of the dos . in quasi-2d case , which is intermediate between 3d and 2d case ,
the diagrams with intersections of impurity lines can also be neglected .
the spatial inhomogeniety of the electron density in @xmath200-direction due to the layered crystal structure also leads to some corrections in the electron self energy , which are equivalent to additional integration over the strength of the impurity potential.@xcite the summation over lls , performed in refs .
@xcite using the poisson summation formula to derive the scba equations , assumes that @xmath201 is independent on ll number @xmath12 , which requires additional proof in q2d limit .
|
the longitudinal interlayer magnetoresistance @xmath0 is calculated in strongly anisotropic layered metals , when the interlayer band width @xmath1 is less than the landau level separation @xmath2 .
the impurity scattering has much stronger effect in this regime than in 3d metals and leads to a linear longitudinal interlayer magnetoresistance @xmath3 in the interval @xmath4 changing to a square - root dependence @xmath5 at higher field or smaller @xmath6 . the crossover field allows to estimate the interlayer transfer integral as @xmath7 .
longitudinal interlayer magnetoresistance , being robust to the increase of temperature or long - range disorder , is easy for measurements and provides a useful tool to investigate the electronic structure of quasi - two - dimensional compounds .
the investigation of angular and field dependence of magnetoresistance ( mr ) provides a powerful tool of studying the electronic properties of various metals , including strongly anisotropic layered compounds , such as organic metals ( see , e.g. , refs .
@xcite for reviews ) , cuprate and iron - based high - temperature superconductors,(see , e.g. , husseynature2003,abdelnature2006,proustnature2007,abdelprl2007amro , mckenzie2007,dvignolle2008,helmnd2009,helmnd2010,bafeas2011,graf2012 ) heterostructures@xcite etc . in layered quasi - two - dimensional ( q2d ) metals with at least monoclinic crystal symmetry the electron dispersion in the tight - binding approximation is given by @xmath8where the 2d electron dispersion in magnetic field perpendicular to conducting layers is quantized in landau levels : @xmath9where the landau level ( ll ) separation @xmath10 , @xmath11 is the effective electron mass , @xmath12 is the ll number , @xmath13 , @xmath14 is out - of - plane electron momentum , and @xmath15 is the interlayer spacing . if the interlayer transfer integral @xmath16 , the standard 3d theory of galvanomagnetic propertiesabrik , shoenberg , ziman can be applied , which is valid in the lowest order of the parameter @xmath17 .
this theory predicts some special features of mr in q2d metals : the angular magnetoresistance oscillations ( amro)@xcite and the beats of the amplitude of magnetic quantum oscillations ( mqo).@xcite in more anisotropic q2d metals , when @xmath18 , several new features appear , such as slow mr oscillations@xcite and the phase shift of mqo beats between transport and thermodynamic quantities.@xcite these two effects are not described by the standard 3d theory @xcite because they appear in the higher orders in the parameter @xmath17 .
the monotonic part of mr also changes when @xmath19 . according to the standard theory,@xcite external magnetic field along the electric current leads only to mqo but does not influence the monotonic ( background ) part of this current.@xcite however ,
the monotonic growth of interlayer mr @xmath20 with the increase of longitudinal magnetic field @xmath21 was observed in various compounds as a general feature of q2d metals.so,coldea,pesotskiijetp95,zuo1999,w3,w4,incoh2009,wosnitza2002,kang its theoretical description is the aim of the present paper . in
very anisoropic and dirty compounds with @xmath22 , where @xmath23 and @xmath24 is electron mean free time in the absence of magnetic field , this monotonic growth of @xmath0 was attributed to the new `` strongly incoherent '' mechanisms of interlayer electron transport , such as metal - insulator transition and variable - range electron hopping with exponential temperature and field dependence of @xmath20,@xcite or the interlayer hopping via local crystal defects with in - series metallic intralayer transport.@xcite the in - plane component of magnetic field in the `` strongly incoherent '' regime leads to weaker mr compared to coherent 3d theory.@xcite however , in the most experiments the longitudinal interlayer magnetoresistance was observed together with the pronounced amro and metallic - type temperature dependence , which is inconsistent with `` strongly incoherent '' mechanisms of interlayer electron transport , which do not conserve the in - plane electron momentum during interlayer electron hopping .
recently it was shown , wiprb2011,wijetp2011,wifnt2011 that at very weak interlayer coupling , @xmath25 , the longitudinal interlayer magnetoresistance has a square - root monotonic growth @xmath26 even within the coherent - tunnelling model .
the angular dependence of mr also changes in this limit,@xcite which contradicts the previous common opinion@xcite that in the `` weakly incoherent '' regime , i.e. at @xmath27 , the interlayer magnetoresistance does not differ from the coherent almost 3d limit @xmath28 . in the present paper
we calculate the longitudinal interlayer magnetoresistance at @xmath29 , which generalizes the result of refs .
@xcite .
this extends the applicability of the square - root dependence @xmath30 to much wider region @xmath31 and gives linear @xmath0 dependence in the interval @xmath32 . at @xmath33
one can use the 3d electron dispersion in eq .
( es3d )
. then the interlayer electron conductivity can be evaluated at finite temperature using the 3d kubo formula,@xcite which gives@xmath34 \,\sigma _ { zz}(\varepsilon ) , \label{sigmazz}\]]where the derivative of the fermi distribution functioncommentchempotosc @xmath35 \ } , \label{nfd}\]]and the zero - temperature conductivity at energy @xmath36 is @xmath37 ^{2}\ , , \label{sigd1}\]]where @xmath38 is the electron charge , @xmath39 is the electron velocity , the sum over the electron quantum numbers @xmath40 ( excluding spin ) is taken in the unit volume , and the retarded electron green s function @xmath41 ^{-1}. \label{gr}\]]both @xmath42 and the electron self - energy is independent of @xmath43 , and the summation over @xmath43 in eq .
( [ sigd1 ] ) gives the factor equal to the ll degeneracy of one conducting layer per spin state @xmath44.@xcite the interlayer conductivity is now given by a sum over lls , @xmath45where the contribution to @xmath46 from the @xmath12-th ll is @xmath47 ^{2}\ , .
\label{sigd2}\ ] ] the electron self - energy @xmath48 depends only on energy @xmath36 and , possibly , on ll number @xmath12 , being independent of @xmath14 ( see appendix ) . substituting eq .
( [ gr ] ) and performing integration over @xmath14 , one obtains@xmath49where @xmath50re@xmath51 , and @xmath52 denotes the interlayer conductivity without magnetic field:@xmath53@xmath54 is the 3d dos at the fermi level in the absence of magnetic field per two spin components , @xmath55 and @xmath56 . to calculate the electron self - energy @xmath57 we start from the standard model of 3d strongly anisotropic metals .
the hamiltonian consists of two terms : @xmath58 .
the first term @xmath59 describes 3d noninteracting electrons in magnetic field with anisotropic dispersion given by eqs .
( [ es3d]),([e2d ] ) .
the second term @xmath60 describes the electron interaction with impurities .
the impurities are taken to be point - like with the potential @xmath61 and randomly distributed with volume concentration @xmath62 . without magnetic field
the broadening @xmath63im@xmath64 of electron levels due to the scattering by impurities in the born approximation is @xmath65 , where @xmath66 is the density of electron states ( dos ) at the fermi level . in q2d metals in strong magnetic field ,
@xmath67 , the electron self - energy @xmath68 depends on the energy deviation @xmath69 from the @xmath12-th ll . in the `` non - crossing '' approximation,@xcite
schematically shown in fig .
[ figse ] , the electron green s function , averaged over impurity configurations , has the form ( see appendix)@xmath70where @xmath71 are the electron wave functions in magnetic field without impurities and @xmath72 is the electron self energy , averaged over impurity positions and given by the equation @xmath73 .
\label{ser}\]]here @xmath74 after the integration over @xmath43 and @xmath14 is given by@xmath75the system of equations ( [ ser ] ) and ( [ gt ] ) allows to calculate the electron self - energy @xmath76 numerically . for
weak impurity potential @xmath77 the self - consistent born approximation ( scba ) is applicable , and eq .
( ser ) reduces to@xmath78the scba is valid when the scattering potential of each impurity is weak compared to fermi energy , but the concentration of impurities can be arbitrary , so that the broadening of electron levels @xmath79 is also arbitrary . in ref .
@xcite it was shown that one needs to apply at least scba to obtain qualitatively correct monotonic growth of interlayer longitudinal mr , which differs in scba and in the non - crossing approximations only by a factor close to unity .
the similar system of scba equations for the electron green s function in q2d metals in magnetic field was written previously ( see eqs . ( 11 ) and ( 15 ) of ref .
@xcite , eqs .
( 11)-(13 ) in ref .
@xcite , eq .
( 29 ) in ref .
@xcite , eq .
( 17 ) in ref .
@xcite or eq .
( 6 ) in ref .
@xcite ) .
there the scba equations have not been solved , and the simple born approximation was applied to calculate the interlayer conductivity . in refs .
@xcite only the limit @xmath80 has been considered , while in ref .
champelmineev a large electron reservoir was introduced to damp the mqo and to make the simple born approximation applicable.@xcite consider the strong magnetic field limit , when @xmath67 and the lls do not overlap .
then one can consider only one ll in the solution of scba eqs .
( [ gt ] ) and ( [ seba ] ) , which simplify to @xmath81where @xmath82 and @xmath83re@xmath84 . at @xmath85
we obtain the 2d result@xcite for the electron self - energy in scba .
note that the scattering enters eq .
( [ eqs ] ) only in the combination @xmath86 .
therefore , at @xmath87 , @xmath88 is the only energy scale in eq .
( eqs ) , and naturally im@xmath89 according to ref .
@xcite , which leads to the similar field dependence of background mr:@xcite @xmath90 .
eq .
( [ eqs ] ) rewrites as an algebraic equation of the fourth power:@xmath91 = \left ( \gamma _ { 0}\hbar \omega _ { c}\right )
^{2}. \label{eqs4}\]]among four solutions of this equation , only one satisfies the physical requirement @xmath92 at @xmath93 its solution @xmath94 is shown in figs .
[ figims],[figres ] for three different values of @xmath95 .
one can see from fig .
1 that im@xmath96 in the finite energy interval of the width @xmath97 at @xmath98 and of the width @xmath99 at @xmath100 .
re@xmath101 has cusps at the boundaries of this energy interval . of electron self energy as function of energy deviation @xmath102 from the center of ll at @xmath103 ( blue solid line),@xmath104
( green dashed line ) , and @xmath105 ( red dotted line).,scaledwidth=49.0% ] at @xmath106.,scaledwidth=49.0% ] now we substitute these solutions into eq .
( [ s0 t ] ) .
the result for monotonic part of interlayer mr @xmath107 is shown in fig .
[ figrzz ] , where @xmath108 is the conductivity averaged over period @xmath109 of mqo . from this figure
we see the crossover from linear to square - root field dependence of interlayer background mr @xmath0 .
the interval of the linear mr is @xmath110 and increases with the increase of @xmath111 .
the crossover from linear to square - root dependence of mr is a general feature of quasi-2d metals and already has been observed in a number of experiments ( see , e.g. refs .
coldea , pesotskiijetp95 ) .
the square - root dependence @xmath3 is obtained at @xmath112 , which is much wider than the applicability region @xmath113 of the calculation in refs . wiprb2011,wijetp2011 . calculated from eqs .
( [ s0 t ] ) and ( [ eqs4 ] ) as function of magnetic field at four different values of @xmath114 ( solid green line ) , @xmath115 ( dashed blue line ) , @xmath116 ( dotted red line ) and @xmath117 ( dash - dotted black line ) . in the interval
@xmath118 the interlayer mr has linear field dependence.,scaledwidth=49.0% ] calculated from eqs .
( [ s0 t ] ) and ( [ eqs4 ] ) as function of the square root of magnetic field at four different values of @xmath114 ( solid green line ) , @xmath119 ( dashed blue line ) , @xmath116 ( dotted red line ) and @xmath117 ( dash - dotted black line ) . the linear dependence @xmath120 is obtained at @xmath121 . only in the interval @xmath122 the interlayer mr has linear field dependence.,scaledwidth=49.0% ] this result on the field dependence of interlayer mr can be analytically obtained as follows .
eq .
( [ eqs4 ] ) simplifies to a quadratic equation at @xmath123 , i.e. at the center of ll , and has two solutions : @xmath124 .
the physical solution does not diverge at @xmath125 , has nonzero imaginary part and , at @xmath85 , agrees with the 2d limit described in refs .
@xcite .
all these criteria are satisfied for sign `` @xmath126 '' , which at @xmath127 gives re@xmath128 and @xmath129substituting this to eq .
( [ s0 t ] ) one obtains the expression for the zero - temperature conductivity in the center of ll , i.e. at @xmath130:@xmath131 in the limit @xmath132 each ll gives an essential contribution to conductivity in the interval @xmath133 , as follows from eq .
( [ eqs4 ] ) and can be seen from fig .
[ figseps ] , where the conductivity as function of energy distance from the nearest ll has a dome shape of the width @xmath99 .
the conductivity , averaged over period @xmath134 of mqo , is then given by @xmath135this predicts a linear background magnetoresistance @xmath136 in the interval @xmath137 of magnetic field in quasi-2d strongly anisotropic compounds .
the extra factor @xmath138 in eq .
( [ sav ] ) comes from the integration of a semicircle .
eq .
( [ sav ] ) also predicts stronger dependence of @xmath108 on @xmath6 : @xmath139 unlike the usual dependence @xmath140 . in the opposite limit @xmath141
the `` width of conducting band '' from each ll is @xmath142 , and for background interlayer conductivity one obtains @xmath143 , or @xmath144 in agreement with refs .
@xcite and fig .
[ figsqrtbz ] . calculated from eqs .
( [ s0 t ] ) and ( [ eqs4 ] ) in the strong - filed limit @xmath145 as function of energy counted from the nearest ll for four different values of @xmath146 ( solid blue line ) , @xmath115 ( dashed green line ) , @xmath147 ( dotted red line ) , and @xmath148 ( dash - dotted magenta line ) . in the limit @xmath149
the interlayer conductivity has approximately a semicircle shape of width @xmath150 . in the opposite limit @xmath151
the `` conducting band '' of each ll has the width @xmath152.,scaledwidth=49.0% ] the magnetic field @xmath153 of the crossover from linear to square - root dependence of mr @xmath0 can be used to estimate the value of the interlayer transfer integral in the compound : @xmath154 , where @xmath155 corresponds to the crossover field .
the obtained monotonic dependence of mr originates from mqo but survives at much higher temperature , when mqo are completely suppressed .
according to eqs .
( [ sigmazz])-([sigd1 ] ) , the temperature smearing of fermi distribution function does not influence the monotonic dependence of mr . only at much higher temperature , when the electron scattering by phonons plays the major role in the relaxation of electron momentum , the above monotonic dependence of mr is weakened .
similarly , the long - range disorder , not included in the above calculation , produces the local variation of the fermi energy along the sample on the scale greater than magnetic length , which damps the mqo but leaves the monotonic part of mr almost unchanged .
therefore , even if mqo are not seen in a compound because of strong disorder or high temperature , the background longitudinal interlayer mr can be observed and provides useful information about the electronic structure of this compound . this feature of background mr is similar to that of slow oscillations in q2d compounds at @xmath156.@xcite to summarize , the calculation of longitudinal interlayer magnetoresistance in q2d metals is performed in the strong - field limit @xmath157 .
it predicts a linear background mr @xmath158 in the interval @xmath159 of magnetic field , where @xmath108 has also unusual dependence on interlayer transfer integral : @xmath139 . at stronger field or at smaller @xmath160 , the usual dependence @xmath161
is recovered , and the mr transforms to square - root dependence @xmath162 , as was recently obtained in the limit @xmath163 in refs . @xcite .
the present calculation generalizes these results to the region @xmath164 .
the magnetic field of the crossover from linear to square - root dependence of @xmath165 allows to estimate the interlayer transfer integral @xmath7 from experimental data .
the obtained longitudinal mr @xmath165 explains numerous experiments on interlayer mr in strongly anisotropic quasi-2d compounds .
the measurement of the monotonic part of longitudinal interlayer mr is much easier than the measurement of mqo or amro , because finite temperature and long - range crystal imperfections do not affect @xmath166 up to much higher temperatures or disorder .
therefore , the experimental study of longitudinal interlayer background magnetoresistance is proposed as a simple additional tool to investigate the electronic structure of strongly anisotropic quasi - two - dimensional compounds in a wide range of parameters .
the work was supported by lea ens - landau exchange program and by rfbr .
|
observationally , due mostly to the inapplicability of high - precision doppler techniques to stars with spectral types earlier than late - f ( which have large rotational velocities and a small number of spectral lines ) , little was known until recently about the frequency of planets around the more - massive stars .
this situation has now changed , as surveys have been extended to searches for planets around more - massive stars in an evolved stage @xcite .
one of the most important trends that these surveys have revealed is the lack of close - in planets orbiting stars with masses @xmath1 @xcite despite the fact that these planets are found around @xmath220 % of the main sequence ( ms ) stars with @xmath3 .
the frequency of planets seems also to be higher around intermediate - mass stars @xcite .
furthermore , contrary to the correlation between metallicity and probability of planet - hosting found for solar - mass , main - sequence stars ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , the more - massive planet - hosting stars do not exhibit higher metallicities ( e.g. , * ? ? ?
* ) . on the theoretical side
, it has been shown that the formation of jupiter - mass planets around m stars may be hindered ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , while the probability that a given star has at least one gas giant increases linearly with the stellar mass up of 3 @xmath0 ( e.g. , * ? ? ?
one possibility is that the observed difference in the orbital distribution of planets found around intermediate and solar - mass primaries is due to the evolution of the star .
since all of the planets orbiting red giants or subgiants with @xmath1 have semi - major axes @xmath4 au it has been suggested that the planets might be engulfed as the star evolves off the ms @xcite .
this possibility is , however , often dismissed out in the literature with the argument that high - mass stars are physically too small to engulf hot jupiters @xcite .
another possibility is that the observed differences in orbital distribution are primordial , and and they are a consequence of the planet formation mechanism around more massive stars . along these lines
, it has been shown by @xcite that the dependence of the lifetime of the gaseous disk on the stellar mass could result in halting the inward migration of planets around high - mass stars , thus explaining the observed lack of short period planets around these stars .
the point we are making in the present paper is that before stellar evolution can be ruled out as the mechanism behind the observed semi - major axis distribution of planets around evolved stars , detailed modeling of the orbital evolution needs to be performed .
in other words , in order to determine the potential role of the stellar mass in the planet formation process , the effects of the evolution of the star on the observed orbit distribution around giants have to be correctly isolated .
this is precisely the goal of this paper .
there are several competing processes that affect the orbital distance between the star and the planet as the star evolves off the ms : the changes in the mass of both the planet and the star , the gravitational and frictional drag , and the tidal force .
to determine the rate of change in the planet s mass , we consider a planet of mass @xmath5 and radius @xmath6 moving with a velocity , @xmath7 , in a circular orbit ( @xmath8 ) around a star of mass @xmath9 . since the planet is moving supersonically through the matter ejected by the giant star , it accretes mass .
the accretion rate onto the planet , @xmath10 , is given approximately by the bondi - hoyle expression ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , @xmath11 where @xmath12 is the density of the environment and @xmath13 is the accretion radius ( @xmath14 with @xmath15 the gravitational constant ) . at very short distances ( where @xmath16 ) we have replaced @xmath17 by @xmath18 to correct the geometrical radius by gravitational focusing effects . at the same time , the planet s surface is being heated by radiation arising from the shock front and from the stellar surface .
this heating can lead to evaporation of surface material .
we estimate the evaporation rate , @xmath19 as in villaver & livio ( 2007 ; eq .
the temperature at the planet s surface has been taken to be the maximum between the radiative equilibrium temperature of the planet ( see e.g. , eq . 5 in @xcite ) and the temperature of the shocked gas @xmath20 ( estimated from the rankine - hugoniot conditions for an adiabatic shock where @xmath21 is boltzman s constant and @xmath22 the mass of the hydrogen atom ) .
the rate of change in the planet mass is thus given by @xmath23 the rate of change of the stellar mass is simply @xmath24 , where @xmath25 is the stellar mass - loss rate . using reimers law for red giants @xcite , @xmath26~~,\ ] ] where @xmath27 , @xmath28 and @xmath9 are the stellar luminosity , radius , and mass respectively ( in solar units ) and @xmath29 is the reimers parameter , which we take throughout this work to be @xmath30 conservation of angular momentum gives the equation for the rate of change in the orbital radius of the planet ( see , e.g. , @xcite ) , @xmath31-\left(\frac{\dot{a}}{a}\right)_{t}~~ , \label{todo}\ ] ] where @xmath32 and @xmath33 are respectively the frictional and gravitational drag forces and @xmath34 is the rate of orbital decay due to the tidal interaction .
the gravitational drag force , @xmath33 , arises from the eddying motions that are set up in the fluid by the passage of the planet .
it is a consequence of the gravitational interaction of the planet with a gaseous medium .
the drag force is given by ( e.g. , * ? ? ? * and references therein ) @xmath35 where @xmath36 is a time - dependent function of the mach number .
the numerical results of @xcite show that for the mach numbers encountered here , @xmath36 is approximately constant and has the value @xmath37 .
the loss of angular momentum associated to the frictional force @xmath32 is proportional to the surface area of the planet exposed to the flow and it can be expressed in the form ( e.g. , * ? ? ?
* ) @xmath38 where @xmath39 is the dimensionless drag coefficient for a sphere . finally , the angular momentum loss associated to the tidal term @xmath34 arises from the additional force ( besides the gravitational pull between the two centers of mass ) resulting from the non - spherical part of the mass distribution from the tidally distorted companion . in giant stars , which have massive convective envelopes ,
the most efficient mechanism to produce tidal friction is turbulent viscosity ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the dissipation timescale is determined by the effective eddy viscosity , with eddy velocities and length scales given approximately by standard mixing length theory if convection transports most of the energy flux @xcite .
the tidal term is given by @xmath40 with @xmath41 being the mass in the convective envelope , @xmath42 , and @xmath43 the eddy turnover timescale , given in the case of a convective envelope @xcite , @xmath44^{1/3}~~ , \label{eddy}\ ] ] where @xmath45 is the radius at the base of the convective envelope .
the term @xmath46 in eq .
( [ tidal ] ) is a numerical factor obtained from integrating the viscous dissipation of the tidal energy across the convective zone .
@xcite used @xmath47 where @xmath48 is the mixing length parameter .
@xcite confirmed that observations are consistent with @xmath491 ( specifically , they obtained values in the range @xmath50 with a preference to values in the upper half of the range ) as long as @xmath51 with @xmath52 being the orbital period .
we therefore used @xmath53 to account only for the convective cells that can contribute to viscosity when @xmath54 , otherwise we take @xmath55 .
the evolution of orbital eccentricities is beyond the scope of the present paper .
the damping of eccentricities due to tidal forces may eventually lead to a narrow distribution of e - values for close - in planets ( e.g. , * ? ? ?
* ; * ? ? ?
note that the initial value of the eccentricity has little effect on the orbital decay rate ( e.g. , * ? ? ?
calculation of the tidal term requires a knowledge of the structure of the star .
we have used detailed stellar models provided to us by lionel siess .
these were calculated based on the stellar evolution code starevol described in @xcite .
we have used stars with ms masses of 1 , 2 , 3 , and 5 , solar metallicity , and a mass - loss prescription with a reimers parameter of @xmath30 .
to determine the evolution of the planet s orbit we integrate eq .
[ todo ] along the rgb . the stellar evolution timescales ( in the absence of significant mass loss )
are set by the rate of consumption of the nuclear fuel . since the nuclear burning has been exhausted in the core , during the rgb hydrogen burning continues in a shell outside the helium core , which now , devoid of energy sources , is contracting and heating up .
as the core contracts the envelope expands and cools .
the stellar effective temperature decreases while the star s radius and luminosity increase .
we have updated the stellar parameters at each time step of the orbital evolution . as the star evolves off the ms the wind velocity decreases ( @xmath5610 ) . the termination shock which marks the position where the stellar wind interacts with the ism cools down and the temperature of the medium at the planet s location has an upper limit provided by the stellar effective temperature ( e.g. , * ? ? ?
we used @xmath57 k as the temperature of the ambient medium .
the mach number @xmath58 is then always @xmath59 for the orbital distances relevant to this work .
the density of the environment has been computed as @xmath60 unless @xmath61 , in which case we took a number density of @xmath62 @xmath63 .
we use a wind velocity of 5 along the rgb .
we have computed the orbital evolution for a range of initial orbital distances and planet and stellar masses in order to determine the minimum initial orbital distance for which a planet will avoid being tidally captured by the expanding star . for illustration , some of the computed orbits ( dash - dotted lines ) are plotted in fig .
[ fig_jupiters ] , where we have used a planet with a mass of @xmath64 @xmath65 orbiting stars with different ms masses to the end of the rgb phase . the evolution of the stellar radius is shown as a solid line .
the panels , from top to bottom and left to right , are for stars with ms masses of 1 , 2 , 3 , and 5 @xmath66 respectively .
note that the scale on the axes is different for each panel . at large distances from the star , the densities involved are low , and the drag terms associated with the forces f@xmath67 and f@xmath68 in eq .
( [ todo ] ) play a negligible role in the evolution of the orbit .
moreover , since the accretion rate onto the planet is always small compared to the stellar mass - loss rate , the first term in eq .
( [ todo ] ) is dominated by @xmath69 .
the temporal behavior of the orbit is then mostly governed by the relative importance of the terms associated with the stellar mass - loss @xmath69 and the tidal interaction @xmath70 .
red giant mass - loss rates are somewhat uncertain .
as noted in 2 we have in the present work estimated the mass - loss rate by the reimers prescription with @xmath30 .
this seems to reproduce fairly well the observations of individual rgb stars .
the peak rgb mass - loss rates are higher for lower - mass stars ( 10@xmath71 for 1 @xmath72 versus 10@xmath73 for the 3 @xmath72 ) , and the lowest mass stars also reach the largest radius at the tip of the rgb .
figure [ fig_jupiters ] demonstrates the three possible outcomes of orbital evolution : ( i ) beyond a certain initial orbital separation , the orbital separation simply increases , due to systemic mass loss .
( ii ) there is a range of initial orbital separations for which the orbit decays , but the planet avoids being engulfed .
( iii ) inward from some critical , initial orbital separation , the planet is engulfed mostly due to tidal interaction . with the purpose of quantifying the influence of the planet s mass , we have also integrated the orbit of planets with masses of 3 and 5 @xmath74 ( see figs .
[ fig_2 m ] and [ fig_3 m ] ) .
the top panel of fig .
[ fig_2 m ] shows the rgb evolution of a 2 @xmath72 star along the hr diagram ( fig .
[ fig_3 m ] is the same , but for a 3 @xmath72 star ) .
the bottom panel shows the orbital evolution for planets with masses 3 @xmath65 ( dotted line ) and 5 @xmath65 ( dash - dotted line ) together with the evolution of the stellar radius ( solid line ) .
note that the bottom panel does not show the maximum extent of the stellar radius , but a zoom of a section at the base of rgb , to better appreciate the details .
we selected small initial orbits to identify at which points during the rgb these planets are swallowed by their stars ( marked by the location of the arrows ) .
for all the initial orbits that satisfy the condition @xmath75 the planet gets engulfed by the star at same point before the end of the rgb phase .
the more massive the planet the stronger is the tidal interaction with the star , and therefore the sooner the orbit decays to meet the stellar radius ( see figs .
[ fig_2 m ] and [ fig_3 m ] ) .
some of our findings are summarized in table [ tableorbit ] , where we list the minimum initial orbital distance for which a planet with a given mass avoids being engulfed by the star .
the second column gives the maximum radius reached by the star on the rgb , @xmath76 ( in au ) , and the following columns give the minimum orbital distances ( in au ) at which planets with masses of 1 , 3 , and 5 m@xmath77 ( respectively ) avoid being engulfed . for the stellar masses of 1 and 2 @xmath66 , we calculated a grid using initial orbits at steps of 0.1 au , for the 3 and 5 @xmath66 stars we used initial orbits at steps of 0.01 and 0.05 au , respectively .
two important conclusions can be extracted from table [ tableorbit ] .
first , the tidal `` capture '' radius increases with the planet s mass ; a jupiter - mass planet is captured by a 2 @xmath66 star if it starts at an initial orbit of 2.1 au while a 5 @xmath65 planet will be engulfed by the star if it has an initial orbit @xmath78 au .
second , the tidal capture radius decreases with increasing stellar mass ; for a 5 @xmath65 planet the initial orbit has to be larger than @xmath79 to avoid tidal capture around a 2 @xmath72 star , while it has to be larger than @xmath80 to avoid tidal capture around a 5 @xmath72 star .
our main goal has been to determine whether stellar evolution could explain the observed distribution of the semi - major axes of planetary orbits around evolved stars ( i.e. , semi - major axis @xmath810.5 au ) .
we found that when the details of the orbital evolution are accurately calculated , tidal interactions constitute a quite powerful mechanism , capable of capturing close - in planets into the envelope of evolved stars . to date
, there are @xmath8220 exoplanets discovered around giant stars with @xmath83 .
the host stars have radii in the range @xmath84 au .
our models are consistent with the existence of these planets at the point in the rgb evolution at which they are observed ( see e.g. , table 8 in @xcite and our figs .
[ fig_2 m ] and [ fig_3 m ] ) .
however , we do not expect to find massive planets with a @xmath850.4 au around a 2 @xmath72 star with a r@xmath86 au ( or 24 @xmath87 ) .
our calculations provide the minimum orbital radius inside of which planets will be engulfed by the star at the end of the rgb evolution .
we find that the evolution of the star alone can quantitatively explain the observed lack of close - in planets around evolved stars even allowing for the uncertainties associated with mechanisms such as mass loss along the rgb or tidal - interaction theory . a mechanism such as the one invoked by currie ( 2009 ; i.e.
, a lifetime stellar - mass dependency of the gas in the planet - forming disk that can halt migration ) is not needed , although it might still be present .
we find that given an initial distance at which tidal capture is possible , the more massive the planet , the earlier it will be captured by the rgb envelope .
observationally , it appears that giant stars host more massive planets than ms stars ( e.g. , * ? ? ?
* ; * ? ? ?
* ) . since we find that more massive planets are expected to be engulfed earlier in the rgb evolution , the fact that they are more frequently observed may point towards a planet - formation mechanism that favors the formation of more - massive planets around intermediate - mass stars
this would be consistent with a scenario in which the disk mass scales with the stellar mass , and more massive disks produce more massive planets ( see also @xcite ) .
along similar lines , since we find a high probability of tidal capture of the planet by evolved stars , the higher frequency of planets observed around intermediate - mass stars @xcite seems to imply that the efficiency of planet formation must be considerably higher for more massive stars , compared to their solar analogous .
@xcite estimated the probability of detecting transits of planetary companions to giant stars to be @xmath88 10 % for several of the known systems .
since tidal oribital decay decreases the initial orbital distance , it may increase the probability for the planet to be observed in transit .
our results suggest that many planets may be accreted by their host star ( see also * ? ? ?
* ; * ? ? ?
although so far the results are based on a fairly limited sample , it appears that giant stars hosting planets have the same metallicity distribution as giant stars without planets @xcite . on the other hand
, we should note that if a giant star engulfs a close - in planet early in the rgb , it will appear to be a giant star without a planet .
this mechanism might perhaps provide for a partial explanation for the lack of correlation with metallicity .
finally , stars that have accreted planets may show a higher spin rate , as the planet s orbital angular momentum is transferred to the star ( e.g. * ? ? ?
this phenomenon has been investigated by @xcite and @xcite , where they estimate the probability of finding rapid rotators among evolved stars .
our quantification of the tidal capture radius may help refine these calculations .
|
recent surveys have revealed a lack of close - in planets around evolved stars more massive than 1.2 @xmath0 .
such planets are common around solar - mass stars .
we have calculated the orbital evolution of planets around stars with a range of initial masses , and have shown how planetary orbits are affected by the evolution of the stars all the way to the tip of the red giant branch ( rgb ) .
we find that tidal interaction can lead to the engulfment of close - in planets by evolved stars .
the engulfment is more efficient for more - massive planets and less - massive stars .
these results may explain the observed semi - major axis distribution of planets around evolved stars with masses larger than 1.5 @xmath0 .
our results also suggest that massive planets may form more efficiently around intermediate - mass stars .
|
for reactions with two incoming spin 1/2 particles one can define the initial state cross sections @xmath8 where @xmath9 denotes the combined spin of the two particles and @xmath10 its projection .
the three possible initial spin state contributions @xmath11 , @xmath12 and @xmath13 are related to @xmath14 and @xmath15 by isospin , parity and angular momentum conservation @xmath16 @xmath17 @xmath18 near threshold only a limited number of partial waves are important and considering the allowed transitions to final states is the angular momentum between the two nucleons and @xmath19 is the angular momentum of the pion with respect to the nucleon pair . ] with @xmath20 and @xmath21 , i.e. assuming @xmath22 , we obtain the following relations is forbidden by parity and angular momentum conservation ] @xmath23 @xmath24 @xmath25 where @xmath26 corresponds to @xmath27 , @xmath28 initial states that contribute to @xmath29 final states but not to @xmath30 . to which extent the ps partial waves contribute to @xmath30 can then be obtained in a model - free way@xcite using equ .
1 - 6 @xmath31 theoretical models for pion production near threshold can thus be tested in an unprecedented manner using measurements of spin observables of the reaction @xmath0@xcite .
the results for the spin - correlation coefficient combinations @xmath32 , and @xmath33 ( the latter equivalent to @xmath34 ) and the analyzing power , @xmath35 , which only needed transverse beam and target polarization , were published recently@xcite and the publication of the the data from the experiment using also longitudinal polarization components is in progress@xcite .
the simultaneous measurements of spin correlation coefficients in @xmath36 will further enhance the usefulness of this experiment for resolving the theoretical issues of @xmath37 reactions@xcite .
section 2 contains a brief description of the experimental method which is described in detail elsewhere@xcite . in section 3
we discuss the preliminary results of the measurements using polarized beam with longitudinal components , and the possible implications .
the vertically polarized beam was injected into the iucf cooler ring at 197 mev and then accelerated to 325 , 350 , 375 or 400 mev .
the longitudinal component of the beam polarization was achieved using one superconducting solenoid and the three existing solenoids in the cooling section operated in a non - compensating mode . at the injection energy the polarization at the target was almost entirely longitudinal but
since the solenoid fields remained constant , after acceleration only partially longitudinal beam polarization could be achieved .
the major components were along and vertical to the beam momentum and one minor component was sideways .
the beam spin direction was reversed at each new injection .
the target was polarized atomic hydrogen in a gas storage cell fed by an atomic beam source@xcite .
the target thickness was of the order of @xmath38 atoms@xmath39 .
the spin direction was changed every two seconds , along the @xmath40 , @xmath41 and @xmath42 axes , by switching weak guide fields ( @xmath43 mt ) , supplied by helmholtz - like coils external to the target chamber .
a total of 12 beam / target spin combinations were used . .
the shaded area indicates the background which has been normalized to the area corresponding to invariant masses below 100 mev / c@xmath44 .
the filled circles represent the data after the background has been subtracted , the statistical errors are shown . the resolution ( fwhm ) of the background
subtracted peak is 7.2 mev / c@xmath44.,width=283 ] .
the shaded area indicates the background which has been normalized to the area corresponding to invariant masses below 100 mev / c@xmath44 .
the filled circles represent the data after the background has been subtracted , the statistical errors are shown .
the resolution ( fwhm ) of the background subtracted peak is 7.2 mev / c@xmath44.,width=283 ] the four - momenta of the two outgoing protons were measured in a cylindrically symmetric forward detector stack , see fig .
[ fig : turkey ] , which allowed for the missing mass , @xmath45 , of the third particle to be calculated . in fig .
[ fig : mx ] the @xmath45 distribution is shown for the data set with longitudinal and transverse beam spin components summed over all spin states at a beam energy of 325 mev .
the shape of the background in the spectrum was well reproduced by data taken with nitrogen gas instead of hydrogen in the target .
full coverage of phase space was achieved with the exception of the loss of events when protons go undetected into the beampipe , i.e. @xmath46 of phase space distributed events were lost at 325 - 400 mev . in an investigation of systematic errors
the spin observables were studied as a function of an artificial @xmath47 cut - off using an effective - range expansion for the @xmath48 final state interaction@xcite .
the corrections applied for the relative spin - dependent cross sections were less than the statistical errors .
the cross - ratio method@xcite which eliminates asymmetries due to detector acceptance and spin - dependent luminosities , and division by the product where @xmath49 is the ratio of the yields , @xmath50 and @xmath51 is the product of the beam and target polarization . ] of the beam and target polarization projections , @xmath52 or @xmath53 , give @xmath54 and @xmath55 from the spinsorted and backgroundsubtracted yields of the @xmath56 .
the values of @xmath52 and @xmath53 are obtained . by detecting @xmath48 elastic scattering at @xmath57 simultaneously with the pion production and measuring the @xmath48 spin correlation coefficients @xmath58 and @xmath59 which are large and known with high precision at this angle@xcite .
the analysis of the experiment described here is near completion@xcite .
preliminary results using equ . 1 - 3 are shown in fig . [
fig : sms ] . also depicted are curves representing predictions by a microscopic model calculation , which is based on direct production and off - shell rescattering diagrams .
contributions from heavy meson - exchange @xmath60 are added and the @xmath61-isobar is included explicitly@xcite .
previous comparison with data yielded qualitative agreement for @xmath62 and @xmath35 , but disagreed completely with the data on @xmath58@xcite .
[ fig : sms ] indicates that @xmath63 , the initial state cross section contributing only to pp final states , contains the amplitudes quantitavely best reproduced by theory so far .
in @xcite it is shown that the energy dependence assumed for the ps and pp partial waves , i.e. the proportionality to @xmath64 and @xmath65 respectively , is valid for the measured energy range .
the magnitude of the ps cross section is calculated using equ .
7 , it increases from about 18% to 30% for bombarding energies in the range from 325 to 400 mev . at 310 mev
there are recent reports of a significant ps contribution to the total cross section@xcite and an sd - ss interference of a few percent@xcite , measuring unpolarized angular distributions .
we conclude that it should be of high priority to proceed with the study of the main inelastic channel in @xmath66 interactions , both experimentally by means of high statistics experiments that allow the extraction of angular distributions of spin - dependent observables , and theoretically by extending the many existing models concerning the ss contribution to the total cross section to also include higher partial waves .
we thank dr .
c. hanhart for making the theoretical predictions of the model of the jlich group available and one of the authors , p.t.e .
, thanks dr .
c. hanhart for useful discussions and comments .
this work has been supported by the us national science foundation under grants phy95 - 14566 , phy96 - 02872 , phy97 - 22556 , and by the us department of energy under grant doe - fg02 - 88er40438
. 9 h.o .
meyer et al . , submitted to phys .
, nucl - ex/9907017 h.o .
meyer , proc .
@xmath67 int . conf . on the structure of baryons , bonn , 22 - 26 sep ( 1998 ) h.o .
meyer et al . , phys .
81 ( 1998 ) 3096 swapan k. saha , w.w .
daehnick et al . , phys .
lett , in print , nucl - ex/9907016 p. thrngren engblom et al .
conf . mesons and light nuclei , prague - pruhonice , aug 31-sep 4 1998 , world scientific , in print , nucl - ex/9810013 t. rinckel et al . , submitted to nucl .
t. wise , a.d .
roberts and w. haeberli , nucl .
inst . meth .
a336 ( 1993 ) 410 h.o .
meyer et al . , nucl .
a 539 ( 1992 ) 633 r.c .
hanna , proc .
polarization phenomena , eds .
p. huber and h. schopper , birkhaeuser , basel ( 1966 ) 280 w. haeberli et al .
c55 ( 1997 ) 597 f. rathmann et al .
c58 ( 1998 ) 658 b. von przewoski et al . ,
c58 ( 1998 ) 1897 b. lorentz et al . , submitted to phys . rev . c. hanhart et al .
b 444 ( 1998 ) 25 , c. hanhart et al . , contribution to these proceedings j. zlomaczuk et al . , contribution to these proceedings j. zlomaczuk et al . , phys . lett . b436
( 1998 ) 251 ,
|
the pintexpintex/ ] facility at the iucf cooler ring , dedicated to the study of spin dependence in nucleon - nucleon interactions , has been used to measure polarization observables of the reaction @xmath0 at beam energies between 325 and 400 mev .
the stored , polarized proton beam had spin projections both in the longitudinal and the transverse directions with respect to the beam momentum .
we report here on the measurements of the relative transverse and longitudinal spin - dependent cross sections - [ \sigma ( \uparrow\uparrow + \sigma ( \downarrow\downarrow)]$ ] where the arrows denote parallel and antiparallel beam / target spin combinations , either transversely ( t ) or longitudinally ( l ) polarized .
] , @xmath1 and @xmath2 , and how from these observables the initial spin singlet and triplet cross sections are obtained .
considering angular momentum states less than or equal to one , the contribution of the ps partial waves to the cross section can be extracted .
thrngren engblom@xmath3 , h.o .
meyer@xmath3 , j.t .
balewski@xmath3 , w.w .
daehnick@xmath4 , j. doskow@xmath3 , + w. haeberli@xmath5 , b. lorentz@xmath5 , p.v .
pancella@xmath6 , r.e .
pollock@xmath3 , b. von przewoski@xmath3 , + f. rathmann@xmath7,t .
rinckel@xmath3 , swapan k. saha@xmath4 , b. schwartz@xmath5 , a. wellinghausen@xmath3 , t. wise@xmath5 @xmath3indiana university cyclotron facility , indiana university , bloomington , in 47405 , u.s.a .
+ @xmath4department of physics , university of pittsburgh , pittsburgh , pa 15260 , u.s.a .
+ @xmath5department of physics , university of wisconsin - madison , madison , wi 53706 , u.s.a .
+ @xmath6department of physics , western michigan university , kalamazoo , mi 49008 , u.s.a .
+ @xmath7friedrich - alexander universitt , erwin - rommel str . 1 , 91508 erlangen , germany .
|
the work of c.s.k . was supported in part by grant no .
2000 - 1 - 11100 - 003 - 1 and src program of the kosef , and in part by the krf grants , project no .
2000 - 015-dp0077 .
the work of s.o . was supported by the krf grants , project no . 2000 - 015-dp0077 and by the bk21 project .
a. b. carter and a. i. sanda , phys .
* 45 * , 952 ( 1980 ) ; phys .
d * 23 * , 1567 ( 1981 ) ; i. i. bigi and a. i. sanda , nucl .
b * 193 * , 85 ( 1981 ) .
m. gronau and d. london , phys .
lett . * 65 * , 3381 ( 1990 ) ; also see k. agashe and n. g. deshpande , phys .
d * 61 * , 071301 ( 2000 ) .
h. j. lipkin , y. nir , h. r. quinn and a. snyder , phys . rev .
d * 44 * , 1454 ( 1991 ) ; m. gronau , phys .
b * 265 * , 389 ( 1991 ) ; h. r. quinn and a. snyder , phys .
d * 48 * , 2139 ( 1993 ) .
a. s. dighe and c. s. kim , phys .
d * 62 * , 111302 ( 2000 ) .
m. gronau and d. london , phys .
b * 253 * , 483 ( 1991 ) ; m. gronau and d. wyler , _ ibid .
_ * 265 * , 172 ( 1991 ) .
i. dunietz , phys .
b * 270 * , 75 ( 1991 ) ; a. soffer , phys .
d * 60 * , 054032 ( 1999 ) ; b. kayser and d. london , phys .
d * 61 * , 116013 ( 2000 ) .
d. atwood , i. dunietz , and a. soni , phys .
lett . * 78 * , 3257 ( 1997 ) ; phys .
d * 63 * , 036005 ( 2001 ) .
m. gronau , phys .
d * 58 * , 037301 ( 1998 ) .
m. gronau and j. l. rosner , phys .
b * 439 * , 171 ( 1998 ) ; j .-
jang and p. ko , phys .
d * 58 * , 111302 ( 1998 ) .
r. aleksan , i. dunietz , b. kayser , and f. le diberder , nucl .
b * 361 * , 141 ( 1991 ) ; r. aleksan , i. dunietz , and b. kayser , z. phys .
c * 54 * , 653 ( 1992 ) ; b. kayser and d. london , phys .
d * 61 * , 116013 ( 2000 ) .
m. athanas _ et al .
_ , cleo collaboration , phys . rev .
lett . * 80 * , 5493 ( 1998 ) . c. c. meca and j. p. silva , phys .
lett . * 81 * , 1377 ( 1998 ) .
particle data group , d. e. groom _ et al _ , eur .
j. c * 15 * , 1 ( 2000 ) .
b. kayser and d. london , phys .
d * 61 * , 116013 ( 2000 ) .
y. nir and d. silverman , phys .
d * 42 * , 1477 ( 1990 ) ; g. c. branco , t. morozumi , p. a. parada , and m. n. rebelo , phys . rev .
d * 48 * , 1167 ( 1993 ) .
f. del aguila , j. a. aguilar - saavedra , and r. miquel , phys
. lett . * 82 * , 1628 ( 1999 ) ; y. liao and x. li , phys .
b * 503 * , 301 ( 2001 ) . c. s. kim and h. yamamoto , hep - ph/0004055 ,
talk given at the 3rd b&cp conference , taipei taiwan , dec .
1999 . z .-
xing , phys .
b * 371 * , 310 ( 1996 ) .
|
we propose a new method to extract the cp violating weak phase @xmath0 ( the phase of @xmath1 ) in the ckm paradigm of the standard model , using @xmath2 and @xmath3 decays , where @xmath4 are final states such as @xmath5 , @xmath6 , @xmath7 , _ etc .
_ we also study the experimental feasibility of our new method .
with possibility of new phases in the ckm matrix , we re - examine some of the previously proposed methods to determine @xmath0 , and find that it would be in principle possible to identify @xmath0 and a new phase angle @xmath8 separately .
epsf.tex ( # 1 width # 2)=#2 23.2 cm 17.0 cm -0.9 in -0.9 in we present a new method for determining @xmath0 , which is similar to the atwood - dunietz - soni ( ads ) method @xcite , but we use @xmath19 decays instead of @xmath18 decays used in the ads method .
the cleo collaboration have observed @xcite that the branching ratio for @xmath20 is much larger than that for @xmath21 , @xmath22 we consider the decay processes @xmath23 , @xmath24 and their cp - conjugate processes , where @xmath25 and @xmath26 decay into common final states @xmath27 , @xmath28 , @xmath7 , and so forth .
we note that the decay mode @xmath29 is severely suppressed relative to the mode @xmath20 , and this fact causes serious experimental difficulties in using @xmath30 decay for the gronau - london - wyler ( glw ) method @xcite .
however , in our method one needs not to perform the difficult task of measuring the branching ratio for @xmath30 , similar to the case of the ads method .
the detailed experimental feasibility for our new method and the ads method is given later .
note that the decay amplitudes of @xmath20 and @xmath31 contain the ckm factors @xmath32 and @xmath33 , respectively , while the amplitudes of @xmath29 and @xmath34 contain the ckm factors @xmath35 and @xmath36 , respectively .
we define the following quantities : @xmath37 @xmath38 \pi^- ) , \label{abcd}\end{aligned}\ ] ] where @xmath39 denotes the relevant decay amplitude and @xmath40 s are the relevant strong rescattering phases . here
@xmath41 $ ] in @xmath42 denotes that @xmath43 originates from a @xmath25 or @xmath44 decay .
similarly , we also define @xmath45 , @xmath46 , @xmath47 , @xmath48 and @xmath49 as the cp - conjugate decay amplitudes corresponding to @xmath50 , @xmath51 , @xmath52 , @xmath53 and @xmath42 , respectively , such as @xmath54 \pi^+)$ ] , _ etc .
_ note that @xmath55 with @xmath56 , but in general @xmath57 , as shown below .
then , the amplitude @xmath42 can be written as @xmath58 thus , @xmath59 and @xmath60 are given by @xmath61 where @xmath62 .
we see that @xmath57 , unless @xmath63 ( @xmath64 ) .
the expressions in eq .
( [ dd ] ) represent four equations for @xmath65 .
now let us assume that the quantities @xmath66 , @xmath67 , @xmath68 , @xmath69 and @xmath70 are measured by experiment , but @xmath71 is unknown .
then there are the four unknowns @xmath71 , @xmath0 , @xmath72 , @xmath73 in the above four equations . by solving the equations one can determine @xmath0 , as well as the other unknowns such as @xmath74 . in the ads method @xcite , @xmath50 , @xmath51 and @xmath42 in eq .
( [ abcd ] ) are replaced by @xmath75 k^- ) .
\label{ads}\end{aligned}\ ] ] then , @xmath59 and @xmath60 can be expressed by the same form as in eq .
( [ dd ] ) .
therefore , the phase @xmath0 can be determined by solving the four equations ( for @xmath76 ) with four unknowns @xmath71 , @xmath0 , @xmath72 , @xmath73 .
the impact on the ads method due to the large @xmath77 mixing from new physics has been studied in ref .
@xcite .
now we study the experimental feasibility of our new method and the ads method , by solving eq .
( [ dd ] ) analytically , @xmath78 to make a rough numerical estimate of the possible statistical error on determination of @xmath0 , we use the experimental result , eq .
( [ cleo ] ) , and the mean values for the ckm elements ; @xmath79 where we used @xmath80 , the color - suppression factor @xmath81 , @xmath82 , and @xmath83 is a wolfenstein parameter . in order to consider the decay parts , @xmath84 , we choose the modes such as @xmath85 , @xmath86 , @xmath87 which makes @xmath88 therefore , if we assume the 1 % level precision in the experimental determination for product of branching ratios , @xmath9 mesons @xcite .
thus , to obtain the data in 1 % level precision , one needs to have about @xmath89 @xmath9 s , which can be achieved at hadronic @xmath9 experiments such as btev and lhc - b , where more than @xmath90 @xmath9 mesons will be produced per year . ]
@xmath86 , @xmath91 = 1 \%$ ] , then we can set the numerical values , for @xmath92 , @xmath93 and for @xmath94 , @xmath95 then , we can make rough estimate for the statistical error from eq .
( [ cosine ] ) as @xmath96 & \sim & 0.1 ~ , \nonumber \\
\delta[\cos(\gamma \pm \delta_i)(b^\pm \to d(\to f_i)k^\pm ) ] & \sim & 0.05 ~. \label{delcosine}\end{aligned}\ ] ] we find the ads method can give approximately twice better precision statistically for determination of @xmath0 , compared to our new method . in fact , there is a general theorem @xcite : @xmath97 where @xmath98 is the number of @xmath9 mesons needed , @xmath99 the branching ratio of a decay mode , @xmath100 , and @xmath101 the relevant asymmetry . now , as shown in eq .
( [ cleo ] ) , @xmath99(ads method)/@xmath99(our method ) @xmath102 . to determine the relevant asymmetry @xmath101 , one has to calculate the following : @xmath103 for simplicity , we have considered the maximum asymmetry in both methods . using the experimental values given in eqs .
( [ cleo ] , [ bbbb ] @xmath104 [ acbck ] ) , we can easily get @xmath101(ads method)/@xmath101(our method ) @xmath105 .
therefore , @xmath98(ads method)/@xmath98(our method ) @xmath106 , which is exactly consistent with the above prediction , eq .
( [ delcosine ] ) , where we have predicted the possible precision with the same number of @xmath9 mesons .
we note that our new method may have other advantages : * the values of @xmath107\pi^\pm)$ ] are an order of magnitude bigger than @xmath108k^\pm)$ ] . therefore ,
if the present asymmetric @xmath9 factories of belle and babar can produce only a handful of such events because of the limited detector and trigger efficiencies , our new method may be the first measurable option .
* systematic errors could be much smaller for our new method due to the final state particle identification , @xmath109 fewer number of the final state pions due to @xmath110 , and the reconstruction of @xmath111 .
now we would like to make comments on new physics effects on determination of weak phase @xmath0 .
there can be two independent approaches to find out new physics beyond the sm , if it exists . *
the unitarity of ckm matrix can be assumed . in this case
, new physics effects can only come out from new virtual particles or through new interactions in penguin or box diagrams in @xmath9 meson decays .
if this is the case , all the methods which we mentioned above will extract the exactly same @xmath0 . *
the ckm matrix can be generalized to the non - unitary matrix . in this case
, new physics effects can appear even in tree diagram decays . and
the values of @xmath0 extracted from each method can be different .
therefore , we will describe in more detail for this second case .
in fact , in models beyond the sm , the ckm matrix may not be unitary ; for instance , in a model with an extra down quark singlet ( or more than one ) , or an extra up quark singlet , or both up and down quark singlets , the ckm matrix is no longer unitary @xcite .
if the unitarity constraint of the ckm matrix is removed , the generalized ckm matrix possesses 13 independent parameters ( after absorbing 5 phases to quark fields ) it consists of 9 real parameters and _ 4 independent phase angles_. the generalized ckm matrix can be parameterized as @xcite @xmath112 with the possibility of the non - unitary ckm matrix , one has to carefully examine the effects from the non - unitarity on the previously proposed methods where the unitarity of the ckm matrix was assumed to test the sm for cp violation .
from now on , we set @xmath113 and @xmath114 . in the parameterization given in eq .
( [ ckm ] ) , using our method , @xmath53 in eq .
( [ abcd ] ) is replaced by @xmath115 this leads to the result that the phase @xmath0 in the expressions for @xmath59 and @xmath60 in eq .
( [ dd ] ) should be replaced by @xmath116 .
therefore , in this case , our method can measure the non - unitary phase @xmath117 . in the ads method ,
besides @xmath53 is changed into the one in eq .
( [ ci ] ) , the phase @xmath0 in @xmath51 is also replaced by @xmath118 . as a result , the new phase @xmath8 is automatically cancelled to disappear in the expressions for @xmath59 and @xmath60 .
thus , the ads method would still measure @xmath0 that _ is the phase of _ @xmath1 .
the glw method @xcite was suggested for extracting @xmath0 from measurements of the branching ratios of decays @xmath119 , @xmath120 and @xmath121 , where @xmath122 is a cp eigenstate .
however , the glw method suffers from serious experimental difficulties , mainly because the process @xmath123 ( and its cp conjugate process @xmath124 ) is difficult to measure in experiment .
that is , the rate for the ckm and color suppressed process @xmath123 is suppressed by about two orders of magnitudes relative to that for the ckm and color allowed process @xmath21 , and it causes experimental difficulties in identifying @xmath44 through @xmath125 since doubly cabibbo suppressed @xmath126 following @xmath21 strongly interferes with @xmath127 following the rare process @xmath128 . with the non - unitary ckm matrix , this method would measure the angle @xmath129 ( see figure 1 ) , instead of @xmath0 as originally proposed in ref .
@xcite . in ref .
@xcite two groups , gronau and rosner ( gr ) , jang and ko ( jk ) , proposed a method to extract @xmath0 by exploiting cabibbo allowed decays @xmath130 and using the isospin relations . in the gr / jk method ,
the decay modes @xmath131 with the quark process @xmath132 contain the ckm factor @xmath133 and their amplitudes can be written as @xmath134 where @xmath135 and @xmath136 denote the amplitude and the strong re - scattering phase for the isospin @xmath137 state .
note that the weak phase angle @xmath118 appears in eq .
( [ aaa ] ) rather than @xmath0 as in ref .
@xcite . in this method ,
three triangles are drawn to extract @xmath138 , using the isospin relation @xmath139 and the following relations @xmath140 where @xmath141 is a cp eigenstate of @xmath142 meson , defined by @xmath143 .
the appearance of @xmath144 in eq .
( [ aaa ] ) results in extraction of @xmath145 , by this method , rather than @xmath138 as in the above reference , as shown in fig
. 1 . in fig
. 1 , @xmath146 , @xmath147 and a thick solid line form a triangle , and @xmath148 , @xmath149 and the other thick solid line form another triangle . these two triangles are those used in the glw method and the thick solid lines correspond to @xmath150 and @xmath151 , respectively , which are very difficult to measure in experiment .
obviously , assuming that the glw method works , the method would extract @xmath145 rather than @xmath138 .
we note that the impact on the glw method from the sizable @xmath77 mixing via new physics effects has been also investigated in ref .
@xcite . in conclusion ,
we have presented a new method to determine @xmath152 in the ckm paradigm of the sm , using @xmath153 and @xmath154 decays , where @xmath27 , @xmath28 , @xmath7 , _ etc . _ the experimental feasibility of our method , comparing with the ads method , has been studied .
although experimentally challenging at present asymmetric @xmath9 factories of belle and babar , the analysis using our method can be carried out in details at hadronic @xmath9 experiments such as btev and lhc - b , where more than @xmath90 @xmath9 mesons will be produced per year . with possibility of new phases in the ckm matrix , we have re - examined some of the previously proposed methods for determining the weak phase @xmath0 using @xmath155 or @xmath19 decays .
we have shown that our method would extract @xmath156 with the new phase @xmath8 .
the ads method would measure @xmath0 , while the glw method or the gr / jk method would measure @xmath118 .
thus , if one uses the above methods independently and compares the results , it would be in principle possible to identify @xmath0 and @xmath8 separately .
if this is the case and @xmath8 is not negligible , this would be a clear indication of the new phase in the ckm matrix , _
i.e. _ an effect from new physics .
we note that , in fact , inconsistencies between the values of @xmath0 can arise from final state interactions ( fsis ) which are known to be important in @xmath111 decays and , consequently , may modify quark - level description of @xmath9 decays .
therefore , there is a possibility that fsis may cause a problem in identifying the new phase @xmath8 by using various methods mentioned above .
|
the stability of differentially rotating gas disks depends sensitively upon whether or not a magnetic field is present .
a weak ( subthermal ) magnetic field undermines the stabilizing influence of coriolis forces ( balbus & hawley 1991 ) , and the resulting turbulence greatly enhances internal angular momentum transport ( hawley , gammie , & balbus 1995 ; balbus & hawley 1998 ) . this important behavior , which we refer to here as the standard magnetorotational instability ( mri ) , probably is the underlying cause of `` anomalous viscosity '' in accretion disks .
while other modes of transport ( e.g. waves ) are certainly possible , the classical enhanced _
turbulent _ transport associated with @xmath0 disks is almost surely mhd turbulence ( balbus & papaloizou 1999 ) .
protostellar disks remain problematic .
they seem too dense and too cool to be magnetically well - coupled over their full radial extent .
understanding how protostellar and protoplanetary disks behave is an ongoing theoretical challenge .
it has also been recently noted ( gammie & menou 1998 , menou 2000 ) that the outer regions of dwarf novae disks may have a very low ionization fractions . that protostellar disks and dwarf novae both show episodic eruptions may be no coincidence ; in both systems the question of magnetic coupling ( and therefore the presence of mhd turbulence ) is a delicate one .
this paper addresses a very small but fundamental part of this problem : the linear effects of hall electromotive forces ( hemfs ) on disk dynamics . while the results are more broadly applicable , protostellar disks will be the focus of this paper .
the importance of the hall effect in modifying the mri was pointed out in a recent paper by wardle ( 1999 ; hereafter w99 ) , who analyzed the linear stability of a protostellar disk in the presence of a vertical magnetic field .
such disks are generally in the hall regime .
that is , their combination of densities , temperatures and ionization fractions places them in a realm of parameter space in which departures from ideal mhd are important , which are strongly influenced by hemfs .
the hall effect tends to be important simultaneously with finite conductivity .
numerical simulations of low ionization disks ( e.g. , fleming , stone , & hawley 2000 ) have not yet included hall terms , but have focused rather exclusively on finite conductivity .
w99 found that standard , locally unstable mri modes may either be stabilized or destabilized depending upon whether @xmath1 is positive or negative .
( here , as usual , @xmath2 is the local angular velocity vector of the disk , and @xmath3 is the magnetic field . ) in other words , hemfs appear to introduce a helicity effect into the problem .
this immediately raises an interesting question : what happens if @xmath4 ? would the hall effect disappear ?
since the w99 analysis was restricted to axial fields and axial wavenumbers , it could not address this problem .
clearly hemfs do not simply vanish when the magnetic field lies in the disk plane , and the identification of @xmath5 as the fundamental hall stability parameter therefore seems suspicious .
identifying this key parameter more precisely is important , for it is at the heart of mhd stability in low ionization disks .
we are motivated , therefore , to examine the stability of hall regime accretion under very general circumstances .
we restrict neither the local field geometry , nor the local wavenumber direction .
furthermore , we have chosen to couch the analysis in a dynamical context , instead of ascribing the inductive hall terms to effective conductivities ( w99 ) .
the effective conductivity approach offers terseness and mathematical economy ; the dynamical approach makes the physics of the coupling between the magnetic field and the disk coupling more transparent .
both methods must of course ultimately be equivalent .
it is enlightening , however , to view this important problem from a different physical perspective , and we have accordingly chosen this route .
the plan of the paper is as follows . in 2 , we discuss the basic formulation of the problem .
this includes the various physical and compositional parameters , the fundamental equations , and an order of magnitude estimate for the relative sizes of the ohmic , hall , and ambipolar diffusion terms in the induction equation .
section 3 presents the linear dispersion relation for a uniformly rotating disk , because this is the simplest setting to see the interplay between rotation and the hall emfs .
section 4 is a detailed and extensive analysis of differentially rotating disks for arbitrary field geometries and perturbation wavenumbers .
section 5 summarizes the findings of this paper .
a typical protostellar disk consists of molecular gas with cosmic abundances , and spans a temperature range from 10 k in its outer regions to a few @xmath6 k near the central star
. the characteristic disk size may be up to of order 100 au , and its total mass will typically be less than roughly 0.1 of the central star s mass .
if 0.01 @xmath7 is spread over a cylinder of radius 1 au and height 0.1 au , this would yield an average mass density @xmath8 g @xmath9 , and an integrated column density of @xmath10 g @xmath11 , fiducial numbers one should bear in mind .
the dynamics of the disk is likely to be dominated by the keplerian orbital velocity @xmath12 : v_k^2 = gmrr^2 ^2 where @xmath13 is the central mass , @xmath14 the cylindrical radius , and @xmath15 the angular velocity .
the disk is assumed to be thin ; the characteristic temperature @xmath16 is much less than any virial temperature .
the scale height @xmath17 ( defined below ) satisfies @xmath18 .
the number density of species @xmath19 is denoted @xmath20 .
the h@xmath21 and he number densities are assumed to be related by n_he = 0.2 n_h_2 , which implies a total neutral number density n = n_h_2 + n_he = 1.2 n_h_2 , and a neutral mass density = 2.8 n_h_2 m_p , where @xmath22 is the proton mass .
this gives a mean mass per particle of = /n = 2.33 m_p , and isothermal and adiabatic sound speeds respectively of c_s^2 = 0.429 k t / m_p , a^2 = 0.6 kt / m_p , where @xmath23 is the boltzmann constant .
for simplicity we will define the disk scale height @xmath17 by @xmath24 . under conditions of interest
, the dominant ion will generally be once - ionized potassium k@xmath25 , and accordingly we take a mean ion mass of m_i = 39m_p in molecular cloud studies , @xmath26 is more typically used ( e.g. , draine , roberge , & dalgarno 1983 ) .
the coupling between ions and neutrals depends upon the rate coefficient , denoted @xmath27 .
in essence , this is the the product of a cross section with the component of the relative velocity along the drift axis , averaged over the electron maxwellian distribution ( spizter 1978 ) .
similarly , there is an electron - neutral coupling rate , @xmath28 .
we follow draine et al .
( 1983 ) and blaes & balbus ( 1994 ) , adopting the values v_in = 1.9 10 ^ -9 cm^3 s^-1 v_en = 10 ^ -15 ( 128kt9m_e ) ^1/2= 8.2810 ^ -10 t^1/2 cm^3 s^-1 the force per unit volume on the neutral fluid due to ion drag is given by an expression of the form ( shu 1992 ) = - _ i ( - ) , where @xmath29 is the ion mass density , @xmath30 the neutral velocity , @xmath31 the ion velocity , and @xmath32 is the so - called drag coefficient , = v_inm_i + = 2.75 10 ^ 13 cm^3 s^-1 g^-1 .
our numerical value differs somewhat from draine et al .
( 1983 ) because we use the potassium value for the ion mass @xmath33 . electron - neutral coupling results in a finite electrical conductivity for the disk gas ( krall & trivelpiece 1973 ) : _ e = n_e e^2m_e _ en = n_e e^2m_e n v_en , where @xmath34 is the magnitude of the electron charge , @xmath35 is the electron number density , and @xmath36 is the electron - neutral collision frequency .
the associated resistivity is = c^24 _ e = 234 ( nn_e ) t^1/2 cm^2 s^-1 , where @xmath37 is the speed of light .
the dimensionless measure of the relative importance of resistivity is the magnetic reynolds number , re_m = v_a h , where @xmath38 is the magnitude of the alfvn velocity = , and @xmath3 is the magnetic field vector .
finally , the cyclotron frequencies of the electrons , ions , and neutrals are denoted _ c = ebm_c where @xmath0 is respectively @xmath39 , @xmath40 , and @xmath41 ( @xmath42 ) .
although the neutrals are of course not directly affected by the magnetic field , the neutral cyclotron frequency is an important characteristic frequency of the analysis .
the fundamental dynamical equations are mass conservation , @xmath43 the dynamical equation of motion , [ mom ] + ( ) = - ( p + b^28 ) - + ( 4 ) and the induction equation , = ( - ) . here
, @xmath30 is the velocity of the neutrals , and @xmath44 is the electron velocity , and @xmath45 is the central gravitational potential . in equation ( [ mom ] ) , which is dominated by the neutral component , we have made implicit use of the low inertia limit for the ions .
ion momentum balance is set by balancing the ion - neutral drag force with the lorentz force , [ i_mom ] 1c = , since all other terms in the ion momentum equation are small .
it is actually @xmath46 that appears on the right hand side of equation ( [ mom ] ) ; the low ion inertia approximation consists of instantaneously balancing this force against the lorentz force .
note the appearance of the electron velocity @xmath44 in the induction equation . in ideal mhd ,
the distinction between the fluid velocity and the electron velocity is unimportant .
but in fact , in the absence of resistance , it is the electron fluid in which the magnetic field is `` frozen , '' since the electrons are the mobile charge carriers .
the difference between ion and neutral velocities is responsible for ambipolar diffusion , and the difference between ion and electron velocities gives rise to the hall effect ( e.g. , wardle & knigl 1993 ) .
in general , [ ve ] = + ( - ) + ( - ) = - n_e e+ _ i c , we are assuming that any ions present are singly ionized ( which is appropriate for a very low ionization gas ) , so that the electron and ion number densities are equal .
in the ideal mhd limit , both @xmath47 and @xmath48 are assumed to be negligibly small .
retaining them in the full induction equation leads to [ ind ] t = ( - 4 c - n_e e + c _ i ) , where we have used the standard mhd approximation = c4 .
the four terms on the right side of the induction equation can be arranged by order of magnitude .
reading from left to right , we denote them @xmath49 ( inductive ) , @xmath50 ( ohmic ) , @xmath51 ( hall ) , and @xmath52 ( ambipolar ) .
assume that typical fluid velocities are of order @xmath38 , and that typical gradients are of order an inverse scale - height @xmath53 .
then , the relative sizes of these terms are ~1re_m , ~_ce _ en , ~_ci the first ratio just reflects the familiar fact that when the magnetic reynolds number approaches unity , ohmic losses are important . the results of 2.1 give ~1re_m = 4.210 ^ -14 ( nn_e ) ( c_sv_a ) ( 1 r)^3/2 ( 10 ^ 3 t ) ^1/2 ( mm_)^1/2 , and ~_ce _ en ( 810 ^ 17n)^1/2 ( v_ac_s ) , ~_ci ( 910 ^ 12n ) ^1/2 ( t10 ^ 3)^1/2 ( v_ac_s ) . under interstellar conditions
@xmath54 is tiny , even though @xmath55 may be large .
( note that in equation [ 13 ] , @xmath17 would be a characteristic interstellar length , not a disk thickness . ) in this case , neither the @xmath50 nor the @xmath51 term is important .
but the ambipolar diffusion @xmath52 term may be non - negligible .
this leads to a regime of ideal mhd in the ions , which then couple the magnetic field to the neutrals via ion - neutral collisions .
in protostellar disks , our interest here , the far smaller ionization fraction can reduce @xmath56 to the point where resistivity can affect the dynamics .
this reynolds number need not be of order unity in an mhd - turbulent fluid before it is significant ; some field configurations are affected when @xmath56 is larger than @xmath57 ( fleming , stone , & hawley 2000 ) . the key point here , however , is not the well - known result that resistivity is important in protostellar disks , it is the less appreciated fact that hemfs will generally be important in a magnetized disk whenever resistivity is .
the two should be treated on the same footing ( w99 ) .
the importance of the hall term relative to ohmic losses is noteworthy because it is surprisingly general . as formulated above , for a given ratio of magnetic to thermal energy , the relative importance of the hall term
is _ independent _ of the ionization level , and depends only upon the number density of the dominant neutrals .
furthermore , the relative importance of the hall term is not a feature unique to gases of low ionization fraction . using a nominal resistivity value of @xmath58 @xmath59 s@xmath60 for a fully ionized plasma ( spitzer 1962 )
, we find ~_ce_ei 4 ( v_ac_s ) ( t10 ^ 5)^2 ( 10 ^ 16n)^1/2 , where @xmath61 is the electron - ion collision rate .
we see that if ohmic dissipation is of interest ( as it would be for magnetic reconnection ) , the temperature and density regimes of ionized accretion disks imply that the hall effect can not be ignored . at typical disk protostellar disk densities ,
ambipolar diffusion can be neglected , though it may become important in the outermost regions of the disk . since = - b^2 ,
if the currents and fields are perpendicular , ambipolar diffusion acts like a field - dependent resistivity .
this is the case for the special vertical field geometry treated by w99 and in 4.2 below : all of ambipolar diffusion may be incorporated into an effective resistivity .
but for the more general field geometries we consider , no such simplification is possible .
we shall accordingly assume a neutral number density much in excess of @xmath62 @xmath9 , and ignore ambipolar diffusion .
the form of the induction equation we shall use henceforth is [ finind ] t = ( - - c 4n_e e ) .
hall currents introduce novel elements into the mhd of astrophysical disks , and to develop an intuitive understanding , we consider a very simple problem : hall - modified alfvn waves in a uniformly rotating disk threaded by a vertical magnetic field , @xmath63 . here
we shall ignore finite resistivity and vertical structure , work in the boussinesq limit , and use standard cylindrical coordinates @xmath64 with the origin at the disk center . finally , we consider plane wave perturbations that depend only upon @xmath65 . linearized quantities ( indicated by @xmath66 notation ) are proportional to @xmath67 , where @xmath68 is the angular frequency and @xmath23 is the vertical wave number
. under these circumstances , pressure , density , vertical velocity and vertical magnetic field perturbations all vanish .
we solve for @xmath69 , @xmath70 , @xmath71 and @xmath72 .
the linearized radial and azimuthal equations of motion are [ aaa ] -i v_r - 2 v_- ikb4 b_r = 0 , [ bbb ] -i v_+ 2 v_r - ikb4 b_=0 .
the same components of the linearized induction equation are -ib_r + k^2 bc4e n_e b_- ikb v_r = 0 , -i b_- k^2 bc4e n_e b_r - ikb v_= 0 , note the symmetry of the problem , in particular the introduction of a `` coriolis term '' into the magnetic field equations .
these , of course , are the electromotive forces of the hall effect .
the combination @xmath73 will be recognized by students of plasma physics as the electron whistler frequency . at characteristic wavenumbers satisfying @xmath74
, this frequency is @xmath75 .
hence , the ionization fraction below which the whistler branch couples with the disk dynamics is @xmath76 .
the `` whistler drift '' of the field lines with respect to the bulk of the fluid is caused by the motion of the current - bearing electron charge carriers .
perturbing the magnetic field induces local currents , and it is the electron motion in the currents relative to the ions that causes the field lines to move relative to the ion - neutral fluid .
( recall that ambipolar diffusion is ignored , so the ions are locked with the neutrals . ) note the sense of the field line drift : if @xmath2 and @xmath3 are both upward ( say ) , the induced whistler circular motion of the field lines is right - handed , following the electrons , _ opposite _ to motion induced by the dynamical coriolis force . how does all this affect the wave response of the fluid ? consider first the case of aligned @xmath3 and @xmath2 .
the sense of the field line motion is to slow the dynamical epicycles . the easiest way to see
this is to note that in the absence of rotation , equations ( [ aaa ] ) and ( [ bbb ] ) show that the perturbed velocity vector is proportional to the perturbed magnetic field vector .
the whistler waves would impart a circularly polarized component to the velocity response .
this response is counter to the direction that the dynamical epicyclic motion is trying to impart , and hence it effectively lowers the coriolis force . as for the magnetic tension , the additional hall response is @xmath77 out - of - phase with respect to this force , and we therefore expect these two to add in quadrature . if , in addition , the dynamical epicycle is slowed by hall currents , the return magnetic tension force should effectively increase .
if the field and angular velocity are counter aligned , the signs of these effects should switch .
these expectations are born out in the dispersion relation , which is most transparently written [ disp1 ] ^2 2(1-k^2v^2_h/4 ^ 2 ) - k^2(v_a^2 + v_h^2 ) = 0 , where we have introduced the _ hall velocity _ defined by v_h^2 . note that
although this quantity has dimensions of ( velocity)@xmath78 , _ it may be positive or negative .
_ it is positive ( negative ) if @xmath2 and @xmath3 are oriented in the same ( opposite ) sense .
we address more general orientations in subsequent sections . in the dispersion relation ( [ disp1 ] ) , the upper @xmath79 sign corresponds to left - handed polarization relative to the magnetic field , the lower @xmath80 sign to right - handed polarization .
whistler waves exist only for the right - handed branch , and correspond to the limit @xmath81 .
there are no instabilities in a uniformly rotating disk , since there is no free energy source .
we postpone until later sections ( which include differential rotation ) a full discussion of the solutions to the dispersion relation , but this is a convenient spot to make contact with waves in a magnetized plasma . in the limit @xmath82 , the dispersion relation becomes ^2 + k^2 v_a^2 = 0 , this is precisely the dispersion relation for a uniformly magnetized plasma with the displacement current and electron inertia ignored ( e.g. , krall & trivelpiece 1973 ) .
( the alfvn speed in the above is dominated by the neutrals , whereas in standard fully ionized plasma treatments it is determined by the ions , of course . ) without loss of generality , we may take @xmath83 , allowing @xmath84 to determine the propagation direction .
the positive frequency solutions are = + ( k^4 b^2 c^264 ^ 2 n_e^2 e^2 + k^2 v_a^2 ) ^1/2 . for small wavenumbers ( low frequencies ) both of the above solutions reduce to alfvn waves . at large wavenumbers , right - handed waves ( @xmath79 sign )
go over to the high frequency whistler wave branch , whereas large @xmath23 left - handed waves ( @xmath80 sign ) are cut off at a frequency given by = ebc ( n_en ) = _ c ( n_en ) , analogous to the ion - cyclotron cut - off in a fully ionized plasma .
the characteristic ionization fraction at which the cut - off frequency drops below the disk rotation frequency is once again @xmath85 .
we consider next the local stability of a differentially rotating disk threaded by a weak vertical field .
we assume that finite resistivity and hall currents are both present ; we neglect ambipolar diffusion . as in the previous section ,
we restrict ourselves to plane wave disturbances of the form @xmath86 . in the boussinesq limit
, this corresponds to fluid displacements in the plane of the disk , so vertical structure is unimportant .
this problem has been considered by w99 , but it bears re - examination for several reasons .
it is an important baseline for understanding the effects of hall currents in protostellar disks , and is worth understanding from more than one perspective .
our approach emphasizes dynamical couplings rather than conductivity tensor formalism , and is physically quite distinct .
it is also somewhat simpler .
finally , we shall examine a qualitatively new feature of the hall effect that has not been discussed before : hemfs can destabilize _ outwardly _ increasing differential rotation . in the process
we show that the maximum growth rate of the instability is always the local oort a value of the rotation ( in magnitude ) , whether the angular velocity is increasing or decreasing outward . the key issue of the analysis is the local stability criterion of a disk with an axial field . whereas the dispersion relation is somewhat complex , the limit @xmath87 is simple , and this is relevant for understanding local stability .
( in principle , local overstability may occur , but for self - consistency we wish to focus on nonpropagating , evanescent , local modes .
for these , the transition between stability and instability must proceed through the point @xmath88 . ) including the effects of differential rotation and finite resistivity along with the hall terms , we find that the linearized , @xmath88 , radial and azimuthal equations of motion are now [ ad ] - 2 v_- ikb4 b_r = 0 , [ ac ] ^2 2 v_r - ikb4 b_=0 , where @xmath89 is the epicyclic frequency , defined by ^2 = 4 ^ 2 + d^2dr .
the same components of the linearized induction equation are [ ab ] k^2b_r + k^2 bc4e n_e b_- ikb v_r = 0 , [ aa ] k^2 b_- ( k^2 bc4e n_e + ddr ) b_r - ikb v_= 0 , notice that in equation ( [ aa ] ) , the whistler frequency hall term introduces a coupling that is equivalent to changing the shear rate @xmath90 .
it arises becauses the @xmath91 component of the perturbed electron velocity ( @xmath92 ) differs from the dominant neutral velocity by a term involving @xmath93 .
this produces a radial field component , @xmath71 .
the effect is present whether or not there is differential rotation , but it is most striking when such motion is present , because it couples exactly like @xmath90 . in effect , it is the magnetic `` epicyclic '' term , combining a coriolis - like coupling with a background angular velocity gradient .
if we ignore resistivity for the moment , our problem decouples into two very simple subproblems , one with @xmath70 coupled only to @xmath71 , and one with @xmath94 coupled only to @xmath95 .
the torque must vanish , so equations ( [ ac ] ) and ( [ ab ] ) combine to give ( k^2 v_a^2 + ^2 k^2 v_h^24 ^ 2 ) b_= 0 , or the tidal and excess centrifugal forces must vanish , in which case equations ( [ ad ] ) and ( [ aa ] ) yield ( d^2dr + k^2 ( v_a^2 + v_h^2 ) ) b_r = 0 . if we are not to have a trivial solution , then ( v_a^2 + ^2 v_h^24 ^ 2 ) ( d^2dr + k^2 ( v_a^2 + v_h^2 ) ) = 0 . in the standard mri ,
the torque is always purely alfvnic , and therefore it always has the same sign : a restoring negative torque for a positive angular displacement .
this means that instability depends only upon the direction of the excess centrifugal force .
the hall effect renders the problem more interesting , allowing an interplay between the torque and centrifugal force .
inwardly _ directed centrifugal force can destabilize if accompanied by a positively directed torque .
the inclusion of resistivity couples @xmath71 and @xmath72 , but gives only a slight modification to the above result : [ stab1 ] ( v_a^2 + ^2 v_h^24 ^ 2 ) ( d^2dr + k^2 ( v_a^2 + v_h^2 ) ) + ^2 ^2 k^2 = 0 .
this is the desired critical stability condition . to determine the sign for instability
, one may take the limit @xmath96 , which returns us to the simple mri .
it then easily follows that the left hand side should be negative for instability , that is [ inequ ] k^2 v_a^2 < - where x is the dimensionless _ hall parameter . _ the terms in square brackets
incorporate the additional physics of hemfs and resistivity ; they become unity for the standard @xmath97 mri ( balbus & hawley 1991 ) . the dimensionless resistivity parameter may be expressed in terms of the magnetic reynolds number : = ( ) ( c_sv_a ) re_m^-1 .
the behavior of the polynomial d(x ) = ( 1+x ) ( 1 + ^2 x4
^ 2 ) + ^2 ^ 2v_a^4 determines the nature of the instability . we may write the instability criterion [ fifty ] sgn(d ) k^2 v_a^2 < - ( 1 + ^2 x/4 ^ 2 ) d^2/dr |d| .
it is not difficult to show that if [ window ] re_m^-1 < 14 ( ^2 ^ 2 ) ( v_ac_s ) | d^2dr| there will be a finite range of @xmath98 for which @xmath99 . in this window ,
all wavenumbers , no matter how large , are unstable , even in the presence of resistivity : the right hand side of the inequality ( [ inequ ] ) is positive , while the left hand side is negative . clearly , the zeroes of @xmath100 are critical for understanding the behavior of disks in the hall regime . for a keplerian disk ,
the window occurs when @xmath98 falls approximately between @xmath101 and @xmath102 for large @xmath56 .
roughly speaking , there is a match between the response frequency of the magnetic tension and the `` epicyclic '' frequency of the field line drift through the fluid caused by hemfs .
these tend to cancel each other s dynamical response .
qualitatively , it is as though the restoring tension were not present as a radial force , leaving the destabilizing dynamical tide to do its work . within the window , going to larger wavenumbers does not change the sign of the radial forces on a fluid element .
the azimuthal torque , however , is still dominated by the alfvn term , and this allows for angular momentum transfer . in figure ( [ wfig1 ] ) , we show a stability plot for an @xmath103 keplerian disk in the @xmath104 plane .
the shaded regions correspond to instability .
the instability window ( i.e. , the region in which all wavenumbers are unstable ) for @xmath105 is evident , together with an abrupt transition for @xmath106 . for @xmath107
the hall currents destabilize by allowing for a larger range of destabilizing wavenumbers than would be present in the magnetorotational instability . for @xmath108 ,
hall currents are a stabilizing influence .
figure ( [ wfig2 ] ) illustrates the modified instability range for large @xmath56 .
the window is narrowed as shown .
figure ( [ wfig3 ] ) illustrates the effect of increased resistivity on stability .
the window has now disappeared , and as @xmath56 diminishes , an ever decreasing portion of wavenumber space is unstable .
this region corresponds to very large wavelengths , which may exceed the global disk scales .
this corresponds to resistive stabilization . in general
, when resistivity becomes dominant , the instability criterion ( [ fifty ] ) leads to @xmath109 .
but the the right hand side of the inequality is a function of @xmath98 , and at its maximum we find @xmath110 .
this is considerably easier to satisfy , though it corresponds to tuning the hall parameter .
this effect was also noted in w99 .
figure ( [ wfig4 ] ) is presented to illustrate a point of principle .
it is the stability diagram of an @xmath111 , @xmath112 ( i.e. , @xmath113 ) disk .
there is an instability window , analogous to its keplerian counterpart , for @xmath114 , abrupt stability for @xmath115 , and instability for an ever diminishing wavenumber domain as @xmath98 falls below @xmath101 . in the presence of sufficiently large hall currents , with counteraligned axial angular velocity and magnetic field vectors ,
the instability criterion is @xmath116 .
any differential rotation law is potentially unstable .
we next obtain the full dispersion relation .
assume that the time dependence of the perturbations takes the form @xmath117 .
( this form keeps the coefficients of the dispersion relation real . )
the linearized radial and azimuthal equations of motion are now v_r - 2 v_- ikb4 b_r = 0 , v_+^2 2 v_r - ikb4 b_=0 .
the same components of the linearized induction equation are ( + k^2 ) b_r + k^2 bc4e n_e b_- ikb v_r = 0 , ( + k^2 ) b_- ( k^2 bc4e n_e + ddr ) b_r - ikb v_= 0 , the resulting dispersion relation is [ basdis ] 0=^4 + 2 k^2 ^ 3 + c_2 ^2 + 2 k^2 ( ^2 + k^2 v_a^2 ) + c_0 where the constants @xmath118 and @xmath119 are given by _ 2 = , and_0 = k^2 + ^2 ^2 k^4 .
balbus & hawley ( 1992a ) conjectured that the maximum growth rate of any instability feeding off the differential rotation in a disk is given by the local oort a value , @xmath120 .
all magnetic field configurations of the magnetorotational instability have this value for their most rapid growth rate .
however , the a value conjecture went beyond alfvn tension as the destabilizing agent , it suggested that whatever the proximate cause , linear perturbations can grow no faster than @xmath121 .
the reasons are rooted in the dynamics of the differential rotation process itself , not in magnetism _
_ now , it is far from obvious that in the absence of resistivity , the rather unwieldy dispersion relation ( [ basdis ] ) has precisely this value as its maximum growth rate .
but it does , as we now demonstrate . to begin , write equation ( [ basdis ] ) in dimensionless form , with all rates normalized to @xmath15 .
we define s = / , = / , x=(kv_a/)^2 , y= ( kv_h/)^2 . note that [ help ] ^2 = 4 + d^2dr .
the @xmath122 dispersion relation is [ dissimp ] s^4 + s^2 + ( x + y^24 ) ( d^2dr + x+y ) = 0 . at the maximum growth rate @xmath123 , partial differentiation of the above with respect to @xmath19 and @xmath124
gives the two equations [ parx ] x + y8(^2 + 4 ) = -s_m^2 - 12 d^2dr [ pary ] x4(^2 + 4 ) + y2
( ^2 + s_m^2 ) = -14d^2dr(^2 + s_m^2 ) . using the identity ( [ help ] ) and eliminating @xmath124 between equations ( [ parx ] ) and ( [ pary ] ) , leads after simplification to a remarkable result , ( x + s_m^2 + ^2)= 0 . since the first factor must be positive definite ,
we conclude that s_m^2 = 116 ( d^2d r)^2 .
this is equivalent to s_m = 12 | ddr | , i.e. , the oort a value !
we must now verify that this value of @xmath125 is a solution of the dispersion relation ( [ dissimp ] ) for @xmath19 and @xmath124 satisfying the system of equations ( [ parx ] ) and ( [ pary ] ) .
it is easily shown that with the oort a value chosen for @xmath126 , the determinant of this system vanishes .
we thus use only one of these equations , which we take to be ( [ parx ] ) .
first we substitute for @xmath125 , and solve for @xmath19 in terms of @xmath124 .
this leads to x= - 12 d^2dr - y8 ( ^2 + 4 ) -116 ( d^2dr)^2 .
note that in the @xmath127 limit of this equation , @xmath128 is possible only for decreasing outward angular velocity profiles .
the magnetorotational instability requires @xmath129 .
the addition of the hall effect allows for the possibility that @xmath130 , and thus for a perfectly well - defined positive @xmath19 parameter , even if @xmath131 . that both senses of angular velocity gradient may be treated on the same footing , accords with the well - known result that the state of minimum energy for a disk of fixed angular momentum is solid body rotation ( lynden - bell & pringle 1974 ) . in principle
_ any _ angular velocity gradient could be unstable , since lower energy states consistent with angular momentum conservation exist . a dynamical path to these low energy states
is required however , and that is what hemfs can provide .
replacing @xmath19 by the above expression wherever it occurs in equation ( [ dissimp ] ) yields , after some simplification , s^4 + c_2 s^2 + c_0 = 0 , with c_2 = 14 ( y-4)^2 - 18 ( d^2dr)^2 , and c_0 = 164 ( d^2dr)^2 the standard solution is s^2 = 18 choosing the @xmath79 sign returns the oort a solution , exactly as desired and consistent with the balbus & hawley ( 1992a ) conjecture .
( w99 obtained the keplerian upper limit by numerical solution of the dispersion relation , and noted that it was the same as the standard magnetorotational instability . )
graphical solutions in the @xmath132 plane are shown in figures ( [ wfig5 ] ) and ( [ wfig6 ] ) .
we now consider the axisymmetric behavior of the instability with more general field geometries and wavenumbers .
we shall ignore buoyancy , so our analysis holds either for a strictly polytropic disk , or locally at the midplane for any constituative relation .
the perturbation wavevector is = k_r + k_z , and disturbances have space - time dependence @xmath133 .
the linearized equations are now that of mass conservation ( boussinesq limit ) : k_r v_r + k_z v_z = 0 , and the equations of motion , v_r - 2 v_- ik_z b_z4 b_r + ik_r4 ( b _ b_+b_zb_z)+ ik_rp = 0 , v_+^22 v_r - i 4 b_=0 , v_z - i k_r b_r 4 b_z + ik_z4 ( b _ b_+b_rb_r ) + i k_z p = 0 . if one considers a polytropic disk , then @xmath134 may be replaced directly by a pure enthalpy perturbation , @xmath135 , say .
otherwise it is to be taken at face value , a pressure perturbation divided by the density @xmath136 .
ultimately , either choice leads to the same result , because the term serves only as a place holder .
the linearized induction equations are ( + k^2 ) b_r + c k_z 4e n_e b_- i v_r = 0 , ( + k^2 ) b_- ( ddr + c k_z 4e n_e ) b_r + c k_r 4e n_e b_z - i v_= 0 , ( + k^2 ) b_z - c k_r 4e n_e b_- i v_z = 0 . the dispersion relation that emerges after a straightforward but lengthy effort is [ gendisp ] ^4 + 2 k^2 ^3 + c_2 ^2 + 2k^2 ( k_z^2k^2 ^2 + ^2 ) + c_0 = 0 , with [ gendispp ] c_2 = k_z^2k^2 ^2 + 2 ^2 + ^2 k^4 + c k_z8e n_e ( d^2dr + c k_z 2 e n_e k^2k_z^2 ) , and [ gendisppp ] c_0 = ^2 k_z^2 k^2 ^2 + ( ( ) ^2 + c k_z ( ) 2 e n_e + k_z^2k^2 d^2dr ) ( ( ) ^2 + ^24 ^ 2 c k_z ( ) 2 e n_e ) [ inst1 ] a sufficient criterion for instability is clearly [ genins ] c_0 < 0 .
[ inst2 ] this would ensure that the sign of the quartic on the left side of equation ( [ gendisp ] ) changes as @xmath137 passes from small to large positive values .
hence there would have to be a positive root somewhere .
the necessity of ( [ genins ] ) as an instability criterion is more difficult to prove , but is very likely to be true .
it is straightforwardly proven in the limit of vanishing resistivity that @xmath138 leads to stability ( since the analysis reduces to the roots of a quadratic ) , and the presence or resistivity will almost certainly cause further stabilization .
numerical analyses are all consistent with ( [ genins ] ) as a necessary and sufficient criterion for instability .
the most important difference between strictly axial geometry and the present calculation lies with the form of the coupling of the hall term and wavenumber . whereas the axial result makes it appear that the hall term stabilizes or destabilizes according to whether @xmath139 is positive or negative , equations ( [ gendisp])([gendisppp ] )
show that the coupling is actually @xmath140 .
the significance of this form of the coupling is that under conditions of marginal stability , if there is any finite radial field component , there will _ always _ be wavenumbers which make this term negative ( destabilizing ) .
this is of particular importance for the nonlinear development of the instability , where power is cascaded throughout the wavenumber spectrum .
consider the case of astrophysical interest , @xmath141 .
if the final factor of equation ( [ gendisppp ] ) ( the torque term ) is negative , the preceding factor ( the centrifugal term ) will also be negative , and obviously no instability will be possible .
let us assume , therefore , that the torque is positive .
then , a necessary condition for instability is [ necc ] ( ) ^2 + c k_z ( ) 2 e n_e < - k_z^2k^2 d^2dr , which differs from the standard mri only by an additive hall term on the left hand side .
( this is a necessary condition because it omits resistivity . )
the effect of this is shown schematically in figure ( 7 ) .
we have introduced the dimensionless hall parameter , ha , the characteristic whistler frequency normalized to @xmath15 .
let us finally note that hemfs will tend to stabilize geometries with vanishing axial field . in the absence of the hall term , the standard @xmath142 mri , has its maximum growth rate for disturbances with @xmath143 . but with hall emfs included , arbitrarily large axial wavenumbers will not allow @xmath144 . instead , these modes become stable waves .
one of the most important reasons for obtaining this general axisymmetric formula ( [ gendisp ] ) is that it provides a basis for understanding the behavior of local _ nonaxisymmetric _ disturbances , to which we now turn .
we shall work in the local shearing coordinate system introduced by goldreich & lynden - bell ( 1965 ) in their pioneering study of local gravitational disk instabilities , and used in a magnetic context by balbus & hawley ( 1992b ) and terquem & papaloizou ( 1996 ) .
the shearing coordinates are defined by r = r , = - ( r ) t , z = z , which leads to = r - tddr , = , = z. the partial time derivative must be taken holding shear coordinates constant , which is just the lagrangian time derivative for the undisturbed disk flow : = t = t + ( r ) .
local disturbances depend on these sheared coordinates as @xmath145 $ ] , where the wavevector @xmath146 is a constant .
this means that we may effect the transformation from eulerian to ( unperturbed ) lagrangian coordinates simply by replacing the static eulerian wavevector @xmath147 with k_r(t ) = k_r - mt ddr , and calculating the time evolution with the lagrangian derivative . throughout this section
, @xmath84 shall denote the shearing wavevector @xmath148 . in the presence of shear ,
the toroidal magnetic field grows linearly with time : b_(t ) = b_(0 ) + t b_r ddr .
note , however , that @xmath149 is time - independent : = k_r b_r + mb_(0)r + k_z b_z .
this has important consequences for nonaxisymmetric behavior .
the linearized dynamical equations are k_r v_r + mr v_+ k_z v_z = 0 , - 2v_+ i k_r ( p + 4 ) - i ( 4 ) b_r = 0 , + i k_z ( p + 4 ) - i ( 4 ) b_z = 0 , + ^22 v_r + i m r ( p + 4 ) - i ( 4 ) b_= 0 . for the induction equations , it is helpful to define an auxiliary vector @xmath150 by = -k^2 + c 4n_e e .
the magnetic field equations are then = i v_r + a_r , = i v_z + a_z , = i v_+ ddr b_r+ a_. the above equations imply that = - k^2 , so that if chosen to vanish initially , @xmath151 will vanish throughout the evolution in a numerically stable manner .
if @xmath152 vanishes , then @xmath153 .
because of the time dependence of the wavevector , the nonaxisymmetric problem is more complex than its axisymmetric counterpart .
the approach we shall use is to reduce the system to two coupled second order differential equations for @xmath71 and @xmath72 , which are then solved numerically ( balbus & hawley 1992b ) .
this is an algebraically tedious process , and we present only the final result : @xmath154 \nonumber \\ + \frac{2mk_r}{k^2 } \frac{d \omega}{dr } \left ( \frac{d \delta b_r}{dt } -a_r \right ) - \frac{2 m \omega k_z}{r k^2 } \left ( \frac{d \delta b_z}{dt } -a_z \right ) - \left ( { \mathbf k } \cdot { \mathbf v}_a \right)^2 \delta
b_r , \label{nax1}\end{aligned}\ ] ] @xmath155 \nonumber \\
+ \frac{2mk_z}{r k^2 } \frac{d \left ( r \omega \right)}{dr } \left ( \frac{d \delta b_r}{dt } -a_r \right ) - \left ( { \mathbf k } \cdot { \mathbf v}_a \right)^2 \delta b_z .
\label{nax2}\end{aligned}\ ] ] these equations are formally the same as equations ( 2.19 ) and ( 2.20 ) of balbus & hawley ( 1992b ) , but with the first - order time derivatives of @xmath156 transformed to @xmath157 although equations ( [ nax1 ] ) and ( [ nax2 ] ) appear rather opaque , it is possible to understand the onset of instability by focusing on the comparatively simple axisymmetric inequality ( [ necc ] ) .
in essence , we find that the nonaxisymmetric evolution unfolds as a series of axisymmetric problems . in going from axisymmetry to nonaxisymmetry , the only term in the @xmath119 stability criterion that acquires a time - dependence is @xmath23 .
consider the evolution of a strongly leading disturbance , @xmath147 large and negative .
as @xmath147 initially moves towards zero , the factor @xmath158 goes from much less than , to nearly unity ( assuming @xmath159 is small ) . the right hand side of ( [ necc ] ) starts out very small , and grows to its maximum value of @xmath160 . at this stage ,
the formal instability criterion must be satisfied , or it will never be satisfied over the entire course of the evolution . then , as @xmath147 grows from zero to large and positive , the right hand side of ( [ necc ] ) once again diminishes with time , and eventually the flow will stabilize .
the situation is summarized in figure ( [ fig1 ] ) , which shows the stability and instability regions of a keplerian disk in the @xmath161 plane .
we have adopted the nominal values @xmath162 and @xmath163 , and for comparison have also displayed the magnetorotational instability results ( @xmath164 ) of balbus & hawley ( 1992b ) . note that despite the complexity of the full system of equations , in the end the hemf simply slides the hyperbolic stability curve up and down along the ordinate . since @xmath165 is constant ,
the evolving path of a perturbation is along a horizontal line in this plane , since only the abscissa coordinate changes with time .
the nature of the unstable response is then determined entirely by how long this horizontal trajectory remains inside the instability region . in figure ( [ fig2 ] ) , both the hall effect and resistivity are included . from equations ( [ inst1 ] ) and ( [ inst2 ] )
, we see that the marginal stability curves in the @xmath166 plane depend on the resistivity only through the parameter @xmath167 .
we have adopted @xmath168 . to orient oneself
, this would correspond to a disk with aspect ratio @xmath169 , @xmath170 , and @xmath171 .
we see clearly in figure ( [ fig2 ] ) that the resistivity stabilizes small radial wavelengths . in figure ( [ fig3 ] ) , to allow for direct comparison , we have plotted both the resistive and nonresistive cases .
figures ( [ fig1])([fig3 ] ) have been obtained strictly by numerically solving equations ( [ nax1 ] ) and ( [ nax2 ] ) . nevertheless , each of these curves is almost indistinguishable from those obtained simply by setting @xmath172 , where @xmath119 is defined by ( [ inst1 ] ) .
therefore , although the criterion ( [ inst2 ] ) has been derived for axisymmetric perturbations , it is very accurate for nonaxisymmetric disturbances also , with @xmath173 varying with time .
note that the curves displayed in figures ( [ fig1])([fig3 ] ) depend on @xmath174 only implicitly , through the ratio @xmath173 .
the role of @xmath174 is limited to determining the evolution of @xmath173 , not the properties of the stability curves . in figure ( [ fig4 ] ) , we show the evolution of the perturbed radial field @xmath71 ( the curves for @xmath175 and @xmath176 are similar ) for the same @xmath177 values as above , and for both @xmath178 and @xmath179 .
the value of @xmath180 is chosen because it lies in the unstable regime for an axisymmetric @xmath122 disk , for all the hall parameters .
initially @xmath173 is very nearly unity , and the perturbation grows until the wavevector moves out of the relevant unstable region in the @xmath181 plane . when @xmath178 , explosive growth is achieved for the three values of @xmath177 considered . as expected , the cases @xmath162 and @xmath163 are respectively more and less unstable than the case @xmath164 .
the effect of the resistivity is to suppress the growth of the perturbation at large wavenumbers , corresponding to very early and very late times in the evolution .
the dramatic reduction of peak amplitudes is evident .
the @xmath182 case , however , remains quite unstable even with @xmath183 . to conclude
, we have found that if @xmath184 is small , the nonaxisymmetric behavior of linear perturbations may be understood on the basis of axisymmetric behavior .
( large @xmath174 disturbances are stable . ) whether growth occurs or not depends upon the instantaneous location of the wavenumber in the @xmath185 plane .
shear causes the wavenumber to retrace a horizontal line in this plane , first from right to left ( for an initially leading distrubance ) , then the reverse .
if this path takes the wavenumber into an _ axisymmetrically _ unstable zone , the disturbance grows .
the effect of resistivity , not surprisingly , is to dampen growth , with the smallest wavelengths being the most affected .
in a gas which is at least partially ionized , the magnetic field is , but for resistivity , frozen into the electron fluid .
the difference in the mean electron velocity and the center - of - mass fluid velocity gives rise to hemfs in the gas .
somewhat surprisingly , in astrophysical environments these can often be more important than ohmic dissipation , even in a fully ionized plasma .
this point is often not appreciated . in protostellar disks ,
the hall effect is small compared with resistivity only if the _ neutral _
density is much in excess of @xmath186 @xmath187 , or if @xmath188 .
the dynamical consequence of hemfs is the appearance of whistler waves as a mode of the gas response .
these waves are carried nominally by the electrons , but inductive electron - ion coupling and ion - neutral collisional coupling together assure that the bulk of the fluid is involved .
it is the interplay between these right - handed circularly polarized whistlers and the coriolis driven epicycles that affects the stability of magnetized disks .
whistler waves become dynamically important when the ionization fraction drops below @xmath189 .
the hall effect allows disks with either decreasing outward or increasing outward angular velocity profiles to become unstable . by way of contrast
, the standard mri affects only those disks with a decreasing outward profile , even though any angular velocity gradient increases the rotation energy for fixed angular momentum .
the maximum growth rate is always found to be the oort a value , @xmath190 . whether hemfs stabilize or destabilize a disk depends not upon @xmath1 , but upon @xmath191 .
the latter should be negative for destabilization .
if @xmath3 has a radial component , then it is always possible to find a wavenumber to make this parameter whatever value is desired .
thus , hall physics will always tend to make some wavenumbers more unstable .
whether or not the disk actually become unstable will depend on the value of the resistivity .
the nonlinear consequences of hemfs are likely to be particularly important for our understanding of transitions between active and quiescent states in both dwarf novae and fu orionis outbursts .
neither of these systems is presently understood at a fundamental level .
numerical simulations which include only ohmic losses have yet to demonstrate that mhd turbulence can be sustained at the ionization fractions that may be present in these cool disks ( fleming et al .
2000 , menou 2000 ) . numerical simulations including both the hall effect and ohmic dissipation should prove most informative .
we thank j. hawley , k. menou , and j. stone for useful discussions , and w. winters for his skillful preparation of several figures .
sab is grateful to the institut dastrophysique de paris for its hospitality and visitor support .
both authors would like to thank prof .
j. papaloizou of the astronomy unit at qmw , where this work was initiated , for his generous support under pparc grant ppa / v / o/1997/00261 .
sab is supported by nasa grants nag 57500 , nag 59266 , and nsf grant ast0070979 .
|
the effects of hall electromotive forces ( hemfs ) on the linear stability of protostellar disks are examined .
earlier work on this topic focused on axial field and perturbation wavenumbers . here
we treat the problem more generally . both axisymmetric and nonaxisymmetric
cases are investigated .
though seldom explicitly included in calculations , hemfs appear to be important whenever ohmic dissipation is .
they allow for the appearance of electron whistler waves , and since these have right - handed polarization , a helicity factor is introduced into the stability problem .
this factor is the product of the components of the angular velocity and magnetic field along the perturbation wavenumber , and it is destabilizing when negative . unless the field and angular velocity are exactly aligned , it is always possible to find destabilizing wavenumbers .
hemfs can destabilize any differential rotation law , even those with angular velocity increasing outward .
regardless of the sign of the angular velocity gradient , the maximum growth rate is always given in magnitude by the local oort a value of the disk , as in the standard magnetorotational instability .
the role of hall emfs may prove crucial to understanding how turbulence is maintained in the `` low state '' of eruptive disk systems .
= 2 # 1to 0pt#1 # 1 2 truein 2 truein
|
throughout all this paper we fix a @xmath0-adically closed field @xmath1 . the reader unfamiliar with this notion may restrict to the special case where @xmath2 or a finite extension of it , and @xmath3 is its @xmath0-adic valuation .
we let @xmath4 denote the valuation ring of @xmath3 , and @xmath5 its valuation group ( augmented with one element @xmath6 ) . in this introductory section
we present informally what we are aiming at .
precise definitions will be given in section [ se : notation ] and at the beginning of section [ se : p - adic ] .
our long - term objective is to set a triangulation theorem which would be an acceptable analogue over @xmath7 of the classical triangulation of semi - algebraic sets over the reals .
polytopes and simplexes in @xmath8 are well known to have the following properties , among others ( see for example @xcite or @xcite ) . (
sim ) : : they are bounded subsets of @xmath8 which can be described by a finite set of linear inequalities of a very simple form .
( fac ) : : there is a notion of `` faces '' attached to them with good properties : every face of a polytope @xmath9 is itself a polytope ; if @xmath10 is a face of @xmath11 and @xmath11 a face of @xmath9 then @xmath10 is a face of @xmath9 ; the union of the proper faces of @xmath9 is a partition of its frontier .
( div ) : : last but not least , every polytope can be divided in simplexes by a certain uniform process of `` barycentric division '' which offers a good control both on their shapes and their faces .
the goal of the present paper is to build a @xmath0-adic counterpart of real polytopes and simplexes having similar properties .
obviously there is no direct translation of concepts like linear _ inequalities _ and _ barycentric _ division to _ non - ordered _ fields , such as the @xmath0-adic ones .
nevertheless we want our @xmath0-adic polytopes and simplexes to be defined by conditions which are as simple as possible , to obtain a notion of faces satisfying all the above properties , and most of all to develop a flexible and powerful division tool .
this is achieved here by first introducing and studying certain subsets of @xmath12 called `` largely continuous precells mod @xmath13 '' , for a fixed @xmath14-tuple @xmath13 of positive integers .
these sets will be defined by a very special triangular system of linear inequalities and congruence relations mod @xmath13 . in particular
they are defined simply by linear inequalities in the special case where @xmath15 ( again , see section [ se : notation ] for precise definitions and basic examples ) .
this paper , which is essentially self - contained , is organised as follows .
the general properties of subsets of @xmath12 defined by conjunctions of linear inequalities and congruence conditions are studied in section [ se : face - proj ] .
property ( fac ) is proved there to hold true for largely continuous precells mod @xmath13 ( while property ( sim ) is a by - product of their definition ) .
section [ se : bound - fun ] is devoted to two technical properties preparing the proof of our main result , a construction analogous to ( div ) in our context .
we call this `` monohedral division '' ( see below )
. section [ se : division ] is devoted to its proof .
we then return to the @xmath0-adic context in the final section [ se : p - adic ] . by taking inverse images of largely continuous precells by the valuation @xmath3 ( which maps @xmath16 onto @xmath12 ) and restricting them to certain subsets of @xmath17 , we transfer all the definitions and results built in @xmath12 in the previous sections to @xmath16 , especially the monohedral division ( which becomes in this context the `` monotopic division '' , theorem [ th : div - p - adic ] ) .
this latter result paves the way towards a triangulation of semi - algebraic @xmath0-adic sets , to appear in a further paper .
[ [ monohedral - division . ] ] monohedral division .
+ + + + + + + + + + + + + + + + + + + + in addition to ( sim ) and ( fac ) , every largely continuous precell @xmath18 mod @xmath13 has one more remarkable property which real polytopes are lacking : its proper faces , ordered by specialisation if and only if @xmath19 is contained in the closure or @xmath18.[fo : spec - ord ] ] , form a rooted tree ( proposition [ pr : face - egale - proj]([it : face - lattice ] ) .
when this tree is a chain , we say that @xmath18 is `` monohedral '' . among real polytopes of a given dimension , the simplexes are those whose number of facets is minimal : a polytope @xmath20 of dimension @xmath21 has at least @xmath22 facets , and it is a simplex if and only if it has exactly @xmath22 facets ( see corollary 9.5 and corollary 12.8 in @xcite ) .
we expect largely continuous precells to fulfill in @xmath12 a function similar to polytopes in @xmath8 , and the monohedral ones ( whose ordered set of faces is in a sense the simplest possible tree ) to fulfill a function similar to simplexes .
indeed our main result , the `` monohedral division '' ( theorem [ th : mono - div ] ) , provides in our context a powerful tool very similar to ( div ) , the barycentric division of real polytopes .
it provides in particular a `` monohedral decomposition '' ( theorem [ th : mono - dec ] ) which says that every largely continuous precell mod @xmath13 in @xmath12 is the disjoint union of a complex of _ monohedral _ largely continuous precells mod @xmath13 .
the latter result is in analogy with the situation in the real case , where every polytope can be divided in simplexes forming a simplicial complex .
but the barycentric division in @xmath8 says much more than this .
roughly speaking , given a polytope @xmath18 and a simplicial complex @xmath23 partitioning the frontier of @xmath18 , it makes it possible to build a simplicial complex @xmath24 by partitioning @xmath18 and `` lifting '' @xmath23 , in the sense that for every @xmath25 in @xmath24 , the faces @xmath26 of @xmath25 which are outside @xmath18 belong to @xmath23 .
moreover , given a positive function @xmath27 ( where @xmath19 is any proper face of @xmath18 ) , the shapes of the elements of @xmath24 can be required to satisfy the following condition : for every @xmath26 in @xmath23 there is a unique @xmath28 such that @xmath26 is the largest proper face of @xmath25 in @xmath24 , and in that case the distance of any point @xmath29 to its projection @xmath30 onto @xmath19 is smaller than @xmath31 ( see figure [ fi : div ] , where the dotted curve shows how the @xmath32 function controls the shapes of the elements of @xmath24 whose largest proper face outside @xmath18 is contained in the lower facet @xmath19 ) .
\(a ) at ( 1,0 ) ; ( b ) at ( 3,0 ) ; ( c ) at ( 5,0 ) ; ( d ) at ( 6,0 ) ; ( e ) at ( 7,1 ) ; ( f ) at ( 6.5,2.5 ) ; ( g ) at ( 6,4 ) ; ( h ) at ( 3,4.5 ) ; ( i ) at ( 0,3 ) ; ( j ) at ( 0.5,1.5 ) ; ( ab ) at ( 2,0.5 ) ; ( cd ) at ( 4.5,1 ) ; ( gh ) at ( 4,3 ) ; ( ij ) at ( 3,2 ) ; ( x ) at ( 3.75,.5 ) ; ( a ) ( d ) ( cd ) ( x ) ( ab ) cycle ; ( a ) ( d ) ( e ) ( g ) ( h ) ( i ) ( a ) ; ( a ) ( ab ) ( x ) ( cd ) ( f ) ( gh ) ( h ) ( ij ) ( x ) ( ij ) ( j ) ( ab ) ( cd ) ( ij ) ( gh ) ( cd ) ( c ) ( cd )
( e ) ( gh )
( g ) ( cd )
( b ) ( ab ) ( ij )
( i ) ; ( 1,.5 ) .. controls + ( 1,.5 ) and + ( -1,0 ) .. ( 3.2,.7 ) .. controls + ( 1,0 ) and + ( -1.2,2.4 ) .. ( d ) ; in a , b , c , d , e , f , g , h , i , j ( ) node @xmath33 ; although all these properties can be derived from the barycentric division in @xmath8 , none of them involves the notion of barycenter . the strength of our monohedral division in @xmath12 ( theorem [ th : mono - div ] ) , and eventually of our monotopic division in @xmath16 ( theorem [ th : div - p - adic ] ) , is that they preserve in @xmath24 has no proper face inside @xmath18 : either @xmath25 has no proper face at all , or its unique facet is outside @xmath18 ( hence so are all its proper faces ) and belongs to @xmath23 . ] all of these properties .
we let @xmath34 , @xmath35 , @xmath36 denote respectively the set of non - negative integers , of integers and of rational numbers , and @xmath37 . for every @xmath38 we let @xmath39\!]$ ]
denote the set of integers @xmath40 such that @xmath41 ( that is the empty set if @xmath42 ) .
recall that a * * @xmath35-group is a linearly ordered group @xmath43 with a smallest positive element such that @xmath44 has @xmath45 elements for every integer @xmath46 .
the reader unfamiliar with @xmath35-groups may restrict to the special but prominent case of @xmath35 itself .
indeed a linearly ordered group is a @xmath35-group if and only if it is elementarily equivalent to @xmath35 ( in the presburger language @xmath47 defined below ) .
@xmath1 is a @xmath0-adically closed field in the sense of @xcite , that is a henselian valued field of characteristic zero whose residue field is finite and whose value group @xmath48 is a @xmath35-group . a field is @xmath0-adically closed if and only if it is elementarily equivalent ( in the language of rings ) to a finite extension of @xmath49 , so the reader unfamiliar with the formalism of model - theory may restrict to this fundamental case .
let @xmath50 be the divisible hull of @xmath51 . by identifying @xmath35 with the smallest non - trivial convex subgroup of @xmath51
, we consider @xmath35 embedded into @xmath51 ( and @xmath36 into @xmath50 ) . for every @xmath52 we let
@xmath53 .
[ re : def - naive-1 ] when @xmath54 , one may naively define polytopes in @xmath55 as intersections @xmath56 where @xmath57 is a polytope of @xmath8 . with other words , a polytope in @xmath55 ( and more generally in @xmath58 )
would be an intersection of finitely many half - spaces , that is the set of solutions of finitely many linear inequalities .
our polytopes are indeed so , but it will soon become clear that we need to be much more restrictive . note first that such a definition does not lead naturally to a good notion of faces , because @xmath59 is closed with respect to the topology inherited from @xmath8 .
in particular if @xmath11 is a face of @xmath9 , @xmath60 is disjoint from the closure of @xmath59 .
in order to carry significant topological properties , a notion of face for subsets of @xmath55 must involve points at infinity . for every @xmath61 in @xmath62 we let @xmath63 .
@xmath64 is endowed the topology generated by the open intervals and the intervals @xmath65a,+\infty]$ ] for @xmath52 .
@xmath66 is equipped with the product topology , and @xmath12 with the induced topology .
the topological closure of any set @xmath18 in @xmath66 is denoted @xmath67 .
thus for example @xmath68 and @xmath69 .
the * * frontier of a subset @xmath18 of @xmath66 is the closure of @xmath70 .
we denote it @xmath71 .
whenever we take an element @xmath72 it is understood that @xmath73 are its coordinates .
we say that @xmath61 is * * non - negative if all its coordinates are .
a subset @xmath18 of @xmath66 is * * non - negative if all its elements are .
a function @xmath74 with values in @xmath64 is * * non - negative ( resp . * * positive ) on a subset @xmath75 of its domain if @xmath76 is non - negative ( resp .
positive ) for every @xmath77 .
if @xmath78 we let @xmath79 ( resp .
@xmath80 ) denote the image of @xmath61 ( resp .
@xmath18 ) under the coordinate projection of @xmath66 onto the first @xmath81 coordinates @xmath82 .
we call it the * * socle of @xmath61 ( resp .
@xmath18 ) .
if @xmath83 is a family of subsets of @xmath12 we also call @xmath84 the socle of @xmath83 . the * * support of @xmath61 , denoted @xmath85 , is the set of indexes @xmath86 such that @xmath87 . when all the elements of @xmath18 have the same support , we call it the * * support of @xmath18 and denote it @xmath88 . for every subset @xmath89 of @xmath90\!]$ ]
we let : @xmath91 when @xmath92 we call it the * * face of @xmath18 of support @xmath93 .
it is an * * upward face if moreover @xmath94 and @xmath95 is non - empty .
note that if @xmath18 is contained in @xmath12 then so are its faces , because @xmath12 is closed in @xmath66 . by construction , @xmath96 hence @xmath67 is the disjoint union of all its faces .
a * * complex in @xmath12 is a finite family @xmath83 of pairwise disjoint subsets of @xmath12 such that for every @xmath97 , @xmath98 is the union of the _ common _ faces of @xmath18 and @xmath19 .
it is a * * closed complex if moreover it contains all the faces of its members , or equivalently if @xmath99 is closed .
note that a finite partition @xmath100 of a subset @xmath18 of @xmath12 is a closed complex if and only if @xmath100 contains the faces of all its members .
the specialisation pre - order ( see footnote [ fo : spec - ord ] ) is an order on the faces of @xmath18 .
the largest proper faces of @xmath18 with respect to this order are called its * * facets .
we say that @xmath18 is * * monohedral if its faces are linearly ordered by specialisation .
note that every subset of @xmath101 is clopen in @xmath101 .
in particular if @xmath102 then @xmath71 is the disjoint union of its proper faces .
note also that @xmath103 whenever @xmath104 .
[ ex : face - non - cvx ] let @xmath105 be defined by @xmath106 , @xmath107 and @xmath108 .
it has four non - empty faces : @xmath18 itself , two facets @xmath109 and @xmath110 , plus @xmath111 .
we let @xmath112 be the natural projection of @xmath12 onto @xmath101 .
when @xmath14 is clear from the context , @xmath113 is simply denoted @xmath114 .
[ re : face - in - proj ] for every @xmath115 note that @xmath116 and @xmath117 ( where @xmath118 ) .
indeed for every @xmath119 there are points in @xmath18 arbitrarily close to @xmath120 .
in particular there is a point @xmath121 such that @xmath122 , which implies that @xmath123 .
this proves the first inclusion , and we leave the second one to the reader . for every @xmath124\!]$ ] and @xmath72 we let @xmath125 ( if @xmath126\!]$ ] we use the convention that @xmath127 ) . again
the superscript @xmath14 is omitted whenever it is clear from the context .
note that for every @xmath128 @xmath129 [ re : delta - dist ] when @xmath54 the topology agrees with the usual one , induced by the euclidian distance .
] on @xmath66 comes from the distance @xmath130 , with the convention that @xmath131 .
thus @xmath132 is just the distance from @xmath61 to its projection @xmath133 . in the general case
the topology on @xmath66 no longer comes from a distance .
nevertheless we will keep this geometric intuition in mind , that @xmath134 measures something like a distance from @xmath61 to @xmath135 : the bigger @xmath134 is , the closer @xmath61 is to @xmath135 .
this intuitive meaning makes the following facts rather obvious .
[ fa : delta - dist ] for every function @xmath136 , given @xmath137 such that @xmath138 we have : 1 .
@xmath139 iff @xmath140 iff @xmath141 , @xmath142 , @xmath123 and @xmath143 .
if @xmath139 then @xmath74 has limit @xmath144 at @xmath120 iff @xmath145 , @xmath146 , @xmath147 , @xmath148 and @xmath149 \rightarrow f(a)\geq\varepsilon$ ] . given a vector @xmath150 we let @xmath151 .
we say that @xmath152 is * * pointing to some @xmath124\!]$ ] if @xmath153 for @xmath154 and @xmath155 for @xmath156 .
[ re : half - line ] let @xmath157\!]$ ] and @xmath9 be any subset of @xmath101 .
using remark [ re : face - in - proj ] and the above facts , one easily sees that if for every @xmath158 there is @xmath150 pointing to @xmath159 such that @xmath160 and @xmath161 then @xmath162 , and in particular @xmath163 .
example [ ex : face - non - cvx ] shows that even if a subset @xmath18 of @xmath55 is defined by finitely many linear inequalities , its faces may not be so .
thus a polytope @xmath18 in @xmath12 must satisfy additional conditions , in order to ensure that some linear inequalities which define @xmath18 also define its faces after passing to the limits ( see proposition [ pr : pres - face ] for a precise statement ) .
it is these conditions that we are going to introduce now .
a function @xmath136 is * * largely continuous on @xmath18 if it can be extended to a continuous function on @xmath67 , which we will usually denote @xmath164 .
if @xmath18 has support @xmath93 , we say that @xmath74 is an * * affine map ( resp . a * * linear map ) if either @xmath74 is constantly equal to @xmath144 , or for some @xmath165 ( resp .
@xmath166 ) and some @xmath167 , we have @xmath168 we call @xmath169 the `` constant coefficient '' in the above expression of @xmath74 .
if such an expression exists for which @xmath170 and @xmath171 for @xmath172 , we say that @xmath74 is * * integrally affine . a affine map which takes values in @xmath173 will be called * * @xmath173-affine .
for example @xmath174 is @xmath173-affine on @xmath175 but is not integrally affine .
[ re : linearity ] affinity and linearity are intrinsic properties because a function @xmath176 is a linear map if and only if for every @xmath177 and every @xmath178 : @xmath179 the symbols of the presburger language @xmath180 are interpreted as usually in @xmath51 : the binary relation @xmath181 says that @xmath182 , and the other symbols have their obvious meanings .
a subset @xmath75 of @xmath183 is * * @xmath47-definable if there is a first order formula @xmath184 in @xmath47 , with parameters in @xmath51 and a @xmath21-tuple @xmath185 of free variables , such that @xmath186 .
a function @xmath187 is * * @xmath47-definable if its graph is .
each @xmath101 can be identified with @xmath183 with @xmath188 .
we say that a subset @xmath18 of @xmath12 is * * definable if for every @xmath189\!]$ ] the set @xmath190 is @xmath47-definable by means of this identification .
we say that a function @xmath136 is * * definable if there is an integer @xmath191 such that @xmath192 and if the restrictions of @xmath193 to each @xmath101 become , after this identification , either an @xmath47-definable map from @xmath194 to @xmath51 or the constant map @xmath144 . note that every affine map is definable in this broader sense .
the next characterisation of definable maps and sets comes directly from theorem 1 in @xcite .
[ th : cluck - piece - lin ] for every definable function @xmath195 on a non - negative set @xmath18 , there exists a partition of @xmath18 in finitely many definable sets , on each of which the restriction of @xmath74 is an affine map .
it is well known that the theory of @xmath35-groups has quantifier elimination and definable skolem functions . at many places , without mentioning , we will use the latter property under the following form .
[ th : skolem ] let @xmath196 and @xmath197 be two @xmath47-definable sets .
let @xmath198 be a first order formula in @xmath47 . if for every @xmath121 there is @xmath199 such that @xmath200 then there is a definable map @xmath201 such that @xmath202 for every @xmath203
. since @xmath51 is elementarily equivalent to @xmath35 in the language @xmath47 , every non - empty @xmath47-definable subset of @xmath51 which is bounded above ( resp
. below ) has a maximum ( resp .
minimum ) element . as a consequence for every @xmath204 there is in @xmath51 a largest element @xmath205 ( resp .
@xmath206 ) which is @xmath207 ( resp .
@xmath208 ) . note that if @xmath209 is definable and @xmath191 is an integer such that @xmath193 is @xmath47-definable , then for every integer @xmath210 the set @xmath211 is @xmath47-definable , and so is the map @xmath212 on @xmath213 .
thus the map @xmath214 is @xmath47-definable , and so is @xmath215 by a symmetric argument .
obviously the same holds true for every definable map from @xmath216 to @xmath64 .
[ le : precomp - borne ] if @xmath217 is a largely continuous definable map on a non - negative set @xmath18 , then it has a minimum in @xmath18 .
it suffices to prove the result separately for each @xmath190 with @xmath189\!]$ ] .
every such piece can be identified with a definable subset of @xmath218 hence we can assume that @xmath196 .
multiplying @xmath74 by some integer @xmath46 if necessary we can assume that @xmath74 takes values in @xmath173 , and even in @xmath51 ( otherwise @xmath74 is constantly @xmath144 and the result is trivial ) .
since @xmath219 , by instantiating the parameters of a definition of @xmath220 it suffices to prove the result for every largely continuous definable function on a non - negative subset @xmath18 of @xmath55 .
but in that case the topology on @xmath12 comes from a metric such that every non - negative subset of @xmath12 is precompact ( that is @xmath67 is compact ) .
so there is @xmath221 such that @xmath222 .
for any @xmath223 close enough to @xmath224 we have @xmath225 ( because @xmath226 ) hence @xmath227 .
[ le : image - compact ] let @xmath228 a continuous definable map
. if @xmath18 is non - negative then @xmath229 is closed . as for lemma
[ le : precomp - borne ] we can reduce to the case where @xmath54 .
but then @xmath67 is compact , hence so is @xmath229 since @xmath74 is continuous .
we extend the binary congruence relations of @xmath51 to @xmath173 with the convention that @xmath230 $ ] for every @xmath231 and every @xmath232 .
a subset @xmath18 of @xmath101 is a * * basic presburger set if it is the set of solutions of finitely many linear inequalities and congruence relations .
although we will not use it , it is worth mentioning that , by the quantifier elimination of the theory of @xmath35 in the language @xmath47 , the definable subsets of @xmath183 , and more generally of @xmath12 , are exactly the finite unions of basic presburger sets .
cluckers has shown in @xcite that every definable subset of @xmath183 is actually the disjoint union of finitely many subsets of a much more restrictive sort , called cells .
the following definition of precells in @xmath12 , more precisely of precells mod @xmath13 for a given @xmath233 , is adapted from his ( see remark [ re : cell - cluc ] below ) . since we only need to consider non - negative precells it is convenient to restrict the definition to this case . if @xmath234 , @xmath235 itself is the only precell mod @xmath13 in @xmath235 . if @xmath78 , for every @xmath189\!]$ ] , a subset @xmath18 of @xmath101 is a * * precell mod @xmath13
if : @xmath80 is a precell mod @xmath236 , and there are non - negative affine maps @xmath237 and an integer @xmath238 such that @xmath239 and @xmath18 is exactly the set of points @xmath240 such that @xmath241 and @xmath242 .
\label{eq : def - cell - mod - n}\ ] ] we call @xmath243 , @xmath244 the * * boundaries of @xmath18 , @xmath238 a * * modulus for @xmath18 , and such a triple @xmath245 a * * presentation of @xmath18 .
we call it a * * largely continuous presentation of @xmath18 if moreover @xmath246 are largely continuous .
@xmath18 is a * * largely continuous precell mod @xmath13 if @xmath234 , or if @xmath80 is largely continuous precell mod @xmath236 and @xmath18 is a precell mod @xmath13 having a largely continuous presentation . [
re : cell - cluc ] cells in @xcite are not required to have non - negative boundaries , but to be of one of these two types : either @xmath247 is not finitely bounded or @xmath248 .
unfortunately this condition seems to be too restrictive for our constructions , we have to relax it .
thus our precells are _ not _ cells in the sense of @xcite but a restriction ( we require non - negative boundaries ) of a slight generalisation ( @xmath247 can be finitely bounded and non - zero ) of them .
we are going to prove in the next section that the faces of every largely continuous precell mod @xmath13 are still largely continuous precells mod _ the same _ @xmath13 ( see proposition [ pr : pres - face ] ) .
thus , if one restricts to the case where @xmath15 , these precells are the best candidate we have for a discrete analogue of real polytopes : they are intersections of finitely many half - spaces and their faces are so . in the @xmath0-adic triangulation that we are aiming at we
will indeed restrict to this case .
however it appears that all the results and constructions that we are going to consider in the present paper remain valid for largely continuous precells mod arbitrary @xmath13 .
since it does not create significant complication , we will then stick to this more general setting .
in this section we consider a non - empty basic presburger set @xmath249 defined by @xmath250 \label{eq : basic - pres}\ ] ] where @xmath251 are integrally linear maps , @xmath252 , @xmath253 and @xmath254 are integers such that @xmath255 .
we prove some basic properties on the faces of @xmath18 and the affine maps on @xmath18 .
finally we derive from these facts that every face of a largely continuous precell @xmath18 mod @xmath13 is a largely continuous precell mod @xmath13 and has a presentation inherited from @xmath18 in a uniform way ( proposition [ pr : pres - face ] ) .
example [ ex : face - non - cvx ] shows that precell mod @xmath13 ( here @xmath256 ) can have a facet which is no longer a precell mod @xmath13 .
but even worse is possible : the next example shows that a precell mod @xmath13 can have a facet which is not even a basic presburger set .
[ ex : face - non - presb ] @xmath105 is defined by @xmath257 and @xmath258 . its unique facet @xmath259 is defined by @xmath260 and either @xmath261 $ ] or @xmath262 $ ] ( and of course @xmath263 ) .
[ le : half - line ] let @xmath102 be defined by ( [ eq : basic - pres ] ) .
let @xmath159 be any subset of @xmath90\!]$ ] .
then @xmath264 if and only if for every @xmath158 there is @xmath150 pointing to @xmath159 such that @xmath160 and @xmath265 .
it suffices to prove the result when @xmath266\!]$ ] .
one direction is general by remark [ re : half - line ] , so let us prove the converse .
assume that @xmath264 and fix any @xmath158 . without lost of generality we can assume that @xmath267 . pick @xmath268 and let @xmath269 . by remark [ re : face - in - proj ] , @xmath270 .
assume that for some @xmath271\!]$ ] we have found a definable subset @xmath272 of @xmath273 such that @xmath274 and for every @xmath275\!]$ ] , either @xmath276 is constant on @xmath272 or @xmath277 tends to @xmath144 as @xmath278 tends to @xmath279 in @xmath272 .
if the same holds true for @xmath280 , let @xmath281 .
otherwise , there is some @xmath282 such that for every @xmath283 there is @xmath284 such that @xmath285 and @xmath286 .
the set @xmath287 of these @xmath288 s is definable , non - empty , and bounded below since @xmath289 on @xmath272 .
hence it has a minimum , say @xmath290 . by minimality of @xmath290
there is @xmath291 such that for every @xmath284 such that @xmath292 , @xmath293 .
thus , for every @xmath283 there is @xmath294 such that @xmath285 and @xmath295 ( because @xmath296 ) .
with other words , the set @xmath297 defined by @xmath298 is such that @xmath299 ( see fact [ fa : delta - dist ] ) .
it obviously has all the other required properties since it is contained in @xmath272 . by repeating the process until @xmath300 we get a definable set @xmath301 as above .
pick any @xmath302 , by construction there is @xmath283 such that for every @xmath303 if @xmath285 then @xmath304 for every @xmath305\!]$ ] .
pick any @xmath306 such that @xmath307 and @xmath308 for every @xmath156 .
it remains to check that @xmath309 gives the conclusion .
for every @xmath310 , @xmath311 because @xmath312 and @xmath313 , hence @xmath314 .
for @xmath156 we have @xmath315 , hence @xmath316 .
in particular @xmath152 points to @xmath159 and @xmath160 . finally let @xmath278 be any element of @xmath301 .
for every @xmath317 we have @xmath318 since @xmath319 , and by linearity of @xmath276 @xmath320 we also have @xmath321 by linearity , and @xmath322 because @xmath307 , hence @xmath323 .
it follows that @xmath324 by ( [ eq : fl - x - plus - u ] ) . on the other hand , for every @xmath325\!]$
] we have @xmath326 $ ] because @xmath319 , @xmath327 $ ] and @xmath328 $ ] for the same reason , hence @xmath329 $ ] .
thus @xmath330 for every @xmath319 , which proves the result .
[ pr : face - egale - proj ] let @xmath102 be a basic presburger set , @xmath159 and @xmath331 be any subsets of @xmath90\!]$ ] such that @xmath332 and @xmath333 are non - empty . 1 . [
it : face - proj ] @xmath334 .
[ it : face - face ] if @xmath335 then @xmath336 .
[ it : mono - supp ] @xmath337 if and only if @xmath335 . in particular the faces of @xmath18 are linearly ordered by specialisation if and only if their supports are linearly ordered by inclusion .
[ it : face - lattice ] @xmath338 is non - empty
. we will refer to the @xmath45-th point of proposition [ pr : face - egale - proj ] as to proposition [ pr : face - egale - proj](@xmath45 ) .
[ re : sub - mono ] proposition [ pr : face - egale - proj]([it : face - lattice ] ) shows that the set of faces of @xmath18 ordered by specialisation is a distributive lower semi - lattice with one smallest element .
if @xmath9 is any monohedral subset of @xmath12 , proposition [ pr : face - egale - proj]([it : mono - supp ] ) implies that every basic presburger subset @xmath18 of @xmath9 is monohedral , and proposition [ pr : face - egale - proj]([it : face - face ] ) that every face of @xmath18 is again monohedral .
the first point @xmath334 follows from lemma [ le : half - line ] , by remark [ re : half - line ] applied to @xmath339 . for the second point
, @xmath335 implies that @xmath340 . since @xmath341 and @xmath334 by the first point , it suffices to prove that @xmath342 . for every @xmath158
there is by lemma [ le : half - line ] a vector @xmath150 pointing to @xmath331 such that @xmath343 and @xmath265 .
then obviously @xmath344 , and the conclusion follows by remark [ re : half - line ] applied to @xmath345 .
for the third point , one direction follows from the second point and the other direction is general since @xmath346 , @xmath347 , and @xmath348 is disjoint from @xmath349 if @xmath331 is not contained in @xmath159 .
it remains to prove the last point .
for every @xmath158 , lemma [ le : half - line ] gives @xmath350 and @xmath351 in @xmath58 pointing to @xmath159 and @xmath331 respectively such that @xmath352 , @xmath353 and similarly for @xmath351 . without lost of generality
we can assume that @xmath267 hence for every @xmath354 , @xmath355 .
in particular @xmath356 points to @xmath357 and @xmath358 . obviously @xmath359 is contained in @xmath18 .
so @xmath360 is non - empty by remark [ re : half - line ] .
[ pr : prol - pi ] let @xmath102 be a basic presburger set defined by ( [ eq : basic - pres ] ) , @xmath361 be an affine map , @xmath362 and @xmath363 .
assume that @xmath19 is not empty and that @xmath74 extends to a continuous map @xmath364 .
then @xmath365 is affine , and if @xmath366 then @xmath367 . in particular if @xmath366 then @xmath368 .
if @xmath74 is @xmath173-affine then so is @xmath365 of course .
however , if @xmath74 is integrally affine we can not conclude that @xmath365 will be integrally affine as well , even if @xmath74 is largely continuous , as the following example shows .
[ ex : prol - non - z - lin ] keep @xmath105 as in example [ ex : face - non - cvx ] .
the map @xmath369 is integrally affine and largely continuous on @xmath18 , with @xmath370 on @xmath71 .
this is no longer an integrally affine map on @xmath371 .
it suffices to prove the result when @xmath266\!]$ ] , @xmath372 is an integrally linear map and @xmath365 is not constantly equal to @xmath144 .
note that @xmath365 is affine by remark [ re : linearity ] ( because equalities satisfied by @xmath74 at every point of @xmath18 pass to the limits ) .
so we only have to prove that @xmath373 for every @xmath319 .
let @xmath374 be an integrally linear map on @xmath58 extending @xmath74 , and @xmath375 such that @xmath376 .
since @xmath377 and @xmath76 tends to @xmath378 at @xmath120 , there exists @xmath158 such that for every @xmath319 , if @xmath379 and @xmath380 then @xmath381 . pick any @xmath121 such that @xmath123 and @xmath143 , hence @xmath382 .
now assume for a contradiction that @xmath383 for some @xmath384 .
let @xmath385 , since @xmath76 tends to @xmath386 at @xmath279 and @xmath387 there exists @xmath283 such that for every @xmath319 , if @xmath388 and @xmath285 then @xmath389 .
lemma [ le : half - line ] gives @xmath150 pointing to @xmath159 such that @xmath390 and @xmath265 .
then @xmath391 and @xmath392 , hence @xmath393 . by linearity
it follows that @xmath394 . on the other hand we have @xmath395 and @xmath396 hence @xmath397 , and thus by linearity @xmath398 , a contradiction .
[ pr : face - socle ] let @xmath102 be a basic presburger set with @xmath78 , and @xmath399 .
then for every face , @xmath102 and @xmath264 imply that @xmath362 . ]
@xmath363 , its socle @xmath400 is a face of @xmath80 .
if moreover @xmath18 is non - negative , then conversely for every face @xmath401 of @xmath75 there is a face @xmath19 of @xmath18 such that @xmath402 . in that case @xmath403 if @xmath404 , and @xmath405 if @xmath406 .
[ re : face - socle ] the last assertion on @xmath19 is general : for every subset @xmath9 of @xmath101 and every face @xmath407 with socle @xmath401 , we have @xmath408 if @xmath409 , and @xmath410 if @xmath411 . indeed @xmath412 and @xmath413 is equal to @xmath414 if @xmath409 and to @xmath415 otherwise .
let @xmath416 , its proper faces are @xmath417 , @xmath418 and @xmath419 .
thus @xmath420 is a facet of @xmath421 which is not the socle of any face of @xmath18 .
this example shows that the assumption that @xmath18 is non - negative is needed for the second part of proposition [ pr : face - socle ] to hold .
note that @xmath422 is not a facet of @xmath421 , which shows that the positivity of @xmath18 is mandatory also in corollary [ co : socle - facet ] .
given that @xmath363 is a face of @xmath18 , hence non - empty , let us prove that @xmath423 . for every @xmath158 we can find a vector @xmath150 pointing to @xmath159 such that @xmath160 and @xmath265 , in particular @xmath424 points to @xmath425 , @xmath426 and @xmath427 .
thus @xmath423 by remark [ re : half - line ] applied to @xmath428 .
since @xmath429 by proposition [ pr : face - egale - proj]([it : face - proj ] ) , and obviously @xmath430 , it follows that @xmath400 .
now assume that @xmath18 is non - negative .
then the socle of @xmath67 is closed by lemma [ le : image - compact ] .
it contains @xmath75 , hence @xmath431 .
in particular it contains @xmath401 , which is non - empty .
so there is @xmath140 whose socle @xmath432 belongs to @xmath401 .
let @xmath433 and @xmath363 .
since @xmath19 contains @xmath120 it is non - empty , hence a face of @xmath18 .
then @xmath434 is a face of @xmath75 by the first point .
since @xmath432 belongs both to @xmath401 and @xmath434 , it follows that @xmath435 .
[ co : socle - facet ] let @xmath102 be a non - closed non - negative basic presburger set with socle @xmath75 .
let @xmath19 be a facet of @xmath18 with socle @xmath401 .
then @xmath436 or @xmath401 is a facet of @xmath75 . by proposition [ pr : face - socle ]
, @xmath401 is a face of @xmath75 . if @xmath437 then there is a facet @xmath438 of @xmath75 whose closure contains @xmath401 .
it remains to show that @xmath439 .
proposition [ pr : face - socle ] gives a face @xmath440 of @xmath18 with socle @xmath438 .
let @xmath159 , @xmath441 , @xmath331 , @xmath442 be the supports of @xmath18 , @xmath443 , @xmath19 , @xmath440 respectively .
obviously @xmath444 and @xmath445 .
if @xmath446 then @xmath447 hence @xmath448 and thus @xmath439 .
now assume that @xmath449 . since @xmath19 is a facet of @xmath18 it is not smaller than @xmath440 ( with respect to the specialisation order ) hence @xmath450 by proposition [ pr : face - egale - proj]([it : mono - supp ] ) .
on the other hand @xmath451 hence @xmath452 ( otherwise @xmath453 is disjoint from the closure of @xmath454 ) .
altogether this implies that @xmath455 .
in particular @xmath456 hence @xmath439 is a facet of @xmath75 .
[ pr : pres - face ] let @xmath102 be a largely continuous precell mod @xmath13 with @xmath78 .
let @xmath245 be a largely continuous presentation of @xmath18 , @xmath159 a subset of @xmath93 , @xmath118 and @xmath457 .
then @xmath264 if and only if either @xmath411 and @xmath458 on @xmath401 , or @xmath409 and @xmath459 on @xmath401 . in any case @xmath460\big\}.\ ] ] in particular , if @xmath332 is non - empty then it is a largely continuous precell @xmath18 mod @xmath13 and @xmath461 is a presentation of @xmath332 .
[ re : mono - cell ] combining the last point of the above result with remark [ re : sub - mono ] , we get that if @xmath18 is a _
monohedral _ largely continuous precell mod @xmath13 in @xmath12 then so is every face of @xmath18 .
let @xmath75 be the socle of @xmath18 . recall that @xmath462 is a face of @xmath75 and of the socle of @xmath332 by proposition [ pr : face - egale - proj]([it : face - proj ] ) .
let @xmath19 be the set of @xmath463 such that @xmath464 , @xmath465 and @xmath466 $ ] .
non - strict inequalities and congruence relations valid on @xmath18 pass to the limits , hence remain valid on @xmath332 .
so @xmath467 , and if @xmath264 then necessarily one of the two alternatives of the first point hold true .
conversely , take any point @xmath121 and let @xmath468 .
assume first that @xmath411 and @xmath458 .
by proposition [ pr : prol - pi ] , @xmath469 on @xmath75 hence @xmath470 .
if @xmath471 then similarly @xmath472 . otherwise @xmath459 and @xmath472
is obvious . since @xmath473 $ ] it follows in both cases that @xmath199 .
now assume that @xmath409 and @xmath459 .
then @xmath474 , hence obviously @xmath475 and @xmath476 $ ] .
thus @xmath199 , which proves that @xmath477 .
in particular @xmath478 , hence @xmath264 since it contains @xmath19 .
this proves the first point .
moreover by proposition [ pr : face - egale - proj]([it : face - proj ] ) it follows that @xmath479 .
hence @xmath480 , which proves the second point .
in particular @xmath332 is a largely continuous precell if @xmath481 is so .
the remaining of the conclusion then follows by a straightforward induction .
we prove here two technical results ( propositions [ pr : maj - z - aff ] and [ pr : min - z - aff ] ) used in the next section . [ pr : maj - z - aff ] let @xmath102 be a definable set , and @xmath482 be definable maps from @xmath18 to @xmath50 .
assume that the coordinates of all the points of @xmath18 are non - negative .
then there exists a largely continuous , positive , integrally affine map @xmath483 such that @xmath484 on @xmath18 and @xmath485 on @xmath71 .
more precisely @xmath74 can be taken of the form @xmath486 on @xmath18 , for some positive @xmath487 and @xmath488 . without lost of generality
we can assume that @xmath266\!]$ ] .
because of theorem [ th : cluck - piece - lin ] , it suffices to consider the case where the @xmath489 s are affine .
each @xmath489 then can be written as @xmath490 for some @xmath491 for @xmath492 and some @xmath493 .
let @xmath494 be an integer greater than @xmath495 for every @xmath496 , and @xmath497 an element of @xmath51 greater the @xmath498 for every @xmath499 .
for every @xmath278 in @xmath18 and every @xmath496 , since @xmath500 we have @xmath501 .
so the function @xmath502 has all the required properties .
[ le : f - hat - c0bar ] let @xmath196 be a largely continuous precell mod @xmath13 , @xmath75 its socle , @xmath245 a largely continuous presentation of @xmath18 and @xmath74 a largely continuous affine map on @xmath18 such that @xmath485 on @xmath71 .
let @xmath503 and @xmath504 be such that @xmath505 on @xmath18 .
extend @xmath74 to @xmath506 by means of this expression .
for every @xmath507 let @xmath508 if @xmath509 , and @xmath510 otherwise . then @xmath511 is a well - defined largely continuous affine map on @xmath75 with limit @xmath144 at every point of @xmath512 , and @xmath513 for every @xmath121 .
the only possible problem in the definition of @xmath511 is when @xmath514 . but then @xmath509 because otherwise , given any @xmath515 we have @xmath516 and @xmath517 for every @xmath319 close enough to @xmath518 , a contradiction since @xmath485 on @xmath71 .
thus @xmath519 is well - defined in this case too .
let @xmath520 if @xmath509 and @xmath521 otherwise .
then @xmath522 is an affine map and @xmath523 is non - negative on @xmath18 by construction , hence @xmath524 for every @xmath515 there is a point @xmath121 such that @xmath525 and @xmath526 .
so there is a definable function @xmath527 such that @xmath528 and @xmath529 for every @xmath515 .
we have @xmath530 in particular @xmath531 on @xmath75 .
it only remains to check that , given any @xmath532 , @xmath533 tends to @xmath144 when @xmath319 tends to @xmath30 . by the above inequality it suffices to prove that @xmath534 tends to @xmath144 when @xmath319 tends to @xmath30 . since @xmath528 for every @xmath515 and @xmath485 on @xmath71 , it is sufficient to show that @xmath535 tends to a limit @xmath536 as @xmath515 tends to @xmath30
indeed , since @xmath537 we will then have that @xmath538 tends to @xmath539 so the conclusion .
we prove it only when @xmath509 , the case where @xmath540 being similar .
if @xmath541 , then obviously @xmath535 tends to @xmath144 since @xmath542 . if @xmath543 then @xmath544 for every @xmath515 close enough to @xmath30 .
hence @xmath535 , which is the smallest element @xmath545 in @xmath173 such that @xmath546 and @xmath547 $ ] , remains constant too .
in particular it has a limit in @xmath51 as @xmath515 tends to @xmath30 .
[ pr : min - z - aff ] let @xmath102 be a largely continuous precell mod @xmath13 , and @xmath482 be largely continuous affine maps on @xmath18 such that @xmath548 on @xmath71 for every @xmath549 .
then there exists a largely continuous affine map @xmath74 on @xmath18 such that @xmath550 on @xmath71 and @xmath551 for every @xmath552 .
if , moreover , each @xmath489 is positive on @xmath18 then @xmath74 can be chosen positive on @xmath18 .
we can assume that @xmath553 and @xmath554 for every @xmath549 .
by lemma [ le : precomp - borne ] there is @xmath555 such that @xmath556 . given an arbitrary @xmath557 in @xmath50 we are going to show that there exists a largely continuous map @xmath558 such that @xmath485 on @xmath71 and @xmath559 on @xmath18 .
this will prove simultaneously the two statements , because if each @xmath489 is positive then @xmath560 hence taking for example @xmath561 will give that @xmath562 on @xmath18 .
the proof goes , needless to say , by induction on @xmath14 .
if @xmath234 , and more generally if @xmath18 is closed , the constant function @xmath563 has the required properties .
so we can assume that @xmath18 is not closed , @xmath78 and the result is proven for smaller integers .
replacing each @xmath489 by @xmath564 we can assume that @xmath565 .
replacing @xmath566 by a bigger one if necessary we can assume that @xmath567 .
let @xmath568 and @xmath569 such that @xmath570 .
let @xmath571 be defined as in lemma [ le : f - hat - c0bar ] , and @xmath572 .
by lemma [ le : f - hat - c0bar ] the induction hypothesis applies to these functions .
given any @xmath573 , it gives a largely continuous affine map @xmath574 such that @xmath575 on @xmath576 and @xmath577 on @xmath75 for @xmath578 .
we do this for @xmath579 . indeed by lemma [ le :
f - hat - c0bar ] , @xmath580 on @xmath75 for @xmath581 hence @xmath582 .
since @xmath583 , replacing @xmath584 by a bigger one if necessary we can assume that @xmath585 . [ [ case-1 ] ] case 1 : + + + + + + + @xmath586 .
then for @xmath578 the coefficient @xmath587 of @xmath588 in the above expression of @xmath489 is positive ( see the proof of lemma [ le : f - hat - c0bar ] ) , hence @xmath589 and @xmath590 is positive .
let @xmath591 on @xmath18 .
for @xmath578 we have @xmath592 every @xmath593 either belongs to @xmath594 or to @xmath595 .
if @xmath596 then @xmath591 tends to @xmath144 as @xmath223 tends to @xmath120 , because @xmath79 then tends to @xmath432 , @xmath597 tends to @xmath144 and @xmath598 . if @xmath599 then @xmath600 tends to @xmath144 as @xmath121 tends to @xmath120 , because @xmath79 then tends to @xmath432 . hence @xmath43 is largely continuous and @xmath601 on @xmath71 .
[ [ case-2 ] ] case 2 : + + + + + + + @xmath602 .
then every @xmath603 belongs to @xmath595 hence @xmath604 tends to @xmath144 as @xmath121 tends to @xmath120 .
moreover @xmath605 for @xmath578 .
[ [ cases-1-and-2 ] ] cases 1 and 2 : + + + + + + + + + + + + + + in both cases , it remains to modify @xmath43 so that its minimum becomes greater than @xmath584 . by construction @xmath606 on @xmath18 .
recall that @xmath579 and @xmath607 are strictly negative _ rational _ numbers .
thus we can define @xmath608 on @xmath18 .
clearly @xmath74 is a largely continuous affine function on @xmath18 with @xmath485 on @xmath71 , and @xmath609 since @xmath610 and @xmath611 on @xmath18 .
moreover @xmath612 hence for every @xmath121 : @xmath613
the next lemma is the technical heart of this paper . loosely speaking , given a precell @xmath216 , a facet @xmath19 of @xmath18 , a function @xmath614 and a family @xmath23 of monohedral precells covering @xmath19
, we are going to inflate each @xmath26 in @xmath23 to a precell @xmath615 in such a way that : 1 .
@xmath616 is a monohedral precell with facet @xmath26 ; 2 .
the shape of @xmath616 is controlled by @xmath74 , in the sense that the distance to @xmath19 of any point @xmath617 is less than @xmath618 ( where @xmath120 is the projection of @xmath61 onto @xmath19 ) ; 3 .
@xmath616 contains a `` neighbourhood of @xmath26 '' , so to say , in the sense that every point of @xmath18 close enough to @xmath26 belongs to @xmath616 ( the `` close enough '' condition will be controlled by a function @xmath619 ) ; 4 .
the various @xmath616 s do not intersect too much ( in particular if @xmath23 is a partition of @xmath19 , we require the various precells @xmath616 to be pairwise disjoint ) .
in addition we construct simultaneously a family @xmath620 of precells partitioning the complement of @xmath621 in @xmath18 , such that the proper faces of every @xmath622 are proper faces of @xmath18 different from @xmath19 .
in particular @xmath623 has less faces than @xmath18 , which will make possible to repeatedly use the next lemma ( first applied to @xmath18 , then to each @xmath622 ) while proving results by induction on the number of faces .
[ le : face - elargie ] let @xmath102 be a non - closed largely continuous precell mod @xmath13 .
let @xmath19 be a facet of @xmath18 , @xmath159 its support , @xmath614 a definable map .
let @xmath23 be a family of largely continuous monohedral precells mod @xmath13 such that @xmath624 .
then there exists a pair @xmath625 of families of largely continuous precells mod @xmath13 contained in @xmath18 and an integrally affine map @xmath619 such that @xmath620 is a finite partition of @xmath626 , the proper faces of every precell in @xmath620 are proper faces of @xmath18 , and @xmath24 is a family @xmath627 of precells with the following properties : ( fac ) : : @xmath616 has a unique facet which is @xmath26 .
( sub ) : : @xmath628 and @xmath629 .
( sup ) : : @xmath630 and @xmath631 .
( diff ) : : for every @xmath632 , @xmath633 .
[ re : face - elargie ] in every application of lemma [ le : face - elargie ] , @xmath23 will be a partition of @xmath19 .
so the condition ( diff ) simply says that the precells in @xmath24 are pairwise disjoint , hence that @xmath634 is a partition of @xmath18 .
however we can not restrict to this case because it may happen that @xmath23 is a partition of @xmath19 and @xmath635 is not a partition of @xmath434 , which will be crippling when proving the result by induction on @xmath14 . before entering in the somewhat intricate proof of this lemma ,
let us make a few preliminary observations .
[ cl : u - pas - b ] with the notation of lemma [ le : face - elargie ] , @xmath19 is not a face of any @xmath622 . for every @xmath199 there is @xmath636 such that @xmath637 . by ( sup ) every point in @xmath18 such that @xmath123 and @xmath638 belongs to @xmath616 , hence not to @xmath623 .
thus @xmath639 , that is @xmath640 .
[ cl : fac ] let @xmath102 be a non - closed largely continuous precell mod @xmath13 , @xmath19 a facet of @xmath18 , @xmath641 .
let @xmath616 , @xmath26 be any precells mod @xmath13 contained in @xmath18 , @xmath19 respectively , satisfying conditions ( sub ) and ( sup ) of lemma [ le : face - elargie ] for some definable maps @xmath642 .
if @xmath74 is largely continuous and @xmath485 on @xmath643 then property ( fac ) of lemma [ le : face - elargie ] follows : @xmath616 has a unique facet which is @xmath26
. for every @xmath644 and every @xmath645 , @xmath140 hence there exists @xmath121 such that @xmath123 and @xmath646 . by ( sup ) this point @xmath61 belongs to @xmath616 , hence @xmath120 is in the closure of @xmath616 .
so @xmath647 , and conversely ( sub ) implies that @xmath648 , hence @xmath649 by proposition [ pr : face - egale - proj]([it : face - proj ] ) .
assume for a contradiction that @xmath616 has a proper face @xmath650 not contained in @xmath651 .
pick any @xmath652 in @xmath650 . by proposition
[ pr : face - egale - proj]([it : mono - supp ] ) , @xmath331 is not contained in @xmath159 so pick any @xmath653 . by proposition [ pr : face - egale - proj]([it : face - lattice ] ) , @xmath654 hence by the remaining of proposition [ pr : face - egale - proj ] , @xmath655 .
so @xmath656 , hence @xmath74 has limit @xmath144 at @xmath657 .
in particular there is @xmath158 such that for every @xmath199 @xmath658 \rightarrow f(b)>c_k \label{eq : fac1}\ ] ] on the other hand @xmath659 hence there is @xmath617 such that @xmath660 and @xmath661 .
let @xmath468 , then @xmath662 and @xmath663 by ( [ eq : fac1 ] ) this implies that @xmath664 , that is @xmath665 . by ( sub ) it follows that @xmath666 , a contradiction since @xmath667 ( because @xmath668 ) and @xmath669 ( because @xmath670 and @xmath660 .
let @xmath245 be a largely continuous presentation of @xmath18 .
let @xmath75 , @xmath401 be the socles of @xmath18 , @xmath19 respectively , @xmath671 and @xmath672 . since @xmath19 is a facet of @xmath18 , by proposition [ pr : face - socle ]
either @xmath436 and @xmath673 , or @xmath401 is a facet of @xmath18 and either @xmath403 or @xmath405 . for each @xmath636 let @xmath674 be a largely continuous presentation of @xmath26 . if @xmath234 the result is trivially true because there is no non - closed precell contained in @xmath235 .
so we can assume that @xmath78 and the result is proven for smaller integers .
if @xmath675 then @xmath676 can be identified with @xmath75 , after which the result follows by the induction hypothesis .
so we can assume that @xmath677 .
proposition [ pr : maj - z - aff ] gives positive @xmath487 and @xmath488 such that @xmath678 on @xmath19 .
without loss of generality we can assume that equality holds on @xmath19 , and we still denote by @xmath74 the corresponding extension of @xmath74 to @xmath135 .
in particular @xmath74 is now largely continuous on @xmath19 with @xmath679 on @xmath680 .
it is sufficient to build a pair @xmath681 of families of largely continuous precells mod @xmath13 contained in @xmath18 and a definable map @xmath682 such that @xmath620 is a finite partition of @xmath626 , that the proper faces of every precell in @xmath620 are proper faces of @xmath18 , and that for each @xmath26 in @xmath23 we have : * @xmath683 ; * @xmath684 on @xmath616 ; * @xmath630 and @xmath685 ; * @xmath686 is disjoint from @xmath687 , for every @xmath632 . indeed , we do not need to require that @xmath688 is integrally affine , because if property ( sup ) holds for a definable map @xmath682 it will then hold for every larger map , and proposition [ pr : maj - z - aff ] provides an integrally affine one . then by ( fac ) , properties ( sub ) and ( diff ) will follow from ( sub ) and ( diff ) . because @xmath485 on @xmath680 , ( fac ) will then follow from ( sup ) and ( sub ) by claim [ cl : fac ] .
we have to distinguish several cases .
for the convenience of the reader most of them are accompanied by a figure representing ( very approximatively ) the general idea of the construction when @xmath689 . in these figures ,
each @xmath690 with positive coordinates takes place at the point of coordinates @xmath691 in the figure , so that the order is preserved .
therefore the set of positive points of @xmath692 is represented by a square whose bottom , left , right and top edges represent @xmath693 , @xmath694 , @xmath695 and @xmath696 respectively in the induction steps , while the left edge is @xmath173 . ] .
a side effect of this compactification is that linear functions are represented by curved lines .
the various precells involved will be represented in this square by gray areas whose union is @xmath18 ( or the auxiliary precell @xmath697 in figure [ fi : div-3-right ] ) .
their socles will take place in the bottom @xmath693 , as we identify it with @xmath173 .
finally @xmath19 will be represented by a thick edge or a corner of the square , depending on the cases .
[ [ case-1 - 1 ] ] case 1 : + + + + + + + @xmath436 . + then @xmath673 hence @xmath514 and @xmath698 , thus @xmath699 and @xmath700 for every @xmath121
. so @xmath19 identifies to @xmath75 and @xmath23 to @xmath635 . roughly speaking
, we are going on one hand to split @xmath18 in two parts by means of a function @xmath701 to be defined such that @xmath702 and @xmath703 , and on the other hand to lift the family @xmath635 of @xmath75 to a family of precells covering of the upper part of @xmath18 .
this will give us @xmath24 .
as figure [ fi : div-1 ] suggests , the lower part of @xmath18 will remain unchanged and give @xmath620 .
# 1plot[domain=#1 ] ( , 2.5*((+2)/12)^2 + 7.5 ) # 1plot[domain=#1 ] ( , 6*((+1)/11)^2 + 4 ) \(a ) at ( 0,7.5694444444444 ) ; ( b ) at ( 2.5,7.85156250000000 ) ; ( c ) at ( 6,8.611111111111111 ) ; ( 0,10 ) cycle ; ( 2.5,10 ) ( 0,10 ) cycle ; ( 6,10 ) ( 2.5,10 ) cycle ; ( 6,10 ) cycle ; ; ; ( 0,0 ) ( a ) ( 2.5,0 ) ( b ) ( 6,0 ) ( c ) ( 10,0 ) ( 10,10 ) ; ( a )
( 0,10 ) ( b ) ( 2.5,10 ) ( c ) ( 6,10 ) ; ( 0,10 )
( 10,10 ) ; ( 0,10 ) node[above]@xmath704 ( 2.5,10 ) ; ( 2.5,10 ) node[above]@xmath705 ( 6,10 ) ; ( 6,10 ) node[above]@xmath706 ( 10,10 ) ; ( 0,0 )
node[above]@xmath707 ( 2.5,0 ) ; ( 2.5,0 ) node[above]@xmath708 ( 6,0 ) ; ( 6,0 ) node[above]@xmath709 ( 10,0 ) ; ( 0,7.4 ) .. controls + ( 5,-.5 ) and + ( -6,-2.8 ) .. ( 10,10 ) ; ( 2,6 ) node @xmath623 ( 1.25,8.9 ) node @xmath710 ( 4.25,9 ) node @xmath711 ( 6.9,9.45 ) node @xmath712 ( 5,5.2 ) node @xmath243 ( 5,7.1 ) node @xmath74 ( 3.65,7.7 ) node @xmath701 ; let us check the details now .
proposition [ pr : maj - z - aff ] gives a largely continuous affine function @xmath713 such that @xmath714 on @xmath75 .
let @xmath623 be the set of @xmath715 such that @xmath716 , @xmath717 and @xmath466 $ ] .
it is clearly a largely continuous precell mod @xmath13 ( with socle @xmath75 since @xmath718 ) . for each @xmath636 let @xmath616 be the set of @xmath240 such that @xmath719 , @xmath720 and @xmath466 $ ]
this is a largely continuous precell mod @xmath13 with socle @xmath721 .
let @xmath722 and @xmath723 .
obviously @xmath724 , @xmath725 and ( fac ) , ( diff ) hold for every @xmath636 .
by construction , for every @xmath636 and every @xmath617 we have @xmath726 which proves ( sub ) .
further , for every @xmath121 such that @xmath727 and @xmath728 , we have @xmath719 , @xmath729 and @xmath466 $ ] hence @xmath617 .
this is property ( sup ) with @xmath730 on @xmath19 .
[ [ case-2 - 1 ] ] case 2 : + + + + + + + @xmath401 is a facet of @xmath75 and @xmath403 .
+ then @xmath731 , @xmath732 on @xmath401 ( otherwise by proposition [ pr : pres - face ] , @xmath733 is a proper face of @xmath18 larger than @xmath19 ) and @xmath734 ( otherwise @xmath594 is a proper face of @xmath18 larger than @xmath19 ) .
in particular @xmath735 for every @xmath736 and every @xmath515 close enough to @xmath30 , so there is a definable map @xmath737 such that for every @xmath515 @xmath738 in the precells @xmath616 that we are looking for , we want to have @xmath739 in order to get condition ( sub ) .
the idea is then to inflate first @xmath635 in a way controlled by @xmath740 ( using the induction hypothesis ) , and then to divide @xmath18 by lifting this division of its socle @xmath75 ( see figure [ fi : div-2 ] ) .
# 1plot[domain=#1 ] ( , 2.5*((+2)/12)^2 + 7.5 ) # 1plot[domain=#1 ] ( , 6*((+1)/11)^2 + 4 ) ( an ) at ( 0,7.5694444444444 ) ; ( bn ) at ( 2.5,7.8515625000000 ) ; ( cn ) at ( 6,8.61111111111111 ) ; ( am ) at ( 0,4.0495867768595 ) ; ( bm ) at ( 2.5,4.6074380165289 ) ; ( cm ) at ( 6,6.4297520661157 ) ; plot[domain=0:2.5 ] ( , 2.5*((+2)/12)^2 + 7.5 ) plot[domain=2.5:0 ] ( , 6*((+1)/11)^2 + 4 ) cycle ; ; plot[domain=2.5:6 ] ( , 2.5*((+2)/12)^2 + 7.5 ) plot[domain=6:2.5 ] ( , 6*((+1)/11)^2 + 4 ) cycle ; ; plot[domain=6:10 ] ( , 2.5*((+2)/12)^2 + 7.5 ) plot[domain=10:6 ] ( , 6*((+1)/11)^2 + 4 ) cycle ; ; ( 0,0 ) ( am ) ( 2.5,0 ) ( bm ) ( 6,0 )
( cm ) ( 10,0 ) ( 10,10 ) ; ( am ) ( an ) ( bm ) ( bn ) ( cm )
( cn ) ; ( an ) ( 0,10 ) ( 10,10 ) ; ; ; ( 10,10 ) node @xmath33 node[above ] @xmath26 ; ( 10,0 ) node @xmath33 node[above ] @xmath721 ; ( 0,0 )
node[above]@xmath741 ( 2.5,0 ) ; ( 2.5,0 ) node[above]@xmath742 ( 6,0 ) ; ( 6,0 ) node[above]@xmath743 ( 9.9,0 ) ; ( 10.4,10 )
node[right ] @xmath74 ( 10.4,6 ) ; ( 5.9,6 ) ( 10.5,6 ) ; ( 1.25,6 ) node @xmath744 ( 4.25,6.7 ) node @xmath745 ( 7.3,8.3 ) node @xmath616 ( 5,5.2 ) node @xmath243 ( 5,8.7 ) node @xmath244 ; the induction hypothesis applies to @xmath75 , @xmath401 , @xmath635 and @xmath746 on @xmath401 .
it gives a definable map @xmath747 and a pair @xmath748 of families of precells . for each @xmath749 ( resp .
@xmath636 ) let @xmath744 ( resp .
@xmath616 ) be the set of @xmath463 such that @xmath750 ( resp .
@xmath79 belongs to the unique precell @xmath751 whose facet is @xmath721 ) , @xmath752 and @xmath466 $ ] .
this is obviously a largely continuous precell mod @xmath13 with socle @xmath741 ( resp .
@xmath753 ) , and exactly the set of @xmath121 such that @xmath750 ) ( resp .
@xmath753 ) . in particular it is contained in @xmath18 , and
if we let @xmath754 and @xmath722 then @xmath620 is a partition @xmath626 by induction hypothesis on @xmath748 . for every @xmath749 ,
every proper face of @xmath741 is a proper face @xmath755 of @xmath75 .
let @xmath331 be its support .
then by proposition [ pr : pres - face ] , @xmath756 is a presentation of @xmath757 , but also of @xmath333 hence @xmath758 is a proper face of @xmath18 .
let us check ( fac ) and ( diff ) .
for every @xmath636 , since @xmath732 on @xmath401 we have @xmath759 by proposition [ pr : pres - face ] , hence @xmath683 by proposition [ pr : face - egale - proj ] .
moreover for every @xmath760 , @xmath761 by induction hypothesis hence @xmath762 \times \{+\infty\ } \subseteq ( \widehat{d}\setminus\widehat{e } ) \times \{+\infty\ } = d\setminus e.\ ] ] now we turn to ( sub ) . for every @xmath617 , since @xmath731 we have @xmath763 and @xmath764 . by the induction hypothesis @xmath765 and @xmath766 because @xmath767 . the first inequality implies that @xmath768 by ( [ eq : le - fa - eta2 ] ) .
together with the second inequality this gives that @xmath769 .
that is @xmath770 .
we finally check ( sup ) with @xmath771 on @xmath19 . since @xmath616 is clearly the set of @xmath121 such that @xmath767 , for every @xmath121 such that @xmath727 ( hence @xmath772 ) and @xmath773 we have @xmath774 by induction hypothesis on @xmath32 and @xmath721 hence @xmath617 .
[ [ case-3 ] ] case 3 : + + + + + + + @xmath401 is a facet of @xmath75 and @xmath405 .
+ then @xmath775 , hence @xmath458 on @xmath401 , @xmath776 for every @xmath636 , and for every @xmath121 : @xmath777 note that @xmath778 for every @xmath636 because , given any @xmath779 , we have @xmath780 and on one hand @xmath781 $ ] , on the other hand @xmath782 $ ] ( using the presentation of @xmath363 given by proposition [ pr : pres - face ] ) . [ [ sub - case-3.1 ] ] _ sub - case 3.1 : _ + + + + + + + + + + + + + + + @xmath783 .
+ this is the most difficult case , because @xmath618 depends both on @xmath432 and @xmath784 .
therefore our construction is done in two steps .
+ _ step 1 _ : intuitively , we are going to remove the top of @xmath18 by introducing a function @xmath785 which will ensure that @xmath597 does nt grow too fast as @xmath121 goes closer to @xmath19 .
the connection between @xmath786 ( a function of @xmath507 ) and @xmath74 ( a function of @xmath199 ) will be made _ via _ an intermediate function @xmath787 defined below .
we need to restrict the socle @xmath75 of @xmath18 to a domain @xmath788 close enough to @xmath401 so as to ensure at least that @xmath789 on @xmath788 . in order to do this we will divide @xmath75 by applying to it the induction hypothesis .
the resulting partition of @xmath75 together with @xmath786 will give us a partition of @xmath18 which might look like figure [ fi : div-3-top ] .
( 0,0 ) rectangle ( 10,10 ) ; # 1plot[domain=#1 ] ( , 6.5*((+3)/13)^3 + 3.5 ) ( 10,2.5 ) ( 7,2.5 ) cycle ; ( 7,5.5 ) cycle ; ( 7,2.5 ) ( 4,2.5 ) cycle ; ( 4,2.5 ) ( 0,2.5 ) cycle ; ( 10.03,2.5 ) node[right]@xmath19 ( 10.03,10 ) ; ( 0,0 ) ( 0,2.5 ) ( 4,0 ) ( 4,2.5 ) ( 7,0 ) ( 7,2.5 ) ; ; plot[domain=0:10 ] ( , 2.5 ) ; ( 4,2.5 ) ( 4,4.5147928994083 ) ( 7,2.5 ) ( 7,6.4585798816568 ) ( 0,2.5 ) ( 0,3.5798816568047 ) ; ; ( 0,0 ) node[above]@xmath790 ( 4,0 ) ; ( 4,0 ) node[above]@xmath791 ( 7,0 ) ; ( 7,0 ) node[above]@xmath788 ( 10,0 ) ; ( 7.6,6.4 ) node@xmath792 ; ( 9,3.4 ) node@xmath697 ; ( 5.5,3.15 ) node@xmath793 ; ( 2,3.15 ) node@xmath794 ; ( 8,4.6 ) node[right ] @xmath74 ; ( 9.2,6.5 ) node@xmath786 ; ( 5.5,2.1 ) node@xmath243 ; ( 5.5 , 5.8 ) node@xmath244 ; let @xmath795 be a positive affine map given by proposition [ pr : maj - z - aff ] such that @xmath796 on @xmath401 and @xmath575 on @xmath797
. given any @xmath798 , since @xmath799 and @xmath800 has limit @xmath144 at @xmath30 , we have @xmath801 for every @xmath515 close enough to @xmath30 .
so there is a definable function @xmath802 such that for every @xmath515 @xmath803 the induction hypothesis applies to @xmath75 , @xmath401 , @xmath804 and @xmath805 .
it gives a definable map @xmath806 and a pair @xmath807 of families of precells . in the present case @xmath808 consists of a single largely continuous precell @xmath788 mod @xmath13 contained in @xmath75 , such that @xmath809 on @xmath788 , and every @xmath515 such that @xmath810 and @xmath811 belongs to @xmath788 .
the family @xmath812 is a finite partition of @xmath813 in largely continuous precells mod @xmath13 .
let @xmath814 where @xmath815 for every @xmath749 .
since @xmath734 , the proper faces of @xmath744 are proper faces of @xmath18 by claim [ cl : u - pas - b ] . for every @xmath816 and every @xmath817
, we have @xmath818 because @xmath816 , and @xmath819 on @xmath788 by the induction hypothesis . thus on one hand @xmath820 , and on the other hand @xmath821 .
in particular @xmath822 is a largely continuous positive affine function on @xmath788 with limit @xmath144 at every point of @xmath823 .
we also have @xmath824 by induction hypothesis , hence @xmath825 by ( [ eq : le - fa - eta31 ] ) .
in particular the restriction of @xmath826 to @xmath788 is a positive affine function with limit @xmath144 at every point of @xmath823 .
proposition [ pr : min - z - aff ] then gives a largely continuous positive affine function @xmath827 such that @xmath828 on @xmath823 , @xmath829 on @xmath788 and @xmath830 for every @xmath816 .
let us quote for further use that in particular @xmath831 note that @xmath832 because @xmath788 has a unique facet which is @xmath401 by claim [ cl : fac ] , hence @xmath828 on @xmath833 .
let @xmath46 an integer such that @xmath834 is integrally affine , so that @xmath835 if and only if @xmath836 for every @xmath837 .
let @xmath838 on @xmath788 , and @xmath792 ( resp .
@xmath697 ) be the set of @xmath240 such that @xmath839 , @xmath840 ( resp .
@xmath841 ) and @xmath466 $ ] . by construction @xmath786
is a largely continuous affine map on @xmath788 with @xmath842 on @xmath823 .
moreover on @xmath788 we have @xmath843 ( because @xmath829 by construction ) hence the socle of @xmath792 is @xmath788 .
obviously @xmath844 ( because @xmath845 by construction ) hence the socle of @xmath697 is @xmath788 .
thus both @xmath792 and @xmath697 are largely continuous precells mod @xmath13 contained in @xmath846 .
moreover @xmath847 if and only if @xmath848 .
thus @xmath792 and @xmath697 form a partition of @xmath849 , or equivalently @xmath850 is a partition of @xmath851 . since @xmath842 on @xmath832 , by proposition [ pr : pres - face ] every proper face @xmath852 of @xmath792
is of type @xmath853 for @xmath755 a face of @xmath401 . in particular
@xmath852 is a proper face of @xmath18 .
+ : intuitively , we are going to build the @xmath616 s by inflating inside @xmath697 each @xmath26 in @xmath23 as suggested by figure [ fi : div-3-right ] ( which zooms in on @xmath697 , the other parts of @xmath18 remaining as in figure [ fi : div-3-top ] ) .
( 8.7,6 ) rectangle ( 10,10 ) ; ( 7,2.5 ) ( 10,2.5 ) cycle ; ( 8.7,4 ) rectangle ( 10,6 ) ; ( 8.7,2.5 ) rectangle ( 10,4 ) ; ( 7.8,0 ) rectangle ( 8.7,10 ) ; ( 7,2.5 ) ( 10,2.5 ) cycle ; ( 7,0 ) rectangle ( 7.8,10 ) ; ( 7,2.5 ) ( 10,2.5 ) cycle ; ( 7,0 ) ( 7,2.5 ) ( 7.8,0 ) ( 7.8,2.5 ) ( 8.7,0 ) ( 8.7,2.5 ) ( 10,0 ) ( 10,2.5 ) ; ( 7,2.5 ) ( 7,5.5 ) ( 7.8,2.5 ) ( 7.8,5.82 ) ( 8.7,2.5 ) ( 8.7,6.45 ) ( 8.7,4 ) ( 10,4 ) + + ( 0.1,0 ) node[right ] @xmath854 ( 8.7,6 ) ( 10,6 ) + + ( 0.1,0 ) node[right ] @xmath855 ; ( 7,10 ) ( 10,10 ) ; ( 6.4,10 ) ( 7,10 ) ; ( 6.4,5.4 ) .. controls + ( .5,.1 ) and + ( -.3,-.1 ) .. ( 7,5.5 ) ; ( 6.4,2.5 ) ( 7,2.5 ) ; ( 6.4,0 ) ( 7,0 ) ; ( 10.03,2.5 )
( 10.03,10 ) ; ( 10.03,2.5 )
node[right ] @xmath706 ( 10.03,4 ) ; ( 10.03,4 )
node[right ] @xmath705 ( 10.03,6 ) ; ( 10.03,6 )
node[right ] @xmath704 ( 10.03,10 ) ; ; plot[domain=7:10 ] ( , 2.5 ) ; ; ( 7,0 ) node[above]@xmath856 ( 7.8,0 ) ; ( 7.8,0 ) node[above]@xmath857 ( 8.7,0 ) ; ( 8.7,0 ) node[above]@xmath9 ( 10,0 ) ; ( 7.4,3.6 ) node@xmath858 ( 8.35,3.5 ) node@xmath859 ( 9.5,3.3 ) node@xmath712 ( 9.5,5 ) node@xmath711 ( 9.5,6.6 ) node@xmath710 ( 8.8,7.5 ) node@xmath74 ( 7.5,6 ) node@xmath786 ( 8.5,2 ) node@xmath243 ; for every @xmath636 let @xmath860 if @xmath861 and @xmath862 otherwise . since @xmath842 on @xmath401 there is a definable function @xmath863 such that for every @xmath817 and every @xmath636 such that @xmath864 we have @xmath865 the induction hypothesis applies to @xmath788 , @xmath401 , @xmath635 and @xmath866 .
it gives a definable map @xmath867 and a pair @xmath868 of families of precells . for each @xmath869
let @xmath870 . clearly the family @xmath871 is a finite partition in largely continuous precells mod @xmath13 of the complement in @xmath697 of the set @xmath872 .
equivalently , @xmath873 is a finite partition of @xmath874 .
since @xmath734 , by claim [ cl : u - pas - b ] the proper faces of @xmath744 are proper faces of @xmath18 for every @xmath869 .
for each @xmath636 let @xmath753 be the precell in @xmath875 given by the induction hypothesis , so that conditions ( fac ) , ( sub ) , ( sup ) , ( diff ) apply to @xmath753 , @xmath866 and @xmath876 . if @xmath877 ( resp .
@xmath861 ) let @xmath616 be the set of @xmath240 such that @xmath767 , @xmath878 ( resp .
@xmath879 ) and @xmath466 $ ] . let us check that @xmath616 is a largely continuous precell mod @xmath13 .
for every @xmath880 we have @xmath864 , because @xmath881 by ( fac ) , and @xmath882 by proposition [ pr : face - egale - proj]([it : face - proj ] ) .
so there is @xmath644 such that @xmath883 , @xmath884 and @xmath782 $ ] .
we can ( and do ) require in addition that @xmath885 , hence @xmath886 .
because @xmath880 we also have @xmath887 by ( sub ) , hence @xmath888 by ( [ eq : le - fa - eta32 ] ) .
altogether this proves that @xmath889 , hence @xmath278 belongs to the socle of @xmath616 .
so the socle of @xmath616 is exactly @xmath753 and @xmath616 is then a largely continuous precell mod @xmath13 .
now we turn to ( fac ) .
the presentation of @xmath890 given by proposition [ pr : pres - face ] is exactly @xmath891 , hence @xmath892 since @xmath893 .
in particular @xmath683 by proposition [ pr : face - egale - proj]([it : face - proj ] ) .
more precisely , the above computations show that we have @xmath894 let @xmath722 and @xmath895 .
we already know that @xmath620 is a finite partition of @xmath874 in largely continuous precells mod @xmath13 whose proper faces are proper faces of @xmath18 , and that each @xmath896 is a largely continuous precell mod @xmath13 contained in @xmath897 with socle @xmath753 and @xmath649 .
let us check that @xmath898 .
in order to do so , we are claiming that @xmath899 assume the contrary and let @xmath900 , @xmath760 be such that @xmath901 and @xmath902 . by ( [ eq : le - fa - cd ] ) this implies that @xmath903 .
but the socle of @xmath897 is @xmath904 , hence @xmath767 for some @xmath636 .
thus @xmath905 belongs to @xmath906 . by the induction hypothesis
the latter is contained in @xmath907 , hence @xmath908 .
but @xmath905 is also the socle of @xmath133 . since @xmath901 it follows that @xmath909 , a contradiction .
that @xmath910 then follows immediately from ( [ eq : le - fa - claim ] ) and the fact that @xmath911 .
so @xmath912 and it only remains to check ( sub ) , ( sup ) and ( diff ) for any fixed @xmath636 .
we start with ( diff ) .
pick any @xmath632 , assume that there is a point @xmath120 in @xmath686 which belongs to @xmath687 .
then @xmath468 for some @xmath913 .
we have @xmath900 and @xmath902 , hence @xmath914 by ( [ eq : le - fa - claim ] ) , that is @xmath915 , a contradiction .
hence @xmath686 is disjoint from @xmath687 .
let us now turn to ( sup ) . for every @xmath199 , since @xmath842 on @xmath916 and @xmath917
, we have @xmath918 whenever @xmath919 is close enough to @xmath432 ( that is whenever @xmath920 and @xmath921 is large enough ) .
so there is a definable function @xmath922 such that for every @xmath923 @xmath924 let @xmath925 . for every @xmath121 such that @xmath727 and @xmath638 , since @xmath926 by ( [ eq : delta - j ] ) we have in particular @xmath927 and @xmath928 , hence @xmath839 by construction .
so @xmath923 and @xmath929 , which implies that @xmath930 by ( [ eq : le - fa - eta33 ] ) , hence @xmath931 by construction .
on the other hand , since @xmath839 , @xmath772 and @xmath932 , we get that @xmath767 by construction . in particular @xmath933 ,
hence @xmath900 since @xmath934 .
altogether we have @xmath900 , @xmath767 and @xmath727 hence that @xmath617 by ( [ eq : le - fa - cd ] ) , which proves ( sup ) .
it only remains to check ( sub ) , that is @xmath684 on @xmath616 .
this is the moment to recall ( [ eq : le - fa - lambda ] ) , which says that @xmath935 on @xmath788 .
recall also that @xmath936 on @xmath401 by definition of @xmath787 .
thus on @xmath788 we have @xmath937 for every @xmath617 , @xmath839 and @xmath931 hence @xmath938 .
we also have @xmath939 by proposition [ pr : prol - pi ] . combining all this with ( [ eq : le - fa - fin - lambda ] ) we get that @xmath940 since @xmath941 by the definition of @xmath74 , and @xmath926 by ( [ eq : delta - j ] ) , we finally get from ( [ eq : le - fa - fin - am ] ) that @xmath942 . [ [ sub - case-3.2 ] ] _ sub - case 3.2 : _ + + + + + + + + + + + + + + + @xmath514 . +
this final case is easy : we simply divide @xmath18 in two pieces , above and below a function @xmath701 to be defined , so that the previous sub - case 3.2 applies to the lower part of @xmath18 .
the upper part @xmath18 does nt require any special treatment : it will simply be incorporated in the family @xmath620 .
let us check the details now .
proposition [ pr : maj - z - aff ] gives a largely continuous integrally affine map @xmath701 on @xmath75 such that @xmath828 on @xmath576 and @xmath943 .
let @xmath944 ( resp .
@xmath945 ) be the set of @xmath240 such that @xmath716 , @xmath946 ( resp .
@xmath720 ) and @xmath466 $ ] .
its socle is @xmath75 ( for @xmath944 we use that @xmath943 ) hence it is a largely continuous precell mod @xmath13 .
since @xmath701 takes values in @xmath51 , @xmath944 and @xmath945 form a partition of @xmath18 .
the presentation of the faces of @xmath18 , @xmath944 @xmath945 given by proposition [ pr : pres - face ] gives that every proper face of @xmath944 and @xmath945 is a proper face of @xmath18 , and @xmath19 is a face of @xmath944 .
the previous sub - case 3.1 applies to @xmath944 , @xmath19 , @xmath23 and @xmath74 .
it gives a pair @xmath947 of families of largely continuous precells mod @xmath13 and an integrally affine map @xmath948 .
then @xmath949 and @xmath950 have all the required properties for @xmath18 , @xmath23 and @xmath74 , except possibly ( sup ) .
we remedy this by replacing @xmath950 by a larger function @xmath688 defined as follows . for every @xmath199
, we have @xmath951 for every @xmath515 close enough to @xmath432 since @xmath828 on @xmath401 .
so there is a definable function @xmath952 such that for every @xmath121 @xmath953 let @xmath954 , then for every @xmath636 and every @xmath121 such that @xmath727 and @xmath955 we have in particular @xmath956 hence @xmath957 by ( [ eq : le - fa - eta - final ] ) , that is @xmath958 .
moreover we have @xmath727 and @xmath959 .
altogether this implies that @xmath61 belongs to @xmath960 , which in turn proves ( sup ) .
[ th : mono - div ] let @xmath102 be a largely continuous precell mod @xmath13 , @xmath961 a definable function , and @xmath23 a complex of monohedral largely continuous precells mod @xmath13 such that @xmath962 .
then there exists a finite partition @xmath24 of @xmath18 in monohedral largely continuous precells mod @xmath13 such that @xmath963 is a closed complex , @xmath24 contains for every @xmath636 a unique precell @xmath25 with facet @xmath26 , and moreover @xmath964 on @xmath25 where @xmath965 .
the proof goes by induction on the number @xmath45 of proper faces of @xmath18 . if @xmath966 then @xmath967 and @xmath18 is monohedral , hence @xmath968 gives the conclusion .
so let us assume that @xmath46 and the result is proved for smaller integers .
let @xmath19 be a facet of @xmath18 .
lemma [ le : face - elargie ] applied to @xmath18 , @xmath19 , @xmath23 and the restriction of @xmath74 to @xmath19 gives a pair @xmath969 of families of precells .
for every @xmath622 , the proper faces of @xmath623 are proper faces of @xmath18 .
so the family @xmath970 is a complex and @xmath971 .
since @xmath19 is not a proper face of @xmath623 by claim [ cl : u - pas - b ] , the induction hypothesis applies to @xmath623 , @xmath972 and the restriction of @xmath74 to @xmath973 .
it gives a family @xmath974 of precells .
let @xmath24 be the union of @xmath975 and @xmath974 for @xmath622 .
this is a family of largely continuous precells mod @xmath13 partitioning @xmath18 . by construction @xmath24 contains for every @xmath636 a unique precell @xmath25 with facet @xmath26 , and @xmath684 on @xmath25 with @xmath965 . in particular
@xmath963 is a partition of @xmath67 which contains the faces of all its members , since @xmath23 is a closed complex ( because @xmath23 is a complex and @xmath976 is closed ) .
so @xmath963 is a closed complex .
[ th : mono - dec ] let @xmath102 be a largely continuous precell mod @xmath13 .
then there exists a complex @xmath24 of monohedral largely continuous precells mod @xmath13 such that @xmath977 .
we are going to show that given any closed complex @xmath83 of largely continuous precells mod @xmath13 in @xmath12 , there is a closed complex @xmath24 of largely continuous _ monohedral _ precells mod @xmath13 such that @xmath978 and @xmath24 refines @xmath83 ( that is every @xmath28 is contained in some @xmath979 ) .
the conclusion for @xmath18 will follow , by applying this to the closed complex consisting of all the faces of @xmath18 .
the proof goes by induction on the cardinality @xmath45 of @xmath83 . if @xmath966 then @xmath980 proves the result .
assume that @xmath46 and the result is proved for smaller integers .
let @xmath18 be a maximal element of @xmath83 with respect to specialisation , and @xmath981 .
by maximality of @xmath18 , @xmath982 is again a closed complex .
the induction hypothesis gives a closed complex @xmath23 of largely continuous monohedral precells mod @xmath13 such that @xmath983 and @xmath23 refines @xmath982 . if @xmath18 is closed then obviously @xmath984 proves the result for @xmath83 .
otherwise let @xmath985 .
the monohedral division theorem [ th : mono - div ] applied to @xmath18 , @xmath986 and the constant function @xmath987 gives a finite partition @xmath988 of @xmath18 in monohedral largely continuous precells mod @xmath13 such that the family @xmath989 is a closed complex .
the family @xmath990 is a partition of @xmath991 .
since @xmath23 is a closed complex and every precell in @xmath988 has a unique facet which belongs to @xmath23 , it follows that @xmath24 is a complex .
we finish this section with another , much more elementary , division result .
contrary to the above ones , it is drastically different from what occurs in the real situation , where the polytopes are connected sets
. [ pr : split - mono ] let @xmath102 be a non - closed monohedral largely continuous precell mod @xmath13 .
for every integer @xmath46 there exists for some @xmath992 a partition @xmath993 of @xmath18 in largely continuous precells mod @xmath994 such that @xmath995 for @xmath996 .
the proof goes by induction on @xmath14 .
the result is trivially true for @xmath234 since there is no non - closed precell in @xmath235 .
assume that @xmath78 and the result is proved for smaller integers .
let @xmath245 be a presentation of @xmath18 .
by induction hypothesis we can assume that @xmath997 hence @xmath677 . if @xmath514 , for @xmath996 let @xmath998 be the set of @xmath240 such that @xmath241 , @xmath999 and @xmath1000 $ ] .
this is obviously a partition of @xmath18 in largely continuous precells mod @xmath1001 having the same boundaries as @xmath18 . on the other hand ,
if @xmath1002 then @xmath80 is not closed ( otherwise @xmath18 would be closed ) hence the induction hypothesis gives for some @xmath1003 a partition @xmath1004 of @xmath80 in largely continuous precells mod @xmath1005 such that @xmath1006 for every @xmath86 .
let @xmath1007 for every @xmath86 .
then @xmath993 is easily seen to give the conclusion , thanks to the description of the faces of @xmath18 and @xmath998 given by proposition [ pr : pres - face ] .
recall that @xmath7 is a @xmath0-adically closed field , @xmath3 its @xmath0-valuation , @xmath4 its valuation ring and @xmath5 .
we still denote by @xmath3 the map @xmath1008 from @xmath16 to @xmath12 .
we are going to define polytopes - adic cells as defined in @xcite . ]
mod @xmath13 in @xmath16 by means of the inverse image by @xmath3 of largely continuous precells mod @xmath13 in @xmath12 .
however , the @xmath0-adic triangulation theorem that we are aiming at requires a more versatile definition .
it involves semi - algebraic subgroups @xmath1009 of the multiplicative group @xmath1010 , where @xmath1011 is a positive integer . in the special case where @xmath7 is a finite extension of @xmath49
, we have @xmath1012 where @xmath1013 is any generator of the maximal ideal of @xmath4 .
since in this paper we will only use that @xmath1014 , we refer the reader to @xcite for a general definition of @xmath1015 for every integers @xmath1016 in arbitrary @xmath0-adically closed fields .
we let @xmath1017 .
given an @xmath14-tuple @xmath233 we call a set @xmath1018 a * * polytope mod @xmath13 in @xmath1019 if @xmath1020 is a largely continuous precell mod @xmath13 in @xmath12 and @xmath1021 . the * * faces and * * facets @xmath1022 of a subset @xmath9 of @xmath1019 are defined as the inverse images , by the restriction of @xmath3 to @xmath1019 , of the faces and facets of @xmath1020
. the * * support of @xmath9 ( resp . of @xmath1023 ) is the support of @xmath1020 ( resp . of @xmath1024 ) , so that : @xmath1025\!]{\mathop{:}}x_i\neq0\big\}\ ] ] @xmath1026
we say that @xmath9 is * * monohedral if @xmath1020 is so , that is if the faces of @xmath9 are linearly ordered by specialisation , in which case we call @xmath9 a * * monotope mod @xmath13 in @xmath1019 .
a family @xmath24 of polytopes mod @xmath13 in @xmath1019 is a * * complex if it is finite and for every @xmath1027 , @xmath1028 is the union of the common faces of @xmath9 and @xmath1029 .
it is a * * closed complex if moreover it contains all the faces of its members .
every complex @xmath100 of polytopes mod @xmath13 is contained in a smallest closed complex , namely the family of all the faces of the members of @xmath100 .
we call it the * * closure of @xmath100 and denote it @xmath1030 .
more generally , all the points of proposition [ pr : face - egale - proj ] , as well as proposition [ pr : face - socle ] , corollary [ co : socle - facet ] , the monohedral decomposition ( theorem [ th : mono - dec ] ) and proposition [ pr : split - mono ] immediately transfer to polytopes mod @xmath13 in @xmath1019 . only the monohedral division ( theorem [ th : mono - div ] ) requires a bit more of preparation .
for the sake of generality we want the @xmath0-adic analogon of the monohedral division theorem in @xmath12 to hold not only with a map @xmath1037 definable in the language of rings ( _ i.e. _ semi - algebraic ) but also with a map definable in various expansions @xmath1038 of the ring structure of @xmath7 .
the proof of theorem [ th : div - p - adic ] below shows that it suffices to make the following assumptions on @xmath1038 : ( ext ) : : for every definable function @xmath1039 , if @xmath74 is continuous and @xmath75 is closed and bounded , then @xmath1040 takes a maximum value at some point @xmath515 . ( pres ) : : the image by the valuation of every subset of @xmath16 definable in @xmath1038 , is @xmath47-definable .
[ re : p - min ] if @xmath7 is a finite extension of @xmath49 then condition ( ext ) holds for every continuous function by the extreme value theorem . but
this condition , when restricted to definable continuous functions , is preserved by elementary equivalence .
hence it will be satisfied whenever the complete theory of @xmath1038 has a @xmath0-adic model ( that is a model whose underlying field is a finite extension of @xmath49 ) . on the other hand , if @xmath1038 is @xmath1041-minimal ( see @xcite ) , theorem 6 in @xcite proves that condition ( pres ) is satisfied .
in particular theorem [ th : div - p - adic ] applies for example to every subanalytic map @xmath32 , and more generally to every map @xmath32 which is definable in a @xmath1041-minimal structure @xmath1038 which has a @xmath0-adic model . for every @xmath1023 we let
@xmath1042 . if @xmath1043 this is the valuative counterpart of the usual norm on @xmath16 , which measures the distance of @xmath278 to the origin ( see also remark [ re : delta - dist ] ) . [
th : div - p - adic ] let @xmath9 be a polytope mod @xmath13 in @xmath1019 , @xmath1044 a definable function ( in some expansion @xmath1038 of the ring structure of @xmath7 satisfying previous conditions ( ext ) and ( pres ) ) .
let @xmath1045 be a complex of monotopes mod @xmath13 in @xmath1019 such that @xmath1046 .
assume that the restriction of @xmath1047 to every proper face of @xmath9 is continuous .
then there exists a finite partition @xmath620 of @xmath9 in monotopes mod @xmath13 in @xmath1019 such that @xmath1048 is a closed complex , @xmath620 contains for every @xmath1049 a unique monotope @xmath623 with facet @xmath1029 , and moreover for every @xmath1050 @xmath1051 where @xmath1052 . for every proper face @xmath1022 of @xmath9 , and every @xmath1053 ,
the function @xmath1054 is continuous on @xmath1055 , which is a closed and bounded domain .
thus it attains a maximum value @xmath1056 ( see remark [ re : p - min ] ) .
so let @xmath1057 this is a definable set hence @xmath1058 is @xmath47-definable ( see remark [ re : p - min ] ) .
moreover by construction @xmath1058 is the graph of a function @xmath1059 , such that @xmath1060 for every @xmath1061 .
let @xmath1062 be the function whose restriction to each @xmath1063 is @xmath1064 .
the monotopic division ( theorem [ th : mono - div ] ) applies to @xmath1031 , @xmath787 and @xmath1065 .
it gives a finite partition @xmath24 of @xmath1031 in monotopes mod @xmath13 such that @xmath1066 is a complex , every non - closed @xmath1067 has a unique facet @xmath26 which belongs to @xmath1068 and @xmath1069 on @xmath25 where @xmath965 .
let @xmath620 be the family of @xmath1070 for @xmath28 .
this is clearly a finite partition of @xmath9 in monotopes mod @xmath13 in @xmath1019 .
every @xmath622 has a unique facet @xmath1049 , and @xmath1071 on @xmath1065 where @xmath1072 .
that is , for every @xmath1050 we have @xmath1073 by construction @xmath1074 and @xmath1075 for every @xmath1076 , hence @xmath1077 together with ( [ eq : div - p - adiv ] ) , this proves the last point .
[ pr : split - p - adic ] let @xmath1078 be a relatively open of a toplogical set is called * * relatively open if it is open in its closure , that is @xmath70 is closed . ] set .
assume that @xmath18 is the union of a complex @xmath83 of monotopes mod @xmath13 in @xmath1019 .
then for every integer @xmath46 there exists a finite partition of @xmath18 in semi - algebraic sets @xmath1079 such that @xmath1080 for every @xmath40 .
thanks to the correspondence between the faces of the monotopes mod @xmath13 in @xmath1019 and their faces , it suffices to prove the result for a relatively open set @xmath216 which is the union of a complex of monotopes mod @xmath13 in @xmath12 .
let @xmath1081 and @xmath1082
. by assumption @xmath18 is relatively open hence @xmath25 is closed , so @xmath24 is a closed complex .
let @xmath1083 be the list of minimal elements of @xmath83 .
every @xmath1084 such that @xmath1085 for some @xmath86 belongs to @xmath83 ( otherwise @xmath1086 which is closed , hence @xmath1087 , a contradiction since @xmath1088 ) .
note further that every @xmath1089 is a proper face of some @xmath1090 ( because @xmath1029 is a face of some @xmath1091 and @xmath1085 for some @xmath86 , hence @xmath1092 or @xmath1093 because @xmath9 is a monotope , and the second case is excluded because @xmath1094 ) .
in particular @xmath1095 is the union of the sets @xmath1096 such @xmath1097 for some @xmath86 , that is @xmath1098 . for
each @xmath1099 let @xmath1100 be the family of @xmath1091 such that @xmath1101 , and @xmath1102 .
the families @xmath1100 are pairwise disjoint , and so are the sets @xmath1103 since @xmath83 is a complex . by the same argument as above ( replacing @xmath83 by @xmath1100 ) @xmath1104 ,
hence @xmath1103 is relatively open and @xmath1105 .
it suffices to prove the result separately for each @xmath1103 .
indeed , assume that for each @xmath1099 we have found a partition @xmath1106 of @xmath1103 in definable sets such that @xmath1107 .
then let @xmath1108 for each @xmath549 . by construction
these sets form a partition of @xmath18 and @xmath1109 thus replacing @xmath18 and @xmath83 by @xmath1103 and @xmath1100 if necessary , we can assume that @xmath83 has a unique smallest element @xmath1110 . if @xmath1110 is closed , then @xmath1111 ( by minimality of @xmath1110 ) , and it suffices to take @xmath1112 and @xmath1113 for @xmath1114 .
so from now on we assume that @xmath1110 is not closed .
proposition [ pr : split - p - adic ] then applies to @xmath1110 and gives for some @xmath994 a partition @xmath1115 of @xmath1110 in largely continuous monotopic cells mod @xmath994 such that @xmath1116 for every @xmath86 . in particular each
@xmath1117 is a basic presburger set .
let @xmath1118 , and for every @xmath1119 and @xmath1120\!]$ ] let @xmath1121 .
note that this is a basic presburger set .
indeed , @xmath9 itself is a basic presburger set , and @xmath1122 with @xmath1123 is a basic presburger set because @xmath1117 is so ( replace every condition @xmath1124 defining @xmath1117 by @xmath1125 ) .
hence their intersection @xmath1126 is a basic presburger set too . for every @xmath1127 let @xmath1128
this defines a partition of @xmath18 . in order to conclude
it only remains to show that @xmath1129 for each @xmath86 , so that @xmath1130 .
since @xmath1131 , it suffices to check that for every @xmath1091 @xmath1132 let @xmath1133 , and @xmath362 be the support of any face of @xmath1126 . note that @xmath1134 implies that @xmath163 ,
thus @xmath159 is the support of a face @xmath1035 of @xmath9 .
this face @xmath1029 belongs to @xmath1135 , hence to @xmath100 if @xmath1136 .
we are claiming that @xmath1137 in that case , and that @xmath1138 if @xmath1139 .
this will finish the proof since @xmath1140 is the union of its faces , and ( [ eq : split - p - adic ] ) then follows immediately .
assume first that @xmath1136 .
since @xmath1126 and @xmath9 are basic presburger set , we know by proposition [ pr : face - egale - proj]([it : face - proj ] ) that @xmath1141 and @xmath162 , that is @xmath1142 .
since @xmath1136 we have @xmath335 hence @xmath1143 .
it follows that @xmath1144 that is @xmath1145 .
conversely , for every @xmath1146 we have on one hand @xmath1147 so there is @xmath1148 such that @xmath1149 , and on the other hand @xmath1150 so @xmath1151 .
thus @xmath1152 , and since @xmath1153 this proves that @xmath1154 .
this proves our claim in this case .
now assume that @xmath1139 .
then @xmath1155 hence @xmath1156 .
we already know that @xmath1157 be the previous case , and that @xmath1158 by construction . in particular @xmath1159 . but
@xmath1160 since @xmath83 is a complex and @xmath1139 .
altogether , using proposition [ pr : face - egale - proj]([it : face - proj ] ) for @xmath1126 and @xmath1117 we get @xmath1161
|
we introduce topological notions of polytopes and simplexes , the latter being expected to fulfill in @xmath0-adically closed fields the function of real simplexes in the classical results of triangulation of semi - algebraic sets over real closed fields .
we prove that the faces of every @xmath0-adic polytope are polytopes and that they form a rooted tree with respect to specialisation .
simplexes are then defined as polytopes whose faces tree is a chain .
our main result is a construction allowing to divide every @xmath0-adic polytope in a complex of @xmath0-adic simplexes with prescribed faces and shapes .
|
probabilistic cellular automata ( pca ) are discrete - time markov processes modelling the time evolution of a multicomponent system .
their main feature is the synchronous update of the states of the components , which take values in a finite set and interact with their neighbours according to a given probabilistic interaction rule .
pca are favourable models to study non - equilibrium phenomena . indeed ,
on the one hand , their definition is simple , as the space of realizations is discrete and interactions are local . on the other hand , despite this simplicity , they show a variety of complex behaviours . one of the interesting phenomena involving probabilistic cellular automata is the transition from ergodic to non - ergodic regime . after setting a free parameter above or below a certain critical threshold , at infinite time the process preserves part of the information on its initial condition ( non - ergodic behaviour ) .
namely , the probability measure at infinite time depends on the initial state of the dynamics . on the contrary , if the process is ergodic , it admits a unique , attracting invariant measure .
in @xcite it has been shown that the non - ergodic regime of a pca is connected to the existence of a phase transition for the pca , considered as a statistical mechanics system . over the last 50 years
, pca have undergone intense analytical and numerical investigations ( e.g. @xcite ) . however , as far as we know , many questions involving the rate of convergence to equilibrium or the characterisation of the invariant measures still remain open , even for the simplest models ( see e.g. @xcite for a survey ) .
in this paper we consider a class of pca that has a correspondence with percolation .
these models are refereed to as _ percolation systems _ in @xcite and as _ percolation operators _ in @xcite . from now on we will refer to them as _ percolation pca_. this class includes the well studied _ stavksaya s process _ ( see e.g. @xcite ) , in which the neighbourhood of every corresponds to the site itself and its right nearest neighbour on the one dimensional lattice . on the contrary in percolation pca the neighbourhood of every site could be any finite ( translation invariant ) set .
the reason why we decided to consider percolation pca is that their simplicity , combined with the presence of a phase transition , make them an interesting test case for attempts to characterise transient behaviour and stationary measures for spatially extended stochastic dynamics .
namely percolation pca are a prominent model for studying _ absorbing state phase transitions _ ( @xcite ) ,
i.e. there exists a phase characterised by almost sure convergence into an `` absorbing state '' ( a realisation where the process remains for ever whenever reached ) and a fluctuating phase , where the process remains active at all times . in this paper
we discuss two distinct aspects of the percolation pca . in section [
sect : criticalprobabilities ] we study how the neighbourhood affects the critical probability .
we provide a lower bound for critical probabilities @xmath10 and our result is stated in theorem [ theo : maintheorem1 ] . with our estimations we improve the previous lower bound @xcite showing that @xmath11 strictly if the neighbourhood @xmath12 . furthermore , we provide new bounds in case of neighbourhoods not considered before ( as far as we know ) . the comparison with numerical estimations , provided in the last section of this article , shows that our bounds are sharp . in order to derive the lower bound
we studied the temporal evolution evolution of `` absorbed sets '' ( sets of adjacent sites all in state `` zero '' ) .
if these sets on average are expanding , the realisation at infinite time is `` all zeros '' almost surely .
this idea comes from ( * ? ? ?
* chapter 6 ) .
our estimations take into account a certain aspect of the dynamics , i.e. absorbed sets can dynamically merge one with the other .
this leads to an improvement of the bound . in section [
sect : timeofconvergence ] we consider percolation pca on a finite one dimensional lattice with periodic boundaries and we study the convergence time of the process into the absorbing state .
our second main result is stated in theorem [ theo : maintheorem2 ] .
we show that at @xmath13 there exists a transition from a fast to a slow convergence regime .
namely we prove that the expected convergence time of the model grows exponentially ( resp .
logarithmically ) as the size of the system grows if @xmath14 ( resp .
@xmath15 ) .
this provides a partial answer to the _ unsolved problem 5.3.3 _ in @xcite and to an open problem mentioned in ( * ? ? ?
80 - 83 ) .
if compared with @xcite , where the fast ( resp .
slow ) convergence behaviour is proved for @xmath4 small enough ( resp .
close enough to @xmath16 ) , our result provides a sharp estimation .
the slow convergence regime can be interpreted as a metastable behaviour of the model , as the process spends an exceptionally long time into a non - stable state before falling into the absorbing state .
similar studies on the metastable behaviour of pca models were recently presented also in @xcite , although the methods used there do not apply in our case , as percolation pca are not reversible and do not have a naturally associated potential .
numerical estimations of @xmath10 ( e.g. @xcite ) are obtained assuming that the metastable regime ( the actual regime observed in numerical simulations , as there is no way to really simulate `` infinite space '' in computers ) is observed only for all values of @xmath4 at which the infinite process is in the fluctuating phase .
although this fact might appear obvious in terms of physical intuition , theorem [ theo : maintheorem2 ] provides a justification for this assumption from a rigorous mathematical point of view .
the proof of our result relies almost entirely on the correspondence between percolation pca and oriented percolation in two dimensions .
this connection has been described for the first time in @xcite .
the proof of the statement of the theorem involving the case of @xmath9 is an application of some percolation estimations presented in @xcite .
we generalize these estimates to the percolation model considered here , which differs from @xcite as here the neighbourhood is an arbitrary finite set and because sites ( instead of bonds ) can be open or closed .
the proof of the statement involving the case of @xmath17 is more technical and is based on * ( 1 ) * the generalization of the dynamic - block argument provided by @xcite to the case of non symmetric neighbourhood with more than two elements and * ( 2 ) * the estimation of the probability of a certain event involving a dual lattice construction provided by @xcite .
we shall end this introductory section by presenting the structure of the paper . in section [
sect : themodel ] we define the model and we present our main results , theorem [ theo : maintheorem1 ] and theorem [ theo : maintheorem2 ] . in section [
sect : criticalprobabilities ] we prove theorem [ theo : maintheorem1 ] . in section [
sect : timeofconvergence ] , divided into three subsections , we prove theorem [ theo : maintheorem2 ] . in subsection
[ sect : ergodicity ] we describe the correspondence between percolation pca and oriented percolation in two dimension , following @xcite . in subsection
[ sect : percolationestimates ] we present several percolation estimations from @xcite used to prove of the theorem .
finally in subsection [ sect : prooftheo2 ] we prove theorem [ theo : maintheorem2 ] .
probabilistic cellular automata ( pca ) are discrete - time markov chains on a product space , @xmath18 . in this paper
we consider both the case of infinite space , @xmath19 , and of finite space , @xmath20 , @xmath21 .
we consider the case of boolean variables , @xmath22 .
realisations of the process are denoted by @xmath23 . for
any @xmath24 and any @xmath25 , use @xmath26 to denote the @xmath1-th component of the vector @xmath27 and @xmath28 to designate the set of components corresponding to the sites of @xmath29 .
we introduce a _ neighbourhood function _ on @xmath30 .
we first fix a finite set @xmath31 , assuming that @xmath32 . if @xmath19 , @xmath33 we define the neighbourhood of @xmath1 as @xmath34 . if @xmath20 we consider periodic boundaries .
namely , for any @xmath35 we define the neighbourhood of @xmath1 as @xmath36 where @xmath37 denotes @xmath1 ( mod @xmath38 ) . for example , if @xmath39 , the neighbourhood of the site @xmath40 is @xmath41 . for any set @xmath25 , we define the neighbourhood of @xmath29 as @xmath42 . in percolation
pca the states of the process are synchronously updated at every site according to the following _ transition probability _ , @xmath43 where @xmath44 $ ] is a free parameter
. corresponds to @xmath45 and zeroes and ones are inverted . ]
the temporal evolution of the process can be represented by introducing a linear operator @xmath46 , which acts on the space of probability measures @xmath47 . for any @xmath48
, we use @xmath49 to denote the measure obtained applying @xmath46 to @xmath50 . by using @xmath51 to denote the cylinder set @xmath52 , with @xmath25 , the measure @xmath49 is defined as @xmath53 in order to characterise the time evolution of pca
, it is useful to introduce the set of _ space - time realisations _ , @xmath54 , where @xmath55 is the _ space - time _ set .
the elements of @xmath56 are the realisations of the process at all times , @xmath57 .
we then introduce an oriented graph @xmath58 , whose edges connect any vertex @xmath59 to the vertices @xmath60 , where @xmath61 . the vertices that can be reached from @xmath59 through a path on @xmath62 constitute the _ evolution cone _ of @xmath63 . with neighbourhood @xmath12 . in this figure only edges between vertices belonging to the evolution cone of @xmath63 have been drawn . ]
we now introduce some definitions that will be used along the whole article .
[ def : evolutionmeasure ] consider the percolation pca ( [ eq : operator ] ) with @xmath19 ( respectively @xmath20 and periodic boundaries ) . for every @xmath64 , we define the _ evolution measure _ @xmath65 ( respectively @xmath66 ) as the joint probability distribution of measures @xmath50 , @xmath67 , @xmath68 , @xmath69 . for example
, we use @xmath70 to denote the evolution measure of the percolation pca on finite space , starting from the realisation `` all ones '' .
[ def : expectationontheevolutiospace ] consider the percolation pca ( [ eq : operator ] ) with @xmath19 ( respectively @xmath20 and periodic boundaries ) .
we use @xmath71 $ ] ( respectively @xmath72 $ ] ) to denote the expectation in relation to the evolution measure @xmath65 ( respectively @xmath73 ) .
[ [ monotonicity ] ] monotonicity + + + + + + + + + + + + it is immediate from the definition of transition probability that the dirac measure @xmath74 , where @xmath75 , is stationary , i.e. @xmath76 .
furthermore , the operator @xmath46 of this stochastic process is _
monotone_. monotonicity of @xmath46 means that it preserves partial order among elements of @xmath47 .
we first introduce partial order `` @xmath77 '' in @xmath78 by defining for any two realizations @xmath79 , @xmath80 @xmath81 .
we then introduce the functions @xmath82 , which only depend on a finite number of sites .
we call @xmath83 _ monotone _ _ iff _ for any @xmath79 , @xmath84 .
we then introduce partial order in @xmath47 by defining @xmath85 for any _ monotone _
function @xmath86 , @xmath87 .
finally , we introduce an order relation between operators and we introduce the notion of monotone operator .
[ def : monotoneoperator ] an operator @xmath88 is called monotone if for any pair of measures @xmath89 , @xmath90 . the operator ( [ eq : operator ] ) of the percolation pca is monotone .
this property follows from the fact that the transition probability ( [ eq : transitionprob ] ) preserves order locally , i.e. for any @xmath24 , @xmath91 ( see for example @xcite for a proof of this ) .
monotonicity of @xmath46 implies that the probability measure , @xmath92 exists and it is invariant .
[ def : critprob ] consider the percolation pca on @xmath93 with finite neighbourhood @xmath94 .
we define the _ critical probability _ as , @xmath95 } \ { \nu_p = \delta_{\mathbf{0 } } \}.\ ] ] [ def : ergodic ] an operator @xmath96 is ergodic if the two following conditions hold : * ( a ) * there exists a unique @xmath97 such that @xmath98 and * ( b ) * @xmath99 , @xmath100 . for any @xmath17 the evolution operator of the percolation pca is not ergodic
. indeed , in this case @xmath74 and @xmath101 ( defined in [ eq : limitmeasure ] ) are two distinct invariant measures . for any @xmath102
, the percolation pca ( [ eq : operator ] ) is ergodic . in @xcite
it has been proved that @xmath103 for the stavskaya s process ( @xmath39 ) and a more general proof in case of general neighbourhood can be found in @xcite .
the proofs are based on two methods widely used in statistical mechanics , namely the counting path method and the peierls argument .
our first result is stated in the following theorem and it involves the estimation of @xmath13 .
[ theo : maintheorem1 ] consider the percolation pca on @xmath93 with finite neighbourhood @xmath104 , where @xmath105 are such that @xmath32 .
define @xmath106 and @xmath107 as the unique solution in the interval @xmath108 of the following equation , @xmath109 where @xmath110 . then @xmath111 .
the proof of the theorem is presented in section [ sect : criticalprobabilities ] . from ( [ eq : theorem1eq1 ] ) it follows that @xmath112 , as @xmath113 is positive in @xmath108 .
our analytical lower bound can be compared with the numerical estimations in the following tables .
the plots of the numerical estimations can be found in the appendix of this article .
the numerical estimation in case @xmath114 has been provided in @xcite . [ cols="^,^,^,^",options="header " , ] our main result
is stated in theorem [ theo : maintheorem2 ] and it involves the convergence time into the absorbing state of the percolation pca with finite space and periodic boundaries , as defined at the beginning of this section .
when @xmath30 is finite , the process is always ergodic ( definition [ def : ergodic ] ) .
indeed , for any realisation of the process @xmath115 at time @xmath116 , the probability that @xmath117 is bounded from below by the constant @xmath118 .
this implies that there exists almost surely a finite time @xmath119 such that @xmath120 for all @xmath121 .
hence , for any @xmath122 , @xmath123 . in order to estimate the convergence time into the absorbing state we define the _ absorption - time _
@xmath124 , representing the first time all sites in the segment @xmath125 are in state zero for @xmath126 .
[ def : absorbtime ] for all @xmath127 , we call the absorption time of the interval @xmath128 the random variable @xmath129 , @xmath130 in case @xmath20 , this random variable is well defined only if @xmath131 .
we recall definitions [ def : evolutionmeasure ] and [ def : expectationontheevolutiospace ] and we state our main result .
[ theo : maintheorem2 ] consider the percolation pca with space @xmath132 , periodic boundaries and finite neighbourhood @xmath133 , where @xmath134 are some distinct elements of @xmath93 .
for every @xmath44 $ ] there exist @xmath135 and some positive constants @xmath136 ( dependent on @xmath4 ) such that for all @xmath137 , 1 .
if @xmath9 , @xmath138 \leq k_2 \log ( c_2 \ , n ) $ ] , 2 . if @xmath17 , @xmath139 \leq k_4 \exp ( c_4 \ , n)$ ] .
the proof of the theorem is presented in section [ sect : timeofconvergence ] .
in this section we prove theorem [ theo : maintheorem1 ] , which provides a lower bound for @xmath13 as a function of the neighbourhood .
the proof of theorem [ theo : maintheorem1 ] requires lemma [ lem : randomwalk ] and propositions [ prop : containedinto ] and [ prop : estimations ] , which are stated in this section .
[ prop : containedinto ] consider two percolation pca in @xmath93 with neighbourhoods respectively @xmath6 and @xmath140 , both finite subsets of @xmath93 , such that @xmath141 .
then @xmath142 from proposition [ prop : percolation ] , stated in section [ sect : ergodicity ] , and from the fact that the edge set of the graph @xmath62 is a subset of the edge set of the graph @xmath143 , it follows that @xmath5 , @xmath144 , @xmath145 ( we added the subscript to the operator in order to distinguish between the two neighbourhoods ) . from definition [ def : critprob ]
it follows that , @xmath146 hence , @xmath147 .
we introduce some notation .
we call a segment of @xmath93 , @xmath148 a _ massif of zeros _ of length @xmath149 for a given @xmath23 , if @xmath150 and @xmath151
. we use @xmath152 $ ] to denote the set of integers in the interval @xmath153 $ ] .
we use @xmath154 to denote the initial realisation of the percolation pca ( namely the initial probability measure is @xmath155 ) and @xmath156 , @xmath157 , @xmath69 the random realisations of the process at different times . for every @xmath158
, we introduce the following notation ( the role of @xmath159 will be clear later ) . for every @xmath160
, we enumerate somehow the massifs of zeros of length larger or equal to @xmath161 .
this means that we assign to every massif a label @xmath162 , where @xmath163 is the set of labels .
we denote by @xmath164 and by @xmath165 respectively the rightmost and the leftmost zero of the @xmath166-th massif .
we observe that , by definition of the transition probability for the process ( [ eq : transitionprob ] ) , such massifs can not have disappeared at time @xmath167 ( see also figure [ fig:2 ] ) . .
if the process starts from a realisation having a massif of zeros in @xmath168 with @xmath169 , as in the figure , then the state of every site in @xmath170 for the random realisation at time @xmath16 and in @xmath171 for the random realisation at time @xmath172 is @xmath173 almost surely . ] for every @xmath162 , we define the random variables @xmath174 and @xmath175 using recursion .
namely , @xmath176 , @xmath177 , @xmath178 , \ , \ , \eta^t_{y } = 0 \ } & \mbox { if } r_k^{t-1 } - l_k^{t-1 } \geq ( s_u - s_1 ) \\ - \infty & \mbox{otherwise}\\ \end{cases}\ ] ] @xmath179 , \ , \ , \eta^t_{y } = 0 \ } & \mbox { if } r_k^{t-1 } - l_k^{t-1 } \geq ( s_u - s_1 ) \\ + \infty & \mbox{otherwise}\\ \end{cases}.\ ] ] namely @xmath180 and @xmath181 keep track of the temporal evolution of two extremal sites of the @xmath166-th massif .
if the distance between such sites at a given time is less than @xmath182 , then at all subsequent times @xmath183 and @xmath184 . instead if at time @xmath185 , then at time @xmath186 the massif still exists almost surely .
note that it might happen that two or more massifs merge at a certain time . in this case
more than one label is used to denote the same massif .
the next lemma shows that if the massifs of zeros are `` on average '' expanding , then the state of the system at infinite time is zero almost surely . as this happens independently on the initial realisation , the process is ergodic .
[ lem : randomwalk ] for every @xmath158 , the following statement holds .
if there exist two families of independent and identically distributed random variables @xmath187 , @xmath188 , such that @xmath189 , @xmath190 , @xmath191 , the conditions ( [ eq : lemcond1 ] ) , ( [ eq : lemcond2 ] ) , ( [ eq : lemcond3 ] ) hold , @xmath192 @xmath193 > e[\xi_1 ^ 1]\ ] ] then @xmath99 , @xmath194 in the statement of the lemma @xmath195 denotes the probability distribution of the random variables @xmath196 or @xmath197 .
such random variables stochastically dominate from below the change of position of the rightmost and leftmost site of the massif every @xmath159 steps .
we also recall that @xmath198 has been defined in definition [ def : evolutionmeasure ] .
the proof of the lemma is similar to the proof of proposition 6.4 in @xcite .
it is sufficient to prove that @xmath189 , @xmath5 , @xmath199 , @xmath200 such that @xmath201 from which condition ( [ eq : lemmaconcl ] ) follows .
we define @xmath202 + e[{\eta}_1 ^ 1]}{4}$ ] and @xmath203 + 3 e[{\eta}_1 ^ 1]}{4}$ ] , where @xmath204 $ ] denotes the expectation , and we observe that if ( [ eq : lemcond3 ] ) holds , then @xmath205 .
then for every @xmath154 , @xmath176 , @xmath206 , @xmath207 such that @xmath208 , there exists two constants @xmath209 such that , @xmath210 in the previous expression @xmath211 denotes the probability measure defined on the space of outcomes of the sum of the increments @xmath212 , @xmath213 .
the first inequality follows from ( [ eq : lemcond1 ] ) and ( [ eq : lemcond2 ] ) .
the second inequality follows from the properties of the one dimensional random walk , observing that by definition @xmath214 < c_1 $ ] and @xmath215 > c_2 $ ] .
the two constants @xmath216 and @xmath217 depend on the probability distribution of the increments of the random walk .
we observe that if for all @xmath116 multiple of @xmath159 , @xmath218 and @xmath219 ( event in the first expression in ( [ eq : lemeq1 ] ) ) , then for all @xmath220 , @xmath221 and @xmath222 .
hence , the state of all sites in the space - time region @xmath223 } : = \ { ( x , t ) \ , \ , : \ , \ , t \in
\mathbb{z}_+ \mbox { and } c_1 t + n + i_0 -t s_1
\leq x \leq c_2 t - m + j_0 - t s_u \}$ ] is zero .
this region is represented in figure [ fig : lemm ] on the left .
this follows from the observation that by definition of transition probability of the percolation pca the following property holds , namely , @xmath224 on the right is defined as @xmath225.,title="fig:",scaledwidth=45.0% ] on the right is defined as @xmath225.,title="fig:",scaledwidth=45.0% ] furthermore we observe that @xmath189 , @xmath5 , @xmath226 , the measure @xmath227 is such that the probability that there exists a massif of zeros of length @xmath228 in @xmath229 $ ] goes to @xmath16 as @xmath230 .
we choose then @xmath231 and @xmath232 such that @xmath233 and @xmath234 large enough such that such probability is larger than @xmath235 for all @xmath236 .
simple geometrical considerations show that for any @xmath237 , @xmath238 , all regions @xmath239}$ ] , where @xmath240 \subset [ y , y+d]$ ] , have a non empty common region ( dark region in figure [ fig : lemm ] - right ) .
we call @xmath241}$ ] this region . from ( [ eq : lemeq1 ] ) and from the previous observations
the following property holds , @xmath242 } , \ , \eta_x^t = 0 \ , \ , \right ) > 1 - \epsilon.\ ] ] choosing @xmath236 and @xmath234 such that @xmath243}$ ] implies ( [ eq : lemtoprove ] ) .
we provide a lower bound for the critical probability of the percolation pca with neighbourhood @xmath244 i.e. all elements between the two extremal ones are present .
our bound is a function of @xmath245 . by proposition [ prop : containedinto ]
such bound holds also for percolation pca with neighbourhood obtained removing some sites from ( [ eq : containedintoex ] ) .
the proof of the theorem is based on an application of lemma [ lem : randomwalk ] .
we fix a value @xmath158 and by using the monotonicity property of the percolation pca , we define the random variables @xmath246 and @xmath247 , whose probability distribution satisfies @xmath248 $ ] the conditions ( [ eq : lemcond1 ] ) and ( [ eq : lemcond2 ] ) of lemma [ lem : randomwalk ] . we define , @xmath249 } \ { e[\pi^1_1 ] > e[\xi^1_1 ] \}.\ ] ] from lemma [ lem : randomwalk ] , for all @xmath250 the percolation pca is ergodic . from definitions [ def : critprob ] and [ def :
ergodic ] , @xmath251 .
we fix first @xmath252 and we derive @xmath253 , later we consider @xmath254 and we derive @xmath107 . both @xmath253 and @xmath107 appear in the statement of the theorem .
higher is the value of @xmath159 considered , more challenging is the estimation of @xmath255 , as this involves the characterization of the increments of @xmath256 over a larger time interval . [ [ section ] ]
we fix now @xmath252 and we provide an estimation for ( [ eq : pik ] ) and ( [ eq : xik ] ) for any @xmath279 . after this
we determine which values of @xmath4 satisfy ( [ eq : lemcond3 ] ) .
we consider the percolation pca starting from initial realisation @xmath280 and we assign assign the label @xmath16 to the unique massif of zeros , namely @xmath281 and @xmath282 .
we recall that by definition , @xmath283 , \ , \ ,
\eta_z^1 = 0\ ] ] ( see also figure [ fig:3 ] ) . hence , @xmath284 , @xmath285 this bound is obtained considering that almost surely @xmath286 $ ] , @xmath287 and that @xmath288 $ ] , independently @xmath289 .
analogously , @xmath284 , @xmath290 thus for all @xmath291 , we define the probability distributions of @xmath292 and @xmath293 respectively as , @xmath294 with this definition , from the relations ( [ eq : stochdomination1 ] ) - ( [ eq : stochdomination4l ] ) , the relations ( [ eq : lemcond1 ] ) and ( [ eq : lemcond2 ] ) are satisfied . it remains to determine for which values of @xmath44 $ ] the second condition of lemma [ lem : randomwalk ] holds . by a simple computation , @xmath295 = \frac{1-p}{p } - s_u , \\ e [ \xi_1 ] = \frac{1-p}{p } - s_1.\end{aligned}\ ] ] and @xmath296 - e[\pi_1 ] \geq 0 \ , \
, \iff \ , \ , p \geq p_1,\ ] ] where @xmath106 appears on the statement of the theorem .
thus we proved that @xmath297 . .
if the process starts from the realisation represented in the figure ( row below ) , then the state of the sites above the small horizontal ball is almost surely @xmath173 at time @xmath16 ( row above ) . ]
[ [ section-1 ] ] we fix now @xmath254 and we use the same argument . namely we consider the percolation pca starting from initial realisation @xmath264 such that @xmath298 and we assign label @xmath16 to the unique massif of zeros of @xmath280 .
we recall that by definition of @xmath299 , @xmath300 , \ , \ , \ , \eta^2_z=0.\ ] ] from definition ( [ eq : operator ] ) it follows that @xmath301 } t ( \eta^{2}_z=0 | \eta^1_{\mathcal{u}(z ) } ) , \end{split}\end{gathered}\ ] ] where @xmath302}$ ] and @xmath303 , \ , \ , \ , \eta^{\prime}_z = \eta^{1}_z \ , \}$ ] .
the sum is reduced to the elements of @xmath304 because the states of the sites in the interval @xmath305 $ ] for @xmath157 depend only on the states of the sites in the finite interval @xmath306 $ ] for @xmath156 .
a similar expression holds for the random variable @xmath307 , @xmath308 } t ( \eta^{2}_z=0 | \eta^1_{\mathcal{u}(z ) } ) , \end{split}\end{gathered}\ ] ] where @xmath309}$ ] and @xmath310 , \ , \ , \ , \eta^{\prime}_z = \eta^{1}_z \}$ ] .
the exact computation of the left hand side of ( [ eq : tau2 - 1 ] ) and of ( [ eq : tau2-left ] ) for any @xmath311 is a difficult combinatorial problem , as for each of the @xmath312 possible realizations one should determine the corresponding product of transition probabilities .
we present our estimations in the following proposition .
[ prop : estimations ] consider the realisation @xmath264 which has zeros in @xmath265 $ ] and ones everywhere else , where @xmath273 are such that @xmath298 .
assign label @xmath16 to the unique massif of zeros of @xmath280 and recall the definition of @xmath313 , @xmath314 .
then , @xmath315 for any @xmath316 , @xmath317 and for any @xmath318 , @xmath319 @xmath320 we postpone the proof of proposition [ prop : estimations ] to the next paragraph and we conclude the proof of theorem [ theo : maintheorem1 ] .
we use the lower bounds provided in the proposition to define the probability distribution of the random variables @xmath321 , @xmath322 .
namely @xmath284 , we define the probability of the event @xmath323 ( respectively @xmath324 ) as the lower bound of the probability of the event @xmath325 ( respectively @xmath326 ) provided in the proposition . with such definition , the expectation of the random variables @xmath327 , @xmath328 is equal to @xmath329 = 2 \ , \frac { ( 1 - p ) } { p } - 2 s_u + \frac{(1-p)^6 + ( 1-p)^{2s_u - 2s_1 + 2}}{p ( 2-p)},\ ] ] @xmath330 = - 2 \ , \frac { ( 1 - p ) } { p } - 2 s_1 - \frac{(1-p)^6 + ( 1-p)^{2s_u - 2s_1 + 2}}{p ( 2-p)}\ ] ] by simple computations , the maximum @xmath44 $ ] such that the inequality @xmath331 - e [ \xi_k ] \geq 0 $ ] is satisfied ( condition [ eq : lemcond3 ] of lemma [ lem : randomwalk ] ) , corresponds to the value @xmath107 defined in the statement of theorem [ theo : maintheorem2 ] . as the function
@xmath332 - e [ \xi^t_k]$ ] intersects the line @xmath333 only in one point of the interval @xmath334 $ ] , @xmath107 is the unique solution of @xmath332 - e [ \xi^t_k]=0 $ ] that falls in this interval . in the proof we present the estimation of the probability @xmath335 of the events @xmath336 .
using the same argument one can estimate the probability of the events @xmath337 . by definition of @xmath299 , @xmath338 , \ , \ , \eta^2_z=0.\ ] ] as observed previously , the state of the sites in @xmath339 $ ] for @xmath157 depends only on the state of the sites in @xmath340 $ ] for @xmath156 .
furthermore we observed that the state of the sites in @xmath341 $ ] is zero almost surely for @xmath157 .
hence , from equation ( [ eq : tau2 - 1 ] ) , we obtain the following estimation ( see also figure [ fig : est12 ] ) , @xmath342 , \ , \ , \eta_x = 0)\\ & = \delta_{\rho(x , y)}\mathcal{p}^2(\eta_{r_1 ^ 0 - 2 s_1 + 1}=0 ) \\ &
= \delta_{\rho(x , y ) } \mathcal{p } ( \eta_{r_1 ^ 0 - 2s_u + 1 } = 0 ) + \delta_{\rho(x , y ) } \mathcal{p } ( \eta_{r_1 ^ 0 - 2s_u + 1 } = 1 ) \ , ( 1-p ) \\ & = ( 1-p ) + p ( 1-p ) = 1 - p^2 , \end{split}\end{gathered}\ ] ] which corresponds to the estimation ( [ eq : r21 ] ) .
similarly we obtain the estimation ( [ eq : r22 ] ) , @xmath343 ) \\ & = \delta_{\rho(x , y)}\mathcal{p}^2 ( \eta_{r_1 ^ 0 - 2s_u + 1 } = \eta_{r_1 ^ 0 - 2s_u + 2 } = 0 ) \\ & = \delta_{\rho(x , y)}\mathcal{p}^1 ( \eta_{r_1 ^ 0 - s_u + 1 } = 0 , \ , \eta_{r_1 ^ 0 - s_u + 2 } = 0 ) \ , \ , 1 \\ & + \delta_{\rho(x , y)}\mathcal{p}^1 ( \eta_{r_1 ^ 0 - s_u + 1 } = 0 , \ , \eta_{r_1 ^ 0 - s_u + 2 } = 1 ) \ , \ , { ( 1-p ) } \\ & + \delta_{\rho(x , y)}\mathcal{p}^1 ( \eta_{r_1 ^ 0 - s_u + 1 } = 1 , \ , \eta_{r_1 ^ 0 - s_u + 2 } = 0 ) \ , \ , { ( 1-p)}^2 \\ & + \delta_{\rho(x , y)}\mathcal{p}^1(\eta_{r_1 ^ 0 - s_u + 1 } = 1 , \ , \eta_{r_1 ^ 0 - s_u + 2 } = 1 ) \,\ , { ( 1-p)}^2 \\ & \geq ( 1-p)^2 \ , + \ , p ( 1-p)^2 \ , + \ , p ( 1-p)^3 \ , + \ , p^2(1-p)^2 . \end{split}\end{gathered}\ ] ] .
if the initial realisation of the percolation pca is the one represented in the figure ( lowest row ) , then almost surely the state of the sites above the short horizontal line is zero .
, scaledwidth=80.0% ] we provide now the estimation ( [ eq : tau2-estj ] ) considering all @xmath344 .
we introduce an index @xmath345 $ ] , and we define the mutually disjoint cylinder sets ( they will be defined later ) , @xmath346}\ , , \ , \ , \ , \
{ c^{b , m } \}_{m \in [ [ 0,j-2]]}\ , , \ , \ , \ , c^{c}.\ ] ] we denote by @xmath347 the set of realisations that are not in sets just defined , namely @xmath348 } c^{a , m } \cup c^{b , m } \cup c^c \cup c^{a , j-1}.\ ] ] for every @xmath349 $ ] we estimate @xmath350 and @xmath351 , and we also estimate @xmath352 and @xmath353 .
furthermore for each of these sets we provide some bounds @xmath354,@xmath355,@xmath356 .
namely for every @xmath357 , where @xmath358 denotes generically @xmath359 , @xmath360 or @xmath361 , the following inequality holds , @xmath362 } t ( \eta^{2}_z=0 | \eta^1_{\mathcal{u}(z ) } ) , \ ] ] we use such estimations to provide a bound for ( [ eq : tau2 - 1 ] ) , as shown in the following expression . @xmath363 } [ & \delta_{\rho(x , y ) } \mathcal{p } ( c^{a , m } ) \ , b^{a , m } + \\ & \delta_{\rho(x , y ) } \mathcal{p } ( c^{b , m } ) \ ,
b^{b , m } ] \ , + \delta_{\rho(x , y ) } \mathcal{p } ( c^{c } ) \ , b^{c } + \\ & \delta_{\rho(x , y ) } \mathcal{p } ( c^{d } ) \end{split}\end{gathered}\ ] ]
we start with the introduction of the cylinder set @xmath364 , @xmath365 \setminus \ { r^0_k - s_u+m+1 \ } , \\ & \eta_z=0 \mbox { and } \eta_{r^0_k - s_u+m+1}=1 \ } , \end{split}\end{gathered}\ ] ] ( see also figure [ subfig : c ] ) . by a simple computation ,
@xmath366 furthermore we observe that @xmath367 , the product over the transition probabilities of equation ( [ eq : tau2 - 1 ] ) satisfies the following bound , @xmath368 } t ( \eta^{2}_z=0 | \eta^1_{\mathcal{u}(z ) } ) \geq b^{a , m},\ ] ] where @xmath369 .
if the initial realisation of the percolation pca ( first row ) is the one represented in the figure , then the sites underlined by a short line on the second and third row have state zero almost surely .
figure ( a ) : the second row represents a realisation @xmath157 belonging to the set @xmath370 , @xmath371 , @xmath372 .
figure ( b ) : the second row from below represents a realisation @xmath157 belonging to the set @xmath373 , @xmath371 , @xmath372 .
figure ( c ) : the second row from below represents a realisation @xmath157 belonging to the set @xmath374 , @xmath372.,scaledwidth=90.0% ] . if the initial realisation of the percolation pca ( first row ) is the one represented in the figure , then the sites underlined by a short line on the second and third row have state zero almost surely . figure ( a ) :
the second row represents a realisation @xmath157 belonging to the set @xmath370 , @xmath371 , @xmath372 . figure ( b ) : the second row from below represents a realisation @xmath157 belonging to the set @xmath373 , @xmath371 , @xmath372 .
figure ( c ) : the second row from below represents a realisation @xmath157 belonging to the set @xmath374 , @xmath372.,scaledwidth=90.0% ] . if the initial realisation of the percolation pca ( first row ) is the one represented in the figure , then the sites underlined by a short line on the second and third row have state zero almost surely . figure ( a ) :
the second row represents a realisation @xmath157 belonging to the set @xmath370 , @xmath371 , @xmath372 .
figure ( b ) : the second row from below represents a realisation @xmath157 belonging to the set @xmath373 , @xmath371 , @xmath372 .
figure ( c ) : the second row from below represents a realisation @xmath157 belonging to the set @xmath374 , @xmath372.,scaledwidth=90.0% ] then we introduce the cylinder sets @xmath375 , where @xmath376 .
@xmath377 , \ , \ , \eta_z=0 , \\
& \eta_{r_1 ^ 0 - s_u + m + 1}=1 , \\ & \exists y \in [ [ r_1 ^ 0 - s_u + m + 2 , r_1 ^ 0 - s_u + j ] ] \mbox { s.t . }
\eta_y = 1 \ } \end{split}\end{gathered}\ ] ] ( see also figure [ subfig : d ] ) . by using the definition of transition probability for the percolation pca we estimate the probability measure of this cylinder set @xmath378 , \ ] ] and we observe that @xmath379 the following bound holds @xmath380 } t
( \eta^{2}_z=0 | \eta^1_{\mathcal{u}(z ) } ) \geq ( 1-p)^{j - m}\ ] ] thus we define @xmath381 the bound ( [ eq : tcbm ] ) is obtained considering that @xmath382 for all @xmath383 $ ] and @xmath384 for all @xmath385 $ ] .
third , we define the cylinder set @xmath386 , @xmath387\}.\ ] ] for this set , @xmath388 and @xmath389 ,
@xmath390 } t ( \eta^{2}_z=0 | \eta^1_{\mathcal{u}(z ) } ) = 1\ ] ] ( see also figure [ subfig : e ] ) .
thus we define @xmath391 finally we recall the definition of @xmath347 provided in equation ( [ eq : c^d ] ) .
we observe that @xmath392 and that @xmath393 , @xmath394 } t ( \eta^{2}_x=0 | \eta^1_{\mathcal{u}(x ) } ) \geq ( 1-p)^j.\ ] ] the inequality is obtained considering that from the definition ( [ eq : transitionprob ] ) it follows that @xmath395 , @xmath396 .
thus we define @xmath397 we finally use the estimations ( [ eq : pcam ] ) , ( [ eq : bcam ] ) , ( [ eq : pcbm ] ) , ( [ eq : bcbm ] ) , ( [ eq : pccm ] ) , ( [ eq : bccm ] ) , ( [ eq : bcdm ] ) , ( [ eq : bcdm ] ) , in ( [ eq : tau2 - 5 ] ) and we derive the lower bounds ( [ eq : tau2-estj ] ) and ( [ eq : tau2-estj2 ] ) .
in this section we prove theorem [ theo : maintheorem2 ] . in section [
sect : ergodicity ] we describe the connection between percolation pca and oriented percolation . we mainly follow @xcite , although propositions and statement have been reformulated emphasising the differences between percolation pca on a finite and infinite space . in section [
sect : percolationestimates ] we list some percolation estimates . some of these percolation estimates have been proved in @xcite in the case of oriented bond percolation with symmetric neighbourhood . in this article
we consider a similar model , namely oriented site percolation with arbitrary neighbourhood .
the proofs of these estimates in our case are substantially the same of those provided in @xcite .
we sketch them illustrating the small differences . in section [
sect : prooftheo2 ] we prove the theorem .
the proof of the right inequality of statement ( a ) of the theorem is an application of the estimates presented in section [ sect : percolationestimates ] .
the proof of the left inequality can be found in @xcite .
the proof of the right inequality of statement ( b ) is trivial .
the proof of the left inequality uses some of the percolation estimates and the estimation provided proposition [ prop : inequality2 ] , which is stated in the same section .
the original contribution of the author consists in the proof and the application of proposition [ prop : inequality2 ] to the proof of the statement ( b ) , in the estimations based on path constructions used in the proof of statement ( b ) and in the generalization of the percolation estimates to the proof of the statement ( a ) . in this section
we describe a connection between the percolation pca and a certain percolation model .
this connection has been pointed out for the first time in @xcite , as far as we know .
we consider a percolation pca with space @xmath20 or @xmath398 , as defined in section [ sect : themodel ] .
we define an auxiliary space @xmath399 , we denote by @xmath400 its elements and we introduce in this space the bernoulli product measure @xmath401 .
namely , the state of every component is @xmath16 with probability @xmath4 and @xmath173 with obability @xmath402 independently .
we declare a vertex @xmath403 `` open '' if @xmath404 and `` closed '' otherwise .
the correspondence between the pca and percolation consists in the fact that the probability that the state of the site @xmath24 is @xmath16 at time @xmath220 for the probabilistic cellular automaton equals the probability that the site @xmath405 is connected by a path of open vertices in @xmath62 to the line @xmath333 .
this is precisely the meaning of the statement of proposition [ prop : percolation ] , which is stated below . in order to describe this connection rigorously ,
we represent the percolation pca starting from an initial realisation @xmath406 by introducing a deterministic mapping @xmath407 for every @xmath405 , the component @xmath408 of @xmath27 is defined as @xmath409 where @xmath410 are elements of @xmath411 .
this mapping defines any @xmath412 , @xmath413 , @xmath158 as a function of the variables @xmath414 associated to vertices belonging to the evolution cone of @xmath415 , and of initial realisation @xmath416 .
one should observe that , recalling ( [ eq : transitionprob ] ) and using independence , for any @xmath24 , @xmath417 , @xmath418 , @xmath419 , @xmath406 , @xmath420 where in the last expression we rewrote the second quantity in a more compact form .
this notation will be used also in the proof of the next proposition .
[ prop : correspondence ] consider the percolation pca with space @xmath421 , represented by the operator @xmath422 . then , for any @xmath406 , @xmath418 , @xmath423 for any @xmath24 , @xmath424 , we define @xmath425 by using equation ( [ eq : correspondence ] ) , we observe that the following equalities hold
. @xmath426 by proceeding with the expansion , by writing the previous term as a sum over all states of the cylinder set of @xmath427 of transition probabilities ( [ eq : transitionprob ] ) and by recalling definition ( [ eq : operator ] ) , one recognizes that this term is equal to the left side of ( [ eq : correspondenceprop ] ) .
the next proposition has been proved in @xcite and it is stated and used also in @xcite . [ prop : percolation ] the function @xmath428 is such that @xmath429 _ iff _ there exists a sequence @xmath430 satisfying the three following properties , 1 .
@xmath431 and @xmath432 for any @xmath433 , 2 .
@xmath434 for any @xmath435 , 3 .
@xmath436 .
we sketch the proof of the proposition .
assume @xmath437 and assume that properties 1 , 2 , 3 hold for a sequence of sites @xmath438 , @xmath439 , @xmath69 @xmath440 .
from ( [ eq : mapping ] ) it follows that @xmath441 and @xmath442 s.t .
this implies that there exists an element @xmath444 such that properties 1 , 2 , 3 hold for the sequence @xmath445 , @xmath438 , @xmath69 @xmath440 .
the proof of the proposition follows by induction .
if we consider the case of infinite space , from the previous proposition it follows that ergodicity for the probabilistic cellular automaton is associated with the existence of an infinite path of open vertices in the auxiliary space .
indeed , recall definitions [ def : critprob ] and [ def : ergodic ] and observe that , @xmath446 thus the probabilistic cellular automaton is non - ergodic if and only if the limit @xmath447 of the probability that a vertex @xmath63 is connected to the line @xmath333 by an open path is positive .
if we consider the case of finite space with periodic boundaries , the previous proposition shows that there is a connection between the absorption time of the probabilistic cellular automaton and the existence of an open path in the auxiliary space .
this connection is clarified in the next proposition . before its statement
we introduce some more definitions . from now on
we use @xmath448 to denote the bernoulli product measure in the finite space and @xmath449 to denote the bernoulli product measure in the infinite space .
[ def : opensitesandpaths ] consider @xmath450 and consider the event , @xmath451 if @xmath19 we denote this event by @xmath452 and if @xmath20 we denote this event by @xmath453 recall the definition of evolution measure ( definition [ def : evolutionmeasure ] ) and of absorption time ( equation [ eq : absorbingtime ] ) .
recall that @xmath454 can be considered as a function @xmath455 , as , from ( [ eq : mapping ] ) , @xmath456 it is a mapping from @xmath457 to @xmath56 .
[ prop : taunconnection ] consider the percolation pca on a finite space with periodic boundaries . for every @xmath220 , @xmath458 \ , \
, \mbox{s.t . } \ , \ , ( x , t ) \overset { \mathcal{g}_{\mathcal{u } } ( n ) } { \longrightarrow } s^0 \ ) , \ ] ] where @xmath459 denotes the set of vertices of @xmath460 belonging to the line @xmath333 . by the definition of @xmath461
( see definition [ def : absorbtime ] ) , @xmath462 if and only if @xmath463 $ ] such that @xmath437 .
from proposition [ prop : percolation ] , it follows that @xmath464 if and only if @xmath465 .
( recall definition [ def : absorbtime ] ) occurs if at least one open path joins one of the sites @xmath427 such that @xmath466 $ ] to one of the sites @xmath467 , @xmath468 . if the neighbourhood is periodic , then the path can leave from one the two vertical lines @xmath469 or @xmath470 and re - appear at the same high on the other line ( e.g. see the path @xmath471 ) . ] * remark . *
recall the definition of the neighbourhood in the case of finite space with periodic boundaries , provided in equation ( [ eq : neighborperiodic ] ) .
as boundaries are periodic , the site @xmath427 is connected to the line @xmath333 also if the path of open vertices leaves one of the vertical boundaries ( @xmath472 or @xmath473 ) from one side and it re - appears at the same high on the other side ( see for example the path @xmath471 in figure [ fig : events1 ] ) . in this section
we list some properties involving the cluster of vertices belonging to an open path in @xmath62 starting from @xmath63 .
these properties have been proved in @xcite in case of a bond percolation model with symmetric neighbourhood of two elements . in this article
we consider a slightly different percolation model , as sites instead of bonds can be open or closed and the neighbourhood is an arbitrary ( translation invariant ) finite set .
the proofs of these propositions in the case considered in this article are similar to those provided in @xcite .
we sketch their proof describing the small differences .
we start with some definitions . from now on we will consider @xmath19 . for every @xmath474 we define the sets , @xmath475 note that @xmath476 .
we define then the variables , @xmath477 and we set @xmath478 , @xmath479 if @xmath480 . as the distributions of @xmath481 , @xmath482 , @xmath483 , @xmath484 , @xmath485 , @xmath486 and @xmath487 depend only on the difference @xmath488 , from now on we will omit the dependence on @xmath116 , that will be some positive integer .
furthermore we consider the space @xmath62 as before , but with vertices @xmath489 instead of @xmath490 . in the former case ,
if we consider only paths starting from @xmath63 , we allow @xmath63 to belong to an infinite open path .
thus we recover the notation of @xcite ( @xmath491 , @xmath492 , @xmath493 , @xmath494 ) , with the difference that in this article paths are oriented from up to down .
equation ( [ eq : proprl1 ] ) is a corollary of equations ( [ eq : proprl2 ] ) and ( [ eq : proprl3 ] ) .
we sketch an argument for ( [ eq : proprl3 ] ) , that can be also found in ( * ? ? ?
* section 3 ) . by reflection
the same argument holds also for ( [ eq : proprl2 ] ) .
it is trivial from the definition that @xmath500 and that @xmath501 $ ] .
we have to show that @xmath502 \subset \xi_m$ ] . in this case
it is clear from figure [ fig : rnrml ] that if there is a path from some site @xmath503 , @xmath504 to @xmath505 , @xmath506 , then there is also a path from @xmath63 to @xmath507 .
then @xmath508 .
we introduce the following quantities , for all integers @xmath509 , @xmath510 @xmath511 the following relations holds .
@xmath512 @xmath513 we prove ( [ eq : subadditive ] ) and a similar argument holds for ( [ eq : subadditivel ] ) .
one should observe that @xmath514 is the rightmost point on the line @xmath515 which can be reached from any of the points @xmath516 with @xmath517 .
instead @xmath518 is the rightmost point on the line @xmath515 which can be reached from any of the points @xmath516 with @xmath517 and with the additional restriction that there exists an open path in @xmath62 from @xmath519 to @xmath516 for some @xmath520 .
see also figure [ fig : rnrm ] .
the next proposition involves the random variables defined above and it corresponds to @xcite[theorem 2.1 ] .
it holds for a class of model called _ growth processes _ that is more general than the class of models considered here .
we refer to @xcite for its proof , which is based on the subadditivity property of ( [ eq : subadditive ] ) and some arguments similar to those used in the proof of the kingman s subadditive ergodic theorem .
[ prop : convergencesubadditive ] let @xmath493 and @xmath521 be the quantities defined above . then there exist two constants @xmath522 $ ] and @xmath523 $ ] such that , @xmath524 @xmath525 let @xmath526 be the following event , @xmath527 then , if @xmath8 , conditioning on @xmath528 , from proposition [ prop : convergencesubadditive ] and from equations ( [ eq : proprl1 ] ) , ( [ eq : proprl2 ] ) , ( [ eq : proprl3 ] ) the following properties hold , @xmath529 @xmath530 @xmath531 if @xmath17 then the event @xmath528 occurs with positive probability . conditioning on @xmath528 , for all @xmath532 @xmath533 .
furthermore , from equations ( [ eq : proprl2 ] ) and ( [ eq : proprl3 ] ) it follows that @xmath534 and @xmath535 .
we define now the variable , @xmath536 which plays the role played by @xmath537 in @xcite .
the proof of the next proposition can be found in ( * ? ? ?
* section 3 ) , in case of bond percolation with symmetric neighbourhood . as the statement is needed for the proof of theorem [ theo : maintheorem2 ] , we sketch its proof , adapting it to the model considered in this article .
[ prop : pcgamma2 ] let @xmath538 be the variable defined in equation ( [ eq : gamma ] ) .
then , @xmath539 observe that equation ( [ eq : r_n ] ) implies that , @xmath540 then , to prove equation ( [ eq : pcgamma0 ] ) , first it is necessary to show that , @xmath541 indeed , equations ( [ eq : pcgamma1 ] ) and ( [ eq : pcgamma2 ] ) imply that @xmath542 hence , it remains to exclude the possibility that the interval @xmath543 has positive length .
this fact is a consequence of the following property , @xmath544 and of the fact that @xmath545 is non - decreasing with @xmath4 . for the proof of ( [ eq : pcgamma4 ] )
we refer to ( * ? ? ?
* section 3 ) , as the symmetry of the neighbourhood does not play any role in the proof .
the proof is based on the construction of two systems with parameter @xmath4 and @xmath546 on the same space by assigning an independent random variable @xmath547 to each vertex @xmath403 which is uniformly distributed on @xmath108 .
the vertex is open if @xmath548 the parameter value and closed otherwise .
the only difference from @xcite is that there these random variables are assigned to bonds and that the set of vertices of the graph is different , i.e. @xmath549 s. t. @xmath550 is even@xmath551 . in the remaining part of the proof we prove equation ( [ eq : pcgamma2 ] ) .
observe that if @xmath552 , then @xmath553 and @xmath554 almost surely .
then there exists an integer @xmath555 such that , @xmath556 secondly we introduce the following notation . if @xmath557 , then we let @xmath558 repeating the proof of ( [ eq : proprl1 ] ) , ( [ eq : proprl2 ] ) , ( [ eq : proprl3 ] ) ( see also ( * ? ? ?
* section 3 , equations 10 ) ) , it follows that @xmath559 } & = \inf\{m \ , : \ , r_m^{[-m , m ] }
< \ell_m^{[-m , m ] } \ } \\ & = \inf\{m \ , : \ , r_m^{[-\infty , m ] } <
\ell_m^{[-m , \infty ] } \ } . \end{split}\ ] ] the previous equality implies that , @xmath560 } = \infty \ } \supset \ { \ell_m^{[-m , \infty ) } \leq \frac{\alpha + \beta}{2 } m \leq r_m^{(-\infty , m ] } , \ , \ , \ , \forall m \}.\ ] ] as @xmath561 } > \frac{\alpha + \beta}{2 } m , \ , \ , \ , \forall m ) = \mathbb{p}_p ( r_m^{(-\infty , 0 ] } > \frac{\alpha + \beta}{2 } m - m,\ , \ , \ , \ , \forall m),\ ] ] and @xmath562 it follows that @xmath563 } \neq \varnothing,\ , \ , \ , \forall m ) \geq 0.02.\ ] ] since , @xmath564 , @xmath565 ) > 0,\ ] ] it follows that @xmath566
. then @xmath17 .
the next estimates have been proved in @xcite .
the proof can be found also in ( * ? ? ?
* section 7 , equations ( 1 ) and ( 2 ) ) .
in particular equation ( [ eq : exponentialspeed ] ) holds for a wide class of percolation models in the subcritical regime ( see @xcite for a proof in a very general setting ) . [
prop : subcritical ] recall definition [ def : opensitesandpaths ] . for every @xmath4 ,
let @xmath567 and @xmath568 .
if @xmath102 there exist some positive constants @xmath569 ( dependent on @xmath4 ) such that , @xmath570 we sketch the proof of ( [ eq : exponentialspeed ] ) , which is similar to the proof of [ eq : exponentialspeed2 ] ) and [ eq : exponentialspeed3 ] ) . if @xmath571 , then from equation ( [ eq : pcgamma0 ] ) @xmath572 .
thus there exists an @xmath573 large enough such that @xmath574 < \frac{\alpha + \beta}{2 } n$ ] , @xmath575 >
\frac{\alpha + \beta}{2 } n$ ] . by using the subadditivity property of @xmath576 and
@xmath577 one can see that , @xmath578 the right hand side of the two previous inequalities is a random walk with expectation respectively @xmath579 < 0 $ ] , @xmath580 < 0 $ ] . as @xmath581 , @xmath582 , then @xmath583 < \infty$ ] and @xmath584 < \infty$ ] for all @xmath585 . from the considerations in @xcite
it follows that we can pick @xmath586 with @xmath587 and @xmath588 such that , @xmath589 = \varphi(\theta_0)^m,\ ] ] @xmath590 = \varphi^{\prime}(\theta_0)^m.\ ] ] this implies that @xmath591 and @xmath592 exponentially fast .
observe also that as @xmath593 , then @xmath594 this implies ( [ eq : exponentialspeed ] ) .
we end this section recalling a property proved in @xcite . as the reference is in russian , we sketch its proof below .
[ prop : mityushin ] recall definition [ def : opensitesandpaths ] . for
every @xmath595 , @xmath596 observe that in @xmath489 all paths of length @xmath116 starting from @xmath63 lie within @xmath597 \times [ 0,t-1]$ ] . at each point
we have a random variable @xmath414 that is equal to @xmath16 with probability @xmath4 and to @xmath173 with probability @xmath402 and these random variables are mutually independent .
we consider the same set @xmath598 but with a different set of random variables @xmath599 . each @xmath599 is equal to @xmath16 with probability @xmath4 and @xmath173 with probability @xmath402 , but these random variables are not independent .
namely , for all @xmath600 , the random variables @xmath601 for all integers @xmath166 such that @xmath602 have the same outcome ( i.e. they are `` synchronized '' ) . this model is equivalent to the model on the cylinder @xmath603 $ ] ( i.e. with periodic boundaries ) , where @xmath414 are independent , because in these two models their probabilistic spaces and sets of open paths starting at @xmath63 are isomorphic .
let then @xmath604 be the set of all possible paths of length @xmath116 from @xmath605 .
we will show that `` synchronization '' does not increase the probability of the existence of an open path of length @xmath116 on @xmath598 .
let then @xmath606 be some random variables with values @xmath173 or @xmath16 associated with @xmath607 .
consider the function @xmath608 , with arguments @xmath606 , @xmath609 then @xmath610 and @xmath611 if and only if there exists an open path .
suppose that at the beginning @xmath612 , for all @xmath607 and at each step we `` synchronize '' the variables @xmath613 for a certain @xmath614 until we get @xmath615 for all @xmath607 .
we will show that each synchronization step does not increase @xmath608 . to do this , we write @xmath616 where @xmath617 is the set of all @xmath618 , i.e. they are independent from the group @xmath613 .
the @xmath619 and @xmath620 are some functions with non - negative integer values . here
we use the fact that a path can contain only one point of the form @xmath621 , so different @xmath613 do nt multiply . before the `` synchronization '' step , @xmath622 and after it , @xmath623 it is easy to show that , fixing any value of the set @xmath617 , @xmath624 .
hence , the same is true when @xmath617 is not fixed .
recall the definitions provided just before the statement of the theorem . along the whole proof
we denote by @xmath625 the bernoulli product measure in @xmath78 , where the space is finite , and by @xmath626 the bernoulli product measure in @xmath78 , where the space is infinite . the proof is based on the estimation of @xmath627 , which gives the expectation , @xmath628 = \sum _ { t = 0 } ^{\infty } \mathcal{e}^n_{{\delta}_{\mathbf{1 } } } ( \tau_n > t ) , \ ] ] we prepare the reader to the proof of the left inequality of the statement ( b ) .
the proofs of the other inequalities do not need an introduction , as they are simpler .
the proof is based on the estimation of the probability of the event @xmath629 . in order to provide this estimation , first we define the event @xmath630 , whose probability is less than the probability of @xmath629 .
the event occurs if a path connects @xmath631 \times \{t\}$ ] to the line @xmath333 without crossing the diagonal sides of a parallelogram ( a rigorous definition is given later ) .
this allows to reduce the estimation of @xmath629 to the estimation of the probability of an event that is simpler to study , as periodic boundaries play no role . as the neighbourhood of the model is in general non symmetric
, the cluster of vertices belonging to an open path starting from @xmath63 ( which is infinite with positive probability , as @xmath8 ) will typically have a drift .
indeed , recall proposition [ prop : convergencesubadditive ] and the fact that @xmath632 and @xmath633 , @xmath634 .
thus , as @xmath4 is slightly larger than @xmath13 , then typically the cluster of vertices will be centred around @xmath635 . hence
, the diagonal sides of the parallelogram is chosen in such a way that in the limit @xmath447 the cluster has typically a non - empty intersection with the parallelogram . with this choice
, the probability of the event @xmath630 does not go to zero too fast as @xmath116 grows .
later we introduce a change of coordinates @xmath636 that allows to simplify the notation , by transforming the graph in a new graph , where the cluster of vertices connected by an open path starting from @xmath63 ( namely in the new graph @xmath637 ) .
we provide a lower bound for the probability of @xmath630 by introducing a new event @xmath638 and by using the fkg inequality to bound the probability of @xmath630 with a product of probabilities of events @xmath638 . in the last part of the proof we estimate the probability of the event @xmath638 , showing that it goes to @xmath16 fast enough with @xmath231 for any @xmath639 .
this estimation is stated in proposition [ prop : inequality2 ] .
[ [ proof - of - part - a . ] ] proof of part ( a ) .
+ + + + + + + + + + + + + + + + + + the proof of the left inequality of the statement ( a ) can be found in ( * ? ? ?
* section 2 ) together with an estimation of the constants , so we do not provide it here . indeed , the left inequality holds for any @xmath44 $ ] .
the proof of the right inequality of statement ( a ) is an application of the estimates presented in section [ sect : percolationestimates ] . starting from ( [ eq : time ] ) , @xmath640 & = \sum _ { t = 1 } ^{\infty } \mathbb{p}_p^n ( \bigcup_{s= -n}^{n-1 } \ { ( s , t ) \overset { \mathcal{g}_{\mathcal{u}}(n)}{\longrightarrow } s^0 \ } ) \\ & \leq \sum _ { t = 1 } ^{\infty } \min \ { 1 , 2 n \ , \mathbb{p}^n ( ( 0,t ) \overset { \mathcal{g}_{\mathcal{u}}(n)}{\longrightarrow } s^0 ) \ } \\ & \leq \sum _ { t = 1 } ^{\infty } \min \ { 1 , 2 n \exp(- h t ) \ } \\ & \leq \frac{\log ( 2\ , n)}{h } + k , \end{split}\ ] ] where @xmath29 is some positive constant . in the first equality we used proposition [ prop : taunconnection ] , in the second inequality we used the union bound and translation invariance , in the second - last inequality we used ( [ eq : exponentialspeed ] ) and ( [ eq : ineqfininf ] ) . the algebraic tricks of ( [ eq : partea ] ) have been used also in the proof of ( * ? ? ?
* proposition 8.6 ) .
[ [ proof - of - part - b . ] ] proof of part ( b ) .
+ + + + + + + + + + + + + + + + + + the proof of the right inequality of the statement ( b ) is trivial .
we define a new process @xmath641 where every @xmath642 is @xmath16 with probability @xmath4 and @xmath173 with probability @xmath402 independently .
observe that for all @xmath643 , @xmath417 , @xmath644 , as long as the two processes are driven by the same random process .
hence , the expected convergence time for the former is larger than the convergence time for the latter . by a simple computation ,
the expected convergence time for the system @xmath645 is @xmath646 .
this implies the inequality .
we start with the proof of the left inequality . for every @xmath647
we define the event , @xmath648 , \ , \ , \mbox { such that } ( x , t ) \mbox { is connected to } s^0 \\ & \mbox { by an open path in } \mathcal{g}_{\mathcal{u } } \mbox { that never crosses the lines } \\ & y = n -1 - a ( x - t ) , \ , \ , y = - n - a ( x - t)\ , \ , \ , x \in \mathbb{r } \ } , \end{split}\ ] ] which is a subset of @xmath411 , recalling that @xmath459 denotes the set of vertices belonging to the line @xmath333 .
see also figure [ fig4 - 5:dn ] - up for a representation . in case of neighbourhood @xmath649 . for graphical reasons only edges belonging to the evolution cone of @xmath650 have been drawn . in the figure @xmath651 .
_ down _ : the same graph of the figure above , transformed via ( [ eq : transf ] ) with parameter @xmath652.,title="fig : " ] in case of neighbourhood @xmath649 . for graphical reasons
only edges belonging to the evolution cone of @xmath650 have been drawn . in the figure @xmath651 .
_ down _ : the same graph of the figure above , transformed via ( [ eq : transf ] ) with parameter @xmath652.,title="fig : " ] recall definition [ def : opensitesandpaths ] and observe that , @xmath653 \ , \ , \mbox{s.t . } \ , \ , ( x , t ) \overset { \mathcal{g}_{\mathcal{u } } ( n ) } { \longrightarrow } s^0 ) = \mathcal{e}^n_{\delta_{\mathbf{1 } } } ( \tau_n > t ) .\
] ] observe that the quantity on the left is defined in the infinite system and the quantities in the middle and on the right are defined on the finite system with periodic boundaries .
we provide a proof of the statement below .
consider two graphs , @xmath654 and @xmath655 .
the former is defined on the infinite space @xmath656 and the latter on the finite space @xmath657 with periodic boundaries , as defined in section [ sect : themodel ] .
let @xmath658 be the region inside the parallelogram identified by the points @xmath659 , @xmath660 , @xmath661 , @xmath662 ( see figure [ fig : proof ] ) .
the event @xmath663 occurs if an open path connects @xmath664 \times \{t\}$ ] to @xmath665 \times \ { 0 \}$ ] without ever crossing the diagonal sides of the parallelograms .
we couple the model on the finite space and the model on the infinite space in the following way .
namely , call @xmath666 , for all @xmath667 , and @xmath599 , for all @xmath668 , the random variables taking values @xmath173 or @xmath16 independently .
the coupling is such that for all @xmath669 , @xmath670 , where @xmath671 , @xmath672 , where @xmath673 denotes @xmath674 .
the random variables @xmath414 associated to sites @xmath600 not contained in @xmath675 are not coupled .
observe that for every @xmath669 there exists a unique @xmath676 in @xmath132 and vice versa .
recalling that boundaries of @xmath655 are periodic , one can observe from figure [ fig : proof ] that , as long as there exists an open path in @xmath654 connecting the top to the bottom of @xmath675 and never crossing its diagonal sites ( e.g. the path represented by a continuous curve in the figure ) , there exists also an open path in @xmath655 connecting @xmath664 \times \{t\}$ ] to @xmath664 \times \{0\}$ ] ( i.e. the path represented by a dashed curve in the figure ) .
this implies the statement .
consider now the following change of coordinates , @xmath677 under which the graph @xmath62 is transformed into the new graph @xmath678 .
we denote by @xmath679 the event @xmath680 , defined for the graph @xmath678 , ( i. e. replace @xmath62 with @xmath678 in the definition of the event above ) .
see figure [ fig4 - 5:dn ] for an example .
the following equation holds , @xmath681 as the change of coordinates preserves connection between vertices .
now we introduce the event @xmath638 , @xmath682 , \ , y^{\prime } \in [ [ 0 , 2n ] ] \\ & \mbox{and } ( -n , y ) \overset{\mathcal{g}_{\mathcal{u}}}{\longrightarrow } ( n , y^{\prime } ) \ } , \end{split}\ ] ] which is represented in figure [ fig6 - 7:events]-right .
the following proposition is about this event .
[ prop : inequality2 ] for any @xmath8 there exist positive constants @xmath683 ( dependent on @xmath4 ) such that for any @xmath684 and for @xmath231 large enough , @xmath685 as before , the event @xmath686 denotes the occurrence of @xmath638 in the graph @xmath687 .
we recall that @xmath537 and @xmath688 are defined in section [ sect : percolationestimates ] .
we first use proposition [ prop : inequality2 ] to conclude the proof of the theorem and later we prove the proposition [ prop : inequality2 ] . define then the new event @xmath689 , which is represented in figure [ fig6 - 7:events ] .
@xmath690 occurs _
iff _ ( a ) and ( b ) hold : 1 . for every odd
@xmath691 $ ] there is a vertex @xmath692 , with @xmath693 $ ] , connected to @xmath694 by an open path in @xmath62 , with @xmath695 $ ] , 2 . for any even @xmath696 $ ]
there is a vertex @xmath697 , with @xmath698 $ ] , connected by an open path in @xmath62 to @xmath699 , with @xmath700 $ ] .
note first that , @xmath701 because if @xmath702 occurs , then the top of the box @xmath703 is connected to the bottom by a path that never goes out from the box ( compare figures [ fig6 - 7:events]-left and [ fig6 - 7:events]-middle ) . secondly , we observe that the event @xmath704 equals the intersection of @xmath705 events of type @xmath706 , represented in figure [ fig6 - 7:events]-right . .
_ center _ : representation of the event @xmath707 .
_ right _ : representation of the event @xmath638 . in all figures
the details on the structure of the graph have been omitted.,title="fig : " ] .
_ center _ : representation of the event @xmath707 .
_ right _ : representation of the event @xmath638 . in all figures
the details on the structure of the graph have been omitted.,title="fig : " ] .
_ center _ : representation of the event @xmath707 .
_ right _ : representation of the event @xmath638 . in all figures
the details on the structure of the graph have been omitted.,title="fig : " ] as the event @xmath638 increasing , the fkg inequality is applicable , i.e. @xmath708 then using ( [ eq : inequality2 ] ) finally we get , @xmath709 then , from ( [ eq : time ] ) and for @xmath231 large enough , @xmath710 & \geq \sum _ { t = 1 } ^{\infty } { \mathbb{p}_p ( t_{\frac{\alpha + \beta}{2}}^t \ , \mathcal{d}_{n , t , 0 } ) } \\ & \geq \sum _ { t = 1 } ^{\infty } { \mathbb{p}_p ( t_{\frac{\alpha + \beta}{2}}^t\ , \mathcal{f}_{n , t } ) ) } \\ & \geq \sum _ { t = 1 } ^{\infty } { ( 1 - a \cdot \exp ( - b n ) ) } ^{\lfloor \frac{t } { n } \rfloor } \\ & \geq j ( 1 - \frac{a \cdot e^{-b n } j } { n } ) , \end{split}\ ] ] where @xmath311 is an arbitrary integer . in the previous expression
we have used proposition [ prop : taunconnection ] , ( [ eq : ineq ] ) and ( [ eq : transformation ] ) for the first inequality , ( [ eq : ineq22 ] ) for the second inequality and ( [ eq : fkg2 ] ) for the third one .
choosing finally @xmath711 , the part ( b ) of the theorem follows .
we prepare the reader to the proof of the proposition and later we present the proof .
we consider two graphs , @xmath712 and @xmath713 , recalling the definitions of @xmath537 and @xmath688 in section [ sect : percolationestimates ] and the definition of the transformation @xmath714 provided in ( [ eq : transf ] ) .
observe that vertices of both graphs could take non integer positions .
the proof is divided in two parts . in the * first part
* we generalize the dynamic - block argument presented in @xcite to the percolation model considered in this article .
the idea of the construction is the same of @xcite , although parameters of the construction have been adapted to the lack of symmetry .
the lack of symmetry involves the structure of the graph @xmath62 and the slope of @xmath493 and @xmath521 , as in general @xmath715 .
two different spatial transformations have been used in order to recover the symmetric setting and simplify the construction , namely @xmath716 and @xmath717 .
the argument is based on a coupling between realisations of the graph @xmath718 and those in @xmath719 .
the construction depends on a rescaling parameter @xmath720 and it is such that the realisation on @xmath721 is a function of the realisation on @xmath719 . in @xmath719
every site is open with probability @xmath4 or closed with probability @xmath402 independently . on the contrary , the states of sites in @xmath721 are not independent .
the construction is such that if the event @xmath638 occurs in the former graph , then the event @xmath722 occurs in the latter graph .
secondly , if @xmath17 , then for every @xmath723 , by choosing @xmath720 is large enough , every site in @xmath721 is open with probability larger than @xmath724 . the * second part * we define a sub - graph of @xmath718 , that we call @xmath725 , for which it is easy to construct a dual graph .
we use peierls argument for the dual graph and we show that @xmath726 . as far as we know , this estimation has not been provided in other works . the dual graph construction can be found in @xcite .
this implies that @xmath727 . recalling the properties of the construction
, it follows that @xmath728 . by rearranging the constants , the statement of the proposition follows .
we start now with the proof of the proposition .
[ [ part-1-dynamic - blocks - construction ] ] part 1 : dynamic blocks construction we divide @xmath719 into _ macro - regions _
@xmath729 centred around the point @xmath730 , where @xmath731 and @xmath732 \times [ 0 , - ( 1 + \delta ) l ] .
\end{split}\ ] ] we recall that from equation ( [ eq : pcgamma0 ] ) @xmath733 for all @xmath17 . the constants @xmath734 and @xmath720 are positive and have to be properly chosen . in order
the argument to work rigorously , @xmath735 and @xmath720 should be even integers . to not complicate the exposition here
we ignore these details , the same as in @xcite .
each vertex @xmath731 is associated to a random variable @xmath736 which takes value @xmath16 if a certain event @xmath737 occurs in the region @xmath729 of @xmath738 or @xmath173 otherwise . in order to define such event
we introduce the following points in space ( see also figure [ fig8:sketchren ] ) , for every @xmath739 , @xmath740 and for every @xmath741 , @xmath742 of the graph @xmath743 , for some positive integer @xmath116 , in case @xmath744 .
, title="fig : " ] of the graph @xmath743 , for some positive integer @xmath116 , in case @xmath744 .
, title="fig : " ] as one can see in the example in figure [ fig8:sketchren ] , these points identify some _
target zones _ on the right and on the left of points @xmath745 , @xmath746 .
consider now the parallelograms obtained connecting the following quadruplets of points ( see also figure [ fig8:sketchren ] ) , @xmath747 for all @xmath741 .
define the translated parallelograms @xmath748 , @xmath749 , @xmath750 for all @xmath741 .
the event @xmath737 occurs if and only if the top of all parallelograms @xmath751 , @xmath752 and @xmath753 , for all @xmath741 , is connected to the the bottom side by an open path in @xmath754 that remains always inside the parallelogram .
this event is represented in figure [ fig8:sketchren ] .
this construction is such that the following properties are satisfied .
namely , 1 . the random variables @xmath755 are @xmath245-dependent . with this
we mean that @xmath736 and @xmath756 , with @xmath600 , @xmath757 , are independent if @xmath758 or @xmath759 .
2 . denote by @xmath760 the vertices of a path in @xmath761 and assume that the path is open , i.e. @xmath762 for all @xmath763 .
then there exists an open path in @xmath754 that connects a vertex in the @xmath764 $ ] to a vertex in @xmath765 $ ] and which remains always inside the parallelograms that connect @xmath766 $ ] to @xmath767 $ ] , for all @xmath763 ( note that @xmath768 $ ] denotes the segment @xmath769 $ ] translated by @xmath770 ) .
if @xmath771 and @xmath17 , we can pick @xmath720 large enough so that for any @xmath772 , @xmath773 .
we sketch the proof of the three properties above .
the proof can be found also in ( * ? ? ?
* section 9 ) in the case of bond percolation and symmetric neighbourhoods .
property 1 follows from the fact that if @xmath729 and @xmath774 have empty intersection , then the variables @xmath736 and @xmath775 are independent .
property 2 follows by construction ( see figure [ fig8:sketchren ] ) . in the example in the figure
we represent the graph @xmath776 assuming @xmath744 as a neighbourhood .
one should observe that if the events @xmath777 and @xmath778 occur , then at least one vertex belonging to the interval @xmath779 $ ] is connected to at least one of the vertices belonging to any of the intervals @xmath780 $ ] , @xmath781 $ ] , @xmath782 $ ] .
we now prove the third property .
recall that proposition [ prop : convergencesubadditive ] implies that in the transformed graph @xmath754 , @xmath783 a.s . and
@xmath784 a. s. we will prove that @xmath785 , the probability that in all the parallelograms in the box there is a connection from the top to the bottom that never crosses the diagonal sides is larger than @xmath45 .
let then @xmath786 be the number of parallelograms in the box @xmath787 .
this number depends on the neighbourhood @xmath6 .
we consider the parallelogram @xmath788 and we prove that for every @xmath723 there exists @xmath720 large enough such that the probability that there is no such open path in the parallelogram is less than @xmath789 . as this probability is the same for all parallelograms , this implies that the probability that such open path is present in all parallelograms is @xmath790 .
consider the parallelogram @xmath788 defined above and recall then the definitions provided in equations ( [ eq : notation1 ] - [ eq : notation4 ] ) .
let then @xmath791}$ ] and observe that @xmath792 $ ] .
let @xmath793}$ ] and observe that @xmath794 .
as @xmath795 a.s . in the transformed graph @xmath796
, then we can pick @xmath720 large enough such that with probability @xmath797 we have that , @xmath798 and for all @xmath799 , @xmath800 the two previous equations imply that there is an open path path from @xmath801 \times \{0\}$ ] to @xmath802 \times \{- ( 1 + \delta ) l \}$ ] which does not cross the line @xmath803 $ ] .
it remains to show that this path does not cross the line @xmath804 $ ] .
we observe that in order a path to travel from the line @xmath805 $ ] to @xmath806 a path must have an average slope @xmath807 .
thus recall equation ( [ eq : exponentialspeed2 ] ) and observe that in the transformed graph @xmath808 , @xmath809 consider then @xmath810 large enough such that , @xmath811 the probability that one of the points on @xmath805 $ ] with @xmath812 is connected to @xmath813 is then @xmath814 . furthermore , observe that the number of points on @xmath805 $ ] with @xmath815 does not depend on @xmath720 and that the distance of any of them from the set @xmath813 is proportional to @xmath720 .
thus we can pick @xmath720 large enough so that the probability that there exists an open path connecting any of these points to @xmath816 $ ] is less then @xmath817 . combining the two estimations ,
we conclude that the probability that the line @xmath805 $ ] is connected by an open path to @xmath816 $ ] is less than @xmath818 .
summarising , we showed that with probability @xmath819 , there is an open path from @xmath801 \times \{0\}$ ] to @xmath802 \times \{- ( 1 + \delta ) l \}$ ] which does not cross the line @xmath803 $ ] and the line @xmath805 $ ] is not connected by an open path to @xmath816 \times \{-(1 + \delta ) l \ } $ ] .
this implies that the probability that there exists a path joining the top to the bottom of @xmath788 without ever crossing its diagonal lines is @xmath819 . repeating the argument for all parallelograms in the box
, we conclude that if @xmath720 is large enough then with probability at least @xmath45 the event @xmath820 occurs .
[ [ part-2-peierls - argument ] ] part 2 : peierls argument now we use the peierls argument for the @xmath821-dependent oriented percolation model to prove that there exists @xmath822 and positive constants @xmath823 , @xmath824 ( dependent on @xmath4 ) such that for all @xmath825 $ ] , @xmath826 let us explain first why this is sufficient to prove the proposition
. later we prove ( [ eq : inequality3 ] ) .
recall the third property of the dynamic - block construction presented above and observe that if @xmath827 , then we can pick @xmath720 large enough such that , for every @xmath600 belonging to the set of vertices of @xmath828 , @xmath829 .
recall that the state of sites belonging to @xmath828 is a function of the realization in the graph @xmath830 . from the second property of the dynamic - block construction ,
if such sites are open with probability @xmath831 , then ( [ eq : inequality3 ] ) implies that with probability not less than @xmath832 the event @xmath638 occurs in @xmath833 .
hence , from the second property of the dynamic - block construction , the event @xmath834 occurs in the graph @xmath754 with probability not less than @xmath832 .
one can rearrange the value of @xmath824 getting rid of the factor @xmath231 , for @xmath231 large enough .
finally , by defining new constants @xmath835 and @xmath836 , the statement of proposition [ prop : inequality2 ] follows .
we start proving ( [ eq : inequality3 ] ) .
we define a new graph @xmath725 , that is a sub - graph of @xmath837 , whose vertices @xmath600 are , @xmath838 and whose edges connect vertices @xmath600 to @xmath839 .
the reason shy we introduce @xmath725 is that , as every site has only two neighbours , it is easier to construct its dual graph .
the new graph @xmath725 is represented in the example in figure [ fig : transformatonpeierls ] on the right .
, in case of @xmath840 .
_ right _ : points correspond to vertices of @xmath841 , arrows represent edges of @xmath725 , points connected by an arrow correspond to vertices of @xmath725 .
the graph @xmath725 , defined in the text , is a subset of @xmath841 . ]
as @xmath725 is a sub - graph of @xmath837 , the following inequality holds , @xmath842 in the previous expression , the superscript @xmath843 is used to denote that event @xmath844 , defined in ( [ eq : evh ] ) , occurs on the graph @xmath725 .
call then @xmath845 the dual graph of @xmath725 .
the graph is represented on the right of figure [ fig9 - 10:eventshn ] and its costruction is due to @xcite . in both graphs .
_ left _ : representation of the graph @xmath725 .
the event @xmath638 occurs iff the side @xmath847 is connected to the side @xmath848 by an open path in @xmath725 that does not cross the sides @xmath849 and @xmath850 .
_ right _ : representation of the graph @xmath851 , as defined in the text .
the event @xmath638 does not occur _ iff _ one of the sides @xmath852 or @xmath853 is connected to one of the sides @xmath854 or @xmath855 by an open path in the dual lattice.,title="fig : " ] in both graphs .
_ left _ : representation of the graph @xmath725 .
the event @xmath638 occurs iff the side @xmath847 is connected to the side @xmath848 by an open path in @xmath725 that does not cross the sides @xmath849 and @xmath850 .
_ right _ : representation of the graph @xmath851 , as defined in the text .
the event @xmath638 does not occur _ iff _ one of the sides @xmath852 or @xmath853 is connected to one of the sides @xmath854 or @xmath855 by an open path in the dual lattice.,title="fig : " ] the dual graph is composed of three types of edges , namely edges pointing down - left , those pointing up - left and those pointing right . every edge pointing right is positioned over a vertex of the original graph @xmath725 .
edges down - left and up - left are always _ open _ , edges pointing right are _ open _ if and only if the corresponding vertex of the original graph is _
closed_. a path in the dual graph is _ open _ if and only if all its edges are open .
the following proposition connects the occurrence of the event @xmath638 in @xmath725 with the occurrence of a second event on the dual lattice .
[ prop : percdual ] consider figure [ fig9 - 10:eventshn ] . for every @xmath856
, there exists an open path in @xmath725 connecting @xmath857 to @xmath858 _ iff _ there is no open path in the dual lattice connecting one of the sides @xmath852 or @xmath853 to one of the sides @xmath854 or @xmath855 .
we provide a graphical proof .
consider figure [ fig9 - 10:eventshn ] .
on the left we have represented the graph @xmath725 and on the right we have represented its dual . consider a realisation in the auxiliary space @xmath411 and recall that if a site is open in @xmath725 , then the corresponding horizontal edge is closed in the dual graph and vice versa .
the reader should observe that , as long as there is an open path connecting @xmath857 to @xmath858 in @xmath725 , no open path in the dual graph connecting one of the sides @xmath852 or @xmath853 to one of the sides @xmath854 or @xmath855 can exist .
on the other hand , as long there exists an open path in the dual graph connecting one of the sides @xmath852 or @xmath853 to one of the sides @xmath854 or @xmath855 , no open path in @xmath725 connecting @xmath857 to @xmath858 can exist . both the proposition and the dual construction are analogous to the one presented in @xcite .
we use this proposition to provide a lower bound for @xmath859 . consider then a vertex @xmath860 on @xmath852 or on @xmath853 . call @xmath861 the set of paths connecting the vertex @xmath860 to one of the sides @xmath854 or @xmath855 and having @xmath862 edges pointing to the right .
call @xmath863 the total number of such paths .
consider one of these paths and call @xmath864 the number of its edges pointing down - left and @xmath865 the number of edges pointing up - left . as the last edge of the path can not be on the left of the first edge , @xmath866 .
this implies that for each of these paths sum @xmath867 is bounded from above by @xmath868 . as there are only @xmath869 different types of steps , for any vertex @xmath860 located on @xmath852 or on @xmath853 , @xmath870 .
thus @xmath870 for every @xmath860 .
denote by @xmath871 the complementary of @xmath872 .
recall proposition [ prop : percdual ] and observe the fact that , in order @xmath873 to be connected to @xmath854 or to @xmath874 , at least @xmath875 horizontal steps to the right are needed .
then , @xmath876 .
observe also that , given a path @xmath877 , @xmath878 , considering only the state of one every two edges to the right , as states of edges located over non - neighbour sites are independent . by using the union bound , we determine an upper bound for @xmath879 , @xmath880 where the second inequality is true with @xmath881 positive constants if @xmath882 .
we consider percolation pca with space @xmath132 and periodic boundaries .
we run the process @xmath883 times and we define @xmath884 where @xmath885 is the number of times the origin has state @xmath16 at time @xmath159 . in figure
[ fig : simulations ] we plot the function @xmath886 for different choices of the neighbourhood for a small range of @xmath4 .
the parameters considered are @xmath887 , @xmath888 , @xmath889 and @xmath890 , @xmath891 and @xmath892 . .
up right : @xmath893 .
middle left : @xmath894 .
middle right : @xmath895 .
down : @xmath896.,title="fig : " ] . up
right : @xmath893 .
middle left : @xmath894 .
middle right : @xmath895 .
down : @xmath896.,title="fig : " ] .
up right : @xmath893 .
middle left : @xmath894 .
middle right : @xmath895 .
down : @xmath896.,title="fig : " ] .
up right : @xmath893 .
middle left : @xmath894 .
middle right : @xmath895 .
down : @xmath896.,title="fig : " ] .
up right : @xmath893 .
middle left : @xmath894 .
middle right : @xmath895 .
down : @xmath896.,title="fig : " ]
the author is grateful to artem sapozhnikov for teaching him many techniques used in the proof of theorem 2.2 .
the author thanks andrei toom , artem sapozhnikov and the anonymous referees whose comments considerably helped to improve the presentation .
the author thanks leonid mityushin for sending him the english version of the proof of proposition 9 .
the author is grateful to jrgen jost for giving him the possibility to study probabilistic cellular automata .
the author also thanks the organizers of the workshop probabilistic cellular automata - eurandom ( june 2013 , eindhoven ) , where many inspiring discussions favoured the development of this work .
h. hinrichsen , nonequilibrium critical phenomena and phase transitions into absorbing states .
_ lectures held at the international summer school on problems in statistical physics xi _ , leuven , belgium , ( september 2005 ) .
a. l. toom , cellular automata with errors : problems for students of probability .
_ topics in contemporary probability and its applications_. ed j. laurie snell .
probability and stochastic _ ed . by richard durret and mark pinsky .
crc press .
pp 117 - 157 , ( 1995 ) .
a. l. toom , n. b. vasilyev , o. n. stavskaya , l. g. mityushin , g. l. kurdyumov and s. a. pirogov , discrete local markov systems . in : _
stochastic cellular systems : ergodicity , memory , morphogenesis _
, r. l. dobrushin , v. i. kryukov and a. l. toom ( eds . ) , manchester university press , 1 - 182 , ( 1990 ) .
|
this paper considers a class of probabilistic cellular automata undergoing a phase transition with an absorbing state . denoting by @xmath0 the neighbourhood of site @xmath1 ,
the transition probability is @xmath2 if @xmath3 or @xmath4 otherwise , @xmath5 . for any @xmath6 there
exists a non - trivial critical probability @xmath7 that separates a phase with an absorbing state from a fluctuating phase .
this paper studies how the neighbourhood affects the value of @xmath7 and provides lower bounds for @xmath7 .
furthermore , by using dynamic renormalization techniques , we prove that the expected convergence time of the processes on a finite space with periodic boundaries grows exponentially ( resp .
logarithmically ) with the system size if @xmath8 ( resp .
@xmath9 ) .
this provides a partial answer to an open problem in toom _ et al . _
( 1990 , 1994 ) .
|
in qcd simulations at small lattice spacings @xmath0 , algorithms typically have severe problems in generating transitions between different topological sectors .
this problem of topology freezing is expected to appear for any lattice discretization of the quark and gluon fields @xcite . for certain discretizations ,
e.g. chirally symmetric quarks , this problem is even present on coarser lattices@xcite . in specific cases it might be motivated to fix topology on purpose .
for example , when using a mixed action setup with light overlap valence and wilson sea quarks , one observes a rather ill - behaved continuum limit @xcite .
this is due to near - zero modes of the dirac operator in the valence sector , which are not compensated by corresponding modes in the sea . a possibility to circumvent
this imbalance could be to restrict the lattice simulation to topological charge @xmath1 , where such near - zero modes are absent , e.g. by employing topology conserving actions @xcite .
methods to extract physically meaningful results from simulations at frozen or fixed topology have been proposed @xcite and tested @xcite in various models and theories . in this work
we extend these methods by also including ordinary finite volume effects .
such a combined treatment of both fixed topology and ordinary finite volume corrections is expected to be particularly important for qcd at light @xmath2 quark masses .
we test our equations in su(2 ) yang - mills theory at fixed topology .
in @xcite an equation has been derived relating a hadron mass @xmath3 obtained at fixed topological charge @xmath4 and finite volume @xmath5 to its counterpart @xmath6 at unfixed topology ( i.e. the physically meaningful hadron mass ) , @xmath7 where @xmath8 denotes the second derivative of @xmath6 with respect to the @xmath9 angle at @xmath10 , and @xmath11 the topological susceptibility .
this equation illustrates that fixed topology corrections are finite volume effects , i.e. effects suppressed by inverse powers of @xmath5 .
it is straightforward to extract a physical hadron mass @xmath6 from computations at fixed topology : one just has to fit eq .
( [ eqn001 ] ) to the available fixed topology and finite volume hadron masses @xmath3 , where @xmath6 , @xmath8 and @xmath11 are the fit parameters ( examples of this procedure can be found in @xcite ) . in fig .
[ fig : result_tfv ] we show recent results for su(2 ) yang - mills theory ( standard plaquette action , gauge coupling @xmath12 , i.e. lattice spacing @xmath13 @xcite ) .
the static potential @xmath14 at separation @xmath15 ( which can be interpreted as a mass ) , has been computed in different topological sectors with topological charges @xmath16 and for different volumes @xmath17 ( @xmath18 gauge link configurations have been generated for each of the four volumes ) .
note that the discrepancies between static potential results @xmath19 at different topological charges @xmath4 are clearly visible , in particular for small volumes @xmath5 .
this demonstrates the necessity of using specific methods to properly deal with topological finite volume effects .
the curves represent a single fit of eq .
( [ eqn001 ] ) to the lattice static potential results @xmath19 .
the fit is of good quality , i.e. @xmath20 .
the resulting @xmath21 is in excellent agreement with a corresponding standard computation at unfixed topology , which yields @xmath22 .
@xmath23 as a function of @xmath24 .
the curves represent the fit of the lattice static potential results to eq .
( [ eqn001]).,width=377 ]
usually lattice simulations are performed at finite spatial volume @xmath25 with periodic boundary conditions .
consequently , a hadron at @xmath26 will interact with images of itself , e.g. at @xmath27 , @xmath28 or @xmath29 .
such interactions cause a shift in the hadron mass compared to infinite spatial volume , as first derived in @xcite . the corresponding equation to describe these ordinary finite volume corrections ( i.e. finite volume corrections not related to fixed topology ) of the static potential in yang - mills theory is @xmath30 where @xmath31 and @xmath32 is the mass of the lightest particle , i.e. the @xmath33 glueball . in fig .
[ fig : oft ] we confront this equation with lattice su(2 ) yang - mills results for @xmath34 and find excellent agreement ( again we have used @xmath12 and generated 4000 gauge link configurations for each of the eight volumes @xmath35 ) . for @xmath36 ordinary finite volume effects are negligible . for smaller @xmath37 , however , there are sizeable corrections , which need to be taken into account , in particular when using such volumes for computations at fixed topology as presented in the previous section . from the fit of eq .
( [ eqn002 ] ) to the lattice results shown in fig .
[ fig : oft ] , one can even extract the @xmath33 glueball mass with remarkable precision , @xmath38 .
this is in perfect agreement with the result obtained by a standard lattice computation of a glueball 2-point correlation function , @xmath39 @xcite . the dependence of @xmath34 on the periodic spatial extension @xmath37 of the lattice ( at unfixed topology).,width=377 ]
in the su(2 ) example discussed in section [ sec001 ] it has been possible to analyze fixed topology results using eq .
( [ eqn001 ] ) in a meaningful way , i.e. without taking ordinary finite volume effects into account .
since the mass of the lightest particle , the @xmath33 glueball , is quite large , ordinary finite volume effects are strongly suppressed for large volumes . as indicated by fig .
[ fig : oft ] and as done in section [ sec001 ] , one just has to discard volumes with @xmath40 . in qcd
the situation is expected to be more difficult , because there the lightest particle , the pion , is much lighter than the @xmath33 glueball of su(2 ) yang - mills theory .
moreover , lattice simulations of qcd , in particular at large volumes , are extremely demanding with respect to high performance computer resources .
therefore , it is highly desirable to combine eqs .
( [ eqn001 ] ) and ( [ eqn002 ] ) , i.e. to obtain an expression describing both topological and ordinary finite volume corrections to hadron masses .
to derive such an expression , one has to consider ordinary finite volume effects also at non - vanishing @xmath9 angles using equations analogous to ( [ eqn002 ] ) .
these equations are the starting point for a lengthy calculation similar to that leading to eq .
( [ eqn001 ] ) ( a detailed derivation of eq .
( [ eqn001 ] ) can e.g. found in @xcite ) .
the resulting expression describing both topological and ordinary finite volume effects takes the form @xmath41 where ordinary finite volume effects for the topological susceptibility @xmath11 have been neglected , since they are expected to be tiny @xcite .
note that in addition to the parameters @xmath6 , @xmath8 and @xmath11 , which are already present in eq .
( [ eqn001 ] ) , there are four more parameters , @xmath32 , @xmath42 , @xmath43 and @xmath44 , characterizing combined topological and ordinary finite volume corrections . in fig .
[ fig003 ] ( top ) we show a plot similar to that from fig .
[ fig : result_tfv ] , this time for @xmath45 .
moreover , also results for small volumes @xmath46 are included .
the curves correspond to eq .
( [ eqn001 ] ) with the fit parameters @xmath6 , @xmath8 and @xmath11 determined by a fit to the large volumes @xmath17 , where ordinary finite volume effects are negligible .
there is a strong discrepancy between these curves and the lattice results for the small volumes @xmath46 .
this is expected , since ordinary finite volume corrections are not part of eq .
( [ eqn001 ] ) in particular for small q. @xmath45 as a function of @xmath24 . * ( top ) : * the curves represent the fit of eq .
( [ eqn001 ] ) to the lattice static potential results for large volumes @xmath17 .
there is a strong discrepancy between these curves and the lattice results for the small volumes @xmath46 . *
( bottom ) : * the curves represent the fit of eq .
( [ eqn003 ] ) to the lattice static potential results for all volumes @xmath47 .
there is almost perfect agreement , even at small volumes and @xmath48.,width=359 ] @xmath45 as a function of @xmath24 . *
( top ) : * the curves represent the fit of eq .
( [ eqn001 ] ) to the lattice static potential results for large volumes @xmath17 .
there is a strong discrepancy between these curves and the lattice results for the small volumes @xmath46 . *
( bottom ) : * the curves represent the fit of eq .
( [ eqn003 ] ) to the lattice static potential results for all volumes @xmath47 .
there is almost perfect agreement , even at small volumes and @xmath48.,width=359 ] .[tab001]results for the static potential @xmath34 , the mass @xmath49 of the @xmath33 glueball , and the topological susceptibility @xmath50 , obtained by a fit of eq .
( [ eqn003 ] ) to fixed topology lattice results @xmath45 . [ cols="<,^,^,^",options="header " , ] in fig .
[ fig003 ] ( bottom ) we show the same lattice results for @xmath45 .
this time , however , the curves correspond to eq .
( [ eqn003 ] ) with the fit parameters @xmath6 , @xmath8 , @xmath50 , @xmath32 , @xmath42 , @xmath43 and @xmath44 determined by a fit to all seven volumes @xmath51 .
there is almost perfect agreement , even at small volumes and for @xmath48 .
the extracted `` hadron mass '' @xmath45 is consistent with a corresponding computation at unfixed topology and also the glueball mass @xmath49 and the topological susceptibility @xmath50 obtained by the fit are in fair agreement with reference values , cf .
table [ tab001 ] . to conclude
, we have incorporated ordinary finite volume corrections into an existing relation between hadron masses at fixed topology and physical hadron masses ( i.e. hadron masses at unfixed topology ) .
we have successfully tested the resulting equation in su(2 ) yang - mills theory by studying the static potential at fixed topology .
as an outlook , we plan to extend these tests to qcd in the near future .
a.d . and m.w .
acknowledge support by the emmy noether programme of the dfg ( german research foundation ) , grant wa 3000/1 - 1 .
this work was supported in part by the helmholtz international center for fair within the framework of the loewe program launched by the state of hesse .
calculations on the loewe - csc high - performance computer of johann wolfgang goethe - university frankfurt am main were conducted for this research .
we would like to thank hpc - hessen , funded by the state ministry of higher education , research and the arts , for programming advice .
m. lscher and s. schaefer , jhep * 1107 * , 036 ( 2011 ) [ arxiv:1105.4749 [ hep - lat ] ] .
s. schaefer , pos lattice * 2012 * , 001 ( 2012 ) [ arxiv:1211.5069 [ hep - lat ] ] .
s. aoki _ et al . _
[ jlqcd collaboration ] , phys .
d * 78 * , 014508 ( 2008 ) [ arxiv:0803.3197 [ hep - lat ] ] .
s. aoki _ et al .
_ , ptep * 2012 * , 01a106 ( 2012 ) .
k. cichy , g. herdoiza and k. jansen , nucl .
b * 847 * , 179 ( 2011 ) [ arxiv:1012.4412 [ hep - lat ] ] .
k. cichy _ et al .
_ , nucl .
b * 869 * , 131 ( 2013 ) [ arxiv:1211.1605 [ hep - lat ] ] .
h. fukaya _ et al .
_ , phys . rev .
d * 73 * , 014503 ( 2006 ) [ hep - lat/0510116 ] .
w. bietenholz _ et al .
_ , jhep * 0603 * , 017 ( 2006 ) [ hep - lat/0511016 ] . f. bruckmann
_ et al . _ ,
j. a * 43 * , 303 ( 2010 ) [ arxiv:0905.2849 [ hep - lat ] ] .
r. brower , s. chandrasekharan , j. w. negele and u .- j .
wiese , phys .
b * 560 * , 64 ( 2003 ) [ hep - lat/0302005 ] .
s. aoki , h. fukaya , s. hashimoto and t. onogi , phys .
d * 76 * , 054508 ( 2007 ) [ arxiv:0707.0396 [ hep - lat ] ] .
w. bietenholz , i. hip , s. shcheredin and j. volkholz , eur .
j. c * 72 * , 1938 ( 2012 ) [ arxiv:1109.2649 [ hep - lat ] ] . w. bietenholz and i. hip , pos lattice * 2008 * , 079 ( 2008 ) [ arxiv:0808.3049 [ hep - lat ] ] .
w. bietenholz and i. hip , j. phys .
* 378 * , 012041 ( 2012 ) [ arxiv:1201.6335 [ hep - lat ] ] .
a. dromard and m. wagner , pos lattice * 2013 * , 339 ( 2014 ) [ arxiv:1309.2483 [ hep - lat ] ] . c. czaban and m. wagner , pos lattice * 2013 * , 465 ( 2013 ) [ arxiv:1310.5258 [ hep - lat ] ] .
i. bautista _ et al .
_ , arxiv:1402.2668 [ hep - lat ] . a. dromard and m. wagner , phys . rev .
d * 90 * , 074505 ( 2014 ) [ arxiv:1404.0247 [ hep - lat ] ] . c. czaban , a. dromard and m. wagner , acta phys .
* 7 * , 551 ( 2014 ) [ arxiv:1404.3597 [ hep - lat ] ] .
a. dromard , c. czaban and m. wagner , pos lattice * 2014 * , 321 ( 2014 ) [ arxiv:1410.4333 [ hep - lat ] ] .
o. philipsen and m. wagner , phys .
d * 89 * , 014509 ( 2014 ) [ arxiv:1305.5957 [ hep - lat ] ] .
d. galletly _ et al .
_ , phys . rev .
d * 75 * , 073015 ( 2007 ) [ hep - lat/0607024 ] .
m. lscher , commun .
phys . * 104 * , 177 ( 1986 ) .
m. j. teper , hep - th/9812187 .
l. del debbio , l. giusti and c. pica , phys .
lett . * 94 * , 032003 ( 2005 ) [ hep - th/0407052 ] .
s. drr , z. fodor , c. hoelbling and t. kurth , jhep * 0704 * , 055 ( 2007 ) [ hep - lat/0612021 ] .
de forcrand , m. garca prez and i. o. stamatescu , nucl .
b * 499 * , 409 ( 1997 ) [ hep - lat/9701012 ] .
|
lattice qcd simulations tend to get stuck in a single topological sector at fine lattice spacing , or when using chirally symmetric quarks . in such cases computed
observables differ from their full qcd counterparts by finite volume corrections , which need to be understood on a quantitative level .
we extend a known relation from the literature between hadron masses at fixed and at unfixed topology by incorporating in addition to topological finite volume effects , also ordinary finite volume effects .
we present numerical results for su(2 ) yang - mills theory .
|
in a quantum computer , it will be necessary to perform gates between distant qubits , whereas the strength of interaction tends to reduce with distance , such that it is impractical to interact them directly .
a typical response is simply to apply a series of swap gates to bring the qubits together so that they can interact .
this , however , is cumbersome and risks introducing significant errors . instead , it has been proposed that an ancillary device be introduced to act as a quantum wire .
this wire would be a chain of qubits , with a fixed interaction , capable of transferring a quantum state from one end of the chain to the other .
following the initial investigation of wires @xcite , and subsequent demonstration that perfect quantum wires exist @xcite , a large number of papers have been published about optimising the schemes over a variety of parameters such as the robustness against errors , or a restricted ability to engineer the state ( see , for example , @xcite ) .
novel modifications of such chains have also been presented for the generation of entangled states or the application of unitary operations during the transfer @xcite . the overhead of local swap gates is thus replaced by an engineering requirement .
such engineering can , however , be tested before the chain is used in a practical situation .
none of these previous works have demonstrated perfect state transfer in a system that has realistic couplings , facilitated by dipole - dipole or coulomb interactions , for example .
once such a coupling is introduced , the fidelity of such schemes is reduced below unity @xcite .
alternatively @xcite , two of these imperfect chains can be used in parallel to perfectly transfer the state . however , the trade - off is that there is no definite time at which the state has arrived .
instead , there is a measurement which will reveal whether the state has arrived or not . in this paper
, we show how to adapt to arbitrary coupling schemes , hence pushing transfer schemes towards physical realisation .
this is achieved by an iterative algorithm founded on the concept of inverse eigenvalue problems ( ieps ) .
the relation between ieps and perfect state transfer has previously been noted in @xcite .
given this relationship , it is also important to understand why we might choose a particular set of eigenvalues , and how this affects the robustness of the chain when it experiences noise or other imperfections .
we explore these issues in the second part of this paper .
in order to make the connection to an iep , we choose to make two assumptions .
firstly , by assuming the hamiltonian is spin preserving , @xmath0=0 $ ] , the problem is reduced to subspaces , and we can concentrate only on the first excitation subspace @xcite . by ensuring a single excitation is correctly transferred , a quantum state is also transferred because the state @xmath1 is an eigenstate of the hamiltonian . in the single excitation subspace ,
the basis states are denoted by @xmath2 , indicating the presence of the excitation on qubit @xmath3 .
secondly , we shall assume the hamiltonian is centrosymmetric ( otherwise known as mirror symmetric ) .
this means that for a chain of @xmath4 qubits , the coupling between qubits @xmath5 and @xmath6 is the same as that between qubits @xmath7 and @xmath8 . throughout this paper ,
we discuss chains of spins because these are the most efficient in terms of the number of qubits used .
further , following @xcite , we are assured that if the hamiltonian can be mapped to a fermionic system ( for example , the xy model can be mapped to a fermionic system by the jordan - wigner transformation ) , then by using a chain we get perfect state transfer in all excitation subspaces .
the assumption about symmetry is useful because it ensures that the eigenvectors of the hamiltonian are always symmetric or antisymmetric @xcite .
it is well - known that a real tridiagonal matrix that exhibits these features has the additional property that when ordered with increasing eigenvalue , the eigenvectors are alternately symmetric or antisymmetric ( see , for example , @xcite ) .
there are no concrete statements of this form for the more generalised problem of non - nearest neighbor couplings .
however , we can invoke the physical restraint that the coupling strengths should drop off with distance , for example @xmath9 . the symmetry of the hamiltonian
is maintained , so the eigenvectors are still symmetric and anti - symmetric . only the eigenvalues are shifted .
however , because the coupling strengths fall off with distance , they act as perturbations in comparison to the tridiagonal case .
provided these perturbations are smaller than the spacing between the original energies , then the hamiltonian will still have the alternating structure of symmetry of the eigenvectors .
this assures us that in this physically reasonable situation , a solution exists to the iep , and that we know enough about the system to be able to generate perfect state transfer .
let us denote the eigenvectors of the hamiltonian by @xmath10 , ordered such that the eigenvalue @xmath11 .
the symmetry condition means that @xmath12 we are particularly interested in the case of @xmath13 , since this relates the initial state , @xmath14 , to the output state , @xmath15 . starting with a single excitation at one end of the chain , we have @xmath16 . after a time @xmath17 , the overlap of the evolved state and the target state , @xmath15 ,
is @xmath18 hence , if we apply the simple constraint on the eigenvalues that @xmath19 for all @xmath3 , we get perfect state transfer in time @xmath20 .
this has previously been observed for tridiagonal structures @xcite , but we emphasise again that this applies to all centrosymmetric chains with coupling strengths that fall off with distance , or any other centrosymmetric network for which we have sufficient information about the ordering of the symmetry of the eigenvectors .
the problem is now reduced to taking a desired eigenvalue spectrum , and a prescribed structure for the hamiltonian , and solving for any free parameters that we might have ( coupling strengths , site spacings , local magnetic fields etc . ) .
some classes of this problem are well - studied topics in the subject of ieps @xcite .
we now present a generalisation of the technique described in @xcite , designed to cope with the arbitrary nature of @xmath21 .
let us assume that we have a hamiltonian @xmath22 which is represented by an @xmath23 matrix in the first excitation subspace .
this hamiltonian depends on @xmath4 parameters @xmath24 , ensuring that there are enough free parameters to be able to find a solution .
our desired eigenvalues are contained in the @xmath23 diagonal matrix @xmath25 . and compare them to our desired values , we need to match up the pairs correctly , such that the differences are minimised . when implementing this , it is readily achieved because eigenvalue - solving algorithms produce the eigenvalues in a well - defined order .
for example , mathematica sorts the eigenvalues in terms of absolute value .
since it is more convenient to have them sorted in numerical order , we can add a large ( in comparison to the maximum eigenvalue ) identity matrix to the hamiltonian , ensuring that all the eigenvalues are positive . ]
we start with a first estimate to @xmath26 , @xmath27 .
the matrix @xmath28 is diagonalised by @xmath29 , @xmath30 where @xmath31 is a diagonal matrix which encapsulates the errors in the energies , and @xmath32 is a small parameter . for our next guess
, we will choose a vector @xmath33 .
again , we can diagonalise the hamiltonian , @xmath34 we choose to parameterise @xmath35 in terms of @xmath32 , @xmath36 where @xmath37 is a hermitian matrix containing information on the change in eigenvectors .
this parameterization ensures that @xmath35 is unitary , and that @xmath38 as @xmath39 .
we can substitute this into eqn .
( [ eqn : iep ] ) , and expand in terms of @xmath32 . the terms for @xmath40 cancel
, so we choose to collect the terms for @xmath41 . @xmath42 the aim of the iteration should be to choose @xmath43 such that @xmath44 is minimised .
note that the diagonal elements of the final term , @xmath45 , are zero .
hence all the eigenvalue information is encapsulated by the diagonal elements of the equations , while changing eigenvectors only affects off - diagonal elements .
therefore , we select @xmath44 to be the zero matrix , and rewrite the previous equation for just the diagonal elements .
@xmath47 is a vector of the diagonal elements of @xmath48 , and the @xmath49 column of the matrix @xmath50 is given by the diagonal elements of @xmath51 the solution to eqn .
( [ eqn : inverse ] ) is the vector @xmath52 , which gives the correct eigenvalues to @xmath53 .
provided @xmath54 is small in comparison to @xmath55 , we can continue to iterate , squaring the error at each step .
hence , to achieve an accuracy of @xmath56 , we only need @xmath57 iterations .
since there are efficient algorithms for solving eqn .
( [ eqn : inverse ] ) , and because the matrices which we want to diagonalise are symmetric ( hence there are efficient diagonalisation procedures , such as householder reductions @xcite ) , the cost of each iteration scales polynomially ( @xmath58 ) with the number of qubits in the chain @xcite .
this means that we can solve for the required parameters with an efficient classical computation . as an example
, we could consider a system of 6 spins , arranged along a line , with separations and local magnetic fields as indicated in fig .
[ fig : example_spins ] .
as we have already seen , exact solution is possible if the couplings do not extend beyond nearest - neighbours .
instead , we shall now consider them to have a dipole interaction , so the potential @xmath59 .
hence the hamiltonian in the first excitation subspace is of the form @xmath60 solving for the eigenvalues in this general form , and then equating these with the desired spectrum is certainly not an easy proposition . as we have seen , however , we can apply the algorithm described here , aiming for a spectrum of @xmath61 , @xmath62 , @xmath63 , @xmath64 , @xmath65 , @xmath66 . the required values which we find
are given in table [ tab : values ] , and are compared to those without the additional couplings .
.table of parameters for perfect state transfer for spin chain of fig .
[ fig : example_spins ] . [ cols="^,^,^ " , ] one implicit assumption that we have made is that the hamiltonian , @xmath22 , is differentiable .
there are some physical systems in which this might not be true . for example
, we may be constrained to having to place spins on lattice sites of another material . choosing which lattice sites to place the spins on is a discretized form of the problem , and is not covered in this formalism .
the best that we can achieve is to allow some additional engineering , such as local magnetic fields , and tune these to give the closest match to a particular spectrum @xcite .
we could envisage a variety of such systems in which we do not have control over a sufficient number of parameters . instead of @xmath4 simultaneous equations for @xmath4 variables , we have @xmath4 equations for @xmath67 variables .
we can solve for these variables in a least - squares sense , minimising the quantity @xmath68 .
hence , while perfect state transfer might not be possible , we can maximise the fidelity of transfer . in fig .
[ fig : large_panel ] we have examined the case of @xmath69 , demonstrating that even without full control of a sufficient number of parameters , we can get higher fidelity than by simply assuming nearest - neighbor couplings .
interaction , without the ability to engineer local magnetic fields .
high fidelity state transfer is possible .
( c ) near - uniform couplings with ability to engineer positions of spins and local magnetic fields , enabling perfect state transfer .
( d ) the original perfect state transfer chain , which has a much wider peak and no periodic revivals at times @xmath70 . ]
we now have an algorithm that takes a desired spectrum as input , and outputs the values of the parameters that we have access to in our hamiltonian .
it is therefore relevant to ask what spectrum we should choose .
unsurprisingly , the preferred spectrum is a trade - off between different properties .
one complaint that has been levelled at the construction of the original quantum wire @xcite is that the coupling strengths at the end of the wire are much smaller than those in the middle ( by a factor of @xmath71 ) .
this issue is relatively easy to correct @xcite by specifying a different spectrum , which closely matches that for the uniformly coupled spin chain .
this selection of eigenvalues is still applicable to the ideas presented here .
for example , we could take a chain of 31 spins , with uniform couplings and no magnetic field .
we can solve for these eigenvalues , and truncate them to some precision ( two decimal places in this example ) . we can then ` nudge ' them slightly to give a spectrum suitable for perfect state transfer .
the results we get require a variation in position of less than @xmath72 about the mean , and require local magnetic fields of the order of @xmath73 .
there are , however , trade - offs in terms of robustness against errors in the distances , and in terms of the time at which the arriving state is removed from the system .
these problems are illustrated in fig .
[ fig : large_panel ] .
here we see that slight errors in the positions ( i.e. the uniformly coupled chain ) give far worse state transfer fidelity .
we also see that the peak is very tightly confined with comparison to the original state transfer chain .
as presented in @xcite , the original state transfer chain has a fixed transfer time , @xmath74
. however , there are two physical restrictions that we might consider applying to the chain which cause us to restate the transfer time .
firstly , we might choose to bound the maximum energy of the chain .
in this case , we have to rescale the spectrum to fit within this bound , making the energy differences smaller . since @xmath75 , when we rescale , we find that @xmath76 .
to minimise this scaling , we therefore want to choose a spectrum whose maximum eigenvalue is @xmath77 , and not @xmath78 , for example . note that the uniformly coupled chain has a spectrum @xmath79 , which has a minimum spacing @xmath80 for large @xmath4 @xcite .
this minimum spacing corresponds to an upper bound to @xmath81 i.e. @xmath82 .
the second approach is the most physically reasonable - we bound the maximum coupling strength in the system .
for the original transfer chain , the maximum coupling strength is at the middle of the chain , @xmath83 .
hence , if we fix this to be constant by rescaling all of the coupling strengths , @xmath76 . however , if we consider the chain with almost - uniform couplings , clearly there is no such variation .
this is a very special case , and indicates that there is no direct correlation between general properties of a spectrum ( such as @xmath84 ) and the coupling strengths .
since we can only choose the spectrum , it is very hard to apply this criterion .
a further concern is the tolerance of the chain to a variety of errors . perhaps the simplest case to consider is a timing error . if the perfect state transfer time is @xmath20 , what is the fidelity of transfer , @xmath85 , at a time @xmath86 ? clearly , it is preferable to have a broad peak , thereby maximising the tolerance to such errors .
the fidelity of transfer at such a time is calculated from @xmath87 ( up to an irrelevant global phase ) by taking the mod - square , @xmath88 given that we have ordered our eigenvalues such that @xmath89 , we can provide a simple lower bound for this quantity , @xmath90 this bound is easily optimised by choosing a spectrum with minimum spread , and it is also clear that this will do a good job ( although not perfect ) of optimising the fidelity . the spectrum that fulfils this condition is just that which is used in the original chain @xcite , with eigenvalues @xmath91 etc .
we refer to this spectrum as the sms ( spectrum of minimal spread ) . one other class of errors that we need to consider are static errors , introduced at the manufacturing stage .
if it is only possible to manufacture coupling strengths to within a certain tolerance , can we select a spectrum that minimises the negative impact on the fidelity of transfer ?
we will consider , independently , the effects when this error propagates to errors in the eigenvalues and eigenvectors of the hamiltonian .
let us first justify that errors in eigenvalues and eigenvectors are equally destructive . if a single eigenvalue is affected by an amount @xmath92 ( which is equivalent to an error in a coupling strength of order @xmath92 , independent of the eigenvalues ) , the fidelity is only affected by @xmath93 - the first order effect only changes the phase of the incoming state , not the arrival probability .
if an eigenvector is altered by errors on the chain , then it can still be written as a sum of a symmetric vector and an antisymmetric vector , by using a discrete form of @xmath94 .
of course , we can not only change one eigenvector , since we need to maintain an orthogonal set of vectors .
we can test what happens when @xmath95 which serves to indicate the general properties of the errors .
in particular , we can calculate that the fidelity of transfer is still only affected to @xmath93 .
hence , errors in the eigenvectors are no more destructive than errors in the eigenvalues .
recall that the condition we imposed upon the eigenvalues was @xmath19 , which means that if there is an error in @xmath96 , we are concerned with how close @xmath97 is to an integer .
hence if @xmath96 is of the form @xmath98 , then @xmath92 , the inaccuracy , must be small in comparison to 1 , not @xmath99 .
when we quote the accuracy to which an eigenvalue must be engineered , we consider it as a fraction of @xmath100 . hence
, this fraction is largest for the spectrum with smallest maximum eigenvalue i.e. the sms . to consider the effect of errors on the eigenvectors , and how to counter them is a more difficult problem because we do not have any direct control - our eigenvectors
are determined by the choice of spectrum and the hamiltonian . however
, we can make some progress .
let us introduce the @xmath23 symmetry operator , @xmath101 , defined by @xmath102 this operator commutes with the centrosymmetric hamiltonian ( this is how we prove the symmetry of the eigenvectors of @xmath21 ) , and hence , should be effective at detecting any symmetry that is broken by the effects of noise , by calculating @xmath103 $ ] .
let us , again , consider the effect described in eqn .
( [ eqn : symm_break ] ) .
@xmath104=2\delta\sqrt{1-\delta^2}(\lambda_1-\lambda_2)\left({\left | \lambda_1 \right\rangle}\!{\left \langle \lambda_2 \right |}-{\left | \lambda_2 \right\rangle}\!{\left \langle \lambda_1 \right |}\right)\ ] ] this tells us that a fault in the coupling strength @xmath105 will translate into an error in the eigenvectors of approximately @xmath106 .
minimising this error is achieved by maximising the energy difference @xmath107 .
this would appear to counter the effect that we just observed for the eigenvalues .
however , we note that in the general case , where we have to consider all of the eigenvectors being affected , the most significant error is related to the smallest energy difference which , in turn , is determined by the state transfer time , which we are holding fixed . hence , the choice of spectrum has limited effect on how the eigenvectors contribute to a reduction in the fidelity .
it is more important to optimise for errors in the eigenvalues .
so far , we have considered only one simple protocol for transferring states with our chain ; we place states on the chain at some time , @xmath108 , and either remove them from the chain , or interact them with another qubit ( leaving them on the chain ) at @xmath109
. however , the rate of transfer suffers at this stage because if we bound the energy of the chain , the state transfer time increases with @xmath4 ( in the case of the sms ) . if we restrict ourselves to only being able to access a fixed number of qubits at each end of the chain , then the transfer rate @xmath110 .
is there something better that we can do , perhaps allowing some fixed reduction in fidelity ?
we restrict ourselves , without loss of generality , to being able to access only a single qubit at each end of the chain .
consider placing an excitation @xmath14 onto a chain with a state transfer hamiltonian , @xmath21 , at a time @xmath108 . at some time @xmath111
, we shall discard the information on qubit 1 , and replace it with a new quantum state .
we need to evaluate the fidelity of arrival of the initial state when the replacement we make is to place the first qubit in the @xmath112 or @xmath14 state .
the first period of evolution gives us @xmath113 where @xmath114 is the amplitude of the state at time @xmath115 on qubit @xmath3 .
if the first qubit then gets indiscriminately reset to the @xmath112 state , the resulting density matrix is @xmath116 after the further evolution for a time @xmath117 , we expect to be able to remove the state from qubit @xmath4 .
the state at this time is @xmath118 given that the chain is centrosymmetric , @xmath119 .
hence , we find that the fidelity of transfer is @xmath120 similar arguments lead to a calculation of the fidelity of transfer when we reset to the @xmath14 state .
this is complicated compared to the previous argument because we now enter the second excitation subspace .
the structure of higher excitation subspaces can be described by the eigenvectors through the slater determinant ( see , for example , @xcite ) .
an alternative approach is to make use of the wedge product , whose application to these systems was described in @xcite . with a probability @xmath121 , we introduce a second excitation into the system , and the state is described by @xmath122 the properties of the wedge product automatically mean that it is impossible to get two excitations on a single spin ( @xmath123 ) , and the evolution of the state is described by applying the single excitation evolution to each component , @xmath124 we conclude that if the reset was made , the state definitely arrives . the other possibility , of having a @xmath14 , occurred with probability @xmath125 and after a time @xmath117 gives an @xmath15 with probability @xmath125 .
hence @xmath126 with this formalism , we can clearly calculate similar transfer fidelities in all excitation subspaces , and we find that the minimum fidelity is @xmath127 , provided @xmath128 is a decreasing function for @xmath70 . by symmetry , the fidelity of arrival the newly added state at a time @xmath129 on qubit @xmath4 is the same , if the first state is removed at time @xmath20 .
these fidelities only depend on @xmath125 , which is the fidelity with which the initial state is still on the first qubit at time @xmath115 . therefore , to maximise the rate of information transfer , we select the chain with the thinnest peak , the opposite of sec .
[ sec : timingerrors ] , but this already assumes that we have control of the timings to within the width of the peak .
[ fig : large_panel ] also serves to illustrate another point that we have to be careful of . for the original state transfer chain ,
@xmath130 is a decreasing function for times @xmath111 , and hence the effect if we also add an extra state at @xmath131 , as well as at @xmath115 , is small , to the point that it becomes insignificant .
however , other spectra can give periodic revivals on the first spin during the transfer .
we have to ensure that we avoid all of these , which potentially reduces the rate of information transfer . for the sms
, this protocol enhances the transfer rate to @xmath132 , for some fixed reduction in fidelity , @xmath32 .
this is an enhancement over @xcite , where it was demonstrated how a uniformly coupled ring , in combination with gaussian wave packets ( to minimise dispersion ) , could be used to give a transfer rate , which was lower bounded by @xmath133 .
one point to note is that the process of removing the state at @xmath109 ( eqn .
( [ eqn : flibble ] ) ) means that only a single excitation is left on the chain .
hence , there are only ever @xmath134 excitations on a chain , and we never saturate the number of excitations on the chain .
in conclusion , we have demonstrated that perfect state transfer is possible in the presence of next - nearest - neighbor couplings by presenting an algorithm that correctly calculates the couplings for any specified system .
if sufficient free parameters are not available , the formalism presented here is easily adapted to find the optimal solution in a least - squares sense .
this technique may generally be useful when , as a first approximation , people have considered the restriction to nearest - neighbor couplings @xcite , and would subsequently like to extend their networks to include more realistic hamiltonians , such as those met in physical implementations .
we have also discussed some of the issues relating to what spectrum should be chosen for the state transfer , demonstrating that the spectrum originally proposed in @xcite is close to optimal is terms of robustness against a range of errors .
|
in this paper we build on the ideas presented in previous works for perfectly transferring a quantum state between opposite ends of a spin chain using a fixed hamiltonian .
while all previous studies have concentrated on nearest - neighbor couplings , we demonstrate how to incorporate additional terms in the hamiltonian by solving an inverse eigenvalue problem .
we also explore issues relating to the choice of the eigenvalue spectrum of the hamiltonian , such as the tolerance to errors and the rate of information transfer .
|
with the advent of grid computing , the classical science of scheduling theory has gained a new sphere of applications .
methods which were originally developed for decision making around limited resources in manufacturing and service industries have been adopted in the areas of computer science , telecommunication and other computationally intensive disciplines such as computational biology , chemistry and finance @xcite .
specifically , in the area of computational finance , where this paper sources its motivation , the problems of portfolio selection , pricing and hedging of complex financial instruments requires an enormous amount of computational resources whose optimal usage is of utmost importance to investment banks .
the prices and risk sensitivity measures of complex portfolios need to be reevaluated daily , for which an overnight batch of calculations is scheduled and performed for millions of financial transactions , utilizing thousands of computing nodes .
each job has a well - defined priority and required completion time for availability of the resulting figures to the trading desk , risk managers and regulators .
the jobs can generally be stopped and resumed at a later point on a different machine which is referred to as _ preemption _ in scheduling theory . for simplicity of modeling the problem ,
machines are generally assumed to be _ identical _ and there is a known , constant number of machines available .
the problem of finding optimal schedules for jobs running on identical machines has been extensively studied over the last three decades .
sahni @xcite presents an @xmath0 algorithm to construct a _ feasible _
schedule , one that meets all deadlines , if one exists , for @xmath1 jobs and @xmath2 machines .
the basic idea of the algorithm is to schedule jobs with earliest due dates first , but fill up machines with smaller jobs if possible .
note that this method allows the development of an algorithm to compute the minimal amount of unit capacity for which a feasible schedule exists .
this result has been extended to machines with identical functionality but different processing speed , termed _ uniform machines _ , and jobs with both starting times and deadlines @xcite .
however , the scheduling task becomes more difficult when a feasible schedule does not exist and the goal is to minimize some measure of delinquency , often termed _
tardiness_. tardiness of an individual job under a given schedule is defined as the amount of time by which the job finishes after its prescribed deadline , and is considered to be zero if the job finishes on or before the deadline . in case of minimizing the maximum tardiness across all jobs
, lawler @xcite shows that the problem is solvable in polynomial time , even with some precedence constraints .
martel @xcite also used his construction to create a polynomial time algorithm to find the schedule which minimizes maximum lateness .
however , if our measure concerns the total tardiness instead of the maximal one , then even the single machine , total tardiness problem ( without weights ) was proven to be np - hard by du et al @xcite . a pseudopolynomial algorithm has been developed by lawler @xcite for this problem , using dynamic programming , but this is for the 1-machine problem and does not have good practical runtime characteristics . in practical applications ,
jobs often have relative priorities associated with them , represented by positive real _ weights _ and the objective becomes minimizing the total weighted tardiness ( twt ) .
once the np - hardness of the twt problem was established , most of the research work on the problem concerned the development of fast , heuristic algorithms .
dogramaci et al .
@xcite propose a simple heuristic for the total ( non - weighted ) tardiness problem without preemption .
rachamadugu et al .
@xcite then studied the identical machine , total weighted tardiness problem without preemption .
they proposed a myopic heuristic and compared this to earliest due date ( edd ) , weighted shortest processing time ( wspt ) and montagne s rule on small problem sizes ( 2 or 5 jobs in total ) .
azizoglu et al .
@xcite worked on an algorithm to find optimal schedule for the unweighted total tardiness problem without preemption , but their branch and bound exponential algorithm is too slow , in practice , for problems with more than 15 jobs .
armentano et al .
@xcite examined the non - weighted problem without preemption , and starting from the kpm heuristic of koulamas @xcite improved upon it , using tabu search .
guinet @xcite applies simulated annealing to solve the problem with uniform and identical machines and a lower bound is presented in order to evaluate the performance of the proposed method .
more recently , sen et al . @xcite surveyed the existing heuristic algorithms for the single - machine total tardiness and total weighted tardiness problems while biskup et al .
@xcite did this for the identical machines total tardiness problem and also proposed a new heuristic .
akyol et al .
@xcite provide an excellent recent review of artificial neural network based approaches to scheduling problems and proposes a coupled gradient network to solve the weighted earliness plus tardiness problem on multiple machines .
the feasibility of the method is illustrated on a single 8-job scheduling problem .
we suggest a novel heuristic for the twt problem , based on the hopfield neural network approach which is shown to perform better than existing simple heuristics and has desirable scaling characteristics .
maheswaran et al .
@xcite applied a similar approach to the single machine twt problem and their results were encouraging for a specific 10-job problem . in this paper , we first map the problem into quadratic optimization and then the hopfield net is used to provide fast polynomial time , heuristic solution .
the topics are organized as follows : in section 2 , we present the problem formulation and the model used , in section 3 existing heuristics are defined and explained , in section 4 our novel hnn approach is introduced , in section 5 numerical results are presented and finally in section 6 some conclusions are drawn and directions for future research are outlined .
in this section , we give a formal presentation for the problem of optimally scheduling jobs on finite number of identical processors under constraints on the completion times .
the basic formalism is the following : * given @xmath3 jobs with sizes @xmath4 .
the processing of the jobs can be stopped and resumed at any time , so the processing time units of each job need not be contiguous . in the literature
this condition is known as preemption and also assumes a task started on one machine can continue on another @xcite .
* for each job a cutoff time is prescribed by @xmath5 .
this constraint defines the time within which the job is to be completed . * the constant number of processors , the capacity of the system
is denoted by @xmath6 .
* we are also given a vector @xmath7 denoting the relative priority ( or weight ) of each job , which can be used in the definition of the objective function .
a schedule is represented by a binary matrix @xmath8 where @xmath9 if job @xmath10 is being processed at time slot @xmath11 , and @xmath12 denotes the length of the schedule .
an example is given in ( [ eq : e21 ] ) where the parameters are the following : @xmath13 , @xmath14 , @xmath15 , @xmath16 .
@xmath17 the first row in ( [ eq : e21 ] ) denotes the fact that under this schedule @xmath18 , the first job is processed in time steps @xmath19 and @xmath20 ( note that preemption is used as the processing of this job is not continuous ) and therefore the prescribed size @xmath21 of this job completes within the prescribed cutoff time of time step @xmath20 .
similarly , the @xmath20 units of the second job complete within the cutoff time of @xmath20 and the third job is completed ahead of the cutoff time on time step @xmath21 .
summing the columns of matrix @xmath18 , we see that the maximal capacity of @xmath22 is fully utilized on each time step . in order to evaluate the effectiveness of a given schedule @xmath18
, we define tardiness of a job as follows : @xmath23 where @xmath24 is the actual finish time of job @xmath10 under schedule @xmath18 : @xmath25 ( the position of the last @xmath19 in the @xmath10th row in scheduling matrix @xmath18 . ) the problem can now be stated formally as follows : @xmath26 under the following constraints : * the sizes of the scheduled jobs in the scheduling matrix are equal to the predefined amounts : @xmath27 * the number of scheduled jobs at any given time instant does not exceed the capacity of the system : @xmath28 we revisit our previous example with a minor change : @xmath13 , @xmath14 , @xmath29 , @xmath16 and a weight vector @xmath30 .
it can be observed that there is no solution , in which all jobs are completed before their cutoff times .
a minimal weighted tardiness solution is the following : @xmath31
given the np - hardness of the scheduling task as shown by du et al .
@xcite , and therefore the amount of time it would take to find the exact , optimal solution , in most real - world settings the pragmatic approach of finding a fast , sub - optimal , but good solution is followed .
as outlined in the introduction , this has been a very active field of research , and most studies use the earliest due date ( edd ) and weighted shortest process time ( wspt ) heuristics as benchmarks for evaluating the proposed algorithms .
in addition to these , we also outline the recently developed load balancing scheduling ( lbs ) heuristic and a newly proposed largest weighted process first ( lwpf ) heuristic .
furthermore , we also outline a random processing method which is used as a low benchmark for our testing results . the edd heuristic orders the sequence of jobs to be executed from the job with the earliest due date to the job with the latest due date .
using the notation of section 2 , we relabel the job indices so that the following inequality holds : @xmath32 once this ordering is determined , the jobs are allocated to the machines in this order , always utilizing the maximum available capacity .
once a job finishes and capacity is freed up , the next job using the above ordering is scheduled on the freed up machine .
it has been shown in @xcite that edd finds the optimal schedule when one wants to minimize the maximum tardiness on a single machine .
however , we note that when the objective function includes the relative priorities of jobs , edd is at a severe disadvantage , as it does not include consideration of the weights .
the wspt method is analogous to the edd method in that it orders the jobs according to well - defined criteria and then schedules the jobs according to this ordering , utilizing the maximum available capacity available .
once a job finishes and capacity is freed up , the next job using the ordering is scheduled . using the notation of section 2 ,
the jobs are ordered such that the below holds : @xmath33 this heuristic is analogous to the edd and the wspt heuristics with the jobs ordered according to the following inequality : @xmath34 this is a simple heuristic which allows us get a sense of the importance of considering the weights versus the cutoff times . we have not seen this simple heuristic mentioned anywhere in the literature , but it turns out to have quite good empirical characteristics which is one of the many contributions of this paper .
laszlo et al @xcite suggest a novel deterministic , polynomial time algorithm for scheduling jobs with cutoff times , in the context of load balancing for wireless sensor networks .
the basic idea of the algorithm is to start scheduling the jobs backwards from the maximal cutoff time , in order of decreasing cutoff times , always ensuring full capacity is utilized . for a more formal definition of the algorithm
, please see @xcite .
we note that whilst the algorithm has proven to be extremely successful for load balancing , it does not take into account the weights of the jobs and therefore will suffer from the same shortcomings in our setting as edd . in this method ,
the jobs are ordered randomly and are scheduled in such a way as to always fully utilize the maximum available capacity .
this unsophisticated heuristic is useful as a low benchmark by repeating it a number of times and taking the best solution over the multiple repeats .
since the number of possible binary matrices grows exponentially with the number of nodes and the length of the schedule , exhaustive search with complexity @xmath35 is generally computationally infeasible to solve the optimization problem at hand .
thus , our goal is to develop a polynomial time approximate solution by mapping the problem to an analogous quadratic optimization problem , similar to the approach of levendovszky et al @xcite and treplan et al @xcite .
we first review the methods available for optimizing equations in quadratic form .
let us assume that matrix @xmath36 is a symmetric matrix of size @xmath37 and vector @xmath38 is of length @xmath1 .
we seek the optimal @xmath1 dimensional vector @xmath39 which minimizes the following quadratic function @xcite : @xmath40 subject to one or more constraints of the form of @xmath41 if the problem at hand contains only linear constraints then it can be solved as presented by murty et al @xcite . in other cases ,
if the matrix @xmath36 is positive definite , then the function @xmath42 is convex and the problem can be solved with the ellipsoid method presented by zhi - quan et al @xcite .
when @xmath43 is indefinite , the problem is np - hard ( for details see @xcite ) . a frequently used powerful , heuristic algorithm to solve quadratic optimization problems is the hopfield neural network ( hnn ) .
this neural network is described by the following state transition rule : @xmath44 where @xmath45 using the lyapunov method , hopfield @xcite proved that the hnn converges to its fixed - point , as a consequence the hnn minimizes a quadratic lyapunov function : @xmath46 thus , the hnn is able to solve combinatorial optimization problems in polynomial time under special conditions as it has been demonstrated by mandziuk @xcite . using this method in practice we have to handle the following problem : in some cases the hnn method may get stuck in local minimal point of its lyapunov function .
these methods motivate us to map the existing optimization problem into quadratic form .
first the binary scheduling matrix @xmath47 is mapped into a binary column vector @xmath48 as follows : @xmath49 @xmath50 the original objective function ( [ eq : e23 ] ) is elaborated as follows : @xmath51 this objective is equivalent to : @xmath52 the minimization problem is thus equivalent to : @xmath53 therefore the mapping required is : @xmath54 the solution is the following : @xmath55 @xmath56 where @xmath57 having transformed the objective function to a quadratic form , we now turn our attention to doing the same with the two constraints outlined in ( [ eq : e24 ] ) and ( [ eq : e25 ] ) . the constraints in ( [ eq : e24 ] ) can be rewritten as follows : @xmath58 this will lead to the following equation : @xmath59 the solution of equation ( [ eq : e416 ] ) is : @xmath60 @xmath61 another constraint ensures that the scheduling does not overload the processing units , described in ( [ eq : e25 ] ) .
the required transformation is as follows : @xmath62 please note the technicality that this constraint may not be relevant for the last few columns where only the remaining jobs have to be scheduled and schedule matrices not exhausting the full capacity should not be penalized .
the total length of scheduling can be written as follows : @xmath63 @xmath64 the number of columns where capacity @xmath65 needs to be fully utilized is the following : @xmath66 taking into consideration ( [ eq : e421 ] ) the mapping of ( [ eq : e419 ] ) can be described by the following equation : @xmath67 the solution of this equation is the following : @xmath68 , \ ] ] @xmath69 where @xmath70 we can combine these mappings into the form of equation ( [ eq : e41 ] ) as follows : @xmath71 and @xmath72 note that the objectives can be controlled with heuristic constants @xmath73 and @xmath74 in order to strike an appropriate balance between the weights of different requirements . having this quadratic form at hand , we can apply the hnn to provide an approximate solution to the optimization problem in polynomial time .
in this section , the performance of the hnn approach is investigated and is compared to the performance of other heuristics . each method outlined in section 3 and
the hnn method outlined in section 4 has been tested by simulation on a large and diverse set of input parameters with the aim to characterize the algorithms empirically on scheduling problems of different size of jobs .
the algorithms were implemented in matlab and tests were run in this simulation environment with randomly generated parameters such as size of jobs , cutoff times , and weights .
the size of each job and its cutoff time and corresponding weight are generated as follows : @xmath75 \right ) , \label{eq : e511}\ ] ] @xmath76 \right ) , \label{eq : e512}\ ] ] @xmath77 \right ) , \label{eq : e513}\ ] ] where @xmath78 produces a uniformly distributed random integer value in range @xmath79 . in our simulations
the constant @xmath80 equals @xmath81 and @xmath82 equals @xmath83 .
the number of processors ( @xmath65 ) is determined as follows : @xmath84 the problem is expected to be solved without tardiness when the capacity is @xmath85 . therefore ( [ eq : e514 ] ) ensures that there is a high likelihood of tardiness associated with the generated problem . for each problem size
, @xmath86 different problems were generated randomly , using ( [ eq : e511])-([eq : e514 ] ) and the results the methods were compared in each case .
the random method was repeated @xmath87 times for each problem and the best solution was used .
furthermore , the hnn method was repeated @xmath87 times for each problem with different random starting points and the best solution was used .
in addition , the heuristical parameters @xmath73 and @xmath74 were adjusted between the simulations in order to provide a good balance between optimizing the objective function , but also meeting the required constraints to produce a valid scheduling matrix . to adjust the heuristical parameters we used algorithm [ alg : a1 ] .
@xmath88 @xmath89 @xmath90 @xmath91 @xmath92 @xmath93 @xmath94 @xmath95 @xmath96 @xmath97 remove @xmath19 from row @xmath11 , where @xmath11 is the row in column @xmath98 with minimal weight that has @xmath99 remove @xmath19 from row @xmath98 from the column @xmath100 , where @xmath100 is the righternmost column in row @xmath98 such that @xmath101 add @xmath19 to row @xmath98 in column @xmath100 where @xmath100 is the lefternmost column where @xmath19 can be added without violating the capacity constraint @xmath65 . @xmath47
even so , in some cases the hnn is unable to provide a valid solution , because the prescribed constraints on job sizes and capacity are violated .
therefore , we introduced an error correction function ( see algorithm [ alg : a2 ] ) in order to cut and replace the unnecessary 1-s in the scheduling matrix . in our simulations
parameter @xmath102 equals @xmath83 .
all of the results presented in the following sections concern valid schedules meeting the required constraints .
the first simulation compares the average total weighted tardiness provided by the algorithms for different problem sizes .
figure [ fig:1 ] demonstrates that the best solution achieved by the hnn produces better average twt for all problem sizes than any of the other heuristics .
the performances of the edd , lbs and random methods are , on average , worse than the wspt and lwpf heuristics and the hnn for all problem sizes .
this shows that consideration of the weights in the optimization problem is critical to reach near - optimal twt schedules .
it is also clearly visible that the hnn consistently outperforms all other methods for all problem sizes .
note also that the simple lwpf heuristic , proposed by us , outperforms the wspt heuristic widely used as a benchmark in the literature . in this section ,
we quantify how much better the solution provided by the hnn is , in comparison to the edd , wspt and lwpf algorithms .
figure [ fig:2 ] depicts the ratio of average twt produced by the hnn method versus the other heuristics .
we conclude that , on average , the best solution of the hnn heavily outperforms the traditional solutions : the total weighted tardiness of the best hnn solution is @xmath103 of the edd , @xmath104 of the wspt , and @xmath105 of the lwpf .
as the problem size increases , the wspt and lwpf heuristics produce solutions which are closer to the hnn . in order to verify that the hnn consistently outperforms the other heuristics on a wide spectrum of problems , not just on a few selected problems for each given problem size , we ran a more detailed experiment . in this case
, we investigated @xmath106 randomly generated problems for each problem size and computed the percentage of times the hnn produced a better solution than lwpf , the next best method .
table [ tab:1 ] shows the result which quite convincingly proves the general applicability of the hnn to different problem types . across the spectrum of problem domain ,
the ratio is higher than @xmath107 for all investigated job sizes .
.percentage of problems in which the hnn provides an improved solution over the next best heuristic , lwpf [ cols="<,^,^,^,^,^,^,^,^,^ " , ] note that this table shows the theoretical runtime for a single run of the specified algorithm , whilst we need to run a constant number of the hnn and random iterations . on the other hand
, these independent runs can be executed in parallel on several computers , multicore machines or even gpu - based architectures , so the overall runtime should not be significantly different due to the need to run multiple iterations .
figure [ fig:3 ] depicts the empirical runtime comparison of the different heuristics which confirm the theoretical limits .
although the hnn is the slowest method of the investigated heuristics , in practice we see that near optimal solution of a @xmath86 job problem takes only around 6 seconds / iteration ( including the iterative adjustment of the heuristic parameters as per algorithm [ alg : a1 ] ) on core i7@3ghz processor , running non - optimized matlab code , which is very promising as far as its practical applicability in real life applications is concerned . with further optimization and parallelization on multicore machines or gpgpu technology , the hnn method provides better quality solutions than the other heuristics within acceptable runtimes .
in this paper , we studied the np - hard problem of scheduling jobs with given relative priorities ( weights ) and deadlines on identical machines , minimizing the twt measure .
we developed a novel heuristic approach , utilizing quadratic programming and the hopfield neural network and we showed that , in general , it outperforms the edd and wspt methods .
furthermore , we have shown that our approach can be applied in real - time settings for small problem sizes and possesses scalability features which are acceptable for real world applications .
more formal methods for the selection of alpha , beta and gamma parameters could be investigated in the future to further improve performance .
another idea to explore is whether a more intelligent selection of the initial point for the hnn algorithm ( eg . the result of the lwpf or wspt heuristics ) would improve the algorithms performance over the random starting point which was used in our tests .
finally , the performance of the hnn algorithm needs to be compared to more complex heuristic methods .
d. e. http://www.deu.edu.tr/deuweb/english/icerik/icerik.php?kod=6933[akyol ] , g. m. http://kisi.deu.edu.tr/mirac.bayhan/[bayhan ] , multi - machine earliness and http://www.springerlink.com/content/j42666764q57j768/fulltext.pdf[tardiness scheduling problem ] : an interconnected neural network approach .
j. of http://www.ingentaconnect.com/content/klu/170;jsessionid=e2q15a1aanbi.alice[adv .
manuf . technol . ]
_ , * 37 * ( 2008 ) 576588
. v. a. http://www.densis.fee.unicamp.br/~vinicius/[armentano ] , d. s. yamashita , http://www.springerlink.com/content/p130j57823581544/fulltext.pdf[taboo search ] for scheduling on identical parallel machines to minimize mean tardiness .
_ j. of http://www.springerlink.com/content/0956-5515/[intell .
_ , * 11 * ( 2000 ) 453460 .
m. http://www.ie.metu.edu.tr/~meral/[azizoglu ] , o. http://www.eng.metu.edu.tr/pages/roster/ie/okirca.html[kirca ] , http://pdn.sciencedirect.com/science?_ob=miamiimageurl&_cid=271692&_user=303407&_pii=s0925527398000346&_check=y&_origin=article&_zone=toolbar&_coverdate=01-jul-1998&view=c&origincontentfamily=serial&wchp=dglzvlt-zskwb&md5=02b6599b4fce34cb16af86e5fab52fa6/1-s2.0-s0925527398000346-main.pdf[tardiness minimization ] on parallel machines .
j. of http://www.journals.elsevier.com/international-journal-of-production-economics/[production econ . ]
_ , * 55 * ( 1998 ) 163168 .
d. biskup , j. herrman , j. n. d. http://cba.uah.edu/is/gupta[gupta ] , scheduling http://pdn.sciencedirect.com/science?_ob=miamiimageurl&_cid=271692&_user=303407&_pii=s0925527308001515&_check=y&_origin=article&_zone=toolbar&_coverdate=30-sep-2008&view=c&origincontentfamily=serial&wchp=dglbvlv-zskwz&md5=7b8fbf6e6ed22800adf81b749691e1d0/1-s2.0-s0925527308001515-main.pdf[identical parallel machines ] to minimize total tardiness .
j. of link : www.elsevier.com / locate / ijpe[production econ . ]
_ , * 115 * ( 2008 ) 134142
. a. http://www.provost.bilkent.edu.tr/rector/dogramac.htm[dogramaci ] , j. surkis , evaluation of a heuristic for http://www.jstor.org/stable/pdfplus/2630795.pdf?accepttc=true[scheduling independent jobs ] on parallel identical processors .
sci . _ ] , * 25 * , 12 ( 1979 ) 12081216 .
j. du , j. y. t. http://web.njit.edu/~leung/[leung ] , http://www.jstor.org/stable/pdfplus/3689992.pdf[minimizing total tardiness ] on one machine is np - hard .
res . _ ] , * 15 * , 3 ( 1990 ) 483495 .
a. guinet , scheduling independent jobs on uniform parallel machines to minimize tardiness criteria .
_ j. of http://www.springer.com/business+%26+management/production/journal/10845[intell .
manuf . ] _ , * 6 * ( 1995 ) 95103 . j. j. http://www.cirs.net/researchers/researchers.php?id=567[hopfield ] , neural networks and physical systems with http://bi.snu.ac.kr/courses/g-ai09-2/hopfield82.pdf[emergent collective computational ] abilities . in : _
* 79 * ( 1982 ) 25542558 . c. p. http://www.isi.gr/en/profile/koulamas_christos.html[koulamas ] , the http://www.jstor.org/stable/pdfplus/171984.pdf[total tardiness problem ] : review and extensions .
res . _ ] , * 42 * ( 1994 ) 10251041 .
e. laszlo , k. http://users.itk.ppke.hu/~kami[tornai ] , g. treplan , j. http://neural.hit.bme.hu/levendov/[levendovszky ] , novel http://www.thinkmind.org/index.php?view=article&articleid=ctrq_2011_3_20_10067[load balancing scheduling ] algorithms for wireless sensor networks . in : _
the fourth int . conf . on communication theory , reliability , and quality of service _ ,
budapest , 2011 , pp .
5449 .
e. l. lawler , preemptive scheduling of precedence - constrained jobs on parallel machines , deterministic and stochastic scheduling .
nato advanced study and research institute on theoretical approaches to scheduling problems _ , 1981 , pp .
101123 .
j. http://neural.hit.bme.hu/levendov/[levendovszky ] , a. olah , k. http://users.itk.ppke.hu/~kami[tornai ] , g. treplan , novel load balancing algorithms ensuring uniform packet loss probabilities for wsn , _
ieee 73rd vehicular technology conf . ,
, 2011 .
r. maheswaran , s. g. ponnambalam , s. d. nithin , a. s. ramkumar , hopfield neural network approach for http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=1460699[single machine scheduling problem ] . in : _ proc .
2004 ieee conf . on http://ieeexplore.ieee.org/xpl/mostrecentissue.jsp?punumber=9872[cybernetics and intelligent systems ] _ , 2004 , pp .
850854 .
j. http://www.mini.pw.edu.pl/~mandziuk/[mandziuk ] , solving the http://www.mini.pw.edu.pl/~mandziuk/prace/tsp_dm.pdf[travelling salesman problem ] with a hopfield type neural network .
http://demmath.mini.pw.edu.pl/index.php[_demonstratio math .
_ ] , * 29 * , 1 ( 1996 ) 219231
. j. http://www.mini.pw.edu.pl/~mandziuk/[mandziuk ] , b. http://www.mini.pw.edu.pl/~macukow/eng/glowna_eng.html[macukow ] , a neural network designed to solve the http://www.mini.pw.edu.pl/~mandziuk/prace/bc92.pdf[n-queens problem ] .
_ ] , * 66 * , 4 ( 1992 ) 375379 . c. http://www.cs.ucdavis.edu/people/faculty/martel.html[martel ] , http://delivery.acm.org/10.1145/330000/322337/p812-martel.pdf?ip=152.66.248.85&acc=active%20service&cfid=68630571&cftoken=15114997&__acm__=1330646036_26f07c187a5ca39c42f2a5dec722e65d [ preemptive scheduling ] with release times , deadlines , and due times _ j. of the http://jacm.acm.org/[acm]_ , * 29 * , 3 ( 1982 ) 812829 .
k. g. http://www-personal.umich.edu/~murty/[murty ] , linear complementarity , http://ioe.engin.umich.edu/people/fac/books/murty/linear_complementarity_webbook/[_linear and nonlinear programming _ ] .
springer - verlag ] , 1988 .
r. m. v. rachamadugu , t. e. morton , _ myopic heuristics for the http://www.ri.cmu.edu/pub_files/pub3/rachamadugu_ram_mohan_v_1983_1/rachamadugu_ram_mohan_v_1983_1.pdf[weighted tardiness problem ] on identical parallel machines_. tech .
rep . , ( cmu - ri - tr-83 - 17 ) carnegie - melon university , the robotics institute , 1983
. s. http://www.cise.ufl.edu/~sahni/[sahni ] , http://www.cise.ufl.edu/~sahni/papers/computationallyrelatedproblems.pdf[computationally related problems ] .
j. on comp .
_ ] * 3 * ( 1974 ) 262279 .
s. http://www.cise.ufl.edu/~sahni/[sahni ] , http://www.cise.ufl.edu/~sahni/papers/preduedates.pdf[preemptive scheduling ] with due dates .
res . _ ] , * 27 * , 5 ( 1979 ) 925934 .
t. sen , j. m. http://70.84.139.130/~sbencat/faculty/index.php?ncat=profile&id=207&depid=2[sulek ] , p. http://www.utc.edu/academic/business/profile_dileepan.php[dileepan ] , http://pdn.sciencedirect.com/science?_ob=miamiimageurl&_cid=271692&_user=303407&_pii=s0925527302002657&_check=y&_origin=article&_zone=toolbar&_coverdate=25-jan-2003&view=c&origincontentfamily=serial&wchp=dglbvlk-zskwa&md5=db4f9a0973601183eb1183d05487bf66/1-s2.0-s0925527302002657-main.pdf[static scheduling research ] to minimize weighted and unweighted tardiness : a state - of - the - art survey .
j. of http://www.journals.elsevier.com/international-journal-of-production-economics/[production econ . ]
_ , * 83 * ( 2003 ) 112
. g. treplan , k. http://users.itk.ppke.hu/~kami[tornai ] , j. http://neural.hit.bme.hu/levendov/[levendovszky ] , http://www.hindawi.com/journals/ijdsn/2011/107062/cta/[quadratic programming ] for tdma scheduling in wireless sensor networks , _ int . j. of distrib . sen .
netws . _ , 2011 .
17 p. l. http://www.ece.umn.edu/~luozq/[zhi-quan ] , m. wing - kin , m. c. s. anthony , y. yinyu , z. shuzhong , semidefinite relaxation of http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=5447068[quadratic optimization ] problems .
http://ieeexplore.ieee.org/xpl/recentissue.jsp?punumber=79[_signal process . mag .
_ ] , ieee * 27 * , 3 ( 2010 ) 2034 .
|
this paper explores fast , polynomial time heuristic approximate solutions to the np - hard problem of scheduling jobs on n identical machines .
the jobs are independent and are allowed to be stopped and restarted on another machine at a later time .
they have well - defined deadlines , and relative priorities quantified by non - negative real weights .
the objective is to find schedules which minimize the total weighted tardiness ( twt ) of all jobs .
we show how this problem can be mapped into quadratic form and present a polynomial time heuristic solution based on the hopfield neural network ( hnn ) approach .
it is demonstrated , through the results of extensive numerical simulations , that this solution outperforms other popular heuristic methods .
the proposed heuristic is both theoretically and empirically shown to be scalable to large problem sizes ( over 100 jobs to be scheduled ) , which makes it applicable to grid computing scheduling , arising in fields such as computational biology , chemistry and finance .
|
single - photon cooling ( spc ) is a general cooling method applicable to most of the periodic table as well as molecules@xcite .
based on one - way wall of light@xcite , it relies on irreversible optical pumping with one photon scattering to trap atoms inside an optical dipole trap .
the irreversibility is achieved by exciting , with a depopulation beam , magnetically trapped atoms to an intermediate hyperfine level from which they decay into an optically trappable state with a finite probability .
this process accumulates atoms inside an optical dipole trap without the need for a cycling transition .
the key mechanism behind spc is that only those atoms with kinetic energy less than a threshold energy are captured by the optical trap i.e. cooling is achieved by `` filtering '' and then isolating colder atoms inside an optical trap .
the optical dipole trap is constructed in the form of a box in spc experiments and hence is simply referred to as the `` box''@xcite .
the nature of spc implies that a high cooling efficiency can be achieved if as many atoms as possible can be made to encounter the one - way wall of the box with the right kinetic energy .
however , various practical constraints exist to limit the cooling efficiency .
for instance , it is not productive to move the box around and stir the atomic cloud with a depopulation beam as this can result in significant loss of atoms there is a high probability that the atoms encountering the box do not have the right energy to be trapped .
the box is therefore assumed to be stationary throughout the evolution .
the finite size of the box means many of the atoms simply will not pass through the position where the box is placed .
the usual isotropic trap is not necessarily the best potential for spc , since the atomic trajectories within an isotropic trap are evenly spread rather than localized near a region where the box can be placed for efficient collection of the atoms .
the energy consideration means that even if the trajectories of the atoms are densely concentrated near one region it is of no use if the atoms are moving too energetically to be trapped . in order to find the best conditions for spc
, we consider in this paper a special system in which various types of atomic trajectories can be demonstrated : the wedge billiard for atoms .
wedge billiard for atoms is a realization of the symmetric gravitational wedge originally introduced by letihet and miller@xcite which , owing to the singularity of the vertex , demonstrates amazingly rich physics
the atomic trajectories can be tuned from stability to chaos with the change in wedge angle .
the wedge billiard for atoms has already been implemented experimentally@xcite where the atomic motion in different regimes of classical chaos were observed and compared with numerical simulations . in this paper
, we consider spc in wedge billiard and , more specifically , look for the best position to place the box for all possible wedge angles .
the best box position is defined as the position at which the most number of atoms is captured .
the overall goal of a cooling scheme is to increase the phase space density .
phase space density in the context of cooling is defined as the number of atoms in a box with sides of one thermal de broglie wavelength@xcite .
this can be written as @xmath0 where @xmath1 is the number of atoms , @xmath2 the volume , and @xmath3 is the thermal de broglie wavelength @xmath4 where @xmath5 is the planck constant .
this implies that the factor by which the phase space density changes is given by @xmath6 where @xmath7 and @xmath8 are the areas of the 2-dimensional wedge and the box , and @xmath9 and @xmath10 denote respectively the initial and final temperature .
the fraction of atoms captured by the box , @xmath11 is therefore an important measure , although it has less impact than the ratio of the final to initial temperature which is raised to the power of @xmath12 .
we shall define in this paper the logarithm of eq .
( [ ratio ] ) as the cooling efficiency @xmath13 we are most interested in how the atoms can be made to attain the right energy and how the cooling efficiency is affected by the regular and chaotic dynamics ; from this , one may gain understanding of how best to implement the spc in general .
the paper is organized as follows : in section ii we discuss how the numerical simulation is carried out , including the parameters used and the three types of boxes we use in our investigation . in particular , we discuss how the results may be generalized , and identify the core parameter space for this system , namely the wedge angle , the initial temperature of the atoms and the box threshold energy . in section iii , we present the numerical results for the case that corresponds to the actual experimental parameters , and lay the foundation for the subsequent section by showing how various quantities such as the fraction of trapped atoms change for different box types and wedge angles . in section
iv we present the result of simulations that cover various combination of the three core parameters to establish the general trend regarding the cooling efficiency of spc in a wedge billiard for atoms .
we conclude in section v.
first , we assume that atomic wedge billiard contains thermal atoms .
the initial positions and momenta of the atoms are assigned according to a gaussian random distribution as is usually done for thermal atoms in equilibrium . in particular , the initial momenta of the atoms are determined from the maxwell - boltzmann distribution @xmath14 $ ] such that the variance of velocity @xmath15 is proportional to the the initial temperature of the atomic cloud , @xmath16 . here
@xmath17 is the boltzmann constant and @xmath18 is the mass of the atom . assuming hard walls for the wedge billiard such that atoms undergo elastic collisions with the walls and taking the scattering cross section for the interatomic collision to be zero , we calculate the expected trajectory of each atom within the wedge billiard using classical kinematic equations as done in ref .
the calculated trajectory for each of the atoms is stored and used later for the analysis involving the box .
the number of atoms in the sample , @xmath1 , is typically be of the order @xmath19 experimentally but in order to simulate a realistic system using a computationally manageable number of atoms we take @xmath1 to be a few hundred and average the final result over several runs with different initial position and velocity ( but with the same velocity variance i.e. the same initial temperature ) . with 200 atoms in each run , averaging over 50 runs gave results similar to a single run with 10000 atoms .
for each run , the @xmath1 trajectories corresponding to that particular set of initial conditions were used to calculate the fraction of atoms trapped and lost by each type of the box discussed below .
these fractions are later averaged over the number of runs to give a better estimate for a realistic atomic sample .
the averaging has the effect of smoothing out any large fluctuations in the result ; in fact , since we have a conservative , closed system , the _ fraction _ of atoms captured by the box was found to be fairly consistent for all @xmath1 .
it was also found from our simulations that a significant proportion of atoms bounce off the walls and escape the wedge billiard , especially in the chaotic regime ; such atoms obviously can not be trapped by the box and are considered lost from the system . in an attempt to avoid such loss and to perhaps improve the performance , we have tried modulating the wedge angle during the time evolution but found that it did nt lead to any better result .
such a set up is also likely to be difficult to implement experimentally .
we therefore consider in our simulations wedge billiard with the walls fixed at one wedge angle at a time throughout the entire duration of the experiment .
numerically , the box is simply modeled as a region in space with a certain implicit threshold energy @xmath20 .
the kinetic energy of individual atoms varies greatly throughout the evolution , with the majority of the atoms taking on kinetic energies far exceeding @xmath20 . for a typical @xmath21th atom with initial position @xmath22 , initial velocity @xmath23 , and initial total energy @xmath24 where @xmath25 denotes gravitational acceleration ,
there are times @xmath26 , @xmath27 during its evolution where the atom reaches a certain height @xmath28 that overlaps with the position of the box such that @xmath29 . at these times
, the atom may be captured by the box .
however not all atoms are captured in reality : based on ref .
@xcite , the efficiency of the box is typically @xmath30 i.e. only @xmath31 of the atoms that pass through the box `` sees '' the box ( converted to the right hyperfine state ) .
this means @xmath31 of the atoms that pass through the box are either lost ( too high energy ) or trapped ( correct energy ) and @xmath32 just passes through .
we include this constraint in our simulations .
it is noted that since the atoms that pass through the box with kinetic energy between zero and @xmath20 are assumed confined by the box , the final equilibrium temperature of the atoms inside the box can be estimated to be of the order @xmath33 .
finding the right position to place a box given that there are @xmath1 atoms with different trajectories is a particularly difficult optimization problem , and is the main goal of this paper . to solve the problem of where to place our box 3 types of boxes were considered in our simulation : the optimum box and two types of fixed box we refer to as type i and type ii box .
we explain these in more detail below . since we are considering a deterministic system , it is , in principle , possible to calculate precisely where and when each atom reaches the required kinetic energy for it to be trapped .
the optimum box is a hypothetical box of vanishing size that , by methodically moving around the trap , captures all theoretically trappable atoms in sequence .
this then gives the upper limit to the number of atoms that can be captured for a given set of parameters . to provide a more useful theoretical upper limit to the number of trappable atoms
, we additionally impose a couple of practical constraints .
one of these constraints is that the optimum box is assumed not able to be at more than one position at one given instant if there are more than one atom attaining trappable energy at the same time only one of them is considered captured and the rest are let evolving .
the optimum box subsequently catches the next available atom that reaches the correct energy , and so on .
the other constraint is , in line with the real experiment , we assume the efficiency of the optimum box to be @xmath31 .
i.e. only @xmath31 of the atoms passing through the optimum box are captured ( _ all _ @xmath31 are captured since , by definition , all atoms intercept the optimum box with the correct kinetic energy ) . in reality , the box is finite in size and can not be moved around freely .
we consider two possible cases in relation to where to place the fixed box .
one is to use the result of the optimum box calculation above to guide us where to place our box .
based on the optimum box calculation which gives a sequence of box positions over the duration of the experiment , we choose to place a real , finite - sized , stationary box at the position where the largest number of the calculated trappable positions fall within the width of box , in both @xmath34 and @xmath28 directions .
we call this stationary box type i box . on the other hand , in the absence of an optimum box calculation , the best one can do is to actually place the finite fixed box at various different places inside the wedge billiard to find , by trial - and - error , the position where the most atoms can be trapped . given the symmetry of the wedge billiard system around @xmath35 , after collecting results corresponding to all possible heights of the box on the axis of symmetry , we found the one height @xmath28 that traps the most atoms for a given wedge angle .
we refer to this box obtained from `` optimization by hand '' as type ii box .
using realistic parameters is obviously crucial for the correct modeling of the system being simulated . with too big or too small dimensions , the atomic trajectories and
hence the calculated cooling efficiency is likely to be unrealistic . to address this issue we base our parameters on the experimental values@xcite and choose our parameters to be within reasonable range of these values .
in addition to simulating a realistic system , one should ideally be able to generalize the result beyond the existing experiment .
however it is noted that , especially in the chaotic regime , it is impossible to write down analytically a general expression of the atomic trajectories as a function of various parameters .
this limits our options to only those of numerical analysis .
since one can not cover every single possible value of various parameters , only a general trend can be identified from a numerical study .
such general trend should , however , provide sufficient information to understand the fundamental physics of the system , and give us a clue as to the range of realistic cooling efficiencies possible with this system .
the interdependence of various parameters means the parameter space boils down to three independent variables : the wedge angle , initial temperature , and the box threshold energy .
it is noted first of all that the regular and chaotic behavior exhibited by the atoms inside the wedge billiard is dependent only on the wedge angle , and is independent of factors such as the atomic species , initial temperature , and the qualities of the box .
we therefore capture all the necessary physics in our simulation by scanning through all possible wedge angles .
also , in our numerical model , different atomic species is represented only via their different atomic mass which shows up in the velocity variance @xmath36 and in the calculation of the kinetic energy @xmath37 .
it is noted that any difference in physics due to different atomic mass is taken care of with an appropriate change in the initial temperature @xmath9 ( @xmath38 ) , and the kinetic energy @xmath39 is automatically scaled accordingly .
this can be shown by setting @xmath40 so that the initial thermal energy @xmath41 becomes unit energy and @xmath42 becomes unit velocity such that the unit length @xmath43 can be defined as @xmath44 where @xmath45 is unit time .
the kinetic energy of an atom traveling with velocity @xmath46 is then given in the scaled unit as @xmath47 i.e. the kinetic energy relative to the initial thermal energy characterized by @xmath9 is what matters .
all the major physics of this system can therefore be covered by studying various combination of wedge angle , initial temperatures and the box threshold energies .
in this section , we illustrate the physics of the system by presenting the results of numerical simulation for select wedge angles using the experimentally verified parameters of refs . @xcite and @xcite .
the results of our simulation for various regimes characterized by the three different wedge angles @xmath48 , @xmath49 , and @xmath50 are presented in figs .
[ fig1]-[fig3 ] respectively . in each of these figures
, we present several sub - figures .
first of all , as one of the sub - figures , we present the poincar section for the wedge billiard typically presented as a plot of @xmath51 vs. @xmath52 where @xmath51 and @xmath53 denote transverse and normal velocity of the atoms immediately after hitting the walls . as is well known , visible geometric structures or `` islands '' within the poincar section represent regions of regular orbit while the regions containing evenly spread dots correspond to chaotic dynamics . the fact that there are no islands for wedge angles @xmath49 and @xmath50 ( figs .
[ fig2 ] and [ fig3 ] ) means that all atoms are expected to undergo chaotic dynamics , while the existence of an island for wedge angle @xmath48 ( fig .
[ fig1 ] ) means atoms with initial conditions that lie in the island undergo regular orbits .
the initial thermal velocity distribution in terms of @xmath51 and @xmath52 after one bounce off the walls are shown as ( thicker ) red dots on the poincar section as a guide to the initial condition used .
next , typical trajectory of one particular atom in the wedge billiard is presented as well as all the positions of the optimum box in small green squares .
the outline of the type i box determined from the optimum box calculation is also shown superposed in the same sub - figure . in the right hand column of the figures ,
the time - varying positions of the optimum box decomposed into @xmath34 and @xmath28 components are presented along with the boundaries of the type i box shown in dashed lines . in fig .
[ allangle ] we present various fractions as a function of wedge angles atoms trapped by the box , atoms remaining in the wedge , and atoms lost for all 3 box types . as mentioned previously , there are two mechanisms by which the atoms are lost one by encountering the box with the wrong energy and the other by escaping the wedge billiard after the collision with one of the walls .
we also show the @xmath34 and @xmath28 components of the type i and type ii boxes as a function of wedge angle .
it is found that the type ii box gives slightly better result in terms of the fraction of atoms trapped .
it is seen that , for the parameters used , from @xmath54 to up to around @xmath32 of the atoms could be trapped by this scheme .
the fraction of atoms trapped has ( discounting fluctuations ) almost linear dependence on the wedge angle with the largest fraction of atoms trapped at @xmath55 .
these observations imply that chaotic dynamics , rather than the regular dynamics , is more conducive for spc .
this makes sense since , to be trapped by the box , the atom has to reach the correct height relative to the initial energy that corresponds to the right kinetic energy .
this is more likely in the regime of chaotic dynamics in which the atoms take on various different trajectories over time rather than that of regular orbit which generally has higher kinetic energy and a limited range of trajectories .
the wide range of trajectories also means that more atoms are likely to encounter the stationary box .
the higher fraction of atoms trapped for larger wedge angle can be explained from the fact that with wider wedge angles , the normal velocity component @xmath53 dominates that of the transverse velocity component @xmath51 and hence one has a higher fraction of atoms in parabolic motion . with parabolic motion
it is easier for the atoms to attain the trappable energy as the kinetic energy goes right down to zero at the turning points .
finally we present here a simple analysis regarding the height of the best fixed box by assuming that the box is able to catch the atom after one bounce from the walls . taking the cusp of the wedge billiard as the origin of position , and considering an atom initially at position @xmath56 and @xmath57 with initial velocity components @xmath58 and @xmath59 , the @xmath34 and @xmath28 components of the velocity on impact with the wall at angle @xmath60 from the perpendicular line at the origin are @xmath61 and @xmath62}$ ] .
energy conservation implies : @xmath63 where @xmath20 is the threshold energy of the box .
this gives the best box height as @xmath64 this result is valid for larger @xmath60 where the first bounce results in a parabolic trajectory e.g. when @xmath65 where @xmath66 is the length of the wedge wall . for smaller
@xmath67 a regular orbit is likely which means any analysis based on one bounce does not hold .
a plot of @xmath5 given by eq .
( [ hbest ] ) was found to closely reproduce the large angle @xmath68 part of fig .
[ allangle](d ) .
the cases where @xmath69 is large enough ( or @xmath60 is small enough ) so that the atom hits the wall due to the horizontal rather than the vertical component of motion require a modified analysis : assuming the time it takes for an atom to hit the wall is @xmath70 , the @xmath28 component of the velocity at the wall is @xmath71 while @xmath34 component of the velocity at the wall is @xmath58 .
the vertical displacement after @xmath70 is @xmath72 and this means the distance traversed horizontally @xmath73 \tan \theta $ ] , i.e. @xmath74 $ ] .
the energy conservation then implies @xmath75 \nonumber \\ & \approx & \frac{e^{i}_k - e_{box } } { mg } - \frac{y_i v_y^i}{v_x^i } \theta + \frac{g y_i^2}{2 v_x^{i \ ; 2 } } \theta^2 + \cdots \label{hbest2}\end{aligned}\ ] ] this reproduces the result for smaller @xmath60 but again , since we are only considering one bounce and ignoring the possibility of a regular orbit , the oscillatory behavior is not recovered in this very approximate result .
in this section , we calculate the cooling efficiency @xmath76 for spc in wedge billiard . in particular , we identify general features of this system by considering various combination of the three core parameters : wedge angle , initial temperature , and the box threshold energy . to give a better sense of the range of parameter values used in this section , we denote the initial temperature and the box threshold energy of existing experiments as @xmath77 and @xmath78 respectively and refer to all the other initial temperatures and threshold energies as some multiples of @xmath77 and @xmath78 . in fig .
[ psdinittemp ] , we present the cooling efficiency for type i and type ii boxes .
figure [ psdinittemp ] is arranged similarly to fig .
[ allangle ] , except that we present the cooling efficiency @xmath76 for the four different initial temperatures of @xmath77 , @xmath79 , @xmath80 , and @xmath81 for the box threshold temperature of @xmath78 .
we also show , as done in fig .
[ allangle ] , the positions of type i and type ii boxes as a function of wedge angle for different initial temperatures .
we chose to look at such widely varying initial temperatures up to @xmath81 since it was found numerically that smaller changes did not noticeably affect the cooling efficiency e.g. the difference in result between say @xmath77 and even @xmath82 was not noticeable . such immunity to temperature changes means the results shown here for each of these four initial temperatures should hold for a wide range of atomic species and length scales .
it is seen that there is , in fact , quite a dramatic increase in phase space density the higher the initial temperature the more significant is the improvement .
it is also seen that due to the interplay of various factors in the calculation of @xmath76 , the curves do not show very strong dependence on the wedge angle , although one can notice slight changes in the ( shallow ) maximum of the curves .
the curves have maximum near @xmath83 for the initial temperature of @xmath77 and near @xmath49 for the initial temperature of @xmath81 .
given that the ratio of the area @xmath84 has maximum at the wedge angle @xmath85 this implies that the fraction of atoms trapped ( i.e. instead of @xmath76 that also compares the temperature difference ) does not change much with the change in the wedge angle for higher initial temperature .
indeed , the fraction of atoms trapped was found to increase almost linearly with the increasing wedge angle , but the gradient of this linear variation was highest for the initial temperature of @xmath77 and the smallest for the initial temperature of @xmath81 .
the fact that the fraction of atoms trapped was the highest for @xmath86 for all initial temperatures ( albeit with a varying degree ) re - confirms the main mechanism by which the atoms get trapped by the box with wider wedge angle , one has a higher fraction of atoms in parabolic trajectory .
the change in phase space density corresponding to the change in initial temperature from @xmath80 to @xmath81 is smaller compared to that with change in initial temperature from @xmath79 to @xmath80 i.e. there is a saturation in the amount of increase in phase space density .
this can be partly explained by examining the fraction of atoms trapped by the box .
it was found that the fraction of atoms trapped did not change much at all between the initial temperatures of @xmath77 and @xmath79 i.e. the temperature ratio has a major effect . over @xmath79 , however , the fraction of trapped atoms began to be noticeably reduced .
this is because after the temperature gets high enough atoms have enough energy to escape the wedge in significant numbers . with many atoms escaping and reducing the pool of atoms inside the wedge ,
the actual fraction of atoms trapped by the box goes down , offsetting the effect of the larger temperature ratio .
finally we note that the results of type i and type ii are very similar .
this is because of two reasons
one is that the maximum fraction of atoms trapped using type ii boxes , while greater than that using type i box , is not significantly different .
the second reason is that the finite box size implies there is actually a significant overlap between the two different types of box positions .
the finite box size also means that the slightly different box positions for different initial temperature ( for both type i and type ii ) is not noticeable the temperature - dependent difference in the position of the box center is smaller than the width of the box itself . similarly to fig .
[ psdinittemp ] , we show in fig .
[ psdboxtemp ] the cooling efficiency @xmath76 and the corresponding positions for type i and type ii boxes .
but this time , we fix the initial temperature at @xmath87 and show the general trend with respect to four different box threshold temperatures of @xmath78 , @xmath88 , @xmath89 , and @xmath90 . the initial temperature of @xmath79 was chosen to make sure that even with box threshold temperature of @xmath90 the atomic sample is actually being cooled down . the observed trend with increasing box threshold energies
is the opposite of the trend observed above with increasing temperature : @xmath76 _ decreases _ with increasing box threshold energy , which is unexpected since with increasing box threshold energy , the actual fraction of atoms trapped is increased as the atoms are more likely to be trapped .
this can be understood from the fact that , with increasing box threshold energy , the ratio of initial to final temperature is decreased and the temperature ratio is the more significant contributor to @xmath76 .
again , there is a relatively weak dependence of @xmath76 on the wedge angle .
the wedge angle at which the maximum @xmath76 occurs is observed to be slightly shifted from near @xmath83 for the box threshold energy of @xmath78 to near @xmath49 for the box threshold energy of @xmath90 .
as before , this can be explained by the considering the ratio of the areas @xmath84 and the fraction of atoms trapped .
in fact , the fraction of the atoms trapped as a function of wedge angle demonstrates quite a different behavior from that seen above for different initial temperatures .
although the fraction of atoms trapped with box threshold energy of @xmath78 has roughly linear dependence on the wedge angle ( with maximum near @xmath50 ) , with higher box threshold energies , the fraction of atoms trapped gradually demonstrates a `` triangular '' shape as a function of wedge angle ( with maximum near @xmath49 ) .
interestingly , with the box threshold energy of @xmath90 , it matches the triangular shape seen for the optimum box result as dotted line in fig .
[ allangle](a ) and ( c ) .
this shows that with large enough box threshold energy , all atoms that can be trapped as calculated by the optimum box are indeed trapped given enough time .
this is a significant result since it shows a potential to `` simulate '' the complicated optimum box calculation experimentally .
a saturation is found to occur as the box threshold energy becomes larger ; changes in @xmath76 due to changes in box threshold energy become smaller as the box threshold energy becomes close to @xmath90 .
this is to be expected since , once the box threshold energy becomes large enough to catch all the atoms that are physically feasible to be caught ( i.e. ones that do not escape the wedge billiard ) , any higher box threshold energy will not give different fraction of atoms trapped .
as regards to the actual box position itself and the use of type i and type ii boxes , similar observation as above is made i.e. the final cooling efficiency for the two different box position is quite similar and that , due to the finite width of the box , difference in box position do not lead to noticeable changes in @xmath76 .
we have found that spc can significantly increase the phase space density of the atoms originally trapped inside a wedge billiard . it was found that even in the very tough scenario of very high initial temperature and very low box threshold energy , enough atoms are expected to be caught to give a high cooling efficiency .
the cooling efficiency @xmath76 showed a relatively weak dependence on the wedge angle .
this is because @xmath76 depends on the interplay of various factors , not just the number of atoms trapped by the box .
for low initial temperature , and also for low box threshold energy , the best angle was near @xmath91 and for higher initial temperature and also for higher box threshold energy , the best angle was near @xmath92 .
various trends could be explained by studying the fraction of atoms trapped as a function of the wedge angle .
both type i and type ii boxes were found to give similar cooling efficiencies , although the type ii box is one that can be obtained experimentally in the absence of any knowledge about the idealized optimum box calculation .
the regime in which the performance of spc in wedge billiard is best was identified to be the regime of chaotic dynamics with wedge angle @xmath93 . in the chaotic regime , atoms take on various different kinetic energies and trajectories , increasing the likelihood of meeting the condition for spc .
although on first sight atoms in regular orbits look more promising , their kinetic energy is less likely to be redistributed , resulting in a smaller cooling efficiency . within the chaotic regime
, it was found that the wider wedge angle is better in terms of capturing a higher fraction of atoms .
this is because , with a wider wedge angle , there are more of atoms undergoing parabolic motion where the kinetic energy becomes zero at the turning point .
these observations indicate that , in general , the best set up for spc is the one where the box is able to access the most number of turning points concentrated within a small region .
somewhat less ideal , but an experimentally more feasible case would be to release the atoms in a rectangular trap@xcite i.e. with a flat bottom or on a magnetic mirror@xcite so that all the atoms are bouncing in parabolic trajectories , and scan the optical box across so as to `` skim '' all the trappable atoms from the top , and slowly move down to capture progressively less energetic atoms . with this arrangement the box is always at the top , so there is no worry about unnecessarily stirring atoms using a depopulation beam and losing them .
this will be presented in our future work@xcite .
1 g. n. price , s. t. bannerman , k. viering , e. narevicius , and m. g. raizen phys .
lett . * 100 * 093004 ( 2008 ) a. ruschhaupt and j. g. muga phys .
rev . a * 70 * , 061604(r ) ( 2004 ) m. g. raizen , a. m. dudarev , qian niu , and n. j. fisch phys .
lett . * 94 * 053003 ( 2005 ) j. j. thorn , e. a. schoene , t. li , and d. a. steck phys .
a * 79 * , 063402 ( 2009 ) h. e. letihet , and b. n. miller physica ( amsterdam ) * 21d * 93 ( 1986 ) v. milner , j. l. hanssen , w. c. campbell , and m. g. raizen phys . rev
. lett . * 86 * 1514 ( 2001 ) c. g. townsend , n. h. edwards , c. j. cooper , k. p. zetie , c. j. foot , a. m. steane , p. szriftgiser , h. perrin , and j. dalibard phys .
a * 52 * , 1423 ( 1995 ) t. p.
meyrath , f. schreck , j. l. hanssen , c .- s .
chuu , and m. g. raizen phys .
a * 71 * , 041604(r ) ( 2005 ) t. m. roach , h. abele , m. g. boshier , h. l. grossman , k. p. zetie , and e. a. hinds phys .
lett . * 75 * , 629 ( 1995 ) .
s. choi , b. sundaram , m. g. raizen ( in preparation )
|
single - photon cooling ( spc ) , noted for its potential as a versatile method for cooling a variety of atomic species , has recently been demonstrated experimentally . in this paper , we study possible ways to improve the performance of spc by applying it to atoms trapped inside a wedge billiard .
the main feature of wedge billiard for atoms , also experimentally realized recently , is that the nature of atomic trajectories within it changes from stable periodic orbit to random chaotic motion with the change in wedge angle .
we find that a high cooling efficiency is possible in this system with a relatively weak dependence on the wedge angle , and that chaotic dynamics , rather than regular orbit , is more desirable for enhancing the performance of spc .
|
current induced magnetization switching in metallic spin valves and magnetic tunnel junctions ( mtjs ) is an intensive area of research for its possible applications in future spintronic / magnetic devices @xcite . during electronic transport ,
inseparable from the charge current the electrons also transfer spins .
if electron transfer occurs between two ferromagnets having a noncollinear magnetic alignment , the transmitted spin transfer torque ( stt ) is directly responsible for the magnetization switching .
the first direct measurement of the in - plane and out - of - plane components of the stt vector in a mtj was performed by sankey _
_ @xcite . on the other hand
, there is a number of different theoretical approaches reported to calculate the stt vector in mtjs .
a brief overview is given below .
free - electron models have been employed by several authors @xcite .
they used the wentzel - kramers - brillouin ( wkb ) approximation for the propagated wave functions .
slonczewski worked out the theory of stt within bardeen s transfer hamiltonian method @xcite .
tight - binding models in combination with keldysh nonequilibrium green s function formalism have been extensively used @xcite .
very recently , the effects of magnetic insulating @xcite and ferroelectric @xcite barriers on the stt have been investigated by the tight - binding method .
a scattering theory of stt has been proposed by xiao _
a fully _ ab initio _ method based on the korringa - kohn - rostoker multiple scattering theory has been provided by heiliger and stiles @xcite . as of particular technological importance , they studied the fe / mgo / fe mtj , and found quantitative agreement with the experimental stt results of sankey _
_ studied the same fe / mgo / fe mtj by a first - principles - based transport method and considered the effect of disorder scattering in mgo @xcite .
the listed theoretical models consider symmetric or asymmetric planar junction geometries to address stt effects in mtjs .
however , there is a growing interest to obtain information on local stt properties in high spatial resolution in asymmetric mtjs .
spin - polarized scanning tunneling microscopy ( sp - stm ) proved to be extremely useful to provide local information on spin - polarized charge transport properties in a wide variety of magnetic surfaces @xcite .
as an important recent application , local current ( and thus stt ) pulses have been used to create and annihilate individual magnetic skyrmion structures in a thin film system @xcite .
based on above , we propose the generalization of high resolution stm charge transport theories @xcite to include the description of vector spin transport in stm junctions . in this work ,
we present a combined theory for charge and vector spin transport in magnetic stm junctions within the three - dimensional ( 3d ) wkb electron tunneling framework , including electronic structures of both the surface and the tip calculated by first principles methods .
we investigate the fe / w(110 ) surface motivated by a recent indirect local stt measurement performed by sp - stm @xcite . based on our results ,
we propose that the stt efficiency in this system can be enhanced by about a factor of seven by selecting a proper bias voltage .
moreover , given the exponentially decaying electron transmission in the mtj with increasing tip - sample distance , we find that the torque - current relationship follows the power law rather than a linear function .
the paper is organized as follows : the 3d - wkb theoretical model of combined charge and vector spin transport in magnetic stm is presented in section [ sec_spstm ] .
computational details of the electronic structure calculations are given in section [ sec_comp ] .
we investigate the electron charge and spin transport in a magnetic tunnel junction consisting of one monolayer fe on the w(110 ) surface in combination with a fe(001 ) tip in section [ sec_res ] .
we focus on the stm contrast formation of the spin - polarized charge current and its determining factors in section [ sec_stm ] , and study vector spin transport characteristics by analyzing the spin transfer torque vector and the longitudinal spin current and their relation to the spin - polarized charge current in section [ sec_torque ] .
our conclusions are found in section [ sec_conc ] .
based on atom - superposition techniques @xcite , the 3d - wkb electron tunneling theory @xcite implemented in the 3d - wkb - stm code @xcite became an established method for the simulation of ( spin - polarized ) scanning tunneling microscopy and spectroscopy .
the model assumes that electron tunneling occurs through a tip apex atom , and individual transitions between this tip apex atom and a suitable number of sample surface atoms , each described by the one - dimensional ( 1d ) wkb approximation , are superimposed .
orbital - dependent tunneling is included in the model by a modified transmission function @xcite . since the 3d geometry of the tunnel junction is considered , the method is , in effect , a 3d - wkb atom - superposition approach .
its main advantages are computational efficiency , see , e.g. , refs .
@xcite , and an enhanced parameter space for modeling tip geometries , i.e. , arbitrary relative tip - sample orientations @xcite .
the 3d - wkb method has been applied to spin - polarized stm , and an analysis of the interplay of orbital - dependent and spin - polarization effects was reported on the stm image contrast formation above the fe(110 ) surface @xcite .
for the combined 3d - wkb model for charge and vector spin transport , we adopt the orbital- and spin - dependent formalism introduced in ref .
the electronic structure of the magnetic surface and the tip is included in the model by taking the atom - projected electron density of states ( pdos ) obtained by _ ab initio _ electronic structure calculations .
the orbital - decomposition of the pdos is necessary for the description of the orbital - dependent electron tunneling @xcite , and both the charge pdos and the magnetization pdos vector are essential for spin - dependent tunneling @xcite .
the energy - dependent orbital - decomposed charge pdos of the @xmath0th sample surface ( @xmath1 ) atom with orbital symmetry @xmath2 and the tip ( @xmath3 ) apex atom with orbital symmetry @xmath4 are denoted by @xmath5 and @xmath6 , respectively . in our model @xmath7
atomic orbitals are considered but this can be extended to treat @xmath8 electron orbitals as well @xcite . similarly to the charge pdos , @xmath9 and @xmath10 denote the corresponding energy - dependent orbital - decomposed magnetization pdos vectors . in the simplest case there
are fixed spin quantization axes of the sample and the tip that can be described by energy - independent @xmath11 and @xmath12 unit vectors , respectively .
we focus on this case in the present work , allowing an arbitrary @xmath13 angle between @xmath11 and @xmath12 .
the corresponding pdos can be calculated from the spin - up ( @xmath14 ) and spin - down ( @xmath15 ) contributions obtained from first principles collinear magnetic calculations for the sample surface and the tip separately @xcite : the orbital - decomposed charge pdos are @xmath16 and the orbital - decomposed magnetization pdos vectors are @xmath17\mathbf{e}_s,\nonumber\\ \mathbf{m}_{t\tau}(e)&=&m_{t\tau}(e)\mathbf{e}_t=\left[n^{\uparrow}_{t\tau}(e)-n^{\downarrow}_{t\tau}(e)\right]\mathbf{e}_t.\end{aligned}\ ] ] the total charge pdos of the @xmath0th sample surface atom and the tip apex atom is the sum of the orbital - decomposed contributions : @xmath18 similarly , the total magnetization pdos vectors can be obtained as the sum of the orbital - decomposed contributions : @xmath19\mathbf{e}_s,\nonumber\\ \mathbf{m}_t(e)&=&\sum_{\tau}\mathbf{m}_{t\tau}(e)=\left[\sum_{\tau}m_{t\tau}(e)\right]\mathbf{e}_t.\end{aligned}\ ] ] note that a similar decomposition of the green s functions was reported within the linear combination of atomic orbitals ( lcao ) framework in ref .
@xcite .
let us define the following energy - dependent orbital - decomposed density matrices in spin space for the @xmath0th sample surface atom and the tip apex atom @xcite : @xmath20 here , @xmath21 is the @xmath22 unit matrix and @xmath23 is the pauli matrix vector .
the main advantage of using the density matrix formalism is that electronic and spin structures calculated either relativistically [ with spin - orbit coupling included , general @xmath24 or nonrelativistically [ @xmath25 , used in the present work ] can be treated within the same theoretical framework @xcite .
the charge ( @xmath26 ) and vector spin ( @xmath27 ) transport of the electron tunneling transition between the @xmath0th sample surface atom and the tip apex atom at energy @xmath28 can be represented by the following traces in spin space : @xmath29,\nonumber\\ \tilde{\mathbf{t}}^a_{\sigma\tau}(e)&=&\frac{1}{2}tr\left(\underline{\underline{\rho}}_{s\sigma}^a(e)\underline{\underline{\rho}}_{t\tau}(e)\underline{\underline{\bm{\sigma}}}\right)\nonumber\\ % & = & \left[n_{s\sigma}^a(e)\mathbf{m}_{t\tau}(e)+\mathbf{m}_{s\sigma}^a(e)n_{t\tau}(e)\right]+i\left[\mathbf{m}_{s\sigma}^a(e)\times\mathbf{m}_{t\tau}(e)\right]\nonumber\\ & = & \left[n_{s\sigma}^a(e)m_{t\tau}(e)\mathbf{e}_t+m_{s\sigma}^a(e)n_{t\tau}(e)\mathbf{e}_s\right]+im_{s\sigma}^a(e)m_{t\tau}(e)\left[\mathbf{e}_s\times\mathbf{e}_t\right ] .
\label{eq_chsp}\end{aligned}\ ] ] since @xmath30 with @xmath13 the angle between the spin quantization axes of the sample surface and the tip , the formula of the charge transport ( conductance ) is formally consistent with the spin - polarized tersoff - hamann model @xcite , except the fact that it explicitly includes the electronic structure of the tip apex @xcite : @xmath31 where the first and second term is respectively the non - spin - polarized and spin - polarized part of the charge current contribution . on the other hand ,
the in - plane and out - of - plane components of the spin transport formula are : @xmath32 here , the in - plane component @xmath33 lies in the plane spanned by the @xmath11 and @xmath12 vectors .
its direction is , however , not necessarily perpendicular to the magnetic moment the torkance is acting on , it rather depends on the electronic structures of both the surface ( @xmath34 ) and the tip ( @xmath35 ) sides of the stm junction .
this implies that @xmath33 can be decomposed to a longitudinal spin conductance part ( parallel to the magnetic moment direction @xmath36 ) and to a spin torkance part ( perpendicular to the magnetic moment direction @xmath36 ) @xcite : @xmath37 , where @xmath38 denotes that this in - plane ( slonczewski ) torkance vector is in the @xmath11-@xmath12 plane . here , the two components are : @xmath39\mathbf{e}_j = t^{a , long - j}_{\sigma\tau}(e)\mathbf{e}_j,\nonumber\\ \tilde{\mathbf{t}}^{a , j\parallel}_{\sigma\tau}(e)&=&\tilde{\mathbf{t}}^{a , in - pl.}_{\sigma\tau}(e)-t^{a , long - j}_{\sigma\tau}(e)\mathbf{e}_j,\end{aligned}\ ] ] with @xmath40 depending on the direction of spin transport , i.e. , @xmath41 for tip @xmath42 sample and @xmath43 for sample @xmath42 tip tunneling . the longitudinal spin conductance term used for @xmath44
is motivated by the limiting case of @xmath45 ( @xmath46 ) , where @xmath47 . since @xmath48
, @xmath49 is perpendicular to the @xmath11-@xmath12 plane and in line with the @xmath50 normal - to - plane unit vector .
its size depends on the magnetization pdos @xmath51 and on the @xmath13 angle .
since this vector is definitely perpendicular to both magnetic moments , we identify it as the out - of - plane ( field - like ) torkance vector @xcite : @xmath52 . from the above formalism
it is clear that this out - of - plane torkance corresponds to conductance - induced torkance solely as a result of electron tunneling through the vacuum barrier and does not include contributions related to the interlayer exchange coupling in planar junctions @xcite that result in the so - called equilibrium torque @xcite , which is non - zero even at zero bias voltage ( @xmath53 ) .
therefore , throughout this work , the out - of - plane torque corresponds to purely current - induced torque , @xmath54 using the notation of ref .
@xcite . the total tunneling charge current , the components of the spin transfer torque ( stt ) vector and the longitudinal spin current are calculated from the charge conductance , the components of the spin torkance vector and the longitudinal spin conductance by the superposition of atomic contributions from the sample surface ( sum over @xmath0 ) and the superposition of transitions from all atomic orbital combinations between the sample and the tip ( sum over @xmath2 and @xmath4 ) integrated in an energy window corresponding to the bias voltage @xmath55 .
assuming elastic electron tunneling at temperature @xmath56 k , the charge current , the stt vector components , the longitudinal spin current , the stt vector and the magnitude of the stt at the tip apex position @xmath57 and bias voltage @xmath55 are given by @xmath58^{1/2 } , \label{eq_current}\end{aligned}\ ] ] with @xmath40 depending on the direction of spin transport , i.e. , @xmath41 for tip @xmath42 sample and @xmath43 for sample @xmath42 tip tunneling .
therefore , @xmath59 naturally labels the side ( magnetic surface or tip ) , where the torque is acting on .
one particular contribution can be calculated as an integral in an energy window corresponding to the bias voltage @xmath55 as @xmath60du,\nonumber\\ \mathbf{t}^{a,\perp}_{\sigma\tau}\left(\mathbf{r}_{tip},v\right)&=&\epsilon^2 e\int_0^v\hat{t}_{\sigma\tau}\left(e_f^s+eu , v,\mathbf{d}_a\right)\nonumber\\ & \times&m_{s\sigma}^a\left(e_f^s+eu\right)m_{t\tau}\left(e_f^t+eu - ev\right)\sin(\phi)\mathbf{e}_{\{\mathbf{e}_s\times\mathbf{e}_t\}}du,\nonumber\\ \mathbf{t}^{a , s\parallel}_{\sigma\tau}\left(\mathbf{r}_{tip},v\right)&=&\epsilon^2 e\int_0^v\hat{t}_{\sigma\tau}\left(e_f^s+eu , v,\mathbf{d}_a\right)\nonumber\\ & \times&n_{s\sigma}^a\left(e_f^s+eu\right)m_{t\tau}\left(e_f^t+eu - ev\right)\left[\mathbf{e}_t-\mathbf{e}_s\cos(\phi)\right]du,\nonumber\\ \mathbf{t}^{a , t\parallel}_{\sigma\tau}\left(\mathbf{r}_{tip},v\right)&=&\epsilon^2 e\int_0^v\hat{t}_{\sigma\tau}\left(e_f^s+eu , v,\mathbf{d}_a\right)\nonumber\\ & \times&m_{s\sigma}^a\left(e_f^s+eu\right)n_{t\tau}\left(e_f^t+eu - ev\right)\left[\mathbf{e}_s-\mathbf{e}_t\cos(\phi)\right]du,\nonumber\\ t^{a , long - s}_{\sigma\tau}\left(\mathbf{r}_{tip},v\right)&=&\epsilon^2 e\int_0^v\hat{t}_{\sigma\tau}\left(e_f^s+eu , v,\mathbf{d}_a\right)\nonumber\\ & \times&\left[n_{s\sigma}^a\left(e_f^s+eu\right)m_{t\tau}\left(e_f^t+eu - ev\right)\cos(\phi)\right.\nonumber\\ & + & \left .
m_{s\sigma}^a\left(e_f^s+eu\right)n_{t\tau}\left(e_f^t+eu - ev\right)\right]du,\nonumber\\ t^{a , long - t}_{\sigma\tau}\left(\mathbf{r}_{tip},v\right)&=&\epsilon^2 e\int_0^v\hat{t}_{\sigma\tau}\left(e_f^s+eu , v,\mathbf{d}_a\right)\nonumber\\ & \times&\left[n_{s\sigma}^a\left(e_f^s+eu\right)m_{t\tau}\left(e_f^t+eu - ev\right)\right.\nonumber\\ & + & \left .
m_{s\sigma}^a\left(e_f^s+eu\right)n_{t\tau}\left(e_f^t+eu - ev\right)\cos(\phi)\right]du .
\label{eq_current_decomp}\end{aligned}\ ] ] here , @xmath61 is the elementary charge , @xmath62 is the planck constant , and @xmath63 and @xmath64 are the fermi energies of the sample surface and the tip , respectively .
the @xmath65 and @xmath66 factors ensure the correct dimensions of the quantities .
the value of @xmath67 has to be determined by comparing the simulation results of the charge current with experiments , or with calculations using standard methods , e.g. , the bardeen approach @xcite . in our simulations
@xmath68 ev was chosen that gives comparable current values with those obtained by the bardeen method @xcite implemented in the bskan code @xcite .
note that the choice of @xmath67 has no qualitative influence on the reported results .
the formal equivalence of the in - plane torque vectors in eq .
( [ eq_current_decomp ] ) with eq .
( 6 ) of ref .
@xcite is proven in appendix . in eq .
( [ eq_current_decomp ] ) , @xmath69 is the orbital - dependent tunneling transmission function that gives the probability of the electron charge / spin tunneling from the @xmath4 orbital of the tip apex atom to the @xmath2 orbital of the @xmath0th surface atom , or vice versa , depending on the sign of the bias voltage .
the conventions of tip @xmath42 sample tunneling at positive bias voltage ( @xmath70 ) and sample @xmath42 tip tunneling at negative bias ( @xmath71 ) are used .
the stt is acting on the magnetic moments where the charge current flows , i.e. , on the sample moments at positive and on the tip moments at negative bias voltage .
the transmission probability depends on the energy of the electron ( @xmath28 ) , the bias voltage ( @xmath55 ) , and the relative position of the tip apex atom and the @xmath0th sample surface atom ( @xmath72 ) .
the following form for the transmission function is considered @xcite : @xmath73 here , the exponential factor corresponds to an orbital - independent transmission , where all electron states are considered as exponentially decaying spherical states @xcite .
this factor depends on the distance between the tip apex atom and the @xmath0th surface atom , @xmath74 , and on the 1d - wkb vacuum decay of electron states , @xmath75 for using this @xmath76 an effective rectangular potential barrier in the vacuum between the sample and the tip is assumed .
@xmath77 and @xmath78 are the electron work functions of the sample surface and the tip , respectively , @xmath79 is the electron s mass and @xmath80 the reduced planck constant .
the remaining factors of eq .
( [ eq_transmission ] ) are responsible for the orbital dependence of the transmission .
they modify the exponentially decaying part according to the real - space shape of the electron orbitals involved in the tunneling , i.e. , the angular dependence of the electron densities of the atomic orbitals of the surface and the tip is taken into account as the square of the real spherical harmonics @xmath81 and @xmath82 , respectively .
the distinction between the local coordinate systems of the sample surface ( without primes ) and the tip apex atom ( denoted by primes ) enables the consideration of arbitrary relative tip - sample geometrical orientations within the orbital - dependent 3d - wkb approach @xcite .
the polar and azimuthal angles @xmath83 given in both real spherical harmonics in eq .
( [ eq_transmission ] ) correspond to the tunneling direction , i.e. , the line connecting the @xmath0th surface atom and the tip apex atom , as viewed from their local coordinate systems , and they have to be determined from the actual tip - sample geometry . it is important to highlight that the magnitude of both stt vector components , and thus the magnitude of the stt , are proportional to @xmath84 . on the other hand ,
the spin - polarized part of the charge current is proportional to @xmath85 .
therefore , if @xmath11 and @xmath12 are in line ( @xmath46 or @xmath86 ) then both torque vector components and the stt vanish and the spin - polarized part of the charge current is largest in magnitude .
we have performed geometry optimization and electronic structure calculations within the generalized gradient approximation ( gga ) of the density functional theory ( dft ) implemented in the vienna ab - initio simulation package ( vasp ) @xcite . a plane wave basis set for electronic wave function expansion in combination with the projector augmented wave ( paw ) method @xcite has been applied .
we have adopted the parametrization of the exchange - correlation functional by perdew and wang ( pw91 ) @xcite .
the electronic structures of the fe / w(110 ) surface and the fe(001 ) tip have been calculated separately . it should be noted that spin - orbit coupling effects have been neglected in our dft calculations , and @xmath87 in eq .
( [ eq_rho ] ) .
we model the fe / w(110 ) surface by a slab of nine atomic layers ( 8 tungsten and 1 iron layers ) with the bulk lattice constant of @xmath88 .
a separating vacuum region of 18 width in the surface normal ( @xmath89 ) direction between neighboring supercell slabs is set up to minimize slab - slab interaction .
after geometry optimization the fe - w interlayer distance between the two topmost layers is reduced by 14% , and the underneath w - w interlayer distance is increased by 1% in comparison to bulk tungsten .
the magnitude of the in - plane spin moment of the surface fe atoms is 2.31 @xmath90 and the induced moments in the neighboring w layer are -0.14 @xmath90 per tungsten atom ( opposite in orientation compared to the surface fe spin moments ) .
the structural parameters and magnetic properties of fe / w(110 ) are in good agreement with refs .
the spin quantization axis ( @xmath11 , magnetic easy axis ) is pointing to the @xmath91 $ ] crystallographic direction .
the average electron work function above the fe / w(110 ) surface is @xmath92 ev calculated from the local electrostatic potential .
we have used an @xmath93 monkhorst - pack @xcite k - point grid for obtaining the orbital - decomposed projected charge and magnetization electron dos onto the surface fe atom , @xmath94 and @xmath95 , respectively .
we model the stm tip as a blunt iron tip , where a single fe apex atom is placed on the hollow position of the fe(001 ) surface , similarly as in ref .
more details of the fe(001 ) tip model can be found in ref .
the orbital - decomposed electronic structure data projected to the tip apex atom , @xmath6 and @xmath96 , have been calculated using a @xmath97 monkhorst - pack k - point grid .
the orientation of the tip spin quantization axis @xmath12 relative to @xmath11 is arbitrarily changed and their angle is denoted by @xmath13 . for calculating the spin - polarized charge current and the spin transfer torque vector , we choose 112 iron atoms on the fe / w(110 ) surface for the atomic superposition @xcite .
figure [ fig1 ] shows the energy - dependent orbital - decomposed charge pdos ( @xmath98 ) and spin - resolved pdos ( @xmath99 ) functions of the fe surface atom in fe / w(110 ) and the fe(001 ) tip apex atom .
we find that the @xmath100 partial pdos is dominating over @xmath101 and @xmath102 partial pdos , and the minority spin character is dominating over the majority spin in the vicinity of the fermi levels .
the obtained results are in good agreement with the total ( orbital - summed ) fe pdos functions reported in refs .
@xcite , where the full potential linearized augmented plane wave ( flapw ) method was employed in different computational implementations ( wien2k and fleur codes ) . [ cols="^,^ " , ]
figure [ fig5 ] shows the tip - sample - distance - dependent in - plane and out - of - plane stt efficiencies and the longitudinal spin current ratio calculated at -0.2 v bias voltage above two surface positions of the fe / w(110 ) surface in seven relative tip - sample magnetization orientations each .
we find that the @xmath89-dependence of the in - plane and out - of - plane stt efficiencies differs considerably , and the in - plane component is always larger by at least a factor of four than the out - of - plane one , in agreement with refs .
this means that the magnitude of the stt is , in effect , determined by the in - plane component at this bias voltage , similarly found in metallic spin valves @xcite .
we also find considerable differences between the @xmath89-dependent stt efficiencies above the top and hollow surface sites of fe / w(110 ) .
the in - plane stt efficiencies ( and thus the magnitudes of the stt ) converge to the same @xmath13-dependent curves above @xmath103 8 for both surface sites .
we recover the already mentioned exponentially increasing stt efficiency [ @xmath104 at large tip - sample distances ( above @xmath103 8 ) .
note that in symmetric tunnel junctions an opposite type of @xmath89-decay ( faster stt than charge current decay ) can occur , which results in an exponentially vanishing stt efficiency as the tip - sample distance ( barrier thickness ) increases @xcite . in our asymmetric stm junction
, we find the maximal stt efficiency values at the investigated upper boundary of @xmath103 10 for all considered @xmath13 values above both surface sites [ fig . [ fig5](a ) and ( b ) ] .
interestingly , we also observe large stt efficiency at small tip - sample distances ( @xmath105 ) above the surface hollow site [ fig .
[ fig5](b ) ] . for the out - of - plane stt efficiency
we find that maximal values can be obtained at small tip - sample distances : the local maximum is found at @xmath103 4.1 in fig .
[ fig5](c ) and at @xmath103 3 in fig .
[ fig5](d ) . however , there are local minima at @xmath103 6.7 and @xmath103 3.9 , respectively , in the latter case quite close to the local maximum .
note that the almost perfect @xmath106-scaling of the stt efficiency curves in fig .
[ fig5](a)-(d ) is recovered .
the longitudinal spin current ratios ( @xmath107 ) shown in fig .
[ fig5](e , f ) are comparable in size with the in - plane stt efficiencies and they show complex @xmath89- and @xmath13-dependences resulting from a complex interplay of orbital - dependent transmission and electronic structure effects , see eq .
( [ eq_current_decomp ] ) .
finally , we consider the possibility of tuning the stt efficiency by the applied bias voltage in the tunnel junction .
figure [ fig6 ] shows in - plane and out - of - plane stt efficiency maps and a longitudinal spin current ratio map 5 above the top surface position of the fe / w(110 ) surface as the function of the relative tip - sample magnetization orientation and the bias voltage .
we find that the obtained stt efficiency maps are qualitatively similar above the top and hollow surface sites with maximal values found close to @xmath108 and @xmath109 , which is also true for the studied @xmath89 range of [ 3 , 10 ] ( not shown ) .
however , the longitudinal spin current ratio map shows a totally different angular dependence with maximal values close to @xmath110 .
we identify @xmath111 v bias voltage ranges for highly enhanced stt efficiencies . a bias - asymmetry in the favorable torque efficiencies
is evident in fig .
[ fig6](a , b ) as reported in other asymmetric tunnel junctions @xcite .
asymmetric bias dependence of the stt has been reported earlier @xcite .
moreover , we find that the angular dependence of the stt efficiencies is symmetric to @xmath108 at low bias but asymmetric at high bias . a similar behavior has been explained by a resonant transmission channel in ref .
@xcite . in our 3d - wkb model , the bias- and angle - asymmetry
are related to the integrated electronic structures in the bias window , where the transport quantities involving the magnetic pdos ( @xmath112 or @xmath113 ) can have a nonmonotonic behavior and can even change sign depending on the bias voltage , see eq .
( [ eq_current_decomp ] ) .
such a behavior is reported for the spin - polarized part of the charge current in ref .
although the torque - current relationship keeps its general power law dependence , the @xmath114 vacuum - decay ratio will surely be bias dependent and can be different than reported for -0.2 v bias .
we obtain a maximal stt efficiency of 1.369 mev/@xmath115a at the experimentally used bias voltage of -0.2 v @xcite at 5 tip - sample distance , whereas the linear fit in table [ table1 ] gave a maximum value of 1.266 mev/@xmath115a , both at @xmath108 .
the direct calculation of the stt efficiency thus provides a better agreement with the experimentally determined 1.5 mev/@xmath115a than the linear fit , and tuning the tip - sample distance would result in a perfect agreement .
this finding already highlights the problem of the linear fit of the torque - current curves .
this problem is even more striking when placing the stm tip above the hollow surface position , where we obtained 1.5 mev/@xmath115a at @xmath116 based on the linear fit but a maximal value of 0.566 mev/@xmath115a is found at @xmath108 and 5 in the direct stt efficiency calculation .
again , by tuning the tip - sample distance a perfect agreement with the experimental value could be achieved .
nevertheless , our most important finding is that the stt efficiency can be enhanced by about a factor of seven from 1.369 ( at -0.2 v and @xmath108 ) to 9.706 mev/@xmath115a ( at 0.8 v and @xmath117 ) by changing the bias voltage and @xmath13 compared to the experimentally used voltage of -0.2 v @xcite .
all reported characteristics of the tip - sample - distance - dependent ( determined by mostly geometric effect ) and the bias - voltage - dependent ( determined by mostly electronic structure effect ) stt efficiencies are due to the complex interplay of the orbital - dependent tunneling ( determined by the geometry of the stm junction ) and the involved spin - polarized electronic structures of the sample surface and the stm tip .
we anticipate that most of the reported effects can generally be found and utilized in other magnetic stm junctions as well .
moreover , we suggest that by changing the tip material the bias - dependent stt efficiencies could be further enhanced . this can be a topic of future research .
we have introduced a method for a combined calculation of charge and vector spin transport of elastically tunneling electrons in magnetic scanning tunneling microscopy ( stm ) .
the method is based on the three - dimensional wentzel - kramers - brillouin ( 3d - wkb ) approach combined with electronic structure calculations using first principles density functional theory . employing the model ,
we have analyzed the stm contrast inversion of the charge current above the fe / w(110 ) surface depending on the bias voltage , tip - sample distance and relative magnetization orientation between the sample and an iron tip .
we have demonstrated the tunability of the atomic contrast inversion in stm images depending on the contributing spin channels .
for the spin transfer torque ( stt ) vector we have found that its in - plane component is generally larger than the out - of - plane component , and we identified a longitudinal spin current component , which , however , does not contribute to the torque . we have shown that the stt vectors depend on the lateral and vertical stm tip positions even above a clean ( defect - free ) planar ferromagnetic surface .
our results suggest that the torque - current relationship in magnetic stm junctions follows the power law rather than a linear function .
consequently , we have shown that the ratio between the stt and the spin - polarized charge current is not constant , and more importantly , it can be tuned by the bias voltage , tip - sample distance and magnetization rotation .
we found that the stt efficiency can be enhanced by about a factor of seven by selecting a proper bias voltage .
thus , we demonstrated the possible enhancement of the stt efficiency in magnetic stm junctions , which can be exploited in technological applications .
we discussed our results in view of the indirect measurement of the stt above the fe / w(110 ) surface reported by krause _
our presented work is expected to inspire future research on high resolution vector spin transport characteristics in magnetic stm junctions .
financial support of the saspro fellowship of the slovak academy of sciences ( project no .
1239/02/01 ) , the hungarian state etvs fellowship and the national research , development and innovation office of hungary under project no .
k115575 is gratefully acknowledged .
here , we prove the formal equivalence of the in - plane torque vectors in eq .
( [ eq_current_decomp ] ) with eq .
( 6 ) of ref .
@xcite . according to theodonis _
_ , the in - plane torque vector acting on @xmath118 is directly obtained from the longitudinal spin currents , and can be written using our notation as @xmath119 using the longitudinal spin currents from eq .
( [ eq_current_decomp ] ) , @xmath120 in the two possible transport directions ( tip @xmath42 sample at @xmath70 and sample @xmath42 tip at @xmath71 ) and the identity , @xmath121 , the in - plane torque vectors following eq .
( [ eq_theodonis ] ) can be obtained as @xmath122,\nonumber\\ \mathbf{t}^{\parallel , s\rightarrow t}=\frac{t^{long - t}(180^{\circ})-t^{long - t}(0^{\circ})}{2}\mathbf{e}_t\times(\mathbf{e}_s\times\mathbf{e}_t)&\propto&-m_s n_t \left[\mathbf{e}_s-\mathbf{e}_t\cos(\phi)\right].\end{aligned}\ ] ] these correspond to the in - plane torque formulas in eq .
( [ eq_current_decomp ] ) : @xmath123,\nonumber\\ \mathbf{t}^{t\parallel}&\propto&m_s n_t \left[\mathbf{e}_s-\mathbf{e}_t\cos(\phi)\right].\end{aligned}\ ] ]
|
we introduce a method for a combined calculation of charge and vector spin transport of elastically tunneling electrons in magnetic scanning tunneling microscopy ( stm ) .
the method is based on the three - dimensional wentzel - kramers - brillouin ( 3d - wkb ) approach combined with electronic structure calculations using first principles density functional theory . as an application
, we analyze the stm contrast inversion of the charge current above the fe / w(110 ) surface depending on the bias voltage , tip - sample distance and relative magnetization orientation between the sample and an iron tip . for the spin transfer torque ( stt ) vector
we find that its in - plane component is generally larger than the out - of - plane component , and we identify a longitudinal spin current component , which , however , does not contribute to the torque .
our results suggest that the torque - current relationship in magnetic stm junctions follows the power law rather than a linear function .
consequently , we show that the ratio between the stt and the spin - polarized charge current is not constant , and more importantly , it can be tuned by the bias voltage , tip - sample distance and magnetization rotation .
we find that the stt efficiency can be enhanced by about a factor of seven by selecting a proper bias voltage .
thus , we demonstrate the possible enhancement of the stt efficiency in magnetic stm junctions , which can be exploited in technological applications . we discuss our results in view of the indirect measurement of the stt above the fe / w(110 ) surface reported by krause _
et al .
_ in phys .
rev .
lett .
107 , 186601 ( 2011 ) .
|
the principle of factorization underlies all theoretical predictions for hadronic processes .
simply put , factorization is the statement that short and long distances contributions to physical processes can be separated , up to corrections suppressed by powers of the relevant large scale in the process .
the predictive power gained from this result stems from the fact that the incalculable long distance effects are universal , defined in an unambiguous way in terms of matrix elements . as a consequence
, the non - perturbative long distance effects can be extracted in one process and then used in another . in general , proving factorization is a difficult task @xcite .
the proof of factorization in drell yan processes , for instance , took several years to sort out @xcite ( for reviews on factorization see @xcite ) .
indeed , there are still some processes such as @xmath7 where a proof of factorization only exists at one - loop @xcite . given that we would like to retain our predictive power over the largest possible range of energies
, we are compelled to understand power corrections to the factorized rates .
these corrections are not necessarily universal , and as such , the relevant size of the power corrections are process dependent . in processes
for which there exists an operator product expansion ( ope ) , there is a systematic way in which to include power corrections .
however , for a majority of observables we do not have an ope at our disposal , and the nature of the power correction is not always known .
for instance , in the case of shape variables there is still some on going discussion about the form of subleading corrections @xcite .
the purpose of this paper is to show that an effective theory framework can be used to simplify proofs of factorization and describe processes with an operator formalism . to do this
we extend the soft - collinear effective theory ( scet ) developed in refs .
@xcite , to high energy processes .
it should be emphasized that there are several other useful advantages in using an effective field theory ( eft ) .
for instance , the eft makes any symmetries which emerge in the @xmath8 limit manifest in the lagrangian and operators , and allow statements to be made to all orders in perturbation theory .
the calculation of hard coefficients reduces to simple matching calculations , where subtracting the eft graphs automatically removes all infrared divergences from the qcd calculation .
perhaps most importantly , it provides a framework for systematically investigating power corrections .
finally , the eft framework allows standard renormalization group techniques to be used for the resummation of logarithms that are often necessary in calculating rates for certain high energy scattering events @xcite .
the factorization formulae that we prove in this paper are not new , but serve to illustrate our approach in familiar settings .
the results are valid to all orders in @xmath9 and leading order in the power expansion .
the simplicity of our approach lies in the fact that factorization occurs at the level of the scet lagrangian and operators , and is facilitated by gauge symmetry in the eft .
this provides the advantage that our proofs do not rely on making use of ward identities and induction , or on specifying a particular gauge .
furthermore , it becomes rather simple to derive factorization formulae for a myriad of processes , since many results are universal .
the examples given here serve to illustrate these simplifications .
developments on the issues of power corrections and resummations are left to future publications . in section [ section_effective ]
we review the construction of the scet .
the formalism developed in refs .
@xcite is extended to include two types of collinear particles moving in opposite directions in section [ section_nnbar ] , and factorization for @xmath10 to two collinear states is discussed as an example . in section [ section_melt ]
we define the non - perturbative matrix elements such as the parton distribution functions that will be needed for the processes presented in the paper , and in section [ section_symm ] we discuss some of the symmetries in scet that may be used to place restrictions on matrix elements . in the remaining sections we give various examples on how factorization theorems emerge in the effective theory language . in section [ section_exclusive ]
we prove factorization theorems for two exclusive processes , namely the @xmath0-@xmath1 form factor , and meson form factors ( @xmath3 ) for arbitrary spin and isospin structure . in section [ section_inclusive ]
two inclusive processes are treated , namely dis ( @xmath4 ) and drell - yan ( @xmath11 ) , and we also give results for deeply virtual compton scattering ( @xmath12 ) . in these processes
we include all leading power contributions in the factorization proofs ( even if the operators are only matched onto at higher orders in perturbation theory such as for the gluon distribution functions ) .
our conclusions are given in section [ section_conclusion ] . in appendix [ app_fact ]
we show how auxiliary fields can be used to prove the simultaneous factorization of soft fields from collinear fields for particles in back - to - back directions .
effective field theories provide a simple and elegant way of organizing physics in processes containing widely disparate energy scales . in constructing an eft ,
some degrees of freedom are eliminated , and the remaining degrees of freedom must reproduce all the infrared physics of the full theory in the domain where the eft is valid . the eft is organized by an expansion in @xmath13 , defined as the ratio of small to large energy scales . as a useful guideline
the following steps are used to identify the infrared degrees of freedom : 1 ) determine the relevant scales in a problem from the size of the momenta and masses of all particles that can make up the initial and final states , 2 ) construct all momenta from these scales whose components correspond to propagating degrees of freedom , and which have offshellness less than the large scale , i.e. @xmath14 .
effective theory fields are then constructed for each unique set of these momenta .
we will be interested in an eft with particles of energy @xmath15 much greater than their mass .
the dynamics of these particles can be described by constructing a soft - collinear effective theory ( scet ) .
this theory is organized as an expansion in powers of @xmath16 , and offshell fluctuations with @xmath17 are integrated out . in section [ section_scet ]
we begin by describing this procedure and comparing the construction to other eft s .
we then give a brief review of the soft - collinear effective theory developed in refs .
we do not attempt to give a comprehensive treatment , but instead emphasize the main results and refer the reader to the literature for details . in section [ section_nnbar ]
we extend the formulation of scet to describe processes with collinear particles moving in back to back directions , and prove the factorization formula for @xmath10 to two collinear states as an example . in section [ section_melt ]
we define the non - perturbative matrix elements that are needed for our examples , then in section [ section_symm ] we discuss some of the symmetry properties of collinear fields and currents . in the standard construction of an eft
one removes the short distance scales and massive fields by integrating them out one at a time .
a classical example is integrating out the @xmath18 boson to obtain the effective electroweak hamiltonian with 4-fermion operators .
however , in some situations we are interested in integrating out large momentum fluctuations without fully removing the corresponding field .
the simplest example of this is heavy quark effective theory ( hqet ) @xcite , which is constructed to describe the low energy properties of mesons with a heavy quark . here
the heavy anti - quarks are integrated out and only heavy quarks with fluctuations close to their mass - shell are retained .
this is accomplished by removing fluctuations of order the heavy quark mass @xmath19 with a field redefinition@xcite @xmath20 where @xmath21 is the heavy quark velocity and @xmath22 is the field in the eft . while @xmath23 , the effective field has @xmath24 , indicating that it no longer describes short - distance fluctuations about the perturbative scale @xmath19 .
instead these effects are encoded in calculable wilson coefficients .
the hqet degrees of freedom with offshellness @xmath25 are the heavy quarks , soft gluons , and soft quarks .
similarly , for collinear particles with energy @xmath26 , one needs to remove momentum fluctuations @xmath27 while retaining effective theory fields to describe smaller momenta .
however , unlike heavy quarks the collinear particles have two low energy scales .
consider the light - cone momenta , @xmath28 and @xmath29 where @xmath30 and @xmath31 . here
@xmath32 parameterizes a light - cone direction close to that of the collinear particle and @xmath33 the opposite direction ( eg . for motion in the @xmath34 direction @xmath35 and @xmath36 ) . for a particle of mass @xmath37
, we have @xmath38 , and a small parameter @xmath39 .
the scaling of the @xmath40 component is then fixed by the equations of motion @xmath41 , so that @xmath42 .
the appearance of two small scales , @xmath43 , is similar to the situation in non - relativistic qcd ( nrqcd ) , which is an eft for systems of two heavy quarks with an expansion in their relative velocity @xmath44 . in a non - relativistic bound state
the momentum of a heavy quark is @xmath45 , but the equations of motion @xmath46 make the energy @xmath47 , giving scales @xmath48 .
the two low energy scales can be distinguished by following eq .
( [ fd1 ] ) with a further field redefinition @xcite @xmath49 , so that derivatives on @xmath50 only pick out the @xmath51 scale .
the on - shell degrees of freedom are then the heavy quarks , soft quarks and gluons with @xmath52 , and ultrasoft quarks and gluons with @xmath53 . * * + for collinear particles the analogous field redefinitions are @xcite @xmath54 where the collinear fields @xmath55 are labelled by light - cone vectors @xmath32 and label momentum @xmath56 . here
@xmath56 contains the @xmath57 and @xmath58 momenta so that @xmath59 .
the field @xmath60 can either be a quark or gluon field .
similar to @xmath50 , the missing @xmath61 fluctuations are described by wilson coefficients and the @xmath62 labels simplify the power counting by distinguishing the @xmath63 and @xmath64 scales .
now @xmath65 so collinear particles and antiparticles are contained in the same effective theory field , but have momentum labels with the opposite sign . in the large energy limit
the four component fermion spinors contain two large and two small components .
one therefore defines collinear quark fields @xmath66 which only retain the large components for motion in the @xmath32 direction and satisfies @xmath67 . for these fields
@xmath68 destroy / create the particles / antiparticles with large momentum @xmath69 @xcite .
for collinear gluons @xmath70 , and @xmath71/@xmath72 destroy / create gluons with @xmath73 . for simplicity we will ignore quark masses and only consider massless @xmath74 and @xmath75 quarks . for the processes considered here scet then requires three types of degrees of freedom : collinear , soft , and ultrasoft ( usoft ) fields .
these are distinguished by the scaling of the light cone components ( @xmath40,@xmath76,@xmath77 ) of their momenta : @xmath78 for collinear modes in the @xmath32 direction ( @xmath79 , @xmath66 ) , @xmath80 for the soft modes ( @xmath81 , @xmath82 ) , and @xmath83 for the usoft modes ( @xmath84 , @xmath85 ) .
the soft modes are labelled by their order @xmath63 momenta , so @xmath81 and @xmath82 are essentially just momentum space fields .
the usoft fields have no labels and depend only on the coordinate @xmath86 .
the fields are assigned a scaling with @xmath13 to make the action for their kinetic terms order @xmath87 @xcite .
for instance @xmath88 , @xmath89 , @xmath90 , and @xmath91 . at leading order only order @xmath87 vertices are necessary to correctly account for all order @xmath87 feynman diagrams
. in hqet only external currents with momenta of order @xmath92 can change the label @xmath21 .
thus the lagrangian has a superselection rule forbidding changes in the four - velocity of the heavy quark @xcite . in nrqcd the @xmath21 labels are also conserved , but the smaller momentum labels @xmath93 are changed by operators in the effective theory such as the coulomb potential .
a novel feature of scet is that interactions in the leading action can change both the large and small parts of the momentum labels @xmath94 .
however , only external currents can change the direction @xmath32 of a collinear particle , so this label is conserved .
thus , for each distinct direction @xmath32 a separate set of collinear fields are needed . in the remainder of this section
we will restrict ourselves to collinear particles with a single @xmath32 .
we will generalize the discussion to the case of two back - to - back directions and discuss the factorization of collinear particles with different @xmath32 s in section [ section_nnbar ] . since in scet interactions can change the order @xmath15 label momenta it turns out to be very useful to introduce a label operator , @xmath95 @xcite , for which the collinear fields satisfy @xmath96 .
more generally , @xmath95 acts on a product of labelled fields as @xmath97 so conjugate field labels come with a minus sign .
the operator @xmath98 acts to the right , while the conjugate operator @xmath99 acts to the left . as explained in ref .
@xcite the label operator allows all large phases to be moved to the front of operators with a factor @xmath100 .
this phase and the label sums can then be suppressed if we impose that interactions conserve label momenta and that the momentum indices on fields are implicitly summed over . basically , for labels @xmath56 and @xmath101 and residual momenta @xmath102 and @xmath103 @xmath104 so that the label and residual momenta are individually conserved .
( although technically the label momenta are discrete we abuse notation and use @xmath105 rather than @xmath106 because it makes the subscripts easier to read . ) for convenience we define the operator @xmath107 to pick out only the order @xmath87 labels on collinear fields , and the operator @xmath108 to pick out only the order @xmath13 labels . for the matrix element of any collinear operator @xmath109 , momentum conservation constrains the sum of field labels @xcite , giving @xmath110 \big| m_{n , p_2}\big\rangle & = & f({{\bar
n}}{\!\cdot\!}(p_2\!-\!p_1))\ \big\langle m_{n , p_1 } \big| { \cal o } \big| m_{n , p_2}\big\rangle \,,\end{aligned}\ ] ] for any function @xmath111 . for a single @xmath32 the lagrangian can be broken up into three sectors : collinear , usoft , and soft .
we therefore write @xmath112 + { \cal l}_{us}[q_{us},a^\mu_{us } ] + { \cal l}_{s}[q_{s , p}\ , , a^\mu_{s , q } ] \ , , \label{sclag}\end{aligned}\ ] ] where we have made the field content of each sector explicit .
we will discuss each of these terms separately .
* * + as explained in detail in ref .
@xcite , gauge invariance in scet restricts the lagrangian and allowed form of operators . only local gauge transformations
whose action is closed on the effective theory fields need to be considered .
these include collinear , soft , and usoft transformations .
each of these vary over different distance scales , with collinear gauge transformations satisfying @xmath113 , soft satisfying @xmath114 , and usoft transformations with @xmath115 .
all particles transform under @xmath116 and usoft gluons act like background fields for collinear particles .
invariance under @xmath117 requires a collinear wilson line built out of the order @xmath87 gluon fields @xcite @xmath118 \,.\end{aligned}\ ] ] here the operator @xmath107 acts only inside the square brackets , the @xmath32 on @xmath119 refers to the direction of the collinear quanta , and @xmath119 is local with respect to @xmath86 ( corresponding to the residual momenta ) .
taking the fourier transform of @xmath120 with respect to @xmath121 gives the more familiar path - ordered wilson line @xmath122 $ ] . under a collinear gauge transformation @xmath119 transforms as @xmath123 .
an invariant under collinear gauge transformations can therefore be formed by combining a collinear fermion @xmath66 and the wilson line @xmath119 in the form @xmath124 this combination still transforms under an usoft gauge transformation , @xmath125 .
we will often suppress the @xmath86 dependence of the combination @xmath126 .
integrating out hard fluctuations gives wilson coefficients in the effective theory that are functions of the large @xmath127 collinear momenta , @xmath128 .
however , collinear gauge invariance restricts these coefficients to only depend on the linear combination of momenta picked out by the order @xmath87 operator @xmath107 @xcite . in general
the wilson coefficients are then functions @xmath129 which must be inserted between gauge invariant products of collinear fields . in general the wilson coefficients also depend on the large momentum scales in a process such as @xmath15 and the renormalization scale @xmath130 . to construct the collinear lagrangian one can match full qcd onto operators with collinear fields that are invariant under usoft and collinear gauge transformations .
the collinear lagrangian at order @xmath87 is @xcite @xmath131^{\,2 } \bigg\ } + { \cal l}_c^{g.f.}\,,\end{aligned}\ ] ] where @xmath132 are gauge fixing terms , @xmath133 , and @xmath134 since usoft gluons act as background fields in the collinear gauge transformation the couplings , @xmath135 , for both types of gluons must be identical . * * + the usoft and soft lagrangians for gluons and massless quarks are the same as those in qcd . from eq .
( [ sclag ] ) we see that collinear quarks and gluons interact with usoft gluons , however at order @xmath87 only the @xmath136 component appears in eq .
( [ lc ] ) . in order to prove factorization formulae it is essential to disentangle the collinear and usoft modes .
this can be done by introducing an usoft wilson line @xmath137 where the subscript @xmath32 on @xmath138 labels the direction of the wilson line ( we emphasize that this is different from the meaning of the subscript on @xmath119 in eq .
( [ momw ] ) ) .
an usoft gauge transformation takes @xmath139 . in ref .
@xcite it was shown that the field redefinitions @xmath140 imply @xmath141 and decouple the usoft gluons from the collinear particles in the leading order lagrangian @xmath142 = { \cal l}_{c , n}[\xi_{n , p}^{(0)},a_{n , q}^{(0)\mu},0 ] \,.\end{aligned}\ ] ] thus , the new collinear fields with superscript @xmath143 no longer interact with usoft gluons or transform under an usoft gauge transformation .
since the field redefinitions do not change physical @xmath144 matrix elements , the new fields give an equally valid parameterization of the collinear modes .
the leading scet lagrangian therefore factors into separate collinear and usoft sectors .
this alone does not guarantee factorization in operators and currents , since after the field redefinition these operators may still contain both usoft and collinear fields .
however , the field redefinition makes factorization transparent since identities such as @xmath145 may be applied directly to the operators .
this will become clear in the examples in sections [ section_exclusive ] and [ section_inclusive ] .
the coupling of soft gluons to collinear particles differs from the usoft - collinear interactions .
interactions of a soft gluon with a collinear particle results in a particle with momentum @xmath146 , so soft gluons can not appear in the collinear lagrangian .
these offshell particles have @xmath147 and since @xmath148 these offshell quarks and gluons can be integrated out .
at leading order in @xmath13 it was shown in ref .
@xcite that in operators with collinear fields this simply builds up factors of a soft wilson line @xmath149 involving the @xmath150 component of the soft gluon field , @xmath151 \,.\end{aligned}\ ] ] the factors of @xmath149 appear outside gauge invariant products of collinear fields , and their location is restricted by soft gauge invariance .
in this section we extend scet to include the possibility of collinear fields moving in different light - cone directions : @xmath152 , @xmath153 , @xmath154 .
these directions can be considered to be distinct provided that @xmath155 for @xmath156 .
this follows from the fact that if @xmath157 then the directions @xmath152 and @xmath153 are too close to be distinguished .
for example , a momentum @xmath158 can be considered to be collinear in the @xmath152 direction if @xmath159 , since this is the correct scaling for the small momentum component of an @xmath152-collinear particle . for simplicity we will only consider the case of back - to - back jets corresponding to collinear particles moving in the @xmath32 and @xmath33 directions .
these are clearly distinct since @xmath160 .
collinear particles in the @xmath33 direction have @xmath161 momenta @xmath162 , and the @xmath163 and @xmath164 momenta appear as labels on the corresponding fields : @xmath165 and @xmath166 .
emission of a collinear particle moving in the @xmath32 direction from a collinear particle in the @xmath33 direction results in a particle with momentum @xmath167 and offshellness @xmath168 .
these offshell modes are integrated out to construct the scet , so collinear modes in the @xmath32 direction do not directly couple to collinear modes in the @xmath33 direction .
a distinct set of collinear gauge transformations is associated with each of @xmath32 and @xmath33 , and fields in one direction do not transform under the gauge symmetry associated with the opposite direction .
two wilson lines @xmath169 and @xmath170 are necessary ( defined as in eq .
( [ momw ] ) ) , and they appear in a way that makes collinear operators gauge invariant .
for instance the combinations @xmath171 are invariant under collinear gauge transformations in the @xmath32 and @xmath33 directions , respectively .
we also require two types of label operators , @xmath107 as before , and an operator @xmath172 to pick out @xmath173 labels that are order @xmath87 .
thus , @xmath107 and @xmath172 act only on the @xmath32 and @xmath33 collinear fields respectively .
( the label operator @xmath108 still picks out order @xmath13 momentum components and therefore acts on both @xmath32 and @xmath33 fields . ) with two collinear directions , decoupling usoft gluons requires introducing both @xmath138 and @xmath174 wilson lines , defined as in eq .
( [ ydef ] ) , but along the @xmath32 or @xmath33 paths respectively . finally , integrating out @xmath175 fluctuations at leading order
induces both @xmath149 and @xmath176 soft wilson lines defined analogous to eq .
( [ sdef ] ) .
this is discussed in greater detail in appendix [ app_fact ] where we show explicitly to all orders in @xmath177 that integrating out the @xmath178 fluctuations causes @xmath179 relations for operators with collinear gluon fields are also derived in appendix [ app_fact ] .
note that we have not included `` glauber gluons '' with momenta @xmath180 , which are kinematically allowed in @xmath181-channel coulomb exchange between @xmath32 and @xmath33 collinear quarks . in determining the relevant degrees of freedom we have assumed that glauber gluons are not necessary to describe the infrared for the processes considered in this paper .
intuitively , this can be seen from the fact these gluons are instantaneous in both time and longitudinal separation , and only could contribute when the @xmath32 and @xmath33 jets overlap for a duration of order @xmath182 in a space - time diagram . in processes with a hard interaction
the overlap scale is always much shorter than this ( however this need not be the case in processes such as forward scattering ) . for the drell - yan process
more quantitative arguments can be found in refs .
@xcite . at order @xmath87
it is not possible to construct a gauge invariant kinetic lagrangian with terms that involve both @xmath32 and @xmath33 fields .
thus , the @xmath32 and @xmath33 collinear modes are described by independent lagrangians ( however @xmath32 and @xmath33 modes may still both appear in an external operator ) .
the collinear sector of the scet lagrangian is therefore @xmath183 + { \cal l}_{c,{{\bar n}}}[\xi_{{{\bar n}},p},a_{{{\bar n}},q}^\mu,{{\bar n}}{\!\cdot\!}a_{us } ] \,.\end{aligned}\ ] ] making the field redefinitions @xmath184 gives @xmath141 , @xmath185 , and usoft degrees of freedom once again decouple from the collinear modes since @xmath183 \!+\ ! { \cal l}_{c,{{\bar n}}}[\xi_{{{\bar n}},p},a_{{{\bar n}},q}^\mu,{{\bar n}}{\!\cdot\!}a_{us } ] \!=\ ! { \cal l}_{c , n}[\xi_{n , p}^{(0)},a_{n , q}^{(0)\mu},0 ] \!+\ ! { \cal l}_{c,{{\bar n}}}[\xi_{{{\bar n}},p}^{(0)},a_{{{\bar n}},q}^{(0)\mu},0 ] .
\ ] ] thus , usoft gluons are removed from the collinear lagrangian at the expense of inducing @xmath138 and @xmath174 factors in operators with collinear fields . in certain cases
the identities @xmath145 and @xmath186 can be used in these operators to cancel usoft gluon interactions .
perturbatively these cancellations would occur by adding an infinite set of feynman diagrams . to see in more detail how this works consider the simple example of the @xmath10-production of back - to - back collinear states @xmath187 and @xmath188 .
the full theory current @xmath189 matches onto an effective theory operator @xmath190 .
naively one might guess that the scet operator mediating this process is @xmath191 however , this operator is not invariant under the collinear gauge transformations @xmath192 and @xmath193 , so the process is instead mediated by the invariant operator @xmath194 a hard matching coefficient @xmath195 can be inserted in any location in the operator that does not break apart the gauge invariant combinations of fields in eq .
( [ invariant2 ] ) .
the operators @xmath107 and @xmath172 in the coefficient only pick out momenta that are order @xmath87 in the power counting .
thus , @xmath172 does not act on fields in the @xmath32 direction and @xmath107 does not act on fields in the @xmath33 direction , and the most general result is @xmath196 next , we integrate out the offshell @xmath178 fluctuations which induces additional soft wilson lines in the operator .
this is discussed in detail in appendix [ app_fact ] and from eq .
( [ sadd ] ) gives @xmath197 note that @xmath107 and @xmath172 do not act on the fields in the soft wilson lines since soft gluons carry only order @xmath13 momenta .
finally , we can make the usoft gluon couplings explicit by switching to the @xmath143 fields using eq .
( [ yfd2 ] ) : @xmath198 this operator is manifestly invariant under collinear gauge transformation in the @xmath32 and @xmath33 direction , as well as under soft and usoft gauge transformations . to separate
the short distance wilson coefficient from the long - distance operator one introduces convolution variables @xmath121 and @xmath199 to give @xmath200 \ , .
{ \nonumber}\end{aligned}\ ] ] the function @xmath201 contains all the short distance physics and is determined by matching the full theory onto this effective theory operator .
@xmath202 contains all the infrared long - distance qcd contributions at leading order in @xmath13 .
now consider the matrix element of the production current between @xmath203 and the vacuum .
taking the @xmath10 to have large time - like momentum @xmath204 ( and zero residual momentum ) we have @xmath205 in the first step the conservation of the large label momentum @xmath206 was made implicit in the matrix element ( c.f .
( [ mcons ] ) ) .
since we are in the center - of - mass frame the @xmath187 has large momentum @xmath207 , and the @xmath188 has momentum @xmath208 . now using translation invariance
, we see that the remaining @xmath86 integral forces the @xmath209 state to have zero residual momentum . using eq .
( [ onnf ] ) this matrix element is equal to @xmath210 where the functions @xmath211 , @xmath212 , @xmath213 , and @xmath144 also depend on the renormalization point @xmath130 . here
we have used the fact that both @xmath187 and @xmath188 can be described entirely by collinear particles in the @xmath32 and @xmath33 directions respectively . since the lagrangians for the collinear ,
soft , and usoft fields are factorized the remaining matrix element splits into distinct matrix elements for each class of modes .
these matrix elements are @xmath214 ( and are matrices whose color , spin , and flavor indices are suppressed ) .
note that @xmath212 , @xmath213 , and @xmath215 are explicitly invariant under the collinear , soft , and usoft gauge transformations @xcite of scet , but still transform globally under a color rotation . now using the momentum conservation identity in eq .
( [ mc ] ) , the large momentum of the @xmath187 and @xmath188 states set @xmath216 and @xmath217 .
label conservation also implies that the total perpendicular momentum of each of @xmath212 , @xmath213 , and @xmath215 is zero .
the sum over @xmath121 and @xmath199 can then be performed to give the final factorized form @xmath218 although rather idealized , the above example illustrates the main steps needed to derive a factorization formula .
taking @xmath187 and @xmath188 to be single quark states the result in eq .
( [ jjrslt ] ) also agrees with the factorization formula for the onshell production form factor for @xmath219 @xcite . and @xmath149 wilson lines in @xmath220 .
for instance if one chooses @xmath221 then the usoft gluons give scaleless loop integrals and can be dropped , so that @xmath222 .
if instead one chooses @xmath223 then the soft gluons give scaleless loop integrals ( they simply act to pull - up the ultraviolet divergences in the usoft integrals to the hard scale @xcite ) , so the soft wilson lines can be suppressed .
this is why one only finds @xmath224 for this operator in the literature . for typical regulator choices the other gluons
are simply not required to reproduce the infrared structure of the full theory result . ] in the above example the factors of @xmath144 and @xmath225 in the operator in eq .
( [ exop ] ) do not cancel . in the examples we will consider in sections [ section_exclusive ] and [ section_inclusive ] there are several operators at leading order
, however the factors of @xmath144 and @xmath225 cancel in observable matrix elements .
the collinear matrix elements of long - distance operators , such as those in eq .
( [ jjsy ] ) , are the ones that have interpretations as structure functions or wave functions . in the next section we give the operator definitions for these functions that will be needed in the remainder of the paper .
predictions for hadronic processes depend on universal matrix elements that are not computable in perturbation theory . for exclusive processes these include light - cone wavefunctions and form factors , while for inclusive processes they include parton distribution functions and fragmentation functions . in this section
we define matrix elements in scet that are needed for our examples .
all the collinear operators considered here decouple from usoft gluons since they are local in the residual coordinate @xmath86 and because @xmath226 .
thus @xmath227 and expressions with and without the @xmath143 superscript are equal . for convenience we will write the collinear fields without the superscript in the remainder of this section .
consider first the light - cone wavefunctions .
for the pion isotriplet @xmath228 , the wavefunction @xmath229 is conventionally defined by @xcite @xmath230}\ , \phi_\pi(\mu , z),\end{aligned}\ ] ] here @xmath231 and the qcd field @xmath232 denotes the isospin doublet @xmath233 .
the coordinates satisfy @xmath234 , which ensures that the path from @xmath235 to @xmath236 is along the light - cone , and @xmath237 is a wilson line along this path . in scet we require the matrix elements of highly energetic pions , which therefore have collinear constituents . boosting the matrix element in eq .
( [ piwfn_rest ] ) and letting @xmath238 , @xmath239 we have @xmath240}\ , \phi_\pi(\mu , z),\end{aligned}\ ] ] where @xmath241 , and @xmath242 is a collinear field with position space label @xmath243 . for our purposes it is more useful to use the operator with momentum space labels @xcite @xmath244\ , \phi_\pi(\mu , z)\ , , { \nonumber}\end{aligned}\ ] ] where @xmath245 . in eq .
( [ pimom ] ) the delta function fixes @xmath121 to the sum of labels picked out by the @xmath246 operator .
the combination picked out by @xmath247 is equivalent to ( @xmath248 ) acting on the entire operator , and using eq .
( [ mc ] ) is fixed to the @xmath249 momentum of the pion state . in some situations
it is convenient to have delta functions which fix the labels of both @xmath250 and @xmath251 . in this case
a useful field is @xmath252 \,.\end{aligned}\ ] ] here @xmath253 is the flavor index and will be omitted if the flavor doublet field is implied .
note that unlike the @xmath56 in @xmath66 the label @xmath121 on @xmath254 is not summed over .
a matrix element with @xmath255 fields is related to a matrix element like the one in eq .
( [ pimom ] ) through @xmath256 where @xmath257 and @xmath258 .
thus , with the @xmath259 notation the momentum conserving delta functions become explicit .
the factor of two appears from treating the @xmath121 s as continuous variables , and in the final results cancels with a factor of @xmath260 from a jacobian . for inclusive processes such as dis
it is the proton parton distribution functions for quarks of flavor @xmath253 , @xmath261 , antiquarks @xmath262 , and gluons , @xmath263 which are needed .
the standard coordinate space definitions @xcite are ( @xmath238 ) @xmath264 and @xmath265 . here
@xmath266 is the gluon field strength , @xmath267 and @xmath268 are path - ordered wilson lines in the fundamental and adjoint representations , and @xmath269 is the proton state with momentum @xmath56 . in scet
these distribution functions can be defined by the matrix elements of collinear fields with collinear proton states @xmath270 \big| p_{n } \big\rangle \ ! & = & -\frac{\omega_+ \,{{\bar n}}{\!\cdot\!}p}{2 } \int_0 ^ 1\!\!\ !
dz \,\delta(\omega_-)\delta(\omega_+ \!-\!2 z \ , { { \bar n}}{\!\cdot\!}p ) f_{g / p}(z ) { \nonumber},\end{aligned}\ ] ] where @xmath271 , and @xmath272 with the collinear gauge invariant field strength @xmath273 w \big ] \,.\end{aligned}\ ] ] both operators in eq .
( [ pdfqg ] ) are order @xmath274 since @xmath275 .
note that the matrix element of a single operator ( @xmath276 ) contains both the quark and antiquark distributions .
this is due to eq .
( [ phi+- ] ) , from which we see that for @xmath277 ( @xmath278 ) this operator reduces to the number operator for collinear quarks ( antiquarks ) with momentum @xmath121 .
processes other than dis sometimes depend on more complicated distribution functions . in deeply virtual compton scattering ( dvcs ) we will need to parameterize the matrix element of an operator between proton states with different momenta . in terms of qcd fields
@xmath232 the nonforward parton distribution function ( nfpdf ) defined by radyushkin in eq.(4.1 ) of ref .
@xcite are ( up to a trivial translation ) @xmath279{\nonumber}\\ & & \qquad\ + \ , h(\sigma^\prime,\sigma ) \int^1_0 dz \big [ e^{i{{\bar n}}\cdot p(2z-\zeta)y}\ : { \cal k}^{(i)}_\zeta(z;t ) - e^{-i{{\bar n}}\cdot p(2z-\zeta)y}\ : \overline{{\cal k}}^{(i)}_\zeta(z;t ) \big ] \,,{\nonumber}\end{aligned}\ ] ] where @xmath280 , and @xmath281 . here
@xmath282 and @xmath283 are matrix elements which respectively preserve or flip the proton spin .
they are defined in terms of the proton spinors @xmath284}\ : u(p,\sigma ) \ , , \end{aligned}\ ] ] where @xmath285 is the proton mass .
the nfpdf for gluons is similarly given by @xcite @xmath286 { \cal f}^{g}_\zeta(z;t ) \nonumber \\ & & \qquad\qquad
\ + \frac{{{\bar n}}{\!\cdot\!}p}{2}\ : h(\sigma^\prime,\sigma ) \int^1_0 dz \left [ e^{i{{\bar n}}\cdot p(2z-\zeta)y } + e^{-i{{\bar n}}\cdot p(2z-\zeta)y}\right ] { \cal k}^{g}_\zeta(z;t ) \ , .
{ \nonumber}\end{aligned}\ ] ] in scet the definition of the nfpdfs in terms of collinear fields are @xmath287 \nonumber \\ & & \qquad\quad + h(\sigma^\prime,\sigma ) \big [ \delta(\omega_+ \!-\ ! ( 2z\!-\!\zeta ) \ , { { \bar n}}{\!\cdot\!}p)\ , { \cal k}^{(i)}_\zeta(z;t)- \delta(\omega_+ \!+\ ! ( 2z\!-\!\zeta ) \ , { { \bar n}}{\!\cdot\!}p)\ , \overline{{\cal k}}^{(i)}_\zeta(z;t ) \big ] \big\ } \ , , { \nonumber}\\ & & \big\langle p_{n}',\sigma ' \big| { \rm tr}\:\big [ b^\mu_{n,\omega } b_\mu^{n,\omega ' } \big ] \big| p_{n},\sigma \big\rangle \
! = -\frac{{{\bar n}}{\!\cdot\!}p}{2}\ : \delta(\omega_-+{{\bar n}}{\!\cdot\!}p\ , \zeta ) \int_0 ^ 1\!\!\
! dz \nonumber \\ & & \qquad\quad \times \big\ { e(\sigma^\prime,\sigma ) \,\delta(\omega_+ \!-\!(2z\!-\!\zeta ) \ , { { \bar n}}{\!\cdot\!}p ) { \cal f}^{g}_\zeta(z;t ) + h(\sigma^\prime,\sigma ) \delta(\omega_+ \!-\!(2z\!-\!\zeta)\,{{\bar n}}{\!\cdot\!}p ) { \cal k}^{g}_\zeta(z , t ) \big\ } \,,{\nonumber}\end{aligned}\ ] ] where the spinors in @xmath288 and @xmath289 are two component effective theory spinors , so that @xmath290 where @xmath291 .
note that for @xmath292 both @xmath293 and @xmath294 . in this limit
the nfpdfs reduce to the standard pdfs : @xmath295 in this section we discuss spin and discrete symmetry constraints on operators involving collinear fields .
the possible spin structures of currents with @xmath66 fields is restricted by the fact that they have only two components , @xmath67 .
the four most general spin structures for currents with two collinear particles moving in the same or opposite directions are @xmath296 other choices for @xmath297 and @xmath298 either vanish between the fields or are related to those in eq .
( [ jspin ] ) .
this result can be expressed in a compact way by the trace formulae @xmath299
-\frac{\bnslash\gamma_5}{8 } { \rm tr}[\nslash\gamma_5 \gamma ] -\frac{\bnslash\gamma_\perp^\mu}{8 } { \rm tr}[\nslash\gamma_\mu^\perp \gamma ] \,,{\nonumber}\\ & & \bar\xi_{{{\bar n}},p ' } \,\gamma \ , \xi_{n , p}\,,\qquad\quad \gamma = \frac{1}{8 } { \rm tr}[\nslash\bnslash \gamma ] + \frac{\gamma_5}{8 } { \rm tr}[\gamma_5\nslash\bnslash \gamma ] + \frac{\gamma_\perp^\mu}{8 } { \rm tr}[\gamma_\mu^\perp\nslash\bnslash \gamma ] \,,\end{aligned}\ ] ] which reduce a general @xmath300 to a linear combination of the terms in eq .
( [ jspin ] ) .
for instance , it implies that @xmath301 , and @xmath302 where @xmath303 .
furthermore , each of the two components of @xmath304 and also @xmath305 can be chosen to be eigenstates of their helicity operators , @xmath306 with eigenvalues @xmath307 .
for these fields @xmath308 is equivalent to the chiral rotation , @xmath309 .
the structures in eq .
( [ jspin ] ) split into two classes depending on whether they conserve or flip the helicity @xmath310 since gluon interactions in qcd preserve helicity , integrating out hard qcd fluctuations results in effective theory operators with the same helicity structure as the original operators at leading order in @xmath13 .
the presence of labels on the effective theory fields makes their transformation properties under the discrete symmetries @xmath211 , @xmath311 , and @xmath312 slightly different than in qcd .
for example under @xmath311 or @xmath312 we have @xmath313 so these transformations relate collinear fields for different directions . under charge
conjugation , parity , and time - reversal the collinear fields transform as @xmath314^t\ , , & & c^{-1 } a^\mu_{n , p}(x ) c = - [ a_{n , p}^\mu(x)]^t \ , , \\[5pt ] & p^{-1 } \xi_{n , p}(x )
p = \gamma_0\ , \xi_{{{\bar n}},\tilde p}(x_p)\ , , & & p^{-1 } a^\mu_{n ,
p}(x ) p = g_{\mu\nu}a_{{{\bar n}},\tilde p}^\nu(x_p ) \,,\\[5pt ] & t^{-1 } \xi_{n , p}(x ) t = { \cal t } \xi_{{{\bar n}},\tilde p}(x_t)\ , , & & t^{-1 } a^\mu_{n , p}(x ) t = g_{\mu\nu } a_{{{\bar n}},\tilde p}^\nu(x_t ) \
, , \end{array}\end{aligned}\ ] ] where @xmath315 and @xmath316 , while if @xmath317 and @xmath318 then @xmath319 , @xmath320 , and @xmath321 . the transformation properties of @xmath119 can be worked out using eq .
( [ cpt ] ) , for instance @xmath322^t$ ] .
the collinear effective lagrangian ( [ lc ] ) is invariant under the transformations in eq .
( [ cpt ] ) ( adding the @xmath323 terms ) .
these symmetries also constrain the form of non - perturbative matrix elements . as an example , for a meson which is an eigenstate of @xmath211 one finds @xmath324 for the isotriplet pion state @xmath325 so combining eq .
( [ psunderc ] ) with eq .
( [ pimom ] ) gives @xmath326 \phi_\pi(x){\nonumber}\\ & & \qquad\qquad = -if_\pi\delta^{ab}\ , { { \bar n}}{\!\cdot\!}p \int_0 ^ 1\!\ !
\mbox{d}x\ , \delta[\omega \!-\ ! ( 2x\!-\!1){{\bar n}}{\!\cdot\!}p ]
\phi_\pi(1\!-\!x ) \,.\end{aligned}\ ] ] together with eq .
( [ pimom ] ) , charge conjugation therefore implies that @xmath327 .
the pion - photon form factor @xmath328 is perhaps the simplest setting for factorization since there is only one hadron in the external state .
the form factor is measurable in single - tagged two photon @xmath329 reactions .
this process involves the scattering of a highly virtual photon and a quark - anti - quark constituent pair off an on - shell photon .
the photon scatters the quark pair away from the incoming photon into a pion , so that @xmath330 . the matrix element for this transition defines the @xmath0-@xmath1 form factor @xmath331 here @xmath332 is the full theory electromagnetic current with isodoublet field @xmath232 and charge matrix @xmath333 , and @xmath334 where @xmath335 is the virtual photon momentum .
it has been shown that the form factor can be written as a one - dimensional convolution of a hard coefficient with the light - cone pion wavefunction @xcite . here
we show how this factorization takes place in the scet . in the breit frame @xmath336 ,
the real photon s momentum is @xmath337 , and the pion is made up of collinear particles with momenta @xmath338 .
the particles exchanged between the two currents in eq .
( [ fdef ] ) have hard momenta and can be integrated out . at leading order in @xmath13
the time ordered product of the two currents in eq .
( [ fdef ] ) matches onto a single operator in the effective theory . for simplicity
we restrict ourselves to the tensor and spin structures that are relevant when the meson is a pion @xmath339\,\gamma c({\bar { \cal p}},{\bar { \cal p}^{\raisebox{0.8mm}{\scriptsize$\dagger$ } } } , \mu)\ , [ w^\dagger \xi_{n , p^\prime } ] \,,\end{aligned}\ ] ] where @xmath66 is an isodoublet collinear quark field , and @xmath340 .
@xmath341 is of dimension two , just like the time - ordered product in eq .
( [ fdef ] ) , and a power of @xmath342 is included to make @xmath343 dimensionless .
the time - ordered product in eq . ( [ fdef ] ) is even under charge conjugation , so the operators in eq .
( [ opig ] ) must also be even .
this implies @xmath344 .
the location of the @xmath18 s in eq .
( [ opig ] ) is fixed by gauge invariance , and @xmath300 contains the spin and flavor structure @xmath345 since the offshellness of the collinear particles in the pion is @xmath346 we can also integrate out offshell modes with @xmath347 which come from soft - collinear interactions . for
the collinear operators @xmath348 , eq . ( [ sadd ] ) implies that factors of the soft wilson line @xmath149 are induced .
however , the location is such that @xmath349 , so no coupling to soft gluons occur at leading order .
the coupling of the collinear fields to usoft gluons can be simplified with the field redefinitions in eq .
( [ def0 ] ) .
as discussed in section [ section_scet ] this moves all couplings into @xmath348 , and using @xmath145 gives @xmath350 \,\gamma\ , c({\bar { \cal p}},{\bar { \cal p}^{\raisebox{0.8mm}{\scriptsize$\dagger$ } } } , \mu)\ , [ w^{(0)\dagger } \xi_{n , p^\prime}^{(0 ) } ] \,.\end{aligned}\ ] ] thus , usoft gluons also decouple . in the breit frame
the pion momentum satisfies @xmath351 , and comparing eq .
( [ fdef ] ) with the scet matrix element @xmath352 , gives @xmath353 defining @xmath354 , the operator @xmath247 is related to @xmath107 acting from the outside on the fields . using eq .
( [ mc ] ) it can therefore be set equal to the momentum label of the state , @xmath355 .
suppressing this dependence we write @xmath356 leaving @xmath357 & = & \frac{2i}{q^3 } \int \!\!{d\omega}\,c_1(\omega,\mu ) \big\langle \pi^0_{n } \big| \bar{\xi}_{n , p}^{(0 ) } w^{(0 ) } \gamma \delta(\omega-{\bar { \cal p}}_+ ) w^{(0)\dagger } \xi^{(0)}_{n , p\prime } \big| 0 \big\rangle \,.\end{aligned}\ ] ] using eq .
( [ pimom ] ) the remaining matrix identity in eq .
( [ pig3 ] ) can be written in terms of the light - cone pion wavefunction @xmath358 this is the final result and is valid to leading order in @xmath13 and all orders in @xmath9 . from eq .
( [ pic ] ) charge conjugation implies that @xmath359 and @xmath360 .
( [ resultpionformfactor ] ) agrees with the brodsky - lepage @xcite result that the form factor can be written as the convolution of a short distance function with the light - cone pion wavefunction .
the scet formalism gives a concise derivation of this result and defines the short distance function in terms of the wilson coefficient of an effective theory operator .
as an illustrative example consider the tree level matching onto @xmath211 illustrated in fig .
[ fig_gamgampi ] .
since the location of the @xmath18 s in @xmath361 are fixed by gauge invariance , @xmath343 can be determined by matching with @xmath362 .
expanding the full theory graphs to leading order gives @xmath363 where we have dropped isosinglet terms , contributions with opposite parity to the pion , as well as those proportional to @xmath364 since @xmath365 .
comparing eq .
( [ 1ab ] ) to eq .
( [ opig ] ) gives @xmath366 so that @xmath367 this result is again in agreement with ref .
@xcite , and the order @xmath368 corrections to this wilson coefficient can be read off from the results in ref .
an identical analysis applies for operators with different spin structures such as the ones contributing to @xmath369 .
another example of an exclusive process which can be treated in the effective theory is the classic case of the electromagnetic pion form factor at large space - like momentum transfer . for generality
we consider in this section the electromagnetic form factors for arbitrary mesons ( pseudoscalar @xmath311 or vector @xmath371 ) , defined as @xmath372\,,{\nonumber}\end{aligned}\ ] ] where @xmath373 . for simplicity
we suppress the dependence of the form factors on the isospin of the two mesons .
we will restrict ourselves in the following to the case of hadrons made up only of @xmath374 quarks .
the electromagnetic current is defined as usual by @xmath375 , with charge matrix @xmath376 , which can be written in terms of the up and down quark charges as @xmath377 .
we will be interested in the asymptotic form of the form factors in the region with @xmath378 , where it can be expanded in a power series in @xmath379 @xcite .
it is convenient to work in the breit frame , where the momentum transfer has the light - cone components @xmath380 . in this frame
the meson momenta are @xmath381 , @xmath382 , so the partons in the incoming / outgoing meson are collinear along the @xmath383/@xmath384 direction .
the electromagnetic current in eq .
( [ piff ] ) is matched in the effective theory onto the most general combination of operators constructed from collinear fields which are compatible with collinear gauge invariance .
operators such as the dimension three current @xmath385 \gamma\ : c(\mu,{{\cal p}}^\dagger,{\bar { \cal p}})\ : [ w^\dagger_n \xi_n ] \,,\end{aligned}\ ] ] can contribute , but only overlap with the asymmetric meson states with one energetic collinear quark and one usoft or soft quark .
often this overlap is referred to as the tail of the wavefunction contribution or the feynman mechanism of generating the form factor @xcite .
there are other operators with significant overlap with more symmetric meson states ( where all the constituents are allowed to be energetic ) .
the leading such operators have the form @xmath386 c(\mu,{\bar { \cal p}},{\bar { \cal p}^{\raisebox{0.8mm}{\scriptsize$\dagger$ } } } , { { \cal p}},{{\cal p}}^\dagger ) \big [ \bar \xi_{{{\bar n}},p_3 } w_{{{\bar n } } } \gamma ' w^\dagger_n \xi_{n , p_4 } \big]\,,\end{aligned}\ ] ] with @xmath211 a dimensionless wilson coefficient . as usual ,
collinear gauge invariance is enforced by the location of the @xmath18 s in eqs .
( [ feyn ] ) and ( [ blmech ] ) .
there is some argument about the relative size of eqs .
( [ feyn ] ) and ( [ blmech ] ) in the literature @xcite .
often it is argued that the tail of the wavefunction is suppressed by an extra @xmath387 @xcite , in which case the operator in eq .
( [ blmech ] ) dominates by two powers of @xmath15 .
an analysis of the tail of wavefunction contributions has not yet been performed in the effective theory framework .
therefore , we choose to ignore the operator in eq .
( [ feyn ] ) , and below only analyze the operator in eq .
( [ blmech ] ) .
we emphasize that we do not claim to have shown that this is justified by the effective theory power counting .
there are two different structures possible for the operator in eq .
( [ blmech ] ) , and we write the general matching for the electromagnetic current as @xmath388 \,,\end{aligned}\ ] ] where @xmath389 .
the scet currents are dimension-6 operators @xmath390 where the @xmath259 fields are defined in eq .
( [ ch1 ] ) . in terms of the charge matrix @xmath391 , the spin and flavor structure is @xmath392 the wilson coefficients @xmath393 can be computed in a power series in @xmath368 .
they are functions of @xmath130 , @xmath15 , and the @xmath394 which are the sum of momentum labels for gauge invariant products of collinear fields in the scet currents . the currents operators in eq .
( [ j12 ] ) are the most general allowed operators which are gauge invariant , transform the same way as @xmath395 under charge conjugation and satisfy current and helicity conservation . to see how these properties constrain the form of the allowed operators , we begin by noting that eq .
( [ jspin ] ) implies that @xmath396 are the most general allowed spin structures . for massless
quarks the electromagnetic and qcd couplings preserve helicity , whereas @xmath397 cause the helicity to flip .
thus , only the structure @xmath398 is allowed .
current conservation @xmath399 , together with @xmath400 implies @xmath401 . under charge conjugation
@xmath402 so the same must be true for the scet currents . in the current operators , charge conjugation switches @xmath403 , @xmath404 , and @xmath405 , as can be seen from eq .
( [ cpt ] ) .
thus , the second term in @xmath406 is required to make these operators odd under charge conjugation .
the operators @xmath406 and the full electromagnetic current are invariant under a combined @xmath407 transformation .
this requires that the wilson coefficients are real .
the operators @xmath408 are responsible for the @xmath409 transition , while the reverse transition @xmath410 is described by similar operators with @xmath323 .
parity invariance requires the wilson coefficients of these operators to be identical to @xmath411 .
demanding hermiticity of the electromagnetic current in the effective theory then gives the relation @xmath412 .
since the coefficients are real they must therefore satisfy @xmath413 . to compute the matrix elements in the effective theory , it is convenient to fierz transform the four - quark operators in eq .
( [ j12 ] ) .
this gives @xmath414 where @xmath415 \big [ { \overline\chi_{{{\bar n}},\omega_3 } } \gamma_j^\prime { \chi_{{{\bar n}},\omega_2 } } \big ] \,.\end{aligned}\ ] ] the spin , flavor , and color structures are @xmath416 \,,{\nonumber}\\ \gamma_2 \otimes \gamma_2^\prime & = & \frac14 \big [ ( q_u+q_d ) ( { \bf 1}\otimes { \bf 1 } + \tau^a \otimes \tau^a ) + ( q_u - q_d ) ( { \bf 1}\otimes \tau^3 + \tau^3 \otimes { \bf 1 } ) \big]{\nonumber}\\ & & \times \big [ \bnslash\otimes \nslash + \bnslash\gamma_5\otimes \nslash\gamma^5 \big ] \,,\end{aligned}\ ] ] while @xmath417 and @xmath418 .
the new wilson coefficients are @xmath419 + ( \omega_{1,2}\leftrightarrow -\omega_{4,3 } ) \,,{\nonumber}\\ c_2(\mu,\omega_j ) & = & \bigg [ \frac18 \big(1-\frac{1}{n_c^2}\big ) c_8(\mu,\omega_j ) + \frac{1}{4n_c } c_0(\mu,\omega_j ) \bigg ] - ( \omega_{1,2}\leftrightarrow -\omega_{4,3 } ) \,,\end{aligned}\ ] ] with similar relations for @xmath420 which are also in terms of @xmath393 .
a few general predictions follow from the form of the operators in eq .
( [ jemc ] ) . ) over those in eq .
( [ feyn ] ) . ] for mesons with spin , only helicity conserving form factors appear , and furthermore no off - diagonal ( e.g. , @xmath421 ) matrix elements are present at leading order in @xmath379 .
these results agree with ref .
we also see that the form factors between arbitrary meson states are determined at leading power by only two hard coefficients , @xmath422 and @xmath423 .
now consider what factorization tells us about the matrix element of the operators in eq .
( [ jemc ] ) . for the decoupling of usoft and soft gluons we will follow section [ section_scet ] .
integrating out offshell modes with @xmath424 induces soft wilson lines @xmath149 and @xmath176 , while the field redefinitions in eq .
( [ def0 ] ) make all couplings to usoft gluons explicit in the operators .
together these give @xmath425\ : \big [ { \overline\chi_{{{\bar n}},\omega_3}}^{(0 ) } y_{{\bar n}}^\dagger s_{{\bar n}}^\dagger \gamma_j^\prime s_{{\bar n}}y_{{\bar n}}{\chi_{{{\bar n}},\omega_2}}^{(0)}\big ] \,.\end{aligned}\ ] ] consider first the color singlet currents @xmath426 . here the @xmath225 s and @xmath144 s all cancel using unitarity of the wilson lines . since the @xmath427 and @xmath428 gluons only interact with fields in the @xmath32 and @xmath33 directions respectively , collinear gluons are not exchanged between the @xmath32 and @xmath33 quark bilinears .
thus , the matrix element between states with particles moving in the @xmath32 and @xmath33 directions factors @xmath429 next consider the currents @xmath430 , which have color structure @xmath431 in @xmath432 . in this case
the usoft and soft gluons do not cancel , but can all be moved into one quark bilinear using the color identity @xmath433 . after this rearrangement
it is clear that the ( u)soft gluons and @xmath427 and @xmath428 gluons only interact with the fields in one of the quark bilinears .
thus , the matrix element @xmath434 factors , similar to eq .
( [ jfactor ] ) . for color singlet states , however ,
the matrix element of an octet operator vanishes identically since @xmath435 thus , the effective theory currents @xmath430 do not contribute to the form factors at any order in perturbation theory .
( [ jfactor ] ) shows that for arbitrary meson states factorization occurs .
it remains to show that the matrix elements in eq .
( [ jfactor ] ) are given by a two - dimensional convolution with the light - cone meson wavefunctions .
to do this we consider the simple example of the @xmath436 form factor for the charged pion .
it should be obvious that the same steps go through for other meson states .
the symmetry of the pion wavefunction @xmath437 under charge conjugation ( @xmath438 ) implies that only the @xmath439 current contributes .
thus , @xmath440 the required matrix elements can be obtained from eqs .
( [ pimom ] ) and ( [ pimom2 ] ) with @xmath441 .
the momentum conserving delta functions fix @xmath442 and @xmath443 , while the @xmath444 and @xmath445 integrations can be done with the delta functions .
this leaves @xmath446 where @xmath447 is defined in terms of @xmath448 as @xmath449 the coefficients @xmath450 , and therefore also @xmath451 , can be obtained at the scale @xmath452 by a matching calculation , as illustrated in fig .
[ fig_pipi ] . for this purpose , it is sufficient to compute the matrix element of the currents with free collinear quarks .
to lowest order in @xmath368 , only @xmath453 is nonvanishing @xmath454 this implies @xmath455 \,,\end{aligned}\ ] ] using the asymptotic light - cone pion wavefunction @xmath456 we find agreement with ref .
@xcite , @xmath457 ^ 2 \to \pm \frac{8\pi f_\pi^2 \alpha_s(q)}{q^2}\,.\end{aligned}\ ] ] the order @xmath458 corrections to eq .
( [ piff20 ] ) can be found in refs .
dis is a process which is both simple and rich in physics . as such it provides an ideal introduction to inclusive factorization in qcd , which we study from an effective field theory point of view in this section .
the aim is to prove that to all orders in @xmath9 and leading order in @xmath13 the dis forward scattering amplitude can be written as an integral over hard coefficients times the parton distribution functions .
this is done by matching onto local operators in scet .
the properties of scet are used to show that matrix elements of the leading local operator can be written as a convolution of a hard coefficient with the parton distribution functions for the proton .
the first step is to understand the kinematics of the process .
the hard scale @xmath459 is set by the invariant mass of the photon , and @xmath460 is the bjorken variable . in the breit frame
the momentum of the virtual photon is @xmath461 , and the incoming proton momentum is @xmath462 up to terms @xmath463 , where @xmath285 is the proton mass . by momentum conservation
the final state momentum is @xmath464 , which gives an invariant mass @xmath465 .
values @xmath466 correspond to the endpoint region where the particles in @xmath467 are collimated into a jet , while values @xmath468 correspond to the resonance region .
we will consider the standard ope region where @xmath469 so that the final state has virtuality of order @xmath370 and can be integrated out .
in contrast , although the incoming proton has a large momentum component in the @xmath470 direction it has a small invariant mass @xmath471 , and therefore is described by collinear fields in the effective theory .
consider the spin - averaged cross section for dis which can be written as @xmath472 where @xmath102 and @xmath103 are the incoming and outgoing lepton momenta with @xmath473 , @xmath474 is the lepton tensor , and @xmath475 .
the hadronic tensor @xmath476 can be related to the imaginary part of the dis forward scattering amplitude : @xmath477\,,{\nonumber}\end{aligned}\ ] ] where for an electromagnetic current @xmath478 we can write @xmath479 as explained above , the intermediate hadronic state has invariant mass @xmath480 .
therefore , one can perform an ope and match @xmath481 onto operators in scet .
all fields in the resulting operators are evaluated at the same residual space time point , however , the presence of wilson lines and label momenta make the operators nonlocal along a particular light cone direction .
these nonlocal operators sum the infinite set of purely local operators of a given twist , however this is built into the formalism automatically . in dis .
[ disopefig],width=528 ] to match we write down the most general leading operator in scet which contains collinear fields moving in the @xmath470 direction , and enforce the condition from current conservation @xmath482 .
this leads to @xmath483 where @xmath484,\end{aligned}\ ] ] where @xmath253 labels the flavor of the fermions and @xmath485 $ ] .
the wilson coefficients are dimensionless functions of @xmath246 , @xmath247 , @xmath15 , and @xmath130 . as in previous sections
we can separate the hard coefficients from the long distance operators by introducing trivial convolutions .
this gives @xmath486 \,,{\nonumber}\\ o_j^g & = & -\int\!\!{d\omega_1\,d\omega_2}\ , c^{g}_j(\omega_+,\omega_-)\ : { \rm tr}\big [ b_{n,\omega_1}^{\mu } b^{n,\omega_2}_\mu \big ] \,,\end{aligned}\ ] ] where @xmath487 , and @xmath488 with @xmath489 defined in eq .
( [ covg ] ) .
next we factor the coupling of usoft gluons from the collinear fields using the field redefinitions in eq .
( [ def0 ] ) .
the operator @xmath490 has the structure in eq .
( [ decoup ] ) so the @xmath225 s cancel trivially , while for @xmath491 we find @xmath492 and the factors of @xmath225 cancel in the trace .
it is easy to see that soft gluons also decouple using eq .
( [ sadd ] ) or by noting that there is no non - trivial soft gauge invariant way of adding soft wilson lines @xmath149 to @xmath490 or @xmath491 . under charge
conjugation the full theory electromagnetic current @xmath493 and therefore the operator @xmath494 .
this implies relations for the effective theory wilson coefficients since the operators @xmath490 must also respect this symmetry .
thus charge conjugation gives @xmath495 \ { \overline\chi_{n,\omega_1}}^{(i ) } { \bnslash } { \chi_{n,\omega_2}}^{(i ) } \,.\end{aligned}\ ] ] in the second line we changed variable @xmath496 and @xmath497 which takes @xmath498 and @xmath499 .
thus , to all orders in perturbation theory @xmath500 .
this relates the wilson coefficients for quarks and anti - quarks .
note that the above results are all independent of the collinear hadron on which dis is performed .
next we take the matrix element between proton states . using the definitions of the nonperturbative matrix elements given in section [ section_melt ] , and picking out the coefficients of the tensor structures we find that the delta functions in eq .
( [ pdfqg ] ) set @xmath501 and @xmath502 . since charge conjugation relates negative and positive values of @xmath503 only coefficients , @xmath504 , with positive @xmath503 are needed in the formulae for dis .
therefore we define @xmath505 where here we have made the dependence on @xmath15 and @xmath130 explicit . combining this with eqs .
( [ tprod ] ) , ( [ tmatch ] ) , ( [ disop1 ] ) , and ( [ pdfqg ] ) , gives the final result @xmath506 + \frac{\xi}{2x } h^{g}_1\big(\frac{\xi}{x}\big)\ , f_{g / p}(\xi ) \bigg\}\ , , { \nonumber}\\ t_2(x , q^2 ) & = & \frac{4x}{q^2 }
\int_0 ^ 1 \!\!d\xi \bigg\ { \big[4\,h^{(i)}_2\big(\frac{\xi } { x}\big ) \!-\ !
h^{(i)}_1\big(\frac{\xi}{x}\big ) \big ] \big [ f_{i / p}(\xi)+ \bar f_{{i}/p}(\xi ) \big ] { \nonumber}\\ & & \qquad\qquad\ \ + \frac{\xi}{2x } \big[4\,h^{g}_2\big(\frac{\xi } { x}\big ) \!-\ !
h^{g}_1\big(\frac{\xi}{x}\big ) \big ] f_{g / p}(\xi ) \bigg\ } \,,\end{aligned}\ ] ] where a sum over @xmath253 is implicit .
the hadronic tensor components @xmath507 and therefore are determined by the imaginary part of the wilson coefficients .
the wilson coefficients are dimensionless and therefore can only have @xmath508 dependence on @xmath15 .
this reproduces the bjorken scaling of the structure functions .
finally , consider the tree level matching onto the wilson coefficients shown in fig .
[ disopefig ] . from these graphs
only the quark coefficient functions @xmath509 can be non - zero and we find @xmath510 where @xmath511 is the charge of parton @xmath253 .
the vanishing of @xmath512 at tree level reproduces the callan - gross relation @xmath513 .
next we will extend the dis analysis to the drell - yan ( dy ) process : @xmath515 .
specifically we consider the @xmath370 distribution , where @xmath370 is the invariant mass of the lepton pair .
drell - yan is more complicated than dis because one has two hadrons in the initial state . in the center - of - mass frame
the incoming proton and anti - proton move in opposite lightlike directions , and to prove factorization we use the fact that collinear modes in different lightlike directions can only couple to each other in external operators in scet .
we take the incoming proton to move in the @xmath470 direction and the incoming antiproton to move in the @xmath516 direction .
the hard scales in dy are @xmath370 and the invariant mass of the colliding @xmath517 pair @xmath518 .
the lepton pair has an invariant mass @xmath370 , and the invariant mass of the final hadronic state is @xmath519 where @xmath520 we are interested in the kinematic region where @xmath521 , which implies that both @xmath522 and @xmath523 are far away from one .
as @xmath524 approaches one the invariant mass becomes too small for the treatment given here to apply .
however , the effective theory can be used to deal with this region as well .
it is also possible to study the @xmath525 distribution , but this again requires a generalization of the discussion given below .
the spin averaged cross section for drell - yan is@xmath526 where @xmath527 the sum over spins refers to the initial hadron spins ( the sum over final hadron spins is included in the sum over @xmath467 ) . integrating eq .
( [ sdy ] ) over the emission angles of the final leptons one obtains @xmath528 where we have neglected the lepton masses and defined the operator @xmath529 as we discussed above in the region of phase space under consideration @xmath530 , so these hard fluctuations can be integrated out .
operationally this means we match @xmath531 onto local operators in the effective theory .
we would like to show that the minimal set of order @xmath532 operators that contribute to drell - yan are @xmath533 \big[{\overline\chi_{{{\bar n}},\omega_3}}^{\,(i ) } \:{\nslash}\ : { \chi_{{{\bar n}},\omega_4}}^{\,(i)}\big ] \ , { \nonumber}\\[3pt ] & - & \frac{1}{q^3 } \int\!\!{d\omega_i}\ : c_{qg}(\omega_i , q)\ , \big[{\overline\chi_{n,\omega_1}}^{\,(i ) } \:{\bnslash}\ : { \chi_{n,\omega_2}}^{\,(i)}\big ] { \rm tr}\big [ { \cal b}_{{{\bar n}},\omega_3}^{\beta } { \cal b}^{{{\bar n}},\omega_4}_{\beta } \big ] \ , { \nonumber}\\[3pt ] & - & \frac{1}{q^3 } \int\!\!{d\omega_i}\ : c_{gq}(\omega_i , q ) { \rm tr}\big [ { \cal b}_{n,\omega_1}^\nu { \cal b}^{n,\omega_2}_{\nu } \big ] \big [ { \overline\chi_{{{\bar n}},\omega_3}}^{\,(i ) } \:{\nslash}\ : { \chi_{{{\bar n}},\omega_4}}^{\,(i)}\big ] \,{\nonumber}\\[3pt ] & + & \frac{1}{q^4 } \int\!\!{d\omega_i}\ : c_{gg}(\omega_i , q)\ , { \rm tr}\big [ { b}_{n,\omega_1}^\nu { b}^{n,\omega_2}_{\nu } \big ] { \rm tr}\big [ { \cal b}_{{{\bar n}},\omega_3}^{\beta } { \cal b}^{{{\bar n}},\omega_4}_{\beta } \big ] \,,\end{aligned}\ ] ] where the powers of @xmath15 are included to make the coefficients dimensionless .
the operators displayed in eq .
( [ dymatching ] ) are just products of the operators that occured in dis , so for these terms the decoupling of soft and usoft gluons occurs in a straightforward manner . to show that the operators in eq .
( [ dymatching ] ) are the most general set needed we must show that all other operators that are order @xmath532 either reduce to these or vanish between the matrix elements in eq . ( [ dyqsqrdist ] ) .
for instance , @xmath532 operators also exist where a @xmath534 field is contracted with a @xmath535 field , or the color structures of the operators in eq .
( [ dymatching ] ) could be arranged in a different way .
we now give a general argument for why we can always rewrite an arbitrary operator in the form of eq .
( [ dymatching ] ) or show that it does not contribute to dy .
all operators relevant for dy contain four order @xmath13 collinear fields chosen from @xmath66 , @xmath165 , @xmath536 , or @xmath537 .
furthemore , two must move in direction @xmath32 and two in the direction @xmath33 ( other possibilities end up vanishing by baryon number conservation or because they involve a set of fields between physical states that can not possibly form a color singlet operator ) . for operators with 4 quark fields , fierz transformations can always be made to arrange the fields such that those in the same direction sit in the same bilinear . using as an example the operator with two collinear quarks in the @xmath32 directions and two gluons in the @xmath33 direction and leaving out the soft wilson lines for the moment , the most general matrix element is @xmath538 where @xmath539 are quark colors , @xmath540 are gluon colors , and @xmath541 are spinor indices for the quarks .
@xmath542 is some tensor that connects the indices in an arbitrary way . in the contraction of @xmath539 and @xmath540
there are two possible ways to make an overall color singlet , one where both the quarks and gluons are in a color singlet , and another where both the quarks and gluons are in a color octet .
we will discuss both of these possibilities in turn .
, width=336 ] in the color singlet case , including the soft and ultrasoft wilson lines is trivial , since using eqs .
( [ sadd ] ) , ( [ yfd2 ] ) , and ( [ eom2 ] ) we see that they cancel due to unitarity / orthogonality of the various wilson lines in the fundamental / adjoint representations .
thus , there are no soft , usoft , or collinear interactions that connect the @xmath32 and the @xmath33 fields . as in previous sections
this leads to a factorization of the matrix element in eq .
( [ generaldy ] ) , namely @xmath543 since the proton spins are summed over , we can write ( with the help of eq .
( [ tracefomulae ] ) ) @xmath544 so that spin and color are summed over in the matrix element .
similarly the antiproton matrix element can be simplified to @xmath545 \
, | \bar{p}_{{\bar n}}\rangle\,.\end{aligned}\ ] ] here we used the fact that the matrix element is symmetric in @xmath130 and @xmath546 , and that only the perpendicular index @xmath130 of the field @xmath547 is order @xmath13 . using eqs .
( [ quarkmatrix ] ) and ( [ gluonmatrix ] ) the original matrix element in eq .
( [ generaldy ] ) can be written as @xmath548 & & \qquad\qquad \propto \ ; { \rm tr}[\delta^{\:\mu}_\mu\ , \nslash]\ , \langle \bar{p}_{{\bar n}}| \ , { \rm tr}\:\big [ { \cal b}_{{{\bar n}},\omega_3}^{(0)\nu } { \cal b}^{(0){{\bar n}},\omega_4}_{\nu } \big ] \ , | \bar{p}_{{\bar n}}\rangle\ , \langle
p_n |\ , { \overline\chi_{n,\omega_1}}^{(0 ) } \bnslash { \chi_{n,\omega_2}}^{(0 ) } \ , |p_n \rangle \,,{\nonumber}\end{aligned}\ ] ] where the trace of @xmath549 is over spin and color , and just gives an overall constant .
the final result in eq .
( [ enddyeg ] ) is identical to the matrix element of the second operator in eq .
( [ dymatching ] ) .
if each of the @xmath32 and @xmath33 field bilinears involve color octet structures , then the soft and usoft wilson lines do nt cancel , since they do nt commute with the su(3 ) generators .
however , one can use the color identity @xmath550 where each @xmath551 , @xmath144 , and @xmath225 factor is in the appropriate representation of the color group .
( [ cident ] ) moves all the soft and ultrasoft interactions between either the @xmath32 or the @xmath33 collinear fields .
thus , again the fields in one bilinear can not be contracted with fields in the other bilinear and the matrix element factors .
however this time the factored matrix element vanishes .
for the example discussed above , @xmath552 since an color octet operator vanishes between color singlet states .
the same holds true for the matrix element of an octet gluon operator .
an identical proof of decoupling goes through for the case of 4 quarks , where we again either have two color singlet or two color octet @xmath32 and @xmath33 bilinears . with 4 gluon fields we can either have the two @xmath32 and two @xmath33 fields coupled as singlets , or coupled in the same higher representation ( an * 8 * , \{*10 * , @xmath553 } , or * 27 * ) . in the latter case the matrix element between color singlet states still vanishes so the proof for the 4 gluon operators also goes through in an identical way .
thus we have shown that the matrix element of an operator with an arbitrary contraction of indices either vanishes or can be written in terms of a product of a matrix element which is related to a proton pdf and a matrix element which is related to an antiproton pdf as in the example in eq .
( [ enddyeg ] ) .
this is the result we want . to see how the final formulae are derived note that we can write the matrix element of eq .
( [ dymatching ] ) in the form of a convolution @xmath554 where @xmath555 and @xmath556 .
the operators here are the same as in dis , with @xmath557 for the quark operator , and @xmath558 for the gluon operator @xmath559 \,,{\nonumber}\\ o^g_n(\omega_+ , \omega_- ) & = & -\frac{1}{q^2 } { \rm tr}\big [ b_{n,\omega_1}^\mu b^{n,\omega_2}_\mu \big ] \,.\end{aligned}\ ] ] apart from the dependence on the labels , the wilson coefficients in eq .
( [ dyconv ] ) can also depend on the renormalization point @xmath130 and the kinematic variable @xmath15 .
using eq .
( [ pdfqg ] ) we see that the matrix elements in eq .
( [ dyop1 ] ) set @xmath560 and @xmath561 , @xmath562 where @xmath563 and @xmath564 are the convolution variables . since all kinematic variables aside from @xmath370 are integrated over in eq .
( [ what ] ) the only other variable that can appear in the wilson coefficient is the center of mass energy which produces the @xmath565 pair , namely @xmath566 .
thus , the wilson coefficients only depend on @xmath567 .
defining new coefficients @xmath568 we can replace the matrix elements in eq .
( [ dyop1 ] ) with parton distribution functions using eq .
( [ pdfqg ] ) to obtain : @xmath569{\nonumber}\\[3pt ] & & \hspace{-1cm}+ h^{(i)(j)}(z_1 z_2 ) \big [ f_{i / p}(z_1 ) f_{j/\bar{p}}(z_2)+ \bar{f}_{i / p}(z_1 ) \bar{f}_{j/\bar{p}}(z_2 ) \big]\nonumber\\[3pt ] & & \hspace{-1 cm } + \frac{z_2}{2\sqrt{\tau}}\ : h^{(i)g}(z_1 z_2 )
f_{i / p}(z_1 ) f_{g/\bar{p}}(z_2)- \frac{z_2}{2\sqrt{\tau}}\ : h^{(i)g}(-z_1 z_2 ) \bar{f}_{i / p}(z_1 ) f_{g/\bar{p}}(z_2 ) { \nonumber}\\[3pt ] & & \hspace{-1 cm } + \frac{z_1}{2\sqrt{\tau}}\ : h^{g(j)}(z_1 z_2 ) f_{g / p}(z_1 )
f_{j/\bar{p}}(z_2 ) -\frac{z_1}{2\sqrt{\tau}}\ : h^{g(j)}(-z_1 z_2 ) f_{g / p}(z_1)\bar{f}_{j/\bar{p}}(z_2 ) { \nonumber}\\[3pt ] & & \hspace{-1 cm } + \frac{z_1 z_2}{4\tau}\ : h^{gg}(z_1 z_2 ) f_{g / p}(z_1 ) f_{g/\bar{p}}(z_2 ) \bigg\ } \,.\end{aligned}\ ] ] this is the final convolution formulae for drell - yan and is valid to all orders in @xmath9 and leading order in the power expansion . at tree level
the matching calculation shown in fig .
[ dyfig ] yields zero for all the wilson coefficients except @xmath570 where @xmath511 is the charge of parton @xmath253 . the coefficients @xmath571 and @xmath572 start at order @xmath368 , while @xmath573 and @xmath574 start at order @xmath458 .
next we examine deeply virtual compton scattering ( dvcs ) . to be more precise
we examine the exclusive reaction @xmath576 , where the incoming photon is highly virtual , the final photon is either off - shell or real , and the incoming and outgoing protons have different momenta .
the reason we have included this process in the inclusive section is that dvcs has the remarkable property that the nonperturbative physics is described by a so called non - forward parton distribution function ( nfpdf ) .
the nfpdf is a more general distribution function that reduces to the standard parton distribution functions ( familiar from dis ) for some values of the momentum fraction , and behaves like a lightcone wavefunction ( familiar from the pion examples ) for other values .
deeply virtual compton scattering was first studied in perturbative qcd in refs @xcite , and proofs of factorization to all orders in perturbation theory were later presented in refs @xcite .
in addition properties of nfpdfs were studied in ref .
@xcite . here
we present a proof of factorization for dvcs based on scet . as with the previous proofs it is important to understand the kinematics of the process .
we take the incoming proton and photon momenta to be @xmath56 and @xmath206 respectively , with @xmath577 and @xmath578 .
the outgoing proton and photon momentum are @xmath101 and @xmath579 respectively , with @xmath580 .
it is convenient to define a parameter @xmath581 , which measures the change to the proton s large momentum .
working in the breit frame and neglecting contributions that are @xmath582 we have : @xmath583 & $ q^\mu=$\hspace{0.8 cm } & $ { \frac{\mbox{\normalsize $ q$}}{\mbox{\normalsize $ 2 $ } } } ( { { \bar n}}^\mu\!-\ ! n^\mu)$ & $ + 0 $ \\[5pt ] & $ p^\mu=$ & $ { \frac{\mbox{\normalsize $ q$}}{\mbox{\normalsize $ 2 x$ } } } \ , n^\mu$ & $ + { \frac{\mbox{\normalsize $ x$}}{\mbox{\normalsize $ 2q$ } } } \ , m_p^2(n^\mu\!+\!{{\bar n}}^\mu)$ \\[5pt ] & $ p^{\prime\mu}=$ & $ { \frac{\mbox{\normalsize $ q$}}{\mbox{\normalsize $ 2 x$ } } } ( 1\!-\!\zeta ) \
, n^\mu \!+\ ! p^{\prime\mu}_\perp$ & $ + { \frac{\mbox{\normalsize $ x$}}{\mbox{\normalsize $ 2q$ } } } \big [ m_p^2(1\!-\ !
\zeta)n^\mu \!+\
! t\}{{\bar n}}^\mu\big]$\hspace{0.3 cm } \\[5pt ] & $ q^{\prime\mu}=$ & $ { \frac{\mbox{\normalsize $ q$}}{\mbox{\normalsize $ 2 $ } } } \big({\frac{\mbox{\normalsize $ \zeta$}}{\mbox{\normalsize $ x$ } } } -\ ! 1 \big ) \ , n^\mu + { \frac{\mbox{\normalsize $ q$}}{\mbox{\normalsize $ 2 $ } } } \ , { { \bar n}}^\mu \!-\ !
p_\perp^{\prime\mu}$\hspace{0.7 cm } & $ + { \frac{\mbox{\normalsize $ x$}}{\mbox{\normalsize $ 2q$ } } } \big [ \zeta m_p^2 n^\mu + ( t\!-\!\zeta m_p^2){{\bar n}}^\mu\big]$ \\[5pt ] \hline \end{tabular } \hspace{0.5 cm } \vspace{0.2cm}\end{aligned}\ ] ] here the label momenta are order @xmath15 or @xmath63 , while the residual momenta are order @xmath64 and depend on @xmath584 and @xmath585 , which are both @xmath586 .
the invariant mass of the intermediate hadronic state is @xmath587 just like dis , so for @xmath588 the intermediate state can be integrated out .
we will proceed in a manner analogous to the analysis for dis .
the amplitude ( up to an overall momentum conserving @xmath589-function ) is given by a time ordered product of currents : @xmath590 \,.\end{aligned}\ ] ] this time ordered product is contracted with a lepton tensor to obtain the amplitude .
now current conservation requires @xmath591 , however the dvcs @xmath592 is not symmetric under @xmath593 . for electromagnetic currents
@xmath478 we have @xmath594 where the functions @xmath595 , and the vectors @xmath596 and @xmath597 are defined so that @xmath598 . in eq .
( [ tdvcs ] ) and below the ellipses denote spin dependent terms .
for simplicity we will show how factorization is achieved for the spin independent contributions shown in eq .
( [ tdvcs ] ) with the understanding that it is no more difficult to also include the other terms .
it is convenient to define a parameter @xmath599 , by @xmath600 .
the dis hadronic time - ordered product is obtained in the limit @xmath601 , where @xmath602 and @xmath293 . from eq .
( [ dvcskin ] ) we see that @xmath603 so these parameters are not independent .
since the intermediate hadronic state has invariant mass @xmath604 we can match @xmath605 onto operators in scet . requiring @xmath606 and @xmath607 for the order @xmath15 label momenta leads to @xmath608 where the ellipses are spin dependent terms and the displayed operators are @xmath609 \,.\end{aligned}\ ] ] we have suppressed the dependence of the wilson coefficients @xmath610 on @xmath15 , @xmath611 , and @xmath130 .
the form of the operators in eq .
( [ dvcsop ] ) looks the same as the dis operators given in eq .
( [ disop ] ) , however the operators here are more general because the wilson coefficients depend on @xmath611 . in the limit @xmath602 the dvcs operators reduce to the dis operators .
however , since the field structure of the dvcs operators is identical to dis several results follow immediately .
for instance , the steps which factorize soft and usoft gluons and leave fields with superscript @xmath143 are the same and are not repeated here , @xmath612 \,.\end{aligned}\ ] ] the restrictions on the dvcs wilson coefficients from charge conjugation are the same as in eq .
( [ dischg ] ) , @xmath613 , however because @xmath614 this is not simply a relation between quark and anti - quark wilson coefficients .
the way in which dvcs is unique is that the matrix elements involve nucleon states with different momenta .
this is what leads to results in terms of non - forward parton distribution functions .
the definition of the nfpdfs are given in eq .
( [ nfpdfqg ] ) , and can be used along with the relations above to obtain expressions for the @xmath615 in terms of the nfpdfs .
before we give this result we note that the wilson coefficients depend on the operators @xmath616 which become the variables @xmath617 after introducing trivial convolutions and the @xmath618 fields in eq .
( [ ch1 ] ) .
the delta functions in eq .
( [ nfpdfqg ] ) then set @xmath619 and @xmath620 , where @xmath621 is the convolution variable .
note that @xmath622 , and just like dis it is the combination @xmath623 which appears .
since charge conjugation relates the wilson coefficients for @xmath624 and @xmath625 it is convenient to define @xmath626 where in the last three arguements we have made the dependence on @xmath15 , @xmath611 , and @xmath130 explicit . combining eqs .
( [ nfpdfqg ] ) , ( [ dvcsop2 ] ) , and ( [ hdvcs ] ) then gives @xmath627 + \frac{1}{2x } { \cal h}^{g}_1 \big(\frac{\xi}{x}\big)\ : { \cal f}^{g}_\zeta(\xi;t ) \bigg\ } + \ldots \ , , { \nonumber}\\[3pt ] t_2 & = & \frac{x^2(1+\alpha)}{q^3}\ : e(\sigma^\prime,\sigma ) \int_0 ^ 1 \!\!d\xi \bigg\ { \big [ 2(1+\alpha ) { \cal h}^{(i)}_2 \ !
\big(\frac{\xi}{x}\big ) \!-\ ! { \cal h}^{(i)}_1 \ !
\big(\frac{\xi}{x}\big ) \big ] \big[{\cal f}^{i}_\zeta(\xi;t)+\overline{{\cal f}}^{i}_\zeta(\xi;t)\big ] { \nonumber}\\ & & + \frac{1}{2x } \big [ 2(1+\alpha)\ , { \cal h}^{g}_2 \big(\frac{\xi}{x}\big ) \!-\ ! { \cal h}^{g}_1 \big(\frac{\xi}{x}\big ) \big ] { \cal f}^{g}_\zeta(\xi;t ) \bigg\ } + \ldots \ , , { \nonumber}\\[3pt ] t_3 & = & 0\ , , \qquad t_4 = 0\ , , \qquad t_5 = 0 \,,\end{aligned}\ ] ] which are the final convolution results valid to all orders in @xmath9 and leading order in the power expansion . the structure functions @xmath628 vanish since the vectors @xmath629 at leading order in the power expansion .
the terms with ellipses are for the spin flip terms and involve the nfpdf @xmath630 defined in eq .
( [ nfpdfqg ] ) .
the results for these terms have a similar form to those in eq .
( [ tmatch ] ) .
finally we match at tree level .
the tree level diagram in qcd is the same as in fig .
[ disopefig ] except the outgoing photon and proton have momenta @xmath631 and @xmath632 respectively . only the quark wilson coefficients are nonzero at tree level .
we find @xmath633 which gives @xmath634 since @xmath635 at tree level , dvcs also obeys a callan - gross relation .
what we hope we have demonstrated here is the power of effective field techniques in the context of factorization for hard scattering processes .
the explicit separation of modes and the implementation of gauge invariance for these modes greatly simplifies seemingly complex problems .
what is normally accomplished by diagramatic ward identities and induction techniques now falls out as a consequence of the gauge symmetry of operators in a low energy soft - collinear effective theory .
as we have emphasized the factorization formulae derived in this paper are not new .
the purpose here was simply to extend the formalism introduced in @xcite to cases with back - to - back collinear particles , and apply these ideas to more general processes than previously considered .
furthermore , the factorization proofs presented are perhaps simpler than those previously given ( certainly they are more concise ) .
we believe that within the confines of the scet more difficult , and unresolved problems can be addressed , such as power corrections in cases without an ope , and proofs of factorization for more complex processes . this work was supported in part by the department of energy under the grant doe - fg03 - 97er40546 .
this appendix discusses the simultaneous factorization of the soft @xmath80 modes , @xmath32-collinear @xmath78 modes , and @xmath33-collinear @xmath636 modes .
these three classes of modes can not interact with each other in a local manner and therefore do not couple through the scet lagrangian .
however , they can couple in a gauge invariant way through external operators and currents . these interactions in currents
are built up by integrating out offshell fluctuations with @xmath637 .
for the special case of factorization of soft from @xmath32-collinear modes this was shown in detail in the appendix of ref .
there it was shown that integrating out certain modes with offshellness @xmath147 causes the wilson lines @xmath119 and @xmath149 to appear in operators in a gauge invariant way . here
we will extend this approach to the factorization of modes for cases involving two classes of collinear particles . for simplicity
we restrict ourselves to the case where the original operators involve only collinear quark or gluon fields .
this type of factorization was used for the pion form factor example discussed in sec .
iii.b and the drell - yan process presented in sec .
.summary of the onshell modes discussed in section [ section_scet ] , and the auxiliary fields introduced to represent the offshell fluctuations that are integrated out in this appendix .
[ table_off ] [ cols="^,<,<,^,^,^,^",options="header " , ] the basic idea is to first match onto a lagrangian with couplings between onshell and offshell modes that give all order @xmath87 diagrams
. the offshell modes ( with @xmath637 ) can then be integrated out , so that all operators are expressed entirely in terms of the onshell degrees of freedom . in table
[ table_off ] a summary is given of the three types of offshell momenta that are induced by adding soft , @xmath32-collinear , and @xmath33-collinear momenta .
for each type auxiliary quark and gluon fields are defined , and for convenience momentum labels are suppressed in this section .
for example , the @xmath638 quarks are created by the interaction of a @xmath32-collinear quark with an @xmath33-collinear gluon , whereas the @xmath639 quarks are created when a collinear quark @xmath304 is knocked offshell by a soft gluon . for the field
@xmath638 we write @xmath640 , where @xmath641 and @xmath642
. then we have @xmath643 and @xmath644 .
various wilson lines are also required and are listed in the table .
it is convenient to define a generic wilson line @xmath645 $ ] along direction @xmath646 with field @xmath647 by the solution of @xmath648 = 0 \,.\end{aligned}\ ] ] with this notation the on - shell wilson lines are @xmath649 $ ] , @xmath650 $ ] , @xmath651 $ ] , and @xmath652 $ ] .
( recall that the subscripts on @xmath18 and @xmath144 mean different things . )
the wilson lines involving offshell fields that we will require are @xmath653 \ , , \qquad x_{{\bar n}}=l[n , a_q\!+\!a_{{\bar n}}^x\!+\!a_{{\bar n } } ] \,,\\ w_n^x & = & l[{{\bar n}},a_n^x\!+\!a_n ] \
, , \qquad\qquad\ ! w_{{\bar n}}^x = l[n , a_{{\bar n}}^x\!+\!a_{{\bar n } } ] \,,{\nonumber}\\ s_n^x & = & l[n , a_n^x\!+\!a_s]\ , , \qquad\qquad\ s_{{\bar n}}^x = l[{{\bar n}},a_{{\bar n}}^x\!+\!a_s ] \,.{\nonumber}\end{aligned}\ ] ] below we discuss the results which allow us to integrate out offshell fluctuations .
the structure of the auxiliary lagrangians and construction of their solutions are very similar to the case presented in ref .
@xcite , to which we refer for a more detailed presentation . from table [ table_off ]
we see that adding @xmath32 and @xmath33-collinear momenta gives @xmath654 , whereas adding soft and collinear momenta gives @xmath655 .
loops that are dominated by offshell momenta only modify wilson coefficients and not the infrared physics .
therefore , to determine the structure of scet fields in an operator it sufficient to integrate out the offshell fields at tree level . for convenience we can integrate out the fluctuations starting with those with the largest offshellness . recall that we only wish to consider offshell propagators connected to external operators .
a subtlety for quarks is that distinct auxiliary fields are needed for the incoming and outgoing offshell propagators .
however , the solution for the outgoing field turns out to be the conjugate of the incoming field , so to avoid a proliferation of notation we simply denote the outgoing terms in the lagrangian by @xmath656 , and present a solution for the incoming fields .
finally , note that for the gluon field @xmath657 the fields @xmath658 , @xmath659 , @xmath660 , @xmath661 , and @xmath662 appear as background fields while for the fields @xmath663 and @xmath661 it is @xmath658 , @xmath659 , and @xmath662 that appear as background fields .
the terms in the auxiliary lagrangian involving the @xmath654 fields are @xmath664 \frac{\bnslash}{2 } \psi^q_n { \nonumber}\\ & & + ( n\leftrightarrow { { \bar n } } ) + \mbox{h.c . } { \nonumber}\\ & & + \frac{1}{2g^2}\mbox{tr}\left\ { [ id_q^\mu + ga_q^\mu\ , , id_q^\nu + ga_q^\nu]^2\right\ } + \frac{1}{\alpha_q}\mbox{tr } \{[id_q^\mu , a_{q\mu}]^2\}\,.\end{aligned}\ ] ] where @xmath665 + \frac12 { { \bar n}}^\mu [ { { { \cal p } } } + gn\cdot ( a^x_{{\bar n}}\!+\ ! a_{{\bar n}})]$ ] .
the solution of the equations of motion for these modes are @xmath666 & & x_{{\bar n}}^\dagger x_n = w_n^x w_{{\bar n}}^{x\dagger}\,.\end{aligned}\ ] ] ( in addition to the last equation a constraint on the components @xmath667 and @xmath668 also comes from the gauge fixing term , but will not be needed . ) the terms in the auxiliary lagrangian involving the @xmath655 fields are @xcite @xmath669 \frac{\bnslash}{2 } \psi^l_n + ( n\leftrightarrow { { \bar n}})+ \mbox{h.c . } \\ & + & \frac{1}{2g^2}\mbox{tr}\big\ { [ id_{nx}^\mu \!+\ !
ga_{n}^{x\mu}\ , , id_{nx}^\nu \!+\ ! ga_n^{x\nu}]^2\big\
} + \frac{1}{\alpha_{n}}\mbox{tr } \big\{[id_{nx}^\mu , a_{n\mu}^{x}]^2\big\ } + ( n\leftrightarrow { { \bar n } } ) \,,{\nonumber}\end{aligned}\ ] ] where @xmath670+\frac12 { { \bar n}}^\mu [ n{\!\cdot\!}{{{\cal p}}}+gn{\!\cdot\!}a_s]$ ] . the solutions for these modes are @xmath671 together eqs .
( [ eom1 ] ) and ( [ eom2 ] ) can be used at leading order to eliminate the fields representing offshell fluctuations with @xmath672 . for collinear quarks
this leads to the rules in eq .
( [ sadd ] ) .
note that we did not need to use the gauge fixing term to resolve the ambiguity in the implicit solution for the @xmath673 and @xmath674 auxillary fields . as an illustration of these results we discuss the soft - collinear factorization for the production of a @xmath675 pair with a large invariant mass @xmath370 .
this process is mediated in the full theory by the electromagnetic current @xmath676 ( @xmath300 a color singlet ) .
this current will match onto a current in scet that is built entirely out of onshell fields . using the results in this appendix this
current can be systematically derived . to start the quark field in @xmath677 matches onto @xmath304 plus all possible fields which the auxiliary lagrangian can create starting from @xmath304 , so @xmath678 integrating out the @xmath654 fluctuations with eq .
( [ eom1 ] ) and inserting a hard wilson coefficient @xmath211 which depends on label operators turns eq .
( [ j1 ] ) into @xmath679 to construct the first operator we used the equations of motion for @xmath680 and @xmath681 , and in the second operator we used the equation of motion identity for the gluons in @xmath187 and @xmath188 . in a similar fashion
we can now integrate out the @xmath147 fluctuations with eq .
( [ eom2 ] ) to give @xmath682 the operator on the right is the final result used in eq .
( [ onnws ] ) , and is soft , collinear , and usoft gauge invariant .
it should be obvious from this example how the equations of motion in eqs . ( [ eom1 ] ) and ( [ eom2 ] ) can be used to determine the factorized form of a general leading order operator .
r. k. ellis , h. georgi , m. machacek , h. d. politzer and g. g. ross , nucl .
b * 152 * , 285 ( 1979 ) .
h. d. politzer , nucl .
b * 129 * , 301 ( 1977 ) ; c. t. sachrajda , phys .
b * 73 * , 185 ( 1978 ) ; g. t. bodwin , s. j. brodsky and g. p. lepage , phys . rev .
lett . * 47 * , 1799 ( 1981 ) ; g. t. bodwin , phys .
d * 31 * , 2616 ( 1985 ) [ erratum - ibid .
d * 34 * , 3932 ( 1985 ) ] ; j. c. collins , d. e. soper and g. sterman , phys .
b * 109 * , 388 ( 1982 ) ; j. c. collins , d. e. soper and g. sterman , phys .
b * 134 * , 263 ( 1984 ) ; j. c. collins , d. e. soper and g. sterman , nucl .
b * 308 * , 833 ( 1988 ) .
j. c. collins and d. e. soper , ann .
nucl . part .
sci . * 37 * , 383 ( 1987 ) .
g. sterman , tasi lectures 1995 [ arxiv : hep - ph/9606312 ] .
r. l. jaffe , [ arxiv : hep - ph/9602236 ] .
m. beneke , g. buchalla , m. neubert and c. t. sachrajda , nucl .
b * 606 * , 245 ( 2001 ) [ arxiv : hep - ph/0104110 ] .
p. nason , [ arxiv : hep - ph/0111024 ] .
g. p. korchemsky and g. sterman , nucl .
b * 555 * , 335 ( 1999 ) [ arxiv : hep - ph/9902341 ] . c. w. bauer , s. fleming and m. luke , phys .
d * 63 * , 014006 ( 2001 ) [ arxiv : hep - ph/0005275 ] . c. w. bauer , s. fleming , d. pirjol and i. w. stewart , phys .
d * 63 * , 114020 ( 2001 ) [ arxiv : hep - ph/0011336 ] . c. w. bauer and i. w. stewart , phys .
b * 516 * , 134 ( 2001 ) [ arxiv : hep - ph/0107001 ] . c. w. bauer , d. pirjol and i. w. stewart , [ arxiv : hep - ph/0109045 ] . c. w. bauer , d. pirjol and i. w. stewart , phys .
lett . * 87 * , 201806 ( 2001 ) [ arxiv : hep - ph/0107002 ] . c. w. bauer , c. w. chiang , s. fleming , a. k. leibovich and i. low , phys .
d * 64 * , 114014 ( 2001 ) [ arxiv : hep - ph/0106316 ] ; for previous work on the resummation of logarithems in @xmath683 and @xmath684 see g. p. korchemsky and g. sterman , phys .
b * 340 * , 96 ( 1994 ) [ arxiv : hep - ph/9407344 ] .
r. akhoury and i. z. rothstein , phys .
d * 54 * , 2349 ( 1996 ) [ arxiv : hep - ph/9512303 ] .
a. k. leibovich and i. z. rothstein , phys .
d * 61 * , 074006 ( 2000 ) [ arxiv : hep - ph/9907391 ] .
a. k. leibovich , i. low and i. z. rothstein , phys .
d * 62 * , 014010 ( 2000 ) [ arxiv : hep - ph/0001028 ] .
j. chay and c. kim , [ arxiv : hep - ph/0201197 ] .
a. v. manohar and m. b. wise , cambridge monogr .
cosmol .
* 10 * , 1 ( 2000 ) .
h. georgi , phys .
b * 240 * , 447 ( 1990 ) .
m. e. luke , a. v. manohar and i. z. rothstein , phys .
d * 61 * , 074025 ( 2000 ) [ arxiv : hep - ph/9910209 ] . j. c. collins , d. e. soper and g. sterman , phys .
b * 438 * , 184 ( 1998 ) [ arxiv : hep - ph/9806234 ] .
g. p. korchemsky , phys .
b * 217 * , 330 ( 1989 ) .
j. collins , in _ perturbative quantum chromodynamics _ , p. 573 .
a. v. manohar and i. w. stewart , phys .
* d63 * , 054004 ( 2001 ) [ arxiv : hep - ph/0003107 ] .
a. h. hoang , a. v. manohar and i. w. stewart , phys .
rev . * d64 * , 014033 ( 2001 ) .
[ arxiv : hep - ph/0102257 ] .
g. p. lepage and s. j. brodsky , phys .
d * 22 * , 2157 ( 1980 ) .
d. e. soper , nucl .
suppl . * 53 * , 69 ( 1997 ) .
[ arxiv : hep - lat/9609018 ] . a. v. radyushkin , phys .
d * 56 * , 5524 ( 1997 ) [ arxiv : hep - ph/9704207 ] .
s. j. brodsky and g. p. lepage , in _ perturbative quantum chromodynamics _ , p. 93 - 240 .
f. del aguila and m. k. chase , nucl .
b * 193 * , 517 ( 1981 ) .
e. braaten , phys .
d * 28 * , 524 ( 1983 ) .
s. j. brodsky , [ arxiv : hep - ph/0102051 ] .
n. isgur and c. h. llewellyn smith , phys .
b * 217 * , 535 ( 1989 ) .
r. d. field , r. gupta , s. otto and l. chang , nucl .
b * 186 * , 429 ( 1981 ) .
f. m. dittes and a. v. radyushkin , sov .
j. nucl .
phys . * 34 * , 293 ( 1981 ) [ yad .
fiz . * 34 * , 529 ( 1981 ) ] .
e. braaten and s. m. tse , phys .
d * 35 * , 2255 ( 1987 ) .
x. d. ji , phys .
lett . * 78 * , 610 ( 1997 ) [ arxiv : hep - ph/9603249 ] .
x. d. ji , phys .
d * 55 * , 7114 ( 1997 ) [ arxiv : hep - ph/9609381 ] .
a. v. radyushkin , phys .
b * 380 * , 417 ( 1996 ) [ arxiv : hep - ph/9604317 ] .
j. c. collins and a. freund , phys .
d * 59 * , 074009 ( 1999 ) [ arxiv : hep - ph/9801262 ] .
x. d. ji and j. osborne , phys .
d * 58 * , 094018 ( 1998 ) [ arxiv : hep - ph/9801260 ] .
|
in this paper we show how gauge symmetries in an effective theory can be used to simplify proofs of factorization formulae in highly energetic hadronic processes .
we use the soft - collinear effective theory , generalized to deal with back - to - back jets of collinear particles .
our proofs do not depend on the choice of a particular gauge , and the formalism is applicable to both exclusive and inclusive factorization . as examples we treat the @xmath0-@xmath1 form factor ( @xmath2 ) , light meson form factors ( @xmath3 ) , as well as deep inelastic scattering ( @xmath4 ) , drell - yan ( @xmath5 ) , and deeply virtual compton scattering ( @xmath6 ) .
|
a great deal of work has been done recently to embed the standard model of particle physics ( sm ) into a brane construction @xcite .
this research is part of the effort , initiated in @xcite , to build a fully realistic four dimensional vacuum out of string theory .
while the original models were formulated in the framework of the heterotic string , the most recent efforts were formulated for type ii strings in order to take advantage of the recent work on moduli stabilization using fluxes .
such brane constructions naturally lead to extra anomalous @xmath3 s in the four dimensional low - energy theory and , in turn , to the presence of possible heavy @xmath4 particles in the spectrum .
these particles should be among the early findings of lhc and besides for the above cited models they are also a prediction of many other theoretical models of the unification of forces ( see @xcite for a recent review ) .
it is then of some interest to know if these @xmath4 particles contribute to the cancelation of the gauge anomaly in the way predicted from string theory or not . in @xcite
some of the present authors have studied a supersymmetric ( susy ) extension of the minimal supersymmetric standard model ( mssm ) in which the anomaly is canceled _ la _ green - schwarz .
the model is only string - inspired and is not the low - energy sector of some brane construction .
the reason of this choice rests in our curiosity to explore the phenomenology of these models keeping a high degree of flexibility , while avoiding the intricacies and uncertainties connected with a string theory construction . for previous work along these lines we refer to @xcite-@xcite . in this work
we perform a consistency check of our model @xcite by evaluating the thermal relic density and comparing it against the wmap data .
the cancelation of the @xmath0 anomaly in our model requires the introduction of an extra complex scalar field whose supersymmetric partner is called the stckelino .
we will see in the following that if the latter is the lightest supersymmetric particle ( lsp ) its interactions are such to define it as an xwimp ( i.e. a weakly interacting particle with couplings at least one order of magnitude less than the standard weak interactions ) : in fact it will also turn out to be a cold relic .
if the stckelino is the lightest supersymmetric particle ( lsp ) , its relic density turns out to be too high with respect to the experimental data .
this is why , following @xcite , we favor a next to lightest supersymmetric particle ( nlsp ) with a mass close to the lsp .
we show that an interesting scenario arises also for three particle coannihilation processes . in these two cases ,
the model is consistent with the experimental data .
moreover in the three particle case we find configurations in which the lsp and the nlsp do not need to be nearly degenerate in mass .
in this case the mass gap between the two can be of the order of @xmath5 .
this is the plan of the paper : in section 2 we describe our model . in section 3 and 4
we find the lsp and study the stckelino interactions .
finally in section 5 we compute the relic density .
section 6 is a summary of our results .
in this section we briefly discuss our theoretical framework .
we assume an extension of the mssm with an additional abelian vector multiplet @xmath6 with arbitrary charges .
the anomalies are canceled with the green - schwarz ( gs ) mechanism and with the generalized chern - simons ( gcs ) terms .
all the details can be found in @xcite .
all the mssm fields are charged under the additional vector multiplet @xmath6 , with charges which are given in table [ qtable ] , where @xmath7 are the left handed quarks and leptons respectively while @xmath8 are the right handed up and down quarks and the electrically charged leptons .
the superscript @xmath9 stands for charge conjugation .
the index @xmath10 denotes the three different families .
@xmath11 are the two higgs scalars .
.charge assignment .
[ cols="^,^,^,^,^",options="header " , ] the key feature of this model is the mechanism of anomaly cancelation .
as it is well known , the mssm is anomaly free . in our mssm extension
all the anomalies that involve only the @xmath12 , @xmath13 and @xmath14 factors vanish identically .
however , triangles with @xmath0 in the external legs in general are potentially anomalous .
these anomalies are u(1)-u(1)-u(1) : & & ^(0 ) + u(1)-u(1)_y - u(1)_y : & & ^(1 ) + u(1)-su(2)-su(2 ) : & & ^(2 ) + u(1)-su(3)-su(3 ) : & & ^(3 ) + u(1)-u(1)-u(1)_y : & & ^(4 ) all the remaining anomalies that involve @xmath0s vanish identically due to group theoretical arguments ( see chapter 22 of @xcite ) .
consistency of the model is achieved by the contribution of a field @xmath15 and its appropriate couplings to the anomalous @xmath0 .
the lagrangian written in terms of superfields is @xcite _s = .
@xmath16 ^ 2 | _ - \ { @xmath17_^2 + h.c . } [ laxion ] where the index @xmath18 runs over the @xmath19 and @xmath12 gauge groups respectively .
the multiplet is a chiral superfield s = s+ i2 _ s + ^2 f_s - i ^|_s + ^2 ||^__s - s [ smult ] the lowest component of @xmath15 is a complex scalar field @xmath20 . in our scenario the scalar @xmath21 is eaten up in the mechanism to give mass to the gauge field .
on the other end , the scalar @xmath22 is the dilaton of string theory and its value must be determined somehow
. we will not investigate the way in which this happens , but we will just retain the final result : @xmath22 drops out of our effective lagrangian . in the string literature
( see for instance @xcite-@xcite ) the fields @xmath21 , @xmath22 and @xmath23 are respectively known as axion , saxion and axino . in this paper , to avoid confusion with the much better known qcd axion / axino system ( see for instance @xcite-@xcite ) , we will adopt the convention of @xcite and , from now on , we will refer to @xmath24 as the scalar and to @xmath23 as the stckelino .
the multiplet @xmath15 transforms under @xmath0 as v^(0 ) & & v^(0 ) + i @xmath25 + s & & s - 4 i b_3 [ u1 ] [ u1trans]where @xmath26 is a constant related to the @xmath27 mass . in our model there are two mechanisms that give mass to the gauge bosons : ( i ) the mechanism and ( ii ) the higgs mechanism . in this extension of the mssm , the mass terms for the gauge fields for @xmath28 . ]
are given by _m = @xmath29 m^2 @xmath30 with @xmath31 being the gauge boson mass matrix m^2= @xmath32 [ bosonmasses ] where @xmath33 is the mass parameter for the anomalous @xmath3 and it is assumed to be in the tev range .
the lower dots denote the obvious terms under symmetrization .
after diagonalization , we obtain the eigenstates a_&= & [ photon ] + z_0 & = & [ z0 ] + z_&= & v^(0)_[zprime ] and the corresponding masses m^2_&=&0 + m^2_z_0 & = & @xmath34 v^2[z0mass ] + m^2_z & = & m_v^(0)^2 [ zpmass ] finally the rotation matrix from the hypercharge to the photon basis is @xmath35 & = & o_ij @xmath36 [ oij ] = @xmath37 @xmath36 where @xmath38 .
we now give the expansion of the lagrangian piece @xmath39 defined in ( [ laxion ] ) in component fields only for the part that is needed in the following sections . using the wess - zumino gauge we get _&= & _ s ^__s -2 b_3 _ s ^(0 ) - _ a=0 ^
2b^(a)_2 @xmath40 _ s + & & - b^(4)_2 @xmath41 + h.c .
[ axinolagr]as it was pointed out in @xcite , the mechanism is not enough to cancel all the anomalies .
mixed anomalies between anomalous and non - anomalous factors require an additional mechanism to ensure consistency of the model : non - gauge invariant gcs terms must be added . in our case ,
the gcs terms have the form @xcite _gcs & = & - d_4 @xmath42 _
+ + & & + d_5 @xmath43 _ + + & & + d_6 _ [ gcs_1 ] where @xmath44 refers to non - abelian completion terms .
the @xmath45 constants in ( [ laxion ] ) and the @xmath46 constants in ( [ gcs_1 ] ) are fixed by the anomaly cancelation procedure ( for details see @xcite ) . for a symmetric distribution of the anomaly , we have & & b^(0)_2 b_3 = - b^(1)_2 b_3 = - b^(2)_2 b_3 = - b^(4)_2 b_3 = - + & & d_4 = - d_5 = d_6= [ bsds ] it is worth noting that the gcs coefficients @xmath47 are fully determined in terms of the @xmath48 s by the gauge invariance , while the @xmath49 s depend only on the free parameter @xmath26 , which is related to the mass of the anomalous @xmath3 .
the soft breaking sector of the model is given by _soft = _soft^mssm- @xmath50 - @xmath51 [ lsoft ] where @xmath52 is the usual soft susy breaking lagrangian while @xmath53 is the gaugino of the added @xmath0 and @xmath23 is the stckelino .
the stckelino soft mass term deserves some comment : from @xcite we know that a fermionic mass term for a chiral multiplet is not allowed in presence of yukawa interactions in which this chiral multiplet is involved .
but in the classical lagrangian the multiplet can not contribute to superpotential terms given that the gauge invariance given from our u(1) symmetry ( [ u1trans ] ) requires non - holomorphicity in the chiral fields .
in fact in our model both the stckelino and the scalar @xmath21 couple only through gs interactions .
it is worth noting that a mass term for the scalar @xmath21 is instead not allowed since it transforms non trivially under the anomalous @xmath0 gauge transformation ( [ u1trans ] ) .
assuming the conservation of r - parity the lsp is a good weak interacting massive particle ( wimp ) dark matter candidate . as in the mssm
the lsp is given by a linear combination of fields in the neutralino sector .
the general form of the neutralino mass matrix is given in @xcite .
written in the interaction eigenstate basis @xmath54 it is a six - by - six matrix . from the point of view of the strength of the interactions the two extra states are not on the same footing with respect to the standard ones .
the stckelino and the extra gaugino @xmath53 dubbed primeino are in fact extremely weak interacting massive particle ( xwimp ) .
thus we are interested in situations in which the extremely weak sector is decoupled from the standard one and the lsp belongs to this sector .
this can be achieved at tree level with the choice = = 0 [ qhuchoice ] the neutralino mass matrix @xmath55 becomes _ n = @xmath56 [ massmatrix ] where @xmath57 are the soft masses coming from the soft breaking terms ( [ lsoft ] ) while @xmath58 is given in ( [ bosonmasses ] ) .
it is worth noting that the d terms and kinetic mixing terms can be neglected in the tree - level computations of the eigenvalues and eigenstates .
moreover , we make the assumption that m_0m_s , m_v^(0 ) [ axinodecoupling ] so that the stckelino is the lsp .
this assumption is motivated by the interaction strengths of the two extra states : the stckelino interacts via the vertex shown in fig .
[ upint]b which can easily be seen ( from ( [ axinolagr ] ) ) to be proportional to the coefficient @xmath59 of ( [ bsds ] ) which , in turn , is inversely proportional to @xmath26 that is the @xmath60 mass given that @xmath61 .
given these considerations the vertex shown in fig .
[ upint]b is then of order @xmath62 .
the primeino interacts via the vertex in fig .
[ upint]a which is of order @xmath63 that is the standard strength of weak interactions . assuming the two decoupling relations ( [ qhuchoice ] ) and ( [ axinodecoupling ] ) we will see in the following sections that the dominant contribution in the ( co)annihilation processes is that of the primeino , which is of the type of a standard gaugino interaction .
the stckelino interactions can be read off from the interaction lagrangian ( [ axinolagr ] ) .
the relevant stckelino - mssm neutralino interaction term , written in terms of four components majorana spinors are in the weyl representation .
we use capital letters for four components spinors and lower case letters for two components spinors .
] , is given by : = i g_1 ^ 2 b_2^(1 ) |^(1 ) _ 5 [ ^,^](_v^(1 ) _ ) _ s + i g_2 ^ 2 b_2^(2 ) |^(2)_3 _ 5 [ ^,^](_v^(2)_3 ) _ s [ lint4d ] where the @xmath49 coefficients are given in ( [ bsds ] ) .
the related interaction vertex feynman rule is c^(a ) _ 5 [ ^ , ^]i k _ [ lpg ] where @xmath64 is the momentum of the outgoing vector and the @xmath65 s are c^(1)=g_1 ^ 2 b_2^(1 ) + c^(2)= g_2 ^ 2 b_2^(2 ) [ vertice ] the interaction lagrangian ( [ lint4d ] ) expressed in the mass eigenstates basis is &= & i _
j @xmath66 _ 5 [ ^,^](_a _ ) _ s + + & & i _
j @xmath67 _ 5 [ ^,^ ] ( _ _ ) _ s [ lintphys ] where @xmath68 is a generic neutralino and @xmath69 is the matrix that diagonalizes ( [ massmatrix ] ) .
we remind that the stckelino @xmath70 is directly a mass eigenstate because of the decoupling ( [ axinodecoupling ] ) .
we stress that these are only interactions between the stckelino and the mssm neutralinos .
all the stuckelino interactions in ( [ axinolagr ] ) include also analogous interactions involving the charged wino or the primeino .
the factors in ( [ vertice ] ) are of naive dimension higher than four .
they then contain the parameters @xmath49 which are inversely proportional to the mass of the @xmath27 ( see ( [ bsds ] ) ) .
given these interactions , our stckelino will be an extremely weak interacting particle , according to the definition we gave in the introduction . the simplifying assumption ( [ qhuchoice ] )
has now decoupled our stckelino and primeino from the standard mssm sectors and , at tree level , the relevant diagrams are given in figs .
[ figura1 ] and [ figuracoann ] .
we can now give a rough estimate of the effective interaction , comparing with the standard fermi coupling , @xmath71 , of weak interactions : given a mass of the @xmath27 boson of the order of the tev , the effective coupling in fig .
[ figura1 ] is @xmath72 and that in fig .
[ figuracoann ] is @xmath73 .
let us go back now to fig .
[ figura1 ] , where we denoted with @xmath74 and @xmath75 the incoming momenta of the stckelinos while @xmath76 and @xmath77 are the two outcoming momenta of the gauge bosons in the final state .
we will concentrate on the case with two photons in the final state . in this case
the result for the differential cross section is given by = _
i , j=1 ^ 2_i ^*_j [ axinoaxinodiffcs ] where @xmath78 and @xmath79 are the energies of the two outcoming photons .
each amplitude @xmath80 is proportional to the relative coefficient @xmath81 whose generic form is given in ( [ vertice ] ) . in our scenario
the stckelino annihilations alone , being the cross section ( [ axinoaxinodiffcs ] ) extremely weak , can not give a relic density in the wmap preferred range .
thus , in the scenario of an xwimp stckelino , we are forced to consider coannihilations between the stckelino and the nlsp @xcite .
several scenarios can be considered for the nlsp .
we can split them into two major classes : one in which the nlsp is either a pure bino or a pure wino , and thus a coannihilation with a third mssm particle is needed in order to recover the wmap result , and one in which the nlsp is a generic mssm neutralino with a non - negligible bino and/or wino component . in both classes in order to have effective coannihilations , the nlsp ( and eventually the other mssm particle involved in the coannihilation process ) must be almost degenerate in mass . furthermore , in the most common applications , the cross sections for the annihilations of the the lsp and the nlsp and that of the coannihilations between the lsp and nslp are roughly of the same order of magnitude . in our case this last condition will be stretched and the cross section we are going to discuss will differ for some orders of magnitude .
this situation is not completely new in literature : already in @xcite in which the nlsp is the stop these differences are of order @xmath82 . in @xcite differences of order @xmath83 are considered , while in @xcite @xmath84 differences are found .
but what assures us that the two species are still in thermal equilibrium and do not decouple separately ?
the existence of interactions of the type @xmath85 is then required to keep the lsp in equilibrium @xcite . @xmath86 are two mssm particles .
furthermore @xmath87 better be relativistic so that its abundance is much larger of any cold particle to foster the above reaction .
the above reaction will then keep the lsp at equilibrium and the formalism of coannihilation can be safely employed . as we will see in the next section , all of these requirements are met in our scenario . as a first example we then consider a pure bino as the nlsp .
the allowed coannihilation processes with the stckelino are those which involve an exchange of a photon or a @xmath88 in the intermediate state and with a sm fermion - antifermion pair , higgses and @xmath89 s in the final state .
the diagram with the fermion - antifermion in the final state is sketched in fig .
[ figuracoann ] .
the differential cross section in the center of mass frame has the following general form ||^2 where @xmath24 is the usual mandelstam variable and @xmath90 is the spatial momentum of the outgoing ( incoming ) particles . on dimensional ground
@xmath91 has at least a linear dependence on @xmath92 and this implies that the dominant contribution comes from the diagram with the sm fermion - antifermion pair @xmath93 and @xmath94 in the final state : _
s ^(a)f |f pair via the exchange of a photon or a @xmath88 . ]
the resulting differential cross section , computed in the center of mass frame , is = _
f c_f ( _ ^2+_z_0 ^ 2+^*__z_0+__z_0^ * ) [ sezcoan ] where the sum is extended to all the sm fermions ( with mass @xmath95 ) while @xmath96 is a color factor
. details of the amplitude computation can be found in appendix [ nnamp ] .
in this section we compute the relic density of the stckelino . the case of the stckelino as a cold dark matter candidate has been studied for the first time in @xcite . as we said in the previous section we study two scenarios : the first in which the stckelino coannihilates with only one nlsp degenerate in mass ( a generic mssm neutralino ) , the second in which there is an additional supersymmetric particle ( either a chargino or a stau ) involved in the coannihilation process with the stckelino and the nlsp . in the following we will be largely following @xcite . since this is a first study and given also the simplifying choice ( [ qhuchoice ] ) we will defer a complete analysis to a future work and will content ourselves with showing that our model can accommodate for wmap data .
then , following this philosophy we will not solve the boltzmann equation numerically but , in agreement with @xcite , we will argue that , if the ratio between the thermally averaged cross sections of the coannihilation of fig . [ figuracoann ] and
that of a typical neutralino annihilation is much less than one , a relic density satisfying the wmap requirements can be found .
just to fix the notation we briefly review the relic density computation for @xmath97 interacting species @xcite .
the requirements discussed in the previous section are met by our model given the lagrangian ( [ axinolagr ] ) and the condition of ( [ qhuchoice ] ) . in this case
all channels are open to interactions and the stckelino has an interaction with the photon and the bino of strength @xmath98 ) : taking also @xmath99 susy terms of the type @xmath100 can be safely added to the lagrangian .
this term induces a vertex between the stckelino , neutrino and higgs field which can be a viable candidate to keep the stckelino in thermal equilibrium . ] . the boltzmann equation for @xmath97 particle species
is given by : = -3hn-_i , j=1^n _
ij v_ij ( n_i n_j - n^eq_i n^eq_j ) [ boltz3 ] where @xmath101 denotes the number density per unit of comoving volume of the species @xmath102 ( @xmath103 refers to the lsp , @xmath104 refers to the nlsp , and so on ) , @xmath105 , @xmath106 is the hubble constant , @xmath107 is the annihilation cross section between a species @xmath108 and a species @xmath109 , @xmath110 is the modulus of the relative velocity while @xmath111 is the equilibrium number density of the species @xmath108 given by : @xmath112 where @xmath113 are the internal degrees of freedom , @xmath114 .
@xmath115 is known once ( [ boltz3 ] ) is solved .
a preliminary estimate of @xmath116 can be obtained by solving the equation @xmath117 which can be obtained from the decoupling condition .
this can be done using the estimate sweeping the temperatures range @xmath118 gev .
we have then fit the results to recover a function in good agreement with ( [ decoupleq ] ) . ]
@xmath119 where the effective coupling @xmath120 can be the @xmath121 introduced in section [ axino interactions ] . in the mass range
@xmath122 gev , by plugging in ( [ decoupleq ] ) @xmath123 we would get @xmath124 ( if we would take our stckelino as a separate species , that is we would use @xmath125 , we would get those values divided by half ) .
( [ boltz3 ] ) can be rewritten in a useful way by defining the thermal average of the effective cross section _ eff v _
i , j=1^n_ij v_ij [ sigmaeff ] obtaining = -3hn-_eff v ( n^2 - ( n^eq)^2 ) [ boltz ] where @xmath126 . it is sensible to use these approximations when the lsp is kept in equilibrium by a relativistic particle in the thermal background as discussed in the previous section . given the typical values of @xmath116 discussed above , the ratio between the number density per comoving volume of this relativistic species and that of a cold relic is @xmath127 , which is sufficient to keep the stckelino coupled until the end of coannihilations .
as a rule of thumb @xcite a first order estimate of the relic density is given by @xmath128 to give a rough idea of the role played by coannihilations we plotted in fig .
[ dmestimate ] the relic density estimate ( [ oh2vssigmaeff ] ) induced by an electro - weak cross section ( ( [ decoupleq ] ) with @xmath123 ) , a stckelino cross section ( ( [ decoupleq ] ) with @xmath129 ) and two coannihilations cross sections estimations .
we can see that the stckelino annihilations can not give a relic density in the wmap data range , while coannihilations can do it .
moreover we can see that the effect of coannihilations on the mssm is to decrease the efficiency of the mssm annihilations ( while they increase the stckelino one ) and to increase the lsp mass value ( according to an increasing mass gap ) in order to agree with wmap data .
we stress that fig .
[ dmestimate ] does not take into account several mssm parameters such as the sfermion masses , neutralino composition etc .
, so it is just a rough estimate that , however , clarifies the role played by the coannihilations .
[ 0cm][0cm]=1.mm ( blue ) and
a coannihilating cross section with @xmath130 ( dashed blue ) .
the wmap data range is the gray strip .
the figure on the right is a zoom of that on the left around the region of interest.,title="fig : " ] ( blue ) and a coannihilating cross section with @xmath130 ( dashed blue ) .
the wmap data range is the gray strip .
the figure on the right is a zoom of that on the left around the region of interest.,title="fig : " ] in the following we will now give a better estimation of @xmath131 in the two cases @xmath132 and @xmath133 using what we have learnt in our scenario and the coannihilation cross section of fig .
[ figuracoann ] presented in detail in appendix a. * @xmath132 case .
assuming that the relative velocities are all equal @xmath134 we get : @xmath135 where @xmath136 .
the first term in the numerator can be neglected because the stckelino annihilation cross section is suppressed by a factor @xmath137 with respect to the mssm neutralino annihilations ( see the previous section ) and thus @xmath138 .
the second term involves the coannihilation cross section .
let us consider the case in which the nlsp is a generic mssm neutralino ( a linear combination of @xmath139 ) with a non - vanishing bino or wino components . as we saw in the previous section
each amplitude is generically proportional to @xmath140 with @xmath141 . without loss of generality
we consider the diagram which involves the bino component @xmath142 and a photon exchange in the intermediate channel , i.e the @xmath143 amplitude in ( [ sezcoan ] ) .
we get c^2_= ( c^(1 ) _ w)^2= 2 ( b_2^(1))^2 g_1 ^ 4 ^2_w [ c ] from the expression of the mixed @xmath144 anomaly ( see @xcite ) and from the ( [ bsds ] ) we have the following relation b_2^(1)= where @xmath145 . with the assumption @xmath146
tev as in @xcite we finally get 5.76 10 ^ -12 ( 3 g_0 q_q + g_0 q_l)^2 gev^-2 where @xmath147 is the electric charge .
we get similar expressions for the other three terms in ( [ sezcoan ] ) .
this result has to be compared to the typical weak cross section @xmath148 .
as long as the charges and the coupling constant of the extra @xmath3 satisfy the perturbative requirement g_0 ^ 2(3 q_q + q_l)^2<16 [ chargebound ] the following upper bound is satisfied : @xmath149 in the case of a pure bino , while @xmath150 in the case of a pure wino . accordingly to eqs .
( [ oh2vssigmaeff ] ) , ( [ sigma2eff ] ) , ( [ binosigma12neg ] ) and ( [ winosigma12neg ] ) the relic density gets rescaled as @xcite @xmath151^(2 ) @xmath152 ^ 2 @xmath151^(1 ) [ 2rescaling ] we performed a random sampling of mssm models in which the nlsp is a pure bino or a mixed bino - higgsino ( the case of a pure wino falls back into the @xmath133 case due to the wino - chargino mass degeneracy ) and we computed the relic density in presence of coannihilations using the darksusy package @xcite .
these two situations are easily realized in some corners of the msugra parameter space .
thus in our scan we assumed this scenario in order to fix the pattern of the supersymmetry breaking parameters at weak scale .
we emphasize here that this choice is completely arbitrary , and it is assumed only for simplicity , since in our model @xcite the supersymmetry breaking mechanism is not specified . in the former case there is no model which satisfies the wmap constraints @xcite : 0.0913h^20.1285 since the annihilation cross section of a pure bino is too low and the rescaling ( [ 2rescaling ] ) is not enough to get the right relic density . in the latter case
the higgsino component tends to increase the annihilation cross section and thus we find models which satisfy the wmap constraints .
the results are summarized in fig .
[ dmbino - higgsino ] for @xmath153 and @xmath154 .
right panel : @xmath154.,title="fig : " ] .
right panel : @xmath154.,title="fig : " ] + in order to fulfill the wmap data ( red ( darker ) points in the plot ( [ dmbino - higgsino ] ) ) the stckelino mass must be in the range @xmath155 in the limit @xmath156 , where the lowest bound is given by the current experimental constraints @xcite . *
@xmath133 case .
this is the case in which there is a third mssm particle almost degenerate in mass with the lsp and the nlsp .
typical situations of this kind arise when the nlsp and the next to next to lightest supersymmetric particle ( nnlsp ) are respectively the bino and the stau or the wino and the lightest chargino .
expanding in an explicit way all the terms in the sum ( [ sigmaeff ] ) we get : _
^(3 ) v & = & _ 11 v _ 1 ^ 2 + _ 12 v _ 1 _ 2 + _ 13 v _ 1 _ 3 + + & & _ 21 v _ 2 _ 1 + _ 22 v _ 2 ^ 2 + _ 23 v _ 2 _ 3 + + & & _ 31 v _ 3 _ 1 + _ 32 v _ 3 _ 2 + _ 33 v _
3 ^ 2 + & = & [ _ 11 v ( _ 1)^2 + _ 12 v _ 1 _ 2 + _ 13 v _ 1 _ 3 + + & & _ 21 v _ 2 _ 1 + _ 22 v ( _ 2)^2 + _ 23 v _ 2 _ 3 + + & & _ 31 v _ 3 _ 1 + _ 32 v _ 3 _ 2 + _ 33 v ( _ 3)^2 ] + & & where in the last line we have neglected the terms @xmath157 , @xmath158 and @xmath159 since these are the thermal averaged cross sections which involve the stckelino . by introducing a new set of variables defined by q_i== ( 1 + _ i)^3/2 e^-x_f _
i i=2,3 where @xmath113 are the internal degrees of freedom of the particle species , @xmath115 and @xmath160 , we obtain _
^(3 ) v & & under the assumption @xmath161 , @xmath162 we finally get _
^(3 ) v _ v [ sigma3eff ] where _ v = _ 22 v + 2 _ 23 v + @xmath163 ^ 2 _ 33 v in order to compute the rescaling factor between the relic density of our model and the mssm relic density we have to express @xmath164 in terms of a two coannihilating species effective cross section .
this is given by _
^(2 ) v & = & + & = & + & = & + & = & where q_23&=&_3/_2= @xmath165 ^ 3/2 e^-x_f + & & since @xmath161 and @xmath166 then @xmath167 .
we remind the reader that the values of @xmath168 , @xmath169 and @xmath170 are different with respect to those in the former case since now there are only two species in the thermal bath .
we then find _
^(2 ) v & & + & & and inserting back this relation into ( [ sigma3eff ] ) we obtain _
^(3 ) v @xmath171 ^ 2 _ ^(2 ) v the rescaling factor between the three and two particle species relic density is given by the following relation @xmath151^(3 ) @xmath172 ^ 2 @xmath151^(2)[rescrelic2to3 ] we performed a random sampling of mssm models with bino - stau and wino - chargino coannihilations .
the first situation is realized in some corners of the msugra parameter space while the second situation is naturally realized in anomaly mediated supersymmetry breaking scenarios . for each model we computed the relic density @xmath173 for the two coannihilating species with the darksusy package @xcite .
we finally computed @xmath174 using ( [ rescrelic2to3 ] ) .
the bino - stau models which satisfy the wmap constraints have a stckelino mass in the range @xmath175 in the limit @xmath156 .
as the mass gap increases the number of allowed models drastically decreases and eventually vanishes for @xmath176 . in the wino - chargino case ,
models which satisfy the wmap constraints are shown in fig .
[ dmwino - chargino ] for four reference values of @xmath177 .
the space of parameters with @xmath178 and a stckelino mass @xmath179 gev is favored while as the mass gap increases lower stckelino masses become favored , e.g. @xmath180 gev ( @xmath181 ) .
upper right panel : @xmath182 .
lower left panel : @xmath154 .
lower right panel : @xmath153.,title="fig : " ] .
upper right panel : @xmath182 .
lower left panel : @xmath154 .
lower right panel : @xmath153.,title="fig : " ] .
upper right panel : @xmath182 .
lower left panel : @xmath154 .
lower right panel : @xmath153.,title="fig : " ] .
upper right panel : @xmath182 .
lower left panel : @xmath154 .
lower right panel : @xmath153.,title="fig : " ]
we studied a possible dark matter candidate in the framework of our minimal anomalous @xmath0 extension of the mssm @xcite . in the decoupling limit ( [ qhuchoice ] ) and under the assumption
@xmath183 the stckelino turns out to be the lsp .
being an xwimp the stckelino annihilation cross section is suppressed with respect to the typical weak interaction cross sections .
this implies that in order to satisfy the wmap constraints on the relic density we must have at least a nlsp almost degenerate in mass with the stckelino .
we considered the case with two and three coannihilating particles and we found some configuration which satisfies the wmap constraints .
the results depend on the mass gap between the stckelino and the nlsp . in the exact degeneracy limit @xmath156 the allowed models have a stckelino mass in the range @xmath184 for the bino - higgsino coannihilation case while @xmath185 for the wino - chargino coannihilation case .
when the mass gap is @xmath186 the allowed models are those with wino - chargino coannihilations and a stckelino mass of @xmath187 gev .
finally let us comment on the differences between our scenario and that studied in the work @xcite . in our framework the @xmath0 does not arise from a hidden sector and thus all the mssm fields can be charged under this extra abelian gauge group .
this is the most relevant feature which could also be detected experimentally ( see for example @xcite ) .
moreover , in our scenario the stckelino interactions are suppressed with respect to the weak interactions due to the gs couplings while in @xcite the mechanism to suppress the couplings and give an xwimp is provided by the kinetic mixing between the @xmath0 and @xmath14 .
in this appendix we give some details of the amplitude computation for the process @xmath188 , = - i k^|v_s _ 5 @xmath189 u_1 @xmath190 where @xmath191 denote the electric charges , @xmath192 and @xmath193 are the vectorial and axial couplings with @xmath88 , @xmath194 , @xmath195 while @xmath196 is the momentum of the intermediate gauge boson .
the corresponding square modulus is ||^2 & = & - 64 @xmath197 [ nunuampl ] with t_v & = & m_f^4 ( p__1 p_s ) + m_1 m_s + + & & -(p_f p_|f ) + m_f^2 + t_a & = & m_f^2-m_1 m_s @xmath198 + + & & -(p_f p_|f ) where @xmath199 , @xmath200 , @xmath92 and @xmath201 are the bino , stckelino and sm fermions 4-momenta respectively .
writing all the momenta in function of @xmath24 and integrating over the solid angle we get & & = c_f @xmath202 ^ 2 + & & + & & a. r. would like to thank prof .
michael green and damtp for hospitality and the marie curie research training network `` superstring theory '' , contract mrtn - ct-2004 - 512194 and contract esf jd164 for financial support during the completion of this paper .
f.fucito wants to thank n.fornengo and p.ullio for discussions .
p. candelas , g. t. horowitz , a. strominger and e. witten , nucl . phys .
b * 258 * ( 1985 ) 46 .
langacker , arxiv:0801.1345 [ hep - ph ] .
p. anastasopoulos , f. fucito , a. lionetto , g. pradisi , a. racioppi and y. s. stanev , phys . rev .
d * 78 * ( 2008 ) 085014 [ arxiv:0804.1156 [ hep - th ] ] .
p. anastasopoulos , m. bianchi , e. dudas and e. kiritsis , jhep * 0611 * , 057 ( 2006 ) [ arxiv : hep - th/0605225 ] .
j. de rydt , j. rosseel , t. t. schmidt , a. van proeyen and m. zagermann , class .
* 24 * ( 2007 ) 5201 [ arxiv:0705.4216 [ hep - th ] ] .
k. rajagopal , m. s. turner and f. wilczek , nucl .
b * 358 * ( 1991 ) 447 .
f. d. steffen , eur .
j. c * 59 * , 557 ( 2009 ) [ arxiv:0811.3347 [ hep - ph ] ] .
a. brandenburg and f. d. steffen , jcap * 0408 * , 008 ( 2004 ) [ arxiv : hep - ph/0405158 ] . l. covi and j. e. kim , arxiv:0902.0769 [ astro-ph.co ] .
l. covi , j. e. kim and l. roszkowski , phys .
* 82 * ( 1999 ) 4180 [ arxiv : hep - ph/9905212 ] .
+ l. covi , h. b. kim , j. e. kim and l. roszkowski , jhep * 0105 * ( 2001 ) 033 [ arxiv : hep - ph/0101009 ] .
b. kors and p. nath , arxiv : hep - ph/0411406 .
l. girardello and m. t. grisaru , nucl .
b * 194 * ( 1982 ) 65 .
k. griest and d. seckel , phys .
d * 43 * ( 1991 ) 3191 .
j. edsjo , m. schelke , p. ullio and p. gondolo , jcap * 0304 * ( 2003 ) 001 [ arxiv : hep - ph/0301106 ] .
j. edsjo and p. gondolo , phys .
d * 56 * ( 1997 ) 1879 [ arxiv : hep - ph/9704361 ] .
g. jungman , m. kamionkowski and k. griest , phys . rept .
* 267 * ( 1996 ) 195 [ arxiv : hep - ph/9506380 ] .
gondolo , j. edsjo , p. ullio , l. bergstrom , m. schelke and e. a. baltz , jcap * 0407 * ( 2004 ) 008 [ arxiv : astro - ph/0406204 ] .
j. dunkley _ et al .
_ [ wmap collaboration ] , arxiv:0803.0586 [ astro - ph ] .
+ e. komatsu _ et al . _
[ wmap collaboration ] , arxiv:0803.0547 [ astro - ph ] . c. amsler
_ et al . _
[ particle data group ] , phys .
b * 667 * ( 2008 ) 1 .
|
we study a possible dark matter candidate in the framework of a minimal anomalous @xmath0 extension of the mssm .
it turns out that in a suitable decoupling limit the stckelino , the fermionic degree of freedom of the multiplet , is the lightest supersymmetric particle ( lsp ) .
we compute the relic density of this particle including coannihilations with the next to lightest supersymmetric particle ( nlsp ) and with the next to next to lightest supersymmetric particle ( nnlsp ) which are assumed almost degenerate in mass .
this assumption is needed in order to satisfy the stringent limits that the wilkinson microwave anisotropy probe ( wmap ) puts on the relic density .
we find that the wmap constraints can be satisifed by different nlsp and nnlsp configurations as a function of the mass gap with the lsp .
these results hold in the parameter space region where the model remains perturbative .
* stckelino dark matter in anomalous @xmath0 models * francesco fucito@xmath1 , andrea lionetto@xmath1 , andrea mammarella@xmath1 , antonio racioppi@xmath2 @xmath1 dipartimento di fisica delluniversit di roma , tor vergata " and i.n.f.n .
- sezione di roma `` tor vergata '' via della ricerca scientifica , 1 - 00133 roma , italy 0.25 cm @xmath2 national institute of chemical physics and biophysics , ravala 10 , tallinn 10143 , estonia
|
flux emergence is one of the key mechanisms in various solar activities .
it is widely accepted that the emerging flux has a form of a twisted flux tube so as not to be collapsed by the convective motions during its ascent in the solar interior .
emerging flux transports magnetic energy and helicity from the convection zone to the atmosphere , which yields active regions including sunspots .
magnetic helicity in the corona stores free energy that can be released in the forms of flares and coronal mass ejections ( cmes ) ( e.g. @xcite ) .
many numerical experiments have revealed the dynamics of the flux emergence .
@xcite carried out two - dimensional magnetohydrodynamic ( mhd ) simulations to study the cross - sectional evolution of the emerging flux tube ( see also @xcite ) .
@xcite calculated the two - dimensional evolution of the undular mode of magnetic buoyancy instability ( parker instability : @xcite ) to reproduce the formation of an @xmath0-shaped coronal loops . @xcite and @xcite gave numerical studies of emerging fluxes from much deeper convection zone ( @xmath1 km ) to the corona .
the three - dimensionality also exerts an influence on emerging process of magnetic flux evolution .
@xcite produced the first three - dimensional work of the parker instability using a magnetic flux sheet and a flux tube .
@xcite compared her numerical results of the twisted tube s emergence with observations of an active region . in this paper , we perform three - dimensional simulations of the twisted emerging flux tube from the uppermost convection zone to the corona . our aim is to study the effect of the initial twist on the emergence process . a series of parametric studies on the flux tube s twist was done by @xcite .
our work is dedicated to further detailed analyses , especially focusing on the effect of the initial twist on the resulting tube s structure ( photospheric lateral expansion and multiple magnetic domes ) and on the consequent coronal magnetic energy . for numerical experiments
, we used the same conditions as those by @xcite ; we calculated ten cases of different twist parameters that cover their three runs . as a result of experiments
, we found that the evolution depends on the initial twist .
when the twist is strong enough , the evolution to the corona reveals two - step way , showing a deceleration and a lateral expansion near the solar surface , although the case with weaker twist spends more time waiting for the secondary emergence to occur @xcite .
this picture of the two - step emergence is highly consistent with recent observations by @xcite , especially its horizontally expanding speed and the rising speed .
in addition to the confirmation of the results by @xcite , it is also found that multiple magnetic domes are built and plasma accumulates in between the domes when the secondary emergence starts . at this moment , the direction of the field lines is almost perpendicular to the alignment of the domes , indicating that the second - step emergence is due to the interchange - mode instability .
if the initial twist is too weak , the tube extends widely near the surface and further evolution never takes place , because the magnetic tension force of the azimuthal component can not hold the tube s coherency .
also , we newly found that the total magnetic energy measured above the surface relies on the initial twist . in the strong twist regime
, the resulting magnetic energy follows the initial relation between the twist and the magnetic energy . in the weak twist regime
, however , the magnetic energy deviates from the initial rule , because the tube with weak twist stores magnetic energy around the photosphere .
the rest of the paper is organized as follows . in section [ sec :
setup ] , we describe the numerical model . the simulation results are shown in section [ sec : results ] .
summary and discussion are given in section [ sec : summary ] and [ sec : discussion ] , respectively .
in numerical simulations , we solve nonlinear , time - dependent , compressible three - dimensional mhd equations . we take a rectangular computation box with three - dimensional cartesian coordinates ( @xmath2 ) , where the @xmath3-coordinate increases upward .
the medium is assumed to be an inviscid perfect gas with a specific heat ratio @xmath4 .
the basic equations in vector form are as follows : @xmath5 @xmath6 @xmath7 @xmath8 -\rho\mbox{\boldmath $ g$}\cdot\mbox{\boldmath $ v$ } = 0,\end{aligned}\ ] ] and @xmath9 @xmath10 @xmath11 where @xmath12 is the internal energy per unit mass , @xmath13 the unit tensor , @xmath14 the boltzmann constant , @xmath15 the mean molecular mass , and @xmath16 the uniform gravitational acceleration .
other symbols have their usual meanings ( @xmath17 is for density , @xmath18 velocity vector , @xmath19 pressure , @xmath20 magnetic field , @xmath21 speed of light , @xmath22 electric field , and @xmath23 temperature ) . to make above equations dimensionless
, we introduce normalizing units of length @xmath24 , velocity @xmath25 , time @xmath26 , and density @xmath27 , where @xmath28 is the pressure scale height , @xmath25 the sound speed , and @xmath27 the density at the photosphere , respectively .
the gas pressure , temperature , magnetic field strength , and energy are normalized by the combinations of the units above , i.e. , @xmath29 , @xmath30 , @xmath31 , and @xmath32 , respectively . the gravity is given as @xmath33 by definition .
for comparison of numerical results with observations , we use @xmath34 , @xmath35 , @xmath36 , and @xmath37 , which are typical values for the solar photosphere .
then , @xmath38 , @xmath39 , @xmath40 , and @xmath41 .
the initial background stratification consists of three regions : an adiabatically stratified convective layer , a cool isothermal photosphere / chromosphere ( afterward , we simply call it photosphere ) , and a hot isothermal corona . the photosphere and the corona
are smoothly connected by the transition region .
we take @xmath42 to be the base height of the photosphere , and the starting height of the transition region and the corona are @xmath43 and @xmath44 , respectively .
the initial temperature of the photosphere and the corona are @xmath45 and @xmath46 , respectively .
the background temperature distribution in the transition region is given by @xmath47 and that of the convection zone being @xmath48 where @xmath49 is the adiabatic temperature gradient ( subscript s is for surrounding distribution ) .
the initial gas pressure and density profiles are defined by solving one - dimensional hydrostatic equation @xmath50 and the equation of state ( [ eq : eos ] ) on the basis of the temperature distribution above .
the initial magnetic flux tube is embedded in the convection zone at @xmath51 .
the longitudinal and azimuthal component of the flux tube are described as follows : for a radial distance from the axis @xmath52^{1/2}$ ] , @xmath53 and @xmath54 where @xmath55 is the tube center , @xmath56 the radius , @xmath57 the twist parameter , and @xmath58 the magnetic field strength at the axis .
we take @xmath59 and @xmath60 , i.e. , these parameters are almost the same as those of @xcite . for pressure balance between the flux tube and the surrounding medium , the gas pressure inside the tube
is obtained as @xmath61 , where @xmath62.\end{aligned}\ ] ] the density inside the tube is also defined as @xmath63 , where @xmath64 and @xmath65 .
that is , the flux tube is most buoyant at the middle of the tube ( @xmath66 ) , and the buoyancy diminishes as @xmath67 increases . here , we investigate ten parameters of @xmath57 , which are @xmath68 , 0.4 , 0.3 , 0.25 , 0.2 , 0.175 , 0.15 , 0.125 , 0.1 , and 0.05 .
the plasma beta ( @xmath69 ) at the tube center is @xmath70 at the initial state .
the initial background stratification ( gas pressure , density , and temperature ) and the magnetic pressure along @xmath71 of the case @xmath72 are indicated in figure [ fig : initial ] .
the simulation domain is taken as @xmath73 , resolved by @xmath74 grids .
the grid spacings for @xmath75 , @xmath76 , and @xmath3 directions are @xmath77 for @xmath78 , and @xmath79 for @xmath80 , respectively . outside this range , the mesh sizes gradually increase .
we assume periodic boundaries for horizontal directions and symmetric for vertical .
a wave - damping region is attached near the top boundary .
we use the modified lax - wendroff scheme version of the cans ( coordinated astronomical numerical software ) code ( see @xcite ) .
figure [ fig : general ] shows the time evolution of the flux tube with the twist @xmath72 . in this figure , we plot the logarithmic field strength @xmath81 and the photospheric magnetogram @xmath82 . in each panel ,
the region @xmath83 and @xmath84 is shown .
initially , the flux tube is embedded at @xmath51 , and is slightly buoyant around the tube center @xmath85 .
the tube rises through the convection zone by magnetic buoyancy , and reaches the surface at @xmath86 ( figure [ fig : general](b ) ) , while the outskirt of the rising portion ( @xmath87 ) begins to sink , because the fluid is drained along the field lines from the apex of the rising tube . due to the isothermal ( i.e. , strongly - subadiabatic ) photosphere ,
the tube is decelerated and expands laterally near the surface to make a `` photospheric tongue '' ( figure [ fig : general](c ) ) .
that is , the convectively stable photosphere inhibits an upward motion of the fluid , and thus , the magnetic field can not penetrate the photosphere only to escape in the horizontal direction .
( it should be noted that the term `` tongue '' here is different from that used in @xcite and @xcite .
the horizontal extension of the magnetic field is also observed in a recent radiative mhd calculation by @xcite .
the lateral expansion speed is @xmath88 , i.e. , a fraction of the photospheric sound speed . at this moment
, the photospheric magnetogram shows a north - south ( @xmath76-directional ) magnetic distribution , and the total field strength is @xmath89 and plasma beta is @xmath90 around the photosphere .
as the magnetic pressure gradient enhances , the second - step emergence takes place . in figure
[ fig : general](d ) , multiple expansions are observed : the rise velocity is about @xmath91 .
this multi - dome structure is more noticeable in a weaker twist case .
figure [ fig : ro - mag](a ) shows the magnetic field structure of the case @xmath92 at @xmath93 .
the corresponding density structure and velocity vectors in the @xmath94 plane , and the field lines are indicated in figure [ fig : ro - mag](b ) . from these figures ,
one can see four magnetic domes are built and the fluid is accumulated between the domes . in this region , the field lines are generally directed in the @xmath76-direction . therefore , this situation can be explained as a consequence of the interchange mode of the magnetic buoyancy instability , i.e. , the wavenumber vector is perpendicular to the field lines .
as time goes on , the flux tube expands both vertically and horizontally , while the photospheric tongue also continues to expand laterally ( figure [ fig : general](e ) ) .
finally , the flux tube makes a single dome of the height @xmath95 on the pancake - like structure at the surface ( figure [ fig : general](f ) ) . at the same time , the sunk part approaches the bottom of the simulation domain ( @xmath96 ) .
overall evolution described above is similar to the observation by @xcite .
they found the lateral expansion with the speed of @xmath97 at the surface before further evolution occurred , which is consistent with our results of @xmath98 .
the gradual rise speed of the secondary emergence was observed to be @xmath99 , which is also consistent with our results of @xmath100 .
figure [ fig : param ] shows the height - time relation of the top of the tube for various twist cases . here , in this figure , we plot the height at the highest portion of the emerging flux tube ( @xmath101 ) .
the evolutions are found to depend on the initial twist , and each line distributes in a continuous fashion .
it can be seen from this figure that almost all the tubes ( @xmath102 ) show the two - step emergence to the corona .
although the rise times within the convection zone are similar to each other , tubes with weaker twists spend more time in the surface waiting for the second - step emergence to be triggered .
the tube with @xmath103 shows only a slight emergence in the atmosphere ( @xmath104 ) . as for the weakest twist case with @xmath105 , further evolution never takes place within the elapse calculated ( failed emergence ) .
these results are consistent with those of @xcite , @xcite , and @xcite .
cross - sections at @xmath66 plane of eight out of ten flux tubes are shown in figure [ fig : cross ] .
these tubes are those who reach @xmath106 , and each figure shows the arrival at that height .
as the initial twist @xmath107 becomes smaller , the lateral expansion at the surface ( tongue - like structure around @xmath108 ) is increasingly remarkable , because the more intense initial twist of the magnetic flux tube yields the stronger azimuthal magnetic tension force , and thus keeps the tube coherent .
therefore , the coronal field intensity also reduces with decreasing @xmath107 , and the coronal structure with a weaker twist is more fragmented compared to the stronger twist cases . in this subsection
, we compare the magnetic field structures of the different emergence cases to study the horizontal expansion and the mechanism of the second - step evolution at the solar surface .
figure [ fig : compare ] shows ( _ top _ ) the cross - sectional configuration of the flux tubes at @xmath66 with velocity vectors , ( _ middle _ ) the horizontal components of forces along the horizontal axis @xmath109 , and ( _ bottom _ ) the vertical components of forces along the vertical axis @xmath71 for cases with @xmath110 ( rapid emergence ) , 0.15 ( slow emergence ) , and 0.05 ( failed emergence ) at the time @xmath111 .
as can be seen from the top figures , the second - step evolution has already begun at this time for the case with @xmath110 , while the vertical expansion can not be seen in @xmath105 case .
it should be noted that the downflow in the uppermost areas in figures [ fig : compare](b ) and ( c ) is a reflected wave from the top boundary .
horizontally , for a strongest twist case ( figure [ fig : compare](d ) ) , an inward magnetic tension force is dominant , and thus the inward total force keeps the tube from a lateral fragmentation . as the twist decreases ,
the magnetic tension reduces so that the total force is outward in a wider range for @xmath105 ( figure [ fig : compare](f ) ) , resulting in the tube s fragmentation and the further expansion never to occur . as for vertical forces ,
magnetic pressure gradient is principal at front of the tube ( figure [ fig : compare](g ) : @xmath112 ) .
however , the total force is about zero within this area .
that is , the second - step expansion is caused by the tube s magnetic pressure , while the tube is almost in a hydrostatic equilibrium with surrounding materials . for a weak twist case ( figure [ fig : compare](i ) ) , the magnetic pressure gradient is much less effective .
therefore , further rise can not occur .
figure [ fig : bvct ] shows the surface magnetogram @xmath82 and the field lines above the surface for the medium twist case ( @xmath92 ) at the time @xmath113 , namely , in the later phase .
we confirm that some undulating field lines connect magnetic patches at the surface , and that , as time goes on , they gradually rise into the corona by forming longer fields . at this time , both near - surface undulating fields and coronal fields are directed almost parallel to the axis of the original flux tube , which is in contrast to the perpendicular fields observed in the earlier phase of the emergence ( see figure [ fig : ro - mag](b ) ) .
this result is reminiscent of the `` sea - serpent '' field lines and the resistive emergence by @xcite .
undulating photospheric fields of a weakly twisted tube are also found by @xcite .
we will discuss this again in section [ sec : discussion ] .
figure [ fig : energy ] shows the initial twist @xmath107 and the total magnetic energy measured above the solar surface @xmath114 when each tube arrives at @xmath106 ( see figure [ fig : cross ] ) .
the magnetic energy above the surface is defined as : @xmath115 when the initial twist is large ( @xmath116 ) , the magnetic energy @xmath117 is found to obey a @xmath118 law . from the field configuration ( [ eq : b1 ] ) and ( [ eq : b2 ] ) , the initial magnetic energy per unit volume can be calculated as @xmath119 that is , the initial tube s magnetic energy depends on @xmath118 . when the twist is strong , the azimuthal magnetic tension force is more effective and the lateral expansion around the photosphere is less efficient that the initial magnetic energy is directly transported into the atmosphere above the surface . as a result of this regime ,
the observed magnetic energy within the atmosphere relies on a @xmath118 function .
contrary to this , when the initial twist is weak ( @xmath120 ) , the consequent magnetic energy deviates from the @xmath118 line and is negatively - correlated with the initial twist .
it is speculated that , for these tubes with weaker twits , it takes a longer time to reach @xmath106 from the surface .
it is because the photospheric field needs more time to satisfy the condition for the second - step emergence ( i.e. @xcite ) , for the weak azimuthal field causes the tube to expand horizontally around the surface .
since the magnetic energy is continuously transported from below and the horizontal expanding velocity is almost the same for @xmath120 cases ( @xmath121 ) , the magnetic energy of the photospheric tongue is expected to depend on the time lag between the tube s arrival at the photosphere and at @xmath106 .
it can be concluded that the magnetic energy in the atmosphere correlates critically with the initial twist . in the strong twist regime
, the energy - twist relation follows the initial @xmath118 rule , because the tightly twisted tube does not exhibit a significant expansion near the photosphere . as for the weak twist cases , they depend on the time lag between reaching the photosphere and the corona .
it is because the weaker twist tube takes more time to rise further , and , therefore , more magnetic energy is stored in the photospheric tongue .
in this paper , we carried out three - dimensional mhd simulations to investigate the effect of the initial twist on the flux tube evolution . here , we summarize the results : * initially , the flux tube rises through the convection zone due to its magnetic buoyancy . reaching the surface , the tube expands laterally to make a `` photospheric tongue . ''
the secondary emergence occurs after sufficient flux accumulates within the photosphere . due to the interchange mode instability ,
the tube builds multiple domes above the surface , between which the fluids piles up .
finally , the flux tube arrives at @xmath95 as a single dome .
the overall emergence is consistent with the recent observations ( e.g. @xcite ) .
* we run ten twist cases to investigate the effect of the initial twist .
nine out of ten reach the coronal height ( @xmath122 ) showing two - step emergence , while the weakest twist case fails to rise further above the surface ( @xmath105 ) .
in the two - step emergence regime , the rise time becomes shorter with increasing initial twist , which is consistent with the previous calculations by @xcite .
the photospheric tongue is more noticeable in weaker twist case .
* we study the force components at the solar surface for different twist cases at the time @xmath111 .
the stronger the initial twist is , the larger the inward magnetic tension is , resulting the tube keeps its coherency . in the weak twist case ,
the magnetic tension is much less effective , causing the tube distorted . at the same time
, the strong twist tube rises further into the atmosphere mainly by the magnetic pressure gradient .
* we found that the photospheric fields of the middle twist case ( @xmath92 ) undulate in the later phase of the emergence .
the field lines gradually rise into the corona as longer loops .
the photospheric and coronal fields are almost parallel to the axis of the initial flux tube .
these features remind us of the resistive emergence model by @xcite . *
we measure the magnetic energy @xmath117 above the surface .
the energy plot follows the initial @xmath118 law when the twist is strong ( @xmath116 ) , while , for weaker twist cases ( @xmath120 ) , the energy depends on the time difference between reaching the surface and the corona .
that is , weakly twisted tube takes more time for magnetic flux to accumulate near the surface and the secondary instability to be triggered .
in section [ sec : results ] , we showed the time - evolution of the twisted flux tube .
when the second - step emergence starts , multiple domes are observed above the surface and fluid is trapped between the expanding magnetic structures ( see figure [ fig : ro - mag ] ) . at this time
, field lines are directed perpendicular to the alignment of the magnetic domes . for middle twist tube
, we also found undulating fields near the surface , emerging into the corona ( see figure [ fig : bvct ] ) . in this section , we discuss these features in connection with future observations . in section [ sec : overview ] , we saw that , as the twist decreases , the interchange mode instability becomes more noticeable .
however , this is contrary to the expectation that the azimuthal field should be less pronounced in a weaker twist case
. it may be because , in a weak twist case , the tube extends laterally near the photosphere and thus the twist increases .
as the tube develops the interchange instability , field lines perpendicular to the alignment of the magnetic domes become more pronounced ( see sections [ sec : param ] and [ sec : compare ] ) .
multiple magnetic structures and the density accumulation between them are also found in previous observations and calculations .
@xcite found that photospheric fields are undulating at its earlier phase of the flux emergence event , and proposed a resistive emergence model that undulating multiple loops reconnect with each other to make larger coronal fields .
@xcite carried out two - dimensional mhd simulation to study the evolution of the serpentine magnetic loops ( resistive emergence model ) , finding that density accumulates in between the magnetic loops ; their elongated vertical plasma structures are similar to our results ( for three - dimensional study , see @xcite ) . however ,
in our model , field lines are directed almost perpendicular to the alignment of the domes , which is against the observations of the undular field lines ( e.g. @xcite ) .
the direction of the field lines are the consequence of the initial tube s twist .
therefore , the difference between the present calculations and the observations indicates that the actual twist of the flux tube beneath the surface may be much weaker than those assumed in our models ( e.g. @xmath92 for figure [ fig : ro - mag ] ) . on the other hand , flux tube with insufficient twist
was found to fail to rise through the convection zone @xcite .
it is because the weak azimuthal field of the flux tube can not hold its coherency during its ascent within the solar interior .
therefore , one of the important problems to be solved is the emergence of the flux tube with much weaker twist ( @xmath123 ) .
we also found that the medium twist tube ( @xmath92 ) reveals the undulating fields at the surface , which gradually rise into the upper atmosphere as longer coronal loops ( section [ fig : bvct ] ) .
these fields are directed parallel to the main axis of the initial flux tube .
this picture seems well accorded with the resistive emergence model .
however , it is in the later phase that this undulatory evolution is observed , and , in the earlier phase , the field lines are perpendicular to the original tube s axis ( see figure [ fig : ro - mag ] ) . therefore , the reproduction of the undulating fields parallel to the original axis @xcite is not achieved .
recently , @xcite have conducted a radiative mhd simulation of the formation of an active region .
they showed that the rising flux tube flattens to make a pancake - like structure near the surface , and that the convective flows create serpentine field lines . in the present study , which does not include the convection
, we also found the sideway expansion of the rising field at the photosphere ( tongue ) and the undulation of the photospheric fields in the later phase of the emergence of the weaker twist tube .
these features are also confirmed by @xcite ; their simulations do not take account of the convection effects , either .
therefore , we can see that other mechanisms , apart from the convection , could also explain the formation of serpentine fields at the photosphere . in this paper , we found some aspects of the flux emergence event .
one is the photospheric tongue , i.e. , the magnetic structure extending horizontally around the surface just before further evolution takes place .
temporally- and spacially - resolved spectroscopic observations of the earlier phase of the flux emergence are required to study this magnetic extension at the photosphere . at the same time
, we found that the initial twist of @xmath124 at @xmath125 is too strong to match the observations ( as mentioned above in section [ sec : twist ] ) .
local and global helioseismology are needed to reveal the flux emergence ( especially on the twist evolution ) within the convection zone .
the key issue is how weakly twisted flux tubes manage to rise through the solar interior .
numerical computations were carried out on nec sx-9 at the center for computational astrophysics , cfca , of the national astronomical observatory of japan , and on m system ( fujitsu fx1 ) of jaxa supercomputer system .
the page charge of this paper is partly supported by cfca .
s. t. and t. y. thank dr .
y. fan of the high altitude observatory , the national center for atmospheric research .
we thank the referee for helpful suggestions for improvements of this paper .
acheson , d. j. 1979 , , 62 23 archontis , v. , moreno - insertis , f. , galsgaard , k. , hood , a. , & oshea , e. 2004 , , 426 , 1047 archontis , v. , & hood , a. w. 2010 , , 508 , 1469 archontis , v. , & hood , a. w. 2010 , , 514 , 56 cheung , m. c. m. , rempel , m. , title , a. m. , & sch@xmath126ssler , m. 2010 , , 720 , 233 emonet , t. , & moreno - insertis , f. 1998 , , 492 , 804 fan , y. 2001 , , 554 , l111 heyvaerts , j. , priest , e. r. , & rust , d. m. 1977 , , 216 , 123 isobe , h. , tripathi , d. , & archontis , v. 2007 , , 657 , l531 li , h. , schmieder , b. , song , m. t. , & bommier , v. 2007 , , 475 , 1081 magara , t. 2001 , , 549 , 608 matsumoto , r. , tajima , t. , shibata , k. , & kaisig , m. 1993 , , 414 , 357 moreno - insertis , f. , & emonet , t. 1996 , , 472 , l53 murray , m. j. , hood , a. w. , moreno - insertis , f. , galsgaard , k. , & archontis , v. 2006 , , 460 , 909 otsuji , k. et al .
2010 , , 62 , 893 pariat , e. aulanier , g. , schmieder , b. , georgoulis , m. k. , rust , d. m. , & bernasconi , p. n. 2004 , , 614 , 1099 parker , e. n. 1966 , , 145 , 811 sch@xmath126ssler , m. 1979 , , 71 , 79 shibata , k. , tajima , t. , steinolfson , r. s. , & matsumoto , r. 1989 , , 345 , 584 toriumi , s. , & yokoyama , t. 2010 , , 714 , 505 toriumi , s. , & yokoyama , t. 2011 , , submitted ( 65.25mm,40.5mm)figure5a.eps2 ( 60.75mm,40.5mm)figure5b.eps2 + ( 65.25mm,40.5mm)figure5c.eps2 ( 60.75mm,40.5mm)figure5d.eps2 + ( 65.25mm,40.5mm)figure5e.eps2 ( 60.75mm,40.5mm)figure5f.eps2 + ( 65.25mm,43.5mm)figure5g.eps2 ( 60.75mm,43.5mm)figure5h.eps2 + ( 46.05mm,6.75mm)figure5i.eps2
|
we present a series of numerical experiments that model the evolution of magnetic flux tubes with a different amount of initial twist . as a result of calculations ,
tightly twisted tubes reveal a rapid two - step emergence to the atmosphere with a slight slowdown at the surface , while weakly twisted tubes show a slow two - step emergence waiting longer the secondary instability to be triggered .
this picture of the two - step emergence is highly consistent with recent observations .
these tubes show multiple magnetic domes above the surface , indicating that the secondary emergence is caused by interchange mode of magnetic buoyancy instability . as for the weakest twist case
, the tube exhibits an elongated photospheric structure and never rises into the corona .
the formation of the photospheric structure is due to inward magnetic tension force of the azimuthal field component of the rising flux tube ( i.e. , tube s twist ) .
when the twist is weak , azimuthal field can not hold the tube s coherency , and the tube extends laterally at the subadiabatic surface .
in addition , we newly find that the total magnetic energy measured above the surface depends on the initial twist .
strong twist tubes follow the initial relation between the twist and the magnetic energy , while weak twist tubes deviates from this relation , because these tubes store their magnetic energy in the photospheric structures .
|
high temperature superconductivity in the lanthanum @xcite , yttrium @xcite and related copper - oxide compounds remains a subject of intensive investigation and controversy .
it was suggested that electron - phonon interaction mechanism , which is very successful in understanding of conventional ( low temperature " ) superconductors within the bardeen - cooper - schriffer scheme @xcite , may not be adequate for high-@xmath2 cuprates , and even the conventional fermi liquid model of metallic state may require reconsideration .
this opens an area for investigation of mechanisms of electron - electron interaction which can be relevant in understanding peculiarities of superconducting , as well as normal state , properties of cuprates .
specific to all of them is the existence of oxide orbitals .
band calculations @xcite suggest that hopping between the oxygen @xmath7 orbitals and between the copper @xmath8 orbitals may be of comparable magnitude . on the experimental side , spectroscopic studies @xcite
clearly show that the oxygen band appears in the same region of oxygen concentration in which superconductivity in cuprates is the strongest .
therefore there exists a possibility that specific features of oxide compounds may be related to oxygen - oxygen hopping , or to the interaction between the copper and the rotational @xmath9 collective modes .
if the oxygen hopping is significant then it immediately follows that intrinsic oxygen carriers ( @xmath7 oxygen holes ) should be different from the more familiar generic @xmath10-orbital derived itinerant carriers .
the difference is related to low atomic number of oxygen such that removing or adding of one electron to atom induces a substantial change in the coulomb field near the remaining ion and therefore results in the change of the effective radius of atomic orbitals near the ion .
this will strongly influence the hopping amplitude between this atom and the atoms in its neighborhood .
such `` orbital contraction '' effect represents a source of strong interaction which does not simply reduce to the coulomb ( or phonon ) repulsion ( or attraction ) between the charge carriers .
it was suggested by hirsch and coauthors @xcite , and by the present authors @xcite that the occupation dependent hopping can have relevance to the appearance of superconductivity in high - temperature oxide compounds . in the present paper
, we investigate the generic occupation - dependent hopping hamiltonians with respect to peculiarities of the normal state , and to the range of existence of the superconducting state .
theoretical investigation of cooper instability is supplemented by numeric study of pairing and diamagnetic currents in finite atomic clusters .
we study the effect of cooper pairing between the carriers and show that at certain values and magnitudes of the appropriate coupling parameters , the system _ is _ actually superconducting .
the properties of such superconducting state are in fact only slightly different from the properties of conventional ( low-@xmath2 ) superconductors . among those we so far can only mention the change in the fluctuation conductivity above or near the critical temperature @xmath2 .
relaxation of the pairing parameter to equilibrium acquires a small real part due to the asymmetry of contraction - derived interaction between the quasi - particles above and below the fermi energy .
oxygen atoms in the copper - oxygen layers of the cuprates ( figure 1 ) have simple quadratic lattice .
we assume that @xmath11 orbitals of oxygen ( @xmath12 is the direction perpendicular to the cuprate plane ) are bound to the near cuprate layers whereas carriers at the @xmath7 orbitals may hop between the oxygen ions in the plane .
let @xmath13 be the hopping amplitude of @xmath14 and @xmath15 the hopping amplitude of @xmath16 oxygen orbitals between the nearest lattice sites in the @xmath17 direction in a square lattice with a lattice parameter @xmath18 .
then the non - interacting hamiltonian is @xmath19 where @xmath20 is the creation ( annihilation ) operator for @xmath21 and correspondingly @xmath22 for @xmath23 orbitals .
the interaction hamiltonian includes the terms @xmath24 + \sum_{<ij > } b_i^{\dagger}b_j \left[v n_i n_j + w ( n_i + n_j ) \right]\ ] ] where @xmath25 , @xmath26 .
this corresponds to the dependence of the hopping amplitude on the occupation numbers @xmath27 of the form @xmath28 + \tau_2 m_i m_j\ ] ] and correspondingly @xmath29 of the same form with @xmath30 replaced with @xmath31 .
the amplitudes @xmath32 correspond to the transitions between the ionic configurations of oxygen : @xmath33 @xmath34 corresponds to the neutral oxygen ion whereas @xmath35 to the single charged and @xmath36 to the double charged negative ions .
since oxygen atom has @xmath37 configuration in its ground state , filling of the @xmath38 shell to the full occupied configuration @xmath39 is the most favorable .
amplitudes @xmath0 and @xmath1 relate to the parameter @xmath32 according to @xmath40 assuming @xmath41 and replacing @xmath42 with @xmath43 with the pseudo - spin indices @xmath44 we write the hamiltonian eq.(1 ) in the form @xmath45 where @xmath46 where we also included the in - site coulomb interaction ( u ) between the dissimilar orbitals at the same site .
@xmath47 can also be considered as a real spin projection of electrons at the site . in that case
, the pairing will originate between the spin - up and spin - down orbitals , rather than between @xmath21 and @xmath23 orbitals .
more complex mixed spin -and orbital- pairing configurations can also be possible within the same idea of orbital contraction ( or expansion ) at hole localization but are not considered in this paper .
the following discussion does not distinguish between the real spin and the pseudo - spin pairing .
the hamiltonian , eq.(6 ) , is a model one which can not refer to the reliable values of the parameters appropriate to the oxide materials .
the purpose of our study is rather to investigate the properties of superconducting transition specific to the model chosen and to find the range of the @xmath48 values which may correspond to superconductivity .
this will be done along the lines of the standard bcs model @xcite in the weak coupling limit , @xmath49 , and by an exact diagonalization of the hamiltonian for a finite atomic cluster at large and intermediate coupling . in the momentum representation ,
the hamiltonian becomes @xmath50 with @xmath51 where @xmath52 and @xmath53 is the chemical potential .
@xmath54 is the zero order vertex part defined as @xmath55 \tau_{\alpha \beta}^x \tau_{\gamma \delta}^x(\delta_{\alpha \gamma } \delta_{\beta \delta } - \delta_{\alpha \delta } \delta_{\beta \gamma } ) \delta_{{\bf p}_1+{\bf p}_2 , { \bf p}_3+{\bf p}_4}\ ] ] where @xmath56 is a pauli matrix @xmath57 for reasons which will be clear later , we separated @xmath58 and put some part of it into the @xmath59 term , while the remaining part is included in the @xmath60 term , thus giving @xmath61 with @xmath62 being the average occupation of the site .
the cooper instability realizes at certain temperature @xmath63 as a singularity in a two - particle scattering amplitude at zero total momentum .
let s introduce a function @xmath64 where @xmath65 , @xmath66 are the imaginary time @xmath67 creation and annihilation operators . at @xmath68 , @xmath69 ,
the kernel of @xmath70 is proportional to @xmath71 @xmath72 ( @xmath73 is one - electron green function ) .
we keep notation @xmath74 for such a reduced green function specifying only momenta @xmath75 and @xmath76 . by assuming temporarily @xmath77 , this hamiltonian results in an equation for the fourier transform @xmath78 @xmath79 corresponding to summation of feynmann graphs shown in figure 2 . in the above formulas , @xmath80 and @xmath81 ( @xmath82 integers ) are the discrete odd and even frequencies of the thermodynamic perturbation theory @xcite .
@xmath83 is a one - particle green function in a fourier representation @xmath84 diagrams of figure 2 are singular since equal momenta of two parallel running lines bring together singularities of both green functions @xmath85 and @xmath86 . 6-vertex interaction , eq.(8 ) , is not generally considered in the theories of strongly - correlated fermionic systems .
such interaction also results in singular diagrams for @xmath87 scattering shown in figure 3 .
since a closed loop in this figure does not carry any momentum to the vertex , it reduces to the average value of @xmath88 which in turn is the average of the number operator , @xmath89 .
taking into consideration of such diagrams is equivalent to replacing one of the @xmath31 s in eq.(8 ) to its thermodynamical average @xmath90 .
then the @xmath0 term can be added to the renormalized value of @xmath1 , @xmath91 we will check by numeric analysis in sec .
iii to which extent such an approximation may be justified .
solution to eq.(16 ) can be received by putting @xmath92 substituting this expression into eq.(16 ) and introducing the quantities @xmath93 we receive a system of coupled equations for @xmath94 @xmath95 where @xmath96 , which are solved to give @xmath97 where @xmath98 is a determinant @xmath99 the determinant becomes zero at some temperature which means an instability in the two - particle scattering amplitude ( @xmath100 ) .
this temperature is the superconducting transition temperature @xmath2 . at @xmath2 , eq.(16 )
is singular , which means that two - particle scattering amplitude gets infinite .
below @xmath2 , the finite value of @xmath4 is established by including the non - zero thermal averages ( the order parameters ) , @xmath101 , @xmath102 .
we first analyze the case of non - retarded , non - contraction interaction @xmath103 , and after that will consider the effect of the occupation - dependent hopping terms , @xmath0 and @xmath1 . neglecting contraction parameters @xmath104 ,
solution to eq.(16 ) reduces to @xmath105 which after the summation over the discrete frequencies reduces to the conventional bcs equation ( at negative @xmath103 ) @xmath106 with @xmath107 . at finite frequency @xmath108 , eq.(23 ) reduces to @xmath109 where we replaced for simplicity an integration over the brillouin , zone @xmath110 , by the integration over the energy assuming that the density of states near the fermi energy @xmath53 is flat .
@xmath111 and @xmath112 are the lower and upper limits of integration equal to @xmath113 and @xmath114 , respectively .
such an approximation is not very bad since most singular contribution to integral comes from the point @xmath115 where the integrand is the largest . above @xmath2 ,
eq.(25 ) determines the frequency of the order parameter relaxation @xcite .
there is a small change in this frequency compared to the bcs model in which limits of the integration @xmath116 are symmetric with respect to the fermi energy , and small in comparison to @xmath117 , therefore we briefly discuss it now . to receive a real - time relaxation frequency ,
eq.(25 ) needs to be analytically continued to a real frequency domain from the discrete imaginary frequencies @xmath118 @xcite . using the identity @xmath119 where @xmath120 is a fermi function @xmath121 gives @xmath122 where @xmath123 @xmath124 is euler constant .
analytic continuation is now simple : we change @xmath108 to @xmath125 , @xmath126 , to receive a function which will be analytic in the upper half plane of complex @xmath127 , @xmath128 .
the order parameter relaxation equation becomes @xmath129 at @xmath130 and @xmath131 , the real and imaginary parts of eq.(29 ) are easily evaluated to give @xmath132 thus , the order parameter relaxation equation at @xmath133 becomes @xmath134 where @xmath135 in comparison to the bcs theory in which @xmath136 ( @xmath137 is the debye frequency ) and therefore @xmath138 , we receive the relaxation which has a non - zero `` inductive '' component , @xmath139 .
typically , @xmath140 and therefore @xmath141 is a small quantity .
it increases however near the low ( @xmath142 ) or near the maximal ( @xmath143 ) occupation where @xmath144 or @xmath112 become small .
such mode of relaxation is specific to a non - retarded ( non - phonon ) interaction which is not symmetric near @xmath117 and spans over the large volume of the @xmath145-space rather than is restricted to a narrow energy @xmath146 near the fermi energy .
neglecting direct interaction , we put @xmath147 in eq.(22 ) and receive @xmath148 where at finite frequency @xmath127 @xmath149 putting @xmath150 we receive from eq.(33 ) a transition temperature @xmath2 .
the equation has a solution at @xmath151 , @xmath152 , or at @xmath153 , @xmath154 ( we assume that @xmath155 ) . plus or minus sign is chosen to receive the maximal value of @xmath2 ( the second solution corresponding to smaller @xmath156 , then , has to be disregarded since at @xmath157 the order parameter will be finite and therefore eqs.(20)-(22 ) do not apply ) .
this gives an expression for @xmath2 @xmath158 \exp \left(-\frac{t}{2|\tilde{w}|n(\varepsilon_f)}\right)\ ] ] where @xmath159 , @xmath160 ( second exponent is dominating the first one in the weak coupling limit @xmath161 )
. real and imaginary parts of @xmath162 are calculated at @xmath130 @xmath163 equation for @xmath164 is received with a value larger than the previous one ( eq.(32 ) ) @xmath165 eigenvalue equation gives the @xmath166-dependence of the two particle correlator @xmath167 near @xmath2 @xmath168.\ ] ] since @xmath124 diverges at @xmath2 , this determines that order parameter becomes macroscopic at @xmath157 .
then , the pair creation operator , @xmath169 , will almost be a number , i.e. , we may decompose eq.(39 ) into a product @xmath170 and , to be consistent with the @xmath166 , @xmath171 dependences , by putting @xmath172 we receive @xmath173 where @xmath174 and @xmath175 is an overall phase which is irrelevant for a single superconductor but is important for calculating currents in multiple or weakly coupled superconductors .
therefore , system undergoes a pairing transition at temperature found from the eq.(35 ) .
since the pairs are charged , the state below @xmath2 can not be non - superconducting .
we have not calculated the meissner response but in the following section we present numerical calculation of flux quantization which supports the above statement .
we calculate the ground state energy of a cubic system as shown in figure 4 . a magnetic flux @xmath176 is produced by a solenoid passing through the cube .
corners of the cube are the lattice sites , which can be occupied by electrons . with the inclusion of the magnetic flux , model hamiltonian , eq .
6 , becomes @xmath177 \exp(i \alpha_{ij } ) + h.c.\end{aligned}\ ] ] where @xmath178 and @xmath179 is the magnetic flux quantum . throughout the calculations we take @xmath180 .
we start with constructing the model hamiltonian . in a hilbert space of one electron @xmath181 with a basis specified as @xmath182 for the ground state ( @xmath183 ) and @xmath184 for the excited state ( @xmath185 ) . in case of @xmath186 states operator of annihilation @xmath187
takes the form @xmath188 where @xmath189 is the unit matrix and @xmath10 is unitary matrix @xmath190 and @xmath191 stands for the kronecker matrix multiplication .
explicitly , we have @xmath192 thus , for example , for two states @xmath193 these matrices , which are annihilation operators , and corresponding hermitian conjugate matrices , which are the creation operators , satisfy the fermi anti - commutation relation .
these operators are sparse matrices with only @xmath194 non - zero elements , which are equal to @xmath195 .
next we solve the schrdinger equation @xmath196 .
we implemented a novel algorithm for solving such sparse systems , which will be described elsewhere .
the cubic cluster within the hubbard hamiltonian and no external flux applied to the system was studied previously by callaway et .
quantum monte carlo methods applicable to large systems within the hubbard model ( both attractive and repulsive ) , but not the occupation - dependent hopping hamiltonians , are reviewed in a paper of dagotto @xcite .
superconductivity reveals itself in the lowering of the ground state energy as electrons get paired .
therefore the energy needs to be minimal for even number of electrons @xmath197 and will attain a larger value when @xmath197 is odd .
we consider a gap " parameter @xcite @xmath198 as a possible signature " of superconductivity ( where @xmath199 corresponds to the ground state energy for @xmath200 fermions ) .
for all interaction parameters set to zero ( @xmath201 ) , no sign of pairing is observed . to check our analytic results of sec .
iib and the argument following eq.(34 ) , we calculated @xmath5 above and below the half - filling ( @xmath202 in case of cubic cluster ) .
below the half - filling chemical potential is negative ( @xmath159 ) and above the half - filling it is positive ( @xmath154 ) .
we first checked that the @xmath203 , @xmath204 and @xmath205 , @xmath206 calculation is consistent with an exact solution available for a non - interacting system of @xmath197 electrons .
we then test our program for the case of negative-@xmath103 hubbard hamiltonian ( @xmath207 ) which is known to be superconducting ( e.g. refs .
positive-@xmath103 hubbard model does not show any sign of superconductivity , in disagreement with some statements in the literature @xcite .
our calculations can not disprove the ( possible ) non - pairing mechanisms of superconductivity but these seem to be unlikely models for the problem of superconductivity in oxides which clearly shows pairing of electrons ( holes ) in the josephson effect and in the abrikosov vortices .
the relation @xmath208 is justified in the first case @xcite and flux quantum of a vortex is @xmath209 in the second @xcite , both with the value of the charge equal to twice the electronic charge , @xmath189 .
figure 5 shows the dependence of the ground state energy upon the number of particles in case of negative-@xmath103 and positive-@xmath103 hubbard models assuming @xmath77 and @xmath210 .
such dependences are typical for any value of @xmath211 .
there clearly is the pairing effect when @xmath212 and there is no sign of pairing at @xmath6 .
tests for pairing in the contraction @xmath0 , @xmath1-models ( @xmath213 and @xmath214 , respectively ) are shown in figs .
the results are in agreement with our perturbative calculation of sec .
ii and with its extension for the intermediate and strong coupling limits @xmath215 , @xmath216 . since
chemical potential is negative below the half - filling and positive above the half - filling , there is no pairing in the former case ( @xmath217 ) and there is a sign of pairing in the latter case ( @xmath218 ) , in accord with the value of the effective coupling constant @xmath219 .
similarly , for @xmath220 below the half - filling there is a sign of pairing ( @xmath221 ) while above the half - filling there is no pairing .
these results are summarized in table 1 . for larger values of the interaction parameters ,
the perturbative results do not remain applicable anymore .
figure 8b shows the dependence of the parity gap @xmath5 on the strength of the interaction . from figure 8
it is understood that the @xmath1 interaction introduces a signature " of pairing in a similar way as the negative-@xmath103 interaction does .
the possibility of the contraction " pairing has been investigated formerly in the papers @xcite .
flux quantization is another signature of superconductivity which is a consequence of the meissner effect .
we also tested for the periodicity of the energy versus flux dependence with the period @xmath222 as compared to the period @xmath179 in the non - interacting system @xcite .
unfortunately , the even harmonics of @xmath223-periodic dependence of the ground state energy ( and related to it , the harmonics of the persistent current @xmath224 @xcite ) may simulate the pairing in a non - superconductive system .
small - size ( mesoscopic ) system can mask the superconducting behavior @xcite .
flux quantization in hubbard hamiltonians was studied formerly in refs .
29 - 31 .
we first demonstrate the behavior of the ground state energy with respect to flux , figure 9 .
a characteristic feature of mesoscopic system suggests that addition of one extra particle to the system changes the sign of the derivative of the ground state energy with respect to magnetic flux at @xmath225 . that is , depending on the parity of the number of particles and on the number of sites , system can change from paramagnetic to diamagnetic state or vice versa . but
this behavior is not always observed for the cubic geometry studied . except the sign change from @xmath226 to @xmath227 and from @xmath228 to @xmath202 ,
no such behavior is seen . as mentioned above
, however , the @xmath229-periodic component of the @xmath230 dependence begins to appear at the higher value of @xmath197 ( figure 9,c ) . for both contraction parameters equal to zero , i.e. @xmath231 , we observe appearance of the @xmath209-periodic component for some values of @xmath103 ( figure 10 ) . even for positive ( repulsive ) values of @xmath103 ,
it is possible to see a local minimum appearing at @xmath232 ( figure 10b ) .
this is in agreement with the authors previous works @xcite . but
this minimum , which does not lead to an exact periodicity of the ground state energy with a period @xmath233 , should not be attributed to superconductivity , this is rather a characteristic behavior in mesoscopic systems . for @xmath212 ( while @xmath231 ) , the expected mesoscopic behavior , that is the change of the sign of the slope of ground state energy at @xmath225 , starts to demonstrate itself ( figure 11 ) .
but this happens at sufficiently large absolute values of ( negative ) @xmath103 .
for other values of @xmath103 , however , there is no such change .
more pronounced @xmath209-periodic components are observed with the introduction of non - zero interaction parameters .
the role of @xmath1 on the ground state energy , when both @xmath103 and @xmath0 are zero , is shown in figure 12 .
meanwhile setting both @xmath103 and @xmath1 to zero and observing the effect of the non - zero @xmath0 shows that @xmath0 does not play a role as significant as the other two interaction parameters do .
there is not much difference in the behavior of the ground state energy upon magnetic flux between the zero and non - zero @xmath0 ( for example @xmath234 ) cases .
we studied the peculiarity of electron conduction in systems in which conduction band is derived from the atomic shells with a small number of electrons ( @xmath235 ) in an atom .
such materials may include oxygen ( @xmath236 ) in the oxides , carbon ( @xmath237 ) in borocarbides ( e.g. @xmath238 ) , hydrogen ( @xmath239 ) in some metals ( e.g. , @xmath240 ) .
some materials of this kind are superconductors .
it was argued that the coulomb effects within the atoms strongly influence the inter - atom wave function overlap between the atomic sites and therefore the electron hopping amplitude between the sites .
the phenomenology of such conduction mechanism results in a novel , to the conventional solid state theory , hamiltonians called the occupation - dependent - hopping ( or contraction ) hamiltonians , specified with the two coupling parameters @xmath0 , @xmath1 .
we then attempted a study of superconductivity in such systems within the bcs - type approach assuming the cooper pairing of electrons .
the weak - coupling limit allows determination of the range of parameters @xmath0 , @xmath1 values and also of the in - site coulomb interaction @xmath103 value which show the cooper instability .
the strong - coupling limit was addressed by a numeric calculation on finite clusters using the novel algorithm ( of non - lanczos type ) for eigenvalues of large sparse matrices .
one of the results of this numeric calculation was that the positive-@xmath103 hubbard model , sometimes believed to be a candidate for high-@xmath2 superconductivity , does not comply with the goal .
j. g. bednorz and k. a. muller , zs .
b * 64 * , 189 ( 1986 ) .
m. k. wu , j. r. ashburn , c. j. torng , p.h .
hor , r. l. meng , l. cao , z. l. huang , y. q. wang and c. w. chu , phys .
lett . * 58 * , 908 ( 1987 ) .
j. bardeen , l. n. cooper and j. r. schriffer , phys . rev . * 108 * , 1175 ( 1957 ) .
m. s. hybertsen , e. b. stechel , m. schlter and d. r. jennison , phys .
b * 41 * , 11068 ( 1990 )
. a. k. mahan , j. a. annett and r. m. martin , phys . rev . * 108 * , 1175 ( 1957 ) .
n. nucker , j. fink , j. c. fugle , p. j. durham and w. m. temmerman , phys .
b * 37 * , 5158 ( 1998 ) .
p. kuiper , g. kruizinaga , j. ghijsen , m. grioni , p. j. w. weijs , f. m. de groot and g. a. sawatsky , phys . rev .
b * 38 * , 6483 ( 1988 )
. j. e. hirsch , phys .
a * 134 * , 451 ( 1989 ) .
j. e. hirsch and f. marsiglio , phys .
b * 39 * , 11515 ( 1989 )
. j. e. hirsch , phys rev .
b * 48 * , 3327 ( 1993 ) ; ibid . , * 48 * , 9815 ( 1993 ) . i. o. kulik , sov . superconductivity : phys .
* 2 * , 201 ( 1989 ) .
i. o. kulik , _ contraction mechanism for pairing interaction in oxides and hydrides _ , in : _ progress in high temperature physics _ , vol .
25 , r.nicholsky , ed . ,
world scientific , singapore ( 1990 ) .
h. boyaci and i. o. kulik , fiz .
nizk . temp .
* 24 * , 316 ( 1998 ) [ low temp
* 24 * , 239 ( 1998 ) ] .
i. o. kulik , tr .
j. phys . *
20 * , 627 ( 1996 ) .
a. a. abrikosov , l. p. gorkov and i. e. dzyaloshinsky , _ methods of quantum field theory in statistical physics _ ,
dover publications , new york ( 1975 ) .
e. abrahams and t. tsuneto , phys .
rev . * 152 * , 416 ( 1966 ) .
l. p. gorkov and g. m. eliashberg , zurn .
. fiz . * 54 * , 612 ( 1968 ) .
i. o. kulik , o. entin - wohlman , r. orbach , j. low temp . phys . * 43 * , 591 ( 1981 ) .
j. callaway , d. p. chen , y. zang , phys . rev . * 36 * , 2084 ( 1987 ) .
e. dagotto , rev .
66 * , 763 ( 1994 ) .
k. a. matveev , a. i. larkin , phys .
lett . * 78 * , 3749 ( 1997 ) .
i. o. kulik , physica b * 126 * , 280 ( 1984 ) ; int . j. mod
b * 1 * , 851 ( 1988 ) .
r. micnas , j. ranninger , s. robaszkiewicz , rev .
phys . * 62 * , 113 ( 1990 ) . c. h. pao , n. e. bickers , phys .
lett . , * 72 * , 1870 ( 1994 ) ; d. j. scalapino , phys .
reports , * 250 * , 329 ( 1995 ) .
t. j. witt , phys .
lett . * 61 * , 1423 ( 1988 ) ; d. esteve , j. m. martinis , c. urbina , m. h. devoret , g. collin , p. monod , r. ribault , a. percolevski , europhys . lett . * 3 * , 1237 ( 1987 ) .
p. l. gammel , d. j. bishop , g. j. dolan , j. r. kwo , c. a. murray , l. f. schneener , j. v. waszczak , phys .
* 59 * , 2592 ( 1987 ) ; l. ya .
vinnikov , l. a. gurevich , g. a. yemelchenko , yu . a. ossipyan , sol .
communs . *
67 * , 253 ( 1988 ) .
i. o. kulik , pisma zh .
11 * , 407 ( 1970 ) .
[ jetp lett .
* 11 * , 275 ( 1970 ) ] .
m. buttiker , y. imry , r. landauer , phys .
a * 96 * , 365 ( 1983 ) .
a. ferretti , i. o. kulik , a. lami , phys .
b * 47 * , 12235 ( 1993 ) .
r. m. fye , m. j. martins , and d. j. scalapino , phys .
b * 44 * , 6909 ( 1991 ) .
a. sudb , c. m. varma , t. giamarchi , e. b. stechel , and r. t. scalettar , phys .
lett . * 70 * , 978 ( 1993 ) . .pairing effect for arbitrarily small values of @xmath0 and @xmath1 , computed by exact diagonalization of the hamiltonian .
the results presented here are in complete agreement with the perturbative calculations . [
cols="^,^,^,^,^ " , ]
|
a generic hamiltonian , which incorporates the effect of the orbital contraction on the hopping amplitude between the nearest sites , is studied both analytically at the weak coupling limit and numerically at the intermediate and strong coupling regimes for finite atomic cluster .
the effect of the orbital contraction due to hole localization at atomic sites is specified with two coupling parameters @xmath0 and @xmath1 ( multiplicative and additive contraction terms ) .
the singularity of the vertex part of the two - particle green s function determines the critical temperature @xmath2 and the relaxation rate @xmath3 of the order parameter at temperature above @xmath2 . unlike in conventional bcs superconductors
, @xmath4 has a non - zero imaginary part which may influence the fluctuation conductivity of superconductor above @xmath2 .
we compute the ground state energy as a function of the particle number and magnetic flux through the cluster , and show the existence of the parity gap @xmath5 appearing at the range of system parameters consistent with the appearance of cooper instability .
numeric calculation of the hubbard model ( with @xmath6 ) at arbitrary occupation does not show any sign of superconductivity in small cluster .
|
with the advent of mesoscopic physics , it has become possible to experimentally investigate quantum phase coherent properties of electrons in solid state conductors in a controlled way . in particular , in ballistic mesoscopic samples at low temperatures , electrons can propagate up to several microns without loosing phase information .
this opens up the possibility to investigate electrical analogs of various optical phenomena and experiments .
an investigation of such analogs is of fundamental interest . on the one hand
, it allows one to establish similarities between the properties of photons and conduction electrons , a consequence of the wave nature of the quantum particles . on the other hand
, it also allows one to investigate the differences between the two types of particles arising from the different quantum statistical properties of fermions and bosons . for many - particle properties , such as light intensity correlations or correspondingly electrical current correlations , noise , the quantum statistical properties are important .
@xcite both the wave - nature of the particles as well as their quantum statistics are displayed in a clearcut fashion in interferometer structures . in this work
we are concerned with the electrical analogs of two well known optical interferometers , the single - particle mach - zehnder ( mz ) interferometer and the two - particle hanbury brown twiss ( hbt ) interferometer .
the mz - interferometer is a subject of most textbooks in optics .
@xcite in the framework of quantum optics , considering individual photons rather than classical beams of light , the interference arises due to the superposition of the amplitudes for two different possible paths of a single photon .
this leads to an interference term in the light intensity .
the mz - interferometer is thus a prime example of a single particle interferometer.@xcite various electronic interferometers with ballistic transport of the electrons have been investigated experimentally over the last decades , as e.g. aharonov - bohm ( ab ) rings@xcite and double - slit interferometers.@xcite detailed investigations of dephasing in ballistic interferometers was carried out in refs .
[ ] . only very recently was the first electronic mz - interferometer realized by ji et al.@xcite in a mesoscopic conductor in the quantum hall regime .
a high visibility of the conductance oscillations was observed , however the visibility was not perfect .
this led the authors to investigate in detail various sources for dephasing .
as a part of this investigation , also shot noise was measured .
still , some aspects of the experiment are not yet fully understood .
theoretically , seelig and one of the authors@xcite investigated the effect of dephasing due to nyquist noise on the conductance in a mz - interferometer .
the effect of dephasing on the closely related four - terminal resistance in ballistic interferometers @xcite was investigated as well .
dephasing in ballistic strongly interacting systems is discussed by le hur .
@xcite following the experimental work of ji et al.,@xcite marquardt and bruder investigated the effect of dephasing on the shot - noise in mz - interferometers , considering dephasing models based on both classical@xcite as well as quantum fluctuating fields .
@xcite very recently , frster , pilgram and one of the authors @xcite extended the dephasing model of refs .
[ ] to the full statistical distribution of the transmitted charge .
the hbt - interferometer @xcite was originally invented for stellar astronomy , to measure the angular diameter of stars .
it is an intensity , or two - particle,@xcite interferometer .
the interference arises from the superposition of the amplitudes for two different two - particle processes .
importantly , there is no single particle interference in the hbt - interferometer .
consequently , in contrast to the mz - interferometer there is no interference in the light intensity , the interference instead appears in the intensity - intensity correlations .
moreover , the intensity - intensity correlation also display the effect of quantum statistics .
photons originating from thermal sources tend to bunch , giving rise to positive intensity cross correlations . for the electronic analog of the hbt - interferometer , it was the corresponding anti - bunching of electrons that originally attracted interest .
it was predicted @xcite that the electrical current cross correlations in mesoscopic conductors would be manifestly negative , i.e. display anti - bunching , as a consequence of the fermionic statistics of the electrons .
negative current cross correlations were subsequently observed in two independent experiments.@xcite recently , anti - bunching for field emitted electrons in vacuum was also demonstrated.@xcite the two - particle interference in the hbt - experiment has received much less attention .
we emphasize that while the bunching of the photons was necessary for obtaining a finite , positive cross correlation signal , it was the two - particle effect that was of main importance to hbt since the angular diameter of the star was determined from the two - particle interference pattern . in electrical conductors , two - particle effects in ab - interferometers
were investigated theoretically in refs .
only very recently two of the authors and sukhorukov@xcite proposed a direct electronic analog of the optical hbt - interferometer which permits to demonstrate two - particle interference in an unambiguous way . in this work
we investigate and compare in detail the current and and zero - frequency noise in electronic mz and hbt interferometers .
we consider interferometers implemented in mesoscopic conductors in the integer quantum hall regime , where the transport takes place along single edge states and quantum point contacts ( qpc s ) serve as controllable beam splitters .
the effect of finite temperature , applied bias and asymmetry , i.e. unequal length of the interferometer arms , is investigated .
the strength of the interference contribution is quantified via the visibility of the phase oscillations .
the dependence of the visibility on the beam splitter transparencies as well as on the temperature , voltage and asymmetry is discussed in detail . of interest is the comparison of visibility of the shot - noise correlation of the mz - interferometer and the hbt - intensity interferometer . shot noise correlations in the mz - interferometer exhibit two contributions , one with the fundamental period of @xmath0 and a second harmonic with period @xmath1 .
the shot noise correlations in the hbt - interferometer , even though they are due to two particle processes , are periodic with period @xmath0 .
thus the aharonov - bohm period can not be used to identify the two particle processes which give rise to the hbt effect .
it is therefore interesting to ask whether the hbt two - particle processes have any other signature , for instance in the temperature or voltage dependence of the visibility of the shot - noise correlation .
we find that this is not the case . to the contrary , we find that the the shot noise correlations in the hbt intensity interferometer behave qualitatively similar to the @xmath0 shot noise correlation in the mz - interferometer .
in contrast the @xmath1 contribution in the shot noise of the mz - interferometer decreases more rapidly with increasing temperature , voltage or dephasing rate than the @xmath0 oscillation in the mz- or hbt - interferometer .
we investigate dephasing of the electrons propagating along the edge states by connecting one of the interferometer arms to a fictitious , dephasing voltage probe . in all cases ,
the current and noise of the mz - interferometer as well as the noise in the hbt - interferometer , the effect of the voltage probe is equivalent to the effect of a slowly fluctuating phase .
in the paper we consider implementations of the mz and hbt interferometers in mesoscopic conductors in strong magnetic fields , in the integer quantum hall regime.@xcite the typical system is a two - dimensional electron gas in a semiconductor heterostructure , with the lateral confinement of the electron gas controllable via electrostatic gating .
the transport between reservoirs@xcite connected to the conductor takes place along edge states.@xcite the edge states , quantum analogs of classical skipping orbits , are chiral , the transport along an edge state is unidirectional .
scattering between edge states is suppressed everywhere in the conductor except at electrostatically controllable constrictions , qpcs.@xcite for a magnetic field that does not break the spin degeneracy of the edge states , each edge state supplies two conduction modes , one per spin .
these properties make conductors in the integer quantum hall regime ideal for realizing analogs of optical experiments .
first , the edge states correspond to single mode waveguides for the light .
the unidirectional motion along the edge states allows for `` beams '' of electrons to be realized .
second , the qpc s work as electronic beam splitters with controllable transparency . moreover
, due to chirality the beamsplitters are reflectionless , a property essential for the mz and hbt interferometers but difficult to achieve for beam splitters in conductors in weak ( or zero ) magnetic fields.@xcite these properties of conductors in the quantum hall regime have been demonstrated experimentally in a number of works , see e.g. [ ] . theoretically , several works have been concerned with the conductance and noise properties of beam splitters and interferometers in quantum hall systems , for a recent reviews see e.g. refs . [ ]
recently , it was proposed to use these appealing properties of edge states in the context of orbital @xcite quasi - particle entanglement in static @xcite and dynamic @xcite systems as well as for quantum state transfer.@xcite it is interesting to note that the edge state description also hold for conductors at even higher magnetic fields , in the fractional quantum hall regime . as examples
, the fractional charge has been determined in shot - noise experiments@xcite and the quantum statistical properties of the fractionally charged quasi - particles have been investigated theoretically in beam - splitter@xcite and hbt@xcite geometries .
various interferometer structures have also been considered.@xcite very recently , a mz - interferometer in the fractional quantum hall regime was proposed.@xcite in this work we however consider only the integer quantum hall effect , where the quasi - particles are noninteracting and the electrical analogs to optical experiments can be directly realized .
this discussion leads us to consider single mode , multi - terminal conductors with noninteracting electrons .
the principle aim of this work is a comparison of the mz and hbt - interferometers . in reality in both interferometers interactions ( screening )
play a role both for the voltage and temperature dependence .
a non - interacting scattering approach is not gauge invariant but requires a treatment of screening .
@xcite however , these effects are expected to be simliar in the two interferometers and will not affect the main conclusions of this work . therefore ,
below we treat non - interacting qausi - particle interferometers .
the conductors are connected to several electronic reservoirs , biased at a voltage @xmath2 or grounded .
the current@xcite and the noise@xcite are calculated within the scattering approach for multi - terminal conductors .
we first introduce the creation and annihilation operators for ingoing , @xmath3 and @xmath4 , and outgoing , @xmath5 and @xmath6 , particles , at energy @xmath7 in terminal @xmath8 . for simplicity
we suppress spin notation .
considering a conductor with @xmath9 terminals , the in- and out - going annihilationoperators are related via the @xmath10 scattering matrix , as @xmath11 where @xmath12 is the amplitude to scatter from terminal @xmath13 to terminal @xmath8 .
the current operator in the lead @xmath8 has the form@xcite @xmath14t/\hbar ) \nonumber \\ & \times & a_{\beta \gamma } ^{\alpha } ( e , e^{\prime } ) \hat{a}_{\beta } ^{\dagger}(e)\hat{a}_{\gamma } ( e^{\prime}),\end{aligned}\ ] ] with the notation @xmath15 the average current is given by@xcite @xmath16 where the spectral current density is @xmath17 here @xmath18)$ ] is the fermi dirac distribution of terminal @xmath19 , with @xmath20 the corresponding applied voltage .
the spectral conductance @xmath21 is given by @xmath22 the zero frequency correlator between current fluctuations in terminals @xmath23 and @xmath19 is defined as @xmath24 where @xmath25 the current correlator is given by@xcite @xmath26 where @xmath27 \label{noisedens}\end{aligned}\ ] ] is the spectral current correlator .
there are several physical mechanisms that might lead to dephasing of the electrons propagating along the edge states ( see e.g. the discussion in ref .
[ ] ) . in this work
we are not interested in any particular mechanism for dephasing but consider instead a phenomenological model , a dephasing voltage probe .
the idea of using a voltage probe to induce dephasing was introduced in refs .
a voltage probe connected to a mesoscopic sample was considered , leading to a suppression of coherent transport due to inelastic scattering .
the probe model , originally considered for the average current , was extended to treat the effect of inelastic scattering on shot noise by bttiker and beenakker@xcite by considering a conservation of current fluctuations at the probe as well . later de jong and beenakker@xcite extended the voltage probe concept and introduced a ( fictitious ) voltage probe which breaks phase but does not dissipate energy .
scattering in the voltage probe is ( quasi-)elastic . this
is achieved with the help of a distribution function in the voltage probe which conserves not only total current like a real voltage probe , but conserves current in every small energy interval .
such a probe provides a model of pure dephasing .
the different probe models have been used as qualitative models in a number of works , see refs .
[ ] for a review . for an application to quantum hall systems ,
[ ] . in this work we consider the dephasing voltage probe model , which conserves the current at each energy .
the model is based on the assumption that the current is conserved on a time scale @xmath28 , much shorter than the time of the measurement but much longer than the time between injection of individual electrons , here of the order of @xmath29 .
one could however consider a more general voltage probe model that takes into account a more complicated dynamics of the probe .
a detailed discussion of such a general model in the light of recent work@xcite is however deferred to a later work .
here we only note that below we find that the voltage probe in both the mz and hbt - interferometers only gives rise to a suppression of the phase dependent terms in conductance and noise , just as one would naively expect to be the effect of pure dephasing .
the condition of zero current into the fictitious probe @xmath30 at each energy is fulfilled by considering a time dependent distribution function of the probe @xmath31 where @xmath32 fluctuates to conserve current on the timescale @xmath28 . as a consequence ,
the spectral current density at each energy in lead @xmath33 fluctuates in time as @xmath34 where the fluctuations @xmath35 consist of two parts , the intrinsic fluctuations @xmath36 and the additional fluctuations due to @xmath37 .
the requirement of zero average current into the probe , @xmath38 , leads to the averaged distribution function at the probe reservoir @xmath30 @xmath39 the average spectral current density @xmath40 is then found from eq .
( [ currspecdens ] ) . the fluctuating part of the distribution function
, @xmath41 is obtained from the requirement of zero current fluctuations into the probe , @xmath42 the total current density fluctuation is then given by @xmath43 as a result , in the presence of dephasing the total spectral current correlation @xmath44 is @xmath45 where @xmath46 is the correlation function between the intrinsic current fluctuations , @xmath47 and @xmath48 , of contact @xmath8 and @xmath19 , given by eq .
( [ noisedens ] ) , and @xmath49 is the conductance , given by eq . ( [ speccond ] ) .
a schematic of the mz - interferometer is shown in fig .
[ mzopt ] .
an incident beam of light from source @xmath50 is divided in two parts at the semitransparent beam splitter @xmath51 .
the two partial beams are reflected at mirrors and later joined at the second beam splitter b. beams of light going out from @xmath52 are detected in @xmath53 and @xmath54 .
the amplitude of the light in an outgoing beam is the sum of the amplitudes for the two partial beams , @xmath55 .
this gives an intensity @xmath56)\right\}$ ] .
the interference term @xmath57)\right\}$ ] thus contains the difference between the geometrical phases , @xmath58 .
importantly , the four terminal geometry together with the reflectionless beam splitters lead to that the incident beam traverses the interferometer only once .
this is a defining property of the mz - interferometer .
we then turn to the electric analog of the mz - interferometer , shown in fig .
as pointed out above , several results for the current and noise are available in the literature.@xcite here we analyze the most general situation possible , with finite voltage , temperature and interferometer arm asymmetry as well as different beam splitters @xmath51 and @xmath52 with arbitrary transparency . when we consider limiting cases for e.g. small temperature , bias or asymmetry ,
known results are recovered .
this detailed analysis of the mz - interferometer is of importance when comparing to the hbt - interferometer below .
we first discuss a fully coherent interferometer , the effect of dephasing is investigated below .
an electric potential @xmath2 is applied at terminal 1 , all other terminals are kept at zero potential .
the injected electrons propagate along single edge states .
scattering between the edge states can take place only at the two qpc s , acting as beam splitters with controllable transparency .
the beam splitters @xmath59 are characterized by the scattering matrices @xmath60 where @xmath61 and @xmath62 are the transmission and reflection probabilities , respectively .
we note that any additional phases of the beam splitters just give rise to a constant phase shift of the oscillations in the interference terms and are therefore not considered .
propagating along the edge states , the electrons pick up geometrical phases @xmath63 and @xmath64 as well as phases @xmath65 and @xmath66 due to the ab - flux @xmath67 through the center of the interferometer .
for example , the amplitude for scattering from terminal 1 to 4 is given by @xmath68 for the geometrical phases , to be specific we consider the case when the potential landscape @xmath69 of the conductor in fig .
[ mzel ] is varying smoothly on the scale of the magnetic length @xmath70 , with @xmath71 the applied magnetic field perpendicular to the plane in fig .
[ mzel ] ( the effect of selfconsistency of the potential @xcite is neglected ) .
this allows for a semiclassical treatment .
@xcite in a high magnetic field the edge states at fermi energy @xmath72 follow equipotential lines determined by @xmath73 where @xmath74 is the cyclotron frequency and @xmath75 the effective electron mass .
we are concerned here with the case where there is only one edge state and thus @xmath76 .
suppose the @xmath77-axis is a line intersecting quantum point contacts @xmath51 and @xmath52 in fig .
[ mzel ] . excluding self - intersections we can express the edge state in terms of functions @xmath78 and @xmath79 for the left and right path of the interferometer . working in the symmetric gauge
, the geometric phases can be written @xcite @xmath80 , where @xmath81 and @xmath82 are the locations of the qpc s .
importantly , @xmath58 corresponds to the total area @xmath51 enclosed by these two paths divided by the magnetic length squared , or @xmath83 where @xmath84 is the total flux through the enclosed area and @xmath85 the elementary flux quantum . note
that the aharonov - bohm flux @xmath67 adds an additional phase @xmath86 and @xmath87 , with @xmath88 , to each of the paths . for the discussion of the temperature and voltage dependence of the current and the noise , we also need to know the energy dependence of the phases .
first , instead of parameterizing the edge state through @xmath77 we introduce the parameter @xmath89 which measures directly the path length , i.e. @xmath90 , @xmath91 .
in addition at @xmath89 we introduce local coordinates @xmath92 along and @xmath93 perpendicular to the equipotential line . in these coordinates ,
an edge state that follows the equipotential line at a small energy @xmath7 away from @xmath72 acquires the additional phase @xmath94 with @xmath95 .
the potential gradient @xmath96 determines the local electric field @xmath97 at @xmath89 .
but @xmath98 where @xmath99 is the drift velocity of the guiding center of the cyclotron orbit at point @xmath89 of the edge state .
thus a small increase in energy leads to a phase increment given by @xmath100e$ ] . a rough estimate using a drift velocity which is constant along the edge gives @xmath101 with @xmath102 the length of the edge state @xmath103 . for the phase - difference of the two interfering paths we have @xmath104 with @xmath105 the equilibrium phase difference .
formally , higher order terms in energy can be neglected for characteristic energies @xmath106 and @xmath2 much smaller than @xmath107 $ ] .
the asymmetry of the two edges thus gives rise to an energy scale @xmath108 - \int ds [ 1/\hbar v_d(s_2)]\}^{-1}$ ] which is due to the mismatch of the edge state path lengths , i.e. @xmath109 with @xmath110 . in principle , for a completely symmetric interferometer one has @xmath111 .
given the scattering amplitudes @xmath112 , the spectral current density is found from eqs .
( [ adef ] ) , ( [ currspecdens ] ) and ( [ speccond ] ) . for e.g. terminal 4 ,
one gets @xmath113 \left [ \mathcal{t}_{a}r_{b}+\mathcal{t}_{b}r_{a}\right .
\notag \\ & & \left.+2\sqrt{\mathcal{t}_{a}\mathcal{t}_{b}r_{a}r_{b}}\cos \left ( e / e_{c}+\theta \right ) \right ] , \end{aligned}\ ] ] where we introduce the total , energy independent phase @xmath114 .
here @xmath115 is the distribution functions of the grounded terminals 2,3 and 4 and @xmath116 the distribution function of terminal 1 .
the current is then given from eq .
( [ avcurr ] ) , as @xmath117 .
\notag \\ & & \label{currmz}\end{aligned}\ ] ] current conservation gives @xmath118 .
the current consists of two physically distinct parts . the first term in eq .
( [ currmz ] ) is the phase independent , incoherent part , the current in the absence of interference , while the second , phase dependent term is the interference contribution .
we note that a bias @xmath2 of the order of the asymmetry energy @xmath119 leads to the phase shifts of the oscillation .
the strength of the interference can conveniently be quantified via the visibility as @xmath120}{\left\langle i\right\rangle},\ ] ] which gives for the current in the mz - interferometer @xmath121 the visibility is a product of a term containing the qpc scattering probabilities and a function depending on the energy scales @xmath122 @xmath2 and @xmath119 .
the scattering probability term is maximum for identical qpc s , @xmath123 the energy scale dependence is shown in fig .
[ figcurrvismz ] where the visibility for identical point contacts is plotted as a function of the normalized temperature , @xmath124 .
we note several interesting features from fig .
[ figcurrvismz ] and eq .
( [ currvismz ] ) .
( i ) the visibility shows decaying oscillations as a function of voltage @xmath125 for arbitrary temperature .
( ii ) a symmetric mz - interferometer , @xmath126 @xmath2 , has unity visibility ( for @xmath127 ) , i.e. shows perfect interference .
( iii ) the visibility decays monotonically with increasing temperature . for large temperatures , @xmath128
, the visibility decays exponentially with the temperature as @xmath129 it is interesting to compare the calculated visibility to the experimentally measured one in ref .
as already shown in ref [ ] , the measured scattering probability dependence of @xmath130 is well reproduced by eq .
( [ currvismz ] ) . for the energy scale dependence ,
no information about the drift velocity @xmath131 or the asymmetry @xmath132 needed to determine @xmath133 is provided in ref .
however , to obtain the order of magnitude of @xmath133 , considering as a rough estimate a typical drift velocity @xcite @xmath134 m / s at a magnetic field @xmath135 and an asymmetry @xmath136 m gives an @xmath133 corresponding to an applied bias @xmath137 or a temperature @xmath138 .
these values are typically of the same order of magnitude as the ones considered in the experiment . as a first approximation
, one would thus expect asymmetry effects to be of importance .
the observed temperature dependence , a strong decrease of the visibility for increased temperature , is also qualitatively described by eq .
( [ currvismz ] ) with an @xmath139 mk .
this is however not the case with the voltage dependence .
ji et al find a differential visibility , i.e. the visibility of @xmath140 , which decays strongly with applied voltage , while eq .
( [ currmz ] ) predicts a constant , voltage independent differential visibility .
there are several possible explanations to why the voltage dependence in contrast to the temperature dependence is not reproduced by the theory .
ji et al themselves point out two voltage dependent dephasing mechanism : low frequency noise of @xmath141 type due to moving impurities , induced by a higher current and fast fluctuations of the potential landscape ( and hence of the phase via the enclosed area ) caused by screening of the additional charges injected at higher current .
screening might also , for the nonlinear current - voltage characteristics predicted by eq .
( [ currmz ] ) , lead to a voltage dependent renormalization of the transmission probabilities , introducing a voltage dependence in the differential visibility .
@xcite we also note that in the model of ref .
[ ] , inducing dephasing by coupling the mz - interferometer to a quantum bath , gives a dephasing rate that increases with increasing voltage .
clearly , further investigations are needed to clarify the origin of the dephasing in the experiment in ref . [ ] .
turning to the noise , we focus on the cross correlator between currents flowing in terminals 3 and 4 ( the auto - correlator can be obtained analogously ) .
this allows for a straightforward comparison to the result of the hbt - interferometer , for which the cross correlator was investigated in ref .
[ ] . from eqs .
( [ noise ] ) and ( [ noisedens ] ) and the expressions for the scattering amplitudes , we arrive at the noise spectral density @xmath142 ^{2 } \nonumber \\ & \times&\left\ { c_{0}+c_{\theta } \cos \left ( \frac{e}{e_{c}}+\theta \right)+c_{2\theta } \cos \left ( 2\left[\frac{e}{e_{c}}+\theta \right ] \right ) \right\ } , \nonumber \\ & & \label{noisespecdens}\end{aligned}\ ] ] with coefficients @xmath143 performing the energy integrals in eq .
( [ noise ] ) we find for the cross correlator @xmath144 \right\ } \label{mznoisetot}\end{aligned}\ ] ] where we introduce the functions @xmath145 @xmath146 . \label{enfcn3}\end{aligned}\ ] ] and @xmath147 .
\label{enfcn2}\end{aligned}\ ] ] containing the dependence on the energy scales @xmath148 @xmath149 and @xmath119 . just as the current in eq .
( [ currmz ] ) , the noise consists of a phase independent , incoherent part and a phase dependent , interference part . however , in contrast to the current , the phase dependent part of the noise contains two terms with different periods in @xmath150 , corresponding to oscillations periodic in @xmath0 and @xmath1 .
these terms result from two - particle scattering processes which enclose the ab - flux one and two times respectively . similarly to the current , the phase of the oscillations are shifted for a bias @xmath2 of the order of the asymmetry energy @xmath119 .
it is important to note that in the mz ( in contrast to the hbt ) interferometer , two particle and higher order scattering processes are just products of single particle scattering processes .
the full distribution of current fluctuations@xcite is thus a function of single particle scattering probabilities only . in particular , the noise spectral density @xmath151 in eq .
( [ noisespecdens ] ) is proportional to @xmath152 , i.e. partition noise@xcite with phase dependent scattering probabilities . as a consequence
, the phase independent , incoherent part of the noise can not be understood as partition noise from incoherent single particle processes , i.e. @xmath153/2 $ ] , from two coherent scattering processes , obviously contribute to the phase independent part of the noise . as a consequence ,
as shown by marquardt and bruder,@xcite a model @xcite with a filled stream of classical particles injected from reservoir @xmath50 correctly reproduces the incoherent part of the current but fails to reproduce the incoherent part of the noise .
in contrast , as found in ref . [ ] and further discussed below , the completely dephasing voltage probe model correctly reproduces the incoherent part of both the current and the noise . to quantify the strength of the oscillations we introduce two separate quantities , @xmath154 and @xmath155 , here simply called visibilities , which in close analogy to the current visibility in eq .
( [ currvismz ] ) are defined as the ratio of the amplitudes of the noise oscillations and the average noise .
they become @xmath156 and @xmath157 similarly to the current , both visibilities are products of a term containing the scattering probabilities and a function of the energy scales @xmath2 , @xmath149 and @xmath119 .
we first focus on the scattering probability dependent term by considering the visibility in the limit of a symmetric interferometer , @xmath158 , @xmath149 , where the energy - scale dependent terms are unity .
this gives @xmath159 and @xmath160 the two visibilities are plotted in fig .
both visibilities are symmetric under the substitutions @xmath161 and @xmath162 .
the visibility @xmath163 is zero for @xmath164 , i.e. for a symmetric setting of any of the qpc s
. the visibility increases for increasing qpc asymmetry , reaching a maximum for @xmath165 and @xmath166 ( unity only in the limit @xmath167 ) and then decreases again toward zero at @xmath168 or @xmath169 .
interestingly , the visibility @xmath170 shows an opposite behavior .
it is maximum , equal to unity , at @xmath171 and then decreases monotonically for increasing qpc asymmetry , reaching zero at @xmath168 or @xmath169 .
this different dependence on the scattering probabilities makes it possible to investigate the two oscillations independently by modulating the qpc transparencies .
turning to the energy scale behavior , we consider for simplicity @xmath163 in the limit @xmath172 and @xmath173 in the limit @xmath171 where respective scattering probability terms are unity . for a symmetric interferometer , i. e. @xmath174 , both visibilities are unity .
considering the situation when the temperature is comparable to the asymmetry energy scale @xmath119 but the voltage is small @xmath175 , we get the visibilities @xmath176\ ] ] and @xmath177.\ ] ] the temperature dependence of the visibilities are shown in fig .
[ noisevis2 ] .
both visibilities decrease monotonically with increasing temperature . for large temperature @xmath128 , the visibilities decay exponentially as @xmath178 and @xmath179 .
the visibility @xmath180 is thus considerably more sensitivity to thermal smearing than @xmath181 . in the opposite limit , for a small temperature but a voltage comparable to @xmath119 , i.e. @xmath182
, we instead get the visibilities @xmath183 and @xmath184 the visibilities as a function of voltage are plotted in fig .
[ noisevis2 ] .
both visibilities show an oscillating behavior , decaying as a power law @xmath185 with increasing voltage .
the period of oscillations , in @xmath2 , is @xmath186 for @xmath187 but @xmath188 for @xmath180 , half the value for @xmath181 .
the different voltage dependence gives an additional possibility to investigate the two visibilities independently . in the experiment of ji
et al.@xcite the noise was measured in the high voltage regime , with the interference terms in both the current and noise completely suppressed .
the dependence of the incoherent noise on the transparencies @xmath189 and @xmath190 was investigated ( @xmath189 was kept at @xmath191 and @xmath190 was varied ) .
a good agreement was found with the first , incoherent term in eq .
( [ mznoisetot ] ) .
taken the open questions on the effect of decoherence on the average current , a detailed experimental investigation on the phase dependent , interference part of the noise would be of great interest . next we consider the effect of dephasing on the current and noise .
as discussed above , dephasing is introduced by connecting one of the two arms of the interferometer to a fictitious , dephasing voltage probe .
the interferometer with the probe , denoted terminal @xmath192 , is shown in fig .
[ mzdeph ] .
the dephasing probe is connected to the edge via a contact described by a scattering matrix @xmath193 where the dephasing parameter @xmath194 varies between @xmath195 ( no dephasing , fully coherent transport ) and @xmath50 ( complete dephasing , fully incoherent transport ) .
the presence of the dephasing probe modifies the amplitudes for scattering between the terminal 1 , 2 , 3and 4 . as an example , the scattering amplitude @xmath196 , given in eq .
( [ s41 ] ) in the absence of dephasing , now becomes @xmath197 in addition , amplitudes for scattering into and out from the probe terminal 5 have to be considered
. the current is obtained from eqs .
( [ avcurr ] ) , ( [ currspecdens ] ) and ( [ fav ] ) . for the current in terminal 4 ,
we find @xmath198.\end{aligned}\ ] ] comparison with the result in the absence of dephasing in eq .
( [ currmz ] ) shows that the effect of the dephasing is to suppress the phase - dependent oscillations by multiplying the phase - dependent interference term with a factor @xmath199 . for complete dephasing @xmath200 ,
the phase dependent term is completely suppressed .
the effect of dephasing can thus be simply incorporated in the visibility as @xmath201 where @xmath202 is the visibility of the current oscillations in the absence of dephasing , given by eq .
( [ currvismz ] ) .
as is clear from the discussion above , to account for the experimental observations in ref .
[ ] , one would have to consider a voltage dependent dephasing parameter @xmath194 . turning to the noise
, we obtain the cross correlator between currents in lead 3 and 4 in the presence of dephasing from eqs .
( [ noise ] ) and ( [ dephnoisedens ] ) , giving @xmath203 \right\}.\end{aligned}\ ] ] here the terms @xmath204 @xmath205 @xmath206 and @xmath207 are defined above in eqs .
( [ ccoeff ] ) and ( [ enfcn1 ] ) to ( [ enfcn2 ] ) .
similarly to the current , the effect of the dephasing is only to suppress the amplitude of the phase - dependent oscillations .
that is what one would naively expect to be the consequence of pure dephasing .
the two phase - dependent terms are however affected differently , the @xmath208 term is suppressed by a factor @xmath199 while the @xmath209 term is suppressed by @xmath210 .
the @xmath209 oscillations are thus more strongly suppressed .
the visibilities of the two oscillations in the presence of dephasing can simply be written @xmath211 and @xmath212 where @xmath181 and @xmath173 are the visibilities for the noise oscillations in the absence of dephasing , given by eqs .
( [ noisevismz1 ] ) and ( [ noisevismz2 ] ) , respectively .
importantly , both oscillating terms are fully suppressed for complete dephasing , @xmath200 .
complete dephasing within the voltage probe model thus gives a noise expression that only consists of the phase independent , incoherent term in eq .
( [ noisespecdens ] ) .
we note already here that the same result is found below for the hbt - interferometer .
since quantum interference by definition is excluded from the model , i.e. all scattering phases are neglected , the completely dephasing voltage probe thus constitutes a classical model that correctly reproduces the incoherent part of the noise . as pointed out above , a more detailed discussion of the physics of the voltage probe and a comparison with refs .
[ ] is deferred to a later work .
it is interesting to note that the effect of dephasing introduced with the voltage probe , both for the current and noise , is for arbitrary dephasing strength identical to a phase averaging .
the result in eqs .
( [ dephvismz1 ] ) and ( [ dephvismz2 ] ) can be obtained by averaging the fully coherent expressions in eqs .
( [ noisevismz1 ] ) and ( [ noisevismz2 ] ) with respect to a lorentzian distribution @xmath213 of slow fluctuations of the phase @xmath150 around the average value @xmath214 , as @xmath215 with the lorentzian distribution @xmath216 we note that , as pointed out in ref .
[ ] , a gaussian distribution of the phase fluctuations gives a different result , not consistent with the dephasing voltage probe approach for arbitrary dephasing stregth .
we emphasize that the results above are independent on to which edge the probe is connected .
moreover , we also point out that the effect of the voltage probes , for arbitrary @xmath194 , is multiplicative , i.e. attaching @xmath217 voltage probes at arbitrary places along the arms can be described by renormalizing @xmath218 . writing @xmath219)=\mbox{exp}(-l / l_{\phi})$ ] , with @xmath220 $ ] and @xmath221 with @xmath222 the distance between two probes , we can quite naturally incorporate the effect of a uniform distribution of probes into a dephasing length @xmath223 . the suppression of the visibilities of the @xmath0 and @xmath1 oscialltions due to dephasing in eqs .
( [ dephvismz1 ] ) and ( [ dephvismz2 ] ) are then modified as @xmath224 and @xmath225
the hbt - interferometer is less well known than the mz - interferometer and deserves some additional comments.@xcite the hbt - interferometer was invented as a tool to measure the angular diameter of stars .
the first measurement @xcite was carried out on a radio star in 1954 .
compared to existing schemes based on michelson interferometers , the hbt - interferometer proved to be less sensitive to atmospheric scintillations , which allowed for a more accurate determination of the angular diameter . after having demonstrated a table - top version of the interferometer in the visual range,@xcite the angular diameter of the visual star sirius was determined.@xcite the experimental results , both the two - particle interference and the positive intensity cross correlations , were successfully explained within a semi - classical framework .
soon after the experiments , it was however shown by purcell@xcite that the positive cross correlations could be explained in terms of bunching of individual photons , emerging from the star , a thermal source of light .
this bunching was also demonstrated explicitly in subsequent photo counting experiments.@xcite the hbt experiment thus laid the foundations for quantum statistical methods in quantum optics.@xcite the hbt approach has also been of importance in experimental particle physics.@xcite it is interesting to note that positive intensity cross correlations between beams of light emerging from a thermal source , according to some contemporary @xcite `` _ would call for a major revision of some fundamental concepts in quantum mechanics _ '' .
purcell,@xcite however , providing an elegant explanation of the bunching phenomena , pointed out that `` _ the hanbury brown twiss effect , far from requiring a revision of quantum mechanics , is an instructive illustration of its elementary principles _ '' .
an optical table - top version@xcite of the hbt - interferometer is shown in fig .
[ hbtopt ] .
a beam of light is emitted from each one of the sources @xmath226 and @xmath53 , completely uncorrelated with each other .
the beams are split in two partial beams at the semitransparent beam splitters @xmath227 and @xmath228 respectively .
the partial beams acquire phases @xmath229 to @xmath230 before scattering at the second pair of beam splitters @xmath51 and @xmath52 .
the resulting beams are collected in detectors at ports @xmath192 to @xmath231 .
importantly , there is no interference pattern in the intensities at the detectors @xmath192 to @xmath231 , instead the interference occurs only in the cross correlations between intensities at @xmath232 and @xmath233 .
the intensity cross correlations are sensitive to the two - particle amplitudes : the interference is thus between two different two - particle scattering events , e.g. ( i ) one particle from @xmath226 scatters to @xmath192 and one particle from @xmath53 scatters to @xmath231 , with an amplitude @xmath234)$ ] and ( ii ) one particle from @xmath226 scatters to @xmath231 and one particle from @xmath53 scatters to @xmath192 , with an amplitude @xmath235)$ ] .
the amplitude to detect one particle in @xmath192 and one in @xmath231 is then the sum of the two two - particle amplitudes .
this is the case since both scattering processes have the same initial and final states and can not be distinguished .
the ( reducible ) cross correlation between intensities in @xmath192 and @xmath231 is directly related to the corresponding two - particle probability @xmath236)+a_2\mbox{exp}(i[\phi_3+\phi_4])|^2$]= @xmath237)\right\}$ ] .
the interference term @xmath238)\right\}$ ] contains the four geometrical phases @xmath229 to @xmath230 .
the hbt - interferometer is thus , in contrast to the mz - interferometer , a two - particle interferometer .
the electrical analog of the hbt - interferometer , presented in ref .
[ ] , is shown in fig .
[ hbtel ] .
it consists of a ( rectangular ) conductor with a hole in the middle , a corbino geometry .
similar to the mz - interferometer , the electrons propagate along single edge states .
scattering between the edge states take place only at the beam splitters @xmath51 to @xmath228 .
the beam splitters are described by scattering matrices given by eq .
( [ smat ] ) .
we first consider the fully coherent case .
in contrast to the mz - interferometer , the scattering amplitudes contain the phases @xmath239 and @xmath240 only via multiplicative phase factors . as an example
, the amplitude to scatter from terminal @xmath226 to terminal 5 is given by @xmath241 as a consequence , the average currents which depend only on the modulus squared of the scattering amplitudes [ see eqs . ( [ avcurr ] ) and ( [ speccond ] ) ] do not contain any scattering phases .
we get the currents at terminals 5 to 8 as @xmath242 turning to the current noise , the correlation between currents in terminals 5,6 and 7,8 is given by eq .
( [ noisedens ] ) .
we find for the spectral density for the correlators between terminal 5 and 8 @xmath243 ^{2 } \notag \\ & \times & \left\{c_{0,58}+c_{\theta } \cos \left ( e / e_{c}+\theta \right ) \right\}\end{aligned}\ ] ] with the coefficients @xmath244 and for the correlator between terminal 5 and 7 @xmath245 ^{2 } \notag \\ & \times & \left\ { c_{0,57}+\bar c_{\theta } \cos \left ( e / e_{c}+\theta \right ) \right\}\end{aligned}\ ] ] with the coefficient @xmath246 performing the energy integrals in eq .
( 9 ) , we obtain the corresponding current cross correlators @xmath247 \label{hbtnoise1}\ ] ] and @xmath248 .
\label{hbtnoise2}\ ] ] here @xmath249 and @xmath250 are given by eqs .
( [ enfcn1 ] ) and ( [ enfcn3 ] ) .
the other two correlators @xmath251 and @xmath252 are given by the substitutions @xmath253 and @xmath254 . here , as for the mz - interferometer we have @xmath255 with @xmath256 and @xmath257
several observation can be made from the results above , put in comparison with the result for the noise correlations for the mz - interferometer in eq .
( [ mznoisetot ] ) . just as for the mz - interferometer , the noise consists of an incoherent , phase independent part , and a coherent , interference part .
the phase dependent part of the noise in eqs .
( [ hbtnoise1 ] ) and ( [ hbtnoise2 ] ) however contains only one term .
the amplitude of the oscillating term is a product of a scattering probability term and an energy - scale dependent function , just as for the mz - interferometer .
this phase dependent term has the same dependence on the phase @xmath150 , the same voltage dependent phase shift as well as the same energy - scale dependence as the second term in eq .
( [ mznoisetot ] ) .
this is the case since they both arise from processes which enclose the ab - flux once .
despite the fact that in the hbt interferometer the ab - effect results from two - particle processes , the periodicity is determined by the single electron flux quantum @xmath0 .
the dependence on the scattering probabilities is however different , a consequence of the mz and hbt interferometer geometries being different .
importantly , there is no term in the noise in eqs .
( [ hbtnoise1 ] ) and ( [ hbtnoise2 ] ) that corresponds to the last term in eq .
( [ mznoisetot ] ) , describing processes which enclose the ab - flux twice .
we note that the elementary scattering processes in the hbt - interferometer , in contrast to the mz - interferometer , are two - particle processes
. an important consequence of this is that the incoherent , phase independent noise term in eqs .
( [ hbtnoise1 ] ) and ( [ hbtnoise2 ] ) can directly be reproduced by a model with filled streams of classical particles incident from reservoirs @xmath226 and @xmath53 .
since there is only one phase - dependent term , the visibility of the phase - dependent oscillations can again be directly defined , giving for @xmath258 and @xmath259 @xmath260 since the energy - scale dependence of the visibilities is identical to @xmath261 for the mz - interferometer in eq .
( [ noisevismz1 ] ) , shown in fig .
[ noisevis2 ] , we focus here only on the scattering probability terms .
we thus consider the limit of a symmetric interferometer , @xmath262 for which the energy - scale dependent part is unity .
several symmetries exists , e.g. all visibilities @xmath263 are unchanged by the substitutions @xmath264 and @xmath265 .
the visibility @xmath266 is unity for scattering probabilities obeying @xmath267 and similar relations hold for the other visibilities .
all visibilities go to zero for any of the transmission probabilities approaching either zero or unity . focusing on the case with @xmath268 ( or equivalently @xmath269 ) , the visibilities are given by @xmath270 and @xmath271 the two different visibilities are plotted in fig .
[ hbtnoisevisfig ] as a function of @xmath272 for different @xmath273 .
the visibility @xmath266has a maximum equal to unity for @xmath274 , while @xmath275 instead has a maximum equal to unity for @xmath276 . just as in the mz - interferometer ,
the dephasing in the hbt - interferometer is introduced by connecting a fictitious voltage probe to an edge between any of the two point contacts .
the hbt - interferometer with the probe , denoted 9 , is shown in fig .
[ hbtdepfig ] . here
the probe is connected to the edge between contact c and a , we emphasize that the results discussed below do not depend on to which edge - state the probe is connected .
the presence of the probe modifies the amplitudes for scattering from terminals 2 , 3 to terminals 5 to 8 . as an example , the scattering amplitude in eq .
( [ s52 ] ) is modified @xmath277 in addition , we also have to consider amplitudes for scattering into and out from the probe terminal 9 .
the average currents in the presence of dephasing , given from eqs .
( [ avcurr ] ) to ( [ speccond ] ) and ( [ fav ] ) , turn out to be given by the same equations as in the absence of dephasing , i.e. eq .
( [ hbtcurr ] ) .
this is what one expects , i.e. that dephasing affects only the phase - dependent parts of the observables . turning to the current correlators , given from eqs .
( [ noise ] ) , ( [ noisedens ] ) and ( [ dephnoisedens ] ) , we find for the correlators between terminal 5 and 8 @xmath278 \label{hbtnoisedeph1}\ ] ] and for the correlators between terminals 5 and 7 @xmath279 . \label{hbtnoisedeph2}\ ] ] the two remaining correlators are again given by the substitutions @xmath280 and @xmath281 .
we see from eq .
( [ hbtnoisedeph1 ] ) and ( [ hbtnoisedeph2 ] ) that just as for the mz - interferometer , the only effect of dephasing is to suppress the phase - dependent term .
the suppression factor is @xmath199 , just the same as for the @xmath282 term in the noise for the mz - interferometer in eq .
( [ mznoisetot ] ) .
we can thus directly write the visibilities in the presence of dephasing as @xmath283 this leads to the conclusion that the voltage probe for the hbt - interferometer , just as for the mz - interferometer , just has the same effect as dephasing due to slow fluctuations of the phase @xmath284 , with the distribution of the phase fluctuations obeying the relation in eq .
( [ dephrel ] ) . moreover , the voltage probes have the same multiplicative property as for the mz - interferometer , allowing one to describe the effect of a continuum of probes along the edges ( of total length @xmath285 ) with a dephasing length @xmath223 . the suppression of the visibilities of the @xmath0 oscillations due to dephasing
are then modified as @xmath224 , just as for the @xmath0 oscillations of the mz - interferometer .
the mz - interferometer is an amplitude interferometer : it exhibits a visibility in the average current with period @xmath0 and exhibits a visibility in the shot noise with periods of both @xmath0 and @xmath1 . in contrast , the hbt interferometer is an intensity interferometer , it exhibits no ab - effect in the current and exhibits only an @xmath0-effect in the shot noise correlations .
interestingly , our investigation shows that the shot noise visibility of the hbt interferometer as a function of temperature , voltage and dephasing rate , is qualitatively similar to that of the @xmath0-component of the shot noise of the mz - interferometer .
this is contrary to the naive expectation that the visibility of the two particle processes which lead to the hbt effect should be similar to the visibility of the two particle processes in the mz - interferometer , that is the @xmath1 component of the shot noise .
instead it is the number of times the ab - flux is enclosed which determines the behavior of the visibility . in this paper
we have investigated and compared in detail the voltage , temperature and asymmetry dependence for the current and noise visibilities in the mz and hbt - interferometers .
the experimental realization of the hbt - interferometer is of large importance since it allows for an unambiguous demonstration of two - particle interference effects with electrons , to date not demonstrated .
moreover , a successful realization of the hbt - interferometer would also enable a first demonstration of orbital entanglement in electrical conductors , a fundamentally important result .
the results presented in this work should prove useful for the experimental work aiming to detect the hbt effect in electrical conductors .
we thank m. heiblum , i. neder , h. frster and e. sukhorukov for stimulating discussions . this work was supported by
the graduate students study abroad program , taiwan national science council and the taiwan nsc93 - 2112-m-009 - 036 , the swedish research council and the swiss national science foundation and the network for materials with novel electronic properties .
m. bttiker , phys .
b * 46 * , 12485 ( 1992 ) . ya .
blanter and m. bttiker , phys . rep . *
336 * 1(2000 ) .
m. born and e. wolf , _ principles of optics _ , 7:th ed .
( cambridge university press , uk , 1999 ) .
l. mandel , rev .
phys . * 71 * , s274 ( 1999 ) .
see e.g. s. pedersen , a.e .
hansen , a. kristensen , c.b .
sorensen and p.e .
lindelof , phys .
b * 61 * 5457 ( 2000 ) and references therein .
e. buks , r. schuster , m. heilblum , d. mahalu and v. umansky , nature * 391 * 871 ( 1999 ) .
a.e . hansen , a. kristensen , s. pedersen , c.b .
sorensen , and p.e .
lindelof , phys .
b * 64 * 045327 ( 2001 )
. k. kobayashi , h. aikawa , s. katsumoto and y. iye , j. phys .
* 71 * 2094 ( 2002 ) .
y. ji , y. chung , d. sprinzak , m. heilblum , d. mahalu and h. shtrikman , nature * 422 * , 415 ( 2003 ) . g. seelig , m. bttiker , phys . rev .
b * 64 * , 245313 ( 2001 ) .
g. seelig , s. pilgram , a.n .
jordan and m. bttiker , phys .
b * 68 * , r161310 ( 2003 ) .
k. le hur , phys .
b * 65 * , 233314 ( 2002 )
. k. le hur , cond - mat/0503652 .
f. marquardt and c. bruder , phys .
lett . * 92 * , 56805 ( 2004 ) .
f. marquardt and c. bruder , phys .
b * 70 * , 125305 ( 2004 ) .
f. marquardt , cond - mat/0410333 .
h. frster , s. pilgram and m. bttiker , cond - mat/0502400 .
r. hanbury brown and r.q .
twiss , philos . mag .
* 45 * , 663 ( 1954 ) .
r. hanbury brown and r.q .
twiss , nature ( london ) * 177 * , 27 ( 1956 ) .
r. hanbury brown and r.q .
twiss , nature ( london ) * 178 * , 1046 ( 1956 ) .
m. henny , s. oberholzer , c. strunk , t. heinzel , k. ensslin , m. holland and c. schnenberger , science * 284 * , 296 ( 1999 ) .
oliver , j. kim , r.c .
liu , and y. yamamoto , science * 284 * , 299 ( 1999 ) .
h. kiesel , a. renz , and f. hasselbach , nature ( london ) * 418 * , 392 ( 2002 ) .
m. bttiker , physica b * 175 * , 199 ( 1991 ) .
m. bttiker , phys .
. lett . * 68 * , 843 ( 1992 ) .
d. loss and e.v .
sukhorukov , phys .
* 84 * , 1035 ( 1992 ) .
p. samuelsson , e.v .
sukhorukov , and m. bttiker , phys .
lett . * 92 * , 26805 ( 2004 ) .
k. v. klitzing , g. dorda , m. pepper , phys .
* 45 * , 494 ( 1980 ) .
m. bttiker , phys .
b * 38 * , 9375 ( 1988 )
. b. i. halperin , phys .
b * 25 * , 2185 ( 1982 ) . b. j. van wees , h. van houten , c. w. j. beenakker , j. g. williamson , l. p. kouwenhoven , d. van der marel , and c. t. foxon , phys .
lett . * 60 * , 848 ( 1988 ) .
d. a. wharam , t. j. thornton , r. newbury , m. pepper , h. ahmed , j. e. f. frost , d. g. hasko , d. c. peacock , d. a. ritchie and g. a. c. jones , j. of phys .
c * 21 * , l861(1988 ) .
liu , b. odom , y. yamamoto , and s. tarucha , nature * 391 * 263 ( 1998 ) .
s. oberholzer , m. henny , c. strunk , c. schonenberger , t. heinzel , k. ensslin , m. holland , physica * 6e * , 314 ( 2000 ) .
m. bttiker , p. samuelsson , e.v .
sukhorukov , physica e * 20 * , 33 ( 2003 ) .
p. samuelsson , e.v .
sukhorukov , and m. bttiker , phys .
lett . * 91 * , 157002 ( 2003 ) .
beenakker , c. emary , m. kindermann , and j.l .
van velsen , phys .
lett . * 91 * , 147901 ( 2003 ) .
beenakker , and m. kindermann , phys .
lett . * 92 * , 056801 ( 2004 ) .
p. samuelsson , and m. bttiker , cond - mat/0410581 ( phys .
b , in press ) . see also related work by c.w.j .
beenakker , m. titov , b. trauzettel , phys .
lett . * 94 * , 186804 ( 2005 ) ; a. v. lebedev , g. b. lesovik , g. blatter , cond - mat/0504583 . t. m. stace , c. h. w. barnes , and g. j. milburn phys .
rev . lett . * 93 * , 126804 ( 2004 ) .
l. saminadayar , d. c. glattli , y. jin and b. etienne , phys .
lett . * 79 * , 2526.2529 ( 1997 ) .
r. de - picciotto , m. reznikov , m. heiblum , v. umansky , g. bunin , d. mahalu , nature * 389 * 162 ( 1997 ) .
i. safi , p. devillard , and t. martin , phys .
lett . * 86 * , 4628 ( 2001 ) .
s. vishveshwara , phys .
lett . * 91 * , 196803 ( 2003 ) .
s.a . kivelson and v.l .
pokrovsky , phys .
b * 40 * , r1373 ( 1989 ) . c. de c. chamon , d.e .
freed , s.a .
kivelson , s.l .
sondhi and x.g .
wen , phys .
b * 55 * , 2331 ( 1997 )
. m.r . geller and d. loss phys .
b * 56 * , 9692 ( 1997 ) .
t. jonckheere , p. devillard , a. crepieux , th .
martin , cond - mat/0503617 .
m. bttiker , j. phys .
condensed matter * 5 * , 9361 ( 1993 ) .
m. bttiker , phys .
* 57 * , 1761 ( 1986 ) .
m. bttiker , phys .
lett . * 65 * , 2901 ( 1990 ) .
m. bttiker , phys .
b * 33 * , 3020 ( 1986 ) .
m. bttiker , ibm j. res .
dev . * 32 * , 63 ( 1988 ) .
beenakker , and m. bttiker , phys .
b * 46 * , 1889 ( 1992 ) .
m. j. m. de jong , and c. w. j. beenakker , physica a * 230 * , 219 ( 1996 )
. s. a. van langen , and m. bttiker , phys .
b * 56 * , r1680 ( 1997 ) . c. texier and m. bttiker , phys
b * 62 * , 7454 ( 2000 ) . *
46 * , 1889 ( 1992 ) .
a.a . clerk and a.d .
stone , phys .
b * 69 * , 245303 ( 2004 ) . c.w.j .
beenakker and b. michaelis , cond - mat/0503347 .
d. b. chklovskii , b. i. shklovskii and l. i. glazman , phys .
b * 46 * , 4026 ( 1992 ) ; n. r. cooper and j. t. chalker , phys . rev .
b * 48 * , 4530 ( 1993 ) . h. a. fertig , phys .
b * 38 * , 996 ( 1988 ) .
see e.g. the discussion in n. c. van der vaart , m. p.
de ruyter van steveninck , l. p. kouwenhoven , a. t. johnson , y. v. nazarov , c. j. p. m. harmans , and c. t. foxon , phys .
lett . * 73 * , 320 ( 1994 ) and references therein .
d. sanchez and m. bttiker , phys .
rev . lett . * 93 * , 106802 ( 2004 ) ; t. christen and m. bttiker , phys . rev .
b * 53 * 2064 , ( 1996 ) . for an interesting historical account , see r. hanbury brown , _ the intensity interferometer _
( london : taylor and francis ) 1974 .
purcell , nature * 178 * , 1449 ( 1956 ) .
twiss , a.g .
little , and r. hanbury brown , nature ( london ) * 180 * , 324 ( 1957 ) . g.a .
rebka , and r.v .
pound , nature ( london ) * 180 * , 1035 ( 1957 ) .
see e.g. r. loudon , rep .
, * 43 * , 58 ( 1980 ) .
g. baym , acta phys . pol .
b * 29 * , 1839 ( 1998 ) .
e. brannen and h.i.s .
ferguson , nature ( london ) * 178 * , 481 ( 1956 ) .
r. hanbury brown and r.q .
twiss , nature ( london ) * 178 * , 1447 ( 1956 ) .
b. yurke and d. stoler , phys .
lett . * 68 * , 1251 ( 1992 ) .
b. yurke and d. stoler , phys .
a * 46 * , 2229 ( 1992 ) .
|
we investigate the visibility of the current and shot - noise correlations of electrical analogs of the optical mach - zehnder interferometer and the hanbury brown twiss interferometer .
the electrical analogs are discussed in conductors subject to high magnetic fields where electron motion is along edge states .
the transport quantities are modulated with the help of an aharonov - bohm flux .
we discuss the conductance ( current ) visibility and shot noise visibility as a function of temperature and applied voltage .
dephasing is introduced with the help of fictitious voltage probes .
comparison of these two interferometers is of interest since the mach - zehnder interferometer is an amplitude ( single - particle ) interferometer whereas the hanbury brown twiss interferometer is an intensity ( two - particle ) interferometer .
a direct comparison is only possible for the shot noise of the two interferometers .
we find that the visibility of shot noise correlations of the hanbury brown twiss interferometer as function of temperature , voltage or dephasing , is qualitatively similar to the visibility of the first harmonic of the shot noise correlation of the mach - zehnder interferometer .
in contrast , the second harmonic of the shot noise visibility of the mach - zehnder interferometer decreases much more rapidly with increasing temperature , voltage or dephasing rate .
|
the basic offline problem of energy - efficient packet transmission scheduling @xcite is to assign code rates ( consequently transmission durations ) to a set of packets whose arrival times are known beforehand , so that they are all transmitted within a given time window with minimum total energy .
the solution needs to strike a tradeoff between energy and delay based on the observation that energy per bit with many ideal and suboptimal coding schemes is convex and monotone increasing with rate .
recently , the problem has been reformulated with a model where energy gets harvested " or replenished at certain known instants @xcite .
while in the former formulations transmission rate needs to be adapted to the arrival rate of information here it is adapted to the generation rate of energy . considering both of theses
effects introduces a richness to the problem on top of the initial model .
the point - to - point problem in @xcite was recast for finite energy storage @xcite and for a wireless fading channel @xcite .
the formulation has been extended to an awgn bc in @xcite , considering a static pool of data to be sent at the beginning of the schedule .
the same bc problem was also studied under a limited battery constraint @xcite .
the problem in @xcite and @xcite is reformulated in @xcite relaxing the assumption that data is ready at the beginning of the schedule .
this paper , extending the work in @xcite , considers the broadcast problem where , given an average transmit power constraint , rates are picked from an achievable rate region which obeys certain structural properties satisfied by the awgn bc .
the sender ( transmitter ) gets replenished with arbitrary amounts of energy as well as data packets of arbitrary length destined to each user at arbitrary points in time .
the choices of power level and the rates to individual receivers across time is called a _ schedule_. an optimal scheduling policy is defined to be one that transmits all the bits that have arrived within a certain time window , in the minimum possible amount of time @xmath0 .
the policy is allowed to use as many energy harvests as it needs , provided it respects causality ( no energy is used before it is harvested . ) the problem considered in this paper is an _ offline _ problem , where data arrival and energy harvest instants and amounts are assumed to be known in advance . although this kind of prior information of data and energy arrivals may not be a widely applicable assumption to real - world problems , results obtained from this work help us understand the nature of an optimal solution and boundaries on the best performance .
online formulations have also appeared in the literature .
notably , @xcite develop online scheduling policies for multihop networks on finite - horizon and infinite horizon problem formulations , respectively . to minimize the overall transmission duration , rates to the individual users should be chosen in a way such that transmission is fast and energy efficiency is satisfied . however , as we choose higher rates , we loose from energy efficiency .
balancing between fast and energy efficient transmission , the decision of rates needs to be based on the sequence of energy harvests and data arrivals .
this paper essentially bridges the work that considered scheduling on a broadcast channel _ data _ that becomes available at arbitrary points in time @xcite and work that considered chunks of _ energy _ becoming available at arbitrary points in time @xcite .
it can also be viewed as the multiuser correspondent of the second problem considered in @xcite .
the main contributions are : ( 1 ) to establish structural properties of the optimal schedule , ( 2 ) the uniqueness of the optimal policy when all weaker user data is ready at the beginning , ( 3 ) an iterative algorithm which returns the optimal schedule under the same condition .
it is shown in section [ sec : sm ] that in an optimal policy , transmit power used is constant within each epoch , and may only rise from one epoch to the next , so that once it starts , the transmitter never lowers its power until it finally goes silent . on the other hand , the transmitter should increase its power only under certain conditions .
these conditions , along with other structural properties of power and rate are established in section [ sec : structure ] .
next , the uniqueness of the optimal policy is established under the condition that all of the weaker user s bits are available at the beginning .
finally , an iterative algorithm ( that we refer to as duopt ) based on the nonlinear block descent method which returns a feasible schedule carrying the same structural properties that the optimal is shown to have , is described .
it has been shown that duopt returns the optimal schedule in case of static pool of weaker user data at the beginning of the schedule .
we start by giving the problem statement in the next section .
consider a broadcast channel with one transmitter and two receivers .
arbitrary amounts of energy , @xmath1 , as well as data for each user @xmath2 become available to the sender at arbitrary times @xmath3 . a possible sequence of data and energy arrivals is illustrated in fig . [ fig : system_model ] .
@xmath4 denotes the total energy that has been _ harvested _ in @xmath5 ( regardless of how much of it has been used . ) similarly , @xmath6 and @xmath7 denote the total number of bits destined to the first and second user , respectively , that the sender has obtained in @xmath5 .
the interval between any two sequential arrival events ( regardless of energy or data ) will be called an inter - arrival _ epoch_. the length of the @xmath8 epoch is @xmath9 . in this offline problem ,
all the future arrival times and amounts of energy and bits are known by the sender at @xmath10 .
it is also assumed that harvested energy and data are available for use instantaneously as they arrive , and code rate and transmission power decisions can be changed instantaneously .
however , codeword block lengths will be chosen such that each codeword is sent completely within a single epoch ( note that starting and ending times of epochs are known ahead of time ) , so that no arrival event occurs during a codeword .
consequently , the power and rate pair decision will be fixed throughout each codeword .
we are interested in minimizing the total transmission time for packets arriving by a certain time @xmath11 , so w.l.o.g .
, set @xmath12 for @xmath13 , @xmath14 . a schedule , which is a sequence of power and rate allocations , is feasible if it sends @xmath15 bits to the @xmath16 user and @xmath17 to the @xmath18 user ( with a certain level of reliability ) , without violating causality ( at any time , using available energy and data by that time ) .
we are interested in finding among all feasible schedules one with the smallest completion time , @xmath0 .
the structure of the achievable rate region will be based on the two - user awgn bc .
the capacity region of a two - user discrete time awgn bc with average power constraint @xmath19 , noise variance @xmath20 , where the @xmath16 user s channel gain ( @xmath21 ) is larger than the @xmath18 user s ( @xmath22 ) , consists of rate pairs @xmath23 satisfying : @xmath24 where @xmath25 , ( @xmath26 ) , denotes the fraction of @xmath19 used for the @xmath16 user . since @xmath27 , the @xmath16 user will be referred as the stronger user " , and the @xmath18 as the weaker user " . from ( [ eq : broadcast_channel ] ) each user s rate can be expressed as a function of the other s and @xmath19 , as @xmath28 , @xmath29 . the rate functions @xmath30 and @xmath31 defined on @xmath32 will be assumed to satisfy the following properties : 1 .
nonnegativity : @xmath33 .
monotonicity : @xmath34 , @xmath35 are both monotone decreasing in @xmath36 , and monotone increasing in @xmath19 .
concavity : @xmath34 , @xmath35 are concave in @xmath19 and @xmath36 : @xmath37,@xmath38 , for @xmath39 .
4 . @xmath40 , @xmath41 5 .
@xmath42 , @xmath43 the results in the rest of the paper will be valid for any rate function satisfying ( 1)-(5 ) , which are also satisfied by the awgn bc @xcite .
it is straightforward to show that one can restrict attention to feasible schedules that do not change their power and rate allocations within epochs .
[ lmm : constantpowerrate ] in an optimal schedule , the power and rate pair remain constant within all epochs , except for the epoch during which the schedule ends .
_ during an epoch , there are no energy or data arrivals and the claim is identical with the one stated and proved in lemma 2 of @xcite .
the power will drop to zero when the schedule ends , which is in general within ( and not necessarily at the end of ) the last epoch used by the schedule . ' '' '' with this , we will take rate and power assignments constant during an epoch .
let @xmath44 be the total transmit power and @xmath45 be the rate assigned to the @xmath46 user during epoch @xmath47 .
similarly , @xmath48 represent the power assigned to @xmath46 user during epoch @xmath47 .
we are now ready to state the problem in terms of power and rate allocations to epochs , more precisely , an assignment of power and the stronger user s rate to each epoch ( the weaker user s rate is thus determined )
. a final technical assumption will be useful in stating the problem : we shall assume that there is some @xmath49 such that there is at least one feasible schedule that ends within the first @xmath50 epochs .
in other words , @xmath50 is an upper bound for epochs to be considered . in problem statement
, @xmath51 denotes the last epoch of an optimal schedule , where @xmath52 .
+ [ pr : multiuserscheduling ] * transmission time minimization of data arriving at arbitrary points on an energy harvesting bc : * @xmath53 we will refer to and as energy and data causality constraints , respectively , as these ensure no energy is consumed and no bit is transmitted before becoming available .
in addition , when the @xmath54 inequality in holds with equality , we shall say that @xmath54 _ energy constraint is active_. similarly , equality case in will be referred as a _
data constraint being active_. finally , the feasibility constraint , ensures all the data bits destined to each user are transmitted . in the next section ,
we investigate structural properties that any optimal schedule has to satisfy .
lemma [ lmm : constantpowerrate ] recorded that in an optimal schedule power can only change upon a data arrival or energy harvest . the next result states that when power changes , it can only increase .
the key to the proof is that more bits per joule " can be sent by evenly distributing energy across a time interval ( , maintaining a constant power level , which is a consequence of the convexity properties of our rate functions . )
if an even distribution of power requires transferring energy or bits to the latter epoch , it can always be done ; hence , total transmit power never decreases in time .
but , power may increase in time , because even distribution of power may result in unmet causality constraints .
we state these results in lemma [ lmm : powernondecreasing ] . due to space constraints
, the proofs of the following results ( lemma [ lmm : powernondecreasing ] through lemma [ lmm : constantrate1 ] ) are omitted , and given in @xcite .
[ lmm : powernondecreasing ] _
( for proof see @xcite ) _ consider an optimal schedule that ends during epoch @xmath51 .
power is non - decreasing with epoch index , i.e , @xmath55 for @xmath56 .
as stated in lemma [ lmm : powernondecreasing ] power can not decrease , yet may rise in time . in the next lemma we note what is necessary condition for such a rise to occur in an optimal policy .
[ lmm : closerpowers ] _
( for proof see @xcite ) _ in an optimal policy , power can only rise at @xmath3 ( end of epoch @xmath47 ) if at least one of the conditions below holds : 1 . _
energy constraint is active at point @xmath3 .
( the @xmath57 energy constraint is active ) _ [ cnd : a ] 2 . _
the data constraints for both users are active at point @xmath3 .
( , the set of constraints in ) _ [ cnd : b ] 3 . _
the weaker user s data constraint is active and data arrival to the weaker user occurs at time @xmath3._[cnd : c ] the next set of results illustrate the structure of _ rate allocation _ in conjunction with the power allocation in an optimal policy .
[ corollary : powerincrease ] _
( for proof see @xcite ) _ in an optimal policy , 1 .
if power increases upon a data arrival for the second user , data to be sent to the weaker user have been finished by this event .
if power rises upon a data arrival for the stronger user , all available bits have been sent by this event .
if power increases upon an energy harvest , all energy available at the beginning of the former constant power band has been consumed by this energy harvest . in the rest
, some properties will be proved under the condition that all weaker user data is available at the beginning .
we shall abbreviate this condition as follows : + _ weaker user full buffer condition ( wufbc ) _ is said to be satisfied whenever all of the data of the weaker user is available at the beginning of transmission .
that is , @xmath58 .
the following lemma states an important feature of the stronger user rate distribution under wufbc .
[ lmm : constantrate1 ] _
( for proof see @xcite ) _ consider two consecutive epochs @xmath47 and @xmath59 of a given schedule , ending at @xmath60 and @xmath61 by definition , and suppose wufbc holds for the problem instance .
the following is necessary for the rate and power allocation to these two epochs of the given schedule to be locally optimal : the stronger user s rate is constant throughout @xmath62 , and @xmath63 .
furthermore , the rate may jump up at @xmath64 ( staying constant otherwise ) if at least one of the below is true : 1 .
there is data arrival to the stronger user at @xmath64 and all the data that arrived before @xmath64 has been transmitted by @xmath60 .
2 . an energy harvest occurs at @xmath64 and all of the power has been used for the stronger user during epoch @xmath47 . we investigate the unique solution of problem [ pr : multiuserscheduling ] in the next section .
in the following lemma we note that an optimal schedule uses all energy harvested by the time the schedule ends completely .
[ lmm : consumed_energy ] the energy consumed by an optimal schedule that ends at @xmath65 is equal to @xmath66 . _ proof . _ to reach contradiction , consider an optimal schedule that consumes less energy than it harvested and has leftover energy in its energy buffer at @xmath65 .
the remaining energy in the buffer could have been used in the last epoch to decrease the transmission completion time , which contradicts the minimality of @xmath65 .
hence , this schedule can not be optimal . ' '' ''
next , we show the uniqueness of the optimal schedule under wufbc .
[ thm : unique_schedule ] _ there is a unique optimum schedule under wufbc , , a unique power - rate allocation achieving @xmath65 . _ _ proof . _
suppose that there are two distinct optimal schedules , @xmath67 and @xmath68 , which have equal power and rate assignments until @xmath69 and differ for the first time at epoch @xmath70 .
consider that the corresponding power allocation vectors , @xmath71 and @xmath72 , also differ at epoch @xmath70 such that @xmath73 and @xmath74 .
first , assume that @xmath71 remains constant after epoch @xmath70 , , @xmath75 . by definition ,
both schedules end at @xmath0 .
the total energy consumption of @xmath67 would be less than that of @xmath68 by @xmath0 , , @xmath76 , which contradicts lemma [ lmm : consumed_energy ] .
hence , total transmit power of @xmath67 can not remain constant after @xmath77 .
since total transmit power is nondecreasing ( see lemma [ lmm : powernondecreasing ] ) , it should increase after epoch @xmath70 and before the end of transmission , , @xmath78 .
since there are no data arrivals for the weaker user , the increase in total transmit power is either due to energy constraint being met or due to _ all _ the packets arrived by the time @xmath79 having been transmitted ( cf . conditions ( a ) or ( c ) in lemma [ lmm : closerpowers ] ) . as @xmath80
, @xmath67 has not consumed all the available energy at the end of epoch @xmath81 .
hence , @xmath67 must have transmitted all the bits arrived until @xmath82 , which means that @xmath67 has transmitted at least the same number of bits to both users while consuming less energy than @xmath68 between @xmath83 and @xmath82 , which contradicts the optimality of @xmath68 .
therefore , if there are two distinct optimal schedules , @xmath67 and @xmath68 , their power allocation vectors can not be different , , @xmath84 .
now , consider two rate pair vectors , @xmath85 and @xmath86 , where @xmath87 and @xmath88 .
let the rate of the stronger user in @xmath67 , @xmath89 stay constant after @xmath69 . by lemma [ lmm :
constantrate1 ] rate of the stronger user can not decrease , hence the rate of the stronger user in @xmath68 would be larger than that of @xmath67 after epoch @xmath70 , , @xmath90 . since both schedules end transmission at the same time
, @xmath67 transmits fewer bits to the stronger user than @xmath68 does , which contradicts the fact that optimal schedule transmits all the packet arrivals by the end of transmission .
therefore , the rate of the stronger user in @xmath67 can not stay constant after epoch @xmath70 .
now suppose that rate of the stronger user in @xmath67 increases at the end of epoch @xmath81 , , @xmath91 .
this increase can not be due to ( 1 ) in lemma [ lmm : constantrate1 ] because @xmath68 has transmitted more bits to the stronger user by @xmath82 , , @xmath92 .
moreover , this increase can not be due to ( 2 ) in lemma [ lmm : constantrate1 ] since rate of the weaker user in @xmath67 is greater than zero in epoch @xmath81 , , @xmath93 .
hence rate of stronger user in @xmath67 can not increase after epoch @xmath70 .
finally , rate of the stronger user in @xmath67 can not _ decrease _ ( see lemma [ lmm : constantrate1 ] ) as this would also contradict optimality .
hence , there can not be two optimal schedules with different rate pair vectors .
as both the power allocation vector and the rate pair vector of an optimal schedule are unique , we conclude that the optimal schedule is _ unique _ under wufbc . ' '' ''
the problem in @xcite which is a special case of problem [ pr : multiuserscheduling ] , where both users data is available at the beginning , was shown to be solved in @xcite by the _
algorithm @xcite .
along similar lines , we develop an algorithm that we call _ duopt _ for solving problem [ pr : multiuserscheduling ] in its general form . as a matter of fact , _ duopt _ simply reduces to _ flowright _ when the given problem instance has all the data arriving at @xmath94 . similarly to _ flowright _ , _ duopt _ starts with any feasible schedule and reduces the transmission completion time iteratively .
let the number of epochs and the transmission completion time of the initial schedule be @xmath95 and @xmath96 respectively .
in each iteration , _ duopt _ sequentially updates rates and powers of two consecutive epochs at a time , , epochs @xmath97 , until all epochs are updated .
then , starting from the first epoch pair , _ duopt _ continues with the next iteration .
_ duopt _ stops after @xmath98 iterations such that @xmath99 , where @xmath100 is the transmission completion time , @xmath101 is the number of epochs used at the end of @xmath102 iteration and @xmath103 is a predefined threshold .
hereafter , we will briefly outline the local optimizations over epoch pairs . in theorem [ thm : duopt_stops
] , it will be shown that local optimizations can only improve the schedule .
we will also prove that under wufbc , successive iterations strictly improves the schedule unless it is optimal .
let @xmath104 denote the energy used during the @xmath8 epoch and @xmath105 denote the number of bits transmitted to the @xmath46 user during epoch @xmath47 at the end of @xmath102 iteration .
suppose that _ duopt _ is at the @xmath102 iteration and running a local optimization over epoch pair @xmath106 .
the values of @xmath107 and @xmath108 , @xmath109 have already been found by previous local optimizations . at the end of this optimization , @xmath104 and @xmath105 will be determined ; @xmath110 , @xmath111 and @xmath112 will be reset to new values that conserve total energy consumption and data transmission in these epochs .
the goal of the local optimization is surely to minimize the total transmission completion time of all the packet arrivals .
hence , it is logical to minimize the transmission time in the local optimization problem , which results in a gap if transmission ends before the end of @xmath113 epoch .
this gap is used in the next local optimization to further reduce the transmission time via transferring bits or energy between epochs @xmath114 and @xmath115 ; hence , a new gap occurs at the end of the next local optimization .
this new gap propagates to the end of the transmission resulting a reduction in the total transmission completion time @xcite .
however , in some cases an epoch long gap occurs and this gap is useless for the next local optimization , , energy or data transfer between epochs in the next local optimization is impossible because of constraints . in that case , it is better to just spread the data out till the end of the second epoch in the local problem and minimize the energy consumption so that the excess energy can be used to further reduce the transmission time in the next local optimization .
this leads to two different local optimization functions : _ time minimization _ and _ energy minimization_. these functions both support the global objective in different ways .
time minimization aims to find the minimum amount of time , @xmath116 , to transmit @xmath117 bits to each user using the energy available in epoch pair @xmath106 , , @xmath118 . on the other hand
, energy minimization aims to find the minimum energy , @xmath119 , to transmit @xmath120 bits to each user in two epoch durations , , @xmath121 , and excess energy , @xmath122 , is transferred to the @xmath123 epoch in order to conserve energy . both of the optimizations respect energy and bit causalities , , @xmath124 and @xmath125 , @xmath126 . for details of the local optimization ,
see @xcite .
+ suppose that all the feasible packets have been transmitted until the end of the @xmath8 epoch and there are still packets to arrive after @xmath60 . then , further minimization of transmission completion time of sequential epochs before @xmath3 will be suboptimal . on the other hand
, we can minimize the energy consumption until @xmath3 and use the excess energy to minimize the transmission completion time . therefore
, utilization of energy minimization for local optimizations in problem [ pr : multiuserscheduling ] is very crucial . if it is guaranteed that current schedule uses at least the same amount of energy as optimal schedule until @xmath3 , _ duopt _ uses the energy minimization function upto @xmath8 epoch pair and the time minimization function for the rest . in order to determine when to switch from energy minimization to time minimization
, a _ flag _ is placed at @xmath8 epoch pair .
initially , the _ flag _ is set to zero and _ duopt _ starts with performing time minimization on epoch pairs . during @xmath102 iteration ,
if all the feasible bits are transmitted by the @xmath8 epoch for @xmath127 , then the _ flag _ is set to @xmath47 @xmath128 . in the following iterations ,
energy minimization function is used up to @xmath8 epoch pair .
[ fig : duoptflag ] illustrates the _ flag _ usage and the pseudo - code in algorithm [ alg : duopt ] outlines the duopt algorithm .
+ initialize ( ) ; @xmath129 @xmath130 @xmath131 @xmath132 @xmath133 ` calculate_t(&`@xmath134 ` ) ` [ thm : duopt_stops ] _ following statements hold : _ 1
. _ successive iterations of duopt can only improve the schedule .
_ duopt stops and returns a schedule with @xmath135 . _ _ proof . _ 1 .
suppose that duopt is running its @xmath102 iteration .
after the local optimization on @xmath8 epoch pair , we obtain @xmath136 and reset the values of @xmath137 .
if the _ flag _ is not placed before @xmath8 epoch pair , , @xmath138 , then the aim of the local optimization will be energy minimization . following the local optimization on @xmath8 epoch pair
, the excess energy will be transferred to @xmath111 . in the next local optimization
this excess energy is either further transferred or is used to reduce the transmit time . on the other hand , if @xmath139 , then the aim of the local optimization on @xmath8 epoch pair will be time minimization . after the local optimization
the transmission completion time of the bits in epochs @xmath106 will either be equal to or before the end of the epoch @xmath114 .
that is , a gap may occur within @xmath8 epoch pair . in the next local optimization , this gap would propagate to the @xmath123 epoch @xcite . during the @xmath102 iteration of _ duopt
_ , if a gap occurs or excess energy is transferred during local optimizations , then the gap ( or the excess energy respectively ) will propagate to the last epoch pair resulting in an ultimate reduction the transmission completion time at the end of the iteration , , @xmath140 .
if neither excess energy nor a gap occurs during local optimizations , then transmission completion time can not be decreased and _ duopt _ will stop by definition . +
both local optimizations are in favor of the next local optimization .
therefore , if in either one of the local optimizations a gap occurs or excess energy is transferred then it would propagate till the last epoch pair and finally the transmission completion time would decrease at the end of @xmath141 iteration , , @xmath140 . if neither excess energy nor gap occurs during local optimizations , then transmission completion time would not be decreased and _ duopt _ would stop . + ' '' '' 2 .
in part-1 we have shown that transmission completion time , @xmath142 , is strictly decreasing in each iteration ; meanwhile it is bounded below by @xmath143 .
therefore , the iterations of _ duopt _ stop and return a schedule @xmath135 .
+ ' '' '' [ thm : duopt_optimal ] _ if wufbc is guaranteed , the schedule returned by _ duopt _ is optimal , , _
@xmath144 .
_ suppose that _ duopt _ stopped and returned a schedule @xmath145 , with completion time @xmath146 .
let @xmath147 be the unique optimal schedule with transmission completion time @xmath148 .
we will now prove that @xmath149 .
let us suppose @xmath150 , then these schedules have to differ in either the power allocation or rate allocation ( or both ) .
first , suppose @xmath151 and @xmath152 for some @xmath70 .
we will show that this case is impossible .
there are two possible cases for epoch @xmath70 : ( i ) @xmath153 , ( ii ) @xmath154 . let us begin with the first case . a. [ prf : item1 ] we assumed @xmath153 .
if @xmath155 stays constant after epoch @xmath70 till the end of transmission , this would mean that @xmath156 consumes more energy than @xmath147 until @xmath148 , which would contradict the fact that optimal schedule consumes all the harvested energy till the end of transmission .
therefore the power of the optimal schedule must increase at the end of epoch @xmath157 for some @xmath158 before the end of transmission .
as @xmath156 has been able to use more energy than the optimal schedule until @xmath159 , the optimal schedule can not have run into an energy constraint at @xmath159 , hence the rise in the power can only be due to a data constraint at @xmath159 , , all the bits arrived have been transmitted by the optimal schedule until @xmath159 . in order to contradict the assumption that @xmath153 , we will now analyze the rate assignments for both schedules .
first let us focus on the case that both schedules use exactly the same rates for the stronger user up to @xmath159 , , @xmath160 .
as we have shown above , @xmath147 should have transmitted all the bits available until @xmath159 . however ,
if we compare the weaker user bits transmitted by both schedules until @xmath159 , we observe that @xmath156 transmits more bits to the weaker user than @xmath147 does , because @xmath161 . on the other hand , _ duopt _ respects _ bit causality _
, , _ duopt _ does not transmit bits that have not arrived yet , so we reach contradiction .
that is , rates can not stay constant up to @xmath159 , , there is some @xmath162 such that @xmath163 for @xmath164 and @xmath165 .
but we shall now show that this is not possible .
first consider the case that @xmath166 . from lemma [ lmm : constantrate1 ]
, the stronger user s rate can not decrease under wufbc .
if @xmath167 , then @xmath156 transmits more bits to the weaker user than @xmath147 does by @xmath159 , , @xmath168 .
however , at the end of @xmath169 epoch , @xmath156 can not send more bits to weaker user because @xmath147 should have transmitted all the weaker user bits .
therefore , @xmath170 should increase before @xmath159 , , at the end of epoch @xmath171 , where @xmath172 .
we have @xmath173 , hence either one of the two conditions in lemma [ lmm : constantrate1 ] must hold . since @xmath174 , until @xmath175 , @xmath147 has transmitted more bits to the stronger user than @xmath156 does ; therefore , all the stronger user s bits arrived have not been transmitted by @xmath156 at the end of epoch @xmath176 . also , @xmath177 . hence neither of the two conditions in lemma [ lmm : constantrate1 ] holds and stronger user s rate can not increase at @xmath175 , which implies @xmath178 .
thus , we are left with the case @xmath179 . if @xmath180 , then @xmath181 , which contradicts the fact that duopt respects bit feasibility .
hence , stronger user s rate in @xmath147 can not remain constant after epoch @xmath182 .
then we should have @xmath183 and @xmath184 . since there is an increase in the stronger user rate , at least one of the conditions in lemma [ lmm : constantrate1 ] should hold at @xmath175 .
however , we have @xmath185 and @xmath186 , which tells us that neither one of the conditions in lemma [ lmm : constantrate1 ] holds , which implies that this final case is also not possible .
hence , we conclude that the case @xmath153 is not possible .
b. now consider the case @xmath154 .
we will prove that this case is also not possible by following a similar method to the one in case ( [ prf : item1 ] ) .
first , suppose that the power of @xmath156 increases after @xmath187 epoch .
this increase can not be due to an energy constraint , since @xmath147 consumes more energy than @xmath156 does until the increase in power .
hence , it should be due to data constraint and under wufbc both user data constraints should be active .
that is , @xmath156 transmits all the feasible data until the increase in power .
this implies that while consuming less energy , @xmath156 transmits at least the same number of bits than @xmath147 does , which contradicts the optimality of @xmath147 .
thus , power of @xmath156 can not increase after epoch @xmath70 .
also , it can not decrease in time , otherwise a local optimization results in either a gap or excess energy that propagates till the end of the schedule and transmission duration decreases .
therefore , we power of @xmath156 should stay constant after epoch @xmath70 until @xmath148 .
now , we will analyze the rate assignments for both schedules .
let the transmission of @xmath147 end in epoch @xmath157 for @xmath158 and suppose that @xmath188 . at the @xmath54 epoch
there are three possible cases : @xmath179 , @xmath166 and @xmath189 .
we will first consider the case @xmath179 and prove that this is not possible .
let @xmath179 and consider the rate of the stronger user in @xmath147 after @xmath54 epoch .
it can not stay constant until @xmath148 , because it contradicts the fact that @xmath147 transmits all the feasible bits before @xmath148 , , @xmath181 .
since the stronger user s rate in @xmath147 can not decrease by lemma [ lmm : constantrate1 ] , it should increase at the end of epoch @xmath176 for @xmath190 , , @xmath184 .
however , we have @xmath186 and @xmath191 which implies that none of the conditions in lemma [ lmm : constantrate1 ] holds and the stronger user s rate in @xmath156 can not increase after epoch @xmath182 . hence , we conclude that @xmath192 .
now we consider the case @xmath166 .
suppose that the stronger user s rate in @xmath147 increase at epoch @xmath176 for @xmath190 .
this increase in stronger user s rate requires that at least one of the conditions in lemma [ lmm : constantrate1 ] should hold .
however , we have @xmath186 and @xmath191 , so the stronger user s rate in @xmath156 can not increase , , @xmath192 . since the stronger user s rate in @xmath147 can not decrease by lemma [ lmm : constantrate1 ]
, it should stay constant until @xmath148 .
+ thus far we have shown that if @xmath156 is different than @xmath147 , then @xmath156 can not have higher power level than @xmath147 until @xmath148 . moreover , if power level of @xmath156 becomes lower than that of @xmath147 , then it should stay constant until @xmath148 and if the stronger user s rate of @xmath156 becomes lower than that of @xmath147 , then it should stay constant until @xmath148 .
these results are shown in the general case in fig .
[ fig : duopt_vs_optimal ] .
+ now , let @xmath193 and @xmath194 be the number of bits transmitted to the weaker user till @xmath195 by @xmath156 and till @xmath148 by @xmath147 , respectively .
then , we have @xmath196 + from ( [ eqn : bitfeasibility_5 ] ) , @xmath156 transmits more bits to the weaker user than @xmath147 does , then this final case also can not happen .
therefore , we conclude that the schedule returned by _
duopt _ can not be different than the unique optimal schedule , , @xmath149 .
+ ' '' ''
consider a two - user awgn bc with 1khz bandwidth and @xmath197 watts / hz . path loss factors on the links of stronger and weaker user are assumed to be 70db and 75db , respectively .
amounts and instants of energy harvests and bit arrivals are depicted in fig .
[ fig : numex ] . under these circumstances
duopt algorithm is run and final schedule is calculated as drawn in fig .
[ fig : numex ]
with the new advances in energy harvesting technologies , optimization of communication systems that depend on renewable energy resources has emerged as an important problem .
a large body of recent research effort in the field has focused on transmission scheduling policies @xcite .
this paper continued the work on to the solution of the previous formulation of the problem stated in @xcite and @xcite .
in particular , it aimed to find the power and rate allocation policy in a broadcast channel with two users , that minimizes the total transmission completion time of data that becomes available at arbitrary points in time , with energy that is harvested at arbitrary points in time .
it should be noted that this line of work relies on offline problem formulations where exact knowledge of data and energy arrival events is assumed .
the approach is useful for obtaining structures and guidelines as well as benchmarks for online problems which may be more applicable in practical transmision scenarios . in this study ,
structural properties about the solution of the general optimal offline broadcast packet scheduling policy for an energy harvesting broadcast channel have been established .
the uniqueness of the optimal policy under the weak - user - full - buffer condition ( wufbc ) has been shown .
an iterative algorithm , duopt , that returns a feasible schedule which possesses the same structural policies that the optimal was shown to have is devised .
moreover , duopt was shown to obtain the unique optimal policy under wufbc .
proving the uniqueness of the optimal policy and optimality of duopt in general are two goals of ongoing work .
the direction set for future work related to the problem presented in this paper also includes online problems .
the claim is that power is non - decreasing with epoch number @xmath47 .
equivalently power is non - decreasing in time . to show this , we will argue that given a schedule in which power decreases at some time @xmath3 , this schedule can only be improved by equating the power levels before and after @xmath3 . consider a time interval @xmath198 , so that power is constant at @xmath199 during @xmath200 , and at @xmath201 during @xmath202 .
as illustrated in fig [ fig : closerpowers ] , let @xmath203 , and the lengths of the constant - power slots be @xmath204 and @xmath205 .
denote the rate pairs in the @xmath16 and @xmath18 slots as ( @xmath206 , @xmath207 ) and ( @xmath208 , @xmath209 ) , respectively .
we will show that keeping the total consumed energy constant , and transferring some amount of energy @xmath210 from the first slot to the second such that power levels are reallocated closer to each other , the sender can transmit at least the same number of bits within the same duration .
let us denote the average rate of the weaker user as @xmath211 .
provided @xmath212 , the sender could transfer some energy and some of user 2 s bits from the first epoch to the second while keeping user 1 s rates @xmath213 and @xmath214 constant . as energy and bits
are simply being deferred for later use , this operation does not violate feasibility .
specifically , let @xmath215 such that the new power allocation to the slots is @xmath216 satisfying @xmath217 . with this new allocation ,
the weaker user s rate in the first slot is @xmath218 .
its new new average rate over the duration of @xmath219 is : @xmath220 this is shown by straightforward application of the properties listed in section ii . in the remaining case which is @xmath221 , we know that @xmath222 must hold ( as @xmath199 . ) in this case , the allocation can similarly be improved by bringing power levels closer and transferring some of the first user s bits to the right , while keeping the rate allocation of the weaker user unchanged .
let @xmath223 be the average rate of the stronger user over the duration @xmath219 .
after the reallocation , the average rate of the stronger user becomes @xmath224 the @xmath16 and @xmath18 order derivatives of @xmath231 with respect to @xmath230 are the following and @xmath232 represent the first and second order partial derivatives of @xmath233 with respect to @xmath19 , respectively . ]
according to the properties listed in section ii , always holds if @xmath236 .
hence @xmath231 is concave in @xmath230 , if @xmath221 .
we conclude that a policy that contains a drop in power level is sub - optimal . to reach contradiction , suppose that power increases at time @xmath3 ( @xmath237 ) .
we will show that if none of the conditions , or hold , then it is possible to improve the schedule by transferring some energy from the @xmath113 epoch to the @xmath8 . assuming the @xmath8 epoch length is @xmath204 and the @xmath113 is @xmath238 , after bringing power levels closer , @xmath239 observe that if we treat @xmath44 as @xmath240 and @xmath241 as @xmath242 , then becomes identical with .
this implies that at least the same number of bits could be transmitted to weaker user , if we can bring power levels closer while keeping the stronger user s rates constant .
in addition to this , the allocation could also be improved by bringing power levels while keeping the weaker user s rates constant , in case @xmath243 .
consequently , equations and hold .
it is straightforward that we can not bring power levels any closer when condition holds , due to the energy causality constraint .
secondly , it also does nt yield a better schedule , if we can not transfer data from the latter epoch to the former(condition ) .
when it is possible to transfer some positive amount of energy from the @xmath113 epoch to the @xmath8 , as shown in , we can always improve allocation while keeping the rates of the stronger user the same .
although bringing power levels closer while keeping the rates of the stronger user the same is not feasible in case weaker user s data constraint is active , we may still improve allocation as proved in . nevertheless , this time we require @xmath243 . as rate of the weaker user can only rise upon a data arrival for the weaker user in case weaker user s data constraint
is active , condition describes the last case that we may not improve allocation by bringing power levels closer .
we have thus shown that this set of three conditions contains all the cases in which power can rise , if none of these hold , then power can not rise .
it is straightforward to show that this set can not be further reduced by finding counterexamples for the claim that if any one of , or is satisfied , then power may rise at time @xmath3 . 1 .
suppose that power increases upon a bit arrival for the weaker user occurring at @xmath3 .
as there is no energy constraint at @xmath3 , bringing power levels closer does not contradict with the energy causality in this case .
this implies that conditions or stated in lemma [ lmm : closerpowers ] must hold .
however , we know that there is a data arrival for the weaker user at @xmath3 , so if were true , then would be true as well .
therefore , condition holds in either case
. + ' '' '' 2 .
suppose that power increases upon a bit arrival for the first user .
with similar reasoning to part-1 , condition of lemma [ lmm : closerpowers ] can not hold . as there is also no data arrival for the weaker user , condition must be satisfied .
+ suppose that @xmath244 .
one can find a better schedule by bringing the rates of the stronger user closer by lemma 6 of @xcite .
therefore , the stronger user s rate can not decrease .
however , the stronger user s rate may increase because it may be against to either bit or energy causality to transfer some stronger user bits from epoch @xmath59 to @xmath47 .
firstly , it is against _ bit causality _ to transfer some stronger user bits from epoch @xmath59 to epoch @xmath47 , if the first condition holds .
secondly , if the second condition is satisfied we can not bring stronger users rates closer to each other as it would violate _ energy causality_. consider the local optimization problem given in fig .
[ fig : local_constraints ] , where @xmath245 is the data arrival for the @xmath8 user , @xmath246 is the energy harvest at the beginning of @xmath46 epoch for @xmath247 .
@xmath248 is the length of the first epoch and @xmath249 is the transmit duration in the second epoch .
let @xmath250 and @xmath251 be the power and rate assigned to the @xmath8 user during @xmath46 epoch after optimization .
the energy and data causality constraints for the local optimization problem are as follows : @xmath252 the structure of the solution changes if either one of the constraints satisfied with equality .
since there are @xmath253 different constraints , after optimization one will encounter one of the @xmath254 results . we have studied all @xmath255 cases and derived solutions to each one of them for both energy minimization and time minimization functions . in each case , the solution can be calculated analytically for energy minimization or it can be found iteratively for time minimization functions . before starting the optimization , if one already knows which constraints should be satisfied with equality , then the result could be obtained solving just that case .
otherwise , one should compute the results for each case and then select the best one that respects energy and bit causalities .
next , we will analyze one of the cases and present the algorithms for both energy and time minimization functions .
the analyzes of the remaining cases can be done in a similar fashion @xcite .
[ [ local - optimization - when - only - stronger - user - data - constraint - is - active ] ] local optimization when only stronger user data constraint is active + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + one of the possible structures of the local optimal solution is when only the stronger user data constraint is active . in this case ,
total transmit power level in local optimal solution should be constant ( see lemma [ lmm : closerpowers ] ) . under this condition , a possible illustration of optimal power allocation is depicted in fig . [
fig : localopt_b1 ] .
since the total power should stay constant during transmission , we should have @xmath256 .
moreover , we should have @xmath257 since only the stronger user bit causality is met .
for a given transmission completion time @xmath258 , the optimum schedule that minimize energy consumption can be found as follows : @xmath259 from and we derive @xmath260.\ ] ] then , total energy consumed in two epochs is calculated by @xmath261 as shown in ( [ b2_derivative2 ] ) , second derivative of @xmath266 is always negative for @xmath268 , which implies that @xmath266 is a strictly concave function of @xmath249 .
as @xmath249 goes to infinity , @xmath266 is as follows , since @xmath266 is a strictly concave function of @xmath249 and goes to a finite number as @xmath249 goes to infinity , @xmath266 is an increasing strictly concave function of @xmath249 and there is a unique @xmath266 for each value of @xmath249 . in time
minimization we have @xmath270 , @xmath271 , @xmath248 , @xmath263 , @xmath272 , @xmath273 ; @xmath274 and @xmath275 are constant terms and @xmath249 is transmission time within the second epoch which is to be minimized . since @xmath266 is an increasing concave function of @xmath249 , we iteratively find @xmath249 that sends exactly @xmath266 bits to weaker user by using bisection method .
minimum time to transmit @xmath262 and @xmath266 bits to the users in these two epochs can be calculated by @xmath276 .
algorithm [ alg : tmin_b1 ] presents a pseudo - code of time minimization algorithm for the case that only the stronger user bit causality event occurs .
+ 10 f. m. ozcelik , h. erkal and e. uysal - biyikoglu , `` optimal offline packet scheduling on an energy harvesting broadcast link , '' _ 2011 ieee int .
symposium on information theory _
, pp.2886 - 2890 , saint- petersburg , russia , aug . 2011 .
e. uysal - biyikoglu and a. el gamal , `` energy - efficient packet transmission over a multi - access channel , '' _ proc .
ieee intl .
symposium on information theory _ ,
p.153 , july 2002 .
r. a. berry and r. g. gallager , `` communication over fading channels with delay constraints , '' _ ieee transactions on information theory _ , vol.48 , pp.1135 - 1149 , may 2002 .
p. nuggehalli , v. srinivashan , and r. r. rao , `` delay constrained energy efficient transmission strategies for wireless devices , '' _ in proc.ieee infocom _ , vol.3 , pp.1765 - 1772 , new york , june 2002 . m. a. zafer and e. modiano , `` a calculus approach to energy - efficient data transmission with quality of service constraints , '' _ ieee / acm transactions on networking _ ,
vol.17 , pp.898 - 911 , june 2009 . j. yang and s. ulukus , `` optimal packet scheduling in an energy harvesting communication system , '' arxiv:1010.1295v1 , 2010 .
k. tutuncuoglu and a. yener , `` optimum transmission policies for battery limited energy harvesting nodes , '' arxiv:1010.6280 ] , 2010 .
o. ozel , k. tutuncuoglu , j. yang , s. ulukus and a. yener , `` transmission with energy harvesting nodes in fading wireless channels : optimal policies , '' _ selected areas in communications , ieee journal on _ , vol.29 , no.8 ,
pp.1732 - 1743 , september 2011 m. a. antepli , e. uysal - biyikoglu , and h. erkal , `` optimal packet scheduling on an energy harvesting broadcast link , '' _ selected areas in communications , ieee journal on _ , vol.29 , no.8 , pp.1721 - 1731 , september 2011 .
j. yang , o. ozel and s. ulukus , `` broadcasting with an energy harvesting rechargeable transmitter , '' arxiv:1010.2993v1 , 2010 .
o. ozel , j. yang and s. ulukus , `` broadcasting with a battery limited energy harvesting rechargeable transmitter , '' _ modeling and optimization in mobile , ad hoc and wireless networks ( wiopt ) , 2011 international symposium on _ , pp.205 - 212 ,
princeton , nj , may 2011 .
s. chen , p. sinha , n. b. shroff and c. joo , `` finite - horizon energy allocation and routing scheme in rechargeable sensor networks , '' _
ieee infocom11 _ , shanghai china , april 2011
. m. gatzianas , l. georgiadis and l. tassiulas , control of wireless networks with rechargeable batteries , " in _ trans . on wireless communications _ ,
vol . 9 , pp .
581 - 593 , february 2010 .
f. m. ozcelik , h. erkal and e. uysal - biyikoglu , `` optimal offline packet scheduling on an energy harvesting broadcast link , '' _ ieee international symposium on information theory ( isit ) _
, pp.2886 - 2890 , st .
petersburg , russia , july 2011 .
e. uysal - biyikoglu and a. el gamal , `` on adaptive transmission for energy efficiency in wireless data networks , '' _ ieee transsactions on information theory _ ,
vol.50 , pp.3081 - 3094 , dec 2004 .
e. uysal - biyikoglu , b. prabhakar , and a. el gamal , `` energy - efficient packet transmission over a wireless link , '' in _
ieee / acm trans . networking _ ,
vol.10 , pp.487 - 499 , aug
zangwill , _ nonlinear programming _ , prentice - hall , engelwood cliffs , new jersey , prentice - hall , 1969 .
h. erkal , f. m. ozcelik and e. uysal - biyikoglu , _ optimal offline broadcasting with an energy harvesting capable transmitter _
, technical report , metu , ankara , turkey , available at http://www.eee.metu.edu.tr/@xmath282cng .
|
we consider an energy harvesting transmitter broadcasting data to two receivers .
energy and data arrivals are assumed to occur at arbitrary but known instants .
the goal is to minimize the total transmission time of the packets arriving within a certain time window , using the energy that becomes available during this time .
an achievable rate region with structural properties satisfied by the two - user awgn bc capacity region is assumed .
structural properties of power and rate allocation in an optimal policy are established , as well as the uniqueness of the optimal policy under the condition that all the data of the weaker " user are available at the beginning .
an iterative algorithm , duopt , based on block coordinate descent that achieves the same structural properties as the optimal is described .
investigating the ways to have the optimal schedule of two consecutive epochs in terms of energy efficiency and minimum transmission duration , it has been shown that duopt achieves best performance under the same special condition of uniqueness .
packet scheduling , energy harvesting , awgn broadcast channel , energy - efficient scheduling .
|
the recent achievements of the vhe @xmath1-ray astronomy offer a great number of diverse and fundamental physics results which make further developments in the field highly physically motivated .
future @xmath1-ray observations will deliver exhaustive knowledge to understanding the origin of the cosmic rays by observing the supernovae remnants ( snrs ) and the physics of the relativistic jets in the active galactic nuclei ( agn ) , the measurements of the infrared ( ir ) absorption of @xmath1-rays propagating in the interstellar medium , the necessary data to prove the mechanisms of a particle acceleration in the pulsars and plerions , clues to the enigma of the gamma - ray bursts _
etc_. at the same time the significant advancement towards the new physics frontiers will be only possible after a significant improvement of the observational instrumentation .
the efficient search and detailed study of the sites of the particle acceleration up to multi - tev energies , which may brighten up in @xmath1-rays , need lowering farther the energy threshold of the instruments down to 50 - 100 gev , ability to localize the @xmath1-ray emission with accuracy better than 1 arcmin , large telescopes field of view for the comprehensive studies of the extended @xmath1-ray emission , large dynamic energy range ( 50 gev - 50 tev ) and good energy resolution ( @xmath2% ) for precise measurements of the @xmath1-ray energy spectra , larger @xmath1-ray statistics at the time scale less than 1 hr for efficient search for the episodic or highly variable sources _ etc_. the forthcoming _ stereoscopic _ arrays of the imaging air erenkov telescopes , like h.e.s.s .
( high energy stereoscopic system ) , will exactly meet all those requirements .
one can reach this conclusion using both the observational experience with the currently operating hegra ( high energy gamma ray astronomy ) instrument ( hereafter by hegra i denote the system of 5 iacts operated by the hegra collaboration ) as well as the specific monte carlo simulations .
all that allows to predict reliably the performance of the h.e.s.s . which is briefly summarized below with the emphasize on the further advancements in the physics analysis of the observational data .
although a good number of hardware parameters is different for h.e.s.s . as compared with hegra one can pick out several of those which are of the utmost importance .
they are _
( i ) _ the larger diameter of the reflector ; _ ( ii ) _ better optics making use of a stable dish and better mirrors alignment ; _ ( iii ) _ smaller angular size of the pixels .
.the hardware parameters of hegra and h.e.s.s . [ cols="<,<,<",options="header " , ] one can calculate the corresponding rate for h.e.s.s . which could be about 150 @xmath1-rays per min
! such enormous statistics will allow the measurements of the spectral shape for the crab like @xmath1-ray sources after less then 1 hr of observations . the maximum @xmath1-ray detection rate for h.e.s.s .
corresponds to the @xmath1-ray energy within @xmath350 - 100 gev ( see figure 1 ) . the uncertainty in the position of the peak is caused by the current uncertainties of the efficiencies along the over - all photon - to - photoelectron propagation chain .
one can break the entire dynamic energy range ( extending up to @xmath4 tev and @xmath5 tev for the observations at high and low elevations , respectively ) into two parts ( see figure 1 ) . as for the _ high performance region _ ( above 100 gev ) h.e.s.s .
will enable the high quality stereoscopic observations with the ability of high precision spectroscopy based on the images of a large size as well as high trigger multiplicity . at such energies the conventional data analysis developed for hegra could be of general use .
it is important to note that the effective collection area of h.e.s.s . is larger then hegra , by a factor of @xmath6 , above 1 tev .
it provides very high sensitivity of the h.e.s.s . in the multi - tev energy region . in the _ sub - threshold region _ ( see figure 1 )
the erenkov light images are substantially affected by the telescope s trigger , have rather small angular size and poorly defined shape due to the large fluctuations .
in addition for these images the night sky light contamination might be a severe problem .
nevertheless , the search for the periodic @xmath1-ray signals ( e.g. , from the pulsars ) in the energy range below 50 gev could well be physically expedient , eventhough it may need to develop a specific analysis algorithms in particular for these observations .
for a system of iacts the _ stereoscopic _ ( geometrical ) _ reconstruction _ of the air showers means the direct measurement of the orientation of the shower axis in space ; determination of the position of the shower core in the observation plane ; measurement of the angular dimensions of a shower , and height of its maximum ( the details of such reconstruction could be found elsewhere ) .
the major parameters of the h.e.s.s .
performance ( phase i ) , comprising 4 iacts in a system , are summarized in table 2 .
the important issue of the h.e.s.s .
performance is a sensitivity to the diffuse @xmath1-rays over the field of view ( fov ) ( see figure 2 ) .
the analysis of the monte carlo events for h.e.s.s . does nt show a significant decrease in the angular resolution for the off - axis events whereas the detection rate drops down rather fast beyond @xmath7 ( see figure 2 ) .
although a numerous attempts have been undertaken in past in order to improve the conventional stereoscopic algorithms , neither of those could gain much .
the hegra , with its ability of the individual event time sampling , did not reveal a dominant use of the time profiles of the erenkov light flashes from the air shower ( at least for the minimum time bins of 8 ns ) . on the contrary , applied after imaging analysis ,
the timing technique does not help at all .
however , one could hope for a 0.5@xmath81 ghz signal processing devices the advantage of the event time sampling might be of great importance ( e.g. , for the night sky noise reduction etc ) .
provided with a 50 hrs observational time looking at a point source , h.e.s.s .
could see the @xmath1-rays of the flux as low as @xmath9 at the 5@xmath10 confidence level with the @xmath1-ray statistics exceeding 10 @xmath1 s .
the minimum detectable flux for an extended @xmath1-ray source was estimated as @xmath11 no doubt such impressive sensitivity of future advanced ground based @xmath1-ray instrument allow to explore very effectively the sources of `` non - thermal emission in the universe '' .
|
at present , the ground - based astronomy of very high energy ( vhe ) ( e@xmath0100 gev ) @xmath1-rays experiences the challenging transition phase caused by the substantial upgrading of its observational instrumentation .
recently the hegra collaboration has evidently demonstrated the power of arrays of imaging atmospheric erenkov telescopes and initiated a number of complementary projects , such as h.e.s.s .
, cangaroo iv , veritas .
the common philosophy of those projects is based on the stereoscopic observations of air showers with a number of imaging telescopes simultaneously .
such observations allow to reduce substantially the effective energy threshold of detected @xmath1-ray showers ; to improve the angular and energy resolution for individual @xmath1-rays ; to gain in suppression of the cosmic ray background . in my talk
i have summarized the general advantages and achieved sensitivity of the stereoscopic observations using the hegra data taken with an array of 5 imaging telescopes having the energy threshold of 500 gev .
these results completed with the relevant simulations may be extrapolated on a firm ground towards the future h.e.s.s .
( phase i ) array of four 12 m imaging telescopes with the estimated energy threshold below 100 gev . here
i discuss in short the anticipated performance of such an array .
|
the main goal of the ultra - relativistic nucleus - nucleus ( a - a ) collisions programme is to create in terrestrial laboratories a new state of matter , the quark - gluon plasma ( qgp ) .
the existence of this phase , where quarks and gluons are deconfined , i.e. can freely move over several hadronic distances , is a definite prediction of quantum chromodynamics ( qcd ) . in a search for qgp signals
a - a collisions at different centre of mass energies per nucleon - nucleon ( nn ) pair have been studied : from few gev to several hundreds of gev recently attained in au - au collisions at rhic .
recently , accurate measurements of hadron production in central pb - pb collisions at 40 , 80 and 158@xmath0 gev of beam energy became available @xcite and also preliminary data at 30@xmath0 gev have been presented @xcite following an energy scan programme carried out by the experiment na49 at cern sps .
this programme is motivated by the hypothesis @xcite that the threshold for creation of qgp in the early stage of pb - pb collisions might be located in the low sps energy range , roughly between 20 and 40@xmath0 gev of beam energy .
one of the main results of the study of high energy a - a collisions is a surprising success of the statistical - thermal models in reproducing essential features of particle production @xcite .
this model succeeds also in describing particle multiplicities in many kinds of elementary collisions @xcite , suggesting that statistical production is a general property of the hadronization process itself @xcite .
furthermore , the statistical hadronization model ( shm ) supplemented with the hydrodynamical expansion of the matter , to a large extent also reproduces transverse momentum spectra of different particle species @xcite .
hence , the shm model proves to be a useful tool for the analysis of soft hadron production and particularly to study strangeness production , whose enhancement has since long been proposed as a signature of qgp formation .
furthermore , anomalies in the energy dependence of strangeness production have been predicted as a signature of deconfinement and have been indeed observed experimentally @xcite , suggesting that the onset of the phase transition could be located around 30@xmath0 gev .
it is thus important to make a systematic analysis , within the framework of shm , of the presently available hadronic multiplicities measured in pb - pb collisions at 30 , 40 and 80@xmath0 gev , which - to our knowledge - is done here for the first time .
along with these intriguing questions , our work is also motivated by issues related to the application of statistical model itself .
in fact , different versions of this model have been used in the past by different authors leading to somewhat different results and conclusions .
these mainly stem from the alternative use of midrapidity and full phase space multiplicities , from the allowance of non - equilibrium abundances of hadrons , from the assumption of exact local vanishing strangeness etc .
therefore , we consider the comparison of these different approaches a worthwhile step .
this has been made it possible by now by the availability of an accurate and large multiplicity sample in pb - pb collisions at 158@xmath0 gev as well as the corresponding data for pp interactions .
the paper is organized as follows : a brief description of our main version of the shm is given in sect . 2 .
the experimental data selected for the analysis are summarized in sect .
3 . in sect .
4 the results of the analysis using the main version and alternative schemes of the shm are given . finally , in sect .
5 we present and discuss the energy dependence of the chemical freeze - out stage .
summary and conclusions are drawn in sect .
the main idea of the shm is that hadrons are emitted from regions at statistical equilibrium , called clusters or fireballs .
no hypothesis is made about how statistical equilibrium is achieved ; this can be a direct consequence of the hadronization process . in a single collision event
, there might be several clusters with different collective momenta , different overall charges and volumes
. however , lorentz - invariant quantities like particle multiplicities are independent of clusters momenta , while they depend on charges and volumes .
if final state interactions among formed hadrons occur , particle multiplicities are frozen when inelastic interactions cease ( chemical freeze - out ) .
thus , by analyzing measured hadron abundances , a snapshot is taken of clusters at that particular stage of the evolution , which may significantly precede the final kinetic freeze - out stage , when also elastic interactions cease .
however , it should be pointed out that chemical and kinetic freeze - out may depend on the hadron species and the assumption of a single chemical freeze - out is certainly an approximation .
most calculations in shm are carried out in the framework of the ideal hadron - resonance gas , that is handling resonances as free particles : this amounts to take a considerable part of the hadronic interactions between strongly stable hadrons into account @xcite .
as has been mentioned , final multiplicities depend on the distribution of initial conserved charges ( baryon number , strangeness and electric charge ) among the produced clusters .
this distribution is determined by the dynamics of the collision and is thus needed as an external input to the statistical model .
however , most analyses , including ours , are carried out by assuming a single fireball .
this is possible provided that one of the two conditions below is fulfilled : 1 .
all clusters are large enough to allow a grand - canonical description and all of them have the same values of relevant intensive parameters , i.e. temperature and chemical potentials ; 2 .
clusters are small and must be treated canonically ( i.e. counting those states having exactly the same charges as the cluster itself ) , yet they have the same temperature and the distribution governing fluctuations of charges is the same as that obtained by splitting one large cluster - the _ equivalent global cluster _
egc - having as volume the sum of all clusters rest frame volumes and charges the sum of all clusters charges ( see appendix a ) . in this case
the overall particle multiplicities turn out to be those calculated in the canonical , perhaps grand - canonical , ensemble of the equivalent global cluster @xcite .
the reduction to egc could be achieved even for micro - canonical clusters with additional requirement on mass fluctuations @xcite .
the first condition sets stronger requirements and applies in the bjorken s boost - invariant scenario , where all clusters are to have the same parameters independently of their rapidity .
the second condition is altogether weaker and leaves room for the compatibility between the single fireball analysis and a variation of net baryon number density in rapidity .
this has been discussed in detail in ref .
the argument can be summarized as follows : particle multiplicities , being lorentz invariants , are unaffected by a shift in rapidity of the clusters ; therefore , clusters arising from the splitting of the egc can be ordered in rapidity according to their net baryon number without affecting fully integrated particle multiplicities and , at the same time , giving rise to an effective variation of the baryon density profile .
although the second condition is certainly more appropriate in the examined energy range , from ags to sps , it must be pointed out that this should not be expected to precisely match physical reality , as well as the first condition in its domain of applicability . in other words , discrepancies
( hopefully small ) between calculations based on this model and measurement are to be expected , so that these analyses shall not provide perfect fits even though the statistical model was the underlying true model .
in this paper we will stick to the picture outlined in the second condition , which implicitely requires the use of full phase space multiplicities in order to ( hopefully ) integrating out correlations between clusters momenta and charges . besides their general fitness , @xmath3 multiplicities also allow to safely enforce overall strangeness neutrality .
as has been mentioned , if the second condition applies , the multiplicity of any hadron @xmath4 can be calculated in the canonical ensemble of the egc .
hence , as the egc has a much larger volume than single clusters , the grand - canonical ensemble , where charges are conserved on average , can be a good approximation ( see appendix a ) .
this is the case for the collisions examined in this paper @xcite . in this case
the mean _ primary _ multiplicity of the @xmath5 hadron with mass @xmath6 and spin @xmath7 reads : @xmath8^{-1}\ ] ] where @xmath9 is the temperature , @xmath10 the egc volume , @xmath11 is a vector having as components the electric charge , baryon number and strangeness of the hadron and @xmath12 is a vector of the corresponding chemical potentials ; the upper sign applies to fermions , the lower to bosons . in order to correctly reproduce the data , it is also necessary to introduce at least one non - equilibrium parameter suppressing hadrons containing valence strange quarks , @xmath13 @xcite . with this supplementary parameter ,
hadron multiplicity is as in eq .
( [ mean ] ) with the replacement : @xmath14
\rightarrow \exp[\muv\cdot\qj / t ] \gamma_s^{n_s}\ ] ] where @xmath15 stands for the number of valence strange quarks _ and _ anti - quarks in the hadron @xmath4 .
the abundances of resonances is calculated convoluting ( [ mean ] ) with a relativistic breit - wigner distribution over a mass interval @xmath16 $ ] , where @xmath17 $ ] .
the minimum mass @xmath18 is needed to open all decay modes .
finally , the overall multiplicity to be compared with the data , is calculated as the sum of primary multiplicity ( [ mean ] ) and the contribution from the decay of heavier hadrons : @xmath19 where the branching ratios are taken from the latest issue of the review of particle physics @xcite and the summation runs over decays which contribute to the experimentaly measured multiplicity . among the hadrons and resonances contributing to the sum in eq .
( [ branching ] ) , in this work all known states quoted in ref .
@xcite up to a mass of 1.8 gev are included ( see discussion in sect . 4 ) .
what we have hitherto described is the main version of the shm used for the data analysis , that will be henceforth referred to as shm(@xmath1 ) . as has been mentioned in the introduction , in this work we also test other schemes and versions of the shm , which will be described in detail in sect .
the bulk of the experimental data consists of measurements made by na49 collaboration in central pb - pb collisions at beam momenta of 30 , 40 , 80 and 158@xmath0 gev , corresponding to @xmath20 7.6 , 8.8 , 12.3 and 17.2 gev respectively .
the acceptance region in rapidity and transverse momentum covers a typical range from midrapidity to projectile rapidity and from 0 to 1.5 gev/@xmath21 respectively .
the overall hadron multiplicities , quoted in referenced papers , were obtained using forward - backward symmetry in rapidity and by extrapolating the yields to full phase space .
all results were corrected for the feed - down from weak decays , e.g. @xmath22 multiplicity does not include pions produced in decays of @xmath23 hyperons and k@xmath24 mesons .
central collisions were selected by a trigger using information from a downstream calorimeter , which measured the energy of the projectile spectator nucleons .
whilst at 30 , 40 and 80@xmath0 gev all published results refer to the 7.2% most central collision sample , at 158@xmath0 gev different centrality selections ( 5% , 10% and 20% most central collisions ) were used to measure various hadronic species . in this analysis
we have rescaled all published multiplicities at 158@xmath0 gev to the corresponding ones at 5% most central collisions assuming that for the considered central collisions the hadron yield is proportional to the mean number of participant nucleona .
the resulting scaling factors are 1.08 and 1.32 for 10% and 20% most central collisions respectively @xcite .
as far as ags data at 11.6@xmath0 gev is concerned , we have used both multiplicities measured by the experiments and extrapolations of measured rapidity distributions made in ref .
@xcite at 3% top centrality .
for @xmath23 we have made a weighted average of the multiplicities measured both at 5% top centrality by e896 @xcite and e891 @xcite .
for the former , the quoted experimental error was only statistical so that we have added a 10% systematic error resulting in a value of [email protected]@xmath251.7 . for the latter , we have used the extrapolated value in ref .
@xcite of [email protected] .
the error on the weighted average has been rescaled by 1.25 ( i.e. @xmath26 ) according to the pdg weighting method in case of discrepancy between different measurements @xcite .
the obtained average has been rescaled by a factor 1.02 to convert it from 5% to 3% top centrality by assuming a linear dependence on the number of participants and by using the tables in ref . @xcite .
since the @xmath27 to @xmath23 ratio has been measured only at midrapidity @xcite , we have obtained a @xmath28 4@xmath29 multiplicity assuming that the double ratio @xmath30 is the same at sps and ags energies . the final experimental multiplicities and ratios used in our analysis
are shown in tables [ ags ] , [ pbpb30 ] , [ pbpb40 ] , [ pbpb80 ] and [ pbpb158 ] . in order to test the effect of the cut in rapidity on the resulting statistical model parameters ( discussed in detail in sect .
4 ) we have also determined the yields integrated over limited ( @xmath31 and @xmath32 ) rapidity windows around midrapidity in pb - pb collisions at 158@xmath0 gev . this has been done by fitting the rapidity distributions measured by na49 to a gaussian or the sum of two gaussians , with area and width as a free parameters and central values set to zero .
the results are shown in table [ yfit ] .
the quality of the fits is quite good , except for pions due to a couple of points near midrapidity ; yet , this discrepancy does not affect significantly the integrated yield .
in fact , it must be stressed that the main goal of these fits is to estimate an integral and not to reproduce accurately the shape of the distributions over the full measured range .
we have also checked that the extrapolations to full phase space are in good agreement with published measurements .
the analysis has been carried out by looking for the minima of the @xmath33 : @xmath34 where @xmath35 is the multiplicity of the @xmath36 measured hadronic species and @xmath37 is the sum in quadrature of statistical and systematic experimental error .
the theoretical multiplicities are calculated according to eq .
( [ branching ] ) with the decay chain stopped to match the experimental definition of multiplicity to properly compare theoretical and experimental values .
this occurs in pb - pb collisions after electromagnetic and strong decays and before weak decays , whilst in au - au collisions at ags the weak decays of @xmath23 , @xmath38 , @xmath39 , @xmath40 and k@xmath24 are included .
the effect of the uncertainties on masses , widths and branching ratios of the involved hadrons on the fit parameters has been studied by the method described in ref .
@xcite and found to be negligible throughout .
in order to cross - check our results and verify their robustness , we have performed the analysis with two independent numerical programs , henceforth referred to as a and b , which mainly differ with regard to the included resonances , their decay modes and branching ratios .
the fitted parameters within the main scheme shm(@xmath1 ) are shown in table [ parameters ] , while the experimental and fitted multiplicities , along with the predicted yields of several hadron species are shown in tables [ ags ] , [ pbpb30 ] , [ pbpb40 ] , [ pbpb80 ] and [ pbpb158 ] and figs .
[ agsf ] , [ pbpb40f ] , [ pbpb80f ] and [ pbpb158f ] .
we do not show any plot for the fit in pb - pb collisions at 30@xmath0 gev because of the low number of data points .
the quality of the fit is good throughout , as proved by the @xmath33 s values and we do not see any clear discrepancy between data and model , with the remarkable exception of the @xmath41 in pb - pb collisions at 158@xmath0 gev . due the 5@xmath42 deviation from the statistical model prediction ,
the measured @xmath41 yield has been removed from the fitted data sample as it could have biased the fit itself .
we argue that this disagreement owes to its short lifetime ( @xmath43 mev ) compared with all other used particles .
if the kinetic freeze - out occurs after some suitable delay from the chemical freeze - out , one can indeed justify the low measured @xmath41 yield as the effect of the elastic reinteractions of its decay products .
the quality of the fits is further illustrated in fig .
[ sqrts ] where the measured and fitted ratios and are plotted as a function of @xmath44 ; these ratios have been chosen as it has been proposed that their energy depedence plays an important role in the search for deconfinement onset at sps energies @xcite .
the observed differences in the fit parameters between a and b are of the order of the fit errors . they may be considered as an estimate of the systematic error due to uncertainties in the implementation of the model .
the first set of parameters in table [ parameters ] have been obtained by using the full data , whilst the second set has been obtained by using the maximal common set of particles measured in the four collisions , that is @xmath45 , k@xmath46 , k@xmath47 , @xmath23 , @xmath28 and the participant nucleons ( net baryon number ) @xmath48 . by comparing fit results in the same analysis ( a with a and b with b )
, it can be seen that the obtained parameters are in good agreement and only in one case a discrepancy larger than one standard deviation ( @xmath1 in pb - pb at 158@xmath0 gev ) is observed ; this demonstrates the robustness of the results .
we have also included in tables [ ags ] , [ pbpb30 ] , [ pbpb40 ] , [ pbpb80 ] and [ pbpb158 ] the prediction for the yield of the recently discovered @xmath49 pentaquark baryon ( uudd@xmath50 ) by using as input mass @xmath51 mev and @xmath52 . according to the shm(@xmath1 ) model , in the boltzmann limit
this simply reads : @xmath53\ ] ] if we disregard feeding from possible excited states .
as the number of data points in pb - pb collisions at 30@xmath0 gev was not sufficient to determine the four free parameters unambiguously , we have forced @xmath9 to lie on the parabolic chemical freeze - out curve ( [ tmuparam ] ) in sect . 5 , interpolating the other four points in the @xmath54 plane .
this method has proved to be able to provide unambiguous solutions for the remaining three free parameters .
the first quoted error beside the best - fit value in table [ parameters ] is the error coming out from the fitting program ( inferred from the analysis of the @xmath55 level contours ) whereas the second error is the fit error rescaled by a factor @xmath56 where @xmath57 is the number of degrees of freedom .
we deem that the latter is a more realistic uncertainty on the parameters with respect to the fit error because of the `` imperfect '' @xmath58 values , expected to be 1 on average if the model correctly matched physical reality . the argument , which is the same used in the particle data book @xcite when averaging discrepant data , is as follows :
if @xmath59 , then the model can not reproduce the data at the level of accuracy relevant to the experimental errors ; on the other hand , this would be the case if experimental errors were larger and , particularly , if they were rescaled by a common factor @xmath60 so that : @xmath61 with this simple rescaling of the @xmath33 , the best - fit parameters would be unchanged , whereas their relevant errors would scale up by a factor @xmath60 .
in fact , the new covariance matrix @xmath62 for the parameter vector @xmath63 is related to the @xmath64 around the minimum through : @xmath65 also : @xmath66 and , being @xmath67 , one finds @xmath68 .
therefore , since this covariance matrix would be the outcome of a standard quality fit , with @xmath69 , the errors rescaled by @xmath60 can be regarded as the sensible minimal uncertainty on the parameters .
these rescaled errors have indeed been used in all of the plots in this paper .
a major issue in the multiplicity fits is where to stop the inclusion of heavy light - flavoured resonances contributing to measured particle yields in eq .
( [ branching ] ) .
the relevance of this cut - off is owing to the peculiar shape of the hadron mass spectrum , which rises almost exponentially between 1 and 1.7 gev and drops thereafter probably due to the missing experimental information ( see fig . [ spectr ] ) .
should the number of states keep on increasing exponentially , the problem is set of the physical meaning of the obtained parameters , which could be heavily affected by the ignorance of further hadronic states .
in fact , although the production of resonances decreases exponentially with the mass , the effect on secondary light particles through the decay chain could be balanced and even exceeded by the increasing number of states .
we have thus checked the stability of the obtained parameters in the four collisions by varying the cut - off on the mass spectrum in a range where we are reasonably confident on the complete experimental knowledge and the number of states apparently rises exponentially ( i.e. up to 1.7 - 1.8 gev ) and repeating the fit . as shown in fig .
[ stabil ] , the fitted @xmath9 , @xmath70 and @xmath1 in pb - pb at 158@xmath0 gev are indeed fairly constant from 1.3 to 1.9 gev .
furthermore , the outcoming primary yields of some measured particles tend to saturate at cut - off masses of about 1.8 gev , implying that the contribution of resonance decays to secondaries ( needed to keep the final multiplicity close to the measured value ) settles down as well and the inclusion of heavier states yields a more and more negligible contribution .
this is a clear indication of the significance of the fit results .
a similar pattern occurs in all other examined collisions .
a major result of these fits is that @xmath1 is significantly smaller than 1 in almost all cases ( with a possible exception at 30@xmath0 gev , see table [ parameters ] ) , that is strangeness seems to be under - saturated with respect to a completely chemically equilibrated hadron gas .
this confirms previous findings @xcite .
there is a considerable interest and ongoing investigations about this deviation of the data from the fully equilibrated hadron gas , particularly motivated by the fact that strangeness production is considered as a possible qgp signature .
it is therefore worth to examine and test , with the presently available large data sample , different scenarios which have been put forward to account for the under - saturation of strangeness .
as we have seen , fits to full phase space multiplicities within the shm yield @xmath71 in most cases .
however , good tests of the same model without extra strangeness suppression ( i.e. assuming @xmath72 ) have been obtained by fitting ratios of hadronic yields within a limited rapidity range around midrapidity at top sps energy @xcite .
this is an appropriate method of estimating the parameters of the sources only if the boost - invariant bjorken scenario holds , at least as a good approximation , over a large rapidity interval ( say @xmath73 ) because , in this case , particle ratios at midrapidity are the same as in full phase space
. however , rapidity distributions of hadrons at sps energies do not feature boost - invariance @xcite and a cut at midrapidity can artificially enhance heavy particle yield with respect to light ones ( see fig .
[ rapid ] ) , as long as their kinetic freeze - out occurs at the same temperature and the leading baryon effect can be neglected . in the statistical model of a single fireball
this can be easily understood , for the width of the rapidity distribution decreases as a function of mass according to ( in the boltzmann approximation ) : @xmath74\ ] ] yet , it is worth testing the effect of the rapidity cut on measured distributions rather than using arguments based solely on the statistical model .
therefore , we have fitted , within the scheme a , the integrated yields over limited rapidity windows measured in central pb+pb collisions at 158@xmath0 gev by na49 ( see sect . 3 for details ) and
quoted in table [ deltay2 ] as well as the yields measured by wa97 @xcite over a @xmath31 window around midrapidity .
we first note that , according to table [ deltay2 ] , the integrated yields over @xmath31 measured by na49 and wa97 are in good agreement with each other . since the fit to the statistical model gave @xmath75 @xcite for wa97 data , the same is expected for the integrated na49 yields over the same rapidity window .
this is indeed what we find , as shown in table [ deltay ] .
while temperature and baryon - chemical potential are essentially unchanged , the best - fit value of @xmath1 is closer to 1 than that obtained in full phase space in fit a ( see table ( [ parameters ] ) and it is also compatible with 1 within the error . we then conclude that the superfluity of @xmath1 in analysis of midrapidity particle yields , at least at top sps energy , is likely to owe to the artificial enhancement of strange particles with respect to lighter non - strange ones , induced by the cut on rapidity . the fact that @xmath75 for midrapidity yields is then not an indication of a fully equilibrated hadron gas at midrapidity ; even if such equilibrated fireball existed at the estimated kinetic freeze - out temperature of @xmath76 mev @xcite , the @xmath31 window would be too narrow for a correct estimation of chemical freeze - out parameters ( see fig . [ rapid ] ) because lighter particles would be cut down significantly . to account for the observed under - saturation of strangeness
, a picture has been put forward in which strangeness is supposed to be exactly vanishing over distances less than those implied by the overall volume @xmath10 @xcite .
we henceforth refer to this version of the statistical model as shm(scv ) . following the description of the model in sect . 2
, this means that the produced clusters or fireballs emerge with @xmath77 and they are not allowed to share non - vanishing net strangeness .
assuming , for sake of simplicity , that all clusters have the same typical volume @xmath78 and that the equivalence of the set of clusters to a global fireball still applies for baryon number and electric charge ( but not to strangeness ) the following expression of the primary average multiplicities can be obtained ( see appendix b ) : @xmath79 \frac{z_c(-ns_j)}{z_c(0)}\ ] ] where @xmath80\ ] ] is the so - called _ strange canonical partition function _ of a single cluster . as usual , in the above equations , the upper sign is for fermions and the lower for bosons . if @xmath78 is sufficiently small , the multiplicities of strange hadrons turn out to be significantly suppressed with respect to the corresponding grand - canonical ones due to the enforcement of exact strangeness conservation in a finite system , an effect called _
canonical suppression_. furthermore , the suppression features hierarchy in strangeness , namely it is stronger for @xmath40 ( @xmath81 ) and @xmath39 ( @xmath82 ) than for kaons and @xmath23 s , so it can be argued that this can account for the actually observed hierarchical pattern of extra strangeness suppression which goes like @xmath83 for open strange particles . the discriminating difference between this picture and our main scheme shm(@xmath1 ) described in sect . 2
is concerned with hidden strange particles such as @xmath84 , which do not suffer canonical suppression , so that its theoretical multiplicity in shm(scv ) turns out to be simply the same as in a cluster with volume @xmath10 , that is given by the formula ( [ mean ] ) without @xmath85 suppression .
we have made a test of this model by fitting the data sample of full phase space multiplicities in pb - pb collisions at 158@xmath0 gev fixing @xmath86 and determining the parameters @xmath9 , @xmath10 , @xmath70 and @xmath87 within the scheme a. the results are shown in tables [ models ] and [ pbpb158 ] . the quality of the fit is worse with respect to the shm(@xmath1 ) model mainly because of the underestimated pion yield and the larger of @xmath84 .
the latter is expected , as has been mentioned . as far as pion discrepancy
is concerned , the deviation stems from the very fact that they are the only non - strange particles in the fit .
the minimization procedure tries to accommodate the relative ratios among strange hadrons by fixing @xmath78 and @xmath9 , then it tries to set the overall normalization @xmath10 and at this stage a competition sets in between the set of strange and non - strange particles . since pions are only two entries , the fit prefers to keep them low rather than raising the whole set of strange particles .
our result suggests that , for the local strangeness correlation to be an effective mechanism , the cluster volume should be of the order of 2.5% of the overall volume .
otherwise stated , strange quarks should have not propagated beyond a distance of about 4 fm from the production point up to chemical freeze - out , if we take the overall volume of about 3 10@xmath88 fm@xmath88 as coming out from this fit where hadrons are pointlike particles . in this picture , henceforth referred to as shm(tc ) , the observed hadron production is approximately the superposition of two components ( tc ) : one originated from one large fireball at complete chemical equilibrium at freeze - out , with @xmath86 , and another component from single nucleon - nucleon collisions .
in fact , according to simulations based on transport models , a significant fraction of beam nucleons interacts only once with target nucleons @xcite . with the simplifying assumption of disregarding subsequent inelastic collisions of particles produced in those primary nn collisions
, the overall hadron multiplicity can be written then as : @xmath90 where @xmath91 is the average multiplicity of the @xmath5 hadron in a single nn collision , @xmath92 is the mean number of single nn collisions giving rise to non - re - interacting particles and @xmath93 is the average multiplicity of hadrons emitted from the equilibrated fireball , as in eq .
( [ mean ] ) , with @xmath86 .
the @xmath91 term can be written in turn as : @xmath94 since it is known that in nn collisions strangeness is strongly suppressed @xcite the idea is to ascribe the observed under - saturation of strangeness in heavy ion collisions to the nn component , leaving the central fireball at complete equilibrium , i.e. with @xmath86 .
of course , this is possible provided that @xmath95 is sufficiently large .
this production mechanism has probably some consequences on the final rapidity and momentum distributions of the different species , whose calculation goes certainly beyond the scope of this paper .
instead , we have confined ourselves to integrated multiplicities and tried to fit @xmath9 , @xmath10 , @xmath70 of the central fireball and @xmath92 by using na49 data in pb - pb collisions at 158@xmath0 gev within the scheme a. to calculate @xmath91 we have used the statistical model and fitted pp full phase space multiplicities measured at @xmath96 gev ( i.e. the same beam energy ) by the same na49 experiment . for np and nn collisions , the parameters of the statistical model determined in pp are retained and the initial quantum numbers are changed accordingly .
theoretical multiplicities have been calculated in the canonical ensemble , which is described in detail in ref .
@xcite . instead of the usual @xmath1 parametrization ,
the new parametrization described in ref .
@xcite has been used in which one assumes that some number of pairs , poissonianly distributed , hadronizes ; the extra strangeness suppression parameter @xmath1 is thus replaced by the mean number of these pairs , @xmath97 .
+ the results of the fit are shown in table [ pp160 ] along with fitted and predicted hadron multiplicities , including the @xmath98 pentaquark baryon , and in fig .
[ pp160f ] .
the temperature value is significantly higher than in pp and collisions at higher energy , an effect already observed for center - of - mass energies below 20 gev @xcite .
we conjecture that this is a possible indication of a beginning inadequacy of the canonical ensemble at low energy , where exact conservation of energy and momentum should start to play a significant role .
perhaps this is the point where the microcanonical hadronization of each cluster is a more appropriate approach .
the results of the fit to pb - pb collisions are shown in table [ models ] . the fit quality , as well as the obtained values of @xmath9 , @xmath70 , are comparable to the main fit within the shm(@xmath1 ) model .
the predicted number of `` single '' nn collisions is about 50 with a 16% uncertainty .
thus , only 260 nucleons out of 360 contribute to the formation of large equilibrated fireballs .
the percentage of primary hadrons stemming from nn collisions varies from 14% for pions to 27% of @xmath99 s and protons and to 0.5% of @xmath40 s .
it should be pointed out that the fitted parameters are affected by a further systematic error owing to the uncertainty on the parameters of the statistical model in nn collisions , which are used as an input in the pb - pb fit . however , because of exceeding computing time needs , it has not been possible to assess these errors . in a simple geometrical picture ,
the single - interacting nucleons are located in the outer corona of the portion of colliding nuclei corresponding to the observed number of participants .
as the projected ( on the collision s transverse plane ) radial nucleon density is : @xmath100 where @xmath101 @xmath102 is the nucleon density and @xmath103 fm is the radius of the portion of colliding nucleus corresponding to a participant number of 180 , the 50 single - interacting nucleons should lie between 4.84 and 6.45 fm .
this simple estimate is in approximate agreement with the calculations performed with the glauber model @xcite .
since the number of single - interacting nucleons is expected to be weakly dependent on center - of - mass energy , the fits to this two - component model should yield consistent values of @xmath92 at 30 , 40 and 80@xmath0 gev collisions .
however , no measurement of hadron production in nn collisions at those energies is available and this question can not be tackled for the present .
an extension of the statistical model has been proposed where qgp hadronization is essentially a statistical coalescence of quarks occurring at an energy density value which does not correspond to a hadron gas at equilibrium @xcite . in this model
two non - equilibrium parameters are introduced for the different types of quarks , @xmath104 for u , d quarks and @xmath105 for strange quarks ( the difference between @xmath105 and @xmath1 is explained below ) .
the multiplicity of each hadron thus reads : @xmath106\ ] ] where @xmath15 is the number of valence s quarks and @xmath107 the number of valence u , d quarks ; @xmath108 and @xmath109 are as in eq .
( [ mean ] ) . by defining :
@xmath110 the boltzmann limit of average multiplicity reads : @xmath111\ ] ] where @xmath112 is the baryon number , as long as mesons have two and baryons have three valence quarks . by comparing this formula with the boltzmann limit of eq .
( [ mean ] ) it can be realized that the introduction of a light - quark non - equilibrium parameter amounts to introduce in the statistical model an overall enhancement ( or suppression ) of baryons with respect to mesons , unlike in the model shm(@xmath1 ) .
we henceforth refer to this model as shm(@xmath113 ) .
the parameter @xmath104 has a definite physical bound for bosons which can be obtained by requiring the convergence of the series @xmath114 for any value of the energy . if the number of u , d quarks to be hadronized is so large that @xmath104 is to attain its bounding value , a bose condensation of particles in the lowest momentum state sets in . for low strangeness and electrical chemical potentials , such as those found in the present analysis ,
the bounding value is @xmath115 where @xmath116 is the neutral pion mass , e.g. @xmath117 for @xmath118 mev . with the introduction of @xmath104 as an additional free parameter , there are 5 parameters to be determined in the model .
this makes the minimization procedure rather unstable because it becomes easier to be trapped in local minima . to avoid this
, we have performed 4 parameter fits with fixed values of @xmath104 varying from 0.6 to 1.7 in steps of 0.1 .
this method allows a clearcut determination of the absolute minimum .
the results of these fits are shown in fig .
[ gammaq ] in terms of the minimum @xmath33 obtained for fixed @xmath104 .
the round dots show the minimum @xmath33 s achieved by using the main sample of multiplicities in pb - pb collisions at 158@xmath0 gev . we find a steady decreasing trend in the value of best - fit temperatures varying from @xmath119 mev at @xmath120 to @xmath121 mev at @xmath122 .
the fitted temperature at @xmath123 reaches the upper limit of 200 mev , which is the maximum allowed in the model to prevent from being critically dependent on the hadron mass spectrum cut - off . on the other hand , the best - fit values of @xmath124 and @xmath1 are rather stable and about the same found in the main fit with @xmath125 .
the number of terms in the series ( [ meangq ] ) has been truncated to 5 for all particles ; the contribution of further terms has been found to be negligible throughout .
it is seen that the absolute @xmath33 minimum falls in the region of pion condensation , marked by a vertical line at @xmath126 , with @xmath127 and @xmath128 mev .
this finding is in agreement with what is found in ref .
@xcite . however , there is also a local minimum at the lower edge @xmath120 , with a temperature of 187 mev , which is only one unit of @xmath33 higher than the absolute minimum .
this indicates that the absolute minimum could be rather unstable against variations of the input data and this is in fact what we find by varying down the pion multiplicities by only 1 @xmath42 . for this case , the minimum @xmath33s
are shown in fig .
[ gammaq ] as triangular dots and the absolute minimum now lies at @xmath120 instead of at the pion condensation point . in view of the instability of the fit , and of the small _
relative _ @xmath33 improvement in comparison with the main fit , we conclude that there is so far no evidence for the need of this further non - equilibrium parameter . however , it is interesting to note that this model predicts an enhanced ( if @xmath129 ) or suppressed ( if @xmath130 ) production of the @xmath49 pentaquark baryon with respect to the other versions of the statistical model , even though they agree in reproducing the multiplicities of all other hadrons .
this is owing to an additional @xmath131 factor for this special hadron having five valence quarks . from eqs .
( [ meangq],[transform ] ) one gets , in the boltzmann limit : @xmath132\ ] ] in fact , as can be seen in table [ pbpb158 ] the predicted yield of @xmath49 at the global minimum @xmath126 is more than a factor 2 higher than for shm(@xmath1 ) . as a final remark
, we stress that minimum @xmath33 fits are very useful tools to get information on the state of the source at chemical freeze - out , but , as already emphasized in sect . 3 ,
the simple multiplicity analysis with global parameters resides on an idealization of the collision ( e.g. the assumed existence of an egc ) which can not _ exactly _ fit physical reality and discrepancies are to be expected anyway .
thus , a new mechanism or of a modification of the basic scheme proves to be relevant only if it leads to a major improvement of the agreement with the data .
slight improvements of the @xmath33 , whenever their significance is beyond its expected statistical fluctuations , can not be seriously taken as a proof of the validity of a particular scheme .
the statistical model does not make any prediction on the energy dependence of hadron production ; its relevant parameters have to be determined separately for each energy and reaction type .
nevertheless , the analysis of the data within this model may help in the study of energy dependence of hadron production because it effectively reduces the full experimental information on numerous hadron yields to only few parameters .
clearly this reduction should be taken with care , were not for the approximate validity of some relevant assumptions , such as the reducibility to egc ( see sect . 2 ) .
furthermore , the reduction procedure may remove or dilute essential physical information . with these caveats in mind , in this section we discuss the energy dependence of the chemical freeze - out parameters extracted from the data ( by using the full data sets ) within our main shm(@xmath1 ) approach .
the chemical freeze - out points in the @xmath54 plane are shown in fig .
the rhic point at @xmath133 gev , obtained fitting particle yield ratios at midrapidity , has been taken from ref .
the four points at beam energies of 11.6 , 40 , 80 and 158@xmath0 gev have been fitted with a parabola : @xmath134 where @xmath9 and @xmath70 are in gev .
the pb - pb point at 30@xmath0 gev has been forced to lie on the above curve , as has been mentioned in the previous section .
the rhic point calculated in ref .
@xcite is in good agreement with the extrapolation of the curve ( [ tmuparam ] ) . in the search for deconfinement phase transition ,
strangeness production is generally believed to be a major item , especially if an anomalous abrupt change was found as a function of centre - of - mass energy or other related quantities .
a possible indication of it in pb - pb collisions at the low sps energies was reported on the basis of the observed energy dependence of several observables @xcite .
particularly , the ratio shows a peaked maximum at about 30@xmath0 gev .
one may expect that this anomaly should be reflected in the energy dependence of @xmath1 parameter fitted within shm(@xmath1 ) scheme .
this dependence is plotted in fig .
[ gs ] and in fact a maximum shows up at 30@xmath0 gev .
although the error bars are large enough so as to make @xmath1 seemingly consistent with a constant as a function of centre - of - mass energy , it is important to note that the dominant systematic errors on experimental data at sps energies are essentially common . therefore , the errors on the model parameters at different sps energies turn out to be strongly correlated , hence fitted @xmath1 s are expected to move up or down together . in order to further study strangeness production features ,
we have also compared the the measured ratio ( including the preliminary rhic result at @xmath135 gev @xcite ) with the theoretical values in a hadron gas along the freeze - out curve ( [ tmuparam ] ) as a function of the fitted baryon - chemical potential for different values of @xmath1 ( see fig .
[ kpi ] ) . the calculated dependence of on @xmath70 is non - monotonic with a broad maximum at @xmath136 mev ( i.e. @xmath137 gev ) @xcite . taking into account that systematic errors at different energies in pb - pb collisions are fully correlated , we can conclude that the data points seem not to follow the constant @xmath1 lines .
in fact , the anomalous increase of relative strangeness production at 30@xmath0 gev can be seen also in the wroblewski variable @xmath138 , the estimated ratio of newly produced strange quarks to u , d quarks at primary hadron level , shown in fig .
[ ls ] and table [ parameters ] .
the calculation of newly produced quark pairs is performed by using the statistical model best fit values of the various hadron multiplicities , so the obtained @xmath139 values are somehow model - dependent .
nevertheless , this variable features a very similar behaviour as the ratio and attains a maximum value of 0.61 at 30@xmath0 gev , very close to that predicted for @xmath86 .
these deviations from a smooth behaviour of strangeness production are certainly intriguing , yet the analysis within the shm will be more conclusive in this regard with a larger data sample at 30@xmath0 gev and at the forthcoming 20@xmath0 gev data .
we presented a detailed study of chemical freeze - out conditions in ultra - relativistic heavy ion collisions at projectile momenta of 11.6 ( au - au at ags ) , 30 , 40 , 80 and 158@xmath0 ( pb - pb at sps ) gev , corresponding to nucleon - nucleon centre of mass energies of 4.8 , 7.6 , 8.8 , 12.3 , 17.2 gev respectively . by analyzing hadronic multiplicities measured in full phase space within the statistical hadronization model ,
we have tested and compared different versions of this model , with special emphasis on possible explanations of the observed strangeness under - saturation at the hadron level .
it is found that version of the model referred to as shm(@xmath1 ) , where a non - equilibrium population of hadron carrying strange valence quarks is allowed , fits all the data analyzed in this paper .
we have also shown that the seeming full chemical equilibrium found for central pb+pb collisions at 158@xmath0 gev by using particle yields integrated over a limited region around mid - rapidity is most likely an artefact of the kinematical cut .
we have tested a model ( shm(tc ) ) in which hadron production is pictured as stemming from two independent components : a fireball ( or a set of fireballs ) at full chemical equilibrium and single nucleon - nucleon interactions .
this model can fit the data at 158@xmath0 gev if the number of collisions is around 50 with a sizeable uncertainty .
so far , it can not be confirmed at other energies due to the lack of the precise data on nn interactions . a model in which strangeness
is assumed to vanish locally @xcite yields a worse fit to the data with respect to shm(@xmath1 ) and shm(tc ) .
finally , we have also tested a model in which it is allowed a non - equilibrium population of hadrons carrying both strange and light valence quarks .
we have found the present set of available data does not allow to establish whether a further non - equilibrium parameter is indeed needed to account for the observed hadron production pattern .
a discriminating prediction of this model with respect to the shm(@xmath1 ) is an enhanced production of the recently discovered @xmath49 pentaquark baryon due to the additional factor @xmath140 .
energy dependence of chemical freeze - out parameters has been discussed based on the results obtained with our main version of the model shm(@xmath1 ) .
the evolution of the freeze - out temperature and baryon - chemical potential is found to be smooth in the ags - sps - rhic energy range .
the strangeness - suppression parameter @xmath1 is found to smaller than one ( @xmath141 ) for most of the studied collisions which confirms previous findings @xcite , with an indication of a maximum at 30@xmath0 gev , where @xmath1 is found to be close to one .
the significance of this maximum is related to the correlation between errors on hadron yield measurements at different energies .
the interpolated dependence of relative strangeness production on energy and baryon - chemical potential , as measured by the ratio and the wroblewski factor @xmath139 , features a broad maximum at about 30@xmath0 of beam energy .
however , the experimental measurement of and the estimated @xmath139 value in pb - pb collisions at this energy seemingly exceed the expected values for a fixed @xmath1 .
the canonical partition function of the @xmath36 cluster can be written as a multiple integral over the interval @xmath142 $ ] @xcite : @xmath143\ ] ] where @xmath144 is a vector having as components the electric charge , baryon number and strangeness of the cluster , @xmath145 and @xmath146 reads : @xmath147 @xmath148 is the volume and @xmath9 the temperature of the cluster ; the sum runs over all hadronic species @xmath4 and @xmath109 the charge vector of the @xmath5 hadron ; the upper sign applies to fermions , the lower to bosons .
the probability distribution required for the reduction to egc to apply reads @xcite : @xmath149 in this case , the overall multiplicity of the @xmath5 hadron is given by @xcite : @xmath150 where @xmath151 is the canonical partition function of the equivalent global cluster : @xmath152 and @xmath153 is a fictitious fugacity parameter .
formally , this turns out to be the same function as in eq .
( [ zcan ] ) with @xmath154 replacing @xmath148 and @xmath155 replacing @xmath156 .
if @xmath10 is large , the canonical partition function can be approximated by the leading term of an asymptotic saddle point expansion .
setting @xmath157 , the canonical partition function can be written as : @xmath158 \exp[f_c({\bf z } ) ] \prod_{k=1}^3 z_k^{-q_k}\ ] ] where : @xmath159 the saddle - point expansion is carried out by requiring the logarithmic derivative of the integrand to vanish : @xmath160 the solutions of this equation are indeed the grand - canonical fugacities @xmath161 .
the function @xmath162 coincides with the logarithm of the grand - canonical partition function @xmath163 , therefore the equation ( [ saddle ] ) expresses the equality between the average charge in the grand - canonical ensemble @xmath164 and the initial value @xmath165 .
the canonical partition function now becomes , at the second order of the expansion : @xmath166 \big [ \prod_{k=1}^3 \lambda_k^{-q_k } \big ] \frac{1}{(2\pi\i)^3 } \big [ \prod_{k=1}^3 \oint \frac{\d z_k}{z_k } \big ] \exp [ -({\bf z}-\lambdav)^t { \sf h } \ , ( { \bf z}-\lambdav)/2]\ ] ] where @xmath167 is the hessian matrix in @xmath168 .
the first exponential factor is just the grand - canonical partition function @xmath169 calculated for the fugacities @xmath170 .
the average multiplicity of the @xmath5 hadron species can now be calculated by using eq .
( [ canmean ] ) taking the approximated expression ( [ cantogc ] ) of the canonical partition function . retaining only the dominant contribution , one just obtains the grand - canonical expression of the average multiplicity as expressed in eq .
( [ mean ] ) .
the argument closely follows the previous one . the main difference is the request of vanishing strangeness for each cluster .
thus , the configurational probabilities ( [ weight ] ) @xmath171 turn to : @xmath172 and the average multiplicity of the @xmath5 hadron species now reads : @xmath173 where @xmath153 is a fictitious fugacity .
let us now work out the expression : @xmath174 by assuming that all the clusters have the same volume @xmath78 and temperature @xmath9 .
this can be done rewriting the canonical partition functions @xmath175 like in eq .
( [ zcan ] ) , using the integral representation of the kronecker s delta and eliminating the redundant strangeness conservation constraints .
thus , expanding the vector @xmath176 in its components , the eq .
( [ global ] ) becomes : @xmath177\end{aligned}\ ] ] where @xmath178 is the function in eq .
( [ func ] ) with @xmath78 replacing @xmath148 .
note that this function is the same for all clusters , being @xmath9 and @xmath78 constant .
we can now carry out the sum over all the integers @xmath179 in eq .
( [ global2 ] ) and get : @xmath180 so that the integration over @xmath181 and @xmath182 in eq .
( [ global2 ] ) can be easily done and one is left with : @xmath183\ ] ] as the function @xmath178 is the same for all clusters , this can be written also as : @xmath184 \right\}^{v / v_c}\ ] ] being @xmath185 .
for large volumes , we can approximate @xmath186 by means of the saddle - point expansion of the integrals over @xmath187 and @xmath188 like in appendix a. thus , similarly to eq .
( [ cantogc ] ) : @xmath189 \right\}^{v / v_c}\ ] ] the function : @xmath190\ ] ] is defined as the _ strange canonical partition function_. the multiplicity of the hadron @xmath4 can now be calculated by means of eq .
( [ means ] ) by using eqs .
( [ global]),([global4 ] ) and the definition ( [ zscan ] ) .
what is obtained is just eq .
( [ scanonical ] ) we would like to thank jens ivar jrdre for his help with ratio at rhic .
this work is part of infn research project fi31 .
s. v. afanasiev _ et al . _ , na49 coll .
c * 66 * ( 2002 ) 054902 .
.comparison between measured and fitted particle multiplicities , in the framework of shm(@xmath1 ) model in central au - au collisions ( 3% ) at a beam energy of 11.6@xmath0 gev .
also shown the predicted multiplicities of the main hadron species . [ cols="^,^,^,^,^",options="header " , ] parameter in pb - pb collisions at a beam energy of 158@xmath0 gev in the fit a ; also quoted the best - fit parameters .
below : residual distribution .
note that the @xmath41 was not used in the fit ( see text ) .
[ pbpb158f ] ] of multiplicity fits in pb - pb collisions at a beam energy of 158@xmath0 gev as a function of a fixed light quark non equilibrium parameter @xmath104 .
the round dots are the @xmath33 s obtained with the main data sample , whilst triangular dots are those obtained with pion multiplicities lowered by one standard deviation and all others unchanged .
the vertical dashed line indicates the condensation point .
[ gammaq ] ] plane in various heavy ion collisions .
the full round dots refer to au - au at 11.6 and pb - pb collisions at 40 , 80 , 158@xmath0 gev obtained in the analysis a , whilst the hollow square dot has been obtained in ref .
@xcite by using particle ratios measured at midrapidity in au - au collisions at @xmath133 gev .
the hollow round dot without error bars refers to pb - pb collisions at 30@xmath0 gev and has been obtained by forcing @xmath9 and @xmath70 to lie on the parabola fitted to the full round dots.[tmu ] ] gev @xcite and the error is only statistical ; the arrow on the left signifies that its associated baryon chemical potential is lower than that estimated at @xmath133 gev @xcite used here . for the sps energy points
the statistical errors are indicated with solid lines , while the contribution of the common systematic error is shown as a dotted line .
also shown the theoretical values for a hadron gas along the fitted chemical freeze - out curve shown in fig .
[ tmu ] , for different values of @xmath1 .
[ kpi ] ] estimated from the fits a ( full dots ) and b ( hollow dots ) as a function of the fitted baryon - chemical potential . also shown the theoretical values for a hadron gas along the fitted chemical freeze - out curve shown in fig .
[ tmu ] , for different values of @xmath1 .
|
we present a detailed study of chemical freeze - out in nucleus - nucleus collisions at beam energies of 11.6 , 30 , 40 , 80 and 158@xmath0 gev . by analyzing hadronic multiplicities within the statistical hadronization approach ,
we have studied the strangeness production as a function of centre of mass energy and of the parameters of the source . we have tested and compared different versions of the statistical model , with special emphasis on possible explanations of the observed strangeness hadronic phase space under - saturation .
we show that , in this energy range , the use of hadron yields at midrapidity instead of in full phase space artificially enhances strangeness production and could lead to incorrect conclusions as far as the occurrence of full chemical equilibrium is concerned .
in addition to the basic model with an extra strange quark non - equilibrium parameter , we have tested three more schemes : a two - component model superimposing hadrons coming out of single nucleon - nucleon interactions to those emerging from large fireballs at equilibrium , a model with local strangeness neutrality and a model with strange and light quark non - equilibrium parameters .
the behaviour of the source parameters as a function of colliding system and collision energy is studied .
the description of strangeness production entails a non - monotonic energy dependence of strangeness saturation parameter @xmath1 with a maximum around 30@xmath0 gev .
we also present predictions of the production rates of still unmeasured hadrons including the newly discovered @xmath2 pentaquark baryon .
|
the single - electron transistor ( set ) is a highly sensitive electrometer , conventionally based on sequential tunneling of electrons in the coulomb blockade regime .
@xcite unfortunately , the @xmath7-time due to cable capacitance ( @xmath8 nf ) and device resistance ( @xmath9 ) limits the bandwidth to a few khz .
this is a major drawback for the direct application of sets as it limits the operation to a regime where @xmath10-noise is strong , either due to background charge noise , @xcite or to variations in the tunneling resistance .
@xcite the limitation on the frequency bandwidth can be bypassed using microwave techniques for reading out the real part of the set impedance . in these so called rf - sets , @xcite an @xmath11-circuit
is used to transform the high impedance of the set to a high - frequency setup with a characteristic impedance of @xmath12 @xmath13 .
the transformation is frequency selective , and a good match is obtained only over a frequency range @xmath14 , where @xmath15 is the resonant frequency of the matching circuit , and @xmath16 under fully matched conditions .
typical charge sensitivities of rf - sets amount to @xmath17 @xmath3 over a signal bandwidth of @xmath18 mhz .
@xcite in order to improve the performance of rf - sets , the operating regime has to be brought from sequential tunneling to co - tunneling , i.e. , from the weak to the strong tunneling regime .
@xcite the energy sensitivity including back action noise has been estimated to be @xmath19 in the sequential tunneling regime .
@xcite in the co - tunneling regime , the sensitivity is expected to approach @xmath20 . @xcite
another benefit of the strong tunneling regime is that a wider bandwidth can be obtained as the @xmath21-factor can be made smaller .
however , the effective coulomb energy diminishes rapidly with lowering resistance @xmath22 of the tunnel barriers .
this behavior has been summarized for single junctions by wang _
@xcite they find , e.g. , that @xmath23 for @xmath24 k@xmath13 . provided that @xmath25 , the strong tunneling sets are expected to operate well as rf - sets .
@xcite this regime of operation has been strived for by brenning _
@xcite they managed to fabricate al / alo@xmath26 tunnel junction sets with @xmath27 k , and a total resistance of @xmath28 k@xmath13 . for the charge sensitivity
, they reached @xmath29 @xmath3 at 4.2 k , which is the best liquid helium result obtained so far .
single - walled carbon nanotubes ( swcnts ) provide an alternative approach to the metallic sets in the strong tunneling regime .
the first reports on single - electron charging effects in individual tubes and bundles were published in 1997 .
@xcite charging energies of about @xmath30 mev were quickly observed .
@xcite since then , contacting techniques have greatly improved , and impedances of @xmath31@xmath18 k@xmath13 can be rather routinely obtained , e.g. , using pd contacts .
@xcite swcnts with large @xmath32 are very promising for rf - sets , especially since their shot noise has been found to be well below the schottky value .
@xcite in this article , we report on rf - sets made from swcnts in the strong tunneling regime .
we find a charge sensitivity of @xmath33 @xmath3 at @xmath0 k , which nearly equals that obtained by brenning _
@xcite compared with previous carbon nanotube rf - sets,@xcite the improvement is by a factor ranging from 7 to 200 . in combination with a bandwidth of ,
our results represent a considerable improvement for broadband charge sensing , e.g. , for fast read - out of single - electron devices such as quantum dots .
our swcnts are grown from patterned catalyst islands , following the approach of kong _ et al . _
@xcite rf - set operation necessitates the use of insulating substrates in order to minimize the shunt capacitance .
we use sapphire for lower losses and charge noise compared to conventional si / sio@xmath34 .
the chemical vapor deposition ( cvd ) takes place in a ceramic tube furnace from a gas mixture of ar , h@xmath34 , and ch@xmath35 at @xmath36 @xmath37 .
after growth , pairs of @xmath38/@xmath39 nm ti / au contacts , @xmath40 @xmath41 m apart , are defined between the catalyst islands by electron beam lithography . a central top - gate ,
@xmath42 @xmath41 m wide , is deposited between the contacts . under the actual @xmath38 nm ti gate ,
an insulating barrier is formed by five @xmath43 nm al layers , each oxidized @xmath43 min at atmosphere .
finally , we deposit a cr / au mask with @xmath44 @xmath41 m bond - pads .
the final device layout is illustrated in fig.1(a ) .
a schematic of the low - temperature measurement setup is shown in fig .
1(b).@xcite the sample connects to an @xmath11-circuit , formed by the inductance of a surface mount inductor and the bond - wires , and the parasitic capacitance of the bond - pads .
the circuit is connected to a coplanar transmission line with a surface mount bias - tee that couples the dc - bias and rf - signal .
the top - gate is connected to a separate coaxial line for high - bandwidth modulation .
we use a home - made low - noise amplifier with a frequency range of @xmath45@xmath46 mhz .
@xcite the rf - input is coupled to the coaxial line through a @xmath47 db directional coupler and a @xmath30 db attenuator that reduces the noise from room temperature .
the rf - output is detected in a fashion dependent of the goal of the measurement : 1 ) by a spectrum analyzer to investigate the carrier modulation spectra , or 2 ) by mixer demodulation for homodyne detection at a particular frequency .
fig . 2(a ) shows the differential conductance ( @xmath48 ) gate and bias voltages ( @xmath49 , @xmath50 ) . for @xmath51 mv , @xmath52@xmath53 @xmath54 , indicating a high - quality swcnt sample with highly transparent contacts .
however , there is a clear but smooth coulomb modulation pattern at @xmath55 . as such
, the sample behaves as a set in the strong tunneling regime . by tracing the coulomb diamonds
, we find the addition energies @xmath56 , with coulomb energy @xmath57 and level spacing @xmath58 .
the values in fig.2(b ) cover the range @xmath59@xmath60 mev , which sets an upper bound of @xmath61 k , corresponding to a total capacitance of the set island of @xmath62 af . from the gate modulation and slopes of the diamonds in fig .
2(a ) , we estimate that the gate capacitance @xmath63 af . in units of @xmath54 as a function of gate voltage @xmath49 and bias voltage @xmath50 .
the lines mark the charge degeneracies .
( b ) values of the addition energy @xmath64 deduced from ( a ) as the degeneracy crossings in @xmath50 . ] for rf - set operation , the optimum operating point was found by searching for points of perfect matching , i.e. , vanishing reflection at the resonance frequency @xmath65 mhz .
the signal was homodyne detected by mixer , and the phase was tuned to be sensitive only to the real part of the set impedance .
we found three points around @xmath66@xmath67 v with maximum differential response , coinciding with perfect matching ( fig .
using a spectrum analyzer , the input carrier power was tuned to obtain maximum signal - to - noise ratio of the sidebands at @xmath68 , while keeping a small gate - charge modulation of @xmath69 at ( fig .
the signal - to - noise ratio ( snr ) of both sidebands yields a charge sensitivity of @xmath70 @xmath3 , corresponding to an uncoupled energy sensitivity of @xmath71 .
the frequency response was mapped out by repeating the sensitivity measurement over a range of modulation frequencies of @xmath72 ( fig .
we found a bandwidth of @xmath73 and observed that @xmath10-noise only contributes significantly below @xmath74@xmath43 mhz .
gate voltage sweeps of each @xmath43 @xmath41s .
circles mark the points of perfect matching used in ( b ) and ( c ) .
( b ) power spectrum ( @xmath40 mhz spectral resolution ) under @xmath1 mhz carrier excitation ( @xmath75 dbm ) and @xmath31 mhz gate modulation of .
( c ) charge sensitivity versus modulation frequency @xmath76 , with / without a @xmath77 mhz low - pass filter on the gate - line ( open / filled symbols ) .
the lines indicate the roll - off limited by the bandwidth and @xmath10-noise at low frequency . ]
optimization of rf - set sensitivity has been treated in several papers .
@xcite the ultimate shot noise limited sensitivity was found in ref .
21 to be @xmath78[1.4em][b]{10.9ex}{$\displaystyle r_{\sigma}c_{\sigma}\frac{k_{\mathrm{b}}t}{e^2/c_{\sigma}}$}},\ ] ] which in our case ( @xmath79 k@xmath13 , @xmath80 af ) amounts to @xmath81 @xmath3 .
when the noise of the amplifier dominates the performance , the sensitivity scales with the noise amplitude and the differential response of the rf reflectance ( @xmath82 ) as @xmath83[0.8em][b]{8.5ex}{$2\,k_{\mathrm{b}}t_{\mathrm{n}}\ , z_0$}}}{\nu_{0}\times|\partial\gamma/\partial q|},\ ] ] where @xmath84 denotes the noise temperature of the amplifier and @xmath85 the amplitude of the rf carrier signal . in our case ,
the amplitude over the set is @xmath86 mv at perfect matching . in ref .
23 , an approximation of eq .
( 2 ) was found from the orthodox theory as @xmath87 with @xmath32 and @xmath84 in units of k , and @xmath88 .
assuming that heating does not play a significant role , we estimate from that @xmath89 @xmath3 , taking @xmath90 k , @xmath91 k@xmath13 , @xmath92 k@xmath13 , and @xmath93 k. all of these estimates are based on the classical orthodox theory for coulomb blockade , @xcite which is only valid in the weak tunneling regime
. however , as discussed above , we are in the strong tunneling regime .
one way around this issue is simply to adopt the effective charging energy @xmath94 , calculated by wang _ _ et al.__@xcite then the estimate for the charge sensitivity becomes @xmath95 @xmath3 , which agrees well with the measured value . in making our estimates ,
we have neglected the effects of quantum level spacing ( see discussion in ref .
we have shown that a swcnt quantum dot in the strong tunneling regime can be operated as an excellent rf - set at liquid helium temperature ( @xmath0 k ) .
we obtain a charge sensitivity of @xmath33 @xmath3 , which is nearly as good as the best al / alo@xmath26 results.@xcite it represents an enhancement of at least a factor of 7 compared with previous carbon nanotube rf - sets.@xcite very recently , comparable sensitivities were achieved with other types of sets : 1 ) inas / inp heterostructured nanowires,@xcite and 2 ) electrostatically defined si quantum dots.@xcite with a bandwidth of @xmath4 mhz , our gain - bandwidth product amounts to @xmath96 @xmath6 , where gain is defined as @xmath97 .
that is by more than one order of magnitude better than for typical rf - sets.@xcite therefore , swcnt based rf - sets may have value in applications where high speed is needed .
one possible application is for charge sensing on quantum dots in the context of quantum computing . here
, the coupling to a nearby quantum dot structure , e.g. , another nanotube , could be achieved via an antenna gate as it was recently demonstrated with sige nanowires.@xcite the best way to enhance the sensitivity further would be to lower the temperature so the coulomb modulation is fully developed . for al / alo@xmath26 junction sets , the improvement at 40 mk was @xmath98 @xmath3 in the superconducting state and @xmath99 @xmath3 in the normal state .
@xcite to improve the bandwidth , the only option is to increase the resonance frequency of the @xmath11-circuit .
the limit set by the bode - fano criterium states that the maximum achievable bandwidth is @xmath100 , which in our case amounts to @xmath101 mhz .
we wish to acknowledge h. i. jrgensen , k. grove - rasmussen , t. heikkil , m. paalanen , p. e. lindelof , and b. placais for fruitful discussions .
this work was supported by the academy of finland grant 213496 and by the eu under contract fp6-ist-021285 - 2 . by studying the coulomb diamond pattern around the optimum sensitivity
we deduce that @xmath102 and @xmath103 .
this suggests that we are in the quantum coulomb blockade regime ( qcb ) , where @xmath104 and only few levels are involved in transport .
however , comparing the conductance peak line - shapes around optimum sensitivity with theory @xcite , we find deviations between the experimental data and qcb predictions . from the experimental data we get @xmath105 at peak maximum and @xmath106 @xmath107 as optimum transconductance .
these values should be compared to @xmath108 and @xmath109 @xmath107 as predicted by theory .
also the predictions from the metallic coulomb blockade regime with @xmath110 and @xmath111 @xmath107 are deviating , but the peak maximum is close to the measured .
this discrepancy is explained by the fact that we are in the strong tunneling regime with @xmath112 which broadens the energy levels .
since we still have strong coulomb blockade we are not quite in the breit - wigner limit , but rather in an intermediate regime @xmath113 which behaves closest to the strong tunneling description .
|
we have investigated radio - frequency single - electron transistor operation of single - walled carbon nanotube quantum dots in the strong tunneling regime . at a temperature of @xmath0 k and with carrier frequency @xmath1 mhz , we reach a charge sensitivity of @xmath2 @xmath3 over a bandwidth of @xmath4 mhz .
our results indicate a gain - bandwidth product of @xmath5 @xmath6 , which is by one order of magnitude better than for typical rf - sets .
|
the renewed interest in the casimir effect @xcite in the last 15 years has fueled numerous complementary studies on different aspects of this long - range interaction . during the last decade or so various papers
have focused on the entropy related to the casimir interactions between two solid bodies @xcite .
an important reason for this interest lies in the fact that the casimir entropy can reach negative values .
this effect has first been theoretically discussed for two plates modelled by drude - type metals @xcite .
its origin lies in the suppression of the reflectivity of the transverse electric mode at low frequencies due to a finite zero - frequency conductivity @xcite .
if the conductivity diverges in the low - frequency limit , as is the case for perfect metals or metals within the plasma model , the entropy in the plane - plane configuration will remain positive for all temperatures @xcite .
the existence of a negative casimir entropy in this geometry is therefore related to the dissipation inside the metal .
however dissipation is not the only mechanism giving rise to negative casimir entropies . for the casimir - polder interaction between an atom and a perfect - metal plate , negative values of the entropy
were found at low temperatures @xcite .
studies of the sphere - plane configuration @xcite and the sphere - sphere configuration @xcite for perfect reflectors demonstrated that negative casimir entropies can be of purely geometric origin . considering the large - distance limit
, we shall see that in these geometries the casimir entropy can become negative for perfect metals , but it remains positive if both objects involved are described by the drude model implying a finite zero - frequency conductivity .
another indication of a negative casimir entropy is given by the non - monotonic behavior of the thermal casimir force . such a behavior has been found in the electromagnetic case in the sphere - plane geometry @xcite as well as for a scalar field satisfying dirichlet boundary conditions in the sphere - plane and the cylinder - plane geometries @xcite . in the present paper we will study in more detail the geometrical origin of negative casimir entropies . before doing so
, we emphasize that negative values of the casimir entropy are not in conflict with well established thermodynamic principles .
while entropies should be positive , the casimir entropy actually is a difference of entropies and can therefore well be negative @xcite .
however , even the casimir entropy has to tend towards zero in the zero - temperature limit .
this is indeed the case for the drude model for any nonvanishing value of the zero - frequency conductivity @xcite .
furthermore , close inspection reveals that strictly positive casimir entropies over the whole temperature range are rather the exception than the rule . in order to explore the geometric origin of the negative casimir entropy ,
a recent study considered the retarded casimir - polder interaction between a nanoobject and a plane or between two nanoobjects @xcite .
the properties of the nanoobjects were described in terms of their electric and magnetic polarizabilities . by allowing also for anisotropic polarizabilities , a variety of scenarios
could be generated , thus providing insight into the conditions under which negative casimir entropies can occur . here ,
we start from a scattering approach for the electromagnetic field and emphasize the contributions of different scattering channels .
first , we observe that the negativity of the casimir entropy becomes most pronounced when the distance between sphere and plane or between the two spheres is large compared to the radius of the spheres .
this observation holds despite the fact that the entropy scales with the third power of the radius for small radii so that its overall value will be strongly suppressed . as a consequence ,
the large - distance limit is appropriate to identify the physical mechanism responsible for the negative casimir entropy . in particular
, this limit permits us to restrict the reflection at the spheres to the dipole modes , @xmath0 .
the large - distance limit considerably simplifies the expression for the casimir free energy within the scattering approach as we shall see in sect .
[ sec : largedistanceapproximation ] .
firstly , the values of the quantum number @xmath1 of the @xmath2-component of the angular momentum , which is conserved for geometries of interest here , are constrained to @xmath3 and @xmath4 .
secondly , it is sufficient to account for one single scattering roundtrip between the two objects involved . in total , we are left with three distinct types of channels .
two of these channels leave the polarization type unchanged and are only distinguished by the value of @xmath5 .
the third channel involves a change of polarization and occurs only for @xmath6 .
the most relevant channel for the negative casimir entropy is the last one .
as we shall see , this scattering channel is characterized by a casimir free energy which increases monotonically with temperature .
its contribution to the casimir entropy vanishes at zero temperature as well as in the high - temperature limit , but is negative for any temperatures in between .
this result points towards the importance of polarization mixing in the appearance of a negative casimir entropy . the scenarios of different geometries and zero - frequency conductivity sketched in the beginning of this section
fit nicely into this picture . in the plane - plane configuration ,
the polarization is conserved at each reflection .
therefore , the negative casimir entropy appearing for drude - type damping in that case can not be of geometric origin . on the other hand , for the sphere - plane and sphere - sphere configurations
, drude - type metals will suppress polarization mixing .
therefore , it is to be expected that at least one of the two scatterers should have a divergent zero - frequency conductivity in order to allow for a negative casimir entropy in the large - distance limit .
in our analysis of the negative casimir entropy , we will concentrate on the sphere - plane configuration and the sphere - sphere configuration depicted in figs . [
fig : geometry]a and b , respectively , together with the corresponding geometric parameters .
the distance between the surfaces of the two objects will always be denoted by @xmath7 . for the plane - sphere geometry ,
the natural length scale within our analysis will turn out to be @xmath8 , which measures the distance between the surface of the plane and the center of the sphere of radius @xmath9 . in the sphere - sphere geometry
, the natural length scale @xmath10 refers to the distance between the centers of the two spheres with radii @xmath11 and @xmath12 .
whenever we refer to the two geometries at the same time , we will denote the natural length scales as @xmath13 .
for example , as a dimensionless temperature we will use @xmath14 which implies @xmath15 in the plane - sphere geometry and @xmath16 in the sphere - sphere geometry . here , @xmath17 , @xmath18 , and @xmath19 are the boltzmann constant , the planck constant and the speed of light , respectively . in this paper , we make use of a formalism based on imaginary frequencies @xmath20 .
the corresponding dimensionless imaginary frequency is defined by @xmath21 while @xmath22 denotes the distance between the plane and the center of the sphere and @xmath23 denotes the distance between the centers of the spheres . ]
the negative casimir entropy for the sphere - plane geometry with perfect metals is shown in fig .
[ fig : sp_entropy_2d_unscaled ] as a function of the ratio of the distance @xmath7 between the sphere s surface and the plane to the radius @xmath9 of the sphere , and of the temperature @xmath24 .
the casimir entropy vanishes along the dashed line . below this line ,
the casimir entropy is negative while it is positive above .
the grey area indicates parameter regions where the casimir entropy has not been evaluated . between sphere and plane and the temperature @xmath24 .
the entropy vanishes along the dashed line . below this line ,
the entropy is negative and changes in steps of 0.0005 .
the minimum of the entropy is marked by a dot . above the dashed line ,
the entropy is positive and changes in steps of 0.001 .
no data have been calculated in the grey region . ] according to fig .
[ fig : sp_entropy_2d_unscaled ] , the casimir entropy becomes minimal at @xmath25 and @xmath26 . at a fixed temperature ,
the casimir entropy becomes positive if the radius @xmath9 is sufficiently large .
in contrast , for small radii , the casimir entropy will always be negative for temperatures below a threshold value to be specified in the discussion of the free energy ( [ eq : freeenergysphereplane ] ) , see below .
it would be expected that the sphere - plane configuration can be obtained from the sphere - sphere configuration by letting the radius of one of the two spheres go to infinity . to illustrate this transition
, we show in fig .
[ fig : ss2sp_extremum ] the position of the minimum of the casimir entropy for perfect metals as a function of the geometric parameters and the temperature for two spheres with radii @xmath27 and @xmath28 for the smaller and larger sphere , respectively . for the filled circles , the corresponding values of the ratio @xmath29 are indicated , ranging from @xmath30 for spheres of equal radius to the extreme case @xmath31 for which the position of the entropy minimum for the plane - sphere geometry is shown .
the data clearly indicate a smooth transition between the sphere - sphere and the plane - sphere configuration . and the temperature @xmath24 .
the points refer to different ratios of the radii @xmath28 and @xmath27 of the larger and smaller sphere , respectively . for the filled circles ,
the ratio of radii is indicated in the plot .
the point marked by @xmath32 corresponds to the position of the minimum of the casimir entropy in fig .
[ fig : sp_entropy_2d_unscaled ] . ] as far as the physical origin of the negative casimir entropy is concerned , the presentation of the data in fig .
[ fig : sp_entropy_2d_unscaled ] is somewhat misleading because for small sphere radius , the entropy decreases with the volume of the sphere .
it is thus appropriate to scale the entropy with @xmath33 .
the result is depicted in fig .
[ fig : sp_entropy_2d_scaled ] where the dashed line again indicates a vanishing casimir entropy and negative values of the casimir entropy are found below the dashed line .
note that in this plot , in contrast to fig .
[ fig : sp_entropy_2d_unscaled ] , small sphere radii are on the left side .
the minimum of the scaled casimir entropy lies at @xmath34 .
we can thus conclude that the large - distance limit @xmath35 is well suited for an analysis of the negative casimir entropy .
is displayed for perfect metals as a function of the inverse of the distance @xmath22 between sphere s center and the plane and the temperature @xmath24 .
the entropy vanishes along the dashed line . below this line ,
the entropy is negative and changes in steps of 0.00125 . above the dashed line ,
the entropy is positive and changes in steps of 0.0025 .
no data have been calculated in the grey region . ] by means of the thermodynamic relation @xmath36 the casimir entropy @xmath37 for the sphere - plane configuration with perfect metals in the large - distance limit can easily be obtained from the expression for the casimir free energy @xmath38 in this limit @xcite @xmath39\,,\ ] ] where we introduced the abbreviation @xmath40 taking the derivative with respect to temperature , one finds negative values for the casimir entropy for temperatures satisfying @xmath41 in agreement with the data shown in fig .
[ fig : sp_entropy_2d_scaled ] for small values of @xmath9 . in order to obtain information about the physical origin of the negative casimir entropy , in sect .
[ subsec : sp_large_distance ] we will decompose the free energy ( [ eq : freeenergysphereplane ] ) and the entropy derived from it into contributions arising from different scattering channels . for the sphere - sphere geometry , the region of negative casimir entropy for perfect - metal spheres of equal radii
is displayed in fig .
[ fig : ss_entropy_2d ] where the entropy is scaled by @xmath42 .
the dashed line separates the regions of negative and positive casimir entropies with negative values appearing inside the region delimited by the dashed line . as an obvious difference to the sphere - plane geometry shown in fig . [
fig : sp_entropy_2d_scaled ] we observe that while in the latter case negative values of the casimir entropy are found down to the lowest temperatures , this is not the case in the sphere - sphere geometry .
this fact has already been noted in @xcite .
furthermore , the region of negative casimir entropies ends before @xmath43 reaches its maximum value of @xmath44 .
is depicted for perfect metals as a function of the inverse of the distance @xmath23 between the centers of the spheres and of the temperature @xmath24 for spheres of equal radii , @xmath45 .
the entropy vanishes along the dashed line . inside the region
bounded by the dashed line , the entropy is negative and changes in steps of 0.025 .
outside of the dashed line , the entropy is positive and changes in steps of 0.005 .
the density of the contour lines has been decreased at higher temperatures to improve the clarity of the plot . ] despite these differences , fig .
[ fig : ss_entropy_2d ] suggests that again interesting insights into the physics of the negative casimir entropy can be obtained from the large - distance limit @xmath46 . the free energy for perfect - metal spheres in this limit is given by @xcite @xmath47\ , , \end{aligned}\ ] ] where @xmath48 was defined in ( [ eq : def_g ] ) .
a decomposition of this result and the corresponding entropy in terms of the scattering channels will be carried out in sect .
[ subsec : ss_large_distance ] .
within the scattering approach , the casimir free energy can be expressed as @xmath49\right)\,.\ ] ] the first sum runs over the matsubara frequencies @xmath50 with @xmath51 .
the prime indicates that only one half of the @xmath52-term should be taken .
the geometries depicted in fig . [ fig : geometry ] are symmetric under rotations around the @xmath2-axis , thus allowing us to decompose the scattering problem into subspaces of fixed eigenvalues @xmath1 of the @xmath2-component of the angular momentum .
since the sign of @xmath1 is irrelevant for the scattering process , we sum only over positive values of @xmath1 , leading to a prefactor 2 except for @xmath53 as again indicated by a prime .
the matrix @xmath54 describes a roundtrip scattering process between the two scattering objects at imaginary frequency @xmath55 in the subspace of the @xmath2-component of the angular momentum characterized by @xmath1 .
the roundtrip scattering operator @xmath56 contains four building blocks , namely the translation operator @xmath57 from the reference frame of object 1 to that of object 2 , the reflection operator @xmath58 on object 2 , the reverse translation operator @xmath59 , and finally the reflection operator @xmath60 on object 1 . while keeping the quantum number @xmath1 unchanged , these operators will in general modify the other parameters of the scattered modes like their polarization and , for spherical waves , their angular momentum quantum number @xmath61 or , for plane waves , their wave vector @xmath62 .
we argued in the previous section that the limit of large distances between the scattering objects is appropriate to analyze the origin of negative casimir entropies . to be specific , we assume that the distance @xmath13 is much larger than the sphere radius @xmath9 . in the case of two spheres , @xmath9 refers to the larger of the two radii . as we shall see now
, this limit allows us to quantify the contribution of each scattering channel to the casimir entropy .
since the matrix elements of the translation operators for imaginary frequencies decay exponentially with @xmath13 , the highest relevant frequencies are of the order of @xmath63 .
the reflection on a sphere will then contribute a factor @xmath64 where @xmath61 denotes the order of the multipole wave scattered at the sphere . as a consequence
, we may restrict the scattering at a sphere to @xmath0 . in the large - distance limit
, the casimir entropy for the sphere - plane geometry and the sphere - sphere geometry thus scales with @xmath65 and @xmath66 , respectively .
these factors are at the origin of the scaling employed in the previous section .
furthermore , in the large - distance limit the matrix elements of the roundtrip operator are very small and we may expand the logarithm in ( [ eq : matsubarafreeenergy ] ) . in view of @xmath67 $ ] with @xmath68 denoting the trace , we then obtain @xmath69\,,\ ] ] where @xmath70 and @xmath71 denote the polarizations on the two scattering objects .
even though we restrict our considerations to @xmath0 on each sphere , the translation operators @xmath72 present in the matrix elements @xmath73 implicitly give rise to a sum over other multipole moments @xmath74 or , in the sphere - plane geometry , to an integral over the projection @xmath75 of the wave vector onto the plane .
the polarizations @xmath70 and @xmath71 in ( [ eq : freeenergysum2 ] ) can be either transverse electric ( te ) or transverse magnetic ( tm ) and correspond to the mode polarizations on the two scatterers .
note that on a sphere , the te and tm polarizations are sometimes referred to as h and e polarizations , respectively . while in general , the polarization may change in the course of the scattering process , this is not the case for @xmath53 . in this case , for a te ( tm ) mode on the sphere the electric ( magnetic ) field will only have a component in the plane perpendicular to the @xmath2 direction . as a consequence ,
the expansion of this mode on the other sphere or the plane will contain exclusively te ( tm ) modes and no polarization change ensues .
we thus have to account for three essentially different scattering channels .
the channels where the polarization is conserved contribute for @xmath53 and @xmath76 .
a third channel involves a change of polarization and is restricted to @xmath76 .
the latter channel is of particular interest for our discussion for the following reason .
focussing on the contribution of the translation operator , it is clear that in the absence of a shift , the polarization of the mode can not change . in ( [ eq : freeenergysum2 ] ) , for dimensional reasons , the shift can only appear in the combination @xmath77 . as a consequence ,
the @xmath52 term will vanish if the polarizations @xmath70 and @xmath71 differ , implying in turn that the free energy contribution of this channel will vanish at high temperatures .
since the contribution to the casimir free energy of the polarization - changing channel at zero temperature is negative and turns out to be monotonically increasing , its contribution to the casimir entropy in view of ( [ eq : defentropy ] ) will always be negative . while the differences between the scattering channels discussed
so far were mainly due to the translation operators @xmath72 , the reflection properties of the sphere(s ) offer an interesting way to select channels .
the reflection at a perfectly conducting ( pc ) sphere of radius @xmath9 to leading order in @xmath78 is dominated by the mie coefficients with @xmath0 . for the tm mode ,
the mie coefficient is then given by @xmath79 and for the te mode it reads @xmath80 we thus have @xmath81 , leading to simple numerical relations between the contributions of the polarization - conserving scattering channels .
in contrast , for spheres made of a metal described by the drude model ( d ) , one finds @xmath82 while @xmath83 is of order @xmath84 and therefore negligible within the large - distance approximation .
switching from a perfectly conducting sphere to a drude metal sphere allows us to suppress the reflection of te modes on the sphere .
in particular , the polarization - changing channel will become irrelevant as we will explain in more detail in sect .
[ sec : drude ] .
these general considerations and their consequences for the casimir entropy will be worked out more explicitly in the following section dealing with the specific geometries shown in fig .
[ fig : geometry ] .
within the large - distance approximation , we only need the matrix elements of the roundtrip operator for @xmath53 and @xmath4 to obtain the free energy by means of ( [ eq : freeenergysum2 ] ) .
we then can make use of ( [ eq : defentropy ] ) to obtain the entropy . for the sphere - plane geometry
, the matrix element of the roundtrip operator can easily be obtained from the expressions given in ref .
@xcite . for @xmath53 , we obtain @xmath85 and @xmath86 with the dimensionless frequency @xmath87 introduced in ( [ eq : dimensionlessfrequency ] ) . for @xmath76 ,
the matrix elements for the roundtrips where polarization is conserved are @xmath88 and @xmath89 for roundtrips involving a change of polarization , one finds @xmath90 and @xmath91 here , the first subscript of the roundtrip operator @xmath92 refers to the polarization on the sphere while the second subscript indicates the polarization on the plane .
the matrix elements ( [ eq : m1tmte ] ) and ( [ eq : m1tetm ] ) for scattering processes involving a change of polarization vanish in the limit of vanishing frequency @xmath87 as discussed in the previous section .
making use of ( [ eq : freeenergysum2 ] ) , we can decompose the casimir free energy for perfect metal sphere and plane into contributions from the different scattering channels according to @xmath93\ , .
\end{aligned}\ ] ] evaluating the corresponding matsubara sums , we find for @xmath53 from ( [ eq : m0tmtm ] ) @xmath94\,.\ ] ] here , we have made use of the dimensionless temperature ( [ eq : dimensionlesstemperature ] ) and the function @xmath95 defined in ( [ eq : def_g ] ) . for @xmath76 , we obtain from ( [ eq : m1tmtm ] ) and ( [ eq : m1tmte ] ) @xmath96\ ] ] and @xmath97 respectively .
the remaining three contributions in ( [ eq : psfreeenergygeneral ] ) are related to the expressions just given by a factor of 1/2 according to the relations ( [ eq : m0tete ] ) , ( [ eq : m1tete ] ) , and ( [ eq : m1tetm ] ) .
summing up all terms , we recover the free energy ( [ eq : freeenergysphereplane ] ) for sphere and plane made of perfect conductors obtained earlier in ref .
@xcite . the decomposition in terms of the different scattering channels
is found to be in agreement with the expressions obtained in ref . @xcite . in order to make the connection ,
one sets the electric and magnetic polarizability equal to @xmath98 and @xmath99 , respectively . for vanishing polarizability in the transverse direction
, one finds the contribution for @xmath53 while an isotropic polarizability yields the sum of the contributions from @xmath53 and @xmath4 .
it is now straightforward to obtain the contributions of the various scattering channels to the casimir entropy . expressing the casimir entropy in terms of a rescaled casimir entropy @xmath100 according to @xmath101 the definition of the entropy ( [ eq : defentropy ] ) turns into @xmath102
we then obtain from ( [ eq : sp_f0tmtm ] ) , ( [ eq : sp_f1tmtm ] ) , and ( [ eq : sp_f1tmte ] ) @xmath103\ ] ] @xmath104 \end{aligned}\ ] ] @xmath105\,.\ ] ] the remaining three contributions to the casimir entropy can be obtained by simple multiplication with a factor @xmath44 as before for the free energy and the matrix elements of the roundtrip operator . in fig .
[ fig : sp_entropy_largedist ] , the temperature dependence of the contributions ( [ eq : sp_s0tmtm ] ) , ( [ eq : sp_s1tmtm ] ) , and ( [ eq : sp_s1tmte ] ) from the channels with tm polarization on the sphere to the casimir entropy are shown . among the polarization - conserving channels ,
the @xmath53 contribution is positive for all temperatures .
the @xmath76 contribution , while being slightly negative at sufficiently small temperatures , in combination with the @xmath53 contribution will still always lead to positive values of the entropy . in order to arrive at a negative entropy
, one needs the polarization - changing mode .
in fact , @xmath106 is negative for all temperatures as was already conjectured in sect .
[ sec : largedistanceapproximation ] . ) , ( [ eq : sp_s1tmtm ] ) , and ( [ eq : sp_s1tmte ] ) to the casimir entropy in the plane - sphere geometry arising from the scattering channels with tm polarization on the sphere are shown as a function of the dimensionless temperature .
while @xmath107 is always positive and @xmath108 becomes only slightly negative for small temperatures , @xmath106 is negative for all temperatures . ]
the negative contribution of the polarization - changing channel is indeed sufficiently large to render the sum of all contributions negative .
this can clearly be seen from the low - temperature expansions of the entropies ( [ eq : sp_s0tmtm])([eq : sp_s1tmte ] ) @xmath109 taking all six channels between perfect conductor plane and sphere into account , we obtain for the low - temperature expansion of the casimir entropy @xmath110 in the large - distance limit , we thus obtain negative values for the casimir entropy at sufficiently low temperatures . in the limit of vanishing temperature , the casimir entropy goes to zero in agreement with the third law of thermodynamics .
we now turn to the discussion of the casimir free energy and entropy of the sphere - sphere configuration depicted in fig .
[ fig : geometry]b . in order to determine the casimir free energy in the limit of large distance , @xmath111 , from the expression ( [ eq : freeenergysum2 ] ) , we first need to determine the matrix elements @xmath73 of the round - trip operator with @xmath53 and @xmath4 . as discussed in sect . [ sec : largedistanceapproximation ] , the large - distance approximation implies that on the spheres , we can restrict the field modes to dipole spherical waves , @xmath0 . for the spheres , the matrix elements of the reflection operators
are thus simply given by the mie coefficients ( [ eq : mie_a1 ] ) and ( [ eq : mie_b1 ] ) , and we have @xmath112 and @xmath113 where @xmath9 is the sphere radius and @xmath87 is the dimensionless imaginary frequency ( [ eq : dimensionlessfrequency ] ) .
in contrast to the reflection on a plane in the preceding subsection , the polarization of a spherical wave remains unchanged by the reflection at a sphere .
however , the translation operator for spherical waves with wave vector @xmath114 from the center of one sphere to the center of the other sphere leads to a mixing of polarizations .
the general expressions for the matrix elements of the translation operator @xcite can be simplified if the translation is performed along the @xmath2-axis @xcite as is the case in the setup displayed in fig .
[ fig : geometry]b . within the large - distance approximation
, we find for the channels conserving the polarization @xmath115(2\ell'+1)\\ & \qquad\times\begin{pmatrix } 1 & 1 & \ell'\\ 0 & 0 & 0 \end{pmatrix } \begin{pmatrix } 1 & 1 & \ell'\\ m & -m & 0 \end{pmatrix } h_{\ell'}^{(1)}(kd)\ , .
\end{aligned}\ ] ] for a change of polarization , @xmath116 , the matrix element vanishes for @xmath53 in agreement with the argument given in sect .
[ sec : largedistanceapproximation ] , while for @xmath76 we have @xmath117 the first two factors in the second lines of ( [ eq : tmpp ] ) and ( [ eq : tmppp ] ) denote wigner @xmath118 symbols while @xmath119 is a spherical bessel function of the third kind .
the overall sign in ( [ eq : tmppp ] ) is positive or negative depending on whether the translation is performed in positive or negative @xmath2-direction , respectively . employing the relation @xmath120 between the spherical bessel function of the third kind and the modified spherical bessel function @xmath121 , it is straightforward to express the matrix elements of the translation operator in imaginary frequency @xmath20 needed in order to evaluate the matsubara sum ( [ eq : freeenergysum2 ] ) . with the explicit expressions for the modified spherical bessel functions
@xmath122 the matrix elements needed in the following are then found to read @xmath123 and @xmath124 for @xmath125 while for differing polarizations @xmath70 and @xmath71 one finds @xmath126 in the last equation , the sign depends on the direction of translation with respect to the @xmath2-axis as discussed in the context of ( [ eq : tmppp ] ) .
we note that even though these matrix elements diverge for vanishing frequency @xmath87 , their products with the matrix elements ( [ eq : mmtmtm ] ) and ( [ eq : mmtete ] ) of the reflection operator remain finite .
while the combination of the two matrix elements yields a nonzero value if the polarization is conserved , the product of the matrix element ( [ eq : t1ppp ] ) for changing polarization with one of the reflection matrix elements ( [ eq : mmtmtm ] ) and ( [ eq : mmtete ] ) goes to zero for vanishing frequency @xmath87 . as a consequence , the contribution of the polarization - changing channels to the casimir free energy and entropy vanishes at high temperatures where the matsubara sum ( [ eq : freeenergysum2 ] ) is dominated by the @xmath52 term .
already at this point , we can therefore expect the same qualitative temperature dependence of the contributions of the polarization - changing channels as in the plane - sphere geometry .
these channels will thus play the same crucial role for an overall negative casimir entropy also for the sphere - sphere geometry .
the difference between the polarization - conserving and polarization - changing channels can be traced back to an extra factor @xmath127 appearing in the front of the right - hand side of ( [ eq : tmppp ] ) compared to ( [ eq : tmpp ] ) .
this factor ensures that for vanishing translation , @xmath128 , no change of polarization occurs as was argued in sect .
[ sec : largedistanceapproximation ] . with the matrix elements listed above , it is straightforward to evaluate the contributions of the various channels to the casimir free energy by means of ( [ eq : roundtripmatrixelement ] ) and ( [ eq : freeenergysum2 ] ) .
we decompose the casimir free energy into the contributions from the various scattering channels @xmath129\ , .
\end{aligned}\ ] ] the contributions to ( [ eq : ssfreeenergygeneral ] ) arising from polarization - conserving channels are given by @xmath130 and @xmath131 \end{aligned}\ ] ] together with @xmath132 the contributions of the polarization - changing channels are @xmath133 \end{aligned}\ ] ] and @xmath134 in these results , we make use of the dimensionless temperature ( [ eq : dimensionlesstemperature ] ) and the function @xmath95 defined in ( [ eq : def_g ] ) .
the total casimir free energy obtained from these expressions agrees with the result ( [ eq : freeenergyspheresphere ] ) and was already given in ref .
@xcite . as in the sphere - plane geometry
, the contributions from the different scattering channels are in agreement with the results presented in ref .
@xcite if the identifications explained above are made . in the high - temperature limit , @xmath135
, the polarization - conserving channels yield a contribution to the casimir free energy linear in temperature while the contributions of the polarization - changing channels vanish . on the other hand ,
all channels give rise to a negative casimir free energy at zero temperature . as a consequence , the contribution to the casimir entropy arising from the polarization - changing channels is negative for all temperatures as already expected above on the basis of the matrix elements of the translation operator . from the expressions listed for the contributions to the casimir free energy ,
the corresponding contributions to the casimir entropy can be obtained from its definition ( [ eq : defentropy ] ) .
introducing a rescaled casimir entropy @xmath100 by means of @xmath136 one finds together with the abbreviation ( [ eq : def_g ] ) @xmath137\\ & = 4s^{(0)}_{te , te } \end{aligned}\ ] ] @xmath138\\ & = 4s^{(1)}_{te , te } \end{aligned}\ ] ] @xmath139\\ & = s^{(0)}_{te , tm}\ , .
\end{aligned}\ ] ] the temperature dependence of the contributions ( [ eq : ss_s0tmtm])([eq : ss_s1tmte ] ) to the casimir entropy is shown in fig .
[ fig : ss_entropy_largedist ] . at first sight , the curves resemble those presented in fig . [
fig : sp_entropy_largedist ] . in both cases ,
the contribution for @xmath53 is positive for all temperatures while the polarization - conserving contribution for @xmath76 starts out negative and becomes positive at higher temperatures .
the polarization - changing channels always yield a negative contribution to the casimir entropy .
a closer look reveals that , in contrast to the plane - sphere configuration , the negative contribution of the polarization - conserving channel with @xmath76 at low temperatures is much bigger than the contribution of the polarization - changing channel .
however , as the dashed curve in fig .
[ fig : ss_entropy_largedist ] shows , the sum of the casimir entropies of the polarization - conserving channels remains positive at all temperatures . ) , ( [ eq : ss_s1tmtm ] ) , and ( [ eq : ss_s1tmte ] ) to the casimir entropy of a sphere - sphere configuration are depicted as a function of the dimensionless temperature .
while @xmath107 is positive for all temperatures and @xmath108 can be positive or negative depending on temperature , @xmath106 always yields a negative contribution to the casimir entropy .
the dashed line represents the sum of the contributions of the two polarization - conserving transverse magnetic channels . ] in order to analyze in more detail the negative contributions to the casimir entropy , we consider the low - temperature expansions of the expressions ( [ eq : ss_s0tmtm])([eq : ss_s1tmte ] ) which are given by @xmath140 the dominant low - temperature contributions of order @xmath141 arise in the polarization - conserving channels .
however , they cancel each other . in view of the fact that according to ( [ eq : sp_s_lowtemp ] ) , the casimir entropy in the plane - sphere configuration contains a leading term of order @xmath141 ,
this may come as a surprise .
indeed , a leading cubic term in the temperature can be obtained for anisotropic objects @xcite . for the isotropic case considered here
, it can be shown that terms of order @xmath142 in the mie coefficients ( [ eq : mie_a1 ] ) and ( [ eq : mie_b1 ] ) lead to such a @xmath141 term in the sphere - sphere configuration .
however , this term is suppressed by a factor of @xmath143 relative to the terms discussed here and therefore negligible in the large - distance limit . with the order @xmath141 not contributing to the entropy ,
the appearance of a negative casimir entropy is a nonperturbative effect @xcite . in the next order , @xmath144 ,
the negative contribution of the polarization - changing channel is not sufficiently strong as compared to the positive contributions of that order .
therefore , at very low temperatures , the casimir entropy of two perfectly conducting spheres will be positive .
on the other hand , it turns out that the terms of order @xmath145 , which in ( [ eq : ss_s0tmtm_lowt])([eq : ss_s1tmte_lowt ] ) are all negative , lead indeed to a negative casimir entropy in an intermediate temperature range .
this was already shown in ref .
@xcite and is also visible in fig .
[ fig : ss_entropy_2d ] .
it is now interesting to study how the different channels can contribute in various ways to obtain either a positive casimir entropy for all temperatures or a negative casimir entropy in a certain temperature regime . to this end , in the next section we will also allow for spheres made of drude - type metals .
so far , we have studied the behavior of the various scattering channels and pointed out the relevance of the polarization - changing channels for the appearance of a negative casimir entropy . the weight
with which the scattering channels contribute can be modified by the physical properties of the objects involved . in ref .
@xcite this was done by choosing objects with anisotropic polarizabilities and by varying their electric and magnetic properties . here , we will do the latter by allowing the objects to be either made of perfectly conducting metals or drude - type metals which have a finite zero - frequency conductivity .
the results will further underline the relevance of the polarization - changing channels . for simplicity
, we will restrict ourselves in the following to setups consisting of two spheres as depicted in fig .
[ fig : geometry]b . then , as pointed out in sect .
[ subsec : ss_large_distance ] , the matrix elements of the reflection operator are given by the mie coefficients for @xmath0 . while for perfectly conducting spheres , the mie coefficients are the same up to a factor -2 , for drude metal spheres with a dc conductivity @xmath146 , the reflection of the te mode can be neglected in the large - distance limit where in addition to the conditions stated earlier , @xmath147 should hold .
these properties of the mie coefficients allow us to construct three different scenarios by choosing two spheres both made of perfect conductors , both made of drude metals , or one made of a perfect conductor and the other of a drude metal . in the first case ,
polarization - conserving as well as polarization - changing channels contribute as discussed in sect .
[ subsec : ss_large_distance ] .
in contrast , in the second case , only modes with tm polarization can complete roundtrips between the two spheres .
therefore , in this case , the polarization - changing channel is completely suppressed . in the third case ,
only one of the two polarization - changing channels and its weight relative to the polarization - conserving channels is modified with respect to two perfectly conducting spheres .
indicating in the superscript on the left - hand side the material of which the two spheres are made , we obtain for the three situations just described the rescaled entropy introduced in ( [ eq : rescaledentropy_ss ] ) with @xmath148 where the components are given in ( [ eq : ss_s0tmtm])([eq : ss_s1tmte ] ) .
these results are consistent with those obtained in ref . @xcite .
the temperature dependence of the casimir entropies ( [ eq : s_pcpc])([eq : s_dd ] ) is displayed in fig .
[ fig : ss_pc_d ] . at high temperatures ,
only the polarization - conserving channels contribute .
the difference between the case of perfectly conducting spheres and the other two cases shows the suppression of the polarization - conserving te channel due to a drude metal sphere . )
([eq : s_dd ] ) for a sphere - sphere setup where the spheres are either made of perfect conductors ( pc ) or drude metals ( d ) .
the inset shows the low - temperature behavior where a positive casimir entropy is found for all three cases .
the curve for d / d corresponds to the dashed curve in fig .
[ fig : ss_entropy_largedist ] . ]
the inset puts a special emphasis on the low - temperature behavior and shows clearly that in all three cases , the casimir entropy takes on positive values for very low temperatures .
however , only when all spheres are made of a drude metal does the casimir entropy remain positive for all temperatures .
this is the case where no polarization - changing channels contribute .
it is sufficient to allow for one polarization - changing channel by making one of the spheres perfectly conducting in order to obtain a temperature window in which the casimir entropy becomes negative .
the effect becomes even more pronounced if both spheres are perfectly conducting as the weight of the polarization - changing channels is doubled while the contribution of the polarization - conserving channels is only increased by a quarter .
the origin of the negative casimir entropy in the plane - sphere and sphere - sphere configuration has been analyzed in the limit where the distance between the objects is much larger than the radius of the sphere(s ) . in this limit , the casimir free energy and the casimir entropy can easily be decomposed into the contributions from various channels describing a roundtrip between the objects .
three kinds of channels have been identified , differing significantly in the temperature dependence of their contribution to the casimir entropy .
the second kind of channels also conserves polarization but involves spherical waves with @xmath76 .
while these channels yield a positive contribution to the casimir entropy at high temperatures , their contribution at sufficiently low temperatures is negative .
however , this negative part is compensated by the polarization - conserving channels with @xmath53 .
the third kind of channels is the most interesting one , because its contribution to the casimir entropy is negative for all temperatures .
this behavior is associated with polarization - changing channels which exist for @xmath76 but not for @xmath53 .
these channels are special , because their casimir free energy vanishes in the high - temperature limit .
this fact and the ensuing negative contribution to the casimir entropy has been traced back to the polarization - changing nature of the channels .
it can thus be concluded that polarization mixing in a scattering process is a crucial ingredient for the appearance of a negative casimir entropy , at least in the plane - sphere and sphere - sphere configuration .
which of the various channels contribute to the casimir entropy can be influenced by appropriately choosing the material out of which the scattering objects are made .
one can use the fact that in the long - distance limit the reflection of transverse electric modes at spheres made of a drude metal becomes negligible .
we have shown that for two drude metal spheres , the casimir entropy is positive for all temperatures .
in this situation , only roundtrip scattering processes involving transverse magnetic modes are relevant .
as soon as at least one of the spheres is perfectly conducting , polarization - changing processes occur and the casimir entropy is found to become negative in a certain temperature window .
kam and gli thank the laboratoire kastler brossel for their hospitality during the period of this work and cnrs and ens for financial support .
kam s work was further supported in part by grants from the simons foundation and the julian schwinger foundation .
i. brevik , j. b. aarseth , j. s. hye , and k. a. milton , in _ proceedings of the sixth workshop on quantum field theory under the influence of external conditions _ , edited by k. a. milton ( rinton press , princeton , 2004 ) , pp .
|
negative values of the casimir entropy occur quite frequently at low temperatures in arrangements of metallic objects .
the physical reason lies either in the dissipative nature of the metals as is the case for the plane - plane geometry or in the geometric form of the objects involved .
examples for the latter are the sphere - plane and the sphere - sphere geometry , where negative casimir entropies can occur already for perfect metal objects . after appropriately scaling out the size of the objects , negative casimir entropies of geometric origin
are particularly pronounced in the limit of large distances between the objects .
we analyze this limit in terms of the different scattering channels and demonstrate how the negativity of the casimir entropy is related to the polarization mixing arising in the scattering process . if all involved objects have a finite zero - frequency conductivity , the channels involving transverse electric modes are suppressed and the casimir entropy within the large - distance limit is found to be positive .
|
sdss j102913.94 + 262317.9 ( sdss j1029 + 2623 ; * ? ? ?
* ) is one of two examples of strongly lensed quasars produced by massive clusters of galaxies .
it was discovered in the sloan digital sky survey quasar lens search ( sqls ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , a survey to identify gravitationally lensed quasars from the spectroscopic sample of quasars in the sloan digital sky survey ( sdss ; * ? ? ?
the image separation of @xmath1 makes it the largest lensed quasar known to date .
@xcite found that the system consists of two images of a radio - loud quasar at @xmath2 created by a massive cluster of galaxies at @xmath3 .
the rareness of such quasar - cluster lens systems ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) validates the importance of understanding this system with extensive follow - up work . in this _ letter _ , we present new imaging and spectroscopic observations of this lens system .
based on these observations we identify a third lensed quasar image , measure the redshift of the lens cluster , and identify several additional lensed arcs .
we discuss the observations in
[ sec : obs ] , model and interpret the lens system in
[ sec : model ] , and summarize our results in [ sec : sum ] .
we obtained deep images of the system using the university of hawaii 2.2-meter ( uh88 ) and keck i telescopes . in the uh88 observations we took 1000 s b - band , 1200 s @xmath4-band , and 400
s @xmath4-band images on the nights of 2006 november 14 , 2007 may 17 , and 2007 november 12 using the tektronix 2048@xmath52048 ccd camera ( tek2k ) .
the typical seeing was @xmath6 and the conditions were photometric .
the photometry was calibrated by the standard star pg 0918 + 029 @xcite .
we obtained additional 720 s @xmath7-band and 880 s @xmath8-band images with the keck i telescope using the low resolution imaging spectrometer with the atmospheric dispersion compensator ( lris - adc ; * ? ? ?
* ) on 2008 january 4 when the seeing was @xmath9 . since no standard star was observed , we calibrated the images using the uh88 and sdss images .
these data were reduced and analyzed using standard iraf tasks .
crrccccccc a & @xmath10 & @xmath10 & 19.20 & 18.72 & 18.55 & 18.01 & 18.72 & 18.46 & 2.197 + b & @xmath11 & @xmath12 & 19.03 & 18.67 & 18.51 & 17.95 & 18.81 & 18.58 & 2.197 + c & @xmath13 & @xmath14 & 20.89 & 20.63 & 20.28 & 19.51 & 20.87 & 20.38 & 2.197 + g1a & @xmath15 & @xmath16 & @xmath1723.5 & 22.18 & 20.92 & 19.64 & 22.98 & 20.93 & 0.596 + g1b & @xmath18 & @xmath19 & @xmath1723.5 & 22.34 & 21.12 & 19.73 & 23.01 & 21.02 & 0.601 + g2 & @xmath20 & @xmath21 & @xmath1723.5 & 22.00 & 20.62 & 19.26 & 22.69 & 20.60 & 0.584 + [ tab : pos ] we show the keck image in figure [ fig : img1029 ] .
the results of the photometry and astrometry are summarized in table [ tab : pos ] .
several lensed arcs are clearly visible in the image . as suggested in @xcite , the central galaxy g1 consists of two components , which we name g1a and g1b .
@xcite argued that object c is not a lensed image on the basis of its different color from image a and b , and our follow - up results confirm that the color of object c is significantly different from those of the other lensed quasar images .
however , both by comparing the uh88 @xmath22-band images and from an ongoing monitoring at the fred lawrence whipple observatory 1.2-meter telescope ( see * ? ? ?
* ; * ? ? ?
* ) , we see that both b and c have faded by approximately 0.1 mag during 2007 .
this suggests that object c is a quasar , as variable blue sources at this magnitude level are typically quasars ( e.g. , * ? ? ?
we obtained spectra of objects b and c with the lris - adc at the keck i telescope on 2007 december 14 , and spectra of galaxies
g1a , g1b , and g2 with the faint object camera and spectrograph ( focas ; * ? ? ?
* ) at the subaru telescope on 2007 january 23 . in the keck lris - adc observation ,
the 800 s exposure was taken with a long slit ( @xmath23 width ) aligned with objects b and c. we used the d560 dichroic with the 400/8500 grating and the 400/3500 grism to achieve a spectral resolution of @xmath24 .
the observations were obtained in non - photometric conditions with poor @xmath25 seeing .
we used the deblending procedure of @xcite in which we assumed gaussian profiles for spatial cross sections to extract independent spectra of the two components .
the standard star feige 34 @xcite was used for flux calibration .
the subaru focas observation was conducted in @xmath26 on - chip binning mode with the 300b grism and the sy47 filter .
the @xmath27 s exposure was taken in @xmath28 seeing , with a long slit ( @xmath23 width ) aligned across the galaxies g1a , g1b , and g2 .
the spectral resolution with this setup is @xmath29 .
figure [ fig : spec1029 ] unambiguously shows that component c is a quasar with the same redshift , @xmath30 , as component b. the shapes of the emission and absorption lines are similar .
note in particular similar absorptions in the wings of the ly@xmath31 , , and lines .
these absorption features are probably intrinsic to the quasar and therefore serve as strong evidence that objects b and c are lensed images of the same quasar .
as we expect from the different broad band colors , image c has a redder continuum than image b. a possible explanation is differential extinction , which is relatively common in lensed quasars ( e.g. , * ? ? ?
* ; * ? ? ?
this requires that the two paths have @xmath32 assuming the extinction is caused by dust ( @xmath33 ) at the cluster redshift , @xmath0 .
we subtracted models for images b and c from the keck images and found no nearby galaxies beyond the red arc to the south of image c. another possibility is microlensing by stars in the lens , which can also modify the color of the continuum ( e.g. , * ? ? ?
* ; * ? ? ?
we note that both dust reddening and microlensing are observed in the other cluster - scale lensed quasar sdss j1004 + 4112 @xcite .
moreover , the hubble space telescope image of this system also shows no obvious candidate galaxy for the effects . in principle
we can distinguish between these possibilities by examining the differences in the flux ratios between the continuum and the emission lines , because dust will affect both equally while microlensing affects the compact continuum emission region more strongly than the larger broad line regions ( e.g. , * ? ? ?
the ratio of the spectra shown in figure [ fig : spec1029 ] shows no strong difference between continuum and emission line regions .
however , we note a possible small difference in the emission line which suggests that microlensing is affecting this system .
finally , the redshifts of g1a , g1b , and g2 are @xmath34 , @xmath35 , and @xmath36 , respectively , confirming the estimated cluster redshift of @xmath37 .
however , the velocity difference between g1b and g2 of @xmath38 @xmath39 is so large that the system may be an on - going merger or a superposition of two smaller clusters .
our identification of the third image has several important implications for the mass distribution of the lensing cluster .
there are four possible explanations for image c : ( i ) it is a so - called central core image @xcite of a `` double '' lens system ; ( ii ) it is a `` double '' lens system , but one of the images is split into a pair of images because of external perturbations ( e.g. , * ? ? ?
* ; * ? ? ?
* ) ; ( iii ) the lens system is a standard cusp - type `` quadruple '' lens , but the fourth ( counter ) image is too faint to be observed ; and ( iv ) the three images are produced by a naked cusp .
the main problem for ( i ) and ( ii ) is that the models require a large offset between the positions of the brightest cluster galaxies and the center of the lens potential , as discussed in @xcite .
in addition , the lensed arcs seem to contradict these scenarios as their shapes and locations suggest the lens potential is centered near g1 and g2 rather than somewhere between a and b. to test ( iii ) , we searched for any blue objects defined by @xmath40 , @xmath41 , and @xmath42 in the uh88 images .
we expect the color cut is conservative as it includes image c which is significantly reddened . except the three confirmed lensed images
, we found no such blue object , to the limiting magnitude of @xmath43 , within @xmath44 of image a. this translates into a constraint on the flux ratio of the putative fourth image d of @xmath45 , which is strong enough to make scenario ( iii ) unlikely .
we also searched for additional variable sources in the monitoring data and found none near the cluster , although with the present data we would not detect variability in a source significantly fainter than the known images .
the remaining scenario , ( iv ) , appears to be a viable solution .
a naked cusp is a rare configuration of caustics in which a tangential caustic extends outside a radial caustic . in the case
that a source is inside a naked cusp , three bright images are created on the same side of the potential center . for definiteness
, we fit the positions of the three images using the _ lensmodel _ package @xcite . we adopt an elliptical @xcite density profile centered on galaxy g1a with a fixed scale of @xmath44 .
we note that the scale radius is degenerate with the estimated mass @xcite .
we assume a positional error of @xmath46 to allow for perturbations from substructures ( e.g. , galaxies ) in the cluster .
this simple model , shown in figure [ fig : cusp ] , works surprisingly well , with @xmath47 for 1 degree of freedom .
the position angle of the model ( @xmath48 east of north ) is reasonable since it is aligned with the g1-g2 axis and the spatial distribution of color - selected cluster members , and puts the lensed arcs on the major axis of the lens potential as is preferred by ray - tracing in @xmath49-body simulations @xcite .
this model predicts time delays between the ab and bc images of @xmath50 days and @xmath51 days , respectively , assuming a cosmological model of @xmath52 and @xmath53 . a caveat for this model , as in many models of gravitational lenses ,
is the flux ratios of the quasar images . given the image geometry , we would generically expect @xmath54 , while we observe @xmath55 . specifically , our model predicts @xmath56 , whereas the observed flux ratios are @xmath57 in the uh88 @xmath58-band image .
the strong wavelength dependence of the flux ratio between images b and c , which is probably because of dust and/or microlensing , needs to be understood in order to fully evaluate this problem .
we also note that the a / b flux ratio in the 6-cm very large array ( vla ) radio map of @xmath59 @xcite is quite different from the ratio in the optical , @xmath60 . the spatial resolution and
signal - to - noise ratio of the radio map were not sufficient to derive robust radio fluxes for all the three components , and thus additional deep , high - resolution radio images will be essential for understanding the anomalous flux ratios .
in @xcite we were left with a puzzle about sdss j1029 + 2623 : the observed geometry seemed inconsistent with the observed cluster center .
our discovery here that object c is a third image of the quasar , despite its redder continuum , solves that problem because a three - image cusp lens configuration is easily produced under these circumstances .
the next problem is to understand the flux ratios of the system , both the wavelength dependences and that we observe @xmath61 rather than the expect @xmath54 .
measuring the radio flux ratios will let us separate the effects of substructures in the cluster from extinction and microlensing , and the differences between continuum and emission line , radio flux ratios , and time variability will allow us to separate the effects of extinction and microlensing .
sdss j1029 + 2623 is probably the second naked - cusp quasar lens . among the @xmath62 galaxy - scale lenses ,
only one candidate is known ( apm 08279 + 5255 ; * ? ? ? * ) .
naked cusps require `` marginal lenses '' , in which the lens has a surface density only moderately above the critical surface density required for the production of multiple images ( see * ? ? ?
the cooling of the baryons and the formation of the stellar component of a galaxy provides such high central surface densities that it is nearly impossible for a galaxy to have a naked cusp unless it is an edge - on disk - dominated system ( see * ? ? ? * ) .
group and cluster halos , where the baryons have not cooled yet , are far less efficient lenses ( e.g. , * ? ? ?
* ) and naked cusp configurations should be far more common . specifically , ( * ? ? ?
* see also @xcite ) predicted that @xmath63 of cluster - scale lensed quasars will be naked - cusp systems depending on the inner density profile .
we can not seriously measure these fractions at present , given one naked cusp lens and one conventional quadruple lens ( sdss j1004 + 4112 ; * ? ? ?
* ) , but the existence of even one naked - cusp cluster lens at this stage supports an important prediction of standard cold dark matter halos . we thank paul schechter for useful discussions , and an anonymous referee for helpful suggestions . this work was supported in part by department of energy contract de - ac02 - 76sf00515 .
csk is supported by nsf grand ast-0708082 .
the authors wish to 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 to have the opportunity to conduct observations from this mountain .
|
we identify a third image in the unique quasar lens sdss j1029 + 2623 , the second known quasar lens produced by a massive cluster of galaxies .
the spectrum of the third image shows similar emission and absorption features , but has a redder continuum than the other two images which can be explained by differential extinction or microlensing .
we also identify several lensed arcs .
our observations suggest a complicated structure of the lens cluster at @xmath0 .
we argue that the three lensed images are produced by a naked cusp on the basis of successful mass models , the distribution of cluster member galaxies , and the shapes and locations of the lensed arcs .
lensing by a naked cusp is quite rare among galaxy - scale lenses but is predicted to be common among large - separation lensed quasars . thus the discovery can be viewed as support for an important theoretical prediction of the standard cold dark matter model .
|
the work has been supported by the european community - access to research infrastructure action of the improving human potential programme , by the daad exchange programme ( ppp - polen ) , by the polish state committe for scientific research ( grants no .
2p03b07123 and pb1060/p03/2004/26 ) and by the research centre jlich .
s. brauksiepe et al . , _ nucl .
inst . & meth .
_ * a376 * , 397 - 410 ( 1996 ) .
p. moskal et al . , nucl - ex/0311003 .
balewski et al . , _ phys.lett .
_ * b420 * , 211 ( 1998 ) .
s. sewerin et al . , _ phys .
lett . _ * 83 * , 682 ( 1999 ) .
p. kowina et al . , e - print archive nucl - ex/0402008 , _ eur . phys
. j. _ * a * , in print .
a. baldini et al . , _ total cross - section for reactions of high - energy particles _ , ( springer , berlin , 1988 ) .
t. roek et al . , cosy proposal no .
117 ( 2002 ) .
a. gasparian et al . , _ phys .
lett . _ * b480 * , 273 ( 2000 ) .
a. gasparian et al . , _ nucl.phys . _
* a684 * , 397 - 399 ( 2001 ) .
r. shyam et al .
rev . _ * c63 * , 022202 ( 2001 ) .
r. shyam hep - ph/0406297 .
a. sibirtsev et al . , _ nucl.phys . _
* a646 * , 427 - 443 ( 1999 ) .
a. sibirtsev et al . , nucl - th/0004022 v2 ( 2000 ) .
|
the @xmath0 , @xmath1 and @xmath2 hyperon production in nn collisions is studied at the installation in order to investigate the production mechanism as well as to extract information about the y - n interaction .
cosy-11 is an internal magnetic spectrometer experiment at the cooler synchroton and storage ring cosy in jlich .
it is equipped with scintillator hodoscopes and drift chambers for charged particle detection @xcite and a scintillator / lead sandwich detector for neutrons @xcite .
the @xmath3 and @xmath0 hyperon production near the kinematical threshold was studied by the cosy-11 collaboration in @xmath4 reactions .
data points , 16 for the @xmath0 and 13 for the @xmath3 channel , were taken in the excess energy range between 0.68 mev and 59.3 mev for the @xmath0 hyperon and between 2.8 mev and 59.1 mev for @xmath3 @xcite .
the cross section ratio @xmath5 below excess energies of 15 mev was measured to be around 28 in contrast to the value of about 2.5 determined for excess energies higher than @xmath6 mev @xcite .
the ratio for higher energies is in good agreement with the @xmath7 isospin relation , which is 3 ( see figure [ all_models ] ) . to explain this unexpected threshold behaviour ,
various theoretical scenarios within meson exchange models were proposed .
calculations have been performed with pion and kaon exchange added coherently @xcite or incoherently @xcite , including the excitation of nucleon resonances @xcite and heavy meson exchange ( @xmath8 , @xmath9 and @xmath10 ) @xcite . although the various descriptions differ even in the dominant basic reaction mechanism , all more or less reproduce the trend of an increase in the cross section ratio in the threshold region .
the present data are not sufficient to definitely exclude possible explanations .
further studies e.g. for the other isospin projections will help to understand the threshold hyperon production . to be more specific lets consider the jlich meson exchange model@xcite where calculations are available for other @xmath11 channels . within this model
the large cross section ratio @xmath12 at threshold is reproduced by a destructive interference of @xmath13 and @xmath14 exchange amplitudes .
calculations of the @xmath15 production in this model predict a factor of three higher cross section compared to the @xmath3 channel for a destructive and a factor of three lower for a constructive interference , a clear experimentally accessible signal .
the ratio between @xmath15 and @xmath3 production will also differ strongly if the dominant production mechanism runs via an intermediate @xmath16 excitation or not .
+ recently the @xmath15 production was measured at the cosy-11 installation via @xmath17 at q = 13 mev and q = 60 mev @xcite .
the @xmath15 is identified via the missing mass technique , by detecting the remaining reaction products - @xmath18 and neutron .
the experimental resolution of the missing mass determination depends on the reconstruction accuracy of the four - momentum vectors for the registered neutrons and kaons .
the momentum vector of the @xmath18 meson can be established by tracking the @xmath18 trajectory reconstructed from signals registered in the drift chambers ( d1 and d2 in the figure [ cosy ] ) through the magnetic field back to the target point .
assuming a hit in the neutron detector being due to a neutron , the four - momentum vector of the neutron is given by the measured velocity , the direction ( given by the first hit module ) of the neutron which can be reconstructed with an accuracy of at least the size of the module ( 9x9 @xmath19 ) and the known mass .
the background from charged particles hitting the neutron detector is discriminated by veto scintillators ( veto in figure [ cosy ] ) .
+ from the monte carlo studies the identification of the @xmath15 events is expected to be comparable to the @xmath3 or @xmath0 channels@xcite .
the data are presently under analysis .
|
we have carried out extensive simulations of a simple lattice model that can be simulated efficiently .
the sites form either a one - dimensional , of length @xmath11 , or two - dimensional @xmath12 square array with periodic boundary conditions . as it is the spreading dynamics at long times that we are interested in , we assume that the local sweeps in a deme are fast compared to migration
. we can then ignore the logistic growth process within demes , so that when jumps occur and establish a new mutant population it is saturated in the new deme by the next time step
. therefore , it is convenient to lump together the probability of an individual to jump , the density of the population from which the jumps occur , and the probability ( proportional to @xmath4 ) that the mutant establishes a new population : we define @xmath13 so that @xmath14 is the rate at which a saturated mutant population near @xmath15 nucleates a mutant population near @xmath16 . in each computational timestep , we pick a source and target site randomly such that their distance @xmath2 is sampled from the ( discretized ) jump distribution @xmath17 with the @xmath18 a lattice site .
if the source site is a mutant and the target site a wild type , the identity of the target site is updated to mutant .
we measure time in units of @xmath19 timesteps .
see si sec .
s1 for more details on the simulation algorithm .
the growth of mutant populations generated by our simulations is best visualized in a space - time portrait .
[ fig : snapshots]a and fig .
[ fig : snapshots]b show the overlaid space - time plots of multiple runs in the regimes @xmath20 and @xmath21 , respectively . fig .
[ fig : v1-spreading ] shows the growth dynamics of the mutant population over large time and length scales for various values of @xmath7 . for @xmath22 ,
the dynamics clearly approaches a power law . for @xmath23 ,
the simulations are consistent with stretched exponentials .
the intermediate regime @xmath24 is elusive , as one can not extract a clear asymptotic behavior on the time scales feasible in simulations .
the behavior in two - dimensions is qualitatively similar , as shown in the si fig .
are @xmath25 in order of increasing darkness .
b ) regime of very fast growth with @xmath7 near 1 .
the values of @xmath7 are @xmath26 in order of increasing darkness .
[ fig : snapshots],title="fig : " ] are @xmath25 in order of increasing darkness .
b ) regime of very fast growth with @xmath7 near 1 .
the values of @xmath7 are @xmath26 in order of increasing darkness .
[ fig : snapshots],title="fig : " ] + to explain these dynamics in detail , we develop an analytical theory that is able to predict not only the asymptotic growth dynamics but also the crucial transients .
traditionally , analyses of spreading dynamics start with a deterministic approximation of the selective and dispersal dynamics ignoring both stochasticity and the discreteness of individuals .
to setup consideration of the actual stochastic dynamics , we first give results in this deterministic approximation and show that these exhibit hints of why they break down .
when the jump rate decreases exponentially or faster with distance , the spread is _ qualitatively _ similar to simple diffusive dispersal and the extent of the mutant population expands linearly in time .
however when the scale of the exponential fall - off is long , the speed , @xmath27 , is faster than the classic result for local dispersal , @xmath28 , which only depends on the diffusion coefficient , @xmath29 : instead , there is a characteristic range of jump distances determined by @xmath4 and @xmath3 that dominate the spread .
as the dispersal range gets longer for fixed @xmath30 , the dominant jumps occur at lower rates suggesting that the stochasticity may become more important . if @xmath3 has a long tail , in particular , @xmath6 , the spread in the deterministic approximation becomes exponentially fast with the radius of the region taken over by the mutant population , @xmath31 .
this is almost as fast as in a fully mixed population , with the growth rate of the total mutant population size slower only by a factor of @xmath32 which approaches unity as @xmath33 , the point at which the spatial structure becomes irrelevant .
the origin of this exponential growth for power - law @xmath3 is the deterministic feeding of the populations far away by jumps from near the origin : this immediately produces a finite population density at all distances , @xmath2 . after a time of order
@xmath34 has passed , the exponential growth of the local population takes off , that is , further jumps to that region no longer matter . as in this time
@xmath34 the expected total number of jumps to the _ whole _ region further than @xmath2 from the origin is only of order @xmath35 , the probability that _ any _ jumps have occurred is very small for large @xmath2 and the deterministic approximation must fail @xcite . with local dispersal , the deterministic approximation is a good starting point with only modest corrections to the expansion speed at high population density , the most significant effect of stochasticity being fluctuations in the speed of the front @xcite . at the opposite extreme of jump rate independent of distance ,
the deterministic approximation is also good with the mutant population growing as @xmath36 and fluctuations only causing stochastic variability and a systematic reduction in the coefficient which arise from early times when the population is small .
surprisingly , in the regimes intermediate between these two , the deterministic approximation is not even qualitatively reasonable .
we assume that , at long times , most of the sites are filled out to some distance scale @xmath8 and that the density decreases sufficiently steeply for larger distances , such that the total mutant population , @xmath37 , is proportional to @xmath38 .
the validity of the this assumption follows from more accurate analyses given in the appendix .
we call the crossover scale @xmath8 the _ core radius _ or size " of the mutant population . in the dynamical regimes of interest , the core population grows primarily because it `` absorbs '' satellite clusters , which themselves were seeded by jumps from the core population .
we now show that the rate of seeding of new mutant satellite clusters and the growth of the core populations by mergers with previously seeded clusters has to satisfy an iterative condition that enables us to determine the typical spreading dynamics of the mutant population . to be occupied at time @xmath39 , a seed typically must become established somewhere in the blue space - time region ( `` target funnel '' ) by means of a long - range jump ( black arrow ) from the red source " region .
this schematic leads to the iterative scaling approximation in eq . .
note that the concavity of the source - funnel geometry , leading to a gap between red and blue regions , is key to our arguments and enables neglecting effects of jumps into the gap region .
[ fig : funnel - antifunnel ] ] it is convenient to illustrate our argument using a space - time diagram , fig .
[ fig : funnel - antifunnel ] , in which the growth of the core has the shape of a funnel .
now consider the edge of this funnel at time @xmath40 ( gray circle in the figure ) .
the only way that this edge can become populated is by becoming part of a population sub - cluster seeded by an appropriate long - range jump at an earlier time . to this end
, the seed of this sub - cluster must have been established somewhere in the inverted blue funnel in fig .
[ fig : funnel - antifunnel ] .
this `` target '' funnel has the same shape as the space - time portrait of the growing total population , but its stem is placed at @xmath41 and the mouth opens backward in time .
notice that if @xmath8 grows faster than linearly , space - time plots of the growing cluster and the funnel have concave boundaries : this necessitates a jump from the source to the funnel of length much longer than @xmath42 , as shown .
in @xmath0 dimensions , the consistency of growth and seeding thus requires @xmath43 where @xmath44 denotes a @xmath0 dimensional ball of radius @xmath45 centered at the origin , and we have taken the point of interest to be @xmath46 with @xmath47 a unit vector in an arbitrary direction . the kernel @xmath17 represents the rate per @xmath48dimensional volume of ( established ) jumps of size @xmath2 . we now show that the asymptotic growth dynamics is essentially constrained by the above iterative scaling argument . specifically , although the argument is more general , we consider a power law jump distribution @xmath49 for large enough @xmath2 . for the _ intermediate - range case
_ @xmath50 , equations and exhibit the asymptotic scaling solution @xmath51 the _ form _ of which could have been guessed by dimensional analysis . inserting
the ansatz into equation determines the pre - factor @xmath52 in this iterative scaling approximation , up to an order - unity coefficient : see details in the appendix .
interestingly , the value @xmath52 depends very sensitively on @xmath7 and runs from 0 to @xmath53 as @xmath7 passes through the interval from @xmath0 to @xmath54 . this can be seen in in si fig .
s2 , where @xmath52 is plotted as a function of @xmath7 for @xmath55 : it drops very steeply , as @xmath56 , for @xmath57 , and diverges as @xmath58 for @xmath59 . as we discuss below
, these singularities are a manifestation of intermediate asymptotic regimes that lead to very slow convergence to the asymptotic behavior .
we now turn to the ( very ) _ long - range case _
, @xmath60 , for which a direct solution to can not be found ( and the dimensional analysis argument gives nonsense ) .
however , much can be learned by approximating using @xmath61\approx g[\ell(t)]$ ] , anticipating the very rapid growth and thus likely smallness of @xmath62 and @xmath63 compared to @xmath64 : @xmath65 \int_0^{t}dt \ell^d(t)\ell^d(t - t ) \approx 1 \;,\ ] ] ( ignoring two factors from the angular integrations ) . with @xmath8 growing sub - exponentially
, the largest contributions will come from @xmath66 .
once we have found the form of @xmath8 , the validity of the ansatz can be tested by checking whether the @xmath64 is much larger than @xmath42 .
indeed , for @xmath67 the solution to is a rapidly growing stretched exponential , [ eq : explaws ] ( t)~(b_t^ ) with = , which can be checked by direct insertion into equation .
notice that as @xmath68 , @xmath69 and @xmath8 grows exponentially as for a flat distribution of long - range jumps that extends out to the size of the system : i.e. the globally - mixed limit . in the opposite limit of @xmath70 small , @xmath71 and
the coefficient @xmath72 diverges as shown below .
we note that the asymptotic stretched - exponential growth for @xmath67 also arises in models of chemical distance " and certain types of spatial spread for network models with a similar power - law distribution of long distance connections : however the pre - factors in the exponent are different @xcite .
we discuss the connections between these in sec .
[ sec : heterogen ] . for the _ marginal case _
, @xmath73 , the asymptotic behavior is similarly found to be : @xmath74 \;,\end{aligned}\ ] ] we now show that this behavior also represents an important _ intermediate _ asymptotic regime which dominates the dynamics over a wide range of times for @xmath7 close to @xmath0 : this is the source of the singular behavior of the coefficients @xmath52 and @xmath72 in this regime .
our source - funnel argument obviously neglects jumps that originate from the not - fully - filled regions outside the core radius @xmath8 .
an improved version of the funnel argument is presented in the si sec .
s2 c , which also allows us to estimate the probability of occupancy outside the core region .
further , we present in the appendix outlines of rigorous proofs of lower and upper bounds for the asymptotic growth laws in one dimension , including the slow crossovers near the marginal case . the linear growth for @xmath75 in one dimension has been proven by mollison @xcite .
after the present paper was essentially complete , we became aware of a recent preprint by chatterjee and dey @xcite who obtained rigorous bounds in all dimensions for the leading asymptotic behaviors in the three super - linear regimes .
our bounds are somewhat tighter than theirs , including the coefficient and leading corrections to the asymptotic behavior in the marginal case , the absence of logarithmic pre factors in the power - law regime , and the full crossovers for @xmath7 near @xmath0 in one dimension : comparisons are discussed in the appendix .
asymptotic laws are of limited value without some understanding of their regime of validity , especially if the approach to the asymptotic behavior is slow . and such knowledge is crucially needed to interpret and make use of results from simulations .
we first consider the short time behavior when the total rate of long jumps is small : i.e. , when @xmath76 , the coefficient of @xmath77 is small . short jumps result in diffusive motion and linear growth @xmath78 with @xmath79 determined by the details of the selective and diffusive dynamics .
long jumps start to become important after enough time has elapsed that there have been at least some jumps of lengths of order @xmath8 : i.e. when @xmath80 : this occurs after a cross - over time @xmath81^{1/(d+1-\mu)}$ ] at which point @xmath82^{1/(d+1-\mu)}$ ] . at longer times and distances
, one can measure lengths and times in units of these crossover scales , defining @xmath83 and time @xmath84 and expect that the behavior in these units will not depend on the underlying parameters .
[ note that this separation in short - time linear growth and long - time regimes can also be done for more general @xmath17 although then the behavior will depend on the whole function the crossover on distances of order @xmath85 , and the super - linear behavior on the longer distance form . ] at times much longer than @xmath86 , there is a slow crossover close to the boundary between the stretched exponential and power law regimes .
thus we must take a closer look at the dynamics in the vicinity of the marginal case , @xmath73 .
as @xmath87 , the integrand in equation develops a sharp peak at @xmath88 : half - way between the bounds of the integral .
laplace s method can then be used to approximate the integral leading to a simplified recurrence relation : @xmath89 this is really only good to a numerical factor of order unity which can be eliminated by rescaling @xmath90 , and to a larger logarithmic factor associated with the narrowness of the range of integration and its dependence on @xmath90 and @xmath91 : this we analyze in si sec .
s2 a. both of these are negligible if we focus on the behavior on logarithmic scales in space and time natural given their relationships .
defining @xmath92 and @xmath93 , and taking the binary logarithm , @xmath94 , of equation , yields a linear recurrence relation that can be solved exactly .
rescaling @xmath95 can be used to make @xmath96 whence @xmath97 where we introduced the variable @xmath98 , which measures the distance to the marginal case .
the asymptotic scaling for @xmath99 reproduces the earlier predicted power law regime , , and yields the pre - factor @xmath100 , up to correction that is subdominant for small @xmath101 ( c.f .
s3 ) . for @xmath102 ,
the asymptotics yields the stretched exponential in , and fixes the pre - factor @xmath103 , which could not be obtained from the basic asymptotic analysis carried out above . the singular pre - factors for @xmath104 give warnings of breakdown of the asymptotic results except at very long times .
this peculiar behavior is the consequence of an intermediate asymptotic regime that dominates the dynamics close to the marginal case .
this leads to slow convergence to the eventual asymptotic behavior for @xmath7 near @xmath0 .
the asymptotic scaling can be observed only on times and length such that @xmath105 on smaller times , the dynamics is similar to the marginal case , .
the rapid divergence of the _ logarithm _ of the time after which the asymptotic results obtain , make it nearly impossible to clearly observe the asymptotic limits : in one - dimensional simulations this problem occurs when @xmath106 , as us clearly visible in fig .
[ fig : v1-spreading ] and it is likely even harder to observe in natural systems .
this underscores the need for the much fuller analysis of the spreading dynamics as via .
while the dynamics at moderate times will be dominated by the initial growth characteristic of the marginal case , we expect to be a good description of the universal dynamics at large @xmath107 even when @xmath101 is small .
the limit @xmath108 while @xmath109 fixed is particularly interesting , as the solution then reduces to a scaling form @xmath110 with ( ) = ( -)+-1 .
this scaling form allows us to test by simulations our analytic results across all intermediate asymptotic regimes , by plotting data obtained for different @xmath101 in one scaling plot , see fig .
[ fig : scaling - plot - a ] . to make the approximation uniformly valid in both the scaling regime and at asymptotically long times outside of it , we can simply replace , for @xmath102 , the scaling variable by @xmath111 ( with @xmath112 defined in ): the scaling form
will then be valid up to corrections that are small compared to the ones given in all regimes : we thus use this form for the scaling fits in the inset of fig .
[ fig : scaling - plot - a ] , plotting data obtained for different @xmath101 in one scaling plot , thereby testing our solution across all intermediate asymptotic regimes . are predicted to collapse on a scaling plot close to the marginal case @xmath73 between the stretched - exponential and power - law growth regimes , here demonstrated for the one - dimensional case ( @xmath55 ) .
the main plot shows the rescaled sizes @xmath113 of the mutant population versus rescaled log - time @xmath114 .
the differently colored data sets correspond to @xmath115 realizations with power law exponents @xmath116 ( same colors as in fig .
[ fig : v1-spreading]a ) .
the simulated data collapse reasonably well with the dashed red line , representing the predicted scaling function @xmath117 .
the inset depicts the same data scaled slightly differently away from the scaling regime such that the horizontal axis for @xmath102 shows @xmath118 ( with @xmath112 defined in ) , upon which the data collapse improves .
note that @xmath119 as @xmath104 so that the stretched exponential form is recovered for @xmath120 large and negative .
[ fig : scaling - plot - a ] ] thus far we have considered spatially uniform systems , in which the jump probability between two points only depends on their separation
. however , long distance transport processes may be very heterogeneous .
an extreme example is airplane travel which occurs on a network of links between airports with mixtures of short and long distance flights , plus local transportation to and from airports .
a simple model is to consider each site to have a number of connections from it , with the probability of a connection between each pair of sites a distance @xmath2 away being @xmath121 , independently for each pair : note that although the network is heterogeneous , statistically , the system is still homogeneous as the connection probability does not depend on position . if the rate at which jumps occur across a connection of length @xmath2 is @xmath122 , then , averaged over all pairs of sites , the rate of jumps of distance @xmath2 is @xmath123 .
how similar is this to a homogeneous model with the same @xmath3 , in particular if @xmath6 ? if there are a large number of possible connections along which the key jumps can occur to get from a source region of size @xmath124 to a funnel of similar size a distance @xmath64 away , then the fact that these only occur to and from a small fraction of the sites should not matter for the large length scale behavior .
the number of such connections is @xmath125 with , making the ansatz that , as in the the homogeneous case , @xmath126 ( ignoring subdominant factors ) one has @xmath127 $ ] .
thus the condition for our results to be valid asymptotically is that @xmath128 with @xmath129 the inverse of the function @xmath8 : i.e. , in the exponential , marginal , and power - law cases , respectively , that @xmath130 , @xmath131 , and @xmath132
. if there are insufficient number of connections for the heterogeneity of the network to be effectively averaged over , the behavior changes .
the extreme situation is when there is a distance - independent rate for jumps along the longest connection out of a site : i.e. @xmath133 . in this case , jumps along the path with the shortest number of steps , @xmath134 , to get from the origin to a point @xmath135 will reach that point in a time proportional to @xmath134 i.e. @xmath136 . the geometrical problem of obtaining the statistics of @xmath137 has been analyzed by biskup @xcite . for @xmath138 , @xmath139 and long jumps do not matter , while for @xmath67 , @xmath140 with the same exponent @xmath112 as in the homogeneous case we have analyzed .
the difference between this result and ours is only in the power - law - of-@xmath40 pre - factor of @xmath64 arising from the integral over time : this does not exist in the extreme network limit . in the marginal case , @xmath73 , @xmath141 with @xmath142
dependent on the coefficient of the power - law decay of the connection probability .
if the probability of a jump along a long distance connection decays with distance but more slowly than @xmath143 , the behavior is similar : for @xmath67 again the ubiquitous stretched exponential behavior occurs , while for @xmath138 there are too few connections only if @xmath144 in which case the number of steps and the time are both proportional to the distance .
the marginal cases we have not analyzed further . for natural transport processes , the probabilities of long dispersal events will depend on both the source and the destination . if the heterogeneities are weak on large length scales , our results still obtain .
but if there are sufficiently strong large scale heterogeneities , either in a spatial continuum or in the network structure ( i.e. location of nodes and links and the jump rates along these , or hub - spoke structure with multiple links from a small subset of sites ) , then the spatial spread will be heterogeneous even on large scales : how this reflects the underlying heterogeneities of the dispersal has to be analyzed on a case by case basis .
we have studied the impact of long - range jumps on evolutionary spreading using the example of mutants that carry a favorable genetic variant .
to this end , we analyzed a simple model in which long - range jumps lead to the continual seeding of new clusters of mutants , which themselves grow and send out more migrant mutants .
the ultimate merging of these satellite clusters limits the overall growth of the mutant population , and it is a balance of seeding and merging of sub - clusters that controls the spreading behavior . to classify the phenomena emerging from this model , we focussed on jump distributions that exhibit a power law tail .
we found that , with power law jumps , four generic behaviors are possible on sufficiently long times : the effective radius of the mutant population grows either at constant speed , as a super - linear power law of time , as a stretched exponential , or simply exponentially depending on the exponent , @xmath145 , of the power law decay of the jump probability .
these predictions are in contrast to deterministic approximations that predict exponential growth for power - law decaying jump kernels @xcite . in dimensions more than one ,
the results also contradict the naive expectation from dynamics of neutral dispersal , that a finite diffusion coefficient is sufficient for conventional behavior ( in this context , finite speed of spreading ) : specifically , for @xmath146 , @xmath147 but the spread is super - linear indeed stretched exponential for @xmath148 . that super - linear dynamics can occur for @xmath149 is not surprising as even a migrating individual undergoes a levy flight
: more surprising is that this occurs even when the dynamics of individuals is , on large scales , like a normal random walk .
the breakdowns of both deterministic and diffusive expectations are indicative of the importance of fluctuations : the dynamics is dominated by very rare but not too rare jumps : roughly , the most unlikely that occur at all up to that time . one of the consequences of this control by the rare jumps is the relatively minor role played by the selective fitness advantage , @xmath4 , of the mutants . with short - range dispersal ,
the speed is @xmath150 , and for global dispersal , @xmath151 .
by contrast , in the regime of power law growth , the asymptotic growth of the mutant population becomes ( to leading order ) independent of the speed of the growth of individual clusters although when individual clusters grow more slowly , the asymptotic regime is reached at a later time . in the stretched exponential regime , the growth of sub - clusters sets the cross - over time from linear to stretched exponential , and thus determines the pre - factor in the power law that characterizes the logarithm of the mutant population size .
an important feature of our results is that the approach to the asymptotic laws is very slow in the vicinity of the marginal cases , as illustrated in fig .
[ fig : phase - diagram ] . consider for instance the two - dimensional case , where we have asymptotically stretched exponential growth for @xmath149 : for @xmath152 ( @xmath153 ) , the epidemics has to run for times @xmath154 @xmath155 to reach the asymptotic regime . by that time , however , the mutant population with @xmath156 @xmath157 would have certainly spread over the surface of the earth . to describe the full cross - over dynamics
, we developed an approximate recurrence relation based on the geometric source - funnel argument described in the illustration fig .
[ fig : funnel - antifunnel ] .
this predicts a universal cross - over function for the transient dynamics near @xmath73 that can be uncovered by simulations for different power law exponents collapsed onto one scaling plot , c.f .
[ fig : scaling - plot - a ] . the good agreement of the predicted scaling function and simulation results provides strong support for the iterative scaling approximation .
the rigorous bounds whose proofs are outlined in the si sec .
s3 , provides further support .
understanding this crossover is essential for making sense of , and extrapolating from , simulations as asymptotic behavior is not visible until enormous system sizes even when the exponent @xmath7 is more than @xmath158 from its marginal value .
another benefit of the ability of the simple iterative scaling argument to capture well non - asymptotic behavior , is that it can be used in cases in which the dispersal spectrum of jumps is not a simple power - law , e.g. with a crossover from one form to another as a function of distance . and the heuristic picture that it gives rise to an exponential hierarchy of time scales separated by roughly factors of two
is suggestive even in more complicated situations . that such a structure should emerge without a hierarchal structure of the underlying space or dynamics
is perhaps surprising .
, the spreading dynamics exhibits three behaviors . on
asymptotically large times , either stretched exponential growth for @xmath159 or superlinear power law growth for @xmath160 occurs .
however , the approach to the asymptotic regime is extremely slow for @xmath7 close to @xmath0 . for log - times @xmath161 ,
the behavior is controlled by the dynamics of the marginal case .
note that cross - over behavior also obtains near to the borderline between linear and super - linear behavior at @xmath162 , cf .
s2 b. ] our primary aim biologically is the qualitative and semi - quantitative understanding that emerges from consideration of the simple models and analyses of these , especially demonstrating how rapid spatial spread of beneficial mutations or other biological novelty can be even with very limited long - range dispersal .
as the models do not depend on any detailed information about the biology or dispersal mechanisms , they can be considered as a basic null model for spreading dynamics in _ physical _ , rather than more abstract network , space .
the empirical literature suggests that power - law spectra of spatial dispersal are wide - spread in the biological world @xcite . although these are surely neither a constant power - law over a wide range of scales , nor spatially homogeneous , our detailed results are not directly applicable .
but , as discussed above , our iterative scaling argument is more general and can be applied with more complicated distance dependence , anisotropy , etc .
furthermore , some of the heterogeneities of the dispersal will be averaged out for the overall spread , while affecting when mutants are likely to arrive at particular locations . for dispersal via hitchhiking on human transport , either of pathogens or of commensals such as fruit flies with food , the apparent heterogeneities are large because of the nature of transportation networks , although data suggest that dispersal of humans can be reasonably approximated by power - law jumps @xcite .
whether or not transport via a network with hubs at many scales fundamentally changes the dynamics of an expanding population of hitchhikers from that with more homogeneous jump processes , depends on the nature of how the population expands . for spread of a human epidemic
, there are several possible scenarios .
if the human population is reasonably uniform spatially , and the chances that a person travels from , say , their home to another person s home is primarily a function of the distance between these rather than the specific locations , then whether or not the properties of the transportation network matter depends on features of the disease .
if individuals are infectious for the whole time the outbreak lasts and if transmission is primarily at end points of journeys rather than enroute for example hiv then the transportation network plays no role except to provide the spatial jumps . at the other extreme but still within an si model , is if individuals living near hubs are more likely to travel ( or even if destinations near hubs are more likely ) , and , more so , if infections are likely to occur enroute , in which case the structure of the transportation network as well as of spectra of city sizes , etc matters a great deal . in between these limits the network ( or lack of it in places ) may matter for initial local spread but at longer times the network structure may effectively average out and the dynamics be more like the homogeneous models .
the two opposite limits and behavior in between these , together with the specific network model we analyzed with jumping probabilities depending on distance even in the presence of a connection as is true from airports all illustrate an important point : geometrical properties of networks alone rarely determine their properties : quantitative aspects , such as probabilities of moving along links and what exists at the nodes , are crucial .
more complicated epidemic models can be discussed within the same framework : the model discussed thus far corresponds to an si model , the most basic epidemic model , which consists of susceptible and infected individuals only .
many important epidemics are characterized by rather short infectious periods , so that one has to take into account the transition from infected to recovered : sir models .
this changes fundamentally the geometry of our space time analysis , illustrated in fig .
[ fig : funnel - antifunnel ] . while the target funnel
remains a full funnel , the source funnel becomes hollow : the center of the population consists mostly of fully recovered individuals , whose long - range jumps are irrelevant .
the relevant source population of infected individuals is primarily near the boundaries of the funnel .
this leads to a break in the time - symmetry of the argument . as a result
, the spreading crosses over from the behavior described above to genuine sir behavior .
the sir dynamics is closely related to the scaling of graph - distance in networks with power - law distributions of link lengths @xcite , as recently shown by one- and two - dimensional simulations @xcite .
in particular , the limited time of infectiousness causes wave - like spreading at a constant speed for @xmath138 .
but , importantly , the spreading velocity is controlled by the si dynamics we have studied until a time of order the infectious period .
analogous crossovers from si to sir occur also with longer range jumps .
more generally , other complications can be discussed in our framework , and we expect new behaviors depending on how they modify the geometry of the source to target - funnel picture . as a last note ,
our analyses naturally provide information on the typical structure of infection ( or coalescent ) trees . for a given site
, the path of jumps by which the site was colonized can be plotted as in fig .
[ fig : infection_trees ] . doing this for many sites yields coalescent ( or infection ) trees , which can reveal the key long - range jumps shared by many lineages . for actual epidemics , combining spatio - temporal sampling of rapidly evolving pathogens with whole genome sequencing is now making it possible to construct such infection trees . for inference purposes , it would therefore be interesting to analyze more about the statistical properties of such infection trees and how they depend on the dispersal properties , network structure , and other features of epidemic models .
this work was partially supported by the dfg via ha 5163/2 - 1 , nsf via dms-1120699 and phy-1305433 .
38 natexlab#1#1bibnamefont # 1#1bibfnamefont # 1#1citenamefont # 1#1url # 1`#1`urlprefix[2]#2 [ 2][]#2 , , , and , , * * ( ) , .
, , . , , , . , and , , * * ( ) , . , and , , * * ( ) , . , , and , , * * ( ) , . , and , , * * ( ) , . , , and , , * * ( ) , . , , * * ( ) , . , , and , , * * ( ) , . , and , , . , , , and , , * * ( ) , . , and , , , . , , , , and , , * * ( ) , . , , , and , , * * ( ) , . , and , , * * ( ) , . , , * * ( ) , . , , and , , * * ( ) , . , , and , , * * ( ) , . , , * *
( ) , . , , * * ( ) , . , , and , , ,
. , , and , , * * , . , , and , , * * ( ) , . , , and , , * * . , , and , , * * , . , , and , , * * ( ) , . , , * * ( ) ,
. , , and , , * * ( ) , .
, , , . , _ et al . _ , , in _ _ ( ) . , and , , * * ( ) , . , , , and , , * * ( ) ,
. , , , , , and , , * * ( ) , . , , , , , and , , * * ( ) , . , , , , , , and , , * * ( ) , . , , and , , * * ( ) , . , , * * ( ) , .
the state of the population is described by a linear array of @xmath163 sites with periodic boundary conditions .
@xmath163 is chosen large enough so that end - effects can be ignored ( typically between @xmath164 and @xmath165 sites ) .
each site has the identity of either mutant or wildtype .
initially , the whole population is wildtype except for the central site , which is occupied by mutants . in each computational timestep ,
a source site @xmath166 and target site @xmath167 are chosen randomly such that their distance @xmath2 is sampled from a probability density function with a tail @xmath168 at large @xmath2 ( see below ) .
if @xmath166 is mutant and @xmath167 a wild type then @xmath167 turns into mutants and seeds a new mutant cluster .
we use the convention that @xmath163 time - steps i.e. an average of one jump attempt per site comprise one unit of time , or effective generation " .
the rate of long - range jumps should be thought of as representing the product of the probability to establish a new cluster per jump and the jump rate per generation per site . in our simulations ,
the distance @xmath169 of a long - range jump was generated as follows .
first , draw a random number @xmath170 within @xmath171 , and calculate the variable @xmath172^{-1/\mu}\;.\ ] ] where @xmath173 is a cutoff ( see below ) and @xmath11 is the system size .
this generates a continuous pdf @xmath174 with @xmath62 values in @xmath175 .
the actual jump distance is obtained from @xmath169 by rounding down to the next integer .
( note that , because the distribution has a tail @xmath168 , we have to choose @xmath176 in equation ( 2 ) of the main text . ) for the one dimensional data in all graphs of this paper , we used @xmath177 and system sizes ranging from @xmath178 for @xmath179 to @xmath180 for @xmath181 . for such large systems , the tail of the distribution
is well approximated by @xmath182 , as stated in the main text .
we also tested variations in the cutoff @xmath173 .
using @xmath183 or @xmath184 only affected the short - time dynamics and had very little influence on the intermediate asymptotic or long distance behavior of the system . for our two - dimensional simulations , we draw jump sizes from the same distribution as the one described above .
we set the lower cutoff to @xmath185 to make sure that jumps reach out of the source lattice point .
after the jump size is drawn , the jump direction is chosen at random .
in analyzing the results from the iterative scaling approximation we have ignored the effects of the range of @xmath39 around @xmath186 that dominates the probabilities of occupation at time @xmath40 .
this is valid for obtaining the leading behaviors of @xmath187 in the large time limit , but there are corrections to these that can be larger than those that arise from the short - time small - length - scale crossovers that we discussed in the text and do so further below . for @xmath7
not much smaller than @xmath54 , when the growth of @xmath45 is a modest power of time , the factor from the range of the time integral is only of order unity and hence no worse than other factors including from the stochasticity that we have neglected .
but when @xmath8 grows very rapidly , the range of @xmath188 that dominates is much smaller than @xmath40 and the corrections are larger . with rapid growth of @xmath8 , a saddle point approximation to the time integral
is valid : @xmath189 with the pre - factor given by @xmath190 which , with the second derivative absorbing the @xmath191 factor , the derivative part can be rewritten as @xmath192 . with the asymptotic growth laws we have derived ,
this gives @xmath193 for @xmath91 small and positive , @xmath194 for @xmath73 , and @xmath195 for @xmath67 .
integrating up the effects of this over the scales yields the following corrections in the various regimes : for @xmath138 , the coefficient , @xmath52 , of @xmath196 is changed by a multiplicative factor which is much less singular for @xmath197 than that already obtained . for @xmath73 , c[^2 t - t t + o(t ) ] with @xmath198 , the second term being new and the smaller correction term including the effects of the small time cross - over . for @xmath60 , b_t^- t with the second term for @xmath199
just that which occurs in the crossover regime analyzed in sec .
crossovers and beyond asymptopia ( main text ) , where we showed that the coefficient @xmath72 diverges proportional to @xmath200 . in the main text
, we mainly focussed on regimes in which the mutant growth is very much faster than linear , i.e. @xmath201 .
this allowed us to approximate the integrals in the iterative scaling approximation of equation ( 1 ) ( main text ) by the use of laplace s method .
this saddle point approximation yields the correct scaling for all exponents @xmath67 , but ( as we see below ) incorrect pre - factors in regimes where the actual growth is close to linear , i.e. in the power law growth regime with @xmath202 . to obtain a better estimate of the pre - factors in this power law regime ,
it is helpful to directly solve for the asymptotics of the iterative scaling argument in equation ( 1 ) ( main text ) .
here , we demonstrate how this can be done for @xmath55 : assume that most of the weight in the integral comes from regions , where the jump kernel is well approximated by its power - law tail described in equation ( 2 ) ( main text ) .
given @xmath203 ( for @xmath55 ) , this always holds at sufficiently long times .
then , one has @xmath204 where @xmath205 for @xmath206 , equation exhibits an asymptotic power law solution @xmath207 by inserting this ansatz into eq.s and , we obtain the following result for the numerical pre - factor @xmath208 with @xmath209 being equal to eq .
( [ eq : oskar2 ] ) with @xmath8 replaced by @xmath210 .
the resulting pre - factor is plotted as a function of @xmath211 in supplementary fig .
[ fig : prefactors ] .
notice that @xmath52 strongly depends on the exponent @xmath7 .
it sharply drops for @xmath7 approaching @xmath212 , where it follows the asymptotics @xmath213 . on the other hand , as @xmath7 approaches the other marginal case at @xmath214 , the pre - factor diverges as @xmath215 .
indicate the importance of intermediate asymptotic regimes , as discussed in the main text sec .
crossovers and beyond asymptopia ( main text ) .
while we have focussed on the marginal case near @xmath73 in this article , it is clear that another case of marginal stability controls the cross - overs near @xmath162 .
simulation results reported in supplementary fig . [ fig : second - marginal - point ] indicate that @xmath216 for @xmath214 in one dimension .
this is consistent with our funnel argument : with nearly constant speed , the gap between the funnels remains roughly constant for a time of order @xmath39 . to ensure that the source emits about one jump to the target funnel
, we must have that , per unit time , the probability of jumping over the gap is of order @xmath217 .
thus , the gap size @xmath218 should be such that @xmath219 . for @xmath214 , we thus have @xmath220 , i.e. , the key jumps span distances of order @xmath39 .
this is ensured when @xmath221 const .
, i.e. , if @xmath222 .
note that a rigorous upper bound of this form follows from the arguments presented in sec .
[ sec : intermediate - range - case ] for the regime @xmath138 .
the jumps of order @xmath223 that drive the logarithmic increase in spreading velocity might be the `` leaps forward '' @xcite recognized by mollison in one of the earliest studies on spreading with long - range jumps . in the main text
, we introduced the notion of a nearly occupied core of the population ( of size @xmath8 ) as the source of most of the relevant seeds in the target funnel .
however , it is clear that outside of this core there is a region of partial occupancy .
this region is potentially broad , in particular for @xmath33 , and may therefore lead to a significant fraction of relevant seeds .
an improved theory should account for those secondary seeds , and should also be able to determine the profile of mean occupancy or , equivalently , the probability that a site is occupied . while we give rigorous bounds on how the total population grows in sec .
[ sec : rigorous ] , we first use an improved version of our funnel argument to describe the occupancy profiles .
we focus on the probability , @xmath224 , that a time @xmath39 after a mutant establishes , it will have taken over the population a distance @xmath2 away .
we expect that @xmath224 will be close to unity out to some core radius @xmath8 , and then decrease for larger @xmath2 , with the average total mutant population proportional to @xmath225 . with only short - range dispersal , @xmath226 and the core
is clearly delineated but when long jumps are important the crossover from mostly occupied core to sparsely occupied halo will not be sharp .
the more important quantity is the average of the total area ( in two - dimensions or linear extent or volume in one or three ) occupied by the mutant population , we denote this @xmath227 . to find out when long jumps _
be important , we first ask whether there are likely to be _ any _ jumps longer than @xmath8 that occur up to time @xmath39 .
the average number of such long jumps is of order @xmath228 .
if @xmath17 decreases more rapidly than @xmath229 , this is much less than @xmath230 for large @xmath39 . as @xmath8 increases at least linearly in time , the probability that there have been any jumps longer than @xmath8 is very small .
the guess that @xmath45 indeed grows as @xmath231 , and consideration of jumps that could advance the front fast enough to contribute substantially to @xmath27 , leads , similarly , to the conclusion that there is a maximum @xmath39-independent jump length beyond which the effects of jumps are negligible , indeed , their effect decreases more rapidly than @xmath17 .
this reinforces the conclusion that there is only linear growth with @xmath232 : a very strong breakdown of the deterministic approximation which yielded exponential growth for any power law . when @xmath233 is longer range , in particular if @xmath234 with @xmath235 , many jumps longer than @xmath8 will have occurred by time @xmath39 .
we now investigate the effects of such long jumps on the density profile .
to do so , we investigate the behavior of @xmath236 for large @xmath135 and @xmath40 in terms of the @xmath237 at shorter times and primarily corresponding distances @xmath238 which can be much less than @xmath135 .
mutants can get to a chosen point , @xmath239 , by one making a long jump at time @xmath39 from a starting point @xmath15 to an end point , @xmath16 , and subsequent spread from there to @xmath239 during the remaining time interval of duration @xmath240 .
the rate ( per volume elements ) of this occurring is @xmath241 . in the approximation that these are independent ,
the probability that this does not occur at any @xmath242 from any @xmath15 to any @xmath16 is simply poisson so that ( r , t)1- e^-(r , t ) [ eq : occupancy - probabili ] with ( r , t ) _ 0^t dt _
z>(t/2 ) d^dd^d ( , t)g(z ) ( |--|,t - t ) .
[ eq : q - definition ] here , we substituted the final point @xmath243 by the sum of the jump start site and a jump vector @xmath244 , over which we integrate .
note that a lower cut - off in the @xmath245 integral is necessary to exclude the many very short jumps that lead to strongly correlated establishments .
the cut - off is also necessary to not count mutants that results from growth in the target area , rather then seeding from the source funnel .
our main assumption here is that if a single seed is sufficiently far from other seeds or occupied regions then the growth from the seed is independent of the rest of the system as long as collisions are unlikely .
when the jump integral is strongly peaked at @xmath246 , as is the case in or close to the stretched exponential regime , the final results will be independent of this cutoff to leading order .
then , we can approximate @xmath247 as @xmath248 where @xmath37 is the expected total size of a population at a time @xmath39 , @xmath249 for a power - law kernel @xmath250 , we make the ansatz that eqs . , and can be approximately solved by a scaling form @xmath251 which leads to the condition @xmath252 where @xmath253 and @xmath254 is given by @xmath255 thus , the above scaling form is a valid solution if the characteristic scale @xmath256 satisfies @xmath257 the resulting condition is similar to our condition in the main text but differs by the numerical factor @xmath258 . in one dimension ,
@xmath259 the divergence @xmath260 as @xmath261 indicates the importance of seeds from the tail regions for small @xmath7 . notice that at long distances , @xmath262 , the length of the jumps that dominate are close to @xmath135 so that our approximation for @xmath263 should be correct even if it is a poor approximation for @xmath264 .
thus the decrease in @xmath265 at large distances is simply proportional to @xmath266 , more specifically , ( r , t ) ~ ( as predicted by the scaling form ) so that it is of order unity at @xmath267 .
note that this implies that , since @xmath17 is integrable , @xmath268 is indeed dominated by @xmath269 as we have assumed .
this long distance form for the density profile is also found in the analyses of upper and lower bounds in the next sections : thus it can be readily proved along the same lines .
as most of our results are based on approximate analyses and heuristic arguments , it is useful to supplement these by some rigorous results .
we focus on the one - dimensional case : extensions to higher dimensions can be done similarly , although with a few complications that will require some care .
we sketch here the arguments that can lead to proofs without all the details filled in . as via the heuristic arguments
, we would like to obtain the behavior at longer times in terms of the behavior at shorter times , in particular times around half as long .
we would like to prove that there exist time dependent length scales , @xmath270 and @xmath271 and functions , @xmath272 and @xmath273 such that the probability , @xmath224 , that a site at @xmath2 from the origin is occupied at time @xmath39 , is bounded above and below by ( r , t ) < ( r , t ) < ( r , t ) with m _ < dr ( r , t ) ( t ) & _ > dr ( r , t ) ( t ) for all times . then @xmath274 and @xmath275 are lower and upper bounds for @xmath8 with some appropriately defined definitions of @xmath8 which differ somewhat , although not significantly , for the upper and lower bounds .
while we would like the upper and lower bounds on @xmath8 to be as close as possible to each other , in practice , we have obtained bounds that are good on a logarithmic scale : i.e. for @xmath187 , rather than on a linear scale .
similarly , we would like to have the bounds be close to the actual expected form of @xmath265 , with @xmath276 very close to unity for @xmath277 and proportional to @xmath278^{\mu+1}$ ] for @xmath279 and similar for the upper bounds .
it is often more convenient to consider the typical time - to - occupation as a function of the distance , @xmath129 and derive upper and lower bounds for this , @xmath280 and @xmath281 , respectively , such that ( t=(r))=r & ( t=(r))=r with ( r ) > ( r ) > ( r ) .
because of the faster - than - power - law growth , bounds on @xmath129 are generally much closer than those on @xmath8 . as we would like to justify the use of the heuristic iterative scaling arguments more generally , it is especially useful to obtain iterative bounds directly of the form used in those heuristic arguments : @xmath64 in terms of @xmath282 for @xmath39 in a range near @xmath186 .
as the heuristic arguments do , in any case , only give @xmath64 up to a multiplicative coefficient of order unity , we will generally ignore such order - unity coefficients in length scales except for coefficients that diverge or vanish exponentially rapidly as @xmath283 , in particular in the intermediate range regime the coefficient , @xmath52 in @xmath284 which vanishes as @xmath285 as @xmath197 .
the simplest bound to obtain is an upper bound for @xmath8 in the short and intermediate range regimes : i.e. , in one dimension , @xmath286 . define @xmath287 to be the right - most edge of the occupied region at time @xmath39 , i.e. @xmath288 for @xmath289 . the probability of a jump that fills a position @xmath290 in @xmath291 is less than @xmath292 . for @xmath293 , the lower extent of the integral can be taken to @xmath294 as the jumps arise , predominantly , from points that are not too far from the edge .
[ in contrast , for @xmath295 jumps from the whole occupied region are important and this bound would yield a total jump probability to long distances that diverged when integrated over @xmath63 , and one would have to instead use a lower extent of the integral of @xmath296 for the left edge . ]
the advancement of the edge is bounded by a translationally and temporally invariant process of jumps of the position of the edge by distances , @xmath297 , whose distribution has a power - law tail . for @xmath298 , the mean @xmath299 , implying that the edge , and hence @xmath8 , can not advance faster than linearly in time .
but for the intermediate regime , @xmath300 so that @xmath287 could advance as fast as a one - sided levy flight with @xmath301 dominated by the largest advance . as this process would yield @xmath302 , this implies that @xmath8 is bounded above by the same form as the heuristic result .
although the simple bound captures some relevant features , in particular the dominance of jumps of length of order @xmath2 to fill up a point at distance @xmath2 , it is otherwise rather unsatisfactory . first , the coefficient does not vanish rapidly as @xmath303 . and second , it suggests that the probability that an anomalously distant point , @xmath304 , is occupied , is , in this crude approximation of full - occupancy out to the edge , simply the probability that @xmath305 which falls - off only as @xmath306 much more slowly than the actual @xmath307 .
nevertheless , for proving better upper - bounds , the levy - flight approximation for the dynamics of the edge is quite useful . as discussed earlier
, one would like to make the heuristic argument of a single long jump from a source region to a target funnel region include also or provide solid reasons to ignore the effects of jumps from the partially filled region outside the core of the source .
very loosely , one would like to write the probability that a point , @xmath135 , is not occupied at time @xmath40 , as 1-(r , t ) with @xmath62 in the source region and @xmath63 in the funnel of @xmath135 .
but for any positive @xmath7 , the spatial integral is dominated by @xmath308 small , so that this does not properly represent the process : there is a drastic over - counting of short jumps .
one can do much better by trying to separate the long jumps from the short ones , and the source region from the funnel ( in the crude approximation these overlap ) . to do this ,
we choose , for the @xmath135 and @xmath40 of interest , a spatio - temporal source region , @xmath134 , around the origin which has a boundary at distance @xmath309 that loosely reflects the growing source : @xmath310 .
we then separate the process of the set of jumps that lead to @xmath135 into three parts .
first , jumps solely inside @xmath311 which lead to a spatio - temporal configuration of occupied sites , @xmath312 ; second , _ bridging _ jumps from these out of @xmath311 , say at time @xmath39 from @xmath62 in @xmath134 to a point @xmath63 in the rest of space - time , @xmath313 , and third all the subsequent dynamics from such seeds in @xmath313 including inside and outside @xmath311 and between these .
this over - counts the possible spatio - temporal routes to @xmath314 especially as returns to inside @xmath311 from outside are included and thus provides an upper bound for @xmath236 .
the probability , @xmath315 , that a single seed , @xmath316 , to @xmath317 at @xmath318 leads to @xmath135 being filled by @xmath40 is @xmath319 .
but the probability that a second seed , @xmath320 , leads to @xmath135 filled by @xmath40 is not independent as the fate of these seeds involves overlapping sets of jumps : indeed , they are positively correlated so that . since for a given @xmath321 , the probability density of a seed at @xmath322 is @xmath323 , and using the generalization of the above bound to many seeds , we have ( r , t ) 1- [ q - up - iter ] where @xmath324 and we have used @xmath325 for any random variable . to derive a useful upper bound on @xmath236 we need to choose appropriately the boundary , @xmath309 , of the source region and put a sufficiently stringent upper bound on @xmath326 . for the long - range case , @xmath295 ,
the integrals over @xmath327 and @xmath328 of @xmath329 are dominated by long distances .
thus the short jumps from inside to outside @xmath311 do not contribute significantly .
we can then simply replace @xmath330 by the larger @xmath265 to obtain a slightly weaker bound which is of exactly the form of the naive estimate except for the strict delineation of the source region which prevents the most problematic over - counting of the effects of short jumps .
a particularly simple choice is @xmath331 independent of @xmath39 .
we now proceed by induction .
take the bound on the scaling function to have the form @xmath332 for @xmath333 while @xmath334^{\mu+1}$ ] for @xmath335 and assume that for some appropriate @xmath271 , this is indeed an upper bound for all @xmath242 : i.e. , @xmath336 $ ] .
we can now use with @xmath330 and @xmath265 both replaced by @xmath337 . when @xmath271 and @xmath338 are both much less than @xmath135 , the integrals over @xmath62 and @xmath63 will be dominated by the regions near the origin and @xmath135 respectively , yielding the spatial convolution @xmath339 .
there are small positive corrections to this from two sources : first , from the regions near @xmath340 and @xmath135 , which , by expanding @xmath308 in @xmath62 and @xmath341 , are seen to be of order @xmath342\lg(t)\lg(t - t)/r^{\mu+3}$ ] ; and , second , from @xmath343 , the region near the source boundary , which are of order @xmath344^{\mu+1}\lb[\frac{\lg(t - t)}{r}\rb]^{\mu+1}r^{1-\mu}$ ] with the last part from the integrals over @xmath62 and @xmath63 .
as the dominant part is exactly of the form in the heuristic treatment , integrating it over time is strongly peaked at @xmath345 ( note that for either @xmath39 or @xmath240 much smaller than @xmath40 , one of the @xmath337 factors will be close to unity near the boundary , but these ranges of time only contribute weakly ) .
if we use @xmath346 then @xmath347 can be chosen as the value of @xmath135 for which @xmath348 , and for @xmath349 the desired @xmath350^{\mu+1}$ ] is obtained . including the small correction factors in the convolutions
necessitates slight modifications of the recursion relations for @xmath275 but these are negligible at long times .
obtaining an upper bound in the intermediate range case is somewhat trickier .
if we again replaced @xmath330 by @xmath265 , then the integrals over @xmath62 and @xmath63 would have a part dominated by both points being near the boundary : with @xmath351 , this contribution to the convolution would be of order @xmath344^{\mu+1}\lb[\frac{\lg(t - t)}{r}\rb]^{\mu+1}$ ] . with @xmath352 and @xmath267 ,
all the lengths should be of order @xmath196 , so that this boundary piece is larger by a factor of @xmath353^{\mu-1 } \sim t$ ] than what - should - be the dominant part from @xmath62 and @xmath63 near @xmath340 and @xmath135 , respectively .
thus we need a better upper bound on the restricted - source @xmath354 which vanishes as @xmath355 . to bound @xmath330 , we can make use of the simple bound for the edge of the occupied region derived above , combined with the restrictive effects of the boundary , @xmath309 . instead of choosing @xmath356 to be constant , we choose it to have constant slope , @xmath357 , of order @xmath358 .
as jumps that contribute to @xmath330 are not allowed to cross the boundary , the distribution of jumps of the edge @xmath287 is cutoff at @xmath359 . since @xmath360 , the speed of spread in the absence of jumps beyond nearest neighboring sites , typically the gap , @xmath361 , will increase with time , decreasing only by jumps .
the sum of all the jumps of @xmath362 in a time interval @xmath363 is dominated by the largest , which is of order @xmath364 .
this would result in the edge moving faster than @xmath365 except for the cutoff .
the typical gap , @xmath366 , is then obtained by balancing its steady decrease against the dominant jump : @xmath367 yielding @xmath368 and hence ~u^1/(2- ) ~^1/(2- ) ~(t ) a_^(-1)/(2- ) using @xmath369 . in the limit of @xmath303 , @xmath370 , vanishing rapidly
a reflection of the strong failure of the simple edge - bound in this limit but sufficient for our present purposes . in a time @xmath371 , the distribution of @xmath120 in this approximation will reach a steady state . the probability that @xmath372 is controlled by the balance between jumps of @xmath362 to near the boundary , and the steady increase in @xmath120 from the boundary motion :
its probability density is hence of order @xmath373 which , since in this approximation all sites are occupied up to @xmath362 , implies that @xmath330 vanishes at least quadratically for small gap @xmath120 .
combining this with the trivial bound of @xmath374 and choosing a convenient normalization of @xmath366 , we thus have ( r , t ) .
it remains to chose @xmath309 so that the bound on @xmath330 remains sufficiently good for @xmath39 small enough that the steady state distribution of @xmath361 has not yet been reached . to keep @xmath287 typically of order @xmath366 from @xmath309
, we can simply chose @xmath375 and @xmath376 .
with our improved bound on @xmath330 , for the convolution @xmath377 , the small @xmath378 parts are no longer dominated by @xmath378 of order unity , but by @xmath378 near the crossover point between the two bounds on @xmath330 .
this yields a contribution to the convolution of order @xmath379 with @xmath380 and @xmath381 and a multiplicative coefficient that does not depend exponentially on @xmath382 because the integral over @xmath62 scales as @xmath383 .
as @xmath303 , @xmath384 , and the boundary contribution is less than the dominant part uniformly in @xmath39 . note that for @xmath385 , the bound on the near - boundary contribution can be somewhat larger for @xmath386 than the dominant parts , but it scales in the same way with @xmath40 and thus only weakens the upper bound on the coefficient , @xmath52 , which is in any case of order unity in this regime .
once the over - counting of short jumps has been sufficiently reduced , as we have now done , the rest of the analysis , in particular the large @xmath387 form of @xmath337 , follows as in the long - range case . the _ marginal case _
@xmath388 can be analyzed similarly to the intermediate - range case , resulting in an additional logarithmic dependence on @xmath135 of the near - boundary contribution which is , nevertheless , still much smaller than the dominant part .
the upper bounds that we have obtained are , except for modifications at small scales and for @xmath7 not much smaller than 2 , essentially the same as given by the heuristic arguments , thus only differing at long scales by order - unity coefficients which , in any case , we did not expect to get correctly .
all the crossover behavior near @xmath388 is in the upper bounds , although that near @xmath214 is not . to obtain lower bounds on the growth of the characteristic length scale @xmath8 and the occupation probability , @xmath224
, a different strategy needs to be employed .
one of the difficulties is the dependence on the behavior at each time scale on all the earlier time scales : for the filled region to grow typically between time @xmath40 and @xmath389 , the stochastic processes that lead to the configuration @xmath390 must not have been atypically slow or ineffective .
as this applies iteratively scale by scale , we must allow for some uncertainty in whether or not the smaller scale regions are typical , leading to some uncertainty at all scales which , nevertheless , we need to bound . because of the stochastic heterogeneity of @xmath391 it is better to focus on a coarse - grained version of the occupation profile rather than on @xmath391 itself , as integrations over @xmath392 at time @xmath39 are what act as the sources of future occupation at larger distances .
we consider the probability that a region is almost full , in particular , with a seed at the origin , we consider the region to one side of the origin and define ( r , t ; ) with @xmath393 close to or equal to unity being of particular interest . in order to keep events sufficiently independent ,
we consider , as for the upper bounds , the probability of events that do not involve any jumps out of some region .
in particular , defining @xmath394 similarly @xmath395 , but with the restriction that jumps do not go out of the interval @xmath396 . for the long - range and marginal cases , we will focus on partial filling , but for the intermediate range case the scale invariance mandates different treatment so we instead analyze full filling i.e. @xmath397 .
the basic strategy is to start with a particular deterministic approximation to @xmath8 , @xmath398 with corresponding times @xmath399 , and then show that at time not too large a multiple of @xmath399 , the region out to @xmath2 will be nearly filled with high probability : i.e. that @xmath400 is close to unity for @xmath401 sufficiently large
. we will be interested in large scales as , in any case , fluctuations at the small scales can only change coefficients by order unity
. we can thus be sloppy with some of the bounding inequalities : these could be improved to include the ignored corrections to the large scale effects to make fully rigorous bounds . as the range of time over which the typical @xmath8 expands significantly plays an important role , it is useful to define ( t ) ^-1 which is small except for @xmath7 substantially larger than one .
the dominant jumps from source to funnel involve an integral over time of @xmath402 , which is primarily from a range of order @xmath403 around @xmath40 as discussed above .
the deterministic - iterative approximation that we use as a base for the lower bounds is the solution to the iterative relation ( ambiguous up to an @xmath404 multiplicative factor which we ignore throughout ) : ^+1= t[(t)]^2 corresponding to roughly one seed into a funnel of width @xmath405 from a jump of distance @xmath406 from the source up to time @xmath40 . for convenience
, we use only half the source @xmath62 from @xmath340 to @xmath405 .
the results , @xmath398 , of this iterative approximation are , up to numerical factors that arise from these modifications and from other from non - asymptotic effects at small scales , equivalent to the upper bounds , @xmath271 from the above .
in particular , the ratio between the corresponding times , @xmath399 and @xmath281 we expect to approach constants that are not singular near the marginal case @xmath407 . for the lower bounds it is convenient to work with a specific set of length scales , @xmath408 , corresponding to a series of time scales , @xmath409 ( dropping a prefactor ) .
to mostly fill out to @xmath410 without jumps going out of @xmath411 from the source of size @xmath412 , most of the @xmath413 bins of size @xmath412 must be mostly filled . to get a lower bound on how long this takes and how likely it is
, we make several simplifications each of which lead to underestimates of the probability that the desired filling has occurred .
first , consider only jumps into each bin that come directly from the source ( rather than from other bins as can occur later ) .
second , ignore all but the first seed jump from the source into the bin ( the effects of later jumps are not independent of those of the first ) . and third , include only jumps that lead from the seed in a bin that do not go outside that bin during the time during which the probability of it being mostly filled is considered .
the last two conditions mean that the probability that the bin is filled to a fraction @xmath393 by a given time , @xmath39 , after the seeding jump , is at least as large as @xmath414 since a seed at the edge of the bin , which corresponds to the definition at the source , is less likely to mostly fill the bin than a seed away from the edge
. at large scales for @xmath415 , the number of bins , @xmath416 , grows with scale : @xmath417 for the marginal case and larger for the long - range case .
thus if the probability that the furthest bin from the source is mostly filled is @xmath418 , with the filling of the others being more probable as they are closer , it is likely that the number that are similarly mostly filled is close to @xmath419 , with significant deviations from this being very unlikely at large scales . in order to iterate while not losing too much in filling fraction
, we chose a series of partial filling fractions , @xmath420 , such that @xmath421 converges to the desired overall filling fraction , @xmath393 , at large @xmath163 , and chose conditions such that @xmath418 is sufficiently large that the fraction of the @xmath416 bins filled to @xmath422 is greater than @xmath423 with high probability : this then implies that the region from the origin to @xmath410 will be filled to greater than @xmath424 with high probability .
a convenient choice is @xmath425 with any positive @xmath142 and @xmath426 . for convenience in dropping @xmath427 factors that otherwise appear in many places , we restrict consideration to @xmath428 not very small , and do not keep careful track of @xmath142 factors that also appear as one can take @xmath429 at the expense of corrections that are down by one extra logarithm .
the filling probability of a bin is at least as large as that obtained from the requirement of the occurrence of both of two independent events : a jump into the bin from the source that occurs before some chosen initial time , @xmath430 , and the bin being filled from that single seed by a time , @xmath431 .
the probability of a jump into a bin is at least @xmath432 in terms of a conveniently chosen lower bound , @xmath433 , on the expected number of jumps from the source into the furthest away bin , and the probability of the bin being filled from the single seed is at least @xmath434 .
we will find iterative bounds on @xmath435 that it is convenient to write in the form ( r , t ; ) 1 - e^-(r , t ; ) so that 1-f_n e^-w_n + e^_b with _ b(_n , t_b;_n ) .
for convenience we chose conditions so that @xmath436 and @xmath437 which , for @xmath438 large , makes the probability that a fraction @xmath423 of the bins are not filled exponentially small .
we henceforth ignore this factor in the probability as it does not matter except on small scales : adjustments to take it into account are straightforward .
we thus require that _ b w_n = ( 1 + ) n + o(1 ) . to obtain a bound on @xmath439 , we must show that a source that can give rise to an average number at least @xmath433 of jumps into the furthest bin by time @xmath430 , occurs with probability that is somewhat larger than the desired bound at the next scale .
the expected effective number of jumps out of the source of size @xmath412 into a bin of the same size a distance up to @xmath410 away before time @xmath440 was assumed in the deterministic iterative approximation to be of order @xmath441 . in order to ensure that the average number from the actual source is sufficiently large
, we can require that it be almost filled by some time @xmath442 and only include jumps that occur between @xmath442 and @xmath430 as the rate of these is bounded below by the filling at @xmath442 .
the required range is t_i - t_s = w_n_n/_n .
we now proceed by induction and show that if ( _ n , t;_n ) _ n(t - u_n_n ) for @xmath39 in a range such that @xmath443 is relatively large the precise range is not crucial but minor modifications are needed to extend out to arbitrary large @xmath39 then a similar bound holds at the next scale with coefficients @xmath444 and @xmath445 with both these varying slowly with @xmath446 at large scales .
note that at the smallest scale the probability that a site is filled by a jump directly from the origin by time @xmath39 converges exponentially to unity for long @xmath39 , thus at the smallest scales there is a trivial bound of this form .
as the scale is increased , @xmath447 will initially change , but once the scale becomes large enough that the width of the distribution of the fraction of the bins mostly filled is small , then @xmath447 saturates and becomes weakly dependent on @xmath446 . in the analysis below , it can be replaced by a constant .
consider a total time @xmath40 to mostly fill out to @xmath410 .
the time for the bins to fill with sufficiently high probability once they have been seeded is @xmath448 . with @xmath449 as above , we have a time for the source to fill t_s t - _
n[u_n + w_n/_n ) ] -w_n/_n .
plugging in the probability that the source is filled in this time gives a bound on @xmath450 of the same form but with , dividing out @xmath451 , u_n+1 u_n + + .
[ u - itern ] as @xmath440 increases rapidly and @xmath433 only slowly , the last term only contributes at small scales . for the _ long - range regime _
, @xmath452 so the second term in is also small except at small scales and we conclude that @xmath365 is bounded above by a @xmath7 dependent constant .
thus the lower bound for @xmath8 , and upper bound for @xmath129 have exactly the same form as the opposite bounds , except with the scale of @xmath39 i.e. @xmath453 different . for the _ marginal case _
, @xmath454 so that @xmath455 changes slowly at large scales . integrating up , one sees that u_n < c n ~ _ n with a coefficient independent of @xmath446 ( but depending on @xmath393 and @xmath142 ) .
one can now solve for the time scale above which mostly filled is likely , @xmath456 , to find a lower bound , @xmath270 , on @xmath8 : ( ( t))((t))= [ t - 2 t - o(t ) ] [ low - bnd - marginal ] which is very close to the upper bound derived above , ( ( t))((t))= [ t - t - o(1 ) ] differing only in the coefficient of the correction term .
one of the advantages of this iterative approach is that the _ crossover regime _ can be handled similarly by integrating up .
the lower bound will be similar to the upper bound throughout this crossover regime and into the asymptotic regimes for the marginal and long - range cases .
the reason that the fluctuation effects are relatively small for the marginal and long - range regimes is that at each successive time scale , more and more roughly independent long jumps are involved in filling up to the next length scale : i.e. @xmath438 continues to grow . for the long - range regime
, it grows so rapidly that almost all the fluctuations come from early times : this is like what occurs for the fully mixed model . for the marginal case ,
the fluctuations are dominated by the smallest scales but the cumulative effects of them over the longer scales does make a difference as found in obtaining the lower - bounds . for the intermediate - range case , the ratio of length scales for each factor of two in time scales saturates ( when out of the crossover regime ) at @xmath457 .
this means that whatever fraction , @xmath423 , of the bins are to be filled at each scale , the probability that this occurs either decreases with scale if the product of the @xmath423 does not go to zero , or saturates to a constant if the @xmath423 s do also , in which case the overall filling fraction @xmath422 tends to zero as a power of time . at each scale
there are a comparable number of long jumps that are needed , thus one should expect that fluctuation effects will be scale invariant and not decrease with scale . to get useful lower bounds on @xmath8 via an upper bound on @xmath129 , @xmath280 , the easiest way is to require that the source be completely full , and that jumps from this completely fill all the bins at the next scale : this avoids the problems with the @xmath423 . as the probability that all the bins are filled is only ( readily ) bounded by @xmath458 if we again chose @xmath459 , @xmath460 must be larger by @xmath461 than for the partially filled analysis above . carrying through similar analysis ,
gives for large @xmath462 a coefficient @xmath463 which means that the filling probability decays for large times as roughly the inverse of the typical time natural as the needed long jumps that occur at rate @xmath464 have a distribution of when they occur on the same time scale .
note the contrast to the rapid decay of the not - mostly - filled probability on a time scale of order unity from the partially filled bound derived above .
the time beyond which the full filling is likely is only bounded by , in this analysis , @xmath465 .
this gives a lower bound on @xmath466 proportional to @xmath467 .
although on a logarithmic scale the additional factor is smaller , we would like to do better .
the bound can be improved by considering a source that is somewhat smaller by a factor of two is sufficient than @xmath412 which increases the probability that it is filled , but means that more extra time , @xmath468 , is needed to produce a mean number of jumps to the furthest bin of at least @xmath460 . using that @xmath365 and @xmath469 vary slowly with scale
, one can expand @xmath470 around @xmath471 and analyze the changes on the bounds at the next scale .
this improves the bounds to @xmath472 and @xmath473 with @xmath474 .
the resulting lower bound on @xmath8 is ( t ) > ^- t^with the coefficient @xmath475 that from the deterministic iteration which is the same , up to an order unity pre factor , as the upper bound .
it is not clear where between the lower and upper bounds on the coefficient will be the typical behavior , nor how broad the fluctuations will be even on a log scale . the behavior as @xmath476 we have not analyzed explicitly , instead focussing on the rapidly growing regime for @xmath303 , but the bounds will be of similar form although more care is needed to get upper and lower bounds reasonably close to one another due to the important jumps being only a modest fraction of the size of the already occupied region . for the marginal case , @xmath388
, one can find an upper bound on the time at which a the region out to @xmath2 is likely to be fully filled by similar methods to that for the intermediate - range power - law regime .
this yields a bound @xmath477 of the same form as that above for partial filling ( @xmath478 ) , , except with the coefficient of the @xmath479 term in equal to 6 instead of 2 .
the convergence of the probability of being fully filled is , however , much slower for this bound on compete filling than for the bound on being mostly filled . while the latter converges for @xmath480 with a rate of order unity
dominated by the small scales the former converges as the time increases above @xmath481 on a time scale , @xmath482 , of order @xmath483 faster than @xmath484 but not much so .
note that the convergence of the probability for being mostly filled to more than a fixed filling fraction , @xmath393 , is a hybrid property : the probability of a fixed site being filled by time @xmath39 , @xmath485 , is bounded below by ( roughly ) the product of @xmath393 and the probability that the region out to @xmath2 is filled to above @xmath393 . to get the convergence of this to unity
, @xmath393 needs to be adjusted and the thus - far ignored @xmath486 factors kept track of .
this also necessitates treating intermediate scales differently as the number of bins that do not need to be filled , @xmath487 is not large .
however a different approach would provide a better bound : focussing on a specific site being filled with high probability can be done by a method more analogous to the funnel picture of the main text : for the site to be filled , it needs to be in a small - scale bin that is mostly but not - necessarily fully filled with high probability , which needs itself to be in a larger bin similarly , etc . but these can be filled from source regions that are not fully filled : being partly filled with high enough probability is sufficient .
we have not carried out such analysis in detail in part because the actual mechanism by which sites that are empty for anomalously long will be filled is more complicated as it will involve filling from nearby regions on a hierarchy of scales that were filled at more typical times .
the analyses here can be immediately extended to give lower bounds on the average density profile at long distances : these will be of the same form as the lower bounds , thus demonstrating that the predicted @xmath488 is essentially correct . as noted in the main text , when this work was essentially complete , a preprint by chatterjee and dey ( cd ) appeared which derives and proves some results closely related to ours in the context of long - range first passage percolation which is essentially equivalent to the lattice dispersal model , with the jump kernel @xmath489 equivalent to the @xmath490 that we use @xcite . while some of the quantities cd focus on are different , the leading asymptotic scaling behaviors they obtain are essentially the same , and their proofs apply in all dimensions .
our @xmath485 corresponds to the probability that the first passage time @xmath491 is less than @xmath39 and their diameter , @xmath492 the maximum distance between any pair of occupied points at time @xmath39 is , with high - probability that decays as a power of @xmath493 not many times @xmath8 , as we both obtain .
however cd s results are sub - optimal . in particular , for the coefficient , @xmath173 , of @xmath494 in the marginal case , they only obtain upper and lower bounds instead of our exact result @xmath198 . indeed ,
in one dimension we obtain upper and lower bounds on the errors : 1-c_- < < 1- with high probability in senses that can be made precise from our analysis with the coefficient @xmath495 either 2 or 6 depending on the definition of @xmath8 used . in the intermediate range power - law growth regime ,
cd s theorems do not appear to exclude @xmath496 pre - factors in @xmath8 , although their analysis might well do so .
but the main difference is our analysis of the whole crossover regime for @xmath7 near @xmath0 , including the divergences and vanishings of coefficients of the asymptotic forms , which they do not consider .
these are crucial for comparisons with simulations because of the very long length scales of the crossovers . to turn our upper and lower bounds into formal proofs in one dimension requires primarily filling in some details associated with the small scale regime . for higher dimensions
some additional work is needed , although the strategies should work without major modifications .
[ ht ! ] of the predicted power law growth in eq .
( 3 ) ( main text ) in one dimension : @xmath497 with @xmath474 for @xmath206 .
the blue curve is obtained numerically from solving equation ( 1 ) ( main text ) with the power - law ansatz ; the red - curve represents an analytic approximation derived in `` crossovers and beyond asymptopia '' ( main text ) .
notice the sharp ( non - analytic ) drop of the pre - factor as @xmath7 approaches @xmath212 .
the reason is very slow cross - over to the power law from an intermediate asymptotic regime controlled by the dynamics of the marginal case . as @xmath7 approaches 2
, the pre - factor diverges as @xmath215 indicative of another slow crossover at @xmath214 , see fig .
[ fig : second - marginal - point ] .
[ fig : prefactors],title="fig : " ] [ ht ! ] . the number of mutant sites , @xmath37 , scaled by time , @xmath498 , is plotted as a function of time , averaged over 10 realizations ( black ) and for two individual realizations ( red and green ) . while the averaged data suggests @xmath499 , the individual realizations indicate strong fluctuations caused by occasional rare jumps , which are of order @xmath39 .
these `` leaps forward '' @xcite are driving the logarithmic increase of the spreading velocity .
[ fig : second - marginal - point],title="fig : " ]
|
the spreading of evolutionary novelties across populations is the central element of adaptation . unless population are well - mixed ( like bacteria in a shaken test tube ) , the spreading dynamics not only depends on fitness differences but also on the dispersal behavior of the species . spreading at a constant speed
is generally predicted when dispersal is sufficiently short - ranged .
however , the case of long - range dispersal is unresolved : while it is clear that even rare long - range jumps can lead to a drastic speedup , it has been difficult to quantify the ensuing stochastic growth process .
yet such knowledge is indispensable to reveal general laws for the spread of modern human epidemics , which is greatly accelerated by aviation .
we present a simple iterative scaling approximation supported by simulations and rigorous bounds that accurately predicts evolutionary spread for broad distributions of long distance dispersal . in contrast to the exponential laws predicted by deterministic mean - field approximations , we show that the asymptotic growth is either according to a power - law or a stretched exponential , depending on the tails of the dispersal kernel . more importantly , we provide a full time - dependent description of the convergence to the asymptotic behavior which can be anomalously slow and is needed even for long times .
our results also apply to spreading dynamics on networks with a spectrum of long - range links under certain conditions on the probabilities of long distance travel and are thus relevant for the spread of epidemics .
humans have developed convenient transport mechanisms for nearly any spatial scale relevant to the globe .
we walk to the grocery store , bike to school , drive between cities or take an airplane to cross continents .
such efficient transport across many scales has changed the way we and organisms traveling with us are distributed across the globe @xcite .
this has severe consequences for the spread of epidemics : nowadays , human infectious diseases rarely remain confined to small spatial regions , but instead spread rapidly across countries and continents by travel of infected individuals @xcite . besides hitchhiking with humans , small living things such as microbes or algae
are easily caught by wind or sea currents , resulting in passive transport over large spatial scales @xcite .
effective long distance dispersal is also wide - spread in the animal kingdom , occurring when individuals primarily disperse locally but occasionally move over long distances . and such animals , too
, can transport smaller organisms .
these active and passive mechanisms of long - range dispersal are generally expected to accelerate the growth of fitter mutants in spatially extended populations . but how can one estimate the resulting speed - up , and the associated spatio - temporal patterns of growth ? when dispersal is only short - range , the competition between mutants and non - mutated ( `` wild - type '' ) individuals is local , confined to small regions in which they are both present at the same time . as a consequence
, a compact mutant population emerges that spreads at a constant speed , as first predicted by fisher and kolmogrov and colleagues @xcite : such selective sweeps are slow and dispersal limited . in the extreme opposite limit in which the dispersal is so rapid that it does not limit the growth of the mutant population , the competition is global and the behavior the same as for a fully - mixed ( panmictic ) population : mutant numbers grow exponentially fast .
it is relevant for our purposes to note that in both the short - range and extreme long - range cases , the dynamics after the establishment of the initial mutant population is essentially deterministic .
when there is a broad spectrum of distances over which dispersal occurs , the behavior is far more subtle than either of the well - studied limits .
when a mutant individual undergoes a long distance dispersal event a jump from the primary mutant population into a pristine population lacking the beneficial mutation , this mutant can found a new satellite sub - population , which can then expand and be the source of further jumps , as shown in fig .
1b and 1c .
thus , long - range jumps can dramatically increase the rate of growth of the mutant population .
potentially , even very rare long - range exceptionally long jumps could be important .
if this is the case , then the stochastic nature of the jumps that drive the dynamics will be essential .
while evolutionary spread with long - range jumps has been simulated stochastically in a number of biological contexts @xcite , few analytic results have been obtained on the ensuing stochastic dynamics @xcite .
most analyses have resorted to deterministic approximations @xcite , which are successful for describing both the local and global dispersal limits . yet , in between these extreme limits , stochasticity drastically changes the spreading dynamics of the mutant population .
this is particularly striking when the probability of jumps decays as a power - law of the distance .
just such a distance spectrum of dispersal is characteristic of various biological systems @xcite .
we will show that the behavior is controlled by a balance between the rarity and the potential effectiveness of long distance jumps and the whole spectrum of jump distances can matter .
the goal of this paper is to develop the theory of stochastic spreading dynamics when the dispersal is neither short range nor global .
long distance dispersal can occur either on a fixed network , or more homogeneously in space . for simplicity ,
we focus on the completely homogeneous case , and then show that many of the results also apply for an inhomogeneous transportation network with hubs between which the long distance jumps occur . for definiteness , we consider for most of the paper the evolutionary scenario of the spread of a single beneficial mutation , but , by analogy , the results can be applied to other contexts , such as the spread of infectious disease or of invasive species . the underlying model of spatial spread of a beneficial mutant is a population in a @xmath0-dimensional space with local competition that keeps the population density constant at @xmath1 and with a probability that any individual jumps to any particular point a distance @xmath2 away of @xmath3 per time per area , per length or per volume . at a very low rate , mutants can appear that have a selective advantage , @xmath4 , over the original population .
a lattice version of this model is more convenient for simulations ( and for aspects of the analysis ) : each lattice site represents a deme " with fixed population size , @xmath5 with the competition only within a deme and the jump migration between demes .
initially , a single mutant occurs and if , as occurs with probability proportional to @xmath4 , it survives stochastic drift to establish , it will take over the local population . when the total rate of migration between demes is much slower than this local sweep time ,
the spatial spread is essentially from demes that are all mutants to demes that are all the original type .
short jumps result in a mutant population that spreads spatially at a roughly constant rate .
but with long - range jumps , new mutant populations are occasionally seeded far away from the place where they came , and these also grow .
the consequences of such long jumps is the key issue that we need to understand .
as we shall see , the interesting behaviors occur when the jump rate has a power - law tail at long distances , specifically , with @xmath6 ( with positive @xmath7 needed for the total jump rate to be finite ) .
crudely , the behavior can be divided into two types : linear growth of the radius of the region that the mutants have taken over , and faster than linear growth . in fig .
[ fig : illustrate - sweep - dynamics ] , these two behaviors are illustrated via simulations on two dimensional lattices .
in addition to the mutant - occupied region at several times , shown are some of the longest jumps that occur and the clusters of occupied regions that grow from these . in fig .
[ fig : illustrate - sweep - dynamics]a , there are no jumps that are of comparable length to the size of the mutant region at the time at which they occur , and the rate of growth of the characteristic linear size @xmath8 of the mutant region loosely its radius is roughly constant in time , i.e. @xmath9 . in fig .
[ fig : illustrate - sweep - dynamics]b and c , @xmath10 is longer range and very long jumps are observed .
these result in faster - than linear growth of the radius of the mutant region , as shown . before developing analytic predictions for the patterns of evolutionary spread ,
we report our simulation results in detail .
|
it is widely accepted that type ia supernovae ( sne ia ) are thermonuclear explosions of accreting c+o white dwarfs ( wds ) , although the nature of the progenitor binary system and the detail of the explosion mechanism are still under debate .
sne ia are good distance indicators , and provide a promising tool for determining cosmological parameters ( e.g. , branch and tammann 1992 ) . from the observations of high redshift sne ia , both the sn cosmology project ( perlmutter et al . 1999 ) and the high - z sn search team ( riess et al .
1998 ) have suggested a statistically significant value for the cosmological constant .
however , sne ia are not perfect standard candles , but show some intrinsic variations in brightness . when determining the absolute peak luminosity of high - redshift sne ia , therefore , these analyses have taken advantage of the empirical relation existing between the peak brightness and the light curve shape ( lcs ) . since
this relation has been obtained from nearby sne ia only ( phillips 1993 ; hamuy et al .
1995 ; riess , press & kirshner 1995 ) , it is important to examine whether it depends systematically on environmental properties such as metallicity and age of the progenitor system .
letter _ addresses the issue of whether a difference in the environmental properties is at the basis of the observed range of peak brightness .
there are some observational indications that sne ia are affected by their environment .
the most luminous sne ia seem to occur only in spiral galaxies , while both spiral and elliptical galaxies are hosts for dimmer sne ia . thus the mean peak brightness is dimmer in ellipticals than in spiral galaxies ( hamuy et al .
the sne ia rate per unit luminosity at the present epoch is almost twice as high in spirals as in ellipticals ( cappellaro et al .
moreover , wang , hflich & wheeler ( 1997 ) and riess et al . ( 1999 ) found that the variation of the peak brightness for sne located in the outer regions in galaxies is smaller .
hflich , wheeler , & thielemann ( 1998 ) examined how the initial composition of the wd ( metallicity and the c / o ratio ) affects the observed properties of sne ia .
umeda et al .
( 1999 ) obtained the c / o ratio as a function of the main - sequence mass and metallicity of the wd progenitors . in this _
we suggest that the variation of the c / o ratio is the main cause of the variation of sne ia brightness , with larger c / o ratio yielding brighter sne ia ( @xmath2 2 ) .
we then show that the c / o ratio depends indeed on environmental properties , such as the metallicity and age of the companion of the wd ( @xmath2 3 ) , and that our model can explain most of the observational trends discussed above ( @xmath3 ) .
we then make some predictions about the brightness of sn ia at higher redshift ( @xmath4 ) .
for the progenitors of sne ia , we adopt the single degenerate ( sd ) chandrasekhar mass model , in which an accreting c - o wd explodes when its mass reaches the critical mass @xmath5 ( nomoto , thielemann , & yokoi 1984 ) .
merging of white dwarfs is likely to lead to accretion - induced - collapse rather than thermonuclear explosion ( saio & nomoto 1998 ) .
chandrasekhar mass models can reproduce well the spectrum and the light curves of sne ia , assuming either the explosion is induced by a deflagration or by a delayed detonation ( hflich & khokhlov 1996 ; nugent et al . 1997 ) . in these models ,
the brightness of sne ia is determined mainly by the mass of @xmath6ni synthesized ( @xmath7 ) .
observational data suggest that @xmath7 for most sne ia lies in the range @xmath8 ( e.g. mazzali et al .
this range of @xmath7 can result from differences in the c / o ratio in the progenitor wd as follows . in the deflagration model , a faster propagation of the convective deflagration wave results in a larger @xmath7 .
for example , a variation of the propagation speed by 15% in the w6 w8 models results in @xmath7 values ranging between 0.5 and @xmath9 ( nomoto et al . 1984 ) , which could explain the observations .
the actual propagation of the deflagration depends on the highly non - linear behavior of the turbulent flame ( niemeyer & hillebrand 1995 ) , and so it may be very sensitive to the c / o ratio .
qualitatively , a larger c / o ratio leads to the production of more nuclear energy and buoyancy force , thus leading to a faster propagation and a larger @xmath7 .
quantitatively , further studies of the turbulent flame are necessary to confirm that the expected range of c / o results in the required 15 - 20% variation of the flame speed . in the delayed detonation model , @xmath7
is predominantly determined by the deflagration - to - detonation - transition ( ddt ) density @xmath10 , at which the initially subsonic deflagration turns into a supersonic detonation ( khokhlov 1991 ) .
we reproduce the relation between @xmath10 and @xmath7 in figure 1 by performing hydrodynamical calculations for several values of @xmath10 , as in nomoto et al .
( 1997 ) , kishimoto et al .
( 1999 ) , and iwamoto et al .
the pre - explosive wd model is model c6 ( nomoto et al .
1984 ) , and the flame speed of the initial deflagration is assumed to be 3% of the local sound velocity .
figure 1 shows that if the transition density varies in the range @xmath11 g @xmath12 , the resulting variation of @xmath7 is large enough to explain the observations .
possible mechanisms for ddt to occur have been studied by arnett & livne ( 1994 ) , niemeyer & woosley ( 1997 ) , and khokhlov , oran & wheeler ( 1997 ) : when the deflagration wave reaches a sufficiently low density , @xmath13 g @xmath12 , the turbulent motion associated with the flame may destroy the burning front .
the resulting turbulent mixing between ashes and fuels efficiently heats up the fuel and could produce a region with a very shallow temperature gradient . in such a region ,
successive spontaneous ignitions cause the over - driven deflagration to propagate supersonically .
this may induce a detonation wave if the mass of the region exceeds a critical mass @xmath14 .
this critical mass is quite sensitive to the carbon mass fraction @xmath0(c ) , e.g. @xmath15 and @xmath16 at @xmath17 g @xmath12 for @xmath0(c ) = 1.0 and 0.5 , respectively ( niemeyer & woosley 1997 ) . though the exact value of @xmath10 is still debated , and its dependence on @xmath0(c ) has not been studied , it is not unlikely that for a larger @xmath0(c ) ddt can occur at larger @xmath10 .
hence a larger @xmath0(c ) is likely to result in a larger @xmath10 and @xmath7 . in this _
, therefore , we postulate that @xmath7 and consequently brightness of a sn ia increase as the progenitors c / o ratio increases , as illustrated in figure 1 , where the range of @xmath18 is the result of an @xmath0(c ) range @xmath19 , which is the range of @xmath0(c ) values of our progenitor models described below ( figure 2 ) . the @xmath0(c )
@xmath7 relation we adopt is still only a working hypothesis , which needs to be proved from studies of the turbulent flame during explosion .
hflich et al .
( 1998 ) considered the dependence on @xmath0(c ) but they assumed in their ddt model that @xmath10 and @xmath0(c ) are independent parameters .
they showed that for the same @xmath10 a smaller @xmath0(c ) leads to a slightly brighter sn ia despite a slightly smaller @xmath7 produced , because a smaller fraction of the explosion energy goes into the kinetic energy . in this _
letter _ we assume that @xmath0(c ) is the primary parameter to determine @xmath10 and thus @xmath7 .
the assumed variation of @xmath7 is so large that a smaller @xmath0(c ) yields an intrinsically dimmer sn .
in this section we discuss how the c / o ratio in the wd depends on the metallicity and age of the binary system . the c / o ratio in c+o
wds depends primarily on the main - sequence mass of the wd progenitor and on metallicity . according to the evolutionary calculations for 3@xmath209 @xmath21 stars by umeda et al .
( 1999 ) , the c / o ratio and its distribution are determined in the following evolutionary stages of the close binary .
\1 ) at the end of central he burning in the 3@xmath209 @xmath21 primary star , c / o@xmath22 in the convective core .
the mass of the core is larger for more massive stars .
2 ) after central he exhaustion , the outer c+o layer grows via he shell burning , where c / o@xmath23 ( umeda et al . 1999 ) .
3a ) if the primary star becomes a red giant ( case c evolution ; e.g. van den heuvel 1994 ) , it then undergoes the second dredge - up , forming a thin he layer , and enters the agb phase .
the c+o core mass , @xmath24 , at this phase is larger for more massive stars . for a larger @xmath24
the total carbon mass fraction is smaller .
4a ) when it enters the agb phase , the star greatly expands and is assumed here to undergo roche lobe overflow ( or a super - wind phase ) and to form a c+o wd .
thus the initial mass of the wd , @xmath25 , in the close binary at the beginning of mass accretion is approximately equal to @xmath24 .
3b ) if the primary star becomes a he star ( case bb evolution ) , the second dredge - up in ( 3a ) corresponds to the expansion of the he envelope .
4b ) the ensuing roche lobe overflow again leads to a white dwarf of mass @xmath25 = @xmath24 .
\5 ) after the onset of mass accretion , the wd mass grows through steady h burning and weak he shell flashes , as described in the wd wind model ( hachisu , kato , & nomoto 1996 , 1999 , and hachisu et al .
1999 ; hereafter , hkn96 , hkn99 , and hknu99 , respectively ) .
the composition of the growing c+o layer is assumed to be c / o=1 .
6 ) the wd grows in mass and ignites carbon when its mass reaches @xmath26 , as in the model c6 of nomoto et al .
because of strong electron - degeneracy , carbon burning is unstable and grows into a deflagration for a central temperature of @xmath27 k and a central density of @xmath28 g @xmath12 . at this stage ,
the convective core extends to @xmath29 and the material is mixed almost uniformly , as in the c6 model . in figure 2
we show the carbon mass fraction @xmath0(c ) in the convective core of this pre - explosive wd , as a function of metallicity ( @xmath30 ) and initial mass of the wd before the onset of mass accretion , @xmath24 .
figure 2 reveals that : 1 ) @xmath0(c ) is smaller for larger @xmath24 .
2 ) the dependence of @xmath0(c ) on metallicity is small when plotted against @xmath24 , even though the relation between @xmath24 and the initial stellar mass depends sensitively on @xmath30 ( umeda et al .
1999 ) .
_ metallicity dependent wind during mass accretion _ : in the sd chandrasekhar mass model for sne ia , a wd explodes as a sn ia only when its rate of the mass accretion ( @xmath31 ) is in a certain narrow range ( e.g. , nomoto & kondo 1991 ) .
hkn96 showed that the accreting wd blows a strong wind if @xmath31 exceeds the rate @xmath32 at which steady burning can process the accreted hydrogen into he .
if the wind is sufficiently _ strong _
( i.e. , the wind velocity @xmath33 exceeds the escape velocity @xmath34 of the wd ) , the wd can avoid the formation of a common envelope and increase its mass continuously at a rate @xmath32 by blowing the extra mass away in a wind . in this model , which is adopted in the present study ,
an interesting metallicity effect has been found ( kobayashi et al .
1998 ; hachisu & kato 1999 ) . the wind velocity is higher for larger @xmath35 and larger fe / h because of higher luminosity and larger opacity , respectively . in order to blow sufficiently strong wind ( i.e. , @xmath36 ) , @xmath37 should exceed a certain mass @xmath38 ( fig .
6 of hkn99 ) . as seen in figure 1 of kobayashi et al .
( 1998 ) , @xmath38 is larger for lower metallicity ; e.g. , @xmath39 0.65 , 0.85 , and 0.95 @xmath21 for @xmath40 0.02 , 0.01 , and 0.004 , respectively . in order for a wd to grow its mass at @xmath41
, its initial mass @xmath24 should exceed @xmath38 .
in other words , @xmath38 is the metallicity - dependent minimum @xmath24 required for a wd to become an sn ia ( _ strong wind condition _ in fig.2 ) .
the upper bound @xmath42 is imposed by the condition that carbon should not ignite and is almost independent of metallicity .
as shown in figure 2 , the range of @xmath24 can be converted into a range of @xmath0(c ) . from this
we find the following metallicity dependence for @xmath0(c ) : 1 ) the upper bound of @xmath0(c ) , which is determined by the lower limit on @xmath24 imposed by the metallicity - dependent conditions for a strong wind , e.g. , @xmath0(c ) @xmath43 , 0.46 and 0.41 , for @xmath30=0.02 , 0.01 , and 0.004 , respectively .
2 ) on the other hand , the lower bound , @xmath0(c ) @xmath44 , does not depend much on @xmath30 , since it is imposed by the maximum @xmath24 .
3 ) assuming the relation between @xmath7 and @xmath45(c ) given in figures 1 and 2 , our model predicts the absence of brighter sne ia in lower metallicity environment .
_ age effects _ :
in our model , the age of the progenitor system also constrains the range of @xmath0(c ) in sne ia . in the sd scenario ,
the lifetime of the binary system is essentially the main - sequence lifetime of the companion star , which depends on its initial mass @xmath46 . hknu99 and hkn99 have obtained a constraint on @xmath46 by calculating the evolution of accreting wds for a set of initial masses of the wd ( @xmath47 ) and of the companion ( @xmath46 ) , and the initial binary period ( @xmath48 ) . in order for the wd mass to reach @xmath49 , the donor star should transfer enough material at the appropriate accretion rates .
the donors of successful cases are divided into two categories : one is composed of slightly evolved main - sequence stars with @xmath50 ( for @xmath30=0.02 ) , and the other of red - giant stars with @xmath51 ( for @xmath30=0.02 ) ( hkn99 , hknu99 ; also li & van den heuvel 1997 ) .
if the progenitor system is older than 2 gyr , it should be a system with a donor star of @xmath52 in the red - giant branch .
systems with @xmath53 become sne ia in a time shorter than 2 gyr .
likewise , for a given age of the system , @xmath46 must be smaller than a limiting mass .
this constraint on @xmath46 can be translated into the presence of a minimum @xmath24 for a given age , as follows : for a smaller @xmath46 , i.e. for the older system , the total mass which can be transferred from the donor to the wd is smaller . in order for @xmath37 to reach @xmath49 , therefore , the initial mass of the wd , @xmath54 , should be larger .
this implies that the older system should have larger minimum @xmath24 as indicated in figure 2 .
using the @xmath0(c)@xmath55 and @xmath56x@xmath57 relations ( figs . 1 and 2 )
, we conclude that wds in older progenitor systems have a smaller @xmath0(c ) , and thus produce dimmer sne ia .
the first observational indication which can be compared with our model is the possible dependence of the sn brightness on the morphology of the host galaxies .
hamuy et al .
( 1996 ) found that the most luminous sne ia occur in spiral galaxies , while both spiral and elliptical galaxies are hosts to dimmer sne ia .
hence , the mean peak brightness is lower in elliptical than in spiral galaxies . in our model ,
this property is simply understood as the effect of the different age of the companion . in spiral galaxies ,
star formation occurs continuously up to the present time .
hence , both wd+ms and wd+rg systems can produce sne ia . in elliptical galaxies ,
on the other hand , star formation has long ended , typically more than 10 gyr ago .
hence , wd+ms systems can no longer produce sne ia . in figure 3
we show the frequency of the expected sn i for a galaxy of mass @xmath58 for wd+ms and wd+rg systems separately as a function of @xmath24 . here
we use the results of hkn99 and hknu99 , and the @xmath59(c ) and @xmath60(c ) relations given in figure 2 . since a wd with smaller @xmath24
is assumed to produce a brighter sn ia ( larger @xmath61 ) , our model predicts that dimmer sne ia occur both in spirals and in ellipticals , while brighter ones occur only in spirals . the mean brightness is smaller for ellipticals and the total sn ia rate per unit luminosity is larger in spirals than in ellipticals .
these properties are consistent with observations .
the second observational suggestion is the radial distribution of sne ia in galaxies
. wang et al .
( 1997 ) and riess et al .
( 1998 ) found that the variation of the peak brightness for sne ia located in the outer regions in galaxies is smaller .
this behavior can be understood as the effect of metallicity . as shown in figure 2 , even when the progenitor age is the same , the minimum @xmath24 is larger for a smaller metallicity because of the metallicity dependence of the wd winds .
therefore , our model predicts that the maximum brightness of sne ia decreases as metallicity decreases . since the outer regions of galaxies are thought to have lower metallicities than the inner regions ( zaritsky , kennicutt , huchra 1994 ; kobayashi & arimoto 1999 ) , our model is consistent with observations .
wang et al .
( 1997 ) also claimed that sne ia may be deficient in the bulges of spiral galaxies .
this can be explained by the age effect , because the bulge consists of old population stars .
we have suggested that @xmath0(c ) is the quantity very likely to cause the diversity in @xmath7 and thus in the brightness of sne ia .
we have then shown that our model predicts that the brightness of sne ia depends on the environment , in a way which is qualitatively consistent with the observations .
further studies of the propagation of the turbulent flame and the ddt are necessary in order to actually prove that @xmath0(c ) is the key parameter .
our model predicts that when the progenitors belong to an old population , or to a low metal environment , the number of very bright sne ia is small , so that the variation in brightness is also smaller . in spiral galaxies ,
the metallicity is significantly smaller at redshifts @xmath1 , and thus both the mean brightness of sne ia and its range tend to be smaller . at @xmath62 sne
ia would not occur in spirals at all because the metallicity is too low . in elliptical galaxies , on the other hand , the metallicity at redshifts @xmath63 is not very different from the present value .
however , the age of the galaxies at @xmath64 is only about 5 gyr , so that the mean brightness of sne ia and its range tend to be larger at @xmath1 than in the present ellipticals because of the age effect .
we note that the variation of @xmath0(c ) is larger in metal - rich nearby spirals than in high redshift galaxies .
therefore , if @xmath0(c ) is the main parameter responsible for the diversity of sne ia , and if the lcs method is confirmed by the nearby sne ia data , the lcs method can also be used to determine the absolute magnitude of high redshift sne ia .
we wish to thank p. mazzali , k. iwamoto and n. kishimoto for useful discussion and suggestions .
this work has been supported in part by the grant - in - aid for scientific research ( 0980203 , 09640325 ) and coe research ( 07ce2002 ) of the ministry of education , science , culture and sports in japan .
kishimoto , n. , iwamoto , k. , umeda , h. , nomoto , k. , brachwitz , f. , hix , w. r. , & thielemann , f - k . , 1999 , in origin of matter and evolution of galaxies 97 ,
eds . s. kubono et al .
( singapore : world scientific ) , 378
|
observations suggest that the properties of type ia supernovae ( sne ia ) may depend on environmental characteristics , such as morphology , metallicity , and age of host galaxies .
the influence of these environmental properties on the resulting sne ia is studied in this paper .
first it is shown that the carbon mass fraction @xmath0(c ) in the c+o white dwarf sn ia progenitors tends to be smaller for lower metallicity and older the binary system age .
it is then suggested that the variation of @xmath0(c ) causes the diversity in the brightness of sne ia : a smaller @xmath0(c ) leads to a dimmer sn ia .
further studies of the propagation of the turbulent flame are necessary to confirm this relation .
our model for the sn ia progenitors then predicts that when the progenitors belong to an older population or to a low metallicity environment , the number of bright sne ia is reduced , so that the variation in brightness among the sne ia is also smaller .
thus our model can explain why the mean sn ia brightness and its dispersion depend on the morphology of the host galaxies and on the distance of the sn from the center of the galaxy .
it is further predicted that at higher redshift ( @xmath1 ) both the the mean brightness of sne ia and its variation should be smaller in spiral galaxies than in elliptical galaxies .
these variations are within the range observed in nearby sne ia .
in so far as the variation in @xmath0(c ) is the most important cause for the diversity among sne ia , the light curve shape method currently used to determine the absolute magnitude of sne ia can be applied also to high redshift sne ia .
|
the short wavelength spectrometer ( sws ) on board the infrared space observatory ( iso ) has allowed us , for the first time , to observe the pure rotational spectrum of h@xmath0 and study the otp ratio in warm ( t@xmath1 100 - 1000 k ) regions .
previous to iso observations , the otp ratio of h@xmath0 had been estimated towards regions heated by uv radiation and shocks to temperatures @xmath2 2000 k using the relative strengths of the vibrational lines . in outflow regions where shock excitation is the main heating mechanism , observations of the vibrational lines typically reveal otp ratios of @xmath1 3 ( smith , davis & lioure 1997 ) .
this is the expected value in thermodynamic equilibrium for gas with temperatures @xmath2 200 k. however , towards the regions heated mainly by the uv radiation ( pdrs ) , the vibrational lines give otp values in the range 1.5 - 2 ( see e.g. chrysostomou et al .
the pure rotational lines of h@xmath0 have been detected by iso toward several galactic regions ( s140 : timmermann et al .
1996 ; cepheus a west : wright et al .
1996 ; bd+40@xmath34124 : wesselius et al . 1996
; hh54 : neufeld et al . 1998 ) .
the observations towards all these regions except hh54 are consistent with the h@xmath0 rotational lines arising in gas with temperature @xmath2 200 k and with an otp ratio of 3 .
this value is not consistent with those derived from the vibrational lines in dense pdrs but is in agreement with theoretical predictions .
recently sternberg & neufeld ( 1999 ) have argued that the low otp ratios measured from vibrational lines do not represent the actual otp ratio in pdrs but it is simply a consequence of the optical depth effects in the fluorescent pumping of the vibrational lines . the non - equilibrium otp ratio measured towards hh54 has been interpreted as arising in hot - shocked excited gas that has not reached the equilibrium yet . in this interpretation
, the observed otp ratio would be the legacy of an earlier stage in the thermal history of the gas when the gas temperature was 90 k. in this _
letter _ we report the observations of the h@xmath0 rotational lines towards the prototypical pdr associated with the reflection nebula ngc 7023 .
an otp ratio in the range 1.5 - 2 is derived from our observations .
this is the first detection of an otp ratio @xmath4 3 in a pdr based on the pure h@xmath0 rotational lines .
the observations of the s(0 ) , s(1 ) , s(2 ) , s(3 ) , s(4 ) and s(5 ) rotational lines of h@xmath0 towards the peak of the pdr associated with ngc 7023 ( r.a . ( 2000 ) : , dec(2000 ) : ) .
were made using the short - wavelength spectrometer ( sws ) ( de grauuw et al .
1996 ) on board the infrared space observatory ( iso ) ( kessler et al . 1996 ) during revolution 514 with a total on - target time of 5149 s. at the spectral resolution of this mode ( @xmath5/@xmath6@xmath5 @xmath1 1000 - 2000 ) all the observed lines are unresolved .
data reduction are carried out with version 7.0 of the off line processing routines and the sws interactive analysis at the iso spectrometer data center at mpe .
further analysis has been made using the isap software package .
the uncertainities in the calibration are of 15 % , 25 % , 25 % , 25 % , 20% and 30% for the s(5 ) , s(4 ) , s(3 ) , s(2 ) , s(1 ) and s(0 ) lines respectively ( salama et al .
1 we present the spectra of the s(0 ) , s(1 ) , s(2 ) , s(3 ) , s(4 ) and s(5 ) h@xmath0 lines towards the peak of the pdr associated with ngc 7023 .
the observed intensitites and some interesting observational parameters are presented in table 1 .
the data have been corrected for dust attenuation using the extinction curve of draine & lee ( 1984 ) , and a value of 0.43 for the dust opacity at 0.55 @xmath7 .
this value has been derived from the iso lws01 spectrum ( fuente et al . 1999 ,
hereafter paper ii ) .
the extinction for the s(0 ) , s(1 ) , s(2 ) , s(3 ) , s(4 ) and s(5 ) lines is very small and amounts to 0.01 , 0.02 , 0.02 , 0.04 , 0.02 and 0.01 respectively . fig .
2 shows the rotational diagram for the h@xmath0 corrected for extinction effects and assuming extended emission .
( note that the errors in fig .
2 are entirely dominated by the calibration uncertainities . )
this diagram shows that the ortho - h@xmath0 levels have systematically lower n@xmath8/g@xmath8 values ( where n@xmath8 and g@xmath8 are the column densities and degeneracies of the upper levels of the transitions ) than the adjacent j1 and j+1 para - h@xmath0 levels producing a @xmath9zig - zag " distribution .
in fact , the ortho - h@xmath0 levels seem to define a curve which is offset from that of the para - h@xmath0 levels ( see fig .
this is the expected trend if the otp ratio is lower than 3 .
the offset between the two set of data is systematic and seems to show the trend of being larger for the low energy levels ( s(0 ) and s(1 ) transitions ) , than for the high energy levels ( s(2 ) , s(3 ) , s(4 ) and s(5 ) transitions ) .
the offset between the ortho- and para- h@xmath0 curves is larger than the observational errors and can not be due to the different apertures for the different lines .
for the case of a point - like source we would have to correct by a factor of 1.93 the intensity of the s(0 ) line and by a factor of 1.35 the intensities of the s(1 ) and s(2 ) lines . in this case , the offset between the para- and ortho- curves would increase . based on the spatial distribution of the hi(21 cm ) line ( fuente et al .
1998 ) and those of the cii ( 158 @xmath7 ) and oi ( 63 @xmath7 ) lines ( chokshi et al . 1988 , paper ii ) , we will assume in the following a beam filling factor of 1 for all the observed h@xmath0 lines .
although the exact shape of the rotational diagram depends on the excitation conditions of the region , the @xmath9zig - zag `` features can not be explained by any model which assumes an equilibrium otp ratio . in thermodynamic equilibrium , the rotational temperature between an ortho- ( para- ) h@xmath0 level and the next j+1 level should always increase with the energy of the upper level giving rise to a smooth curve in the rotational diagram regardless of the temperature profile of the region .
a @xmath9zig - zag '' distribution implies that for some pair of levels the rotational temperature decreases with the energy of the upper level .
one can only get this effect with a non - equilibrium otp ratio .
we have compared our data with the models by burton et al .
( 1992 ) in order to estimate the incident uv field , density and otp ratio .
burton et al . ( 1992 ) assumed a constant value of 3 for the otp ratio in their calculations .
the best fit to our data is for an incident uv field of g@xmath10 = 10@xmath11 in units of the habing field and a density of n = 10@xmath12 @xmath13 , but it systematically underestimates the intensity of the para - h@xmath0 transitions and overestimates the intensities of the ortho - h@xmath0 transitions ( see open triangles in fig .
this is the expected behavior if one assumes an otp ratio of 3 and the actual otp ratio in the region is @xmath4 3 .
we have corrected the line intensities predicted by the model for different values of the otp ratio assuming that the total amount of h@xmath0 molecules at a given temperature and the line ratios between levels of the same symmetry are not affected by the otp ratio .
these assumptions are correct for the low - j transitions where the collisional excitation dominates . in fig .
3 we compare our data with the predicted diagram for an otp ratio of 3 which correspond to the model without any correction ( open triangles ) , 2 ( open circles ) and 1 ( open squares ) .
the diagram for an otp ratio of 3 clearly does not fit any of our observational points .
we find the best fit to our data with an otp ratio of 2 . in this case , the predicted intensities for the s(1 ) , s(2 ) and s(3 ) lines are in agreement with the observed values .
only the predicted intensity for the s(0 ) line is not consistent with the observations . to fit the intensity of the s(0 ) line
it is necessary to assume an otp ratio of @xmath14 1.5 ( see the case otp = 1 in fig .
3 ) , but in this case , we will have a worse fit for the s(1 ) , s(2 ) and s(3 ) lines than for an otp ratio of 2 . as discussed in section 4
, this suggests a possible variation of the otp ratio with the temperature . the cooler gas emitting only in the s(0 ) line could have an otp ratio of @xmath4 1.5 , while the warmer gas emitting in the s(1 ) , s(2 ) and s(3 ) lines could have an otp ratio of 2 . however , to establish this variation unambigously , it is necessary to have a very accurate knowledge of the excitation conditions in the region . to be conservative
, we will conclude that the otp ratio is in the range 1.5 - 2 in this pdr .
martini et al .
( 1997 ) found an otp ratio of [email protected] using the near - ir h@xmath0 vibrational lines towards their position 1 which is 20@xmath16 offset from ours .
taking into account that because of the optical depth effects in the vibrational lines this value is just a lower limit to the actual otp ratio , it proves that the otp ratio is close to 3 for the gas with kinetic temperatures t@xmath17 @xmath2 2000 k. this difference between the otp ratios derived from the rotational and vibrational lines argues in favor of a variaton of the otp ratio with the kinetic temperature .
since the otp ratio in this source is different from the equilibrium value , the rotation temperature between an ortho- and a para- level does not represent an estimate of the gas kinetic temperature . to estimate the gas kinetic temperature we have calculated the rotation temperature between levels of the same symmetry . for the para - h@xmath0 levels , we have derived a rotation temperature of @xmath1 290 k from the s(0 ) and s(2 ) lines and of @xmath1 500 k from the s(2 ) and s(4 ) lines .
for the ortho - h@xmath0 levels , the derived temperature is @xmath1 440 from s(1 ) and s(3 ) lines and @xmath1 700 k from the s(3 ) and s(5 ) lines .
based on these calculations we conclude that the otp ratio is @xmath1 1.5 - 2 in the gas with kinetic temperatures @xmath1 300 - 700 k. we derive a total h@xmath0 column density of 5 10@xmath18 @xmath19 assuming a rotation temperature of 290 k between the j = 0 and j = 2 para - levels and of 440 k between the j = 1 and j = 3 ortho - levels .
based on the h@xmath0 rotational lines data we have derived an otp ratio in the range of 1.5 - 2 for the gas with kinetic temperatures @xmath1 300 - 700 k towards the reflection nebula ngc 7023 .
this is the second object with a non - equilibrium otp ratio measured from the h@xmath0 pure rotational lines .
the first non - equilibrium otp ratio was detected towards the outflow source hh54 , and interpreted as arising in shock - heated gas that has not reached the equilibrium yet .
this interpretation is not plausible for ngc 7023 .
the high dust temperature , t@xmath20@xmath21 40 k , and the detection of the siii ( 34.8 @xmath7 ) and cii ( 157.7 @xmath7 ) lines ( paper ii ) proves the existence of a pdr in this region .
although a shock component might also exist ( see martini et al .
1997 ) , the heating is dominated by uv photons . it is then expected that most of the warm h@xmath0 arises in the pdr .
a non - equilibrium value of the otp ratio is not expected for the physical conditions prevailing in a dense pdr .
the initial otp ratio after the h@xmath0 formation is very uncertain . because of the large exothermicity of the formation process
, it is expected to be 3 .
however , this otp ratio can change if the h@xmath0 molecules remain on the grain surface long enough to reach the equilibrium at the dust temperature . in this case , the otp ratio after h@xmath0 formation will be that of the equilibrium at the grain temperature .
after the ejection of the h@xmath0 molecules to the gas phase , exchange reactions with h and h@xmath22 change the otp ratio until achieving the equilibrium at the gas temperature . for the gas temperatures traced by the h@xmath0 rotational lines ( @xmath2 200 k ) ,
the equilibrium otp ratio is 3 .
one possibility to explain the low otp ratio observed in ngc 7023 is to suppose that the otp ratio in the h@xmath0 formation is lower than 3 , and once in the gas phase , the h@xmath0 molecules are destroyed before attaining the equilibrium value at the gas temperature .
the dust temperature in a pdr with g@xmath10 = 10@xmath11 and n = 10@xmath12 @xmath13 is @xmath4 100 k. in particular , the dust temperature measured towards the pdr peak in ngc 7023 using our lws01 spectrum is 40 k which corresponds to an equilibrium otp ratio of @xmath1 0.1 .
the otp conversion in the atomic region is dominated by h - h@xmath0 collisions with a rate of 10@xmath23 s@xmath24 where @xmath25 is the atomic hydrogen density ( see sternberg & neufeld 1999 and references therein ) .
the hydrogen density derived from the hi image published by fuente et al .
( 1998 ) is @xmath1 5 10@xmath26 @xmath13 . a similar value ( @xmath1 0.5 - 1 10@xmath11 @xmath13 ) was derived by chokshi et al . ( 1988 ) based on the oi and cii lines .
however , densities larger than 10@xmath27 @xmath13 are derived from molecular data ( fuente et al .
1996 , lemaire et al .
1996 , gerin et al .
the density of atomic hydrogen in the region with t@xmath17 @xmath1 300 k is expected to be about an order of magnitude lower than the h@xmath0 density . then we assume an atomic hydrogen density of @xmath1 10@xmath11 @xmath13 in our calculations . with this value ,
the otp conversion rate due to h@xmath0 - h collisions is 10@xmath28 s@xmath24 . at the cloud surface
the unshielded photodissociation rate is @xmath1 5 10@xmath29 s@xmath24 @xmath1 5 10@xmath30 s@xmath24 , i.e. , 2 orders of magnitude larger than the otp conversion rate .
then , in the cloud surface , before self - shielding is important ( a@xmath31 @xmath14 0.3 mag , t@xmath17@xmath2 500 k ) , an otp ratio lower than 3 can be explained by assuming that the otp ratio in the h@xmath0 formation is the equilibrium value at the grain temperature .
deeper into the molecular cloud , when h@xmath0 is self - shielded , one can consider that the h@xmath0 destruction rate is similar to the h@xmath0 reformation rate which is given by @xmath1 3 10@xmath32 s@xmath24 . in this region , a lower limit to the otp conversion rate is given by the conversion rate due to h@xmath0 - h@xmath22 collisions which is @xmath1 10@xmath32 s@xmath24 ( sternberg & neufeld 1999 ) .
the otp conversion rate is of the same order than the destruction rate and the otp ratio is expected to be close to the equilibrium . according to these estimates one expects to have an otp ratio close to 3 at low temperatures and a non - equilibrium otp ratio at higher temperatures .
the contrary trend is derived from our observations .
the high energy s(1 ) , s(2 ) and s(3 ) lines are fitted with an an otp ratio of 2 while the low energy s(0 ) line is better fitted with an otp ratio of 1.5 .
another possibility to explain the non - equilibrium otp value in ngc 7023 is to consider the case of a dynamic pdr , i.e. the dissociation front is advancing into the molecular cloud . in this case
we do not have to assume a non - equilibrium otp ratio in the h@xmath0 formation .
the pdr is being fed continuously by the cool gas of the molecular cloud in which the equilibrium value of the otp ratio is lower than 3 .
this gas is heated by the stellar uv radiation to temperatures @xmath2 200 k but leaves the pdr before attaining the equilibrium otp ratio at this temperature . in this scenario ,
the gas which is expected to have an otp ratio smaller than the equilibrium value is the gas that has more recently been incorporated into the pdr , which is also the gas at lowest temperature .
the otp ratio is expected to increase and reach values close to 3 at high temperatures .
this behavior is consistent with the trend observed in our data . as discussed above
, the otp h@xmath0 conversion is mainly due to h - h@xmath0 collisions with a conversion rate of @xmath1 10@xmath28 s@xmath24 . to have a significant fraction of the gas with a non - equilibrium otp ratio the photodissociation front must advance about 1 - 2 mag in the conversion time , i.e.
, it must penetrate into the molecular cloud at a velocity of @xmath1 10@xmath33 kms@xmath24 .
skinner et al . (
1993 ) proposed the existence of an anisotropic ionized wind associated with this star based on radio continuum observations .
later , fuente et al . (
1998 ) detected an hi outflow with a velocity of @xmath1 7.5 kms@xmath24 .
they proposed that this outflow is formed when the gas that has been photodissociated by the uv radiation , is accelerated by the stellar winds along the walls of a biconical cavity .
the velocity of the hi outflow can not account for the velocity at which the photodissociation front must advance into the molecular cloud to have an otp ratio lower than 3 unless the otp conversion rates are severely overestimated ( by a factor of @xmath2 10 ) and/or the density of the outflow is @xmath2 10@xmath27 @xmath13 . in spite of this problem
, a photodissociation front penetrating into the cloud because of the outflow seems to be the most plausible explanation for the low otp ratio measured in this region .
the existence of an outflow can also explain the difference between this region and other pdrs like s140 in which an otp ratio of 3 has been derived from the h@xmath0 rotational lines .
we would like to thank the referee for his / her helpful comments .
this work has been partially supported by the spanish dges under grant number pb93 - 0048 and the spanish pnie under grant number esp97 - 1490-e .
n.j.r - f acknowledges conserjeria de educacin y cultura de la comunidad de madrid for a pre - doctoral fellowship .
l ccc + & & & + & & & + + h@xmath0 s(5 ) & 6.909 & 14 x 20 & [email protected] 10@xmath34 + h@xmath0 s(4 ) & 8.025 & 14 x 20 & [email protected] 10@xmath34 + h@xmath0 s(3 ) & 9.665 & 14 x 20 & [email protected] 10@xmath34 + h@xmath0
s(2 ) & 12.279 & 14 x 27 & [email protected] 10@xmath34 + h@xmath0 s(1 ) & 17.035 & 14 x 27 & [email protected]
10@xmath34 + h@xmath0 s(0 ) & 28.219 & 20 x 27 & [email protected] 10@xmath34 +
|
we have observed the s(0 ) , s(1 ) , s(2 ) , s(3 ) , s(4 ) and s(5 ) rotational lines of molecular hydrogen ( h@xmath0 ) towards the peak of the photodissociation region ( pdr ) associated with the reflection nebula ngc 7023 .
the observed h@xmath0 line ratios show that they arise in warm gas with kinetic temperatures @xmath1 300 - 700 k. however , the data can not be fitted by an ortho- to para- ( otp ) ratio of 3 .
an otp ratio in the range @xmath1 1.5 - 2 is necessary to explain our observations .
this is the first detection of a non - equilibrium otp ratio measured from the h@xmath0 pure - rotational lines in a pdr .
the existence of a dynamical pdr is discussed as the most likely explanation for this low otp ratio .
# 1 = " 10i ' 2h@xmath0
|
the interacting one - dimensional electron gas ( 1deg ) has intrigued physicists since the pioneering work of tomonaga@xcite . through the collective effort of many of them , and especially due to the advancement of the bosonization technique@xcite , a large number of its properties have been discovered .
the past two decades have witnessed an experimental effort to identify and study , using various probes , quasi - one - dimensional systems . among the materials
studied one finds organic compounds such as polyacetylene@xcite and the bechgaard salts@xcite as well as several families of inorganic quasi - one - dimensional materials@xcite ( for example , the transition metal bronzes ) .
the edge states in the quantum hall effect constitute yet another realization of one - dimensional interacting systems@xcite .
progress made in nanoscale fabrication techniques have made it possible to artificially define one - dimensional channels in two - dimensional electronic heterostructures@xcite .
most recently , experimental evidence has emerged for the existence of stripe phases in doped antiferromagnets , such as the copper - oxide high temperature superconductors@xcite . in these phases the doped holes segregate into quasi - one - dimensional metallic regions embedded in a predominantly antifferomagnetic background . in recent years angle resolved photoemission spectroscopy ( arpes ) has matured into a powerful experimental method for probing the single particle properties of strongly correlated systems .
arpes measurements on the bechgaard salts@xcite , the transition metalchalcogenides @xcite , the blue bronzes@xcite and the cuprates@xcite have provided evidence in favor of correlated one - dimensional physics although full agreement with the predicted theoretical picture is still lacking .
the zero temperature correlation functions of the gapless tomonaga - luttinger liquid , and in particular the single hole spectral function which is measured by arpes , have been calculated previously for the spinless@xcite as well as for the spinfull case@xcite .
extending these results to finite temperatures is desirable on the following grounds .
first , real experiments are carried out at finite temperatures and should be contrasted with theoretical predictions relevant for such conditions .
secondly , the tomonaga - luttinger liquid is a quantum critical system which consequently exhibits scaling behavior .
it is a rare example where not only the scaling exponents but indeed the entire scaling functions can be computed explicitly . in section
ii we obtain the scaling form of the single hole spectral function and the spin dynamic structure factor of the spinfull tomonaga - luttinger liquid . closed form analytical expressions are derived for a number of special cases .
these results complement previous numerical evaluations of the finite temperature spectral functions @xcite and allow for easy investigation of their properties in few limits which are of physical interest .
integrals of these spectral functions are considered as well .
when backward scattering of electrons becomes relevant the 1deg develops a spin gap and is described by the luther - emery liquid@xcite ( umklapp processes may create a charge gap ) . while the spectrum of the luther - emery liquid can be readily derived at the special `` free fermion '' point @xmath0 , the evaluation of electronic correlation functions there is non - trivial due to their highly non - local form in terms of the pseudo- fermions .
several authors have made progress in this direction , see refs .
[ voitref],[wiegmannref ] , and in section iii we extend and correct their study and obtain exact expressions for the single hole spectral function at the free fermion point for temperatures much smaller than the spin gap .
some technical details are relegated to appendices .
the tomonaga luttinger model embodies the low energy and long wavelength physics of the 1deg .
it is composed of two branches of left @xmath1 and right @xmath2 moving massless dirac fermions constructed around the left and right fermi points of the 1deg .
we will consider the case where the fermions carry spin @xmath3 and denote by @xmath4 their spin polarization . in the absence of backward scattering and umklapp
processes the hamiltonian density is given by @xmath5 @xmath6 where @xmath7 is the noninteracting fermi velocity and @xmath8 with being the pauli matrices . despite its appearance , the last term in ( [ hamiltonian ] ) does not break the @xmath9 spin symmetry , since @xmath10 ; the @xmath11 term is the only term which breaks this symmetry . in order to compute the single particle properties of the model one uses the bosonization identity@xcite @xmath12 \ ; , \ ] ] which expresses the fermionic fields in terms of self - dual fields @xmath13 obeying @xmath14=-i\pi\delta_{\eta,\eta'}\delta_{\sigma,\sigma'}{\rm sign}(x - x')$ ] .
the klein factors @xmath15 are responsible for reproducing the correct anti - commutation relations between different fermionic species and @xmath16 is a short distance cutoff that is taken to zero at the end of the calculation . here and throughout the paper we consider the limit where the size of the system @xmath17 is taken to infinity and correspondingly ignore terms of the order @xmath18 ( for a discussion of finite size effects in the finite temperature case see refs .
[ mattsson97ref]-[eggert97ref ] ) .
the demonstration of the celebrated charge - spin separation in the model is facilitated by expressing @xmath19 in terms of the bosonic fields @xmath20 and their conjugated momenta @xmath21 @xmath22 \ ; , \ ] ] in terms of which the charge and spin densities are given by @xmath23 the decomposition ( [ decomp ] ) also facilitates the diagonalization of the hamiltonian .
it becomes a sum of two independent pieces describing noninteracting charge and spin density waves which are the elementary excitations of the system @xmath24 \ ; .\ ] ] the velocities of the collective modes are @xmath25 and the parameters @xmath26 , which determine the power - law behavior of the correlation functions , read @xmath27 the effective parameters that enter the hamiltonian depend on the specifics of the model for which ( [ hamiltonian ] ) is a low energy fixed point .
we note , however , that spin - rotation invariance dictates @xmath28 and consequently @xmath29 . the bosonized expression for the fermion field operators ( [ bidentity ] ) and the fact that the theory reduces to that of free bosons make it straightforward to derive expressions for the space - time response functions ( which we denote with a tilde ) .
we will focus on the finite temperature single hole green function @xmath30 we compute this green function , rather than the more usual time ordered or retarded green functions , because it includes only the one - hole states .
this fact makes it relevant to arpes whose cross section is directly proportional to the fourier transform of @xmath31 .
other green functions can be easily obtained from it , since for the model ( [ hamiltonian ] ) @xmath32 .
we will also consider the @xmath33 component of the transverse spin dynamic structure factor , which is measured by polarized neutron scattering @xmath34 where @xmath35 and the @xmath36 are the pauli matrices . the tomonaga - luttinger liquid ( [ hamiltonian ] ) is a quantum critical system .
it also exhibits spin - charge separation .
this implies a scaling form for the response functions with separate spin and charge pieces .
specifically one finds@xcite @xmath37 where we introduced the thermal lengths ( @xmath38 ) @xmath39 and the exponents @xmath40 defined so that @xmath41 for noninteracting fermions . since the spin and charge sectors are formally invariant under separate lorentz transformations the functions @xmath42 also spilt into right and left moving parts @xmath43 where @xmath44^{-\gamma } \ ; .\ ] ] similarly , for the spin correlation function one finds@xcite @xmath45 where @xmath46 here we introduced the exponents @xmath47 we note that the perpendicular component of the spin dynamic structure factor @xmath48 is obtained from the result for the transverse part ( [ scalesr ] ) after multiplying it by an overall factor of @xmath3 and using the exponents @xmath49 instead of ( [ betadef ] ) .
this holds true also for the spectral functions calculated in the following subsections . of coarse , in the spin rotation invariant
case @xmath50 .
it is a general feature of the field - theoretic approach to this problem that relatively simple expressions are obtained for the space - time dependent correlation functions . for comparison with experiments , however , we are typically interested in the fourier transform of these correlation functions . conceptually , this is simple , and indeed , to evaluate the fourier transform , we need only to perform a two dimensional integral .
however , it is not generally simple to carry out this calculation analytically .
below we consider the cases where this can be done . as we noted before , the critical nature of the model leads to a scaling form for the spectral functions .
a simplifying feature introduced by the spin - charge separation is the ability to express these scaling functions as a convolution of spin and charge parts @xmath51}\tilde g^<_{\eta}(x , t;t ) \\ \nonumber & = & \frac{a}{(2\pi)^3 v_c}\left(\frac{a}{\lambda_{t , c}}\right)^{2\gamma_c- \frac{1}{2}}\left(\frac{a}{\lambda_{t , s}}\right)^{2\gamma_s-\frac{1}{2 } } \\ \nonumber & \times&\int_{-\infty}^{\infty } dq\ , d\nu\ , g_c(q,\nu)\,g_s ( \tilde k - rq,\tilde\omega-\nu ) \\\end{aligned}\ ] ] where we introduce the velocity ratio @xmath52 , define the scaling variables @xmath53 and @xmath54 @xmath55 , the fourier transform of @xmath56 , is evaluated in appendix a , where some of its properties are listed as well .
for non - integer values of @xmath57 it is given by @xmath58 \ ; , \ ] ] where @xmath59 is the beta function . in a similar fashion
we obtain for the spin susceptibility @xmath60}\tilde { \cal s}(x , t;t ) \\ \nonumber & = & \frac{1}{(2\pi)^4 v_c}\left(\frac{a}{\lambda_{t , c}}\right)^ { 2\beta_c-1}\left(\frac{a}{\lambda_{t , s}}\right)^{2\beta_s-1 } \\
\nonumber & \times & \int_{-\infty}^{\infty } dq\ , d\nu\ , c_c(q,\nu)\,c_s ( \tilde k - rq,\tilde\omega-\nu ) \\\end{aligned}\ ] ] where @xmath61 henceforth @xmath62 is measured relative to @xmath63 and @xmath33 when computing @xmath64 and @xmath65 , respectively .
we also define the fourier transform with respect to @xmath66 .
this has the effect of making @xmath62 positive outside the fermi surface and negative inside ; thus giving the same expression for the spectral function of left and right moving holes .
here we would like to note the non - commutativity of the limits @xmath67 and @xmath68 in calculating the zero temperature dc response of the system . due to the scaling form of the spectral functions taking the zero temperature limit
first gives a result which is determined by the @xmath69 behavior of the scaling functions , while it is the @xmath70 behavior which is relevant in case @xmath62 or @xmath71 is set to zero from the outset .
further analytic progress in calculating the spectral functions can be achieved in the following cases : since there is no symmetry between the charge and spin sectors , we generally expect that the spin and charge velocities are different .
this greatly complicates the explicit calculation of the fourier transforms , as the model is not `` lorentz invariant '' under transformations that involve both the spin and charge sectors .
such an invariance is restored if @xmath73 . in this case
the correlation functions have the same form as those of a related model of spinless electrons , for which the zero temperature spectral functions have been calculated by luther and peschel @xcite .
the major simplification that follows directly from the fact that the charge and spin velocities are equal is that the two - dimensional fourier transform reduces to a product of two one - dimensional transforms - one for the right moving piece and one for the left moving piece .
specifically one finds , with @xmath74 , that @xmath75 a similar analytic form can be obtained for the spin correlation function @xmath76 where @xmath77 .
we already noted that when the system is invariant under spin rotations @xmath29 ( @xmath79 ) . at this important special point
there is no mixing between left and right moving spin excitations . as a result
the expression for the hole spectral function is simplified , due to the appearance of the factor @xmath80 in the spin part of ( [ gk ] ) , and can be expressed as a single integral @xmath81 h_{\gamma_c}\left[\frac{\tilde\omega-\tilde k}{2 } -(1-r)q\right ] \ ; . \\\end{aligned}\ ] ] when @xmath82 as well , the integral in ( [ gkgamma0 ] ) is trivial and we obtain @xmath83 ( 3.2,4.2)(0,-1.2 ) ( -0.15,-1 ) the case @xmath84 is unique in the sense that there is no mixing between left and right moving excitations . as a result
there are severe kinematic constraints on @xmath64 that make it non - vanishing , in the zero temperature limit , only in a wedge in @xmath85 plane defined by the lines @xmath86 and @xmath87 for @xmath88 . while @xmath29 reflects a symmetry of the problem it is unlikely that @xmath82 ( @xmath89 ) is realized in any interacting system .
however , if the effective interactions are not too strong then @xmath90 may be small .
for instance , for the hubbard model@xcite , even in the @xmath91 limit , @xmath92 .
for such systems we expect ( [ gk00 ] ) to be qualitatively correct with the exception of the behavior of @xmath64 outside the above mentioned wedge .
in contrast to the result for @xmath89 the @xmath93 support of @xmath64 , for non - zero @xmath90 , extends up to the line @xmath94 with @xmath95 . nevertheless , if the mixing between left and right moving charge excitations ( i.e @xmath90 ) is small the amplitude of @xmath64 outside the wedge is small and its gross features resemble the @xmath89 result . the customary way to present arpes data
is by plotting momentum distribution curves ( mdcs ) and energy distribution curves ( edcs ) .
these curves are cuts in @xmath96 of constant @xmath71 and constant @xmath62 respectively . in fig .
1 we present mdcs at the fermi energy ( @xmath97 ) and edcs at the fermi wave - vector @xmath98 for a spin rotationally invariant tomonaga - luttinger liquid for various values of the parameter @xmath90 . in the limit where one of the velocities is much smaller than the other
the calculation becomes again more tractable .
in many physical systems @xmath100 and we will consider this case ( for example for the @xmath101 model away from half filling we have @xmath102 ) . in the limit @xmath103 the spin piece in eq .
( [ scalegk ] ) becomes @xmath104-independent and any singularities in @xmath64 disperse with the slow velocity @xmath105 .
the @xmath104 integral is then readily evaluated by expressing the factors in @xmath106 as fourier transforms of their real - space counterparts .
the result is @xmath107 a similar calculation gives for the spin spectral function @xmath108 at the spin - rotationally invariant point @xmath78 the integral in ( [ gkr0 ] ) can be performed to give a result for @xmath64 in the limit @xmath99 but arbitrary @xmath90 @xmath109 integrals of the spectral functions , such as the density of states and the momentum occupation number , are of interest too .
these simpler quantities give us a qualitative view of the spectrum , without too many complicated details .
the calculation of the density of states @xmath110 and its analogous quantity for the spin susceptibilities @xmath111 is straightforward .
the results are @xmath112 @xmath113 the evaluation of the momentum occupation number @xmath114 and the spin structure factor @xmath115 is complicated by the fact that they can not be expressed in terms of the functions @xmath55 . contrary to the scaling functions considered above the scaling functions of these quantities depend on the cut - off @xmath116 .
it is , however , possible to compute them in the case where the charge and spin velocities are equal .
@xmath117^{\gamma_g+1 } } \!\right\ } \ ; , \\\end{aligned}\ ] ] where again @xmath74 .
since the integrand in ( [ nk1 ] ) is imaginary in the interval @xmath118 $ ] of the imaginary axis , we can add the integral along this interval to the one already present in ( [ nk1 ] ) without affecting @xmath114 .
@xmath119^{\gamma_g+1 } } \!\right\ } \ ; . \\\end{aligned}\ ] ] formally , for @xmath120 , eq . ( [ nk2 ] )
is related to the integral representation of the associated legendre function ( defined with a branch cut from @xmath121 to 1 ) @xmath122 , see ref .
[ erdref ] .
since @xmath114 is regular in @xmath123 we can use analytic continuation to obtain for all non - integer @xmath123 @xmath124^{-(\gamma_g+1/2 ) } \\ \nonumber & & \hspace{0.25in}\times\bigl [ e^{-ia/\lambda_t}q_{-i\tilde k/2 - 1 } ^{\gamma_g+1/2}\left[\cos(2a/\lambda_t)\right ] \\ & & \hspace{0.32in}-e^{ia/\lambda_t}q_{-i\tilde k/2 } ^{\gamma_g+1/2}\left[\cos(2a/\lambda_t)\right]\bigr]\biggr\ } \ ; .\end{aligned}\ ] ] it can be shown that for @xmath125 and @xmath126 this expression reduces , in the limit @xmath127 , to @xmath128}{\gamma[(1 - 2\gamma_g - i\tilde k)/2 ] } \biggr\ } \ ; . \\\end{aligned}\ ] ] using similar manipulations we obtain , for non - integer @xmath129 , the spin structure factor @xmath130 ^{1/2-\beta_{\cal s } } \\ & & \hspace{0.3 in } \times q_{-i\tilde k/2 - 1/2}^{\beta_{\cal s}-1/2 } \left[\cos(2a/\lambda_t)\right]\biggr\ } \ ; , \end{aligned}\ ] ] which for @xmath131 , @xmath132 and @xmath133 tends to @xmath134
in the tomonaga - luttinger liquid both the charge and spin excitations are gapless .
spin and charge gaps may open up as a result of adding to the tomonaga - luttinger hamiltonian ( [ hamiltonian ] ) terms describing backward and umklapp scattering respectively . in the following
we will assume that the 1deg is sufficiently incommensurate so that umklapp scattering may be neglected .
consequently the charge sector remains gapless and continues to be described by a free bosonic theory [ the charge part of eq .
( [ diagh ] ) ] . including a backward scattering term @xmath135 results , after bosonization , in a spin hamiltonian density of the sine - gordon type @xmath136+\frac{2g_1}{(2\pi a)^2 } \cos(\sqrt{8\pi}\phi_s ) \ ; .\ ] ] the @xmath137 perturbation is relevant for @xmath138 , in which case a spin gap is dynamically generated according to the scaling relation @xcite @xmath139^{1/(2 - 2k_s)}$ ] and the excitations are massive spin solitons . in the spin gapped phase the problem is most simply treated in terms of spin fermion fields and their mode decomposition @xmath140 ( obeying the usual fermionic anti - commutation relations ) in momentum space ( 3.375,0 ) ( 0,0)(1,0)3.375 ( 3.375,0)(0,1)0.08 @xmath141 \\ & = & \sqrt{\frac{\delta_s}{4\pi v_s}}\int_{-\infty}^{\infty}d\theta e^{\eta\theta/2}\left[c_+(\theta)e^{i(x/\xi_s)\sinh\theta}- \eta c_-^{\dagger}(\theta)e^{-i(x/\xi_s)\sinh\theta}\right ] \ ; , \end{aligned}\ ] ] ( 3.375,0 ) ( 0,0)(1,0)3.375 ( 0,0)(0,-1)0.08 2 where we introduce the spin correlation length @xmath142 and use the rapidity representation @xmath143 . for @xmath0 , which is known as the free fermion or luther - emery point@xcite , the refermionized hamiltonian is non - interacting and massive with a gap @xmath144 @xmath145 \\ \nonumber & = & \int_{-\infty}^{\infty}d\theta e_s(\theta)\left [ c_+^{\dagger}(\theta)c_+(\theta)+c_-^{\dagger}(\theta)c_-(\theta)-1 \right ] \ ; , \\\end{aligned}\ ] ] where the spin excitation spectrum is @xmath146 in the following we will concentrate on computing the correlation functions at the luther - emery point .
we will comment briefly on the effects of deviations from this point .
in contrast to the tomonaga - luttinger model the calculation of the spin part of most spectral functions for the luther - emery liquid is no longer trivial .
the difficulty lies in the fact that generically the refermionized form for the spectral functions involves highly non - local operators .
we start by evaluating the spin contribution to the transverse spin form factor , which fortunately has a simple representation in terms of the pseudo - fermions .
we then consider the single hole spectral function that belongs to the wider class of functions which do not admit such a simple form .
interestingly , it is still possible to obtain an exact expression for this function too at the luther - emery point .
this exact result extends and corrects earlier work of voit@xcite and wiegmann@xcite . in computing the various spectral properties of the system we can distinguish between two temperature regions . at temperatures
large compared to @xmath147 the spin gap can be ignored , and the results for the tomonaga - luttinger liquid apply . if the temperature is small compared to the spin gap we can evaluate the spin contribution to the correlation functions in the zero temperature limit while introducing errors of order @xmath148 .
henceforth we will be concerned with temperatures in the second region only .
the spin piece of the transverse spin correlation function has the following simple form in terms of the spin fermion fields @xmath150 since the theory reduces , at the luther - emery point , to a theory of free massive fermions , the corresponding spectral function can be readily computed with the result , for @xmath93 , @xmath151 \ ; , \ ] ] where the spin excitation spectrum @xmath152 is given by eq.([spinexcspec ] ) and @xmath153 are the solutions to the quadratic equation @xmath154 , that is @xmath155 the refermionized form of the single hole green function ( below we consider the case @xmath156 and @xmath157 ) @xmath158 involves the non - local vertex operators @xmath159 \ ; , \ ] ] with @xmath160 from eqs .
( [ spinfields ] ) , ( [ vertex ] ) and ( [ refphis ] ) it is evident that @xmath161 and @xmath162 create and destroy a single spin soliton or an integer number of soliton pairs , respectively .
therefor , the green function consists of a coherent one spin soliton piece and an incoherent multi - soliton piece@xcite @xmath163 + g_{s}^{(multi)}(k,\omega ) \ ; , \ ] ] where the multi - soliton piece is proportional ( at @xmath93 ) to @xmath164 $ ] .
deviations from the luther - emery point in the case @xmath165 will result in the formation of a spin soliton - antisoliton bound state , a `` breather '' , which can shift the threshold energy for multi - soliton excitations somewhat . at the luther - emery point the form factors of the vertex operators , _
i.e. _ their matrix elements between the vacuum and various multi - soliton states , are known exactly .
this fact enables us to obtain exact results for the different parts of the spectral function .
the details of the calculation are presented in appendix b. for the spectral weight of the coherent piece we find @xmath166 \ ; , \ ] ] where @xmath167 .
we would like to note here that a simple scaling argument@xcite allows us to obtain the dependence of @xmath168 on @xmath169 for arbitrary @xmath138 .
it follows from the observation that the sine - gordon theory is asymptotically free and hence the dependence of @xmath170 on the short distance cutoff @xmath16 is unaffected by the opening of a spin gap . since in the absence of a gap @xmath170
is proportional to @xmath171 it is a matter of dimensional analysis to see that @xmath172 where @xmath173 is an undetermined scaling function .
the incoherent piece of @xmath174 consists of contributions from processes involving intermediate states containing @xmath175 spin solitons , @xmath176 .
each such contribution starts at an energy threshold @xmath177 @xmath178 \,\ ! .\ ] ] at the vicinity of the threshold _
i.e. _ for @xmath179 and @xmath180 the behavior of @xmath181 is given by @xmath182^{n^2+n-1}$ ] .
in particular the 3 soliton part reads then @xmath183 \ ; .\ ] ] as discussed above , for large @xmath184 the incoherent piece should asymptotically approach the tomonaga - luttinger result @xmath185 .
once the spin part is calculated there still remains the task of convolving it with the corresponding charge part in order to obtain an expression for the full spectral function .
analytically this is difficult , and like in the case of the tomonaga - luttinger liquid , progress can be made only in special cases .
below we carry out the convolution for the single hole spectral function .
we consider temperatures well below the spin gap scale and correspondingly use the above derived zero temperature results for the spin part . for the gapless charge degrees of freedom
we continue to use the finite temperature tomonaga - luttinger expressions .
the spectral function can be written as a sum of two contributions @xmath186 coming from the convolution of the charge part with the coherent ( single soliton ) and incoherent ( multi - soliton ) pieces of @xmath187 .
the frequency integral in @xmath188 is readily evaluated with the result @xmath189 h_{\gamma_c+\frac{1}{2}}\left[\frac{\omega+e_s(k - q)+v_c q}{2\pi t}\right ] \\
\times & & h_{\gamma_c}\left[\frac{\omega+e_s(k - q)-v_c q}{2\pi t}\right ] \ ; , \end{aligned}\ ] ] where here @xmath190 . in fig .
2 we present representative contributions of the single spin soliton piece , @xmath188 , to the mdcs and edcs of a luther - emery liquid with various values of the charge exponent @xmath90 . using some of the results derived below
, we also indicate the asymptotic behavior of the three spin soliton contribution , @xmath191 , to the edcs in the vicinity of its zero temperature threshold @xmath192 .
( 3.2,4.2)(0,-1.2 ) ( -0.15,-1 ) we now restrict the discussion to the following special circumstances : in this case the remaining integral in ( [ g1general ] ) is straightforward since the @xmath104 dependence of the spin part and @xmath193 disappears and we are left with @xmath194h_{2\gamma_c+\frac{1}{2 } } \left[\frac{\omega+e_s(k)}{\pi t}\right ] \ ; , \end{aligned}\ ] ] which in the limit of zero temperature reduces to @xmath195^{2\gamma_c-\frac{1}{2 } } \theta[-\omega - e_s(k ) ] \ ; .\end{aligned}\ ] ] for the 3-soliton contribution to @xmath196 we obtain , assuming @xmath179 and @xmath197 , @xmath198 \ ; , \\\end{aligned}\ ] ] which at @xmath93 equals @xmath199^{2\gamma_c+\frac{3}{2 } } \theta[-\omega-3e_s(k/3 ) ] \ ; . \\\end{aligned}\ ] ] for this case the @xmath175 soliton contribution to @xmath196 is proportional , in the vicinity of its threshold , to @xmath200^ { 2\gamma_c+n^2+n-1/2}\theta[-\omega - ne_s(k / n)]$ ] .
it is also possible to derive closed expressions for the spectral function when @xmath89 but @xmath52 is arbitrary . @xmath201 \\ & \times&\frac{r\omega - v_s k+\sqrt{(r\omega - v_s k)^2+(1-r^2)\delta_s^2 } } { \sqrt{(r\omega - v_s k)^2+(1-r^2)\delta_s^2 } } \ ; .\end{aligned}\ ] ] using the asymptotic form ( [ ztlimit ] ) of @xmath55 one can obtain the zero temperature limit of @xmath188 . when @xmath202 we find @xmath203}{\sqrt{-\omega+r v_s k- { \sqrt{(r\omega - v_s k)^2+(1-r^2)\delta_s^2 } } } } \ ; .
\\ \end{aligned}\ ] ] the kinematics in the case @xmath204 is more involved .
in particular we find that for sufficiently large @xmath205 it is possible to distribute the total momentum between the spin soliton and the gapless charge modes in a way that gives a contribution at frequencies smaller than @xmath152 .
( in this case the effect exists only for @xmath88 .
this is a special property of the point @xmath89 which precludes mixing of left and right moving charge excitations . )
( 3.375,0 ) ( 0,0)(1,0)3.375 ( 3.375,0)(0,1)0.08 @xmath206 ( 3.375,0 ) ( 0,0)(1,0)3.375 ( 0,0)(0,-1)0.08 2 for @xmath179 and @xmath197 the 3-soliton contribution to @xmath196 reads @xmath207}{\delta_s}\right ] \\ & \times&\theta\{v_c q-\omega-3e_s[(k - q)/3]\ } \ ; . \end{aligned}\ ] ] when @xmath202 the zero temperature limit of the above is readily evaluated giving @xmath208^{\frac{3}{2 } } \theta[-\omega-3e_s(k/3 ) ] \ ; .\end{aligned}\ ] ] the zero temperature limit when @xmath204 is , once again , more complicated .
however , it is possible to show that in the range of validity of eq .
( [ g2g03sol ] ) , _ i.e. _ for @xmath179 , and for @xmath209 it coincides with the expression for the case @xmath202 eq .
( [ g23solt0rl ] ) .
deviations from this behavior may occur only if the velocity ratio is large and satisfies @xmath210 .
the function @xmath212 is real @xmath213^{\gamma}}\right\ } \ ; .\end{aligned}\ ] ] although the imaginary part of the integral diverges as @xmath214 for @xmath215 , its real part , and hence also @xmath212 , are analytic for all values of @xmath57 . substituting @xmath216
we find @xmath217 the integral is analytic in the limit @xmath127 for @xmath218 . in this range
the exponential factor in ( [ hk1 ] ) can be dropped and it reduces to the integral representation of the beta function @xcite .
we can then use analytical continuation to obtain for all non - integer values of @xmath57 @xmath219 \ ; .\ ] ] for the integers , @xmath220 can be calculated as follows .
first , we obtain using standard residue technique @xmath221 where @xmath222 are the fermionic and bosonic occupation functions
. we then integrate ( [ hk ] ) by parts twice to find the recursion relation @xmath223 which implies @xmath224f_+(\pi k ) \ ; , \\
\right]f_-(\pi k ) \ ; . \end{aligned}\ ] ] we also note that @xmath225 finally , the asymptotic behavior of @xmath212 for large @xmath226 is easily evaluated with the result @xmath227 since the spectral functions have a scaling form in the variables @xmath228 and @xmath229 eq . ( [ ztlimit ] ) also determines the zero temperature limit of these functions .
inserting the resolution of the identity into eq .
( [ refg ] ) one obtains @xmath231 the matrix elements of the vertex operators appearing in ( [ relrefg ] ) are known as the form factors of these operators . they have been derived , at the luther - emery point ( see also ref . [ lukref ] ) , by a variety of ways .
they were first obtained by schroer and truong@xcite who normal ordered the vertex operator with respect to the spin fermions .
smirnov@xcite derived them for @xmath232 using bootstrap axioms .
most recently they were calculated using monodromy relations by bernard and leclair @xcite . since the fields @xmath233 are neutral with respect to the topological @xmath234 charge of the solitons @xmath235 the form factors are non - vanishing only for @xmath234 neutral states @xmath236 \frac{\prod_{1\leq k <
j\leq n}\sinh\left(\frac{\theta_k-\theta_j}{2}\right ) \sinh\left(\frac{\theta_{n+k}-\theta_{n+j}}{2}\right ) } { \prod_{1\leq k , j\leq n}\cosh\left(\frac{\theta_{n+k}-\theta_j}{2}\right ) } \ ; , \end{aligned}\ ] ] furthermore , since @xmath161 creates and destroys a single soliton the state @xmath240 can differ from @xmath241 by @xmath242 solitons only .
it is also easy to check that @xmath174 is real . using eqs .
( [ spinfields ] ) and ( [ formfact ] ) one obtains that each term in the fourier transform of eq .
( [ relrefg ] ) is proportional to @xmath243 times a real expression .
thus only terms with @xmath244 an even integer should be considered .
combining these two observations we conclude that the contribution to @xmath174 comes solely from the terms @xmath245 ( the amplitudes for the cases @xmath246 are finite but cancel each other ) .
the coherent piece of the spectral function is due to the terms @xmath247 and @xmath248 .
the evaluation of the first is straightforward with the result @xmath249\delta[\omega+e_s(k ) ] \ ; , \ ] ] where @xmath250 .
the cutoff dependence of the above result is in conflict with the general scaling argument that was given following eq .
( [ zs ] ) .
it represents corrections to scaling coming from irrelevant operators .
the leading scaling behavior is recovered by considering the contribution of the @xmath248 term to the coherent piece @xmath251 ^ 2 \!\int d\theta\,e^ { -\theta+i[(x/\xi_s)\sinh\theta- \delta_s t\cosh\theta ] } \ ; .\ ] ] the @xmath252 integral is divergent and a lower cutoff @xmath253 ( corresponding to a cutoff @xmath254 ) should be introduced .
we then find for the leading contribution to the coherent piece @xmath255 \ ; , \end{aligned}\ ] ] where @xmath256 is a modified bessel function .
fourier transforming it one obtains @xmath257 $ ] with the spectral weight @xmath258 given by eq .
( [ zs ] ) .
the @xmath259 soliton contribution to the incoherent piece of the spectral function comes from the terms @xmath260 and @xmath261 .
as is the case with the coherent piece the former contains corrections to scaling while the latter is responsible for the leading behavior which is proportional to @xmath262 ^ 2 \delta\left(k\xi_s-\sum_{i=1}^{2n+1}\sinh\theta_i\right ) \delta\left(\frac{\omega}{\delta_s}+\sum_{i=1}^{2n+1}\cosh\theta_i\right ) \\
\nonumber \times&&\exp\left(\frac{1}{2 } \sum_{j=1}^{n+1}\theta_j-\frac{1}{2}\sum_{j = n+2}^{2n+1}\theta_j\right ) \frac{\prod_{1\leq k <
j\leq n+1}\sinh^2\left(\frac{\theta_k-\theta_j}{2 } \right)\prod_{n+2\leq k < j\leq 2n+1}\sinh^2\left(\frac{\theta_k-\theta_j}{2 } \right)}{\prod_{n+2\leq k\leq 2n+1\ , , \ , 1\leq j\leq n+1 } \cosh^2\left(\frac{\theta_k-\theta_j}{2}\right ) } \ ; .\end{aligned}\ ] ] 2 the exact evaluation of the above integrals is difficult but their behavior near the threshold @xmath263 $ ] may be extracted in the following way .
writing @xmath264-\delta\omega$ ] one finds that for @xmath179 and @xmath265 the second @xmath266-function restricts the integration region over @xmath267 to a small ball near the origin .
it is then legitimate to expand the integrand to lowest order in these variables .
the integral over @xmath252 gives ( together with the factor @xmath268 ) the overall cutoff dependence @xmath269 in accord with the general scaling argument . by changing variables to spherical
coordinates it is easy to check that the remaining integral over @xmath267 is proportional to @xmath270 . a detailed evaluation of the 3 solitons case results in eq . ( [ 3sol ] ) .
s. tomonaga , prog .
. phys . * 5 * , 544 ( 1950 ) .
k. d. schotte and u. schotte , phys . rev .
* 182 * , 479 ( 1969 ) .
d. c. mattis , j. math .
15 * , 609 ( 1974 )
. a. luther and i. peschel , phys . rev .
b * 9 * , 2911 ( 1974 ) . for a review , see a. j. heeger , s. a. kivelson , j. r. schrieffer , and w. p. su , rev .
* 60 * , 781 ( 1988 ) .
for a review , see c. bourbonnais and d. jerome in _ advances in synthetic metals , twenty years of progress in science and technology _ , edited by p. bernier , s. lefrant , and g. bidan ( elsevier , new york , 1999 ) , p. 206 .
p. monceau , _ electronic properties of inorganic quasi - one - dimensional compounds _ , ( d. reidel publ .
co. , dordrecht , boston , 1985 ) .
g. gruner , _ density waves in solids _ , ( addison - wesley , 1994 ) .
x. g. wen , phys .
b * 41 * , 12838 ( 1990 ) ; f. p. milliken , c. p. umbach , and r. a. webb , sol . state comm .
* 97 * , 309 ( 1996 ) ; a. m. chang , l. n. pfeiffer and k. w. west , phys .
77 * , 2538 ( 1996 ) ; m. grayson , d. c. tsui , l. n. pfeiffer , k. w. west , and a. m. chang , _ ibid .
_ * 80 * , 1062 ( 1998 ) .
s. tarucha , t. honda , and t. saku , solid state comm .
* 94 * , 413 ( 1995 ) ; a. yacoby , h. l. stormer , n. s. wingreen , k. w. baldwin , and k. w. west , phys . rev .
lett . * 77 * , 4612 ( 1996 ) . for a recent perspective on stripe phases in high temperature superconductors
see , v. j. emery , s. a. kivelson , and j. m. tranquada , proc .
usa * 96 * , 8814 ( 1999 ) .
f. zwick , s. brown , g. margaritondo , c. merlic , m. onellion , j. voit , and m. grioni , phys .
lett . * 79 * , 3982 ( 1997 ) .
y. hwu , p. almares , m. marsi , h. berger , f. lvy , m. grioni , d. malterre , and g. margaritondo , phys .
b * 46 * , 13624 ( 1992 ) ; a. terassi , m. marsi , h. berger , g. margaritondo , r. j. kelly , and m. onellion , _ ibid . _ * 52 * , 5592 ( 1995 ) .
m. grioni , h. berger , m. garnier , f. bommeli , l. degiorgi , and s. schlenker , physica scripta t * 66 * , 172 ( 1996 ) .
x. j. zhou , p. bogdanov , s. a. kellar , t. noda , h. eisaki , s. uchida , z. hussain , and z .- x .
shen , _ science _ * 286 * , 268 ( 1999 ) .
v. meden and k. schnhammer , phys .
b * 46 * , 15753 ( 1992 ) .
j. voit , phys . rev .
b * 47 * , 6740 ( 1993 ) .
k. schnhammer and v. meden , j. electron spectrosc .
relat . phenom .
* 62 * , 225 ( 1993 ) .
n. nakamura and y. suzumura , prog .
theor . phys . *
98 * , 29 ( 1997 ) .
a. luther and v. j. emery , phys .
lett . * 33 * , 589 ( 1974 ) .
j. voit , euro .
j. * b * 5 , 505 ( 1998 ) .
[ voitref ] p. b. wiegmann , * 59 * , 15705 ( 1999 ) .
[ wiegmannref ] v. j. emery , in _ highly conducting one - dimensional solids _ , edited by j. t. devreese , r. p. evrard and v. e. van doren ( plenum , new york 1979 ) , p. 247 .
j. von delft and h. schoeller , annalen phys .
* 7 * , 225 ( 1998 ) .
a. e. mattsson , s. eggert and h. johannesson , phys .
b * 56 * , 15615 ( 1997 ) .
[ mattsson97ref ] s. eggert , a. e. mattsson and j. m. kinaret , phys .
b * 56 * , r15537 ( 1997 ) .
[ eggert97ref ] n. kawakami and s. k. yang , phys .
a * 148 * , 359 ( 1990 ) .
a. erdlyi _
_ , _ higher transcendental functions _ ( mcgraw hill , new york , 1953 ) vol .
i , p. 155 .
[ erdref ] see , for example , a. o. gogolin , a. a. nersesyan and a. m. tsvelik , _ bosonization and strongly correlated systems _ , ( cambridge university press , cambridge , 1998 ) . e. w. carlson , d. orgad , s. a. kivelson and v. j. emery , phys .
rev b * 62 * , 3422 ( 2000 ) .
i.s . gradshteyn and i. m. ryzhik , _ table of integrals , series and products _
( academic , new york , 1994 ) p. 957 . an integral representation for the form factors away from the luther - emery point
is suggested in s. lukyanov , phys .
a * 12 * , 2543 ( 1997 ) .
[ lukref ] b. schroer and t. t. truong , nucl .
b * 144 * , 80 ( 1978 ) .
f. a. smirnov , _ form factors in completely integrable models of quantum field theory , advanced series in mathematical physics 14 _ ( world scientific , 1992 ) .
d. bernard and a. leclair , nucl .
b * 426 * , 534 ( 1994 ) .
s. lukyanov and a. zamolodchikov , nucl .
b * 493 * , 571 ( 1997 ) .
|
we calculate the finite temperature single hole spectral function and the spin dynamic structure factor of spinfull one - dimensional tomonaga - luttinger liquid .
analytical expressions are obtained for a number of special cases .
we also calculate the single hole spectral function of a spin gapped luther - emery liquid and obtain exact results at the free fermion point @xmath0 .
these results may be applied to the analysis of angle resolved photoemission and neutron scattering experiments on quasi - one - dimensional materials . 2
|
rare earth transition metal oxy - arsenides adopting the zrcusias structure - type ( 1111 materials ) contain square nets of transition metal atoms coordinated by arsenic in an edge - sharing tetrahedral geometry @xcite .
these materials , and others containing the same transition metal
arsenic layers , have been a main focus of experimental and theoretical studies in the solid state chemistry and condensed matter physics communities since the discovery of high temperature superconductivity in many materials , primarily containing iron , and several structural families @xcite .
a common theme , and perhaps necessary condition , for high temperature superconductivity in the iron compounds is the suppression of magnetism , which is accomplished by chemical substitutions or application of pressure .
while aleovalent substitution ( doping ) is most common , replacing some fe with isovalent ru has been shown to produce superconductivity in the family of layered iron arsenides adopting the thcr@xmath5si@xmath5 structure type ( 122 materials ) @xcite . however
, this substitution does not produce superconductivity in 1111 materials , although it does suppress the magnetism @xcite .
indeed , partial replacement of fe with ru in already superconducting compositions ( smfeaso@xmath6f@xmath7 ) decreases the superconducting critical temperature @xcite .
the response to ru substitution is one of the few striking differences between the behavior of 1111 and 122 materials .
the natural extension of these substitution studies is the analysis of the pure ruthenium compounds .
the 122 materials srru@xmath5as@xmath5 and baru@xmath5as@xmath5 have been studied and are diamagnetic metals @xcite .
interestingly , the 122 phosphide laru@xmath5p@xmath5 , with trivalent lanthanum in place of the ususal divalent alkaline earth , is superconducting below 4.1 k @xcite .
literature reports for @xmath0ruaso ( @xmath0 = lanthanide ) are limited to lattice constants @xcite and resistivity for @xmath0 = la and ce @xcite .
a more thorough investigation of the ru materials is important in developing a full understanding of the behavior of these interesting chemical systems .
the current study aims to examine the evolution of structural and basic physical properties of some 1111 materials of composition @xmath0ruaso as the lanthanide ( @xmath0 ) is varied .
full crystal structure refinements at room temperature and thermal expansion , heat capacity , electrical transport , and magnetic behavior below 300 k are reported for laruaso , ndruaso , smruaso , and gdruaso .
the results are compared to similar 1111 and 122 materials containing fe and ru .
ruas was made by reacting reduced ru powder with as pieces in an evacuated silica ampoule at 1000 @xmath8c and used as a starting material for the target compounds .
polycrystalline samples of laruaso , ndruaso , smruaso , and gdruaso were synthesized from thoroughly ground mixtures of ruas with fresh @xmath0 filings and dry @xmath9o@xmath10 powder .
the starting materials were handled and mixed inside a he glove box .
samples ( @xmath11 2 g each ) were pressed into 1/2 inch diameter pellets and placed in covered alumina crucibles inside silica tubes .
the tubes were evaculated , back - filled with @xmath110.2 atm of ultra - high - purity ar , and flame sealed .
the samples were heated at 12001250 @xmath8c for 1236 hours several times , and were ground and pelletized between the heating cycles . surface contamination from reaction with vapor from the sio@xmath5 tubes
was removed at each step .
powder x - ray diffraction ( panalytical xpert pro mpd , monochromatic cu - k@xmath12 radiation ) was used to determine phase purity and refine the crystal structures using the program fullprof @xcite .
low temperature powder diffraction was performed with an oxford phenix closed - cycle cryostat .
a quantum design physical property measurement system was used for transport and heat capacity measurements , and a quantum design magnetic property measurement system squid magnetometer was used for magnetization measurements .
room temperature rietveld refinements are shown in figure [ fig : rietveld ] , and indicate all samples are @xmath13 90 % pure .
impurities including @xmath9o@xmath10 and ruas were observed in the samples with @xmath0 = la , nd , and gd ; however , no significant impurity peaks were observed in the smruaso sample .
agreement factors for the fits ranged from @xmath14 = 2.97.1 , @xmath15 = 3.710.1 .
@xmath16 = 1.52.2 .
reitveld refinement of room temperature powder x - ray diffraction data .
measured intensities are shown as circles , calculated intensities as grey lines , and difference curves as black lines .
the upper set of ticks in each panel correspond to the main phase .
middle and lower ticks in ( a , b , d ) represent @xmath9o@xmath10 and ruas , respectively .
, width=312 ]
these materials adopt the zrcusias structure type shown in figure [ fig : bonds]a , with @xmath0 and as at wyckoff positions 2c ( @xmath17 @xmath17 @xmath18 ) , ru at 2a ( @xmath19 @xmath17 0 ) , and o at 2b ( @xmath19 @xmath17 @xmath20 ) . room temperature structural parameters are listed in table [ lattice - table ] .
the lattice constants are in good agrement with the original report of these materials , and reflect the expected lanthanide contraction , as previously noted @xcite .
table [ lattice - table ] also lists refined atomic positions @xmath21 and @xmath22 , which have not been previously reported .
.[lattice - table ] room temperature lattice constants and z - coordinates for the @xmath0 and as positions from powder x - ray diffraction .
uncertainties on the last digit of refined parameters are listed in parentheses , and are those reported by the refinement program ( fullprof ) . hall coefficient ( @xmath23 ) measured at 2 k. effective moment ( @xmath24 ) , weiss temperature ( @xmath25 ) , and neel temperature ( t@xmath26 ) for compounds with magnetic rare earth elements determined from temperature dependent magnetization and heat capacity measurements . [ cols="<,^,^,^,^",options="header " , ] based on the full structure refinements , the evolution of interatomic distances and angles can be examined as @xmath0 is varied . selected distances and angles
are plotted in figure [ fig : bonds ] . as expected , the ru as distance is least sensitive to the identity of @xmath0 ( figure [ fig : bonds]a ) , changing by less than 0.4 % across this series . the @xmath0as and @xmath0o distances ( figure [ fig : bonds]a ) increase smoothly as the @xmath0 ionic radius increases . figure
[ fig : bonds]c shows the two as ru as bond angles , as defined in the inset of figure [ fig : bonds]a .
the observation that @xmath27 indicates that the as tetrahedron around ru is compressed along the c - axis . as the @xmath0 radii is decreased along the series from la to gd , @xmath28 decreases and @xmath29 increases , moving toward ideal tetrahedral coordination .
it is interesting to note that a strong correlation between these angles and superconducting critical temperatures has been observed in the related fe compounds , where the highest critical temperatures occur when the tetrahedra are closest to ideal @xcite .
a systematic structural study of the analogous fe compounds using single crystal x - ray diffraction results has been reported @xcite .
comparison with table [ lattice - table ] and figure [ fig : bonds ] shows that the same structural trends exist in @xmath0feaso and @xmath0ruaso . in the fe compounds , all the interatomic distances are shorter than those in the corresponding ru analogues . @xmath0as and @xmath0o distances in the fe compounds are shorter by about 0.30.4 and 0.2 @xmath30 for the lanthanides studied here .
this reflects the compression of the overall structure when ru is replaced by smaller fe , and is often referred to as chemical pressure .
the fe as distances in @xmath0feaso vary from 2.412 to 2.392 @xmath30 from @xmath0 = la to gd , showing a much stronger dependence on @xmath0 than the ru as distances in figure [ fig : bonds ] which vary by less than 0.01 @xmath30 across the same series .
these results suggest that the ru compounds are in some sense less compressible than the fe analogues , at least in the @xmath31-plane .
this may indicate the ruas layers are stiffer than the feas layers due to increased ru
ru repulsion between the larger ru atoms .
this is consistent with the changes in lattice constants across the @xmath0 series . from laruaso to gdruaso @xmath32 changes by 2.0 % and @xmath33 by 4.3 % . from lafeaso to gdfeaso @xmath32 changes by 2.9 % and @xmath33 by 3.2 % .
the @xmath0ruaso lattice response is much more anisotropic than that of @xmath0feaso , and it is stiffer in the @xmath32 direction than in @xmath33 . in addition , the tetrahedral coordination around ru ( figure [ fig : bonds]c ) is significantly more flattened along @xmath33 than that found around fe in @xmath0feaso @xcite , also pointing to ru - ru repulsion in the @xmath31-plane .
finally , it is interesting to note that the deviation from a linear trend in ru - as distance at @xmath0 = nd in @xmath0ruaso ( figure [ fig : bonds]a ) is also seen in @xmath0feaso @xcite , although its origin and significance remain unclear .
structural parameters of _ _ ln__ruaso plotted as a function of the ionic radius of @xmath34 in eight - fold coordination @xcite .
( a ) ru as distance , with the structure shown in the inset .
( b ) @xmath0as and @xmath0o distances .
( c ) coordination angles around ru as defined in the inset of ( a ) .
the two angles are not independent due to the ru site symmetry ( -4__m__2 ) .
, width=312 ] the temperature dependence of the lattice parameters and unit cell volume between @xmath1120 and 300 k are presented in figure [ fig : thermal - expansion ] .
the data are normalized to 300 k values .
all four compounds show similar volume thermal expansion behavior .
a small divergence in @xmath35v / v among materials occurs at the lowest temperatures , but no simple relationship to the composition is observed .
the temperature dependence of the lattice constants shows strong anisotropy in all materials , with @xmath33 changing more rapidly than @xmath32 .
this reflects the layered nature of the crystal structure which contains covalently bonded ruas layer extending in the @xmath31-plane , and is consistent with the trends in lattice constant with @xmath0 at room temperature discussed above .
the magnitude of this anisotropy ( [ @xmath36 / [ @xmath37 ) depends strongly on the composition , and reaches a value of 1.4 , 2.2 , 2.9 , and 1.7 at 20 k for la , nd , sm , and gd , respectively .
the reason for the maximum for smruaso is not clear .
however , among the lanthanides studied here sm does possess two unique properties : ( 1 ) a tendency toward mixed - valent behavior ( sm@xmath38 ) , ( 2 ) two low energy , closely spaced magnetic configurations in the trivalent state ( j = 5/2 , 7/2 ) .
these can be expected to affect the temperature dependence of physical and structural properties .
it is interesting to speculate whether this increased `` flexibility '' of sm could be the reason that the smruaso synthesis produced the cleanest sample ( figure [ fig : rietveld ] ) .
temperature dependence of ( a ) the unit cell volume ( v ) and ( b ) the lattice constants @xmath32 ( solid symbols ) and @xmath33 ( open symbols ) determined by powder x - ray diffraction and normalized to values at 300 k. uncertainties based on the rietveld fitting are smaller than the data markers .
, width=288 ] results of magnetization and electrical resistivity measurements are shown in figure [ fig : mag - res ] .
magnetism in these compounds is dominated by the rare - earth element .
the very small and nearly temperature independent susceptibility of laruaso suggest that ru does not have a local magnetic moment in these materials .
the small upturn at low temperatures corresponds to an effective moment of 0.09 @xmath39 per formula unit , and is likely due to magnetic impurities .
the magnetic susceptibility ( @xmath40 ) has a cusp at low temperatures for smruaso and gdruaso , and displays curie - wiess ( cw ) behavior at higher temperatures .
ndruaso shows cw behavior over the entire temperature range investigated .
effective moments ( @xmath24 ) and weiss temperatures ( @xmath25 ) determined from cw fits to the data from 50300 k are listed in table [ lattice - table ] .
the usual cw model is not sufficient for fitting the susceptibility data for smruaso .
this is not surprising since sm@xmath41 is known to have closely space energy levels with different magnetic moments and strongly temperature dependent populations in the temperature ranges studied here .
effective moments in table [ lattice - table ] are consistent with the free ion values of 7.94 @xmath39 for gd@xmath41 and 3.62 for nd@xmath41 , and similar to previous reports for the relevant fe - compounds ndfeaso ( 3.60 @xmath39 @xcite ) and gdfeaso ( 7.83 @xmath39 @xcite ) .
weiss temperatures are negative , indicating antiferromagnetic interactions , and the cusps in @xmath40(t ) near 5 k for smruaso and gdruaso are attributed to antiferromagnetic ordering of rare earth moments . no evidence of magnetic ordering of nd moments is observed above 2 k. electrical resistivity measurements ( figure [ fig : mag - res]c ) indicate metallic behavior for these materials , similar to previous reports for laruaso @xcite , cefeaso @xcite , srru@xmath5as@xmath5 @xcite , and baru@xmath5as@xmath5 @xcite .
the magnitude of the resistivity ( @xmath42 ) at room temperature decreases across the series from la to gd .
the inset in figure [ fig : mag - res]c shows the effect of the magnetic ordering on the electrical resistivity in smruaso and gdruaso .
reduced spin - disorder scattering below the transition is expected to be responsible for the decrease in @xmath42 .
hall coefficients ( @xmath23 ) measured at 2 k are listed in table [ lattice - table ] .
the values are negative , indicating conduction dominated by electrons , and the magnitude decreases from @xmath0 = la to gd .
carrier concentrations ( @xmath43 ) inferred from these hall coefficients from the simple one band formula @xmath23 = 1/@xmath43e range from @xmath44 for laruaso to @xmath45 for gdruaso . the inferred carrier concentrations trend with the resistivity values near room temperature , and imply hall mobilities of about 12 @xmath46/vs at room temperature .
( a , b ) temperature dependence of the magnetic susceptibility ( @xmath47 ) per mole of formula units ( note the different y - axis scales ) .
the insets show the data for gdruaso ( a ) and smruaso ( b ) near their neel temperatures .
( c ) temperature dependence of the resistivity ( @xmath42 ) with the low temperature behavior for smruaso and gdruaso in the inset .
, width=288 ] to confirm the bulk nature of the magnetic phase transitions , heat capacity measurements were performed .
the results are shown in figure [ fig : hc ] .
sharp anomalies are observed at the ordering temperatures for smruaso and gdruaso .
an upturn below about 4 k in the ndruaso data suggest the nd magnetic moments may undergo a long range ordering transition near or below 2 k. ndfeaso shows similar heat capacity behavior in this temperature range , with nd moments ordering near 2.1 k @xcite .
a small anomaly is observed near 11 k in the ndruaso heat capacity data ( figure [ fig : hc ] ) .
although this may be intrinsic , it is perhaps more likely due to the presence of a very small amount of ndas , which has magnetic phase transition near 11 k with a large heat capacity anomaly @xcite
. based on the heat capacity data along with the resistivity and magnetic susceptibility results in figure [ fig : mag - res ] the antiferromagnetic ordering temperatures ( @xmath48 ) are estimated to be 4.5 k for smruaso and 5.0 k for gdruaso ( table [ lattice - table ] ) .
these are close to the rare - earth ordering temperatures in the related fe materials : 4.1 k for gdfeaso @xcite and 56 k in smfeaso @xcite .
the entropy change associated with the magnetic ordering in smruaso and gdruaso can be determined by the integral @xmath49 . the magnetic heat capacity @xmath50 is estimated by subtracting the heat capacity of laruaso for the total heat capacity .
integrating up to t = 15 k gives an entropy per mole of lanthanide of 5.4 j / k@xmath51/mol for smruaso and 16.2 j / k@xmath51/mol for gdruaso .
the value obtained for gdruaso is close to r@xmath52 = 17.3 j / k@xmath51/mol expected for @xmath53 .
the experimental value is likely underestimated due to the large value of @xmath54 remaining at the lowest temperature investigated here ( figure [ fig : hc ] ) . for smruaso ,
the entropy value is close to r@xmath55 = 5.8 j / k@xmath51/mol .
this result is somewhat surprising , since @xmath56 is expected for the ground state of sm .
however , similar values have been obtained for smfeaso@xmath6f@xmath57 and attributed to crystal field splitting which results in an @xmath58 state at the lowest temperatures @xcite .
temperature dependence of the heat capacity per mole of formula unit below 100 k. the upper inset shows the sharp anomalies occurring in smruaso and gdruaso at the magnetic ordering temperatures , and an upturn at the lowest temperatures for ndruaso suggesting ordering below 2 k. the lower inset shows @xmath59 vs. @xmath60 for laruaso , and the linear fit used to determine the electronic heat capacity coefficient .
, width=336 ] for laruaso , the electronic specific heat coefficient ( @xmath61 ) and the debye temperature ( @xmath62 ) can be determined from the plot of @xmath54/t vs. t@xmath51 shown in the lower inset of figure [ fig : hc ] . assuming @xmath63 at low temperature , @xmath61 = 2.2 mj / k@xmath51/mol - f.u . or 0.55 mj / k@xmath51/mol - atom is obtained .
this is not a very high value , suggesting electron correlations are weak in these materials .
the debye temperature determined from @xmath29 is 330 k. these values can be compared to those reported for baru@xmath5as@xmath5 ( @xmath61 = 0.98 mj / k@xmath51/mol - atom , @xmath62 = 271 k ) and srru@xmath5as@xmath5 ( @xmath61 = 0.82 mj / k@xmath51/mol - atom , @xmath62 = 271 k ) @xcite .
the present study of the structural and physical properties of the ru - based 1111 materials @xmath0ruaso ( @xmath0 = la , nd , sm , gd ) allows some interesting comparisons with isostructural and isoelectronic @xmath0feaso series .
structural differences include increased distortion of the transition metal coordination environment , and increased anisotropy in the response of the lattice to changing @xmath0 ionic radii .
both observations may be related to ru - ru repulsion in the @xmath31-plane .
results of magnetization and heat capacity measurements indicate antiferromagnetic ordering of @xmath0 magnetic moments at 5.0 k in gdruaso , 4.5 k in smruaso , and below 2 k in ndruaso , similar to values reported for fe analogues .
evidence of the magnetic transitions in smruaso and gdruaso is seen in the electrical resistivity .
together , resistivity and hall effect measurements indicate metallic conduction dominated by electrons .
no clear sign of ru magnetism is observed in laruaso , and analysis of low temperature heat capacity data suggest weak electron correlations in these materials .
estimates of entropy associated with the magnetic phase transitions give values close to r@xmath55 for smruaso and r@xmath52 for gdfeaso .
a. kreyssig , m. a. green , y. lee , g. d. samolyuk , p. zajdel , j. w. lynn , s. l. budko , m. s. torikachvili , n. ni , s. nandi , j. b. leo , s. j. poulton , d. n. argyriou , b. n. harmon , r. j. mcqueeney , p. c. canfield , a. i. goldman , phys .
b 78 ( 2008 ) 184517 .
|
polycrystalline samples of laruaso , ndruaso , smruaso , and gdruaso have been synthesized and studied using powder x - ray diffraction , electrical transport , magnetization , and heat capacity measurements .
variations in structural properties across the series reveal a trend toward more ideal tetrahedral coordination around ru as the size of the rare earth element is reduced .
the lattice parameters of these ru compounds show a more anisotropic response to variation in @xmath0 than their fe analogues , and significant anisotropy in thermal expansion is also observed .
transport measurements show metallic behavior , and carrier concentrations near 10@xmath110@xmath2 electrons per @xmath3 are inferred from simple analysis of hall effect measurements .
anomalies in resistivity , magnetization , and heat capacity indicate antiferromagnetic ordering of rare earth moments at 5 k for gdruaso , 4.5 k for smruaso , and @xmath42 k for ndruaso .
magnetization measurements on laruaso show no evidence of a magnetic moment on ru .
observed behaviors are compared to those reported for similar fe and ru compounds .
|
the coalescence and merging of two compact stars into a single object is a very common end - point of close binary evolution .
dissipation mechanisms such as friction in common envelopes , tidal dissipation , or the emission of gravitational radiation , are always present and cause the orbits of close binary systems to decay .
this review will concentrate on the coalescence of compact binaries containing either two neutron stars ( hereafter ns ) or two white dwarfs ( wd ) .
many theoretical models of gamma - ray bursts ( grbs ) rely on coalescing ns binaries to provide the energy of grbs at cosmological distances ( e.g. , eichler et al .
1989 ; narayan , paczyski , & piran 1992 ; mszros & rees 1992 ; for recent reviews see mszros 1999 and piran 1999 ) .
the close spatial association of some grb afterglows with faint galaxies at high redshifts may not be inconsistent with a ns binary merger origin , in spite of the large recoil velocities acquired by ns binaries at birth ( bloom , sigurdsson , & pols 1999 ; but see also bulik & belczynski 2000 ) .
currently the most popular models all assume that the coalescence of two ns leads to the formation of a rapidly rotating black hole ( bh ) surrounded by a torus of debris .
energy can then be extracted either from the rotation of the kerr bh or from the material in the torus so that , with sufficient beaming , the gamma - ray fluxes observed from even the most distant grbs can be explained ( mszros , rees , & wijers 1999 ) .
however , it is important to understand the hydrodynamic processes taking place during the final coalescence before making assumptions about its outcome .
in particular , as will be argued below ( 2.2 ) , it is not clear that the coalescence of two @xmath0 ns forms an object that will collapse to a bh on a dynamical timescale , and it is not certain either that a significant amount of matter will be ejected during the merger to form an outer torus around the central object ( faber & rasio 2000 ) . coalescing ns binaries are also important sources of gravitational waves that may be directly detectable by the large laser interferometers currently under construction , such as ligo ( abramovici et al . 1992 ;
see barish & weiss 1999 for a recent pedagogical introduction ) and virgo ( bradaschia et al .
in addition to providing a major new confirmation of einstein s theory of general relativity ( gr ) , including the first direct proof of the existence of black holes ( see , e.g. , flanagan & hughes 1998 ; lipunov , postnov , & prokhorov 1997 ) , the detection of gravitational waves from coalescing binaries at cosmological distances could provide accurate independent measurements of the hubble constant and mean density of the universe ( schutz 1986 ; chernoff & finn 1993 ; markovi 1993 ) .
expected rates of ns binary coalescence in the universe , as well as expected event rates in laser interferometers , have now been calculated by many groups . although there is some disparity between various published results , the estimated rates are generally encouraging ( see kalogera 2000 for a recent review ) .
many calculations of gravitational wave emission from coalescing binaries have focused on the waveforms emitted during the last few thousand orbits , as the frequency sweeps upward from @xmath1hz to @xmath2hz .
the waveforms in this frequency range , where the sensitivity of ground - based interferometers is highest , can be calculated very accurately by performing high - order post - newtonian ( pn ) expansions of the equations of motion for two _ point masses _ ( see , e.g. , owen & sathyaprakash 1999 and references therein ) . however , at the end of the inspiral , when the binary separation becomes comparable to the stellar radii ( and the frequency is @xmath3khz ) , hydrodynamics becomes important and the character of the waveforms must change .
special purpose narrow - band detectors that can sweep up frequency in real time will be used to try to catch the last @xmath4 cycles of the gravitational waves during the final coalescence ( meers 1988 ; strain & meers 1991 ) .
these `` dual recycling '' techniques are being tested right now on the german - british interferometer geo 600 ( danzmann 1998 ) . in this terminal phase of the coalescence ,
when the two stars merge together into a single object , the waveforms contain information not just about the effects of gr , but also about the interior structure of a ns and the nuclear equation of state ( eos ) at high density . extracting this information from observed waveforms
, however , requires detailed theoretical knowledge about all relevant hydrodynamic processes .
if the ns merger is followed by the formation of a bh , the corresponding gravitational radiation waveforms will also provide direct information on the dynamics of rotating core collapse and the bh `` ringdown '' ( see , e.g. , flanagan & hughes 1998 ) .
coalescing wd binaries have long been discussed as possible progenitors of type ia supernovae ( iben & tutukov 1984 ; webbink 1984 ; paczyski 1985 ; see branch et al .
1995 for a recent review ) . to produce a supernova ,
the total mass of the system must be above the chandrasekhar mass . given evolutionary considerations , this requires two c - o or o - ne - mg wd .
yungelson et al . (
1994 ) showed that the expected merger rate for close wd pairs with total mass exceeding the chandrasekhar mass is consistent with the rate of type ia supernovae deduced from observations .
alternatively , a massive enough merger may collapse to form a rapidly rotating ns ( nomoto & iben 1985 ; colgate 1990 ) .
chen & leonard ( 1993 ) speculated that most millisecond pulsars in globular clusters might have formed in this way . in some cases planets
may also form in the disk of material ejected during the coalescence and left in orbit around the central pulsar ( podsiadlowski , pringle , & rees 1991 ) . indeed
the very first extrasolar planets were discovered in orbit around a millisecond pulsar , psr b1257@xmath512 ( wolszczan & frail 1992 ) . a merger of two magnetized wd might lead to the formation of a ns with extremely high magnetic field , and this scenario has been proposed as a source of grbs ( usov 1992 ) .
close wd binaries are expected to be extremely abundant in our galaxy , even though their direct detection remains very challenging ( han 1998 ; saffer , livio , & yungelson 1999 ) .
iben & tutukov ( 1984 , 1986 ) predicted that @xmath6% of all binary stars produce close wd pairs at the end of their stellar evolution .
more recently , theoretical estimates of the double wd formation rate in the galaxy have converged to a value @xmath7 , with an uncertainty that may be only a factor of two ( han 1998 ; kalogera 2000 ) .
the most common systems should be those containing two low - mass helium wd .
their final coalescence can produce an object massive enough to start helium burning .
bailyn ( 1993 and references therein ) and others have suggested that some `` extreme horizontal branch '' stars in globular clusters may be such helium - burning stars formed by the coalescence of two wd .
planets in orbit around a massive wd may also form following the binary coalescence ( livio , pringle , & saffer 1992 ) .
coalescing wd binaries are also important sources of low - frequency gravitational waves that should be easily detectable by future space - based laser interferometers .
the currently planned lisa ( laser interferometer space antenna ; see folkner 1998 ) should have an extremely high sensitivity ( down to a characteristic strain @xmath8 ) to sources with frequencies in the range @xmath9hz . han ( 1998 ) estimated a wd merger rate @xmath10 in our own galaxy .
individual coalescing systems and mergers may be detectable in the frequency range @xmath4@xmath11mhz .
in addition , the total number ( @xmath12 ) of close wd binaries in our galaxy emitting at lower frequencies @xmath13@xmath14mhz ( the emission lasting for @xmath15@xmath16yr before final merging ) should provide a continuum background signal of amplitude @xmath17@xmath18 ( hils et al .
the detection of the final burst of gravitational waves emitted during an actual merger would provide a unique opportunity to observe in `` real time '' the hydrodynamic interaction between the two degenerate stars , possibly followed immediately by a supernova explosion , nuclear outburst , or some other type of electromagnetic signal .
the final hydrodynamic merger of two ns is driven by a combination of relativistic and fluid effects . even in newtonian gravity , an innermost stable circular orbit ( isco )
is imposed by _ global hydrodynamic instabilities _ , which can drive a close binary system to rapid coalescence once the tidal interaction between the two stars becomes sufficiently strong .
the existence of these global instabilities for close binary equilibrium configurations containing a compressible fluid , and their particular importance for binary ns systems , were demonstrated for the first time by rasio & shapiro ( 1992 , 1994 , 1995 ; hereafter rs13 ) using numerical hydrodynamic calculations .
these instabilities can also be studied using analytic methods .
the classical analytic work for close binaries containing an incompressible fluid ( e.g. , chandrasekhar 1969 ) was extended to compressible fluids in the work of lai , rasio , & shapiro ( 1993a , b , 1994a , b , c , hereafter lrs15 ) .
this analytic study confirmed the existence of dynamical instabilities for sufficiently close binaries .
although these simplified analytic studies can give much physical insight into difficult questions of global fluid instabilities , fully numerical calculations remain essential for establishing the stability limits of close binaries accurately and for following the nonlinear evolution of unstable systems all the way to complete coalescence .
a number of different groups have now performed such calculations , using a variety of numerical methods and focusing on different aspects of the problem .
nakamura and collaborators ( see nakamura & oohara 1998 and references therein ) were the first to perform 3d hydrodynamic calculations of binary ns coalescence , using a traditional eulerian finite - difference code .
instead , rs used the lagrangian method sph ( smoothed particle hydrodynamics ) .
they focused on determining the isco for initial binary models in strict hydrostatic equilibrium and calculating the emission of gravitational waves from the coalescence of unstable binaries .
many of the results of rs were later independently confirmed by new & tohline ( 1997 ) and swesty , wang , & calder ( 1999 ) , who used completely different numerical methods but also focused on stability questions , and by zhuge , centrella , & mcmillan ( 1994 , 1996 ) , who also used sph .
zhuge et al .
( 1996 ) also explored in detail the dependence of the gravitational wave signals on the initial ns spins .
davies et al .
( 1994 ) and ruffert et al .
( 1996 , 1997 ) have incorporated a treatment of the nuclear physics in their hydrodynamic calculations ( done using sph and ppm codes , respectively ) , motivated by models of grbs at cosmological distances .
all these calculations were performed in _
newtonian gravity _
, with some of the more recent studies adding an approximate treatment of energy and angular momentum dissipation through the gravitational radiation reaction ( e.g. , janka et al . 1999 ; rosswog et al .
1999 ) , or even a full treatment of pn gravity to lowest order ( ayal et al . 2000 ; faber & rasio 2000 ) .
all recent hydrodynamic calculations agree on the basic qualitative picture that emerges for the final coalescence ( see fig . 1 ) . as the isco is approached , the secular orbital decay driven by gravitational wave emission is dramatically accelerated ( see also lrs2 , lrs3 ) .
the two stars then plunge rapidly toward each other , and merge together into a single object in just a few rotation periods . in the corotating frame of the binary ,
the relative radial velocity of the two stars always remains very subsonic , so that the evolution is nearly adiabatic .
this is in sharp contrast to the case of a head - on collision between two stars on a free - fall , radial orbit , where shock heating is very important for the dynamics ( rs1 ; shapiro 1998 ) . here
the stars are constantly being held back by a ( slowly receding ) centrifugal barrier , and the merging , although dynamical , is much more gentle .
after typically @xmath19 orbital periods following first contact , the innermost cores of the two stars have merged and a secondary instability occurs : _ mass shedding _ sets in rather abruptly . material ( typically @xmath4% of the total mass ) is ejected through the outer lagrange points of the effective potential and spirals out rapidly . in the final stage , the spiral arms
widen and merge together , forming a nearly axisymmetric thick disk or torus around the inner , maximally rotating dense core .
= 5.truein in gr , strong - field gravity between the masses in a binary system is alone sufficient to drive a close circular orbit unstable . in close ns binaries , gr effects
combine nonlinearly with newtonian tidal effects so that the isco is encountered at larger binary separations and lower orbital frequency than predicted by newtonian hydrodynamics alone , or gr alone for two point masses .
the combined effects of relativity and hydrodynamics on the stability of close compact binaries have only very recently begun to be studied , using both analytic approximations ( basically , pn generalizations of lrs ; see , e.g. , lai & wiseman 1997 ; lombardi , rasio , & shapiro 1997 ; shibata & taniguchi 1997 ) , as well as numerical calculations in 3d incorporating simplified treatments of relativistic effects ( e.g. , baumgarte et al .
1998 ; marronetti , mathews & wilson 1998 ; wang , swesty , & calder 1998 ; faber & rasio 2000 ) .
several groups have been working on a fully general relativistic calculation of the final coalescence , combining the techniques of numerical relativity and numerical hydrodynamics in 3d ( baumgarte , hughes , & shapiro 1999 ; landry & teukolsky 2000 ; seidel 1998 ; shibata & uryu 2000 ) .
however this work is still in its infancy , and only very preliminary results of test calculations have been reported so far . the final fate of a ns
ns merger depends crucially on the ns eos , and on the extraction of angular momentum from the system during the final merger . for a stiff ns eos ,
it is by no means certain that the core of the final merged configuration will collapse on a dynamical timescale to form a bh .
one reason is that the kerr parameter @xmath20 of the core may exceed unity for extremely stiff eos ( baumgarte et al .
1998 ) , although newtonian and pn hydrodynamic calculations suggest that this is never the case ( see , e.g. , faber & rasio 2000 ) .
more importantly , the rapidly rotating core may in fact be dynamically stable .
take the obvious example of a system containing two identical @xmath21 ns .
the total baryonic mass of the system for a stiff ns eos is then about @xmath22 .
almost independent of the spins of the ns , all hydrodynamic calculations suggest that about @xmath23 of this mass will be ejected into the outer torus , leaving at the center a _ maximally rotating _ object with baryonic mass @xmath24 ( any hydrodynamic merger process that leads to mass shedding will produce a maximally rotating object since the system will have ejected just enough mass and angular momentum to reach its new , stable quasi - equilibrium state ) .
most stiff ns eos ( including the well - known `` au '' and `` uu '' eos of wiringa et al . 1988 ; see akmal et al.1998 for a recent update ) allow stable , maximally rotating ns with baryonic masses exceeding @xmath22 ( cook , shapiro , & teukolsky 1994 ) , i.e. , well above the mass of the final merger core .
differential rotation ( not taken into account in the calculations of cook et al .
1994 ) can further increase this maximum stable mass very significantly ( see baumgarte , shapiro , & shibata 2000 ) .
thus the hydrodynamic merger of two ns with stiff eos and realistic masses is not expected to produce a bh .
this expectation is confirmed by the preliminary full - gr calculations of shibata & uryu ( 2000 ) , for polytropes with @xmath25 , which indicate collapse to a bh only when the two ns are initially very close to the maximum stable mass . for _ slowly rotating _ stars , the same stiff ns eos give maximum stable baryonic masses in the range @xmath26 , which may or may not exceed the total merger core mass .
therefore , collapse to a bh could still occur on a timescale longer than the dynamical timescale , following a significant loss of angular momentum .
indeed , processes such as electromagnetic radiation , neutrino emission , and the development of various secular instabilities ( e.g. , r - modes ) , which may lead to angular momentum losses , take place on timescales much longer than the dynamical timescale ( see , e.g. , baumgarte & shapiro 1998 , who show that neutrino emission is probably negligible ) .
these processes are therefore decoupled from the hydrodynamics of the coalescence .
unfortunately their study is plagued by many fundamental uncertainties in the microphysics .
the question of the final fate of the merger could also depend crucially on the ns spins and on the evolution of the fluid vorticity during the final coalescence .
close ns binaries are likely to be _ nonsynchronized_. indeed , the tidal synchronization time is almost certainly much longer than the orbital decay time ( kochanek 1992 ; bildsten & cutler 1992 ) . for ns binaries that are far from synchronized
, the final coalescence involves some new , complex hydrodynamic processes ( rasio & shapiro 1999 ) .
consider for example the case of an irrotational system ( containing two nonspinning stars at large separation ; see lrs3 ) . because the two stars appear to be counter - spinning in the corotating frame of the binary , a _ vortex sheet _
( where the tangential velocity jumps discontinuously by @xmath27 ) appears when the stellar surfaces come into contact .
such a vortex sheet is kelvin - helmholtz unstable on all wavelengths and the hydrodynamics is therefore extremely difficult to model accurately given the limited spatial resolution of 3d calculations .
the breaking of the vortex sheet generates some turbulent viscosity so that the final configuration may no longer be irrotational . in numerical simulations
, however , vorticity is quickly generated through spurious shear viscosity , and the merger remnant is observed to evolve rapidly ( in just a few rotation periods ) toward uniform rotation .
the final fate of the merger could be affected drastically by these processes .
in particular , the shear flow inside the merging stars ( which supports a highly triaxial shape ; see rasio & shapiro 1999 ) may in reality persist long enough to allow a large fraction of the total angular momentum in the system to be radiated away in gravitational waves during the hydrodynamic phase of the coalescence . in this case
the final merged core may resemble a dedekind ellipsoid , i.e. , it will have a triaxial shape supported entirely by internal fluid motions , but with a stationary shape in the inertial frame ( so that it no longer radiates gravitational waves ) .
this state will be reached on the gravitational radiation reaction timescale , which is no more than a few tens of rotation periods .
on the ( much longer ) _ viscous timescale _ , the core will then evolve to a uniform , slowly rotating state and will probably collapse to a bh .
in contrast , in all 3d numerical simulations performed to date , the shear is quickly dissipated , so that gravitational radiation never gets a chance to extract more than a small fraction ( @xmath28% ) of the angular momentum , and the final core appears to be a uniform , maximally rotating object ( stable to collapse ) exactly as in calculations starting from synchronized binaries .
however this behavior is most likely an artefact of the large spurious shear viscosity present in the 3d simulations .
in addition to their obvious significance for gravitational wave emission , these issues are also of great importance for models of grbs that depend on energy extraction from a torus of material around the central bh .
indeed , if a large fraction of the total angular momentum is removed by the gravitational waves , rotationally - induced mass shedding may not occur at all during the merger , eventually leaving a bh with no surrounding matter , and no way of extracting energy from the system .
note also that , even without any additional loss of angular momentum through gravitational radiation , pn effects tend to reduce drastically the amount of matter ejected during the merger ( faber & rasio 2000 ) .
the results of rs3 for polytropes with @xmath29 show that hydrodynamics also plays an important role in the coalescence of two wd , either because of dynamical instabilities of the equilibrium configuration , or following the onset of dynamically unstable mass transfer .
systems with mass ratios @xmath30 must evolve into deep contact before they become dynamically unstable and merge . instead , equilibrium configurations for binaries with @xmath31 sufficiently far from unity never become dynamically unstable .
however , once these binaries reach their roche limit , dynamically unstable mass transfer occurs and the less massive star is completely disrupted after a small number ( @xmath32 ) of orbital periods ( see also benz et al .
1990 ) . in both cases ,
the final merged configuration is an axisymmetric , rapidly rotating object with a core
thick disk structure similar to that obtained for coalescing ns ( rs2 , rs3 ; see also mochkovitch & livio 1989 ) . for two massive enough wd
, the merger product may be well above the chandrasekhar mass @xmath33 .
the object may therefore explode as a ( type ia ) supernova , or perhaps collapse to a ns .
the rapid rotation and possibly high mass ( up to @xmath34 ) of the object must be taken into account for determining its final fate .
unfortunately , rapid rotation and the possibility of starting from an object well above the chandrasekhar limit have not been taken into account in most previous theoretical calculations of `` accretion - induced collapse '' ( aic ) , which consider a nonrotating wd just below the chandrasekhar limit accreting matter slowly and quasi - spherically ( e.g. , canal et al . 1990 ; nomoto & kondo 1991 ; see fryer et al .
1999 for a recent 2-d sph calculation including rotation ) . under these assumptions
it is found that collapse to a ns is possible only for a narrow range of initial conditions . in most cases ,
a supernova explosion follows the ignition of the nuclear fuel in the degenerate core .
however , the fate of a much more massive object with substantial rotational support and large deviations from spherical symmetry ( as would be formed by dynamical coalescence ) may be very different .
if a ns does indeed form , and later accretes some of the material ejected during the coalescence , a millisecond radio pulsar may emerge .
planets around this millisecond pulsar may be formed at large distances @xmath35au following the viscous evolution of the remaining material in the outer disk ( podsiadlowski , pringle & rees 1991 ; phinney & hansen 1993 ) .
this is one of the possible formation scenarios for the extraordinary planetary system discovered around the millisecond pulsar psr b1257@xmath512 ( see wolszczan 1999 for a recent update ; podsiadlowski 1993 for alternative planet formation scenarios ) .
this system contains three confirmed earth - mass planets in quasi - circular orbits ( wolszczan & frail 1992 ; wolszczan 1994 ) .
the planets have masses of @xmath36 , @xmath37 , and @xmath38 , where @xmath39 , @xmath40 and @xmath41 are the inclinations of the orbits with respect to the line of sight , and are at distances of 0.19au , 0.36au , and 0.47au , respectively , from the pulsar .
in addition , the unusually large second and third frequency derivatives of the pulsar suggest the existence of a fourth , more distant and massive planet in the system ( wolszczan 1999 ) .
the simplest interpretation of the present best - fit values of the frequency derivatives implies a mass of about @xmath42 ( i.e. , comparable to saturn s mass ) for the fourth planet , at a distance of about @xmath43 ( i.e. , comparable to pluto s distance from the sun ) , and with a period of about @xmath44 in a circular , coplanar orbit ( wolszczan 1996 ; joshi & rasio 1997 ) .
however , if , as may well be the case , the first pulse frequency derivative is not entirely acceleration - induced , then the fourth planet can have a wide range of masses ( joshi & rasio 1997 ) . in particular , it can have a mass comparable to that of mars ( at a distance of @xmath45au ) , uranus ( at a distance of @xmath46au ) or neptune ( at a distance of @xmath47au ) .
the presence of this fourth planet , if confirmed , would place strong additional constraints on possible formation scenarios , as both the minimum mass and minimum angular momentum required in the protoplanetary disk would increase considerably ( see phinney & hansen 1993 for a general discussion ) .
this work was supported by nsf grant ast-9618116 , nasa atp grant nag5 - 8460 , and by an alfred p. sloan research fellowship .
our computational work is supported by the national computational science alliance .
wolszczan , a. 1999 , in pulsar timing , general relativity and the internal structure of neutron stars , ed .
z. arzoumanian , f. van der hooft , & e.p.j .
van den heuvel ( amsterdam : koninklijke nederlandse akademie van wetenschappen ) , 101
|
coalescing compact binaries are thought to be involved in a wide variety of astrophysical phenomena . in particular , they are important sources of gravitational radiation for both ground - based and space - based laser - interferometer detectors , and they may be sources of supernova explosions or gamma - ray bursts .
mergers of two white dwarfs may produce neutron stars with peculiar properties , including perhaps millisecond radio pulsars sometimes accompanied by planets ( as observed in psr 1257 + 12 ) . according to a widely held belief
, the coalescence of two neutron stars should produce a rapidly rotating black hole surrounded by an accretion disk or torus , but this is by no means certain .
this review paper focuses on the final hydrodynamic coalescence and merger of double neutron stars and double white dwarfs , and addresses the question of the nature of the final merger products .
|
let @xmath0 be a set of @xmath1 distinct points ( called _ nodes _ or _ sample points _ ) of @xmath2 , and let @xmath3 be a function defined on a domain @xmath4 containing @xmath5 .
the classical shepard operators ( first introduced in @xcite in the particular case @xmath6 ) are defined by@xmath7 \left ( \boldsymbol{x}\right ) : = { \displaystyle\sum \limits_{i=1}^{n}}a_{\mu , i}\left ( \boldsymbol{x}\right ) f\left ( \boldsymbol{x}_{i}\right ) , \quad\mu>0 , \label{shepoper}\ ] ] where the weight functions @xmath8 in barycentric form are@xmath9 and @xmath10 denotes the euclidean norm in @xmath11 .
the interpolation operator @xmath12 $ ] is stable , in the sense that@xmath13 \left ( \boldsymbol{x}\right ) \leq\max_{i}f\left ( \boldsymbol{x}_{i}\right)\ ] ] but for @xmath14 the interpolating function @xmath15 \left ( \boldsymbol{x}\right ) $ ] has flat spots in the neighborhood of all data points .
moreover , the degree of exactness of the operator @xmath16 $ ] is @xmath17 , i.e. if it is restricted to the polynomial space @xmath18 , then @xmath19 \right\vert _ { \mathcal{p}_{\boldsymbol{x}}^{m}}=id_{\mathcal{p}_{\boldsymbol{x}}^{m}}$ ] ( the identity function on @xmath20 ) only for @xmath21 . these drawbacks , in particular , can be avoided by replacing each value @xmath22 in ( [ shepoper ] ) with an interpolation operator in @xmath23 , applied to @xmath3 , with a degree of exactness @xmath24 @xcite .
in addition , to make the shepard method a local one , according to @xcite we multiply euclidean distances @xmath25 in ( [ sheweighfun ] ) by franke - little weights @xcite @xmath26 where @xmath27 is the positive part of the argument . as a result ,
functions @xmath8 are replaced by compact support functions @xmath28 @xmath29 defined as follows : @xmath30 where@xmath31 and @xmath32 is the radius of influence about node @xmath23 : in practice @xmath32 is taken to be just large enough to include @xmath33 nodes in the open ball @xmath34 @xcite .
more precisely , if @xmath35 $ ] denotes an interpolation operator in @xmath23 based on data in the ball @xmath36 , @xmath37 , the _ local combined shepard operators _ are then defined as follows : @xmath38(\boldsymbol{x}):={\displaystyle\sum\limits_{i=1}^{n}}\widetilde{w}_{\mu , i}\left ( \boldsymbol{x}\right ) p[f,\boldsymbol{x}_{i}]\left ( \boldsymbol{x}\right ) .
\label{improvshep}\ ] ] we emphasize that weight functions @xmath39 have the same basic properties as weight functions @xmath8 , but the value of the improved shepard operator ( [ improvshep ] ) at a point @xmath40 is influenced only by values @xmath41\left ( \boldsymbol{x}\right ) $ ] with @xmath42 .
starting from shepard himself various combinations have been proposed in order to improve the approximation properties of the shepard operators ( see , for example , @xcite ) : shepard - bernoulli operators introduced recently in @xcite represent a further attempt .
these univariate operators are obtained by replacing each value @xmath43 in ( [ shepoper ] ) with the generalized taylor polynomial @xmath44(x)$ ] , proposed by f. costabile in @xcite . in @xcite the rate of convergence of the shepard - bernoulli operators is studied in depth and numerical examples demonstrate the accuracy of the proposed combination in special situations , in particular , when it is applied to the problem of interpolating the discrete solutions of initial value problems for ordinary differential equations . in the conclusion of @xcite the possibility of extending the shepard - bernoulli operator to higher dimension
was hypothesized by using the expansions studied in @xcite . in 2007
t. catinas @xcite combined classic shepard operators with the tensorial extension of the generalized taylor polynomial discussed in @xcite .
the resulting combination has separated degree of exactness @xmath45 with respect to @xmath46 and @xmath47 with respect to @xmath48 when applied to sufficiently smooth functions in the convex hull of data ; it uses @xmath49 specially structured three - dimensional data ( in the situation depicted in @xcite each point of @xmath5 is the vertex of a rectangle with vertices in @xmath5 ) but , on account of the nature of the polynomial , it interpolates only @xmath1 of them .
the numerical results provided in the paper , on some of the test functions provided in @xcite , show that the accuracy of the operator can be improved by using compact support basis functions @xmath50 instead of the global basis functions @xmath51 . in this paper
we extend the shepard - bernoulli operators to the bivariate case using local basis functions @xmath50 and the three point interpolation polynomials discussed in @xcite and introduce a new combination which interpolates on all data used for its definition .
we do this by associating to each sample point @xmath23 a triangle with a vertex in it and other two vertices in @xmath36 ; the association is done in order to reduce the error of the three point interpolation polynomial based on the three vertices of the triangle . for fixed values of @xmath33
@xcite this choice allows us to reduce the error of the proposed combination . as a consequence , the resulting operator
not only interpolates at each sample point and increases by @xmath52 the degree of exactness of the shepard - taylor operator @xcite which uses the same data , but also improves its accuracy .
the paper is organized as follows .
we start section [ sec1 ] by briefly recalling the definition of the generalized taylor polynomial
. then we deal with the extension @xmath53 } [ f]$ ] of the generalized taylor polynomial to a generic simplex @xmath54 of vertices @xmath55 .
in particular we provide new results concerning : error of approximation , limit behaviour and interpolation conditions of the given extension . in section [ sec:4 ]
we use these results to define the bivariate shepard - bernoulli operators and to study their remainder terms and rate of convergence . in section [ sec:5 ]
we apply the bivariate shepard - bernoulli operators to the scattered data interpolation problem .
the numerical results on some commonly used test functions for scattered data approximation @xcite show that the bivariate interpolation scheme proposed here is comparable well with the better known operators qshep2d @xcite and cshep2d @xcite .
finally , in section [ sec:6 ] we draw conclusions .
the generalized taylor polynomial @xcite is an expansion in bernoulli polynomials , i.e. in polynomials defined recursively by means of the following relations @xcite @xmath56{ll}b_{0}\left ( x\right ) = 1 , & \\
b_{n}^{\prime}\left ( x\right ) = nb_{n-1}\left ( x\right ) , & n\geq1,\\ { \displaystyle\int\limits_{0}^{1}}b_{n}\left ( x\right ) dx=0 , & n\geq1 .
\end{array } \right . \label{berricrel}\ ] ] for functions in the class @xmath57 \right ) , \ a , b\in\mathbb{r},\ a < b$ ] , this expansion is realized by the equation @xmath58(x)+r_{m}\left [ f , a , b\right ] ( x),\quad x\in\left [ a , b\right ] , \label{taygenappfor}\ ] ] the polynomial approximant is defined by @xmath59(x)=f\left ( a\right ) + { \displaystyle\sum\limits_{k=1}^{m}}\frac{s_{k}\left ( \dfrac{x - a}{h}\right ) } { k!}h^{k-1}\left ( f^{(k-1)}\left ( b\right ) -f^{\left ( k-1\right ) } \left ( a\right ) \right ) \ \label{poltaygen}\ ] ] and the remainder term is @xmath60 ( x)=\frac{h^{m-1}}{m!}{\displaystyle\int \limits_{a}^{b}}f^{\left ( m\right ) } \left ( t\right ) \left ( b_{m}\left ( \frac{b - t}{h}\right ) -b_{m}\left ( \frac{\left ( x - t\right ) -\left [ x - t\right ] } { h}\right ) \right ) dt , \label{remtaygen}\ ] ] where we have set @xmath61 , @xmath62 and we have denoted by @xmath63 $ ] the integer part of the argument and @xmath64 .
the polynomial @xmath65 $ ] can be extended in a natural way to the whole real line ; in this case peano s kernel theorem @xcite provides an integral expression for the remainder ( [ remtaygen ] ) @xcite .
the main properties of the generalized taylor polynomial have been extensively studied in @xcite . in @xcite
the univariate expansion ( [ taygenappfor ] ) has been extended to a bivariate expansion for functions of class @xmath66 in the standard simplex @xmath67 which interpolates at the vertices of the simplex and it is exact in @xmath20 .
as mentioned in @xcite , this expansion can be generalized to a generic simplex of @xmath68 by means of a linear isomorphism . in this paper
we require this general expansion and , in order to formalize it , let us denote by @xmath69 the set of all pairs with non - negative integer components in the euclidean space @xmath70 . for @xmath71 , we use the notations @xmath72 , @xmath73 and @xmath74 if and only if @xmath75 for all @xmath76 .
moreover , we assume that @xmath4 is a compact convex domain and @xmath55 . to fix the ideas we set @xmath77 .
we denote by @xmath54 the simplex of vertices @xmath78 , i.e. the convex hull of the set @xmath79 .
the baricentric coordinates @xmath80 , of a generic point @xmath81@xmath82 , relative to the simplex @xmath54 , are defined by @xmath83{ccc}\lambda_{1}\left ( \boldsymbol{x}\right ) = \dfrac{a\left ( \boldsymbol{x},v_{2},v_{3}\right ) } { a\left ( v_{1},v_{2},v_{3}\right ) } , & \lambda _
{ 2}\left ( \boldsymbol{x}\right ) = \dfrac{a\left ( v_{1},\boldsymbol{x},v_{3}\right ) } { a\left ( v_{1},v_{2},v_{3}\right ) } , & \lambda_{3}\left ( \boldsymbol{x}\right ) = \dfrac{a\left ( v_{1},v_{2},\boldsymbol{x}\right ) } { a\left ( v_{1},v_{2},v_{3}\right ) } \end{array } \label{coorbar}\ ] ] where @xmath84 is the signed area of the simplex of vertices @xmath78@xmath85{ccc}1 & 1 & 1\\ x_{1 } & x_{2 } & x_{3}\\ y_{1 } & y_{2 } & y_{3}\end{array } \right\vert .\ ] ] if @xmath3 is a differentiable function , and @xmath86 and @xmath87 are two distinct vertices of the simplex @xmath54 , the derivative of @xmath3 along the directed line segment from @xmath86 to @xmath87 ( side of the simplex ) at @xmath88 is denoted by@xmath83{cc}d_{ij}f\left ( \boldsymbol{x}\right ) : = \left ( v_{i}-v_{j}\right ) \cdot\nabla f\left ( \boldsymbol{x}\right ) , & i , j=1,2,3,i\neq j , \end{array } \label{d_ij(f)}\ ] ] where @xmath89 is the dot product and @xmath90 .
the composition of derivatives along the directed sides of the simplex ( [ d_ij(f ) ] ) are denoted by @xmath83{ccc}d_{1}^{\beta}=d_{21}^{\beta_{1}}d_{31}^{\beta_{2 } } , & d_{2}^{\beta}=d_{12}^{\beta_{1}}d_{32}^{\beta_{2 } } , & d_{3}^{\beta}=d_{13}^{\beta_{1}}d_{23}^{\beta_{2}}. \end{array } \label{d0d1d2}\ ] ] the following theorem holds : [ gensimexp]let @xmath3 be a function of class @xmath91 .
then for each @xmath92 we have@xmath93 } [ f](\boldsymbol{x})+r_{m}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{1}\right ] } \left [ f\right ] \left ( \boldsymbol{x}\right ) \label{gspoltayexp}\ ] ] where@xmath94{l}p_{m}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{1}\right ] } [ f](\boldsymbol{x})=f\left ( v_{1}\right ) \\ \quad\quad+{\displaystyle\sum\limits_{j=1}^{m } } \left ( d_{1}^{\left ( 0,j-1\right ) } f\left ( v_{3}\right ) -d_{1}^{\left ( 0,j-1\right ) } f\left ( v_{1}\right ) \right ) \dfrac{s_{j}\left ( \lambda _ { 2}+\lambda_{3}\right ) } { j!}\\ \quad\quad+{\displaystyle\sum\limits_{i=1}^{m } } { \displaystyle\sum\limits_{j=1}^{m - i+1 } } \left ( \left ( -1\right ) ^{i+j}\left ( d_{2}^{\left ( j-1,i-1\right ) } f\left ( v_{2}\right ) -d_{2}^{\left ( j-1,i-1\right ) } f\left ( v_{1}\right ) \right ) \right .
\\ \quad\quad\left .
+ \left ( -1\right ) ^{j}\left ( d_{3}^{\left ( j-1,i-1\right ) } f\left ( v_{3}\right ) -d_{3}^{\left ( j-1,i-1\right ) } f\left ( v_{1}\right )
\right ) \right ) \\ \quad\quad\times\dfrac{\left ( \lambda_{2}+\lambda_{3}\right ) ^{i-1}s_{i}\left ( \frac{\lambda_{2}}{\lambda_{2}+\lambda_{3}}\right ) } { i!}\dfrac{s_{j}\left ( \lambda_{2}+\lambda_{3}\right ) } { j ! } , \end{array } \label{gspoltaygen}\ ] ] and @xmath95 } \left [ f\right ] \left ( \boldsymbol{x}\right ) $ ] is the remainder term .
[ proof sketch]rather than obtaining expansion ( [ gspoltaygen ] ) by using a linear transformation which maps points @xmath96 in points @xmath97 respectively , we proceed by adapting , in this general case , the extension technique to the simplex specially developed in @xcite for expansions in bernoulli and lidstone polynomials ( see also @xcite for extensions of asymmetric expansions ) .
we denote by @xmath98the line through @xmath99 , by @xmath100 the line through @xmath101 and by @xmath102 the line through @xmath103 .
let @xmath104 be an interior point .
we denote by @xmath105 the line through @xmath106 parallel to the line @xmath107 and by @xmath108 the intersection points between @xmath105 and @xmath109 , respectively , where @xmath110 .
we assign point @xmath106 to the line segment @xmath111 $ ] parameterized by@xmath56{cc}\begin{array } [ c]{c}x\left ( \lambda\right ) = x-\lambda_{2}h_{1}+\lambda\left ( \lambda _ { 2}+\lambda_{3}\right ) h_{1}\\ y\left ( \lambda\right ) = y-\lambda_{2}k_{1}+\lambda\left ( \lambda _ { 2}+\lambda_{3}\right ) k_{1}\end{array } , & \lambda\in\left [ 0,1\right ] .
\end{array } \right.\ ] ] the restriction of @xmath3 to @xmath112 is the univariate function @xmath113 in @xmath114 $ ] and we expand it by generalized taylor expansion ( [ taygenappfor ] ) . in this expansion
we replace parameter @xmath115 by the value @xmath116 which corresponds to point @xmath106 .
this results in an expansion in terms of the values at @xmath117 of @xmath3 and its derivatives in direction of the side @xmath99 .
we reach the vertices of the simplex @xmath54 by assigning points @xmath117 to the segments with end points @xmath101 and @xmath103 respectively and by repeated use of the expansion ( [ taygenappfor ] ) on @xmath3 and its directional derivatives ( see fig .
[ simplesso ] ) .
[ ptb ] simplesso.eps if @xmath92 an expression for the remainder @xmath95 } \left [ f\right ] \left ( \boldsymbol{x}\right ) $ ] can be obtained by a repeated use of ( [ remtaygen ] ) as in @xcite .
if @xmath40 the following theorem provides an expression for @xmath118 } \left [ f\right ] \left ( \boldsymbol{x}\right ) $ ] under only slightly stronger hypothesis on @xmath3 . more precisely , we consider the class @xmath119 of functions @xmath120 @xcite with partial derivatives @xmath121 lipschitz - continuous in @xmath4 for each @xmath122 and we set @xcite@xmath123 [ gentaypol]let @xmath3 be a function of class @xmath119 .
then for each @xmath40 we have@xmath124 } [ f](\boldsymbol{x})=t_{m}\left [ f , v_{1}\right ] \left ( \boldsymbol{x}\right ) + \delta_{m}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{1}\right ] } [ f](\boldsymbol{x } ) \label{ptplusdgen}\ ] ] where @xmath125 \left ( \boldsymbol{x}\right ) $ ] is the taylor polynomial of order @xmath45 for @xmath3 at @xmath126 ( * ? ? ?
1 ) and@xmath94{l}\delta_{m}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{1}\right ] } [ f](\boldsymbol{x})=\\ \quad{\displaystyle\sum\limits_{j=1}^{m } } { \displaystyle\int_{0}^{1 } } \frac{d_{31}^{m - j+2}d_{1}^{\left ( 0,j-1\right ) } f\left ( v_{1}+t\left ( v_{3}-v_{1}\right ) \right ) \left ( 1-t\right ) ^{m - j+1}}{\left ( m - j+1\right ) ! } dt\dfrac{s_{j}\left ( \lambda_{2}+\lambda_{3}\right ) } { j!}\\ \quad+{\displaystyle\sum\limits_{i=1}^{m } } { \displaystyle\sum\limits_{j=1}^{m - i+1 } } \left ( \left ( -1\right ) ^{i+j}{\displaystyle\int_{0}^{1 } } \frac{d_{21}^{m - j - i+3}d_{2}^{\left ( j-1,i-1\right ) } f\left ( v_{1}+t\left ( v_{2}-v_{1}\right ) \right ) \left ( 1-t\right ) ^{m - j - i+2}}{(m - j - i+2)!}dt\right .
\\ \quad+\left .
\left ( -1\right ) ^{j}{\displaystyle\int_{0}^{1 } } \frac{d_{31}^{m - j - i+3}d_{3}^{\left ( j-1,i-1\right ) } f\left ( v_{1}+t\left ( v_{3}-v_{1}\right ) \right ) \left ( 1-t\right ) ^{m - j - i+2}}{(m - j - i+2)!}dt\right ) \\ \quad\times\dfrac{\left ( \lambda_{2}+\lambda_{3}\right ) ^{i-1}s_{i}\left ( \frac{\lambda_{2}}{\lambda_{2}+\lambda_{3}}\right ) } { i!}\dfrac{s_{j}\left ( \lambda_{2}+\lambda_{3}\right ) } { j!}. \end{array } \label{dgen}\ ] ] by applying the taylor theorem with integral remainder ( * ? ? ?
* ch . 7 ) to all derivative differences in ( [ gspoltaygen ] ) we find@xmath124 } [ f](\boldsymbol{x})=t_{m}\left [ f , v_{1}\right ] \left ( \boldsymbol{x}\right ) + \delta_{m}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{1}\right ] } [ f](\boldsymbol{x})\ ] ] where @xmath127 } [ f](\boldsymbol{x})$ ] is as in ( [ dgen ] ) and @xmath83{l}\widetilde{t}_{m}[f , v_{1}](\boldsymbol{x}):=f\left ( v_{1}\right ) + { \displaystyle\sum\limits_{j=1}^{m } } { \displaystyle\sum\limits_{k=1}^{m - j+1 } } \dfrac{1}{k!}d_{1}^{\left ( 0,k+j-1\right ) } f\left ( v_{1}\right ) \dfrac{s_{j}\left ( \lambda_{2}+\lambda_{3}\right ) } { j!}\\ \quad+{\displaystyle\sum\limits_{i=1}^{m } } { \displaystyle\sum\limits_{j=1}^{m - i+1 } } { \displaystyle\sum\limits_{k=1}^{m - i - j+2 } } \dfrac{1}{k!}\left ( \left ( -1\right ) ^{i+j+k}d_{2}^{\left ( j-1+k , i-1\right ) } f\left ( v_{1}\right ) + \left ( -1\right ) ^{j+k}d_{3}^{\left ( j-1+k , i-1\right ) } f\left ( v_{1}\right ) \right ) \\
\quad\times\dfrac{\left ( \lambda_{2}+\lambda_{3}\right ) ^{i-1}s_{i}\left ( \frac{\lambda_{2}}{\lambda_{2}+\lambda_{3}}\right ) } { i!}\dfrac{s_{j}\left ( \lambda_{2}+\lambda_{3}\right ) } { j!}. \end{array}\ ] ] since @xmath127 } [ f](\boldsymbol{x})\equiv0 $ ] for each @xmath128 and @xmath129 } \left [ \cdot\right ] \right ) = m$ ] , the polynomial operator @xmath130 $ ] reproduces exactly polynomials up to the degree @xmath45 .
for this reason we can affirm that @xmath131 $ ] is the @xmath45-th order taylor polynomial @xmath125 $ ] for @xmath3 centered at @xmath126@xmath132 \left ( \boldsymbol{x}\right ) : = \sum \limits_{\substack{\left\vert \alpha\right\vert \leq m\\\alpha\in\mathbb{n } _ { 0}^{2}}}\frac{1}{\alpha!}\frac{\partial^{\left\vert \alpha\right\vert } f}{\partial x^{\alpha_{1}}\partial y^{\alpha_{2}}}\left ( v_{1}\right ) \left ( x - x_{1}\right ) ^{\alpha_{1}}\left ( y - y_{1}\right ) ^{\alpha_{2}}. \label{taypolv0}\ ] ] in fact , @xmath131 $ ] can be expressed , after some computation and rearrangement , in terms of some partial derivatives of @xmath3 up to the order @xmath45 at a point @xmath133 @xmath134 \left ( \boldsymbol{x}\right ) = \sum\limits_{\substack{\left\vert \alpha\right\vert \leq m\\\alpha\in\mathbb{n } _ { 0}^{2}}}\frac{\partial^{\left\vert \alpha\right\vert } f}{\partial x^{\alpha_{1}}\partial y^{\alpha_{2}}}\left ( v_{1}\right ) p_{\alpha}\left ( \boldsymbol{x}\right ) , \ ] ] where @xmath135 are polynomials of degree at most @xmath45 . since @xmath131 = f$ ] for each @xmath128 then @xmath136 generates @xmath20 , and therefore @xmath137 for each @xmath138 .
since @xmath130 $ ] reproduces exactly all polynomials in @xmath20 it follows that@xmath83{cc}\widetilde{t}_{m}\left
[ t_{m}\left [ f , v_{1}\right ] , v_{1}\right ] \left ( \boldsymbol{x}\right ) = t_{m}\left [ f , v_{1}\right ] \left ( \boldsymbol{x}\right ) , & \boldsymbol{x}\in d ; \end{array } \label{t_tilde_kequalt_k}\ ] ] on the other hand@xmath83{cc}\widetilde{t}_{m}\left [ t_{m}\left [ f , v_{1}\right ] , v_{1}\right ] \left ( \boldsymbol{x}\right ) = \widetilde{t}_{m}\left [ f , v_{1}\right ] \left ( \boldsymbol{x}\right ) , & \boldsymbol{x}\in d , \end{array } \label{t_tilde_k_f_v0equalt_tilde_kv0}\ ] ] since for each @xmath138 , such that @xmath139@xmath140 } { \partial x^{\alpha_{1}}\partial y^{\alpha_{2}}}\left ( v_{1}\right ) = \frac{\partial^{\left\vert \alpha\right\vert } f}{\partial x^{\alpha_{1}}\partial y^{\alpha_{2}}}\left ( v_{1}\right ) .\ ] ] therefore , by equaling the right hand side terms of ( [ t_tilde_kequalt_k ] ) and ( [ t_tilde_k_f_v0equalt_tilde_kv0 ] ) we get@xmath134 \left ( \boldsymbol{x}\right ) = t_{m}\left [ f , v_{1}\right ] \left ( \boldsymbol{x}\right ) .\ ] ] [ remgentaypol]in the hypothesis of theorem [ gentaypol ] equation ( [ gspoltayexp ] ) extends to all @xmath40 by setting@xmath141 } \left [ f\right ] \left ( \boldsymbol{x}\right ) = r_{m}^{t}\left [ f , v_{1}\right ] \left ( x , y\right ) \left ( \boldsymbol{x}\right ) -\delta _ { m}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{1}\right ] } [ f](\boldsymbol{x } ) \label{rrtminusdgen}\ ] ] where @xmath142 \left ( \boldsymbol{x}\right ) $ ] is the remainder in the taylor expansion ( * ? ? ?
1 ) . we are now able to give a bound for the remainder ( [ rrtminusdgen ] ) . from here on , we shall use the notation @xmath143 for the absolute value of the real number @xmath115 and @xmath144 for the modulus of the vector @xmath145 .
we set:@xmath146 and@xmath147 by settings ( [ coorbar ] ) and by the cauchy - schwarz inequality , we get the following inequalities@xmath94{lll}\left\vert \lambda_{i}\right\vert & \leq rs\left\vert \boldsymbol{x}-v_{1}\right\vert _ { 2 } , & i=2,3 , \end{array } \label{ineqmodl2l3}\]]@xmath94{ll}\left\vert \lambda_{2}+\lambda_{3}\right\vert & \leq rs\left\vert \boldsymbol{x}-v_{1}\right\vert _ { 2}\end{array } \label{ineqmodl2pl3}\ ] ] and@xmath83{ccc}\left\vert \dfrac{\partial\lambda_{i}}{\partial x}\right\vert \leq rs , & \left\vert \dfrac{\partial\lambda_{i}}{\partial y}\right\vert \leq rs , & i=2,3 . \end{array } \label{moddxlambdaidylambai}\ ] ] the following lemma provides bounds for the derivatives of @xmath3 along the directed sides of the simplex ( [ d0d1d2 ] ) .
[ lemma : moddjf]let @xmath148 .
the derivatives of @xmath3 along the directed sides of the simplex @xmath54 satisfy @xmath83{cccc}\left\vert d_{j}^{\beta}f\left ( \boldsymbol{x}\right ) \right\vert \leq2^{\left\vert \beta\right\vert } r^{\left\vert \beta\right\vert } \left\vert f\right\vert _ { \left\vert \beta\right\vert -1,1 } , & j=1,2,3 & \left\vert \beta\right\vert = m+1 , & \boldsymbol{x}\in d. \end{array } \label{seminorm2di}\ ] ] let us consider the case @xmath149 .
by ( [ d0d1d2 ] ) and ( [ d_ij(f ) ] ) we have @xmath83{l}d_{1}^{\beta}f\left ( \boldsymbol{x}\right ) = d_{21}^{\beta_{1}}d_{31}^{\beta_{2}}f\left ( \boldsymbol{x}\right ) = \sum\limits_{i=0}^{\beta_{1}}\binom{\beta_{1}}{i}(x_{2}-x_{1})^{\beta_{1}-i}\left ( y_{2}-y_{1}\right ) ^{i}\\ \quad\sum\limits_{j=0}^{\beta_{2}}\binom{\beta_{2}}{j}(x_{3}-x_{1})^{\beta _ { 2}-j}\left ( y_{3}-y_{1}\right ) ^{j}\frac{\partial^{\left\vert \beta\right\vert } f\left ( \boldsymbol{x}\right ) } { \partial x^{\left\vert \beta\right\vert -i - j}\partial y^{i+j}}. \end{array}\ ] ] since @xmath150 , at least one between @xmath151 and @xmath152 is greater than or equal to @xmath52 , @xmath153 , @xmath154 . if @xmath155@xmath156 otherwise@xmath157 @xmath153 , @xmath154 .
therefore we have@xmath83{l}\left\vert d_{1}^{\beta}f\left ( \boldsymbol{x}\right ) \right\vert = \left\vert \sum\limits_{i=0}^{\beta_{1}}\binom{\beta_{1}}{i}(x_{2}-x_{1})^{\beta_{1}-i}\left ( y_{2}-y_{1}\right ) ^{i}\right .
\\ \qquad\left .
\sum\limits_{j=0}^{\beta_{2}}\binom{\beta_{2}}{j}(x_{3}-x_{1})^{\beta_{2}-j}\left ( y_{3}-y_{1}\right ) ^{j}\frac{\partial ^{\left\vert \beta\right\vert } f\left ( \boldsymbol{x}\right ) } { \partial x^{\left\vert \beta\right\vert -i - j}\partial y^{i+j}}\right\vert \\
\qquad\leq\sum\limits_{i=0}^{\beta_{1}}\binom{\beta_{1}}{i}\left\vert \left\vert x_{2}-x_{1}\right\vert \right\vert _ { 2}^{\beta_{1}-i}\left\vert \left\vert y_{2}-y_{1}\right\vert \right\vert _ { 2}^{i}\\ \qquad\sum\limits_{j=0}^{\beta_{2}}\binom{\beta_{2}}{j}\left\vert \left\vert x_{3}-x_{1}\right\vert \right\vert _ { 2}^{\beta_{2}-j}\left\vert \left\vert y_{3}-y_{1}\right\vert \right\vert _ { 2}^{j}\left\vert \frac{\partial ^{\left\vert \beta\right\vert } f\left ( \boldsymbol{x}\right ) } { \partial x^{\left\vert \beta\right\vert -i - j}\partial y^{i+j}}\right\vert \\
\qquad\leq2^{\left\vert \beta\right\vert } \left\vert \left\vert v_{2}-v_{1}\right\vert \right\vert _ { 2}^{\beta_{1}}\left\vert \left\vert v_{3}-v_{1}\right\vert \right\vert _
{ 2}^{\beta_{2}}\left\vert f\right\vert _
{ \left\vert \beta\right\vert -1,1}\\ \qquad\leq2^{\left\vert \beta\right\vert } r^{\left\vert \beta\right\vert } \left\vert f\right\vert _ { \left\vert \beta\right\vert -1,1}. \end{array}\ ] ] the cases @xmath158 are analogous .
now , we are able to prove the following theorem .
[ deltabound]let @xmath3 be a function of class @xmath119 .
then for each @xmath40 we have the following bound for the derivative @xmath159 } \left [ f\right ] \left ( \boldsymbol{x}\right ) } { \partial x^{\alpha}\partial y^{\beta}}$ ] of the difference ( [ dgen ] ) , for each @xmath160 @xmath161@xmath162 } \left [ f\right ] \left ( \boldsymbol{x}\right ) } { \partial x^{\alpha}\partial y^{\beta}}\right\vert \leq\left\vert f\right\vert _ { m,1}c_{\alpha,\beta}\left ( m\right ) { \displaystyle\sum\limits_{l=\max\left\ { 1,\alpha+\beta\right\ } } ^{m } } r^{m+1-l}\left ( r^{2}s\right ) ^{l}\left\vert \boldsymbol{x}-v_{1}\right\vert _ { 2}^{l-\left ( \alpha+\beta\right ) } \label{deltab}\ ] ] where @xmath163 is a constant independent of @xmath3 or @xmath88 explicitly computable . from properties of bernoulli polynomials @xcite , @xcite@xmath164 it follows that@xmath94{ll}s_{n}\left ( t\right ) & = b_{n}\left ( t\right ) -b_{n}\\ & = { \textstyle\sum\limits_{k=0}^{n-1 } } \dbinom{n}{k}b_{k}t^{n - k},\quad n=1,2,\ldots \end{array } \label{snassum}\ ] ] moreover , we express the derivative @xmath165 as a combination of derivatives along the directed line segments from @xmath166 to @xmath126 and from @xmath167 to @xmath126 .
in fact , from relations ( [ d_ij(f ) ] ) we obtain@xmath83{c}\frac{\partial}{\partial x}=\frac{\partial\lambda_{3}}{\partial x}d_{31}+\frac{\partial\lambda_{2}}{\partial x}d_{21},\\ \frac{\partial}{\partial y}=\frac{\partial\lambda_{3}}{\partial y}d_{31}+\frac{\partial\lambda_{2}}{\partial y}d_{21 } , \end{array}\ ] ] and then@xmath83{c}\dfrac{\partial^{\alpha+\beta}\delta_{m}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{1}\right ] } \left [ f\right ] \left ( \boldsymbol{x}\right ) } { \partial x^{\alpha}\partial y^{\beta}}={\displaystyle\sum\limits_{p=0}^{\alpha } } { \displaystyle\sum\limits_{q=0}^{\beta } } \dbinom{\alpha}{p}\dbinom{\beta}{q}\left ( \dfrac{\partial\lambda_{2}}{\partial x}\right ) ^{p}\left ( \dfrac{\partial\lambda_{3}}{\partial x}\right ) ^{\alpha - p}\\ \times\left ( \dfrac{\partial\lambda_{2}}{\partial y}\right ) ^{q}\left ( \dfrac{\partial\lambda_{3}}{\partial y}\right ) ^{\beta - q}d_{21}^{p+q}d_{31}^{\alpha+\beta - p - q}\delta_{m}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{1}\right ] } \left [ f\right ] \left ( \boldsymbol{x}\right ) . \end{array } \label{dxalphadybetaf}\ ] ] by taking the modulus of both sides of ( [ dxalphadybetaf ] ) and by using relations ( [ moddxlambdaidylambai ] ) we get@xmath162 } [ f](\boldsymbol{x})}{\partial x^{\alpha}\partial y^{\beta}}\right\vert \leq{\displaystyle\sum\limits_{p=0}^{\alpha } } { \displaystyle\sum\limits_{q=0}^{\beta } } \dbinom{\alpha}{p}\dbinom{\beta}{q}\left ( rs\right ) ^{\alpha+\beta } \left\vert d_{21}^{p+q}d_{31}^{\alpha+\beta - p - q}\delta_{m}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{1}\right ] } \left [ f\right ] \left ( \boldsymbol{x}\right ) \right\vert . \label{moddxalphadybetaf}\ ] ] therefore we need to calculate and to bound @xmath168 } [ f](\boldsymbol{x } ) \label{d1gammadelta}\ ] ] for each @xmath169 , @xmath170 , @xmath171 ; @xmath172 . in order to calculate ( [ d1gammadelta ] )
we substitute relations ( [ snassum ] ) in equation ( [ dgen ] ) and , by the binomial theorem , we get@xmath94{l}\delta_{m}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{1}\right ] } [ f](\boldsymbol{x})=\\ \quad{\displaystyle\sum\limits_{j=1}^{m } } { \displaystyle\int_{0}^{1 } } \frac{d_{31}^{m - j+2}d_{1}^{\left ( 0,j-1\right ) } f\left ( v_{1}+t\left ( v_{3}-v_{1}\right ) \right ) \left ( 1-t\right ) ^{m - j+1}}{\left ( m - j+1\right ) ! } dt\dfrac{{\textstyle\sum\limits_{k=0}^{j-1 } } \binom{j}{k}b_{k}\left ( \lambda_{2}+\lambda_{3}\right ) ^{j - k}}{j!}\\ \quad+{\displaystyle\sum\limits_{i=1}^{m } } { \displaystyle\sum\limits_{j=1}^{m - i+1 } } \left ( \left ( -1\right ) ^{i+j}{\displaystyle\int_{0}^{1 } } \frac{d_{21}^{m - j - i+3}d_{2}^{\left ( j-1,i-1\right ) } f\left ( v_{1}+t\left ( v_{2}-v_{1}\right ) \right ) \left ( 1-t\right ) ^{m - j - i+2}}{(m - j - i+2)!}dt\right .
\\ \quad+\left .
\left ( -1\right ) ^{j}{\displaystyle\int_{0}^{1 } } \frac{d_{31}^{m - j - i+3}d_{3}^{\left ( j-1,i-1\right ) } f\left ( v_{1}+t\left ( v_{3}-v_{1}\right ) \right ) \left ( 1-t\right ) ^{m - j - i+2}}{(m - j - i+2)!}dt\right ) \\ \quad\times\dfrac{{\textstyle\sum\limits_{k=0}^{i-1 } } \binom{i}{k}b_{k}}{i!}\dfrac{\sum\limits_{l=0}^{j-1}\binom{j}{l}b_{l}\sum\limits_{v=0}^{k+j-1-l}\binom{k+j-1-l}{v}\lambda_{2}^{i+j-1-l - v}\lambda_{3}^{v}}{j!}. \end{array } \label{deltamsem}\ ] ] the application of the operator @xmath173 to @xmath127 } [ f](\boldsymbol{x})$ ] causes the disappearing of some addenda on the right hand side of ( [ deltamsem ] ) therefore , to highlight this operation we make some changes of dummy index .
firstly we set @xmath174 in the first sequence of sums @xmath175 and @xmath176 in the second sequence of sums @xmath177 and we get , by writing @xmath178 instead of @xmath179 and @xmath180 instead of @xmath181 @xmath94{l}\delta_{m}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{1}\right ] } [ f](\boldsymbol{x})=\\ \quad{\displaystyle\sum\limits_{j=1}^{m } } { \displaystyle\int_{0}^{1 } } \frac{d_{31}^{m - j+2}d_{1}^{\left ( 0,j-1\right ) } f\left ( v_{1}+t\left ( v_{3}-v_{1}\right ) \right ) \left ( 1-t\right ) ^{m - j+1}}{\left ( m - j+1\right ) ! } dt\dfrac{{\textstyle\sum\limits_{k=1}^{j } } \binom{j}{j-\kappa}b_{j - k}\left ( \lambda_{2}+\lambda_{3}\right ) ^{k}}{j!}\\ \quad{\displaystyle\sum\limits_{i=1}^{m } } { \displaystyle\sum\limits_{j = i}^{m } } \left ( \left ( -1\right ) ^{j+1}{\displaystyle\int_{0}^{1 } } \frac{d_{21}^{m - j-1 + 3}d_{2}^{\left ( j - i , i-1\right ) } f\left ( v_{1}+t\left ( v_{2}-v_{1}\right ) \right ) \left ( 1-t\right ) ^{m - j+1}}{(m - j+1)!}dt\right . \\
\quad+\left .
\left ( -1\right ) ^{j - i+1}{\displaystyle\int_{0}^{1 } } \frac{d_{31}^{m - j+2}d_{3}^{\left ( j - i , i-1\right ) } f\left ( v_{1}+t\left ( v_{3}-v_{1}\right ) \right ) \left ( 1-t\right ) ^{m - j+1}}{(m - j+1)!}dt\right ) \\
\quad\times\dfrac{{\displaystyle\sum\limits_{k=0}^{i-1 } } { \displaystyle\sum\limits_{l=0}^{j - i } } \binom{j - i+1}{l}b_{l}{\displaystyle\sum\limits_{v=0}^{k+j - i - l } } \binom{k+j - i - l}{v}\lambda_{2}^{j - l - v}\lambda_{3}^{v}}{i!(j - i+1)!}. \end{array}\ ] ] secondly we set @xmath182 in the sequence of sums @xmath183 and we get , by writing @xmath184 instead of @xmath115,@xmath94{l}\delta_{m}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{1}\right ] } [ f](\boldsymbol{x})=\\ \quad{\displaystyle\sum\limits_{j=1}^{m } } { \displaystyle\int_{0}^{1 } } \frac{d_{31}^{m - j+2}d_{1}^{\left ( 0,j-1\right ) } f\left ( v_{1}+t\left ( v_{3}-v_{1}\right ) \right ) \left ( 1-t\right ) ^{m - j+1}}{\left ( m - j+1\right ) ! } dt\dfrac{{\textstyle\sum\limits_{k=1}^{j } } \binom{j}{j - k}b_{j - k}\left ( \lambda_{2}+\lambda_{3}\right ) ^{k}}{j!}\\ \quad+{\displaystyle\sum\limits_{i=1}^{m } } { \displaystyle\sum\limits_{j = i}^{m } } \left ( \left ( -1\right ) ^{j+1}{\displaystyle\int_{0}^{1 } } \frac{d_{21}^{m - j-1 + 3}d_{2}^{\left ( j - i , i-1\right ) } f\left ( v_{1}+t\left ( v_{2}-v_{1}\right ) \right ) \left ( 1-t\right ) ^{m - j+1}}{(m - j+1)!}dt\right . \\ \quad+\left .
\left ( -1\right ) ^{j - i+1}{\displaystyle\int_{0}^{1 } } \frac{d_{31}^{m - j+2}d_{3}^{\left ( j - i , i-1\right ) } f\left ( v_{1}+t\left ( v_{3}-v_{1}\right )
\right ) \left ( 1-t\right ) ^{m - j+1}}{(m - j+1)!}dt\right ) \\
\quad\times\dfrac{\sum\limits_{k=0}^{i-1}\binom{i}{k}b_{k}\sum\limits_{l = i}^{j}\binom{j - i+1}{j - l}b_{j - l}\sum\limits_{v=0}^{k - i - l}\binom{k - i - l}{v}\lambda_{2}^{l - v}\lambda_{3}^{v}}{i!(j - i+1)!}. \end{array}\ ] ] thirdly we change the order of summations @xmath185 and @xmath186 and we get@xmath94{l}\delta_{m}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{1}\right ] } [ f](\boldsymbol{x})=\\ \quad{\displaystyle\sum\limits_{k=1}^{m } } { \displaystyle\sum\limits_{j = k}^{m } } { \displaystyle\int_{0}^{1 } } \frac{d_{31}^{m - j+2}d_{1}^{\left ( 0,j-1\right ) } f\left ( v_{1}+t\left ( v_{3}-v_{1}\right ) \right ) \left ( 1-t\right ) ^{m - j+1}}{\left ( m - j+1\right ) ! } dt\dfrac{\tbinom{j}{j - k}b_{j - k}\left ( \lambda_{2}+\lambda _ { 3}\right ) ^{k}}{j!}\\ \quad+{\displaystyle\sum\limits_{l=1}^{m } } { \displaystyle\sum\limits_{i=1}^{l } } { \displaystyle\sum\limits_{j = l}^{m } } \left ( \left ( -1\right ) ^{j+1}{\displaystyle\int_{0}^{1 } } \frac{d_{21}^{m - j-1 + 3}d_{2}^{\left ( j - i , i-1\right ) } f\left ( v_{1}+t\left ( v_{2}-v_{1}\right ) \right ) \left ( 1-t\right ) ^{m - j+1}}{(m - j+1)!}dt\right . \\ \quad+\left . \left ( -1\right ) ^{j - i+1}{\displaystyle\int_{0}^{1 } } \frac{d_{31}^{m - j+2}d_{3}^{\left ( j - i , i-1\right ) } f\left ( v_{1}+t\left ( v_{3}-v_{1}\right ) \right ) \left ( 1-t\right ) ^{m - j+1}}{(m - j+1)!}dt\right )
\\ \quad\times\dfrac{\sum\limits_{k=0}^{i-1}\binom{i}{k}b_{k}\binom{j - i+1}{j - l}b_{j - l}\sum\limits_{v=0}^{k - i - l}\binom{k - i - l}{v}\lambda_{2}^{l - v}\lambda_{3}^{v}}{i!(j - i+1)!}. \end{array}\ ] ] now it is easy to calculate ( [ d1gammadelta ] ) by using the relations@xmath83{cc}d_{21}\lambda_{2}=1 , & d_{21}\lambda_{3}=0\\ d_{31}\lambda_{2}=0 , & d_{31}\lambda_{3}=1 .
\end{array}\ ] ] in fact@xmath187{cc}\dfrac{k!}{\left ( k-\left\vert \gamma\right\vert \right ) ! } \left ( \lambda_{2}+\lambda_{3}\right ) ^{k-\left\vert \gamma\right\vert } , & k\geq\left\vert \gamma\right\vert , \\ 0 , & \text{otherwise,}\end{array } \right.\]]@xmath188{cc}\dfrac{\left ( l - v\right ) ! } { \left ( l - v-\gamma_{1}\right ) ! } \dfrac { v!}{\left ( v-\gamma_{2}\right ) ! } \lambda_{2}^{l - v-\gamma_{1}}\lambda _ { 3}^{v-\gamma_{2 } } , & \gamma_{2}\leq v\leq l-\gamma_{1},\\ 0 , & \text{otherwise,}\end{array } \right.\ ] ] and by the change of dummy index @xmath189 we get , by writing @xmath190 instead of @xmath191 and @xmath184 instead of @xmath178 in the first sequence of sums @xmath94{l}d_{21}^{\gamma_{1}}d_{31}^{\gamma_{2}}\delta_{m}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{1}\right ] } [ f](\boldsymbol{x})=\\ \quad{\displaystyle\sum\limits_{l=\max\{1,\left\vert \gamma\right\vert \}}^{m } } { \displaystyle\sum\limits_{j = l}^{m } } { \displaystyle\int_{0}^{1 } } \frac{d_{31}^{m - j+2}d_{1}^{\left ( 0,j-1\right ) } f\left ( v_{1}+t\left ( v_{3}-v_{1}\right ) \right ) \left ( 1-t\right ) ^{m - j+1}}{\left ( m - j+1\right ) ! } dt\dfrac{\binom{j}{j - l}b_{j - l}\dfrac{l!}{\left ( l-\left\vert \gamma\right\vert \right ) ! } \left ( \lambda_{2}+\lambda_{3}\right ) ^{l-\left\vert \gamma\right\vert } } { j!}\\ \quad{\displaystyle\sum\limits_{l=\max\{1,\left\vert \gamma\right\vert \}}^{m } } { \displaystyle\sum\limits_{i=1}^{l } } { \displaystyle\sum\limits_{j = l}^{m } } \left ( \left ( -1\right ) ^{j+1}{\displaystyle\int_{0}^{1 } } \frac{d_{21}^{m - j-1 + 3}d_{2}^{\left ( j - i , i-1\right ) } f\left ( v_{1}+t\left ( v_{2}-v_{1}\right ) \right ) \left ( 1-t\right ) ^{m - j+1}}{(m - j+1)!}dt\right . \\ \quad+\left .
\left ( -1\right ) ^{j - i+1}{\displaystyle\int_{0}^{1 } } \frac{d_{31}^{m - j+2}d_{3}^{\left ( j - i , i-1\right ) } f\left ( v_{1}+t\left ( v_{3}-v_{1}\right ) \right ) \left ( 1-t\right ) ^{m - j+1}}{(m - j+1)!}dt\right ) \\
\quad\times\dfrac{\sum\limits_{k=0}^{i-1}\binom{i}{k}b_{k}\binom{j - i+1}{j - l}b_{j - l}\sum\limits_{v=\left\vert \gamma\right\vert } ^{l}\binom { k - i - l}{v-\gamma_{1}}}{i!(j - i+1)!}\\ \quad\times\dfrac{\left ( l - v+\gamma_{1}\right ) ! } { \left ( l - v\right ) ! } \dfrac{\left ( v-\gamma_{1}\right ) ! } { \left ( v-\left\vert \gamma \right\vert \right ) ! } \lambda_{2}^{l - v}\lambda_{3}^{v-\left\vert \gamma\right\vert } .
\end{array } \label{d21d31fin}\ ] ] now , by taking the modulus of both sides of ( [ d21d31fin ] ) and by using relations ( [ ineqmodl2l3 ] ) , ( [ ineqmodl2pl3 ] ) and ( [ seminorm2di ] ) we get@xmath94{l}\left\vert d_{21}^{\gamma_{1}}d_{31}^{\gamma_{2}}\delta_{m}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{1}\right ] } [ f](\boldsymbol{x})\right\vert \leq2^{m+1}\left\vert f\right\vert _ { m,1}\\{\displaystyle\sum\limits_{l=\max\left\ { 1,\left\vert \gamma\right\vert \right\ } } ^{m } } \left ( { \displaystyle\sum\limits_{j = l}^{m } } \dfrac{\dbinom{j}{j - l}\left\vert b_{j - l}\right\vert \dfrac{l!}{\left ( l-\left\vert \gamma\right\vert \right ) ! } } { \left ( m - j+2\right )
! j!}\right .
\\ \quad\left .
+ { \displaystyle\sum\limits_{i=1}^{l } } { \displaystyle\sum\limits_{j = l}^{m } } \dfrac{{\displaystyle\sum\limits_{k=0}^{i-1 } } 2\dbinom{i}{k}\left\vert b_{k}\right\vert \binom{j - i+1}{j - l}\left\vert b_{j - l}\right\vert { \displaystyle\sum\limits_{v=\left\vert \gamma\right\vert } ^{l } } \binom{k - i - l}{v-\gamma_{1}}}{(m - j+2)!i!(j - i+1)!}\right . \\
\quad\left .
\times\dfrac{\left ( l - v+\gamma_{1}\right ) ! } { \left ( l - v\right ) ! } \dfrac{\left ( v-\gamma_{1}\right ) ! } { \left ( v-\left\vert \gamma\right\vert \right ) ! } \right ) r^{m+1}\left ( rs\right ) ^{l-\left\vert \gamma\right\vert } \left\vert \boldsymbol{x}-v_{1}\right\vert _ { 2}^{l-\left\vert \gamma\right\vert } \end{array } \label{deltaless}\ ] ] finally by using inequality ( [ moddxalphadybetaf ] ) we get@xmath83{l}\left\vert \dfrac{\partial^{\alpha+\beta}\delta_{m}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{1}\right ] } \left [ f\right ] \left ( \boldsymbol{x}\right ) } { \partial x^{\alpha}\partial y^{\beta}}\right\vert \leq2^{m+1}\left\vert f\right\vert _ { m,1}{\displaystyle\sum\limits_{p=0}^{\alpha } } { \displaystyle\sum\limits_{q=0}^{\beta } } \dbinom{\alpha}{p}\dbinom{\beta}{q}\\{\displaystyle\sum\limits_{l=\max\left\ { 1,\alpha+\beta\right\ } } ^{m } } \left ( { \displaystyle\sum\limits_{j = l}^{m } } \dfrac{\dbinom{j}{j - l}\left\vert b_{j - l}\right\vert \dfrac{l!}{\left ( l-\alpha-\beta\right ) ! } } { \left ( m - j+2\right ) ! j!}\right . \\
\quad\left .
+ { \displaystyle\sum\limits_{i=1}^{l } } { \displaystyle\sum\limits_{j = l}^{m } } \dfrac{{\displaystyle\sum\limits_{k=0}^{i-1 } } 2\dbinom{i}{k}\left\vert b_{k}\right\vert \dbinom{j - i+1}{j - l}\left\vert b_{j - l}\right\vert { \displaystyle\sum\limits_{v=\alpha+\beta}^{l } } \dbinom{k - i - l}{v-\alpha}}{(m - j+2)!i!(j - i+1)!}\right . \\
\quad\left .
\times\dfrac{\left ( l - v+p+q\right ) ! } { \left ( l - v\right ) ! } \dfrac{\left ( v - p - q\right ) ! } { \left ( v-\alpha-\beta\right ) ! } \right ) r^{m+1-l}\left ( r^{2}s\right ) ^{l}\left\vert \boldsymbol{x}-v_{1}\right\vert _ { 2}^{l-\left ( \alpha+\beta\right ) } \end{array}\ ] ] we get the bound ( [ deltab ] ) by the equality @xmath192 and by setting@xmath83{l}c_{\alpha,\beta}\left ( m\right ) = 2^{m+1+\alpha+\beta}\left ( { \displaystyle\sum\limits_{j = l}^{m } } \dfrac{\dbinom{j}{j - l}\left\vert b_{j - l}\right\vert \dfrac{l!}{\left ( l-\alpha-\beta\right ) ! } } { \left ( m - j+2\right ) ! j!}\right .
\\ \quad\left .
+ { \displaystyle\sum\limits_{i=1}^{l } } { \displaystyle\sum\limits_{j = l}^{m } } \dfrac{{\displaystyle\sum\limits_{k=0}^{i-1 } } 2\dbinom{i}{k}\left\vert b_{k}\right\vert \dbinom{j - i+1}{j - l}\left\vert b_{j - l}\right\vert { \displaystyle\sum\limits_{v=\alpha+\beta}^{l } } \dbinom{k - i - l}{v-\alpha}}{(m - j+2)!i!(j - i+1)!}\right . \\
\quad\left .
\times\dfrac{\left ( l - v+p+q\right ) ! } { \left ( l - v\right ) ! } \dfrac{\left ( v - p - q\right ) ! } { \left ( v-\alpha-\beta\right ) ! } \right ) .
\end{array } \label{c_alpha_beta}\ ] ] let us observe that the bound in theorem ( [ deltabound ] ) implies that for @xmath193 @xmath194 } [ f](\boldsymbol{x})=t_{m}\left [ f , v_{1}\right ] \left ( \boldsymbol{x}\right ) .\ ] ] [ theobound]in the hypothesis of theorem [ deltabound ] we have for each @xmath40 the following bound for the error ( [ rrtminusdgen]):@xmath83{l}\left\vert \dfrac{\partial^{\alpha+\beta}r_{m}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{1}\right ] } \left [ f\right ] \left ( \boldsymbol{x}\right ) } { \partial x^{\alpha}\partial y^{\beta}}\right\vert \leq\left\vert f\right\vert _ { m,1}\left ( \dfrac{2^{m-\left ( \alpha + \beta\right ) } } { \left ( m-\left ( \alpha+\beta\right ) -1\right ) ! } \left\vert \boldsymbol{x}-v_{1}\right\vert _ { 2}^{m+1-\left ( \alpha + \beta\right ) } \right .
\\ \left .
\qquad+c_{\alpha,\beta}\left ( m\right ) \sum\limits_{l=\max\left\ { 1,\alpha+\beta\right\ } } ^{m}r^{m+1-l}\left ( r^{2}s\right ) ^{l}\left\vert \boldsymbol{x}-v_{1}\right\vert _ { 2}^{l-\left ( \alpha+\beta\right ) } \right ) \end{array } \label{rembound}\ ] ] the thesis follows from ( [ rrtminusdgen ] ) by an application of the triangle inequality , by ( [ deltab ] ) bounding the derivatives of taylor remainder as in @xcite .
the previous result gives a positive answer to a conjecture made in @xcite .
therefore , we can refer to polynomial @xmath195 } [ f]$ ] as the generalized taylor polynomial on @xmath196 with respect to the vertex @xmath197 .
this polynomial joins other bivariate polynomials with similar properties ( see for example @xcite ) .
[ theoshape]in the hypothesis of theorem [ gentaypol ] for each @xmath138,@xmath198 , @xmath199 we have@xmath200 } [ f]}{\partial x^{\alpha}\partial y^{\beta}}(v_{1})=\frac{\partial^{\alpha+\beta}f}{\partial x^{\alpha}\partial y^{\beta}}\left ( v_{1}\right ) + o\left ( r^{m+1-\left ( \alpha+\beta\right ) } \right ) .
\label{derohmmimjp1}\ ] ] by theorem [ gentaypol ] and by interpolation conditions satisfied by the taylor polynomial @xcite it follows that@xmath200 } [ f]}{\partial x^{\alpha}\partial y^{\beta}}(v_{1})-\frac{\partial^{\alpha+\beta}f}{\partial x^{\alpha}\partial y^{\beta}}\left ( v_{1}\right ) = \frac{\partial^{\alpha+\beta}\delta _ { m}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{1}\right ] } \left [ f\right ] \left ( v_{1}\right ) } { \partial x^{\alpha}\partial y^{\beta}}.\ ] ] now we use the bound ( [ deltab ] ) in the particular case @xmath201 to get the thesis , recalling that @xmath202 depends only on the shape of the triangle . by rearranging the terms in the sums on the right hand side of ( [ gspoltaygen ] ) we note that@xmath203 } [ f](\boldsymbol{x})=f\left ( v_{1}\right ) \lambda_{1}+f\left ( v_{2}\right ) \lambda_{2}+f\left ( v_{3}\right ) \lambda_{3}\ ] ] is the lagrange interpolant at the nodes @xmath78 and therefore it does not depend on the choice of the vertex @xmath126 ; the polynomial @xmath83{c}p_{2}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{3}\right ] } [ f]\left ( \boldsymbol{x}\right ) = p_{1}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{1}\right ] } [ f](\boldsymbol{x})\\ \quad+\frac{1}{2}\lambda_{1}\lambda_{2}\left ( d_{2}^{\left ( 1,0\right ) } f\left ( v_{2}\right ) -d_{2}^{\left ( 1,0\right ) } f\left ( v_{1}\right ) \right ) \\ \quad+\frac{1}{2}\lambda_{1}\lambda_{3}\left ( d_{1}^{\left ( 0,1\right ) } f\left ( v_{1}\right ) -d_{1}^{\left ( 0,1\right ) } f\left ( v_{3}\right ) \right ) \\ \quad+\frac{1}{2}\lambda_{2}\lambda_{3}\left ( d_{3}^{\left ( 0,1\right ) } f\left ( v_{3}\right ) -d_{3}^{\left ( 0,1\right ) } f\left ( v_{2}\right ) \right ) \end{array}\ ] ] satisfies the same property and therefore joins well known quadratic triangular finite elements @xcite . for @xmath204 , @xmath124 } [ f]\left ( \boldsymbol{x}\right ) = p_{2}^{\left [ \delta_{2}\left ( v_{1},v_{2},v_{3}\right ) ; v_{3}\right ] } [ f]\left ( \boldsymbol{x}\right ) + \text{terms of degree at least } 3,\ ] ]
depends on the choice of the referring vertex .
the polynomial @xmath205 } [ f]$ ] can be used to improve the accuracy of approximation of the triangular shepard method @xcite .
let @xmath206 be fixed points of @xmath4 ; we set @xmath207 .
we now associate to each point @xmath86 a simplex with a vertex in @xmath86 for each @xmath29 . taking into account the bound ( [ deltab ] ) in theorem [ deltabound ] , for each fixed radius of influence @xmath32 about node @xmath86 @xcite we associate to @xmath86 the simplex @xmath208 which minimizes the quantity @xmath209 where , as above
, @xmath210 is the length of the longest side of the simplex @xmath211 and @xmath212 is twice the area of @xmath211 .
if @xmath213 denote the adjacent angles to the side of length @xmath210 , then @xmath214 depends only on the form of the triangle @xmath211 .
such a procedure can be well - defined if the following steps are followed : 1 .
enumerate the @xmath215 nodes in the closed ball @xmath216 according to increasing distance from @xmath86 using the induced order of the given set of interpolation nodes ; 2 . enumerate the triangles according increasing order of the vertices ; 3 .
get the first useful triangle .
[ defshepber]for each fixed @xmath217 and @xmath218 the bivariate shepard - bernoulli operator is defined by @xmath83{cc}s_{b_{m}}\left [ f\right ] \left ( \boldsymbol{x}\right ) = \sum\limits_{i=1}^{n}\widetilde{w}_{\mu , i}\left ( \boldsymbol{x}\right ) p_{m}^{\delta_{2}(i)}[f](\boldsymbol{x } ) , & \boldsymbol{x}\in d \end{array } \label{bshep}\ ] ] where @xmath219(\boldsymbol{x}),\ i=1,\ldots , n$ ] is the generalized taylor polynomial ( [ gspoltaygen ] ) over @xmath4 .
the remainder term is@xmath94{ll}r_{b_{m}}\left [ f\right ] \left ( \boldsymbol{x}\right ) = f\left ( \boldsymbol{x}\right ) -s_{b_{m}}\left [ f\right ] \left ( \boldsymbol{x}\right ) , & \boldsymbol{x}\in d. \end{array } \label{rshep}\ ] ] the following statements can be checked without any difficulty . the operator @xmath220 $ ] is an interpolation operator in @xmath221 @xmath222 .
in fact @xmath219(\boldsymbol{x})$ ] interpolates at @xmath221 @xmath29 and the assertion follows in view of the fact that the shepard basis is cardinal:@xmath223 the degree of exactness of the operator @xmath220 $ ] is @xmath45 , i.e. @xmath224 = p$ ] for each bivariate polynomial @xmath225 .
the assertion follows from the fact that the shepard basis is a partition of unity : @xmath226 since the degree of exactness of @xmath219 $ ] is @xmath45 for @xmath227 as for the continuity class of the shepard operator , and consequently the continuity class of the shepard - bernoulli operators , there is the following result @xcite .
if @xmath228,\ i=1, ... ,n$ ] are polynomial interpolation operators in @xmath229 , then the continuity class of the operator ( [ bshep ] ) depends upon @xmath230 and , for @xmath217 , is as follows : * if @xmath230 is an integer , then @xmath231 \in c^{\mu-1}$ ] ; * if @xmath230 is not an integer , then @xmath231 \in c^{[\mu]}$ ] ; here @xmath232 $ ] is the largest integer @xmath233 . for each @xmath234 s.t .
@xmath235 we have@xmath83{ll}\dfrac{\partial^{\alpha+\beta}}{\partial x^{\alpha}\partial y^{\beta}}s_{b_{m}}\left [ f\right ] \left ( v_{k}\right ) & = \dfrac{\partial ^{\alpha+\beta}}{\partial x^{\alpha}\partial y^{\beta}}p_{m}^{\delta_{2}(k)}[f]\left ( v_{k}\right ) \\ & = \dfrac{\partial^{\alpha+\beta}f}{\partial x^{\alpha}\partial y^{\beta}}\left ( v_{k}\right ) + o\left ( r_{k}^{m+1-\left ( \alpha+\beta\right ) } \right ) \end{array}\ ] ] for each @xmath236 .
it follows from the known relation @xcite @xmath237 which holds for @xmath238 , @xmath239 @xmath233 by applying the leibniz rule and by using relations ( [ cardinality ] ) , ( [ parunity ] ) and finally equation ( [ derohmmimjp1 ] ) .
convergence results can be obtained by following the known approaches @xcite , @xcite .
we set : 1 .
[ 1]@xmath240 , @xmath40 ; 2 .
@xmath241 ; 3 . @xmath242 , @xmath37 , 4 .
[ 4]@xmath243 , @xmath244 @xmath245 constants satisfying @xmath246 [ boundderf]let @xmath3 be a function of class @xmath119 . then for each @xmath244 @xmath247 the following bound holds@xmath83{l}\sup\limits_{\boldsymbol{x}\in d}\left\vert \dfrac{\partial^{\alpha+\beta } r_{b_{m}}\left [ f\right ] } { \partial x^{\alpha}\partial y^{\beta}}\right\vert \leq\left\vert f\right\vert _ { m,1}m{\displaystyle\sum\limits_{\substack{0\leq\gamma_{1}\leq\alpha\\0\leq \gamma_{2}\leq\beta } } } \dbinom{\alpha}{\gamma_{1}}\dbinom{\beta}{\gamma_{2}}\\ \quad\max\limits_{i\in\mathcal{i}_{\boldsymbol{x}}}\left\ { \dfrac { k_{\alpha-\gamma_{1},\beta-\gamma_{2}}}{d_{i}^{\alpha+\beta-\gamma_{1}-\gamma_{2}}}\left ( \dfrac{2^{m-\left ( \gamma_{1}+\gamma_{2}\right ) } } { \left ( m-\left ( \gamma_{1}+\gamma_{2}\right ) -1\right ) ! } r_{w_{i}}^{m+1-\left ( \gamma_{1}+\gamma_{2}\right ) } \right . \right . \\
\qquad+c_{\gamma_{1},\gamma_{2}}\left ( m\right ) \sum\limits_{l=\max\left\ { 1,\gamma_{1}+\gamma_{2}\right\ } } ^{m}r_{i}^{m+1-l}\left ( r_{i}^{2}s_{i}\right ) ^{l}r_{w_{i}}^{l-\left ( \gamma_{1}+\gamma_{2}\right ) } \right ) \right\ } \end{array}\ ] ] with @xmath248 defined in ( [ c_alpha_beta ] ) . by differentiating @xmath138 times with respect to @xmath46 and @xmath249 times with respect to @xmath48 , @xmath247 , both sides of ( [ rshep ] ) , by using leibniz rule
, we get@xmath83{l}\dfrac{\partial^{\alpha+\beta}r_{b_{m}}\left [ f\right ] \left ( \boldsymbol{x}\right ) } { \partial x^{\alpha}\partial y^{\beta}}=\\{\displaystyle\sum\limits_{i\in\mathcal{i}_{\boldsymbol{x } } } } { \displaystyle\sum\limits_{\substack{0\leq\gamma_{1}\leq\alpha\\0\leq \gamma_{2}\leq\beta } } } \dbinom{\alpha}{\gamma_{1}}\dbinom{\beta}{\gamma_{2}}\dfrac{\partial ^{\alpha+\beta-\gamma_{1}-\gamma_{2}}\widetilde{w}_{\mu , i}\left ( \boldsymbol{x}\right ) } { \partial x^{\alpha-\gamma_{1}}\partial y^{\beta -\gamma_{2}}}\dfrac{\partial^{\gamma_{1}+\gamma_{2}}r_{m}^{\delta_{2}\left ( i\right ) } \left [ f\right ] \left ( \boldsymbol{x}\right ) } { \partial x^{\gamma_{1}}\partial y^{\gamma_{2}}}. \end{array}\ ] ] therefore @xmath83{l}\left\vert \dfrac{\partial^{\alpha+\beta}r_{b_{m}}\left [ f\right ] \left ( \boldsymbol{x}\right ) } { \partial x^{\alpha}\partial y^{\beta}}\right\vert \leq\\ \qquad{\displaystyle\sum\limits_{i\in\mathcal{i}_{\boldsymbol{x } } } } { \displaystyle\sum\limits_{\substack{0\leq\gamma_{1}\leq\alpha\\0\leq \gamma_{2}\leq\beta } } } \dbinom{\alpha}{\gamma_{1}}\dbinom{\beta}{\gamma_{2}}\left\vert \dfrac { \partial^{\alpha+\beta-\gamma_{1}-\gamma_{2}}\widetilde{w}_{\mu , i}\left ( \boldsymbol{x}\right ) } { \partial x^{\alpha-\gamma_{1}}\partial y^{\beta -\gamma_{2}}}\right\vert \left\vert \dfrac{\partial^{\gamma_{1}+\gamma_{2}}r_{m}^{\delta_{2}\left ( i\right ) } \left [ f\right ] \left ( \boldsymbol{x}\right ) } { \partial x^{\gamma_{1}}\partial y^{\gamma_{2}}}\right\vert \\ \quad\leq{\displaystyle\sum\limits_{i\in\mathcal{i}_{\boldsymbol{x } } } } { \displaystyle\sum\limits_{\substack{0\leq\gamma_{1}\leq\alpha\\0\leq \gamma_{2}\leq\beta } } } \dbinom{\alpha}{\gamma_{1}}\dbinom{\beta}{\gamma_{2}}\dfrac{k_{\alpha -\gamma_{1},\beta-\gamma_{2}}}{d_{i}^{\alpha+\beta-\gamma_{1}-\gamma_{2}}}\left\vert f\right\vert _ { m,1}\left ( \dfrac{2^{m-\left ( \gamma_{1}+\gamma_{2}\right ) } } { \left ( m-\left ( \gamma_{1}+\gamma_{2}\right ) -1\right ) ! } \left\vert \boldsymbol{x}-v_{i}\right\vert _ { 2}^{m+1-\left ( \gamma_{1}+\gamma_{2}\right ) } \right . \\
\qquad+c_{\gamma_{1},\gamma_{2}}\left ( m\right ) \sum\limits_{l=\max \left\ { 1,\gamma_{1}+\gamma_{2}\right\ } } ^{m}r_{i}^{m+1-l}\left ( r_{i}^{2}s_{i}\right ) ^{l}\left\vert \boldsymbol{x}-v_{i}\right\vert _ { 2}^{l-\left ( \gamma_{1}+\gamma_{2}\right ) } \right ) \\
\quad\leq\left\vert f\right\vert _ { m,1}m{\displaystyle\sum\limits_{\substack{0\leq\gamma_{1}\leq\alpha\\0\leq \gamma_{2}\leq\beta } } } \dbinom{\alpha}{\gamma_{1}}\dbinom{\beta}{\gamma_{2}}\max\limits_{i\in \mathcal{i}_{\boldsymbol{x}}}\left\ { \dfrac{k_{\alpha-\gamma_{1},\beta -\gamma_{2}}}{d_{i}^{\alpha+\beta-\gamma_{1}-\gamma_{2}}}\left ( \dfrac{2^{m-\left ( \gamma_{1}+\gamma_{2}\right ) } } { \left ( m-\left ( \gamma_{1}+\gamma_{2}\right ) -1\right ) ! } r_{w_{i}}^{m+1-\left ( \gamma _ { 1}+\gamma_{2}\right ) } \right .
\right . \\
\qquad+c_{\gamma_{1},\gamma_{2}}\left ( m\right ) \sum\limits_{l=\max\left\ { 1,\gamma_{1}+\gamma_{2}\right\ } } ^{m}r_{i}^{m+1-l}\left ( r_{i}^{2}s_{i}\right ) ^{l}r_{w_{i}}^{l-\left ( \gamma_{1}+\gamma_{2}\right ) } \right ) \right\ } .
\end{array}\ ] ] in the following section we present numerical results that testify to the accuracy of the proposed operator .
( left ) , @xmath250 ( middle ) , and @xmath251 ( right ) nodes.,title="fig : " ] ( left ) , @xmath250 ( middle ) , and @xmath251 ( right ) nodes.,title="fig : " ] ( left ) , @xmath250 ( middle ) , and @xmath251 ( right ) nodes.,title="fig : " ] [ sec:5]to test the accuracy of approximation of the bivariate shepard - bernoulli operator we apply it to different sets of nodes in the rectangle @xmath252 \times\left [ 0,1\right ] $ ] ( see figure [ fig : set - of - nodes ] ) and @xmath253 test functions ( see figure [ fig : test - functions ] ) generally used in the multivariate interpolation of large sets of scattered data @xcite . in the following
we report the results of some of these experiments . in a first series of experiments
, we consider the case in which at each node @xmath86 function evaluations and derivative data up to the order @xmath254 are given . in this case , for each function @xmath255 we compare the numerical results obtained by applying the approximation operator @xmath256 $ ] with those obtained by the local version of the famous shepard - taylor operator @xcite@xmath257 \left ( x , y\right ) = \sum\limits_{i=1}^{n}\widetilde{w}_{\mu , i}\left ( x , y\right ) t_{2}\left [ f , v_{i}\right ] \left ( x , y\right ) \label{shepard - taylor}\ ] ] which uses the same data .
we report the results for the first four functions in figure [ fig : bshepvstshep ] , where we show the maximum interpolation errors , computed for the parameter value @xmath258 .
the remaining six functions have a similar behaviour and for this reason we omit them .
numerical results show that the operator @xmath259 improves the accuracy of the operator @xmath260 .
operator and bshep2d(3,2 ) operators when function evaluations and derivative data up to the order @xmath254 are given.,title="fig : " ] operator and bshep2d(3,2 ) operators when function evaluations and derivative data up to the order @xmath254 are given.,title="fig : " ] operator and bshep2d(3,2 ) operators when function evaluations and derivative data up to the order @xmath254 are given.,title="fig : " ] operator and bshep2d(3,2 ) operators when function evaluations and derivative data up to the order @xmath254 are given.,title="fig : " ] in a second series of experiments , we consider the case in which at each node @xmath86 only function evaluations are given . in this case , the second order derivatives @xmath261 are usually replaced by the coefficients @xmath262 of the quadratic polynomial @xmath263 which fits data values @xmath264 on a set of nearby nodes in a weighted least - square sense , as in the definition of operator qshep2d @xcite .
the procedure for computing these coefficients is well detailed in @xcite and it is based on the choice of another radius of influence about node @xmath86 , @xmath265 , which varies with @xmath266 and is taken to be just large enough to include @xmath267 nodes in @xmath268 . at the same time the derivative data ( [ derivativedata ] ) at @xmath269 can be replaced by the coefficients @xmath270 of the cubic polynomial @xmath271 which fits the data values @xmath264 on a set of nearby nodes in a weighted least - square sense , as in the definition of operator cshep2d in @xcite . in the following we denote by bshep2d(3,2 ) the shepard - bernoulli operator obtained by substituting the partial derivatives in @xmath256 $ ] with linear combinations of @xmath262 and by bshep2d(3,3 ) the shepard - bernoulli operator obtained by substituting the partial derivatives in @xmath256 $ ] with linear combinations of @xmath270 .
therefore the operator bshep2d(3,2 ) has degree of exactness @xmath254 as the operator qshep2d , while the operator bshep2d(3,3 ) has degree of exactness @xmath272 as the operator cshep2d .
we report the results for the first four functions in figure [ fig : bshepvsqshep ] , where we show the maximum interpolation errors , computed for the parameter value @xmath258 and @xmath273 for the operator qshep2d and @xmath274 when we replace the derivative data by using the cubic polynomial @xmath275 .
the remaining six functions have a similar behaviour and for this reason we omit them .
numerical results show that the operator @xmath259 improves the accuracy of the operator qshep2d and is comparable with the operator cshep2d . with regard to the computational cost we note that the point - triangle associations which reduce the error of the three point interpolation polynomials ( [ gspoltayexp ] ) involves an additional cost of @xmath276 calculations that do not modify the computational cost of operator qshep2d which is @xmath276 for uniform distributions of nodes and @xmath277 in worst cases @xcite .
on the other hand , local basis ( [ wtbasisfunctions ] ) used to define our operators , containing a considerably lower number of nodes compared to the choices recommended by renka , involve a better localization of the combined operator .
in this paper we propose a new definition of the bivariate shepard - bernoulli operators which avoids the drawbacks of catinas extension @xcite .
these new interpolation operators are realized by using local support basis functions introduced in @xcite instead of classical shepard basis functions and the bivariate three point extension @xcite of the generalized taylor polynomial introduced by f. costabile in @xcite .
their definition requires the association , to each sample point , of a triangle with a vertex in it and other two vertices in its neighborhood .
the proposed point - triangle association is carried out to reduce the error of the three point interpolation polynomial . as a consequence ,
the resulting operator not only inherits interpolation conditions that each three point local interpolation polynomial satisfies at the referring vertex and increases by @xmath52 the degree of exactness of the shepard - taylor operator @xcite which uses the same data , but also improves its accuracy . in this sense ,
the shepard - bernoulli operators belong to a recently introduced class of operators for enhancing the approximation order of shepard operators by using supplementary derivative data @xcite . for the general problem of the enhancement of the algebraic precision of linear operators of approximation see the papers @xcite and the references therein .
moreover , when applied to the scattered data interpolation problem , the shepard - bernoulli operator @xmath259 improves the accuracy of the operator qshep2d and is comparable with the operator cshep2d by renka @xcite . finally , the quadratic triangular finite element @xmath278 } [ f]\left ( \boldsymbol{x}\right ) $ ] can be used to improve the accuracy of approximation of the triangular shepard method @xcite .
costabile , f.a . ,
dellaccio , f. , luceri , r. : explicit polynomial expansions of regular real functions by means of even order bernoulli polynomials and boundary values .
journal of computational and applied mathematics , 176 , 77 - 90 ( 2005 ) xuli , h. : multi - node higher order expasions of a functions .
journal of approximation theory 124 , 242 - 253 ( 2003 ) hoschek , j. , lasser , d. : fundamentals of computer aided geometric design .
a. k. peters , usa ( 1989 )
|
in this paper we extend the shepard - bernoulli operators introduced in @xcite to the bivariate case .
these new interpolation operators are realized by using local support basis functions introduced in @xcite instead of classical shepard basis functions and the bivariate three point extension @xcite of the generalized taylor polynomial introduced by f. costabile in @xcite .
the new operators do not require either the use of special partitions of the node convex hull or special structured data as in @xcite .
we deeply study their approximation properties and provide an application to the scattered data interpolation problem ; the numerical results show that this new approach is comparable with the other well known bivariate schemes qshep2d and cshep2d by renka @xcite .
|
in the past decade , the study of systems of two - dimensional electron gas ( 2deg ) in semiconductors@xcite has been extended by the application of spatially inhomogeneous magnetic fields .
the inhomogeneity of the magnetic field can be realized experimentally either by varying the topography of the electron gas @xcite , by using ferromagnetic materials @xcite , depositing a superconductor on top of the 2deg @xcite .
numerous theoretical works also show an increasing interest in the study of electron motion in inhomogeneous magnetic field ( see , e.g. , refs . ) . in the experimental works mentioned above , for gaas heterostructures , on the one hand ,
the electron dynamics is confined to two - dimensions . on the other hand , the coherence length and
the mean free path of the electron can be much larger than the size of the system , while the fermi wavelength is comparable to the size of the 2deg .
moreover , the electron system can be described to a good approximation as a free electron gas with an effective mass@xcite .
therefore , the quantum mechanical treatment of these systems is of some physical interest . in this paper , as an example , we consider the energy levels of a two - dimensional non - interacting electron gas in a magnetic field that is zero inside a circular region and constant outside . this system ( shown in fig .
[ geo_bantidot : fig ] ) will be called a magnetic anti - dot ; it was first studied by solimany and kramer @xcite . solving the schrdinger equation
it was shown that there are bound states . introducing an effective angular momentum , the schrdinger equation of the particle in symmetric gauge
can be mapped to the landau model .
this effective angular momentum is a sum of the angular momentum in a uniform magnetic field and the flux ( in units of the flux quantum ) missing from the uniform field .
recently , sim et al .
@xcite have renewed the study of this system and pointed out the crucial role of the magnetic edge states in the magnetoconductance .
the classical counterparts of these states correspond to trajectories of the charged particle that consist of straight segments inside the non - magnetic region and arcs outside . and constant outside the circle .
[ geo_bantidot : fig ] ] although it is difficult to measure directly the density of states of a quantum system , it affects many observable quantities such as the magnetoconductance , the magnetization or the susceptibility . in the interpretation of the experimental results the semiclassical approximation proved to be a useful tool .
several semiclassical approaches @xcite are known in the literature and an excellent overview of the subject can be found in the textbook by brack and bhaduri@xcite .
different semiclassical theories for magnetic systems have successfully been applied , for example , in works @xcite . for integrable systems
berry and tabor @xcite have shown that the oscillating part of the density of states can always be expressed in terms of classical periodic orbits .
this formula is commonly called the berry - tabor trace formula .
one of our aims in this paper is to apply , for the first time , the berry - tabor trace formula for a magnetic anti - dot . to illustrate the power of the method
, we also calculate the exact eigenvalues of the single particle schrdinger equation and find a very good agreement between the two results .
we should mention here that the statement by sim et al .
@xcite on the relation between the quantum states and the corresponding periodic orbits is somewhat misleading .
their condition for a given periodic orbit is not necessarily satisfied at the value of the corresponding exact energy level as they claimed however , including more and more periodic orbits with the proper weights in the trace formula the sum converges to the correct quantum density of states . in practice only a few of the shortest orbits
are enough to get a rough estimate of the positions of the exact energy levels .
the power of the semiclassical approach can also be demonstrated by applying the bohr - sommerfeld approximation .
we shall show that the energy levels obtained from the bohr - sommerfeld quantization rules also agree very well with the numerically exact levels even for the lowest eigenstates .
note that in this semiclassical treatment the quantization should be applied to the classical motion on a two - dimensional torus parametrized by the action variables and their canonically conjugate angle variables ( for details see , eg , ref . ) .
the classical orbits can be classified by their cyclotron radius @xmath0 and their guiding center @xmath1 ( distance of the center of the orbits from the origin ) . in the quantum mechanical treatment
one can calculate the average of operators defining the cyclotron radius and the guiding center . for circular magnetic billiards ,
lent @xcite has derived approximate expressions for these averages for a given quantum state . following ref .
, one may derive the corresponding relations for magnetic anti - dot systems .
thus , these relations are the basis for classifying the different quantum states in terms of classical orbits for our system
. we will show that the quantum states can be described by six different types of classical orbits .
in addition , this classification enables us to draw a so - called ` phase diagram ' which shows a clear one - to - one correspondence between classical orbits and quantum states . to complete our semiclassical study , we finally present results for the probability current density calculated from quantum calculations .
we shall argue that the current flow patterns can be understood qualitatively from the corresponding classical trajectories .
recently , halperin @xcite has been shown that the total current ( the integral of the current density along the radial direction ) can be related to the dispersion of the energy levels ( their angular momentum dependence ) . as it will be shown this general relation works in our magnetic anti - dot system , too . regarding the numerical calculations
, we should mention that the semiclassical approach presented in this paper proves to be a very effective method .
moreover , it provides a better understanding of the nature of the quantum system .
our semiclassical method applied to magnetic anti - dot systems may be an important tool to understand the role of the magnetic edge states in the density of states or the magnetization ( both are experimentally accessible physical qauntities ) .
we believe that our semiclassical analysis can be extended to other types of inhomogeneous magnetic fields such as studied , eg , in refs . as well as non - circular dot systems .
the rest of the text is organized as follows . in sec .
[ quantum : seq ] the exact quantization condition ( secular equation ) is derived from the matching conditions of the wave functions at the boundary of the magnetic and non - magnetic regions . in sec .
[ berry - tabor : ref ] the semiclassical approximation is presented including the description of the classical motion of the particle in subsec .
[ class_dyn : sec ] , the characterization of the possible periodic orbits in subsec .
[ periodic_orbit : sec ] , some numerical results in subsec . [ berry_tabor_results : sec ] , and the phase diagram in subsec . [
phase_diag : sec ] .
the current flow patterns of the system are discussed in sec .
[ current : sec ] .
finally , the conclusions are given in sec .
[ conclusion : sec ] .
in this section , we present the quantum mechanical treatment of the magnetic anti - dot . the magnetic field with a constant @xmath2 outside a circle of radius @xmath3
is assumed to be perpendicular to the plane of the 2deg .
the hamiltonian of the electron of mass @xmath4 and charge @xmath5 is given by @xmath6 \frac{{\left({\bf p } - e{\bf a}\right)}^2}{2 m } , & { \rm if } \,\,\,\
, r > r , \end{array } \right .
\label{h_op - start : eq}\ ] ] where @xmath7 is the canonically conjugate momentum , and the vector potential in polar coordinates @xmath8 and symmetric gauge is given by @xcite @xmath9 a_\varphi(r,\varphi ) & = & b\ , \frac{r^2 -r^2}{2r}\,\theta(r - r ) , \nonumber\end{aligned}\ ] ] and @xmath10 is the unit vector in the @xmath11 direction .
here @xmath12 is the heaviside step function .
the energy levels of the system are the eigenvalues @xmath13 of the schrdinger equation : @xmath14 rotational symmetry of the system implies a separation ansatz for the wave function as a product of radial and angular parts .
we choose for the angular part the appropriate angular momentum eigenfunctions @xmath15 with quantum number @xmath16 ( here @xmath16 is an integer ) .
thus the wave function for a given @xmath16 is separated as @xmath17 , where the radial wave functions @xmath18 satisfy a one - dimensional schrdinger equation in the normal region : @xmath19 in which the radial hamiltonian takes the form @xmath20 here we introduce the dimensionless variable @xmath21 , where @xmath22 is the magnetic length , @xmath23 is the cyclotron frequency , @xmath24 is the dimensionless energy , and the radial potential is given by @xmath25 \frac{{\left(\frac{\tau^2}{2 } -m_{\rm eff}\right)}^2}{\tau^2 } , & { \rm if } \,\,\,\ ,
r > r , \end{array } \right .
\label{vm - pot : eq } \\
m_{\rm eff } & = & s + m\ , \rm{sgn}(eb ) , \label{m_eff : def}\end{aligned}\ ] ] where @xmath26 is the magnetic flux @xmath27 ( in units of the magnetic flux quantum @xmath28 ) missing inside the circle of radius @xmath3 .
the function @xmath29 stands for the sign function . in the numerical results presented in this paper
, we always assume that the particle is an electron moving in a magnetic field along the positive @xmath30-axis , i.e. , @xmath31 . however , our theoretical results are not restricted in such a way . introducing the new variable @xmath32 and transforming the wave functions in the magnetic region ( @xmath33 ) as @xmath34 eq .
( [ rad_sch : eq ] ) results in a kummer differential equation @xcite @xmath35 thus the ansatz for the radial wave function in the magnetic region can finally be written as @xmath36 where @xmath37 is the confluent hypergeometric function @xcite .
note that the function @xmath37 tends to zero as @xmath38 .
it is easy to show that the radial wave function inside the circle of radius @xmath3 ( where the magnetic field is zero ) satisfies the bessel differential equation@xcite .
thus the radial wave function is given by @xmath39 where @xmath40 is the bessel function of order @xmath16 .
matching the radial wave functions inside and outside the circle gives a secular equation whose solutions are the eigenvalues of the system .
the matching conditions at @xmath41 yield @xmath42{.2pt}{4ex } \;\raisebox{-1.5ex}{$\scriptstyle \tau = r / l$ } = \frac{d}{d \tau } \ , \ln f_m ( \tau)\ , \rule[-1.6ex]{.2pt}{4ex } \;\raisebox{-1.5ex}{$\scriptstyle \tau = r / l$}. \label{secular : eq}\ ] ] for a given @xmath16 this secular equation depends only on the dimensionless missing flux @xmath43 .
we now turn to the semiclassical treatment of the system .
generally , in @xmath44 dimensions , a system is integrable if there are @xmath44 independent constants of the motion .
usually this is the result of the separability of the hamiltonian : in a suitably chosen coordinate system the hamiltonian depends only on separate functions @xmath45 of the coordinates and the conjugate momenta .
this means that the dynamics can be viewed as a collection of independent one dimensional dynamical systems .
the function @xmath46 plays the role of the hamiltonian in each subsystem .
the one dimensional semiclassical quantization procedure can be carried out in each subsystem separately @xmath47 where @xmath48 is the action variable and @xmath49 is the maslov index ( for details see , eg , ref . ) .
the maslov index is the sum of the maslov indices of the turning points of the classical motion .
smooth or `` soft '' classical turning points ( zeros of @xmath50 ) contribute @xmath51 to the maslov index , while `` hard '' classical turning points ( infinite potential walls ) contribute @xmath52 .
equation ( [ action ] ) , the bohr - sommerfeld quantization condition , is widely used to approximate the energy levels of classically integrable systems . alternatively , from eq .
( [ action ] ) a semiclassical trace formula known as the berry - tabor formula @xcite can be derived for the oscillating part of the density of states . for two - dimensional systems ,
this formula can be written as @xmath53 ( for the detailed derivation of this expression see appendix [ berry_tabor : app ] ) . here
@xmath54 is the average density of states .
the @xmath55-summation runs over the primitive periodic orbits of the system , the @xmath56-summation runs over their repetitions ; @xmath57 and @xmath58 denote the classical action , the time and the maslov - index of orbit @xmath55 , respectively ; @xmath59 is the number of cycles in the motion projected to the action variable @xmath60 under one cycle of the orbit ; and @xmath61 denotes the action variable @xmath62 as a function of the energy and @xmath60 .
it is easy to show that the classical hamiltonian in polar coordinates @xmath8 , inside and outside the non - magnetic region is : @xmath63 where @xmath64 and @xmath65 are the canonically conjugate momenta , and the radial potential @xmath66 is the same as in ( [ vm - pot : eq ] ) with the following replacements : @xmath67 [ replace_m : eq ] note that here @xmath16 and @xmath68 are continuous classical variables . as we shall see below in the semiclassical approximation , the canonical momentum @xmath65 is quantized according to the bohr - sommerfeld quantization rules ( [ action ] ) .
since the hamiltonian does not depend explicitly on @xmath11 ( the system is rotationally invariant ) , the conjugate momentum @xmath65 is a constant of motion .
thus the angular action variable becomes @xmath69 the conjugate momentum inside the anti - dot is in fact the angular momentum .
outside the anti - dot there is an additional term due to the non - zero vector potential . from @xmath70 one
finds @xmath71 mr^2 \dot{\varphi } + eb\ , \frac{r^2-r^2}{2 } , & { \rm if } \,\,\,\ ,
r > r. \end{array } \right .
\label{pfi_phi_dot : eq}\ ] ] we now choose @xmath72 and @xmath73 in eq .
( [ action ] ) . to calculate the radial action variable @xmath74 one needs to perform the integral of @xmath75 between the classical turning points of the radial potential . for
a given @xmath13 these turning points can be obtained from @xmath76 . using the same dimensionless variables as in sec .
[ quantum : seq ] , we have one turning point for the potential inside the non - magnetic circle : @xmath77 and for the potential valid outside the circle there are two turning points : @xmath78[turnpoints - in_out : eq ] where the upper / lower sign of @xmath79 distinguishes the first and second turning points . note that @xmath80 , and the turning points are real if either @xmath81 or @xmath82 for @xmath83 . for a given energy @xmath13 and momentum @xmath65 one can calculate the cyclotron radius @xmath0 and the guiding center @xmath1 : @xmath84 c & = & l \ , \sqrt{\varepsilon + 2\ , m_{\rm eff}}. \label{guiding_center : eq}\end{aligned}\ ] ] [ rho_c : def ] the derivation in the frame of classical mechanics
is outlined in appendix [ guiding : app ] .
following ref .
the relation between these quantities and the corresponding quantum states of the system can be derived from quantum mechanics .
it turns out that the same relations hold for the cyclotron radius and the guiding center provided @xmath65 is quantized as @xmath85 , where @xmath16 now is an integer .
then , @xmath68 is the same as that defined by eq .
( [ m_eff : def ] ) .
the same results were found by sim et al .
using eq .
( [ rho_c : def ] ) we shall discuss in detail the correspondence between classical orbits and quantum states in secs .
[ phase_diag : sec ] and [ current : sec ] .
the turning points given in eq .
( [ turnpoints - out : eq ] ) can be expressed in terms of the cyclotron radius and the guiding center : @xmath86 \varrho -c , & \text{if } \,\,\ , m_{\rm eff }
< 0 , \end{array } \right . \\[1ex ] l\ , \tau_{2}^{\rm out } & = & c+\varrho .
\label{turnpoint_guiding : eq}\end{aligned}\ ] ] we can now classify the classical orbits according to the relation between the values of the turning points given by ( [ turnpoints - in_out : eq ] ) and the corresponding radius of the circular non - magnetic region ( in units of @xmath87 ) @xmath88 there are two different cases listed in table [ class_cond : table ] . for orbits of type
@xmath89 the particle outside the non - magnetic region moves along a cyclotron orbit and then passes through the magnetic field free region as a free particle . in the case of orbits of type
@xmath2 the particle does not penetrate the non - magnetic region . in this case one can further distinguish two additional types of cyclotron orbits depending on the sign of @xmath68 . the condition @xmath90 listed in table [ class_cond : table ] and eq .
( [ 1st_turnig_guiding : eq ] ) imply that @xmath91 for @xmath92 , and @xmath93 for @xmath94 . from a simple geometrical consideration
it follows that in the first case the cyclotron orbits ( denoted by @xmath95 ) lie outside the circle of radius @xmath3 , while in the latter case the orbits ( denoted by @xmath96 ) completely encircle the non - magnetic region .
these conditions can be rewritten as @xmath97 for both types @xmath95 and @xmath96 , @xmath98 is valid . in the case of orbits of type @xmath89
we have @xmath99 .
we now turn to the calculation of the radial action variable @xmath100 . using the radial potential given by ( [ vm - pot : eq ] ) inside the non - magnetic circle
we find @xmath101 and similarly outside , we have @xmath102 & & - \frac{1}{2}\ , |m_{\rm eff}|\arcsin\left ( \frac{\tau^2(\varepsilon + m_{\rm eff } ) -2 m_{\rm eff}^2}{\tau^2 \sqrt{\varepsilon \left(\varepsilon + 2m_{\rm eff}\right ) } } \right ) .
\label{theta_outside : eq}\end{aligned}\ ] ] the radial action variables for the orbits of types @xmath89 and @xmath2 can be expressed in terms of the functions @xmath103 and @xmath104 , and are listed in table [ class_cond : table ] .
.[class_cond : table ] classification of the orbits and the corresponding radial action variables .
see also the text .
[ cols="<,^ , > " , ] a similar classification has been made for electronic states of a circular ring in magnetic field @xcite and for ring - shaped andreev billiards @xcite .
it may also be useful to present the conditions listed in table [ table : orbits_a - b - conds ] graphically . in @xmath105-@xmath16 space ,
the classically allowed regions corresponding to the different orbits look like a ` phase diagram ' . in fig .
[ fazis_diag : fig ] such a phase diagram is plotted in the space of @xmath105 and @xmath16 in @xmath105-@xmath16 space , for a given magnetic field . and @xmath16 : the classically allowed regions defined by the conditions in table [ table : orbits_a - b - conds ] for the different orbits
. here @xmath106 and @xmath107 .
[ fazis_diag : fig ] ] this phase diagram should be compared with fig .
[ exact_levels - bs : fig ] , the plot of the energy levels from the quantum mechanical calculation . in this way
, the different exact levels can be classified in terms of the corresponding classical orbits .
we should stress again that these orbits are not necessarily periodic .
examples will be shown in sec .
[ current : sec ] .
an apparent correspondence between the classical orbits and the quantum states can also be made by calculating the current flow patterns in the magnetic anti - dot system .
the particle ( probability ) current density @xcite in magnetic field is given by @xmath108 using the vector potential ( [ vector - pt : eq ] ) the current density ( in our units ) for states @xmath109 can be written as @xmath110 , where @xmath111 , \label{current : eq}\ ] ] and @xmath10 is the unit vector in the @xmath11 direction .
solving the secular equation ( [ secular : eq ] ) and then determining the normalized eigenstates , the related current densities can be calculated from ( [ current : eq ] ) .
figures [ nyilas_0_-1:fig]-[nyilas_0_1:fig ] , [ nyilas_a3:fig ] and [ nyilas_a4:fig ] show the current flow patterns for given eigenstates and the corresponding classical trajectories of the particle .
the missing flux quanta is @xmath106 and @xmath107 in all figures . in these figures
the current density @xmath112 at point @xmath113 is represented by an arrow with length proportional to the magnitude of the current density and the midpoint of the arrow is at the point @xmath113 . at a given energy and quantum number @xmath16 corresponding to the classical canonical momentum @xmath114 the cyclotron radius and the guiding center of the classical orbit
is calculated from eq .
( [ rho_c : def ] ) .
hence the classical trajectory of the particle can be determined and are shown in the figures ( scaling is in units of the magnetic length @xmath87 ) . in figs .
[ aram_n0:fig ] and [ aram_a3_a4:fig ] the radial dependence of the current density is plotted for the corresponding eigenstates .
states @xmath115 and @xmath116 were called magnetic edge states in ref . . in fig .
[ nyilas_0_-1:fig ] the current inside the non - magnetic region flows clockwise ( the magnitude of the current density is negative in accordance with fig . [ aram_n0:fig ] ) , while outside the magnetic dot it flows counterclockwise .
the classical trajectories inside the dot form a ` caustic ' and the current is enhanced here in the clockwise direction . from table [
table : orbits_a - b - conds ] we find that the orbit is of type @xmath117 .
the trajectory apparently also satisfies the conditions given in table [ table : orbits ] ( without the requirement for the periodicity of the orbits ) . in fig .
[ nyilas_0_0:fig ] for state @xmath118 and @xmath119 the current is zero inside the magnetic dot .
this can easily be seen from eq .
( [ current : eq ] ) , while the direction of the current flow is counterclockwise outside .
the orbit is a limiting case between types @xmath117 and @xmath120 .
figure [ nyilas_0_1:fig ] for state @xmath121 and @xmath119 shows a counterclockwise current flow both inside and outside the magnetic dot .
the current @xmath122 is positive everywhere as it can be also seen in fig .
[ aram_n0:fig ] .
the orbit is again of type @xmath120 , in accordance with the conditions given in tables [ table : orbits ] and [ table : orbits_a - b - conds ] . ) for state @xmath123 and @xmath119 .
the classical orbit is of type @xmath117 . here and in figs .
[ aram_n0:fig]-[aram_a3_a4:fig ] the missing flux quanta is @xmath106 , and we choose @xmath107 .
[ nyilas_0_-1:fig ] ] ) for state @xmath118 and @xmath119 .
the classical orbit is of type @xmath120 .
[ nyilas_0_0:fig ] ] ) for state @xmath121 and @xmath119 .
the classical orbit is of type @xmath120 .
[ nyilas_0_1:fig ] ] as functions of @xmath124 ( in units of @xmath87 ) for states shown in figs .
[ nyilas_0_-1:fig]-[nyilas_0_1:fig ] .
the vertical line is at @xmath125 .
[ aram_n0:fig ] ] one can observe that the current density @xmath122 is not differentiable at @xmath126 .
this is because the theta function in eq .
( [ current : eq ] ) .
physically , this is a consequence of the step function behaviour of the magnetic field .
nevertheless , the divergence of the current density vector @xmath127 is still zero everywhere .
finally , orbits of types @xmath128 and @xmath129 are shown in figs .
[ nyilas_a3:fig ] and [ nyilas_a4:fig ] for states @xmath130 and @xmath131 , respectively
. similarly , for these states the corresponding current densities as functions of the distance from the origin are plotted in fig .
[ aram_a3_a4:fig ] . ) for state @xmath121 and @xmath132 .
the classical orbit is of type @xmath128 .
[ nyilas_a3:fig ] ] ) for state @xmath133 and @xmath134 .
the classical orbit is of type @xmath129 .
[ nyilas_a4:fig ] ] as functions of @xmath124 ( in units of @xmath87 ) for states shown in figs .
[ nyilas_a3:fig]-[nyilas_a4:fig ] .
the parameters are the same as in fig .
[ nyilas_0_-1:fig ] .
the vertical line is at @xmath125 .
[ aram_a3_a4:fig ] ] for both states the current is very small inside the magnetic dot . in the case of state @xmath130
, the trajectories almost cross the origin ( @xmath16 is small ) , while for state @xmath135 only a small portion of the trajectory penetrates into the magnetic anti - dot regions . the qualitative agreement between the current flow patterns and the classical trajectories
is , again , clearly visible .
note that in all of these figures the classical trajectories are not periodic orbits with energy corresponding to the given eigenstate .
this is not surprising , since the quantization should not be applied to periodic orbits in real space but to the motion on the two - dimensional torus parametrized by the action variables and their canonically conjugate angle variables ( for details see , eg , ref . ) .
in fact , the berry - tabor formula ( [ berrytabor2d ] ) suggests that an infinite number of periodic orbits with proper weights can only result in the correct quantum mechanical density of states .
it has been shown by halperin @xcite that the probability current can be related to the derivative of the energy levels with respect to the angular momentum quantum number : @xmath136 in fig .
[ total_aram : fig ] the integral of the current densities given in eq .
( [ halperin_current : def ] ) is plotted as functions of @xmath16 for @xmath137 .
one can see that for @xmath138 the total
current is zero ( these are the orbits of type @xmath95 in our classification ) and that for @xmath139 it tends to a constant value ( type @xmath96 ) .
the current is negative for states corresponding to orbits of type @xmath117 , while it increases monotonically for states corresponding to orbits of types @xmath140 .
( in units of @xmath141 ) obtained from the integral of @xmath122 given in ( [ halperin_current : def ] ) as functions of @xmath16 for @xmath137 .
the parameters are the same as in fig .
[ nyilas_0_-1:fig ] .
[ total_aram : fig ] ] this is in accordance with the right hand side of eq .
( [ halperin_current : def ] ) using the energy dispersion ( the @xmath16 dependence of the energy levels ) plotted in fig .
[ exact_levels - bs : fig ] .
in this paper we investigated the energy spectrum of the circular magnetic anti - dot systems obtained from exact quantum and semiclassical calculations . in the proper
dimensionless variables the only relevant parameter of the system is the missing flux quantum .
the system is separable , and in the quantum calculation the energy levels are the solutions of the secular equation derived from the matching conditions of the wave functions inside and outside the dot . in the semiclassical treatment we presented two different methods . on the one hand
, the density of states was calculated using the berry - tabor formula . on the other hand , the energy levels obtained from the bohr - sommerfeld quantization rules .
the main difference between the two methods is that in the first case one needs to characterize the possible periodic orbits in real space , while in the latter , the motion of the particle is on a torus in the space of the action variables . in our numerical results we compared the quantum energy spectrum and that obtained in the semiclassical approach .
we showed that the energy levels of the magnetic anti - dot systems obtained from the two semiclassical methods were in good agreement with the numerically exact quantum results for weak magnetic fields .
however , by increasing the magnetic field , a slight deviation between the exact and the semiclassically approximated energy levels can be observed .
we argued that the reason for this discrepancy may be traced back to the fact that the maslov index should be magnetic field dependent .
a thorough investigation of this problem might be an interesting future work .
a classification of the energy spectrum for arbitrary magnetic fields was presented in terms of the classical orbits defined by their cyclotron radius and guiding center .
such identifications are based on the explicit relations between these classical parameters of the orbits and the quantum states .
the correspondence between the quantum states and the classical trajectories can be made transparent by drawing a phase diagram with regions corresponding to six different types of orbits in the space of energy and angular momentum quantum number . finally , we calculated the current flow patterns for eigenstates that correspond to orbits with trajectories penetrating into the field - free region .
the related classical trajectories were also shown for the sake of comparisson . from these results
one can see the close correspondence between the structure of the trajectories and the distribution of the current densities obtained from the quantum calculations . from the energy spectrum of the magnetic anti - dot systems one can determine the free energy .
the good agreement between the semiclassical and quantum treatment of the system allows us to use semiclassical methods in the weak field limit for calculating the energy spectrum .
therefore , the semiclassical approach provides a useful starting point for successive studies of thermodynamic properties , such as magnetization .
moreover , the semiclassical approximation can be an effective tool for investigating arbitrarily shaped magnetic anti - dots ( which would be a very difficult task in the quantum case ) or systems with more complicated magnetic field profiles .
we would like to thank b. kramer , a. nogaret and a. pirth for useful discussions .
this work is supported in part by the european community s human potential programme under contract no .
hprn - ct-2000 - 00144 , nanoscale dynamics , the hungarian - british intergovernmental agreement on cooperation in education , culture , science and technology , and the hungarian science foundation otka to34832 and fo47203 .
the quantized energies can be recovered if we express the hamiltonian in terms of @xmath48 @xmath142 the semiclassical density of states is the density of these energies : @xmath143 the density of states can be rewritten via the poisson resummation technique @xmath144 here , we used the fourier expansion of the delta spike train .
the term @xmath145 can be evaluated directly and yields the non - oscillatory average density of states .
other terms can be evaluated by the saddle point method , when @xmath146 .
the saddlepoint conditions select the periodic orbits of the system , and the result of the integration is @xmath147 here @xmath55 is the index of the primitive periodic orbits , @xmath56 is the number of repetitions , @xmath148 is the classical action along the orbit , @xmath149 is the time period of the orbit , and @xmath58 is the maslov index .
the quantity @xmath150 is the number of action variables of the periodic orbit whose saddle point value is zero @xmath151 , since in this case the gaussian saddle point integral is only one - sided , and its contribution is @xmath152 of the full gaussian integral .
the matrix @xmath153 is related to the second derivative matrix @xmath154 equation ( @xmath155 ) is the generic form of the semiclassical density of states in terms of periodic orbits , known as the berry - tabor formula @xcite . in two dimensions ,
very often the hamiltonian can not be expressed with the action variables explicitly , only the implicit function @xmath156 is available . in this case
it is more useful to express the quantities in the berry - tabor trace formula in terms of the derivatives of @xmath157 . taking the partial derivative of ( [ gi ] ) with respect to @xmath62 yields @xmath158 while the partial derivative of ( [ gi ] ) with respect to @xmath60 gives @xmath159 the frequencies
can be expressed from these equations as @xmath160 periodic orbits are recovered from @xmath161 and @xmath162 .
the action @xmath62 for a periodic orbit at energy @xmath13 can be obtained by solving equation @xmath163 where we introduced @xmath164 and @xmath165 corresponding to the primitive orbit .
then the period can be expressed simply as @xmath166 the main determinant to be calculated reads @xmath167 now , the second derivatives of @xmath168 can be expressed with the second derivatives of @xmath157 by taking further partial derivatives of ( [ gi1 ] ) and ( [ gi2 ] ) with respect to @xmath62 and @xmath60 .
then we can express the second derivatives as @xmath169 using these expressions , the determinant becomes @xmath170 the density of states in two dimensions is then @xmath171
the cyclotron radius can be determined from the energy @xmath13 of the particle .
the energy is conserved , and obviously @xmath172 , thus @xmath173 the guiding center may be calculated as follows . as we have seen , the conjugate momentum given by eq .
( [ pfi_phi_dot : eq ] ) is a constant of motion , therefore , e.g. , for @xmath33 the right hand side of ( [ pfi_phi_dot : eq ] ) should also be a constant at any point of the orbit .
at first apply this equation for points @xmath174 and @xmath175 , which are the points closest to and farthest from the origin ( the center of the circle of radius @xmath3 ) of an orbit lying outside the anti - dot . these are special points of the orbits for which the right hand side of eq .
( [ pfi_phi_dot : eq ] ) has a simpler form .
then the distances of points @xmath174 and @xmath175 from the origin are @xmath176 and @xmath177 ( we assume that point @xmath175 is farther from the origin ) . from a simple geometrical argument one finds that the angular velocity at points @xmath174 and @xmath175 satisfies the following equations @xmath178 substituting , for example , @xmath176 and @xmath179 from ( [ r_p_app : eq ] ) into eq .
( [ pfi_phi_dot : eq ] ) , and using ( [ cycl - rad_app : eq ] ) , we find @xmath180 the same results can be obtained by using ( [ r_q_app : eq ] ) for point @xmath175 .
if the oribit encompasses the anti - dot then the right hand side of ( [ r_p_app : eq ] ) should be multiplied by a factor of @xmath181 .
the case of orbits with trajectories penetrating into the anti - dot can be treated similarly .
however , the expressions for the cyclotron radius and the guiding center are the same as above for all cases .
in the case of the cyclotron orbits , the integral in @xmath182 in eq .
( [ altal ] ) has to be calculated directly rather than using the saddlepoint method .
as @xmath182 is constant , the integrand does not depend on the integration variable and therefore the integral is equal to the measure of the interval of the possible @xmath182s . without loss of generality , we take @xmath183 in this section .
cyclotron orbits which do not encompass the anti - dot ( type @xmath95 ) are possible at any value of @xmath184 and any negative angular momentum @xmath65 ( see table [ table : orbits_a - b - conds ] ) . at point @xmath174 of these orbits ( points @xmath174 and @xmath175 of a cyclotron orbit
are defined in appendix [ guiding : app ] ) , from eqs .
( [ angular_action : eq ] ) and ( [ pfi_phi_dot : eq ] ) we obtain @xmath185 this is minimal when @xmath186 ( and the cyclotron orbit touches the boundary of the anti - dot ) , and is maximal when the orbit is placed as far as possible from the anti - dot . by denoting the radius of the system with @xmath187 ( in units of @xmath87 ) the intergation in ( [ altal ] ) with respect to @xmath182
yields a factor @xmath188 . using ( [ i_phi_b_1;eq ] ) and ( [ r_p_app : eq ] )
, one finds @xmath189 the @xmath100- and @xmath190-integrals can be evaluated with the saddle point method , just as in the case of a one - dimensional system .
the determinant of the second derivative matrix is @xmath191 from eqs .
( [ gi2 ] ) and ( [ i_r_b - type : eq ] ) , we find @xmath192 \det\begin{bf } d \end{bf } & = & - \omega_c^2.\end{aligned}\ ] ] thus , the total amplitude of these orbits in the periodic orbit sum is @xmath193 .
\label{eq : a_c_pl}\end{aligned}\ ] ] the action can be calculated from eq .
( [ i_r_b - type : eq ] ) , and for @xmath194 we have @xmath195 and their maslov index is @xmath196 , therefore the contribution to the semiclassical level density from these orbits reads @xmath197 the sum has dirac delta peaks at @xmath198 , where @xmath199 and @xmath200
. these are the familiar landau levels of an electron for @xmath200 . for cyclotron orbits encompassing the anti - dot ( type @xmath96 ) the angular momentum satisfies the condition
@xmath201 ( see table [ table : orbits_a - b - conds ] ) . at point @xmath175 of these orbits , using ( [ angular_action : eq ] ) and ( [ pfi_phi_dot : eq ] ) , we can write @xmath202 the minimum and the maximum of @xmath203 are @xmath0 and @xmath204 , respectively . between these values , @xmath205 , as a function of @xmath203 ,
is monotonic , thus the integration in ( [ altal ] ) over @xmath182 gives @xmath206 . using ( [ i_phi_b_2;eq ] ) and ( [ r_q_app : eq ] ) , we have @xmath207 similarly to ( [ eq : a_c_pl ] ) , the amplitude of the orbits becomes @xmath208 using ( [ i_r_b - type : eq ] ) , the action for @xmath209 is @xmath210 finally , the contribution to the periodic orbit sum of these orbits is @xmath211.\end{aligned}\ ] ] the sum has dirac delta peaks at @xmath212 , where @xmath16 and @xmath213 are non - negative integers , and @xmath214 .
these are again the familiar landau levels of an electron for @xmath214 .
|
we study the energy level structure of two - dimensional charged particles in inhomogeneous magnetic fields . in particular , for magnetic anti - dots the magnetic field is zero inside the dot and constant outside .
such a device can be fabricated with present - day technology .
we present detailed semiclassical studies of such magnetic anti - dot systems and provide a comparison with exact quantum calculations . in the semiclassical approach we apply the berry - tabor formula for the density of states and the borh - sommerfeld quantization rules . in both cases we found good agreement with the exact spectrum in the weak magnetic field limit .
the energy spectrum for a given missing flux quantum is classified in six possible classes of orbits and summarized in a so - called phase diagram .
we also investigate the current flow patterns of different quantum states and show the clear correspondence with classical trajectories .
|
since the discovery of heat engines people were struggling to optimize their performance @xcite
. one of the first results in the field was due to sadi carnot @xcite and rudolf clausius @xcite .
they showed that the maximum efficiency attainable by any heat engine operating between the hot reservoir at the temperature @xmath0 and the cold one at the temperature @xmath1 is given by the carnot efficiency @xmath2 .
alas their result was of limited practical importance for the corresponding heat engine must work reversibly and thus its output power is zero .
optimization of the power of _ irreversible _ carnot cycles working under finite - time conditions turned out to be much more practical .
it was pioneered by novikov @xcite , chambadal @xcite and later by curzon and ahlborn @xcite .
the obtained result @xmath3 for the efficiency at maximum power ( emp ) seemingly exhibits the same degree of generality as the carnot s formula and also it describes well the efficiency of some actual thermal plants @xcite .
although it turned out that @xmath4 is not a universal result , neither it represents an upper or lower bound for the emp @xcite , its close agreement with emp for several model systems @xcite ignited search for universalities in performance of heat engines . for thermochemical engines it was shown that the form of the emp @xcite is controlled by the symmetries of the underlying dynamics ( see also @xcite for a similar study , but with finite - sized reservoirs ) . this result was further precised @xcite and recently a similar argumentation was successfully used for general thermodynamic devices @xcite .
further universalities were obtained for the class of heat engines working in the regime of `` low dissipation '' @xcite , where the work dissipated during the isothermal branches of the carnot cycle grows in inverse proportion to the duration of these branches .
namely a general expression for the emp for these engines has been published @xcite and subsequently lower and upper bounds on the emp were derived @xcite .
all these results have been confirmed within a minimally nonlinear irreversible thermodynamics framework @xcite .
another universal result recently appeared in the realm of brownian motors @xcite .
engines so small that their output work and input heat become stochastic due to thermal fluctuations .
the ratio of these variables yields the efficiency which is thus itself random variable .
large deviation form of the corresponding probability distribution exhibits the following universal features @xcite : ( i ) the most probable value of efficiency corresponds to its traditional definition using averages of work and heat , ( ii ) in case of time - reversal symmetric driving the least probable value of efficiency equals to @xmath5 . in the present paper
we reveal a new universality in optimization of _ trade - off between power and efficiency _ for low - dissipation carnot cycles @xcite . as our main result
, we show that any trade - off measure expressible in terms of efficiency and the ratio of power to its maximum value exhibits the same degree of universality as the emp ( see sec .
[ sec : tradeoff ] ) .
consequently , for many systems these trade - off measures can be optimized independently of most details of the dynamics and of the coupling to thermal reservoirs .
investigation of the trade - off between power and efficiency is immensely important for engineering practice , where not only powerful , but also economical devices should be developed .
indeed , it was already highlighted @xcite that actual thermal plants and heat engines should not work at the maximum power , where the corresponding efficiency can be relatively small , but rather in a regime with slightly smaller power and considerably larger efficiency .
our results strongly support these findings .
examples of the low - dissipation carnot cycles are brownian motors based on two - level quantum systems @xcite or on driven overdamped systems @xcite , given that the used driving minimizes the work dissipated during the isothermal branches and that all thermodynamic variables are defined in the traditional manner , i.e. , via averages .
we illustrate our results using carnot cycles based on an overdamped particle diffusing in an externally controlled confining potential . due to the recent advances in micromanipulation techniques @xcite , such engines
were already realized experimentally @xcite and hence are of vital theoretical interest @xcite .
the paper is organized as follows . in sec .
[ sec : model ] we introduce the diffusion - based heat engines and standard thermodynamic variables used to characterize their performance @xcite . the scheme @xcite for minimization of the work dissipated during the isothermal branches which leads to the low - dissipation behavior is described in sec .
[ sec : max_eta ] . in sec .
[ sec : maxp ] , we review the results @xcite concerning the emp for low - dissipation carnot cycles . in the both sections we discuss consequences of the optimisation procedure on specific model systems .
our main conclusions are : ( i ) under the restrictions of the optimization procedure , the protocol which minimizes the dissipation yields , at the same time , the maximum power and efficiency .
( ii ) when the exact optimal protocol can not be obtained , a numerical optimization within a certain class of `` test '' protocols yields performance comparable to that of the optimal protocol .
( iii ) the scheme for minimization of the work dissipated during the isothermal branches leads to the low - dissipation behavior only in case the optimization procedure has enough freedom to modify the potential . in sec .
[ sec : tradeoff ] we utilize the findings on the emp reviewed in sec . [ sec : maxp ] to demonstrate the universality in the trade - off between power and efficiency .
as particular examples we optimize two specific trade - off measures .
the first one ( [ eq : tradeoff ] ) is suitable for finding optimal efficiency for a given output power .
the results of the optimization procedure are illustrated on the model of log - harmonic potential introduced in secs .
[ sec : model]-[sec : max_eta ] .
the performed optimization indeed shows that engines operating near maximum power exhibit only slightly smaller power than the maximal one , but , at the same time , work with considerably larger efficiency than the emp . within the second example we derive universal bounds ( [ eq : boundsmv ] ) on the efficiency at maximum trade - off ( [ eq : lambda ] ) given by the product of power and efficiency . by comparison with data from actual nuclear power plants
these bounds give promising results .
induce the periodic response of the pdf to find the particle at time @xmath6 at position @xmath7 ( filled brown curve . ) ] consider the carnot - like model @xcite sketched in fig .
[ fig : cycle ] .
the working medium of the heat engine is an overdamped particle diffusing in an externally driven confining potential .
both the potential , @xmath8 , and the bath temperature , @xmath9 , are periodically altered .
the cycle consists of two isotherms ( branches i and iii ) and two adiabats ( branches ii and iv ) .
state of the engine is described by the probability density function ( pdf ) to find the particle at time @xmath6 at position @xmath7 . under the time - periodic driving the engine will eventually attain a time - periodic steady state , the so called limit cycle .
the corresponding pdf @xmath10 follows from the fokker - planck equation @xmath11 \right\}p(x , t)\ , , \label{eq : current}\end{aligned}\ ] ] supplemented by the condition @xmath12 here @xmath13 stands for the mobility , which is assumed to depend on temperature , @xmath14 is the boltzmann constant , @xmath15 denotes the probability current and @xmath16 is the duration of one operational cycle of the engine . in the following subsections [ subs : iia ] and
[ subs : iib ] we will specify the individual branches of the heat engine . in subsection [ subs : iic ] we will discuss on examples under which conditions the heat engine is capable of producing positive work .
other performance characteristics of the engine like its efficiency and output power will be introduced in section [ sec : max_eta ] and discussed in the rest of the paper .
[ cols= " < , < , < , < " , ] although the low - dissipation assumption may not be fulfilled in actual thermal plants , it is tempting to confront the efficiency intervals ( [ eq : boundsmv])-([eq : boundsmo ] ) with real data .
this is done in fig .
[ fig : power_plants ] , where we compare the individual efficiency intervals with experimental data measured for various actual thermal plants ( tab .
[ tab : data ] , data taken from @xcite ) . within the bounds for the efficiency at maximum @xmath17 fall five experimental points while within those at maximum power lie seven of them .
the points belonging to the @xmath17 sector lie closer to its middle , while the points which fall into the maximum power sector lie closer to its borders .
hence this ( very limited ) set of data suggests that at least some real thermal plants operate close to optimum @xmath17 conditions rather than close to maximum power .
notably , the data which fits well into the interval corresponding to the optimal @xmath17 were all obtained from nuclear power plants , see tab .
[ tab : data ] .
we have utilized the known results @xcite on the emp for low - dissipation carnot cycles to derive new universality in optimization of trade - off between power and efficiency .
our main results are eqs .
( [ eq : reldp ] ) and ( [ eq : reldeta ] ) which demonstrate that any trade - off measure expressible in terms of efficiency @xmath18 and the ratio @xmath19 of power to its maximum value exhibits the same degree of universality as the emp . consequently , for many systems
the trade - off measure can be optimized independently of most details of the dynamics and of the coupling to thermal reservoirs . in order to demonstrate this result
we have introduced two specific trade - off measures .
the first one ( [ eq : tradeoff ] ) is designed for finding optimal efficiency for a given output power .
the performed optimization clearly reveals diseconomy of engines working at maximum power .
engines operating in regimes near the maximum power with slightly smaller power and considerable larger efficiency are more economical ( see figs . [
fig : universal2]-[fig : universal4 ] ) . within the second example
we have derived universal lower and upper bounds ( [ eq : boundsmv ] ) on the efficiency at maximum trade - off given by the product of power and efficiency ( [ eq : lambda ] ) .
comparison of the obtained bounds ( [ eq : boundsmv ] ) with data from actual nuclear power plants gives promising results , see fig [ fig : power_plants ] .
the data points outside the interval ( [ eq : boundsmv ] ) could result from thermal plants not operating at maximum value of the product of power and efficiency ( [ eq : lambda ] ) or not operating in the low - dissipation regime at all . all the results have been illustrated on a diffusion - based heat engine specified in secs [ sec : model]-[sec : max_eta ] .
these engines operate in the low - dissipation regime given that the used driving minimizes the work dissipated during the isothermal branches .
the peculiarities of the corresponding optimization procedure have been reviewed and thoroughly discussed in secs.[sec : max_eta ] and [ sec : maxp ] .
let us note that this analysis has been performed for overdamped models only and it could be interesting to consider also the underdamped case . in this respect
only few studies are available in the literature so far @xcite .
the low - dissipation assumption allows performing quite general analyzes . while this assumption may be reasonable for many systems @xcite , it is definitely not general
. therefore it is important to investigate also cases where the work dissipated during the isothermal branches behaves differently than in the low - dissipation regime .
the present study has been performed for carnot cycles with infinitely fast adiabatic branches .
similar research of carnot cycles with adiabatic branches performed in finite - time @xcite and also of other thermodynamic cycles like the joule cycle or the stirling cycle could provide further insight .
68ifxundefined [ 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 https://books.google.de/books?id=u13kiglz2zcc[__ ] ( , ) https://books.google.cz/books?id=j3jzxjxyzfic[__ ] , acadmie internationale dhistoire des sciences .
collection des travaux ( , ) link:\doibase 10.1080/14786445608642141 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/0891-3919(58)90244-4 [ * * , ( ) ] @noop _ _ , vol .
( , ) link:\doibase 10.1119/1.10023 [ * * , ( ) ] http://books.google.cz/books?id=uoizb_2y5pic[__ ] , student edition ( , ) https://books.google.cz/books?id=5h1ur1r_ebcc[__ ] , wiley - interscience publication ( , ) http://www.degruyter.com/view/j/jnet.1997.22.issue-4/jnet.1997.22.4.311/jnet.1997.22.4.311.xml [ * * , ( ) ] http://eu.wiley.com/wileycda/wileytitle/productcd-0471967521.html[__ ] ( , ) \doibase http://dx.doi.org/10.1016/s0360-5442(00)00059-1 [ * * , ( ) ] \doibase http://dx.doi.org/10.1119/1.14240 [ * * , ( ) ] \doibase http://dx.doi.org/10.1063/1.362674 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.98.130602 [ * * , ( ) ] http://stacks.iop.org/0295-5075/81/i=2/a=20003 [ * * , ( ) ] http://stacks.iop.org/0295-5075/83/i=6/a=60003 [ * * , ( ) ] link:\doibase 10.1103/physreve.80.021121 [ * * , ( ) ] link:\doibase 10.1103/physreve.77.041118 [ * * , ( ) ] http://stacks.iop.org/1751-8121/41/i=31/a=312003 [ * * , ( ) ] http://stacks.iop.org/0295-5075/85/i=6/a=60010 [ * * , ( ) ] link:\doibase 10.1103/physrevb.80.235122 [ * * , ( ) ] link:\doibase 10.1103/physreve.81.041106 [ * * , ( ) ] link:\doibase 10.1103/physreve.82.011120 [ * * , ( ) ] http://stacks.iop.org/1674-1056/21/i=2/a=020513 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.102.130602 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.112.180603 [ * * , ( ) ] link:\doibase 10.1103/physreve.91.022136 [ * * , ( ) ] link:\doibase 10.1140/epjst / e2015 - 02433 - 8 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.105.150603 [ * * , ( ) ] link:\doibase 10.1143/jpsj.66.3326 [ * * , ( ) ] , \doibase http://dx.doi.org/10.1063/1.4885277 [ * * , ( ) ] link:\doibase 10.1103/physreve.87.012105 [ * * , ( ) ] http://stacks.iop.org/0295-5075/112/i=2/a=20002 [ * * , ( ) ] http://stacks.iop.org/0295-5075/97/i=1/a=10004 [ * * , ( ) ] http://stacks.iop.org/0295-5075/101/i=1/a=10005 [ * * , ( ) ] link:\doibase 10.1063/1.1535005 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.81.387 [ * * , ( ) ] link:\doibase 10.1002/andp.200410121 [ * * , ( ) ] http://stacks.iop.org/0034-4885/75/i=12/a=126001 [ * * , ( ) ] link:\doibase 10.1038/ncomms5721 [ * * , ( ) ] link:\doibase 10.1103/physreve.90.052145 [ * * , ( ) ] link:\doibase 10.1103/physreve.90.042146 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/j.physa.2015.10.095 [ ( ) ] http://stacks.iop.org/0295-5075/109/i=2/a=20004 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.114.050601 [ * * , ( ) ] \doibase doi:10.1038/nphys3518 [ ( ) , doi:10.1038/nphys3518 ] http://stacks.iop.org/1367-2630/16/i=10/a=102003 [ * * , ( ) ] link:\doibase 10.1103/physrevb.91.115417 [ * * , ( ) ] link:\doibase 10.1103/physreve.91.042127 [ * * , ( ) ] link:\doibase 10.1103/physreve.89.062119 [ * * , ( ) ] http://stacks.iop.org/1751-8121/46/i=40/a=402001 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/s0196-8904(00)00055-8 [ * * , ( ) ] https://books.google.cz/books?id=sae5aqaamaaj[__ ] , oxford science publications ( , ) http://stacks.iop.org/0022-3727/27/i=6/a=011 [ * * , ( ) ] link:\doibase 10.1103/physreve.92.032113 [ * * , ( ) ] link:\doibase 10.1103/physreve.92.032117 [ * * , ( ) ] `` , '' in link:\doibase 10.1002/9780470238080.ch2 [ _ _ ] ( , ) pp .
link:\doibase 10.1038/nphys2163 [ * * , ( ) ] http://stacks.iop.org/1742-5468/2014/i=5/a=p05022 [ * * , ( ) ] link:\doibase 10.1103/physreve.77.051132 [ * * , ( ) ] link:\doibase 10.1103/physreve.89.052148 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.98.108301 [ * * , ( ) ] link:\doibase 10.1007/978 - 3 - 319 - 07091 - 9 [ _ _ ] , springer theses ( , ) http://arxiv.org/abs/cond-mat/9912121 [ ( ) ] , http://www.scielo.br/scielo.php?script=sci_arttext&pid=s0103-97332006000700029&nrm=iso [ * * , ( ) ] link:\doibase 10.1103/physreve.88.062115 [ * * , ( ) ] link:\doibase 10.1103/physreve.62.7759 [ * * , ( ) ] link:\doibase 10.1103/physreve.77.041105 [ * * , ( ) ]
|
a new universality in optimization of trade - off between power and efficiency for low - dissipation carnot cycles is presented .
it is shown that any trade - off measure expressible in terms of efficiency and the ratio of power to its maximum value can be optimized independently of most details of the dynamics and of the coupling to thermal reservoirs .
the result is demonstrated on two specific trade - off measures .
the first one is designed for finding optimal efficiency for a given output power and clearly reveals diseconomy of engines working at maximum power . as the second example we derive universal lower and upper bounds on the efficiency at maximum trade - off given by the product of power and efficiency .
the results are illustrated on a model of a diffusion - based heat engine .
such engines operate in the low - dissipation regime given that the used driving minimizes the work dissipated during the isothermal branches .
the peculiarities of the corresponding optimization procedure are reviewed and thoroughly discussed .
|
let @xmath1 , @xmath2 , be independent and identically distributed random vectors , and assume that the cumulative distribution function , @xmath0 , of @xmath3 is continuous .
by sklar s theorem ( sklar @xcite ) , there exists a unique copula , @xmath4 , such that @xmath5 where @xmath6 is the @xmath7th marginal distribution function of @xmath0 . for
@xmath2 and @xmath8 , let @xmath9 be the rank of @xmath10 among @xmath11 ; namely , @xmath12 the vector of ranks is denoted by @xmath13 .
the basic nonparametric estimator for the copula @xmath4 is the empirical copula ( deheuvels @xcite ) , and we use the version given by @xmath14^d.\ ] ] slightly different definitions are employed for instance in fermanian et al .
@xcite and tsukahara @xcite , but the supremum distance between these empirical copula variants is at most @xmath15 . in section [ sec : simul ] , we will also use the so - called empirical checkerboard copula ; see .
let @xmath16 be independent random variables which are uniformly distributed on the unit interval and are independent of the sample @xmath17 . for each @xmath8 ,
let @xmath18 be the order statistics based on @xmath19 .
define @xmath20 one of the authors conceived that @xmath21 could be interpreted as a sample from some version of the empirical copula .
although this was not quite correct , we found that picking one vector randomly from these @xmath22 vectors is equivalent to sampling from a smoothed version of the empirical copula , which we call the _ empirical beta copula _ in view of the marginal distribution of uniform order statistics .
this idea may be regarded as baker @xcite s copula construction based on uniform order statistics , where @xmath23-tuples of order statistics are determined by the rank vectors @xmath24 , and the probability of choosing among them is equal to @xmath25 .
the empirical beta copula arises as a particular case of the empirical bernstein copula when the degrees of all bernstein polynomials are set equal to the sample size .
the bernstein copula and the empirical bernstein copula are introduced in sancetta and satchell @xcite , and the asymptotic behavior of the latter is studied in janssen et al .
we give necessary and sufficient conditions for the empirical bernstein copula to be a genuine copula , conditions which hold for the empirical beta copula .
we show asymptotic results for the empirical bernstein copula process under weaker assumptions than those in janssen et al .
@xcite .
the advantages of the empirical beta copula are that it is a genuine copula , that it does not require the choice of a smoothing parameter , and that simulating random samples from it is straightforward .
furthermore , the corresponding empirical process converges weakly to a gaussian process under standard smoothness conditions on the underlying copula . for small samples ,
the empirical beta copula outperforms the empirical copula both in terms of bias and variance .
compared with the empirical bernstein copula with polynomial degrees as suggested in janssen et al .
@xcite , the empirical beta copula is still more accurate for several copula models , especially in terms of the bias .
the paper is organized as follows . in section [ sec : cop ] , we define the empirical beta copula and prove its relation to the empirical bernstein copula together with some auxiliary results for the bernstein transformation . in section [ sec : asym ] , we formulate and prove the asymptotic results for the empirical bernstein copula process . a monte carlo simulation study is presented in section [ sec : simul ] , and we conclude the paper with some remarks in section [ sec : concl ] .
some additional proofs are given in the appendix .
we define the empirical beta copula and we show that the sampling scheme introduced in the preceding section amounts to the first stage of sampling from it .
the empirical beta copula turns out to be a particular case of the empirical bernstein copula .
necessary and sufficient conditions are given for the empirical bernstein copula to be a genuine copula , conditions which imply that the empirical beta copula is a genuine copula .
finally , we provide a non - asymptotic bound for the difference between the empirical copula and the empirical beta copula .
we continue to use the notation given in the introduction .
to express mathematically the idea stated above , we replace each indicator function in the definition of the empirical copula by the cumulative distribution function of @xmath26 in conditionally on @xmath27 .
since @xmath28 is the @xmath29th order statistic of an independent random sample of size @xmath22 from the uniform distribution on @xmath30 $ ] , its distribution is @xmath31 .
thus we define the _ empirical beta copula _ by @xmath32^d,\ ] ] where , for @xmath33 $ ] and @xmath34 , @xmath35 here and henceforth , @xmath36 generically denote the order statistics based on @xmath22 independent random variables @xmath37 , uniformly distributed on @xmath30 $ ] . picking one element randomly from among the points @xmath38 , @xmath2 , in amounts to sampling from the empirical beta copula @xmath39 conditionally on @xmath40 . in the absence of ties ,
all @xmath23 margins of @xmath39 are equal to the uniform distribution on @xmath30 $ ] , so that @xmath39 is a genuine copula .
indeed , for @xmath41 and @xmath42 $ ] , @xmath43 = { \operatorname{e}}\biggl[\frac{1}{n}\sum_{r=1}^n { \mathbf{1}}\{u_r\leq u_j\}\biggr]=u_j.\end{aligned}\ ] ] before we show that the empirical beta copula is a particular case of the empirical bernstein copula , we need to state and prove some auxiliary results on bernstein polynomials in general . put @xmath44,\ m\in{\mathbb{n}},\ s\in\{0,1,\ldots , m\}.\ ] ] for @xmath45 and a real array @xmath46 , consider the following bernstein polynomial : @xmath47^d.\ ] ] our objective in this subsection is to derive a ( necessary and ) sufficient condition on the coefficients @xmath48 for @xmath49 to be a copula . to this end , we need the following two lemmas . [ prop : prelim ] let @xmath50 and @xmath51 . 1
. we have @xmath52 for all @xmath53 $ ] if and only if @xmath54 for all @xmath55 .
we have @xmath56 for all @xmath53 $ ] if and only if @xmath57 for all @xmath55 .
proof the ` if ' parts follow from direct computation .
the ` only if ' parts are an immediate consequence of the fact that the bernstein polynomials @xmath58 form a basis of the linear space of all real polynomials with degree not larger than @xmath22 .
[ prop : bernstein-0 ] @xmath59 for all @xmath60^d$ ] if and only if @xmath61 .
proof the ` if ' part is trivial .
the ` only if ' part can be proven by induction on @xmath23 , using lemma [ prop : prelim](i ) both for the case @xmath62 as for the induction step .
we leave the details to the reader . for @xmath41 ,
define the difference operator @xmath63 mapping a given array @xmath48 to the new array @xmath64 given by @xmath65 difference operators @xmath63 with distinct indices @xmath7 can be composed in the obvious way . in particular , @xmath66 where @xmath67 for each @xmath41 . furthermore , consider the following three conditions on a real array @xmath48 : 1 .
@xmath68 as soon as @xmath69 for some @xmath41 ; 2 .
@xmath70 for each @xmath41 and each @xmath71 ; 3 .
@xmath72 for all @xmath73 .
[ prop : bernstein - polynomial - copula ] if the conditions ( c.1 ) , ( c.2 ) and ( c.3 ) hold , then the bernstein polynomial @xmath49 is a copula . moreover , ( c.1 ) and ( c.2 ) are necessary for @xmath49 to be a copula . before commencing the proof ,
recall that a function @xmath74^d{\to}[0,1]$ ] is a copula if and only if the following conditions hold : a. @xmath4 is _ grounded _ ; that is , @xmath75 whenever at least one of the @xmath76 equals @xmath77 .
b. @xmath78 for all @xmath8 .
c. @xmath4 is _ @xmath23-increasing _ ; that is , for all @xmath79^d$ ] with @xmath80 for all @xmath8 , we have @xmath81 where the sum is taken over all @xmath82 such that @xmath83 for all @xmath8 .
see nelsen @xcite or more explicitly , theorem 1.1 of mai and scherer @xcite .
proof of proposition [ prop : bernstein - polynomial - copula ] since @xmath84 , we have @xmath85 it then follows from lemma [ prop : bernstein-0 ] that ( c.1 ) is equivalent to the groundedness of @xmath49 . since @xmath86
, we have @xmath87 by lemma [ prop : prelim](ii ) , condition ( c.2 ) is equivalent to the marginal condition ( ii ) above for @xmath49 . using the identity and by induction
, one can easily find that @xmath88 if ( c.3 ) is satisfied , then it is obvious from the above expression that we have @xmath89 for all @xmath90 .
since @xmath49 is infinitely differentiable , it thus follows from proposition [ prop : d - increasing - partial - derivative ] that @xmath49 is @xmath23-increasing .
therefore ( c.1 ) , ( c.2 ) and ( c.3 ) imply that @xmath49 is a copula .
the necessity of ( c.1 ) and ( c.2 ) for @xmath49 to be a copula has already been proved .
the condition ( c.3 ) is not necessary for @xmath49 to be a copula .
take @xmath91 and @xmath92 , and let @xmath93 for @xmath94 , @xmath95 , and @xmath96 , @xmath97 , @xmath98 , @xmath99 , @xmath100 , @xmath101 .
then @xmath102 satisfies ( c.1 ) and ( c.2 ) , but not ( c.3 ) since @xmath103 . by direct inspection , it is easy to see that @xmath104 for all @xmath60 ^ 2 $ ] because @xmath105 is linear in @xmath106 . for a function @xmath107^d{\to}{\mathbb{r}}$ ] , the bernstein polynomial of order @xmath108 of @xmath109 is defined by @xmath110 when @xmath4 is a copula , @xmath111 is called the _ bernstein copula _ of @xmath4 . for the empirical copula @xmath112 , we call @xmath113 the _ empirical bernstein copula_. for a copula @xmath4 , the real array @xmath48 defined by @xmath114 satisfies conditions ( c.1)(c.3 ) . by proposition [ prop : bernstein - polynomial - copula ] ,
the bernstein copula @xmath111 of @xmath4 is therefore a copula itself . on the other hand , the empirical bernstein copula @xmath113 is not necessarily a copula . [
prop : bernstein - genuine - copula ] let @xmath112 be the empirical copula of a sample @xmath17 of @xmath23-variate vectors without ties in any of the @xmath23 components .
let @xmath115 .
the empirical bernstein copula @xmath113 is a copula if and only if all the polynomial degrees @xmath116 are divisors of @xmath22 .
proof the empirical copula @xmath112 in is a cumulative distribution function on @xmath30^d$ ] without mass on the lower boundary of the unit hypercube .
the real array @xmath48 defined by @xmath117 therefore satisfies conditions ( c.1 ) and ( c.3 ) above . by proposition
[ prop : bernstein - polynomial - copula ] , @xmath113 is a copula if and only if that array @xmath48 also satisfies ( c.2 ) .
let @xmath118 denote the integer part of the real number @xmath119 .
for @xmath120 and @xmath67 , we have @xmath121 condition ( c.2 ) is fulfilled if and only if the right - hand side is equal to @xmath122 for all @xmath67 . setting @xmath123 shows that this requires @xmath124 to be a divisor of @xmath22 , and the latter condition
is easily seen to be sufficient as well .
now we show that the empirical beta copula @xmath39 in is a particular case of the empirical bernstein copula .
[ lem : betabernstein ] let @xmath112 be the empirical copula of a sample @xmath17 of @xmath23-variate vectors without ties in any of the @xmath23 components
. then @xmath125 .
proof the beta cumulative distribution function @xmath126 in can be written in terms of bernstein polynomials @xmath127 as @xmath128.\ ] ] as a consequence , for @xmath129^d$ ] , @xmath130\prod_{j=1}^d p_{n , s_j}(u_j).\end{aligned}\ ] ] invoking the definition of the empirical copula @xmath112 in finally yields @xmath131 as required .
proposition [ prop : bernstein - genuine - copula ] and lemma [ lem : betabernstein ] confirm that the empirical beta copula is itself a copula .
we had already reached this conclusion in subsection [ subsec : beta ] by a direct argument .
sancetta and satchell @xcite first introduced the bernstein copula .
in that paper , it is falsely claimed ( p. 537 ) that a function @xmath4 on @xmath132^d$ ] is a copula if and only if @xmath4 is nondecreasing in all its arguments and satisfies the frchet bounds @xmath133 [ for a counterexample , see nelsen ( * ? ? ?
* exercise 2.11 ) . ]
our results above correct both the statement and the proof of their theorem 1 .
we provide a deterministic , non - asymptotic bound for the difference between the empirical copula @xmath112 and the empirical beta copula @xmath39 .
[ prop : nonasymp - bd ] let @xmath112 and @xmath39 be the empirical copula and the empirical beta copula , respectively , of a sample @xmath17 of @xmath23-variate vectors without ties in any of the @xmath23 components .
we have @xmath134^d}|{\mathbb{c}}_n({\boldsymbol{u}})-{\mathbb{c}}_n^\beta({\boldsymbol{u}})| \leq d[n^{-1/2}(\log n)^{1/2}+n^{-1/2}+n^{-1}].\ ] ] proof consider the ranks @xmath9 as in . using the identity @xmath135
we have , for @xmath129^d$ ] , combining and , @xmath136 let @xmath137 and fix @xmath129^d$ ] .
let @xmath138 be the set of indices @xmath2 such that @xmath139 $ ] for some @xmath8 .
the set @xmath138 contains at most @xmath140 elements because for each @xmath8 , the ranks @xmath141 constitute a permutation of @xmath142 .
it follows that @xmath143 let @xmath144 and @xmath33 $ ] .
let @xmath145 denote a binomial random variable with @xmath22 trials and success probability @xmath146 . if @xmath147 , then , in view of and by hoeffding s inequality , @xmath148 similarly , if @xmath149 , then @xmath150 the two displays taken together imply that for @xmath151 $ ] , we have @xmath152 combine and to see that @xmath153 the right - hand side does not depend on @xmath154 and we still have the liberty of choosing @xmath155 . choosing @xmath156 yields . by proposition
[ prop : nonasymp - bd ] , the empirical beta copula is uniformly consistent whenever the empirical copula is .
nevertheless , the bound in is too weak to conclude that the empirical processes @xmath157 and @xmath158 are asymptotically the same .
this is the topic of the section [ sec : asym ] .
we provide asymptotic theory for the empirical bernstein copula process @xmath159 under weaker assumptions than those in janssen et al .
@xcite . here
@xmath160 is a sequence of multi - indices such that @xmath161 as @xmath162 . for @xmath115 and @xmath129^d$ ]
, let @xmath163 be the law of the random vector @xmath164 , where @xmath165 are independent random variables , the law of @xmath166 being binomial@xmath167 for each @xmath41 . for asymptotic analysis ,
it is convenient to write the empirical bernstein copula as a mixture : @xmath168^d}{\mathbb{c}}_n({\boldsymbol{w}})\,{{\mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w } } ) , \qquad { \boldsymbol{u}}\in [ 0 , 1]^d.\ ] ] the empirical copula process and the empirical bernstein copula process are defined by @xmath169 in view of lemma [ lem : betabernstein ] , the special case @xmath170 yields the empirical beta copula process : @xmath171 noting that @xmath172^d}c({\boldsymbol{w}})\,{{\mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w}})$ ] , we have @xmath173^d}{\mathbb{g}}_n({\boldsymbol{w}})\,{{\mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w } } ) + \sqrt{n}\bigl(b_{{\boldsymbol{m}}}(c)({\boldsymbol{u}})-c({\boldsymbol{u}})\bigr).\ ] ] we call the first and second term on the right - hand side of the stochastic term and the bias term , respectively .
we deal with these terms in subsections [ subsec : stoch ] and [ subsec : bias ] .
the combination of both analyses in subsection [ subsec : thm ] then yields our main result , theorem [ thm : empbernstein ] below , on the asymptotic distribution of the empirical bernstein copula process .
let @xmath174^d)$ ] be the banach space of real - valued , bounded functions on @xmath132^d$ ] , equipped with the supremum norm @xmath175 .
the arrow @xmath176 denotes weak convergence in the sense used in van der vaart and wellner @xcite .
if @xmath177 are independent and identically distributed random vectors from a common distribution with copula @xmath4 , then , provided @xmath4 satisfies condition [ cond : s ] below , we have @xcite @xmath178 in @xmath179^d)$ ] , where @xmath180 and where @xmath181 is a centered , gaussian process with continuous trajectories and covariance function @xmath182^d.\ ] ] more generally , if @xmath177 is a strictly stationary time series whose stationary @xmath23-variate distribution has copula @xmath4 satisfying condition [ cond : s ] , then , provided certain mixing or weak dependence conditions hold , we still have weak convergence in @xmath179^d)$ ] of the empirical copula process , as shown in bcher and volgushev @xcite : @xmath183 with @xmath184 a tight , centered gaussian process on @xmath30^d$ ] whose covariance structure depends on the full distribution of the time series . in the analysis of the stochastic term in , the weak convergence in is all that we need to know of the empirical copula process . [
lem : stochastic - term ] if @xmath185 in @xmath174^d)$ ] as @xmath186 , and if the limiting process @xmath184 has continuous trajectories almost surely , then , whenever @xmath187 as @xmath186 , @xmath188^d}\left|\int_{[0,1]^d}{\mathbb{g}}_n({\boldsymbol{w}})\,{{\mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w } } ) -{\mathbb{g}}_n({\boldsymbol{u}})\right|={o_{\scriptscriptstyle p}}(1).\ ] ] proof let @xmath189 denote the maximum norm on @xmath190 . for @xmath191 , we have @xmath192^d } \lvert { \mathbb{g}}_n({\boldsymbol{w}})-{\mathbb{g}}_n({\boldsymbol{u } } ) \rvert \,{{\mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w}})\\ \leq 2\ , \lvert{\mathbb{g}}_n\rvert_\infty \ , \mu_{{\boldsymbol{m}},{\boldsymbol{u}}}(\{{\boldsymbol{w}}\colon \lvert { \boldsymbol{w}}-{\boldsymbol{u}}\rvert_\infty>{\varepsilon}\ } ) + \sup _ { \lvert { \boldsymbol{x}}-{\boldsymbol{y}}\rvert_\infty\leq{\varepsilon } } \lvert { \mathbb{g}}_n({\boldsymbol{x}})-{\mathbb{g}}_n({\boldsymbol{y } } ) \rvert.\end{gathered}\ ] ] for a given @xmath193 , we can find @xmath194 sufficiently small such that @xmath195 moreover , the weak convergence @xmath185 implies that @xmath196 as @xmath186 . finally , chebyshev s inequality yields @xmath197 which converges to zero as @xmath198 , uniformly in @xmath60^d$ ] .
since @xmath199 was arbitrary , the conclusion follows .
now we turn to the bias term on the right - hand side of : @xmath200^d } \bigl ( c({\boldsymbol{w}})-c({\boldsymbol{u } } ) \bigr ) \,{{\mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w}}).\ ] ] one - dimensional bernstein polynomials are well studied , and results on the accuracy of the approximation of a function by its associated bernstein polynomial are presented in lorentz @xcite and devore and lorentz ( * ? ? ?
* chapter 10 ) . we need to extend these to the multidimensional case .
[ lem : bias - term1 ] for a @xmath23-variate copula @xmath4 and a multi - index @xmath115 we have , writing @xmath201 , @xmath202 proof any copula is lipschitz with respect to the @xmath203-norm with lipschitz constant equal to @xmath204 .
we obtain @xmath205^d } \lvert c({\boldsymbol{w}})-c({\boldsymbol{u } } ) \rvert \,{{\mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w } } ) \leq\sum_{j=1}^d\int_{[0,1]^d } \lvert w_j - u_j \rvert \,{{\mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w}}).\ ] ] fix @xmath41 and let @xmath166 be a binomial@xmath167 random variable . by the cauchy
schwarz inequality , @xmath206^d } \lvert w_j - u_j \rvert \,{{\mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w } } ) \leq { \operatorname{e } } [ \lvert s_j / m_j - u_j \rvert ] \le \sqrt { { \operatorname{var}}(s_j / m_j ) } \le \frac{1}{2\sqrt{m_j}},\ ] ] which yields the conclusion .
according to lemma [ lem : bias - term1 ] , the uniform convergence rate of @xmath111 to @xmath4 is at least @xmath207 as @xmath198 . with a bit of additional smoothness , this can be strengthened to @xmath208 .
the following condition originates from segers @xcite .
[ cond : s ] for each @xmath120 , the copula @xmath4 has a continuous first - order partial derivative @xmath209 on the set @xmath210^d\colon 0<u_j<1\}$ ] . [
lem : bias - term2 ] if the copula @xmath4 satisfies condition [ cond : s ] , then @xmath211 proof by monotonicity and lipschitz continuity of @xmath4 , we have @xmath212 . fix @xmath213^d$ ] and write @xmath214 for @xmath53 $ ] .
the function @xmath215 is continuous and , by condition [ cond : s ] , continuously differentiable on @xmath216 with derivative @xmath217 for each @xmath7 such that @xmath218 , the corresponding term in the sum is zero no matter how @xmath219 is defined . if @xmath220 , then @xmath221 whenever @xmath222 , so that the partial derivative @xmath223 is well defined at @xmath224 by condition [ cond : s ] . by the fundamental theorem of calculus ,
we get @xmath225 hence , by fubini s theorem , @xmath226^d } \bigl ( c ( { \boldsymbol{w } } ) - c ( { \boldsymbol{u } } ) \bigr ) \ , { { \mathrm{d } } } \mu_{{\boldsymbol{m } } , { \boldsymbol{u } } } ( { \boldsymbol{w } } ) \\ & = \sum_{j=1}^d\int_0 ^ 1 \left ( \int_{[0,1]^d}(w_j - u_j ) \ , \dot{c}_j({\boldsymbol{w}}(t))\,{{\mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w } } ) \right ) \ , { { \mathrm{d}}}t.\end{aligned}\ ] ] we need to bound its absolute value uniformly in @xmath60^d$ ] .
fix @xmath8 and @xmath53 $ ] .
choose @xmath227 .
first , let @xmath60^d$ ] be such that @xmath228 $ ] .
since @xmath229 , we find @xmath230^d}(w_j - u_j ) \
, \dot{c}_j({\boldsymbol{w}}(t))\,{{\mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w}})\right\rvert \leq \int_{[0,1]^d } \lvert w_j - u_j \rvert \,{{\mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w } } ) \leq \sqrt{\frac{u_j(1-u_j)}{m_j } } < \sqrt{\frac{\delta}{m_*}}.\ ] ] second , let @xmath60^d$ ] be such that @xmath231 $ ] .
then @xmath232 is well defined . since @xmath233^d}(w_j - u_j)\,{{\mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u } } } ( { \boldsymbol{w}})=0 $ ] , we obtain @xmath234^d}(w_j - u_j ) \bigl ( \dot{c}_j({\boldsymbol{w}}(t))-\dot{c}_j({\boldsymbol{u } } ) \bigr ) \,{{\mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w } } ) \right ) { { \mathrm{d}}}t.\ ] ] choose @xmath235 .
let @xmath189 denote the maximum norm on @xmath190 .
split the integral over @xmath236^d$ ] into two parts , according to whether @xmath237 is larger than @xmath238 or not .
we have @xmath239^d}(w_j - u_j ) \bigl ( \dot{c}_j({\boldsymbol{w}}(t))-\dot{c}_j({\boldsymbol{u } } ) \bigr ) \,{{\mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w}})\right\rvert \\
\leq \sup _ { \lvert { \boldsymbol{w}}-{\boldsymbol{u}}\rvert_\infty\leq{\varepsilon } } \lvert \dot{c}_j({\boldsymbol{w}}(t))-\dot{c}_j({\boldsymbol{u } } ) \rvert \int_{[0,1]^d } \lvert w_j - u_j \rvert \ , { { \mathrm{d } } } \mu_{{\boldsymbol{m } } , { \boldsymbol{u}}}({\boldsymbol{w } } ) \\ + \int_{[0,1]^d } { \mathbf{1}}\{\lvert{\boldsymbol{w}}-{\boldsymbol{u}}\rvert_\infty > { \varepsilon}\ } \ , \lvert w_j - u_j \rvert \,{{\mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w}}).\end{gathered}\ ] ] here we have used the inequality @xmath240 again .
we now treat both terms on the right - hand side separately .
* first , fix @xmath193 . by condition [ cond : s ] , we can choose @xmath241 sufficiently small such that , for each @xmath8 ,
@xmath242^d\\ \delta\leq u_j\leq 1-\delta } } \sup_{\lvert { \boldsymbol{w}}-{\boldsymbol{u}}\rvert_\infty\leq{\varepsilon } } \lvert \dot{c}_j({\boldsymbol{w}})-\dot{c}_j({\boldsymbol{u } } ) \rvert \leq \eta.\ ] ] furthermore , we have @xmath243^d } \lvert w_j - u_j \rvert \ , { { \mathrm{d } } } \mu_{{\boldsymbol{m } } , { \boldsymbol{u}}}({\boldsymbol{w } } ) \le 1 / ( 2\sqrt{m_*})$ ] by . *
second , by the cauchy
schwarz inequality , @xmath244^d } { \mathbf{1}}\ { \lvert { \boldsymbol{w}}-{\boldsymbol{u}}\rvert_\infty > { \varepsilon}\ } \ , \lvert w_j - u_j \rvert \ , { { \mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w } } ) \leq \sqrt { \frac { \mu_{{\boldsymbol{m}},{\boldsymbol{u}}}(\{{\boldsymbol{w}}\colon \lvert { \boldsymbol{w}}- { \boldsymbol{u}}\rvert_\infty > { \varepsilon}\ } ) } { m _ * } } .\ ] ] by the markov inequality and , the numerator of the right - hand side converges to zero as @xmath198 , uniformly in @xmath60^d$ ] . by the above bounds
, it follows that @xmath245 .
since @xmath246 and @xmath193 were arbitrary , the result follows .
condition [ cond : s ] on @xmath4 is significantly weaker than the assumption that @xmath4 has bounded third - order partial derivatives , which janssen et al .
( 2012 ) require to prove their asymptotic results for the empirical bernstein copula .
the latter assumption is violated by many common parametric copula families .
janssen et al .
( 2012 ) in fact proved a pointwise @xmath247 convergence rate of the bias term under the assumption that @xmath4 has bounded third - order partial derivatives , but to obtain such a rate , the assumption is still too strong .
in fact , a lipschitz condition on the first - order partial derivatives already suffices .
[ lem : bias - term3 ] let @xmath4 be a copula satisfying condition [ cond : s ] .
let @xmath248^d$ ] and suppose that there exists @xmath249 such that , for each @xmath41 such that @xmath250 , we have @xmath251 and @xmath223 is lipschitz on @xmath252^d : \lvert { \boldsymbol{w}}- { \boldsymbol{u}}\rvert_\infty \le { \varepsilon}\}$ ] . then @xmath253 proof if @xmath254 for some @xmath41 , then , since the binomial@xmath255 distribution is concentrated at @xmath124 , we can omit the @xmath7th coordinate altogether and pass to the appropriate @xmath256-dimensional margins of @xmath111 and @xmath4 .
in this way , we can eliminate all margins @xmath7 such that @xmath254 .
it follows that , without loss of generality , we can assume that @xmath257 .
as in the proof of proposition [ lem : bias - term2 ] , we have the representation @xmath258^d } ( w_j - u_j ) \ , \bigl ( \dot{c}_j({\boldsymbol{w}}(t ) ) - \dot{c}_j({\boldsymbol{u } } ) \bigr ) \ , { { \mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w } } ) \right ) { { \mathrm{d}}}t.\ ] ] we fix @xmath259 and @xmath41 and split the integral over @xmath82 into two parts , according to whether @xmath260 is larger than @xmath238 or not .
this yields the following bound : @xmath261^d } ( w_j - u_j ) \ , \bigl ( \dot{c}_j({\boldsymbol{w}}(t ) ) - \dot{c}_j({\boldsymbol{u } } ) \bigr ) \
, { { \mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w } } ) \right\rvert \\
\le \int_{[0 , 1]^d } m _ * \ , \lvert w_j - u_j \rvert \
, \lvert \dot{c}_j({\boldsymbol{w}}(t ) ) - \dot{c}_j({\boldsymbol{u } } ) \rvert \ , { \mathbf{1}}\ { \lvert { \boldsymbol{w}}- { \boldsymbol{u}}\rvert_\infty \le { \varepsilon}\ } \ , { { \mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w } } ) \\ + \int_{[0 , 1]^d } m _ * \ , \lvert w_j - u_j \rvert \ , { \mathbf{1}}\ { \lvert { \boldsymbol{w}}- { \boldsymbol{u}}\rvert_\infty > { \varepsilon}\ } { { \mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w}}).\end{gathered}\ ] ] for the second term on the right - hand side , we used the fact that @xmath240 .
we now bound the two integrals separately . *
recall that @xmath262 , a convex combination of @xmath154 and @xmath82 .
the lipschitz assumption on @xmath223 implies that there exists a constant @xmath263 such that @xmath264 we find that the first term on the right - hand side of is bounded by @xmath265^d } m _ * \ , \lvert w_j - u_j \rvert \ ,
\lvert w_k - u_k \rvert \ , { { \mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w } } ) \le \frac{1}{4 } tld.\ ] ] here we applied the cauchy
schwarz inequality , by which @xmath266^d } m _ * \ , \lvert w_j - u_j \rvert \
, \lvert w_k - u_k \rvert \ , { { \mathrm{d}}}\mu_{{\boldsymbol{m}},{\boldsymbol{u}}}({\boldsymbol{w } } ) & = { \operatorname{e } } [ m^ * \lvert s_j / m_j - u_j \rvert \ , \lvert s_k / m_k - u_k \rvert ] \\ & \le m^ * \ , \left ( { \operatorname{e } } [ ( s_j / m_j - u_j)^2 ] \ , { \operatorname{e } } [ ( s_k / m_k - u_k)^2 ] \right)^{1/2 } \\ & \le m^ * \left ( m_j^{-1 } u_j(1-u_j ) \ , m_k^{-1 }
u_k(1-u_k ) \right)^{1/2 } \\ & \le \left ( u_j(1-u_j ) \ , u_k(1-u_k ) \right)^{1/2 } \le \frac{1}{4}.\end{aligned}\ ] ] * for the second term on the right - hand side of , we can bound the indicator function in the integrand by @xmath267 the resulting integrals have already been bounded above in the previous term .
this finishes the proof of proposition [ lem : bias - term3 ] . by lemma [ lem : betabernstein ]
, the empirical beta copula @xmath39 is equal to the empirical bernstein copula @xmath113 with @xmath268 for all @xmath41
. propositions [ lem : stochastic - term ] , [ lem : bias - term2 ] and [ lem : bias - term3 ] then immediately imply the following theorem and corollary . recall the empirical copula process @xmath269 , the empirical bernstein copula process @xmath270 , and the empirical beta copula process @xmath271 .
in theorem [ thm : empbernstein ] , we assume weak convergence of @xmath272 to some limit process @xmath184 . see the discussion before proposition [ lem : stochastic - term ] for a justification of this condition and for the form of the limit process . in particular , the condition is automatically satisfied in the iid case when @xmath4 satisfies condition [ cond : s ] , and then the limit process is given by @xmath273 in .
[ thm : empbernstein ] suppose that @xmath4 satisfies condition [ cond : s ] and that @xmath185 in @xmath174^d)$ ] as @xmath186 , the limiting process @xmath184 having continuous trajectories almost surely .
let @xmath274 be multi - indices such that @xmath161 as @xmath162 .
a. if @xmath275 , then , for all @xmath248^d$ ] that satisfy the condition in proposition [ lem : bias - term3 ] , we have @xmath276 b. if @xmath277 , then , in @xmath179^d)$ ] , @xmath278 proof recall the decomposition of @xmath279 in .
the stochastic term is equal to @xmath280 , uniformly in @xmath129^d$ ] , by proposition [ lem : stochastic - term ] .
\(i ) since @xmath281 , the bias term converges to zero at the given point @xmath154 .
\(ii ) since @xmath282 , the bias term converges to zero uniformly in @xmath129^d$ ] thanks to proposition [ lem : bias - term2 ] . setting @xmath283 in theorem [ thm : empbernstein](ii ) yields the following corollary .
[ cor : empbeta ] suppose that @xmath4 satisfies condition [ cond : s ] and that @xmath185 in @xmath174^d)$ ] as @xmath186 , the limiting process @xmath184 having continuous trajectories almost surely . then , in @xmath179^d)$ ] , @xmath284 the conclusion of this section is that under reasonable conditions , the empirical bernstein and empirical beta copulas have the same large - sample distribution as the empirical copula .
the real advantage of the use of the smoothed versions is visible mainly for small samples .
the difference can perhaps be quantified via higher - order asymptotic theory .
instead , we assess the finite - sample performance through monte carlo simulations .
in this section , we compare the finite - sample performance of the empirical beta copula with those of various other estimators by a monte carlo experiment .
the simulations were performed in @xcite using the package @xcite .
we show the results for five copulas , three bivariate ones and two trivariate ones : * the bivariate farlie gumbel
morgenstern ( fgm ) copula with negative dependence ( @xmath285 ) ; * the bivariate independence copula ; * the bivariate gaussian copula with positive dependence ( @xmath286 ) ; * the trivariate @xmath287 copula with degrees - of - freedom parameter equal to @xmath288 and pairwise correlation parameters equal to @xmath289 , @xmath290 and @xmath291 ; * the trivariate nested archimedean copula with frank generators and with kendall s tau equal to @xmath292 at the upper node and @xmath293 at the lower node .
see , e.g. , joe @xcite and nelsen @xcite for the explicit functional forms and properties of these copulas .
we have run simulations for many other copula models , but the results were qualitatively the same as the ones showed here .
we compared the following five estimators : * the empirical copula @xmath112 in ; * the empirical checkerboard copula @xmath294 in below ; * the empirical beta copula @xmath39 in ; * the empirical bernstein copula @xmath295 with @xmath296 and @xmath297 , the smallest integer not smaller than @xmath298 ; * the empirical bernstein copula @xmath295 with @xmath296 and @xmath299 as recommended by janssen et al .
@xcite , see below . for each possible combination of estimator and model and for sample sizes @xmath22 from @xmath300
up to @xmath301 , we computed the integrated ( over the unit cube ) squared bias , the integrated variance , and the integrated mean squared error by monte carlo simulation .
see appendix [ app : estim ] for the definitions of the three performance measures and a description of our computation method . as the simplest version of smoothed empirical copula
, we introduce the _ empirical checkerboard copula _ defined by @xmath302 ( li et al .
@xcite for @xmath303 ; carley and taylor @xcite for general @xmath23 ) .
we note that this is a genuine copula , just like the empirical beta copula . in each of the graphs in figure [ fig : emp - checker - beta ] ,
the horizontal axis indicates the sample size ; the vertical axis indicates the integrated squared bias , the integrated variance and the integrated mean squared error for the left , middle and right panels , respectively ( the same comment also applies to the other figures ) . from the figures ,
one sees that the empirical copula performs worst with respect to all three measures .
the empirical beta copula and the empirical checkerboard copula are comparable in terms of bias , but the variance of the former is smaller than the one of the latter , resulting in a smaller mean squared error for the empirical beta copula in all cases considered .
note that the empirical checkerboard copula can be thought of as smoothing at bandwidth @xmath304 , whereas the empirical beta copula involves smoothing at bandwidth @xmath305 . in this subsection
, we consider the empirical bernstein copula with all the orders @xmath306 equal to an integer @xmath299 .
our first choice for the order of bernstein polynomial is @xmath297 , where @xmath307 denotes the smallest integer not smaller than @xmath119 . with this choice ,
we see the consequence of proposition [ prop : bernstein - genuine - copula ] because @xmath299 is a divisor of @xmath22 only when @xmath22 is a multiple of 3 .
our second choice for @xmath299 is based on janssen et al .
@xcite , who recommend , in the bivariate case , the following choice for @xmath299 : @xmath308 ( or the integer part thereof ) , where @xmath309 and with @xmath223 and @xmath310 the first- and second - order partial derivatives of @xmath4 with respect to @xmath76 , for @xmath311 .
the following remarks on this choice @xmath312 are in order . *
this choice requires the knowledge of @xmath4 through the derivatives @xmath223 and @xmath310 . to estimate these derivatives in small samples
is not easy .
* for the independence copula , we have @xmath313 , so that it is not clear how to define @xmath312 . * in lemma 3(i ) in @xcite , an @xmath314 term is written but subsequently neglected . however , the order of magnitude of this term could be larger than the remainder terms that are analyzed later on . in the simulations ,
we take the true ( but in practice unknown ) value for @xmath315 for the bivariate fgm and gaussian copulas . from figure
[ fig : beta - bern3-bernjsv ] , it is observed that the performance of the empirical bernstein copulas are strongly affected by the periodicity due to the divisibility relation between @xmath22 and @xmath299 , especially in terms of bias .
compared to the empirical bernstein copulas , the empirical beta copula has a smaller bias and a larger variance in every case .
there is no clear ordering of the three in terms of the mean squared error .
it is interesting to note that while the mean squared error of the empirical beta copula is smaller in positively dependent cases , it is larger in negatively dependent cases . at this point , we have no explanation for this phenomenon .
the empirical beta copula can be considered as a simple but effective way of correcting and smoothing the empirical copula .
no smoothing parameter needs to be chosen .
the empirical beta copula is a special case of the empirical bernstein copula , but in contrast to the latter , it is always a genuine copula .
moreover , it is extremely simple to simulate samples from it .
the asymptotic distribution of the empirical beta copula is the same as that of the empirical copula , but in small samples , it performs better both in terms of bias and variance .
moreover , there seems little to be gained from using bernstein smoothers at other polynomial degrees @xmath299 , except in special cases such as the independence copula .
we also note that the only properties of the kernels @xmath163 that we needed in the asymptotic analysis were moment properties and concentration inequalities of their margins around the points @xmath76 .
this implies that the approach may perhaps be extended to products of other families of smoothing kernels than beta kernels .
our smoothing procedure might also have a beneficial effect on the accuracy of resampling schemes for the empirical copula process ( bcher and dette @xcite ) . more specifically ,
testing procedures based on the empirical copula typically rely on the bootstrap for the computation of the critical values of the test statistic . for finite samples ,
the accuracy is often not very good : the true type i error of the test may differ greatly from the nominal one . then , the question is how to construct a bootstrap or multiplier resampling scheme for the empirical beta and bernstein copulas .
this is a topic for future research , together with the question of higher - order asymptotics for the various nonparametric copula estimators .
j. segers gratefully acknowledges funding by contract `` projet dactions de recherche concertes '' no .
12/17 - 045 of the `` communaut franaise de belgique '' , by iap research network grant p7/06 of the belgian government ( belgian science policy ) , and by the `` projet de recherche '' no . frfc pdr t.0146.14 of the `` fonds de la recherche scientifique fnrs '' ( belgium ) .
h. tsukahara is supported by jsps kakenhi grant number 15h03337 .
we derive a convenient expression for the derivative of the bernstein - type polynomial ( making no claim of originality ) .
. then we have @xmath317\\[5pt ] & = \begin{cases } -n(1-t)^{n-1 } & \text{if $ r=0$}\\[5pt ] \displaystyle\frac{n!}{(r-1)!(n - r)!}t^{r-1}(1-t)^{n - r } -\frac{n!}{r!(n - r-1)!}t^r(1-t)^{n - r-1 } & \text{if $ r=1,\ldots , n-1$}\\[10pt ] nt^{n-1 } & \text{if $ r = n$ } \end{cases}\end{aligned}\ ] ] using the above expression and summation by parts , it is easy to see that for any sequence @xmath318 , @xmath319 , the following identity holds : @xmath320 when a function is differentiable , we can give a condition for @xmath23-increasing property in terms of its partial derivative .
[ prop : d - increasing - partial - derivative ] suppose that a function @xmath107^d{\to}[0,1]$ ] is infinitely differentiable on @xmath321
. then @xmath109 is @xmath23-increasing if and only if @xmath322 for all @xmath90 . _
proof_. suppose that @xmath109 is @xmath23-increasing .
then , since @xmath109 is infinitely differentiable on @xmath321 , it follows that for @xmath90 and @xmath323 , @xmath324 , @xmath325 where the sum is taken over all @xmath82 such that @xmath218 or @xmath326 for all @xmath324 .
conversely , suppose that @xmath322 . by the standard calculus , for all @xmath79^d$ ] with @xmath80 for all @xmath324 , we have @xmath327 where the sum is taken over all @xmath82 such that @xmath218 or @xmath328 for all @xmath324 . @xmath329
given a copula estimator @xmath330 , we consider in section [ sec : simul ] the following three performance measures : @xmath331^d } \bigl\ { { \operatorname{e } } [ { \widehat}{c}_n ( { \boldsymbol{u } } ) - c ( { \boldsymbol{u } } ) ] \bigr\}^2 \ , { \mathrm{d}}{\boldsymbol{u } } ; \\ \text{integrated variance : } \quad & \int_{[0 , 1]^d } { \operatorname{e}}\bigl [ \bigl\ { { \widehat}{c}_n ( { \boldsymbol{u } } ) - { \operatorname{e } } [ { \widehat}{c}_n ( { \boldsymbol{u } } ) ] \bigr\}^2 \bigr ] \ , { \mathrm{d}}{\boldsymbol{u } } ; \\ \text{integrated mean squared error : } \quad & \int_{[0 , 1]^d } { \operatorname{e}}\bigl [ \bigl\ { { \widehat}{c}_n ( { \boldsymbol{u } } ) - c ( { \boldsymbol{u } } ) \bigr\}^2 \bigr ] \ , { \mathrm{d}}{\boldsymbol{u}}.\end{aligned}\ ] ] to compute these , we apply the following trick .
let @xmath332 and @xmath333 be independent replications of the estimator , and let the random vector @xmath334 be independent of the two estimators and uniformly distributed on @xmath30^d$ ] .
then each of the three performance measures above can be written as a single expectation with respect to @xmath335 : @xmath336 ; \\ \text{integrated variance : } \quad & { \operatorname{e}}\left [ \frac{1}{2 } \bigl\ { { \widehat}{c}_n^{(1 ) } ( { \boldsymbol{v } } ) - { \widehat}{c}_n^{(2 ) } ( { \boldsymbol{v } } ) \bigr\}^2 \right ] ; \\
\text{integrated mean squared error : } \quad & { \operatorname{e}}\left [ \frac{1}{2 } \sum_{j=1}^2 \bigl\ { { \widehat}{c}_n^{(j ) } ( { \boldsymbol{v } } ) - c ( { \boldsymbol{v } } ) \bigr\}^2 \right].\end{aligned}\ ] ] to see these identities , first condition on @xmath334 and then use the fact that @xmath332 and @xmath333 are independent of @xmath334 and are independent random copies of @xmath330 . to compute the above three expectations , we rely on monte carlo simulation . for a given large integer @xmath337 , we simulate the independent random vectors @xmath338 where @xmath339 .
the random vectors @xmath340 are sampled from @xmath4 and the random vectors @xmath334 are sampled from the uniform distribution on @xmath30^d$ ] .
then we compute the copula estimator under consideration based on the samples @xmath341 and @xmath342 and evaluated at @xmath343 , yielding @xmath344 and @xmath345 , respectively .
we also compute the true copula , @xmath4 , at @xmath343 .
then we compute the desired function of @xmath344 , @xmath345 and @xmath346 as in the three expectations above . finally , we average over the @xmath337 samples to obtain a monte carlo estimate of the desired performance measure . 99 baker , r. ( 2008 ) . an order - statistics - based method for constructing multivariate distributions with fixed marginals , _ journal of multivariate analysis
, * 99 * , 23122327 .
bcher , a. and dette , h. ( 2010 ) . a note on bootstrap approximations for the empirical copula process , _ statistics and probability letters _ ,
* 80 * , 19251932 .
bcher , a. and volgushev s. ( 2013 ) .
empirical and sequential empirical copula processes under serial dependence , _ journal of multivariate analysis _ , * 119 * , 6170 .
carley , h. and taylor , m. d. ( 2002 ) . a new proof of sklar s theorem , in : cuadras , c. m. , fortiana , j. and rodrguez - lallena , j. a. ( eds . ) ( 2002 ) . _ distributions with given marginals and statistical modelling _ , p.p .
2934 , kluwer academic publishers , dordrecht .
deheuvels , p. ( 1979 ) .
la fonction de dpendence empirique et ses proprits , un test non paramtrique dindpendance , _ bulletin de la classe des sciences , acadmie royale de belgique , 5e srie _ , * 65 * , 274292 .
devore , r. a. and lorentz , g. g. ( 1993 ) . _ constructive approximations _
, springer , berlin - heidelberg .
fermanian , j .- d
. , radulovi , d. and wegkamp , m. j. ( 2004 ) .
weak convergence of empirical copula processes .
_ bernoulli _ , * 10 * , 847860 .
li , x. , mikusiski , p. , sherwood , h. and taylor , m. d. ( 1997 ) . on approximation of copulas , in : bene , v. and tpn , j. ( eds . ) ( 1997 ) .
_ distributions with given marginals and moment problems _ , p.p . 107116 , kluwer academic publishers , dordrecht .
marius hofert , ivan kojadinovic , martin maechler and jun yan ( 2015 ) .
_ copula : multivariate dependence with copulas .
_ r package version 0.999 - 14 .
url http://cran.r-project.org/package=copula .
janssen , p. , swanepoel , j. and veraverbeke , n. ( 2012 ) .
large sample behavior of the bernstein copula estimator , _ journal of statistical planning and inference _ , * 142 * , 11891197 .
joe , h. ( 1997 ) .
_ multivariate models and dependence concepts _ , chapman and hall , london .
lorentz , g. g. ( 1986 ) , _ bernstein polynomials _ , 2nd ed . , chelsea publishing company , new york .
mai , j .- f . and
scherer , m. ( 2012 ) .
_ simulating copulas : stochastic models , sampling algorithms , and applications _ , imperial college press , london .
nelsen , r. b. ( 2006 ) .
_ an introduction to copulas _ , 2nd ed . , springer - verlag , new york .
r core team ( 2016 ) .
_ r : a language and environment for statistical computing . _
r foundation for statistical computing , vienna , austria .
url https://www.r - project.org/. sancetta , a. and satchell , s. ( 2004 ) .
the bernstein copula and its applications to modeling and approximations of multivariate distributions , _ econometric theory _ , * 20 * , 535562 .
segers , j. ( 2012 ) .
asymptotics of empirical copula processes under nonrestrictive smoothness assumptions , _ bernoulli _ , * 18 * , 764782 .
sklar , m. ( 1959 ) .
fonctions de rpartition n dimensions et leurs marges , _ publ .
statist .
paris _ , * 8 * , 229231 .
tsukahara , h. ( 2005 ) .
semiparametric estimation in copula models , _ canadian journal of statistics _ , * 33 * , 357375 [ erratum : _ canadian journal of statistics _ , * 39 * , 734735 ( 2011 ) ] .
van der vaart , a. w. and wellner , j. a. ( 1996 ) .
_ weak convergence and empirical processes : with applications to statistics _ , springer - verlag , new york .
|
given a sample from a multivariate distribution @xmath0 , the uniform random variates generated independently and rearranged in the order specified by the componentwise ranks of the original sample look like a sample from the copula of @xmath0 .
this idea can be regarded as a variant on baker s [ j. multivariate anal .
99 ( 2008 ) 23122327 ] copula construction and leads to the definition of the empirical beta copula .
the latter turns out to be a particular case of the empirical bernstein copula , the degrees of all bernstein polynomials being equal to the sample size . necessary and sufficient
conditions are given for a bernstein polynomial to be a copula .
these imply that the empirical beta copula is a genuine copula .
furthermore , the empirical process based on the empirical bernstein copula is shown to be asymptotically the same as the ordinary empirical copula process under assumptions which are significantly weaker than those given in janssen , swanepoel and veraverbeke [ j. stat . plan . infer . 142 ( 2012 )
11891197 ] .
a monte carlo simulation study shows that the empirical beta copula outperforms the empirical copula and the empirical checkerboard copula in terms of both bias and variance .
compared with the empirical bernstein copula with the smoothing rate suggested by janssen et al .
, its finite - sample performance is still significantly better in several cases , especially in terms of bias . _
ams 2010 subject classifications _ : 62g20 , 62g30 , 62h12 + _ keywords _ : copula , empirical copula , bernstein polynomial , empirical bernstein copula , checkerboard copula
|
over the past decade , the physical characteristic of the entanglement of quantum - mechanical states , both pure and mixed , has been recognized as a central resource in various aspects of quantum information processing .
significant settings include quantum communication @xcite , cryptography @xcite , teleportation @xcite , and , to an extent that is not quite so clear , quantum computation @xcite . given the central status of entanglement , the task of quantifying the degree to which a state is entangled is important for quantum information processing and , correspondingly , several measures of it have been proposed .
these include : entanglement of formation @xcite , entanglement of distillation @xcite , relative entropy of entanglement @xcite , negativity @xcite , and so on .
it is worth remarking that even for the smallest hilbert space capable of exhibiting entanglement , i.e. , the two - qubit system ( for which wootters has determined the entanglement of formation @xcite ) , there are aspects of entanglement that remain to be explored . among the family of mixed quantum mechanical states , special status should be accorded to those that , for a given value of the entropy @xcite , have the largest possible degree of entanglement @xcite .
the reason for this is that such states can be regarded as mixed - state generalizations of the bell states , the latter being known to be the maximally entangled 2-qubit pure states .
the notion of maximally entangled mixed states was introduced by ishizaka and hiroshima @xcite in a closely related setting , i.e. , that of 2-qubit mixed states whose entanglement is maximized at fixed eigenvalues of the density matrix ( rather than at fixed entropy of the density matrix ) .
evidently , the entanglement of the maximally entangled mixed states of ishizaka and hiroshima can not be increased by any _
global _ unitary transformation . for these states , it was shown by verstraete et al .
@xcite that the maximality property continues to hold if any of the following three measures of entanglement entanglement of formation , negativity , and relative entropy of entanglement is replaced by one of the other two .
the question of the ordering of entanglement measures was raised by eisert and plenio @xcite and investigated numerically by them and analytically by verstraete et al .
it was proved by virmani and plenio @xcite that all good asymptotic entanglement measures are either identical or fail to uniformly give consistent orderings of density matrices .
this implies that the resulting maximally entangled mixed states ( mems ) may depend on the measures one uses to quantify entanglement .
moreover , in finding the form of mems , one needs to quantify the _ mixedness _ of a state , and there can also be ordering problems for mixedness .
this implies that the mems may depend on the measures of mixedness as well .
this paper is organized as follows .
we begin , in secs .
[ sec : mems ] and [ sec : measofmix ] , by reviewing several measures of entanglement and mixedness . in the main part of the paper , sec .
[ sec : frontier ] , we consider various entanglement - versus - mixedness planes , in which entanglement and mixedness are quantified in several ways . our primary objective , then
, is to determine the _ frontiers _ ,
i.e. , the boundaries of the regions occupied by physically allowed states in these planes , and to identify the structure of these maximally entangled mixed states . in sec .
[ sec : conclude ] we make some concluding remarks .
it is well known that there are a large number of entanglement measures @xmath0 . for a state described by the density matrix
@xmath1 a good entanglement measure must satisfy , at least , the following conditions @xcite : + c1 .
( a ) @xmath2 ; ( b ) @xmath3 if @xmath1 is not entangled @xcite ; and ( c ) @xmath4
. + c2 . for any state @xmath1 and any local unitary transformation ,
i.e. , a unitary transformation of the form @xmath5 , the entanglement remains unchanged .
local operations , classical communication and postselection can not increase the expectation value of the entanglement .
entanglement is convex under discarding information : @xmath6 .
+ the entanglement quantities chosen by us satisfy the properties c1c4 . here
, we do not impose the condition that any good entanglement measure should reduce to the _ entropy of entanglement _ ( to be defined in the following ) for pure states .
the first measure we shall consider is the entanglement of formation @xmath7 @xcite ; it quantifies the amount of entanglement necessary to create the entangled state .
it is defined by @xmath8 where the minimization is taken over those probabilities @xmath9 and pure states @xmath10 that , taken together , reproduce the density matrix @xmath11 ( usually called the _ entropy of entanglement _ ) measures the entanglement of the pure state @xmath12 and is defined to be the von neumann entropy of the reduced density matrix @xmath13 , i.e. , @xmath14 for two - qubit systems , @xmath7 can be expressed explicitly as @xcite @xmath15 \right ) , \label{eqn : wootters } \\
h(x ) & \equiv & -x \log_2 x - ( 1-x ) \log_2(1-x ) , \label{eqn : entropy}\end{aligned}\ ] ] where @xmath16 , the _ concurrence _ of the state @xmath1 , is defined as @xmath17 in which @xmath18 are the eigenvalues of the matrix @xmath19 in nondecreasing order and @xmath20 is a pauli spin matrix .
@xmath21 , @xmath16 , and the _ tangle _ @xmath22 are equivalent measures of entanglement , inasmuch as they are monotonic functions of one another .
a measure associated with the entanglement of formation is the entanglement cost @xmath23 @xcite which is defined via @xmath24 this is the asymptotic value of the average entanglement of formation .
@xmath23 is , in general , difficult to calculate .
related to the entanglement of formation is the entanglement of distillation @xmath25 @xcite , which characterizes the amount of entanglement of a state @xmath1 as the fraction of bell states that can be distilled using the optimal purification procedure : @xmath26 , where @xmath27 is the number of copies of @xmath1 used and @xmath28 is the maximal number of bell states that can be distilled from them .
the difference @xmath29 can be regarded as _
undistillable entanglement_. @xmath25 is a difficult quantity to calculate , but the relative entropy of entanglement @xmath30 @xcite , which we shall define shortly , provides an upper bound on @xmath25 and is more readily calculable than it .
for this reason , it is the second measure that we consider in this paper .
it is defined variationally via @xmath31 where @xmath32 represents the ( convex ) set of all separable density operators @xmath33 . in certain ways ,
the relative entropy of entanglement can be viewed as a _ distance _ @xmath34 from the entangled state @xmath1 to the closest separable state @xmath35 .
we remark that for pure states of two - qubit systems the relative entropy has the same value as the entanglement of formation .
the third measure that we shall consider is the _ negativity_.
the concept of the negativity of a state is closely related to the well - known peres - horodecki condition for the separability of a state @xcite .
if a state is separable ( i.e. , not entangled ) then the partial transpose of its density matrix is again a valid state , i.e. , it is positive semi - definite .
it turns out that the partial transpose of a non - separable state has one or more negative eigenvalues .
the negativity of a state @xcite indicates the extent to which a state violates the positive partial transpose separability criterion .
we will adopt the definition of negativity as twice the absolute value of the sum of the negative eigenvalues : @xmath36 where @xmath37 is the sum of the negative eigenvalues of @xmath38 .
in @xmath39 ( i.e. , two - qbit ) systems it can be shown that the partial transpose of the density matrix can have at most one negative eigenvalue ( see app .
[ app : oneneg ] ) .
it was proved by vidal and werner @xcite that negativity is an entanglement _ monotone _ , i.e. , it satisfies criteria c1c4 and , hence , is a good entanglement measure .
we remark that for two - qubit pure states the negativity gives the same value as the concurrence does .
the bures metric of entanglement is defined as @xmath40where @xmath41 is the _
fidelity_. in the same way that relative entropy can , this entanglement measure can be viewed as the distance from the closest separable state to the entangled state considered , where the distance is now defined by @xmath42 @xcite .
we remark that for two - qubit pure states the bures metric reduces to the tangle defined in sec .
[ sec : ef ] .
it was shown by lewenstein and sanpera @xcite that any density matrix @xmath1 has a decomposition into two parts : @xmath43 where @xmath44 is separable , @xmath45 is entangled , and the weight @xmath46 is maximal , in which case the decomposition is unique .
they refer to @xmath44 as the _ best separable approximation _ ( bsa ) to @xmath1 .
it should be pointed out that , in general , it is nontrivial to establish the decomposition , even in the simplest relevant setting of @xmath39 systems
. evidently , @xmath46 and @xmath45 contain information about the entanglement of @xmath1 .
karnas and lewenstein @xcite later showed that the quantity @xmath47 , which we will call _ the ls entanglement _ , satifies the above criteria , and hence is a good entanglement measure . for @xmath39 systems it turns out that @xmath48 is a pure state , i.e. , @xmath49 , and in this case it suggests that the quantity @xmath50 , defined via @xmath51 may also be a good entanglement measure .
we remark that the for two - qubit case the entanglement measure ( @xmath52 ) is known to be equal to the schmidt measure introduced in ref .
@xcite . even though the ls decomposition is not , in general , straightforward to find for the states in eq .
( [ eqn : ansatz ] ) , the ls decomposition reads @xmath53 where @xmath54 ( with @xmath55 ) , @xmath56 , and we consider only @xmath57 ( as for @xmath58 the whole density matrix is separable ) . if we compute the concurrence of the state in eq .
( [ eqn : ansatz ] ) , we find @xmath59 .
it is interesting to note that the ls entanglement gives , for the ansatz states in eq .
( [ eqn : ansatz ] ) , the same value as the concurrence does ; but for an arbitrary state this is not , in general , true @xcite .
we now pause to touch on certain difficulties posed by the task of ordering physical states using entanglement . as first discussed and explored numerically by eisert and plenio @xcite , and subsequently investigated analytically by verstraete et al .
@xcite , different entanglement measures can give different orderings for pairs of mixed states .
this can be seen , e.g. , from the plot of concurrence versus negativity , fig .
[ fig : cvsn ] .
the upper boundary is readily seen to be @xmath60 whereas the lower boundary can be derived , giving @xmath61 ; see ref .
@xcite . hence , when we wish to explain maximally entangled mixed states we need to be very explicit about the measure of entanglement ( and also mixedness
; see the following section ) .
different measures are likely to lead to different classes of mems states .
we end this section by mentioning the three entanglement measures that we shall use to compute the entanglement - mixedness frontiers : entanglement of formation , negativity , and relative entropy of entanglement .
the first two of these are straightforward to compute , at least in two - qubit settings . for the third ,
certain results are available @xcite that ease the computation of mems .
we have also reviewed four additional measures ( entanglement cost , entanglement of distillation , ls entanglement , and the bures metric ) .
of these , however , the first two are rather difficult to compute , let alone maximize ; the third is also difficult to compute , at least in practice . as for the fourth , calculating the entanglement involves finding the closest separable states ( as is required for the case of relative entropy ) .
in the entanglement - measure literature , two measures of mixedness have basically been used : @xmath62\big)$ ] and the von neumann entropy . whereas the latter has an natural significance stemming from its connections with statistical physics and information theory ,
the former is substantially easier to calculate .
of course , for density matrices that are almost completely mixed , the two measures show the same trend .
the von neumann entropy , the standard measure of randomness of a statistical ensemble described by a density matrix , is defined by @xmath63 where @xmath64 are the eigenvalues of the density matrix @xmath1 and the @xmath65 is taken to base @xmath66 , the dimension of the hilbert space in question .
it is straightforward to show that the extremal values of @xmath67 are zero ( for pure states ) and unity ( for completely mixed states ) . to compute the von neumann
entropy it is necessary to have the full knowledge of the eigenvalue spectrum .
as we shall mention in the following subsection , there is a linear entropy threshold above which all states are separable .
qualitatively identical behavior is encountered for the von neumann entropy . in particular , as we shall see in sec .
[ sec : efsv ] , for two - qubit systems all states are separable for @xmath68 .
the second measure that we shall consider is called the _ linear entropy _ and is based on the purity of a state , @xmath69 $ ] , which ranges from 1 ( for a pure state ) to @xmath70 for a completely mixed state with dimension @xmath66 .
the linear entropy @xmath71 is defined via @xmath72 ) , \label{eqn : linearentropy}\end{aligned}\ ] ] which ranges from 0 ( for a pure state ) to 1 ( for a maximally mixed state ) .
the linear entropy is generally a simpler quantity to calculate than the von neumann entropy as there is no need for diagonalization . for @xmath39 systems
the linear entropy can be written explicity as @xmath73).\end{aligned}\ ] ] a related measure , which we shall not use in this paper ( but mention for the sake of completeness ) , is _ the inverse participation ratio_. defined via @xmath74 $ ] , it ranges from @xmath75 ( for a pure state ) to @xmath66 ( for the maximally mixed state ) .
an attractive property of the inverse participation ratio is that all states with @xmath76 are separable @xcite , which implies all states with a linear entropy @xmath77 ( which is @xmath78 when @xmath79 ) are separable .
the aim of this subsection is to illustrate the difference between the linear and von neumann entropies .
we shall do this by considering the @xmath79 hilbert space , and seeking the highest and lowest von neumann entropies consistent with a given value of linear entropy . before restricting @xmath66 to 4
, the corresponding stationarity problem reads : @xmath80 where @xmath81 and @xmath82 are , respectively , langrange multipliers that enforce the constraints that linear entropy be fixed and that @xmath1 be normalized .
thus , we arrive at the engaging self - consistency condition @xmath83 in which @xmath82 and @xmath81 can be fixed upon implementing the constraints . by hypothesizing certain relationships among the eigenvalues of the density matrix , we have been able , for the case of @xmath79 , to find certain solutions of the self - consistency condition and , hence , a candidate for the boundary of the region of @xmath71 vs. @xmath67 plane ( see fig . [ fig : svsl ] ) corresponding to legitimate density matrices . these solutions , when given in terms of eigenvalues , read @xmath84 [ eqn : svslsolutions ] and they correspond to the upper boundary , and the lowest , middle , and highest pieces of the lower boundary , respectively .
note that the lower boundary comprises three ( in general , @xmath85 ) segments that meet at cusps .
we remark , parenthetically , that the solutions with zero eigenvalues correspond to extrema within some subspace spanned by those eigenvectors with nonzero eigenvalues , and therefore only obey the stationarity condition ( [ eqn : stationarity ] ) within the subspace .
is there any significance to the boundary states ?
boundary segment ( a ) includes the werner states defined in eq .
( [ eqn : werner ] ) .
boundary segment ( b ) includes the first branch of the mems for @xmath7 and @xmath71 specified below in eq .
( [ eqn : mems ] ) .
the segment ( c ) includes the states @xmath86states on segment ( d ) are all unentangled .
of course , the boundary segments include not only the specified states but also all states derivable from them by global unitary transformation . as for the interior
, we have obtained this numerically by constructing a large number of random sets of eigenvalues of legitimate density matrices , and computing for each the two entropies . as fig .
[ fig : svsl ] shows , no points lie outside the boundary curve , providing confirmatory evidence for the forms given in eq .
( [ eqn : svslsolutions ] ) .
the fact that the bounded region is two - dimensional indicates the lack of precision with which the linear entropy characterizes the von neumann entropy ( and vice versa , if one wishes ) .
in particular , the figure reveals an ordering difficulty : pairs of states , @xmath87 and @xmath88 , exist for which @xmath89 and @xmath90 differ in sign .
worse still , states having a common value of @xmath67 have a continuum of values of @xmath71 , and vice versa .
we now attempt to identify regions in the plane spanned by entanglement and mixedness that are inhabited by physical states ( i.e. , characterized by legitimate density matrices ) .
we shall consider the various measures of entanglement and mixedness discussed in the previous section .
of particular interest will be the structure of the states that inhabit the _ frontier _ , i.e. , the boundary delimiting the region of physical states .
frontier states are maximal in the following sense : for a given value of mixedness they are maximally entangled ; for a given value of entanglement they are maximally mixed .
the aim of this subsection is to derive the general form of the maximal states given in eq .
( [ eqn : rhomems ] ) , which is what we will use to parametrize maximal states . in ref .
@xcite , it is shown that , given a fixed set of eigenvalues , all states that maximize one of the three entanglement measures ( entanglement of formation , negativity or relative entropy ) automatically maximize the other two .
it was further shown that the global unitary transformation that takes arbitrary states into maximal ones has the form @xmath91 where @xmath92 and @xmath93 are arbitary local unitary transformations , @xmath94 @xmath95 is a unitary diagonal matrix , and @xmath96 is the unitary matrix that diagonalizes the density matrix @xmath1 , i.e. , @xmath97 , where @xmath98 is a diagonal matrix , the diagonal elements of which are the four eigenvalues of @xmath1 listed in nonincreasing order ( @xmath99 ) .
hence , the general form of a density matrix that is maximal , given a set of eigenvalues , is ( up to local unitary transformations ) @xmath100 this matrix is locally equivalent to the form @xmath101with @xmath102 , @xmath103 , @xmath104 , and @xmath105 .
the above derivation justifies the ansatz form ( [ eqn : ansatz ] ) used in ref .
@xcite to derive the entanglement of formation vs. linear entropy mems .
we remark that one may as well use the four eigenvalues @xmath64 s as the parametrization .
nevertheless , the form ( [ eqn : rhomems ] ) , as well as ( [ eqn : ansatz ] ) , can be nicely viewed as a mixture of a bell state @xmath106 with some diagonal separable mixed state .
we begin by measuring mixedness in terms of the linear entropy , and comparing the frontier states for various measures of entanglement .
the characterization of physical states in terms of their entanglement of formation and linear entropy was introduced by munro et al . in ref .
( strictly speaking , they considered the tangle rather than the equivalent entanglement of formation . ) here , we shall consider yet another equivalent quantity : concurrence ( see sec .
[ sec : ef ] ) . in order to find the frontier , munro et al . proposed ansatz states of the form @xmath107 where @xmath108 and @xmath109 .
they found that , of these , the subset @xmath110 [ eqn : mems ] lies on the boundary in the tangle vs. linear - entropy plane and , accordingly , named these mems , in the sense that these states have maximal tangle for a given linear entropy .
we remark that at the crossing point of the two branches , @xmath111 , the density matrices on either side coincide . in fig .
[ fig : csl ] we plot the entanglement of formation / concurrence vs. linear entropy for the family of mems ( [ eqn : mems ] ) ; this gives the frontier curve .
for the sake of comparison , we also give the curve associated with the family of werner states of the form @xmath112 evidently , for a given value of linear entropy these mems ( which we shall denote by @xmath113 ) achieve the highest concurrence . as the tangle @xmath114 and entanglement of formation @xmath7 are monotonic functions of the concurrence , eq .
( [ eqn : mems ] ) also gives the boundary curve for these measures .
this raises an interesting question : is ( [ eqn : mems ] ) optimal for other measures of entanglement ? to find the frontier states for the relative entropy of entanglement we again turn our attention to the maximal density matrix ( [ eqn : rhomems ] ) . for this form of density matrix
the linear entropy is given ( with @xmath115 expressed in terms of @xmath116 ) by @xmath117 to calculate the relative entropy of entanglement , we need to determine the closest separable state to ( [ eqn : rhomems ] ) .
it is simpler to do this analysis via several cases .
we begin by considering the rank-2 and rank-3 cases of ( [ eqn : rhomems ] ) .
we set @xmath118 ( @xmath119 ) and express @xmath115 in terms of @xmath120 and @xmath121 in the density matrix , obtaining @xmath122 and thus find that the closest separable density matrix @xmath123 is given by @xcite @xmath124 the relative entropy of entanglement is now simply given by @xmath125 with the linear entropy being given by @xmath126 subject to the constraint @xmath127 . for the rank-2 case @xmath128 ( @xmath129 ) , and the resulting solution is the rank-2 matrix @xmath130 given in eq .
( [ eqn : mems ] ) with @xmath131 .
we remark that this rank-2 solution is always a candidate mems for the three entanglement measures that we consider in this paper . in order to determine whether or in what range the rank-2 solution achieves the global maximum
, we need to compare it with the rank-3 and rank-4 solutions . by maximizing @xmath132 for a given value of @xmath71
, we find the following stationary condition : @xmath133 given a value of @xmath71 , we can solve eqs .
( [ eqn : sl3 ] ) and ( [ eqn : stationaryresl ] ) , at least numerically , to obtain the parameters @xmath120 and @xmath121 , and hence , from eq .
( [ eqn : rank2n3 ] ) , the rank-3 mems . however , if the constraint inequality @xmath134 turns out to be violated , the solution is invalid .
we now turn to the rank-4 case .
it is straightforward , if tedious , to show that the werner states , eq .
( [ eqn : werner ] ) , obey the stationarity conditions appropriate for rank 4 .
however , it turns out that this solution is not maximal . to summarize , the frontier states , which we denote by @xmath135 , are states of the form ( [ eqn : rank2n3 ] ) ; the dependence of the parameters @xmath120 and @xmath121 on @xmath71 is shown in fig .
[ fig : arsl ] . in fig .
[ fig : rsl ] , we show the resulting frontier , as well as curves corresponding to non - maximal stationary states .
the frontier states have the following structure : ( i ) for @xmath136 they are the rank-2 mems of eq .
( [ eqn : mems ] ) but with @xmath121 restricted to the range from @xmath75 ( at @xmath137 ) to approximately @xmath138 ( at @xmath139 ) ; ( ii ) for @xmath140 the mems are rank 3 , with parameters @xmath120 and @xmath121 satisfying eqs .
( [ eqn : stationaryresl ] ) and ( [ eqn : sl3 ] ) at each value of @xmath71 , and @xmath141 ranging between approximately @xmath142 ( at @xmath139 ) and @xmath143 ( at @xmath144 ) .
as noted previously , beyond @xmath144 there are no entangled states . as the inset of fig . [
fig : arsl ] shows , the parameter @xmath120 can be regarded as a continous function of parameter @xmath145 $ ] .
the two branches of the solution , ( i ) and ( ii ) , cross at @xmath146 ; at this point , the states on the two branches coincide , @xmath147 just as in the case of entanglement of formation vs. linear entropy , the density matrix is continuous at the transition between branches .
we remark that the curve generated by the states @xmath113 , when plotted on the @xmath30 vs. @xmath71 plane , falls just slightly below that generated by the states @xmath135 for @xmath148 ( and coincides for smaller values of @xmath71 ) .
we also remark that the parameter @xmath121 turns out to be the concurrence @xmath149 of the states , so that fig .
[ fig : arsl ] can be interpreted a plot of the concurrence of the frontier states vs. their linear entropy . by comparing this concurrence vs. linear entropy curve to that in fig .
[ fig : csl ] , we find that the former lies just slightly below the latter for @xmath148 ( and the two coincide for smaller values of @xmath71 ) , the maximal difference between the two being less than @xmath150 .
it is evident that , for a given linear entropy , the relative entropies of entanglement for both @xmath151 and @xmath152 are significantly less than the corresponding entanglements of formation .
in fact , for small degrees of impurity , the entanglements of formation for the two mems states are quite flat ; however , the relative entropies of entanglement fall quite rapidly . more specifically
, for a change in linear entropy of @xmath153 near @xmath137 we have that @xmath154 ( see fig . [
fig : csl ] ) and @xmath155 ( see fig . [
fig : rsl ] ) . as the curves of the states @xmath113 and @xmath135 are very close on the two planes , @xmath7 vs. @xmath71 and @xmath30 vs. @xmath71 , we show , in fig .
[ fig : desl ] , the entanglement difference @xmath156 for the states @xmath113 , and compare it with the corresponding difference for the werner states .
while it is clear that @xmath157 , for certain values of the linear entropy the difference turns out to be quite large , this difference being uniformly larger for @xmath113 than for the werner state ; see fig .
[ fig : desl ] .
this raises several interesting questions , e.g. , whether , for a given value of @xmath7 , the undistillable entanglement ( the difference between the entanglement of formation and the relative entropy of entanglement is the lower bound ) increases as states become more mixed .
as we have seen , werner states are not frontier states either in the case of entanglement of formation or in the case of relative entropy of entanglement .
by contrast , as we shall see in the next section , if we measure entanglement via negativity , then for a given amount of linear entropy , the werner states ( as well as another rank-3 class of states ) achieve the largest value of entanglement .
said equivalently , the werner states belong to @xmath158 . in order to derive the form of the mems in the case of negativity
, we again consider the density matrix of the form ( [ eqn : rhomems ] ) , for which it is straightforward to show that the negativity @xmath159 is given by @xmath160 furthermore , because we aim to find the entanglement frontier , we can simply restrict our attention to states satisfying @xmath161 , i.e. , to states that are entangled @xcite .
then , by making @xmath159 stationary at fixed @xmath71 and with the constraint @xmath162 , we find two one - parameter families of stationary states ( in addition to the rank-2 mems , which are common to all three entanglement measures ) . the parameters of the first family obey @xmath163 when expressed in terms of parameter @xmath121 , the density matrix takes the form @xmath164 which are precisely the werner states in eq .
( [ eqn : werner ] ) . for the second solution ,
the parameters obey @xmath165 when expressed in terms of parameter @xmath121 , the density matrix takes the form @xmath166 we remark that the two solutions give the same bound on the negativity for a given value of linear entropy . the resulting frontier in the negativity vs. linear entropy plane is shown in fig .
[ fig : nsl ] .
we have shown by numerical exploration @xcite that the family states found in the previous paragraph do indeed achieve the maximal negativity .
the results of this exploration are indicated by points in the inset in fig .
[ fig : nsl ] , and compelling evidence for the maximality of the negativity is furnished by the fact that no points lie beyond the frontier .
thus , the states @xmath167 on the boundary include , up to local unitary transformations , both werner states in eq .
( [ eqn : mems_n_sl1 ] ) and states in eq .
( [ eqn : mems_n_sl2 ] ) .
we also plot in fig .
[ fig : nsl ] the curve belonging to @xmath113 ; note that it falls slightly below the curve associated with @xmath167 and that it has a cusp , due to the structure of the states , at the value @xmath168 for the parameter @xmath121 in eq .
( [ eqn : mems ] ) . here , we see that maximally entangled mixed states change their form when we adopt a different entanglement measure .
we continue this section by choosing to measure mixedness in terms of the von neumann entropy , and comparing the frontier states for various measures of entanglement . to find this frontier , we consider states of the form ( [ eqn : rhomems ] ) , and compute for them the concurrence and the von neumann entropy : @xmath169 [ eqn : cnsv ] note that the parameters obey the normalization constraint @xmath170 .
as we remarked previously , the rank 2 mems is always a candidate . for the rank-3 case , we can set @xmath118 in eq .
( [ eqn : cnsv ] ) . by maximizing @xmath149 at fixed @xmath67
, we find a stationary solution : + ( i ) @xmath171 , @xmath172 , and @xmath173 ; the resulting density matrix is @xmath174for the rank-4 case ( @xmath175 ) , the stationarity condition can be shown to be @xmath176 where @xmath177 , @xmath178 , and @xmath179 .
there are two solutions , due to the two - to - one property of the function @xmath180 for @xmath181 .
the first one is ( @xmath182 ) .
+ ( ii ) @xmath183 , and @xmath184 , which can readily be seen to be a werner state as in eq .
( [ eqn : werner ] ) or , equivalently , @xmath185being the concurrence , @xmath149 is restricted to the interval @xmath186 $ ] .
the second solution is transcendental , but can be solved numerically . in fig .
[ fig : csv ] we compare the four possible candidate solutions , and find that the global maximum is composed of only ( i ) and ( ii ) .
we summarize the states at the frontier as follows : @xmath187 note the crossing point at @xmath188 , at which extremality is exchanged , so the true frontier consists of two branches .
it is readily seen that @xmath189 is the solution of the equation @xmath190 , and the approximate numerical values of @xmath189 and the corresponding @xmath191 are @xmath192 and @xmath193 , respectively .
the resulting form of mems states is peculiar , in that , even at the crossing point of two branches on the entanglement - mixedness plane , the forms of matrices on the two branches are not equivalent ( one is rank 3 , the other rank 4 ) .
this is in contrast to the @xmath113 .
this peculiarity can be partially understood from the plot of the two mixedness measures , fig .
[ fig : svsl ] : as the value of the von neumann entropy rises , there are fewer and fewer rank-3 entangled states , and above some threshold , no more rank-3 states exist , let alone entangled rank-3 states .
there are , however , still entangled states of rank 4 .
hence , if rank-3 states attain higher entanglement than rank-4 states do when the entropy is low , a transition must occur between mems states of rank 3 and rank 4 . from fig .
[ fig : csv ] it is evident that beyond a certain value of the von neumann entropy no entangled states exist .
this value can be readily obtained by considering the mems state ( [ eqn : rhoii ] ) at @xmath194 , @xmath195for which @xmath196 . as an aside
, we mention a tantalizing but not yet fully developed analogy with thermodynamics @xcite . in this analogy ,
one associates entanglement with energy and von neumann entropy with entropy , and it is therefore tempting to regard the mems just derived as the analog of thermodynamic equilibrium states .
if we apply the jaynes principle to an ensemble in equilibrium with a given amount of entanglement then the most probable states are those mems shown above .
let us now find the frontier states for the case of relative entropy of entanglement . to do this
, we first consider the rank-3 states in eq .
( [ eqn : rank2n3 ] ) , for which the relative entropy is given by eq .
( [ eqn : er3 ] ) . for these states ,
the von neumann entropy is given by @xmath197 where the @xmath65 function is taken to be base four . even though the @xmath65 functions in @xmath67 and @xmath30 use different bases , the stationary condition for the parameters @xmath121 and @xmath120 does not change , because the difference can be absorbed by a rescaling of the constraint - enforcing lagrange multiplier .
thus , in maximizing @xmath30 at fixed @xmath67 , we arrive at the stationarity condition @xmath198 we can solve for the parameter @xmath120 as a function of @xmath199 $ ] , at least numerically ; the result is shown in fig .
[ fig : asverofr ] , along with @xmath67 and @xmath30 . turning to the rank-4 case , it is straightforward , if tedious , to show that the werner states satisfy the corresponding stationarity conditions . in order to ascertain
which rank gives the mems for a given @xmath67 , we compare the stationary states of rank 2 , 3 , and 4 in fig .
[ fig : resv ] .
thus , we see that for @xmath200 the frontier states are given by the rank-3 states , whereas for @xmath201 the frontier states are given by the werner states ( [ eqn : werner ] ) with the parameter @xmath121 ranging from approximately 0.6059 down to 0 . at the crossing point , @xmath202 , the mems undergo a discontinuous transition ; recall that we encountered a similar phenomenon in the case of entanglement of formation vs. von neumann entropy .
we saw in sec .
[ sec : nsl ] that there is a pair of families of mems which differ in rank but give the identical frontier in the @xmath159 vs. @xmath71 plane .
it is interesting to see what happens for the combination of negativity and von neumann entropy .
once again , we begin with states of form ( [ eqn : rhomems ] ) , for which the negativity and the von neumann entropy are given in eqs .
( [ eqn : negativity ] ) and ( [ eqn : s_v ] ) , respectively . by making @xmath159 stationary at fixed @xmath67
, we are able to find only one solution ( in addition to the rank-2 candidate ) : @xmath203 . expressing the resulting density matrix , as we may , in terms of the single parameter @xmath121
, we arrive at the following candidate for the frontier states : @xmath204 where @xmath205 , i.e. , the werner states . the resulting frontier in the negativity vs. von - neumann - entropy plane is shown in fig .
[ fig : nsv ] which , for comparison , also shows the curve for the rank-2 candidate .
in this paper we have determined families of maximally entangled mixed states ( mems , i.e. , frontier states , which possess the maximum amount of entanglement for a given degree of mixedness ) . these states may be useful in quantum information processing in the presence of noise , as they have the maximum amount of entanglement possible for a given mixedness .
we considered various measures of entanglement ( entanglement of formation , relative entropy , and negativity ) and mixedness ( linear entropy and von neumann entropy ) .
we found that the form of the mems depends heavily on the measures used .
certain classes of frontier states ( such as those arising with either entanglement of formation or relative entropy of entanglement vs. the von neumann entropy ) behave discontinously at a specific point on the entanglement - mixedness frontier . under most of the settings considered ,
we have been able to explicitly derive analytical forms for the frontier states . for entanglement of formation and relative entropyand for most values of mixedness
, we have found that the rank-2 and rank-3 mems have more entanglement than werner states do . on the other hand , at fixed entropy no states have higher negativity than werner states do . at small amounts of mixedness , the @xmath113 states lose entanglement with increasing mixedness at a substantially lower rate than do the werner states .
however , when the entanglement is measured by the relative entropy , the difference in loss - rate is significantly smaller . from eq .
( [ eqn : lsdecomp ] ) it is tempting to assert that in the case of ls entanglement vs. mixedness the frontier states should be the same as those in the case of entanglement of formation vs. mixedness .
however , as we do not know whether eq .
( [ eqn : ansatz ] ) ( up to local unitary transformations ) exhausts all maximal states for ls entanglement , further investigation of this point is needed .
having characterized the mems for various measures , it is worthwhile considering them from the perspective of bell - inequality violations . to quantify the violation of bell s inequality
, it is useful to consider the quantity @xmath206 where @xmath207 and the vectors @xmath208 and @xmath209 ( @xmath210 and @xmath211 ) are two different measuring apparatus settings for observer a ( observer b ) . if @xmath212 then the corresponding state violates bell s inequality . for the density matrix of the form ( [ eqn : rhomems ] ) it is straightforward @xcite to show that the quantity @xmath88 is given by @xmath213 in fig . [ fig : bell ] , we plot @xmath88 vs. linear and von neumann entropies for several families of frontier states . as a comparison
, we also draw the violation by the following rank-2 state ( which is diagonal in the bell basis ) : @xmath214.\end{aligned}\ ] ] this state , although not belonging to any of the families of frontier states derived previously , turns out to achieve the maximum possible violation , as a function of linear entropy . on the other hand ,
the werner states appear to achieve maximal violation in the case of von neumann entropy .
ekert s application of bell s inequalities to quantum cryptography @xcite , together with the discussions of the present paragraph , suggests that mems may be relevant to quantum communication .
another natural application for which entanglement is known to be a critical resource is quantum teleportation .
how do these frontier mems teleport , compared with the werner and rank-2 bell diagonal states ?
if we restrict our attention to high purity situations ( i.e. , to states with only a small amount of mixedness ) then it is straightforward to show that , e.g. , @xmath113 states teleport average states better than the werner states do , but worse than the rank-2 bell diagonal state does .
part of the explanation for this behavior is that standard teleportation is optimized for using bell states as its core resource .
it is also interesting to note that for certain combinations of entanglement and mixedness measures , as well as the bell inequality violation , the rank-2 candidates fail to furnish mems .
thus , these states seem to be less useful than other mems .
however , from the perspective of distillation , these states are exactly quasi - distillable @xcite , and can be useful in the presence of noise because they can be easily distilled into bell states .
this material is based on work supported by nsf eia01 - 21568 ( tcw , pmg and pgk ) and by the u.s .
department of energy , division of material sciences , under award no .
defg02 - 91er45439 , through the frederick seitz materials research laboratory at the university of illinois at urbana - champaign ( tcw and pmg ) .
kn and wjm acknowledge financial support from the european projects quiprocone and equip .
pmg gratefully acknowledges the hospitality of the unversity of colorado at boulder , where part of this work was carried out .
tcw gratefully acknowledges the receipt of a mavis memorial fund scholarship .
in this appendix we address the result that for @xmath39 systems the partial transpose of any density matrix @xmath1 has at most one negative eigenvalue .
in fact , we shall consider the result from two perspectives .
first , we build upon results ( theorem 3 , in particular ) contained in ref .
@xcite , from which it follows that it is sufficient to consider ( i ) bell diagonal states and ( ii ) states of the form @xmath215 for the latter case , straightforward calculation shows that the partially transposed matrix can have one negative eigenvalue when @xmath216 and that it does not have negative eigenvalues when @xmath217 . for the former case , suppose that the bell diagonal state has the four eigenvalues @xmath218 , @xmath219 , @xmath220 , @xmath221 , in nonincreasing order .
then it is straightforward to see that the corresponding partially transposed matrix has the four eigenvalues @xmath222 , @xmath223 , @xmath224 , @xmath225 .
thus , it can have at most one negative eigenvalue . a second perspective is provided by first invoking the ls decomposition ( [ eqn : lsdec ] ) and then making a schmidt decomposition , via the local unitary transformation @xmath226 , of the pure entangled part @xmath227 . in this way ,
the pure part becomes @xmath228 where @xmath229 .
meanwhile , @xmath1 is transformed into @xmath230 where @xmath231 is still separable , and its partial transpose remains positive semi - definite .
taking , then , the partial transpose , the density matrix is transformed into @xmath232 and we note that the last matrix has eigenvalues @xmath233 , @xmath234 , @xmath235 , @xmath236 in ( as @xmath237 ) nonincreasing order . as the hermitian matrix @xmath238 retains positive - semi - definiteness , we can employ a well - known result in matrix analysis @xcite that for @xmath239 hermitian matrices @xmath87 and @xmath88 , with @xmath88 being positive semi - definite , the eigenvalues of @xmath240 and @xmath87 , when arranged in non - ascending order , obey @xmath241 hence , identifying @xmath88 with @xmath242 and @xmath87 with the product of @xmath243 and the matrix in eq .
( [ eqn : rhotb ] ) , we immediately see that @xmath244 and , thus , @xmath245 ( or equivalently @xmath246 ) can have at most one negative eigenvalue .
thus , the negativity for @xmath39 systems can then be written as @xmath247 .
99 b. schumacher , phys .
a * 54 * , 2614 ( 1996 ) .
a. k. ekert , phys .
lett . * 67 * , 661 ( 1991 ) . c. h. bennett , g. brassard , c. crpeau , r. jozsa , a. peres , and w. k. wootters , phys .
. lett . * 70 * , 1895 ( 1993 ) .
s. lloyd , science * 261 * 1589 ( 1993 ) ; d. p. divincenzo , science * 270 * , 255 ( 1995 ) ; m. nielsen and i. chuang , _ quantum computation and quantum information _
( cambridge university press , cambridge , 2000 ) . c. h. bennett , d. p. divincenzo , j. a. smolin , and w. k. wootters , phys
* a54 * , 3824 ( 1996 ) .
w. k. wootters , phys .
lett . * 80 * , 2245 ( 1998 ) . c. h. bennett , g. brassard , s. popescu , b. schumacher , j. smolin , and w. k. wootters , phys .
lett . * 76 * , 722 ( 1996 ) .
v. vedral , m. b. plenio , k. jacobs , and p. l. knight , phys .
a * 56 * , 4452 ( 1997 ) .
k. yczkowski , p. horodecki , a. sanpera , and m. lewenstein , phys .
a * 58 * , 883 ( 1998 ) . g. vidal and r. f. werner , phys .
a * 65 * , 032314 ( 2002 ) . by entropy
we mean a measure of how mixed a state is ( i.e. , its mixedness ) .
we shall focus on two measures : the linear entropy and the von neumann entropy , defined in sec .
[ sec : measofmix ] .
w. j. munro , d. f. v. james , a. g. white , and p. g. kwiat , phys .
a * 64 * , 030302 ( 2001 ) .
s. ishizaka and t. hiroshima , phys .
a * 62 * , 022310 ( 2000 ) .
f. verstraete , k. audenaert , and b. de moor , phys .
a * 64 * , 012316 ( 2001 ) .
j. eisert and m. b. plenio , j. mod
. opt . * 46 * , 145 ( 1999 ) .
f. verstraete , k. audenaert , j. dehaene , b. de moor , j. phys .
a * 34 * , 10327 ( 2001 ) .
s. virmani and m. b. plenio , phys .
a * 268 * , 31 ( 2000 ) .
v. vedral , m. b. plenio , m. a. rippin , and p. l. knight , phys .
lett . * 78 * , 2275 ( 1997 ) ; v. vedral and m. b. plenio , phys .
a*57 * , 1619 ( 1998 ) . m. horodecki , p. horodecki , and r. horodecki , phys .
lett . * 84 * , 2014 ( 2000 ) . a separable ( or unentangled ) state ( bi - partite ) @xmath248 can be expressed as @xmath249 , whereas an entangled state has no such decomposition .
a. peres , phys .
77 * , 1413 ( 1996 ) ; m. horodecki , p. horodecki , and r. horodecki , phys .
a * 223 * , 1 ( 1996 ) .
suppose a bi - partite density matrix @xmath1 is expressed in the following form : @xmath250 .
then the partial transpose @xmath38 of the density matrix @xmath1 is defined via @xmath251 we remark that the partial transpose depends on the basis chosen but the eigenvalues of the partial transposed matrix do not . m. lewenstein and a. sanpera , phys .
80 * , 2261 ( 1998 ) .
s. karnas and m. lewenstein , j. phys .
a : math . gen . * 34 * , 6919 ( 2001 ) .
j. eisert and h. j. briegel , phys . rev .
a * 64 * , 022306 ( 2001 ) .
t. wellens and m. ku , phys .
a * 64 * , 052302 ( 2001 ) . by numerical exploration
we mean the creation of a large set of random density matrices , for each of which we compute the relevant characteristics , in this case negativity and linear entropy . m. horodecki , r. horodecki , and p. horodecki , acta phys .
slov . * 48 * , 133 ( 1998 ) .
p. k. aravind , phys .
a * 200 * , 345 ( 1995 ) .
m. horodecki , p. horodecki , and r. horodecki , phys .
a * 60 * , 1888 ( 1999 ) .
f. verstraete , j. dehaene , and bart demoor , phys .
a * 64 * , 010101 ( 2001 ) .
r. horn and c. johnson , _ matrix analysis _
( cambridge university press , cambridge , 1985 ) .
|
maximally entangled mixed states are those states that , for a given mixedness , achieve the greatest possible entanglement . for two - qubit systems and for various combinations of entanglement and mixedness
measures , the form of the corresponding maximally entangled mixed states is determined primarily analytically . as measures of entanglement , we consider entanglement of formation , relative entropy of entanglement , and negativity ; as measures of mixedness , we consider linear and von neumann entropies .
we show that the forms of the maximally entangled mixed states can vary with the combination of ( entanglement and mixedness ) measures chosen .
moreover , for certain combinations , the forms of the maximally entangled mixed states can change discontinuously at a specific value of the entropy .
|
pulsars are effective astrophysical laboratories for quantum theory and gravitation theory . however , how to reproduce the observed radiation theoretically is still one of the most essential challenges in pulsar study .
it is well known that the polarization observations are very important to provide much information about pulsar physics , but there are still many troubles in explaining the polarization data .
one of the difficulties in understanding pulsar polarization observations is the polarization position angle jumps in mean ( or integrated ) pulses as well as in individual pulses [ 1 - 3 ] . _ for mean pulses _ , it is generally found that position angles would have discontinuities about 90@xmath0 at some longitudes where the linear polarization intensities are near zero ( totally depolarized ) . _ for individual pulses
_ , the position angles would be dispersed or have two @xmath1 separated distributions at some observational longitude bins where the linear polarization percentages are remarkably small . a famous example to display the polarization position angle jumps in individual pulses
is shown in fig.1 for psr b2020 + 28[1 ] . in case ` a ' and ` c ' , the two position angle distributions are clear , and the linear polarization percentages are obviously low . in case
` b ' , there is only one position angle distribution , and the emission in each longitude point is highly linearly polarized .
it means that such orthogonally distributed position angles are usually observed only in the part of the profiles where the linear polarization is low [ 4 ] .
a conclusion from the observation is that , _ both for mean profiles and for individual pulses , the position angle jumps are related to low linear polarization ( percentage ) at all time_. based on above observational facts , there should be two possibilities logically : one is that ` position angle jump ' causes ` low linear polarization ' , another is that ` low linear polarization ' causes ` position angle jump ' .
many authors believe , without justification , that the position angle jumps should be attribute to the appearance of opm [ 1 - 3 ] , in - coherent superposition of the opm is the origin of depolarization . however , why the another possibility is impossible ?
we investigate this possibility in this paper . previously , stinebring et al .
[ 1 ] concluded that most of pulsar emission occurs in one orthogonal mode or the other , which is called orthogonal polarization modes ( opm ) . at
a given longitude , the plane of polarization can be of two perpendicular or nearly perpendicular states . which one can operate was governed by some variability as yet not understood .
the opm can explain many things : the sharp jumps of position angles , the depolarization of linear as well as circular polarization due to the existence of both modes at the same time [ 2 ] .
however , there is no acceptable theory to reproduce such orthogonal modes , which is called as ` opm problem ' . for mean pulses
, it is suggested that the depolarization and position angle jumps might be attribute to the relative longitude shifts of pulsar beams [ 5,6 ] .
such kind of longitude shifts of pulsar beams is natural in the inverse compton scattering ( ics ) model [ 7 ] .
for individual pulses , many authors believe that there are two orthogonal modes at a given longitude .
nevertheless , there may be in fact two possibilities to produce such orthogonal modes .
one is that the emission for a given frequency is emitted at different heights , and another is that there is an unknown emission mechanism to produce orthogonal modes at a same emission point [ 8 ] .
the first possibility has been studied already [ 6 , 7 ] .
but for the second one , no acceptable theory has been found to produce such orthogonal modes hitherto known .
thus , the ` opm problem ' still confuses the pulsar world .
we ask a question here : are really such ` orthogonal modes ' the reason for the ` low linear polarization ' ?
otherwise , might the ` low linear polarization ' be responsible for the observed ` orthogonal modes ' ?
many authors believe that the reduction in the percentage of linear polarization is caused by in - coherent superposition of the opm .
contrary to the above idea , another possibility ( i.e. , the ` low linear polarization ' is the reason for producing the observed ` position angle jumps ' ) is suggested in this paper . _
our analysis and simulations show that , when the linear polarization percentages are low enough , the position angles would be distributed in two areas separated nearly ninety degrees_. a suggestion to check this idea is presented .
we show how ` low linear polarization ' causes so - called ` position angle jumps ' in mean - pulses and in single - pulses in section 2 and 3 , respectively .
some troubles faced by opm radiation mechanism are summed up in section 4 .
finally , conclusion and discussion are given in section 5 .
almost certainly , for observed mean - pulses , the smoothly changing position angle curves will suddenly jump at some longitudes where the linear polarization is highly depolarized .
these facts can be understood under the properties of stokes parameters .
it could be verified mathematically that the position angle would jump 90@xmath0 when the line of sight travels across a _ singular point _ [ 5 , 6 ] where the linear polarization intensity is zero .
the four stokes parameters @xmath2 , from which one can obtain linear polarization intensity @xmath3 and position angle @xmath4 ( see equ.(1 ) below ) , are functions of observational longitude @xmath5 . for the sake of simplicity ,
we let @xmath6 , as the linear polarization is focused on here . at a singular point ( @xmath7 )
, @xmath8 means @xmath9 expanding @xmath10 and @xmath11 near singular point , we come to @xmath12 @xmath13 assuming @xmath14 and @xmath15 are the lowest non - zero power terms of @xmath10 and @xmath11 , respectively , and @xmath16 $ ] , one could find that @xmath17 should be @xmath18 as long as @xmath19 is an odd number ) . ] .
it is very possible that @xmath20 , thus , position angle _ naturally _ jumps @xmath21 if @xmath8 [ 6 ] .
therefore , _ the reason that position angle jumps in integrated profiles might be why the beamed radiation is depolarized_. depolarization should be the cause of position angle jumps in mean - pulses .
there are many ways to cause depolarization .
first of all , depolarization may have an intrinsic origin .
as emission beams are formed in different heights , and each of them has different position angle , depolarization must take place by incoherent superposition of such emission beams .
in the ics model [ 7 ] , different emission beams are formed in different heights , hence , the retardation and aberration effects could make the apparent emission beams be superposed incoherently [ 5,6 ] .
secondly , depolarization might be originated from propagation process , such as the scattering by interstellar medium or magnetospheric plasma [ 4 ] , and the propagating properties of different radiation modes in plasma .
the third way might be the result of observational effect .
since the stokes parameters are added from many frequency channels after de - dispersion , the emissions in each frequency channel are incoherent superposed .
such kind of treatment in observation should also depolarize the original radiation .
there are many factors to reduce the precision of observational results , such as noises from the observational system and the sky background . usually , we put thresholds for the total intensity ( @xmath22 ) and the linear polarization intensity ( @xmath23 ) in each longitude bin in order to exclude fake polarization due to the observational error .
for example , we select observational data whose @xmath22 and @xmath23 are greater than 5 to 10 times of off - pulse rms . however , as will be discussed in this section , some fake polarization data , which may responsible for the observed ` position angle jumps ' in individual pulses [ 5 ] , do survive from such selection .
some observational uncertainties can cause the observed position angles to ` jump ' in individual pulses , such as the ` error transference ' ( section 3.1 ) , the unequal rms of stokes parameter @xmath10 and @xmath11 ( section 3.2 ) , and the fake linear polarization ( section 3.3 ) .
all these uncertainties can bring wrong polarization information . if @xmath24 is a random number , then the function @xmath25 is also random .
the random distribution of @xmath26 is known as long as the distribution of @xmath24 is given .
for example , let the distribution function of @xmath24 as a gaussian distribution , with the expectation value @xmath27 .
the distribution function of @xmath26 depends on the function @xmath28 .
if @xmath28 is a monotonous function near @xmath27 , the distribution function of @xmath26 is approximately a gaussian . whereas , if @xmath28 is a very complex function near @xmath27 , the distribution function of @xmath26 is also complicated .
it is known that the position angle @xmath4 is a function of the stokes parameters @xmath10 and @xmath11 , depending on the signs of @xmath10 and @xmath11 . from @xmath29
, one can get a general expression for @xmath4 , where the value region of @xmath4 is from @xmath30 to @xmath31 . ]
@xmath32 , \eqno(1)\ ] ] which is ` singular'(unusual ) near @xmath33 . here
@xmath34 ( @xmath35 ) if @xmath36 ( @xmath37 ) .
this property of singularity of @xmath4 would cause two error distribution peaks ( see fig.2 ) , which will be discussed later in section 3.2 . in a word ,
the observational uncertainty ( error ) of @xmath10 and @xmath11 could bring an error distribution of @xmath4 in two regions with 90@xmath0 separation by the error transference effect .
such observational results might be mistakenly considered as position angle ` jumps ' in a real beamed radiation . usually , the rms of the stokes parameter @xmath10 and that of @xmath11 are not equal for an astronomical polarimetry , i.e. , @xmath38 .
the difference between @xmath39 and @xmath40 can be as large as several percentages in observations .
the observed linear polarization position angles could ` jump ' during different observing time as long as @xmath38 . as demonstrated in fig.3
, we see that @xmath41 is about @xmath30 , and @xmath42 is about @xmath21 .
the reason that @xmath43 might be diversity .
for example , for some polarimetry , the stokes parameters @xmath44 and @xmath45 are computed from the hybrid networks out - put signals . here
@xmath46 and @xmath47 are the observed intensity from two orthogonal dipole antenna , @xmath48 and @xmath49 are the intensity received from a system which has been rotated 45 degrees . for a dipole antenna , @xmath46 and
@xmath47 are obtained directly .
however , the @xmath48 and @xmath49 are yielded through a turnstile junction where the phase mis - alignment can make the rms of @xmath48 ( and @xmath49 ) larger than the rms of @xmath46 ( and @xmath47 ) .
so , the rms of @xmath10 and that of @xmath11 can not be equal because of the imperfection of the turnstile junction .
simulations of this kind of polarimetry are given in fig .
4 ( see the appendix ) , from which we see that * the observed percentages of polarization @xmath50 are much greater than the true percentage @xmath51 , if @xmath51 is small enough . observed linear polarization may be larger than that of the true value . *
the position angle ` jumps ' takes place when linear polarization percentage @xmath52 .
when @xmath53 , there is few possibilities to make position angle jump .
because the observational uncertainty is of random , position angle jumps that come of this kind of errors discussed above can be avoided by more time observation .
it is almost impossible that position angle jump due to observational uncertainty appears in integrated profiles . for a telescope with an effective detection area @xmath54 , a frequency bandwidth @xmath55 , a time constant @xmath56 , and a systematical noise temperature @xmath57 , then the systematical noise flux @xmath58 is @xmath59 , where @xmath60 is the boltzmann s constant , and
the off - pulse rms @xmath61 is @xmath62 whereas there is a signal with intensity flux @xmath63 , the rms of the signal flux is @xmath64 ( the on - pulse rms ) @xmath65 where @xmath63 can be one of the out - put signals from the hybrid networks , such as @xmath46 , @xmath47 , @xmath48 , @xmath49 , @xmath66 , @xmath67 , which are correspondent to the linearly polarized components of the input signals at position angle @xmath30 , @xmath21 , @xmath68 , and @xmath69 , and to the right - hand and left - hand circularly polarized components . because @xmath70 , the rms of @xmath63 is @xmath71 .
if we assume that @xmath58 is accurate enough ( i.e. , we have enough time to measure @xmath58 ) , then @xmath72 , which will be used in the following discussion .
if we let @xmath73 square meters , @xmath74mhz , @xmath75ms , and @xmath76k , then @xmath61 is about 0.2 jy . for @xmath77jy[9
] , then @xmath78 which is more than three times that of @xmath61 , and the fake linear polarization percentage could be as large as @xmath79 .
thus , some data , whose linear polarization is originated from such uncertainty , _ can also exceed _ the observational threshold level with an un - negligible possibility .
we might be in a dilemma if there are really so - called orthogonal polarization modes in pulsars beams .
first , how does the opm radiative mechanism produce ?
no reasonable theory has been appeared in literature .
furthermore , the two orthogonal modes should be in - coherent , which makes the opm more difficult to be set up .
the suggestion for observed ` orthogonal modes ' in mean - pulses by xu at al .
[ 6 ] is not a real one .
( in their calculations , two components are emitted in - coherently from different regions . ) furthermore , if the two modes are coherent , the total radiation is elliptically polarized , thus no position angle ` jump ' appears .
secondly , why havent we seen that single - pulses are highly polarized , but the position angle distribution is till separated by 90@xmath0 ( like fig.5 ) ?
individual pulses , which could be highly polarized , are generally conjectured to be from single radiation elements .
since two orthogonal modes are incoherent , a radiation element might emit only one of the opr modes at one time .
therefore , it is possible to observe some highly linearly polarized individual pulses , while the position angles of which are 90@xmath0 separated .
unfortunately , observation result similar to that of fig.5 has _ never _ been found .
thirdly , how to explain the non - orthogonal separation of position angles in the regime of opm ?
non - orthogonal emission modes have proverbially been found in observation [ 1 ] .
these facts are rigorous for anyone to theorize opm models .
from analysis above some conclusions and discussions are reached : \1 . another possible way to solve the problem of position angle ` _ jumps _ ' in pulsars beamed radio emission was proposed .
there might be _ no real _ ` orthogonal polarization modes ' in the emission at all .
position angle jumps due to the observational uncertainties could appear in observed individual pulses when the linear polarization _
percentages _ are small ( not only the linear polarization intensity to be small ) .
at least part of the observed position angle jumps in individual pulses and mean - pulses can be explained by depolarization and observational uncertainty . for a real pulsar , we must put together these two possible factors to investigate the position angle variation in the individual pulses as well as in the integrated pulses .
for example , observational uncertainty might be the main reason of position angle separation near point ` a ' in fig.1 . nevertheless , near point ` c ' , orthogonal and non - orthogonal separations are clear , which might be the result of the relative longitude shifts of pulsar beams [ 6 ] and the observational uncertainties .
rathnsree & rankin[10 ] pointed out that , for psr b1929 + 10 , lower degree of polarization is seen simultaneously with the presence of ` orthogonal ' modes whereas the polarized power is not seen to be highly correlated with the position angle flip .
also , they have got dynamic pictures of the orthogonal polarization mode changes for psr b2110 + 27 at 430 mhz , and they found the transition from the dominant mode to the other orthogonal one and back are _
rapid_. most of the transition is achieved over time scales of a individual period , and the change of modes does not seem to be any periodicity in time evolution ( like a _ stochastic _ process ) .
all this observational facts have the properties of observational uncertainties discussed in section 3 . for psr b0525 + 21[1 ] at 1404 mhz ,
the position angle sweep is s - shaped in the averaged profile .
the polarization position angles in individual pulses are also an ` s ' shape distribution , but two weak patches of ` orthogonal modes ' on the outside edges of the profile where more individual pulses have very small percentages of linear polarization .
these two patches appearance should come of the observational uncertainty according to our analysis .
this statement _ can be checked _ by future expriment of observation .
\3 . from simulations
, we see that the jumped position angles are distributed near @xmath68 and @xmath69 .
in fact , there are observational data which does show that the position angles distribute near @xmath68 and @xmath69 in the scatter plots , such as fig.26 , fig.37 in stinebring et al .
if the rms of @xmath11 less than that of @xmath10 , the jumped position angles should be near @xmath30 and @xmath21 , like psr 0525 + 21 ( fig.2 in [ 1 ] ) in observations .
there are observational data where the jumped position angles are not distributed near @xmath30 , @xmath68 , @xmath21 , and @xmath69 , which could be intrinsic in polarimetry or resultant from the longitude shift of beam phases [ 6 ] .
the idea suggested in this paper can be checked experimentally .
we can input the polarimetry a simulated lower polarized and pulsed signal to see if two @xmath21 separated position angle distribution can appear in the output . if such distribution can also be obtained , the opm in pulsar emission should be doubted .
for the kind of polarimetry discussed in section 3 , let s study a partially polarized wave , with the total intensity @xmath22 , the un - polarized intensity @xmath80 , the percentage of linear polarization @xmath81 , the percentage of circular polarization @xmath82 , and the position angle @xmath4 .
if we measure this wave by a telescope with an effective area @xmath54 , systematic noise temperature @xmath83 , a bandwidth @xmath84 , a time constant @xmath56 , the angle between two the dipole antenna ( parasitism polarization ) @xmath85 ( whose expectation value is @xmath86 , @xmath87 , the rms of @xmath88 is @xmath89 ) , and the phase misalignment @xmath90 ( the rms of @xmath90 is @xmath91 ) , then the six intensity for the stokes parameters can be deduced as @xmath92 + { 1\over 2}i^{unp},\\ s_{135 } & = & { 1\over 4 } [ x^2 + y^2\cos^2\delta\alpha + \\ & & x^2\sin^2\delta\alpha + \\ & & 2xy\cos(\delta-\delta\phi-\pi)\cos\delta\alpha+\\ & & 2xy\sin\delta\alpha\cos\delta\alpha\cos\delta + \\ & & 2x^2\sin\delta\alpha\cos(\delta\phi+\pi ) ] + { 1\over 2}i^{unp},\\ s_r & = & { 1\over 4 } [ x^2 + y^2\cos^2\delta\alpha + \\ & & x^2\sin^2\delta\alpha + \\ & & 2xy\cos(\delta+\delta\phi+{3\over 2}\pi)\cos\delta\alpha+\\ & & 2xy\sin\delta\alpha\cos\delta\alpha\cos\delta + \\ & & 2x^2\sin\delta\alpha\cos(\delta\phi+{3\over 2}\pi)]+{1\over 2}i^{unp},\\ s_l & = & { 1\over 4 } [ x^2 + y^2\cos^2\delta\alpha + x^2\sin^2\delta\alpha + \\ & & 2xy\cos(\delta+\delta\phi+{\pi\over 2})\cos\delta\alpha+\\ & & 2xy\sin\delta\alpha\cos\delta\alpha\cos\delta + \\ & & 2x^2\sin\delta\alpha\cos(\delta\phi+{\pi\over 2})]+{1\over 2}i^{unp } , \end{array}\ ] ] here , @xmath93 the observed stokes parameters should be @xmath94 so that , the observed linear polarization intensity @xmath95 , the observed percentages of linear polarization @xmath96 , the observed percentages of circular polarization @xmath97 , and the observed linear polarization position angle @xmath98 would be @xmath99 .
\end{array}\ ] ] considering this kind of observational uncertainty , we have obtained some simulation results to show the position angle ` jumps ' in individual pulse observations .
one of the simulations is shown in fig.4 , where we have chosen @xmath100 the scatter plots in fig.4 are resemble to observations , especially the @xmath101 plot , which is similar to the observed linear polarization versus position angle scatter plots for position angle jumps at a fixed longitude .
based on this simulation and other simulations for different parameters , we found that the position angles ` jump ' if @xmath52 , whereas , there is few possibilities of position angle jump if @xmath53 .
[ 5 ] xu , r.x . ,
qiao , g.j .
, two possibilities of observed position angle jumps in pulsar radio emission , _ in _ : cheng k.s .
( ed ) , proc . of the 21st chinese astronomy conf .
, world scientific , 1997 , 197@xmath102300
|
the orthogonal polarization modes ( opm ) have been reported observationally ( see e.g. [ 1 ] ) and accepted widely by pulsar researchers ( see e.g. [ 2,3 ] )
. however , no acceptable theory can show the origin of the opm , which becomes a mystery in pulsar research field . here
a possible way to solve this mystery is presented .
we ask a question : does there exist any real so - called opm in pulsar radiation ? it is proposed in this paper that the ` observed opm ' in individual pulses could be the results of depolarization of pulsar radiation and the observational uncertainties originated from polarimeter in observation . a possible way to check this idea
is suggested .
if the idea is verified , the pulsar research would be influenced significantly in theory and in observation .
|
dwarf galaxies play a key role in the enrichment of the icm or igm , not only as far as heavy elements are concerned , but possibly also regarding the magnetisation .
according to the standard bottom - up scenario of galaxy formation , primeval galaxies must have injected much of their ( enriched ) interstellar medium ( ism ) into the igm during the initial bursts of star formation , thereby polluting large volumes of intergalactic space because of their high number density . by the same token ,
the following two properties render dwarf galaxies potentially very efficient in injecting a relativistic plasma into their surroundings : first , they exist in large numbers , and second , they possess low escape velocities , making it easier to expel their interstellar gas , as compared to massive spiral galaxies .
such galactic winds are in fact seen in some prototypical low - mass galaxies in the local universe .
galaxy clusters are known to be pervaded by a relativistic plasma , i.e. particles with ( mildly ) relativistic energy and magnetic fields . in the centres of some rich clusters ,
radio halos have been found , produced by either primary or secondary electrons ( see fig .
[ fig1 ] , left ) .
the problem with primary electrons is their relatively short lifetime , limited by synchrotron and inverse - compton losses : @xmath0^{-1/2 } \,\ , \rm yr \;,\ ] ] where @xmath1 is the equivalent magnetic field strength of the cosmic microwave background .
this latter equation results from the similar dependences of synchrotron and inverse - compton losses , @xmath2 @xmath3 where @xmath4 is the magnetic energy density @xmath5 , and @xmath6 is the energy density of the radiation field ( which for the perfect black - body cmb radiation is @xmath7 ) .
( [ eqn1 ] ) then tells us that , for instance , with a magnetic field of @xmath8 g , particles radiating at 1.4 ghz will be rendered invisible after @xmath9 yr in the local universe ( @xmath10 ) .
hence , primary electrons require continuous injection .
however , diffuse radio halos are never associated with any obvious `` fresh '' sources , one therefore has to invoke secondary electrons .
these are produced by hadronic collisions of relativistic protons with the thermal gas in a pion - muon chain .
owing to their much larger mass , protons have synchrotron lifetimes exceeding a hubble time . finally , it should pointed out that completely independent evidence for the magnetisation of ( at least the ) central regions of galaxy clusters comes from the observation of faraday rotation ( e.g. @xcite ) .
clusters undergoing large - scale merging frequently exhibit so - called radio relics , mostly on their periphery ( fig .
[ fig1 ] , right ) .
in contrast to radio halos , these are significantly polarised , probably reflecting magnetic - field enhancement in the compression zones where the subclusters produce large - scale shockwaves during their mutual penetration .
naturally , this also provides an efficient acceleration mechanism .
particles that were formerly highly relativistic ( electrons in particular ) have cooled via synchrotron and inverse - compton losses . during a cluster merger , they are boosted to high energies again , and the acceleration regions shine up in the radio regime .
kronberg et al .
( @xcite ) were the first to raise the question whether low - mass galaxies could have made a significant contribution to the magnetisation of the igm ( apart from more massive starburst galaxies and agn ) .
owing to their large number ( observed and predicted in a cdm cosmology ) and their injection of relativistic particles as described above , they could have played a cardinal role in the context of this cosmologically important scenario .
if true , it is to be expected that dwarf galaxies are `` wrapped '' in large envelopes of previously highly relativistic particles - and magnetic fields , which are pushed out of them during epochs of vigorous star formation . @xcite
have discussed this more quantitatively and made predictions for the strengths of magnetic seed fields , to be then amplified by large - scale dynamos over cosmic time . in particular
, they also predict the existence of magnetic voids .
of course , in magnetising the icm / igm , low - mass galaxies have been competing with agn .
judging from the radio luminosities of the `` culprits '' , it is clear that nevertheless low - mass galaxies may have contributed significantly .
a typical starburst dwarf galaxy emits a monochromatic radio luminosity of @xmath11 w hz@xmath12 , while this figure is @xmath13 w hz@xmath12 for radio galaxies in the fri / ii transition regime .
hence , the radio power produced by agn is some 15000 times larger than that of dwarf galaxies .
however , ( @xmath14cdm ) cosmology helps at this point , since dwarf galaxies have come in huge numbers if the bottom - up scenario of structure formation holds .
furthermore , the lifetime of radio galaxies is limited ( @xcite ) , @xmath15 yr , with duty cycles of @xmath16 yr .
hence , this yields an effective activity period of @xmath17 yr @xmath18 .
finally , it should be noted that central so - called `` mini - halos '' , which are bright extended radio sources located in the centres of cooling - flow clusters ( perseusa , hydraa , virgoa ) can not do the magnetisation job : they are pressure - confined .
measurements of the radio continuum radiation of dwarf galaxies over a large frequency range have shown that ongoing star formation in them is accompanied by enhanced radio continuum emission ( @xcite , @xcite ) .
owing to their shallow gravitational potentials , the containment of relativistic cosmic - ray particles in such galaxies is , however , low , as inferred from studies of the spectral index and magnetic - field structure .
[ fig2 ] demonstrates this : in contrast to massive spiral galaxies , whose radio continuum is dominated by synchrotron radiation at cm wavelengths ( @xcite ) there is a lack of it for dwarf galaxies .
the lack is more pronounced the lower the luminosity ( hence mass ) of the dwarf galaxy ( @xcite ) .
these relativistic particles streaming into the halos of dwarf galaxies lose their energy on a time scale of @xmath19 yr according to eqn .
( [ eqn1 ] ) via synchrotron and inverse - compton radiation .
they are thus quickly rendered invisible at cm wavelengths , while their synchrotron emission will still be detectable at metre waves where their lifetime is 5 to 10 times longer .
the containment of the relativistic plasma and its possible transport out of a galaxy is obviously influenced by the overall magnetic - field configuration . in massive disk galaxies ( spirals )
the face - on view shows that the magnetic field closely follows the spiral arms , while the edge - on view discloses a plane - parallel magnetic field close to the galactic planes , while the halo field flares at large galacto - centric distances , rendering its overall appearance x - shaped ( @xcite ) . the existence of winds in low - mass galaxies is inferred from the observed kinematics of the gas ( measured with slit spectroscopy ) , but can arguably be also inferred from measurements of the temperature of the hot ( x - ray - emitting ) gas .
for instance , @xcite found the outflow velocities in ngc1569 to exceed the escape speed , and @xcite derived a temperature of its hot , x - ray - emitting gas to exceed the virial temperature .
the transport of a relativistic plasma out of this galaxy is strongly suggested by two observations ( @xcite ) .
first , ngc1569 possesses a radio halo , extending out to about 2 kpc at 1.4 ghz ( see fig .
[ fig3 ] , left ) .
second , the projected orientation of its magnetic field as deduced from measurements of the linear radio polarisation is radial throughout ( fig .
[ fig3 ] , right ) .
this magnetic field is just dragged along with the wind , as the energy density of the wind is about 30 times that of the magnetic field . a similar finding
was made by @xcite for the dwarf irregular ngc4449 , which also possesses a low - frequency radio halo ( @xcite ) .
the lifetime of low - frequency halos around dwarf galaxies can be estimated using eqn .
( [ eqn1 ] ) . with @xmath20
g magnetic fields in their surroundings the dominant loss mechanism will be inverse compton , the lifetime of relativistic electrons radiating at 120 mhz then being of order 500 myr , much longer than for those seen at cm wavelengths .
this is also illustrated in fig .
[ fig4 ] , which sketches the temporal variation of an `` aging '' synchrotron spectrum .
assume that the energy spectrum of the particles initially extended to infinity .
as time elapses , high - energy particles quickly lose their energy , producing a break in the spectrum .
this gives rise to a corresponding break in the radiation spectrum , which wanders more and more slowly towards lower frequencies as time elapses .
hence , it takes much more time for this break to arrive at frequencies observed , e.g. , with lofar .
the break frequency measures the time elapsed since the last starburst with its supernova activity has ceased . while aging , particles that have left a dwarf galaxy by a wind will slowly diffuse away from it .
this can happen with a maxmimum speed corresponding to the alfvenic one , i.e. @xmath21 where @xmath22 is the number density of thermal electrons ( or protons ) of the surrounding medium . taking @xmath23 g and @xmath24 @xmath25 ,
the relativistic particles radiating at 120 mhz could move out to @xmath26 kpc within 500 myr , or correspondingly further when caught at still lower frequencies . hence
, such halos could have total sizes of @xmath27 kpc .
lofar will easily detect low - frequency halos of dwarf galaxies out to large ( 100 mpc , coma cluster ) distances .
the lofar survey , which will cover about half the sky and will be conducted within a year s time , will provide a 5-@xmath28 flux limit of 0.1 mjy / b.a . at 120 mhz in 1 hour ( with a synthesized beam of 13 , for long baselines ) . using the synchrotron halo of ngc4449 as a template
, we can estimate the expected brightness at lofar frequencies , which is about 0.4 mjy / b.a . at 120 mhz .
its detection should hence be feasible .
assuming a 60 kpc diameter as estimated above and placing it at @xmath29 mpc it would be seen with a @xmath30 2 angular extent . in a cluster or group environment , such halos are likely to trail behind the galaxy as it moves in the cluster potential with speeds much larger than the diffusion speed of the relativistic particles .
this will actually provide some information on the proper motion of dwarf galaxies having undergone star formation over the past @xmath31 yrs .
an impressive example of such trailing low - frequency radio structures is seen in the perseus cluster , where the head - tail radio source ngc1265 leaves behind a huge radio tail that can be traced over a projected path of 1.1 mpc at low frequencies ( @xcite ) so , low - frequency radio emission `` memorises '' starburst activity up to about 1 gyr after its termination , a time scale much longer than that of any other integral property of galaxies ( h@xmath32 and fir luminosity , or broad - band optical / uv / ir colours ) .
this emission therefore also provides a powerful diagnostic tool to search for post - starburst galaxies and to explore their recent - to - past star - formation rate ( sfr ) and magnetic field evolution . in this context , lofar will also deliver important new information to address the long - standing question pertaining to possible evolutionary links between the two main classes of late - type dwarf galaxies , dwarf irregulars ( dis ) and blue compact dwarf galaxies ( bcdgs ) .
bcdgs in a post - starburst phase will be easily identifiable by their weak radio emission at cm wavelengths in conjunction with their bright radio halos at meter wavelengths .
by contrast , non - starbursting late - type dwarf galaxies will be rendered undetectable at cm wavelengths , while their low - frequency radio continuum emission should still exhibit the synchrotron halos created during past activity phases .
hence , if the standard evolutionary scenario , i.e. from bcdgs to dis , is correct , then lofar will discover a large population of quiescent dis with low - frequency radio halos .
uk expresses his deep gratitude for the warmhearted hospitality he experienced at _ the orthodox academy of crete _ and , in particular , to prof .
alexandros k. papaderos and his family .
|
with the operation of lofar , a great opportunity exists to shed light on a problem of some cosmological significance .
diffuse radio synchrotron emission not associated to any obvious discrete sources as well as faraday rotation in clusters of galaxies both indicate that the intergalactic or intracluster medium ( igm , icm ) is pervaded by a weak magnetic field , along with a population of relativistic particles . both , particles and fields must have been injected into the igm either by active galactic nuclei ( agn ) or by normal star - forming galaxies .
excellent candidates for the latter are starburst dwarf galaxies , which in the framework of hierarchical structure formation must have been around in large numbers
. if this is true , one should be able to detect extended synchrotron halos of formerly highly relativistic particles around local starburst or post - starburst dwarf galaxies .
with lofar , one should easily find these out to the coma cluster and beyond .
|
the light variability of the short - period contact binary sw lac ( p 0.@xmath032 , v@xmath1=8.@xmath291 ) is very well known and studied by several investigators since its discovery by ( * ? ? ?
* miss ashall ( 1918 ) ) .
the first photoelectric ubv light curves of the system were obtained by ( * ? ? ?
* brownlee ( 1956 ) ) , who also pointed out the light curve asymmetries from cycle to cycle .
these asymmetries were confirmed and attributed to the existence of cool spot regions by several authors ( see ( * ? ? ? * albayrak et al .
2004 ) , ( * ? ? ?
* alton & terrell 2006 ) and references there in ) .
the spectral studies of the system including spectral classification , mass ratio determination and uv / x - ray region spectral analysis were carried out by several investigators , who revealed that the system is a w - type contact binary showing chromospheric and coronal activity ( see ( * ? ? ?
* enavci et al . 2011 ) and references there in , for details ) .
the aim of this study is to perform the light curve analysis with the spot modeling , using the 2009 and 2010 light curves of the system with the help of the information coming from the doppler maps obtained by ( * ? ? ?
* enavci et al . ( 2011 ) ) .
the 2009 and 2010 bv band light curves of the contact binary sw lac were obtained at the ankara university observatory , using an apogee alta u47 ccd camera attached to a 40 cm schmidt - cassegrain telescope .
bd+37@xmath3 4715 and bd+37@xmath3 4711 were chosen as comparison and check stars , respectively .
the nightly extinction coefficients for each passband were determined by using the observations of the comparison star .
a total of 700 and 995 data points were obtained in each passband , while the probable error of a single observation point was estimated to be @xmath4 and @xmath5 for 2009 and 2010 bv bands , respectively .
the 2009 and 2010 bv light curves were analysed simultaneously with the radial velocity curves of the system obtained by ( * ? ? ?
* rucinski et al .
( 2005 ) ) using the interface version of the wilson - devinney code ( ( * ? ? ?
* wilson & devinney 1971 ) ) , phoebe ( ( * ? ? ?
* prsa & zwitter 2005 ) ) .
since the surface reconstructions of the system were performed using the time series spectra obtained in 2009 , we first adopted the spot modeling , as three main circular spot regions , to 2009 light curves and carried out the lc modeling ( see fig.[fig1 ] ) .
the results from the lc and spot modeling were represented in fig.[fig2 ] .
the analysis showed that the theoretical light curves are compatible with the observed ones , though the circular spot modeling was performed .
however , in order to perform more reliable spot modeling , a code with a none circular shaped spot approximation is needed as the doppler maps clearly show us the spots are not circular .
|
in this study , we present the preliminary light curve analysis of the contact binary sw lac , using b , v light curves of the system spanning 2 years ( 2009 - 2010 ) . during the spot modeling process , we used the information coming from the doppler maps of the system , which was performed using the high resolution and phase dependent spectra obtained at the 2.1 m otto struve telescope of the mcdonald observatory , in 2009 .
the results showed that the spot modeling from the light curve analysis are in accordance with the doppler maps , while the non - circular spot modeling technique is needed in order to obtain much better and reliable spot models .
|
with the rise of micropattern gas gain devices like gem @xcite , micromegas @xcite , cat @xcite or microcat @xcite the field of application of gas filled detectors has been widened up .
these structures accept high rates and due to their parallel plate geometry they produce short signals and good time - resolution . + several attempts have been made in the past to enhance the gain and the reliability by combining micropattern structures .
the performance of the combination of msgcs with gem structures is reported in refs .
@xcite . the combinations of msgcs and mgcs with gem , micromegas and plate avalanche chambers are discussed in ref .
@xcite with a special emphasis on rate and gain limitations .
a gem , used as a preamplification stage , has also been successfully combined with a groove chamber @xcite .
first results of a device combining a micromegas and a gem have been presented in ref .
gem structures have been combined to double gem @xcite , triple gem and even to quad gem configurations @xcite , which reach an enormous gas gain in the order of @xmath0 . + adding a gem to the microcat structure a stable operation with respect to sparks is obtained at moderate potentials ; nevertheless quite high gas gains are achievable at higher potentials . for this reason
the detector can be operated in applications where higher gas pressures ( @xmath1 ) and high-@xmath2 gases ( like e.g. xenon ) are required , i.e. in x - ray detection , as will be shown in this paper .
+ the intention of this investigation is to optimise microcat- and gem - based detector systems for x - ray imaging ( typical photon energy range : @xmath3 ) , where the drift gaps are usually several @xmath4 large and the drift fields are in the range of @xmath5 or even less .
in the past several microcat structures have been investigated @xcite . recently , an optimised microcat mesh has been produced by stork screens . in this section
we compare the charge transfer derived from current measurements to simulations , and we demonstrate that the new device is superior to previously investigated microcat structures with respect to maximum electron transparency ( to maximise the effective gain ) and minimum ion feedback ( to avoid field perturbations in the drift region ) .
the microcat structure ( see fig . [ fig_opmcat ] ) has been optimised with respect to maximum electron transparency and minimum ion feedback while retaining a good mechanical stability @xcite .
the parameters of the newly built device are given together with those of previously used microcats in table [ table_mcattypes ] .
.dimensions of the microcat structures .
[ cols="^,^,^,^,^,^",options="header " , ] the holes which have a hexagonal shape with edges rounded by the production process form a hexagonal lattice .
the cathode distance between the microcat and the subjacent anode amounts to @xmath6 ( see fig .
[ fig_gemmcat ] ) , .
the drift cathode is mounted at a distance of @xmath7 above the gem structure.,width=245 ] which is in fact very close to the optimum distance of about @xmath8 with respect to maximum time - resolution .
the anode is realised by an interpolating resistive readout structure @xcite .
the drift region has a height of @xmath9 ( since the gem structure is not mounted ) .
for all measurements in this section the detector volume is filled with a gas mixture of @xmath10 @xmath11 ( 90/10 ) .
+ the charge transfer behaviour of the microcat depends only on the ratio @xmath12 of the fields above and below the mesh .
this has been confirmed by simulations @xcite . for the determination of the charge
transfer the measurement of at least two of the following three currents is neccessary : @xmath13 with @xmath14 as the incoming current : @xmath15 here , @xmath16 denotes the electron transparency , @xmath17 the ion feedback , which describes the fraction of ions drifting back to the drift cathode , @xmath18 the gas gain , @xmath19 the photon rate , @xmath20 the mean energy deposited by the photons in the detector , @xmath21 the mean ionisation potential of the gas atoms / molecules and @xmath22 the elementary charge . + since eqs .
[ eq_idrift]-[eq_ianode ] fulfill current conservation , equivalent to @xmath23 , only two of the three relations are independent and thus not sufficient to determine all three unknown variables @xmath16 , @xmath17 and @xmath18 .
therefore , one additional condition is required , which we determine by the simulation presented in the following paragraph .
the electric fields for the microcat structures have been calculated in three dimensions using the maxwell package @xcite .
the charge drift was calculated using the garfield program @xcite .
the gas properties used for the garfield simulation like electron diffusion or drift velocity are calculated by the magboltz program @xcite .
the distance between the lower side of the mcat and the anode has been set to @xmath8 , the mcat voltage to @xmath24 while the anode structure has zero potential , leading to an average electric field in the amplification gap of @xmath25 .
all potentials are assumed to be constant during the simulation since the change of the potential during one multiplication process due to charge motion in the real detector is calculated to be in the order of @xmath26 .
+ diffusion has been included for the drifting electrons but neglected for the ions since this contribution is expected to be small .
charge multiplication has not been considered .
the ion feedback simulation is based on an effective size of the ion cloud of about @xmath27 .
a more detailed description of the charge transfer simulation can be found in ref .
+ the results of the simulation for the electron transparency @xmath16 and ion feedback @xmath17 are shown in fig . [ fig_mcat_chargetransfersimulation ] and ion feedback @xmath17 for the optimised mcat215 structure.,width=283 ] as a function of the ratio @xmath12 for fixed @xmath28 . as expected
, the ion feedback increases with @xmath29 but the electron transparency decreases .
we have measured the electric currents at all electrodes as a function of the ratio @xmath12 with the settings @xmath30 and @xmath31 . due to technical reasons the measurements of @xmath16 and @xmath17
are restricted to small values of @xmath29 below about @xmath32 . in section [ combination ] the range of @xmath29
is extended to larger values of about @xmath33 .
+ from the measured currents we compute the effective gain @xmath34 using the relations [ eq_idrift][eq_ianode ] : @xmath35 the gain variation due to the increasing drift field in the investigated range of @xmath29 has been calculated with garfield for electrons starting @xmath8 above the mcat along the symmetry axis of a hole to be less than @xmath36 . therefore , the measured product @xmath37 is expected to follow the simulation of @xmath16 . in fig .
[ fig_mcat_epsiloncomparison ] and comparison of the normalised product with the simulated electron transparency for the optimised mcat215 structure.,width=283 ] where we compare the measurement with the simulation , we observe that contrary to the simulation the measured product drops at small values of @xmath29 .
we attribute this effect to attachment of the primary electrons and to recombination with back drifting positive ions .
these effects which have not been considered in the simulation seem to be negligible at higher drift fields corresponding to @xmath38 .
we derive an effective electron transparency @xmath16 ( see fig . [ fig_mcat_epsiloncomparison ] ) by normalising the measured @xmath37 to the simulation of @xmath16 at @xmath39 .
the systematic uncertainty of this procedure is indicated in the figure by the enlarged error bars .
+ we also compare the ion feedback @xmath17 with the measurement using the reasonable approximation @xmath40 : @xmath41 the measurement deviates considerably from the simulation at small @xmath29 illustrated in fig .
[ fig_mcat_deltacomparison ] . as a function of @xmath29 for the optimised microcat structure.,width=283 ] obviously , our model is too coarse in this region of @xmath29 .
[ fig_mcat_comparison1 ] of four microcat structures as a function of @xmath29.,width=283 ] shows a comparison of the measured effective electron transparency @xmath16 of four microcat structures with dimensions summarised in table [ table_mcattypes ] .
the optimised mesh offers the largest electron transparency .
the drop of the effective electron transparency at low @xmath29 is different for the different microcats because the amount of gas impurities slightly varies from measurement to measurement .
+ the comparison of the ion feedback @xmath17 of the microcat structures ( see fig .
[ fig_mcat_comparison2 ] ) of four microcat structures as a function of @xmath29.,width=283 ] shows , that the fraction of back drifting ions of the optimised mcat215 structure is nearly as small as for the old mcat215 structure and lower than that of the mcat155 device , although the electron transparency @xmath16 of the optimised mesh is much larger than that of the competing devices ( compare to fig .
[ fig_mcat_comparison1 ] ) . knowing the transparency @xmath16 , the gas gain @xmath18 can be calculated from eq .
[ eq_transparency1 ] : @xmath42 fig .
[ fig_mcat_gasgain ] as a function of the applied mcat voltage.,width=283 ] shows the gain @xmath18 as a function of the mcat voltage .
it has been obtained from the observed pulse heights of an @xmath43-source where the gain had been calibrated at one voltage to the current measurement . the maximum gain , which we define as the gas amplification , that can be reached within a discharge limit of less than @xmath44 , is in the order of @xmath45 .
the comparison of the optimised and the three previously investigated microcat structures shows that the new mesh is superior with respect to electron transparency .
the ion feedback of the optimised structure is in a reasonable range ( @xmath46 up to @xmath47 ) . in order to reach maximum effective gain
the optimised device should be operated with ratios @xmath48 of the fields above and below the mcat
. a maximum gain of about @xmath45 can be obtained at atmospheric pressure for argon gas mixtures .
the charge transfer and gas gain behaviour of the gem is investigated by simulations and current measurements to obtain the optimum operation parameters in view of a combination of the gem with the microcat .
the influence of different voltage settings and gas mixtures and pressures has been studied .
the gem structure is fixed above the microcat at a distance of @xmath49 ( see fig . [ fig_gemmcat ] ) .
the microcat serves as anode with a slightly positive voltage ( @xmath50 ) . in this geometry
the drift region above the gem amounts to about @xmath7 , the region below the gas electron multiplier corresponds to the distance to the microcat structure , consequently @xmath49 .
we denote the electric field above and below the gem with drift field and transfer field , respectively .
+ the gem @xcite ( see fig . [ fig_gemmodel2 ] ) , itself , has a hexagonal hole arrangement with a pitch of @xmath51 , an optical transparency of about @xmath52 and a total thickness of @xmath53 ( see fig .
[ fig_gemmodel ] ) .
+ the current measurements have been carried out with a @xmath43-source , collimated to @xmath54 , emitting photons with a rate of about @xmath55 .
the charge transfer behaviour of the gem can be described by four current equations .
two electron transparencies are introduced : @xmath56 is the probability , that the electrons , coming from the drift / conversion region , reach the centre of a gem hole ; @xmath57 is the probability , that these electrons reach the anode .
this approach has already been introduced by other groups ( see for example in ref .
the ion feedback is again denoted by @xmath17 . @xmath58 current conservation reduces these equations to three independent relations . in the following we make use of the effective gain which is defined by @xmath59 . like in the previous section the three - dimensional electric field in the gem geometry and the charge transfer behaviour
have been investigated with maxwell and garfield , respectively . for the transparency simulation
the charge multiplication process has not been considered ; ion feedback simulations have not been carried out .
+ if not stated differently , we have assumed in the simulations a drift field of @xmath60 , a transfer field of @xmath61 , a gem voltage of @xmath62 and a gas mixture of @xmath10 and @xmath63 @xmath11 ( 90/10 ) .
[ [ influence - of - the - drift - field ] ] influence of the drift field : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the dependency of the electron transparencies @xmath56 and @xmath57 on the applied drift field ( @xmath64 ) is shown in fig .
[ fig_gem_driftsim1 ] . and @xmath57 and normalised gain as a function of the drift field.,width=340 ]
all electrons reach the centre of the hole in the gem structure ( @xmath65 ) , independent of the chosen gas pressure .
the parameter @xmath57 is not influenced noticeably by the drift field .
the deviation of @xmath57 between the two gas pressures can be explained by the smaller electron diffusion at higher pressure .
+ the situation is very different from that in particle physics applications where usually due to rather high drift fields @xmath56 is smaller than @xmath66 . in our case
the overall electron transparency is obviously limited by diffusion effects which cause a large fraction of electrons to be caught by the lower electrode .
+ we have also studied the effect of drift field variations on the gain @xmath18 .
the gain was calculated with garfield for one electron starting @xmath8 above the gem along the symmetry axis of a hole .
the smallest gain @xmath67 , which is obtained at a drift field of @xmath68 , is normalised to @xmath69 .
the gain increases by about @xmath70 in the investigated drift field range .
[ [ influence - of - the - transfer - field ] ] influence of the transfer field : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the electron transparencies @xmath56 and @xmath57 as a function of the applied transfer field ( @xmath71 ) are shown in fig .
[ fig_gem_driftsim2 ] . and @xmath57 and normalised gain as a function of the transfer field .
, width=340 ] the transparency @xmath56 is constant @xmath72 for all applied transfer fields , whereas @xmath57 shows a strong increase with rising transfer fields .
smaller electron diffusion at higher pressure leads to a larger @xmath57 .
[ fig_gem_driftsim2 ] also shows the normalised gain @xmath73 as a function of the transfer field .
the change at low transfer fields is about @xmath74 ; identical as observed with a variation of the drift field ( see fig .
[ fig_gem_driftsim1 ] ) . at transfer fields
@xmath75 the parallel plate amplification starts in the gap between the gem and the anode plane , resulting in an additional contribution to the gain @xmath18 .
[ [ influence - of - the - gem - voltage ] ] influence of the gem voltage : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the dependence of the electron transparencies on the applied gem voltage ( @xmath76 ) has been studied ( see fig . [ fig_gem_driftsim3 ] ) . and @xmath57 and gas gain as a function of the gem voltage.,width=340 ] the transparencies @xmath56 and @xmath57 depend only weakly on the gem voltage .
for very low gem voltages @xmath56 decreases because the drift channel through the gem holes becomes wider . smaller diffusion at a gas pressure of @xmath63 results in a larger electron transparency than for atmospheric pressure .
the calculated gem gain @xmath18 has the typical exponential shape with increasing operation voltage .
a gas mixture of @xmath11 ( 90/10 ) at a pressure of @xmath10 was chosen for most of the measurements if not stated differently .
in addition , the charge transfer and multiplication behaviour was investigated for several argon , krypton , xenon and @xmath77-quench gas mixtures at various pressures up to @xmath63 .
+ from eqs .
[ eq_gidrift][eq_gianode ] the product @xmath78 , the transparency @xmath57 and the ion feedback @xmath17 ( using the reasonable approximation @xmath40 ) can be calculated : @xmath79 @xmath80 @xmath81 [ [ influence - of - the - drift - fielddriftfield ] ] influence of the drift field:[driftfield ] + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + fig . [ fig_gem_eps1g5 ] as a function of the drift field for several gas pressures of @xmath82 ( 90/10).,width=283 ] shows the influence of the drift field on the product @xmath78 for various gas pressures of @xmath82 ( 90/10 ) .
the decrease at small drift fields is again attributed to recombination and attachment effects which are expected to increase with gas pressure .
the simulation ( see fig . [ fig_gem_driftsim1 ] ) indicates that the gain @xmath18 slightly depends on the drift field , whereas the transparency @xmath65 is constant .
therefore the measured electron transparency @xmath56 can be assumed to be very close to @xmath66 .
+ the measurements show that the transparency @xmath57 does not change noticeably with the varying drift field ( see fig .
[ fig_gem_eps27 ] ) .
this is in line with expectation and simulation .
however , the transparency @xmath57 is apparently overestimated in the simulation , which means that in reality more electrons move to the gem - bottom electrode .
possible explanations of this effect could be : 1 .
neglection of the avalanche development in the simulation : repulsive forces and uv - photons widen the electron cloud . 2
. underestimation of the electron diffusion at high electric fields in magboltz .
the measured ion feedback @xmath17 of the gem is presented in fig .
[ fig_gem_delta4 ] for several gas pressures of @xmath82 ( 90/10 ) .
it increases nearly linearly with rising drift field .
[ [ influence - of - the - transfer - field-1 ] ] influence of the transfer field : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the measured increase of @xmath78 with rising @xmath83 ( see fig . [ fig_gem_eps1g8 ] ) as a function of the transfer field.,width=283 ] is in a good agreement with the increase of the gain @xmath18 which was predicted by the simulation ( compare to fig .
[ fig_gem_driftsim2 ] ) .
thereby the simulated gain had to be renormalised because of the inaccurate knowledge of the corresponding townsend coefficients .
we conclude that @xmath56 is not sizeably influenced by the transfer field .
+ however , the transparency @xmath57 increases strongly with rising transfer fields ( see fig .
[ fig_gem_eps28 ] ) .
the deviation of the absolute values of @xmath57 between simulation and measurement has already been discussed in section [ driftfield ] . for transfer fields
@xmath75 the parallel plate amplification underneath the gem leads to an apparent rise of @xmath57 which is more pronounced in the measurement due to the longer multiplication path of @xmath49 compared to @xmath8 in the simulation .
+ the measured ion feedback @xmath17 ( see fig . [ fig_gem_delta2 ] ) is approximately constant at lower transfer fields .
when the parallel plate amplification produces a noticeable amount of gain ( at large transfer fields ) the ion feedback rises .
we conclude that the ion feedback is larger for ions which are produced below the gem rather than in the gem holes .
[ [ influence - of - the - gem - voltage-1 ] ] influence of the gem voltage : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the expected exponential dependence of @xmath78 on the gem voltage is shown in fig .
[ fig_gem_eps1g6 ] for @xmath11 ( 70/30 ) at various gas pressures . + with increasing @xmath84 the transparency @xmath57 decreases ( see fig .
[ fig_gem_eps26 ] ) . for high gem fields
the electron drift channel becomes very narrow . as a consequence the probability that electrons diffuse to one of the dense drift lines that end on the gem bottom electrode
is increased .
+ the ion feedback as a function of the gem voltage is shown in fig .
[ fig_gem_delta5 ] . due to very small currents at the drift cathode at low gains
the error bars are very large .
the deviations from a constant behaviour are not significant .
we have investigated the gain behaviour of the gem for different gas mixtures and pressures . the maximum gas amplification , which can be reached within a discharge limit of less than @xmath44 , decreases drastically with gas pressure ( shown as an example for a @xmath85 ( 90/10 ) gas mixture in fig . [ fig_gem_geff1 ] ) .
this effect is studied in more detail in ref .
@xcite .
the approach of the introduction of the two electron transparencies @xmath56 and @xmath57 has been sensibly confirmed by our measurements .
+ the systematic studies of the currents at the electrodes in a wide range of voltages and electric fields show the following results : * the transparencies @xmath56 and @xmath57 do not depend significantly on the drift field for @xmath86 .
the ion feedback @xmath17 rises nearly linearly with increasing drift field . in order to avoid electron losses due to recombination and attachment in the conversion region the drift field
should not be chosen too small , especially for higher gas pressures .
the gain @xmath18 rises slightly with increasing drift field .
* the transfer field has no influence on the electron transparency @xmath56 in the investigated field range @xmath87 .
the transparency @xmath57 strongly rises with increasing transfer fields .
the ion feedback @xmath17 does not change noticeably for small transfer fields and increases slightly for higher transfer fields .
also the gain @xmath18 rises with increasing transfer field .
* very small gem voltages lead to a decrease of @xmath56 .
the transparency @xmath57 decreases slightly for large potential differences , whereas the ion feedback @xmath17 is nearly constant .
the combination of the optimised mcat215 and the gem has been investigated by current measurements to determine the optimum voltage settings with respect to maximum effective gain and minimum ion feedback . the maximum gas gain in different gas environments has beed studied .
no simulations have been carried out .
however , the results of the individual simulations from sections [ mcat ] and [ singlegem ] can be combined .
[ fig_gemmcat ] shows the schematic setup of the combination of microcat and gem .
we call the electric field between the drift cathode and the gem by @xmath88 , the field between the gem and the mcat by @xmath83 and between the mcat and the anode by @xmath89 , where @xmath90 denotes the voltage applied to the mcat and @xmath31 the distance between mcat and anode
. the ratio of the fields above and below the microcat structure is denoted by @xmath91 .
the electron transparency , ion feedback and gain of the microcat are denoted by @xmath92 , @xmath93 and @xmath94 , respectively .
all gem transparencies are denoted as in the previous section , the gem gain is denoted by @xmath95 .
the total effective gain is defined as follows : @xmath96 fig .
[ fig_mcatgem_relpulseheight1 ] shows the relative signal pulse heights as a function of the transfer field for a drift field of @xmath60 , a gem voltage of @xmath97 , an mcat voltage of @xmath98 at a gas pressure of @xmath10 @xmath11 ( 90/10 ) .
up to transfer fields of @xmath99 the signal pulse height rises .
the subsequent plateau reaches up to @xmath100 , before the signals get smaller again . at very high transfer fields of @xmath101 parallel plate amplification between gem and microcat starts , which again leads to an increase of the signals pulse heights .
+ the shape of this curve is affected by two opposite effects : on the one hand the transparency @xmath57 of the gem rises for increasing transfer fields ( compare to fig .
[ fig_gem_eps28 ] ) , and on the other hand the electron transparency @xmath92 of the microcat drops for larger values of @xmath29 ( compare to fig .
[ fig_mcat_chargetransfersimulation ] ) . with the knowledge of @xmath102 , determined in section [ singlegem ]
, the electron transparency of the mcat can be calculated up to values of @xmath103 .
the result is shown in fig .
[ fig_mcatgem_relpulseheight3 ] .
it agrees perfectly with the transparency measured directly with the microcat alone ( see fig .
[ fig_mcat_epsiloncomparison ] ) .
measurements at different gem voltages show the same results . however , a large deviation from the simulation is obvious for @xmath104 .
the measured transparency @xmath92 is significantly smaller than what the simulation predicts .
we have no explanation for this effect . + by means of current relations and the reasonable assumptions , that @xmath65 , @xmath105 and @xmath106 the ion feedback @xmath93 of the microcat can be determined : @xmath107 the result ( see fig . [ fig_mcatgem_deltamcat ] ) shows a good agreement to the direct measurement ( compare to fig .
[ fig_mcat_deltacomparison ] ) . despite the coarse model for the ion feedback simulation the simulation and the measurement match
very well for larger ratios @xmath29 .
additional measurements have shown that the field in the conversion region @xmath86 above the gem has no influence on the ion feedback @xmath93 .
[ fig_mcatgem_deltatotal ] shows , that the total ion feedback @xmath108 is getting more favourable for higher transfer fields .
the gem voltage has no strong effect on the total ion feedback @xmath17 ( see fig .
[ fig_mcatgem_ionfeedback1 ] ) .
the increase of the total ion feedback @xmath17 with decreasing mcat voltage ( see fig .
[ fig_mcatgem_ionfeedback2 ] ) can be explained by the rising contribution of the ion feedback @xmath93 , which gets larger for higher ratios @xmath91 ( compare to fig .
[ fig_mcatgem_deltamcat ] ) . the effective gain @xmath34 ( see eq .
[ eq_mggeff ] ) of the gem / microcat - combination can be calculated from the measured currents @xmath14 ( see eq .
[ eq_i0 ] ) and @xmath109 as follows : @xmath110 the effective gain as a function of the transfer field rises to maximum at @xmath111 in @xmath10 @xmath11 ( 90/10 ) ( see fig . [ fig_mcatgem_geff1 ] ) . + in order to study the maximum gas gain a systematic investigation of the effective gain @xmath34 as a function of the gas pressure has been carried out for different noble gas / quench gas mixtures .
the gem voltages are chosen such , that this fragile pre - amplification device always works in a safe range , whereas the mcat produces the main amplification .
this method has the advantage that all discharges appear below the microcat structure , which is not easily damaged by sparks .
all sparks appear at a defined avalanche size depending on the gas mixture and pressure used , but independent on the voltage settings on gem or on mcat .
+ the measurements of the maximum gas gain are carried out for gas pressures up to @xmath63 .
[ fig_mcatgem_geffmax1 ] shows the measured maximum gain for several pressures of @xmath82 ( 90/10 ) and @xmath85 ( 90/10 ) . in a first approximation
the drop of the gain for higher pressures is approximately exponential .
the results of the exponentially fitted data are shown in fig .
[ fig_mcatgem_geffmax2 ] .
the largest gas gain of about @xmath112 can be achieved with the gas mixture of @xmath11 ( 90/10 ) . at atmospheric pressure
the gain decreases with the atomic number of the nobel gas .
we estimate that for gas pressures up to @xmath113 gains of at least @xmath114 can be reached , even for xe - mixtures .
the investigation of the optimum voltage settings for a gem / microcat - combination leads to the following results : * to avoid recombination and attachment effects the drift field above the gem should be chosen larger than @xmath115 .
higher gas pressure requires larger drift fields ( compare to section [ singlegem ] and fig .
[ fig_gem_eps1g5 ] ) .
* a transfer field in the range of @xmath116 ensures a maximum effective gain and a small total ion feedback .
* since the combined gain of the gem and the microcat is limited and because the gem is easily destroyed by discharges the microcat should be operated at much higher gain than the gem .
the maximum gas gain decreases with pressure and atomic number of the noble gas . a stable operation with gains in the order of @xmath117
can be obtained even for xenon gas mixtures up to @xmath63 .
the new optimised microcat structure is superior to all other microcat devices with respect to electron transparency .
the ion feedback of the new mesh is smaller than for the mcat155 and nearly as good as for the old mcat215 structure .
gas gains of larger than @xmath45 can be achieved in argon gas mixtures at standard pressure . + in our special detector setup a gem has been investigated with respect to its charge transfer and gas gain behaviour with different voltage settings and different gas environments .
the charge transfer simulations are well confirmed by the measurements .
a sub - division of the overall electron transparency into the two transparencies @xmath56 and @xmath57 is reasonable .
the achieved gains of the gem are comparable to published results .
+ the combination of both gas gain devices leads to a very high gas gain at atmospheric pressure and a reliable detector operation with gains in the order of @xmath117 in xenon gas mixtures at higher pressures of about @xmath63 .
the gem can be operated in a very safe gas gain mode , whereas the the microcat produces the largest amount of gain ; due to the robustness of the nickel mesh and its insensitivity against sparking no destruction of this device needs to be feared .
the fraction of back drifting ions is in the order of a few percent when microcat and gem are combined .
therefore , spatial reconstruction distortions due to space charge effects in the conversion region are expected to be small .
we are indebted to g. claassen and t. van de mortel from stork screens . only due to their effort the prototype production of
the optimised microcat mesh was possible .
we are grateful to g. zech for valuable comments that have contributed to this publication .
9 f. sauli .
_ gem : a new concept for electron amplification in gas detectors _ , nucl .
instr . and meth . * a 386 * ( 1997 ) 531 y. giomataris , ph .
rebourgeard , j.p .
robert , g. charpak .
_ micromegas : a high - granularity position - sensitive gaseous detector for high particle - flux environments _ , nucl .
instr . and meth .
* a 376 * ( 1996 ) 29 f. bartol , m. bordessoule , g. chaplier , m. lemonnier , s. megtert . _
the c.a.t .
pixel proportional gas counter detector _
, j. phys .
iii france ( 1996 ) 337 a. sarvestani , h.j .
besch , m. junk , w. meissner , n. sauer , r. stiehler , a.h .
walenta , r.h .
_ study and application of hole structures as gas gain devices for two dimensional high rate x - ray detectors _ , nucl .
. and meth . ( 1998 ) 238 y. benhammou , j.m .
brom , j.c .
fontaine , d. huss , f. jeanneau , a. lounis , i. ripp - baudot , a. zghiche .
_ comparative studies of msgc and msgc - gem detectors _ , nucl .
. and meth . ( 1998 ) 400 o. baruth , s. keller , u. werthenbach , g. zech , t. zeuner . _ sparks in msgc and gem detectors and a robust alternative _ , nucl .
instr . and meth .
( 2000 ) 272 y. bagaturia , o. baruth , h.b .
dreis , f. eisele , i. gorbunov , s. gradl , w. gradl , s. hausmann , m. hildebrandt , t. hott , s. keller , c. krauss , b. lomonosov , m. negodaev , c. richter , p. robmann , b. schmidt , u. straumann , p. trul , s. visbeck , t. walter , c. werner , u. werthenbach , g. zech , t. zeuner , m. ziegler .
_ studies of aging and hv break down problems during development and operation of msgc and gem detectors for the inner tracking system of hera - b _ , accepted for publication in nucl .
instr . and meth .
p. fonte , v. peskov , b.d .
. _ rate and gain limitations of msgcs and mgcs combined with gem and other preamplification structures _ , nucl . instr . and meth . (
1998 ) 405 i. reichwein , u. werthenbach , g. zech .
_ properties of groove chambers _ , accepted for publication in nucl . instr . and meth
. s. kane , j. may , j. miyamoto , i. shipsey . _ a study of micromegas with preamplification with a single gem _ , proceeding for the international conference on advanced technology and particle physics , villa olmo , como , italy , oct 15 - 19 ( 2001 ) , submitted to world scientific ( 2002 ) s. bachmann , a. bressan , l. ropelewski , f. sauli , a. sharma , d. mrmann .
_ charge amplification and transfer processes in the gas electron multiplier _ , nucl .
instr . and meth . ( 1999 ) 376 a. buzulutskov , a. breskin , r. chechik , g. garty , f. sauli , l shekhtman .
_ the gem photomuliplier operated with noble gas mixtures _ , nucl
. instr . and meth .
( 2000 ) 164 a. buzulutskov , a. breskin , r. chechik , g. garty , f. sauli , l shekhtman .
_ further studies of the gem photomultiplier _ , nucl .
instr . and
( 2000 ) 68 a. orthen , h. wagner , h.j .
besch , r.h .
menk , a.h .
_ charge transfer considerations of microcat - based detector systems _ , nucl .
. and meth . ( 2002 ) 160 h.j .
besch , m. junk , w. meissner , a. sarvestani , r. stiehler , a.h .
_ an interpolating 2d pixel readout structure for synchrotron x - ray diffraction in protein crystallography _ , nucl .
instr . and meth . ( 1997 ) 244 h. wagner , h.j .
besch , r.h .
menk , a. orthen , a. sarvestani , a.h .
walenta , h. walliser .
_ on the dynamic two - dimensional charge diffusion of the interpolating readout structure employed in the microcat detector _ , nucl .
instr . and meth . ( 2002 ) 334 maxwell 3d field simulator version 5.0.04 , ansoft corporation , pittsburgh , pa usa garfield , cern wire chamber field and transport computation program written by r. veenhof , version 7.05 ( 2001 ) s.f .
_ monte carlo simulation of electron drift and diffusion in counting gases under the influence of electric and magnetic fields _ , nucl .
instr . and meth . * a 421 * ( 1999 ) 234 magboltz , cern transport of electrons in gas mixtures computation program written by s.f .
biagi , version 2 j. benlloch , a. bressan , c. buttner , m. capeans , m. gruwe , m. hoch , j .- c .
labb , a. placci , l. ropelewski , f. sauli , a. sharma , r. veenhof .
_ development of the gas electron multiplier ( gem ) _ , ieee trans . nucl .
( 1998 ) 234 - 243 c. richter , a. breskin , r. chechik , d. mrmann , g. garty , a. sharma .
_ on the efficient electron transfer through gem _ , nucl .
instr . and meth . ( 2002 ) 528 a. bondar , a. buzulutskov , f. sauli , l. shekhtman .
_ high- and low - pressure operation of the gas electron multiplier _ , nucl .
instr . and meth . ( 1998 ) 418
|
a microcat ( micro compteur trous ) structure which is used for avalanche charge multiplication in gas filled radiation detectors has been optimised with respect to maximum electron transparency and minimum ion feedback .
we report on the charge transfer behaviour and the achievable gas gain of this device .
a three - dimensional electron and ion transfer simulation is compared to results derived from electric current measurements .
similarly , we present studies of the charge transfer behaviour of a gem ( gas electron multiplier ) by current measurements and simulations .
finally , we investigate the combination of the microcat and the gem by measurements with respect to the performance at different voltage settings , gas mixtures and gas pressures .
+ _ pacs : _ 02.60.cb ; 51.10.+y ; 51.50.+v ; 29.40.cs , , , , , micro pattern gaseous detectors ; gas gain devices ; gem ; gas electron multiplier ; microcat ; gas gain ; charge transfer ; electron transparency ; ion feedback
|
the large area telescope ( lat ) @xcite is the primary instrument on glast , and so it is of utmost importance to extract as much information as possible from the response of the lat to incident photons and particles . while the quantities of primary interest for each event are few ( namely the azimuth , elevation and energy of the incident photon / particle ) , the rich physics of the interactions of the particles / photons with the lat makes a principled reconstruction algorithm complex . whilst the objects of primary interest are photons , the interaction of a photon with the lat is essentially that of an electron - positron pair .
therefore we concentrate on the basic building blocks of event reconstruction , the analysis of the interaction of charged particles with the lat .
figure [ fig : tree ] ( left ) shows the schematic of an interaction between a charged particle and the lat .
visible in the figure are 1 ) multiple coulomb scattering in the tungsten foils ; 2 ) the production of secondary photons ; and 3 ) the production of secondary charged particles .
the geant4 toolkit @xcite is designed to simulate these physics processes in the _ forward _ direction .
the task in event reconstruction is the _ inverse _ problem estimating , from the data of the microstrip responses , which physics processes actually occurred in a particular event .
the result is an estimate not only for the original particle and its properties , but also of all secondary particles and photons . to accurately estimate the primary particle , it is necessary to estimate accurately all secondaries .
figure [ fig : tree ] ( right ) shows the tree of hypotheses for the physical processes at the first two layers of interaction of a charged particle with the lat .
the final leaves describe the _ structure _ of the hypothesized event reconstructions . clearly as we descend the layers the number of branches in this tree explodes .
but only a small number of their leaves will be consistent with the instrument data - with the pattern of microstrips that fired - and , further , this consideration can be used to exclude branches as the tree is descended , further limiting the number of leaves that must be considered .
a full event reconstruction consists of two stages : 1 . the enumeration of the possible event structures consistent with the microstrip data .
2 . the computation of the parameters of each event structure and their relative probabilities .
computation of the relative probabilities allows the final event reconstruction to be a weighted average of event structures , weighted according to the probability that the microstrip data actually came from that structure .
we concentrate first on the simplest type of event . if the only physical process that actually occurred was multiple coulomb scattering , then there is only a single ( x - y ) pair of microstrips at each layer ( excluding noisy strips ) .
this is the case for muons of moderate energy ( up to a few hundred mev ) , where the probability of producing secondary electrons or photons is extremely small , and hence can be neglected .
we parameterize the trajectory of the particle by 1 ) its origin ( a point outside the lat ) ; 2 ) the position at which it traverses each conversion layer ; and 3 ) its endpoint ( also outside the lat ) ; and from these we derive the incident directions @xmath0 and the scattering angles , @xmath1 , at each layer .
finally , we add the incident energy , @xmath2 , and the energy deposited in each conversion layer @xmath3 . denoting by @xmath4 the microstrips at each layer ,
we can write @xmath5 the first term on the right hand side is the likelihood .
it takes one of two values
one if the trajectory described by @xmath6 intersects all the microstrips that fired , and zero otherwise .
it serves to limit the region of the state space that is of interest .
the second term contains all the physics of the interactions of the particle with the lat .
we use conditional independence and decompose it as follows . @xmath7
these scattering distributions are the known distributions for particles of a specified energy incident on a lat foil @xcite .
the non - gaussian tails of the scattering angle distributions were modeled by a second gaussian component . for muons ,
the energy loss was parameterized as a landau distribution , with the distribution s parameters being functions of energy .
the parameters of primary interest , however , are the azimuth , elevation and energy , and so the distribution of primary interest is @xmath8 .
this is obtained from ( [ eq : bayes ] ) by marginalization .
this is performed numerically using markov chain monte carlo ( mcmc ) .
originally developed in physics @xcite , it has been extensively developed in statistics in the past 20 years , and is now a standard tool for use in the analysis of complex , high - dimensional probability distributions @xcite .
it works by simulating a markov chain whose equilibrium distribution is constructed to be the distribution of interest ( in this case , the distribution in ( [ eq : bayes ] ) ) .
averages over the distribution can then be made by forming averages over the states of the simulated markov chain .
for example , the mean energy is estimated by forming the mean of the energy variables over a length of the simulated chain , while ignoring all the other variables .
collecting all the variables into @xmath9 , and initializing @xmath10 the mcmc algorithm is iteration of 1 .
propose a change , @xmath11 with some proposal distribution @xmath12 2 .
accept the change with @xmath13 , and set @xmath11 , else retain @xmath9 the proposal distribution at each stage may be chosen to only change some of the elements of @xmath9 .
for this work , we use a cycle of proposals that successively proposes changes to @xmath14 . for electrons incident on the lat foils , as well as multiple coulomb scattering ,
there is appreciable probability of producing a secondary photon , and a small probability of producing a secondary electron .
( at 100mev these probabilities are @xmath15 and @xmath16 respectively . )
we restrict the discussion here to events which contain at most secondary photons .
typically , for electrons of a few hundred mev the secondary photons are not detected .
they carry energy away from the electron which is `` lost '' to the tracker . in the forward direction
this is modeled by a mixture distribution . with probability
@xmath17 no secondary is produced , and the energy loss follows a landau distribution . with probability @xmath18 , a photon is produced and the energy loss has two components , a landau distributed component from multiple coulomb scattering , plus a component distributed as @xmath19 representing the energy carried away by the photon .
the samples generated by the mcmc algorithm represent the distribution over the trajectories parameters . to compute the probability for an event structure
it is necessary to compute the normalizing factor that was omitted from equation ( [ eq : bayes ] ) .
this can be done by using the mcmc output to construct an importance sampling distribution , and using samples from that distribution to compute the normalizing factor .
this will be discussed elsewhere .
one thousand events were simulated @xcite for each of four sources muons and electrons of 100 and 200 mev .
the incidence directions were chosen randomly within a 45 degree cone .
figure [ fig : results ] shows the energy estimates of the reconstruction .
we do not show direction estimates for reasons of space and also because , for charged particles , the accuracy of the direction estimate is determined almost entirely by multiple coulomb scattering in the top foil . for electron events ,
those events where two charged tracks were detected by the reconstruction algorithm were not analyzed . in all four cases
the estimates are unbiased ; the histograms are centered accurately on either 100mev or 200mev .
the histograms for electrons show more dispersion than those for muons , due to the effect of energy being transferred into photons which are not detected .
note that these estimates were made using only the first 12 regular glast layers of the tracker and did not use any information from the calorimeter . in an upcoming paper
we will present detailed results and comparisons with current methodology .
|
we describe a statistical reconstruction methodology for the glast lat .
the methodology incorporates in detail the statistics of the interactions of photons and charged particles with the tungsten layers in the lat , and uses the scattering distributions to compute the full probability distribution over the energy and direction of the incident photons .
it uses model selection methods to estimate the probabilities of the possible geometrical configurations of the particles produced in the detector , and numerical marginalization over the energy loss and scattering angles at each layer .
preliminary results show that it can improve on the tracker - only energy estimates for muons and electrons incident on the lat .
address = usra - riacs , 444 , castro st , suite 320 , mountain view , ca 94041 address = slac / kipac , stanford university , stanford , ca
|
particles in anharmonic potentials is a well studied example of chaotic system .
a simple but good model of a system which exhibits parametric transitions from regular to chaos and vice - versa is the quartic oscillator .
the parameters which control the transitions are the coupling strengths of the cross terms .
the quartic oscillator besides being a case study of chaotic system , plays a very important role in the guise of higg s field , through which all the known fundamental particles acquire finite masses .
the same potential also appears in optical lattices , where quantum phase transition from superfluid to mott insulator has been observed @xcite .
optical lattices are regular intensity patterns of light created with counter propagating laser beams .
the laser beams have a gaussian profile and create an overlapping anharmonic potential across the optical lattice .
recent theoretical studies show that realization of the quantum phase transition can be more robust with a quartic potential @xcite . besides the quantum phase transitions , introducing a quartic confining potential stabilizes fast rotating vortices in bose - einstein condensates @xcite . in a recent work @xcite ,
the condensation temperature and thermodynamic properties of a rotating ideal bose gas in an anharmonic trap has been studied . however , in the experimental realization of the quadratic plus quartic confining potential , the observation of the fast rotating vortices eludes an unambiguous detection @xcite . in this paper
we calculate the critical temperature @xmath0 at which bosons confined by a quartic potential condense .
bec in such a potential was studied for low - dimensional systems @xcite .
this calculation requires the density of states . in our calculations we use the semiclassical expression of the density of states , which is valid at higher energies .
it is however not appropriate to study low lying states .
in particular , the energy of the ground state is essential to estimate @xmath0 for finite number of particles . to estimate the correction to @xmath0 in the finite particle case
, we calculate the ground state energy analytically @xcite .
the calculation is based on a method which optimizes the matrix elements of the quartic potential hamiltonian in the harmonic oscillator basis .
our calculations show that @xmath0 in quartic confining potential is higher than quadratic potential .
this is perhaps to be expected , since @xmath0 varies as @xmath4 when the density of states is proportional to @xmath5 . in case of 3d harmonic oscillator potential @xmath6 whereas it is 9/4 in the case of 3d quartic potential .
hence , in the quartic potential case @xmath0 varies as @xmath7 compared to @xmath8 in the case of harmonic potential . from our calculations ,
it is evident that the cross terms increase @xmath0 in 2d as well as 3d quartic potential .
we find that @xmath0 rises by factor of 1.2 and 1.1 in the 3d and 2d potentials respectively .
however , the experimental realizations of the trapping potentials , which are created from laser beams , are more appropriately described without the cross terms .
the general form of the quartic oscillator potential is @xmath9 .
it is homogeneous and has cross terms in the cartesian coordinate representation which couple motions along different axes .
potentials of this form occur in optical lattices , where counter propagating lasers create undulating patterns of standing radiation field . in one dimension
, a pair of counter propagating gaussian laser beams along @xmath10-axis of intensity profile @xmath11 creates an array of periodic intensity minima and maxima .
these are located along the @xmath10-axis .
depending on the detuning of the laser , the atoms are attracted to the intensity minima or maxima .
usually the wavelength of the laser @xmath12 is much smaller than the beam width @xmath13 . to a very good approximation , the potential across a surface normal to the laser beam @xmath14 tuning the parameters of the laser beams , it possible to retain only the quartic term . for simplicity , neglecting the cross terms , in three dimension @xmath15 the eigen energies of the corresponding hamiltonian is the sum of eigen values corresponding to each dimension .
for the one dimensional quartic oscillator , eigen values can be calculated by minimizing the expectation of the hamiltonian in the basis states of harmonic oscillator of appropriately chosen frequency@xcite .
generalizing the result to three dimensional case , the eigen energy @xmath16 since @xmath17 has the dimensions of energy we can represent this factor by @xmath18 , then @xmath19 we now determine the number of states @xmath20 with energy less than a given value @xmath21 . for energies large compared to @xmath18
, we may treat @xmath22 s as continuous variables and neglect the ground state energy . to calculate @xmath20
, we introduce a coordinate system in terms of the three variables @xmath23 . in this coordinate system @xmath24
defines a surface of constant energy @xmath21 .
then @xmath20 is proportional to the volume in the first octant bounded by the surface @xmath25 to evaluate the integral we use the relation @xmath26 , where @xmath27 @xcite .
then the density of states @xmath28 this expression for density of states is used to calculate @xmath0 . for bosons ,
the total number of particles occupying the excited states is @xmath29 at critical temperature @xmath30 in the case of bosons and @xmath31 is equal to the total number of bosons @xmath32 .
evaluating the integral gives @xmath0 in terms of the number of bosons @xmath33 ^{4/9}}.\ ] ] the corresponding expression for three dimensional isotropic harmonic oscillator potential is @xcite @xmath34^{1/3}}.\ ] ] from eq .
( [ eq.a.h ] ) and eq .
( [ eq.c.c ] ) , the ratio of the critical temperatures in the two potentials is @xmath35^{4/9 } } \nonumber \\ & = & 1.006n^{1/9}.\end{aligned}\ ] ] where we have taken @xmath36 to obtain the ratio .
the ratio is proportional to @xmath37 , which means that for @xmath38 atoms , @xmath0 in the case of 3d quartic potential trap is approximately six times higher than that of the 3d isotropic harmonic trap .
the expression of @xmath0 in the previous section is with the approximation that the ground state energy is zero , which is a valid approximation when the system has large number of bosons . for finite number of bosons ,
the zero point energy causes a change in the value of @xmath0 . for the 3d quartic potential ,
the ground state energy is @xcite @xmath39 this should be equal to the change in the chemical potential at the critical temperature , that is @xmath40 . as the total number of bosons
is fixed @xcite @xmath41 this implies @xmath42 using the expression of @xmath32 at temperatures slightly above @xmath0 @xmath43 this relation is obtained by substituting the general expression for the density of states i.e. @xmath44 in eq .
( [ eq.a.g ] ) . here
@xmath45 is a constant whose value depends on the form of the trapping potential . then from eq .
( [ eq.a.l ] ) and eq .
( [ eq.a.m ] ) we get @xmath46 in this expression @xmath47 should be greater than 2 , otherwise the relation is not valid since @xmath48 diverges . using this expression , the change in the critical temperature due to
the finite particle number is @xmath49 in the case of 3d quartic potential @xmath47 is equal to @xmath50 . then @xmath51 a relative measure of the effect of zero point energy on the critical temperature is the fractional change of the critical temperature .
it is the ratio between @xmath52 and @xmath0 , for the present case @xmath53 noting that @xmath54 is equivalent to @xmath55 we get @xmath56 @xmath57 for the 3d isotropic harmonic potential , the fractional change of the critical temperature is @xmath58 if we compare eq .
( [ eq.a.s ] ) and eq .
( [ eq.c.a ] ) , we find that the percentage decrease in @xmath0 is larger in the case of 3d quartic potential trap for number of particles @xmath59 . but the scenario is reversed for number of particles @xmath60 .
this is also evident from fig .
( [ fig.2 ] ) where cross over point corresponds to the number of particles @xmath61 .
consider the general form of the quartic potential , as mentioned earlier @xmath62 in optical traps , it is possible to create confining potentials which are approximately close to this form but a truly spherically symmetric one is not realizable .
the difficulty is in producing the cross terms of the potential , for example , terms like @xmath63 in cartesian coordinate representation .
the absence and presence of the cross terms in quadratic and quartic potentials respectively introduce a key difference between the dynamics in the two potentials . in absence of the cross terms like @xmath64 , in quadratic potential , a perturbation to the dynamics of a particle along an axis remains confined along that axis . in contrast , it propagates to other axes in the case of quartic potential . for condensates in traps ,
an important parameter which reflects the effects of these terms is the critical temperature .
semiclassically , total number of states available to the system can be obtained by dividing the total phase space volume by @xmath65 , the volume of a single state @xmath66 in the above equation @xmath67 and @xmath68 are radial coordinate and momentum corresponding to the classical turning point respectively . transforming the variable of integration from @xmath69 to @xmath21 ( using the relation @xmath70 ) we get @xmath71 thus the density of states is @xmath72 substituting @xmath73 we can transform the integral into a form which can be evaluated analytically @xcite @xmath74 using this expression for the density of states in eq .
( [ eq.a.g ] ) we get @xmath75^{4/9 } } \left [ \frac{h^3\lambda^{3/4}\gamma(9/4 ) } { \gamma(3/2)\gamma(3/4)4\sqrt{2}\pi^2m^{3/2}}\right]^{4/9}.\ ] ] comparing with eq .
( [ eq.a.h ] ) we can obtain the ratio of @xmath0 in the two cases , with and without the cross terms , for the 3d isotropic quartic potential .
it is found that @xmath0 with the cross terms is 1.2 times higher .
this rise in @xmath0 can be attributed to the contribution from the cross terms which were neglected while deriving eq .
( [ eq.a.h ] ) .
the reason for the difference is , when @xmath76 then @xmath0 varies as @xmath77 .
hence , the lower @xmath0 in 3d quartic potentials without the cross terms is due to the higher value of @xmath45 . in the 2d case , neglecting the cross terms , the potential is of the form @xmath78 using the same approach as adopted in the 3d case , the total number of states available to the system is @xmath79 thus the density of states @xmath80 substituting this expression of @xmath81 in eq .
( [ eq.a.g ] ) we get @xmath82^{2/3 } } .\ ] ] the corresponding expression when cross terms are considered is @xcite @xmath83^{2/3}.\ ] ] comparing eq .
( [ eq.b.f ] ) and eq .
( [ eq.b.g ] ) we find that @xmath0 with the cross terms in eq .
( [ eq.b.g ] ) is approx .
1.12 times higher than @xmath0 without the cross terms in eq .
( [ eq.b.f ] ) .
thus the cross terms increase @xmath0 in 2d as well as 3d case .
our calculations show that @xmath0 in the case of the 3d quartic potential trap is higher than that of the isotropic harmonic potential trap .
this is due to the form of the density of states @xmath81 , which varies as @xmath84 and @xmath85 in 3d isotropic quartic and quadratic trapping potentials respectively .
this implies lower density of states in quartic oscillator potential compared to isotropic harmonic oscillator potential . however , more interesting is the effect of the cross terms . in the 3d isotropic
harmonic potential trap the cross terms are absent , which is not the case for the 3d quartic potential trap .
the cross terms tend to decrease the density of states and raise @xmath0 .
these terms increase @xmath0 by factor of 1.2 and 1.1 in the 3d and 2d quartic trap potentials respectively as compared to the case without the cross terms .
experimentally , in optical traps , the potentials without the cross terms are more appropriate .
we find that the effect of finite particle number is more pronounced in the 3d quartic potential when the number of particles is @xmath86 .
the situation is reversed when the number of particles is @xmath3 .
the cause of the reversal lies in the form of the fractional change @xmath1 for the two potentials .
the ratio of the fractional change between 3d isotropic quartic potential to harmonic potential is @xmath87 .
it is @xmath88 for @xmath89 , @xmath90 for @xmath91 and @xmath92 for @xmath93 .
thus , when @xmath94 the constant factor is dominant in eq .
( [ eq.a.s ] ) and is responsible for the larger value of @xmath1 in quartic potential .
but when @xmath95 , @xmath96 dominates and @xmath1 of the quartic potential is lower than that of the harmonic potential .
99 m. greiner , o. mandel , t. esslinger , t. w. hnsch and i. bloch , nature * 415 * , 39 ( 2002 ) .
o. gygi , h. g. katzgraber , m. troyer , s. wessel and g. g. batrouni , phys .
a * 73 * , 063606 ( 2006 ) .
a. l. fetter , phys .
a. * 64 * , 063608 ( 2001 ) .
t. k. ghosh , phys .
a. * 69 * , 043606 ( 2004 ) .
j. kim and a. l. fetter , phys .
a. * 72 * , 023619 ( 2005 ) .
i. danaila , phys .
a. * 72 * , 013605 ( 2005 ) .
a. l. fetter , b. jackson and s. stringari , phys .
a. * 71 * , 013605 ( 2005 ) .
s. bargi , g. m. kavoulakis and s. m. reimann , phys .
a. * 73 * , 033613 ( 2006 ) .
s. kling and a. pelster , phys .
a * 76 * , 023609 ( 2007 ) .
v. bretin , s. stock , y. seurin and j. dalibard , phys .
. lett . * 92 * , 050403 ( 2004 ) .
v. bagnato and d. kleppner , phys .
a. * 44 * , 7439 ( 1991 ) .
p. m. mathews , m. seetharaman and s. raghavan , j. phys .
a * 15 * , 103 ( 1981 ) .
i. s. gradshteyn and i. m. ryzhik , _ table of integral series and products_. academic press , sixth edition , 2001 .
c. j. pethick and h. smith , _ bose - einstein condensation in dilute gases _ cambridge university press , 2002 .
|
the quartic confining potential has emerged as a key ingredient to obtain fast rotating vortices in bec as well as observation of quantum phase transitions in optical lattices .
we calculate the critical temperature @xmath0 of bosons at which normal to bec transition occurs for the quartic confining potential .
further more , we evaluate the effect of finite particle number on @xmath0 and find that @xmath1 is larger in quartic potential as compared to quadratic potential for number of particles @xmath2 .
interestingly , the situation is reversed if the number of particles is @xmath3 .
|
observational and theoretical studies of low - luminosity active galactic nuclei ( llagn ; agn with @xmath2 ) suggest that these objects differ substantially from higher - luminosity agn . at the low accretion rates typical of llagn , the inner part of the accretion disk
is thought to take the form of a geometrically thick , optically thin advection - dominated accretion flow ( adaf ) , with the standard thin accretion disk limited to a region exterior to the adaf .
if llagn are analogous to x - ray binaries , the radius of the transition between the inner adaf and outer disk should increase as the accretion rate declines @xcite , and the thin disk may disappear altogether in quiescent objects .
strong winds and outflows are expected from adafs ( e.g. * ? ? ?
* ) , and indeed llagn are radio - loud compared to the general agn population @xcite .
the spectral energy distribution ( sed ) has been particularly useful in expanding our understanding of llagn .
it has been instrumental , for instance , in the debate about whether llagn lack the `` big blue bump '' characteristic of the thin accretion disk in luminous agn ( e.g. * ? ? ?
? * ; * ? ? ?
the high - resolution data needed to separate the nuclear and host galaxy emission have been available for some time at radio , optical / ultraviolet and x - ray frequencies , but the first compilations of comparable infrared ( ir ) observations of substantial samples of llagn have only recently been published ( * ? ? ?
* ; * ? ? ?
* hereafter m12 ) .
the data show considerable variation among these objects .
however , when considered in terms of eddington ratio and radio loudness , they turn out to have some common characteristics . at low eddington ratios ( log @xmath3 ) ,
the ir - optical seds of radio - loud llagn ( e.g. m87 ) are relatively flat ( in @xmath4 ) and may be dominated by synchrotron radiation . at higher @xmath5 , the seds display a prominent peak in the mid - ir , and their ir - optical slopes tend to be within the range reported for higher - luminosity seyferts . the origin of the ir emission of llagn , and the process(es ) responsible for the seyfert - like ir peak in the higher-@xmath5 objects ,
are currently open questions .
several mechanisms may contribute : thermal emission from dust and/or the truncated accretion disk , and nonthermal emission from the jet and/or adaf .
the nuclear ir emission of radio - quiet seyferts and quasars comes at least partly from dust in the torus @xcite .
however , some models predict that the torus should disappear in llagn , either because of a lack of material reaching the central engine from the host galaxy , or because of a disk wind switching off " at low accretion rates @xcite .
the seyfert - like ir - optical seds of some llagn , and the observation that these objects lie on the mid - ir / x - ray correlation established for higher - luminosity agn , suggests that the torus could persist to fairly low luminosities ( * ? ? ? * m12 ) .
the strong silicate emission features observed in a number of llagn , though , may indicate comparatively large amounts of optically thin dust , as expected if little material is reaching or being expelled from the nucleus . while the accretion disk emission peaks in the uv in luminous agn @xcite , that of the truncated disk is expected to shift to longer wavelengths .
consequently the ir emission of certain llagn has been suggested to arise in this component @xcite , and several sets of adaf+disk+jet models predict a prominent peak from the disk between @xmath61 10 ( e.g. * ? ? ? * ; * ? ? ?
* ; * ? ? ?
those models are barely constrained in the ir , however , because at the time of their publication , little high - resolution photometry was available . )
points are indicated by open symbols .
the gray lines in each figure indicate 1 @xmath0 m and 20 @xmath0 m .
lower panels : as for upper panels , for ngc 3998 . ]
synchrotron emission from a jet may also play a rle in some objects .
mid - ir interferometric observations of the radio - loud llagn , centaurus a , show two components : a resolved , disk - like structure and an unresolved core @xcite . while the disk component , which accounts for 20 - 40% of the 813 @xmath0 m parsec - scale flux , is interpreted as dust heated by the agn ,
the de - reddened core is well fit by a synchrotron model that turns over in the mid - ir .
many llagn have now been modeled in terms of a jet , adaf and truncated thin disk .
the models often provide a good fit to the high - resolution radio , optical , uv and x - ray data and have been used to estimate quantities such as mass accretion rates in llagn ( nemmen et al .
2013 , submitted ; hereafter n13 ) . as a first step towards understanding the ir emission of these objects , in this letter
we compare a representative set of these existing adaf+disk+jet model fits with seds that include the newly - available high - resolution ir data .
of the 22 llagn in m12 , adaf+disk+jet models have been published for 14 @xcite .
we select seven for further study ( figures [ f1 ] [ f3 ] ) .
the remaining galaxies seds may contain a strong contribution from dust heated by nuclear star formation ( ngc 1097 ; * ? ? ?
* ) , are strongly affected by foreground extinction ( ngc 4261 ; * ? ? ?
* ) , or are very sparsely sampled and/or may contain significant host galaxy emission .
ngc 4258 is discussed by @xcite .
the most extensive and uniform sets of model fits are those of n13 and ( * ? ? ?
* hereafter y11 ) , so we use those for this comparison ( see * ? ? ?
* n13 for details of the underlying models ) .
the adaf lies within the transition / truncation radius , , outside which is found the truncated thin disk ( see figure 1 of n13 ) .
the adaf spectrum depends on the black hole mass , accretion rate m , and other factors such as the viscosity parameter @xmath7 and the fraction of the turbulent dissipation heating the electrons @xmath8 .
the accretion rate varies with radius , with denoting the accretion rate at . the mass loss rate in the jet , m@xmath9 , is required to be a reasonable fraction of the accretion rate in the adaf s inner region ( y11 ) , or not to exceed ( n13 ) .
the thin disk emits locally as a blackbody and is described by ( estimated from stellar velocity dispersions or gas kinematics ) and the inclination angle @xmath10 ( often constrained by radio jet observations , modelling of the broad h@xmath7 line or other measurements ) , along with and . and are correlated , but the main effect of is to determine the spectral cutoff of the disk emission ( smaller implying a higher - frequency cutoff ) .
has little effect on the adaf sed , as most of the adaf emission comes from the inner region of the flow .
y11 make initial estimates of @xmath11 and m@xmath12 , then combine these values with the other parameters ( @xmath13 , @xmath7 , @xmath8 , etc . ) , which are fixed , to derive the adaf solution and sed .
the disk spectrum is calculated using these initial values of @xmath11 and m@xmath12 , the jet sed computed , and the sum of the emission of these components compared to the observed sed .
the input parameters are adjusted until a model is obtained that is roughly consistent with the data .
the radio and/or x - ray data constrain m@xmath12 , which is correlated with .
n13 first fit a jet model to the radio data , then fit the truncated disk to the optical photometry to obtain @xmath11 and m@xmath12 .
the x - ray data are then used to refine the values of @xmath11 and m@xmath12 , estimate @xmath8 , and derive the adaf model .
the emission from the various components is then summed and compared to the overall broad - band sed . both n13 and y11 refer to two families of models : adaf - dominated " ( ad ) and jet - dominated " ( jd ) , depending on which is the dominant contributor to the x - ray emission .
the models are generally fit only to radio , optical , uv and x - ray data ; existing low - resolution ir points are treated as upper limits . figures [ f1 ] - [ f3 ] show the comparison between the observed seds ( from m12 and references therein , plus herschel far - ir data for ngc 3031 from * ? ? ?
* ) and published adaf+disk+jet models for these objects .
the m12 seds rely heavily on those of @xcite , with the addition of high - resolution ir data obtained from ( near-)diffraction - limited mir , adaptive optics and hst imaging . outside the ir region
there is a large degree of overlap between the m12 seds and those fitted by n13 and y11 , but they are not identical .
the n13 optical / uv data use the x - ray - based extinction corrections of @xcite while those of m12 and y11 are not corrected for extinction , and the x - ray ( and some other ) points chosen by each set of authors are not the same in all cases . in view of the heterogeneity of the data collections and extinction corrections ,
our focus here is simply to compare the model fits and seds in the ir region , rather than to re - examine in detail the model results over all frequencies .
the reader is encouraged to consult y11 and n13 to see the model fits to the original seds .
y11 give their preferred ( ad or jd ) model for each nucleus , whereas n13 show both ad and jd versions unless one results in physically unreasonable parameters .
where possible , then , we show one ad and one jd model for each object .
the comparison between the models and nuclear ir data shows that : \1 ) the models are usually consistent with the nuclear ir data , in that they do not predict stronger emission than is observed .
ngc 4579 may be an exception .
although its sed lacks the high - resolution 1 3 @xmath0 m data needed to make a definitive statement , the truncated disk in the n13 ad model ( fig .
[ f2 ] ) will produce too much nir emission unless the spectral shape of this object is highly unusual .
similarly , @xcite fit an ad model to ngc 4258 and find that the disk is over - luminous in the nir .
conversely , the models never account for _ all _ of the emission over the whole 1 20 @xmath0 m range .
there is always room for another component to contribute , particularly in the mid - ir .
\2 ) the spectral shape of the truncated disk emission does not bear a good resemblance to the ir seds , tending to be much narrower than the broad ir peaks in the observations .
it is conceivable that the disk and adaf emission could combine so as to reproduce the broad ir peak in the seds , but this is unlikely to happen in every object . in any case
the luminosity of the truncated disk component in most of the models is much lower than the observed ir luminosity .
\3 ) the adaf synchrotron emission , which peaks at @xmath14 hz , is accompanied by a compton scattering bump at @xmath15 hz .
these two components usually bracket the nuclear ir emission .
\4 ) the contribution of the combined adaf+disk+jet to the observed ir emission is rather model - dependent . in ngc 3031 the y11 jd model accounts for very little of the ir luminosity , whereas the n13 ad model predicts that most of the emission at @xmath16 m comes from the truncated disk .
the ad models tend to predict an sed whose luminosity increases sharply towards optical wavelengths , while the jd models are flatter .
however , different models from the same family can also give very different results . the n13 ad model for ngc 4579
predicts strong nir emission from the truncated disk , but the disk emission in the y11 ad model peaks at optical wavelengths . in both models
is consistent with constraints from the broad h@xmath7 line @xcite .
, for ngc 4486 and ngc 4579 .
the truncated disk spectrum for ngc 4486 is not given by n13 as its luminosity is negligible .
y11 and n13 do not present jd models for ngc 4579 , so both ad models are shown here . ]
perhaps the most notable aspect of the comparison between the adaf+disk+jet model fits and high - resolution seds is the presence of an ir `` excess '' above the predictions of the models . except in one or two cases ,
the truncated disk is not luminous enough to account for the ir emission even over a limited range of wavelengths .
the fits presented in y11 and n13 are not unique , and it may be possible to identify adaf models that fit the radio and x - ray emission while allowing a more luminous disk component . however , the spectral shape of the disk emission remains a poor match to the overall 1 20 @xmath0 m data .
we conclude that the truncated accretion disk is unlikely to be responsible for the ir `` bump '' in llagn , contrary to some recent suggestions ( e.g. * ? ? ?
* ; * ? ? ?
* ) . the contribution of the adaf can be significant around @xmath17 m , but it drops off sharply towards longer wavelengths in almost all of the models .
the peak frequency of the adaf synchrotron emission depends on and @xmath18 @xcite , and occurs at @xmath19 hz for these llagn .
the characteristic compton boost parameter is such that the first peak due to the upscattering of seed synchrotron photons is typically around @xmath20 hz .
the synchrotron and compton peaks bracket the 1 - 20 @xmath0 m ( @xmath21 hz ) observations , disfavoring an adaf origin for the nuclear ir emission .
the models do include synchrotron radiation from a jet , but the jet generally does not account for all of the ir emission . in the y11 model for the radio galaxy ngc 1052
, the jet dominates the emission at @xmath17 m but is negligible at longer ir wavelengths .
however , @xcite point out that the optical ir emission of that object resembles a power - law , and are able to fit the entire radio
uv sed ( not corrected for extinction ) with a simple self - absorbed synchrotron spectrum .
this may suggest that the nuclear ir data of radio - loud objects should be included when fitting their seds with the adaf+disk+jet models , potentially leading to very different derived model parameters .
a number of the llagn studied here , including ngc 1052 , have remarkably strong silicate emission features in their spectra ( m12 ) .
many also have ir sed slopes within the range spanned by `` conventional '' seyferts ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the silicate features are an unambiguous indicator of dust , and their strength in these llagn suggests that they arise in dust heated by the agn rather than in dust associated with circumstellar shells in the host galaxy ( c.f . * ? ? ?
these observations imply that dust must contribute at some level to the nuclear ir emission of llagn . in figure [ f4 ]
we show two dusty model fits to the ir sed of ngc 3998 : optically thin dust emission produced with the dusty radiative transfer code @xcite , and a clumpy torus spectrum obtained using the models of @xcite and the bayesclumpy fitting tool @xcite .
clumpy torus models have been successful in fitting the seds of numerous seyfert galaxies ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , while the presence of the strong silicate emission features in this object ( * ? ? ?
* m12 ) suggests the use of the optically thin dust model . , for ngc 4594 , plus two objects , ngc 1052 and ngc 4374 , for which only a single adaf+disk+jet model has been published .
the truncated disk spectrum for ngc 4374 and ngc 4594 is not given by n13 as its luminosity is negligible .
the y11 model for ngc 1052 is neither ad nor jd . ] in the torus model in figure [ f4 ] , the bolometric luminosity required to account for the ir emission of ngc 3998 , @xmath22 , is comparable to that estimated from the x - rays ( @xmath23 , assuming the bolometric correction factor of * ? ? ?
the @xmath24 implied by the visual extinction through the clumpy torus , @xmath25 mag , is quite close to the observed @xmath26 @xcite , and the escape probability of an agn - produced optical photon , @xmath27 , is within the range spanned by the type 1 objects in @xcite .
the torus models can produce a wide range of sed shapes and the photometry alone does not restrict the parameter space to a high degree .
for that reason , the spitzer spectrum was included in the fits .
although the angular resolution of the spectroscopy is a factor of @xmath2810 lower than that of the ground - based photometry , its spectral shape matches the photometric data well . with the constraints imposed by the spectroscopy
, the torus model does predict silicate features in emission , although not as strongly as observed .
the torus appears under - luminous at @xmath29 m .
given the uncertainty in the luminosity of the adaf+disk+jet in the near - ir ( [ models ] ) , it is not yet clear whether the torus model is deficient in hot dust emission , or whether the observed emission at these wavelengths contains a contribution from sources other than dust . in the adaf+disk+jet models considered for ngc 3031
, the largest contribution to the observed ir emission arises in the y11 model around 3 4 @xmath0 m , where the model would account for @xmath30% of the observed luminosity . at longer wavelengths the adaf+disk+jet contribution would be within the photometric errors .
we have subtracted the y11 model from the data and refit the residual emission , finding only minor changes to the derived parameters ( @xmath22 , @xmath31 mag , @xmath32 ) .
the basic features of the optically thin dust model are described in @xcite .
briefly , the dust is distributed in a spherical shell around a broken power law agn spectrum , from a minimum radius corresponding to the dust sublimation temperature ( 1500 k ) .
dust that is optically thin in the mid - ir naturally produces silicate features in emission , so the spitzer irs spectrum was not included in these fits .
the best - fitting model ( figure [ f4 ] ) provides a reasonable description of the sed of ngc 3998 , and the strength of the silicate emission features is comparable to those observed in the spitzer spectrum .
more sophisticated treatment of grain chemistry / size / structure and/or radiative transfer effects may be necessary to improve the match between the observed and modelled feature profiles ( e.g. * ? ? ?
* ; * ? ? ?
the v - band optical depth of the model , @xmath33 , means that the agn is heavily obscured in the optical .
this is not consistent with the detection of broad @xmath34 emission in this object @xcite .
a toroidal dust distribution , not considered in this simple model , would allow direct views of the broad line region without significantly affecting the reprocessed spectrum .
more in - depth modeling of this and other nuclei will certainly be necessary for a full understanding of the ir emission mechanisms of llagn .
if confirmed , optically thin dust emission would be consistent with a `` disappearing torus '' scenario in which clouds forming in the accretion disk wind of a low - luminosity agn do not become sufficiently optically thick to remain coherent and form an obscuring torus @xcite .
additional high - resolution near - ir imaging and mid - ir spectroscopy would allow firmer constraints to be placed on the dusty models , and ir polarimetry ( with mmt - pol , for example ; * ? ? ?
* ) may help disentangle the contributions of dust and jet emission to the ir luminosity of radio - loud llagn .
however , on the basis of the ir excess " above the adaf+disk+jet model fits examined here , the mismatch between the shape of the observed ir sed and that of the adaf+disk+jet models , and the ability of dusty models to account for the ir sed , silicate emission and other characteristics of ngc 3998 , we suggest that dust contributes strongly to the emission of many low - luminosity agn between @xmath28 1 30 @xmath0 m .
the dashed line shows the clumpy torus model fit produced using the maximum a - posteriori values that represent the best fit to the data , while the shaded region denotes the range of models compatible with a 68% confidence interval .
the filled circles show the high - resolution photometric data points , while the red line is the spitzer spectrum of this object ( @xmath35 0.6 ) . ]
we thank the referee for a useful report that helped improve this work . supported by the gemini observatory , operated by the association of universities for research in astronomy , inc .
, on behalf of the international gemini partnership of argentina , australia , brazil , canada , chile , and the usa .
h . acknowledges support from the spanish plan nacional de astronoma y astrofsica under grant aya2009 - 05705-e , cra from pn - aya2010 - 21887-c04.04 . ,
a. , ramos almeida , c. , mason , r. , asensio ramos , a. , roche , p. f. , levenson , n. a. , elitzur , m. , packham , c. , rodrguez espinosa , j. m. , young , s. , daz - santos , t. , & prez - garca , a. m. 2011 , , 736 , 82 , g. j. , boselli , a. , dariush , a. , pohlen , m. , roussel , h. , sauvage , m. , smith , m. w. l. , wilson , c. d. , baes , m. , cooray , a. , clements , d. l. , cortese , l. , foyle , k. , galametz , m. , gomez , h. l. , lebouteiller , v. , lu , n. , madden , s. c. , mentuch , e. , ohalloran , b. , page , m. j. , remy , a. , schulz , b. , & spinoglio , l. 2012 , , 419 , 1833 , c. , jones , t. j. , warner , c. , krejny , m. , shenoy , d. , vonderharr , t. , lopez - rodriguez , e. , & dewahl , k. 2012 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol .
8446 , society of photo - optical instrumentation engineers ( spie ) conference series , e. , schweitzer , m. , lutz , d. , contursi , a. , genzel , r. , lehnert , m. d. , tacconi , l. j. , veilleux , s. , rupke , d. s. , kim , d .-
c . , sternberg , a. , maoz , d. , lord , s. , mazzarella , j. , & sanders , d. b. 2005 , , 629 , l21
|
we use recent high - resolution infrared ( ir ; 1 20 @xmath0 m ) photometry to examine the origin of the ir emission in low - luminosity agn ( llagn ) .
the data are compared with published model fits that describe the spectral energy distribution ( sed ) of llagn in terms of an advection - dominated accretion flow ( adaf ) , truncated thin accretion disk , and jet .
the truncated disk in these models is usually not luminous enough to explain the observed ir emission , and in all cases its spectral shape is much narrower than the broad ir peaks in the data .
synchrotron radiation from the jet appears to be important in very radio - loud nuclei , but the detection of strong silicate emission features in many objects indicates that dust must also contribute .
we investigate this point by fitting the ir sed of ngc 3998 using dusty torus and optically thin ( @xmath1 ) dust shell models .
while more detailed modeling is necessary , these initial results suggest that dust may account for the nuclear mid - ir emission of many llagn .
|
heterogeneous cellular networks with dense deployment of access points ( aps ) are anticipated to be a major component of 5 g networks .
a challenge with such dense deployments is interference management .
coordinated radio resource allocation across multiple cells is one approach for mitigating inter - cell interference . that includes joint scheduling across multiple cells over a fast timescale @xcite as well as the assignment of resources across cells over a slower timescale @xcite .
whereas the former approach requires instantaneous knowledge of channel gains , the latter approach relies on statistical knowledge of interference .
we consider the joint allocation of spectrum and user association in a large network with many aps .
the objective is to optimize a network utility , such as average delay , given traffic statistics over a geographic region that change slowly relative to channel fading .
our approach builds on our prior work @xcite in which for a network of @xmath0 aps and @xmath3 mobiles ( or user equipments ( ues ) ) , the spectrum is partitioned into @xmath7 _ patterns _ , corresponding to all possible subsets of active aps .
the problem is to optimize the widths of spectrum segments , associated with the different patterns , along with the user association under each pattern .
this has been shown to provide significant performance improvement in throughput enhancement and delay reduction @xcite .
the original convex problem formulation in @xcite is not useful for large networks because the number of patterns grows as @xmath1 . in prior work @xcite , we have reduced the number of variables to @xmath8 by recognizing that each link rate depends only on _ local _ patterns of active aps .
the problem can then be redefined over sets of overlapping _ interference neighborhoods _ , associated with those local patterns .
each ap has its own interference neighborhood , which captures the interference from nearby aps .
the challenge with the approach in @xcite is to ensure that the spectrum assigned to each particular ap is consistent across the neighborhoods to which it belongs . to accomplish that , the spectrum is discretized and a coloring algorithm is proposed to ensure that the local patterns of active aps are globally consistent .
here we take a different approach to addressing the scalability problem .
this is based on the fact that the solution is _ sparse _ , meaning that at most @xmath4 out of the @xmath7 possible patterns appear in the optimal allocation for a general network utility function .
hence we reformulate the problem by dividing the spectrum into @xmath4 segments , rather than @xmath7 , and attempt to identify the pattern that should be associated with each segment .
this effectively reverses the approach in @xcite , which attempts to assign a segment of spectrum to each pattern . in this reformulation
, we initially assume that any combination of patterns can be assigned to each of the @xmath4 segments .
this problem is a convex relaxation of our original problem .
the one - to - one mapping of spectrum segments to patterns is then enforced with an @xmath5 ( cardinality ) constraint .
an algorithm for finding an approximate solution to this problem is presented based on a reweighted @xmath6 approximation of this constraint @xcite .
the approach to scalability presented here has the advantage of eliminating the combinatorial coloring problem that arises in @xcite . instead
, a new @xmath5 constraint is introduced .
although this does not simplify the original problem , it helps in finding an approximate solution , since the reweighted @xmath6 approximations for the @xmath5 constraints are known to perform well .
numerical results indicate that this method generally gives better performance for a fixed computational complexity than the method in @xcite .
as in @xcite , we consider a network containing the set of aps @xmath9 and the set of ues @xmath10 .
each `` ue '' is actually associated with a particular location , and could refer to a group of nearby mobiles . the @xmath0 aps share @xmath11 hz of spectrum . for convenience ,
we normalize @xmath12 .
each ap can transmit on any part(s ) of the spectrum , which is homogeneous ( has the same distance attenuation ) .
aps sharing the same spectrum interfere at ues within range of both transmitters .
all transmissions are assumed to be omni - directional , although a similar set of problems can be reformulated with directional transmissions .
we define a subset of active ( transmitting ) aps @xmath13 as a _
pattern_. there are @xmath7 patterns , and a particular allocation of spectrum maps each pattern to a slice of spectrum .
we denote an allocation as @xmath14 , where @xmath15 denotes the amount of bandwidth assigned to pattern @xmath16 .
an example with three aps is depicted in fig .
[ fig : mod ] .
ap 1 owns @xmath17 exclusively ; shares @xmath18 with ap 2 ; shares @xmath19 with ap 3 ; and shares @xmath20 with both ap 2 and ap 3 .
user association is determined by how the spectrum assigned to a particular ap is allocated to different ues .
specifically , denote the fraction of total bandwidth used by ap @xmath21 to serve ue @xmath22 under pattern @xmath16 as @xmath23 , for @xmath24 .
ue @xmath22 is then assigned to ap @xmath21 if @xmath25 for some @xmath26 .
since the total bandwidth assigned to pattern @xmath16 is @xmath15 , we have @xmath27 the total bandwidth allocated to all patterns is then @xmath28 let @xmath29 denote the spectral efficiency of the link from ap @xmath21 to ue @xmath22 under pattern @xmath16 .
this is measured over a _
slow timescale _ , and is therefore an average over short - term fading .
the value of @xmath29 is therefore determined by the distance between ap @xmath21 and ue @xmath22 , shadowing , and similar long - term characteristics of the interference links from aps in @xmath16 to ue location @xmath22 . for concreteness
we assume @xmath30 where @xmath31 if @xmath24 and @xmath32 otherwise , @xmath33 is the transmit power spectral density ( psd ) at ap @xmath21 , @xmath34 is the power gain of link @xmath35 , and @xmath36 is the noise psd at ue @xmath22 .
we assume fixed , flat transmit psds over the slow timescale considered .
the factor @xmath37 , where @xmath38 is the average packet length ( bits ) , gives the units in packets / sec .
the link gain @xmath34 includes pathloss and shadowing effects .
clearly , @xmath39 if @xmath40 , i.e. , ap @xmath21 does not transmit on pattern @xmath16 . in practice
, the spectral efficiencies can be measured as time - averaged channel gains .
the total rate received by ue @xmath22 is therefore @xmath41
the problem is to maximize a network utility function over the spectrum allocation and user association designated by @xmath42 , @xmath43 : p0 [ eq : opt ] @xmath44 and constraints , , and , where the network utility function @xmath45 depends on the service rates to all ues .
the optimization problem is convex if @xmath45 is concave in @xmath46 $ ] .
[ line - through]*and element - wise non - decreasing . * as in @xcite , we will take @xmath45 to be the average packet delay , given by @xmath47where @xmath48 equals @xmath49 if @xmath50 and @xmath51 otherwise , and @xmath52 is the poisson packet arrival rate for ue @xmath22 .
this assumes exponential packet lengths and backlogged interference @xcite . solving [ eq : opt ]
becomes prohibitively expensive as the network size grows , due to the inherit complexity from the @xmath7 global patterns .
however , a key property of solution(s ) to is that at least one is _ sparse _
, i.e. , contains at most @xmath4 patterns .
( @xcite ) [ thm : spec ] [ eq : opt ] has a solution that divides the spectrum into at most @xmath4 segments , i.e. , @xmath53furthermore , if @xmath45 is element - wise nondecreasing in @xmath54 , @xmath4 in is reduced to @xmath3 .
determining which @xmath4 active patterns appear in a solution is then the key challenge in solving [ eq : opt ] .
we first reduce the number of variables in from @xmath1 to @xmath8 by reformulating over _ interference neighborhoods _ based on local patterns @xcite . due to pathloss , we assume interference vanishes beyond a certain distance .
let @xmath55 denote the set of links with nonzero gains .
hence each ue only receives power from aps within its neighborhood : @xmath56 from the ap side , each ap can only transmit to a collection of ues in a ap neighborhood with positive rates : @xmath57 we define the interference neighborhood for each ap @xmath21 as : @xmath58 i.e. , @xmath59 includes ap @xmath21 and all aps that interfere with it .
an example with three aps ( denoted by @xmath60 ) and 2 ues ( denoted by @xmath61 ) is shown in fig [ fig : localpattern ] .
the set of nonzero links are @xmath62 .
the ap neighborhoods are @xmath63 , @xmath64 , and @xmath65 ; and the ue neighborhoods are @xmath66 and @xmath67 . ap 1
s interference neighborhood is @xmath68 , as ap 2 interferes with it at ue @xmath69 .
ap 3 s interference neighborhood is @xmath70 , as ap 2 interferes with it at ue @xmath71 .
ap 2 s interference neighborhood is @xmath72 , since ap 1 and ap 3 interfere with ap 2 at ue @xmath69 and ue @xmath71 , respectively .
the spectral efficiency of link @xmath35 only depends on the local patterns in @xmath73 , according to the definition of ue neighborhoods in . therefore , for any link @xmath35 , the spectral efficiency under a global pattern @xmath16 is equivalent to its intersection with ue @xmath22
s neighborhood @xmath73 : @xmath74 we next express each rate in terms of local patterns .
we define bandwidth allocation variables within a local interference neighborhood @xmath75 as : @xmath76 [ line - through]*if we treat an interference cluster @xmath75 as a partition of all the global patterns , many global patterns are equivalent .
* because all global patterns sharing the same overlap with @xmath75 contribute to the same local pattern of ap @xmath21 , @xmath77 represents the bandwidth allocated to link @xmath35 under local pattern @xmath78 .
hence from ap @xmath21 s perspective , the bandwidth assigned to a local pattern @xmath78 must be the sum of the bandwidths assigned to all global patterns containing @xmath78 . the service rate of link @xmath35 , as defined in ,
can then be calculated as @xcite : @xmath79where we have used and , and the fact that only aps in @xmath73 transmit to ue @xmath22 with positive rate . motivated by proposition [ thm : spec ] , we reformulate by dividing the spectrum into @xmath4 segments , and seek to assign a single pattern to each segment .
proposition [ thm : spec ] implies that by optimizing this assignment we obtain a solution to [ eq : opt ] .
[ line - through]*the number of local patterns in each local interference cluster is bounded despite the size of the network .
the general idea is provided here .
* let @xmath80 denote the bandwidth of the @xmath81th segment ( to be optimized ) .
we now associate a set of variables @xmath82 ( for local neighborhoods ) and @xmath83 _ for each segment _ @xmath81 .
we can therefore rewrite the rate in as @xmath84 where the first sum is over the @xmath4 spectrum segments .
apart from the addition of segment index , the @xmath82 variable is the same local bandwidth allocation variable defined in .
the amount of bandwidth assigned to local pattern @xmath78 in ap @xmath21 s interference neighborhood @xmath75 within segment @xmath81 is @xmath85 the total amount of spectrum assigned to @xmath75 satisfies @xmath86 we also introduce the following _ consistency constraint _ to ensure that the amount of spectrum allocated to an ap is consistent across any two neighborhoods @xmath75 and @xmath87 that contain it @xcite .
that is , for any nonempty @xmath88 , @xmath89 see , for example , @xmath90 and @xmath91 in fig .
[ fig : localpattern ] . in interference cluster @xmath90 , ap 2 transmits under pattern @xmath92 and @xmath93 ; in interference cluster @xmath91 , ap 2 transmits under pattern @xmath92 and @xmath94 . the total bandwidth used by ap 2 must be consistent across neighborhoods so that @xmath95 , where @xmath92 is the overlapping pattern .
this example can be extended to a _ set _ of aps , which are members of two interference neighborhoods , giving . to ensure a one - to - one mapping of patterns to the @xmath4 segments
, we add the @xmath5-norm constraint @xmath96 where @xmath97 if @xmath98 , and @xmath99 if @xmath100 .
that is , we constrain each ap to use at most one active pattern in each segment .
we can now reformulate [ eq : opt ] in terms of the local interference variables @xmath101 across the @xmath4 spectrum segments : p1 [ eq : opt3 ] @xmath102 and constraints - for each @xmath103 .
[ thm : main ] [ eq : opt ] and [ eq : opt3 ] are equivalent ( have the same solutions ) given the local neighborhood definitions - .
the proof is omitted due to limited space . hence a solution to [ eq : opt3 ] always satisfies proposition [ thm : spec ] .
this reformulation leads to a computationally efficient approximation algorithm .
the number of variables is reduced from @xmath1 in [ eq : opt ] to @xmath2 in [ eq : opt3 ] .
however , the @xmath5 norm constraint makes the problem non - convex and difficult to solve .
we apply the reweighted @xmath6 approach , described in @xcite , to approximate the @xmath5 constraint .
specifically , the @xmath5 norm is approximated by the weighted @xmath6 norm , @xmath104 , where @xmath105 is iteratively adapted , and taken to be inversely proportional to @xmath106 computed at the preceding iteration .
this has the effect of suppressing small nonzero entries with large weights .
the @xmath6 approximation can not be directly applied to [ eq : opt3 ] since the @xmath5 norm constraints are coupled through and . hence we present an iterative algorithm based on the @xmath6 reweighted heuristic , where the weights depend on both the @xmath6 norm and the bandwidth of each segment @xmath80 . in each iteration of the algorithm , shown in algorithm [ alg : l1 ]
, we solve [ eq : opt3 ] , but with the @xmath5 norm constraint replaced by the weighted sum @xmath107 in algorithm [ alg : l1 ] , we refer to this reweighted @xmath6 version of [ eq : opt3 ] as p2 .
algorithm [ alg : l1 ] shows a random initialization of the weights to introduce asymmetry in the first iteration .
otherwise , e.g. , if all @xmath108 s are initialized to a constant , the solution stays symmetric over all segments , because the constraints and objectives are identical for each segment . a symmetric solution , i.e. , @xmath109 , [ line - through]*is obviously not optimal for an arbitrary network topology*is generally not a solution to the original problem .
in each iteration , we solve p2 with the current weights @xmath110 to obtain the current @xmath111 and @xmath112 $ ]
. then the weights are updated as shown .
the iterations terminate when the variables converge or the maximum number of iterations @xmath113 is reached . the weight update in algorithm [ alg : l1 ]
is obtained by approximating the @xmath5 norm with @xmath114 for small @xmath115 ( see @xcite and the references therein ) .
adding @xmath115 in the denominator allows small components to be propagated to the next iteration . here
the goal is to obtain a _
nonzero @xmath116 , so we take @xmath117 .
this is different from the fixed @xmath115 proposed in @xcite in that @xmath118 changes with @xmath80 in each iteration .
that is , algorithm [ alg : l1 ] simultaneously searches for the optimal reuse pattern and its associated bandwidth .
it is possible that algorithm [ alg : l1 ] produces multiple reuse patterns for a segment at termination . in those cases
ap @xmath21 chooses the dominant pattern @xmath119 . the global reuse pattern assigned to segment @xmath81
is then given by @xmath120 .
the spectral efficiencies for each segment are determined by the corresponding global reuse pattern ( set of interfering aps ) : @xmath121 . given the assignment of patterns to each spectrum segment , the widths of the segments , @xmath122 along with the assignment of spectrum to mobiles , @xmath123 , can then be re - optimized by solving the convex problem : p3 [ eq : opt7 ] @xmath124 @xmath125 and for @xmath103 .
solve p2 with the current weights @xmath110 .
2 . update @xmath126 .
3 . @xmath127 .
the simulation results assume one macro ap is located at the center of the area , and the remaining @xmath128 pico aps are randomly dropped around it .
the @xmath3 ues are placed on a rectangular lattice .
link gains include both pathloss and shadowing .
additional simulation parameters are shown in the footnote.w / hz , pico transmit psd is 1 @xmath129w / hz , noise psd is @xmath130 @xmath129w / hz , total bandwidth is 20 mhz , and average packet length is 1 mb . ] we first compare solutions to [ eq : opt ] and [ eq : opt3 ] for a small network with @xmath131 and @xmath132 .
since the number of variables is relatively small , we solve both versions of [ eq : opt ] with and without the local neighborhood approximation using a standard convex optimization solver .
the local neighborhoods are constructed by including the strongest four aps for each ue .
the solution to [ eq : opt3 ] is obtained using algorithm [ alg : l1 ] .
we compare those with full spectrum reuse where each user is assigned to the ap with the strongest signal ( maxrsrp association ) , and also the optimal orthogonal allocation , i.e. , only @xmath133 are active . fig . [
fig : delay - n10k23 ] shows delay versus traffic arrival rate for all schemes .
the end of each curve represents the maximum arrival rate the scheme can support .
the curves obtained by solving [ eq : opt ] with and without local neighborhood approximation are very close , which indicates considering the four strongest interferers is enough in such a small network .
the solution to [ eq : opt3 ] incurs slightly larger delay than the solution to [ eq : opt ] .
the jointly optimized spectrum allocations and user associations achieve substantial delay reduction as well as eight times throughput compared to full frequency reuse with maxrsrp association .
the optimized spectrum allocation and user associations obtained solving [ eq : opt3 ] are depicted in fig .
[ fig : sol3 ] .
the macro and pico aps are represented by the bigger and smaller towers ; each handset represents a ue .
each solid line shows a connection between the corresponding ap and ue .
the grid for each ue shows the spectrum used to serve that ue . the traffic arrival rate ( in ( 0 , 100 ) ) for each ue
is shown under each grid .
the user association in fig .
[ fig : sol3 ] is close but not identical to that obtained by solving [ eq : opt ] ( not shown ) due to the approximation in algorithm [ alg : l1 ] , which explains the performance difference shown in fig .
[ fig : delay - n10k23 ] . and
@xmath134.,height=245 ] [ fig : sol3 ] fig .
[ fig : delay - n100k195 ] compares the performance of different allocation schemes in a large network with @xmath135 aps and @xmath136 ues . to facilitate the simulations ,
we reduce the size of each ue neighborhood from four to three , i.e. , each ue can only be served by the three strongest aps .
we compare full - spectrum - reuse with maxrsrp association , full - spectrum - reuse with optimized associations and the solution to [ eq : opt3 ] with 50 segments .
the solution to [ eq : opt3 ] achieves 1.5 times the throughput of the full - spectrum - reuse with optimized association .
the solutions to [ eq : opt3 ] use no more than 23 active patterns ( @xmath137 available segments ) . the throughput gain achieved by the solution to [ eq : opt3 ] in this large network is smaller than that in section [ subsec : small ] .
this is mainly because we only consider the three strongest interferers , which compromises the benefits from interference management . and
@xmath138.,height=226 ] [ fig : delay - n100k195 ]
an approach to joint allocation of spectrum with user association has been presented which exploits the sparsity of the optimal solution .
the proposed algorithm has been observed to achieve near - optimal performance for small to medium - size networks for which the optimal solution can be computed , and provides substantial gains relative to full frequency reuse .
although not considered here , the formulation can be extended to accommodate spatial selectivity and different power levels . the performance - complexity tradeoff for such extensions
is left for future work .
f. wang , l. song , z. han , q. zhao , and x. wang , `` joint scheduling and resource allocation for device - to - device underlay communication , '' in _ proc .
wireless comm . and
networking _ , pp . 134139 , ieee , 2013 .
b. zhuang , d. guo , e. wei , and m. l. honig , `` scalable spectrum allocation and user association in networks with many small cells , '' _
https://arxiv.org/abs/1701.03247 , _ submitted to _
ieee trans .
commun._. b. zhuang , d. guo , and m. l. honig , `` energy - efficient cell activation , user association , and spectrum allocation in heterogeneous networks , ''
_ ieee j. sel .
areas commun .
_ , vol .
pp , no .
99 , pp .
11 , 2016 .
|
joint allocation of spectrum and user association is considered for a large cellular network .
the objective is to optimize a network utility function such as average delay given traffic statistics collected over a slow timescale .
a key challenge is scalability : given @xmath0 access points ( aps ) , there are @xmath1 ways in which the aps can share the spectrum .
the number of variables is reduced from @xmath1 to @xmath2 , where @xmath3 is the number of users , by optimizing over local overlapping neighborhoods , defined by interference conditions , and by exploiting the existence of sparse solutions in which the spectrum is divided into @xmath4 segments .
we reformulate the problem by optimizing the assignment of subsets of active aps to those segments .
an @xmath5 constraint enforces a one - to - one mapping of subsets to spectrum , and an iterative ( reweighted @xmath6 ) algorithm is used to find an approximate solution .
numerical results for a network with 100 aps serving several hundred users show the proposed method achieves a substantial increase in total throughput relative to benchmark schemes .
|
viscous discs can acquire large dimensions under the effect of angular momentum redistribution .
this is corroborated by the observation of young stellar objects ( yso ) which reveals the emission of wide gaseous discs with an outer radius reaching @xmath6 au ( beckwith et al .
, 1990 ; pudritz et al . ,
1996 ; duvert et al . , 1998 ; guilloteau & dutrey , 1998 ; shepherd & kurtz , 1999 ) . also ,
discs hosted by active galactic nuclei ( agn ) , although yet unresolved |except in the active galaxy ngc4258 ( herrnstein et al . , 1999)| could probably be of much larger size .
the existence of strong similarities between ysos and agns , like jets and outflows ( falcke , 1998 ; frank , 1998 ) , some excesses in the spectral energy distribution at infrared wavelengths ( adams & shu , 1986 ; zdziarski , 1986 ; bertout , 1989 ; sanders et al . , 1989 ; voit 1991 ; kenyon , yi & hartmann , 1996 ) , a surrounding ( dusty ) torus ( gsten , chini & neckel , 1984 ; th & molster , 1994 ; drinkwater , combes & wiklind , 1996 ; sandqvist , 1999 ) indicates that the same physical mechanisms should take place in discs , despite the difference in the central mass scale .
this constitutes an interesting challenge for the disc theory , especially to understand the transport of angular momentum .
many models are based on the @xmath0-theory of thin discs ( shakura & sunyaev , 1973 ; pringle , 1981 ) and assume a vertically averaged structure ( e.g. , collin & dumont 1990 ; ruden & pollack , 1991 ; cannizzo & reiff , 1992 ; hur et al .
1994a ; artemova et al . 1996
; burderi , king & szuszkiewicz , 1998 ; drouart et al . 1999 ; aikawa et al . , 1999 ) .
a physically more satisfactory approach is the investigation of the vertical structure in detail , as done for stars .
such a problem has been discussed by several groups already , in different contexts , with various goals and degrees of sophistication : at the scale of agns ( cannizzo , 1992 ; wehrse , strzer & shaviv , 1993 ; drrer et al . , 1996 ; siemiginowska , czerny & kostyunin , 1996 ; hubeny & hubeny , 1998 and references therein ; sincell & krolik , 1997 ; raska et al . , 1999 ) , in cataclysmic variables ( cvs ) ( smak , 1984 ; mineshige & osaki , 1983 ; meyer & meyer - hofmeister , 1982 ; pojmaski , 1986 ; see cannizzo , 1993 for a review ; milsom , chen & taam ; 1994 ; dubus et al . , 1999 ) and in ysos ( malbet & bertout , 1991 ; bell et al . , 1997 ; dalessio et al .
, 1998 ) including the primitive solar nebula ( psn ) ( lin & papaloizou , 1980 ; papaloizou & terquem , 1999 ) .
basic considerations show that the outermost regions of ( low mass ) discs are expected to be regulated by self - gravity ( goldreich & lynden - bell , 1965 ; shlosman & begelman , 1987 )
. this may be the case of agn and yso discs .
a few constraints on the disc thickness , mass and rotation motion at large radii are set by observations ( guilloteau & dutrey , 1998 ; mundy , looney & welch , 2000 ; herrnstein et al . , 1999 ) .
so , it appears essential to construct models for as accurate and realistic as possible , despite the lack of knowledge regarding turbulent viscosity which causes accretion . to our knowledge ,
no self - consistent 2d - model accounting for vertical self - gravity in the outermost parts of discs has been published yet .
this is the aim of this paper .
the present model includes simultaneously vertical convection , self - gravitation , turbulent pressure , and realistic equation of state ( eos ) and opacities . of particular interest here
is the transition from the `` classical '' disc to the self - gravitating disc which is predicted to be as close as a few @xmath7 ( @xmath8 is the radius of the central object ) by vertically averaged models ( ruden & pollack , 1991 ; dalessio , calvet & hartmann , 1997 ; hur , 1998 ) .
the hypothesis of the model and relevant equations are developed in sect .
two versions of the @xmath0-prescription are tested .
the ingredients ( eos and opacities ) as well as the numerical method are presented in sect .
3 . we discuss in sect .
4 a few properties of the model , namely the effect of turbulent pressure and depth dependent viscosity , and the position of the inner edge of the self - gravitating disc , for prototypal circumstellar and agn discs , for various accretion rates . an appendix contains a note concerning the treatment of convection and a high precision formula fitting the eos .
a distinctive feature of steady state keplerian accretion discs is the absence of coupling between the vertical structure and the radial structure ( e.g. frank , king & raine , 1992 ) .
this attractive property is undoubtedly an oversimplification and probably does not match reality .
however , it is particularly advantageous from a modeling point of view because any annulus can be treated individually , whatever the state of neighbors , unlike in thick discs where pressure gradients and energy transport in the radial direction gain in importance with respect to their vertical counterparts ( maraschi , reina & treves , 1976 ; abramowicz , calvani & nobili , 1980 ; abramowicz et al . , 1988 ;
narayan , madevan & quataert , 1998 ) .
note that the keplerian assumption which fixes the rotation law to the value @xmath9 where @xmath10 is the central mass and @xmath11 the polar radius , requires that the gas remains confined at altitudes @xmath12 such as @xmath13 , and small radial pressure gradients too . self - gravity may influence and even dominate the equilibrium structure and dynamical evolution of almost any kind of disc , not only in massive or thick discs or tori where effects are global ( bodo & curir , 1992 ; hashimoto , eriguchi , mller , 1995 ; boss , 1996 ; laughlin & ryczka , 1996 , masuda , nishida & eriguchi , 1998 ) , but also in low mass keplerian discs ( paczyski , 1978 ; shlosman & begelman , 1987 ) as soon as the mass density of the accreted gas exceeds @xmath14 locally , @xmath15 being the rotation frequency . in that latter case which is of interest here ,
self - gravity increases vertical pressure gradients and gathers matter closer to the midplane ( sakimoto & coroniti , 1981 ; shore & white , 1982 ; cannizzo & reiff , 1992 ; hur et al . , 1994 ; hur , 1998 ) .
so , only the outermost regions of keplerian discs where @xmath15 reaches low values can be affected by vertical self - gravity , except in very special situations ( e.g. sincell & krolik , 1997 ) .
accounting correctly for the disc gravity requires the resolution of the poisson equation ( hunter , 1963 ; strzer , 1993 ; bertin & lodato , 1999 ) @xmath16 where @xmath17 is the gas mass density and @xmath18 is the gravitational potential due to the bare disc ( the @xmath19-invariance is assumed ) .
this is a very difficult task , in particular because the disc surface has a non trivial form which is not known a priori and the deviation from sphericity is extreme . as eq.([eq : fish ] ) connects all annuli together , the method of `` independent rings '' no longer applies , unless an iterative scheme in which the potential would be step by step improved from the actual density field until convergence ( stahler , 1983 ) .
there is however no guarantee that such a scheme effectively converge and the computational time might be prohibitive ( eriguchi & mller , 1991 ) . here , we follow another , more simple approach : we adopt the infinite and @xmath11-homogeneous slab approximation ( paczynski , 1978 ; sakimoto & coroniti , 1981 ; shore & white , 1982 ; cannizzo & reiff , 1992 ; liu , xie & ji , 1994 ; hur , 1998 ) which yields the gravity due to the disc @xmath20 where the surface density @xmath21 is defined by @xmath22 but other assumptions are possible ( mineshige & umemura , 1997 ) .
this enables again to investigate the disc structure annulus by annulus , but introduces a bias : it tends to overestimate ( underestimate ) self - gravity in regions of high ( respectively low ) densities .
we expect that the present approximation gives acceptable results , at least in regions where the surface density radial gradients remain low with respect to density .
anyway , the error made with respect to a self - consistent model is not known and would be interesting to estimate .
when the upward transport of heat is treated in the diffusion approximation , the vertical structure of a keplerian disc , possibly irradiated , can be determined from the resolution of a system of four first order coupled ordinary differential equations ( odes ) in between the midplane and the top of the disc ( pojmaski , 1986 ; tuchman , mineshige & wheeler , 1990 ; cannizzo , 1992 ; meyer & meyer - hofmeister , 1982 ; milsom , chen & taam , 1994 ; dubus et al . , 1999 ) .
the complexity of the problem rises when a multi - frequency radiative transfer is performed , as required to compare theoretical spectra with observations and make key predictions ( ross , fabian & mineshige , 1992 ; wehrse , strzer & shaviv , 1993 ; drrer et al . , 1996 ; sincell & krolik , 1997 ; dalessio et al .
1998 ; el - khoury & wickramasinghe , 1998 ; hubeny & hubeny , 1998 ; de kool & wickramasinghe , 1999 ) .
the four equations specify respectively ( frank , king & raine , 1992 ) : * the pressure gradient @xmath23 which describes the hydrostatic equilibrium of each fictitious slab @xmath24 * the heat flux gradient @xmath25 due to viscous heating @xmath26 where @xmath27 is the @xmath12-dependent viscosity law ( see sect .
2.4 ) , * the temperature gradient @xmath28 which is determined by the heat net flux transported upwards through radiation and convection @xmath29 where @xmath30 is the pressure height scale and @xmath31 is the actual gradient ( see the appendix , sect .
a ) , * the surface density gradient @xmath32 @xmath33 note that this last equation is not relevant when self - gravity is left apart ( the total surface density of the disc + atmospheres @xmath34 can be easily computed a posteriori from eq.([eq : sigma ] ) , once the mass density distribution is known ) .
the above odes must be supplemented by a closure relation , an equation of state . for a mixture of radiation and perfect gas undergoing atomic ionization and molecular dissociation at lte
, the total pressure is linked to the density and temperature of matter by @xmath35 where @xmath36 is the mean mass per particle which depends on the density and temperature ( see sect . 3.1 and the appendix , sect .
the above expression for radiation pressure is compatible with an optically thick disc only .
turbulence is the main mechanism driving accretion in discs .
it is an extra source of pressure . according to the standard theory of @xmath0-discs ( shakura & sunyaev , 1973 ; pringle , 1981 ) , the typical velocity of turbulent eddies can be taken as @xmath37 if one assumes the equipartition between length and velocity turbulent scales ( @xmath38 is the adiabatic sound speed ) .
so , the turbulent pressure is @xmath39 where @xmath40 is the first adiabatic exponent in the total ( gas plus radiation ) pressure .
the effect of turbulent pressure on the hydrostatic equilibrium is therefore expected if @xmath0 is not too small , as simulations shall confirm .
if we include @xmath41 into eq.([eq : eh1 ] ) and rewrite it in terms of a density gradient equation , we find @xmath42 where @xmath43 and @xmath44 are respectively the temperature and density exponents of the total pressure , and we have assumed @xmath45 ( this is justified since @xmath46 , see for instance the appendix , fig.([fig_coefs ] ) ) .
this expression clearly shows the existence of a density inversion ( that is , a zone where @xmath47 ) each time the actual gradient satisfies @xmath48 ( raska et al . , 1999 ) .
such an inversion is likely to occur in radiative pressure dominated layers where @xmath43 is the largest or / and in convectively unstable zones where @xmath49 may be large .
note that the inversion still remains if turbulent pressure plays a role , but with a weaker amplitude .
it is likely that turbulent pressure plays a role , not only on the hydrostatic equilibrium as considered here , but also on advection of matter and energy both radially and vertically .
we ignore these effects .
in vertically averaged disc models , the anomalous viscosity takes locally a mean value defined from midplane quantities , namely @xmath50 ) . here , we need to specify how @xmath27 varies with the altitude .
this point is critical since our knowledge of turbulent viscosity remains extremely limited , if not null .
there are however two standard ways to do this within the framework of the @xmath0-prescription .
the most common one is to assume that the shear stress is proportional to some pressure ( the so - called `` @xmath51-formalism '' ; e.g. cannizzo , 1992 ) .
it leads to a viscosity law @xmath52 where @xmath53 can be either gas pressure , or total pressure or some combination of the two ( siemiginowska , czerny & kostyunin , 1996 ; artemova et al . , 1996 ;
hameury et al .
, 1998 ; papaloizou & terquem , 1999 ) , with different consequences on the disc stability ( camenzind , demole & straumann 1986 ; clarke , 1988 ) .
the other way uses the local version of the @xmath0-prescription ( meyer & meyer - hofmeister , 1982 ) @xmath54 where @xmath55 is the reduced pressure height scale ( see the appendix , sect .
whatever the expression the authors adopt , the @xmath0-parameter is most often assigned to a fixed value in a disc .
however , theoretical arguments , numerical simulations as well as observational constraints indicate that @xmath0 should vary , not only with the radius ( shakura & sunyaev , 1973 ; cannizzo , 1993 ; lasota & hameury , 1998 ) , but also with the altitude via the temperature or density , or other quantities ( brandenburg , 1998 ) .
strictly speaking , the need for strong variations of the @xmath0-parameter means the failure of the @xmath0-viscosity model . since @xmath56 and @xmath57 are formally different , even with the same @xmath0-parameter , they lead to different vertical structures and consequently to different discs , as we shall see below . besides , the correspondence between the two laws ( a multiplying factor @xmath58 from @xmath56 to @xmath57 , if @xmath59 ) is much more subtile than shifting @xmath0 : in general , @xmath60 at any altitude ( this is specially true at the equatorial plane where @xmath61 ; see sect .
the volumic energy production associated to @xmath57 is @xmath62 note that both eqs.([eq : nu1 ] ) and ( [ eq : nu2 ] ) satisfy @xmath63 above the equatorial plane , meaning that the gas is accreted much faster at the equatorial plane than at the top of the disc .
so , these laws implicitly suggest the existence a plan parallel shear which might be able to trigger turbulence and to expand turbulent eddies in the radial direction .
there are still no physical arguments to decide if @xmath57 is better than @xmath56 .
that is why we use both expressions in the following ( in particular , for @xmath56 we take @xmath64 ) .
in fact , a wide class of functions @xmath65 should be considered and their effects compared .
it is possible that turbulent viscosity shows a weaker dependence with @xmath12 than considered so far and even does not depend on the altitude at all .
this is precisely the case with the @xmath66-viscosity prescription which is suggested by laboratory experiments ( pringle & rees , 1972 ; lynden - bell & pringle , 1974 ; richard & zahn , 1999 ; duschl , biermann & strittmatter , 2000 ; hur , richard & zahn , 2000 ) . in the eddington approximation , the structure of the atmosphere
is governed by three ( only two in the absence of self - gravity ) first orders odes .
two of these ( namely eqs.([eq : eh1 ] ) and ( [ eq : sd2 ] ) ) are basically the same as for the disc interior , and the third one controls the optical depth @xmath67 in the atmosphere @xmath68 where @xmath69 is a grey absorption coefficient . in order
to prevents any temperature runaway which would lead to the formation of a hot corona ( shaviv & wehrse , 1991 ; raska et al . , 1999 )
, we assume that there is no viscous energy generation in this layer and no active turbulence .
it means that the atmosphere is not accreted .
we are aware that the validity of all these approximations , including the eddington approximation , is easily open to criticism .
the temperature within the atmosphere follows the law ( mihalas , 1978 ) @xmath70 where @xmath71 and @xmath72 is the effective temperature which is fixed by the outgoing flux due to internal and external heating sources , at the photosphere s base .
actually , the global effect of the disc irradiation can be taken into account at this level by a suitable definition of @xmath72 ( kenyon & hartmann , 1987 ; hubeny , 1990 ; robinson , marsh & smak , 1993 ; dubus et al . , 1999 ) .
wee see from eqs.([eq : eh1 ] ) , ( [ eq : tau ] ) and ( [ eq : tdistrib ] ) that the density gradient is @xmath73 and may also become positive .
as suggested above , we do not take into account turbulent pressure in this layer .
this induces a discontinuity in the density ( or gas pressure ) gradient at the altitude where the disc and its atmosphere join together , but not in the density itself .
there are two problems to solve simultaneously : the structure of the disc interior from eqs.([eq : dfoverdz])([eq : temp])([eq : sd2 ] ) and ( [ eq : eh2 ] ) and the structure of its atmosphere from eqs.([eq : sd2])([eq : tau ] ) and ( [ eq : eh3 ] ) .
these differential equations are subject to dirichlet boundary conditions . at the midplane
, the symmetry imposes @xmath74 and @xmath75 like for a star . at the top atmosphere located at @xmath76 ( to be determined )
, conditions are @xmath77 the ambient pressure ( which may include the pressure due to external illumination ) , @xmath78 and @xmath79 , half the total surface density of the disc and atmosphere .
some authors specify the density ( for instance @xmath80 ) instead of the pressure at the boundary ( e.g. pojmaski , 1986 ; drrer et al . 1993 ; el - khoury & wickramasinghe , 1999 ) .
we choose a value typical of the interstellar medium , namely @xmath81 k.@xmath82 ( duley & williams , 1988 ) , significantly smaller than in dalessio et al .
( 1998 ) . at the base of the atmosphere located at @xmath83 ( to be determined too ) ,
the matching conditions are @xmath84 and @xmath85 ( see sect .
it is worthy of note that boundary conditions at the disc surface have a very weak effect on midplane quantities as long as the disc is optically thick and has a large surface density , locally .
such an insensitivity is well known in stellar structure computations ( kippenhahn & weigert , 1990 ) .
let us quote also that , contrary to a general belief , models that do not consider the atmosphere and use @xmath86 as a boundary condition at @xmath83 always violate the eddington approximation since this relation does not guarantee that the atmosphere has the right optical thickness ( i.e. @xmath87 ) .
further , the approximation @xmath88 becomes bad in regions where the surface density of the atmosphere is comparable to that of the disc , when the mean absorption within the atmosphere is low , or when the disc is not optically very thick .
for these reasons , we sustain the idea that the use of the eddington approximation has sense only if the structure of the atmosphere is computed together with that of the disc interior .
the physical problem depends on a certain amount of thermodynamical and radiative data . for the present application ,
the gas has cosmic abundances and is subject to atomic ionizations and molecular dissociations .
local thermodynamic equilibrium ( lte ) is assumed ( hur et al . , 1994b ) .
the mean mass per particle @xmath89 , coefficients @xmath43 and @xmath44 , adiabatic gradient @xmath90 , heat capacity @xmath91 ( needed to treat convective transport ) and adiabatic exponent @xmath92 have been computed accurately from chemical abundances at thermal equilibrium ( hur , 1998 ) .
grids of data with @xmath17 and @xmath93 as the inputs have been generated .
interpolations between mesh points are performed with bicubic splines .
we have computed a high precision analytical expression fitting @xmath94 ( and subsequently coefficients @xmath43 and @xmath44 ) as functions of the temperature and density with an accuracy less than 4@xmath95 relative to raw data in the case of a zero metallicity gas ( see the appendix , sect . b1 ) .
this fit can be used for a cosmic gas as well without producing important errors ( since metals in a cosmic mix have very low abundances relative to hydrogen , with almost no influence on the eos and related quantities ) .
opacity is probably the most important ingredient of the model and the choice for the grey absorption coefficient @xmath69 is critical , specially near the surface ( mihalas , 1978 ) . here , we take ( hameury et al .
, 1998 ) @xmath96 where @xmath97 and @xmath98 are the rosseland and planck means respectively , and @xmath99 is a function which varies continuously from 0 to 1 as the optical depth increases from 0 to infinity .
a suitable choice is @xmath100 where the index @xmath101 sets the stiffness of the transition from @xmath98 to @xmath97 ( @xmath102 is taken in the following applications ) .
the @xmath99-function is arbitrary and should strongly influence physical quantities at the disc surface . also , we are aware that the best grey coefficient is probably neither given by @xmath97 , nor by @xmath98 , nor by eq.([eq : kgrey ] ) .
a flux weighted opacity mean could be better ( mihalas , 1978 ; hubeny , 1990 )
. tables of rosseland and planck means have been taken from various sources ( pollack et al . , 1994 ;
hur et al .
, 1994b ; alexander & fergusson , 1994 ; henning & stognienko , 1996 , seaton et al . , 1994 ) . but
planck absorption means published by the opacity project ( op ) ( seaton et al . , 1996 )
have not be selected due to apparently large errors of unknown origin ( alexander , 1998 ; zeippen , 1999 ) .
two grids of data have been generated after passing through filters ( mainly running averages ) to smooth out discontinuities .
finally , opacity grids cover the temperature range @xmath103 k and extend from extremely high to extremely low densities where the ideal gas and lte assumptions are expected to fail .
an example of rosseland and planck means versus the temperature is displayed in fig.([fig_opac ] ) .
as done for the equation of state and related quantities , opacity values between mesh points are interpolated using bicubic splines . working with analytical opacities
may present some advantages ( burgers & lamers , 1989 ; collin & dumont , 1990 ; bell & lin , 1994 ) , but they can also produce numerical instabilities in the integration of the disc structure equations since they mostly consist in continuous but non derivable piecewise functions . the altitude of the top atmosphere being not known a priori , many numerical methods for solving two boundary value problems ( tbvps ) are almost unusable in the actual situation .
it is however possible to apply variable changes , namely @xmath104 in the atmosphere and @xmath105 in the disc interior ( cannizzo & cameron , 1988 ) in order to recover a common tbvp with fixed boundaries ( e.g. @xmath106 $ ] ) , but this drastically increases vertical gradients near the surface and make the problem more unstable from a numerical point of view .
some groups work with relaxation algorithms ( cannizzo , 1992 ; milsom , chen & taam , 1994 ; dalessio et al . , 1998 ;
hameury et al . ,
1998 ) which require ( good ) vertical profiles as starting guesses and are known to be rapidly converging methods .
however , zones with steep gradients generally need a local mesh refinement .
other authors prefer straight integrators and algorithms based on shooting methods ( lin & papaloizou , 1980 ; meyer & meyer - hofmeister , 1982 ; mineshige & osaki , 1983 ; smak , 1984 ; mineshige , tuchman & wheeler , 1990 ; raska et al . , 1998 ; papaloizou & terquem , 1999 ) typical of initial values problems ( ivps ) , with only a few quantities to guess but with possible troubles regarding the precision . in the present case
, we have written a fortran code named vs@xmath107kad which performs the integration of the disc and atmosphere equations using single side shooting and numerical routines specially designed for stiff ivps .
integration steps are variable , internally to the routines and the accuracy can be as small as the machine precision . in its present version ,
the code works with the following variables : @xmath108 , @xmath109 , @xmath110 , @xmath111 and @xmath67 , but more clever choices are possible , specially to reduce artificially the stiffness of the equations .
we have noticed that the computational time is considerably smaller than with runge - kutta algorithms , even with a variable step . like in many stiff problems ,
the numerical integration may turn out to be non conservative : integrating from the top atmosphere down to the midplane can give a solution which , in the surface neighborhood , can be very different than that obtained when integrating in the opposite direction ( mineshige , tuchman & wheeler , 1990 ; press et al . , 1992 ; de kool & wickramasinghe , 1999 ) .
this has been observed in the present case .
it reminds the influence of both boundary conditions and underlying methodology .
the best stability and reliability of the results are obtained by starting from the boundary layer where almost all the stiffness of the problem is concentrated . in practical ,
given a central mass @xmath10 , accretion rate @xmath112 , @xmath0-parameter , radius @xmath11 and total energy deposition @xmath113 at the bottom atmosphere ( see sect . 4.1 ) , the computation starts at an arbitrary altitude @xmath114 ( the top atmosphere ) where the surface density @xmath115 is guessed .
the integration then proceeds downwards .
once the bottom atmosphere is reached , at an altitude @xmath116 , the equations for the disc interior are integrated down to the midplane where the net flux and surface density generally differ from zero . by successive iterations on @xmath117 and
@xmath118 |performed with a newton - raphson method| , quantities @xmath119 and @xmath120 can be driven to very small values .
the problem has converged after @xmath121 iterations , when simultaneously @xmath122 and @xmath123 with the requested precisions @xmath124 @xmath125 and @xmath126 where @xmath127 , @xmath127 , @xmath128 , @xmath129 , @xmath130 and @xmath131 are very small , positive values .
since our treatment of self - gravity is very approximate , it is legitimate to allow a much lower ( but sufficient ) precision on the surface density ( this accelerates convergence and lowers the computational time ) . on average
, it takes a few milliseconds cpu on a single user personal computer and @xmath132 with @xmath133 where @xmath134 , @xmath135 and @xmath136 .
this precision is sufficient in most cases .
when self - gravity is neglected , eq.([eq : sd2 ] ) becomes obsolete .
the iteration is on the flux only and the code then runs much faster ( by a factor @xmath137 ) , with @xmath138 . the code has been tested and compared in many situations ( e.g. without self - gravity , without turbulent pressure , with and without external irradiation ) with reference models ( raska , 1998 ; hameury et al . , 1998 ) and appears to behave very well .
it automatically scales to the central mass , and is able to model any kind of disc : agn discs , cv discs , yso discs and sub - nebulae as well .
a simulation of d / h enrichments in the psn with this code is currently under way ( hersant , gautier , hur , 2000 ) .
vertically averaged disc models show that the non self - gravitating and gas pressure dominated parts of @xmath0-discs have an aspect ratio @xmath139 , depending on the central mass and accretion rate mainly .
there is a slight flaring ( i.e. @xmath140 ) which eventually may vanish in the outermost regions ( e.g. ruden & pollack , 1991 ; bell et al . , 1997 ) .
the existence of a flaring inner disc is supported by observations : the spectrum of t - tauri stars and agn contains a prominent infrared component which is currently interpreted as light re - processing in the superficial layers of the disc ( adams & shu , 1986 ; voit , 1991 ; chiang & goldreich , 1997 ) .
it is possible that self - irradiation also plays a role ( fukue , 1992 ) .
for circumstellar discs , the flaring angle of bare @xmath0-disc is too low to enhance the infrared spectral component to the observed level ( kenyon & hartmann , 1987 ) . and
the heating by the central star should not produce a significant increase of the flaring ( e.g. dalessio et al .
but the long wavelength emission can efficiently be boosted if the disc gets into a warped configuration ( terquem & bertout , 1993 ; miyoshi et al . , 1995 ; bachev , 1999 ) . beyond a certain radius ,
self - gravity becomes important and reduces the disc thickness such that @xmath141 ( sakimoto & coroniti , 1981 ; shore & white , 1982 ; cannizzo & reiff , 1992 ; liu , xie & ji , 1994 ; hur et al .
, 1994a ; hur , 1998 ) .
as we shall see below , the same effect is observed from vertical structure computations .
it means that outer parts should not receive directly light emitted at the center , contrary to the inner parts .
note that a disc is generally not isolated but embedded into a `` warm '' environment which may in turn heat it up .
discs in yso are surrounded by an envelope of gas and dust resulting from the cloud core collapse ( mundy , looney & welch , 2000 ; chick & cassen , 1997 ; dalessio , calvet & hartmann , 1997 ) . in the case of agn ,
clouds moving above the disc ( blr clouds ) can scatter light back onto it ( shields 1977 ; collin - souffrin , 1987 ; osterbrock , 1993 ) .
it follows that , even outer regions which are not directly exposed to the luminous central source can be substantially irradiated .
the disc response to irradiation is a complex problem to solve self - consistently since it depends on many parameters ( size and location of the ionizing source(s ) , shape of the ionizing spectrum , intensity of irradiation , disc flaring angle , surface albedo , disc optical thickness , gas metalicity , etc . ) which are neither well known , nor well constrained by current observations . for discs having a large total surface density , mostly the superficial layers are heated up and ionized , and the midplane is essentially not affected ( sincell & krolik , 1997 ; collin & hur , 1999 ; nayakshin , kazanas & kallman , 1999 ; igea & glassgold , 1999 ) . deep structural changes ( for instance , a temperature inversion or an iso - thermalization )
may however occur in some cases ( dalessio et al .
although discs should experience some external heating , even in their outermost ( self - gravitating ) parts , we do not consider irradiation in this paper , for simplicity .
so , considering internal viscous heating only , the disc effective temperature , far from the central object , is given by @xmath142 where @xmath112 is the mass accretion rate . in the following ,
we apply the code to the computation of the vertical structure of two different systems subject to vertical self - gravity : circumstellar discs and agn discs .
we restrict to central masses of @xmath143 m@xmath2 and @xmath1 m@xmath2 respectively , and we show on a few examples how self - gravity and turbulent pressure which are the main specificities of this model affect the disc structure .
computations are performed discarding any kind of instabilities the disc could undergo .
these are stopped when the temperature at the top atmosphere ( i.e. @xmath144 ) attains 10 k , for both physical and practical reasons . to see clearly the influence of turbulent pressure ( see eq.([eq : pt ] ) ) , we have chosen a simulation with @xmath145 which corresponds to the very upper limit regarding supersonic turbulence .
in addition , convection and self - gravitation are turned off . under these circumstances ,
we have computed the vertical structure with viscosity law @xmath56 for an disc around a massive black hole with @xmath146 m@xmath2 , at two different radii : @xmath147 ( @xmath148 is the schwarzchild radius of the black hole ) and @xmath149 .
the accretion rate is @xmath150 m@xmath2/yr .
the temperature and density profiles so obtained are plotted in fig.([fig : effecttp ] ) .
we see that turbulent pressure makes the disc thicker , as expected intuitively . actually , according to eq.([eq : eh2 ] ) , the density gradient is lowered with turbulent pressure and the transition from the ambient medium to the disc interior is softer .
the flux gradient depending on the density to a positive power |in the absence of convection at least| , a larger integration range is needed to obtain a zero net flux at the midplane , hence a thicker disc .
the effect is specially visible at @xmath149 , with a @xmath151 disc thickening ( measured on @xmath152 ; we have @xmath153 on @xmath154 ) .
the disc flaring therefore increases by the same relative quantity .
since we have not explored the whole parameter space , it is probably easy to find cases where the effect is more important .
the total surface density is almost unchanged ( a @xmath155 increase only ) . at @xmath156 ,
the disc thickening is less important , about @xmath157 .
note in this example the presence of a density inversion which occurs well below the base of the atmosphere .
it tends to be washed out by turbulent pressure , as argued in sect.2.3 .
the way turbulent pressure acts on both the temperature and density at the midplane temperature is however less straightforward and depends intimately on the vertical stratification . for a given effective temperature ,
a disc with a larger surface density would theoretically have a hotter core .
but this happens only in the calculation at @xmath156 , and the effect is very minor .
we have checked that the conclusions derived here - above are qualitatively unchanged using @xmath57 ( see sect .
4.5 for another effect of turbulent pressure ) .
we will not discuss the case of circumstellar discs because values of the @xmath0-parameter currently adopted for these systems are rather very low ( typically @xmath158 ; dalessio , calvet & hartmann , 1997 ) and so , no effect of turbulent pressure is indeed observable .
this has been verified . in this paragraph , we show a few differences between @xmath56 and @xmath57 . convection is taken into account and turbulent pressure is left aside . the midplane temperature , total surface density , disc thickness and disc mass are plotted versus radius in fig.([fig : rprofyso ] ) for @xmath159 m@xmath2 , @xmath160 m@xmath2/yr and @xmath161 .
these parameter values are typical of discs around t - tauri stars ( dalessio , calvet & hartmann , 1997 ) .
the disc mass @xmath162 is @xmath163 where @xmath164 is the inner edge of the disc ( unimportant as long as @xmath165 ) .
results computed without self - gravity are also shown in comparison .
we see that , in the non self - gravitating region , differences on the geometrical thickness and midplane temperature are small , the disc being slightly thicker and hotter with @xmath56 than with @xmath57 .
the most important effect is on the surface density and consequently on the disc mass : the @xmath56-prescription leads to slightly more massive disc than the @xmath57-prescription ( by a factor @xmath166 here ) .
we see that the disc is definitely affected by self - gravity from approximately @xmath167 au whatever the viscosity law ( in fact , a little bit less with @xmath56 which agrees with the fact that that prescription leads to more massive discs , as long as @xmath168 ) .
let us remind that the importance of self - gravity can be measured by the quantity ( see eq.([eq : eh1 ] ) ) @xmath169 which remains finite at @xmath170 as a taylor expansion shows .
it follows that a `` natural '' definition for the limit between the standard disc and the self - gravitating disc can be @xmath171 at the midplane ( if @xmath172 then the disc contribution to gravity exceeds that due to the central object , vertically ) . on the example
, the disc thickness goes through a maximum near 10 au for both viscosity laws . beyond this radius ,
the disc becomes thinner and thinner .
regions located farther away can therefore not intercept photons emitted at the center .
interestingly , the temperature and the surface density show very different behaviors . with law
@xmath56 , the temperature decreases monotonically as @xmath11 increases , almost as in the absence of self - gravity .
so does the total surface density .
the midplane density , not shown on graphs , gently increases with the radius .
note that @xmath173 ( and @xmath162 ) are surprisingly not affected by self - gravity .
the origin of this insensitivity has not been identified at the time being .
conversely , with law @xmath57 , the surface density and the density violently increase with the radius .
the disc thickness falls in also very rapidly .
the temperature reaches a plateau .
in fact , this somewhat `` singular '' behavior is predicted by the vertically averaged model ( shlosman & belgelman , 1987 ; hur et al .
1994a ; hur , 1998 ; see also duschl , biermann & strittmatter , 2000 ) . in particular , the temperature @xmath174 at the plateau is mainly fixed by the ratio @xmath175 , namely @xmath176 for @xmath177 m@xmath2/yr as in fig.([fig : rprofyso ] ) , eq.([psgtg ] ) yields @xmath178 k ( assuming @xmath179 ) . although this value is derived from the one zone model , it is in good agreement with the vertical structure computation which gives a plateau at about @xmath180 k. note that the drastic increase of the surface density produces an important increase of the disc mass which even exceeds the value obtained with law @xmath56 . when @xmath181 ( this occurs at @xmath182 au on the example ) , the disc becomes very self - gravitating , probably gravitationally unstable ( goldreich & lynden - bell , 1965 ; shu et al . ,
1990 ) and should therefore not be well described with current viscosities ( e.g. lin & pringle , 1987 ) nor by steady state solutions ( goldreich & lynden - bell , 1965 ) .
one reaches here the limit of the model .
it is however interesting to note that , asymptotically , there is either a hot very dense solution or a cold more diffuse solution ( see paczyski , 1978 ) . the same quantities obtained under the same conditions but for an agn disc with @xmath146 m@xmath2
, @xmath183 m@xmath2/yr and @xmath184 are displayed in fig.([fig : rprofagn ] ) .
globally , we find similar trends .
the singular behavior observed in the previous example with @xmath57 inside the self - gravitating regime seems even stronger .
self - gravitation becomes important from about @xmath156 .
the midplane temperature stabilizes radially at @xmath185 k. this value again compares quite well with eq.([psgtg ] ) which gives @xmath186 k ( assuming @xmath187 ) . the presence of extremely steep surface density and density radial gradients probably means that the assumption made on the disc potential ( see sect .
2.1 ) is no longer valid .
this would be worthwhile to check .
the physical picture there is then that of a disc surrounded by a very dense ring ( @xmath21 increases by more than one order of magnitude over a relative length @xmath188 ) which contains almost all the disc mass and angular momentum . when @xmath189 ( beyond a few @xmath190 for law @xmath57 ) , effects of self - gravity are expected to be global , with a change in the rotation law .
all these results require a few comments .
first , it is important to note that the effect of self - gravity is slightly underestimated when it is neglected in the computations .
this is specially true with law @xmath57 .
besides , self - gravity becomes important before @xmath191 reaches unity , like in the one zone model ( hur , 1998 ) . at the inner edge of the self - gravitating disc ( that is at @xmath171 following our definition ) ,
the disc mass is much less than the central mass : we find @xmath192 ( depending on the viscosity law ) in the yso case , and @xmath193 in the agn case .
this suggests that low mass discs must not be automatically classified as non self - gravitating discs as often asserted .
in particular , it is not excluded that t - tauri discs we observe be subject to vertical self - gravity , despite their relatively low mass ( beckwith et al . ,
1990 ) .
it has been stated in sect .
2.4 that the relation between @xmath56 and @xmath57 is not trivial . in the non self - gravitating parts however , midplane temperatures @xmath194 show a striking quasi - parallel variation with the radius in logarithmic scales ( this is also true for quantities @xmath173 and @xmath152 ) , which could be attributed to the fact that @xmath56 is indeed proportional to @xmath57 ( or one goes from one solution to the other by a change of @xmath0 ; see power law solutions for @xmath0-discs ) . to demonstrate that these two viscosity prescriptions are intrinsically different ( as suggested by what happens in the self - gravitating regions ) ,
we have computed the ratio @xmath195 with @xmath56 , for the examples discussed above .
we have chosen two radii , the one lies in the classical disc part , the other is in the self - gravitating part .
the results are shown in fig.([fig : deltanu ] ) . [ cols="^,^ " , ] we show in fig.([fig : psn ] ) the density field within a circumstellar disc for the same parameter values as in fig.([fig : rprofyso ] ) .
this simulation has been carried out using @xmath56 and @xmath57 as well , including convection , self - gravity and turbulent pressure .
we also indicate iso - contours of the temperature , the limit between gas pressure dominated zones and radiative pressure dominated ones , and lines where @xmath196 and @xmath197 .
we notice that the density and the temperature show strong variations in between the midplane to the photosphere s base and we are far from a vertically isothermal , homogenous disc as often considered .
note , specially for viscosity law @xmath56 , the re - increase of the midplane density in the radial direction ( another density inversion ) as self - gravity gains in importance .
it is remarkable that self - gravity does not appear suddenly ( i.e. on a small radial range ) but installs gently and steadily over a very extended domain .
for instance , with law @xmath57 , it contibutes by @xmath198 in the hydrostatic equilibrium near @xmath199 au and by @xmath200 at about 7 au , and all disc quantities are significantly modified over this domain .
we see clearly that the disc density is slightly smaller with @xmath57 than with @xmath56 ( same color code for both maps ) .
we notice that the whole disc is optically very thick , due to its large surface density .
it is then likely that a mean external heating should not change noticeably the temperature and density at the midplane . from this point of view
, the neglect of irradiation is entirely justified .
it is interesting to see that , as long as an @xmath0-model can be applied to the solar nebula ( e.g. ruden & pollack , 1991 ; papaloizou & terquem , 1999 ; drouart et al , 1999 ) , our giant planets ( displayed on graphs ) lie inside the self - gravitating part of the disc ( this is true whatever the accretion rate , but depends on the @xmath0-parameter ; see the next section ) . as already noticed ( ruden & pollack , 1991 ) ,
this coincidence is somewhat striking and we do not know to what extent self - gravity could have played a role in the process of the formation of outer planet and other objects ( lissauer , 1993 ) .
similar 2d - maps are displayed in fig.([fig : agn ] ) for an agn disc with @xmath146 m@xmath2 , @xmath183 m@xmath2/yr and @xmath201 , using the same color table .
note the existence of an inner radiative pressure dominated region at @xmath202 , common for agn discs as well as the great steepness of density gradients near the surface which may cause numerical difficulties .
also visible is the density inversion at @xmath203 which has less amplitude and is less extended radially in the presence of convection and turbulent pressure .
we see on both figs.([fig : psn ] ) and ( [ fig : agn ] ) that @xmath204 and iso - values of @xmath205 form very curved lines ( `` diabolo''-shape surfaces in 3d ) .
this is important if one wishes to estimate correctly the position of the self - gravitating regime .
for instance , in the case of the law @xmath56 discussed in fig.([fig : psn ] ) , @xmath171 in the midplane at 7 au ( we have @xmath206 in the agn case depicted in fig.([fig : agn ] ) ) whereas this occurs in the disc atmosphere at about twice the distance . for many purposes , it is interesting to know the position of the ( vertically ) self - gravitating disc .
that is why we have performed a systematic computation to find the radius @xmath207 satisfying @xmath208 as a function of the accretion rate , for 4 values of the @xmath0-parameter in the range @xmath209 , and for @xmath159 m@xmath2 and @xmath146 m@xmath2 .
these parameter values should cover a wide variety of circumstellar discs and agn discs .
the calculations have been performed with @xmath57 including convection , self - gravitation , with and without turbulent pressure ( the use of @xmath56 would systematically give a slightly lower value of @xmath207 ) .
the results are plotted in fig.([fig : zetayso ] ) for parameter values typical of circumstellar discs .
we see that the location of self - gravitating regime is very sensitive to the viscosity parameter .
the larger the value of @xmath0 , the further away the self - gravitating region . regarding the sensitivity to the accretion rate
, there are three different trends . for @xmath210 , the lower @xmath112 , the larger @xmath207 . for @xmath211 ,
the dependence is rather weak : @xmath212 au . for higher values of the viscosity parameter ( generally not appropriate to fit properties of observed discs ) ,
@xmath207 is an increasing function of @xmath112 .
note also that turbulent pressure , when important , slightly increases @xmath207 ( @xmath213 in the example for @xmath145 and the highest accretion rate ) .
results for agn discs are shown in fig.([fig : zetaagn ] ) .
as above , large viscosity parameters imply wider non self - gravitating inner disc ( because discs are less dense with large values of the @xmath0-parameter ) , and turbulent pressure pushes further away the self - gravitating regime . for @xmath214 , we have @xmath215 cm .
the sensitivity to the accretion rate remains weak . for higher values of the viscosity parameter and for very low to moderate accretion rates ( @xmath216 m@xmath2/yr , depending on @xmath0 ) , @xmath207 globally decreases as @xmath112 increases . for higher accretion rates , @xmath207 increases as @xmath112 increases and self - gravity is important within the radiative pressure dominated region .
it is important to notify that the `` toomre s criterion '' ( @xmath217 ) is an helpful tool to trace regions where some gravitational instabilities occur in a stellar disc ( toomre , 1965 ) , but it differs significantly from the criterion derived by goldreich & lynden - bell ( 1965 ) for gaseous discs ( @xmath218 ) . besides , @xmath219 is commonly written in different forms which are not strictly equivalent ( e.g. , ruden & pollack , 1991 ; sincell & krolik , 1997 ; dalessio , calvet & hartmann , 1997 , papaloizou & terquem , 1999 ) , specially in a 2d - model .
althought these @xmath220-parameters are undisputably related to the @xmath191-parameter in some ways ( for instance , @xmath221 and @xmath222 ) , they must be used with care if one wishes to check the importance of self - gravity .
they may lead to noticeable over - estimates of @xmath207 , specially if , as it is always the case , the @xmath219-parameter is computed from models that discards self - gravity .
this is illustrated in fig.([fig : zetaq ] ) where we have plotted values of @xmath191 at three key altitudes and @xmath219 versus the radius .
it seems therefore preferable to make the check on @xmath223 rather than on any other quantities , otherwise the importance of self - gravity is under - estimated .
in this article , we have presented the equations accounting for the vertical structure of steady state keplerian accretion discs , including simultaneously turbulent pressure , convective transport in the framework of the mixing length theory , and the disc self - gravity within the infinite slab approximation .
the emphasis has been placed on outer regions where vertical self - gravity becomes increasingly influencal .
we have shown that turbulent pressure makes the disc thicker , but it can be neglected if @xmath224 .
also , self - gravity may be important even if the disc mass is very low compared to the central object , contrary to usual assertions .
further , the transition from the classical disc to the self - gravitating disc takes place gradually and so , concerns a large radial domain . another important conclusion of this work is that the model of mechanical energy deposition towards the vertical direction is crucial .
this is not surprising since viscosity is the source of heating and accretion .
two different behaviors are predicted inside the self - gravitating ( and gravitationally unstable ) region . on the one hand ,
the @xmath225-formalism yields a solution which is asymptotically cold and diffuse , whereas the local version of the standard prescription gives a more `` singular '' solution , asymptotically hot and dense , which is also predicted by the vertically averaged model .
we can imagine that a suitable choice for the function @xmath226 should provide any intermediate solution , meaning that this model for outer regions has almost no predictive power .
although the present stationary keplerian disc model is probably irrelevant to describe the disc structure in its strongly self - gravitating parts , the existence of two asymptotically distinct solutions might indicate that the outer disc can get into , either a flat and cold configuration where fragmentation and formation of indiviual clouds ( kumar , 1999 ) and compact objects like planets and stars ( collin & zahn , 1999 ) could occur , or into a thick diffuse configuration ( a torus ) ( paczynski , 1978 ) .
time dependent simulations should shed light on this question .
this study confirms that discs around super - massive black holes are self - gravitating close to the center , beyond a few hundreds schwarzchild radii , depending on the central mass accretion rate and @xmath0-parameter .
we have found that the disc surface density remains high when self - gravity dominates and that the disc mass definitely rises outwards .
let us remind that a current problem with the fueling of active nuclei is that the accretion time scale @xmath227 is usually much too long at large radii , that is why other more efficient mechanisms are invoked ( frank , 1990 ) .
however , given the sensitivity of the disc structure to viscosity , it is not excluded that depth dependent viscosity laws other than considered here would lead to smaller surface density distributions , and consequently to shorter accretion time scales .
it is therefore important to test other models for the energy deposition along the @xmath12-axis as well as other kinds of viscosity prescriptions ( lin & pringle , 1987 ; cannizzo & cameron , 1988 ; richard & zahn , 1999 ; see hur & richard , 2000 ) . with accretion rates and values of the viscosity parameter
usually considered to model discs in t - tauri stars , we conclude that our giant planets ( if not jupiter , the other ones ) were probably formed within the self - gravitating part of the solar nebula ( drouart et al .
this is a fortiori true if these objects experienced any inward migration .
more generally , as long as the @xmath0-theory may be applied to describe the inner parts of circumstellar discs around forming stars , self - gravity is expected to play a role at about ten au from the center .
this is not uncompatible with the discs we observed ( beckwith et al . ,
1990 ) .
i specially thank a. raska for many helpful discussions and for providing some reference vertical profiles during the writing and test of the computational code , s. collin for reading the manuscript .
i thank also d. gautier for highlights on circumstellar discs and the primitive solar nebular , j .-
zahn for pointing out the role of turbulent pressure .
i am personally grateful to j.m .
hameury for his initiation , some years ago , to the physics of stellar interiors and vertical structure computations .
i also thank d. richard for interesting comments on numerical aspects as well as f. hersant for widely testing the code .
also , i thank for their hospitality , people at the ita - heidelberg where this work was completed , and specially w.j . duschl .
abramowicz m.a .
, czerny b. , lasota j.p .
, szuszkiewicz e. , 1988 , apj , 332 , 646 abramowicz m.a . , calvani m. , nobili , l. , 1980 , apj , 242 , 772 adams f.c . ,
shu f.h . , 1986 , apj , 308 , 836 aikawa y. , umebayashi t. , nakano t. , miyam s.m . , 1999 , apj , 519 , 705 alexander d.r . 1998 , private communication alexander d.r . , ferguson j.w . , 1994 ,
apj , 437 , 879 artemova i.v .
, bisnovatyi - kogan g.s . , bjrnsson g. , novikov i.d .
1996 , apj , 456 , 119 bachev , 1999 , a&a,348,71 beckwith s.v.w . ,
sargent a.i . ,
chini r.s .
, gsten r. , 1990 , aj , 99,924 bell , k.r .
, cassen p.m. , klahr h.h .
, henning th . , 1997 , apj , 486 , 372 bell k.r . , lin d.n.c . , 1994 , apj , 427 , 987 bertin g. , lodato g. , 1999 , a&a , 350 , 694 bertout c. , 1989 , ara&a , 27 , 351 bodo g. , curir a. , 1992 , a&a , 253 , 318 boss a. , 1996 , apj , 469 , 906 brandenburg a. , 1998 , in _
`` theory of black hole accretion discs '' _ , p61 , eds .
abramowicz , bjrnsson & pringle j.e .
, cambridge university press burgers p. , lamers h.j.g.l.m , 1989 , a&a , 218 , 161 burderi l. , king a.r . , szuszkiewicz e. , 1998 , apj , 509 , 85 camenzind m. , demole f. , straumann n. , 1986 , a&a , 158 , 212 cannizzo j.k .
, 1992 , apj , 385 , 94 cannizzo j.k .
, 1993 , in _
`` accretion disks ionm compact stellar systems '' _ , p6 , ed . j.c .
weehler , world scientific cannizzo j.k . ,
reiff c.m .
1992 , apj , 385 , 87 cannizzo j.k . ,
cameron a.g.w . , 1988 , 330 , 327 cannizzo k.k . , wheeler j.c . , 1984 , apj sup
ser . , 55 , 367 collin - souffrin s. , 1987 , a&a , 179 , 60 collin - souffrin s. , dumont a .- m .
1990 , a&a , 229 , 292 collin s. , zahn j.p . , 1999 , a&a , 344 , 433 collin s. , hur j.m . , 1999 , a&a , 341 , 385 cox j.p . , giuli r.t . , 1968 , in _ `` principles of stellar structure ''
_ , new york , gordon and breach chiang e.i . , goldreich p. , 1997 ,
apj , 490 , 368 chick k.m .
, cassen p. , 1997 , apj , 477 , 398 clarke c.j .
1988 , mnras , 235 , 881 dalessio p. , calvet n. , hartmann , 1997 , apj , 474 , 397 dalessio p. , canto g. , calvet n. , lizano s. , 1998 , apj , 500 , 411 dalessio p. , cant j. , hartmann l. , calvet n. , lizano s. , 1999 , apj , 511 , 896 drrer t. , riffert h. , staubert r , ruder h. , a&a 1996 , 311 , 69 de kool m. , wickramasinghe d. , 1999 , mnras , 307 , 449 drinkwater m.j . , combes f. , wiklind t. , 1996 , a&a , 312 , 771 drouart a. , dubrulle , b. , gautier , d. , robert f. , 1999 , i , 140 , 129 dubus g. , lasota j.p . , hameury j.m .
, charles p. , 1999 , mnras , 303,139 duley w.w .
, williams d.a . , in _ `` interstellar chemistry '' _ , london , england and orlando , fl , academic press , 1984 duschl w.j , strittmatter p.a .
, biermann p.l . , 2000 , in press duvert g. et al .
, 1998 , a&a , 332 , 867 el - khoury walid , wickramasinghe d. , 1999 , mnras , 303 , 380 eriguchi y. , mller , 1991 , 248 , 435 falcke h. , 1998 , rev . mod . astr .
, 11 , schielicke r.e . ed .
frank a. , 1998 , in _ accretion processes in astrophysical systems : some like it hot
_ , proceedings of the 8th aip conference , ed .
holt , kallman t.r . , 513 frank j. , king a. , raine d. 1992 , in _
`` accretion power in astrophysics '' _
, cambridge uni . press .
fukue j. , 1992 , pasj , 44 , 663 goldreich p. , linden - bell d. 1965 ,
mnras , 130 , 97 guilloteau s. , dutrey a. , 1998 , a&a , 339 , 467 gsten r. , chini r. , neckel t. , 1984 , 138 , 205 hameury j.m .
et al . , 1998 , mnras , 298 , 1048 hashimoto m. , eriguchi y. , mller e. , 1995 , a&a , 297 , 135 henning & stognienko r. , 1996 , a&a 311 , 291 herrnstein j.r .
et al . , 1999 ,
nature , 400 , 539 hersant f. , gautier d. , hur j.m . , 2000 , in preparation hubeny i. 1990 apj , 351 , 632 hubeny i. , hubeny v. , 1998 , apj , 505 , 558 hunter c. , 1963 , mnras , 126 , 23 hur j.m .
, collin - souffrin s. , le bourlot j. , pineau des forts g. 1994a , a&a , 290 , 19 hur j.m . , collin - souffrin s. , le bourlot j. , pineau des forts g. 1994b , a&a , 290 , 34 hur j.m . 1997 , _ accretion processes in astrophysical systems : some like it hot
_ , proceedings of the 8th aip conference , ed .
holt , kallman t.r .
, 137 hur j.m .
1998 , a&a , 337 , 625 hur j.m .
, richard d. , to appear in _
`` agn in their cosmic environment '' _ , eds .
b. rocca - volmerange , sol h. , edps conference series in astronomy & astrophysics hur j.m . ,
richard , zahn j.p .
, 2000 , in preparation igea j. , glassgold e , 1999 , apj , 518 , 848 kenyon s.j .
, hartmann l. , 1987 , apj , 323 , 714 kenyon s.j . , yi i. , hartmann l. , 1996 , apj , 462 , 439 kippenhahn r. , weigert a. , 1990 , in _ `` stellar structure and evolution '' _ , springer berlin kumar p. , 1999
, apj , 519 , 599 lasota j .-
, hameury j .-
1998 , in _ accretion processes in astrophysical systems : some like it hot _ , proceedings of the 8th aip conference , ed .
holt , kallman t.r . , 351 laughlin g. , ryczka m. , 1996 , apj , 456 , 279 lin d.n.c . ,
papaloizou j.c.b . , 1980 , mnras , 191 , 37 lin d.n.c . , pringle j.e . , 1987 , mnras , 225 , 607 liu b.f . , xie g.z .
, ji k.f . , 1994 , a&ss , 220 , 75 lynden - bell d. , pringle j. 1974 mnras , 168 , 603 lissauer j.j . , 1993 , araa , 31 , 129 malbet f. , bertout c. , 1991 , 383 , 814 maraschi l. , reina c. , treves a. , 1976 , apj , 206 , 295 masuda n. , nishida s. eriguchi y. , 1998 , mnras , 297 , 1139 meyer f. , meyer - hodmeister e. , 1982 , a&a , 106 , 34 miyoshi m. et al .
1995 , nature , 373 , 127 mihalas d. , 1978 , in _ `` stellar atmospheres '' _ ( san fransisco : freeman ) milsom j. a. , chen x. , taam r. e. , 1994 , apj , 421 , 668 mineshige s. , osaki y. , 1983 , pasj , 35 , 377 mineshige s. , tuchman y. , wheeler j.c . , 1990 , apj , 359 , 176 mineshige s. , umemura .
m , 1997 , apj , 480 , 167 mundy l.g .
, looney l.w . and
welch , w.j .
, 2000 , in _
`` protostars and planets iv '' _ , ed . mannings v. , boss a.p . , russell s.s .
( tucson : university of arizona press ) , in press narayan r. , madevan r. , quataert e. , 1998 , in _
`` the theory of black hole accretion disks '' _ , eds .
abramowicz m.a .
, bjornsson g. & pringle j.e . , cambridge uni .
press nayakshin s. , kazanas d. , kallman , t. , 1999 , aas , 195 , 3903 osterbrock d.e .
, 1993 , ap.j . ,
404 , 551 paczyski b. 1978 , aca , 28 , 91 papaloizou j.c.b . , terquem c. , 1999 , apj , 521 , 823 pojmaski g. , 1986 , acta a , 36 , 69 pollack j.b .
1994 , ap .
j. , 421 , 615 press w.h .
, teukolsky s.a . , vetterling w.t . ,
flannery b.p .
, 1992 , in _ `` numerical recipes in fortran . the art of scientific computing '' _ , cambridge : university press , p727 pringle j. 1981 ara&a , 19 , 137 pringle j.e .
, rees m.j .
1972 , astr .
, 21 , 1 pudritz r.e .
et al . , 1996 , apj , 470 , l123 richard d. , zahn j .- p . 1999 ,
a&a , 347 , 734 ross r.r . , fabian a.c . ,
mineshige s. , 1992 , mnras , 258 , 189 robinson e.l .
, marsh t.r .
, smak j.i . , 1993 , in _ `` accretion disks ionm compact stellar systems '' _ ,
weehler , world scientific rzaska a. , czerny b. , zycki p.t . ,
pojmaski g. , 1999 , mnras , 305 , 481 rzaska a. 1998 , private communication ruden s.p . , pollack j.b . , 1991 , apj , 375 , 740 sakimoto p.j . ,
coroniti f.v .
1981 , apj , 247 , 19 sanders d.b .
1989 , apj , 347 , 29 sandqvist aa .
, 1999 , a&a , 343 , 367 seaton m.j . , yan y. , mihalas d. , pradhan a.k .
, 1994 , mnras , 266 , 805 shakura n.i .
, sunyaev r.a .
1973 , a&a , 24 , 337 shaviv g. , wehrse r. , 1991 , a&a , 251 , 117 shepherd d.s . , kurtz s.e . , 1999 , apj , 523 , 690 shields g.a . , 1977 , ap.j .
, 18 , 119 shlosman i. , begelman m.c .
1987 , nature , 329 , 810 shore s.n .
, white r.l .
256 , 390 shu f. h. , tremaine s. , adams f. c. , ruden s. p. , 1990 ,
apj , 358 , 495 siemiginowska a. , czerny b. , kostyunin v. 1996 , apj , 458 , 491 sincell m. , krolik j. , 1997 , apj , 476 , 605 smak j. , 1984 , acta astron . , 34 , 161 stahler s.w . , 1983 , apj , 268 , 155 strzer h. , 1993 , apj , 271 , 25 terquem c. , bertout c. , 1993 , a&a , 274 , 291 toomre a. 1964 , apj , 139 , 1217 th p.s . ,
molster f.j . , 1994 ,
a&ss , 212 , 125 tuchman y. , mineshige s. , wheeler j.c . , 1990 , apj , 359 , 164 voit g.m . , 1991 , apj , 379 , 122 wehrse r. , strzer h. , shaviv g. , , 1993 , a&ss , 205 , 163 zeippen c.j . , 1998 , private comminucation zdziarski a.a . 1986 , apj , 305 , 45
convection is treated with in the framework of the mixing length theory ( mlt ) ( cox & giuili , 1968 ) .
it is modified in order to include turbulent pressure in the determination of the pressure scale height , and self - gravititation . following the original version of the mlt
, we neglect the variation of gravity with the altitude @xmath12 .
it is likely that turbulent pressure modify significantly the velocity of rising elements within convectively unstable zones , specially if the @xmath0-parameter is close to unity , but this effect is not considered here .
the total pressure height scale writes @xmath228 where @xmath229 is the total ( gas plus radiation ) pressure and @xmath41 is the turbulent pressure .
this expression being singular at the midplane ( @xmath230 ) ( see eq.([eq : eh1 ] ) ) , the pressure scale height is usually limited to the disc thickness @xmath231 \label{eq : llambdap}\ ] ] where @xmath17 is the gas density , @xmath15 is the rotation frequency , @xmath21 is the surface density and @xmath152 is the altitude of the bottom atmosphere .
according to eq.([eq : temp ] ) , the temperature gradient is @xmath232 where the actual gradient @xmath31 depends on the adiabatic gradient @xmath233 ( computed from the eos ; see the next section ) with respect to the fictitious radiative gradient @xmath234 defined as @xmath235 where @xmath236 is the flux to be transported upwards and @xmath69 is a grey absorption coefficient which must approach the rosseland mean @xmath97 at great optical depth ( see sect .
3.1 ) . when @xmath237 ( i.e. in convectively stable zones ) , heat is transported through radiation only and we have @xmath238 .
conversely , when @xmath239 , @xmath49 is computed from @xmath240 where @xmath241 is the real root of the third degree equation @xmath242 with @xmath243^{1/3}\ ] ] and @xmath244 where @xmath245 is the mixing length parameter ( @xmath246 here ) and @xmath91 is the constant pressure specific heat capacity .
within the assumption of lte , the pressure of a mixture of radiation and ideal gas undergoing atomic ionization molecular dissociation is given by eq.([eq : eh1 ] ) .
in particular , the mean mass per particle @xmath94 , in units of the proton mass , is defined as @xmath248 where @xmath249 and @xmath250 are respectively individual mass and partial pressure of the chemical compounds , and the sum extends over the total number of compounds , including electrons . we have computed equilibrium abundances for a mixture of hydrogen and helium ( h : he=1:@xmath251 ) only as function of @xmath17 and @xmath93 ( see hur , 1998 ) .
data relative to the eos so obtained are plotted versus the temperature in fig.([fig_coefs ] ) for three values of the gas density .
in particular , the resulting function @xmath252 has then been fitted with great accuracy by the following expression @xmath253 with @xmath254 where coefficients @xmath255 and @xmath256 and @xmath257 are listed in tab.([tab_coefformu ] ) .
hyperbolic functions account successively for the transitions h@xmath258/hi , hi / hii , he / hei and hei / heii .
relative errors on raw data @xmath259 are displayed in fig.([fig_muerrors ] ) and never exceed 4 % over the whole domain of temperature and density .
besides , the standard deviations with respect to raw data are less than 1.3 % .
we have tried to fit the residuals with a few gaussians , but no major improvement has been obtained . note that , by construction , the fitting formula shows no singular behavior at high of low densities and temperature .
the fit can be used for a cosmic gas without producing large errors ( the effect of heavy elements on the eos is very weak as long as @xmath260 ) .
ll i & @xmath255 + 0 & @xmath261 + 1 & @xmath262 + 2 & @xmath263 + 3 & @xmath264 + 4 & @xmath265 + + i & @xmath266 + 1 & @xmath267 + 2 & @xmath268 + 3 & @xmath269 + 4 & @xmath270 + + i & @xmath257 + 1 & @xmath271 + 2 & @xmath272 + 3 & @xmath273 + 4 & @xmath274 + the temperature and density exponents of the total pressure are repectively given by @xmath275 \label{eq : chit}\ ] ] and @xmath276 \label{eq : chirho}\ ] ] where @xmath66 is the ratio of gas pressure to total ( gas plus radiation ) pressure .
it follows from eq.([eq : fitofmu ] ) @xmath277 and @xmath278 and the precision with respect to raw data is also of the order of a few percents .
[ [ low - precision - expressions - for - coefiicients - chi_t - chi_rho - nabla_rm - ad - gamma_1-and - c_p ] ] low precision expressions for coefiicients @xmath43 , @xmath44 , @xmath90 , @xmath92 and @xmath91 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ the adiabatic gradient @xmath90 , heat capacity at constant pressure @xmath91 and adiabatic exponent @xmath92 are respectively defined by @xmath279^{-1}\ ] ] where @xmath280 is the specific internal energy of gas and radiation , @xmath281 and @xmath282 since disc models are still very uncertrain , one may use , as a first approximation , low precision expressions for all coefficients related to the eos discarding changes in the internal energy due to ionizations and dissociations , for instance @xmath283 @xmath284 and then @xmath285 @xmath286 and @xmath287
|
the equations governing the vertical structure of a stationary keplerian accretion disc supporting an eddington atmosphere are presented .
the model is based on the @xmath0-prescription for turbulent viscosity ( two versions are tested ) , includes the disc vertical self - gravity , convective transport and turbulent pressure .
we use an accurate equation of state and wide opacity grids which combine the rosseland and planck absorption means through a depth - dependent weighting function . the numerical method is based on single side shooting and incorporates algorithms designed for stiff initial value problems .
a few properties of the model are discussed for a circumstellar disc around a sun - like star and a disc feeding a @xmath1 m@xmath2 central black hole . various accretion rates and @xmath0-parameter values
are considered .
we show the strong sensitivity of the disc structure to the viscous energy deposition towards the vertical axis , specially when entering inside the self - gravitating part of the disc . the local version of the @xmath0-prescription leads to a `` singular '' behavior which is also predicted by the vertically averaged model : there is an extremely violent density and surface density runaway , a rapid disc collapse and a temperature plateau .
with respect , a much softer transition is observed with the `` @xmath3-formalism '' .
turbulent pressure is important only for @xmath4 .
it lowers vertical density gradients , significantly thickens the disc ( increases its flaring ) , tends to wash out density inversions occurring in the upper layers and pushes the self - gravitating region to slightly larger radii .
curves localizing the inner edge of the self - gravitating disc as functions of the viscosity parameter and accretion rate are given .
the lower @xmath0 , the closer to the center the self - gravitating regime , and the sensitivity to the accretion rate is generally weak , except for @xmath5 .
this study suggests that models aiming to describe t - tauri discs beyond about a few to a few tens astronomical units ( depending on the viscosity parameter ) from the central protostar using the @xmath0-theory should consider vertical self - gravity , but additional heating mechanisms are necessary to account for large discs .
the primitive solar nebula was probably a bit ( if not strongly ) self - gravitating at the actual orbit of giant planets . in agreement with vertically averaged computations , @xmath0-discs hosted by active galaxies are self - gravitating beyond about a thousand schwarzchild radii .
the inferred surface density remains too high to lower the accretion time scale as requested to fuel steadily active nuclei for a few hundred millions years .
more efficient mechanisms driving accretion are required .
|
recent discovery of luminous quasars at @xmath6 suggests the existence of black holes with mass exceeding @xmath7 when the age of the universe was less than one billon years ( e.g. , * ? ? ?
* ; * ? ? ?
the formation process of such supermassive black holes ( smbhs ) is largely unknown , and poses a serious challenge to the theory of structure formation .
the so - called direct collapse scenario @xcite provides an attractive pathway for the early bh formation via a peculiar mode of the primordial star formation .
this scenario supposes the gravitational collapse of a primordial gas cloud forming a supermassive star with a mass of @xmath8 .
specifically , the formation process proceeds in a two - step manner ; a very small ( @xmath9 ) protostar first forms at the densest part of the cloud , and then grows in mass by rapid accretion from the surrounding envelope ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . the accretion rate onto the protostar is expected to be as large as @xmath10 , with which the stellar mass will exceed @xmath11 in less than the stellar lifetime
. such a supermassive star , if successfully formed , eventually collapses into a bh either by the core - collapse or by the general relativistic ( gr ) instability ( e.g. , * ? ? ?
* ; * ? ? ?
given the massive remnant bh as a seed , a smbh with more than @xmath7 can be assembled through further growth by accretion and/or mergers in less than one billion years ( e.g. , * ? ? ?
the size and the internal structure of an accreting protostar ( e.g. , * ? ? ?
* ; * ? ? ?
* ) and its resulting radiative feedback effect , which potentially terminates the mass accretion ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , are key ingredients to set the final stellar mass ( e.g. , * ? ? ?
recent calculations show that the protostar structure is substantially modified if the accretion rate exceeds a few @xmath12 ; the star inflates to have an extended radius of @xmath13 au ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
since such a fluffy protostar has low effective temperature and hence very low uv luminosity , the radiative feedback does not disturb the accretion flow at least until the stellar mass exceeds @xmath14 as long as continuous and efficient gas accretion ensues . in this _ letter _ , we extend our stellar evolution calculations until the smss begin to collapse .
our calculations determine the final mass of the accreting smss by incorporating the realistic conditions within the context of the direct collapse scenario .
previous studies show that the gr instability causes the collapse of a non - rotating sms with @xmath15 a few @xmath16 but without considering its evolution under rapid mass accretion ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
we show that gas accretion significantly alters the stellar structure . with a very large accretion rate
, a star continues to grow in mass without collapsing by the gr instability until the stellar mass reaches @xmath17 .
our calculations show the final fate of such a sms .
.final stellar mass and composition of the inner core [ cols="<,^,^,^,^",options="header " , ] [ tb : t1 ] we solve the stellar evolution equations using a modified code developed in @xcite , incorporating the gr effects in the post - newtonian approximation @xcite .
we adopt a primordial abundance pattern for an initial model with hydrogen and helium mass fractions @xmath18 and @xmath19 , and the same nuclear reaction networks as in @xcite and @xcite .
a major difference of the code from that of @xcite is that no entropy generated at the accretion shock is assumed to be incorporated into the stellar interior .
this corresponds to @xmath20 models in @xcite , where @xmath21 is the fraction of the accretion energy deposited in the stellar interior .
however , varying this setting only affects the early evolution for the total stellar mass @xmath22 .
we follow the evolution of an accreting protostar with four different rates of @xmath23 = 0.1 , 0.3 , 1 , and @xmath4 , starting with a @xmath24 initial model which has an inflated structure .
we first checked that our calculations well reproduce the previous results by @xcite , who study the evolution before the gr effect becomes important .
we terminate the calculations when the stellar central temperature reaches @xmath25 , after which the evolutionary timescale is so short that the stellar mass hardly increases thereafter .
the final mass and thus the final fate of the star are determined at this epoch .
, 1 , and @xmath4 .
the gray - shaded area denotes the electron - positron pair - unstable region .
the star becomes gr - unstable in the dot - shaded area under the assumptions of the @xmath26 polytropic structure and primordial composition .
, width=302 ] in figure [ fig : f1 ] we show the evolution of the central temperature and density for the cases with @xmath27= 0.1 , 1 and @xmath4 . although the evolutionary tracks in this figure are quite similar for all the models , there are some differences in the actual evolution , including the cause of the final collapse , as described below . from a to b in figure
[ fig : f1 ] , the stellar core contracts monotonically over the kelvin - helmholtz timescales until the hydrogen burning via the cno cycle starts to sustain the star .
although the evolutions from a to b are almost identical with different @xmath23 values in figure [ fig : f1 ] , figure [ fig : f2 ] shows that the stellar mass at point b is larger for higher @xmath23 .
as shown in these figures , the central temperature at point b , i.e. , when the hydrogen burning begins , is @xmath28 with only very weak dependence on @xmath23 . during the hydrogen burning , i.e. , from b to c in figure [ fig :
f1 ] , the central temperature remains almost constant while the density decreases by more than a factor of 10 .
this is because the central entropy is in general higher for more massive stars , and thus an accreting star evolves toward higher entropy ( lower density ) regions as its mass increases .
after either the exhaustion of nuclear fuel at the center ( for the cases with 0.1 @xmath29 ) or the onset of the gr instability ( for 1 and 10 @xmath29 ) , the stellar core contracts again toward d. figure [ fig : f2 ] shows that this occurs at higher stellar mass for higher accretion rate . in all the models ,
the stars eventually enter the electron - positron pair - unstable region .
it is known that stars with @xmath30 collapse directly to form bhs without explosion once entering this region ( e.g. , * ? ? ?
presumably , our smss also will follow the same fate ( see section [ sec : discuss ] ) .
table [ tb : t1 ] summarizes the stellar masses and mass fractions @xmath31 or @xmath32 for the different accretion rates when the central temperature reaches @xmath33 .
we here show the mass fractions not at the very center but at the slightly outer part @xmath34 , where @xmath35 is the core mass , because the central composition changes relatively fast in the final collapse stage .
the central hydrogen has been exhausted by the onset of the collapse in the cases with @xmath36 , for which the values of the helium fraction @xmath32 are described in table [ tb : t1 ] . on the other hand
, we show the hydrogen mass fraction @xmath31 for the model with the highest rate @xmath37 , where the hydrogen still remains at the center .
the gr instability of smss has been studied for stellar models with the polytropic equation of state with @xmath26 assuming that the entire star is radiation - pressure dominant . for the @xmath26 polytropic star
, the critical central temperature above which the gr instability occurs can be written as @xcite , @xmath38 where @xmath39 and @xmath40 is the mean molecular weight .
as shown in figure [ fig : f2 ] , the central temperature during the stable hydrogen and helium burning phase is roughly constant at @xmath41 for all the models .
for this temperature , the critical mass for the gr instability is given by @xmath42 where @xmath43 is normalized with @xmath44 .
with the mean molecular weight @xmath45 for the initial primordial abundance ( 1.33 for the pure helium matter ) , the critical mass is @xmath46 ( 0.06 , respectively ) .
as shown in table [ tb : t1 ] , the final mass in our case with @xmath47 is @xmath48 , smaller than the critical mass estimated above .
in fact , the collapse in this case is not triggered by the gr instability .
the star just collapses after exhausting the central hydrogen and helium . for @xmath49 ,
on the other hand , the collapse occurs while the nuclear fusion is still in operation , suggesting that the gr instability causes the collapse .
the final masses in these cases are , however , again much larger than the critical masses for the polytropic stars given by equation ( [ eq : mcr ] ) . to understand these differences ,
we show in figure [ fig : f3 ] ( top left ) the entropy distribution inside the star at three different epochs for the case with @xmath50 , where effects of the gr instability is the most prominent .
recall that the polytrope models presume the spatially constant specific entropy in the interior ( e.g. , * ? ? ?
as seen in figure [ fig : f3 ] , however , this is not the case for our accreting stellar models , for which the outer envelope has higher entropy than in the core .
since such an envelope is gravitationally loosely bound , higher total stellar mass is required for the gr instability in the accreting models than in the polytropic model .
another useful expression for the gr - instability criterion is given by @xcite for a polytropic star with the index @xmath51 , which states that the star is gr unstable if @xmath52 , where @xmath53 is the stellar radius .
for a highly radiation - dominant state as in smss , the adiabatic index in the equation of state can be approximated as @xmath54 , where @xmath55 is the ratio of the gas pressure to total pressure .
the criterion for the gr instability can thus be written as @xmath56 note that this condition is obtained for a spatially constant @xmath57 , i.e. , for a constant @xmath58 , which is inversely proportional to the specific entropy @xmath59 as @xmath60 for @xmath61 . in our case
, this condition should be applied to the isentropic core . in figure
[ fig : f3 ] we show the distributions of @xmath62 , @xmath63 , and @xmath40 for @xmath50 at three different epochs with @xmath64 and 0.8 .
we define the surface of the isentropic core ( with mass @xmath65 ) as the shell where the specific entropy becomes larger than the central value by 1% ; @xmath66 and @xmath67 for @xmath68 and 0.8 , respectively . as seen in the figure , as the stellar mass @xmath69 increases , @xmath58 becomes smaller while @xmath70 becomes larger at the same mass shell @xmath71 .
owing to the high value of the specific entropy outside the core , @xmath58 becomes small in this region with the relation @xmath60 .
the mean molecular weight @xmath40 increases toward the center as the hydrogen burning proceeds .
the rise of @xmath70 with the total stellar mass indicates that the more massive star is more concentrated , as in the case for the ordinary primordial stars ( e.g. , * ? ? ?
* ; * ? ? ?
figure [ fig : f3 ] shows that , for @xmath72 , the condition ( [ eq : gr ] ) is not satisfied anywhere and the star is gr - stable . for @xmath73 ,
the isentropic core is still gr - stable as condition ( [ eq : gr ] ) holds for @xmath74 , only outside the isentropic core @xmath75 .
finally , when the stellar mass reaches @xmath76 , equation ( [ eq : gr ] ) is satisfied at the surface where @xmath77 , and the core begins to collapse .
similarly , we find that the gr instability begins to operate satisfying the condition ( [ eq : gr ] ) at some moment for the cases with @xmath78 and @xmath79 , but never with @xmath47 . from this analysis ,
we conclude that the gr instability causes the gravitational collapse for the models with @xmath49 .
we first compare our results with the previous work by @xcite , who study the gr stability of hydrogen - burning stars with fixed masses .
they define a star @xmath80 if the gr instability does not occur during the hydrogen burning stage .
the critical stellar mass for the stability @xmath81 lies between @xmath69 = 0.1 and 0.25 .
however , our calculations show that all the models except for @xmath82 are `` stable '' and begin collapsing after the exhaustion of the hydrogen at the center , when the stellar mass is well above @xmath81 .
specifically , the star becomes gr - unstable during the helium - burning stage for @xmath83 with the final mass @xmath84 .
the main difference from @xcite is that accreting stars that we consider here have a high - entropy envelope .
hence the final masses ( including the outer envelope ) are larger than @xmath85 .
the final stellar masses obtained in this _ letter _ should be regarded as upper mass limits for the remnant bhs after the collapse .
@xcite argue that energetic supernova explosions occur at the collapse of smss in a very narrow mass strip around @xmath86 .
since none of our models end up in this mass range , it is unlikely that such explosions are the final fate of our models
. we will provide further results for this issue using a 1d - gr hydrodynamical code in a forthcoming paper .
preliminary results suggest that the bhs with approximately the same mass as the progenitors will be left behind after the collapse .
if the progenitor star is a rapid rotator , some mass ejection or even explosion may occur at the final stellar collapse .
an accretion disk can appear around a rotating bh and some outflows or jets can be launched .
@xcite , for instance , show by way of a gr hydrodynamics simulation that some portion of the stellar envelope is ejected while most of the stellar matter is drawn into a newborn bh .
we conclude that an accreting protostar at a very large rate of @xmath87 grows to become a sms with mass @xmath88 .
bhs with roughly the same mass will be left as remnants of such smss , which might have sown seeds for the formation of supermassive bhs in the early universe .
we thank masaru shibata and koh takahashi for valuable comments .
this work has been supported by grants - in - aid for scientific research ( 26400220 : hu , 16h05996 , 15h00776 , 25800102 : th , 25287040 : ko , 25287050 : ny ) from japan society for the promotion of science and grant - in - aid for scientific research on innovative areas ( 26104007 : hu ) from the mext in japan .
|
the formation of supermassive stars ( smss ) via rapid mass accretion and their direct collapse into black holes ( bhs ) is a promising pathway for sowing seeds of supermassive bhs in the early universe .
we calculate the evolution of rapidly accreting smss by solving the stellar structure equations including nuclear burning as well as general relativistic ( gr ) effects up to the onset of the collapse .
we find that such smss have less concentrated structure than fully - convective counterpart , which is often postulated for non - accreting ones .
this effect stabilizes the stars against gr instability even above the classical upper mass limit @xmath0 derived for the fully - convective stars .
the accreting sms begins to collapse at the higher mass with the higher accretion rate .
the collapse occurs when the nuclear fuel is exhausted only for cases with @xmath1 . with @xmath2 ,
the star becomes gr - unstable during the helium - burning stage at @xmath3 . in an extreme case with @xmath4 ,
the star does not collapse until the mass reaches @xmath5 , where it is still in the hydrogen - burning stage .
we expect that bhs with roughly the same mass will be left behind after the collapse in all the cases .
|
in simplest terms , spiral galaxy disks can be described as geometrically thin , flat , and circular . we understand that spiral disks are geometrically thin because the gas of which they are composed is cold ( the sound speed of the gas is much smaller than its circular orbital velocity ) ; and they are both circular and flat because , being dissipative , the gas is fairly efficient at both minimizing out - of - the - plane motions and radial excursions that would lead to departures from circular orbits .
describing spiral disks as _ perfectly _ circular and flat is clearly an oversimplification , however .
in addition to the nonaxisymmetric structures that are obvious in optical photographs of many spiral disks , 21-cm maps of the projected velocity fields of spiral disks often reveal isovelocity contours that are significantly twisted ( @xcite ; @xcite ; @xcite ; @xcite , 1980b ; @xcite ; @xcite ) .
kinematical tilted - ring models have been constructed in an effort to explain the presence of such twists in the velocity maps of disks .
the models indicate that the outer regions of many normal spiral disks are significantly warped out of the principal plane that is defined by the optically visible , central portion of each galaxy .
the line of nodes that defines the intersection of adjacent rings of gas in these kinematical models also usually must be twisted significantly as a function of radius in order to explain the observed contour maps ( for recent reviews , see @xcite ; @xcite ; @xcite , hereafter ctsc ) .
it is not particularly surprising that many galaxies are observed to possess extended , rotationally flattened disks because such structures appear to be fairly ubiquitous in gravitationally bound , astrophysical systems .
( there is strong evidence , for example , that rotationally flattened disks either exist now or have existed in the past around our sun , individual planets within our solar system , numerous protostars , the primary star of many mass exchanging binary star systems , and active galactic nuclei . )
what is peculiar about the disks of many galaxies is that the disks are significantly warped .
it is not clear why natural dissipative processes similar to those which work effectively to minimize out - of - the - plane motions in stellar or protostellar `` accretion '' disks are unable to suppress warps in the gaseous disks of galaxies . as binney ( 1992 )
has reviewed , there still is no generally accepted dynamical model of spiral galaxies that satisfactorily explains either the origin or the current structure of warped disks .
almost thirty years ago , hunter & toomre ( 1969 ) examined whether normal , infinitesimal bending oscillations might be exhibited by thin , rotating disks of self - gravitating material , permitting them to sustain coherent warps for more than a hubble time .
they concluded that `` for any disk whose density tapers sufficiently gradually to zero near its edge , '' the frequency spectrum of such oscillations is at least partly continuous and , as a result , coherent warps can not be sustained .
over the subsequent decade , a number of other ideas surfaced to explain the persistence of warps in galaxies , each one taking advantage of the demonstrated existence of dark matter halos around galaxies .
as toomre ( 1983 ) has reviewed , however , models proposing to use the halo as an active agent to excite warps in otherwise flat disks for example , via the mathieu instability ( binney 1981 ) or via a flapping instability ( bertin & mark 1980 ) each present significant difficulties . toomre ( 1983 ) and dekel & shlosman ( 1983 ) proposed , instead , that untwisted steady - state warps may be sustained as a result of steady forcing by a nonspherical , tilted halo .
building on the early work of hunter & toomre ( 1969 ) and the idea that forcing by a nonspherical dark halo can influence the dynamics of the visible disks of galaxies , sparke ( 1986 ) and sparke & casertano ( 1988 ) have shown that a discrete warping mode can persist if the disk is sufficiently self - gravitating and if it is embedded in a nonspherical halo whose equatorial plane is tilted with respect to the centralmost regions of the disk . adopting this model of galaxy warps ,
the following evolutionary picture emerges . during the galaxy formation process , gas which falls into a spheroidal dark matter halo
generally will find that its angular momentum vector is tipped at some nonzero angle , @xmath0 , away from the symmetry axis of the halo .
if the gas is cold , it will settle into a rotationally flattened disk that is tilted at the same angle @xmath0 with respect to the equatorial plane of the halo . in the centralmost regions of the galaxy , where the self - gravity of the gas ( or , ultimately , the combined gas / star system ) dominates over the gravitational influence of the halo , the gas will be content to remain in orbits that preserve this original tilt . in the outermost regions of the galaxy
where the gravitational field of the halo dominates , however , the gas should settle into the halo s equatorial plane .
as toomre ( 1983 ) and dekel & shlosman ( 1983 ) both sketched in their original concept papers , there also should be an intermediate region where the gas will be influenced significantly by the nonspherical gravitational fields of both the halo and the central gas ( or gas / star ) disk . through their modeling efforts , sparke ( 1986 ) and sparke & casertano ( 1988 ) confirmed this earlier suspicion that in the intermediate region , the gas can reside in a steady - state , `` warped disk '' structure that provides a smooth radial transition between the separate `` flat disk '' orientations of the inner and outer regions of the galaxy . concentrating on the dynamics of the centralmost and intermediate regions that is , by building models in which there was effectively no gas in the outermost regions
sparke & casertano ( 1988 ) showed that , in steady - state , the warped disk exhibits a straight line of nodes which precesses slowly and coherently in a direction retrograde to the orbital motion of the gas . in a time - dependent simulation , furthermore , hofner & sparke ( 1994 ) showed that settling to this steady - state warped disk structure occurs from the inside , out , and is driven not by dissipative processes that are similar to those which are thought to drive settling in most stellar or protostellar accretion disks but , rather , because `` bending waves carry energy associated with transient disturbances out toward the disk edge . ''
they also showed that , during an evolution as the bending waves propagate outward through the intermediate region of the disk , a twisted structure can develop and persist until the gas has had sufficient time to settle into the steady - state ( constant line of nodes ) configuration .
sparke & casertano ( 1988 ) and hofner & sparke ( 1994 ) have demonstrated that , with an appropriate choice of parameters , this model of disk warping matches well the observed properties of several galaxies with warped disks .
( see also kuijken 1991 and dubinski & kuijken 1995 . ) as hofner & sparke ( 1994 ) have pointed out , in galaxies with extended disks `` the outermost gas can not be expected to form part of a coherent warping mode .
'' they did not include normal dissipative forces in their simulations and therefore were unable to comment on how such forces might influence the settling process . in this paper
, we examine the disk - settling process from the other extreme , ignoring the self - gravity of the gas but introducing an effective kinematical viscosity into the dynamical equations in order to simulate the effects of dissipative forces . hence
, our effort is complementary to the work of hofner & sparke ( 1994 ) and is most relevant to galaxies with extended disks although there is one galaxy used for model comparisons ( ngc 2841 ) that is shared by both works .
we adopt the view that warps in extended disks which exhibit substantial twists are transient features .
independent of precisely what physical process was responsible for initially placing the gas in an orbit that is inclined to the halo s equatorial plane ( _ e.g. _ , gas infall at the time of formation or a recent tidal encounter with another galaxy ) , the twisted structure can be understood as the result of differential precession in the gaseous disk as it dissipatively settles toward that `` preferred plane . '' in the past , there has been considerable concern ( first enunciated by kahn & woltjer [ 1959 ] , but reiterated in the reviews by both toomre [ 1983 ] and binney [ 1991 ] ) that differential precession will destroy any warped disk structure on a time scale that is short compared to a hubble time and , therefore , that the mechanism we are examining can not reasonably be used to explain the persistence of such structures . in the outermost regions of disks , however , precession times are relatively long and , as was first pointed out by tubbs & sanders ( 1979 ) , a warped gas layer can persist for a hubble time if the dark halo in which the disk is embedded deviates only slightly from spherical symmetry . by modeling carefully the process of disk settling that is driven by normal dissipative forces and comparing the models to the observed kinematical properties of galaxies with extended , warped layers , we hope to be able to more carefully examine the viability of such models .
steiman - cameron & durisen ( 1988 , hereafter scd88 ) have developed this idea rather extensively .
they have adopted a numerical , cloud - fluid model to simulate the time - dependent evolution of a galaxy disk that intially is tilted out of the equatorial plane of an underlying , spheroidal dark halo .
the disk is assumed to be composed of a set of annular mass elements , or `` clouds , '' which act like atoms in a viscous fluid .
the scd88 model has offered some valuable physical insight into the time - dependent settling process that is driven by normal dissipative forces and their dynamically generated model of a twisted galaxy disk has been surprisingly successful at matching the peculiar optical image of one particular galaxy , ngc 4753 ( @xcite ) .
our model is analogous to the one developed by scd88 but it derives from an analytical prescription of the viscous settling process .
more specifically , we employ the `` twisted - disk '' equation formalism first developed by bardeen & petterson ( 1975 ) and petterson ( 1977 , 1978 ) to describe the time - dependent settling of a thin , viscous disk in a nonspherical dark halo potential .
this is a rather natural formalism to adopt because , as numerous kinematical `` tilted - ring '' models have demonstrated , warped galaxy disks display a structure that resembles , at least qualitatively , the twisted geometry that had once been thought to be important in accretion disks which surround certain compact stellar objects ( bardeen & petterson 1975 ; see a recent rejuvenation of this idea put forward by maloney & begelman 1997 ) . in adapting the model to galaxy disks ,
we have replaced the approximate keplerian gravitational potential used in earlier accretion disk work with a logarithmic potential appropriate to galaxy halos .
( pringle [ 1992 ] also recently described how the twisted disk formalism may be adapted to galaxies . ) in the limit of stress - free precession , our model reproduces the analytical prescription of disk settling first presented by scd88 , but our model is not constrained to this limit .
a more general solution to the governing equations predicts an exponential settling rate that depends on time to the first power , rather than on time to the third power as has been derived in the limit of stress - free precession .
furthermore , an analytical , steady - state solution to the governing equations produces a twisted - disk structure that is very similar to previously constructed , kinematical models .
we demonstrate that projected surface density maps and radial velocity maps derived from our analytical model match published maps of five well - studied warped disk galaxies ( m83 , ngc 300 , ngc 2841 , ngc 5033 , and ngc 5055 ) very well .
bardeen & petterson ( 1975 ) and petterson ( 1978 ) have written the hydrodynamic equations determining the structure and evolution of nonplanar , thin accretion disks in a `` twisted - coordinate system . '' in this system , the position @xmath1 on each ring of radius @xmath2 is referenced to a local cylindrical coordinate frame @xmath3 which has been rotated with respect to the cartesian coordinate system of the central reference ring by the two orientation angles @xmath4 and @xmath5 ( see fig . 1 ) .
for an appropriate choice of the two functions @xmath6 and @xmath7 , including specifically the assumption that @xmath8 , the equations separate into a set of the usual hydrodynamic equations for a flat disk and a pair of coupled `` twist - equations '' governing the orientation of the disk . because of its ability to describe an apparently complicated , fully three - dimensional dynamical problem with a relatively simple and elegant mathematical model , a number of different groups have subsequently also adopted this formalism . however , until recently there has been some disagreement over what terms may be dropped from the fully three - dimensional , nonlinear partial differential equations ( due to their `` smallness '' relative to other terms ) when deriving the governing twist - equations .
( for a complete discussion of the various simplifications and assumptions made in deriving this formalism see petterson [ 1978 , 1979 ] ; hatchett , begelman , & sarazin [ 1981 ] ; and papaloizou & pringle [ 1983 ] . ) as the following brief review points out , the disagreements have not been over the general _ form _ of the equations but , rather , over the precise value of certain order - unity coefficients . ultimately ,
as explained below , we have adopted the derivation presented by papaloizou & pringle ( 1983 ) . according to petterson ( 1978 ) , the twist - equations take the following form : @xmath9\ ; c_1\frac{\beta^\prime}{r}-\gamma^{\prime2}\beta\biggr]\ ; + \;\frac{1}{2\pi v_{\psi}}\int_0^{2\pi } f_{t}\cos\psi\,d\psi , \nonumber \\ \\ \dot \gamma+v_{r}\gamma^\prime&= & \frac{\nu}{2}c_0 \biggl[\gamma^{\prime\prime}+
c_1\frac{\gamma^\prime}{r}+ \frac{2\gamma^\prime\beta^\prime}{\beta}\biggr ] \;+\;\frac{1}{2\pi\beta v_{\psi}}\int_0^{2\pi } f_{t}\sin\psi\,d\psi,\nonumber \label{p78teqs}\end{aligned}\ ] ] with @xmath10 and @xmath11 . in these expressions , dots and primes denote differentiation with respect to time and space , respectively , @xmath12 and @xmath13 are the radial and azimuthal components of the fluid velocity , @xmath14 is the vertically averaged kinematical viscosity , and @xmath15 is an externally supplied `` twisting force '' .
hatchett , begelman , & sarazin ( 1981 , hereafter hbs ) pointed out some inconsistencies between petterson s ( 1978 ) derivation and the earlier presentation by bardeen & petterson ( 1975 ) . through an independent derivation
, they determined that the twist - equations do take the form of equation ( 1 ) , but they concluded that the two coefficients should have the values @xmath16 and @xmath17 .
most significantly , hbs showed that the pair of twist - equations can be written as a single , _
complex _ twist - equation of the form @xmath18 where the complex variable _ w _ is defined in terms of the angles @xmath5 and @xmath4 as @xmath19 and @xmath20}\,d\psi\;. \label{ldef}\ ] ] written in this form , the twist - equation readily submits to analytical solution for certain driving forces .
papaloizou and pringle ( 1983 , hereafter pp ) have suggested that the hbs derivation also has shortcomings ( for example , as viscous evolution occurs according to the hbs equations , angular momentum does not appear to be conserved globally ) .
they have derived a set of twist - equations that more consistently ties in with traditional @xmath21-disk models and have demonstrated explicitly that their derived equations conserve angular momentum .
as can most easily be deduced from the presentations of kumar & pringle ( 1985 ) and pringle ( 1992 ) , for a disk whose inner radius goes to zero and whose surface density is independent of radius , the pp derivation also leads to a complex twist - equation that takes the form of equation ( [ hbsteq ] ) . however , adopting pp s `` naive approach '' in order to gain physical insight into the nature of the problem ( i.e. , setting kumar and pringle s function @xmath22 ) , one concludes from these three papers that the proper coefficients are : @xmath23 @xmath24 , \eqnum{5b } \label{ppcoefs}\ ] ] where @xmath25 for a keplerian disk @xmath26 ; therefore , @xmath27 and @xmath28 . interestingly , these are the same coefficients that appear in bardeen & petterson s ( 1975 ) earliest presentation of the twist equations .
henceforth , we will adopt and build upon the twist - equations derived by pp because it is clear that in a viscous disk evolution that is governed by the pp twist - equations , angular momentum is conserved .
petterson ( 1978 ) , hbs , and pp each used an approximately keplerian potential to represent the underlying gravitational potential well that governs disk dynamics . while the keplerian potential correctly describes the gravitational field in which a thin accretion disk surrounding a compact object sits , it is inappropriate for a thin galaxy disk sitting in the potential well of a nonspherical dark - matter halo . utilizing kumar & pringle s ( 1985 ) and pringle s ( 1992 ) extensions of the pp derivation ,
it is possible to generalize the complex twist - equation to incorporate an arbitrary power - law form for the gravitational potential .
specifically , adopting a rotation law of the form @xmath29 expression ( [ ppcoefs ] ) becomes @xmath30.\ ] ] furthermore , as petterson ( 1978 ) has pointed out , the radial velocity can be expressed in terms of the vertically averaged kinematical viscosity as @xmath31 employing our generalized rotation law , we therefore can write @xmath32 using this expression in conjunction with equations ( 5a ) and ( 7 ) , the complex twist - equation ( 2 ) takes the form : @xmath33 + i{\it l}\;. \label{btfteq1}\ ] ]
as outlined in @xmath34 1 , our objective is to examine the evolution of a cold gaseous disk that is embedded in a spheroidal , dark halo potential . in general , we assume that when the gas is initially introduced into the halo potential well , the orientation of its angular momentum vector is not aligned with the symmetry axis of the halo . because the halo is not spherically symmetric , the halo s gravitational field will exert a finite `` twisting force '' on the gaseous disk ; in simplest terms , this field forces rings of gas at different radii to precess about the symmetry axis ( or symmetry plane ) of the halo . in the present context , a spheroidal dark - matter halo can be satisfactorily represented by a scale - free logarithmic potential ( cf .
@xcite ; @xcite , 1990 ) .
this potential is a useful approximation because a circular disk orbiting inside such a potential will exhibit a flat rotation curve ( i.e. , @xmath35 and @xmath13 is independent of radius ) , as seen in real galaxies . in a halo potential of this type ,
the precession frequency about the symmetry axis of the potential @xmath36 is @xmath37 where @xmath38 is a constant which measures the strength of the quadrupole distortion of the halo and the minus ( plus ) sign indicates that the spheriodal halo is oblate ( prolate ) . for a twisted disk embedded in a spheroidal potential well , the twisting force that enters
the derived twist equation assumes precisely the same algebraic form as the twisting force that bardeen & petterson ( 1975 ) used when modeling the evolution of a disk orbiting a nonspherical , compact stellar object , namely , @xmath39 hence , from equation ( [ ldef ] ) , @xmath40 and we deduce that the evolution of a thin galaxy disk in a scale - free logarithmic halo potential can be described by the expression @xmath41
\pm i\biggl(\frac{3}{4}\eta v_{\psi}\biggr)\frac{w}{r}\;. \label{btfteq3}\ ] ] unless explicitly stated otherwise , we henceforth will adopt the minus sign in front of the last term of this expression , thereby assuming that the halo is _ oblate _ in shape . before the character of _ general _ solutions to this complex ordinary differential equation is examined , it is worthwhile to demonstrate that when certain limiting physical conditions are imposed , familiar analytical expressions for the time variation of the inclination angle @xmath42 and the `` twist '' angle @xmath43 are derivable from ( [ btfteq3 ] ) .
this will give us additional confidence that ( [ btfteq3 ] ) offers a valid description of the dynamics of a twisted galaxy disk .
we present , first , the limit of viscous - free precession , then the limit of stress - free precession . in the limit where viscous effects are completely negligible
, @xmath44 and equation ( [ btfteq3 ] ) reduces to the case of free precession , @xmath45 with no associated evolutionary damping of orbit inclinations , that is , @xmath46 equation ( 15a ) can be integrated to give @xmath47 where @xmath48 prescribes the twisting of the disk at time @xmath49 . if the viscous stresses due to twisting have negligible effect on differential precession during the settling process a condition that scd88 refer to as `` stress - free precession ''
one may set @xmath50 in the governing , complex twist - equation . under this condition ,
equation ( [ btfteq3 ] ) predicts the following evolutionary behavior : @xmath51 @xmath52 this pair of equations matches equations [ 37a ] and [ 37b ] of scd88 except that the first term on the rhs of equation [ 17a ] is a factor of 2 larger than the equivalent term in equation [ 37b ] of scd88 .
we suspect that the scd88 result is in error , but we have been unable to identify where scd88 dropped the factor of 2 because their published derivation was not sufficiently clear . if we furthermore adopt the `` stress - free '' precession condition introduced by scd88 , the first term in equation ( 17a ) is set to zero ( hence the discrepant factor of 2 is not an issue ) and the equation reduces to the form of equation ( 15a ) hence , @xmath53 with @xmath54 , equation ( 17b ) specifically integrates to give @xmath55\;,\ ] ] where @xmath56^{-\frac{1}{3}}\;,\ ] ] and @xmath57 is the value of the inclination at @xmath49 . hence the analytical settling model presented by scd88
( see also steiman - cameron , kormendy , & durisen 1992 ) can be straightforwardly derived from our generalized twist - equation .
it is not at all clear how widely applicable the scd88 analytical settling model is to the warped disks of normal spiral galaxies because the simplifying assumptions ( e.g. , @xmath58 ) are quite limiting .
however , with the analytical twist - equation in hand , we can perform a less restrictive examination of such systems .
we begin by rewriting equation ( [ btfteq3 ] ) in the form @xmath59 where derivatives are now taken with respect to the dimensionless time and space variables @xmath60 and @xmath61 if we assume that the spatial and temporal parts of the complex angle @xmath62 are separable , i.e. , @xmath63 then @xmath64,\ ] ] where @xmath65 is in general complex , and @xmath66 must be a solution of @xmath67 the above equation can be transformed by the substitutions @xmath68 and @xmath69 , into a form which shows its relationship to bessel s equation @xmath70 the character of the solutions and the spectrum of @xmath65 are determined by boundary conditions , but some general comments can be made without reference to specific boundary conditions or initial state . the fact that one can do so is important here since we do not have a well - posed problem in which we know the initial state and the relevant boundary conditions . for each @xmath65
there is a radius @xmath71 inside which the third term inside the parantheses in equation ( 27 ) may be neglected . in this region
the solution behaves as @xmath72 .
we rule out @xmath73 because it is not finite at the origin ( abramowicz & stegun 1972 ) whereas for plausible initial states @xmath62 must be finite everywhere . in general , the spectrum of @xmath65 may contain both growing and decaying modes .
we do not consider growing modes since their presence would indicate instability .
the fact that one can fit a number of observational cases with the simpler solution in which only decaying modes are present may be considered a justification _ a posteriori _ of this assumption . in any case , it is clear that if an arbitrary initial state can be written as a superposition containing many values of @xmath65 satisfying appropriate boundary conditions , the behaviour in the central region will approach the solution with @xmath74 .
this tendency manifests itself as a function of time first near the center and later further out as any contribution with large @xmath75 dies away rapidly .
since the solution with @xmath74 is formally the steady state solution , we would expect it to apply more or less universally to the interior regions of twisted disks , no matter what the initial conditions were .
the memory of the initial conditions is erased inside out leaving at most an exponentially decaying twist in the innermost regions . because the functions of primary interest to us ultimately are @xmath76 and @xmath77
, we note that @xmath78^{\frac{1}{2}}\exp \biggl[-re\bigl(k^2)\tau\biggr ] , \label{beta}\\ \gamma&=&\bigl(\theta + \phi\bigr)+\frac{\pi}{2},\end{aligned}\ ] ] where : @xmath79 because @xmath80 , we recognize also that for a given @xmath65 , the time dependent solution of equation ( 21 ) presents a wave ( with a `` twisted '' spiral character ) that propagates azimuthally through the disk with a pattern speed @xmath81 drawing on the hbs discussion , we realize that an analytical solution to the steady - state problem can be derived . specifically , setting @xmath82 in equation ( [ btfteq4 ] ) or @xmath83 in equation ( [ zeq ] ) , we obtain formally @xmath84\ ; , \label{btfteq5}\ ] ] where @xmath85 is any first - order bessel function of the first kind ( see gradshteyn & ryzhik 1965 , 8.491 ,
equation 3 ) .
the physically most relevant solution appears to be @xmath86 because in this case @xmath5 is finite as @xmath87 and as @xmath71 increases , @xmath5 increases monotonically ( abramowicz & stegun 1972 ) . as a point of reference for all subsequent models ,
then , we define the steady - state twisted - disk function @xmath88 = \frac{1}{\sqrt{x}}[{\rm ber}_1(2\sqrt{x } ) + i { \rm bei}_1(2\sqrt{x})]\ ; ; \label{wss}\ ] ] where @xmath89 and @xmath90 are the bessel real and bessel - imaginary or kelvin functions ( see e.g. mclachlan 1941 ; abramowicz & stegun 1972 ) .
this solution leads to the functions @xmath91 and @xmath92 shown in fig . 2 .
from ( [ wss ] ) one can obtain @xmath93 and @xmath94 in closed form as follows : @xmath95 where @xmath96 and @xmath97 are the modulus and phase functions for @xmath98 and @xmath99 . over the interval @xmath100 , @xmath93 and @xmath94
are approximated well by the expressions : @xmath101.\end{aligned}\ ] ] as we shall illustrate presently , the _ form _ of the function @xmath102 over the interval @xmath100 defines a warped disk model whose features match the observed properties of a number of individual disks .
the steady - state solution just derived is formally inappropriate for real galaxies especially because the outermost regions of real galaxy disks are unlikely to be described well by time - independent conditions .
however , over the ( innermost ) regions of a galaxy disk where differential precession and viscous dissipation have been able to effect appreciable settling in a hubble time , we expect the spatial structure of the disk to assume a form that is very similar to the one portrayed by the function @xmath103 .
more specifically , the discussion in 3.3 and equations ( [ beta])-([phi ] ) suggest a general time - dependent behavior of the form : @xmath104 , \label{gammass}\end{aligned}\ ] ] where @xmath105 is a spatial constant that describes the amplitude of the warp at a given time and whose time - evolution follows the above form with @xmath106 being the lowest @xmath75 compatible with boundary conditions . again , since we do not have a well - posed problem , all we can say is that whatever restrictions the boundary conditions impose on the spectrum of @xmath65 , only the lowest @xmath75 will survive after some time .
when comparing our model to the properties of real galaxy disks , we will treat @xmath107 as a free parameter that measures the overall amplitude of a galaxy s warp and not be immediately concerned about the values of @xmath57 , @xmath106 , or the age of each disk .
the other free parameter that may be adjusted before our twisted disk model is compared to real galaxy disks is @xmath108 , the radial cutoff to the function @xmath103 that will correspond to the maximum radial extent @xmath109 of a galaxy s disk . although @xmath109 can be specified for individual galaxies ( see table 1 ) , the length scale @xmath110 as defined in equation ( [ dimr ] ) can not be specified _ a priori _ because neither of the physical parameters @xmath14 or @xmath38 is known for individual galaxies . hence , initially it was unclear to us what cutoff radius @xmath111 ( _ i.e. _ , what range @xmath112 ) should be adopted for our model .
fortunately , from previously published tilted
ring models of warped disks we were able to determine how rapidly the warp angle @xmath5 and the twist angle @xmath4 vary with radius in real galaxies . more specifically , we measured the pitch angles " @xmath113 and @xmath114 in several systems , then looked for the range(s ) of @xmath71 in our steady
state model over which both pitch
angle values occurred simultaneously .
we found good matches only when @xmath115 , that is , galaxy models which match the observations can not be produced with values of @xmath111 that are orders of magnitude less than or greater than @xmath116 .
previously constructed tilted ring models also directed the choices of the maximum warping angles @xmath107 in our models . by selecting a single value of @xmath111 ( specifically , @xmath117 ) and a rather narrow range of warp amplitudes @xmath107 ,
we have been able to match many of the qualitative features that are frequently seen in the published maps of galaxies .
for example , using the model parameters identified in table 1 , we have produced the surface brightness maps and radial velocity maps depicted in fig . 3 for comparison with published maps of the galaxies m83 , ngc300 , ngc2841
, ngc5033 and ngc5055 . figure 3 actually has been pieced together from five separate frames of a @xmath118 600-frame digital animation sequence in which the viewer `` flies around '' our model disk , examining it from a variety of different lines of sight . the two angles @xmath119 and @xmath120 uniquely define the orientation of the observer s line of sight with respect to the @xmath121 cartesian coordinate system ( see fig .
1 ) that is fixed in the body of the galaxy .
( when the inclination angle @xmath122 or @xmath123 the galaxy is being viewed face - on with its angular momentum vector pointing directly at the observer ; when @xmath124 the galaxy is seen edge - on with its angular momentum vector pointing down . when the azimuthal orientation angle @xmath125 the @xmath126 axis is pointing to the right in each image ; when @xmath127 the @xmath128 axis is pointing to the right . ) in each frame of fig . 3 ,
the disk is displayed in three different ways : the righthand image is the disk s projected 2d velocity contour map ; the central image is the disk s projected 2d surface brightness map ; and the lefthand image presents a 3d rendering of an isodensity surface that encloses virtually the entire disk .
each of the velocity contour maps and surface brightness maps in fig .
3 were produced with the 3d radiative transfer routine developed by vth ( 1994 ) ; each 3d isodensity surface was generated using the idl imaging package . in the lefthand image of each frame ,
the light source reflecting from the disk s surface originates from the same position as the viewer s eye .
the five frames displayed in fig . 3 were selected from our animation sequence because they identify observer lines of sight from which our model s projected 2d maps match well the published maps of the five indicated galaxies .
ngc 2841 : to produce the frame of fig .
3 that is identified as ngc 2841 , we set @xmath129 in our model disk which corresponds to a maximum warp at the edge of the disk of @xmath130 . in fig . 3 ,
the galaxy s angular momentum vector is pointed almost directly straight up ( the disk is tipped slightly so that the viewer is seeing the `` top '' side of the galaxy ) , so velocity contours to the right of the kinematic minor axis are red - shifted and contours to the left are blue - shifted .
our 2d projected surface - brightness and velocity maps should be compared , respectively , with the maps published as figs .
7a and 7b in bosma ( 1981 ) .
bosma s maps have been reprinted here in fig .
in order to make the relevant comparisons , the observational maps have been rotated counterclockwise approximately @xmath131 in order to properly orient them with respect to our model images .
after this rotation has been made , as our 3d rendering illustrates , the southern edge of our model disk is farther away from the observer than is its northern edge ; this is consistent with the tilt interpretation one derives from an optical photograph of ngc 2841 and dust obscuration arguments ( cf .
plate 14 in sandage 1961 and table 3 of ctsc ) .
the velocity contour map of our model of ngc 2841 displays the same gentle , counter
clockwise twist as the observed velocity map .
furthermore , the north - south extensions seen in bosma s surface - brightness map of ngc 2841 can be identified with the `` leading arm '' features that appear in the projected surface - brightness image of our twisted disk model .
m83 : to produce the frame of fig .
3 that is identified as m83 , we set @xmath132 in our model disk , which corresponds to a maximum warp at the edge of the disk of @xmath133 .
the model is oriented so that , in fig . 3 , the galaxy s angular momentum vector is pointed into the page , almost directly away from the observer . ( with this orientation , m83 s optically visible spiral arms cf .
plate 28 in sandage 1961 are `` trailing '' spiral features . )
the disk is not being viewed precisely face - on ; as displayed in fig .
3 , it is oriented such that , even if the disk were perfectly flat , its upper half would be tipped slightly away from the observer and its lower half toward the observer . as a result of this orientation ,
velocity contours to the right of the kinematic minor axis are blue - shifted and contours to the left are red - shifted .
our 2d projected surface - brightness and velocity maps should be compared , respectively , with the maps published as figs . 2 and 3 in rogstad ,
lockhart , & wright ( 1974 ; hereafter rlw ) .
rlw s maps have been reprinted here in fig .
in order to make the relevant comparisons , the observational maps have been rotated counterclockwise approximately @xmath134 in order to properly orient them with respect to our model images .
the velocity contour map produced by our twisted disk model of m83 displays most of the broad features that appear in the velocity map published by rlw .
the surface - brightness map published by rlw displays a pair of faint , `` leading spiral '' arms that are not noticeable in the image we have produced in fig . 3 from our model .
( the features are actually present in our model at a very low amplitude level and can be enhanced somewhat by adjusting the model parameter @xmath107 .
see , for example , the tilted ring model developed by rlw and the model published by ctsc . )
analogous features _ are _ present in our surface - brightness image of ngc 300 , presented below . as the 3d isodensity surface of our model of m83 suggests , these spiral arm features probably appear in projected surface - brightness maps of m83 because the galaxy s twisted disk is bending toward the observer in the northwestern quadrant and away from the observer in the southwestern quadrant .
ngc 5033 : to produce the frame of fig .
3 that is identified as ngc 5033 , we set @xmath135 in our model disk which corresponds to a maximum warp at the edge of the disk of @xmath136 . in fig . 3 ,
( see also fig . 5 and the related discussion in 4.3 , below ) ,
the galaxy s angular momentum vector is pointed down and away from the observer , so velocity contours to the right of the kinematic minor axis are blue - shifted and contours to the left are red - shifted .
our 2d projected surface - brightness and velocity maps should be compared , respectively , with the maps published as figs .
4a and 4b in bosma ( 1981 ) .
bosma s maps have been reprinted here in fig .
in order to make the relevant comparisons , the observational maps have been rotated counterclockwise approximately @xmath137 in order to properly orient them with respect to our model images .
both the velocity contour image and the surface - brightness image produced by our model of ngc 5033 bear a striking resemblance to bosma s published maps .
the relatively steep density gradients that are visible at the northeast and southwest edges of bosma s surface - brightness maps are , according to our model , clearly due to the disk bending toward the observer at one edge and away from the observer at the other .
ngc 5055 : to produce the frame of fig .
3 that is identified as ngc 5055 , we set @xmath138 in our model disk which corresponds to a maximum warp at the edge of the disk of @xmath139 .
this galaxy s angular momentum vector , like that of ngc 5033 ( notice that @xmath140 is the same in these two systems ) , is pointed down and away from the observer , so velocity contours to the right of the kinematic minor axis are blue - shifted and contours to the left are red - shifted .
our 2d projected surface - brightness and velocity maps should be compared , respectively , with the maps published as figs .
6a and 6b in bosma ( 1981 ) .
these maps are reprinted in fig .
4d of this manuscript . in this case
, the observational maps did not need to be rotated more than @xmath141 clockwise in order to make the relevant comparisons .
again , the velocity contour image and the surface - brightness image produced by our model bear a strong resemblance to bosma s published maps .
notice , in particular , that from this line - of - sight , thin arcs are displayed in the upper lefthand and lower righthand regions of our model s projected surface - brightness map .
bosma s surface - brightness map definitely displays similar features .
ngc300 : to produce the frame of fig .
3 that is identified as ngc300 , we set @xmath138 in our model disk , which corresponds to a maximum warp at the edge of the disk of @xmath139 .
the model is oriented so that the galaxy s angular momentum vector is pointed out of the page , almost directly at the observer .
the disk is not being viewed precisely face - on ; as displayed in fig .
3 , it is oriented such that , even if the disk were perfectly flat , its upper half would be tipped slightly toward the observer and its lower half away from the observer . as a result of this orientation ,
velocity contours to the right of the kinematic minor axis are blue - shifted and contours to the left are red - shifted .
our 2d projected surface - brightness and velocity maps should be compared , respectively , with the maps published as figs . 1 and 2 in rogstad , crutcher , & chu ( 1979 ; hereafter rcc ) .
rcc s maps have been reprinted here in fig .
in order to make the relevant comparisons , the observational maps have been rotated counterclockwise approximately @xmath142 in order to properly orient them with respect to our model images .
notice that after this rotation has been made , our 3d rendered image of ngc300 closely resembles the perspective drawing of rcc s tilted - ring model ( see their fig .
notice , first , that the velocity contour image of our model of ngc 300 closely resembles rcc s published velocity map .
in addition , our surface - brightness map displays a pair of faint , `` leading spiral '' arms that touch the left and right edges of the image shown in fig .
we believe that these arms provide an explanation for the faint features that extend to the northwest and , particularly , southeast regions of the rcc map .
the model that was used to generate images for fig . 3 was constructed assuming that the precessional frequency @xmath36 was negative ( see eq .
that is , the underlying halo was assumed to be _ oblate _ in shape and accordingly , the precessional motions retrograde with respect to the orbital motion of the gas . because the magnitude of the precessional frequency _ decreases _ with increasing radius in each disk , retrograde precession results in a physical twist that has a `` leading '' appearance in the 3d , rendered image of each system .
( in contrast , optically visible spiral arms are generally thought to be `` trailing '' features . ) knowing this , one can discern whether the angular momentum vector of each galaxy is pointing into ( out of ) the page by simply noticing whether the warped structure exhibits an overall clockwise ( counterclockwise ) twist .
as ctsc pointed out , an identically good fit to the maps can be obtained with a disk that exhibits a retrograde twist ( i.e. , one that settles into a prolate spheroidal halo ) as with a disk that exhibits a prograde twist .
after switching the sign of @xmath143 in eq .
( 11 ) to reverse the sense of the twisting in the 3d model , one need only adjust one s line - of - sight viewing angle according to the prescription @xmath144 in order to generate projected velocity and surface - density maps from a `` prolate '' model that are identical to the maps generated from and `` oblate '' model . fig .
5b illustrates what the `` prolate '' model for ngc 5033 looks like once this transformation is employed .
if two quite different 3d disk structures ( one with a prograde twist and one with a retrograde twist ) can generate identical 2d maps when projected onto the sky , how can one decipher which offers the physically more realistic model ?
as ctsc pointed out , information gleaned from optical photographs can often provide the supplementary information that is needed to identify which model is more realistic .
for example , one must insist that the angular momentum vector of m83 points away from the observer ( _ into _ the page on fig .
3 ) if its optical spiral arms are to be interpreted as _ trailing _ features ; hence a warped disk model with a prograde twist ( and settling in an _ oblate _ spheriodal halo ) as illustrated in fig .
3 is the preferred solution .
similar arguments ( or ones based on dust obscuration ) lead one to conclude that the disks of ngc 2841 , ngc 5055 , and ngc 300 are also settling in oblate spheriodal halos . however , for ngc 5033 , the retrograde twisting ( prolate spheriodal halo ) model is clearly preferrable to the prograde twisting model .
this statement is supported by two arguments : first , dust obscuration in an optical photograph of ngc 5033 ( cf .
4a of ctsc and panel 127 of sandage & bedke 1994 ) identifies the western edge of the galaxy as nearer to the observer than its eastern edge ; second , model 5b must be chosen over 5a if the optical spiral arms are to be interpreted as trailing features .
hence , the 3d image of ngc 5033 shown in fig .
5b is the correct depiction of the galaxy as inferred from our model fitting .
it should be noted that the animation sequence from which the frames shown in fig . 3 were selected was originally constructed with the expressed intention of flying through lines of sight that we previously had determined would produce good fits to m83 , ngc 5033 and ngc 300 ; the only intrinsic model parameter that was adjusted during the fly - by sequence was the overall warp amplitude @xmath107 . during our viewing of the fly - by sequence , we also spotted images that matched the published observational maps of ngc 2841 and ngc 5055 .
the frames presented in fig .
3 have been taken directly from the original animation sequence ; absolutely no attempt has been made to fine - tune the model parameters to fit these two additional galaxies separately . clearly , then
, it is primarily the line - of - sight viewing angle to each disk and not the parameters defining the underlying structural properties of each disk that had to be adjusted from galaxy to galaxy in order to match the observations .
considering its simplicity in particular , the significantly reduced number of free parameters as compared to previous tilted - ring models the agreement between the projected maps of our model and observed maps is excellent .
we conclude , therefore , that the function @xmath103 and the corresponding expressions for @xmath93 and @xmath94 given by eqs .
( 36a ) and ( 36b ) provide a good template for matching the observed kinematical structure of disks . because a single value of the parameter @xmath111 and little variation in the parameter @xmath107 are required to achieve a good model fit to five separate galaxies ,
we conclude that these five systems intrinsically have very similar warped disk structures .
most importantly , however , since the function @xmath103 derives from a physically reasonable dynamical picture of disk settling , there is an expectation that our model provides insight into the underlying physical properties of these systems .
for example , since @xmath117 for each galaxy , we conclude from definition ( [ dimr ] ) that in each galaxy , the ratio @xmath145 where @xmath109 is the outer radius of the observed disk .
( table 1 shows the values of @xmath146 calculated individually for each modeled system . )
also , if our model is relevant at all , the existence of each warped layer indicates that viscous settling has been effective in a hubble time ( @xmath147 ) and @xmath148 . without a proper understanding of the boundary conditions , we are unable to determine the correct physical value of @xmath106 . assuming for the moment that @xmath149 and setting @xmath150 , we conclude from definition ( [ dimt ] ) that @xmath151 these two ratios can only hold if , in each system , @xmath152 and @xmath153 it is worth noting that the values obtained for @xmath14 in all cases ( see table 1 ) are consistent with a rough estimate of the viscosity due to dissipational cloud cloud collisions in the solar neighborhood where @xmath154 km / s kpc ( lynden - bell & pringle 1974 ) .
very modest variations in cloud dimensions , r.m.s . velocity and filling factor from galaxy to galaxy
are sufficient to reproduce the range of values shown on table 1 .
the fact that reasonable viscosity estimates are obtained by assuming @xmath155 may be considered as a posteriori justification for our choice .
utilizing the formalism originally introduced by petterson to describe warped and twisted accretion disk structures in keplerian potentials , we have derived a single , complex ode to describe time dependent settling of an disk in the logarithmic potential that appears to be typical of normal spiral galaxies . over the interval @xmath100 , the analytical function @xmath103 derived from an analysis of the steady - state limit of the general twisted - disk equation appears to describe quite accurately the general warped and twisted structure that is exhibited by a number of galaxy disks .
it should be noted again that our analysis is based on the assumption that the warping angle @xmath5 is @xmath156 . for galaxies with larger warps ( including some that we have modeled )
this is a simplifying assumption and should be disregarded in more proper treatments ( see pringle [ 1992 ] ) . from our model
fits we conclude that , quite generally , the effective kinematical viscosity in these neutral hydrogen disks is @xmath157 km s@xmath158 kpc .
that is , the effective reynolds number in these systems is @xmath159 according to equation ( [ radvsub ] ) , this also implies that the ratio of the radial inflow velocity of the gas to its orbital velocity is @xmath160 this model also provides a mechanism by which the parameter @xmath38 the quadrupole moment of the underlying dark halo potential well can be measured in spiral galaxies .
our fits to five normal spirals with well - studied warped disks specifically indicate that @xmath161 .
( this value can be increased to @xmath162 , with a corresponding factor of ten increase in viscosity , only if the age of these disks is assumed to be 0.1@xmath163 which seems unlikely or @xmath106 in expression 38c is found to be @xmath164 . )
hence , we conclude that the dark halos in which these warped disks sit are , to quite high accuracy , spherically symmetric
. this particular conclusion should not come as a surprise because some time ago tubbs and sanders ( 1979 ) pointed out that if warped disks are identified as transient structures , the warp can only be sustained for a hubble time if the underlying halo potential is very nearly spherical .
although we have examined in detail here only the steady - state solution , the derived time - dependent twisted - disk equation provides a tool that can be utilized to model the time - evolution of warped disks without resorting to elaborate numerical techniques . the twisted - disk formalism in general and the analytical function @xmath103 in particular provides an avenue through which models of warped disks can advance from purely kinematical fits ( e.g. tilted - ring models ) to dynamical models based on a reasonable physical model .
in the future , we also expect to use the time - dependent form of the twisted - disk equations as a point of comparison for fully three - dimensional gas dynamic simulations of disks settling into distorted halo potentials .
we appreciate the assistance that h. cohl provided us in rendering the 3d galaxy images shown in figs
. 3 and 5 , and also would like to thank dr .
a. bosma and dr .
d. rogstad for granting permission to reprint figures from their earlier papers .
we also would like to thank the anonymous referee for suggestions that resulted in substantial improvements in this manuscript . this work has been supported in part by nasa through grants nagw-2447 and nag5 - 2777 , and in part by the u.s . national science foundation through ast-9528424 .
abramowicz , m. & stegun , i. a. 1972 , handbook of mathematical functions ( new york : dover ) bardeen , j. m. , & petterson , j. a. 1975 , , 195 , l65 bertin , g. , & mark , j. w .- k .
1980 , , 88 , 289 binney , j. 1981 , , 196 , 455 binney , j. 1992 , , 30 , 51 bosma , a. 1981 , , 86 , 1791 bosma , a. 1991 , in warped disks and inclined rings around galaxies , eds . s. casertano , p. d. sackett , and f. h. briggs , ( new york : cambridge univ . press ) , 181 briggs , f. h. 1990 , , 352 , 15 christodoulou , d. m. , tohline , j. e. , & steiman - cameron , t. y. 1993 , , 416 , 74 ( ctsc ) dekel , a. , & shlosman , i. 1983 , in internal kinematics and dynamics of galaxies , ed .
e. athanassoula , ( boston : reidel ) , 187 dubinski , j. , & kuijken , k. 1995 , apj , 442 , 492 gradshteyn , i. s. , & ryzhik , i. m. 1965 , table of integrals , series , and products ( new york : academic press ) hatchett , s. p. , begelman , m. c. & sarazin , c. l. 1981 , , 247 , 677 hofner , p. , & sparke , l. s. 1994 , , 428 , 466 hunter , c. , & toomre , a. 1969 , , 157 , 183 kahn , f. d. & woltjer , l. 1959 , , 130 , 705 kuijken , k. 1991 , , 376 , 467 kumar , s. & pringle , j. e. 1985 , , 213 , 435 lynden - bell , d. & pringle , j. e. p. 1974
, , 168 , 603 mclachlan , n. w. 1941 , bessel functions for engineers ( london : oxford univ . press ) maloney , p. r. & begelman , m. c. 1997 , , 491 , l43 newton , k. 1980a , , 191 , 169 newton , k. 1980b , , 191 , 615 papaloizou , j. c. b. , & pringle , j. e. 1983 , , 202 , 1181 petterson , j. a. 1977 , , 214 , 550 petterson , j. a. 1978 , , 226 , 253 pringle , j. e. 1992 , , 258 , 811 reakes , m. l. , & newton , k. 1978 , , 185 , 277 .
richstone , d. o. 1980 , , 238 , 103 rogstad , d. h. , crutcher , r. m. , & chu , k. 1979 , , 229 , 509 rogstad , d. h. , lockhart , i. a. , & wright , m. c. h. 1974 , , 193 , 309 sandage , a. 1961 , the hubble atlas of galaxies ( washington , d. c. : carnegie institute of washington ) sandage , a. , & bedke , j. 1994 , the carnegie atlas of galaxies , ( washington , d. c. : carnegie institute of washington ) , vol . 1 schwarz , u. j. 1985 , , 142 , 273 sparke , l. s. 1986 , , 219 , 657 sparke , l. s. , & casertano , s. 1988 , , 234 , 873 steiman - cameron , t. y. & r. h. durisen 1988 , , 325 , 26 steiman - cameron , t. y. & r. h. durisen 1990 , , 357 , 62 steiman - cameron , t. y. , kormendy , j. , & durisen , r. h. 1992 , , 104 , 1339 toomre , a. 1983 , in internal kinematics and dynamics of galaxies , ed .
e. athanassoula , ( boston : reidel ) , 177 tubbs , a. d. & sanders , r. h. 1979 , , 230 , 736 vth , h. m. 1994 , , 284 , 319 fig .
1. the twisted coordinate system .
the position @xmath1 on each ring of radius @xmath2 is referenced to a local cylindrical coordinate frame @xmath3 which has been rotated with respect to the cartesian coordinate system @xmath165 of the central reference ring by the two orientation angles @xmath4 and @xmath5 .
( adapted from petterson 1977 ) fig .
3. models of the five galaxies ngc2841 , m83 , ngc5033 , ngc5055 , and ngc300 , based on the steady state solution of the complex twist equation . in each frame , the lefthand image is a 3d isodensity surface which encompasses virtually the entire disk ; the central image is the projected 2d surface brightness map of the disk ; and the righthand image is the projected 2d velocity contour map of the disk . fig .
4. observational maps of five galaxies are reprinted here for comparison with the models shown in fig .
3 . as detailed in the text , each of these maps has been rotated in order to properly orient it with respect to the corresponding model image .
observed surface brightness and velocity contour maps appear on the left and right , respectively , of each panel ( a)-(e ) .
( a ) maps of ngc 2841 ( bosma 1981 ) ; ( b ) maps of m83 ( rlw ) ; ( c ) maps of ngc 5033 ( bosma 1981 ) ; ( d ) maps of ngc 5055 ( bosma 1981 ) ; and ( e ) maps of ngc 300 ( rcc ) . lcrccrcll @xmath166galaxy & r@xmath167 & v@xmath168@xmath169 & x@xmath167 & g & @xmath170 & @xmath14/@xmath171 & @xmath172 @xmath14 & @xmath38 + & ( kpc ) & ( km s@xmath158 ) & & & @xmath173 & ( km s@xmath158 kpc ) & ( km s@xmath174 kpc ) & @xmath175 + ngc2841 & 32 & 300@xmath176 & 13 & 1.2e-2 & 4.5 & 1100 & @xmath176 1.0 & 0.92 + m83 & 23 & 180@xmath176 & 13 & 2.2e-2 & 8.3 & 480 & @xmath176 0.53 & 1.1 + ngc5033 & 24 & 220@xmath176 & 13 & 3.2e-2 & 12.0 & 610 & @xmath176 0.58 & 0.94 + ngc5055 & 28 & 213@xmath176 & 13 & 5.7e-2 & 21.3 & 690 & @xmath176 0.78 & 1.1 + ngc300 & 15 & 94@xmath176 & 13 & 5.7e-2 & 21.3 & 160 & @xmath176 0.22 & 1.4
|
warped gas layers in the outer regions of spiral galaxies usually display a noticeably twisted structure .
this structure almost certainly arises primarily as a result of differential precession in the disk as it settles toward a preferred orientation in an underlying dark halo potential well that is not spherically symmetric . in an attempt to better understand the structure and evolution of these twisted , warped disk structures , we have adopted the `` twist equation '' formalism originally developed by petterson ( 1977 ) to study accretion onto compact objects .
utilizing more recent treatments of this formalism , we have generalized the twist equation to allow for the treatment of non keplerian disks and from it have derived a steady state structure of twisted disks that develops from free precession in a nonspherical , logarithmic halo potential .
we have used this steady state solution to produce maps of five galaxies ( m83 , ngc 300 , ngc 2841 , ngc 5033 , ngc5055 ) , which match the general features of the observed maps of these galaxies quite well .
in addition , the model provides an avenue through which the kinematical viscosity of the disk and the quadrupole distortion of the dark halo in each galaxy can be quantified .
this generalized equation can also be used to examine the time - evolutionary behavior of warped galaxy disks .
|
the odderon is the partner of the pomeron carrying negative charge parity quantum number . in high energy scattering
processes it gives the leading contribution to processes in which negative @xmath3-parity is exchanged in the @xmath4-channel .
after the concept of the odderon had been proposed in @xcite , it was for a long time almost exclusively discussed in the context of elastic or inclusive processes .
these have the disadvantage that the odderon gives only one of many contributions to the scattering amplitude and a clean identification of the odderon is rather difficult .
the only experimental evidence of the odderon so far has been found as a small difference between the differential cross sections of elastic proton - proton and antiproton - proton scattering at the cern isr @xcite . due to the low statistics of the data and the difficulty of extracting the odderon contribution , however , it is not possible to interpret this as an unambiguous signal of the odderon . for a more detailed review of the phenomenological and theoretical status of the odderon
we refer the reader to @xcite .
recently an important change of direction in the search for the odderon has taken place .
now the search concentrates on processes in which it basically gives the only contribution to the cross section .
the cross section for such processes is in general smaller than for elastic or inclusive processes , but here already the observation of the process as such would establish the existence of the odderon .
examples of such exclusive processes are the double - diffractive production of vector mesons in proton-(anti)proton scattering @xcite or the diffractive production of pseudoscalar or tensor mesons in electron - proton scattering @xcite-@xcite . in all of these processes odderon exchange
gives the main contribution to the cross section at high energies .
other possible contributions can only arise due to photon or reggeon exchange , but both of these contributions are under good theoretical control .
another interesting possibility is to study the interference between pomeron and odderon exchange .
this is possible in the diffractive production of final states that can be produced both in a @xmath5 and in a @xmath6 state like for example a pair of charged pions .
the interference term between the two corresponding production mechanisms can be isolated in suitable asymmetries like for example the charge or spin asymmetry .
asymmetries of this kind have been studied in @xcite-@xcite .
also here already the experimental observation of an asymmetry could firmly establish the existence of the odderon .
the first experimental search for one of these exclusive processes was performed for the case of diffractive pion photoproduction in @xmath7 scattering at hera in @xcite .
this process is the one for which the largest cross section is expected , but its theoretical description obviously has to rely on nonperturbative techniques . such a calculation was performed in @xcite making use of the stochastic vacuum model @xcite in the framework of the functional approach to high energy scattering developed in @xcite . in @xcite
the experimental results have been compared to the expectations based on that calculation , and no signal of odderon exchange has been found .
the failure of the theoretical prediction for this process is currently not understood . in order to avoid the large theoretical uncertainties of nonperturbative calculations in diffractive pion production one can consider the diffractive production of heavy pseudoscalar or tensor mesons . in that case
the large mass of the meson provides a hard scale , and one can hope that perturbation theory is applicable even for real photons . here in particular
the production of @xmath8 mesons has been considered , see @xcite . the expected cross section for that process was in the range of several tens of picobarns . in a study of elastic @xmath9 scattering it
has subsequently been found that the choice of parameters in the odderon - proton coupling in those calculations was very optimistic @xcite , and a realistic estimate of the cross section should be even smaller by at least an order of magnitude . due to the small cross section ,
the process is not of immediate phenomenological interest , but it has turned out to be quite interesting from a theoretical perspective . that interest is related to the occurrence of large logarithms of the energy in the perturbative series . in the simplest possible perturbative picture
the exchange of an odderon is described by the exchange of three noninteracting gluons in a symmetric color state . in higher orders in perturbation
theory large logarithms of the energy can compensate the smallness of the strong coupling constant , @xmath10 , and one needs to resum these logarithms . for the case of the odderon
this leads to the generalized leading logarithmic approximation ( glla ) which is encoded in the bartels - kwieciski - praszaowicz ( bkp ) equation .
recently two different solutions of this equation have been found explicitly in @xcite and in @xcite .
the former solution does not couple to the @xmath11 impact factor in leading order and is hence not relevant for the production of @xmath8 mesons .
the latter solution , the so - called bartels - lipatov - vacca ( blv ) solution , on the other hand does couple to that impact factor .
its intercept exactly equals one , and it hence leads to a cross section which is constant with the energy up to logarithmic corrections .
the blv solution was recently also found in the dipole picture of high energy scattering @xcite .
although the intercept of the blv solution is equal to the intercept of the simple three - gluon exchange model for the odderon it has quite different properties .
so far the phenomenological consequences of using the blv solution in the scattering amplitude have been considered only in the diffractive production of @xmath8 mesons in @xcite .
interestingly , in @xcite it was found that for real photons the resulting cross section is by about a factor of five larger than the one obtained in @xcite by using a simple three - gluon exchange for describing the odderon .
it is a very interesting question whether that enhancement is a general property of the blv solution or whether and how strongly it depends on the couplings of the odderon to the proton and to the @xmath11 impact factor .
it is one of the aims of the present paper to address this question .
all of the processes mentioned above involve the coupling of the odderon to the proton .
this coupling is known to be rather sensitive to the internal structure of the proton @xcite , and it is therefore possible that due to nonperturbative effects this coupling is small . in that case
it could be quite difficult to find the odderon in these processes .
it is therefore interesting to study also processes which do not involve the uncertainties of the odderon - proton coupling . from a theoretical point of view the quasidiffractive process
@xmath0 is the cleanest possible probe of the odderon . due to the large mass of the charm
quark the coupling of the odderon to the @xmath11 impact factor can be calculated perturbatively even for small photon virtualities .
again , already the observation of this process at high energies would firmly establish the existence of the odderon .
more generally , one can study the quasidiffractive processes @xmath12 and @xmath13 with @xmath14 being a heavy pseudoscalar or tensor meson .
also these can occur at high energies only due to odderon exchange .
such processes have first been studied in @xcite and more recently for the case of @xmath8 meson production in @xcite . in these studies
the odderon has been modeled as a simple exchange of three noninteracting gluons . in the present paper we study in detail the process @xmath0 at high energies . in particular ,
we take into account the effects of resumming large logarithms of the energy in perturbation theory by using the blv solution of the bkp equation .
that allows us to perform a detailed study of the properties of the blv odderon solution in a completely perturbative process , that is in a clean theoretical setting which does not involve model assumptions about the impact factors .
the properties of the blv odderon can easily be compared to those of an odderon modeled by the exchange of three noninteracting gluons .
furthermore , by comparing the behavior of the blv solution in this process and in the process @xmath15 we can draw some conclusions about the possible origin of the enhancement obtained in the latter process for the blv solution as compared to simple three - gluon exchange .
another important motivation for the present study is to estimate the chances of finding the odderon in the process @xmath0 in @xmath16 scattering at a future linear collider . in section [ sec : xsect ]
we provide the cross section formulae for the process @xmath0 .
in particular we discuss the blv solution and the @xmath11 impact factor . in section [ cresults ]
we study the resulting cross section and its dependence on the different parameters . after discussing some technical details of the calculation
, we start with the case of real photons and calculate the differential and total cross sections in section [ scrosssection ] . the applicability of the saddle point approximation for the blv odderon solution in this process is considered in [ ssaddleresults ] .
we investigate the energy dependence of the cross section in section [ senergy ] .
in section [ se+e- ] we address the possibility to observe this process at a future linear collider .
the case of virtual photons is studied in section [ svirtual ] .
finally , we discuss our results in the light of results obtained for the blv solution in the process @xmath15 in section [ scompare ] .
our main results are summarized in section [ sec : concl ] .
we consider the process @xmath0 at high energy and relatively small momentum transfer , that is @xmath17 in terms of mandelstam variables .
the photons in the initial state can both be real or virtual .
the large mass of the charm quark provides a justification for treating the process in perturbation theory . at high energies
the process is dominated by odderon exchange .
diagrams involving quark exchange in the @xmath4-channel are suppressed by powers of the energy and can be neglected at the energies which we will consider below . due to high energy factorization
the scattering amplitude for odderon exchange can be written in the form illustrated in figure [ amplitudefig ] .
[ amplitudefig],width=264 ] the amplitude is a convolution of the odderon green function @xmath18 with two impact factors @xmath19 coupling the odderon to the external particles .
symbolically , [ symbconv ] a ~_u |g | _ l , and the convolution includes the integration over the unconstrained transverse momenta of the gluons as will be described further below .
the subscripts @xmath20 and @xmath21 of the impact factors stand for the upper and lower impact factor , respectively .
for the odderon green function @xmath18 one can insert either the blv odderon solution or the propagation of three noninteracting gluons as the simplest possible model for the perturbative odderon . in order to make
the paper self - contained we collect in the following sections the known results for the impact factors and for the odderon green function , and we bring them into a form which can then be used to compute the above amplitude and the resulting cross section .
we first consider the impact factor that describes the transition of a photon ( real or virtual ) into an @xmath8 meson and three @xmath4-channel gluons in a color singlet state as it has been calculated in @xcite .
there it was found that the impact factor has only transverse components .
the resulting expression reads @xmath23 where the sum runs over all cyclic permutations and the index @xmath24 corresponds to the two possible transverse polarizations of the incident photon .
furthermore , @xmath25 is the total momentum transfer , and we will have @xmath26 .
@xmath27 is the virtuality of the photon , @xmath28 the mass of the charm quark , @xmath29 the totally antisymmetric tensor in two dimensions , and we have @xmath30 here @xmath31 and @xmath32 are the strong and electromagnetic coupling constants , respectively . the charm quark carries the charge @xmath33 , and the @xmath8 meson has a radiative ( photon ) width of @xmath34 and a mass of @xmath35 .
the factor @xmath36 is the totally symmetric structure constant for the color su(@xmath37 ) group .
the scattering amplitude contains the contraction @xmath38 where the last equality holds for @xmath39 .
the bartels - lipatov - vacca ( blv ) odderon solution @xmath40 found in @xcite is constructed from the known eigenfunctions @xmath41 of the bfkl equation @xcite .
these eigenfunctions are labeled by a discrete quantum number @xmath42 , called the conformal spin , and a continuous quantum number @xmath43 .
the functions @xmath41 were found in impact parameter space in @xcite , and they can be obtained in transverse momentum space via a fourier transformation .
we will use the same symbol @xmath41 for both representations of the bfkl eigenfunctions . in the blv solution
the eigenfunctions of the bkp integral operator are constructed as @xmath44 where the sum runs over cyclic permutations , and @xmath45 needs to be an odd integer .
we use the same normalization convention as in @xcite , so that the odderon states have the same norm as the pomeron eigenfunctions of which they are constructed .
this leads to @xmath46 the functions ( [ blve ] ) are eigenfunctions of the bkp integral operator with eigenvalues @xmath47 \,,\ ] ] where @xmath48 is the logarithmic derivative of the euler gamma function @xmath49 .
the odderon green function in spectral representation is constructed as a superposition of all states with odd integer numbers @xmath45 and general ( real ) @xmath50 , @xmath51 [ \nu^2+\frac{(n+1)^2}{4 } ] } \psi^{(\nu , n ) } ( \mathbf k_1 , \mathbf k_2 , \mathbf k_3 ) { \psi^{*(\nu , n ) } } ( \mathbf k_1 ' , \mathbf k_2 ' , \mathbf k_3 ' ) \,.\ ] ] here @xmath52 is the rapidity and @xmath53 is a fixed energy scale .
the scale @xmath53 is undetermined in leading logarithmic approximation , and we will discuss possible choices for @xmath53 in section [ senergy ] below .
the normalization in ( [ blvg ] ) is chosen in such a way that in the limit of vanishing coupling , @xmath54 , the green function reduces to the exchange of three noninteracting gluons . in order to calculate the momentum integral for the scattering amplitude , we need to know the bfkl eigenfunctions in momentum space . in impact parameter space
the eigenfunctions are @xmath55 where @xmath56 . in the r.h.s .
we use complex coordinates @xmath57 for describing positions in the two - dimensional impact parameter space .
further we have the conformal weights @xmath58 and @xmath59 .
the fourier transform of these eigenfunctions was calculated in @xcite .
it was found that the momentum space functions have the form @xmath60 where @xmath61 denotes an analytic contribution and @xmath62 a part containing @xmath63-functions .
the analytic part reads @xmath64 \,,\ ] ] where the coefficient @xmath3 is @xmath65 the expression @xmath66 can be given in terms of the hypergeometric function @xmath67 , @xmath68 the two - dimensional momenta are denoted as complex numbers on the r.h.s .
the @xmath63-function part in ( [ decanadel ] ) is simpler . denoting the total momentum transfer ( in complex notation ) by @xmath69
, it can be written as @xmath70 \frac{i^n}{2 \pi } 2^{1-h-\bar h } \frac{\gamma(1-\bar h)}{\gamma(h ) } q^{\bar h-1 } q^ { * h-1 } \,.\ ] ] we want to calculate the scattering amplitudes @xmath71 for different transverse polarizations @xmath72 of the incoming photons , where we have distributed the constant factors as in @xcite . in order to compute these expressions
, we have to evaluate integrals over the independent transverse momenta , e.g. @xmath73 however , this four - dimensional integral reduces to a two - dimensional one @xcite , @xmath74 where the reduced impact factor is @xmath75 let us now consider the infinite sum over odd values of @xmath45 in ( [ blvg ] ) which needs approximation in order to be evaluated numerically . in the full green function ( [ blvg ] )
the exponential factor @xmath76 clearly is of special importance to the integrand , and an expansion of its argument can help us determine the dominant values of @xmath45 . expanding ( [ chi ] ) up to second order around @xmath77 yields @xmath78 \,.\ ] ] for values of @xmath45 other than @xmath79
we therefore get a constant part in the taylor expansion of the argument of the exponential which grows with @xmath45 .
we have in fact checked numerically the contribution of @xmath80 and find that this term is already of relative size @xmath81 compared to the leading term .
therefore we can reduce the sum to one over @xmath82 .
now we can further simplify the integral that we have to calculate numerically .
the analytic part reduces for @xmath82 to @xmath83 \,.\ ] ] for @xmath84 , ( [ x ] ) leads to x^(n=1 ) ( ) & = & ( ) ^i -2 ( ) ^i -1 f(-i , 1-i ; 2;- ) + & & f(1-i , 2-i ; 2 ; - ) . for @xmath85 ,
we get @xmath86 and thus obtain @xmath87 from now on we choose the coordinate axes such that the total momentum transfer @xmath88 is in 1-direction , so that in complex notation @xmath89 is purely real . as we have an integral over both components of the two - dimensional momentum vector @xmath90
, we can replace the second component by its negative value , which in complex notation corresponds to complex conjugation . in the reduced impact factor ( [ reducedimpfac ] ) we have to switch the sign of the @xmath91 component due to the @xmath92 symbol .
the @xmath93 expression remains unchanged because it includes only the first component of the momentum in its numerator . for @xmath94 in ( [ scatampl ] )
the coherent sum over @xmath82 thus gives two equal but opposite results .
for parallel photon polarizations , @xmath95 , the resulting expression in ( [ scatampl ] ) is the same for @xmath84 and @xmath85 .
thus , we can work with the expression @xmath96 , which we will denote as @xmath97 from now on .
let us now turn to the @xmath63-function part for @xmath82 .
for the @xmath63-function part one also obtains the same result for @xmath98 and @xmath85 , which we will denote by @xmath99 , @xmath100 \frac{i}{2 \pi } 4^{- i \nu } \frac{\gamma(1-i \nu)}{\gamma(1 + i \nu)}\frac{q^{2 i \nu } } { q } \,.\ ] ] with our specific choice @xmath101 we can easily evaluate the convolution of the @xmath63-function part with the impact factor .
we can see from ( [ reducedimpfac ] ) that due to the @xmath92 tensor the @xmath91 component of the impact factor vanishes for @xmath102 or @xmath103 , that is when the @xmath63-functions in ( [ enudeltasimp ] ) are applied . hence the only non - vanishing contribution to the convolution of the @xmath63-function part with the impact factor is the one for @xmath104 , @xmath105 in @xcite diffractive production of an @xmath8 meson in @xmath106 scattering was calculated in the saddle point approximation ( spa ) . for our study this approximation will not be needed , and our results given below will not make use of the spa . nevertheless , it is interesting for us to study the same approximation also for our scattering process in order to discuss our results in the context of those of @xcite . for this purpose
the @xmath50 integral in ( [ blvg ] ) is approximated by expanding the argument of the exponential and other @xmath50-dependent factors in the integrand ( in particular the bfkl eigenfunctions ) in taylor series .
in lowest non - vanishing order the lipatov characteristic function ( [ chi ] ) is quadratic in @xmath50 , @xmath107 the rest of the integrand is also expanded in @xmath50 .
the first non - vanishing term in the expansion of the analytic part is of first order , but this term vanishes in the momentum integral because it is orthogonal to the @xmath8 impact factor , see @xcite .
the first term that survives this convolution is of second order .
the @xmath63-function part leads to a non - vanishing contribution already in zeroth order in @xmath50 , @xmath108 therefore in @xcite only the @xmath63-contribution is calculated .
we will study in section [ ssaddleresults ] below how that approximation affects our results .
as we will show , the spa turns out to be inadequate for our process .
in the following sections we will present our numerical results for the cross section for @xmath0 and for its dependence on various parameters . in the present section we make several general remarks relevant for those results . the differential cross section is obtained from the amplitudes @xmath109 involving the blv odderon solution as @xmath110 as we have seen in the previous section , the mixed polarization amplitudes vanish .
furthermore , we find numerically that the @xmath111 contribution to the cross section is only @xmath112 of that coming from @xmath104 .
since the numerical error of our calculation is also on the percent level we neglect this contribution .
we hence have @xmath113 where @xmath114 . according to the discussion above we obtain the scattering amplitude @xmath115 as the integral @xmath116 where we have the reduced impact factor @xmath117 , and the convolution of @xmath118 with @xmath119
is defined in ( [ reducedintegrals ] ) .
further , @xmath120 has two parts as in ( [ decanadel ] ) , and we have set @xmath98 here while multiplying by 2 to take into account the contribution of @xmath85 as explained in the previous section .
our results below are obtained from a numerical evaluation of the integral ( [ ampl ] ) .
we emphasize that we compute that integral without further approximations .
in particular we do not use the saddle point approximation for our main results .
the outcome of using the spa is included below only in order to discuss the applicability of that approximation .
in fact we will show that the saddle point approximation is not applicable to our process in the phenomenologically relevant kinematical region .
recall that the convolutions @xmath121 in the integrand of ( [ ampl ] ) involve only two - dimensional integrations , see ( [ reducedintegrals ] ) .
the reduction from a four - dimensional to a two - dimensional integral in these convolutions occurred due to the special structure of the blv odderon solution .
the integral ( [ ampl ] ) involves hypergeometric functions which are expensive to evaluate in terms of computer time .
but because of the reduction to two - dimensional integrations in @xmath121 it is still possible to perform the integral ( [ ampl ] ) using mathematica .
as already pointed out in section [ sscatteringamplitude ] , we do not take into account contributions to the blv solution with quantum number @xmath122 .
we have in fact calculated numerically the @xmath80 contribution for a variety of values of the parameters @xmath27 , @xmath123 and @xmath4 and find it to be negligible .
one of the most interesting questions which we want to study is how the blv odderon solution compares to the exchange of three noninteracting gluons in our process .
the latter exchange is the simplest possible perturbative model for the odderon . technically speaking it amounts to replacing the odderon green function @xmath18 in ( [ scatampl ] ) by three free gluon propagators , _ 3 g = ^(2)(_1- _ 1 ) ^(2)(_2- _ 2 ) .
the cross section for @xmath0 with a simple three - gluon exchange was calculated in @xcite .
we have reproduced the results of that paper in order to compare them with our calculations .
our calculation is based on the blv odderon solution which results from the resummation of leading logarithms .
it should be pointed out that there are several uncertainties which are inevitable in that approximation scheme .
the first of these uncertainties concerns the choice of the appropriate value of the strong coupling constant . strictly speaking
the scale of @xmath31 is undetermined in glla . in all our calculations we use @xmath124 to allow for an easy comparison with the results from @xcite and @xcite where the same value had been chosen .
it should be emphasized , however , that already a small change in @xmath31 implies a considerable change in the cross section .
this is due to the simple fact that the cross section contains a factor @xmath125 already from the coupling of the three gluons to the impact factors .
another important uncertainty is the choice of the energy scale @xmath53 in @xmath52 .
also this scale is , strictly speaking , undetermined in glla and has to be chosen as a typical energy scale for the process
. we will discuss several possible choices in detail in section [ senergy ] .
it will turn out that choices which appear equally natural can lead to quite different results .
of course there is a minimal momentum transfer required for the transition from a real photon in the initial state to a @xmath8 meson in the final state .
that minimal momentum transfer @xmath126 can be estimated to be @xmath127 . at the energies which we will consider its numerical value will be very small and will not have any quantitative effect visible in our figures . for the calculations in this section we choose a center - of - mass energy @xmath128 gev .
this choice is somewhat arbitrary and our main motivation for it is that later on we want to compare the behavior of the blv solution in the process @xmath129 to that in the process @xmath15 .
the latter was calculated for the hera energy in @xcite , and we therefore use the same energy here . as the process
@xmath129 is phenomenologically interesting mainly as a subprocess of electron - positron scattering for example , in an actual collider setup there is a continuous energy range available for the @xmath130 scattering , rather than a fixed center - of - mass energy .
we give an estimate on the size of the cross section for the process @xmath131 in section [ se+e- ] , but in the first part we are only concerned with the properties of the differential cross section of the subprocess .
the squared energy @xmath123 enters in the argument of the exponential as @xmath132 , with the scale factor @xmath53 on which we will comment later . for this section
we use a scale @xmath133 .
our numerical results for the differential cross section for real photons ( virtuality @xmath134 ) are shown in figure [ dsigdt ] , together with the corresponding results for the noninteracting three - gluon process from @xcite . for real photons ( @xmath135 ) and center - of - mass energy @xmath136 [ dsigdt],width=529 ] comparing our numerical results for the blv solution to the three noninteracting gluons in the @xmath4-channel , we find a huge enhancement of the differential cross section at small momentum transfer .
the cross section calculated with the blv solution reaches a maximal value of @xmath137 fb / gev@xmath138 at @xmath139 , and then quickly falls to @xmath140 fb / gev@xmath138 already at @xmath141 gev@xmath138 .
the simple solution exhibits a fundamentally different @xmath4-dependence that can fairly well be described by an exponential decay .
its maximal value at @xmath142 is only @xmath143 fb / gev@xmath138 , so there is a maximal enhancement factor of about @xmath144 that comes from the interaction of the three gluons . as the blv curve has a much steeper @xmath4-dependence , the two curves intersect at @xmath145 .
we have also estimated the total cross sections for the two cases .
for the calculation involving the blv solution we get a cross section of about @xmath146 , whereas the simple three - gluon process yields @xmath147 . in both cases ,
no cutoff for the integral was needed as the differential cross section falls off sufficiently quickly with growing @xmath148 to allow for a reasonable estimate of the contribution from large @xmath148 .
we notice that the minimal momentum transfer @xmath126 is sufficiently small at @xmath149 so that it does not affect the calculation of the total cross section . in summary , we see that the blv solution enhances the total cross section of @xmath129 by a factor of about @xmath150 compared to the noninteracting three - gluon calculation of @xcite .
the dependence of the differential cross section on the momentum transfer changes significantly , and the region of small momentum transfer becomes more important in the case of the blv solution .
next we want to study the reliability of the saddle point approximation for our process .
this question is primarily of theoretical interest .
however , it turns out that due to the hypergeometric functions in the blv solution it is numerically extremely challenging to calculate processes in which the blv odderon solution is coupled to a proton . for these one has to make use of the spa , and it is therefore interesting to study the reliability of that approximation . for this
our process is well suited since here we can compare the spa to the exact result .
figure [ compsaddle ] shows our exact results together with the result obtained by using the saddle point approximation as described in section [ saddlepoint ] .
again we have chosen @xmath151 .
, width=529 ] we find a maximal enhancement factor of the exact calculation over the spa of order 25 at @xmath142 , but quickly shrinking with increasing @xmath148 to an enhancement of order 2 . for the total cross section the spa calculation leads to @xmath152 , hence underestimating the total cross section approximately by a factor 5 .
so we find that the spa in its simplest form should not be applied to the scattering process at hand , at least if one is interested in small values of @xmath148 .
this result appears surprising at first sight , as it was found that in calculations involving the bfkl pomeron the saddle point approximation typically overestimates the cross section , the deviation from the actual cross section usually being of the order of @xmath153 , see for example @xcite for the case of the total hadronic cross section in virtual photon collisions .
we therefore find it instructive to discuss the origin of the large deviation and its direction in the case of the blv solution in our process . as was mentioned in section [ saddlepoint ] , in the spa the analytic part @xmath154 of the bfkl eigenfunction
is completely neglected . in figure [ compsaddle ]
we also show how this affects the @xmath4-dependence .
going to small @xmath148 , the curve resulting from the full calculation ( including the analytic part ) has a much steeper @xmath4-dependence than the spa curve , leading to the large enhancement .
we have also included in this figure the cross section obtained by neglecting the analytic part while calculating the @xmath50 integral numerically without the spa .
that curve is very similar to the spa curve in its @xmath4-dependence , but is higher by a factor of about 1.5 .
thus we see that the crucial difference is not caused by the approximation of the argument of the exponential , but by the fact that the analytic part is neglected .
the omission of the analytic part @xmath97 in the spa is due to the fact that the first non - vanishing term in the taylor expansion ( in @xmath50 ) of @xmath97 is of second order , whereas the delta function contribution @xmath155 already has a non - vanishing zeroth order term .
but as the factor @xmath156 in the exponent is not very large , the analytic piece nevertheless contributes substantially to the @xmath50 integral .
thus a numerical investigation of the @xmath50-dependence of the different parts of the solution is needed to obtain a clearer picture of the importance of the analytic part . in figure [ fnudep ]
we show the real and imaginary parts of both contributions to the expression @xmath157 at @xmath141 gev@xmath138 . ; the solid line is the gauss - like function @xmath158 representing the momentum - independent factors of the @xmath50-integrand ( the curve is scaled and shifted , see text ) [ fnudep],width=529 ] we have not included the constant factors in this calculation , so that the figure shows only the relative size of the different terms .
we recall that , before being integrated , the expression @xmath157 gets multiplied with the corresponding lower part and the rest of the @xmath50 integrand in ( [ ampl ] ) . the latter is given by @xmath159 to better understand the significance of a particular contribution to the overall result , we have included in figure [ fnudep ] also this momentum - independent contribution @xmath158 to the integral ( solid curve ) .
it basically gives a gauss - like - curve with a maximum @xmath160 at @xmath77 .
it is scaled and shifted in such a way that the lower horizontal axis in the figure is the @xmath161 level and the upper one the @xmath162 level ( see r.h.s .
of the figure ) .
it now becomes clear from the figure , why the spa can not lead to good results in our calculation .
the analytic part ( dashed lines ) gets comparable in size to the @xmath63-function part ( dotted ) already at @xmath163 , where the gauss - like factor ( solid line ) is still at about 70% of its maximal value . at larger values of @xmath50
the analytic part even contributes dominantly to the amplitude .
compared to the approximated integrand , where the analytic part is neglected but the @xmath50 integration is done numerically ( no spa ) , this gives an enhancement by a factor of 12 in the differential cross section ( which can be understood when keeping in mind that the expression @xmath164 gets squared when the result of the lower part is multiplied and again squared when the cross section is calculated ) .
the analytic part falls off much faster with @xmath148 than the @xmath63-function part , so the spa improves with increasing momentum transfer . for values of @xmath165
it reproduces the @xmath4-dependence fairly well .
nevertheless , even for larger values of @xmath148 the size of the differential cross section is significantly underestimated by the spa .
we find that the analytic part @xmath166 becomes negligible at small @xmath4 only for extremely large @xmath156 above @xmath167 .
the intercept of the blv solution equals unity , so there should be no power - like energy dependence of the cross section .
but as the continuous quantum number @xmath50 of the blv solution leads to a cut in the complex angular momentum plane ( instead of a simple pole ) , we expect a @xmath168 dependence .
this can be numerically verified by keeping @xmath4 fixed and calculating the cross section as a function of @xmath123 .
again , the results of this calculation can be compared to the spa to check the significance of the latter .
the comparison with the noninteracting three - gluon exchange process does not give any new insight , as there is no energy dependence in that cross section .
the saddle point approximation gives an inverse logarithmic dependence on the energy , as can be easily seen when keeping in mind that @xmath123 only appears in the gaussian exponential @xmath169 in the factor @xmath52 ( with @xmath170 ) .
similar results are expected for the numerical calculation if this approximation should be reasonable . however , as pointed out in the previous section , in the domain of the momentum transfer that gives the largest contribution to the total cross section ( i.e. the small @xmath148 domain ) the applicability of the saddle point approximation is very questionable .
again , we have calculated the differential cross section numerically , this time varying @xmath123 . to the results a function of the form @xmath171
is fitted with fitting parameters @xmath172 and @xmath173 that depend only on @xmath4 . to see the change of the @xmath123-dependence with varying @xmath4 , we have performed the calculation for @xmath174 and @xmath175 .
the results for a wide range of squared center - of - mass energies @xmath123 together with the fitted curves are shown in figure [ fdiffs ] .
-dependence of the differential cross section for real photons for @xmath176 ( top to bottom ) , the points are numerical results , the lines represent fits ( see text ) [ fdiffs],width=529 ] one can see that the expected behavior is reproduced quite well by the numerical results .
this should however not be understood as a possibility to determine the value of @xmath173 in ( [ fitfunc ] ) as a function of @xmath4 . from the fits we get values for @xmath173 ranging from @xmath177 at @xmath178 to @xmath179 at @xmath180 , but in the limit of asymptotically large @xmath123
, all these curves eventually have to approach the saddle point approximation , because it does become valid at some large value of @xmath156 when the gaussian factor gets so narrow that only the region of small @xmath50 gives a sizeable contribution to the integrand .
therefore , in the limit @xmath181 , we know that @xmath182 . in the case of center - of - mass energies accessible at present and
planned accelerators ( up to @xmath183 ) , however , the fitted curves give a reasonable description of the energy dependence . again , we find the spa to be inappropriate for the parameter range of small @xmath148 and realistic @xmath123 . in order to keep the figure readable , we did not include the spa curves , but again they fail completely in quantitatively reproducing the numerical results .
this can be easily seen from the fact that the exponent of the logarithm is determined to be around @xmath184 , whereas the spa gives an exponent @xmath185 .
it is only at values of @xmath123 well beyond any realistic size ( @xmath186 ) that the saddle point approximation becomes acceptable already for small @xmath148 .
we now turn to the question of the dependence of our results on the choice of the scale @xmath53 . for all of the previous calculations
we had chosen the fixed scale @xmath133 in @xmath52 .
the scale @xmath53 of the energy can not be determined in lla since a change in @xmath53 is formally sub - leading in the expansion of logarithms .
the numerical results , however , do naturally depend on the specific choice .
it is therefore interesting to see the influence of different choices of @xmath53 on the cross section .
if @xmath123 , @xmath4 and @xmath187 are fixed , a change of the scale @xmath53 by a factor @xmath188 clearly has the same effect as replacing @xmath123 by @xmath189 with fixed scale .
it is then straightforward to obtain the resulting cross sections from the results obtained above .
as long as the scale @xmath53 remains small compared to @xmath123 , the change does not qualitatively alter the results .
yet , the overall magnitude of the cross section is significantly changed by a change in the scale factor . for example , in comparison with @xmath133 the choice @xmath190 , which is the typical mass scale for hadronic processes , changes the result by an approximately constant factor of about @xmath191 .
compared to this uncertainty , the numerical and systematical errors in our calculations are definitely negligible . strictly speaking ,
the numerical values of our results should only be taken as an indication for the order of magnitude .
the appropriate choice for @xmath53 and hence the absolute results could only be determined in a next - to - leading order calculation .
the qualitative dependence on the momentum transfer and on the center - of - mass energy , however , is quite stable under changes of the scale factor .
as soon as we consider non - vanishing virtualities of one or both of the photons , the choice of the scale @xmath53 becomes even more ambiguous .
as the virtuality of the photon provides another momentum scale for the reaction , it is natural to include it into the scale factor .
the inclusion of the virtuality in @xmath53 will lead to different results for the dependence of the differential cross section on the virtuality of the photons , as will be discussed in more detail in section [ svirtual ] . in order to relate our results to phenomenology
, we want to give some estimates for the cross section in possible future collider setups .
so far , we have been concerned with the process @xmath129 with real photons at the fixed center - of - mass energy @xmath192 . that scale was chosen having in mind a later comparison to other works concerning the process @xmath15 .
a possible realization of quasidiffractive double @xmath8 production would be the photon collider option at tesla . in this section
we use the definitions and numbers given in the tesla design report @xcite . from an electron - positron center - of - mass beam energy of @xmath193 a beam of real photons with a maximum center - of - mass energy @xmath194 can be produced .
however , due to the production mechanism of inverse compton scattering , the resulting beam is not very narrowly peaked , but has a maximum at about @xmath195 with a width at half maximum of about 15 % .
the luminosity for photons with an energy in this peak region is estimated as @xmath196 . as the dependence on the momentum transfer at this energy does not give any new insight
, we do not show a figure of the differential cross section ( the curve looks exactly like figure [ dsigdt ] ) .
the total cross section for @xmath195 is @xmath197 .
this would lead to a total number of events of the order of @xmath198 in five years of continuous running .
another possibility of realizing the process @xmath199 is directly in electron - positron collisions . here
the process occurs as a subprocess in @xmath200 at high energies .
the @xmath130 subsystem in such a collision has a continuous spectrum and we have to integrate over the energy fractions . in @xcite
this calculation was performed in the equivalent photon approximation ( see also @xcite ) for the noninteracting three - gluon exchange process .
there , it was much easier to calculate the convolution integral , as the simple solution does not exhibit any energy dependence .
because of the large numerical effort of calculating total cross sections in our approach , we use the results from the previous section to estimate the energy dependence of the total cross section .
there we found fairly good fits of the form @xmath201 with values of @xmath173 around 2 .
therefore , we approximate the total cross section by @xmath202 where @xmath203 is determined from the value of @xmath204 at @xmath192 . the center - of - mass energy @xmath205 in the @xmath130 subprocess is related to the beam energy @xmath206 as @xmath207 where @xmath208 , @xmath209 are the fractions of the electron and positron energies carried by the two photons . in the equivalent photon approximation the energy distribution of the photons is given by the flux - factor @xmath210 . for untagged @xmath211
it reads ( for details see equations ( 23)-(25 ) in @xcite ) : @xmath212 where we use @xmath213 .
the cross section for the process @xmath200 , which corresponds to the collision of almost real photons , is given by : @xmath214 here @xmath215 denotes the minimal squared center - of - mass energy for the process .
again , @xmath216 is used .
if we integrate over the complete domain of possible values for @xmath123 , that is @xmath217 we get a total cross section for the electron - positron scattering process of @xmath218 , as opposed to @xmath219 that was obtained for the simple three - gluon exchange in @xcite
. however , with a squared center - of - mass energy @xmath220 one is clearly not in the high energy limit . in particular ,
the requirement @xmath221 is not met .
therefore , we have performed the calculation again for a minimal squared energy of @xmath222 and obtain for the @xmath16 total cross section a value of @xmath223 .
the large difference between these values comes about because the total photon cross section rises as one goes to small values of @xmath123 . compared to the value at @xmath224 the total cross section for @xmath225 is larger by a factor of 50 . stating it very cautiously , we estimate the total cross section of the process @xmath200 to be of the order of @xmath226 .
the planned luminosity at tesla is @xmath227 @xcite .
the resulting order of magnitude of the number of events is similar as in the case of the photon collider option .
a realistic assessment of the feasibility of a measurement of our process at a future linear collider would clearly require a more detailed study of the process including detector cuts and tagging efficiencies .
such a study is beyond the scope of the present paper .
so far we have been concerned with the case of real photon scattering . in this section
we consider the cross section for virtual photons . for definiteness
, we choose a virtuality @xmath228 , again motivated by the choice in @xcite .
the comparison of the differential cross section for real and virtual photons is presented in figure [ fdifftvarq ] . )
photons for @xmath229 .
[ fdifftvarq],width=529 ] we would like to point out that now and in the following we are again considering cross sections for @xmath130 scattering rather than for the corresponding process in @xmath16 scattering .
we have included in the figure the cross sections for the case of two real photons , one real photon and one virtual photon , and two virtual photons . in the latter two processes
there is an additional natural scale in the process that can be included in the scale factor @xmath53 .
in addition to the choices discussed in section [ senergy ] this clearly gives another range of reasonable choices .
for the case of two real photons we have used again @xmath133 , for the processes in which at least one virtual photon is involved , we have chosen @xmath230 . that choice appears natural in particular in the light of the typical energy scales occurring in deep inelastic scattering . in the case of a virtual photon scattering on a real photon , one could also think of using some kind of average momentum scale ( for example the geometric or the arithmetic mean ) .
this would change the results only by a factor which is almost independent of @xmath148 .
in figure [ fdifftvarq ] it can be seen that the basic dependence on the squared momentum transfer is qualitatively similar for all three processes , but the absolute size of the result is quite different .
the virtuality of a photon leads to a suppression of the differential cross section .
if both photons are virtual , the cross section is further suppressed .
the relative enhancement of the real photon scattering over the processes including virtual photons decreases with growing @xmath148 .
this behavior does not come as a surprise if one looks at the reduced impact factor ( [ reducedimpfac ] ) @xmath231 the dominant region for the momentum integration is where the gluon momentum @xmath90 is small . for small values of @xmath148
the suppression by the virtuality is therefore basically given by a factor @xmath232 .
for a value of @xmath148 comparable in size to @xmath27 , the suppression is only @xmath233 , and if @xmath234 the effect of the virtuality becomes negligible . for the total cross section this gives a strong suppression with respect to @xmath27 . for one virtual ( @xmath235 ) and one real photon , we get a total cross section @xmath236 , for both being virtual @xmath237 ( for both real , the cross section was @xmath238 ) .
the three - gluon approximation leads to cross sections @xmath239 for one virtual photon and @xmath240 if both photons are virtual .
thus we see that the enhancement of the total cross section of the blv odderon over the three - gluon approximation gets amplified when one considers virtual photons .
next , we want to investigate the dependence of the differential cross section on the virtuality of one photon ( figure [ fqcomp ] ) when the second photon is real . for this
we keep @xmath4 fixed ( @xmath180 ) and vary @xmath27 in one impact factor . as was already mentioned above ,
the results of the numerical calculations depend on the specific choice of @xmath53 which can now include the virtuality @xmath27 . in that case
the @xmath27-dependence of the cross section will strongly be affected by the specific choice of @xmath53 .
in figure [ fqcomp ] we plot the results for three different calculations : the approximation by three noninteracting gluons and the numerical blv odderon calculation with two choices for the scale factor .
-dependence of the differential cross section for one virtual and one real photon at @xmath241 : comparison of blv odderon with two different choices for the scale factor @xmath53 and noninteracting three - gluon exchange [ fqcomp],width=529 ] these are : the scale that was used in @xcite , @xmath230 , which we will call ` full ' scale factor because of the inclusion of the virtuality , and a ` reduced ' version without the virtuality of the photon , @xmath133 . as we have pointed out above
, there is no way to determine in lla which scale choice is the correct one . consequently the terms ` full ' and ` reduced ' are not meant in the sense that the ` full ' scale factor is ` better ' in any sense .
we see that the reduced scale factor leads to a steeper slope than the full scale factor .
again , we find that the three - gluon approximation exhibits a yet steeper slope with respect to @xmath27
. nevertheless , all curves have a somewhat similar dependence on the virtuality .
a remark is in order here concerning the applicability of the bkp odderon in the case of a virtual photon scattering on a real one .
if the virtuality of the former is large , there is an evolution in transverse momentum along the exchanged gluons in the @xmath4-channel .
the bkp equation , on the other hand , takes into account only evolution in energy but not in transverse momentum .
the situation of two largely different photon virtualities would hence require the inclusion of dglap type evolution @xcite and is not appropriately described by bkp evolution .
this limitation of our results for largely different virtualities should be kept in mind when interpreting the results of the present section . ignoring that limitation for the moment , we now want to investigate the ` asymptotic ' @xmath27-dependence . that dependence is of theoretical rather than of phenomenological interest , but can be useful for gaining insight into the blv odderon solution .
again we consider the case of one virtual and one real photon . in figures
[ fqnoy ] ( for the reduced scale factor ) and [ fqay ] ( for the full scale factor ) we plot the cross section as a function of @xmath27 for @xmath141 and @xmath242 .
the points are our numerical results , the curves are certain fits on which we will comment below .
-dependence of the differential cross section for one virtual and one real photon , @xmath243 ( upper points ) and @xmath241 ( lower points ) with reduced scale factor @xmath53 ; the points are numerical results , the curves are fits ( see text ) [ fqnoy],width=529 ] -dependence of the differential cross section for one virtual and one real photon , @xmath244 ( upper points ) and @xmath245 ( lower points ) with full scale factor @xmath53 ( @xmath27 included ) ; the points are numerical results , the curves are fits ( see text ) [ fqay],width=529 ] if the @xmath27-dependence is not included in @xmath53 , the result is simple . in the region
@xmath246 we find a ( @xmath4-independent ) scaling behavior of the cross section of @xmath247 ( the fitted curves in figure [ fqnoy ] ) .
this can be easily understood .
the virtuality of the incident photon appears explicitly only in the impact factors ( [ redimpfac ] ) , as the blv green function itself does not exhibit a @xmath27-dependence .
the momentum integrals @xmath164 receive the dominant contribution from the region of small transverse momenta .
as long as all momentum components are small compared to @xmath248 , the error we make by pulling the numerator out of the integral as a factor @xmath249 is small . in the cross section
this gives a contribution of @xmath250 .
we have added to the figures a fit with exactly that @xmath27-dependence to show the agreement with the numerical results .
if we include a @xmath27-dependence in the scale @xmath53 via the ` full ' scale factor @xmath230 , the overall dependence on @xmath27 changes qualitatively . in the domain @xmath251 we basically expect a curve that resembles the @xmath250 result from the simple choice @xmath252 multiplied by the dependence that we get from section [ senergy ] , @xmath253 . in figure [ fqay ] we show the numerical results together with a fit representing that expectation .
the fit fails when @xmath27 approaches the boundary of the domain in which it is expected to apply but gives a fairly good description of the slope within that domain .
the slope with respect to @xmath248 is about @xmath254 for @xmath255 and about @xmath256 for @xmath245 , indicating that the slope is not entirely independent of @xmath4 .
a remark is in order here concerning the question of the twist of the different solutions to the bkp equation , and in particular of the blv solution .
this issue has been addressed in @xcite , see also @xcite .
the twist of the odderon solutions is an interesting quantity from a theoretical point of view and crucial for a thorough understanding of the different solutions .
it should be kept in mind , however , that at high energies ( or small bjorken-@xmath257 ) the operator product expansion becomes problematic and eventually breaks down , see for example @xcite .
our results illustrate an additional problem of the phenomenological aspects of the odderon twist .
a potential measurement of the odderon twist is likely to be obscured by the effects of different but theoretically equivalent choices of the energy scale @xmath53 , at least as long as one works in glla . in @xcite
the process @xmath15 was studied using the blv odderon solution .
the coupling of the blv odderon to the proton is more complicated than the @xmath11 impact factor , and here the reduction of the four - dimensional integral over two transverse momenta is not reduced to a two - dimensional one .
one therefore has to make use of the saddle point approximation for a numerical calculation of the cross section .
the quality of the saddle point approximation is difficult to assess in that process .
a comparison with our process can be helpful in this respect , since there we can compare the spa with the exact results and are hence able to trace the effects of the approximation .
in our calculations we have used kinematical parameters similar to those of @xcite , and we can therefore hope that our results can give us an idea of what the effects of the exact calculation in @xmath15 can be . before we discuss possible implications of our results for the process
@xmath15 we want to point out some general differences between the results for that process found in @xcite and the results found here for the @xmath130 case . as was already found in the approximation of three noninteracting gluons @xcite the differential cross section for the process @xmath15 vanishes as @xmath258 .
this is due to the coupling of the odderon to the proton impact factor and is not expected in our scattering process .
in addition , the calculation involving the blv solution leads to a pronounced dip in the cross section at @xmath259 .
also this is due to the coupling of the odderon to the proton as was pointed out in @xcite , and such a dip is not expected in @xmath129 . in @xcite
a sizable difference between the blv cross section and the exchange of three noninteracting gluons @xcite was found in @xmath260 . for real photons ,
the cross section is enhanced by a factor of 5 due to resummation , that is for the blv solution .
we find a much smaller enhancement in our process .
in addition , when calculating our process in spa the results are actually smaller than for three - gluon exchange .
therefore , it appears that in @xmath106 scattering the enhancement is caused solely by the coupling of the odderon to the proton . in the case of virtual photons
( with @xmath235 ) the @xmath106 calculation in spa gives an enhancement of the total cross section by one order of magnitude .
this means that there is an additional enhancement factor of 2 ( as compared to the respective three - gluon calculation ) caused by the virtuality of the photon .
we find approximately the same additional enhancement factor in the case of one virtual photon in our process . in section [ ssaddleresults ] we found that for our process the saddle point approximation significantly underestimates the cross section .
as the convolution of the odderon wave function with the @xmath11 impact factor in the case of photon - proton scattering was calculated in saddle point approximation , too , it is plausible to expect that using the full numerical calculation also in that process would lead to an additional enhancement that is approximately the square root of the one found here , i.e. by about a factor 2 for the total cross section .
we should point out , however , that it is by no means clear that the saddle point approximation also underestimates the odderon - proton coupling . instead
, also a completely different behavior of the full 4-dimensional integral is conceivable .
it would be important to study this problem in more detail and to determine at least the direction of the effect of the saddle point approximation in that coupling .
we have studied the quasidiffractive process @xmath0 at high energies which is mediated by the exchange of an odderon .
this process is considered to be the theoretically cleanest probe of the odderon in perturbative qcd since it does not involve the uncertainties typically associated with the coupling of the odderon to a proton .
we have taken into account the effects of resummation of large logarithms of the energy by using the blv odderon solution .
this is the only solution of the bkp equation which couples in leading order to the @xmath11 impact factor .
we have investigated in detail the effect of resummation in this process by comparing our results to the exchange of three noninteracting gluons which is the simplest possible model for a perturbative odderon . for real photons
we find that resummation strongly enhances the differential cross section at small @xmath148 , but leads to a faster decrease with increasing @xmath148 .
the total cross section is consequently only slightly enhanced due to resummation .
the enhancement due to resummation is more pronounced when one considers virtual photons .
we find a logarithmic decrease of the cross section with the energy in agreement with the intercept of the blv solution being exactly one .
we have discussed in detail the effects of different possible choices for the energy scale @xmath53 which is undetermined in leading logarithmic approximation .
we have investigated this uncertainty and find that the cross section is rather sensitive to the choice of the energy scale , in particular in the case of virtual photons when the scale @xmath53 can naturally involve also the virtuality of the photons .
we have estimated the expected event rates for the process @xmath129 at a future linear collider for @xmath261 scattering as well as for a photon collider option . in both cases
the observation of the odderon in this process appears feasible , but more detailed studies accounting for detector cuts and tagging efficiencies will be required to obtain a conclusive assessment of this process .
all previous phenomenological studies of the blv odderon solution were done for processes involving protons . due to the complicated odderon - proton coupling
the numerical effort for an exact calculation of the cross section is prohibitively large in these cases .
these processes have therefore been studied in the saddle point approximation . in our process
a considerable simplification occurs due to the special structure of the @xmath11 impact factor and an exact numerical calculation becomes possible .
we have used our exact results to test the quality of the saddle point approximation in our process .
we find that for realistic values of the kinematical parameters the saddle point approximation fails and underestimates the actual cross section by about an order of magnitude .
we have identified the origin of this large deviation and have indicated possible implications of our result for processes in which the blv odderon solution is coupled to a proton .
finally , we would like to point out that also other final states can be produced via odderon exchange in quasidiffractive photon - photon scattering .
an interesting example among them is the single - inclusive process @xmath262 which can be treated in a similar way as the process discussed in the present paper . in that process
, however , the situation is similar to the processes in which the blv odderon is coupled to a proton .
the coupling of the blv odderon to the @xmath263 impact factor does not allow one to reduce the four - dimensional integral over transverse momenta to a two - dimensional one .
an exact numerical evaluation of the integral is therefore not feasible and again one has to make use of the saddle point approximation . in the light of our results
, however , it seems likely that also here that approximation gives only a relatively poor estimate of the actual cross section . despite this difficulty it would be very interesting to study the effects of resummation also in @xmath262 and in related processes involving tensor mesons in more detail since they might offer a good chance to observe the odderon .
we would like to thank o. nachtmann and g.p .
vacca for helpful discussions . l. lukaszuk and b. nicolescu , lett .
nuovo cim .
* 8 * ( 1973 ) 405 .
a. breakstone _ et al .
_ , phys . rev . lett .
* 54 * ( 1985 ) 2180 . c. ewerz , arxiv : hep - ph/0306137 .
a. schfer , l. mankiewicz and o. nachtmann , phys .
b * 272 * ( 1991 ) 419 .
v. v. barakhovsky , i. r. zhitnitsky and a. n. shelkovenko , phys .
b * 267 * ( 1991 ) 532 .
a. schfer , l. mankiewicz and o. nachtmann , in proc . of the workshop `` physics at hera '' , hamburg 1991 , vol . 1 ,
p. 243 .
j. czyzewski , j. kwieciski , l. motyka and m. sadzikowski , phys .
b * 398 * ( 1997 ) 400 [ erratum - ibid .
b * 411 * ( 1997 ) 402 ] [ arxiv : hep - ph/9611225 ] .
r. engel , d. y. ivanov , r. kirschner and l. szymanowski , eur .
j. c * 4 * ( 1998 ) 93 [ arxiv : hep - ph/9707362 ] .
m. g. ryskin , eur .
j. c * 2 * ( 1998 ) 339 .
w. kilian and o. nachtmann , eur .
j. c * 5 * ( 1998 ) 317 [ arxiv : hep - ph/9712371 ] .
e. r. berger , a. donnachie , h. g. dosch , w. kilian , o. nachtmann and m. rueter , eur .
j. c * 9 * ( 1999 ) 491 [ arxiv : hep - ph/9901376 ] .
e. r. berger , a. donnachie , h. g. dosch and o. nachtmann , eur .
j. c * 14 * ( 2000 ) 673 [ arxiv : hep - ph/0001270 ] .
j. bartels , m. a. braun , d. colferai and g. p. vacca , eur .
j. c * 20 * ( 2001 ) 323 [ arxiv : hep - ph/0102221 ] .
j. bartels , m. a. braun and g. p. vacca , arxiv : hep - ph/0304160 .
j. p. ma , nucl .
a * 727 * ( 2003 ) 333 [ arxiv : hep - ph/0301155 ] .
s. j. brodsky , j. rathsman and c. merino , phys .
b * 461 * ( 1999 ) 114 [ arxiv : hep - ph/9904280 ] .
i. p. ivanov , n. n. nikolaev and i. f. ginzburg , in proc .
9th international workshop on deep inelastic scattering ( dis 2001 ) , bologna , italy , 2001 , arxiv : hep - ph/0110181 .
p. hgler , b. pire , l. szymanowski and o. v. teryaev , phys .
b * 535 * ( 2002 ) 117 [ erratum - ibid .
b * 540 * ( 2002 ) 324 ] [ arxiv : hep - ph/0202231 ] .
p. hgler , b. pire , l. szymanowski and o. v. teryaev , eur .
j. c * 26 * ( 2002 ) 261 [ arxiv : hep - ph/0207224 ] .
i. f. ginzburg , i. p. ivanov and n. n. nikolaev , eur .
j. directc * 5 * ( 2003 ) 02 [ arxiv : hep - ph/0207345 ] . c. adloff
[ h1 collaboration ] , phys .
b * 544 * ( 2002 ) 35 [ arxiv : hep - ex/0206073 ] .
h. g. dosch , phys . lett .
b * 190 * ( 1987 ) 177 .
h. g. dosch and y. a. simonov , phys .
b * 205 * ( 1988 ) 339 .
y. a. simonov , nucl .
b * 307 * ( 1988 ) 512 .
o. nachtmann , annals phys .
* 209 * ( 1991 ) 436 .
h. g. dosch , c. ewerz and v. schatz , eur .
j. c * 24 * ( 2002 ) 561 [ arxiv : hep - ph/0201294 ] .
j. bartels , nucl .
b * 175 * ( 1980 ) 365 .
j. kwieciski and m. praszaowicz , phys .
b * 94 * ( 1980 ) 413 .
r. a. janik and j. wosiek , phys .
* 82 * ( 1999 ) 1092 [ arxiv : hep - th/9802100 ] .
j. bartels , l. n. lipatov and g. p. vacca , phys .
b * 477 * ( 2000 ) 178 [ arxiv : hep - ph/9912423 ] .
y. v. kovchegov , l. szymanowski and s. wallon , arxiv : hep - ph/0309281 . i. f. ginzburg , d. y. ivanov and v. g. serbo , phys . atom .
* 56 * ( 1993 ) 1474 [ yad .
* 56n11 * ( 1993 ) 45 ]
. i. f. ginzburg and d. y. ivanov , nucl .
* 25b * ( 1992 ) 224 .
l. motyka and j. kwieciski , phys .
d * 58 * ( 1998 ) 117501 [ arxiv : hep - ph/9802278 ] .
e. a. kuraev , l. n. lipatov and v. s. fadin , sov .
jetp * 45 * ( 1977 ) 199 [ zh .
* 72 * ( 1977 ) 377 ] .
i. i. balitsky and l. n. lipatov , sov .
j. nucl .
* 28 * ( 1978 )
822 [ yad . fiz .
* 28 * ( 1978 ) 1597 ] .
l. n. lipatov , sov .
jetp * 63 * ( 1986 ) 904 [ zh .
. teor . fiz .
* 90 * ( 1986 ) 1536 ] .
j. bartels , a. de roeck and h. lotter , phys .
b * 389 * ( 1996 ) 742 [ arxiv : hep - ph/9608401 ] .
b. badelek _ et al . _
[ ecfa / desy photon collider working group collaboration ] , `` tesla technical design report , part vi , chapter 1 : photon collider at tesla , '' arxiv : hep - ex/0108012 .
p. aurenche _ et al .
_ , `` gamma - gamma physics at lep2 , '' arxiv : hep - ph/9601317 . v.
n. gribov and l. n. lipatov , sov
. j. nucl .
phys . * 15 * ( 1972 ) 438 [ yad
. fiz . *
15 * ( 1972 ) 781 ] .
g. altarelli and g. parisi , nucl .
b * 126 * ( 1977 ) 298 .
y. l. dokshitzer , sov .
jetp * 46 * ( 1977 ) 641 [ zh .
* 73 * ( 1977 ) 1216 ] .
h. j. de vega and l. n. lipatov , phys .
d * 66 * ( 2002 ) 074013 [ arxiv : hep - ph/0204245 ] .
g. p. korchemsky , j. kotanski and a. n. manashov , phys .
b * 583 * ( 2004 ) 121 [ arxiv : hep - ph/0306250 ] .
a. g. shuvaev , arxiv : hep - ph/0310344 .
a. h. mueller , phys .
b * 396 * ( 1997 ) 251 [ arxiv : hep - ph/9612251 ] .
|
we study the perturbative odderon in the quasidiffractive process @xmath0 . at high energies this process is dominated by odderon exchange and can be viewed as the theoretically cleanest test of the perturbative odderon .
we calculate the differential and total cross section , as well as the dependence on the energy and on the photon virtualities taking into account the effects of resummation of logarithms of the energy .
the results are compared with those obtained with a simple exchange of three noninteracting gluons .
we present the expected cross section for this process at a future linear collider and discuss implications for other processes involving the perturbative odderon .
hd - thep-04 - 09 + hep - ph/0403197 + + * stefan braunewell@xmath1 , carlo ewerz@xmath2 * _ institut fr theoretische physik , universitt heidelberg + philosophenweg 16 , d-69120 heidelberg , germany _ ' '' '' width 5.cm
|
the simultaneous flow of immiscible fluids through porous media has been studied for a long time @xcite .
it is a problem that lies at the heart of many important geophysical and industrial processes .
often , the length scales in the problem span numerous decades ; from the pores measured in micrometers to researvoir scales measured in kilometers . at the largest scales ,
the porous medium is treated as a continuum governed by effective equations that encode the physics at the pore scale .
the problem of tying the pore scale physics together with the effective description at large scale is the upscaling problem . in 1936 ,
wycoff and botset proposed a generalization of the darcy equation to immiscible two - phase flow @xcite .
it is instructive to reread wycoff and botset s article .
this is where the concept of relative permeability is introduced .
the paper is eighty years old and yet it is still remarkably modern .
capillary pressure was first considered by richards as early as 1931 @xcite . in 1940 , leverett combined capillary pressure with the concept of relative permeability , and the framework describing essentially all later analysis of immiscible multiphase flow in porous media was in place @xcite .
the introduction of the concepts of relative permeability and capillary pressure as solution to the upscaling problem , dominates still today but other theories exist .
these theories are , as is relative permeability , based on a number of detailed assumptions concerning the porous medium and concerning the physics involved .
it is the aim of this paper to present a new theory for flow in porous media that is solely based on a thermodynamic balance between work done on the flowing fluids and the disspation in them . in the same way as buckley and leverett s analysis based on the conservation of the mass of the fluids in the porous medium led to their buckley - leverett equation @xcite , the balancing of work per time against dissipation leads to the new equations we present in the following .
the theory we present rests on non - equilibrium thermodynamics @xcite which combines conservation laws with the laws of thermodynamics .
the structure of the theory is reminiscent of the structure of thermodynamics itself : we have a number of variables that are related through general thermodynamic principles leaving an equation of state to account for the detailed physics of the problem .
we only consider here _ isotropic _ porous media where the local flow always points in the opposite direction of the pressure gradient .
it will be defered to later to consider more general systems .
we also consider for now the pressure gradient as the only driving force in the system . in section [ system ]
we describe the porous medium system we consider .
we review the key concepts that will be used in the subsequent discussion .
in particular , we discuss the relation between average seepage velocity and the seepage velocities of each of the two fluids .
we then ask the central question : given the average seepage velocity and the saturation , is this enough to determine the seepage velocities of each of the two fluids ?
the answer seems to be no , " and we demonstrate why .
however , in section [ diss ] we use non - equilibrium thermodynamics to balance dissipation in the two - fluid system against the power supplied to it by the pressure difference across it .
we demonstrate that the splitting of the dissipation into that of each fluid is unique . in section [ scaling ]
we use a scaling argument resting on the euler theorem for homogeneous functions to set up the framework allowing us to derive a set of equations relating the average seepage velocity , the saturation and the seepage velocities of each of the two fluids .
section [ saturation ] contains the derivation of these equations . in section [ non - eq ]
we summarize briefly the theory and point out the role played by the constitutive equation .
we derive in section [ fractional ] a fractional flow equation , which essentially is a rewriting of one of the two central equations in section [ saturation ] .
section [ solving ] contains solutions to the two equations of section [ saturation ] .
we obtain integral expressions for the two fluid seepage velocities which are closed by combining them with the constitutive equation . in section [ 2examples ] , we demonstrate the use of the formulas derived in the previous section to relate a given average seepage velocity as function of saturation to the seepage velocities of each of the two fluids . in section
[ numerical ] we analyze data obtained through a numerical pore scale network model @xcite . section [ relperm ] uses the relative permeability equations as constitutive equations demonstrating that the equations derived in section [ saturation ] lead to equations between the relative permeabilities and between the capillary pressure and the relative permeabilities .
this leads to problems for the relative permeability formulation as the equations predict that at least one of the relative permeabilities must depend on the ratio between the viscosities of the two fluids .
we conclude that relative permeability theory is not thermodynamically consistent . in section [ conclusion ]
we write down the full set of equations to describe immiscible two - phase flow in porous media , equations ( [ conc-1 ] ) to ( [ conc-7 ] ) based on our theory .
it reduces the problem from a two - phase fluid problem to a one - phase fluid problem with the saturation as a conserved , extra variable .
the aim of this paper is to derive a set of equations on the continuum level where differentials make sense from non - equilibrium thermodynamics .
we define a _ representative elementary volume _ rev as an isotropic block of porous material filled with two immiscible fluids with no internal structure : it is fully described by a small set of parameters which we will now proceed to define .
we will then imagine the volume of the rev shrinking to zero so that the equations we derive become pointwise equations .
these equations may then be supplemented by conservation laws to produce a full set of flow equations in the same way as is done with the relative permeability equations .
we show in fig .
[ fig1 ] the rev .
it is a block of homogeneous porous material of length @xmath0 and area @xmath1 .
we seal off the surfaces that are parallel to the @xmath0 direction .
the two remaining surfaces , each with area @xmath1 , are kept open and act as inlet and outlet for the fluids that are injected and extracted from the rev .
the porosity is @xmath2 where @xmath3 is the pore volume .
is the _ effective _ pore volume that exclude any irreducible wetting fluid or residual non - wetting fluid . ] due to the homogeneity of the porous medium , any cut orthogonal to the axis along the @xmath0 direction ( named the @xmath4 axis for later ) will reveal a _ pore area _ that fluctuates around the value @xmath5 .
the homogeneity assumption consists in the fluctuations being so small that they can be ignored .
there is a pressure drop @xmath6 across the rev as shown in fig .
this leads to a time averaged volumetric flow rate @xmath7 .
we assume that the rev is isotropic so that @xmath7 and other volumetric flow rates are in the direction of the negative pressure gradient .
the volumetric flow rate consists of two components , @xmath8 and @xmath9 , which are the volumetric flow rates of the more wetting ( @xmath10 for wetting " ) and the less wetting ( @xmath11 for non - wetting " ) fluids with respect to the porous medium .
we have @xmath12 in the porous medium , there is a volume @xmath13 of _ incompressible _ wetting fluid ( excluding the irreducible contents of wetting fluid ) and a volume @xmath14 of _ incompressible _ non - wetting fluid so that @xmath15 .
we define the wetting and non - wetting saturations @xmath16 and @xmath17 .
we have that @xmath18 we define the wetting and non - wetting pore areas @xmath19 and @xmath20 as the parts of the pore area @xmath21 which is filled with the wetting or the non - wetting liquids respectively . as
the porous medium is homogeneous , we will find the same averages @xmath19 and @xmath20 on any cut through the cylindrical porous medium orthogonal to the axis .
this is illustrated in fig .
we have that @xmath22 so that @xmath23 likewise , @xmath24 hence , we have @xmath25 we define the seepage velocities for the two immiscible fluids , @xmath13 and @xmath14 as @xmath26 and @xmath27 hence , equation ( [ eqn1 ] ) may be written @xmath28 we finally define an average seepage velocity associated with the total flow rate @xmath7 as @xmath29 by using equations ( [ eqn3 ] ) to ( [ eqn5 ] ) and ( [ eqn5.1 ] ) , ( [ eqn5.2 ] ) and ( [ eqn5.3 ] ) we transform ( [ eqn14.9 ] ) into @xmath30 it is important to note that we assume the rev to be small enough so that the saturations and seepage velocities may be taken as constant throughout the sample . there is no internal structure in the rev .
we pose here the central question that will be answered positively in the next two sections : _ from knowledge of @xmath31 and @xmath32 in equation ( [ eqn5.4 ] ) , is it possible to determine @xmath33 and @xmath34 , defined in equation ( [ eqn5.1 ] ) and ( [ eqn5.2 ] ) ?
_ seemingly , the answer is no . "
given a constant reference velocity @xmath35 , we may define the new velocities @xmath36 and @xmath37 and we have @xmath38 hence , there is seemingly no way to determine which pair of seepage velocities corresponds to equation ( [ eqn5.1 ] ) and ( [ eqn5.2 ] ) without knowing @xmath8 or @xmath9 .
this turns out not to be correct . by balancing the dissipation in the rev against the power delivered to the rev by the action of the pressure difference @xmath6
, we arrive at unique split of the average seepage velocity @xmath31 into @xmath33 and @xmath34 based on the basic laws of thermodynamics .
this is the subject of the next two sections .
the following discussion is based on non - equilibrium thermodynamics @xcite .
let @xmath39 be the entropy density , @xmath40 the internal energy density and @xmath41 the temperature of the fluids in the rev .
the fluids have chemical potentials @xmath42 and @xmath43 .
their molar concentrations are @xmath44 and @xmath45 where @xmath46 and @xmath47 are their respective molar numbers in the rev .
the gibbs relation between these variables is @xmath48 this leads to a rate of change of entropy density with time given by @xmath49 where @xmath50 is the time variable .
we now introduce conservation equations for entropy , energy and mass of each fluid , @xmath51 @xmath52 and @xmath53 where the @xmath4 axis is oriented along the positive flow direction of the rev , see fig . [ fig1 ] . here
@xmath54 is the current density of energy .
the other two are the current densities of the wetting fluid and non - wetting fluid molar numbers given by @xmath55 and @xmath56 where @xmath57 and @xmath58 are the velocities associated with these currents . by combining the conservation laws ( [ eqn14 - 3 ] ) to ( [ eqn14 - 5 ] ) with ( [ eqn14 - 2 ] ) we find @xmath59+j_u\ \frac{\partial}{\partial x}\ \frac{1}{t}\ -j_w\ \frac{\partial}{\partial x}\ \frac{\mu_w}{t } -j_n\ \frac{\partial}{\partial x}\ \frac{\mu_n}{t}\nonumber\\\ ] ] if we now compare this expression to the entropy balance equation @xmath60 where @xmath61 is the entropy current density and @xmath62 is the entropy production density , we find @xmath63\;,\ ] ] and @xmath64 the system is in a steady state so that @xmath65 .
we also assume that the temperature @xmath41 is constant throughout the rev .
the steady state assumption also leads to @xmath66 through equations ( [ eqn14 - 3 ] ) to ( [ eqn14 - 5 ] ) .
we integrate the expression for the entropy production ( [ eqn14 - 9 ] ) over the pore volume of the rev , giving the dissipation @xmath67 in the system , @xmath68= a_p\left[c_w\delta\mu_w \overline{v}_w+c_n\delta\mu_n \overline{v}_n\right]\;,\ ] ] where @xmath69 and @xmath70 are the chemical potential differences between the entrance and the exit of the rev for the fluids
. the dissipation @xmath67 must equal the work done on the system by the pressure difference @xmath6 per time unit , @xmath71 .
the power supplied to the wetting and the non - wetting fluids by the pressure gradient is @xmath72 and @xmath73 respectively , where @xmath74 and @xmath75 .
the power supplied to each fluid must balance the dissipation in each fluid if the system is to be in a steady state .
hence , @xmath76 and @xmath77 combining these two equations with ( [ eqn14 - 10 ] ) gives @xmath78\delta p\;,\ ] ] where we have used that @xmath79 and @xmath80 .
hence , we have @xmath81\;,\ ] ] in analogy to equation ( [ eqn5.4 ] ) .
what has been accomplished here is ( 1 ) a unique split of the average seepage velocity @xmath31 into two velocities @xmath57 and @xmath58 associated with the dissipation in each fluid and ( 2 ) we have shown that @xmath7 is a thermodynamic variable .
as we shall see in section [ numerical ] , in small systems @xmath57 and @xmath58 defined in equations ( [ eqn14 - 11 ] ) and ( [ eqn14 - 12 ] ) , are not identical to the seepage velocities defined in equations ( [ eqn5.1 ] ) and ( [ eqn5.2 ] ) since they are different averages . however , in the thermodynamic limit which is the continuum limit , they will be equal due to the system being self averaging , and we have @xmath82 and @xmath83
the volumetric flow rate @xmath7 , which we now recognize as a thermodynamic function , is a _ homogeneous function of order one _ of the two areas @xmath19 and @xmath20 defined in equations ( [ eqn3 ] ) and ( [ eqn4 ] ) .
hence , if @xmath84 is a scale factor , we may scale the two areas @xmath85 and @xmath86 .
we illustrate this in fig .
this leads to the scaling relation @xmath87 this scaling property is essentially self evident .
it may be implemented in pratice by using different porous media samples with different areas @xmath1 .
is , [ fig2 ] we take the derivative with respect to @xmath84 on both sides of ( [ eqn10 ] ) and set @xmath88 .
this gives @xmath89 where it is understood that we keep the pressure drop @xmath6 constant during the partial differentiations .
equation ( [ eqn11 ] ) is essentially the euler theorem for homogeneous functions of order one . by dividing this equation by @xmath21
, we have @xmath90 by comparing this equation to ( [ eqn15 - 1 ] ) using ( [ eqn15 - 2 ] ) and ( [ eqn15 - 3 ] ) , we find @xmath91 and @xmath92
let us return to the scaling equation ( [ eqn10 ] ) which we combine with the expression for the total flow in terms of the seepage velocities @xmath33 and @xmath34 , equation ( [ eqn14.9 ] ) .
in order for @xmath7 in this equation to obey the scaling relation ( [ eqn10 ] ) , neither @xmath33 nor @xmath34 can be extensive in the variables @xmath19 and @xmath20 , i.e. , @xmath93 they are homogeneous functions of @xmath19 and @xmath20 of order zero .
hence , they must depend on the areas @xmath19 and @xmath20 through their ratio @xmath94 , where we have used equations ( [ eqn3 ] ) ( [ eqn5 ] ) .
hence , @xmath95 the scaling relation ( [ eqn10 ] ) implies changing the wetting and non - wetting pore areas @xmath19 and @xmath20 by changing @xmath21 , but without changing the saturations @xmath96 and @xmath97 .
we now change the saturation while keeping the total pore area @xmath98 constant .
mathematically , this is accomplished by changing our variables from @xmath99 to @xmath100 where @xmath101 and @xmath102 .
we calculate @xmath103\;,\ ] ] where we have used ( [ eqn10 - 2 ] ) and ( [ eqn10 - 3 ] ) .
we divide by the area @xmath21 which is kept constant and find @xmath104 this equation is a direct consequence of scaling assumption ( [ eqn10 ] ) .
this equation is one of the two fundamental equations that constistute the main result presented in this paper .
the second equation we find by taking the derivative of equation ( [ eqn5.4 ] ) with respect to @xmath32 , @xmath105 = v_w - v_n+s_w \frac{dv_w}{ds_w}+(1-s_w)\frac{dv_n}{ds_w}\;.\ ] ] combining this equation with equation ( [ eqn18 ] ) , we have the equation @xmath106 which is our second main result .
also this equation is a consequence of the scaling assumption ( [ eqn10 ] ) alone and does not entail any further assumpions about the properties of the flow problem . equation ( [ eqn20 ] ) is in fact the _ gibbs - duhem equation , _ in disguise , see e.g. reference @xcite .
let us lastly in this section remark that the darcy or superficial velocities @xmath107 , @xmath13 and @xmath14 are related to the seepage velocities by @xmath108 , @xmath109 and @xmath110 .
hence , equations ( [ eqn18 ] ) and ( [ eqn20 ] ) will not change if expressed in terms of these velocities instead of the seepage velocities .
we note that the three equations ( [ eqn5.4 ] ) , ( [ eqn18 ] ) and ( [ eqn20 ] ) are related in such a way that given any two of them , the third follows .
we have accomplished to construct a non - equilibrium thermodynamic theory for immiscible two - phase flow in porous media @xcite .
there are three velocities , @xmath31 , @xmath33 and @xmath34 .
these are the responses to the pressure difference @xmath6 .
lastly , there is the saturation .
the three velocities are bound together by the two equations ( [ eqn18 ] ) and ( [ eqn20 ] ) .
hence , three variables remain : say @xmath31 , @xmath32 and @xmath111 .
they will be related through a _ constitutive equation _ @xmath112
it is through this equation that the detailed physics enters the description .
we note that this equation plays the same role as an _ equation of state _ in thermodynamics .
the equations we have derived refer only to velocities , saturation and pressure gradient .
hence , the non - zero volume of the rev has dropped out and the two equations may be seen as applying pointwise in a continuous porous medium . if the three equations ( [ eqn18 ] ) , ( [ eqn20 ] ) and ( [ eqn20.1 ] ) are supplied with a conservation law for the saturation , a complete descpription of porous media in the continuum limit ensues .
we elaborate on this in the conclusion and discussion section [ conclusion ] .
we define the wetting and non - wetting fractional flows as @xmath113 and @xmath114 where we have used equations ( [ eqn5.1 ] ) and ( [ eqn5.2 ] ) .
we now combine these expressions for @xmath115 and @xmath116 with equation ( [ eqn18 ] ) to find @xmath117\;.\ ] ] by solving for @xmath115 , we find @xmath118 hence , we have derived a new _ fractional flow equation _ relating the fractional flow to the total flow rate and the saturation .
again , no further assumptions have been made concerning the flow apart from the scaling relation given in ( [ eqn10 ] ) .
all that is needed to determine the fractional flow is contained in the average seepage velocity @xmath119 , i.e. , in the constitutive law ( [ eqn20.1 ] ) .
equations ( [ eqn18 ] ) and ( [ eqn20 ] ) may be integrated to find @xmath33 and @xmath34 as functions of @xmath32 .
we start by transforming equation ( [ eqn18 ] ) into @xmath120 equation ( [ eqn20 ] ) may then be subtracted from this equation to give @xmath121 by using equation ( [ eqn20 ] ) once again , we find @xmath122 we integrate equations ( [ eqn36 ] ) and ( [ eqn37 ] ) finding @xmath123 and @xmath124 as with the fractional flow equation ( [ eqn27 ] ) , only the average seepage velocity @xmath119 is needed to determine seepage velocities for each of the fluids . this is no surprise as @xmath115 and @xmath33 convey the same information about the system .
we now give a couple of analytically tractable examples of the use of equations ( [ eqn40 ] ) and ( [ eqn41 ] ) . first a trivial example .
suppose that surface tension between the two fluids is negligible .
the total average seepage velocity ( [ eqn5.4 ] ) is @xmath125 for @xmath126 and it is @xmath127 for @xmath128 . if the wetting fluid has a viscosity @xmath42 and the non - wetting fluid a viscosity @xmath43 , we have @xmath129 .
@xmath130 must be linear in @xmath32 : @xmath131 and we must have @xmath132 and @xmath133 so that @xmath134\;.\ ] ] equations ( [ eqn40 ] ) and ( [ eqn41 ] ) give since @xmath135 @xmath136 and @xmath137 as it should be . we may also test the fractional flow equation ( [ eqn27 ] ) for this simple case . using @xmath130 from ( [ eqn42 ] )
, we find @xmath138 which is equal to @xmath139 as it should be .
we illustrate @xmath115 in fig .
[ fig3 ] . we know from equation ( [ eqn18 ] ) that when @xmath140 , @xmath141 . from equation ( [ eqn5.4 ] ) this implies furthermore that @xmath142 . as a consquence of ( [ eqn27 ] )
, we also have that @xmath143 when @xmath140 .
let us name this particular saturation @xmath144 .
when we move away from @xmath144 , we have that @xmath145 where @xmath146 .
hence , to first order in @xmath147 the two fluids behave as if they were miscible and both having the same effective viscosity
. however , the constitutive equation ( [ eqn20.1 ] ) describing the mixture may still be more complicated than the darcy equation even if the two fluids each are newtonian @xcite .
in the example just discussed , there is no minimum in the average seepage velocity @xmath31 , as it is a constant with respect to @xmath148 .
hence , as long as @xmath149 , @xmath33 is different from @xmath34 and as seen in fig .
[ fig3 ] , @xmath150 for all values of @xmath32 except the trivial values 0 and 1 .
we now move on to our next example .
let us suppose that the average seepage velocity @xmath31 may be parametrized as a fourth order polynomial in @xmath32 , @xmath151 we find from the integrals ( [ eqn40 ] ) and ( [ eqn41 ] ) @xmath152 and @xmath153 a concrete example where @xmath154 is @xmath155\;,\ ] ] see fig .
integrals ( [ eqn40 ] ) and ( [ eqn41 ] ) give @xmath156\;,\ ] ] and @xmath157\;.\ ] ] fig .
[ fig4 ] shows @xmath31 , @xmath33 and @xmath34 as function of @xmath32 for @xmath158 .
we note that the three seepage velocities meet as a point .
such a point must exist when @xmath33 is an increasing function and @xmath34 is a decreasing function of @xmath32 as they must cross somewhere , and at the point they meet , @xmath31 must be equal to the two others due to equation ( [ eqn5.4 ] ) .
however , equation ( [ eqn18 ] ) dictates that this point is also the point at which @xmath31 is minimal . in this case
this happens for @xmath159 .
we see that this is the case in fig .
we show in fig .
[ fig5 ] the fractional flow @xmath115 calculated from equation ( [ eqn27 ] ) as a function of @xmath32 for the example given in ( [ eqn45 ] ) . for the value @xmath159 , @xmath143 .
as described above , the two immiscible fluids with different viscosities act as if they were miscible and have the same viscosity .
typically , part of the wetting fluid will be stuck and does not contribute to the flow properties apart from lowering the available pore volume .
the volume of this bound wetting fluid divided by the pore volume is the irreducible wetting saturation @xmath160 and the residual non - wetting saturation @xmath161 .
it is then convenient to define an effective saturation @xmath162 we define the corresponding effective non - wetting saturation as @xmath163 . the network simulator first proposed by aker et al .
@xcite has been refined over the years and is today a versatile model for immiscible two - phase flow under steady - state conditions .
the model tracks the interfaces between the immiscible fluids by solving the kirchhoff equations with a capillary pressure given by @xmath164 where @xmath165 is the position of the @xmath166th interface in a link measured in units of the length @xmath167 of the link , @xmath168 its average radius and @xmath169 is the surface tension between the two fluids . in the following , we will use this model to test the fractional flow equation ( [ eqn27 ] ) .
our parameters are chosen as follows : we have set @xmath170 mm and @xmath171 n / m .
the link radii @xmath168 were drawn from an approximately lognormal distribution with an average of @xmath172 and a standard deviation of @xmath173 .
the viscosities of the fluids were equal , @xmath174 pa s. steady state is obtained by implementing the network on a torus .
hence , the system is closed and the saturation @xmath32 does not fluctuate .
we have used a two - dimensional hexagonal lattice consisting of @xmath175 nodes .
the system is driven by a constant pressure difference of 15 kpa .
the capillary number was hovering around 0.02 .
we show in fig .
[ fig6 ] the volumetric flow rate @xmath7 as a function of @xmath32 and in fig .
[ fig7 ] the corresponding fractional flow rate @xmath115 as a function of @xmath32 .
our aim is to compare the fractional flow equation ( [ eqn27 ] ) to the calculated fractional flow . in the derivation of the fractional flow equation ( [ eqn27 ] )
, the velocities defined in equations ( [ eqn15 - 2 ] ) and ( [ eqn15 - 3 ] ) , @xmath57 and @xmath58 are used . in a network model , one would measure e.g. @xmath57 as follows : let @xmath176 be the volumetric flow rate of the wetting fluid in the @xmath166th link and @xmath177 is the area of this link .
then , the wetting fluid velocity in this link is @xmath178 .
we then average these velocities over the links .
hence , we have @xmath179 where the sum runs over the all the links in the network . there are @xmath180 links in each layer ( row of links in the direction orthogonal to the flow direction ) and @xmath181 layers .
the seepage velocity , @xmath33 , defined in ( [ eqn5.1 ] ) on the other hand , is calculated in the network as @xmath182 where the sum index runs over layers in the network in the flow direction . in the limit of an infinitely large network
the two averages ( [ eqn25 - 1 ] ) and ( [ eqn25 - 2 ] ) will be equal as stated in [ diss ] .
in a small system , such as the one we consider here , they are not , @xmath183
. we measure in fig .
[ fig7 ] @xmath33 and not @xmath57 .
is measured by keeping track of how much of the wetting fluid has crossed a cross section of the system over a time step .
@xmath115 is then the ratio of how much non - wetting fluid has passed that cross section divided by how much total fluid has passed . ]
if we assume that the difference between @xmath33 and @xmath57 is of the form ( [ eqn5.5 ] ) , @xmath184 , we find the following modified fractional flow equation @xmath185\;.\ ] ] we have defined @xmath186 .
furthermore , we have used the effective saturation @xmath187 rather than the saturation @xmath32 .
we have set @xmath188 , @xmath189 and @xmath190 @xmath191/s .
the derivative @xmath192 was calculated by forwards difference with @xmath193 .
the difference between the measured and calculated fractional flow curves in fig .
[ fig7 ] is then due to finite - size effects in the averaging process .
theories of immiscible two - phase flow in porous media attempt to provide a set of equations describing the flow at the continuum scale while anchoring the key concepts at the pore level . as mentioned in the introduction
, there are several different theories in existence .
there is the percolation - based approach @xcite , there is the approach based on continuum mixture theory where the interface energies play a major role @xcite , and there is the approach where additional variables describing the theormodynamics of the interfaces are introduced @xcite .
the classical , and by far dominating theory in use today is relative permeability theory @xcite even though it is well known to have serious weaknesses . the two equations ( [ eqn18 ] ) and ( [ eqn20 ] ) are based on general non - equilibrium thermodynamical arguments that only invoke scale symmetry ( [ eqn10 ] ) and a balance of dissipation and power
, they must apply to all these theories . in the following ,
we analyze relative permeability theory in light of our equations .
the relative permability approach consists in making explicit assumptions about the functional form of @xmath33 and @xmath34 through the generalized darcy equations .
returning to the rev and the definitions of section [ system ] , we have @xmath194 and @xmath195 here @xmath196 is the permeability of the porous medium , and @xmath197 and @xmath198 are the relative permeabilities of the wetting and non - wetting fluids .
one distinguishes between the pressure gradient in the wetting fluid @xmath199 and in the non - wetting fluid @xmath200 .
they are related through the capillary pressure @xmath201 by @xmath202 the assumptions concerning the physical properties of the flow made in the relative permeability description are that the relative permeabilities and the capillary pressure are functions of the wetting ( or equivalently , the non - wetting ) saturation _ only , _
@xmath203 , @xmath204 and @xmath205 .
these are strong assumptions , and there is growing evidence that they do not hold in general , see @xcite : the relative permeabilities and the capillary pressure are _ not _ functions of the saturation alone .
let us now turn to the second fundamental equation ( [ eqn20 ] ) , combining it with the relative permeability and capillary pressure equations ( [ eqn7])([eqn9 ] ) . by inserting the expressions for @xmath33 and @xmath34 in terms of the relative permeabilities , and eliminating @xmath206 by equation ( [ eqn9 ] ) ,
we find @xmath207\ \delta p_n & \nonumber\\ -\left[\frac{s_w}{\mu_w}\ \left(\frac{d}{ds_w}\right ) \left(\frac{k_{r , w}}{s_w}\ p_c\right ) \right]&=0\;.\nonumber\\\end{aligned}\ ] ] this equation must be valid for all values of @xmath208 .
hence , it splits into two equations , @xmath209 and @xmath210 this latter equation may in turn be integrated to give @xmath211 where @xmath212 is a reference pressure .
we see here that the fundamental assumptions of relative permeability concept are challenged : equation ( [ eqn32 ] ) _ contains the viscosities _ @xmath42 and @xmath43 .
hence , at least one of the relative permeabilities @xmath197 and @xmath198 _ must _ depend of the viscosity ratio @xmath213 .
we also see from equation ( [ eqn34 ] ) that the capillary pressure @xmath201 _ can not become negative .
_ however , under mixed wetting conditions , it does @xcite .
there are equations relating the capillary pressure to the relative permeability in the literature such as the brooks - corey relation @xcite or the van genuchten equation @xcite .
these equations all build on the work of purcell @xcite who generalized to pore networks the relation between the capillary pressure and the permeability of a single capillary tube via the young formula for interfacial tension .
equation ( [ eqn34 ] ) is different from these earlier equations in that it does not rely on assumptions concerning the physics of the problem apart from the balance of dissipation in and work on the fluids .
we also note that popular parametrizations of the relative permeabilities such as that of corey @xcite and that of lomeland et al .
@xcite ( let "- type ) do not obey equation ( [ eqn32 ] ) .
hence , they are thermodynamically inconsistent , as are the basic assumption that the relative permeabilities @xmath214 and @xmath198 are functions of the saturation @xmath32 alone .
we have in this paper constructed a theory based on non - equilibrium thermodynamics that reduces the immiscible _ two - phase flow _ in porous media to a _
one - phase _ flow problem .
let us now consider a three - dimensional isotropic porous medium .
let @xmath215 be a point somewhere in this porous medium .
the theory that we have developed can then be summarized by the following set of equations , @xmath216 @xmath217 and @xmath218 here ( [ conc-1 ] ) is the conservation law for the wetting saturation .
this expression becomes the buckley - leverett equation @xcite if we set @xmath219 , see equation ( [ eqn21 ] ) and take the incompressibility of the fluids into account .
the non - wetting saturation @xmath220 has been eliminated by using the incompressibility of the two fluids .
this implies that @xmath221 is build into the equation set .
equation ( [ conc-6 ] ) is the three - dimensional version of equation ( [ eqn20 ] ) .
the three - dimensional version of equation ( [ eqn18 ] ) is @xmath222 .
this equation follows by taking the derivative of ( [ conc-5 ] ) with respect to @xmath32 and using ( [ conc-6 ] ) .
hence , this equation is also contained in ( [ conc-1 ] ) to ( [ conc-6 ] ) .
it may replace either of the equations ( [ conc-5 ] ) or ( [ conc-6 ] ) .
these equations are all conservation laws , the last two express power input equals dissipation .
thus , they transcend the details of the porous medium .
equations ( [ conc-1 ] ) to ( [ conc-6 ] ) are 7 equations .
there are ten variables @xmath32 , @xmath223 , @xmath224 and @xmath225 .
the three equations that close the system of equations are the _ constitutive equations _
@xmath226\;,\ ] ] containing the detailed physics of the system .
we see that the constitutive equation does not contain the seepage velocities of the immiscible fluids , @xmath224 and @xmath225 , explicitly .
only the saturation @xmath32 enters .
hence , the constitutive equation ( [ conc-7 ] ) can be interpreted as that of a single fluid depending on one extra variable , @xmath227 .
hence , the equation set ( [ conc-1 ] ) ( [ conc-7 ] ) reduces the immiscible two - phase flow problem in porous media to a one - phase flow problem involving a complex fluid .
this viewpoint permeates recent work on the effective permeability of immicible two - phase systems where it is suggested that two fluids behave as if they were a single bingham plastic @xcite .
we may clarify this point even further by eliminating the two fluid velocities @xmath224 and @xmath225 in equations ( [ conc-1 ] ) to ( [ conc-6 ] ) .
the equation set then reduces to a single equation @xmath228 together with the constitutive equation ( [ conc-7 ] ) , we now have a closed set describing effectively a single fluid with a velocity field @xmath223 which transports an active scalar @xmath32 . in this manuscript
we have considered a single driving force that induces the flow : the pressure gradient .
other forces such as buoyancy , temperature gradients and chemical driving forces can be incorporated in the non - equil - ibrium thermodynamics formalism used in section [ diss ] .
the same is true for the introduction of more immiscible fluids than two .
hence , equations ( [ conc-1 ] ) to ( [ conc-6 ] ) may be generalized to include these additional complications .
the authors thank eirik grude flekky , knut jrgen mly , thomas ramstad , per arne slotte and marios valavanides for interesting discussions on this topic .
ah , sk and is thank vista , a collaboration between statoil and the norwegian academy of sciences , for financial support .
ss thanks the norwegian research council , nfr and the beijing computational science research center csrc for financial support .
k. t. tallakstad , h. a. knudsen , t. ramstad , g. lvoll , k. j. mly , r. toussaint and e. g. flekky , steady - state two - phase flow in porous media : statistics and transport properties , phys . rev .
lett . * 102 * , 074502 ( 2009 ) .
k. t. tallakstad , g. lvoll , h. a. knudsen , t. ramstad , e. g. flekky and k. j. mly , steady - state simultaneous two - phase flow in porous media : an experimental study , phys .
e * 80 * , 036308 ( 2009 ) .
e. m. rassi , s. l. codd and j. d. seymour , nuclear magnetic resonance characterization of the stationary dynamics of partially saturated media during steady - state infiltration flow , n. j. phys .
* 13 * , 015007 ( 2011 ) .
s. sinha , a. t. bender , m. danczyk , k. keepseagle , c. a. prather , j. m. bray , l. w. thrane , j. d. seymor , s. l. codd and a. hansen , effective rheology of two - phase flow in three - dimensional porous media : experiment and simulation , submitted to trans .
( 2016 ) .
r. g. bentsen and j. trivedi , on the construction of an experimentally based set of equations to describe cocurrent or countercurrent , two - phase flow of immiscible fluids through porous media , transp . in por
. media , * 99 * , 251 ( 2013 ) .
m. moura , e. -a .
florentino , k. j. mly , g. schfer and r. toussaint , impact of sample geometry on the measurement of pressure - saturation curves : experiments and simulations , water res .
51 * , 8900 ( 2015 ) .
|
based on non - equilibrium thermodynamics we derive a set of general equations relating the partial volumetric flow rates to each other and to the total volumetric flow rate in immiscible two - phase flow in porous media .
these equations together with the conservation of saturation reduces the immiscible two - phase flow problem to a single - phase flow problem of a complex fluid .
we discuss the new equation in terms of the relative permeability equations .
we test the equations on model systems , both analytically and numerically .
|
work is currently in progress to upgrade the cylindrical ooty radio telescope ( ort ) ) so that it functions as a linear interferometric array the ooty wide field array ( owfa ; prasad & subrahmanya 2011a , b ; ram marthi & chengalur 2014 ) .
this telescope operates at a nominal frequency of @xmath9 which corresponds to the neutral hydrogen ( hi ) @xmath10 radiation from a redshift @xmath11 .
observations of the fluctuations in the contribution from the hi to the diffuse background radiation are a very interesting probe of the large - scale structures in the high-@xmath12 universe ( bharadwaj , nath & sethi 2001,bharadwaj , & sethi 2001 ) .
in addition to the power spectrum ( bharadwaj & pandey 2003 , bharadwaj & srikant 2004 ) this is also a sensitive probe of the bispectrum ( ali , bharadwaj and pandey 2006 , guha sarkar & hazra 2013 ) .
there has been a continued , growing interest towards the detection of the 21 cm signal from the lower redshifts @xmath13 to probe the post - reionization era ( chang et al .
2008 ; visbal et al . 2009 ; bharadwaj et al . 2009
; wyithe & loeb 2009 ; bagla , khandai & datta 2010 ; seo et al . 2010 ; mao 2012 ; ansari et al . 2012 ; bull et al . 2014 ; villaescusa - navarro et al .
recently , ali & bharadwaj ( 2014 ) ( henceforth , paper i ) have studied the prospects for detecting the hi signal from redshift @xmath11 using owfa .
the owfa provides an unique opportunity to study the large scale structures at @xmath11 . a number of similar upcoming packed radio interferometer ( chime , bandura et al .
2014 ; baobab and the kzn array ) have been proposed to probe the expansion history of the low - redshift universe ( @xmath14 ) with an unprecedented precision using bao measurements from the large - scale hi fluctuations .
even more innovative designs are being planned for the future low frequency telescope ska .
this promises to yield highly significant measurements of the hi power spectrum over a large redshift range spanning nearly the entire post - reionization era @xmath15 .
however , the detection of the faint @xmath16 hi signal ( @xmath17 ) is extremely challenging due to the presence of different astrophysical foregrounds .
the foregrounds are four to five orders of magnitude brighter than the post - reionization hi signal ( ghosh et al .
2011a,2011b ) . in this paper
, we have considered the visibility correlation ( bharadwaj , & sethi 2001 , bharadwaj , & ali 2005 ) which essentially is the data covariance matrix that is necessary to calculate the fisher matrix .
we have employed the fisher matrix technique to predict the expected signal - to - noise ratios ( snr ) for detecting the hi signal . in our analysis
we have assumed that the hi traces the total matter with a linear bias , and the matter power spectrum is precisely as predicted by the standard lcdm model with the parameter values mentioned later .
the hi power spectrum is then completely specified by two parameters @xmath4 , which sets the overall amplitude of the power spectrum , and @xmath7 the redshift distortion parameter .
the parameter @xmath4 here is the product of the mean neutral hydrogen fraction ( @xmath18 ) and the linear bias parameter ( @xmath19 ) . for a detection , we focus on measuring the amplitude @xmath4 , marginalizing over @xmath7 .
we also consider the joint estimation of @xmath4 and @xmath7 .
our entire analysis is based on the assumption that the visibility data contains only the signal and the noise , and the foregrounds and radio - frequency interference have been completely removed from the data .
the bao feature is within the baseline range covered by owfa ( paper i ) .
however , the frequency coverage ( @xmath20 ) is rather small .
further , for the present analysis we have only considered observations in a single field of view .
all of these result in having very few fourier modes across the @xmath21 range relevant for the bao , and we do not consider this here .
the rest of the paper is organized as follows .
section 2 briefly discusses some relevant system parameters for owfa . in section 3 , we present the theoretical model for calculating the signal and noise covariance , and predict their respective contributions . here
we also estimate the range of k - modes which are probed by the owfa . in section 4
we use the fisher matrix analysis to make predictions for the snr as a function of the observing time .
finally , we present summary and conclusions in section 5 . in this paper , we have used the ( planck @xmath22 wmap ) best - fit @xmath23cdm cosmology with cosmological parameters ( ade et al .
2013 ) : @xmath24 .
we have used the matter transfer function from the fitting formula of eisenstein & hu ( 1998 ) incorporating the effect of baryonic features .
[ tab : array ] .system parameters for phases i , ii , iii and iv of the owfa . [ cols="<,^,^,^,^",options="header " , ] we find ( figure [ fig : snr ] ) that for small observing times @xmath25 , where the visibilities are dominated by the system noise , the conditional and the marginalized snr both increase as @xmath26 .
the increase in the snr is slower for larger observing times , and it is expected to subsequently saturate at a limiting value which is set by the cosmic variance for very large observing times not shown here .
we see ( table [ tab : error ] ) that @xmath27 of observation are needed for a @xmath3 detection with phase i. the corresponding observing time for phase ii falls drastically to @xmath28 and @xmath29 for the conditional and the marginalized cases respectively . for phase
ii , the hi signal is largely dominated by the low wave numbers @xmath30 ( discussed later ) .
phase i which has a larger antenna spacing and smaller frequency bandwidth does not cover many of the low @xmath21 modes which dominate the signal contribution for phase ii .
the required observing times are @xmath31 and @xmath32 of those for phase ii for phases iii and iv respectively .
the marginalized snr are somewhat smaller than the conditional ones , the difference however is not very large . the required observing time does not differ very much except for phase ii where it increases from @xmath28 to @xmath29 for a @xmath3 detection .
.2 cm [ c][c][0.8][0]@xmath33 [ c][c][0.8][0]@xmath34 we have considered the joint estimation of the two parameters @xmath4 and @xmath7 using owfa . figure [ fig : contour ] shows the expected @xmath35 confidence intervals for @xmath36 and @xmath37 with three different observing times ( 630 , 1600 and 4000 hr ) for phases ii , iii and iv . for phase ii , a joint estimation of the parameters @xmath4 and @xmath7 is possible with 15% and 60% errors respectively using 1600 hr of observation .
the errors on the parameters @xmath4 and @xmath7 for 4000 hr are @xmath38 times smaller as compared to 1600 hr .
the constraints are more tight in case of phases iii and iv .
a joint detection of @xmath4 and @xmath7 with 3% and 15% errors respectively is feasible with 1600 hr of observation with phase iv .
[ c][c][1.0][0]@xmath39 [ c][c][1.0][0]@xmath21 mpc@xmath40 [ c][c][0.7][0]@xmath41 [ c][c][0.7][0]@xmath42 [ c][c][0.7][0]@xmath43 [ c][c][0.8][0]phase ii [ c][c][0.7][0]@xmath44=150 hr [ c][c][0.7][0]@xmath44=800 hr 0.01 cm it is interesting to investigate the @xmath21 range that contribute most to the hi signal at owfa .
we have seen that the fourier modes @xmath21 sampled by owfa are predominantly in the range @xmath45 , and there are relatively few modes in the range @xmath46 ( figure [ fig : pkhi2 ] ) .
however , the hi signal ( figure [ fig : convo1 ] ) is much stronger at the smaller modes , whereas the larger modes have a weaker hi signal and are dominated by the noise .
it is therefore not evident as to which @xmath21 range contributes the most to the owfa hi signal detection .
figure [ fig : pkbin ] shows the relative contributions to the fisher matrix from the different @xmath21 modes .
we see that for @xmath47 , which corresponds to a @xmath3 detection , the bulk of the contribution is from the range @xmath48 .
the larger modes do not contribute much to the signal .
we have also considered @xmath49 .
here we have a slightly larger range @xmath2 and the contribution peaks around @xmath50 . in a nutshell , the owfa hi signal is predominantly from the @xmath21 range @xmath51 . the larger modes ,
though abundant , do not contribute much to the hi signal .
we have considered four different phases of ofwa , and studied the prospects of detecting the redshifted 21-cm hi signal at @xmath52 which corresponds to redshift @xmath0 .
phases i and ii are currently under development and are expected to be functional in the near future .
phases iii and iv are two hypothetical configurations which have been considered as possible future expansions . we have used the fisher matrix analysis to predict the accuracy with which it will be possible to estimate the two parameters @xmath4 and @xmath7 using owfa . here @xmath4 is the amplitude of the 21-cm brightness temperature power spectrum and @xmath7 is the linear redshift space distortion parameter . for the purpose of this work
we make the most optimistic assumption that the foreground contributions are not changing across different frequency channels , and hence they only contribute to the @xmath53 mode . in reality
the foreground contamination will extend to other modes also .
further , the chromatic response of the interferometer , calibration errors , systematics in the receivers and radio - frequency interference ( rfi ) have not been considered in the paper .
focusing first on just detecting the hi signal , we have marginalized @xmath7 and considered the error estimates on @xmath4 alone .
we find that a @xmath3 detection of the hi signal is possible with @xmath54 and @xmath29 of observation for phases i and ii respectively .
the observing time is reduced by factor @xmath31 and @xmath32 compared to phase ii for phases iii and iv respectively .
we find that there is a significant improvement in the prospects of a detection using phase ii as compared to phase i , and we have mainly considered phase ii for much of the discussion in the paper .
we have also considered the joint estimation of the parameters @xmath4 and @xmath7 . for phase ii , a joint estimation of the parameters @xmath4 and @xmath7 is possible with 15% and 60% errors respectively using 1600 hr of observation . to estimate @xmath7
it is necessary to sample fourier modes @xmath8 of a fixed magnitude @xmath21 which are oriented at different directions to the line of sight . in other words
, @xmath55 should uniformly span the entire range @xmath56 .
however , the @xmath57 values are much larger than @xmath58 , and the fourier modes are largely concentrated around @xmath59 , for phase ii ( section [ sec : vc ] ) .
the restriction arises from the limited owfa frequency bandwidth ( table [ tab : array ] ) which is restricted by the anti - aliasing filter .
multi - field observations and larger bandwidth ( @xmath60 of the owfa hold the potential to probe of the expansion history and constrain cosmological parameters using bao measurements from the large - scale hi fluctuations at @xmath0 .
anisotropies in the clustering pattern in redshifted 21-cm maps at this redshift produced by alcock - paczyski effect has the possibility of probing cosmology and structure formation .
it is also possible to constrain neutrino masses of using owfa and compare among different fields of cosmology ( lss , cmbr , bbn ) .
thus the owfa could provide highly complementary constraints on neutrino masses .
we leave investigation of such issues for future studies .
the present work has assumed that the shape of the hi power spectrum is exactly determined by the @xmath23cdm model , and has only focused estimating the overall amplitude @xmath4 from owfa observations .
the owfa hi signal is predominantly from the @xmath21 range @xmath61 .
it is possible to use owfa observations to estimate @xmath62 in several separate bins over this @xmath21 range , without assuming the anything about the shape of the hi power spectrum . in a forthcoming paper ,
we plan to calculate fisher matrix estimates for the binned power spectrum .
the authors acknowledge jayaram n. chengalur , jasjeet s. bagla , tirthankar roy choudhury , c.r .
subrahmanya , p.k .
manoharan and visweshwar ram marthi for useful discussions .
aks would like to acknowledge rajesh mon- dal and suman chatterjee for their help .
ssa would like to acknowledge cts , iit kharagpur for the use of its facilities and the support by dst , india , under project no .
sr / ftp / ps-088/2010 .
ssa would also like to thank the authorities of iucaa , pune , india for providing the visiting as- sociateship programme .
|
we have used the fisher matrix formalism to quantify the prospects of detecting the @xmath0 redshifted 21-cm hi power spectrum with the upcoming radio - imterferometric array owfa .
owfa s frequency and baseline coverage spans comoving fourier modes in the range @xmath1 .
the owfa hi signal , however , is predominantly from the range @xmath2 . the larger modes ,
though abundant , do not contribute much to the hi signal . in this work
we have focused on combining the entire signal to achieve a detection .
we find that a @xmath3 detection of @xmath4 is possible with @xmath5 of observations , here @xmath6 is the amplitude of the hi power spectrum .
we have also carried out a joint analysis for @xmath4 and @xmath7 the redshift space distortion parameter .
our study shows that owfa is very sensitive to the amplitude of the hi power spectrum .
however , the anisotropic distribution of the @xmath8 modes does not make it very suitable for measuring @xmath7 .
cosmology : large scale structure of universe - intergalactic medium - diffuse radiation
|
superconductivity appears in iron pnictides when the antiferromagnetic phase is suppressed by chemical doping of the parent compound or by applying pressure.@xcite it is now well known that an isovalent chemical dilution ( e.g. , the partial replacement of as by p , or fe by ru)@xcite presents notable differences with charge doping ( e.g. , by partially replacing o by f in , where r is a rare earth element,@xcite or ba by k in bafe@xmath0as@xmath0)@xcite .
while isovalently - diluted compounds are closer to the clean limit,@xcite charged dopants create strong scattering potentials that affect superconducting properties like the vortex pinning,@xcite the upper critical field,@xcite and even the superconducting gap symmetry.@xcite in fact , many studies indicate that iron - pnictides with charged dopants present a fully gapped fermi surface ( see e.g. , smfeaso@xmath1f@xmath2,@xcite prfeaso@xmath6 @xcite , bafe@xmath7co@xmath2as@xmath0,@xcite ba@xmath1k@xmath2fe@xmath0as@xmath0,@xcite and bafe@xmath7ni@xmath2as@xmath0 @xcite ) , while isovalently substituted compounds like ba(fe@xmath1ru@xmath2)@xmath0as@xmath0 and bafe@xmath0(as@xmath1p@xmath2)@xmath0 present evidences of nodes.@xcite it is expected that fluctuation effects around the superconducting transition may also present differences in these two types of iron pnictides.@xcite these effects have been already investigated in presence of charge - doping,@xcite but to our knowledge still remain unexplored in isovalently substituted iron - pnictides .
the main aim of the present work is twofold : on the one hand , we will probe the nature of superconducting fluctuation effects in isovalently - diluted compounds , and the applicability of the existing ginzburg - landau ( gl ) approaches in the different regions of the @xmath8 phase diagram . on the other hand
, these effects will be used to obtain precise information about superconducting parameters like the coherence lengths ( or the upper critical fields ) , the anisotropy factors , and the system dimensionality ( a technique sometimes referred to as _
_ fluctuation spectroscopy__@xcite ) .
fluctuation effects may also provide information about the multiband nature of these materials .
in fact , it has been shown that a two - band model predicts a change in the relative amplitude of different fluctuation observables in relation with the single - band case , mainly through a renormalization of the @xmath9-axis coherence length.@xcite among the isovalently - substituted iron pnictides , bafe@xmath0(as@xmath1p@xmath2)@xmath0 with @xmath3 presents the highest critical temperature ( @xmath10 k ) and thus is the best candidate to investigate superconducting fluctuations in these materials . in this work
we present a systematic study of fluctuation effects around @xmath11 in the electrical conductivity and magnetization of several high - quality single crystals of this compound .
the measurements are performed in magnetic fields up to 9 t , which close to @xmath11 is well inside the finite - field ( or prange ) fluctuation regime.@xcite the results will allow to probe the applicability of gl approaches for finite applied magnetic fields , and to investigate possible differences in the coherence length amplitudes associated with the multiband nature of this material .
finally , the low - field ( @xmath12 t ) behavior of the magnetization for temperatures around @xmath11 will be studied to check the relevance of phase fluctuations in bafe@xmath0(as@xmath1p@xmath2)@xmath0 , which were claimed to be important in other iron pnictides.@xcite a ) example of a typical edx spectrum .
b ) example of x - ray diffraction pattern on a single crystal , obtained by using the geometry to observe the reflections by the @xmath13 layers .
inset : rocking curve associated to the @xmath14 reflection , showing that the dispersion in the orientation of the crystal @xmath9-axis is about 0.1@xmath15 . ]
the bafe@xmath0(as@xmath1p@xmath2)@xmath0 single crystals used in this work were grown by using the ba@xmath0as@xmath16 / ba@xmath0p@xmath16 self - flux method described in ref . .
some details of their characterization may be seen in ref . . they are plate - like , with typical surfaces of several mm@xmath17 and thicknesses up to @xmath18 mm .
their stoichiometry was checked with a scanning electron microscope ( zeiss fesem ultra plus ) equipped with a edx ( energy dispersive x - ray ) spectroscope . a typical edx spectrum is shown in fig . [ edxrx](a ) .
the average stoichiometry resulted to be ba@xmath19fe@xmath20as@xmath21p@xmath22 , with a variation smaller than 0.4% between the different crystals and the different studied areas .
the partial substitution of as by p is about @xmath23 % , which is close to the value that maximizes @xmath11 .
the crystallographic structure was studied in some of the crystals by x - ray diffraction ( xrd ) by using a rigaku miniflex ii diffractometer with a cu - target .
a typical example of the reflections by the @xmath13 planes is presented in fig .
[ edxrx](b ) . the absence of reflections other than the @xmath24 indicates an excellent structural quality of the crystals .
an example of the rocking curve for the ( 004 ) reflection is presented in the inset of fig .
[ edxrx](b ) .
it confirms that the crystal @xmath9 axis presents a dispersion of only @xmath25 .
powder x - ray diffraction in some grounded crystals allowed to determine the lattice constants of the tetragonal structure , which resulted to be @xmath26 nm and @xmath27 nm , in agreement with data in the literature.@xcite .[tablecrystals ] physical parameters of the crystals .
r1 and r2 were used in the resistivity measurements , and the stack in the magnetization measurements .
[ cols="^,^,^,^ " , ] as it is shown in table [ tableparameters ] , the @xmath5 values determined from the resistivity measurements are in excellent agreement with the one derived from the fluctuation magnetization .
however , there is a significant difference between the corresponding @xmath4 values .
this effect can not be attributed to an incomplete superconducting volume fraction in the stack used for the magnetization measurements , because @xmath5 and @xmath28 ( or @xmath4 ) affect the amplitudes of @xmath29 and @xmath30 , but also their dependence on @xmath31 .
the @xmath5 and @xmath4 values in the literature for bafe@xmath0(as@xmath1p@xmath2)@xmath0 with @xmath11 values near the maximum one , are summarized in table [ tableliterature ] .
these data were obtained from the @xmath32 dependence , as probed by dc and ac electrical transport and by a static property like the specific heat .
the @xmath5 values in these works are close to each other and to the values obtained here ( see table [ tableparameters ] ) .
the @xmath4 value obtained from dc resistivity@xcite is also in excellent agreement with our data obtained from the same observable .
however , a notable difference is found with the @xmath4 values derived from other observables ( 0.9 - 1.75 nm ) .
these results may be somewhat affected by the uncertainty associated to the use of a criterion to determine @xmath32 , but could indicate a possible dependence of @xmath4 on the observable used to determine it
. our present results may be due to the presence of several bands contributing to the superconductivity in the material under study.@xcite in fact , a recently proposed two - band model predicts that the functional forms of dominating divergences of the fluctuation - induced specific heat and electrical conductivity do not change with respect to the single - band case , but their relative amplitudes are affected mainly by a renormalization of the @xmath9-axis coherence length.@xcite it has been suggested that this effect could be behind a dependence of the upper critical fields of fese@xmath33te@xmath33 on the observable used to determine them . in particular , the @xmath34 values derived from measurements of the specific heat are systematically larger than the ones obtained from the electrical resistivity , and these last are in turn larger than the ones obtained from magnetic torque.@xcite it would be then very interesting to check whether an expression for @xmath35 in a two - band superconductor ( still unavailable ) would also quantitatively account for the differences found here with the data obtained from resistivity measurements .
-dependence of the fluctuation magnetization for temperatures just above @xmath11 and for both @xmath36 ( a ) and @xmath37 ( b ) . as a reference ,
the dashed lines represent the ideal diamagnetic response ( evaluated taking into account the demagnetizing factors of the sample under study ) .
the anomalous upturn appears at temperatures within the transition width ( 0.8 k full width ) .
inset : temperature dependence of the upturn magnetic field , @xmath38 . ] for completeness , we have also explored this region of the phase diagram .
in contrast with the linear behavior predicted by gaussian gl approaches under low fields ( @xmath39 ) , the @xmath40 isotherms just above @xmath11 present an anomalous upturn at a field @xmath41 t for both field orientations ( see fig . [ anomaly ] ) .
below @xmath38 the magnetic susceptibility is orders of magnitude larger than the one associated to gaussian superconducting fluctuations . above
@xmath38 the @xmath42 amplitude decreases and , consistently with fig .
[ fluc ] , for fields about @xmath43 t the conventional ggl behavior is recovered .
a qualitatively similar anomalous behavior has been observed in high-@xmath11 cuprates like la@xmath7sr@xmath2cuo@xmath44@xcite and y@xmath1ca@xmath2ba@xmath0cu@xmath16o@xmath45,@xcite and more recently in fe - based superconductors like smfeaso@xmath1f@xmath2 ( sm1111 ) @xcite and ba(fe@xmath1rh@xmath2)@xmath0as@xmath0 ( rh122)@xcite . the similarities between our results and these previous works in iron pnictides are even quantitative . for instance ,
the isotherms measured at temperatures near @xmath11 present very similar @xmath38 and @xmath46 values ( in the case of sm1111 one has to take into account its 2d nature and that the sample is granular ) .
so , to a large extent the effect in bafe@xmath0(as@xmath1p@xmath2)@xmath0 has the same origin than in rh122 and sm1111 . in refs .
and it has been proposed that , like in the htsc , the anomaly is due to the presence of important phase fluctuations .
alternatively , it has been proposed that it is a consequence of a @xmath11 distribution , which in cuprate or fe - based superconductors would be associated to their non - stoichiometric nature.@xcite in the case of bafe@xmath0(as@xmath1p@xmath2)@xmath0 some indications support this last explanation : on the one side , below @xmath38 the magnetic susceptibility is a significant fraction of the perfect diamagnetism ( indicated as a dashed line in fig . [ anomaly ] ) , as if a significant volume fraction of the sample were actually in the meissner region .
on the other , the effect is limited to a range of temperatures above @xmath11 of the order of the transition width .
finally , the @xmath38 value at the average @xmath11 is consistent with the lower critical field of the highest-@xmath11 domains , which may be approximated by @xmath47 , and that is in the @xmath48 t range.@xcite an argument supporting the possible presence of phase fluctuations in sm1111 and rh122 is that @xmath38 increases with the temperature above @xmath11 , while the lower critical field decreases.@xcite such an @xmath38 increase with @xmath49 is not observed in bafe@xmath0(as@xmath1p@xmath2)@xmath0 ( see inset in fig . [ anomaly ] ) , in spite that the transition width of our sample is comparable or even smaller ( in relative terms ) than those in refs . and . as a consequence
, we may assume that phase fluctuations do not play a role in this compound , or at least they are much less relevant than in sm1111 and rh122 .
we have presented measurements of the magnetization and electrical resistivity around the superconducting transition of the isovalently - substituted iron pnictide bafe@xmath0(as@xmath1p@xmath2)@xmath0 with @xmath3 , near the optimal ( maximum-@xmath11 ) value .
the measurements were performed with magnetic fields up to 9 t in the conductivity and 6 t in the magnetization , in both cases applied in the two main crystal directions ( parallel and perpendicular to the @xmath13 layers ) . due to the relatively large normal - state electrical conductivity near @xmath11 ,
the fluctuation effects in this observable are almost unobservable . as a consequence ,
the resistive transitions are very well defined , allowing to determine with accuracy the temperature dependence of the parallel and perpendicular upper critical fields , and the in - plane @xmath5 and out - of - plane @xmath4 coherence lengths . on the contrary
, the magnetization presents an appreciable diamagnetic contribution just above @xmath11 that may be attributed to superconducting fluctuations .
these data were consistently analyzed in the different fluctuation regimes ( gaussian and critical ) in terms of existing ginzburg - landau approaches for single - band superconductors .
while the resulting @xmath5 is in excellent agreement with the one obtained from the field dependence of the resistive transition , @xmath4 is significantly larger ( @xmath50% ) .
a dependence of the coherence lengths with the observable used to determine them was also found in other fe - based superconductors.@xcite recent theoretical approaches show that the presence of several bands contributing to the superconductivity affects the fluctuation effects mainly through a renormalization of the @xmath9-axis coherence length.@xcite thus , we suggest that our present results could be a consequence of the multiband nature of the compound under study . an expression for the fluctuation magnetization in two - band superconductors would be desirable to confirm this proposal . finally ,
from the analysis of the anomalous upturn in the @xmath51 isotherms just above @xmath11 , we find that phase fluctuations may play a negligible role in this compound , contrary to what is proposed for other fe - based superconductors.@xcite this work was supported by the xunta de galicia ( grant no .
gpc2014/038 ) , and cost action mp1201 ( nanosc ) .
ar - a acknowledge financial support from spain s micinn through a fpi grant ( no .
bes-2011 - 046820 ) .
the work at iop , cas in china is supported by nsfc program ( no .
11374011 ) , most of china ( 973 project : 2011cba00110 ) and cas ( sprp - b : xdb07020300 ) .
h. shishido , a. f. bangura , a. i. coldea , s. tonegawa , k. hashimoto , s. kasahara , p. m. c. rourke , h. ikeda , t. terashima , r. settai , y. nuki , d. vignolles , c. proust , b. vignolle , a. mccollam , y. matsuda , t. shibauchi , and a. carrington , phys .
104 * , 057008 ( 2010 ) .
d. watanabe , t. yamashita , y. kawamoto , s. kurata , y. mizukami , t. ohta , s. kasahara , m. yamashita , t. saito , h. fukazawa , y. kohori , s. ishida , k. kihou , c.h .
lee , a. iyo , h. eisaki , a.b .
vorontsov , t. shibauchi and y. matsuda , phys .
b * 89 * , 115112 ( 2014 ) .
t. yoshida , s. ideta , t. shimojima , w. malaeb , k. shinada , h. suzuki , i. nishi , a. fujimori , k. ishizaka , s. shin , y. nakashima , h. anzai , m. arita , a. ino , h. namatame , m. taniguchi , h .
kumigashira , k. ono , s. kasahara , t. shibauchi , t. terashima , y. matsuda , m. nakajima , s. uchida , y. tomioka , t. ito , k. kihou , c. h. lee , a. iyo , h. eisaki , h. ikeda , r. arita , t. saito , s. onari and h. kontani , sci .
rep . * 4 * , 7292 ( 2014 ) . for reviews on the theoretical and experimental aspects of the superconducting fluctuations see m. tinkham , _ introduction to superconductivity _
( mcgraw - hill , new york , 1996 ) , chap . 8 ; f. vidal and m. v. ramallo , _ the gap symmentry and fluctuations in high-@xmath11 superconductors _ ( plenum , london , 1998 ) , p. 443 ; a. larkin and a. varlamov , _ theory of fluctuations in superconductors _ ( oxford university press , oxford , 2005 ) .
m. putti , i. pallecchi , e. bellingeri , m. r. cimberle , m. tropeano , c. ferdeghini , a. palenzona , c. tarantini , a. yamamoto , j.jiang , j. jaroszynski , f. kametani , d. abraimov , a. polyanskii , j. d. weiss , e. e. hellstrom , a. gurevich , d. c. larbalestier , r. jin , b. c. sales , a. s. sefat , m. a. mcguire , d. mandrus , p. cheng , y. jia , h. h. wen , s. lee , and c. b. eom , supercond .
sci . technol . * 23 * , 034003 ( 2010 ) .
d. hu , x. lu , w. zhang , h. luo , s. li , p. wang , g .
chen , f. han , s. r. banjara , a. sapkota , a. kreyssig , a. i. goldman , z. yamani , c. niedermayer , m. skoulatos , r. georgii , t. keller , p. wang , w. yu , and p. dai , phys .
lett . * 114 * , 157002 ( 2015 ) . c. putzke , p. walmsley , j. d. fletcher , l. malone , d. vignolles , c. proust , s. badoux , p. see , h. e. beere , d. a. ritchie , s. kasahara , y. mizukami , t. shibauchi , y. matsuda , and a. carrington , nat
. commun . * 5 * , 5679 ( 2014 ) .
|
the nature of superconducting fluctuation effects in the isovalently - substituted iron pnictide bafe@xmath0(as@xmath1p@xmath2)@xmath0 ( @xmath3 ) is probed through measurements of the magnetization and magnetoconductivity around the superconducting transition .
the results , obtained with magnetic fields up to 9 t applied in the two main crystal directions , are consistent with anisotropic ginzburg - landau ( gl ) approaches for finite applied magnetic fields .
the analysis allowed to determine with accuracy the out - of - plane , @xmath4 , and in - plane , @xmath5 , gl coherence lengths .
significant differences are found between the @xmath4 values resulting from electrical transport and magnetization data . according to recent theoretical approaches
, these differences could be interpreted in terms of the multiband nature of this material .
the analysis of data in the low field region around the transition temperature also suggests that phase fluctuations , although possibly relevant in other fe - based superconductors , may play a negligible role in this compound .
|
* characteristic oscillations in the photon number distribution of dynamically displaced number states for @xmath68 , @xmath108 and @xmath109 . in ( a ) and ( b ) the quantum limit , eq . ( [ 8 ] ) , is displayed for different values of @xmath57 . in ( c ) and
( d ) the semi - classical limit , eq .
( [ 9 ] ) , is shown for different values of @xmath110 .
the full line shows the poisson distribution with @xmath70 . *
coherent destruction of tunneling monitored in the cavity field .
displayed is eq .
( [ 9 ] ) for @xmath111 , @xmath108 as a function of @xmath72 : upper figure @xmath112 , lower figure @xmath113 . the first root of @xmath91 occurs at @xmath114 .
the effect is exhibited by a decrease of the amplitude modulation in the cavity mode oscillations .
|
* abstract . * the mechanism of coherent destruction of tunneling found by grossmann _
et al . _ [ phys .
rev . lett . * 67 * , 516 ( 1991 ) ] is studied from the viewpoint of quantum optics by considering the photon statistics of a single mode cavity field which is strongly coupled to a two - level tunneling system ( ts ) . as a function of the interaction time between ts and cavity the photon statistics displays the tunneling dynamics . in the semi - classical limit of high photon occupation number @xmath0 , coherent destruction of tunneling
is exhibited in a slowing down of an amplitude modulation for certain parameter ratios of the field .
the phenomenon is explained as arising from interference between displaced number states in phase space which survives the large @xmath0 limit due to identical @xmath1 scaling between orbit width and displacement .
pacs number(s ) : 42.50.ct , 42.50.lc , 05.30.-d 2 in recent years the idea of modulation and therefore control of tunneling by a monochromatic electromagnetic field has been subject of considerable interest .
the typical hamiltonian describes a particle in an isolated double - well potential ( dwp ) which is periodically driven by an external force @xmath2 here @xmath3 is the hamiltonian of the dwp , @xmath4 is the amplitude and @xmath5 the driving frequency .
the attention has mainly focused on a possible enhancement or suppression of coherent tunneling : lin and ballentine@xcite have demonstrated that the tunneling probability is highly enhanced due to periodic modulation for high field strengths and driving frequencies close to the classical oscillation frequency at the bottom of each well . in the opposite limit grossmann , hnggi and coworkers@xcite found complete suppression of tunneling such that a particle initially localized in one of the two wells will never escape to the other well .
they termed this effect `` coherent destruction of tunneling '' .
it has since been of continuing interest.@xcite the most surprising feature is the _ periodicity _ of the destruction of tunneling for certain parameter ratios of @xmath4 and @xmath5 . so far there is no clear understanding of this phenomenon .
by using the floquet formalism it has been shown that many characteristic features of the tunneling suppression can already be described in a two - level approximation of the dwp.@xcite many different aspects of the effect have been illuminated in this framework : makarov@xcite , plata and gomez llorente@xcite quantized the electromagnetic field and recovered the effect in the limit of large number of photons in the field .
wang and shao@xcite mapped the driven two - level dynamics to a classical one of a charged particle moving in a harmonic potential plus a magnetic field in a plane .
kayanuma@xcite explained the suppression of tunneling as an effect arising from interference at periodic level crossings . in the present rapid communication ,
an alternative explanation is proposed which uses the concept of _ phase space interference _ known from squeezed states@xcite and displaced number states@xcite in quantum optics .
let us start with a description of the physical situation we have in mind . at @xmath6 a single two - level tunneling system ( ts ) prepared in a state localized in
say the left well @xmath7 is injected in a cavity and starts to tunnel between its left and right state @xmath7 and @xmath8 .
the cavity contains a single mode which has been prepared in a number state @xmath9 .
we consider an ideal cavity , i.e. , we neglect any kind of dissipation . after an interaction time
@xmath10 the ts leaves the cavity and the photon number distribution is measured irrespective of the state of the ts .
thus our aim will be to calculate the transition probability from the product state @xmath11 at @xmath6 to another product state @xmath12 for any @xmath13 at time @xmath14 , @xmath15 where @xmath16 . for fixed @xmath10
this is the photon number distribution with @xmath17 . for fixed @xmath18
, @xmath19 describes the dynamics of the cavity mode interacting with a ts .
an important aspect of the present problem is the strong coupling between the cavity mode and the ts , and the separation of timescales between the _ slow _ tunneling motion and the _ fast _ field oscillations .
if we denote the coupling energy by @xmath20 and the tunneling frequency by @xmath21 , this means that we are interested in the limit @xmath22 and @xmath23 for low , and @xmath24 for high number of photons in the field . in this regime
the field strongly dresses the ts and both ts + field have to be treated as a single unit .
in contrast to the situation where @xmath25 , we are confronted in the present case with a situation in which the rotating - wave approximation is not applicable .
thus instead of using the jaynes - cummings hamiltonian@xcite , we must include counter - rotating terms so that our hamiltonian reads @xmath26 where @xmath27 , @xmath28 are the bosonic creation and annihilation operators of the cavity mode
. we have identified @xmath29 and @xmath30 and @xmath31 , the usual formulation of the hamiltionian in quantum optics is recovered .
we also have added a constant energy shift for later convenience .
it has to be noted that suppression of tunneling does not occur in the jaynes - cummings model.@xcite hence the situation here is fundamentially different from the one usually encountered in quantum optics .
owing to the separation of timescales between the tunneling and the oscillation dynamics , @xmath19 can be calculated in perturbation theory in @xmath32 by introducing dressed states @xmath33 and @xmath34\ ] ] where @xmath35 .
equivalently , one may perform the polaron transformation @xmath36 and continue to use the product state instead of the dressed state basis . in the dressed state basis
the hamiltonian can be written as @xmath37 with @xmath38 and @xmath39 where @xmath40 $ ] is the shift operator and @xmath41 an associate laguerre polynomial ( @xmath42 ) .
the diagonal part @xmath43 builds a ladder of tunneling doublets with intra - doublet spacing @xmath44 and inter - doublet spacing @xmath45 . because @xmath46 , @xmath47 induces only mixing between dressed states belonging to _ different _ doublets
hence , corrections to the dressed states are @xmath48 . neglecting @xmath47 for this reason we find for the transition probability if @xmath42 @xmath49 and @xmath50 if @xmath51 .
_ we conclude that the tunneling dynamics can be seen in the spectrum of the transition probability of the cavity field . _ in addition to harmonics of @xmath5 its power spectrum also contains resonances at @xmath52 arising from the tunneling motion .
this behavior strongly depends on the initial preparation of the ts .
injecting it in its ground state @xmath53 , for instance , yields only resonances at @xmath54 where @xmath55 depending on whether @xmath56 is even or odd , respectively , and hence only rapid oscillations .
the effect of a strong coupling between the cavity mode and the ts is to mix a coherent amplitude @xmath57 with the intial number state of the mode .
this happens by displacing the oscillator wave function @xmath58 like @xmath59 in dimensionless coordinates @xmath60 .
thus by injecting a ts which strongly couples to the cavity field , it is possible to realize displaced number states@xcite .
the statistical properties of displaced number states have been discussed in ref . . the photon number distribution is simply given by @xmath61 because @xmath62 is the probability amplititude of finding @xmath18 photons in a displaced number state @xmath63 .
the photon distribution ( [ 8 ] ) for number states which are _ dynamically _ displaced by the tunneling process resembles @xmath64 .
the displaced number states can either be shifted into the same well ( first term in ( [ 8 ] ) ) or opposite wells ( second term in ( [ 8 ] ) ) . as expected for @xmath65 , @xmath66 independent of the interaction time . in fig .
1(a),(b ) we have plotted @xmath19 for @xmath67 , @xmath68 and @xmath69 0.25 and 0.5 .
whereas a coherent state with @xmath70 obeys the familiar poisson distribution@xcite shown by the line , the dynamically displaced number states exhibit oscillations shown by the histogram . increasing @xmath71 results in further oscillations
this effect is independent of the specific value of @xmath72 for @xmath73 , though the absolute value of @xmath19 depends on @xmath72 .
these modulations are clearly exhibited in the asymptotic expansion of ( [ 8 ] ) in the semi - classical limit ( bohr s correspondence principle ) : @xmath74 , @xmath75 with @xmath76 finite .
if we scale the coupling constant between the field and the ts as @xmath77 and note that associate laguerre polynomials asymptotically approach bessel functions@xcite one finds in the semi - classical limit @xmath78 here @xmath79 is the renormalized tunneling frequency , and @xmath80 is the rabi frequency . in fig .
1(c),(d ) , eq .
( [ 9 ] ) is plotted for @xmath81 1 and 3.3 .
it shows the same oscillations as the exact expression ( [ 8 ] ) .
the expression ( [ 9 ] ) is easily understood .
the cosine and the sine factors represent the probability for the tunneling particle to stay in the well where it has been prepared initially , or to escape to the other well , respectively.@xcite the bessel functions represent the probability for the corresponding displaced number state to contain @xmath18 photons if it had @xmath0 initially ( note that @xmath82 in the semi - classical limit ) . hence the bessel functions represent the photon statistics of displaced number states with an effective displacement @xmath83 or @xmath84 . from ( [ 10 ] ) and the cosine and sine factors in ( [ 9 ] ) we further see that for specific parameter values of the driving field where @xmath85 hits the roots of the zero order bessel function
the tunneling process is completely suppressed .
note that localization can not occur for small photon number in the field because there is no value of @xmath71 which is simultaneously a root of all laguerre polynomials .
we conclude that in the large @xmath0 limit _ the tunneling dynamics is displayed in the photon statistics by an amplitude modulation of the cavity mode oscillations .
_ _ coherent destruction of tunneling manifests itself in the slowing down of this amplitude modulation _ as depicted in fig .
2 .
we now show how the present picture can give a quantitative understanding of this effect .
first recall that the oscillation in the photon number statistics in fig .
1 originate from phase space interference:@xcite the @xmath0th number state can be associated with a circular band of width @xmath86 around its orbit with radius @xmath87 and centered around the origion .
analogously so can the @xmath18th displaced number state which is shifted by @xmath88 . according to the area - of - overlap concept@xcite ,
the transition amplitude between two states is governed by the sum of all possible overlap areas weighted with appropriate phases .
this results in the interference between contributions from different overlaps which can be constructive or destructive , thus , giving rise to oscillations in the photon number distribution .
another way of understanding the oscillations is that they arise from the possibility of a constructive and destructive overlap between displaced harmonic oscillator wave functions by noting that @xmath89 is a polynomial in @xmath18 of order min@xmath90 , i.e. , has min@xmath90 roots . from this consideration
it becomes apparent that varying the relative radius of the two bands by changing @xmath18 vs. @xmath0 , or varying the relative displacement by changing @xmath71 for fixed @xmath18 and @xmath0 should result in similar effects .
if one notices further that the factor @xmath91 in ( [ 10 ] ) is the probability amplitude of finding in a number state displaced by @xmath92 again exactly the same number of photons , one expects that the periodic suppression of tunneling arises from destructive and constructive interference of displaced harmonic oscillator wave functions in phase space . to verify this argument
, the @xmath93 limit has to be considered with care .
first take the number of photons in the cavity mode to be finite . the factor @xmath94 which scales the doublet splitting ( [ 6 ] ) is the overlap between a number state @xmath9 shifted into the right well with the same number state shifted into the left well .
if we note that @xmath94 is a polynomial in @xmath95 of degree @xmath0 , and consequently has @xmath0 zeros , we expect @xmath96 oscillations of @xmath94 as a function of @xmath71 .
the @xmath0 zeros between the maxima result from the @xmath0 possible ways of destructive overlap between displaced harmonic oscillator wave functions .
this explains fig . 2 in ref .
however , contrary to the claim there , this picture remains also valid if the oscillator energy is larger than the reorganization energy @xmath97 , and even in the limit @xmath93 .
the simple reason for this is that in the semi - classical limit @xmath93 with @xmath98 fixed , both the distance between the nodes of the harmonic oscillator wave function ( as a function of @xmath99 ) _ and _ the displacement @xmath92 scale as @xmath1 .
based on this argument , we expect oscillation to become noticable as soon as the displacement exceeds the bandwidth , i.e. , @xmath100 .
finally , we note that with the scaling chosen above both displaced orbits will always intersect no matter how large @xmath101 is .
this results in the infinite number of oscillations seen in the zero order bessel function ( [ 10 ] ) .
summarizing coherent destruction of tunneling is explained as a _ quantum effect arising from the destructive interference of displaced harmonic oscillator wave functions in phase space_. the effect is strictly speaking only observable in the semi - classical limit for the reasons mentioned above , and survives the large @xmath0 limit since both _ the bandwidth of the orbit and the displacement scale in the same way to zero as _ @xmath102 . in closing , we note that experimental study of the effects described here is presently still out of range . in a microcavity
the driving frequency @xmath5 is @xmath103(ghz ) whereas typical values of the one - photon rabi frequency @xmath104 are @xmath103(khz).@xcite hence a number state with large @xmath0 is needed which is difficult to realize experimentally . possibly a trap is more suited because of its lower driving frequency @xmath105(mhz).@xcite damping of the cavity mode which is too strong will also render an observation impossible .
furthermore the tunneling dynamics must still be coherent .
finally the superposition state @xmath106 is difficult to prepare because the particle experiences strong electromagnetic fields when it enters and leaves the cavity@xcite ( for a discussion of this point if one uses quantum wells , see ref . ) .
@xmath107 @xmath107 this research has been supported in part by the nsf and by the alexander von humboldt foundation .
we also thank prof .
peter hnggi and prof .
herbert walther , as well as dr .
min cho and dr .
jochen rau for discussions .
w. a. lin and l. e. ballentine , phys .
rev .
lett .
* 65 * , 2927 ( 1990 ) ; phys .
rev .
a * 45 * , 3637 ( 1992 ) .
f. grossmann , t. dittrich , p. jung , and p. hnggi , phys .
rev .
lett . * 67 * , 516 ( 1991 ) ; f. grossmann and p. hnggi , z. phys .
b * 85 * , 293 ( 1991 ) ; f. grossmann , t. dittrich , p. jung , and p. hnggi , j. stat . phys . *
70 * , 229 ( 1993 ) .
f. grossmann and p. hnggi , europhys . lett . * 18 * , 571 , ( 1992 ) .
j. m. gomez llorente and j. plata , phys .
rev .
a * 45 * , r6954 ( 1992 ) .
d. e. makarov , phys .
rev .
e * 48 * , r4146 ( 1993 ) .
j. plata and j. m. gomez llorente , phys .
rev .
a * 48 * , 782 ( 1993 ) .
l. wang and j. shao , phys .
rev .
a * 49 * , r637 ( 1994 ) .
y. kayanuma , phys .
rev .
a * 50 * , 843 ( 1994 ) .
yu .
dakhnovskii and h. metiu , phys .
rev .
a * 48 * , 2342 ( 1993 ) ; yu .
dakhnovskii and r. d. coalson , j. chem . phys . * 103 * , 2908 ( 1995 ) ; d. e. makarov and n. makri , phys .
rev .
e * 52 * , 5863 ( 1995 ) ; d. g. evans , r. d. coalson , h. kim , and yu .
darkhnovskii , phys .
rev .
lett . * 75 * , 3649 ( 1995 ) .
w. schleich and j. a. wheeler , nature * 326 * , 574 ( 1987 ) ; w. schleich , d. f. walls , and j. a. wheeler , phys .
rev .
a * 38 * , 1177 ( 1988 ) .
d. f. walls and g. j. milburn , _ quantum optics _ , ( springer , berlin , heidelberg , new york , 1995 ) .
s. m. roy and v. singh , phys .
rev .
d * 25 * , 3413 ( 1982 ) ; m. venkata satyanarayana , phys . rev .
d * 32 * , 400 ( 1985 ) .
f. a. m. de oliveira , m. s. kim , p. l. knight , and v. buzek , phys .
rev .
a * 41 * , 2645 ( 1990 ) .
e. t. jaynes and f. w. cummings , proc .
ieee * 51 * , 126 ( 1963 ) .
m. abramowitz and i. a. stegun , _ handbook of mathematical functions _ , ( dover , new york , 1965 ) .
g. rempe , h. walther , and n. klein , phys .
rev .
lett . * 58 * , 353 ( 1987 ) .
h. walther , private communication .
r. bavli and h. metiu , phys .
rev .
lett . * 69 * , 1986 ( 1992 ) ; phys .
rev .
a * 47 * , 3299 ( 1993 ) .
|
we thank all members of the flavianet kaon working group ( http://www.lnf.infn.it/wg/vus ) for comments , discussion , and suggestions .
this work is supported in part by eu contract mtrn - ct-2006 - 035482 ( flavianet ) .
|
we update the numerical results for the s - wave @xmath0 scattering phase - shift difference @xmath1 at @xmath2 from a previous study of isospin breaking in @xmath3 amplitudes in chiral perturbation theory .
we include recent data for the @xmath4 and @xmath5 decay widths and include experimental correlations .
la - ur-07 - 8412 the authors of refs . and have shown the importance of taking into account radiative corrections in @xmath6 decays both in experimental measurements and theoretical predictions .
in particular , these corrections are enhanced in the extraction of the phase shifts @xmath1 by the @xmath7 rule by a factor of @xmath822 . on the experimental side
, kloe has measured with high accuracy the ratio of branching ratios ( brs ) for the decay @xmath9 to the decay @xmath10 @xcite .
this measurement is fully inclusive of radiation for the @xmath11 channel , allowing an unambiguous comparison with theoretical predictions . for the extraction of the phase shift @xmath1 from @xmath6
we follow ref . .
in the presence of electromagnetic interactions , the usual isospin decomposition of amplitudes becomes : @xmath12 where @xmath13 are the infrared - finite amplitudes for the decays @xmath14 and @xmath15 , respectively , and @xmath16 is the amplitude for the decay @xmath17 .
these amplitudes are related to the decay rate by : @xmath18 where @xmath19 .
the factor @xmath20 is associated with the effect of real and virtual photons @xcite .
the following notation is introduced in ref .
: @xmath21 where in the absence of electromagnetic interactions the @xmath22 are the standard isospin amplitudes and the phases @xmath23 are identified with the s - wave @xmath0-scattering phase shifts @xmath24 .
otherwise , we have , by eqs .
( 7.20 ) and ( 7.33 ) of ref .
: @xmath25 and : @xmath26 where @xmath27 are the coefficients of the leading ( @xmath28 ) octet and 27-plet weak non - leptonic chiral operators ( as defined in eqs .
( 2.6 ) and ( 2.7 ) of ref . ) .
note that in order to arrive at eqs .
( [ sviluppo ] ) , a number of higher - order chiral effective couplings are needed .
here we adopt the estimates given in section 5 of ref . .
the coefficients @xmath29 , @xmath30 , and @xmath31 depend on the chiral renormalization scale @xmath32 ( from the matching uncertainty in the low - energy constants ) and the tree - level @xmath33 mixing angle @xmath34 given by : @xmath35 we have obtained the values of these coefficients from the authors of ref . .
eqs .
( [ finiteampl ] ) , ( [ expampl ] ) , and ( [ isospinampl ] ) can be combined to obtain @xmath36 where @xmath37 . using the expansion of eq .
( [ sviluppo ] ) , this system of equations can be written in terms of the three unknowns , @xmath38 , @xmath39 , and @xmath40 .
we solve this system using a numerical minimization procedure .
the experimental inputs used to obtain the amplitudes by eq .
( [ expampl ] ) are listed in table [ tab : exp ] .
as noted above , the brs for @xmath41 and @xmath42 are obtained from the kloe measurement of their ratio , which accounts for the large value of the correlation coefficient .
the br for @xmath5 is weakly correlated with the value of the @xmath43 lifetime by the fit performed in ref . .
these correlations are taken into account in the minimization procedure . while the kloe measurement of the ratio of @xmath4 brs is fully inclusive of radiation in the @xmath11 channel , the inclusiveness of the value for @xmath44 from the fit in ref .
is less well defined .
however , this is of lesser concern because of the dominance of the @xmath45 amplitudes .
.experimental inputs used to obtain @xmath46 , @xmath40 , and @xmath47 .
[ cols="<,^,^,^",options="header " , ] with @xmath48 gev and @xmath49 we obtain : @xmath50 the error matrix is : @xmath51 we compute the systematic error by varying the chiral renormalization scale parameter @xmath32 from 0.5 gev to 1 gev , and the tree - level @xmath33 mixing angle @xmath34 from @xmath52 to @xmath53 , and taking half of the total variation as the uncertainty .
the systematic error matrix is : @xmath54 including the systematic errors and using eq .
( [ deltacorr ] ) we have : @xmath55 the results in eqs .
( [ phaseresult ] ) have been obtained using the central values for the higher - order chiral couplings estimated in ref .
within the large-@xmath56 expansion .
the quoted uncertainty reflects only the uncertainty in the matching scale . as already pointed out in ref .
( section 7.3 ) , the extraction of @xmath46 and @xmath40 is rather sensitive to the input on the chiral couplings .
for instance , even a simple variant of the large-@xmath56 procedure leads to changes in @xmath46 and @xmath40 at the @xmath57 level @xcite .
this implies that the values for @xmath27 quoted in eq .
( [ phaseresult ] ) are affected by an unknown systematic offset of at least @xmath57 .
( see also the analysis of ref . . ) on the other hand , the extraction of the phase difference @xmath47 is quite insensitive to the input on the effective chiral couplings .
the result for the phase difference @xmath1 instead depends quite sensitively on the estimate of the isospin - breaking correction of eq .
( [ deltacorr ] ) .
it is interesting to compare the present result for @xmath1 with predictions from phenomenological evaluations .
the roy equations @xcite determine the @xmath0 scattering amplitude in terms of its imaginary part at intermediate energies , up to two subtraction constants : the s - wave scattering lengths @xmath58 and @xmath59 .
colangelo , gasser , and leutwyler @xcite obtain values for @xmath58 and @xmath59 by matching a representation of the @xmath0 scattering amplitude from @xmath60 calculations in chiral perturbation theory with a phenomenological representation based on the roy equations .
they obtain @xmath61 , which differs from our result by @xmath62 .
kamiski , pelez , and yndurin @xcite fit experimental @xmath0 scattering amplitudes at both low and high energies with parameterizations that satisfy analyticity at low energy , constrained to satisfy the forward dispersion relations and the roy equations .
they obtain @xmath63 , which differs from our result by @xmath64 .
it is important to note that their input data includes low - energy s - wave phase shift determinations from @xmath65 and @xmath3 decays , including a preliminary result for @xmath66 from this update , differing very little from the value presented here .
the significant discrepancies between our result , based on @xmath67 and @xmath43 br measurements , and the results of phenomenological analyses of @xmath0 scattering ( with or without constraints from chiral symmetry ) are puzzling and deserve further investigation .
|
quantum nonlocality in quantum mechanics ( qm ) manifests when combining quantum entanglement of two or more parties with quantum measurement on one of these parties . in the simplest scenario a quantum measurement
_ locally _ ( in a subspace of the full hilbert space ) on one party of an entangled pair ( say , @xmath0 and @xmath1 ) will lead to an instantaneous projection or collapse of the quantum state of both parties such that the other party is also affected .
the situation becomes more intriguing if we add the assumption that both @xmath0 and @xmath1 are local in position space and they are separated at a distance .
then the quantum state of @xmath1 will be projected instantaneously by a local measurement on @xmath0 , no matter how far @xmath1 is away from @xmath0 @xcite . while this appears to have some kind of superluminal signal
, causality will not be violated since no meaningful information can be communicated using such an instantaneous wavefunction collapse .
aharonov and albert have further shown that indirect measurement on quantum objects localized in space and time is consistent with relativistic qm @xcite , where the measurement process has no covariant description in terms of the time evolution of quantum states @xcite and quantum states make sense only within a given frame .
nevertheless , local quantum objects such as atoms or charged particles are inevitably coupled with quantum fields defined on the whole spacelike hypersurface ( the time - slice ) and evolving in time as the environment .
if a relativistic quantum field is involved in the system that we are looking at , will the above scenario of wavefunction collapse still be consistent ?
or more generally , is instantaneous wavefunction collapse in a projective measurement local in position space consistent with relativistic quantum field theory ( rqft ) ?
rqft is very powerful in solving the scattering problem . in particle colliders , measurements on an ensemble of particles
are done in the huge detectors surrounding the collision point , effectively in the future asymptotic region .
so it is sufficient to calculate the scattering amplitude between the in- and out - states defined in past infinity and future infinity , respectively , where all particles are free .
this is called the in - out " formalism , which gives the statistics of the outgoing particles against the incoming particles @xcite .
our questions raised here , however , concern the _ single - shot _ projective measurements on a quantum field at some moment _ in the interaction region _ rather than in the asymptotic region .
so we have to go beyond the in - out " formalism to answer our questions .
one is tempted to generalize the indirect measurement scheme in @xcite to rqft to study this issue .
nevertheless , there each indirect measurement process is modeled by an interaction localized in space and time between a quantum probe and the fields , and the projective measurements on the quantum probes are still performed in the future asymptotic region ( more discussion will be given in appendix a ) , while the wavefunction collapse we are looking at is not described by any interaction hamiltonian .
thus we turn to the standard schrdinger picture of rqft to watch the discontinuous and continuous evolutions of the whole system @xcite , whose quantum state at each moment is represented as the wavefunctional of the fields ( and the sources , if any ) living on the whole associated time - slice . in this formalism
one needs to specify an initial state on the time - slice associated with some non - infinity initial moment in some coordinate where the hamiltonian is defined ( we assume that it is always possible to prepare such an initial state ) , then the schrdinger equation will give the continuous evolution of the quantum state from the initial moment and between the events of the projective measurement .
suppose the postulate of the projective measurement in qm can be extended to rqft , then the wavefunctional of the fields on the whole time - slice would be collapsed by a measurement local in position space if the relevant degrees of freedom have nonzero correlations .
since the dynamical variables of quantum fields can be nonlocal in position space or separable in a quantum state , such a scenario of local measurement must be carefully formulated .
one simple way to achieve this is by going back to the atom - field interacting system : to measure an unruh - dewitt detector @xcite or similar object locally coupled with quantum fields @xcite , analogous to an optical system with a photodiode coupled to em field , as is described in our setup .
below we are looking at , but not limited to , a detector model similar to the unruh - dewitt detector theory . before getting into detailed calculation , in section
ii we give an alternative frame in minkowski space to make our discussion more economic and precise .
then in section iii we introduce the raine - sciama - grove detector theory in ( 1 + 1)d minkowski space as our toy model .
we perform explicit calculations for one - detector and two - detector cases in sections iv and v , respectively , then the results will be summarized in section vi .
finally , we compare our model with those in the indirect measurement scheme in appendix a , and apply an argument similar to that for our linear toy model to nonlinear detector - field models in appendix b.
it has been shown by aharonov and albert that , in relativistic qm , quantum states _ in the non - asymptotic region _ defined on two different time - slices intersecting at some spacetime points could be very different even for the sectors of the dynamical variables defined right on the intersections of the two time - slices @xcite .
thus in the interaction region one can compare two wavefunctionals of a field defined in two different frames only if the whole time - slices that they are living on are exactly the same .
moreover , since the initial state of the detector - field system must be specified on the whole fiducial time - slice associated with the initial moment , if this was done by an observer at rest in minkowski space but not in past infinity , for an observer moving with constant velocity the initial data far enough from the detectors will appear to be specified at some times _ after _ the measurement event on the detector when the wavefunctional was collapsed ( figure [ nucoord ] ( left ) ) .
to avoid this situation and make the quantum states comparable , we have to go beyond the linear lorentz transformation and inertial frames . a good example of the reference frame for our discussion is the following alternative coordinates in ( 1 + 1 ) dimensional minkowski space , @xmath2 with constant @xmath3
. then @xmath4 @xmath5 , where @xmath6 @xmath7 with @xmath8 and @xmath9 according to @xmath10 . here
@xmath11-slices and @xmath12-slices will overlap at @xmath13 with integer @xmath14 , where quantum states can be compared . off
those moments , @xmath11-slices are the wavy ones in figure [ nucoord ] ( right ) , where @xmath12-slices would be the horizontal straight lines . now the questions can be put more precisely .
suppose a pointlike detector coupled with a quantum field is located at @xmath15 and started at its proper time @xmath16 .
if a local measurement is done on the detector at some moment @xmath17 , then which time - slice , a @xmath18-slice or an @xmath19-slice ( @xmath20 ) , will the wavefunctional of the combined system collapse on
? if both collapses occur for different observers , will the two post - measurement states ( pms ) be identical " ? what happens if two measurement events on two detectors are spacelike separated ? and @xmath21 on @xmath22 and @xmath23 , respectively , can be altered by a lorentz transformation : @xmath21 is earlier than @xmath24 in @xmath12 , but later in @xmath25 . here
the dotted lines denote the @xmath12-slices , while the dot - dashed line is a constant @xmath25 hypersurface .
note that the initial data of the quantum field on the dashed - line part of the @xmath26-slice appear to be specified _ after _
@xmath24 and @xmath21 for an inertial observer in coordinate time @xmath25 .
( right ) the alternative coordinates @xmath27 given by @xmath10 with @xmath28 in the @xmath12-@xmath29 diagram of ( 1 + 1)d minkowski space .
the solid and dashed curves are constant @xmath11 and @xmath30 hypersurfaces , respectively.,title="fig:",width=188 ] and @xmath21 on @xmath22 and @xmath23 , respectively , can be altered by a lorentz transformation : @xmath21 is earlier than @xmath24 in @xmath12 , but later in @xmath25 . here
the dotted lines denote the @xmath12-slices , while the dot - dashed line is a constant @xmath25 hypersurface .
note that the initial data of the quantum field on the dashed - line part of the @xmath26-slice appear to be specified _
@xmath24 and @xmath21 for an inertial observer in coordinate time @xmath25 .
( right ) the alternative coordinates @xmath27 given by @xmath10 with @xmath28 in the @xmath12-@xmath29 diagram of ( 1 + 1)d minkowski space .
the solid and dashed curves are constant @xmath11 and @xmath30 hypersurfaces , respectively.,title="fig:",width=188 ]
to answer these questions , let us consider one or more pointlike raine - sciama - grove(rsg ) detectors @xcite coupled to a massless scalar field in ( 1 + 1)d minkowski space , described by the action @xmath31 + \right .
\nonumber\\ & & \left .
\lambda \int d\tau^{}_{\bf d } \partial^{}_{{\bf d } } q^{}_{\bf d } \int d^2x \phi(x^0,x^1 ) \delta^2(x^\alpha- z^\alpha_{\bf d}(\tau^{}_{\bf d } ) ) \right\ } , \label{rsgact}\end{aligned}\ ] ] where @xmath32 or @xmath33 , @xmath34 is the determinant of the metric tensor @xmath35 of the background spacetime , @xmath36 is a massless scalar field , @xmath37 is the internal degree of freedom acting like a harmonic oscillator ( ho ) in the detector @xmath38 with @xmath39 for the one - detector case , @xmath40 for the two - detector case , and also @xmath41 and @xmath42 are the proper time and the prescribed trajectory of the detector @xmath38 , respectively , and @xmath43 .
the momenta conjugate to @xmath44 and @xmath36 read @xmath45 where @xmath46 .
then one can write down the hamiltonian @xmath47 ^ 2 + \omega_{\bf d}^2 q^2_{\bf d}(\tau^{}_{\bf d}(x^0))\right\ } + \nonumber\\ & & { 1\over 2}\int dx^1 \sqrt{-g}\left\{{-1\over g^{00}}\left[{\pi_{x^1}(x^0)\over \sqrt{-g } } + g^{01}\partial_1 \phi_{x^1}(x^0)\right]^2 + g^{11 } \left[\partial_1\phi_{x^1}(x^0)\right]^2\right\},\end{aligned}\ ] ] parameterized by the time variable @xmath48 of the observer s frame and defined on the whole time - slice @xmath49 associated with that time
. solutions of the schrdinger equation with the quantized hamiltonian with @xmath50 and @xmath51 are the wavefunctionals of the detectors and the field @xmath52 $ ] .
suppose that at @xmath53 ( when @xmath54 for all detectors ) the combined system of the detectors and the field is initially in a gaussian state , which could be pure or mixed ( e.g. the direct product of the ground states of the detectors and the minkowski vacuum of the field . )
then the quantum state will always be gaussian by virtue of the linearity of the system . since the explicit form of the wavefunctional @xmath55 or the density matrix @xmath56 $ ] is not quite simple enough to be solved directly , we will work with the equivalent gaussian state in the @xmath57-representation of the density matrix @xcite , which is the double fourier transform of the conventional wigner functional : @xmath58 = \int { \cal d}{\bf \sigma } e^{{i\over\hbar } { \bf k}\cdot { \bf \sigma } } \bar{\rho}\left [ { \bf \sigma } - { { \bf \delta}\over 2 } , { \bf \sigma } + { { \bf \delta}\over 2 } ; x^0 \right ] \nonumber\\ & = & \exp -{1\over 2\hbar^2 } \left [ k^\mu { \cal q}_{\mu\nu}(x^0 ) k^\nu -2 \delta^\mu { \cal r}_{\mu\nu}(x^0 ) k^\nu + \delta^\mu { \cal p}_{\mu\nu}(x^0 ) \delta^\nu \right ] , \label{qstate}\end{aligned}\ ] ] where dewitt notation has been used , @xmath59 run over all the detector and field degrees of freedom defined on the whole time - slice , and the time - dependent factors @xmath60 , @xmath61 , and @xmath62 are actually the symmetric two - point correlators @xmath63 evaluated on the @xmath48-slice , for they are obtained from @xmath64 \right|_{{\bf \delta}={\bf k}=0 } = { \cal q}_{\mu\nu},\nonumber\\ \langle \hat{\pi}^{}_\mu , \hat{\pi}^{}_\nu \rangle & = & \left .
{ i\hbar\delta\over \delta \delta^\mu}{i\hbar\delta\over \delta \delta^\nu } \rho[{\bf k } , { \bf \delta } ] \right|_{{\bf \delta } = { \bf k}=0 } = { \cal p}_{\mu\nu},\nonumber\\ \langle \hat{\pi}^{}_\mu , \hat{\phi}^{}_\nu \rangle & = & \left .
{ i\hbar\delta\over \delta \delta^\mu}{\hbar\delta\over i\delta k^\nu } \rho[{\bf k } , { \bf \delta } ] \right|_{{\bf \delta } = { \bf k}=0 } = { \cal r}_{\mu\nu},\end{aligned}\ ] ] where we denote @xmath65 and @xmath66 by @xmath67 and @xmath68 , respectively .
thus , looking at the evolution of the gaussian state @xmath69 is equivalent to looking at the dynamics of those symmetric two - point correlators , which would be obtained more easily in the heisenberg picture .
the heisenberg equations of motion for the operators @xmath70 and @xmath71 read @xmath72 where @xmath73 . by virtue of the linearity of the system , operators at each moment are linear combinations of the operators defined at the initial moment @xcite : @xmath74}_{{\bf d } ' } + f^{{\bf d}'}_{\bf d}(\tau^{}_{\bf d})\hat{p}^{[0]}_{{\bf d } ' } \right ] \nonumber\\ & & + \int dy \left [ \phi^{y}_{\bf d}(\tau^{}_{\bf d})\hat{\phi}^{[0]}_{y } + f^{y}_{\bf d}(\tau^{}_{\bf d})\hat{\pi}^{[0]}_{y } \right ] , \label{qexp}\\ \hat{\phi}^{}_{x^1}(x^0 ) & = & \sum_{{\bf d } ' } \left[\phi^{{\bf d}'}_{x^1}(x^0)\hat{q}^{[0]}_{{\bf d } ' } + f^{{\bf d}'}_{x^1}(x^0)\hat{p}^{[0]}_{{\bf d}'}\right ] \nonumber\\ & & + \int dy \left [ \phi^{y}_{x^1}(x^0)\hat{\phi}^{[0]}_{y } + f^{y}_{x^1}(x^0)\hat{\pi}^{[0]}_{y } \right ] , \label{phiexp}\end{aligned}\ ] ] from which @xmath75 and @xmath76 can be derived according to @xmath77 .
here @xmath78 }
\equiv \hat{\cal o}_{\mu}(t_n)$ ] and all the mode functions " @xmath79 and @xmath80 are real functions of time , which can be related to those in @xmath81-space in @xcite ( with different initial conditions , though . ) inserting the above expansions into @xmath82 , one has @xmath83 which gives @xmath84 where , for proper initial conditions , the homogeneous solutions are @xmath85 , \label{zerophi}\end{aligned}\ ] ] with @xmath86 , while the inhomogeneous solutions read @xmath87 with the retarded green s function for the massless scalar field in ( 1 + 1)d , @xmath88 $ ] @xmath89/2 $ ] in the @xmath12-@xmath29 frame .
now @xmath90 can be interpreted as vacuum fluctuations of the field propagating from @xmath91 to @xmath92 , while @xmath93 behave like retarded relativistic fields sourced by the pointlike detectors @xmath94 .
inserting the solutions of @xmath95 into @xmath96 , one obtains @xmath97 with @xmath98 . for the cases with a single detector ,
the right hand side of @xmath99 for @xmath100 is simply zero . from the equations of motion @xmath99
one learns that @xmath100 behave like damped hos , while @xmath101 behave like damped hos driven by vacuum fluctuations of the field @xmath102 at the position of the detector .
both are living in the pointlike detectors and not extended in space .
both would also be driven by retarded mutual influences from the other detectors , if the last term of @xmath99 is non - vanishing . in figure [ phidiag ]
these mode functions are represented in diagrams . }
$ ] , @xmath103}$ ] , and @xmath104}$ ] , which generate the retarded fields @xmath105}$ ] , @xmath106}$ ] , and @xmath107}$ ] through @xmath108 and @xmath109 represented in the diagrams from left to right in the lower row , respectively . here
the gray horizontal lines denote the @xmath12-slices at @xmath110 , @xmath111 , and @xmath112 , the dashed wavy lines are @xmath11-slices with @xmath113 ( here we choose @xmath28 in @xmath10 ) , and the dotted vertical lines are the worldlines of detector @xmath0 at @xmath114 .
all the above mode functions are independent of the data on the @xmath18- or @xmath19-slices except those right at the position of the pointlike detector @xmath115 where the local measurement was done .
the long - dashed lines in the two plots for @xmath104}$ ] ( upper right ) and @xmath116}$ ] ( lower right ) denote the vacuum fluctuations started from @xmath117 at @xmath118 ( cf .
@xmath119).,title="fig:",width=149 ] } $ ] , @xmath103}$ ] , and @xmath104}$ ] , which generate the retarded fields @xmath105}$ ] , @xmath106}$ ] , and @xmath107}$ ] through @xmath108 and @xmath109 represented in the diagrams from left to right in the lower row , respectively . here
the gray horizontal lines denote the @xmath12-slices at @xmath110 , @xmath111 , and @xmath112 , the dashed wavy lines are @xmath11-slices with @xmath113 ( here we choose @xmath28 in @xmath10 ) , and the dotted vertical lines are the worldlines of detector @xmath0 at @xmath114 .
all the above mode functions are independent of the data on the @xmath18- or @xmath19-slices except those right at the position of the pointlike detector @xmath115 where the local measurement was done .
the long - dashed lines in the two plots for @xmath104}$ ] ( upper right ) and @xmath116}$ ] ( lower right ) denote the vacuum fluctuations started from @xmath117 at @xmath118 ( cf .
@xmath119).,title="fig:",width=149 ] } $ ] , @xmath103}$ ] , and @xmath104}$ ] , which generate the retarded fields @xmath105}$ ] , @xmath106}$ ] , and @xmath107}$ ] through @xmath108 and @xmath109 represented in the diagrams from left to right in the lower row , respectively . here
the gray horizontal lines denote the @xmath12-slices at @xmath110 , @xmath111 , and @xmath112 , the dashed wavy lines are @xmath11-slices with @xmath113 ( here we choose @xmath28 in @xmath10 ) , and the dotted vertical lines are the worldlines of detector @xmath0 at @xmath114 .
all the above mode functions are independent of the data on the @xmath18- or @xmath19-slices except those right at the position of the pointlike detector @xmath115 where the local measurement was done .
the long - dashed lines in the two plots for @xmath104}$ ] ( upper right ) and @xmath116}$ ] ( lower right ) denote the vacuum fluctuations started from @xmath117 at @xmath118 ( cf .
@xmath119).,title="fig:",width=149 ] + } $ ] , @xmath103}$ ] , and @xmath104}$ ] , which generate the retarded fields @xmath105}$ ] , @xmath106}$ ] , and @xmath107}$ ] through @xmath108 and @xmath109 represented in the diagrams from left to right in the lower row , respectively .
here the gray horizontal lines denote the @xmath12-slices at @xmath110 , @xmath111 , and @xmath112 , the dashed wavy lines are @xmath11-slices with @xmath113 ( here we choose @xmath28 in @xmath10 ) , and the dotted vertical lines are the worldlines of detector @xmath0 at @xmath114 .
all the above mode functions are independent of the data on the @xmath18- or @xmath19-slices except those right at the position of the pointlike detector @xmath115 where the local measurement was done .
the long - dashed lines in the two plots for @xmath104}$ ] ( upper right ) and @xmath116}$ ] ( lower right ) denote the vacuum fluctuations started from @xmath117 at @xmath118 ( cf .
@xmath119).,title="fig:",width=149 ] } $ ] , @xmath103}$ ] , and @xmath104}$ ] , which generate the retarded fields @xmath105}$ ] , @xmath106}$ ] , and @xmath107}$ ] through @xmath108 and @xmath109 represented in the diagrams from left to right in the lower row , respectively . here
the gray horizontal lines denote the @xmath12-slices at @xmath110 , @xmath111 , and @xmath112 , the dashed wavy lines are @xmath11-slices with @xmath113 ( here we choose @xmath28 in @xmath10 ) , and the dotted vertical lines are the worldlines of detector @xmath0 at @xmath114 .
all the above mode functions are independent of the data on the @xmath18- or @xmath19-slices except those right at the position of the pointlike detector @xmath115 where the local measurement was done .
the long - dashed lines in the two plots for @xmath104}$ ] ( upper right ) and @xmath116}$ ] ( lower right ) denote the vacuum fluctuations started from @xmath117 at @xmath118 ( cf .
@xmath119).,title="fig:",width=149 ] } $ ] , @xmath103}$ ] , and @xmath104}$ ] , which generate the retarded fields @xmath105}$ ] , @xmath106}$ ] , and @xmath107}$ ] through @xmath108 and @xmath109 represented in the diagrams from left to right in the lower row , respectively . here
the gray horizontal lines denote the @xmath12-slices at @xmath110 , @xmath111 , and @xmath112 , the dashed wavy lines are @xmath11-slices with @xmath113 ( here we choose @xmath28 in @xmath10 ) , and the dotted vertical lines are the worldlines of detector @xmath0 at @xmath114 .
all the above mode functions are independent of the data on the @xmath18- or @xmath19-slices except those right at the position of the pointlike detector @xmath115 where the local measurement was done .
the long - dashed lines in the two plots for @xmath104}$ ] ( upper right ) and @xmath116}$ ] ( lower right ) denote the vacuum fluctuations started from @xmath117 at @xmath118 ( cf .
suppose an rsg detector ( detector @xmath0 ) is at rest in minkowski space with the worldline @xmath120 in the @xmath12-@xmath29 frame and @xmath121 in the @xmath11-@xmath30 frame , and the detector - field system is initially in a gaussian state at @xmath110 .
suppose a gaussian measurement is done on detector @xmath0 at @xmath122 when the quantum state on the @xmath18-slice collapses to @xmath123 for the observer in the @xmath12-@xmath29 frame , such that @xmath124 \label{rhoa}\ ] ] is a gaussian state of detector @xmath0 with @xmath125}_a)^2\rangle_1 = \hbar g_a /2 $ ] , @xmath126}_a)^2\rangle_1 = \hbar/2g_a$ ] , and @xmath127},\hat{p}_a^{[1]}\rangle_1=0 $ ] with some constant @xmath128 .
here @xmath129 denotes the expectation values taken from the quantum state right after @xmath130 . from @xmath131
one obtains @xmath132 \nonumber\\ & = & \exp -{1\over 2\hbar^2 } \left [ k^{\bar{\mu } } \tilde{\cal q}_{\bar{\mu}\bar{\nu } } k^{\bar{\nu } } - 2\delta^{\bar{\mu } } \tilde{\cal r}_{{\bar{\mu}}{\bar{\nu } } } k^{\bar{\nu } } + \delta^{\bar{\mu } } \tilde{\cal p}_{{\bar{\mu}}{\bar{\nu } } } \delta^{\bar{\nu } } \right],\end{aligned}\ ] ] for the rest of the system . here
@xmath133 is the normalization factor , @xmath134 , @xmath135 } , \hat{\phi}_{\bar{\nu}}^{[1 ] } \rangle_{1}$ ] , @xmath136 } , \hat{\pi}_{\bar{\nu}}^{[1 ] } \rangle_{1}$ ] , and @xmath137},$ ] @xmath138 } \rangle_{1}$ ] , where @xmath139 } \hat{\theta}_{\bar{\nu}}^{[1 ] } \rangle_{1 } = \langle \hat{\theta}_{\bar{\mu}\ ; , } ^{[10 ] } \hat{\theta}_{\bar{\nu}}^{[10 ] } \rangle_{0 } + % % % { i^{[1,0]}_a ( \hat{\theta}_{\bar{\mu}\ ; , } ^{[10 ] } \hat{\theta}_{\bar{\nu}}^{[10 ] } ) \over j_a^{[1,0]}}. \label{cort1t0}\ ] ] with @xmath140 or @xmath141 , @xmath142}_{\mu}\equiv { \cal o}^{}_\mu(t_n - t_s)$ ] , and @xmath143}_{\bf d } \left(\hat{\cal o},\hat{\cal o}'\right ) \equiv \nonumber\\ & & -\langle \hat{\cal o } _ , \hat{q}_{{\bf d}}^{[ns ] } \rangle_{s } \langle\hat{\cal o ' } _ , \hat{q}_{{\bf d}}^{[ns ] } \rangle_{s}\ell^{[1,0]}_{\bf d } -\langle\hat{\cal o } _ , \hat{p}_{{\bf d}}^{[ns ] } \rangle_{s } \langle \hat{\cal o ' } _ , \hat{p}_{{\bf d}}^{[ns ] } \rangle_{s}\bar{\ell}^{[1,0]}_{\bf d } \nonumber\\ & & + \langle \hat{q}_{{\bf d}\ ; , } ^{[ns ] } \hat{p}_{{\bf d}}^{[ns ] } \rangle_{s } % \times \nonumber\\ & & \hspace{1 cm } \left [ \langle \hat{\cal o } _ , \hat{q}_{{\bf d}}^{[ns ] } \rangle_{s } \langle \hat{\cal o ' } _ , \hat{p}_{{\bf d}}^{[ns ] } \rangle_{s } + ( \hat{\cal o}\leftrightarrow \hat{\cal o } ' ) \right],\end{aligned}\ ] ] with @xmath144}_{\bf d } \equiv \langle ( \hat{q}_{\bf d}^{[1]})^2\rangle_1 + \langle ( \hat{p}^{[10]}_{\bf d})^2\rangle_0 $ ] , @xmath145}_{\bf d } \equiv \langle(\hat{p}_{\bf d}^{[1]})^2 \rangle_1+$ ] @xmath146}_{\bf d})^2\rangle_0 $ ] , and @xmath147 } \equiv \ell^{[n , s]}_{\bf d } \bar{\ell}^{[n , s]}_{\bf d } -\langle\hat{q}^{[ns]}_{\bf d\ ; , } \hat{p}^{[ns]}_{\bf d}\rangle_s^2 $ ] . for the observer in the @xmath11-@xmath30 frame , according to the postulate of the projective measurement in that frame , we assume that a similar projection occurs but the wavefunctional collapses on the @xmath19-slice instead .
so the pms and the factors therein have the same form as the above ones in the @xmath12-@xmath29 frame except that the correlators are evaluated in the @xmath11-@xmath30 frame .
then , starting at @xmath18 and @xmath19 , both the pms in the @xmath12-@xmath29 frame and the pms in the @xmath11-@xmath30 frame evolve to @xmath148 , when @xmath12 and @xmath11-slices overlap and two observers can make a comparison on these two quantum states . in the conventional @xmath149 coordinates of minkowski space , the two - point correlators at @xmath150 determining the wavefunctional can be expressed as combinations of the mode functions evolving from @xmath18 to @xmath150 , together with the initial data on the @xmath18-slice in the form of the correlators of the field at space points on the slice , e.g. from @xmath151 and @xmath152 , @xmath153 } , \hat{\phi}_y^{[2 ] } \rangle_{2 } = { \rm tr } \hat{\phi}_x^{[21 ] } \hat{\phi}_y^{[21 ] } \tilde{\rho } \nonumber\\ & \sim & \int dx ' dy ' \phi_x^{x'[21]}\phi_y^{y'[21 ] } \langle \hat{\phi}_{x'}^{[1 ] } , \hat{\phi}_{y'}^{[1 ] } \rangle_{1 } + \cdots , \label{qxyt1}\end{aligned}\ ] ] where @xmath154 and @xmath155 are points on the @xmath18-slice .
apparently @xmath156 } , \hat{\phi}_{y'}^{[1 ] } \rangle_{1}$ ] depends on the data on the @xmath18-slice , and as mentioned below @xmath108 , @xmath157}$ ] is a superposition of vacuum fluctuations @xmath158}$ ] propagating from the point @xmath159 on the @xmath18-slice to @xmath160 and the retarded field @xmath161}$ ] sourced by the pointlike detector driven by vacuum fluctuations from the @xmath18-slice . on the other hand , in the alternative
coordinates @xmath162 , the form of the correlator is similar to @xmath163 , except that the dependence is on the @xmath19-slice .
so here the two - point correlators , or equivalently , the wavefunctionals at @xmath164 , appear to depend on the time - slicing scheme .
nevertheless , by comparing the expansions @xmath151 and @xmath152 of two equivalent evolutions without considering any measurement : one from @xmath26 to @xmath18 then from @xmath18 to @xmath150 , the other from @xmath26 all the way to @xmath150 , one can see that the mode functions have the following identities , @xmath165}_\nu & = & % \sum_{{\bf d}'}\left[\phi_\nu^{{\bf d}'[21]}\phi_{{\bf d}'}^{\mu[10 ] } + % f_\nu^{{\bf d}'[21 ] } \pi^{\mu[10]}_{{\bf d}'}\right ] \nonumber\\ & & % + \int dx ' \left [ \phi^{x'[21]}_\nu \phi^{\mu[10]}_{x ' } + f^{x'[21]}_\nu \pi^{\mu[10]}_{x'}\right ] , \phi_\nu^{\sigma [ 21]}\phi_{\sigma}^{\mu[10 ] } + f_\nu^{\sigma [ 21 ] } \pi^{\mu[10]}_{\sigma},\label{id1}\\ f^{\mu[20]}_\nu & = & % \sum_{{\bf d}'}\left[\phi_\nu^{{\bf d}'[21]}f_{{\bf d}'}^{\mu[10 ] } + % f_\nu^{{\bf d}'[21 ] } p^{\mu[10]}_{{\bf d}'}\right ] \nonumber\\ & & % + \int dx ' \left [ \phi^{x'[21]}_\nu f^{\mu[10]}_{x ' } + f^{x'[21]}_\nu p^{\mu[10]}_{x'}\right ] , \phi_\nu^{\sigma [ 21]}f_{\sigma}^{\mu[10 ] } + f_\nu^{\sigma [ 21 ] } p^{\mu[10]}_{\sigma } \label{id2 } \end{aligned}\ ] ] where dewitt notation is understood , @xmath166 , @xmath167 , @xmath168 and @xmath169 with @xmath170 according to @xmath77 .
similar identities for @xmath171 and @xmath172 can be derived straightforwardly from @xmath173 and @xmath174 .
such identities can be interpreted as the huygens principle of the mode functions , and can be verified by inserting particular solutions of the mode functions into the identities .
the dependence of the @xmath18-slice in @xmath163 turns out to be removable by expressing the @xmath175 correlators as @xmath176 with @xmath151 and @xmath152 inserted , then using @xmath173 , @xmath174 and similar identities for @xmath171 and @xmath172 to replace the @xmath177 and @xmath178 terms to reach @xmath179 } \phi_y^{a[21]}+{\hbar\over 2 g^{}_a } f_x^{a[21 ] } f_y^{a[21 ] } \nonumber\\ & & + \langle\hat{\upsilon}_x^{[0 ] } , \hat{\upsilon}_y^{[0]}\rangle_0 + { 1\over j_a^{[1,0 ] } } i_a^{[1,0 ] } \left(\hat{\upsilon}_x^{[0 ] } , \hat{\upsilon}_y^{[0 ] } \right)\end{aligned}\ ] ] where @xmath180 } & \equiv & \hat{\phi}_{\mu}^{[0 ] } \left(\phi_x^{\mu[20]}-\phi_x^{a[21]}\phi_a^{\mu[10 ] } - f_x^{a[21]}\pi_a^{\mu[10]}\right ) + \nonumber\\ & & \hat{\pi}_{\mu}^{[0]}\left(f_x^{\mu[20]}-\phi_x^{a[21]}f_a^{\mu[10 ] } - f_x^{a[21]}p_a^{\mu[10]}\right ) .
\end{aligned}\ ] ] then all the two - point correlators , and therefore the @xmath181-representation of the quantum state on the @xmath150-slice , end up with functionals of @xmath182}$ ] , @xmath183}$ ] , @xmath184}$ ] , @xmath185}$ ] , @xmath186}$ ] with @xmath187 , and the initial data in the two - point correlators evaluated at @xmath26 .
the quantum state at @xmath188 in the @xmath11-@xmath30 frame has exactly the same functional form .
now @xmath184}$ ] , @xmath185}$ ] , and @xmath186}$ ] are hos in the pointlike detector , @xmath183}$ ] and @xmath189}$ ] are retarded fields sourced from the detector , while @xmath190}$ ] are superpositions of vacuum fluctuations and retarded fields , as indicated from @xmath108 to @xmath99 .
all of them are explicitly covariant and independent of the data on the @xmath18- or @xmath19-slice outside the detector ( see figure [ phidiag ] ) .
thus the quantum states collapsed in two different frames are identical up to a coordinate transformation when compared on the same time - slice at @xmath191 .
the case with two successive projective measurements on a detector at @xmath192 and @xmath150 , @xmath193 has also been calculated .
the two - point correlators at @xmath194 can similarly be written in functionals of the covariant functions @xmath195}$ ] , @xmath196}$ ] , @xmath197}$ ] , @xmath198}$ ] with @xmath199 , and the initial data at @xmath26 .
they are independent of the data on @xmath18- , @xmath19- , @xmath150- , or @xmath188-slice outside the detector , so the quantum state at @xmath194 is still independent of the time - slices on which the wavefunctional collapsed .
it is straightforward to generalize this result to the cases with many successive projective measurements on the detector .
now consider the case with two spatially separated detectors . for simplicity ,
@xmath22 is put at @xmath200 and @xmath23 at @xmath15 , so both are at rest in the @xmath12-@xmath29 and @xmath11-@xmath30 frames .
suppose a local measurement like @xmath201 is done on @xmath22 at @xmath18 and @xmath202 , and a similar local measurement on @xmath23 at @xmath203 but @xmath204 : the two measurement events are spacelike separated , so the time order of these two events can be altered . in this case , the two - point functions at @xmath205 can still be written in functionals independent of the data on the @xmath18- , @xmath19- , @xmath150- , or @xmath188-slice outside the detectors .
for example , in the @xmath12-@xmath29 frame , @xmath206 } , \hat{\phi}_y^{[3 ] } \rangle_{3 } \nonumber\\ & = & { \hbar\over 2}g^{}_b \phi_x^{b[32 ] } \phi_y^{b[32 ] } + { \hbar\over 2 g^{}_b } f_x^{b[32 ] } f_y^{b[32 ] } + { \hbar\over 2}g^{}_a \phi_x^{a[31 ] } \phi_y^{a[31]}+{\hbar\over 2 g^{}_a } f_x^{a[31 ] } f_y^{a[31 ] } \nonumber\\ & & + \langle\hat{\upsilon}_x^{[0 ] } , \hat{\upsilon}_y^{[0]}\rangle_0 + { 1\over j_a^{[1,0 ] } } i_a^{[1,0 ] } \left(\hat{\upsilon}_x^{[0 ] } , \hat{\upsilon}_y^{[0 ] } \right ) + { 1\over j_b^{[2,1 ] } } i_b^{[2,1 ] } \left(\hat{\upsilon}_x^{[1 ] } , \hat{\upsilon}_y^{[1 ] } \right),\end{aligned}\ ] ] with @xmath207 } \equiv \hat{\phi}_{\mu}^{[n]}\tilde{\phi}_x^{\mu[3n ] } + \hat{\pi}_{\mu}^{[n]}\tilde{f}_x^{\mu[3n]}$ ] and @xmath208 } \equiv \phi_x^{\mu[30]}-\phi_x^{a[31]}\phi_a^{\mu[10 ] } - f_x^{a[31]}\pi_a^{\mu[10]}-\phi_x^{b[32]}\phi_b^{\mu[20]}- f_x^{b[32]}\pi_b^{\mu[20 ] } ,
\\ \tilde{f}_x^{\mu[30 ] } \equiv f_x^{\mu[30]}-\phi_x^{a[31]}f_a^{\mu[10 ] } - f_x^{a[31]}p_a^{\mu[10 ] } -\phi_x^{b[32]}f_b^{\mu[20 ] } - f_x^{b[32]}p_b^{\mu[20]},\\ % \label{up1 } \tilde{\phi}_x^{\mu[31 ] } \equiv \phi_x^{\mu[31]}-\phi_x^{b[32]}\phi_b^{\mu[21]}- f_x^{b[32]}\pi_b^{\mu[21 ] } , \\ \tilde{f}_x^{\mu[31 ] } \equiv f_x^{\mu[31]}-\phi_x^{b[32]}f_b^{\mu[21 ] } - f_x^{b[32]}p_b^{\mu[21]}. % \label{up2}\end{aligned}\ ] ] on expressing all the correlators @xmath209 in @xmath210}$ ] and @xmath211}$ ] in terms of @xmath212 with the help of @xmath176 , both @xmath210}$ ] and @xmath211}(\hat{\upsilon}_x^{[1 ] } , \hat{\upsilon}_y^{[1]})$ ] will become functionals of @xmath213}$ ] , @xmath214}$ ] , and the correlators evaluated on the initial time - slice .
so the final expression for the wavefunctional at @xmath194 depends only on the covariant mode functions corresponding to the hos in the detectors , the retarded fields sourced by detectors a and b , vacuum fluctuations , as well as the initial data on the @xmath26-slice .
further , using computer algebra it is straightforward to verify that whenever the retarded mutual influence @xmath215}$ ] is zero , @xmath216 } ( \hat{\upsilon}_{\xi}^{[2 ] } , \hat{\upsilon}_{\xi'}^{[2]})/j_a^{[1,2]}+ i_b^{[2,0 ] } ( \hat{\upsilon}_{\xi}^{[0 ] } , \hat{\upsilon}_{\xi'}^{[0]})/j_b^{[2,0]}$ ] obtained in the @xmath11-@xmath30 frame has the same functional form of the mode functions and correlators as @xmath217}(\hat{\upsilon}_x^{[0 ] } , \hat{\upsilon}_y^{[0]})/j_a^{[1,0 ] } + i_b^{[2,1 ] } ( \hat{\upsilon}_x^{[1 ] } , \hat{\upsilon}_y^{[1]})/j_b^{[2,1]}$ ] in the @xmath12-@xmath29 frame .
thus the time order of the spacelike separated measurement events does not matter .
the wavefunctionals collapsed in two different frames are identical at @xmath205 up to a coordinate transformation .
if the two measurement events are timelike separated , however , different observers in different frames will recognize the same time order of the two events .
in this case the retarded mutual influences will come into play , so the resulting pms will be different from those in the reversed time order , while the consistency of quantum states at the final time - slice is obvious when one knows the results for the previous cases .
in all of the above simple cases , we found that the quantum states of the rsg detector - field system started with the same initial state defined on the same fiducial time - slice , then collapsed on different time - slices in different reference frames by one or two measurements on the pointlike detectors , perhaps in different time - orders provided the measurement events are spacelike separated , will evolve to the same state of the combined system up to a coordinate transformation when compared on the same final time - slice .
our analysis is actually independent of any specific choice of coordinates for the spacetime between the initial and final time - slices or any specific property of the field theories in ( 1 + 1)d such as conformal symmetry , and the identities interpreted as the huygens principle for the mode functions such as @xmath173 are consequences of the linearity of quantum theory and independent of the details of the initial state . thus our calculations for the gaussian states of the combined system with one or two detectors collapsed by one or two spatially local measurements in ( 1 + 1)d minkowski space can be generalized straightforwardly to the cases with more detectors , more measurement events , and more general states , in higher dimensions .
while we took advantage of the linearity of our model to perform explicit calculations , the consistency observed here is actually more generic in rqft . in [ nlindft ]
a similar argument has been applied to the detector - field systems with nonlinear , spatially local interactions .
one can see that the multi - point correlators of the combined system evaluated on the final time - slice are still independent of the data on the time - slice outside the detector at the moment of the measurement on the detector .
so the quantum states in these nonlinear local detector - field theories started with the same initial state and then collapsed by the same spatially local projective measurements in different reference frames will be consistent with each other when compared on the same final time - slice .
the spatial locality of interactions ( which can be linear or nonlinear ) and the general covariance in a relativistic system , together with the linearity of quantum dynamics and the spatial locality of quantum measurements in its quantum theory , guarantee this consistency . + * acknowledgment * the author thanks bei - lok hu for illuminating discussions and lajos disi for helpful comments .
this work was supported by the nsc taiwan under the grant 99 - 2112-m-018 - 001-my3 and in part by the national center for theoretical sciences , taiwan .
in the indirect measurement scheme for relativistic quantum systems @xcite , there is no projective measurement between the initial and final time - slices . instead
, a measurement event is modeled by linearly coupling , and so entangling , a quantum probe or device locally in space and time with the field to be observed .
each local coupling process is described by an interaction hamiltonian with a compact support localized in spacetime between the initial and final time - slices .
after all these couplings or interactions have been finished , a set of the projective measurements on the quantum probes will be performed on the final time - slice associated with some moment essentially in the asymptotic region .
in contrast , the rsg detectors and the field in our model are interacting continuously at all times , and the projective measurements on the detectors , if any , will be performed at some moments between the initial and final time - slices , which will always be in the interaction region . then we compare the quantum states of the combined system in different frames on the same final time - slice , still in the interaction region , without any further projective measurement there
. in @xcite aharonov and albert concluded that there is no covariant description of measurement in terms of time evolution of quantum states when the whole time - slice which the quantum state is defined on is not in the asymptotic region where all interactions have been finished . in the simplest case that they considered with no wavefunction collapse between the initial and the final time - slices , they have found that quantum states have to be parameterized by many local times if the system consists of many local quantum objects .
disi pointed out that in this case it would be simpler to describe such processes using heisenberg operators , which are naturally covariant in a relativistic model @xcite . in our model @xmath218 and other linear detector theories ,
this can be understood as follows .
a quantum state in the schrdinger picture for our model is represented as a wavefunctional consisting of the dynamical variables ( @xmath37 , @xmath219 , @xmath220 , and @xmath221 , or equivalently , @xmath222 and @xmath223 in @xmath69 in the @xmath181-representation ) and the time - dependent factors .
the latter can be expressed as combinations of the correlators of the dynamical variables ( for gaussian states , only two - point correlators are involved ; see @xmath69 , and also see @xcite for the conventional wigner functionals . ) without any projective measurement , the form of these combinations of the correlators will remain unchanged from the initial all the way to the final time - slices , while the time dependence of the factors is due to the evolving correlators of the corresponding time - dependent , covariant operators with respect to the initial state of the combined system in the heisenberg picture . here the time evolution of those covariant operators ( or equivalently , the mode functions in section [ rsgdetector ] in our linear system ) are governed by the heisenberg equations of motion , which are also covariant in a relativistic system with a well defined hamiltonian .
the evolutions of the operators corresponding to the dynamical variables local in position space , such as @xmath37 and @xmath219 living inside the pointlike detectors , are naturally parameterized by their own proper times , like those in @xmath96 .
suppose the time evolutions of all the operators have been solved . in a given reference frame
started with some initial time - slice , the quantum state on a time - slice associated with some moment in this frame is obtained by inserting the operators for the dynamical variables on that time - slice into the correlators composing the time - dependent factors in the wavefunctional . for the dynamical variables local in position space ( @xmath37 , @xmath219 , @xmath224 , and @xmath225 ) those operators are the ones with their own proper times or coordinate times at the intersections of their world lines and the time - slice that the correlators are evaluated on .
different frames have different time - slicing schemes , so the histories of the wavefunctionals in different frames can be very different , though the history of each operator corresponding to a spatially local dynamical variable is uniquely determined by the covariant heisenberg equations of motion @xcite .
note that the above argument is not restricted to just the indirect measurement scheme , but is valid for all relativistic systems with well defined hamiltonians without any projective measurement between the initial and final time - slices .
the only thing special to the indirect measurement scheme is that all the couplings will eventually be turned off , then a set of the projective measurements will be performed on the quantum probes or devices on some final time - slice in the future asymptotic region .
it is obvious that , if the covariant interaction hamiltonians describing two individual local measurement events " ( entanglement processes , actually ) in the indirect measurement scheme commute , then the time order of these events being possibly different in different frames will not matter and the quantum states will be consistent when compared on _ the same _ final time - slice in different frames in the future asymptotic region @xcite .
this is guaranteed by the covariance of the operator histories and the form - invariance of the time - dependent factors in terms of the correlators in a quantum state never being projected .
if there is a projective measurement performed at some moment during the evolution of a detector - field system in the interaction region , however , the dependence on the correlators of the time - dependent factors in the wavefunctional of the whole detector - field system will be suddenly changed like @xmath176 , so the argument in this subsection will not be valid .
suppose each quantum device in the indirect measurement scheme is associated with a pointlike object in space so that the dynamics of the device can be parameterized in its proper time , which is invariant under coordinate transformations .
following @xcite , suppose the device - field system is started with the initial state @xmath226 defined on the initial time - slice in some reference frame , where @xmath227 is a gaussian state of the @xmath14-th device and @xmath228 is the initial state of the field .
right after all the local device - field interactions are finished , say , at @xmath18 in this frame , the device - field system evolves into an entangled state of the devices and the field , denoted by @xmath229 . while there is no device - field or device - device coupling after @xmath18 , the backreactions from the devices to the field will keep evolving in the field sector , and so do the quantum state of the combined system as well as the correlations between the devices and the field .
after some time a set of the local projective measurements on all the devices are performed simultaneously on some final time - slice associated with , say , @xmath150 in the same reference frame ( @xmath203 ) with the outcomes @xmath230 , then the post - measurement state of the combined system becomes @xmath231 with the normalization factor @xmath232 and the un - normalized final state of the field @xmath233 defined on the time - slice associated with @xmath150 .
the joint probability of the set of the outcomes @xmath234 is proportional to the squared norm of @xmath235 , namely , @xmath236 where the reduced state of the devices is given by @xmath237 thus the joint probability @xmath238 taken at @xmath150 corresponds to the reduced state of the devices right before the projective measurements , @xmath239 , where the field operators defined on the time - slice associated with @xmath150 are traced out ( or coarse - grained ) . by measuring the devices in an ensemble of many copies of the state @xmath240
, one could reconstruct the reduced state @xmath239 of the device , rather than the quantum state of the field or the combined system , using the outcomes of the measurements . here
the choice of the final time - slice actually specifies a class of reference frames with all kinds of the final time - slices intersecting the same set of the projective measurement events localized in spacetime .
the reduced states of the devices @xmath241 right before the projective measurements in all the reference frames in this particular class are the same . similarly , since the harmonic oscillators in the pointlike rsg detectors in our model are scalars parameterized in their proper times , the outcomes of the projective measurements on the detectors are automatically invariant under coordinate transformations .
if the projective measurement events on the detectors are spacelike distributed on a time - slice in some reference frame , the outcomes of the measurements on an ensemble of many copies of the quantum state of the combined system will correspond to the joint probability as well as the reduced state of the detectors defined on that time - slice .
this reduced state of the detectors is identical to all those observed in a class of reference frames in which each frame has a time - slice simultaneously intersecting all the projective measurement events . indeed ,
in our model , the gaussian reduced state of the detectors at some moment @xmath242 in some reference frame without any projective measurement between @xmath110 and @xmath48 reads @xmath243 = \exp \left\ { -{1\over 2\hbar^2 } \times\right .
\nonumber\\ & & \left .
\sum_{\bf d , d'}\left [ k^{\bf d } { \cal q}_{\bf d d'}(x^0 ) k^{\bf d ' } -2 \delta^{\bf d } { \cal r}_{\bf d d'}(x^0 ) k^{\bf d ' } + \delta^{\bf d } { \cal p}_{\bf d d'}(x^0 ) \delta^{\bf d ' } \right]\right\},\end{aligned}\ ] ] where @xmath38 and @xmath244 run over all the detectors , and the factors @xmath245 , @xmath246 , and @xmath247 are the symmetric two - point correlators of the detectors tr@xmath248 @xmath249 $ ] with @xmath250 ( cf .
( [ qstate ] ) ) . from @xmath151 , @xmath119 and
@xmath99 one can see that only the operators in the past lightcones of the measurement events are relevant in @xmath251 $ ] , so the reduced state of the detectors is independent of the choice of the time - slice outside the measurement events .
in contrast to the detector sector , the field sector of the collapsed or projected state of the combined system ( e.g. the @xmath235 above ) defined on different time - slices passing through the same set of events of the projective measurements on the devices or the detectors can appear very differently but can not be compared directly .
nevertheless , our result in this paper implies that if these collapsed states can be compared later on some time - slice in common in those different frames , they will be identical to each other up to a coordinate transformation .
note that in the position space representation of the field , each field amplitude defined at a space point corresponds to one degree of freedom , and there are infinitely many degrees of freedom for the field , as many as the number of the space points in the universe .
it is impossible to obtain the full information of the field through finitely many measurement events on a finite number of the devices or the detectors .
so our result in this paper on the consistency of the quantum state of the whole detector - field system is certainly stronger then those on the reduced state of the detectors .
a similar argument can be applied to the nonlinear detector - field theories where all the detector - field and field - field couplings are spatially local . here
we illustrate the single - detector case with one measurement on the detector .
the more general cases are straightforward .
let @xmath252 be the covariant operators in a nonlinear relativistic local detector - field theory .
given the hamiltonian @xmath253 defined at the initial moment @xmath254 in the minkowski frame , the time evolution of the operator @xmath255 from @xmath26 to some @xmath256 can be expressed as @xmath257 } % = \exp\left({i\over \hbar}\int_{t_0}^{t_1}dt\hat{h}(t)\right)\hat{\theta}_\mu^{[0 ] } % \exp\left(-{i\over \hbar}\int_{t_0}^{t_1}dt\hat{h}(t)\right ) \nonumber\\ % & & \hspace{-.4 cm } = \hat{\theta}_\mu^{[0 ] } + { i\over\hbar}(t_1-t_0)[\hat{h}^{[0 ] } , \hat{\theta}_\mu^{[0 ] } ] + { 1\over 2}\left({i\over\hbar}(t_1-t_0)\right)^2 [ \hat{h}^{[0 ] } , [ \hat{h}^{[0 ] } , \hat{\theta}_\mu^{[0 ] } ] ] % \nonumber\\ & & + \cdots , \label{psi10}\ ] ] or @xmath258 } \equiv \varphi_\mu^{\nu[10]}\hat{\theta}_\nu^{[0 ] } + \varphi_\mu^{\nu\sigma [ 10]}\ { \hat{\theta}_{\nu}^{[0 ] } , \hat{\theta}_{\sigma}^{[0 ] } \ } + \varphi_\mu^{\nu\sigma\rho [ 10]}\ { \hat{\theta}_{\nu}^{[0 ] } , \hat{\theta}_{\sigma}^{[0 ] } , \hat{\theta}_{\rho}^{[0]}\ } + \cdots . \label{varphimf}\ ] ] if the operators keep evolving from @xmath18 to some @xmath259 , then one also has @xmath260 } = \varphi_\mu^{\nu[21]}\hat{\theta}_\nu^{[10 ] } + \varphi_\mu^{\nu\sigma [ 21]}\ { \hat{\theta}_{\nu}^{[10 ] } , \hat{\theta}_{\sigma}^{[10 ] } \ } + \varphi_\mu^{\nu\sigma\rho [ 21]}\ { \hat{\theta}_{\nu}^{[10 ] } , \hat{\theta}_{\sigma}^{[10 ] } , \hat{\theta}_{\rho}^{[10]}\ } + \cdots , \label{psi20}\end{aligned}\ ] ] which is a consequence of the linearity of quantum theory . here
the brace bracket @xmath261 denotes the fully symmetrized ordering of the operators @xmath262 , the indices @xmath263 , and dewitt notation is understood .
the c - number factors @xmath264 are the counterparts of the mode functions in ( [ qexp ] ) and ( [ phiexp ] ) in the linear rsg detector theory . a general quantum state of the combined system at the moment @xmath12 in the @xmath57-representation can be written as @xmath265 where @xmath266 . at @xmath192 , a measurement on
detector @xmath0 projects the quantum state from @xmath267 to the pms @xmath268 where @xmath133 is the normalization factor , @xmath269 or @xmath270 , and @xmath271 .
starting with this new initial state , the combined system keeps evolving to the final time - slice associated with @xmath272 , where it can be compared with the quantum states evolving in other reference frames . at @xmath150 the quantum state
can again be written in the general form ( [ gqs ] ) , where the factors @xmath273}_{\mu_1,\cdots,\mu_n}\equiv { \cal s}_{\mu_1,\cdots,\mu_n}(t_2)$ ] in @xmath274 are functions of the @xmath33- , @xmath275- , @xmath276 , @xmath14-point symmetric correlators for the pms right after @xmath18 such as @xmath277 } , \hat{\theta}_{\mu^{}_2}^{[21]},\cdots,\hat{\theta}_{\mu^{}_n}^{[21 ] } \}\rangle^{}_1 \nonumber\\ & = & \left < \left\ { \left[\varphi_{\mu^{}_1}^{\nu[21]}\hat{\theta}_{\nu}^{[1 ] } + \varphi_{\mu^{}_1}^{\nu\sigma [ 21 ] } \{\hat{\theta}_{\nu}^{[1 ] } , \hat{\theta}_{\sigma}^{[1]}\}+ \cdots\right],\cdots \right.\right.\nonumber\\ & & \,\,\ , \left.\left .
\cdots , \left[\varphi_{\mu^{}_n}^{\nu'[21]}\hat{\theta}_{\nu'}^{[1 ] } + \varphi_{\mu^{}_n}^{\nu'\sigma ' [ 21 ] } \{\hat{\theta}_{\nu'}^{[1 ] } , \hat{\theta}_{\sigma'}^{[1]}\ } + \cdots\right]\right\}\right>^{}_1 \nonumber\\ & = & \left\ { \left(\varphi_{\mu^{}_1}^{\nu[21]}{\delta\over\delta{\cal k}^\nu } + \varphi_{\mu^{}_1}^{\nu\sigma [ 21 ] } { \delta\over\delta{\cal k}^\nu}{\delta\over\delta{\cal k}^\sigma } + \cdots\right)\cdots \right.\nonumber\\ & & \,\,\ , \left.\cdots \left(\varphi_{\mu^{}_n}^{\nu'[21]}{\delta\over\delta{\cal k}^{\nu ' } } + \varphi_{\mu^{}_n}^{\nu'\sigma ' [ 21]}{\delta\over\delta{\cal k}^{\nu'}}{\delta\over\delta{\cal k}^{\sigma ' } } + \cdots\right)\tilde{\rho}({\cal k } ) \right\}_{{\cal k}={\bf 0}}\nonumber\\ & = & n\int{d\tilde{\cal k}^a d\tilde{\cal k}^{a^ * } \over 2\pi\hbar } % \left .
\rho_a^*(\tilde{\cal k}^a)\cdot \rho({\cal k}^x=0,\tilde{\cal k}^a ; t_1 ) % \right|^{{\cal k}^x_{}={\cal k}^{x^*}_{}=0}%_{({\cal k}^a , { \cal k}^{a^*})\to ( \tilde{\cal k}^a_{},\tilde{\cal k}^{a^ * } _ { } ) } \left\ { \left[\langle \hat{\upsilon}^{}_{\mu^{}_1:\mu^{}_n}\rangle^{}_0 + \right.\right.\nonumber\\ & & \tilde{\cal k}^a \left(\langle \hat{\upsilon}^{}_{\mu^{}_1:\mu^{}_n } , \hat{\theta}_a^{[10 ] } \rangle^{}_0 -\langle \hat{\upsilon}^{}_{\mu^{}_1:\mu^{}_n}\rangle^{}_0 \langle\hat{\theta}_a^{[10]}\rangle^{}_0\right)+\nonumber\\ & & \tilde{\cal k}^a \tilde{\cal k}^b \left ( \langle \hat{\upsilon}^{}_{\mu^{}_1:\mu^{}_n } , \hat{\theta}_a^{[10 ] } , \hat{\theta}_b^{[10 ] } \rangle^{}_0 -\langle \hat{\upsilon}^{}_{\mu^{}_1:\mu^{}_n}\rangle^{}_0 \langle\hat{\theta}_a^{[10 ] } , \hat{\theta}_b^{[10 ] } \rangle^{}_0 \right .
\nonumber\\ & & \hspace{.5 cm } \left . -2
\langle \hat{\upsilon}^{}_{\mu^{}_1:\mu^{}_n}\hat{\theta}_a^{[10]}\rangle^{}_0 \langle \hat{\theta}_b^{[10 ] } \rangle^{}_0 + 2 \langle \hat{\upsilon}^{}_{\mu^{}_1:\mu^{}_n}\rangle^{}_0 \langle \hat{\theta}_a^{[10]}\rangle^{}_0 \langle \hat{\theta}_b^{[10 ] } \rangle^{}_0 \right ) + \nonumber\\ & & \left.\left .
\cdots \right ] \rho^{}_a({\cal k}^a)\right\}_{{\cal k}^a=0 } \label{serrep}\end{aligned}\ ] ] with @xmath278 , where @xmath279}\tilde{\theta}_\nu + \varphi_\mu^{\nu\sigma [ 21]}\ { \tilde{\theta}_{\nu } , \tilde{\theta}_{\sigma}\ } + \varphi_\mu^{\nu\sigma\rho [ 21]}\ { \tilde{\theta}_{\nu } , \tilde{\theta}_{\sigma } , \tilde{\theta}_{\rho } \ } + \cdots , \label{up}\ ] ] which looks similar to @xmath280}$ ] in ( [ psi20 ] ) but now @xmath281}$ ] and @xmath282 . from ( [ psi20 ] ) and ( [ up ] ) , one has @xmath283 } = \varphi_\mu^{a [ 21 ] } \left ( { \delta\over\delta{\cal k}^a}-\hat{\theta}_a^{[10]}\right ) - 2 \varphi_\mu^{ax [ 21]}\ { \hat{\theta}_{a}^{[10 ] } , \hat{\theta}_{x}^{[10]}\ } \nonumber\\ & & \,\,\,\,+ \varphi_\mu^{ab [ 21 ] } \left({\delta\over\delta{\cal k}^a}{\delta\over\delta{\cal k}^b } + 2 \hat{\theta}_a^{[10 ] } { \delta\over\delta{\cal k}^b}-\ { \hat{\theta}_{a}^{[10 ] } , \hat{\theta}_{b}^{[10]}\}\right ) % \nonumber\\ & & + \cdots . \label{updef}\end{aligned}\ ] ] comparing ( [ psi10 ] ) with ( [ varphimf ] ) , one can see that the factors @xmath284}$ ] with @xmath285 vanish whenever @xmath286 or @xmath287 , @xmath288 , because the detector - field coupling is spatially local .
this implies that no operator defined on the @xmath18-slice outside of detector @xmath0 is present on the right hand side of ( [ updef ] ) . since @xmath280}$ ] is independent of the @xmath18-slice , @xmath289 is independent of the operators on the @xmath18-slice outside of detector @xmath0
. thus @xmath290},\hat{\theta}_{\mu^{}_2}^{[21]},\cdots , \hat{\theta}_{\mu^{}_n}^{[21]}\ } \rangle^{}_1 $ ] are independent of the data on the @xmath18-slice outside of detector @xmath0 , and so is the quantum state @xmath291 of the combined system .
lin , b. l. hu , accelerated detector - quantum field correlations : from vacuum fluctuations to radiation flux , phys .
d 73 ( 2006 ) 124018 . s
lin , b. l. hu , backreaction and unruh effect : new insights from exact solutions of uniformly accelerated detectors , phys .
d 76 ( 2007 ) 064008
. w. g. unruh , notes on black - hole evaporation , phys .
d 14 ( 1976 ) 870 .
|
suppose the postulate of measurement in quantum mechanics can be extended to quantum field theory , then a local projective measurement at some moment on an object locally coupled with a relativistic quantum field will result in a projection or collapse of the wavefunctional of the combined system defined on the whole time - slice associated with the very moment of the measurement , if the relevant degrees of freedom have nonzero correlations .
this implies that the wavefunctionals in the same hamiltonian system but defined in different reference frames would collapse on different time - slices passing through the same local event where the measurement was done .
are these post - measurement states consistent with each other ?
we illustrate that the quantum states of the raine - sciama - grove detector - field system started with the same initial gaussian state defined on the same initial time - slice , then collapsed by the measurements on the pointlike detectors on different time - slices in different frames , will evolve to the same state of the combined system up to a coordinate transformation when compared on the same final time - slice .
such consistency is guaranteed by the spatial locality of interactions and the general covariance in a relativistic system , together with the spatial locality of measurements and the linearity of quantum dynamics in its quantum theory . quantum field theory in curved spacetime , relativistic quantum information , nonequilibrium quantum field theory , quantum measurement
|
interleavers for turbo codes @xcite have been extensively investigated . however , the design of interleavers for turbo codes is complex enough and we believe there are still several relevant open questions .
recently , sun and takeshita @xcite suggested the algebraic approach of using permutation polynomial - based interleavers over integer rings . in this paper , we expand the theory in @xcite by adding a geometric and group - theoretic perspective .
several new results and important and interesting new questions arise with this framework .
we believe that the approach in @xcite has major advantages over any earlier interleaver constructions ( either pseudo - random or structured ) because it simultaneously provides : * excellent error performance with practical code - lengths @xcite . * completely algebraic structure with elegant and relevant properties @xcite . * efficient implementation @xcite with high - speed , low - power consumption , and little memory requirements .
an interleaver is a device that permutes a sequence of @xmath0 symbols .
let each symbol be indexed by an element in the set @xmath1 .
then the interleaver can be represented by a one - to - one onto ( permutation ) function @xmath2 given by @xmath3 .
permutation polynomials over @xmath4 are functions that belong to the ring @xmath5 $ ] , i.e. , polynomials of the form @xmath6 such that @xmath7 and @xmath8 is a permutation function . in earlier work
, we only treated _ constant - free _ pps , i.e. , @xmath9 ; in this paper , we also consider @xmath10 .
the set of permutation functions ( not necessarily permutation polynomials ) for a given interleaver length @xmath0 will be denoted by @xmath11 .
the set @xmath11 is no longer a ring but its elements form a group under function composition .
permutation polynomials over @xmath12 naturally generate interleavers where @xmath13 gets equipped with a finite ring structure @xmath12 , which is not necessary for the definition of an arbitrary interleaver but allows a useful algebraic treatment of pp interleavers .
replacing @xmath13 by @xmath12 in the case of arbitrary interleavers brings no loss of generality , therefore from here on we will always use @xmath12 .
the algebraic structure of pps over finite integer rings has already been proven very successful in producing turbo codes @xcite and low density parity check ( ldpc ) codes @xcite with excellent error performance compared with the best known constructions and similar parameters .
conditions for the coefficients of a polynomial over @xmath4 to be a pp for an arbitrary integer @xmath0 were studied in @xcite . however , the conditions are not the simplest for polynomials of degrees larger than two ; to the best of our knowledge , simple necessary and sufficient conditions for arbitrary degrees are only known when @xmath0 is a power of two @xcite .
if @xmath14 is of second degree and @xmath0 is arbitrary then a simple necessary and sufficient condition was proved in @xcite and summarized in @xcite .
interleavers constructed using second degree or quadratic permutation polynomials will be called qpp interleavers , linear interleavers if the pp is of first degree , and pp interleavers for pps of arbitrary degrees .
because the number of distinct permutations of length @xmath0 is @xmath15 , it is very important that effective interleaver measures for turbo codes are defined ; this considerably reduces the number of interleavers that still need to be filtered by costly analysis and computer simulations for a complex turbo codec system .
the main drawback of the theory in @xcite was that despite it provided some rules for choosing good permutation polynomials ( pp ) , the procedures were too complex when handling input weights larger than two . in this paper , we propose a new simple but effective measure @xmath16 for interleavers .
two typical measures for interleavers in turbo coding are the spread factor @xmath17 @xcite and `` randomness . '' in this paper , the notion of `` randomness '' is replaced by a more principled concept of a degree of non - linearity @xmath18 of an interleaver .
an interleaver is represented by what we call an _ interleaver - code _ , which is the geometric representation of an interleaver by pairs of coordinates @xmath19 forming _ points _ in @xmath20 .
the degree of non - linearity @xmath18 measures the number of disjoint orbits ( a set of points ) of the action of an isometry group of the interleaver - code .
the new measure is simply the product of the logarithm of the spread factor by the new non - linearity metric , i.e. , @xmath21 .
the algebraic - geometric nature of pps allows a very efficient selection of pps that maximizes the new metric .
drp interleavers @xcite are among the best known interleavers for turbo codes with a combined excellent error rate performance ( exceeding that of @xmath22random interleavers @xcite ) and simplicity but are not fully algebraic .
the efficacy of the new metric is shown by simulation curves of several turbo codes using pp interleavers with impressive frame error rate performance , similar to the ones with drp interleavers . however , the selection of good pp interleavers ( particularly qpp interleavers ) is much simpler with the results presented this paper .
the non - linearity metric @xmath18 is shown to be inversely related to the degree of shift - invariance @xmath23 of an interleaver . for qpp interleavers , the shift - invariance @xmath24 is computed in closed form as function of the second degree coefficient , which gives us a complete control of this parameter . when tail - biting codes are used as constituent codes ,
turbo codes using qpp interleavers become quasi - cyclic ; for those codes , it is predicted that the multiplicity of many low - weight codewords is typically multiples of the shift - invariance @xmath24 .
it is also shown that a lower bound on the spread factor @xmath17 constrained to points within an orbit of an interleaver - code has a closed form for qpp interleavers . an immediate tie between @xmath17 and @xmath18
is then established for qpp interleavers where one would need to trade for either larger @xmath17 or @xmath18 .
this paper is organized as follows . in section
[ sec : ms ] , maximum - spread interleavers are investigated and a list of good permutation polynomials for turbo codes for several lengths that maximizes the spread factor @xmath17 is found in table [ tab : maxqpp ] .
a new metric for interleavers is developed in section [ sec : qpp_nl_si ] ; we provide a table ( table [ tab : maxomega ] ) of good permutation polynomials for turbo codes according to the new metric .
numerical results are shown in section [ sec : nr ] via the computation of the distance spectra and computer simulations for several codes .
finally , conclusions and possible future directions are discussed in section [ sec : conclusions ] .
in this section , the spread factor of an interleaver is revisited .
an infinite sequence of qpp interleavers achieving the upper bound on the spread factor is presented .
further , a list of qpp interleavers with optimal spread factors for several interleaver lengths reported in the literature is provided . in the algebraic - geometric treatment in this paper
, it will be convenient to view an interleaver represented by a permutation function @xmath25 ( not necessarily a permutation polynomial ) as an interleaver - code @xmath26 ( not to be confused with a turbo code ) under the natural mapping @xmath27 given by @xmath28 , where @xmath29 .
we will call a pair @xmath19 as a _ point _
let a linear interleaver be given by the pp @xmath31 ; the corresponding interleaver - code @xmath32 is plotted in fig .
[ fig : ms512_0_31 ] . viewed as an interleaver - code @xmath32 over @xmath33 . ] the notion of _ distance _ or _
metric _ between points in an interleaver - code @xmath26 will be of central role .
we will mainly use two different metrics .
let us define a distance @xmath35 in @xmath33 to form a metric space @xmath36 , @xmath37 i.e. , the metric is the @xmath34 or manhattan metric .
let us define a distance @xmath38 in @xmath33 to form a metric space @xmath39 , @xmath40 where @xmath41 is the lee distance @xcite between @xmath42 and @xmath43 modulo @xmath0 .
therefore , @xmath38 is a two - dimensional lee metric .
the spread factor @xcite of an interleaver is a popular measure of merit in turbo coding applications .
the spread factor of an interleaver @xmath25 over the metric space @xmath44 is defined as @xmath45 the root of this measure is the early recognition in turbo coding that self - terminating information sequences of weight two create low - weight codewords @xcite .
divsalar and dolinar then proposed a construction of linear interleavers achieving spread factors @xmath46 equal to or close to @xmath47 . they have concluded that by using linear interleavers , the minimum distance of turbo codes conditioned to weight - two self - terminating information sequences grows roughly as @xmath47 .
however , recent results show that the true minimum distance grows asymptotically only at most logarithmically @xcite for all interleavers .
if we use the metric space @xmath48 then the following definition of spread factor is also possible @xcite : @xmath49 the spread factor @xmath17 is mathematically more convenient because the metric space @xmath48 is `` isotropic '' ( in the sense of having no boundaries ) .
it is also more suitable when designing turbo codes with tail - biting convolutional constituent codes @xcite .
an analytical proof on the upper bound @xmath50 of @xmath17 was recently shown to be @xmath47 @xcite .
clearly @xmath51 by the definition of @xmath35 .
we show @xmath52 is close to @xmath50 by computing a new upper bound @xmath52 .
it is a constructive bound for certain values of @xmath0 .
see appendix [ sec : bound ] for a sketch of a proof .
@xmath53 the difference @xmath54 goes to 1 as @xmath0 grows ( see fig .
[ fig : ubounds ] ) . .
] only one example was found ( for a limited search ) of an interleaver ( up to symmetries ) that has a spread factor @xmath46 exceeding @xmath55 . for @xmath56 , the interleaver @xmath57 defined by @xmath58 , @xmath59 , @xmath60 , and @xmath61
has @xmath62 our new bound ( [ eq : debound ] ) is not defined at @xmath56 but for this case clearly an upper bound is @xmath63 by inspection , i.e. , an upper bound on @xmath46 strictly larger than @xmath55 is achievable .
the main message learned is that in practice considering @xmath17 is good enough . for the remaining of the paper
, we will only use @xmath17 instead of @xmath46 , unless otherwise noted , because @xmath17 lets us use the algebra of @xmath12 .
an interleaver of length @xmath0 is a _ maximum - spread _ interleaver if it achieves a spread factor @xmath17 equal to the upper bound .
however , this will be of little relevance in this paper and we keep @xmath47 for simplicity , unless otherwise noted . ]
@xmath47 .
once more , dolinar and divsalar @xcite have reported that linear interleavers either achieve or closely approximate a spread factor of @xmath47 for any @xmath0 . in particular ,
if @xmath0 is twice of a perfect square @xmath64 they have given _ all _ maximum - spread linear interleavers of the form @xmath65 by an explicit simple condition for @xmath66 . however , they pointed out that the resulting interleavers were not good for turbo coding because of their high regularity ( see fig .
[ fig : ms512_0_31 ] , which corresponds to a plot of a maximum - spread linear interleaver ) .
this fact is also theoretically addressed by the linear interleaver asymptote in @xcite , which implied the existence of low - weight codewords of input - weight four and a high multiplicity , close to @xmath0 .
therefore they proposed a semi - random interleaver construction algorithm @xcite , generating the so - called @xmath67-random interleavers with a parameter @xmath67 , but the algorithm sacrificed the spread factor @xmath17 .
the obtained spread factor is typically @xmath68 , i.e. , smaller than about 50% of the upper bound @xmath69 .
an @xmath67-random interleaver yield turbo codes with very good error rate performances and became the typical benchmark interleaver .
one of the main drawbacks of @xmath67-random interleavers is a costly storage of a sequence of @xmath0 integers needed to specify the interleaver . because the construction algorithm relies heavily on a pseudo - random number generator , the sequence has little margin for compression .
this characteristic also makes more difficult for an accurate reproduction of the results by others because typically only the parameter @xmath67 is reported in the literature ; however , the repeatability problem is not so critical because in general , for a given parameter @xmath67 , instances of @xmath67-random interleavers perform similarly for error rates that are not extremely low , which also reflects a good minimum distance of the associated turbo code .
crozier proposed two interleaver constructions @xcite that attempt maximization of the spread factor @xmath17 but avoid or minimize the regularity of linear interleavers : the high - spread construction and the dithered - diagonal construction .
the dithered - diagonal interleavers @xcite are reported to be maximum - spread for interleaver sizes as in ( [ eq:2perfsq ] ) and have large spread factors for others .
dithered - diagonal interleavers require @xmath70 integer parameters for their definition .
an impressive error performance , exceeding the performance of @xmath67-random interleavers was shown for @xmath71 @xcite .
this was a great progress in the field of interleaver constructions for turbo codes because it produced a large spread factor @xmath17 combined with sufficient irregularity to produce an excellent error performance and a much smaller number of integer parameters specifying the interleaver compared with @xmath67-random interleavers .
subsequently , crozier and guinard proposed dithered relatively prime ( drp ) interleavers @xcite .
drp interleavers also perform very well but require a much smaller number of integer parameters for their specification .
their typical choice for a good performance requires about 10 integers . ] for their dither parameters @xmath72 . for an excellent performance ,
their dither parameters @xmath73 require about 18 integers .
interleavers based on qpp require only 2 integers as parameters with the simplicity of only a simple quadratic congruence as the algorithm to generate the permutation sequence .
in fact , qpp interleavers can be implemented in hardware with only additions and comparisons @xcite . if we fix @xmath0 , what is the largest spread for a qpp interleaver ?
an exhaustive search for the largest achievable spread @xmath74 for @xmath75 is shown in fig .
[ fig : dmax ] . due to several algebraic and geometric properties of qpp interleavers explained in section [ sec : qpp_nl_si ] , the exhaustive search is efficiently completed in a very short time on a regular personal computer using theorem [ th : qpp_search ] . with qpp interleavers for a given @xmath0 .
] a few of the polynomials for some commonly found interleaver lengths in the literature are reported in table [ tab : maxqpp ] .
some of the qpps in table [ tab : maxqpp ] generate very good turbo codes , however , better qpps that do not simply attempt maximization of the spread factor are provided in section [ sec : new_metric ] .
.qpps with the largest spread factors @xmath74 .
[ cols=">,^,^,^,^,^",options="header " , ] [ tab : maxomega ] we have mentioned that while there are linear _ maximum - spread _
interleavers for every @xmath76 , this is not true for quadratic permutation polynomials . for example , if @xmath77 there are no _ maximum - spread _ interleavers of second degree .
however , the cubic permutation polynomial @xmath78 generates a _
maximum - spread _ interleaver .
further , interleaver lengths not admitting quadratic permutation polynomials at all may have permutation polynomials of larger degrees .
for example , if @xmath79 , there are no quadratic permutation polynomials but @xmath80 is a cubic permutation polynomial of irreducible degree .
this means maybe it is worth investigating polynomials of larger degrees because the complexity of evaluating the polynomials only grows linearly with the degree using horner s rule @xcite .
the cubic polynomial @xmath78 can be evaluated at every point @xmath81 with three multiplications and two additions using @xmath82 .
further , it would be interesting if an efficient implementation that sequentially generates all evaluations @xmath83 that requires only additions and comparisons for arbitrary polynomial degrees , generalizing the idea in @xcite , is possible . to determine the coefficients for a permutation polynomial of an arbitrary degree and an arbitrary @xmath0
, there is an easy sufficient condition by using theorem 2.3 and corollary 2.5 in @xcite .
algorithms for computing the true distance spectra of turbo codes such as in @xcite and @xcite are very useful to help analysis . we used the algorithm in @xcite to compute the first 20 smallest distances and respective multiplicities for a few turbo codes using interleavers in tables [ tab : ms128]-[tab : ms512 ] .
we constrained the search for input weights at most 10 .
the turbo codes are parallel concatenated codes of nominal rate 1/3 .
we used two types of constituent codes : the 8-state constituent convolutional code with generator matrix @xmath84 $ ] , also denoted in octal notation as ( 13,15 ) and the 16-state constituent convolutional code with generator matrix @xmath85 $ ] , also denoted in octal notation as ( 23,35 ) .
the trellis termination method was the same as for the 3rd generation partnership project ( 3gpp ) standard @xcite , i.e. , udt . |c |c |c |c |c |c |c |c |c |c |c |c |c |c |c |c |c |c |c |c |c | + + @xmath42 & * 0 * & 1 & 2 & 3 & * 4 * & 5 & 6 & 7 & 8 & 9 & 10 & 11 & 12 & 13 & 14 & 15 & 16 & 17 & 18 & 19 + @xmath86 & * 16 * & 18 & 19 & 20 & * 21 * & 22 & 23 & 24 & 25 & 26 & 27 & 28 & 29 & 30 & 31 & 32 & 33 & 34 & 35 & 36 + @xmath87 & * 1 * & 1 & 1 & 2 & * 55 * & 2 & 7 & 8 & 9 & 16 & 27 & 120 & 53 & 359 & 526 & 391 & 775 & 1368 & 1865 & 2542 + + @xmath42 & * 0 * & 1 & 2 & 3 & 4 & 5 & 6 & 7 & * 8 * & 9 & 10 & 11 & 12 & 13 & 14 & 15 & 16 & 17 & 18 & 19 + @xmath86 & * 14 * & 15 & 16 & 18 & 19 & 20 & 21 & 22 & * 23 * & 24 & 25 & 26 & 27 & 28 & 29 & 30 & 31 & 32 & 33 & 34 + @xmath87 & * 2 * & 1 & 1 & 2 & 4 & 3 & 8 & 8 & * 33 * & 37 & 24 & 123 & 72 & 198 & 111 & 296 & 578 & 731 & 1240 & 1822 + + @xmath42 & * 0 * & 1 & 2 & 3 & 4 & 5 & 6 & * 7 * & 8 & 9 & 10 & 11 & 12 & 13 & 14 & 15 & 16 & 17 & 18 & 19 + @xmath86 & * 12 * & 16 & 18 & 19 & 20 & 21 & 22 & * 23 * & 24 & 25 & 26 & 27 & 28 & 29 & 30 & 31 & 32 & 33 & 34 & 35 + @xmath87 & * 1 * & 1 & 3 & 2 & 1 & 7 & 8 & * 34 * & 32 & 15 & 117 & 61 & 184 & 119 & 289 & 532 & 651 & 1166 & 1731 & 3023 + + + @xmath42 & * 0 * & 1 & 2 & 3 & 4 & 5 & 6 & 7 & 8 & 9 & 10 & 11 & * 12 * & 13 & 14 & 15 & 16 & 17 & 18 & 19 + @xmath86 & * 17 * & 18 & 19 & 20 & 21 & 22 & 23 & 24 & 25 & 26 & 27 & 28 & * 29 * & 30 & 31 & 32 & 33 & 34 & 35 & 36 + @xmath87 & * 1 * & 3 & 2 & 3 & 1 & 1 & 3 & 1 & 6 & 9 & 14 & 13 & * 52 * & 28 & 199 & 78 & 79 & 154 & 235 & 422 + + @xmath42 & * 0 * & 1 & 2 & 3 & 4 & 5 & * 6 * & 7 & 8 & 9 & 10 & 11 & 12 & 13 & 14 & 15 & 16 & 17 & 18 & 19 + @xmath86 & * 17 * & 21 & 23 & 24 & 25 & 26 & * 27 * & 28 & 29 & 30 & 31 & 32 & 33 & 34 & 35 & 36 & 37 & 38 & 39 & 40 + @xmath87 & * 1 * & 2 & 2 & 2 & 4 & 5 & * 31 * & 18 & 55 & 23 & 97 & 103 & 133 & 253 & 234 & 596 & 722 & 1261 & 1923 & 3119 + + @xmath42 & * 0 * & 1 & 2 & 3 & * 4 * & 5 & 6 & 7 & 8 & 9 & 10 & 11 & 12 & 13 & 14 & 15 & 16 & 17 & 18 & 19 + @xmath86 & * 23 * & 24 & 25 & 26 & * 27 * & 28 & 29 & 30 & 31 & 32 & 33 & 34 & 35 & 36 & 37 & 38 & 39 & 40 & 41 & 42 + @xmath87 & * 3 * & 1 & 4 & 5 & * 31 * & 9 & 56 & 16 & 93 & 101 & 109 & 233 & 190 & 577 & 655 & 1099 & 1731 & 2920 & 4407 & 6408 + [ tab : ms128 ] |c |c |c |c |c |c |c |c |c |c |c |c |c |c |c |c |c |c |c |c |c | + + @xmath42 & * 0 * & 1 & 2 & 3 & 4 & 5 & * 6 * & 7 & 8 & 9 & 10 & 11 & 12 & 13 & 14 & 15 & 16 & 17 & 18 & 19 + @xmath86 & * 27 * & 28 & 29 & 30 & 31 & 32 & * 33 * & 34 & 35 & 36 & 37 & 38 & 39 & 40 & 41 & 42 & 43 & 44 & 45 & 46 + @xmath87 & * 1 * & 2 & 5 & 1 & 8 & 3 & * 126 * & 4 & 14 & 14 & 17 & 960 & 1923 & 41 & 304 & 99 & 730 & 1113 & 539 & 406 + + @xmath42 & * 0 * & 1 & 2 & 3 & 4 & 5 & 6 & 7 & 8 & 9 & 10 & 11 & 12 & 13 & 14 & 15 & * 16 * & 17 & 18 & 19 + @xmath86 & * 16 * & 21 & 22 & 24 & 25 & 26 & 27 & 28 & 29 & 30 & 31 & 32 & 33 & 34 & 35 & 36 & * 37 * & 38 & 39 & 40 + @xmath87 & * 1 * & 1 & 1 & 1 & 4 & 2 & 2 & 3 & 5 & 7 & 11 & 13 & 18 & 16 & 36 & 30 & * 114 * & 200 & 665 & 543 + + @xmath42 & * 0 * & 1 & 2 & 3 & 4 & 5 & 6 & 7 & 8 & 9 & 10 & 11 & 12 & 13 & 14 & * 15 * & 16 & 17 & 18 & 19 + @xmath86 & * 20 * & 21 & 24 & 25 & 26 & 27 & 28 & 29 & 30 & 31 & 32 & 33 & 34 & 35 & 36 & * 37 * & 38 & 39 & 40 & 41 + @xmath87 & * 1 * & 1 & 1 & 5 & 2 & 1 & 2 & 5 & 9 & 15 & 9 & 13 & 22 & 27 & 34 & * 114 * & 196 & 669 & 530 & 430 + + @xmath42 & * 0 * & 1 & 2 & 3 & 4 & 5 & 6 & 7 & 8 & * 9 * & 10 & 11 & 12 & 13 & 14 & 15 & 16 & 17 & 18 & 19 + @xmath86 & * 27 * & 28 & 29 & 30 & 31 & 32 & 33 & 34 & 35 & * 36 * & 37 & 38 & 39 & 40 & 41 & 42 & 43 & 44 & 45 & 46 + @xmath87 & * 2 * & 2 & 5 & 1 & 5 & 2 & 9 & 3 & 14 & * 246 * & 79 & 505 & 563 & 58 & 76 & 371 & 617 & 432 & 591 & 1324 + + + @xmath42 & * 0 * & 1 & 2 & 3 & 4 & 5 & 6 & 7 & 8 & 9 & 10 & 11 & 12 & 13 & 14 & 15 & 16 & 17 & * 18 * & 19 + @xmath86 & * 26 * & 27 & 29 & 31 & 33 & 34 & 35 & 36 & 37 & 38 & 39 & 40 & 41 & 42 & 43 & 44 & 45 & 46 & * 47 * & 48 + @xmath87 & * 2 * & 1 & 1 & 1 & 3 & 1 & 4 & 3 & 3 & 4 & 2 & 7 & 14 & 9 & 18 & 25 & 28 & 41 & * 151 * & 184 + + @xmath42 & * 0 * & 1 & 2 & 3 & 4 & 5 & 6 & 7 & 8 & 9 & 10 & 11 & 12 & 13 & 14 & 15 & 16 & 17 & 18 & 19 + @xmath86 & * 17 * & 19 & 23 & 25 & 26 & 27 & 28 & 29 & 30 & 31 & 32 & 33 & 34 & 35 & 36 & 37 & 38 & 39 & 40 & 41 + @xmath87 & * 1 * & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 4 & 1 & 3 & 3 & 2 & 12 & 5 & 14 & 16 & 13 & 17 + + @xmath42 & * 0 * & 1 & 2 & 3 & 4 & 5 & 6 & 7 & 8 & 9 & 10 & 11 & 12 & 13 & 14 & 15 & 16 & 17 & 18 & 19 + @xmath86 & * 18 * & 19 & 23 & 25 & 26 & 27 & 28 & 29 & 31 & 32 & 33 & 34 & 35 & 36 & 37 & 38 & 39 & 40 & 41 & 42 + @xmath87 & * 1 * & 1 & 1 & 1 & 1 & 2 & 1 & 1 & 4 & 1 & 6 & 2 & 3 & 7 & 6 & 14 & 19 & 11 & 16 & 16 + + @xmath42 & * 0 * & 1 & 2 & 3 & 4 & 5 & * 6 * & 7 & 8 & 9 & 10 & 11 & 12 & 13 & 14 & 15 & 16 & 17 & 18 & 19 + @xmath86 & * 27 * & 28 & 29 & 30 & 31 & 33 & * 35 * & 36 & 37 & 38 & 39 & 40 & 41 & 42 & 43 & 44 & 45 & 46 & 47 & 48 + @xmath87 & * 2 * & 1 & 2 & 1 & 1 & 3 & * 63 * & 1 & 3 & 6 & 3 & 1 & 6 & 12 & 10 & 19 & 21 & 33 & 45 & 48 + [ tab : ms512 ] we make the following observations : * improving the corner merit @xmath88 has mixed effects .
it improves the minimum distance ( highlighted in the tables ) for certain cases but it degrades for others as commented in section [ sec : cm ] .
* the first spectral line with a high multiplicity has a multiplicity very close to the degree of shift invariance @xmath24 as predicted in section [ sec : cm ] .
* the @xmath89 interleavers appear in general to improve the first spectral line that has a high multiplicity . however , the comparison is not straightforward because the distance spectra is truncated to the first 20 lines .
the error performances of turbo codes using the _ maximum - spread _ qpp interleavers in table [ tab : msinterleavers ] as well as interleavers in tables [ tab : maxqpp ] and [ tab : maxomega ] are examined .
the structure of the simulated turbo codes is the same as in the previous section .
however , we simulated for both udt and idt .
the true code - rates for 8-state constituent codes are @xmath90 and @xmath91 for udt and idt , respectively .
the true code - rates for 16-state constituent codes are @xmath92 and @xmath93 for udt and idt , respectively .
the true code - rate is used to compute the signal - to - noise ratio @xmath94 .
we used bpsk modulation and assumed an additive white gaussian noise ( awgn ) channel .
the decoding was performed with log - map decoding and simulated until at least 50 frame errors had been counted , unless otherwise noted .
we compare first length @xmath71 pp interleavers under similar conditions with drp interleavers in @xcite .
the iterative decoding method used in @xcite is simpler than log - map but worse by about 0.1db at lower snrs .
the drp curves were obtained directly from @xcite but adjusted in snr to the true code - rate .
we used 16 log - map decoding iterations to match their number of iterations .
the simulation curves are compared in fig .
[ fig : drpvsqpp ] .
pp interleavers compared with drp interleavers in turbo codes with 8-state constituent convolutional codes and 16 decoding iterations . ]
there is a close match in performance with the drp interleaver with parameter @xmath95 at higher snrs .
the similarity in performance is not quite surprising if we consider that both pp and drp constructions are based on number congruences .
a deeper study of their resemblance is left for a future work , however , we provide a few insights of why qpp interleavers are expected to be close to drp interleavers with @xmath95 for this interleaver length of @xmath71 . the parameter @xmath96 in a drp interleaver is the length of an arbitrary permutation of @xmath96 letters . explaining only the essence , a drp interleaver can be understood as a modification of a linear interleaver by using the permutation of @xmath96 letters .
we show that pp and drp constructions are in general different , except when @xmath97 . when @xmath97 , drp interleavers are what we call linear interleavers , i.e. , pps of first degree because no modification of a linear interleaver is possible if @xmath97 .
as @xmath96 grows , drp interleavers can become arbitrarily `` random '' in the sense that when @xmath98 , arbitrary interleaving functions are possible .
the class of pp interleavers can not generate arbitrary interleaving functions regardless of the degree of the permutation polynomial .
this is easily proven by counting arguments but also because _ all _ interleavers generated by pps are maximum contention - free @xcite but an arbitrary interleaving function is generally not ( maximum ) contention - free @xcite .
factoring out the construction algorithm , which is similar in complexity and nature , one may characterize pp and drp interleavers by their input parameters ( two vectors of length @xmath96 for drp interleavers and the coefficients of the polynomials for pp interleavers ) .
we propose to measure the `` randomness '' of drp interleavers by computing the entropy @xmath99 of the two input permutations of @xmath96 letters .
considering that an arbitrary sequence of @xmath96 letters can be chosen , then the entropy is @xmath100 bits , which is independent of the interleaver length . for an interleaver length of @xmath0 ,
the entropy of the coefficients in @xmath101 is @xmath102 bits assuming that @xmath0 is a power of 2 and @xmath103 s are either even or odd @xcite . now observing the previous matching in fig .
[ fig : drpvsqpp ] for drp @xmath104 and qpp , we have @xmath105 bits and @xmath106 bits with a perfect match .
higher because it requires the specification of a base linear interleaver that includes a constant coefficient . ]
would this be a coincidence ?
the drp interleaver with @xmath107 has an entropy of @xmath108 bits , which would correspond to a degree-6 pp .
we obtained a very close error performance with a degree six permutation polynomial .
@xmath109 the polynomial in ( [ eq : sixth ] ) has a degree of non - linearity @xmath110 , a refined degree of non - linearity @xmath111 and the orbit size is @xmath112 .
its spread factor is @xmath113 and has a refined product merit of @xmath114 .
the corresponding performance curve is shown in fig . [
fig : drpvsqpp ] .
we used interleaved dual termination ( idt ) @xcite for this code , which gave a small advantage over udt at high snrs ( it is not shown in the plot but about a factor of 2 at 2.0db ) . in summary :
* the maximum - spread interleaver ( @xmath115 ) is mostly indifferent to the termination methods and very similar in performance to the drp @xmath95 and idt . *
the @xmath89 interleaver ( @xmath116 ) has a significant improvement with idt over udt and is a little better than the maximum - spread code . * the degree-6 interleaver with idt and drp @xmath107 interleaver
have mostly the same performance .
we plot next the fer curves for the first four _ maximum - spread _ interleavers in table [ tab : msinterleavers ] using udt and the corresponding constant - free codes in table [ tab : maxomega ] using idt for turbo codes using the 8-state ( 13,15 ) convolutional codes and eight decoding iterations .
similarly as in @xcite , excellent error performance is obtained down to at least an fer of @xmath117 .
there is noticeable improvement in fer for codes using @xmath89 interleavers as the block length increases .
also , a truncated union bound is plotted using the 20 first spectral lines for two of the codes using maximum - spread interleavers ( ms qpp @xmath118 and ms qpp @xmath71 ) .
the refined product merit effectively avoids the selection of bad interleavers .
for example ,
the permutation polynomial @xmath119 has a spread factor @xmath120 ( a maximum - spread interleaver ) , a degree of non - linearity @xmath110 , a simple product measure of @xmath121 but a refined degree of non - linearity @xmath122 , and a refined product measure of @xmath123 .
the simulation performance at 1.75db ( not shown in plots ) gives an fer of @xmath124 , clearly inferior to the maximum - spread interleaver of second degree that has a larger @xmath125 .
an impressive error performance @xcite with fers around @xmath126 at an snr of 1.25db was demonstrated for a drp interleaver with @xmath107 in a turbo code using the ( 23,35 ) 16-state constituent code , idt , and 16 decoding iterations .
we show that a similar result is obtained with the qpp interleaver @xmath127 in table [ tab : maxomega ] with the same constituent code , udt , and 16 log - map decoding iterations .
we have not simulated the constant - free case with idt but it should provide an even better error performance . other codes with interleavers in tables [ tab : maxqpp ] ( ms ) and tables [ tab : maxomega ] ( @xmath89 ) were also simulated .
a genie stopper for the decoding iterations ( the iterations are stopped when the decoded information sequence completely agrees with the transmitted one ) was used when simulating points at very low fers . for the length @xmath128 @xmath89 qpp code , at 0.5db , close to 6 million frames
were simulated with no frame errors with a regular turbo decoder ; additionally , almost 40 million frames were simulated with a genie stopper for the iterations , resulting in only three frame errors .
similarly , other points at very low fers were simulated with the decoding parameters in table [ tab : sim ] .
& frames & frame errors + regular & 74,151,303 & 1 + genie & 657,736,347 & 7 + total & 731,887,650 & 8 + + & frames & frame errors + regular & 56,273,401 & 10 + genie & 0 & 0 + total & 56,273,401 & 10 + + & frames & frame errors + regular & 195,667,074 & 11 + genie & 0 & 0 + total & 195,667,074 & 11 + + & frames & frame errors + regular & 25,603,477 & 1 + genie & 180,873,414 & 3 + total & 206,476,891 & 4 + + & frames & frame errors + regular & 5,887,701 & 0 + genie & 39,323,725 & 3 + total & 45,211,426 & 3 + + & frames & frame errors + regular & 4,299,436 & 0 + genie & 29,305,054 & 3 + total & 33,604,490 & 3 + [ tab : sim ] although there was an evident improvent by using @xmath89 polynomials for larger block lengths using 8-state constituent codes , there is not a clear difference for the length @xmath128 using 16-state constituent codes within the reach of simulations .
in addition , we observe the following for length @xmath71 : * codes using both the maximum - spread interleaver ( @xmath115 ) and the @xmath89 ( @xmath116 ) interleaver benefit significantly with idt . *
the code with the maximum - spread interleaver and idt is better than the code with @xmath89 ( @xmath116 ) . *
the code with the maximum - spread interleaver and idt has a performance very close to the best interleaver in @xcite , a dithered diagonal interleaver with identical spread factor of @xmath120 . in summary , for resource - constrained applications , the combination of 8-state constituent codes , a maximum of 8 decoding iterations and qpp interleavers give excellent performance down to fers around @xmath117 or smaller from short to medium information block sizes of up to 8192 bits .
if 16-state constituent codes and a maximum of 16 decoding iterations are acceptable then qpp interleavers give impressive performance down to fers close to @xmath129 with little or no signs of error - floors .
the recently proposed construction of pp interleavers @xcite and the subclass of qpp interleavers yield very good error rate performance for turbo codes in a number of practical examples @xcite .
interesting properties of qpp interleavers have been studied earlier @xcite ; however , their characterization is still in its infancy with many open questions . in this paper ,
the spread factor of qpp interleavers was studied .
an infinite sequence of qpp interleavers achieving the _ maximum - spread _ was given .
moreover , several properties of pps and effective measures that are relevant for turbo coding were investigated .
we proposed a new _ refined product measure _
@xmath89 for pp interleavers that is easily and quickly computed ; a table of good interleavers according to the new measure is provided for several of the interleaver lengths reported in the literature .
qpps are convenient because of their small footprint ; however , we provided some evidence that pps with larger degrees than two may be necessary for very low fers .
although , turbo codes are limited by a logarithmic growth on the minimum distance @xcite , that result is asymptotic : good codes at practical lengths achieving fers close to @xmath129 without signs of an error floor have been demonstrated earlier @xcite using drp interleavers .
similar results are demonstrated in this paper with qpp interleavers . the work in @xcite opened the doors to a class of interleavers for turbo coding that provides the combination of an excellent performance , elegant algebraic and geometric analysis , and simplicity of implementation ; to the best of our knowledge
, there are no other interleavers sharing simultaneously these three characteristics to their fullest extent .
many problems in turbo coding that were earlier treated empirically may now be recast to be solved systematically and rigorously : the design of multiple turbo codes @xcite for an improved asymptotic minimum distance by classifying the properties of pp interleavers under function composition ; and puncturing of code bits for higher data rates using the @xmath130 theory @xcite ( a stronger generalization of an even - odd interleaver @xcite ) are only a couple of relevant examples .
* appendix *
a technique inspired from @xcite is used to compute the upper bound @xmath52 .
we use the same distance @xmath35 in section [ sec : ms ] but over the space @xmath131 and forming the metric space @xmath132 . in this metric space , a sphere of radius @xmath133 has the shape of a square when projected over the euclidean metric space as shown in fig .
[ fig : tspace ] . if two spheres of radii @xmath133 and @xmath134 touch each other then the distance between their centers is @xmath135 . . ] before proceeding with the upper bound , it is straightforward to realize that in the metric space @xmath136 there are only two types of densest packings for spheres with identical radius : aligned packing and unaligned packing as shown in fig .
[ fig : tpacking ] .
the unaligned packing is obtained by arbitrarily `` sliding '' stripes of the aligned packing . .
] the upper bound on @xmath46 is constructive . given a square whose area is @xmath137 , we attempt to pack @xmath0 spheres of identical radii @xmath138 and maximize the radii of the spheres @xmath138 .
we do so for values of @xmath0 for which clearly a subset of the aligned packing gives the densest packing as shown in fig .
[ fig : tbound ] .
we end up with 1 ) several points touching the boundaries of @xmath139 and 2 ) excess cover areas . .
] there are two classes of values for @xmath0 .
the cases @xmath140 and @xmath141 belong to @xmath142 where @xmath143 is the number of points on the bottom edge .
the upper bound is the distance between adjacent points , i.e. , @xmath144 .
after some simple algebra , the upper bound @xmath52 is @xmath145 the cases @xmath146 and @xmath147 belong to @xmath148 the upper bound @xmath52 is @xmath149
although not explained in @xcite , it is clear that many candidates for an interleaver length @xmath0 achieving maximum - spread @xmath47 are of the form @xmath76 because those allow the optimal packing in @xmath150 .
this is illustrated in fig .
[ fig : perfect ] for @xmath151 .
the dots show the potential places that may be occupied by a point of the interleaver - code . without loss of generality ,
we place first a point in @xmath152 .
next , we place a point along the line @xmath153 and @xmath154 with the constraint that a linear curve that interpolates the first two points is a permutation polynomial
. then we follow a similar but simpler technique as we do to prove theorem [ th : sequence ] in appendix [ sec : sequence ] .
this leads to an alternate geometric construction for all maximum - spread linear interleavers ( up to symmetries ) .
the authors of @xcite explain the so - called dithered diagonal construction using the same geometric approach .
the difference with linear interleavers is that the placement of a second point does not fix all the remaining points . therefore a different slide ( or dither )
is allowed for the other stripes provided that an interleaver constraint is maintained . .
the set of linear curves @xmath155 is simply passing through the center of the stripes in either the aligned or an unaligned packing as show in fig . [
fig : perfect ] , i.e. , the smallest distance @xmath38 between two points that belong to any two disjoint curves is @xmath156 .
if we show that the intersection of each of the curves @xmath157 with @xmath25 has @xmath158 solutions that are equally spaced by @xmath159 then we are done .
we start by rewriting the infinite sequence ( [ eq : sequence ] ) as function of @xmath160 .
@xmath161 the intersection of ( [ eq : n_form ] ) and a linear curve @xmath157 is the solution to @xmath162 or @xmath163 we then transform the quadratic congruence ( [ eq : change_var ] ) into a linear congruence @xmath164 whose set of solutions is @xmath165 i.e. , we have @xmath166 solutions that are equally spaced for ( [ eq : lin_reduction ] ) . to solve ( [ eq : change_var ] ) , we simply need to find the solution to @xmath167 when @xmath42 and @xmath43 are fixed .
observing that @xmath168 is a qpp , the solution for @xmath81 is unique .
further , from the proof of the @xmath130 property of permutation polynomials ( * ? ? ?
* theorem 1 ) , we know that if @xmath96 divides @xmath0 then a set of @xmath96 equally spaced ( modulo @xmath0 ) values by @xmath169 is always mapped to an equally spaced set of values by a pp and also by its inverse pp . therefore we establish the desired result .
if @xmath170 is an isometry of the interleaver - code @xmath171 then we must have @xmath172 . developing it we have there are exactly @xmath179 distinct solutions , which means that each point in @xmath180 belongs to an orbit of size @xmath179 ,
i.e. , the degree of shift - invariance is @xmath181 .
hence , from proposition [ prop : shift - nl ] , the degree of non - linearity for qpps is @xmath182 .
a. nimbalker , t. k. blankenship , b. classon , t. e. fuja , and d. j. costello , jr .
, `` contention - free interleavers , '' in _ proc . 2004 ieee international symposium on information theory _ , chicago , il , 2004 , p. 54 . c. berrou , s. k. y. saouter , c. douillard , and m. jzquel , `` designing good permutations for turbo codes : towards a single model , '' in _ proc .
international conference on communications _ , vol . 1 , paris , france , june 2004 , pp . 341345 .
c. j. corrada - bravo and i. rubio , `` deterministic interleavers for turbo codes with random - like performance and simple implementation , '' in _ proc .
3rd international symposium on turbo codes _ , brest , france , sept .
a. giulietti , l. van der perre , and m. strum , `` parallel turbo coding interleavers : avoiding collisions in access to storage elements , '' _ iee electronics letters _ , vol .
38 , no . 5 , pp .
232234 , feb .
2002 .
s. crozier and p. guinand , `` high - performance low - memory interleaver banks for turbo - codes , '' in _ proc .
54th ieee vehicular technology conference ( vtc 2001 fall ) _ , atlantic city , new jersey , oct . 2001 , pp .
23942398 .
f. daneshgaran and m. mondin , `` design of interleavers for turbo codes : iterative interleaver growth algorithms of polynomial complexity , '' _ ieee trans .
inform . theory _
45 , no . 6 , pp . 18451859 , sept . 1999 .
o. y. takeshita and d. j. costello , jr .
, `` on deterministic linear interleavers for turbo - codes , '' in _ proc .
35th annual allerton conference on communication , control , and computing _ , sept .
1997 , pp . 711712 . c. berrou , a. glavieux , and p. thitimajshima , `` near shannon limit error - correcting coding and decoding : turbo - codes , '' in _ proc .
conf . on commun .
_ , geneve , switzerland , may 1993 , pp
. o. y. takeshita , `` on maximum contention - free interleavers and permutation polynomials over integer rings , '' _ ieee trans .
inform . theory _
2006 , to be published .
[ online ] .
available : http://arxiv.org/abs/cs.it/0506093/ j. ryu and o. y. takeshita , `` on quadratic inverses for quadratic permutation polynomials over integer rings , '' _ ieee trans .
inform . theory _
2006 , to be published .
[ online ] .
available : http://arxiv.org/abs/cs.it/0511060/ o. y. takeshita , `` a new construction for ldpc codes using permutation polynomials over integer rings , '' _ ieee trans .
inform . theory _
, submitted for publication .
[ online ] .
available : http://arxiv.org/abs/cs.it/0506091/ r. garello , p. pierloni , and s. benedetto , `` computing the free distance of turbo codes and serially concatenated codes with interleavers : algorithms and applications , '' _ ieee j. select .
areas commun .
_ , vol . 19 , no . 5 , pp .
800812 , may 2001 .
s. crozier and p. guinand , `` distance upper bounds and true minimum distance results for turbo - codes designed with drp interleavers , '' in _ proc .
3rd international symposium on turbo codes & related topics _ , brest , france , sept .
2003 , pp . 169172 .
|
an interleaver is a critical component for the channel coding performance of turbo codes .
algebraic constructions are important because they admit analytical designs and simple , practical hardware implementation
. the spread factor of an interleaver is a common measure for turbo coding applications .
maximum - spread interleavers are interleavers whose spread factors achieve the upper bound .
an infinite sequence of quadratic permutation polynomials over integer rings that generate maximum - spread interleavers is presented .
new properties of permutation polynomial interleavers are investigated from an algebraic - geometric perspective resulting in a new non - linearity metric for interleavers .
a new interleaver metric that is a function of both the non - linearity metric and the spread factor is proposed .
it is numerically demonstrated that the spread factor has a diminishing importance with the block length .
a table of good interleavers for a variety of interleaver lengths according to the new metric is listed .
extensive computer simulation results with impressive frame error rates confirm the efficacy of the new metric .
further , when tail - biting constituent codes are used , the resulting turbo codes are quasi - cyclic . algebraic , geometry , interleaver , permutation polynomial , quadratic , quasi - cyclic , spread , turbo code .
|
state - space models , also known as hidden markov models , are a very popular class of time series models that have found numerous of applications in fields as diverse as statistics , ecology , econometrics , engineering and environmental sciences ; see @xcite , @xcite , @xcite , @xcite .
formally a state - space model is defined by two stochastic processes @xmath0 and @xmath1 .
the process @xmath0 is a @xmath2-valued latent markov process of initial density @xmath3 and markov transition density @xmath4 , that is @xmath5 whereas the @xmath6-valued observations @xmath1 satisfy @xmath7 where @xmath8 denotes the conditional marginal density ,
@xmath9 the parameter of the model and @xmath10 denotes components @xmath11 of a sequence @xmath12 .
the spaces @xmath2 and @xmath6 can be euclidean but what follows applies to more general state spaces as well .
the popularity of state - space models stems from the fact that they are flexible and easily interpretable .
applications of state - space models include stochastic volatility models where @xmath13 is the volatility of an asset and @xmath14 its observed log - return @xcite , biochemical network models where @xmath13 corresponds to the population of various biochemical species and @xmath14 are imprecise measurements of the size of a subset of these species @xcite , neuroscience models where @xmath13 is a state vector determining the neuron s stimulus - response function and @xmath14 some spike train data @xcite . however , nonlinear non - gaussian state - space models are also notoriously difficult to fit to data and it is only recently , thanks to the advent of powerful simulation techniques , that it has been possible to fully realize their potential . to illustrate the complexity of inference in state - space models , consider first the scenario where the parameter @xmath15 is _
known_. on - line and off - line inference about the state process @xmath16 given the observations @xmath17 is only feasible analytically for simple models such as the linear gaussian state - space model .
in nonlinear non - gaussian scenarios , numerous approximation schemes such as the extended kalman filter or the gaussian sum filter @xcite have been proposed over the past fifty years to solve these so - called optimal filtering and smoothing problems , but these methods lack rigor and can be unreliable in practice in terms of accuracy , while deterministic integration methods are difficult to implement .
markov chain monte carlo ( mcmc ) methods can obviously be used but they are impractical for on - line inference ; and even for off - line inference it can be difficult to build efficient high - dimensional proposal distributions for such algorithms . for nonlinear non - gaussian state space models _ particle algorithms _ have emerged as the most successful .
their widespread popularity is due to the fact that they are easy to implement , suitable for parallel implementation @xcite and , more importantly , have been demonstrated in numerous settings to yield more accurate estimates than the standard alternatives ; e.g. see @xcite , @xcite , @xcite , @xcite
. in most practical situations , the model ( [ equ : statespacesystemlitrev1])-([equ : statespacesystemlitrev-2 ] ) depends on an _ _ unknown _ _
parameter vector @xmath15 that needs to be inferred from the data either in an on - line or off - line manner . in fact inferring the parameter @xmath15 is often the primary problem of interest ; e.g. for biochemical networks , we are not interested in the population of the species per se , but we want to infer some chemical rate constants , which are parameters of the transition prior @xmath4 . although it is possible to define an extended state that includes the original state @xmath13 and the parameter @xmath15 and then apply standard particle methods to perform parameter inference , it was recognized very early on that this naive approach is problematic @xcite due to the parameter space not being explored adequately .
this has motivated over the past fifteen years the development of many particle methods for the parameter estimation problem , but numerically robust methods have only been proposed recently . the main objective of this paper is to provide a comprehensive overview of this literature .
this paper thus differs from recent survey papers on particle methods which all primarily focus on estimating the state sequence @xmath18 or discuss a much wider range of topics , e.g. @xcite , @xcite , @xcite , @xcite .
we will present the main features of each method and comment on their pros and cons .
no attempt however is made to discuss the intricacies of the specific implementations .
for this we refer the reader to the original references .
we have chosen to broadly classify the methods as follows : bayesian or maximum likelihood ( ml ) and whether they are implemented off - line or on - line . in the bayesian approach ,
the unknown parameter is assigned a prior distribution and the posterior density of this parameter given the observations is to be characterized . in the ml approach ,
the parameter estimate is the maximizing argument of the likelihood of @xmath15 given the data .
both these inference procedures can be carried out off - line or on - line .
specifically , in an off - line framework we infer @xmath15 using a fixed observation record @xmath19 .
in contrast , on - line methods update the parameter estimate sequentially as observations @xmath20 become available .
the rest of the paper is organized as follows . in section [ sec : computationalissues ] , we present the main computational challenges associated to parameter inference in state - space models . in section [ sec : filteringandparticle ] , we review particle methods for filtering when the model does not include any unknown parameters whereas section [ sec : smoothingandparticle ] is dedicated to smoothing . these filtering and smoothing techniques are at the core of the off - line and on - line ml parameter procedures described in section [ sec : mlestimation ] . in section [ sec : bayesianestimation ] , we discuss particle methods for off - line and on - line bayesian parameter inference .
the performance of some of these algorithms are illustrated on simple examples in section [ sec : experimentalresults ] .
finally , we summarize the main advantages and drawbacks of the methods presented and discuss some open problems in section [ sec : conclusion ] .
a key ingredient of ml and bayesian parameter inference is the likelihood function @xmath21 of @xmath15 which satisfies @xmath22 where @xmath23 denotes the joint density of @xmath24 which is given from equations ( [ equ : statespacesystemlitrev1])-([equ : statespacesystemlitrev-2 ] ) by @xmath25 the likelihood function is also the normalizing constant of the posterior density @xmath26 of the latent states @xmath18 given data @xmath27 @xmath28 this posterior density is itself useful for computing the score vector @xmath29 associated to the log - likelihood @xmath30 , as fisher s identity yields @xmath31 the main practical issue associated to parameter inference in nonlinear non - gaussian state - space models is that the likelihood function is intractable . as performing ml
parameter inference requires maximizing this intractable function , it means practically that it is necessary to obtain reasonably low variance monte carlo estimates of it , or of the associated score vector if this maximisation is carried out using gradient - based methods .
both tasks involve approximating high dimensional integrals , and , whenever @xmath32 is large . on - line
inference requires additionally these integrals to be approximated on the fly , ruling out the applications of standard computational tools such as mcmc .
bayesian parameter inference is even more challenging , as it requires approximating the posterior density @xmath33 where @xmath34 is the prior density . here
not only @xmath35 but also @xmath36 are intractable and , once more , these integrals must be approximated on - line if one wants to update the posterior density sequentially .
we will show in this review that particle methods are particularly well - suited to these integration tasks .
in this section , the parameter @xmath15 is assumed known and we focus on the problem of estimating the latent process @xmath0 sequentially given the observations .
an important byproduct of this so - called filtering task from a parameter estimation viewpoint is that it provides us with an on - line scheme to compute @xmath37 . as outlined in section [ sec : computationalissues ] , the particle approximation of these likelihood terms is a key ingredient of numerous particle - based parameter inference techniques discussed further on .
filtering denotes usually the task of estimating recursively in time the sequence of marginal posteriors @xmath38 , known as the filtering densities .
however we will adopt here a more general definition and will refer to filtering as the task of estimating the sequence of joint posteriors @xmath39 recursively in time but we will still refer to the marginals @xmath38 as the filtering densities .
it is easy to verify from and that the posterior @xmath26 and the likelihood @xmath21 satisfy the following fundamental recursions : for @xmath40 , @xmath41 and @xmath42 where @xmath43 there are essentially two classes of models for which @xmath26 and @xmath21 can be computed exactly : the class of linear gaussian models , for which the above recursions may be implemented using kalman techniques , and when @xmath2 is a finite state - space ; see for example @xcite . for other models
these quantities are typically intractable , i.e. the densities in - can not be computed exactly .
particle filtering methods are a set of simulation - based techniques which approximate numerically the recursions to .
we focus here on the apf ( auxiliary particle filter @xcite ) for two reasons : first , this is a popular approach , in particular in the context of parameter estimation ( see e.g. section [ sub : using - mcmcsteps - within - smc ] ) ; second , the apf covers as special cases a large class of particle algorithms , such as the bootstrap filter @xcite and sisr ( sequential importance sampling resampling @xcite , @xcite ) .
let @xmath44 where @xmath45 is a probability density function which is easy to sample from , and @xmath46 is not necessarily required to be a probability density function but just an non - negative function of @xmath47 one can evaluate .
( for @xmath48 , remove the dependency on @xmath49 , i.e. @xmath50 . ) the algorithm relies on the following importance weights @xmath51 in order to alleviate the notational burden we omit the dependence of the importance weights on @xmath15 ; we will do so in the remainder of the paper when no confusion is possible . the auxiliary particle filter can be summarized as follows . * _ at time _
@xmath48 , for all @xmath52 : 1 .
@xmath53 2 .
@xmath54 @xmath55 3 . @xmath56 . *
_ at time _
@xmath40 , for all @xmath52 : 1 .
@xmath57 @xmath58 2 .
@xmath59 @xmath60 3 .
@xmath61 one recovers the sisr algorithm as a special case of algorithm [ alg : apf ] by taking @xmath62 ( or more generally , by taking @xmath63 , some arbitrary positive function ) .
further , one recovers the bootstrap filter by taking @xmath64 .
this is an important special case , as some complex models are such that one may sample from @xmath65 , but not compute the corresponding density ; in such a case the bootstrap filter is the only implementable algorithm . for models such that the density @xmath65 is tractable , @xcite recommend selecting @xmath66 and @xmath67 when these quantities are tractable , and using approximations of these quantities in scenarios when they are not .
the intuition for these recommendations is that this should make the weight function nearly constant .
the computational complexity of algorithm [ alg : apf ] is @xmath68 per time step ; in particular see e.g. @xcite for a @xmath68 implementation of the resampling step . at time @xmath32 , the approximations of @xmath69 and @xmath70 presented earlier in and respectively are given by @xmath71 where @xmath72 and @xmath73 in practice , one uses mostly to obtain approximations of posterior moments @xmath74\ ] ] but expressing particle filtering as a method for approximating distributions ( rather than moments ) turns out to be a more convenient formalization .
the likelihood is then estimated through @xmath75 the resampling procedure is introduced to replicate particles with high weights and discard particles with low weights . it serves to focus the computational efforts on the promising regions of the state - space .
we have presented above the simplest resampling scheme .
lower variance resampling schemes have been proposed in @xcite , @xcite , as well as more advanced particle algorithms with better overall performance , e.g. the resample - move algorithm @xcite . for the sake of simplicity ,
we have also presented a version of the algorithm that operates resampling at every iteration @xmath32 .
it may be more efficient to trigger resampling only when a certain criterion regarding the degeneracy of the weights is met ; see @xcite , ( * ? ? ?
* and 74 ) .
many sharp convergence results are available for particle methods @xcite .
a selection of these results that gives useful insights on the difficulties of estimating static parameters with particle methods is presented below . under minor regularity assumptions
, one can show that for any @xmath76 , @xmath77 and any bounded test function @xmath78 $ ] , there exist constants @xmath79 such that for any @xmath80 @xmath81\leq\frac{a_{\theta , n , p}}{n^{p/2}}\label{eq : weakresult1}\ ] ] where the expectation is with respect to the law of the particle filter .
in addition , for more general classes of functions , we can obtain for any fixed @xmath32 a central limit theorem ( clt ) as @xmath82 @xcite , ( * ? ? ?
* proposition 9.4.2 ) .
such results are reassuring but weak as they reveal nothing regarding long time behavior .
for instance , without further restrictions on the class of functions @xmath83 and the state - space model , @xmath84 typically grows exponentially with @xmath32 .
this is intuitively not surprising as the dimension of the target density @xmath69 is increasing with @xmath32 .
moreover the successive resampling steps lead to a depletion of the particle population ; @xmath85 will eventually be approximated by a single unique particle as @xmath86 increases .
this is referred to as the _ degeneracy _ problem in the literature ( * ? ? ?
* figure 8.4 , p. 282 ) .
this is a fundamental weakness of particle methods : given a fixed number of particles @xmath87 , it is impossible to approximate @xmath26 accurately when @xmath32 is large enough .
fortunately , it is also possible to establish much more positive results .
many state - space models possess the so - called _ exponential forgetting _ property ( * ? ? ?
* chapter 4 ) .
this property states that for any @xmath88 and data @xmath27 , there exist constants @xmath89 and @xmath90 such that @xmath91 where @xmath92 is the total variation distance ; that is the optimal filter forgets exponentially fast its initial condition .
this property is typically satisfied when the signal process @xmath93 is a uniformly ergodic markov chain and the observations @xmath94 are not too informative ( * ? ? ? * chapter 4 ) or when @xmath94 are informative enough that it effectively restricts the hidden state to a bounded region around it @xcite .
weaker conditions can be found in @xcite .
when exponential forgetting holds , it is possible to establish much stronger uniform - in - time convergence results _ for functions @xmath83 that depend only on recent states_. specifically , for an integer @xmath95 and any bounded test function @xmath96 $ ] , there exist constants @xmath97 such that for any @xmath80 , @xmath98 , @xmath99\leq\frac{c_{\theta , l , p}}{n^{p/2}},\label{eq : lp}\ ] ] where @xmath100 this result explains why particle filtering is an effective computational tool in many applications such as tracking , where one is only interested in @xmath101 , as the approximation error is uniformly bounded over time .
similar positive results holds for @xmath102 .
this estimate is unbiased for any @xmath103 ( * ? ? ?
* theorem 7.4.2 , page 239 ) and , under assumption , the _ relative _ variance of the likelihood estimate @xmath102 , that is the variance of the ratio @xmath104 , is bounded above by @xmath105 @xcite , @xcite .
this is a great improvement over the exponential increase with @xmath32 that holds for standard importance sampling techniques , see for instance @xcite .
however , the constants @xmath106 and @xmath107 are typically exponential in @xmath108 , the dimension of the state vector @xmath13 . we note that non - standard particle methods designed to minimize the variance of the estimate of @xmath21 have recently been proposed @xcite .
finally we recall the theoretical properties of particles estimates of the following so - called smoothed additive functional ( * ? ? ?
* section 8.3 ) , @xcite @xmath109 such quantities are critical when implementing ml parameter estimation procedures ; see section [ sec : mlestimation ] .
if we substitute @xmath110 to @xmath111 to approximate @xmath112 , then we obtain an estimate @xmath113 which can be computed recursively in time , see e.g. ( * ? ? ?
* section 8.3 ) . for the remainder of this paper
we will refer to this approximation as the _ path space _ approximation .
even when holds , there exists @xmath114 such that the asymptotic bias @xcite and variance @xcite satisfy @xmath115 for @xmath116 $ ] where the variance is w.r.t the law of the particle filter . the fact that the variance grows at least quadratically in time follows from the degeneracy problem and makes @xmath113 unsuitable for some on - line likelihood based parameter estimation schemes discussed in section [ sec : mlestimation ] .
in this section , the parameter @xmath15 is still assumed known and we focus on smoothing , that is the problem of estimating the latent variables @xmath117 given a fixed batch of observations @xmath19 . smoothing for a fixed parameter @xmath15 is at the core of the two main particle ml parameter inference techniques described in section [ sec : mlestimation ] as these procedures require computing smoothed additive functionals of the form .
clearly one could unfold the recursion ( [ eq : jointupdate ] ) from @xmath48 to @xmath118 to obtain @xmath119 .
however , as pointed out in the previous section , the path space approximation ( [ eq : smcfullposterior ] ) suffers from the degeneracy problem and yields potentially high variance estimates of ( [ eq : expectationadditivefunctionals ] ) as ( [ eq : sufficientstatsdegrade ] ) holds .
this has motivated the development of alternative particle approaches to approximate @xmath119 and its marginals . for state - space models with
good forgetting properties ( e.g. ( [ eq : ergod ] ) ) , we have @xmath120 for @xmath121 large enough ; that is observations collected at times @xmath122 do not bring any significant additional information about @xmath18 . in particular , when having to evaluate @xmath123 of the form we can approximate the expectation of @xmath124 w.r.t @xmath125 by its expectation w.r.t @xmath126 .
algorithmically , a particle implementation of means not resampling the components @xmath127 of the particles @xmath128 obtained by particle filtering at times @xmath122 .
this was first suggested in @xcite and used in @xcite , @xcite .
this algorithm is simple to implement but the main practical problem is the choice of @xmath121 .
if taken too small , then @xmath129 is a poor approximation of @xmath130 .
if taken too large , the degeneracy remains substantial .
moreover , even as @xmath131 , this particle approximation will have a non - vanishing bias since @xmath132 .
the joint smoothing density @xmath119 can be expressed as a function of the filtering densities @xmath133 using the following key decomposition @xmath134 where @xmath135 is a backward ( in time ) markov transition density given by @xmath136 a backward in time recursion for @xmath137 follows by integrating out @xmath138 and @xmath139 in while applying @xmath140 this is referred to as forward - backward smoothing as a forward pass yields @xmath133 which can be used in a backward pass to obtain @xmath137 .
combined to @xmath141 , this allows us to obtain @xmath123 .
an alternative to these forward - backward procedures is the generalized two - filter formula @xcite .
the decomposition suggests that it is possible to sample approximately from @xmath119 by running a particle filter from time @xmath48 to @xmath142 storing the approximate filtering distributions @xmath143 , i.e. the marginals of , then sampling @xmath144 and for @xmath145 sampling @xmath146 where this distribution is obtained by substituting @xmath147 for @xmath148 in : @xmath149 this forward filtering backward sampling ( ffbsa ) procedure was proposed in @xcite .
it requires @xmath150 operations to generate a single path @xmath117 as sampling from costs @xmath151 operations .
however , as noted in @xcite , it is possible to sample using rejection from an alternative approximation of @xmath152 in @xmath153 operations if we use an unweighted particle approximation of @xmath154 in and if the transition prior satisfies @xmath155 . hence , with this approach , sampling a path @xmath117 costs only on average @xmath156 operations .
a related rejection technique was proposed in @xcite . in practice
, one may generate @xmath87 such trajectories to compute monte carlo averages that approximates smoothing expectations @xmath157 $ ] . in that scenario ,
the first approach costs @xmath158 , while the second approach costs @xmath150 on average . in some applications ,
the rejection sampling procedure can be computationally costly as the acceptance probability can be very small for some particles ; see for example section 4.3 in @xcite for empirical results .
this has motivated the development of hybrid procedures combining ffbsa and rejection sampling @xcite .
we can also directly approximate the marginals @xmath137 .
assuming we have an approximation @xmath159 where @xmath160 then by using and , we obtain the approximation @xmath161 with @xmath162 this forward filtering backward smoothing ( ffbsm , where ` m ' stands for ` marginal ' ) procedure requires @xmath158 operations to approximate @xmath137 instead of @xmath150 for the path space and fixed - lag methods .
however this high computational complexity of forward - backward estimates can be reduced using fast computational methods @xcite .
particle approximations of generalized two - filter smoothing procedures have also been proposed in @xcite , @xcite .
whenever we are interested in computing the sequence @xmath163 recursively in time , the forward - backward procedure described above is cumbersome as it requires performing a new backward pass with @xmath164 steps at time @xmath165 an important but not well - known result is that it is possible to implement exactly the forward - backward procedure using only a forward procedure .
this result is at the core of @xcite but its exposition relies on tools which are non - standard for statisticians .
we follow here the simpler derivation proposed in @xcite which simply consists of rewriting as @xmath166 where @xmath167 it can be easily checked using that @xmath168 satisfies the following forward recursion for @xmath76 @xmath169 with @xmath170 and where @xmath135 is given by . in practice , we shall approximate the function @xmath171 on a certain grid of values @xmath172 , as explained in the next section .
we can easily provide a particle approximation of the forward smoothing recursion .
assume you have access to approximations @xmath173 of @xmath174 at time @xmath32 , where @xmath175 then when updating our particle filter to obtain @xmath176 we can directly compute the particle approximations @xmath177 by plugging and @xmath147 in ( [ eq : additivesmoothfunctionalsasfunctionoft])-([eq : recursionadditivefunctionalvn ] ) to obtain @xmath178 this approach requires @xmath179 operations to compute @xmath113 at iteration @xmath32 .
a variation over this idea recently proposed in @xcite and @xcite consists of approximating @xmath180 by sampling @xmath181 for @xmath182 to obtain @xmath183 when it is possible to sample from @xmath184 in @xmath153 operations using rejection sampling , provides a monte carlo approximation to of overall complexity @xmath185 .
empirically , for a fixed number of particles , these smoothing procedures perform significantly much better than the naive path space approach to smoothing ( that is , simply propagating forward the complete state trajectory within a particle filtering algorithm ) .
many theoretical results validating these empirical findings have been established under assumption ( [ eq : ergod ] ) and additional regularity assumptions .
the particle estimate of @xmath112 based on the fixed - lag approximation ( [ eq : forgetting ] ) has an asymptotic variance in @xmath186 with a non - vanishing ( as @xmath131 ) bias proportional to @xmath32 and a constant decreasing exponentially fast with @xmath121 @xcite . in @xcite , @xcite , @xcite
, it is shown that when holds , there exists @xmath187 such that the asymptotic bias and variance of the particle estimate of @xmath112 computed using the forward - backward procedures satisfy @xmath188 the bias for the path space and forward - backward estimators of @xmath112 are actually equal @xcite .
recently , it has also been established in @xcite that , under similar regularity assumptions , the estimate also admits an asymptotic variance in @xmath186 whenever @xmath189 .
we describe in this section how the particle filtering and smoothing techniques introduced in sections [ sec : filteringandparticle ] and [ sec : smoothingandparticle ] can be used to implement maximum likelihood parameter estimation techniques .
we recall that @xmath190 denote the log - likelihood function associated to data @xmath19 introduced in section [ sec : computationalissues ] .
so as to maximize @xmath191 one can rely on standard nonlinear optimization methods , e.g. using quasi - newton or gradient - ascent techniques .
we will limit ourselves to these approaches even if they are sensitive to initialization and might get trapped in a local maximum .
we have seen in section [ sec : filteringandparticle ] that @xmath190 can be approximated using particle methods , for any fixed @xmath9 .
one may wish then to treat ml estimation as an optimization problem using monte carlo evaluations of @xmath190 . when optimizing a function calculated with a monte carlo error , a popular strategy is to make the evaluated function continuous by using common random numbers over different evaluations to ease the optimization .
unfortunately , this strategy is not helpful in the particle context .
indeed , in the resampling stage , particles @xmath192 are resampled according to the distribution @xmath193 which admits a piecewise constant and hence discontinuous cumulative distribution function ( cdf ) .
a small change in @xmath15 will cause a small change in the importance weights @xmath194 and this will potentially generate a different set of resampled particles . as a result , the log - likelihood function estimate will not be continuous in @xmath15 even if @xmath190 is continuous . to bypass this problem ,
an importance sampling method was introduced in @xcite but it has computational complexity @xmath158 and only provides low variance estimates in the neighborhood of a suitably preselected parameter value . in the restricted scenario where @xmath195 , an elegant solution to the discontinuity problem was proposed in @xcite .
the method uses common random numbers and introduces a continuous version of the resampling step by finding a permutation @xmath196 such that @xmath197 and defining a piecewise linear approximation of the resulting cdf from which particles are resampled , i.e. @xmath198 this method requires @xmath199 operations due to the sorting of the particles but the resulting continuous estimate of @xmath190 can be maximized using standard optimization techniques .
extensions to the multivariate case where @xmath200 ( with @xmath201 ) have been proposed in @xcite and @xcite
. however , the scheme @xcite does not guarantee continuity of the likelihood function estimate and only provides log - likelihood estimates which are positively correlated for neighboring values in the parameter space whereas the scheme in @xcite has @xmath202 computational complexity and relies on a non - standard particle filtering scheme .
when @xmath15 is high - dimensional , the optimization over the parameter space may be made more efficient if provided with estimates of the gradient .
this is exploited by the algorithms described in the forthcoming sections .
the log - likelihood @xmath190 may be maximized with the following steepest ascent algorithm : at iteration @xmath203 @xmath204 where @xmath205 is the gradient of @xmath206 w.r.t @xmath15 evaluated at @xmath207 and @xmath208 is a sequence of positive real numbers , called the step - size sequence .
typically , @xmath209 is determined adaptively at iteration @xmath210 using a line search or the popular barzilai - borwein alternative .
both schemes guarantee convergence to a local maximum under weak regularity assumptions ; see @xcite for a survey .
the _ score _ vector @xmath211 can be computed by using fisher s identity given in .
given , it is easy to check that the score is of the form .
an alternative to fisher s identity to compute the score is presented in @xcite but this also requires computing an expectation of the form .
these score estimation methods are not applicable in complex scenarios where it is possible to sample from @xmath4 but the analytical expression of this transition kernel is unavailable @xcite . for those models ,
a naive approach is to use a finite difference estimate of the gradient ; however this might generate too high a variance estimate .
an interesting alternative presented in @xcite , under the name of iterated filtering , consists of deriving an approximation of @xmath205 based on the posterior moments @xmath212 of an artificial state - space model with latent markov process @xmath213 , @xmath214 and observed process @xmath215 . here
@xmath216 is a zero - mean white noise sequence with variance @xmath217 , @xmath218 , @xmath219 @xmath220 .
it is shown in @xcite that this approximation improves as @xmath221 and @xmath222 .
clearly as the variance @xmath223 of the artificial dynamic noise @xmath224 on the @xmath225component decreases , it will be necessary to use more particles to approximate @xmath226 as the mixing properties of the artificial dynamic model deteriorates
. gradient ascent algorithms can be numerically unstable as they require to scale carefully the components of the score vector .
the expectation maximization ( em ) algorithm is a very popular alternative procedure for maximizing @xmath206 @xcite . at iteration @xmath203 , we set @xmath227 where @xmath228
the sequence @xmath229 generated by this algorithm is non - decreasing .
the em is usually favored by practitioners whenever it is applicable as it is numerically more stable than gradient techniques . in terms of implementation , the em consists of computing a @xmath230-dimensional summary statistic of the form ( [ eq : expectationadditivefunctionals ] ) when @xmath231 belongs to the exponential family , and the maximizing argument of @xmath232 can be characterized explicitly through a suitable function @xmath233 , i.e. @xmath234 the path space approximation ( [ eq : smcfullposterior ] ) can be used to approximate the score and the summary statistics of the em algorithm at the computational cost of @xmath235 ; see @xcite , @xcite , @xcite .
experimentally the variance of the associated estimates increases typically quadratically with @xmath236 @xcite . to obtain estimates whose variance increases only typically linearly with @xmath236 with similar computational cost
, one can use the fixed - lag approximation presented in section [ sub : fixedlagapproxparticle ] or a more recent alternative where the path space method is used but the additive functional of interest , which is a sum of terms over @xmath237 , is approximated by a sum of similar terms which are now exponentially weighted w.r.t @xmath32 @xcite .
these methods introduce a non - vanishing asymptotic bias difficult to quantify but appear to perform well in practice . to improve over the path space method without introducing any such asymptotic bias , the ffbsm and forward smoothing discussed in sections [ sub : forward - backwardsmoothing ] and [ sub : forward - only - smoothing ] as well as the generalized two - filter smoother
have been used @xcite , @xcite , @xcite , @xcite , @xcite .
experimentally the variance of the associated estimates increases typically linearly with @xmath236 @xcite in agreement with the theoretical results in @xcite , @xcite , @xcite .
however the computational complexity of these techniques is @xmath238 . for a fixed computational complexity of order @xmath239 ,
an informal comparison of the performance of the path space estimate using @xmath240 particles and the forward - backward estimate using @xmath87 particles suggest that both estimates admit a mean square error ( mse ) of order @xmath241 , but the mse of the path space estimate is variance dominated whereas the forward - backward estimates are bias dominated .
this can be understood by decomposing the mse as the sum of the squared bias and the variance and then substituting appropriately for @xmath240 particles in for the path space method and for @xmath87 particles in for the forward - backward estimates .
we confirm experimentally this fact in section [ sub : maximum - likelihood - methods ] .
these experimental results suggest that these particle smoothing estimates might thus of limited interest compared to the path based estimates for ml parameter inference when accounting for computational complexity .
however , this comparison ignores that the @xmath242 computational complexity of these particle smoothing estimates can be reduced to @xmath151 by sampling approximately from @xmath243 with the ffbsa procedure in section [ sub : forward - backwardsmoothing ] or by using fast computational methods @xcite .
related @xmath151 approaches have been developed for generalized two - filter smoothing @xcite , @xcite .
when applicable , these fast computational methods should be favored . for a long observation sequence
the computation of the gradient of @xmath206 can be prohibitive , moreover we might have real - time constraints . an alternative would be a recursive procedure in which the data is run through once sequentially . if @xmath244 is the estimate of the model parameter after the first @xmath32 observations , a recursive method would update the estimate to @xmath245 after receiving the new data @xmath246 . several on - line variants of the ml procedures described earlier are now presented . for these methods to be justified , it is crucial for the observation process to be ergodic for the limiting averaged likelihood function @xmath247 to have a well - defined limit @xmath248 as @xmath249 an alternative to gradient ascent is the following parameter update scheme at time @xmath76 @xmath250 where the positive non - increasing step - size sequence @xmath251 satisfies @xmath252 and @xmath253 @xcite , @xcite ; e.g. @xmath254 for @xmath255 . upon receiving @xmath246 ,
the parameter estimate is updated in the direction of ascent of the conditional density of this new observation . in other words
, one recognizes in the update of the gradient ascent algorithm , except that the partial ( up to time @xmath32 ) likelihood is used .
the algorithm in the present form is however not suitable for on - line implementation , because evaluating the gradient of @xmath256 at the current parameter estimate requires computing the filter from time @xmath257 to time @xmath32 using the current parameter value @xmath244 .
an algorithm bypassing this problem has been proposed in the literature for a finite state - space latent process in @xcite .
it relies on the following update scheme @xmath258 where @xmath259 is defined as @xmath260 with the notation @xmath261 corresponding to a ` time - varying ' score which is computed with a filter using the parameter @xmath262 at time @xmath263 .
the update rule ( [ eq : rml ] ) can be thought of as an approximation to the update rule ( [ eq : update_theta_partial_gradient ] ) .
if we use fisher s identity to compute this ` time - varying ' score , then we have for @xmath264 @xmath265 the asymptotic properties of the recursion ( [ eq : rml ] ) ( i.e. the behavior of @xmath244 in the limit as @xmath32 goes to infinity ) has been studied in @xcite for a finite state - space hmm . it is shown that under regularity conditions this algorithm converges towards a local maximum of the average log - likelihood @xmath248 , @xmath248 being maximized at the ` true ' parameter value under identifiability assumptions .
similar results hold for the recursion ( [ eq : update_theta_partial_gradient ] ) .
it is also possible to propose an on - line version of the em algorithm .
this was originally proposed for finite state - space and linear gaussian models in @xcite , @xcite ; see @xcite for a detailed presentation in the finite state - space case .
assume that @xmath266 is in the exponential family .
in the on - line implementation of em , running averages of the sufficient statistics @xmath267 are computed @xcite , @xcite .
let @xmath268 be the sequence of parameter estimates of the on - line em algorithm computed sequentially based on @xmath269 .
when @xmath246 is received , we compute @xmath270 where @xmath251 needs to satisfy @xmath252 and @xmath253
. then the standard maximization step ( [ eq : maximiem ] ) is used as in the batch version @xmath271 the recursive calculation of @xmath272 is achieved by setting @xmath273 , then computing @xmath274 and finally @xmath275 again , the subscript @xmath276 on @xmath277 indicates that the posterior density is being computed sequentially using the parameter @xmath262 at time @xmath278 .
the filtering density then is advanced from time @xmath279 to time @xmath32 by using @xmath280 , @xmath281 and @xmath282 in the fraction of the rhs of . whereas the convergence of the em algorithm towards a local maximum of the average log - likelihood @xmath248 has been established for i.i.d .
data @xcite , its convergence for state - space models remains an open problem despites empirical evidence it does @xcite , @xcite , @xcite .
this has motivated the development of modified versions of the on - line em algorithm for which convergence results are easier to establish @xcite , @xcite . however , the on - line em presented here performs empirically usually better @xcite .
both the on - line gradient and em procedures require approximating terms ( [ eq : timevaryingscore ] ) and ( [ eq : suffstatonline ] ) of the form ( [ eq : expectationadditivefunctionals ] ) , except that the expectation is now w.r.t the posterior density @xmath277 which is updated using the parameter @xmath262 at time @xmath278 . in this on - line
framework , only the path space , fixed lag smoothing and forward smoothing estimates are applicable , the fixed lag approximation is also applicable but introduces a non - vanishing bias .
for the on - line em algorithm , similarly to the batch case discussed in section [ sub : discussion - of - particle - batch ] , the benefits of using the forward smoothing estimate @xcite compared to the path space estimate @xcite with @xmath240 particles are rather limited , as experimentally demonstrated in section [ sub : maximum - likelihood - methods ] .
however for the on - line gradient ascent algorithm , the gradient term @xmath259 in ( [ eq : rml ] ) is a difference between two score - like vectors ( [ eq : timevaryingscore ] ) and the behavior of its particle estimates differs significantly from its em counterpart .
indeed the variance of the particle path estimate of @xmath259 increases linearly with @xmath32 , yielding an unreliable gradient ascent procedure , whereas the particle forward smoothing estimate has a variance uniformly bounded in time under appropriate regularity assumptions and yields a stable gradient ascent procedure @xcite . hence the use of a procedure of computational complexity @xmath202
is clearly justified in this context .
the very recent paper @xcite reports that the computationally cheaper estimate ( [ eq : olssonmcapproximation ] ) appears to exhibit similar properties whenever @xmath189 and might prove an attractive alternative .
in the bayesian setting , we assign a suitable prior density @xmath283 for @xmath15 and inference is based on the joint posterior density @xmath284 in the off - line case , or the sequence of posterior densities @xmath285 in the on - line case .
using mcmc is a standard approach to approximate @xmath284 . unfortunately designing efficient mcmc sampling algorithms for non - linear non - gaussian state - space models
is a difficult task : one - variable - at - a - time gibbs sampling typically mixes very poorly for such models , whereas blocking strategies that have been proposed in the literature are typically very model - dependent ; see for instance @xcite .
particle mcmc are a class of mcmc techniques which rely on particle methods to build efficient high dimensional proposal distributions in a generic manner @xcite .
we limit ourselves here to the presentation of the particle marginal metropolis hastings ( pmmh ) sampler , which is an approximation of an ideal mmh sampler for sampling from @xmath284 which would utilize the following proposal density @xmath286 where @xmath287 is a proposal density to obtain a candidate @xmath288 when we are at location @xmath15 .
the acceptance probability of this sampler is @xmath289 unfortunately this ideal algorithm can not be implemented as we can not sample exactly from @xmath290 and we can not compute the likelihood terms @xmath291 and @xmath292 appearing in the acceptance probability .
the pmmh sampler is an approximation of this ideal mmh sampler which relies on the particle approximations of these unknown terms .
given @xmath15 and a particle approximation @xmath293 of @xmath291 , we sample @xmath294 then run a particle filter to obtain approximations @xmath295 and @xmath296 of @xmath297 and @xmath292 .
we then sample @xmath298 , that is we choose randomly one of @xmath87 particles generated by the particle filter , with probability @xmath299 for particle @xmath300 , and accept @xmath301 ( and @xmath296 ) with probability @xmath302 the acceptance probability ( [ eq : acceptprobapmmh ] ) is a simple approximation of the ideal acceptance probability ( [ eq : acceptprobamh ] ) .
this algorithm was first proposed as an heuristic to sample from @xmath303 in @xcite .
its remarkable feature established in @xcite is that it does admit @xmath284 as invariant distribution whatever being the number of particles @xmath87 used in the particle approximation @xcite .
however the choice of @xmath87 has an impact on the performance of the algorithm . using large values of @xmath87
usually results in pmmh averages with variances lower than the corresponding averages using fewer samples but the computational cost of constructing @xmath293 increases with @xmath87 .
a simplified analysis of this algorithm suggests that @xmath87 should be selected such that the standard deviation of the logarithm of the particle likelihood estimate should be around @xmath304 if the ideal mmh sampler was using the perfect proposal @xmath305 @xcite and around @xmath306 if one uses an isotropic normal random walk proposal , the target is a product of @xmath307 i.i.d .
components and @xmath308 @xcite . for general proposal and target densities , a recent theoretical analysis and empirical results suggest that this standard deviation should be selected around @xmath309 @xcite .
as the variance of this estimate typically increases linearly with @xmath236 , this means that the computational complexity is of order @xmath310 by iteration .
a particle version of the gibbs sampler is also available @xcite which mimicks the two - component gibbs sampler sampling iteratively from @xmath311 and @xmath119 .
these algorithms rely on a non - standard version of the particle filter where @xmath312 particles are generated conditional upon a `` fixed '' particle .
recent improvements over this particle gibbs sampler introduce mechanisms to rejuvenate the fixed particle , using forward or backward sampling procedures @xcite , @xcite , @xcite .
these methods perform empirically extremely well but , contrary to the pmmh , it is still unclear how one should scale @xmath87 with @xmath236 . in this context , we are interested in approximating on - line the sequence of posterior densities @xmath285 .
we emphasize that , contrary to the on - line ml parameter estimation procedures , none of the methods presented in this section bypass the particle degeneracy problem .
this should come as no surprise .
as discussed in section [ sec : convergenceparticle ] , even for a _ fixed _
@xmath15 , the particle estimate of @xmath21 has a relative variance that increases linearly with @xmath32 under favorable mixing assumptions . the methods in this section attempt to approximate @xmath313 .
this is a harder problem as it implicitly requires having to approximate @xmath314 for all the particles @xmath315 approximating @xmath316 . at first sight
, it seems that estimating the sequence of posterior densities @xmath285 can be easily achieved using standard particle methods , by merely introducing the extended state @xmath317 , with initial density @xmath318 and transition density @xmath319 i.e. @xmath320 .
however , this extended process @xmath321 clearly does not possess any _ forgetting _ property ( as discussed in section [ sec : filteringandparticle ] ) , so the algorithm is bound to degenerate .
specifically , the parameter space is explored only in the initial step of the algorithm .
then , each successive resampling step reduces the diversity of the sample of @xmath15 values ; after a certain time @xmath32 , the approximation @xmath322 contains a single unique value for @xmath15 .
this is clearly a poor approach . even in the much simpler case
when there is no latent variable @xmath18 , it is shown in ( * ? ? ?
* theorem 4 ) that the asymptotic variance of the corresponding particle estimates diverges at least at a polynomial rate , which grows with the dimension of @xmath15
. a pragmatic approach that has proven useful in some applications is to introduce artificial dynamics for the parameter @xmath15
@xcite @xmath323 where @xmath324 is an artificial dynamic noise with decreasing variance .
standard particle methods can now be applied to approximate @xmath325 .
a related kernel density estimation method also appeared in @xcite , which proposes to use a kernel density estimate @xmath316 from which one samples from .
as before the static parameter is transformed to a slowly time - varying one , whose dynamics is related to the kernel bandwidth . to mitigate the artificial variance inflation , a shrinkage correction is introduced .
an improved version of this method has been recently proposed in @xcite .
it is difficult to quantify how much bias is introduced in the resulting estimates by the introduction of this artificial dynamics .
additionally , these methods require a significant amount of tuning , e.g. choosing the variance of the artificial dynamic noise or the kernel width .
however they can perform satisfactorily in practice @xcite , @xcite .
the practical filtering approach proposed in @xcite relies on the following fixed - lag approximation @xmath326 for @xmath121 large enough ; that is observations coming after @xmath279 presumably brings little information on @xmath327 . to sample
approximately from @xmath328 , one uses the following iterative process : at time @xmath32 , several mcmc chains are run in parallel to sample from @xmath329 where the @xmath330 have been obtained at the previous iteration , and are such that ( approximately ) @xmath331 .
then one collects the first component @xmath332 of the simulated sample @xmath333 , increments the time index and runs several new mcmc chains in parallel to sample from @xmath334 and so on .
the algorithm is started at time @xmath335 , with mcmc chains that target @xmath336 .
like all methods based on fixed - lag approximation , the choice of the lag @xmath121 is difficult and this introduces a non - vanishing bias which is difficult to quantify .
however , the method performs well on the examples presented in @xcite . to avoid the introduction of an artificial dynamic model or of a fixed - lag approximation , an approach originally proposed independently in @xcite and @xcite
consists of adding mcmc steps to re - introduce diversity among the particles .
assume we use an auxiliary particle filter to approximate @xmath285 then the particles @xmath337 obtained after the sampling step at time @xmath32 are approximately distributed according to @xmath338 we have @xmath339 if @xmath66 and @xmath67 . to add diversity in this population of particles ,
we introduce an mcmc kernel @xmath340 with invariant density @xmath341 , and replace , at the end of each iteration , the set of resampled particles , @xmath342 with @xmath87 `` mutated '' particles @xmath343 simulated from , for @xmath344 @xmath345 if we use the sisr algorithm , then we can alternatively use an mcmc step of invariant density @xmath346 after the resampling step at time @xmath32 .
contrary to standard applications of mcmc , the kernel does not have to be ergodic .
ensuring ergodicity would indeed require one to sample an increasing number of variables as @xmath32 increases this algorithm would have an increasing cost per iteration , which would prevents its use in on - line scenarios but it can be an interesting alternative to standard mcmc and was suggested in @xcite . in practice one
therefore sets @xmath347 and only sample @xmath348 and @xmath349 , where @xmath121 is a small integer ; often @xmath350 ( only @xmath15 is updated ) .
note that the memory requirements for this method do not increase over time if @xmath351 is in the exponential family and thus can be summarized by a set of fixed dimensional sufficient statistics @xmath352 .
this type of methods was first used in to perform on - line bayesian parameter estimation in a context where @xmath351 is in the exponential family @xcite , @xcite .
similar strategies were adopted in @xcite and @xcite . in the particular scenario where @xmath66 and @xmath67 , this method was mentioned in @xcite , @xcite and is discussed at length in @xcite who named it particle learning .
extensions of this strategy to parameter estimation in conditionally linear gaussian models , where a part of the state is integrated out using kalman techniques @xcite , @xcite , is proposed in @xcite .
as opposed to the methods relying on kernel or artificial dynamics , these mcmc - based approaches have the advantage of adding diversity to the particles approximating @xmath316 without perturbing the target distribution . unfortunately , these algorithms rely implicitly on the particle approximation of the density @xmath353 even if algorithmically it is only necessary to store some fixed - dimensional sufficient statistics @xmath354 . hence in this respect
they suffer from the degeneracy problem .
this was noticed as early as in @xcite ; see also the word of caution in the conclusion of @xcite , @xcite and @xcite .
the practical implications are that one observes empirically that the resulting monte carlo estimates can display quite a lot of variability over multiple runs as demonstrated in section [ sub : bayesianmethodsexperiments ] .
this should not come as a surprise as the sequence of posterior distributions does not have exponential forgetting properties , hence there is an accumulation of monte carlo errors over time .
the smc@xmath355 algorithm introduced simultaneously in @xcite and @xcite may be considered as the particle equivalent of particle mcmc .
it mimics an `` ideal '' particle algorithm proposed in @xcite approximating sequentially @xmath356 where @xmath357 particles ( in the @xmath15-space ) are used to explore these distributions .
the @xmath357 particles at time @xmath32 are reweighted according to @xmath358 at time @xmath164 .
as these likelihood terms are unknown , we substitute to them @xmath359 where @xmath360 is a particle approximation of the partial likelihood @xmath35 , obtained by a running a particle filter of @xmath361 particles in the @xmath362dimension , up to time @xmath32 , for each of the @xmath357 @xmath225particles . when particle degeneracy ( in the @xmath225dimension ) reaches a certain threshold , @xmath225particles are refreshed through the succession of a resampling step , and an mcmc step , which in these particular settings takes the form of a pmcmc update .
the cost per iteration of this algorithm is not constant and , additionally , it is advised to increase @xmath361 with @xmath32 for the relative variance of @xmath360 not to increase , therefore it can not be used in truly on - line scenarios .
yet there are practical situations where it may be useful to approximate jointly all the posteriors @xmath363 , for @xmath364 , for instance to assess the predictive power of the model .
we focus on illustrating numerically a few algorithms and the impact of the degeneracy problem on parameter inference .
this last point is motivated by the fact that particle degeneracy seems to have been overlooked by many practitioners . in this way
numerical results may provide valuable insights .
we will consider the following simple scalar linear gaussian state space model : @xmath365 where @xmath366 are independent zero - mean and unit - variance gaussians and @xmath367 $ ] .
the main reason for choosing this model is that kalman filter recursions can be implemented to provide the exact values of the summary statistics @xmath112 used for ml estimation through the em algorithm , and to compute the exact likelihood @xmath21 .
hence , using a fine discretization of the low - dimensional parameter space , we can compute a very good approximation of the true posterior density @xmath328 . in this model
it is straightforward to present numerical evidence of some effects of degeneracy for parameter estimation and show how it can be overcome by choosing an appropriate particle method .
( top panel ) , empirical variance ( middle panel ) and mse ( bottom panel ) for the estimate of @xmath368 . left column : @xmath369 method using @xmath370 particles .
right column : @xmath242 method using @xmath371 particles . in every subplot
, the top line corresponds to using @xmath372 the middle for @xmath373 and the lower for @xmath374,scaledwidth=100.0% ] as ml methods require approximating smoothed additive functionals @xmath112 of the form ( [ eq : expectationadditivefunctionals ] ) , we begin by investigating the empirical bias , variance and mse of two standard particle estimates of @xmath112 , where we set @xmath375 for the model described in .
the first estimate relies on the path space method with computational cost @xmath369 per time , which uses @xmath110 in ( [ eq : smcfullposterior ] ) to approximate @xmath112 as @xmath113 ; see ( * ? ? ?
* section 8.3 ) for more details .
the second estimate relies on the forward implementation of ffbsm presented in section [ sub : forward - only - smoothing ] using ( [ eq : additivesmoothfunctionalsasfunctionoft])-([eq : smcapproxadditivefunctionals-1 ] ) ; see @xcite . recall that this procedure has a computational cost that is @xmath242 per time for @xmath87 particles and provides the same estimates as the standard forward - backward implementation of ffbsm . for the sake of brevity
we will not consider the remaining smoothing methods of section [ sec : smoothingandparticle ] ; for the fixed - lag and the exponentially weighted approximations we refer the reader to @xcite respectively @xcite for numerical experiments .
we use a simulated dataset of size @xmath376 obtained using @xmath377 and then generate 300 independent replications of each method in order to compute the empirical bias and variance of @xmath378 when @xmath15 is fixed to @xmath379 . in order to make a comparison that takes into account the computational cost , we use @xmath240 particles for the @xmath369 method and @xmath87 for the @xmath242 one .
we look separately at the behavior of the bias of @xmath380 and the variance and mse of the rescaled estimates @xmath381 .
the results are presented in figure [ fig : smoothing ] for @xmath371 . for both methods the bias grows linearly with time , this growth being higher for the @xmath242 method . for the variance of @xmath381 ,
we observe a linear growth with time for the @xmath369 method with @xmath240 particles whereas this variance appears roughly constant for the @xmath242 method .
finally , the mse of @xmath381 grows for both methods linearly as expected . in this particular scenario ,
the constants of proportionality are such that the mse is lower for the @xmath369 method than for the @xmath242 method . in general
, we can expect that the @xmath369 method be superior in terms of the bias and the @xmath242 method superior in terms of the variance .
these results are in agreement with the theoretical results in the literature @xcite , @xcite , @xcite but additionally show that the lower bound on the variance growth of @xmath380 for the @xmath369 method of @xcite appears sharp . for various @xmath236 using 25 iterations of off - line em and 150 realizations of the algorithms .
top panels : @xmath369 method using @xmath382 particles .
bottom panels : @xmath242 with @xmath383 .
the dotted horizontal lines are the ml estimate for each time @xmath236 obtained using kalman filtering on a grid.,scaledwidth=105.0% ] we proceed to see how the bias and variance of the estimates of @xmath112 affect the ml estimates , when the former are used within both an off - line and an on - line em algorithm ; see figures [ fig : offlineem ] and [ fig : onlineem ] respectively .
for the model in ( [ eq : dlm1 ] ) the e - step corresponds to computing @xmath112 where @xmath384 and the m - step update function is given by @xmath385 we compare the estimates of @xmath379 when the e - step is computed using the @xmath369 and the @xmath242 methods described in the previous section with @xmath386 and @xmath387 particles respectively .
a simulated dataset for @xmath388 will be used . in every case
we will initialize the algorithm using @xmath389 and assume @xmath390 is known . in figures [ fig : offlineem ] and [ fig : onlineem ] we present the results obtained using 150 independent replications of the algorithm . for the off - line em , we use @xmath391 iterations for @xmath392 . for the on - line em
, we use @xmath393 with the step size set as @xmath394 and for the first @xmath395 iterations no m - step update is performed .
this `` freezing '' phase is required to allow for a reasonable estimation of the summary statistic ; see @xcite , @xcite for more details .
note that in figure [ fig : onlineem ] we plot only the results after the algorithm has converged , i.e. for @xmath396 . in each case , both the @xmath369 and the @xmath242 methods yield fairly accurate results given the low number of particles used .
however we note , as observed previously in the literature , that the on - line em as well as the on - line gradient ascent method requires a substantial number of observations , i.e. over 10000 , before achieving convergence @xcite , @xcite , @xcite , @xcite . for smaller datasets ,
these algorithms can also be used by going through the data say @xmath397 times .
typically this method is cheaper than iterating ( [ eq : batchgradient ] ) or ( [ eq : qfunction])-([eq : maximiem ] ) @xmath397 times the off - line algorithms and can yield comparable parameter estimates @xcite .
experimentally , the properties of the estimates of @xmath112 discussed earlier appear to translate into properties of the resulting parameter estimates : the @xmath369 method provides estimates with less bias but more variance than the @xmath242 method . for @xmath398 using 150 realizations of the algorithms .
we also plot the ml estimate at time @xmath32 obtained using kalman filtering on a grid ( black).,scaledwidth=105.0% ] for more numerical examples regarding the remaining methods discussed in section [ sec : mlestimation ] , we refer the reader to @xcite , @xcite for iterated filtering , to @xcite , @xcite , @xcite for comparisons of the @xmath369 and @xmath242 methods for em and gradient ascent , to @xcite for the @xmath369 on - line em , to @xcite and ( * ? ? ?
10 ) for smooth likelihood function methods and to ( * ? ? ?
10 - 11 ) for a detailed exposition of off - line em methods .
we still consider the model in ( [ eq : dlm1 ] ) but simplify it further by fixing either @xmath399 or @xmath400 .
this is done in order to keep the computations of the benchmarks that use kalman computations on a grid relatively inexpensive .
for those parameters that are not fixed , we shall use the following independent priors : a uniform on @xmath401 $ ] for @xmath399 , and inverse gamma for @xmath402 with the shape and scale parameter pair being @xmath403 and @xmath404 respectively with @xmath405 .
in all the subsequent examples , we will initialize the algorithms by sampling @xmath15 from the prior . at @xmath406 .
top right : relative variance , i.e. empirical variance ( over independent runs ) for the estimator of the mean of @xmath407 using particle method with mcmc normalized with the true posterior variance computed using kalman filtering on a grid .
bottom left : average ( over independent runs ) of the estimated variance of @xmath407 using particle method with mcmc normalized with the true posterior variance .
bottom right : variance of the @xmath408 ; all plots are computed using @xmath409 and over 100 different independent runs.,title="fig:",scaledwidth=50.0% ] at @xmath406 .
top right : relative variance , i.e. empirical variance ( over independent runs ) for the estimator of the mean of @xmath407 using particle method with mcmc normalized with the true posterior variance computed using kalman filtering on a grid .
bottom left : average ( over independent runs ) of the estimated variance of @xmath407 using particle method with mcmc normalized with the true posterior variance .
bottom right : variance of the @xmath408 ; all plots are computed using @xmath409 and over 100 different independent runs.,title="fig:",scaledwidth=50.0% ] + at @xmath406 .
top right : relative variance , i.e. empirical variance ( over independent runs ) for the estimator of the mean of @xmath407 using particle method with mcmc normalized with the true posterior variance computed using kalman filtering on a grid .
bottom left : average ( over independent runs ) of the estimated variance of @xmath407 using particle method with mcmc normalized with the true posterior variance .
bottom right : variance of the @xmath408 ; all plots are computed using @xmath409 and over 100 different independent runs.,title="fig:",scaledwidth=50.0% ] at @xmath406 .
top right : relative variance , i.e. empirical variance ( over independent runs ) for the estimator of the mean of @xmath407 using particle method with mcmc normalized with the true posterior variance computed using kalman filtering on a grid .
bottom left : average ( over independent runs ) of the estimated variance of @xmath407 using particle method with mcmc normalized with the true posterior variance .
bottom right : variance of the @xmath408 ; all plots are computed using @xmath409 and over 100 different independent runs.,title="fig:",scaledwidth=50.0% ] we proceed to examine the combination of particle method with mcmc methods described in section [ sub : using - mcmcsteps - within - smc ] .
we focus on an efficient implementation of this idea discussed in @xcite which can be put in practice for the simple model under consideration .
we investigate the effect of the degeneracy problem in this context .
the numerical results obtained in this section have been produced in matlab ( code available from the first author ) , and double - checked using the r program available on the personal web page of the first author of @xcite .
we first focus of the estimate of the posterior of @xmath410 given a long sequence of simulated observations with @xmath411 . in this scenario , @xmath266 admits the following two - dimensional sufficient statistics , @xmath412 , and @xmath15 can be updated using gibbs steps .
we use @xmath413 and @xmath409 .
we ran the algorithm over 100 independent runs over the same dataset .
we present the results only for @xmath414 and omit the ones for @xmath223 as these were very similar .
the top left panel of figure [ flo : relvar ] shows the box plots for the estimates of the posterior mean , and the top right panel shows how the corresponding relative variance of the estimator for the posterior mean evolves with time . here
the relative variance is defined as the ratio of the empirical variance ( over different independent runs ) of the posterior mean estimates at time @xmath32 over the true posterior variance at time @xmath32 , which in this case is approximated using a kalman filter on a fine grid .
this quantity exhibits a steep increasing trend when @xmath415 and confirms the aforementioned variability of the estimates of the posterior mean . in the bottom left panel of figure [ flo : relvar ]
we plot the average ( over different runs ) of the estimators of the variance of @xmath416 .
this average variance is also scaled / normalized by the actual posterior variance .
the latter is again computed using kalman filtering on a grid .
this ratio between the average estimated variance of the posterior over the true one decreases with time @xmath32 and it shows that the supports of the approximate posterior densities provided by this method cover on average only a small portion of the support of the true posterior .
these experiments confirm that in this example the particle method with mcmc steps fails to adequately explore the space of @xmath15 .
although the box plots provide some false sense of security , the relative and scaled average variance clearly indicate that any posterior estimates obtained from a single run of particle method with mcmc steps should be used with caution .
furthermore , in the the bottom right panel of figure [ flo : relvar ] we also investigate experimentally the empirical variance of the marginal likelihood estimates @xmath408 .
this variance appears to increase quadratically with @xmath32 for the particle method with mcmc moves instead of linearly as it does for state - space models with good mixing properties .
this suggests that to one should increase the number of particles quadratically with the time index to obtain an estimate of the marginal likelihood whose relative variance remains uniformly bounded with respect to the time index .
although we attribute this quadratic variance growth to the degeneracy problem , the estimate @xmath417 is not the particle approximation of a smoothed additive functional , thus there is not yet any theoretical convergence result explaining rigorously this phenomenon . ; estimated marginal posterior densities for @xmath418 over 50 runs ( red ) versus ground truth ( blue ) .
, title="fig:",scaledwidth=95.0% ] one might argue that these particle methods with mcmc moves are meant to be used with larger @xmath87 and/or shorter data sets @xmath236 .
we shall consider this time a slightly different example where @xmath419 is known and we are interested in estimating the posterior of @xmath420 given a sequence of observations obtained using @xmath421 and @xmath422 . in that case , the sufficient statistics are @xmath423 , and the parameters can be rejuvenated through a single gibbs update .
in addition , we let @xmath424 and use @xmath425 particles . in figures [ flo : dlm_drift_snapshot ] we display the estimated marginal posteriors @xmath426 and @xmath427 obtained from @xmath395 independent replications of the particle method . on this simple problem ,
the estimated posteriors seem consistently rather inaccurate for @xmath399 , whereas they perform better for @xmath223 but with some non - negligible variability over runs which increases as @xmath236 increases .
similar observations have been reported in @xcite and remain unexplained : for some parameters this methodology appears to provide reasonable results despite the degeneracy problem and for others it provides very unreliable results .
we investigate further the performance of this method in this simple example by considering the same example for @xmath428 but now consider two larger numbers of particles , @xmath429 and @xmath430 , over 50 different runs .
additionally we compare the resulting estimates with estimates provided by the particle gibbs sampler of @xcite using the same computational cost , that is @xmath431 particles with @xmath432 and @xmath433 iterations respectively .
the results are displayed in figure [ flo : compare_low ] and [ flo : compare_high ] .
as expected , we improve the performance of the particle with mcmc moves when @xmath87 increases for a fixed time horizon @xmath236 . for a fixed computational complexity ,
the particle gibbs sampler estimates appear to display less variability . for a higher dimensional parameter @xmath15 and/or very vague priors
, this comparison would be more favorable to the particle gibbs sampler as illustrated in @xcite . with @xmath434 over 50 runs ( black - dotted ) versus ground truth ( green ) .
top : particle method with mcmc , @xmath429 .
bottom : particle gibbs with @xmath432 iterations and @xmath431.,title="fig:",scaledwidth=90.0% ] with @xmath434 over 50 runs ( black - dotted ) versus ground truth ( green ) .
top : particle method with mcmc , @xmath430 .
bottom : particle gibbs with @xmath433 iterations and @xmath431.,title="fig:",scaledwidth=90.0% ]
most particle methods proposed originally in the literature to perform inference about static parameters in general state - space models were computationally inefficient as they suffered from the degeneracy problem .
several approaches have been proposed to deal with this problem by either adding an artificial dynamic on the static parameter @xcite , @xcite , @xcite or introducing a fixed - lag approximation @xcite , @xcite , @xcite .
these methods can work very well in practice but it remains unfortunately difficult / impossible to quantify the bias introduced in most realistic applications .
various asymptotically bias - free methods with good statistical properties and a reasonable computational cost have recently appeared in the literature . to perform batch ml estimation , the forward filter backward sampler / smoother and generalized two filter procedures
are recommended whenever the @xmath435 computational complexity per iteration of their direct implementations can be lowered to @xmath436 using , for example , the methods described in @xcite , @xcite , @xcite , @xcite . otherwise , besides a lowering of memory requirements , not much can be gained from these techniques compared to simply using a standard particle filter with @xmath240 particles . in an on - line ml context ,
the situation is markedly different .
whereas for the on - line em algorithm , the forward smoothing approach in @xcite , @xcite of complexity @xmath242 per time step will be similarly of limited interest compared to a standard particle filter using @xmath240 particles , it is crucial to use this approach when performing on - line gradient ascent as demonstrated empirically and established theoretically in @xcite . in on - line scenarios where one can admit a random computational complexity at each time step ,
the method presented in @xcite is an interesting alternative when it is applicable .
empirically these on - line ml methods converge rather slowly and will be primarily be useful for large datasets . in a bayesian framework
, batch inference can be conducted using particle mcmc methods @xcite , @xcite .
however these methods are computationally expensive as , for example , an efficient implementation of the pmmh has a computational complexity of order @xmath310 per iteration @xcite . on - line
bayesian inference remains a challenging open problem as all methods currently available , including particle methods with mcmc moves @xcite , @xcite , @xcite suffer from the degeneracy problem .
these methods should not be ruled out but should be used cautiously as they can provide unreliable results even in simple scenarios as demonstrated in our experiments .
very recent papers in this dynamic research area have proposed to combine individual parameter estimation techniques so as to design more efficient inference algorithms .
for example , @xcite suggests to use the score estimation techniques developed for ml parameter estimation to design better proposal distributions for the pmmh algorithm whereas @xcite demonstrates that particle methods with mcmc moves might be fruitfully used in batch scenarios when plugged into a particle mcmc scheme .
n. kantas was supported by the engineering and physical sciences research council ( epsrc ) under grant ep / j01365x/1 and programme grant on control for energy and sustainability ( ep / g066477/1 ) .
singh s research is partly funded by epsrc under the first grant scheme ( ep / g037590/1 ) .
a. doucet s research is partly funded by epsrc ( ep / k000276/1 and ep / k009850/1 ) .
n. chopin s research is partly by the anr as part of the investissements davenir program ( anr-11-labex-0047 ) .
del moral , p. and doucet , a. and singh , s.s .
uniform stability of a particle approximation of the optimal filter derivative .
preprint arxiv:1106.2525 .
_ siam j. control optimization _ , to appear .
doucet , a. , pitt , m.k . ,
deligiannidis , g. and kohn , r. ( 2012 ) .
efficient implementation of markov chain monte carlo when using an unbiased likelihood estimator .
preprint arxiv:1210.1871 .
_ biometrika _ , to appear .
ford , j.j .
adaptive hidden markov model estimation and applications .
phd thesis , department of systems engineering , australian national university .
available at http://infoeng.rsise.anu.edu.au/files/jason_ford_thesis.pdf lopes , h.f .
, carvalho , c. , johannes , m. and polson , n. ( 2010 ) .
particle learning for sequential bayesian computation . in _ bayesian statistics
_ 9 ( bernardo et al .
eds ) , oxford university press , to appear .
nemeth , c. , fearnhead , p. and mihaylova , l. ( 2013 ) .
particle approximations of the score and observed information matrix for parameter estimation in state space models with linear computational cost .
preprint arxiv:1306.0735 .
paninski , l. , ahmadian , y. , ferreira , d.g . ,
koyama , s. , rad .
s.r , vidne , m. , vogelstein , j. and wu , w. ( 2010 ) . a new look at state - space models for neural data . _
j. computational neuroscience _ , 29 , 107126 .
poyiadjis , g. , doucet , a. and singh , s.s .
particle approximations of the score and observed information matrix in state - space models with application to parameter estimation . _
biometrika _ , 98 , 6580 .
vercauteren , t. , toledo , a. and wang , x. ( 2005 ) .
online bayesian estimation of hidden markov models with unknown transition matrix and applications to ieee 802.11 networks . _ proc .
ieee icassp _ ,
iv , 1316 .
|
nonlinear non - gaussian state - space models are ubiquitous in statistics , econometrics , information engineering and signal processing .
particle methods , also known as sequential monte carlo ( smc ) methods , provide reliable numerical approximations to the associated state inference problems .
however , in most applications , the state - space model of interest also depends on unknown static parameters that need to be estimated from the data . in this context ,
standard particle methods fail and it is necessary to rely on more sophisticated algorithms . the aim of this paper is to present a comprehensive review of particle methods that have been proposed to perform static parameter estimation in state - space models .
we discuss the advantages and limitations of these methods and illustrate their performance on simple models . , , ,
|
given a domain @xmath0 in @xmath1 , @xmath2 , with a sufficiently regular boundary @xmath3 , let @xmath4 be a relatively open subset of @xmath5 ( in its relative topology ) , and set @xmath6 .
we consider the solution of the problem @xmath7,\\ \label{eq : signor - v-2 } v\geq { \varphi},\quad \partial_\nu v\geq 0,\quad ( v-{\varphi})\partial_\nu v=0&\quad\text{on } { \mathcal{m}}_t:={\mathcal{m}}\times(0,t],\\ \label{eq : signor - v-3 } v = g&\quad\text{on } { \mathcal{s}}_t:={\mathcal{s}}\times(0,t],\\ v(\cdot , 0)={\varphi}_0&\quad\text{on } \omega_0:=\omega\times\{0\ } , \label{eq : signor - v-4}\end{aligned}\ ] ] where @xmath8 is the outer normal derivative on @xmath3 , and @xmath9 , @xmath10 , and @xmath11 are prescribed functions satisfying the compatibility conditions : @xmath12 on @xmath13 , @xmath14 on @xmath15 $ ] , and @xmath16 on @xmath17 , see fig . [
fig : par - sig - prob ] .
the condition is known as the _ signorini boundary condition _ and the problem as the _ ( parabolic ) signorini problem _ for the heat equation .
the function @xmath18 is called the _ thin obstacle _ , since @xmath19 is restricted to stay above @xmath18 on @xmath20 .
classical examples where signorini - type boundary conditions appear are the problems with unilateral constraints in elastostatics ( including the original signorini problem @xcite ) , problems with semipermeable membranes in fluid mechanics ( including the phenomenon of osmosis and osmotic pressure in biochemistry ) , and the problems on the temperature control on the boundary in thermics .
we refer to the book of duvaut and lions @xcite , where many such applications are discussed and the mathematical models are derived .
another historical importance of the parabolic signorini problem is that it serves as one of the prototypical examples of evolutionary _ variational inequalities _ , going back to the foundational paper by lions and stampacchia @xcite , where the existence and uniqueness of certain weak solutions were established . in this paper
, we work with a stronger notion of solution .
thus , we say that a function @xmath21 solves if @xmath22 where @xmath23 .
the reader should see section [ sec : parab - funct - class ] for the definitions of the relevant parabolic functional classes .
the existence and uniqueness of such @xmath19 , under some natural assumptions on @xmath18 , @xmath24 , and @xmath25 can be found in @xcite .
see also section [ sec : existence - regularity ] for more details .
two major questions arise in the study of the problem : * the regularity properties of @xmath19 ; * the structure and regularity of the _ free boundary _
@xmath26 where @xmath27 indicates the boundary in the relative topology of @xmath20 .
( 150,150 ) ( 0,0 ) ( 84,130)@xmath28 ( 58,110 ) ( 80,95 ) ( 75,80)@xmath29 ( 25,75 ) ( 20,60 ) ( 80,30)@xmath30 ( 140,90 ) @xmath31 ( 88,59)@xmath32 ( 87,62)(-2,1)10 concerning the regularity of @xmath19 , it has long been known that the spatial derivatives @xmath33 , @xmath34 , are @xmath35-hlder continuous on compact subsets of @xmath36 , for some unspecified @xmath37 . in the time - independent case ,
such regularity was first proved by richardson @xcite in dimension @xmath38 , and by caffarelli @xcite for @xmath39 . in the parabolic case ,
this was first proved by athanasopoulos @xcite , and subsequently by uraltseva @xcite ( see also @xcite ) , under certain regularity assumptions on the boundary data , which were further relaxed by arkhipova and uraltseva @xcite .
we note that the hlder continuity of @xmath33 is the best regularity one should expect for the solution of . this can be seen from the example @xmath40 which is a harmonic function in @xmath41 , and satisfies the signorini boundary conditions on @xmath42 , with thin obstacle @xmath43 .
this example also suggests that the optimal hlder exponent for @xmath33 should be @xmath44 , at least when @xmath4 is flat ( contained in a hyperplane ) .
indeed , in the time - independent case , such optimal @xmath45 regularity was proved in dimension @xmath38 in the cited paper by richardson @xcite .
the case of arbitrary space dimension ( still time - independent ) , however , had to wait for the breakthrough work of athanasopoulos and caffarelli @xcite .
very recently , the proof of the optimal regularity for the original signorini problem in elastostatics was announced by andersson @xcite .
one of the main objectives of this paper is to establish , in the parabolic signorini problem , and for a flat thin manifold @xmath4 , that @xmath46 , or more precisely that @xmath47 , see theorem [ thm : opt - reg ] below .
our approach is inspired by the works of athanasopoulos , caffarelli , and salsa @xcite and caffarelli , salsa , and silvestre @xcite on the time - independent problem .
in such papers a generalization of the celebrated almgren s frequency formula established in @xcite was used , not only to give an alternative proof of the optimal @xmath45 regularity of solutions , but also to study the so - called _ regular set _ @xmath48 of the free boundary @xmath32 .
this approach was subsequently refined in @xcite by the second and third named authors with the objective of classifying the free boundary points according to their separation rate from the thin obstacle @xmath18 .
in @xcite the authors also introduced generalizations of weiss s and monneau s monotonicity formulas , originally developed in @xcite and @xcite , respectively , for the classical obstacle problem .
such generalized weiss s and monneau s monotonicity formulas allowed to prove a structural theorem on the so - called _ singular set _ @xmath49 of the free boundary , see @xcite . for an exposition of these results in the case
when the thin obstacle @xmath43 we also refer to the book by the third named author , shahgholian , and uraltseva @xcite*chapter 9 .
in closing we mention that , as far as we are aware of , the only result presently available concerning the free boundary in the parabolic setting is that of athanasopoulos @xcite , under assumptions on the boundary data that guarantee boundedness and nonnegativity of @xmath50 . in that paper
it is shown that the free boundary is locally given as a graph @xmath51 for a lipschitz continuous function @xmath52 . in this paper
we extend all the above mentioned results from the elliptic to the parabolic case .
we focus on the situation when the principal part of the diffusion operator is the laplacian and that the thin manifold @xmath4 is flat and contained in @xmath53 .
one of our central results is a generalization of almgren s frequency formula @xcite , see theorem [ thm : thin - monotonicity ] below .
as it is well known , the parabolic counterpart of almgren s formula was established by poon @xcite , for functions which are caloric in an infinite strip @xmath54 $ ] . in section [ sec : gener - freq - funct ] , we establish a truncated version of that formula for the solutions of the parabolic signorini problem , similar to the ones in @xcite and @xcite . the time dependent case presents , however , substantial novel challenges with respect to the elliptic setting .
these are mainly due to the lack of regularity of the solution in the @xmath55-variable , a fact which makes the justification of differentiation formulas and the control of error terms quite difficult .
to overcome these obstructions , we have introduced ( steklov - type ) averaged versions of the quantities involved in our main monotonicity formulas .
this basic idea has enabled us to successfully control the error terms .
similarly to what was done in @xcite , we then undertake a systematic classification of the free boundary points based on the limit at the point in question of the generalized frequency function .
when the thin obstacle @xmath18 is in the class @xmath56 , @xmath57 , this classification translates into assigning to each free boundary point in @xmath32 ( or more generally to every point on the _ extended free boundary _
@xmath58 , see section [ sec : classes - solutions ] ) a certain _ frequency _
@xmath59 , see sections [ sec : exist - homog - blow ] and [ sec : class - free - bound ] . at the points for which @xmath60 , the separation rate of the solution @xmath19 from the thin obstacle can be `` detected '' , in a sense that it will exceed the truncation term in the generalized frequency formula . at those points
we are then able to consider the so - called blowups , which will be parabolically @xmath61-homogeneous solutions of the signorini problem , see section [ sec : exist - homog - blow ] .
next , we show that , similarly to what happens in the elliptic case , the smallest possible value of the frequency at a free boundary point is @xmath62 , see section [ sec : homog - glob - solut ] .
we emphasize that our proof of this fact does not rely on the semiconvexity estimates , as in the elliptic case ( see @xcite or @xcite ) .
rather , we use the monotonicity formula of caffarelli @xcite to reduce the problem to the spatial dimension @xmath38 , and then study the eigenvalues of the ornstein - uhlenbeck operator in domains with slits ( see fig .
[ fig : slit - dom ] in section [ sec : homog - glob - solut ] ) .
the elliptic version of this argument has appeared earlier in the book @xcite*chapter 9 .
the bound @xmath63 ultimately implies the optimal @xmath64 regularity of solutions , see section [ sec : optim - regul - solut ] .
we next turn to studying the regularity properties of the free boundary .
we start with the so - called _ regular set _
@xmath48 , which corresponds to free boundary points of minimal frequency @xmath65 .
similarly to the elliptic case , studied in @xcite , the lipschitz regularity of @xmath48 with respect to the space variables follows by showing that there is a cone of spatial directions in which @xmath66 is monotone .
the @xmath44-hlder regularity in @xmath55 is then a consequence of the fact that the blowups at regular points are @xmath55-independent , see theorem [ thm : lip - reg - reg - set ] .
thus , after possibly rotating the coordinate axes in @xmath67 , we obtain that @xmath48 is given locally as a graph @xmath68 where @xmath25 is parabolically lipschitz ( or @xmath69 in alternative terminology ) . to prove the hlder @xmath70 regularity of @xmath71 , @xmath72
, we then use an idea that goes back to the paper of athanasopoulos and caffarelli @xcite based on an application of the boundary harnack principle ( forward and backward ) for so - called domains with thin lipschitz complement , i.e , domains of the type @xmath73 see lemma [ lem : bhp ] . this result
was recently established in the work of the third named author and shi @xcite .
we emphasize that , unlike the elliptic case , the boundary harnack principle for such domains can not be reduced to the other known results in the parabolic setting ( see e.g. @xcite for parabolically lipschitz domains , or @xcite for parabolically nta domains with reifenberg flat boundary ) .
another type of free boundary points that we study are the so - called _ singular points _
, where the coincidence set @xmath74 has zero @xmath75-density in the thin manifold with respect to the thin parabolic cylinders .
this corresponds to free boundary points with frequency @xmath76 , @xmath77 .
the blowups at those points are parabolically @xmath61-homogeneous polynomials , see section [ sec : free - bound - sing ] . following the approach in @xcite , in section [ sec : weiss - monneau - type ]
we establish appropriate parabolic versions of monotonicity formulas of weiss and monneau type . using such formulas we are able to prove the uniqueness of the blowups at singular free boundary points @xmath78 , and consequently obtain a taylor expansion of the type @xmath79 where @xmath80 is a polynomial of parabolic degree @xmath61 that depends continuously on the singular point @xmath78 with frequency @xmath61 .
we note that such expansion holds only for @xmath81 and may fail for @xmath82 ( see remark [ rem : taylor - cntrex ] ) .
nevertheless , we show that this expansion essentially holds when restricted to singular points @xmath83 , even for @xmath84 .
this is necessary in order to verify the compatibility condition in a parabolic version of the whitney s extension theorem ( given in appendix [ sec : parab - whitn - extens ] ) . using the latter we are then able to prove a structural theorem for the singular set . for the elliptic counterpart of this result see @xcite .
it should be mentioned at this moment that one difference between the parabolic case treated in the present paper and its elliptic counterpart is the presence of new types of singular points , which we call _ time - like_. at such points the blowup may become independent of the space variables @xmath85 . we show that such singular points are contained in a countable union of graphs of the type @xmath86 where @xmath25 is a @xmath87 function .
the other singular points , which we call _ space - like _ , are contained in countable union of @xmath88-dimensional @xmath89 manifolds ( @xmath90 ) .
after a possible rotation of coordinates in @xmath67 , such manifolds are locally representable as graphs of the type @xmath91 with @xmath25 and @xmath71 , @xmath92 , continuous . the time - independent version of the signorini problem is closely related to the obstacle problem for the half - laplacian in @xmath67 @xmath93 more precisely , if we consider a harmonic extension of @xmath94 from @xmath95 to @xmath41 ( by means of a convolution with the poisson kernel )
, we will have that @xmath96 thus , the extension @xmath94 will solve the signorini problem in @xmath41 . more generally , in the above problem instead of the half - laplacian one can consider an arbitrary fractional power of the laplacian @xmath97 , @xmath98 , see e.g. the thesis of silvestre @xcite .
problems of this kind appear for instance in mathematical finance , in the valuation of american options , when the asset prices are modelled by jump processes .
the time - independent problem corresponds to the so - called perpetual options , with infinite maturity time . in such framework , with the aid of an extension theorem of caffarelli and silvestre @xcite , many of the results known for @xmath99 can be proved also for all powers @xmath98 , see @xcite .
the evolution version of the problem above is driven by the fractional diffusion and can be written as @xmath100 in @xmath101 with the initial condition @xmath102 this problem has been recently studied by caffarelli and figalli @xcite .
we emphasize that , although their time - independent versions are locally equivalent , the problem studied in @xcite is very different from the one considered in the present paper . in relation to temperature control problems on the boundary , described in @xcite , we would like to mention the two recent papers by athanasopoulos and caffarelli @xcite , and by allen and shi @xcite .
both papers deal with two - phase problems that can be viewed as generalizations of the one - phase problem ( with @xmath103 ) considered in this paper .
the paper @xcite establishes the phenomenon of separation of phases , thereby locally reducing the study of the two - phase problem to that of one - phase .
a similar phenomenon was shown earlier in the elliptic case by allen , lindgren , and the third named author @xcite . in what follows
we provide a brief description of the structure of the paper . * in section [ sec : notat - prel ] we introduce the notations used throughout the paper , and describe the relevant parabolic functional classes . * in section [ sec : existence - regularity ] we overview some of the known basic regularity properties of the solution @xmath19 of the parabolic signorini problem .
the main ones are : @xmath104 ; and , @xmath105 for some @xmath106 .
such results will be extensively used in our paper . * in section [ sec : classes - solutions ] we introduce the classes of solutions @xmath107 of the parabolic signorini problem with a thin obstacle @xmath18 , and show how to effectively `` subtract '' the obstacle by maximally using its regularity . in this process
we convert the problem to one with zero thin obstacle , but with a non - homogeneous right hand side @xmath108 in the equation . in order
to apply our main monotonicity formulas we also need to to extend the resulting functions from @xmath109 to the entire strip @xmath110 .
we achieve this by multiplication by a a cutoff function , and denote the resulting class of functions by @xmath111 .
* section [ sec : estimates - w2 - 1_2s_1 + ] contains generalizations of @xmath112 estimates to the weighted spaced with gaussian measure .
these estimates will be instrumental in the proof of the generalized frequency formula in section [ sec : gener - freq - funct ] and in the study of the blowups in section [ sec : exist - homog - blow ] . in order not to distract the reader from the main content ,
we have deferred the proof of these estimates to the appendix [ sec : est - gauss - proofs ] .
* section [ sec : gener - freq - funct ] is the most technical part of the paper .
there , we generalize almgren s ( poon s ) frequency formula to solutions of the parabolic signorini problem . * in section [ sec : exist - homog - blow ] we prove the existence and homogeneity of the blowups at free boundary points where the separation rate of the solution from the thin obstacle dominates the error ( truncation ) terms in the generalized frequency formula . * in section [ sec : homog - glob - solut ] we prove that the minimal homogeneity of homogeneous solution of the parabolic signorini problem is @xmath113 . * section [ sec : optim - regul - solut ] contains the proof of the optimal @xmath64 regularity of the solutions of the parabolic signorini problem . * in the remaining part of the paper we study the free boundary .
we start in section [ sec : class - free - bound ] by classifying the free boundary points according to the homogeneity of the blowups at the point in question .
we also use the assumed regularity of the thin obstacle in the most optimal way . * in section [ sec : free - bound - regul ] , we study the so - called regular set @xmath48 and show that it can be locally represented as a graph with @xmath70 regular gradient . * in section
[ sec : free - bound - sing ] we give a characterization of the so - called singular points .
* section [ sec : weiss - monneau - type ] contains some new weiss and monneau type monotonicity formulas for the parabolic problem .
these results generalize the ones in @xcite for the elliptic case . * in section
[ sec : struct - sing - set ] we prove the uniqueness of blowups at singular points and the continuous dependence of blowups on compact subsets of the singular set .
we then invoke a parabolic version of whitney s extension theorem ( given in appendix [ sec : parab - whitn - extens ] ) to prove a structural theorem on the singular set .
to proceed , we fix the notations that we are going to use throughout the paper .
@xmath114 & & \text{(parabolic cylinder)}\\ & q_r^\pm(x_0,t_0 ) & & = b_r^\pm(x_0)\times(t_0-r^2,t_0]&&\text{(parabolic half - cylinders)}\\ & \tilde q_r(x_0,t_0 ) & & = b_r(x_0)\times(t_0-r^2,t_0+r^2)&&\text{(full parabolic cylinder)}\\ & q_r'(x_0,t_0 ) & & = b_r'(x_0)\times(t_0-r^2,t_0]&&\text{(thin parabolic cylinder)}\\ * & q_r''(x_0,t_0 ) & & = b_r''(x_0)\times(t_0-r^2,t_0]\\ & s_r & & = { \mathbb{r}}^n\times(-r^2,0 ] & & \text{(parabolic strip)}\\ & s_r^\pm & & = { \mathbb{r}}^n_\pm\times(-r^2,0 ] & & \text{(parabolic half - strip)}\\ & s'_r & & = { \mathbb{r}}^{n-1}\times(-r^2,0 ] & & \text{(thin parabolic strip ) } \intertext{when $ x_0=0 $ and $ t_0=0 $ , we routinely omit indicating the centers $ x_0 $ and $ ( x_0,t_0)$ in the above notations . } & \partial_e u , u_e&&\text{partial derivative in the direction $ e$ } \\ & \partial_{x_i}u , u_{x_i } & & = \partial_{e_i}u,\quad \text{for standard coordinate}\\ * & & & \text{vectors } e_i,\ i=1,\ldots , n\\ & \partial_t u , u_t&&\text{partial derivative in $ t$ variable}\\ & u_{x_{i_1}\cdots x_{i_k}}&&=\partial_{x_{i_1}\cdots x_{i_k}}u=\partial_{x_{i_1}}\cdots\partial_{x_{i_k}}u\\ & \partial_x^\alpha u&&=\partial_{x_1}^{\alpha_1}\cdots\partial_{x_n}^{\alpha_n}u,\\ * & & & \text{for $ \alpha=(\alpha_1,\ldots,\alpha_n)$ , $ \alpha_i\in{\mathbb{z}}_+$}\\ & \nabla u , \nabla_{x}u&&=(\partial_{x_1}u,\ldots,\partial_{x_{n}}u)&&\text{(gradient)}\\ & \nabla'u , \nabla_{x'}u&&=(\partial_{x_1}u,\ldots,\partial_{x_{n-1}}u)&&\text{(tangential or thin gradient)}\\ & \nabla''u,\nabla_{x''}u&&=(\partial_{x_1}u,\ldots,\partial_{x_{n-2}}u)\\ & d^ku , d^k_x u&&=(\partial_x^\alpha u)_{|\alpha|=k},\quad k\in{\mathbb{z}}_+,\\ * & & & \text{where } |\alpha|=\alpha_1+\cdots+\alpha_n\\ & \delta u,\delta_xu&&=\sum_{i=1}^n \partial_{x_ix_i}u&&\text{(laplacian)}\\ & \delta'u,\delta_{x'}u&&=\sum_{i=1}^{n-1 } \partial_{x_ix_i}u&&\text{(tangential or thin laplacian)}\end{aligned}\ ] ] we denote by @xmath115 the _ backward heat kernel _ on @xmath116 @xmath117 we will often use the following properties of @xmath115 : @xmath118 besides , to simplify our calculations , we define the differential operator : @xmath119 which is the generator of the parabolic scaling in the sense that @xmath120 using , the operator @xmath121 can also be defined through the following identity : @xmath122 for the parabolic functional classes , we have opted to use notations similar to those in the classical book of ladyzhenskaya , solonnikov , and uraltseva @xcite .
let @xmath123 be an open subset in @xmath1 and @xmath124 $ ] for @xmath125 .
the class @xmath126 is the class of functions continuous in @xmath28 with respect to parabolic ( or euclidean ) distance .
further , given for @xmath127 we say @xmath128 if for @xmath129 @xmath130 , and define the norm @xmath131 the parabolic hlder classes @xmath132 , for @xmath133 , @xmath127 , @xmath134 are defined as follows .
first , we let @xmath135 then , we define @xmath132 as the space of functions @xmath94 for which the following norm is finite : @xmath136 the parabolic lebesgue space @xmath137 indicates the banach space of those measurable functions on @xmath28 for which the norm @xmath138 is finite .
the parabolic sobolev spaces @xmath139 , @xmath140 , denote the spaces of those functions in @xmath137 , whose distributional derivative @xmath141 belongs to @xmath142 , for @xmath129 .
endowed with the norm @xmath143 @xmath139 becomes a banach space .
we also denote by @xmath144 , @xmath145 the banach subspaces of @xmath146 generated by the norms @xmath147 let @xmath148 for some @xmath149 .
the weighted lebesgue space @xmath150 , @xmath151 , with gaussian weight @xmath152 , will appear naturally in our proofs .
the norm in this space is defined by @xmath153 when @xmath154 is a relatively open subset of @xmath155 , one may also define the respective weighted sobolev spaces .
we will also consider weighted spatial lebesgue and sobolev spaces @xmath156 , and @xmath157 with gaussian weights @xmath158 on @xmath1 for some fixed @xmath159 . for instance , the norm in the space @xmath156 is given by @xmath160
in this section we recall some known results about the existence and the regularity of the solution of the parabolic signorini problem that we are going to take as the starting point of our analysis . for detailed proofs
we refer the reader to the works of arkhipova and uraltseva @xcite . for simplicity we state the relevant results only in the case of the unit parabolic half - cylinder @xmath109 .
suppose we are given functions @xmath161 , @xmath162 , @xmath163)$ ] , and @xmath164 obeying the compatibility conditions @xmath165.\end{aligned}\ ] ] given @xmath18 and @xmath25 as above , we introduce the following closed subset of @xmath166 @xmath167\}.\ ] ] we say that @xmath168 satisfies @xmath169,\\ \label{eq : u - init - q1 } u(\cdot,-1)={\varphi}_0&\quad\text{on } b_1^+,\end{aligned}\ ] ] if @xmath94 solves the variational inequality @xmath170\geq
0\quad\text{for any } v\in{\mathfrak{k}},\\ u\in{\mathfrak{k } } , \quad \partial_t u\in l_2(q_1^+),\\ u(\cdot , -1)={\varphi}_0 \quad\text{on } b_1^+.\end{gathered}\ ] ] under the assumptions above there exists a unique solution to the problem . moreover , the solution will have hlder continuous spatial gradient : @xmath171 for any @xmath172 with the hlder exponent @xmath106 depending only on the dimension , see @xcite .
below , we sketch the details in the case @xmath103 and @xmath173 , that would be most relevant in our case . for any @xmath174
we denote by @xmath175 a mollifications of @xmath108 and consider the solution @xmath176 to the approximating problem @xmath177,\\ u^{\varepsilon}(\cdot,-1)={\varphi}_0&\quad\text{on } b_1^+,\end{aligned}\ ] ] where the penalty function @xmath178 is such that @xmath179 the solutions @xmath176 to the penalization problems are shown to be smooth in @xmath109 up to @xmath180 , and it can be proved that they are uniformly bounded @xmath181 . to this end , for any @xmath182 vanishing a.e . on
@xmath183 $ ] and @xmath184\subset ( -1,0]$ ] , one writes the integral identity @xmath185}(\nabla u^{\varepsilon}\nabla \eta+u^{\varepsilon}_t\eta+f^{\varepsilon}\eta)dx dt+\int_{b_1'\times(t_1,t_2]}\beta_{\varepsilon}(u^{\varepsilon})\eta dx'dt=0.\ ] ] taking in first @xmath186 , and then @xmath187 , one obtains the following global uniform bounds for the family @xmath188 : @xmath189}\|u^{\varepsilon}(\cdot , t)\|_{l_2(b_1^+)}^2+\|\nabla u^{\varepsilon}\|_{l_2(q_1^+)}^2&\leq c_n \big(\|{\varphi}_0\|^2_{l_2(b_1^+)}+\|f\|_{l_2(q_1^+)}^2\big),\\ \sup_{t\in(-1,0]}\|\nabla u^{\varepsilon}(\cdot , t)\|_{l_2(b_1^+)}^2+\|\partial_t u^{\varepsilon}\|_{l_2(q_1^+)}^2&\leq c_n \big(\|\nabla { \varphi}_0\|^2_{l_2(b_1^+)}+\|f\|_{l_2(q_1^+)}^2\big).\end{aligned}\ ] ] thus , the family @xmath188 is uniformly bounded in @xmath181 and passing to the weak limit as @xmath190 one obtains the existence of solutions of the signorini problem in the case @xmath103 , @xmath173 .
besides , by choosing the test functions @xmath191 , @xmath151 , where @xmath192 solves the boundary value problem @xmath193 one can show the global uniform bound @xmath194 for complete details , see the proofs of lemmas 4 and 5 in @xcite . next we have a series of local estimates . with @xmath195 ,
we take the function @xmath196 $ ] , @xmath197 in . integrating by parts , we obtain the following local , uniform , second order estimate @xmath198 one
should compare with our proof of lemma [ lem : w212 ] in appendix [ sec : est - gauss - proofs ] which is the weighted version of this estimate .
furthermore , with more work one can establish the following locally uniform spatial lipschitz bound @xmath199 see lemma 6 in @xcite .
finally , one can show that there exists a dimensional constant @xmath106 such that @xmath200 for any @xmath172 , with the estimate @xmath201 see theorem 2.1 in @xcite .
we summarize the estimates above in the following two lemmas .
[ lem : known - w22 ] let @xmath202 be a solution of the signorini problem with @xmath203 , @xmath204 , @xmath103 , and @xmath173 .
then , @xmath205 for any @xmath172 and @xmath206 [ lem : known - halpha ] let @xmath202 be a solution of the signorini problem with @xmath207 , @xmath208 , @xmath103 , and @xmath173 .
then , for any @xmath172 , @xmath209 , @xmath210 with a dimensional constant @xmath106 and @xmath211 we will also need the following variant of lemma [ lem : known - halpha ] that does not impose any restriction on the boundary data @xmath25 and @xmath24 .
[ lem : known - halpha-2 ] let @xmath212 be a solution of the signorini problem with @xmath161 , and @xmath213 . then
, for any @xmath172 , @xmath214 with a universal @xmath106 , and @xmath215 consider the function @xmath216 \eta(x , t)\ ] ] with @xmath217 , such that @xmath218 in particular , @xmath219 on @xmath180 .
then @xmath94 satisfies the conditions of lemma [ lem : known - halpha ] with @xmath103 , @xmath173 , @xmath220 and with @xmath108 replaced by @xmath221\eta+(v-{\varphi})(\delta-\partial_t)\eta+2(\nabla v-\nabla{\varphi})\nabla \eta.\ ] ] the assumptions on @xmath19 and @xmath18 now imply the required estimate from that in lemma [ lem : known - halpha ] . in section [ sec :
estimates - w2 - 1_2s_1 + ] we generalize the estimates in lemma [ lem : known - w22 ] for the appropriate weighted gaussian norms .
the proof of these estimates are given in appendix [ sec : est - gauss - proofs ] .
one of our main results in this paper is the optimal value of the hlder exponent @xmath35 in lemma [ lem : known - halpha-2 ] .
we show that @xmath222 , or slightly stronger , that @xmath223 , when @xmath108 is bounded , see theorem [ thm : opt - reg ] .
in this paper we are mostly interested in local properties of the solution @xmath19 of the parabolic signorini problem and of its free boundary . in view of this , we focus our attention on solutions in parabolic ( half-)cylinders .
furthermore , thanks to the results in section [ sec : existence - regularity ] , we can , and will assume that such solutions possess the regularity provided by lemmas [ lem : known - w22 ] and [ lem : known - halpha ] .
given @xmath224 , we say that @xmath225 if @xmath226 , @xmath227 for some @xmath228 , and @xmath19 satisfies @xmath229 and @xmath230 where @xmath231 is the boundary in the relative topology of @xmath180 .
we call the set @xmath58 the _ extended free boundary _ for the solution @xmath19 . recall that the _ free boundary _ is given by @xmath232 note that by definition @xmath233 .
the reason for considering this extension is that parabolic cylinders do not contain information on `` future times '' .
this fact may create a problem when restricting solutions to smaller subcylinders .
the notion of extended free boundary removes this problem .
namely , if @xmath234 and @xmath235 is such that @xmath236 , then @xmath237 .
sometimes , we will abuse the terminology and call @xmath58 the free boundary . replacing @xmath109 and @xmath180 by @xmath238 and @xmath239 respectively in the definition above
, we will obtain the class @xmath240 .
note that if @xmath241 then the parabolic rescaling @xmath242 where @xmath243 can be arbitrary ( but typically chosen to normalize a certain quantity ) , belongs to the class @xmath244 .
having that in mind , we will state most of the results only for the case @xmath245 .
the function @xmath246 allows a natural extension to the entire parabolic cylinder @xmath247 by the even reflection in @xmath248 coordinate : @xmath249 then @xmath19 will satisfy @xmath250 where @xmath251 is the so - called _ coincidence set_. more generally , @xmath252 in the sense of distributions , where @xmath253 is the @xmath254-dimensional hausdorff measure and by @xmath255 we understand the limit from the right @xmath256 on @xmath180 .
we next show how to reduce the study of the solutions with nonzero obstacle @xmath18 to the ones with zero obstacle . as the simplest such reduction we consider the difference @xmath257 , which will satisfy the signorini conditions on @xmath180 with zero obstacle , but at an expense of solving nonhomogeneous heat equation instead of the homogeneous one .
one may further extend this difference to the strip @xmath258 $ ] by multiplying with a cutoff function in @xmath259 variables . more specifically ,
let @xmath260 be such that @xmath261 and consider the function @xmath216\psi(x)\quad\text{for } ( x , t)\in s_1^+.\ ] ] it is easy to see that @xmath94 satisfies the nonhomogeneous heat equation in @xmath110 @xmath262+[v(x , t)-{\varphi}(x',t)]\delta\psi+2\nabla v\nabla\psi,\end{aligned}\ ] ] and the signorini boundary conditions on @xmath263 @xmath264 moreover , it is easy to see that @xmath108 is uniformly bounded .
[ def : sf ] we say that @xmath265 , for @xmath266 if @xmath267 , @xmath268 , @xmath94 has a bounded support and solves @xmath269 and @xmath270 if we only assume @xmath213 , then in the construction above we can only say that the function @xmath271 . for some of the results that we are going to prove ( such as the optimal regularity in theorem [ thm : opt - reg ] ) this will be sufficient .
however , if we want to study a more refined behavior of @xmath94 near the origin , we need to assume more regularity on @xmath18
. thus , if we assume @xmath272 with @xmath273 , @xmath274 , @xmath134 , then for its parabolic taylor polynomial @xmath275 of parabolic degree @xmath276 at the origin , we have @xmath277 for a certain @xmath278 , and more generally @xmath279 for @xmath280 .
to proceed , we calorically extend the polynomial @xmath275 in the following sense .
[ lem : calor - ext ] for a given polynomial @xmath281 on @xmath282 there exists a caloric extension polynomial @xmath283 in @xmath284 , symmetric in @xmath248 , i.e. , @xmath285 moreover , if @xmath281 is parabolically homogeneous of order @xmath61 , then one can find @xmath286 as above with the same homogeneity .
it is easily checked that the polynomial @xmath287 is the desired extension . here
@xmath288 is taken so that the parabolic degree of the polynomial @xmath281 does not exceed @xmath289 .
let now @xmath290 be the extension of the parabolic taylor polynomial @xmath291 of @xmath18 at the origin and consider @xmath292 it is easy to see that @xmath293 solves the signorini problem with the thin obstacle @xmath294 , i.e. @xmath295 , now with an additional property @xmath296 then if we proceed as above and define @xmath297\psi(x)\\ & = [ v(x , t)-{\varphi}(x ,
t)-(\tilde q_k(x , t)-q_k(x',t))]\psi(x)\end{aligned}\ ] ] then @xmath298 will satisfy with the right - hand side @xmath299 + [ v_k(x , t)-{\varphi}_k(x',t)]\delta\psi+2\nabla v_k\nabla\psi,\ ] ] which additionally satisfies @xmath300 for @xmath301 depending only on @xmath302 , @xmath303 , and @xmath304 . moreover , for @xmath305 one will also have @xmath306 we record this construction in the following proposition . [ prop : uk - def ] let @xmath307 with @xmath308 , @xmath273 , @xmath309 , @xmath134 .
if @xmath291 is the parabolic taylor polynomial of order @xmath276 of @xmath18 at the origin , @xmath290 is its extension given by lemma [ lem : calor - ext ] , and @xmath302 is a cutoff function as in , then @xmath310\psi(x)\ ] ] belongs to the class @xmath311 with @xmath312 and more generally , for @xmath313 , @xmath306 furthermore , @xmath314 , @xmath315 in @xmath316 and therefore @xmath317
in this section we state @xmath112-estimates with respect to the gaussian measure @xmath318 in the half - strips @xmath319 .
the estimates involves the quantities that appear in the generalized frequency formula that we prove in the next section . since the computations are rather long and technical , to help with the readability of the paper , we have moved the proofs to appendix [ sec : est - gauss - proofs ] .
[ lem : w212 ] let @xmath265 with @xmath271 .
then , for any @xmath320 we have the estimates @xmath321 [ rem : w212-alt ] even though the estimates above are most natural for our further purposes , we would like to note that slightly modifying the proof one may show that @xmath322 i.e. , without the weights @xmath323 and @xmath324 in the integrals for @xmath325 and @xmath326 . more generally , the same estimate can be proved if @xmath327 in @xmath328 , with the half - strips @xmath319 and @xmath110 replaced by the full strips @xmath329 and @xmath328 .
[ lem : w112-diff ] suppose @xmath330 , @xmath331 , with @xmath332 .
then for any @xmath320 we have the estimate @xmath333 g. \end{aligned}\ ] ]
in this section we will establish a monotonicity formula , which will be a key tool for our study .
the origins of this formula go back to almgren s big regularity paper @xcite , where he proved that for ( multiple - valued ) harmonic functions in the unit ball , the frequency function @xmath334 is monotonically increasing in @xmath335 .
versions of this formula have been used in different contexts , most notably in unique continuation @xcite and more recently in the thin obstacle problem @xcite .
almgren s monotonicity formula has been generalized by poon to solutions of the heat equation in the unit strip @xmath328 .
more precisely , he proved in @xcite that if @xmath336 in @xmath328 , then its caloric frequency , defined as @xmath337 is monotone non - decreasing in @xmath335 .
this quantity , which to a large extent plays the same role as almgren s frequency function , differs from the latter in that it requires @xmath94 to be defined in an entire strip .
thereby , it is not directly applicable to caloric functions which are only locally defined , for instance when @xmath94 is only defined in the unit cylinder @xmath247 .
one possible remedy to this obstruction is to consider an extension of a caloric function in @xmath247 to the entire strip @xmath328 by multiplying it with a spatial cutoff function @xmath302 , supported in @xmath338 : @xmath339 such extension , however , will no longer be caloric in @xmath328 , and consequently the parabolic frequency function @xmath340 is no longer going to be monotone .
however , there is a reasonable hope that @xmath340 is going to exhibit properties close to monotonicity .
in fact , to be able to control the error terms in the computations , we will need to consider a `` truncated '' version of @xmath288 .
moreover , we will be able to extend this result to functions @xmath265 , and to functions @xmath246 , via the constructions in proposition [ prop : uk - def ] .
to proceed , we define the following quantities : @xmath341 for any function @xmath94 in the parabolic half - strip @xmath110 for which the integrals involved are finite .
note that , if @xmath94 is an even function in @xmath248 , then poon s parabolic frequency function is given by @xmath342 there are many substantial technical difficulties involved in working with this function directly . to overcome such difficulties , we consider the following averaged versions of @xmath343 and @xmath344 : @xmath345 one further obstruction is represented by the fact that the above integrals may become unbounded near the endpoint @xmath346 , where @xmath115 becomes singular . to remedy this problem
we introduce the following truncated versions of @xmath347 and @xmath348 . for a constant @xmath349 ,
let @xmath350 the following lemma plays a crucial role in what follows .
[ lem : differentiation - formulae - smooth ] assume that @xmath351 satisfies @xmath352 then , for any @xmath353 we have the following differentiation formulas @xmath354 where the vector field @xmath121 is as in .
the main step in the proof consists in establishing the following differentiation formulas for @xmath355 : @xmath356 and @xmath357 once this is done , then noting that @xmath358 we find @xmath359 using we thus obtain @xmath360 the formula for @xmath361 is computed similarly .
we are thus left with proving and .
[ step:1-diff - form - smooth](@xmath362 ) we start with claiming that @xmath363 indeed , noting that @xmath364 in @xmath365 , and keeping in mind that the outer unit normal to @xmath41 on @xmath67 is given by @xmath366 , we integrate by parts to obtain @xmath367 where in the first integral of the last equality we have used .
this proves the claim .
[ step:2-diff - form - smooth](@xmath362
) we now prove the formula for @xmath368 .
note that for @xmath369 we have @xmath370 here , we have used the identity @xmath371 . differentiating with respect to @xmath372 at @xmath373 , and using
, we therefore obtain @xmath374 or equivalently @xmath375 using now the formula for @xmath376 in ( [ step:1-diff - form - smooth]@xmath362 ) , and the fact that @xmath377 , we obtain @xmath378 [ step:3-diff - form - smooth](@xmath362 ) to obtain the differentiation formula for @xmath376 , note that using the scaling properties of @xmath379 , similarly to what was done for @xmath380 , we have @xmath381 differentiating with respect to @xmath372 at @xmath373 , we obtain @xmath382 we now use the following easily verifiable identity @xmath383 which , after substitution in the latter equation and integration by parts , yields @xmath384 hence , we obtain @xmath385 which establishes . with lemma [ lem : differentiation - formulae - smooth ] in hands , we turn to establishing the essential ingredient in the proof of our main monotonicity result .
[ prop : diff - form - signor ] let @xmath265 .
then , @xmath347 and @xmath348 are absolutely continuous functions on @xmath386 and for a.e .
@xmath335 we have @xmath387 we note that , thanks to the estimates in lemma [ lem : w212 ] above , all integrals in the above formulas are finite . the idea of the proof of the proposition is to approximate @xmath94 with smooth solutions @xmath176 to the signorini problem , apply lemma [ lem : differentiation - formulae - smooth ] , and then pass to the limit in @xmath388 .
the limit process is in fact more involved then one may expect .
one complication is that although we have the estimates in lemma [ lem : w212 ] for the solution @xmath94 , we do not have similar estimates , uniform in @xmath388 , for the approximating @xmath176 .
this is in fact the main reason for which we have to consider truncated quantities @xmath389 and @xmath390 , let @xmath190 first , and then @xmath391 . however , the main difficulty is to show that the integrals over @xmath392 vanish .
this is relatively easy to do for @xmath347 , since @xmath393 on @xmath263 . on the other hand , proving
the formula for @xmath394 is considerably more difficult , since one has to justify that @xmath395 on @xmath263 .
furthermore , the vanishing of this term should be interpreted in a proper sense , since we generally only know that @xmath396 , and thus its trace may not even be well defined on @xmath263 . with this being
said , in the sequel we justify only the formula for @xmath394 , the one for @xmath397 being analogous , but much simpler .
[ step:1-diff - form](@xmath362 ) assume that @xmath94 is supported in @xmath398 $ ] , @xmath399 .
multiplying @xmath94 with a cutoff function @xmath400 such that @xmath401 on @xmath402 $ ] and @xmath403 on @xmath404 $ ] for @xmath405 , without loss of generality we may assume that @xmath406 . we then approximate @xmath94 in @xmath407 $ ] with the solutions of the penalized problem @xmath408,\\ \partial_{x_n}u^{\varepsilon}=\beta_{\varepsilon}(u^{\varepsilon})&\quad\text{on } b_r'\times(-1,0],\\ u^{\varepsilon}=0&\quad\text{on } ( \partial b_r)^+\times(-1,0],\\ u^{\varepsilon}(\cdot , -1)=0&\quad\text{on } b_r^+ , \end{aligned}\ ] ] where @xmath175 is a mollification of @xmath108 . for any @xmath409 ,
let @xmath410,\quad q_{r-1,\rho } ' = b_{r-1}'\times(-\rho^2,0].\ ] ] from the estimates in section [ sec : existence - regularity ] we have @xmath411 uniformly in @xmath412 .
[ step:2-diff - form](@xmath362 ) now , in order to extend @xmath176 to @xmath110 , pick a cutoff function @xmath413 such that @xmath414 and define @xmath415 note that since @xmath94 is supported in @xmath416 $ ] , @xmath417 will converge to @xmath94 and we will have the uniform estimates @xmath418 for @xmath417 we have that @xmath419 it is easy to see that @xmath420 converges strongly to @xmath108 in @xmath421 for @xmath349 .
besides , we also have @xmath422 and therefore @xmath423 [ step:3-diff - form](@xmath362 ) we now fix a small @xmath424 , apply the differentiation formulas in lemma [ lem : differentiation - formulae - smooth ] to @xmath417 and pass to the limit .
we have @xmath425 [ step:3.1-diff - form](.@xmath362 ) to pass to the limit in @xmath426 , we note that @xmath427 converges to @xmath428 weakly in @xmath429 .
indeed , this follows from the uniform @xmath112 estimates on @xmath417 in ( [ step:2-diff - form]@xmath362 ) and the boundedness of @xmath115 in @xmath430 .
thus , in the limit we obtain @xmath431 note that here we can not claim equality , as we do not have a strong convergence of @xmath427 to @xmath428 .
[ step:3.2-diff - form](.@xmath362 ) in @xmath432 , the weak convergence of @xmath427 to @xmath428 , combined with the strong convergence of @xmath420 to @xmath108 in @xmath433 is enough to conclude that @xmath434 moreover , the convergence will be uniform in @xmath435\subset ( 0,1)$ ] .
[ step:3.3-diff - form](.@xmath362 ) finally , we claim that @xmath436 as @xmath190 , i.e. , @xmath437 indeed , we have @xmath438 we then estimate the integrals @xmath439 and @xmath440 separately .
[ step:3.3.1-diff - form]( .. @xmath362 ) we start with @xmath439 . recall that @xmath441 in @xmath442 . by , this implies that @xmath443 in @xmath442 and therefore @xmath444 [ step:3.3.2-diff - form]( .. @xmath362 ) a similar estimate holds also for @xmath440 , but the proof is a little more involved . to this end
consider @xmath445 from , it is easy to see that @xmath446 with @xmath447 . note that the uniform bound @xmath448 in @xmath442 implies that @xmath449 to use this fact , note that @xmath450 g\\ & = \int_{s_r'\setminus s_{\delta r}'}z[t\zeta^2{\mathcal{b}}_{\varepsilon}(u^{\varepsilon } ) ] g-\int_{s_r'\setminus s_{\delta r}'}z(t\zeta^2){\mathcal{b}}_{\varepsilon}(u^{\varepsilon } ) g\\ & = : e_{22}-e_{21}.\end{aligned}\ ] ] [ step:3.3.2.1-diff - form]( ... @xmath362 ) the estimate for @xmath451 is straightforward : @xmath452 [ step:3.3.2.2-diff - form]( ... @xmath362 ) to estimate @xmath453 , denote @xmath454 .
then , substituting @xmath455 , @xmath456 , we have @xmath457g(y',-1)dy'\\ & \qquad-2\int_{\delta r}^r\int_{{\mathbb{r}}^{n-1 } } u^{\varepsilon}(\lambda y',-\lambda^2 ) g(y',-1)dy'd\lambda\end{aligned}\ ] ] and consequently @xmath458 combining the estimates in ( [ step:3.3.1-diff - form]@xmath362)([step:3.3.2-diff - form]@xmath362 ) above , we obtain @xmath459 [ step:4-diff - form](@xmath362 ) now , writing for any @xmath460 @xmath461 collecting the facts proved in ( [ step:3.1-diff - form]@xmath362)([step:3.3-diff - form]@xmath362 ) , and passing to the limit as @xmath190 , we obtain @xmath462 next , note that by lemma [ lem : w212 ] the integrands are uniformly bounded with respect to @xmath463 ( for fixed @xmath464 and @xmath465 ) .
therefore , we can let @xmath391 in the latter inequality , to obtain @xmath466 this is equivalent to the sought for conclusion for @xmath467 . to state the main result of this section , the generalized frequency formula , we need the following notion .
we say that a positive function @xmath468 is a @xmath469-convex function of @xmath470 on @xmath471 if @xmath472 is a convex function of @xmath55 .
this simply means that @xmath473 this is equivalent to saying that @xmath474 is locally absolutely continuous on @xmath475 and @xmath476 is nondecreasing .
for instance , @xmath477 is a @xmath469-convex function of @xmath470 for any @xmath61 .
the importance of this notion in our context is that almgren s and poon s frequency formulas can be regarded as @xmath469-convexity statements in @xmath470 for the appropriately defined quantities @xmath478 .
[ thm : thin - monotonicity ] let @xmath479 with @xmath108 satisfying the following condition : there is a positive monotone nondecreasing @xmath469-convex function @xmath468 of @xmath470 , and constants @xmath480 and @xmath481 , such that @xmath482 then , there exists @xmath483 , depending only on @xmath484 , @xmath485 and @xmath254 , such that the function @xmath486 is nondecreasing for @xmath487 .
note that on the open set where @xmath488 we have @xmath489 which coincides with @xmath490 , when @xmath491 .
the purpose of the `` truncation '' of @xmath478 with @xmath468 is to control the error terms in computations that appear from the right - hand - side @xmath108 .
first , we want to make a remark on the definition of @xmath492 , for @xmath493 .
the functions @xmath478 and @xmath468 are absolutely continuous and therefore so is @xmath494 .
it follows that @xmath495 is uniquely identified only up to a set of measure zero .
the monotonicity of @xmath495 should be understood in the sense that there exists a monotone increasing function which equals @xmath495 almost everywhere .
therefore , without loss of generality we may assume that @xmath496 on @xmath497 and @xmath498 in @xmath499 .
following an idea introduced in @xcite we now note that it will be enough to check that @xmath500 in @xmath501 .
indeed , from the assumption on @xmath474 , it is clear that that @xmath495 is monotone on @xmath502 .
next , if @xmath503 is a maximal open interval in @xmath501 , then @xmath504 and @xmath505 unless @xmath506 .
besides , if @xmath495 is monotone in @xmath503 , it is easy to see that the limits @xmath507 and @xmath508 will exist and satisfy @xmath509 and therefore we will have @xmath510 with the latter inequality holding when @xmath511 .
this will imply the monotonicity of @xmath495 in @xmath386 .
therefore , we will concentrate only on the set @xmath499 , where the monotonicity of @xmath512 is equivalent to that of @xmath513 the latter will follow , once we show that @xmath514 in @xmath501 .
now , from proposition [ prop : diff - form - signor ] we have @xmath515 we then estimate the derivatives of each of the quantities @xmath516 , @xmath331 .
[ step:1-mon - form](@xmath362 ) using the differentiation formulas in proposition [ prop : diff - form - signor ] , we compute @xmath517\\&\qquad-\int_{s_r^+ } u^2 g\int_{s_r^+ } t^2 f^2 g+\big(\int_{s_r^+ } t u f g\big)^2 .
\end{aligned}\ ] ] applying the cauchy - schwarz inequality we obtain that the term in square brackets above is nonnegative .
therefore , @xmath518 [ step:2-mon - form](@xmath362 ) we next estimate the derivative of @xmath519 .
@xmath520 [ step:3-mon - form](@xmath362 ) combining together the estimates for @xmath521 and @xmath522 , we have @xmath523 we estimate the third term separately
. first , from @xmath524 we have @xmath525 from here we see that @xmath526 and therefore we also have @xmath527 substituting into the inequality above , we then have @xmath528 on the set @xmath529 , we easily have @xmath530 and consequently , @xmath531 note that in the last step we have again used the fact that @xmath532 .
the desired conclusion follows readily .
in this section , we show how the generalized frequency formula in theorem [ thm : thin - monotonicity ] can be used to study the behavior of the solution @xmath94 near the origin .
the central idea is to consider some appropriately normalized rescalings of @xmath94 , indicated with @xmath533 ( see definition [ def : resc - h ] ) , and then pass to the limit as @xmath534 ( see theorem [ thm : exist - homogen - blowups ] ) .
the resulting limiting functions ( over sequences @xmath535 ) are known as _
blowups_. however , because of the truncation term @xmath468 in the generalized frequency function @xmath512 , we can show the existence of blowups only when the growth rate of @xmath94 can be `` detected , '' in a certain proper sense to be made precise below .
finally , as a consequence of the monotonicity of @xmath512 , we obtain that the blowups must be parabolically homogeneous solutions of the signorini problem in @xmath536 $ ] .
henceforth , we assume that @xmath265 , and that @xmath468 be such that the conditions of theorem [ thm : thin - monotonicity ] are satisfied . in particular , we assume that @xmath537 consequently , theorem [ thm : thin - monotonicity ] implies that the function @xmath538 is nondecreasing for @xmath493 . hence , there exists the limit @xmath539 since we assume that @xmath476 is nondecreasing , the limit @xmath540 also exists .
we then have the following basic proposition concerning the values of @xmath61 and @xmath541 .
[ lem : hu - mu ] let @xmath265 and @xmath474 satisfy the conditions of theorem [ thm : thin - monotonicity ] . with @xmath61 , @xmath541 as above
, we have @xmath542 moreover , if @xmath543 , then there exists @xmath544 such that @xmath545 for @xmath546 . in particular , @xmath547 as a first step we show that @xmath548 indeed , if the implication claimed in fails , then for a sequence @xmath549 we have @xmath550 .
this implies that @xmath551 and therefore @xmath552 which contradicts @xmath553 .
we have thus proved .
now , implies that , if @xmath553 , then @xmath554 passing to the limit , we conclude that , if @xmath553 , then @xmath555 however , in this case we also have @xmath556 indeed , recall that ( see proposition [ prop : diff - form - signor ] ) @xmath557 and will follow once we show that @xmath558 using the cauchy - schwarz inequality , we have @xmath559 where in the last inequality before the limit we have used . summarizing , the assumption @xmath560 implies above .
therefore , the proof will be completed if we show that the case @xmath561 is impossible .
so , assume towards a contradiction that @xmath561 , and fix @xmath562 .
for such @xmath388 choose @xmath563 so that @xmath564 integrating these inequalities from @xmath565 to @xmath566 , we obtain @xmath567 since by our choice of @xmath174 we have @xmath568 , the above inequalities imply that @xmath569 for small enough @xmath565 , contrary to the established conclusion of above .
hence , the case @xmath561 is impossible , which implies that we always have @xmath570 . to proceed
, we define the appropriate notion of rescalings that works well with the generalized frequency formula .
[ def : resc - h ] for @xmath479 and @xmath235 define the _ rescalings _ @xmath571.\ ] ] it is easy to see that the function @xmath533 solves the nonhomogeneous signorini problem @xmath572 with @xmath573 in other words , @xmath574 .
further , note that @xmath533 is normalized by the condition @xmath575 and that , more generally , we have @xmath576 we next show that , unless we are in the borderline case @xmath577 , we will be able to study the so - called blowups of @xmath94 at the origin .
the condition @xmath543 below can be understood , in a sense , that we can `` detect '' the growth of @xmath94 near the origin .
[ thm : exist - homogen - blowups ] let @xmath265 , @xmath474 satisfy the conditions of theorem [ thm : thin - monotonicity ] , and @xmath578 then , we have : 1 .
for any @xmath149 , there is @xmath579 such that @xmath580 2 .
there is a sequence @xmath549 , and a function @xmath581 in @xmath582 $ ] , such that @xmath583 we call any such @xmath581 a _ blowup _ of @xmath94 at the origin .
3 . @xmath581 is a nonzero _ global solution _ of signorini problem : @xmath584 in the sense that it solves the signorini problem in every @xmath238 .
@xmath581 is parabolically homogeneous of degree @xmath61 : @xmath585 the proof of theorem [ thm : exist - homogen - blowups ] is based on the following lemmas .
[ lem : hu - est]if @xmath586 and @xmath544 is such that @xmath587 and @xmath545 for @xmath588 , then @xmath589 from the assumptions we have @xmath590 for @xmath591 , which implies that @xmath592 integrating from @xmath593 to @xmath565 and exponentiating , we find @xmath594 which implies that @xmath595 similarly , integrating from @xmath565 to @xmath596 ( under the assumption that @xmath597 ) we find @xmath598 [ lem : fr - est ] under the notations of the previous lemma , for any @xmath599 and @xmath600 , we have @xmath601 note that from the assumptions we have @xmath602 now take @xmath599 .
then , making the change of variables and using the inequalities above , we have @xmath603 thus , if @xmath604 , then @xmath605 and therefore @xmath606 this completes the proof
. we will also need the following well - known inequality ( see @xcite ) and one of its corollaries .
for any @xmath607 one has @xmath608 [ lem : cor - log - sob ] for any @xmath609 , let @xmath610 .
then , @xmath611 let @xmath612 for @xmath613 and @xmath614 .
then , the @xmath469-sobolev inequality can be rewritten as @xmath615 on the other hand , since @xmath302 is convex on @xmath616 , by jensen s inequality we have @xmath617 combining these inequalities and using the identity @xmath618 , we arrive at the claimed inequality .
\i ) from lemmas [ lem : hu - est ] and [ lem : fr - est ] as well as lemma [ lem : w212 ] , for @xmath599 , @xmath619 we have @xmath620 since @xmath599 is arbitrary , this implies the claim of part i ) .
\ii ) note that , in view of lemma [ lem : w112-diff ] , it will be enough to show the existence of @xmath581 , and the convergence @xmath621 from lemma [ lem : hu - mu ] it follows that @xmath622 , , but the information @xmath622 will suffice in this proof . ]
and therefore we obtain @xmath623 integrating , we obtain that for small @xmath424 @xmath624 which gives @xmath625 and consequently @xmath626 next , let @xmath627 be a cutoff function , such that @xmath628 we may take @xmath629 so large that @xmath630 for @xmath631 . then from lemma [ lem : cor - log - sob ] , we have that @xmath632 for small enough @xmath565 , where in the last step we have used the uniform estimate from part i ) .
next , notice that on @xmath633 $ ] the function @xmath379 is bounded below and above and therefore the estimates in i ) imply that the family @xmath634 is uniformly bounded in @xmath635 and thus we can extract a subsequence @xmath636 converging strongly in @xmath637 and consequently in @xmath638 . letting @xmath391 and @xmath639 , combined with the estimates above , by means of the cantor diagonal method , we complete the proof of this part .
\iii ) we first start with the signorini boundary conditions . from the estimates in ii ) we have that @xmath640 is uniformly bounded in @xmath641)$ ] for any @xmath642 .
we thus obtain that @xmath643 strongly , and @xmath644 weakly in @xmath645)$ ] .
this is enough to pass to the limit in the signorini boundary conditions and to conclude that @xmath646 besides , arguing similarly , and using lemma [ lem : fr - est ] we obtain that @xmath647 thus , to finish the proof of this part it remains to show that @xmath581 is in the unweighted sobolev class @xmath648 for any @xmath149 . because of the scaling properties , it is sufficient to prove it only for @xmath649 .
we argue as follows .
first , extend @xmath581 by even symmetry in @xmath248 to @xmath650 .
we then claim that @xmath651 are subcaloric functions in @xmath650 . indeed ,
this would follow immediately , if we knew the continuity of @xmath581 , since @xmath581 is caloric where nonzero .
but since we do not know the continuity of @xmath581 at this stage , we argue as follows . by continuity of @xmath533
we easily obtain @xmath652.\ ] ] then , passing to the limit as @xmath653 , we conclude that @xmath651 are subcaloric , since @xmath654 in @xmath655)$ ] by lemma [ lem : fr - est ] .
further , we claim that @xmath651 satisfy the sub mean - value property @xmath656 for any @xmath657 .
the proof of this fact is fairly standard , since , by the estimates in part i ) , @xmath581 satisfies an integral tychonoff - type condition in the strips @xmath658 , @xmath424 .
nevertheless , for completeness we give the details below . for large @xmath149
let @xmath659 be a cutoff function such that @xmath660 , @xmath661 on @xmath662 , @xmath663 , @xmath664 .
let now @xmath665 in @xmath666 .
from the fact that @xmath651 are subcaloric , we have that @xmath667 the advantage of @xmath192 now is that it has a bounded support , and therefore we can write @xmath668 to proceed , fix @xmath669 large and @xmath670 small and consider @xmath671 and @xmath672 . we want to show that the second integral above will vanish as we let @xmath673 .
this will be done with suitable estimates on the kernel @xmath115 .
[ clm : rho - est ] let @xmath671 , @xmath674 , and @xmath675 .
then @xmath676 with @xmath677 , @xmath678 .
[ step:1-g - est](@xmath362 ) @xmath679 .
choose @xmath680 and let @xmath681 .
then @xmath682 and therefore @xmath683 where we have used that the function @xmath684 is uniformly bounded on @xmath685 . [ step:2-g - est](@xmath362 ) suppose now @xmath686 .
then @xmath687 hence , the claim follows .
using the claim , the facts that @xmath688 and @xmath689 , and letting @xmath673 , we then easily obtain @xmath690 then , using the second estimate in claim [ clm : rho - est ] , for @xmath691 we obtain @xmath692 more generally , changing the initial point @xmath693 to arbitrary point @xmath694 $ ] we will have @xmath695 and by applying cauchy - schwarz @xmath696 the energy inequality applied to @xmath651 then yields @xmath697 . further , applying the estimate in lemma [ lem : known - w22 ] for @xmath698 , where @xmath699 is a smooth cutoff function in @xmath700 , equal to @xmath701 on @xmath702 , with @xmath220 and @xmath703 , we obtain that @xmath704 . as remarked earlier , the scaling properties imply that @xmath705 for any @xmath149 .
\iv ) finally , we show that @xmath581 is parabolically homogeneous of degree @xmath61 .
let @xmath549 be such that @xmath643 as in ii ) .
then by part ii ) again we have for any @xmath320 @xmath706 moreover , since by lemma [ lem : hu - est ] @xmath707 for sufficiently small @xmath565 , we also have @xmath708 hence , we obtain that for any @xmath320 @xmath709 by lemma [ lem : hu - mu ] . thus the ratio @xmath710 is constant in the interval @xmath386 .
further , notice that passing to the limit in the differentiation formulas in proposition [ prop : diff - form - signor ] , we will obtain the similar formulas hold for @xmath581 for any @xmath711 .
thus , from computations in step ( [ step:1-mon - form]@xmath362 ) in theorem [ thm : thin - monotonicity ] , before the application of the cauchy - schwarz inequality , we have @xmath712.\ ] ] note here that @xmath713 is never zero , since @xmath714 for @xmath715 and @xmath716 . and
since we know that the above derivative must be zero it implies that we have equality in cauchy - schwarz inequality @xmath717 which can happen only if for some constant @xmath718 we have @xmath719 or that @xmath581 is parabolically homogeneous of degree @xmath718 .
but then , in this case it is straightforward to show that @xmath720 and therefore @xmath721 this implies that @xmath722 and completes the proof of the theorem .
in this section we study the homogeneous global solutions of the parabolic signorini problem , which appear as the result of the blowup process described in theorem [ thm : exist - homogen - blowups ] .
one of the conclusions of this section is that the homogeneity @xmath61 of the blowup is @xmath723 see theorem [ thm : min - homogen ] below .
this will have two important consequences : ( i ) the fact that @xmath63 will imply the optimal @xmath724 regularity of solutions ( see theorem [ thm : opt - reg ] ) and ( ii ) the `` gap '' @xmath725 between possible values of @xmath61 will imply the relative openness of the so - called regular set ( see proposition [ prop : reg - set - rel - open ] ) .
we start by noticing that @xmath726 .
[ prop : kappa>1alpha ] let @xmath265 be as in theorem [ thm : exist - homogen - blowups ] . then , @xmath727 , where @xmath35 is the hlder exponent of @xmath728 in definition [ def : sf ] . for the proof we will need the following fact . [
lem : holder - hu - above ] let @xmath479 .
then , @xmath729 where @xmath35 is the hlder exponent of @xmath728 in definition [ def : sf ] .
since @xmath730 , we must have @xmath731 . recalling also that @xmath94 has a bounded support , we obtain that @xmath732 let us show that for @xmath483 @xmath733 because of the gradient estimate above , it will be enough to show that @xmath734 first , observe that since @xmath735 on @xmath736 , we readily have @xmath737 to show the estimate from above , it will be enough to establish that @xmath738 note that since @xmath94 is bounded , it is enough to show this bound for @xmath739 . assuming the contrary , let @xmath740 be such that @xmath741 with large enough @xmath742 .
then , from the bound on the gradient , we have @xmath743 on @xmath744 . in particular , @xmath745 also , let @xmath301 be such that @xmath746 in @xmath110 .
then , consider the function @xmath747 we will have @xmath748 besides , @xmath749 then , from the parabolic harnack inequality ( see e.g.@xcite*theorems 6.176.18 ) @xmath750 or equivalently , @xmath751 but then from the bound on the gradient we will have @xmath752 if @xmath742 is sufficiently large , a contradiction .
this implies the claimed estimate @xmath753 the estimate for @xmath478 is then a simple corollary : @xmath754 the proof of proposition [ prop : kappa>1alpha ] now follows easily .
let @xmath755 be arbitrary .
then , by lemma [ lem : hu - est ] we have @xmath756 on the other hand , by lemma [ lem : holder - hu - above ] ( proved below ) we have the estimate @xmath757 hence , @xmath758 .
since this is true for any @xmath759 , we obtain that @xmath760 , which is the sought for conclusion .
we will also need the following technical fact .
let @xmath581 be a nonzero @xmath61-parabolically homogeneous solution of the signorini problem in @xmath761 , as in theorem [ thm : exist - homogen - blowups](iii ) .
then , @xmath762 for some @xmath228 .
note that this does not follow from lemmas [ lem : known - halpha ] or [ lem : known - halpha-2 ] directly , since they rely on @xmath763-regularity of @xmath24 ( which is given by the function @xmath581 itself ) , or @xmath764-regularity of @xmath581 , which has to be properly justified .
note that because of the homogeneity , it is enough to show that @xmath765 is in @xmath766 .
indeed , @xmath767 for any @xmath149 and since @xmath768 , we obtain that @xmath25 solves the signorini problem @xmath769 but now , the known results for the elliptic signorini problem for operators with variable coefficients ( see e.g. @xcite ) imply that @xmath770 , as claimed .
[ prop : homogen - glob - sol-1 - 2 ] let @xmath581 be a nonzero @xmath61-parabolically homogeneous solution of the signorini problem in @xmath771 $ ] with @xmath772 .
then , @xmath65 and @xmath773 for some tangential direction @xmath774 .
extend @xmath581 by even symmetry in @xmath248 to the strip @xmath775 , i.e. , by putting @xmath776 take any @xmath774 , and consider the positive and negative parts of the directional derivative @xmath777 @xmath778 we claim that they satisfy the following conditions @xmath779 the last two conditions are obvious .
the first one follows from the fact that @xmath780 are continuous in @xmath650 ( by lemma [ lem : holder - hu - above ] ) and caloric where positive .
hence , we can apply caffarelli s monotonicity formula to the pair @xmath780 , see @xcite .
namely , the functional @xmath781 is monotone nondecreasing in @xmath565 . on the other hand , from the homogeneity of @xmath94 , it is easy to see that @xmath782 since @xmath783 , @xmath784 can be monotone increasing if and only if @xmath785 and consequently @xmath786 for all @xmath235 .
if fact , one has to exclude the possibility that @xmath787 as well .
this can be seen in two different ways .
first , by remark [ rem : w212-alt ] , one has @xmath788 alternatively , from theorem [ thm : exist - homogen - blowups ] ( i ) and ( iv ) it follows that @xmath789 then , @xmath790 since @xmath726 .
from here it follows that one of the functions @xmath780 is identically zero , which is equivalent to @xmath791 being ether nonnegative or nonpositive on the entire @xmath792 $ ] .
since this is true for any tangential direction @xmath793 , it thus follows that @xmath581 depends only on one tangential direction , and is monotone in that direction .
therefore , without loss of generality we may assume that @xmath38 and that the coincidence set at @xmath794 is an infinite interval @xmath795\times\{0\}=:\sigma_a^-.\ ] ] besides , repeating the monotonicity formula argument above for the pair of functions @xmath796 , where @xmath797 is a caloric function in @xmath798\setminus\lambda$ ] , parabolically homogeneous of degree @xmath799 , we obtain that also @xmath192 does not change sign in @xmath798 $ ] .
noting also that @xmath800 on @xmath801 $ ] , we see that @xmath800 everywhere(unless @xmath802 identically on @xmath803 $ ] , in which case @xmath192 is a polynomial of degree @xmath799 , which is impossible , since @xmath61 is noninteger ) .
hence , we get @xmath804.\ ] ] further if @xmath805 and using that @xmath806 is caloric we obtain that @xmath807 since also @xmath808 is nonnegative and not identically zero , then @xmath808 is the ground state for the ornstein - uhlenbeck operator in @xmath809 and @xmath810 on the other hand , let @xmath811 in @xmath812 and @xmath813 for @xmath814
. then , we have @xmath815 with @xmath816 .
( 100,100 ) ( 0,0 ) , @xmath817.,title="fig : " ] ( 40,43)@xmath818 ( 78,40 ) @xmath819 ( 10,40)@xmath820 ( 86,90)@xmath821 thus , this time @xmath822 is the ground state for the ornstein - uhlenbeck operator in @xmath823 , and therefore @xmath824 observe now that @xmath825 and therefore from the above equations we have @xmath826 on the other hand , it is easy to see that the function @xmath827 is strictly monotone and therefore the above equality can hold only if @xmath828 .
in particular , @xmath829 we now claim that @xmath830 .
indeed , consider the function @xmath831 which is harmonic and homogeneous of degree @xmath44 : @xmath832 therefore , @xmath833 also , since @xmath19 is nonnegative , we obtain that @xmath19 is the ground state of the ornstein - uhlenbeck operator in @xmath834 .
this implies in particular that @xmath835 , and consequently @xmath836 moreover , @xmath805 must be a multiple of the function @xmath19 above and from homogeneity we obtain that @xmath837 from here it is now easy to see that necessarily @xmath838 combining propositions [ prop : kappa>1alpha ] and [ prop : homogen - glob - sol-1 - 2 ] we obtain the following result . [ thm : min - homogen ] let @xmath265 and @xmath474 satisfies the conditions of theorem [ thm : thin - monotonicity ]
. assume also @xmath839 .
then @xmath840 more precisely , we must have @xmath841 very little is known about the possible values of @xmath61 .
however , we can say that the following values of @xmath61 do occur : @xmath842 this can be seen from the following explicit examples of homogeneous solutions ( that are actually @xmath55-independent ) @xmath843 it is known ( and easily proved ) that in @xmath55-independent case and dimension @xmath38 , the above listed values of @xmath61 are the only ones possible . in all other cases , finding the set of possible values of @xmath61 is , to the best of our knowledge , an open problem .
using the tools developed in the previous sections we are now ready to prove the optimal regularity of solutions of the parabolic signorini problem with sufficiently smooth obstacles .
in fact , we will establish our result for a slightly more general class of functions solving the signorini problem with nonzero obstacle and nonzero right - hand side .
[ thm : opt - reg ] let @xmath844 , @xmath161 .
assume that @xmath845 be such that @xmath227 for some @xmath228 , and satisfy @xmath846 then , @xmath847 with @xmath848 the proof of theorem [ thm : opt - reg ] will follow from the interior parabolic estimates and the growth bound of @xmath94 away from the free boundary @xmath32 .
[ lem : humr3 ] let @xmath265 with @xmath849 , @xmath850 .
then , @xmath851 the @xmath852 bound on @xmath108 allows us to apply theorem [ thm : thin - monotonicity ] with the following specific choice of @xmath474 in the generalized frequency function . indeed ,
fix @xmath853 and let @xmath854 . then , @xmath855 and therefore @xmath856 is monotone for @xmath857 . by theorem [ thm : min - homogen ]
we have that @xmath858 .
now , for @xmath478 we have two alternatives : either @xmath859 or @xmath488 . in the first case
the desired estimate is readily satisfied , so we concentrate on the latter case .
let @xmath503 be a maximal interval in the open set @xmath860 .
then , for @xmath861 we have @xmath862 we thus have , @xmath863 which , after integration , implies @xmath864 and therefore @xmath865 now , for @xmath464 we either have @xmath506 or @xmath505 .
note that @xmath866 from the @xmath852 bound on @xmath94 , and thus in both cases we have @xmath867 .
we thus have the desired conclusion @xmath868 to apply the results of the previous sections , we will need the following @xmath869 type estimates .
[ lem : l2linf - lower ] let @xmath192 be a nonnegative function with at most polynomial growth at infinity in the strip @xmath155 , and such that for some @xmath870 @xmath871 then , @xmath872 choosing a constant @xmath873 we can guarantee that @xmath874 is still nonnegative , has a polynomial growth at infinity and satisfies @xmath875 moreover , @xmath876 from the scaling properties of @xmath877 it is easily seen that @xmath878 and therefore we may assume that @xmath879 from the beginning .
the rest of the proof is now similar to that of theorem [ thm : exist - homogen - blowups](iii ) .
indeed , for @xmath880 and @xmath881 $ ] we have the sub mean - value property @xmath882 this can be proved as in theorem [ thm : exist - homogen - blowups](iii ) by combining the fact that @xmath883 for any @xmath884 , with the polynomial growth assumption on @xmath192 , and the energy inequality . on the other hand , for our choice of @xmath55 and @xmath885
, we have @xmath886 .
thus , arguing as in claim [ clm : rho - est ] , we have @xmath887 hence , we obtain that @xmath888 now , applying the cauchy - schwarz inequality , we will have @xmath889 where in the last step we have used that @xmath890 and @xmath891 . integrating over @xmath892
$ ] , we obtain @xmath893 this completes the proof .
[ lem : l2linf - upper ] let @xmath192 be a nonnegative bounded function in @xmath155 satisfying @xmath894 and let @xmath895 , @xmath896 be such that for some constant @xmath629 @xmath897 then , @xmath898 with @xmath899 . without loss of generality we may assume that @xmath900 .
then , from the assumptions on @xmath192 we have that @xmath901 and more generally @xmath902,\ ] ] where @xmath899 . to propagate the estimate to @xmath903 we use the sub - mean value property .
namely , for @xmath904 and @xmath905 we have @xmath906 to obtain the desired conclusion , we notice that the integral @xmath907 is parabolically homogeneous of degree @xmath113 in the sense that @xmath908 and therefore we immediately obtain that @xmath909 consequently , this yields @xmath910,\ ] ] which implies the statement of the lemma .
[ lem : growth - est ] let @xmath265 and @xmath911 .
then , @xmath912 with @xmath913 this is simply a combination of lemma [ lem : humr3 ] for @xmath94 and lemmas [ lem : l2linf - lower][lem : l2linf - upper ] applied for @xmath914 . as in section
[ sec : classes - solutions ] , let @xmath915 be a cutoff function satisfying , and consider @xmath216\psi(x).\ ] ] we may also assume that @xmath916
. we will thus have @xmath917 with @xmath918 in the remaining part of the proof , @xmath919 will denote a generic constant that has the same form as the right - hand side in the above inequality . for @xmath691 , and @xmath920),\ ] ] let @xmath921 we first claim that @xmath922 indeed , if @xmath923 , this follows from boundedness of @xmath94 . if instead @xmath924 , then there exist @xmath925 and by lemma [ lem : growth - est ] we have the desired estimate .
next , we claim that @xmath926 this will follow from the interior gradient estimates , applied to the even or odd extension of @xmath94 in @xmath248 variable .
more specifically , consider the intersection @xmath927 .
since there are no points of @xmath928 in this set , we have a dichotomy : either ( i ) @xmath929 on @xmath927 , or ( ii ) @xmath930 on @xmath931 .
accordingly , we define @xmath932 in either case @xmath933 satisfies a nonhomogeneous heat equation @xmath934 for an appropriately defined @xmath935 .
the claimed estimate for @xmath936 now follows from parabolic interior gradient estimates , see e.g. @xcite*chapter iii , theorem 11.1 .
moreover , by @xcite*chapter iv , theorem 9.1 , one also has that @xmath937 for any @xmath938 . to be more precise
, we apply the latter theorem to @xmath939 , where @xmath699 is a cutoff function supported in @xmath940 , such @xmath941 on @xmath942 , @xmath943 , @xmath944 , @xmath945 . from the estimates on @xmath933 and @xmath946
, we thus have @xmath947 which provides the estimate @xmath948 then , from the sobolev embedding of @xmath949 into @xmath950 when @xmath951 , we obtain the estimates for hlder seminorms @xmath952 for any @xmath228 , see @xcite*chapter ii , lemma 3.3 . in particular , we have @xmath953 now take two points @xmath954 , @xmath331 , and let @xmath955 . without loss of generality
we may assume @xmath956 .
let also @xmath957 .
consider two cases : \1 ) @xmath958 .
in this case , we have by @xmath959 \2 ) @xmath960 . in this case , both @xmath961 , and therefore by @xmath962 this gives the desired estimate for the seminorm @xmath963 .
arguing analogously , we can also prove a similar estimate for @xmath964 , thus completing the proof of the theorem .
after establishing the optimal regularity of the solutions , we are now able to undertake the study of the free boundary @xmath965 we start with classifying the free boundary points and more generally points in @xmath966 as we will see the higher is the regularity of @xmath18 , the finer is going to be the classification .
let @xmath246 with @xmath272 , @xmath273 , @xmath967 , @xmath968 and @xmath969 be as constructed in proposition [ prop : uk - def ] .
in particular , @xmath970 satisfies @xmath971 this implies that @xmath972 thus , if we choose @xmath973 then @xmath474 , @xmath970 and @xmath298 will satisfy the conditions of theorem [ thm : thin - monotonicity ] . in particular , we will have that @xmath974 is monotone increasing in @xmath335 and consequently there exists the limit ( see above ) @xmath975 recalling the definition of @xmath541 , we note that in the present case we have @xmath976 therefore , by lemma [ lem : hu - mu ] we infer that @xmath977 generally speaking , the value of @xmath61 may depend on the cutoff function that we have chosen to construct @xmath298 .
however , as the next result proves , it is relatively straightforward to check that this is not the case .
[ lem : kappa - indep - psi ] the limit @xmath978 does not depend on the choice of the cutoff function @xmath302 in the definition of @xmath298 . indeed ,
if we choose a different cutoff function @xmath979 , satisfying and denote by @xmath980 the function corresponding to @xmath298 in the construction above , and by @xmath981 the corresponding value as in , then by simply using the fact that @xmath982 on @xmath983 $ ] and that @xmath984 for @xmath985 and @xmath986 , we have @xmath987 to show now that @xmath988 , we consider several cases .
\1 ) if @xmath989 , then we are done .
\2 ) if @xmath990 , then lemma [ lem : hu - mu ] implies that @xmath991 integrating these inequalities we obtain @xmath992 for some ( generic ) positive constants @xmath993 , @xmath994 .
this will also imply @xmath995 now , if @xmath388 is so small that @xmath996 , we will have that @xmath997 for @xmath998 .
but then , we also have @xmath999 , and therefore @xmath1000 for arbitrarily small @xmath174 .
obviously , the above estimates imply that @xmath1001 .
\3 ) if @xmath1002 , we argue as in 2 ) above . to stress in the above construction the dependence only of the function @xmath19
, we will denote the quantity @xmath978 by @xmath1003 more generally , for @xmath1004 we let @xmath1005 which translates @xmath78 to the origin .
then , @xmath1006 for some small @xmath235 .
the construction above has been carried out in @xmath109 , rather than @xmath1007 . however , a simple rescaling argument generalizes it to any @xmath235 .
thus , we can define @xmath1008 which we will call the _ truncated homogeneity _ of @xmath19 at an extended free boundary point @xmath78 .
suppose now for a moment that the thin obstacle @xmath18 , that was assumed to belong to the class @xmath1009 , has a higher regularity .
to fix the ideas , suppose @xmath1010 , for some @xmath1011 , with @xmath1012 , @xmath1013 , @xmath1014 , @xmath1015 .
we may thus define @xmath1016 for any @xmath1017 .
it is natural to ask about the relation between @xmath1018 and @xmath1019 .
the following proposition provides an answer to this question .
[ prop : kappa - compat ] if @xmath1020 , @xmath1021 are as above , then @xmath1022 this proposition essentially says that @xmath1023 is the truncation of @xmath1024 by the value @xmath1025
. it will be sufficient to prove the statement for @xmath1026 . to simplify the notation in the proof
, we are going to denote @xmath1027 so we will need to show that @xmath1028 first , we fix a cutoff function @xmath302 is the definition of the functions @xmath298 and @xmath1029 , and note that @xmath1030 this implies that @xmath1031 arguing as in the proof of lemma [ lem : kappa - indep - psi ] , we obtain that , in fact , @xmath1032 using this information , form now on in this proof we will abbreviate @xmath1033 with @xmath1034 .
we consider two cases .
[ step:1-trunc - kappa](@xmath362 ) assume first that @xmath1035 .
in this case we need to show that @xmath1036 .
from the assumption we will have that @xmath1037 and by lemma [ lem : hu - mu ] @xmath1038 therefore , for any @xmath174 we obtain @xmath1039 integrating , we find @xmath1040 in particular , if @xmath174 is so small that @xmath1041 , then @xmath1042 and therefore @xmath1043 [ step:2-trunc - kappa](@xmath362 ) assume now that @xmath1044 .
we need to show in this case that @xmath1045 . in general , we know that @xmath1046 , so arguing by contradiction , assume @xmath990 .
we thus know by lemma [ lem : hu - mu ] that @xmath1047 for @xmath1048 , and @xmath1049 but then , we also have @xmath1050 for @xmath1048 and therefore @xmath1051 contrary to the assumption .
[ def : trunc - hom - ii ] in view of proposition [ prop : kappa - compat ] , if @xmath308 we can push @xmath1025 in the definition of the truncated homogeneity up to @xmath1052 by setting @xmath1053 indeed , proposition [ prop : kappa - compat ] guarantees that @xmath1054 is monotone increasing .
moreover , we have @xmath1055 [ lem : kappa - upp - semi ] the function @xmath1056 is upper semicontinuous on @xmath58 ( with respect to euclidean or , equivalently , parabolic distance ) , i.e. , for any @xmath1004 one has @xmath1057 suppose first @xmath1058 and fix @xmath1059 .
then , for any @xmath174 there exists @xmath563 such that @xmath1060 , where @xmath1061 .
this implies that that @xmath1062 for @xmath998 .
since the mapping @xmath1063 is continuous on @xmath58 , we will have @xmath1064 if @xmath1065 , @xmath1066 , provided @xmath566 and @xmath1067 are small enough .
in particular , this implies the explicit formula @xmath1068 therefore , taking @xmath1067 small , we can guarantee @xmath1069 if @xmath1065 , @xmath1066 .
it follows that , for such @xmath83 , one has @xmath1070 which implies the upper semicontinuity of @xmath1071 and @xmath1072 at @xmath78 . if @xmath1073 , the upper continuity follows immediately since @xmath1074 .
the truncated homogeneity @xmath1072 gives a natural classification of extended free boundary points .
let @xmath246 , with @xmath272 . for @xmath1075
$ ] , we define @xmath1076 as a direct corollary of proposition [ prop : kappa - compat ] , we have the following consistency for the above definition .
if @xmath1077 with @xmath1078 , then @xmath1079 the latter identity essentially means that , if @xmath18 is more regular than @xmath56 , then @xmath1080 is an `` aggregate '' of points with higher homogeneities @xmath61 .
we conclude this section with the following description of the free boundary , based on the fact that the function @xmath1072 never takes certain values .
we also characterize the points that are in the extended free boundary @xmath58 , but not in the free boundary @xmath32 .
[ prop : free - bound - struct ] if @xmath246 with @xmath308 , @xmath57 , then for any @xmath1081 , either we have @xmath1082 as a consequence , @xmath1083 moreover , @xmath1084 the first part is nothing but theorem [ thm : min - homogen ] .
suppose now @xmath1085 and that @xmath1058 .
then , there exists a small @xmath424 such that @xmath1086 on @xmath1087 .
next , consider the translate @xmath1088 and let @xmath1061 be obtained from @xmath1089 as in proposition [ prop : uk - def ] .
since @xmath60 , by theorem [ thm : exist - homogen - blowups ] , there exists a blowup @xmath581 of @xmath94 over some sequence @xmath535 . since @xmath930 on @xmath1090
, @xmath581 will vanish on @xmath1091 .
hence , extending it as an odd function @xmath1092 of @xmath248 from @xmath761 to @xmath775 , we will obtain a homogeneous caloric function in @xmath775 .
then , by the liouville theorem , @xmath1092 must be a caloric polynomial of degree @xmath61 .
thus , @xmath61 is an integer .
we further claim that @xmath61 is odd .
indeed , @xmath1092 solves the signorini problem in @xmath761 and therefore we must have that @xmath1093 is a nonnegative polynomial on @xmath1094 of homogeneity @xmath799 .
the latter is possible when either @xmath799 is even or if @xmath1095 vanishes on @xmath1094 .
however , the latter case is impossible , since otherwise holmgren s uniqueness theorem would imply that @xmath581 is identically zero , contrary to theorem [ thm : exist - homogen - blowups ] .
thus , the only possibility is that @xmath799 is even , or equivalently , @xmath61 is odd .
since we also have @xmath1096 , we obtain that @xmath1097 .
it is easy to construct @xmath1098 such that @xmath1099 and @xmath1100 .
the simplest example is perhaps @xmath1101 it is easy to verify that @xmath1102 , and @xmath1103 on @xmath180 .
thus , @xmath1099 . however , @xmath1104 $ ] , and because of the @xmath1105-homogeneity of @xmath19 with respect to any point on @xmath58 , if we choose @xmath1106 , we have that @xmath1107 .
in this section we study a special subset @xmath48 of the extended free boundary .
namely , the collection of those points having minimal frequency @xmath65 . let @xmath246 with @xmath272 , @xmath57 .
we say that @xmath1108 is a _ regular _ free boundary point if it has a minimal homogeneity @xmath65 , or equivalently @xmath1109 .
the set @xmath1110 will be called the _ regular set _ of @xmath19 .
we have the following basic fact about @xmath48 . [
prop : reg - set - rel - open ] the regular set @xmath48 is a relatively open subset of @xmath32 . in particular , for any @xmath1111 there exists @xmath1112 such that @xmath1113 first note that , by proposition [ prop : free - bound - struct ] , we have @xmath1114 .
the relative openness of @xmath48 follows from the upper semicontinuity of the function @xmath1115 and form the fact that it does not take any values between @xmath113 and @xmath1116 .
we will show in this section that , if the thin obstacle @xmath18 is sufficiently smooth , then the regular set can be represented locally as a @xmath1117-dimensional graph of a parabolically lipschitz function .
further , such function can be shown to have hlder continuous spatial derivatives .
we begin with the following basic result .
[ thm : lip - reg - reg - set ] let @xmath246 with @xmath272 , @xmath1118 and that @xmath1119 .
then , there exist @xmath1120 , and @xmath1121 ( i.e. , @xmath25 is a parabolically lipschitz function ) , such that possibly after a rotation in @xmath67 , one has @xmath1122 ( 150,150 ) ( 12,0 ) in @xmath1090 , given by the graph @xmath1123 with @xmath1121 and @xmath1124 by theorems [ thm : lip - reg - reg - set ] and [ thm : h1alpha - reg - set],title="fig : " ] ( 50,80 ) ( 45,100 ) ( 92,65)@xmath1125 ( 90,105)@xmath48 ( 100,70)(-1,1)10 for an illustration , see fig [ fig : reg - set ] .
following the well - known approach in the classical obstacle problem , see e.g.@xcite*chapter 4 , the idea of the proof is to show that there is a cone of directions in the thin space , along which @xmath66 is increasing .
this approach was successfully used in the elliptic signorini problem in @xcite , @xcite , see also @xcite*chapter 9 , and in the arguments below we generalize the constructions in these papers to the parabolic case .
this will establish the lipschitz regularity in the space variables . to show the @xmath44-hlder regularity in @xmath55 ( actually better than that )
, we will use the fact that the @xmath113-homogeneous solutions of the parabolic signorini problem are @xmath55-independent ( see proposition [ prop : homogen - glob - sol-1 - 2 ] ) . however
, in order to carry out the program outlined above , in addition to ( i ) and ( ii ) in theorem [ thm : exist - homogen - blowups ] above , we will need a stronger convergence of the rescalings @xmath533 to the blowups @xmath581 .
this will be achieved by assuming a slight increase in the regularity assumptions on the thin obstacle @xmath18 , and , consequently , on the regularity of the right - hand side @xmath108 in the construction of proposition [ prop : uk - def ] .
[ lem : blowups - h1alpha - conv ] let @xmath265 , and suppose that for some @xmath1126 @xmath1127 and @xmath1128 then , for the family of rescalings @xmath1129 we have the uniform bounds @xmath1130 in particular , if the sequence of rescalings @xmath636 converges to @xmath581 as in theorem [ thm : exist - homogen - blowups ] , then over a subsequence @xmath1131 for any @xmath1132 and @xmath149 . because of theorem [ thm : opt - reg ] , it is enough to show that @xmath533 , @xmath1133 , and @xmath1134 are bounded in @xmath238 .
we have @xmath1135 then , the functions @xmath1136 satisfy @xmath1137 by lemma [ lem : l2linf - lower ] we thus obtain @xmath1138 for small @xmath565 .
then , by the energy inequality for @xmath1139 in @xmath1140 , we have @xmath1141 on the other hand , using that for @xmath197 , @xmath1142 satisfy @xmath1143 then from @xmath869 estimate for subcaloric functions , we obtain @xmath1144 thus , @xmath533 , @xmath1133 and @xmath1134 are uniformly bounded in @xmath1145 for small @xmath1146 , and this completes the proof of the lemma .
the next lemma will allow to deduce the monotonicity of the solution @xmath94 in a cone of directions in the thin space , from that of the blowup .
it is the parabolic counterpart of @xcite*lemma 4 and @xcite*lemma 7.2 .
[ lem : glob - local ] let @xmath1147 be a closed subset of @xmath803 $ ] , and @xmath1148 a continuous function in @xmath1149
. for any @xmath1112 there exists @xmath1150 , depending only on @xmath1151 and @xmath254 , such that if 1 .
@xmath1152 on @xmath1153 , 2 .
@xmath1154 in @xmath1155 , 3 .
@xmath1156 in @xmath247 , 4 .
@xmath1157 in @xmath1158 , @xmath1159 , then @xmath1152 on @xmath1160 .
it is enough to show that @xmath1152 on @xmath1161 .
arguing by contradiction , let @xmath1162 be such that @xmath1163 .
consider the auxiliary function @xmath1164 where @xmath1165 .
it is immediate to check that @xmath1166 now , consider the function @xmath192 in the set @xmath1167 . by the maximum principle
, we must have @xmath1168 analyzing the different parts of @xmath1169 we show that this inequality can not hold : \1 ) on @xmath1170 we have @xmath800 .
\2 ) on @xmath1171 we have @xmath1172 if @xmath1173 .
\3 ) on @xmath1174 we have @xmath1175 with @xmath1176 if we choose @xmath1177 , we conclude that @xmath1178 on this portion of @xmath1169 .
\4 ) on @xmath1179 we have @xmath1180 for @xmath1181 . in conclusion ,
if @xmath1182 is sufficiently small , we see that we must have @xmath1183 , thus arriving at a contradiction with the assumption that @xmath1163 .
this completes the proof .
let @xmath1184 and @xmath1185 be as in proposition [ prop : uk - def ] . from the assumption @xmath1118
, we have that @xmath1186 in @xmath110 and @xmath1187 in @xmath1188 .
we also choose @xmath1189 .
we thus conclude that @xmath1190 for @xmath591 . in view of theorem [ thm : exist - homogen - blowups ] , the rescalings @xmath636 converge ( over a sequence @xmath535 ) to a homogeneous global solution @xmath581 of degree @xmath113 .
furthermore , we note that lemma [ lem : blowups - h1alpha - conv ] is also applicable here . in view of proposition
[ prop : homogen - glob - sol-1 - 2 ] , after a possible rotation in @xmath67 , we may assume that @xmath1191 it can be directly calculated that for any @xmath1192 @xmath1193 thus , if for any given @xmath1194 we consider the thin cone around @xmath1195 @xmath1196 then it is immediate to conclude that for any @xmath1197 , @xmath1198 , @xmath1199 where @xmath1159 is the dimensional constant in lemma [ lem : glob - local ] .
we next observe that , by lemma [ lem : blowups - h1alpha - conv ] , for any given @xmath174 we will have for all directions @xmath1192 @xmath1200 provided @xmath1201 is sufficiently large . moreover ,
note that in view of proposition [ prop : uk - def ] we can estimate @xmath1202 thus , the function @xmath1203 ( evenly reflected to @xmath247 ) will satisfy the conditions of lemma [ lem : glob - local ] , and therefore we conclude that @xmath1204 for @xmath1205 . scaling back ,
we obtain that @xmath1206 where @xmath1207 .
now a standard argument ( see @xcite*chapter 4 , exercise 4.1 ) implies that @xmath1208 where , for every fixed @xmath1209 $ ] , @xmath1210 is a lipschitz continuous function with @xmath1211 we are now left with showing that @xmath25 is @xmath1212-hlder continuous in @xmath55 .
in fact , we are going to show that @xmath1213 , uniformly in @xmath1214 .
suppose , towards a contradiction , that for @xmath1215 , @xmath1216 , @xmath1217 , we have for some @xmath483 @xmath1218 let @xmath1219 and @xmath1220 let also @xmath1221 note that @xmath1222 moreover , we claim that @xmath1223 .
indeed , we may assume that the sequences @xmath1224 , @xmath1225 , @xmath1226 , @xmath1227 converge to some @xmath85 , @xmath1228 , @xmath55 , @xmath463 respectively .
if @xmath424 then we obtain that @xmath1229 . but
@xmath1230 , which can not happen since @xmath32 is given as a graph @xmath1231 in @xmath1232 .
thus , @xmath1223 .
consider now the rescalings of @xmath94 at @xmath1233 by the factor of @xmath1227 : @xmath1234 we want to show that the sequence @xmath1235 converges to a homogeneous global solution in @xmath775 , of homogeneity @xmath113 .
for that purpose , we first assume that @xmath1236 is so small that @xmath1237 this is possible by the upper semicontinuity of the mapping @xmath1238 on @xmath928 as in lemma [ lem : kappa - upp - semi ] , the equality @xmath1239 , and theorem [ thm : min - homogen ] . moreover , arguing as in the proof of lemma [ lem : kappa - upp - semi ] , we may assume that @xmath1240 this assumption implies @xmath1241 otherwise , we would have @xmath1242 , a contradiction . as a consequence ,
the functions @xmath1243 will have an explicit representation through @xmath1244 and its derivatives , and therefore will be continuous .
we thus have a monotone family of continuous functions @xmath1245 on a compact set @xmath1246 such that @xmath1247 by the theorem of dini the convergence @xmath1248 is uniform on @xmath1249 .
this implies that @xmath1250 for the functions @xmath1235 defined in above this implies @xmath1251 for any @xmath235 .
now , analyzing the proof of theorem [ thm : exist - homogen - blowups ] , we realize that the same conclusions can be drawn about the sequence @xmath1235 as for the sequence of rescalings @xmath636 .
in particular , over a subsequence , we have @xmath1252 in @xmath1253 , where @xmath1254 is a @xmath113-homogeneous global solution of the signorini problem . by proposition [ prop : homogen - glob - sol-1 - 2 ]
we conclude that for some direction @xmath1255 it must be @xmath1256 further , since @xmath1257 for unit @xmath1258 , we must have @xmath1259 therefore , @xmath1260 further , note that since @xmath479 with @xmath1186 in @xmath110 , and @xmath1187 in @xmath1188 , we can repeat the arguments in the proof of lemma [ lem : blowups - h1alpha - conv ] ( with @xmath1189 ) to obtain for any @xmath149 @xmath1261 going back to the construction of the functions @xmath1235 , note that @xmath1262 , in addition to @xmath1263 .
without loss of generality we may assume that @xmath1264 .
but then the convergence implies that @xmath1265 and @xmath1266 . from the explicit formula for @xmath1254
it follows that @xmath1267 or equivalently , @xmath1268 . since @xmath1269
, we must have @xmath1270 .
thus , we have proved that @xmath1271 which is equivalent to @xmath1272 contrary to our assumption .
we next show that , following an idea in @xcite , the regularity of the function @xmath25 can be improved with an application of a boundary harnack principle .
[ thm : h1alpha - reg - set ] in the conclusion of theorem [ thm : lip - reg - reg - set ] , one can take @xmath424 so that @xmath1273 for some @xmath106 . to prove this theorem we first show the following nondegeneracy property of @xmath1274 .
[ prop : nondeg - deu ] let @xmath307 and @xmath1184 be as theorem [ thm : lip - reg - reg - set ] . then , for any @xmath1194 there exist @xmath424 and @xmath1275 such that @xmath1276 where @xmath1277 is the parabolic distance from the point @xmath83 to the coincidence set @xmath1278 . the proof is based on the following improvement on lemma [ lem : glob - local ] , which is the parabolic counterpart of @xcite*lemma 7.3 .
[ lem : h - nondeg ] for any @xmath1112 there exist @xmath1150 and @xmath1279 , depending only on @xmath1151 and @xmath254 , such that if @xmath52 is a continuous function on @xmath1280 , @xmath1159 , satisfying 1 .
@xmath1281 in @xmath1282 , 2 .
@xmath1152 in @xmath1282 , 3 .
@xmath1283 , on @xmath1284 , then @xmath1285 the proof is very similar to that of lemma [ lem : glob - local ] .
let @xmath1286 , and consider the auxiliary function @xmath1287 with @xmath1165 . as before
, we have @xmath1288 in @xmath1282 .
we now claim that @xmath800 on @xmath1289 .
this will follow once we verify that @xmath800 on @xmath1290 .
we consider several cases : \1 ) on @xmath1291 , we clearly have @xmath800 .
\2 ) on @xmath1292 , one has @xmath1293 provided @xmath1294 and @xmath1295 .
\3 ) on @xmath1296 we have @xmath1297 provided @xmath1298 .
\4 ) on @xmath1299 we have @xmath1300 provided @xmath1301 . in conclusion , for small enough @xmath1182 and @xmath1302 , we have @xmath1178 in @xmath1303 , and in particular @xmath1304 .
this implies that @xmath1305 , as claimed .
considering the rescalings @xmath533 as in the proof of theorem [ thm : lip - reg - reg - set ] , and applying lemma [ lem : h - nondeg ] , we obtain @xmath1306 for @xmath1307 small .
( here , we assume that @xmath94 has been extended by even symmetry in @xmath248 variable to @xmath247 . ) besides , by choosing @xmath565 small , we can also make @xmath1308 for a dimensional constant @xmath1309 to be specified below .
let now @xmath1310 and @xmath1311 be the parabolic distance from @xmath83 to @xmath1312 . consider the lowest rightmost point on the boundary @xmath1313 @xmath1314 we have @xmath1315 by the parabolic harnack inequality ( see , e.g. , @xcite*theorems 6.176.18 ) @xmath1316 if we take @xmath1317 sufficiently small .
scaling back to @xmath94 , we complete the proof of the proposition .
a key ingredient in the proof of theorem [ thm : h1alpha - reg - set ] is the following version of the parabolic boundary harnack principle for domains with thin lipschitz complements established in @xcite*section 7 . to state the result
, we will need the following notations . for a given @xmath1318 and @xmath235
denote @xmath1319 [ lem : bhp ] let @xmath1320 for a parabolically lipschitz function @xmath25 in @xmath1321 with lipschitz constant @xmath1318 such that @xmath1322 .
let @xmath1323 , @xmath1324 be two continuous nonnegative functions in @xmath1325 such that for some positive constants @xmath1302 , @xmath1326 , @xmath301 , and @xmath331 , 1 .
@xmath1327 in @xmath1328 and @xmath1329 on @xmath1147 , 2 .
@xmath1330 in @xmath1331 , 3 .
@xmath1332 in @xmath1333 , where @xmath1334 .
assume additionally that @xmath1323 and @xmath1324 are symmetric in @xmath248 .
then , there exists @xmath37 such that @xmath1335 furthermore , @xmath35 and the bound on the corresponding norm @xmath1336 depend only on @xmath254 , @xmath1337 , @xmath1302 , @xmath1326 , and @xmath301 .
we note that , unlike the elliptic case , lemma [ lem : bhp ] above can not be reduced to the other known results in the parabolic setting ( see , e.g. , @xcite , @xcite for parabolically lipschitz domains , or @xcite for parabolically nta domains with reifenberg flat boundary ) .
we also note that this version of the boundary harnack is for functions with nonzero right - hand side and therefore the nondegeneracy condition as in iii ) is necessary .
the elliptic version of this result has been established in @xcite .
we are now ready to prove theorem [ thm : h1alpha - reg - set ] . fix @xmath1194 and let @xmath1338 be such that @xmath1339 for @xmath1340 .
consider two functions @xmath1341 then , by proposition [ prop : nondeg - deu ] , the conditions of lemma [ lem : bhp ] are satisfied for some rescalings of @xmath1323 and @xmath1324 .
hence , applying lemma [ lem : bhp ] and scaling back , we obtain that for a small @xmath424 , and @xmath37 , @xmath1342 this gives @xmath1343 hence , the level surfaces @xmath1344 are given as graphs @xmath1345 with uniform in @xmath388 estimates on @xmath1346 for small @xmath174 .
consequently , we obtain @xmath1347 and this completes the proof of the theorem .
we now turn to the study of a special class of free boundary points , called _ singular points _ , that are characterized by the property that the coincidence set has a zero density at those points with respect to the @xmath253 measure in the thin space . in the time - independent signorini problem
the analysis of the singular set was carried in the paper @xcite .
[ def : parab - sing - points ] let @xmath246 with @xmath272 , @xmath57 .
we say that @xmath1108 is _ singular _ if @xmath1348 we will denote the set of singular points by @xmath1349 and call it the _ singular set_. we can further classify singular points according to the homogeneity of their blowup , by defining @xmath1350 since we are going to work with the blowups , we will write @xmath1351 , for @xmath274 , @xmath134 , and construct the functions @xmath1352 as in proposition [ prop : uk - def ] . by abusing the notation
, we will write @xmath1353 whenever @xmath1354 .
also , for technical reasons , similarly to what we did in the study of the regular set , we will assume that @xmath1118 for most of the results in this section .
the following proposition gives a complete characterization of the singular points in terms of the blowups and the generalized frequency .
in particular , it establishes that @xmath1355 [ prop : char - sing - point ] let @xmath479 with @xmath1356 in @xmath110 , @xmath1357 in @xmath1188 , @xmath1118 and @xmath1358 with @xmath60 .
then , the following statements are equivalent : 1 . @xmath1359 .
any blowup of @xmath94 at the origin is a nonzero parabolically @xmath61-homogeneous polynomial @xmath1360 in @xmath775 satisfying @xmath1361 we denote this class by @xmath1362 .
3 . @xmath76 , @xmath77 .
\(i ) @xmath1363 ( ii ) recall that the rescalings @xmath533 satisfy @xmath1364 in the sense of distributions , after an even reflection in @xmath248 variable .
since @xmath533 are uniformly bounded in @xmath1365 for small @xmath565 by theorem [ thm : exist - homogen - blowups ] , @xmath1366 are uniformly bounded in @xmath1367 . on the other hand , if @xmath1368 , then @xmath1369 and therefore @xmath1370 in the sense of distributions .
further , the bound @xmath1371 implies that @xmath1372 where @xmath1059 and we have used the fact that @xmath1190 for @xmath591 .
hence , any blowup @xmath581 is caloric in @xmath1373 for any @xmath149 , meaning that it is caloric in the entire @xmath536 $ ] . on the other hand , by proposition [ thm : exist - homogen - blowups](iv )
, the blowup @xmath581 is homogeneous in @xmath775 and therefore has a polynomial growth at infinity .
then by the liouville theorem we can conclude that @xmath581 must be a homogeneous caloric polynomial @xmath1360 of a certain integer degree @xmath61 .
note that @xmath1374 by construction .
the properties of @xmath94 also imply that that @xmath1375 for all @xmath1376 and and @xmath1377 for all @xmath1378 .
\(ii ) @xmath1363 ( iii ) let @xmath1360 be a blowup of @xmath94 at the origin . since @xmath1360 is a polynomial , clearly @xmath1379 .
if @xmath61 is odd , the nonnegativity of @xmath1360 on @xmath1380 implies that @xmath1360 vanishes there identically , implying that @xmath1381 on @xmath1094 .
on the other hand , from the even symmetry in @xmath248 we also have that @xmath1382 on @xmath1094 . since @xmath1360 is caloric in @xmath1 and @xmath1094
is not characteristic for the heat operator , holmgren s uniqueness theorem implies that @xmath1383 in @xmath1 , contrary to the assumption .
thus , @xmath1384 .
\(iii ) @xmath1363 ( ii ) the proof of this implication is stated as a separate liouville - type result in lemma [ lem:2m - homogen ] below .
\(ii ) @xmath1363 ( i ) suppose that @xmath1385 is not a singular point and that over some sequence @xmath535 we have @xmath1386 .
by lemma [ lem : blowups - h1alpha - conv ] , taking a subsequence if necessary , we may assume that @xmath636 converges locally uniformly to a blowup @xmath581 .
we claim that @xmath1387 indeed , otherwise there exists an open set @xmath1303 in @xmath1094 with @xmath1388 such that @xmath1389 .
then for large @xmath1201 we must have @xmath1390 , which is a contradiction , since @xmath1391 . since @xmath1392 is a polynomial , vanishing on a set of positive @xmath253-measure on @xmath1094
, it follows that @xmath581 vanishes identically on @xmath1094 .
but then , applying holmgren s uniqueness theorem one more time , we conclude that @xmath581 must vanish on @xmath775 , which is a contradiction .
this completes the proof of the theorem .
the implication ( iii ) @xmath1363 ( ii ) in proposition [ prop : char - sing - point ] is equivalent to the following liouville - type result , which is the parabolic counterpart of lemma 1.3.3 in @xcite .
[ lem:2m - homogen ] let @xmath19 be a parabolically @xmath61-homogeneous solution of the parabolic signorini problem in @xmath775 with @xmath76 for @xmath77 .
then @xmath19 is a caloric polynomial .
this , in turn , is a particular case of the following lemma , analogous to lemma 1.3.4 in @xcite in the elliptic case , which stems from lemma 7.6 in @xcite .
[ lem : monn - homogen - harm ] let @xmath1393 be such that @xmath1394 in @xmath775 and @xmath1395 in @xmath1396 .
if @xmath19 is parabolically @xmath1397-homogeneous , @xmath77 , and has a polynomial growth at infinity , then @xmath1395 in @xmath775 .
consider @xmath1398 in @xmath650 . by the assumptions
, @xmath474 is a nonpositive measure , supported on @xmath1399 .
we are going to show that in fact @xmath1400 . to this end , let @xmath1401 be a parabolically @xmath1397-homogeneous caloric polynomial , which is positive on @xmath1399 .
for instance , one can take @xmath1402 it is straightforward to check that @xmath1403 is caloric .
moreover on @xmath1399 we have @xmath1404 so it is positive on @xmath1399 .
further , let @xmath1405 , with @xmath1406 , and define @xmath1407 note that we have the following identity ( similar to that of @xmath1408 ) @xmath1409 we have @xmath1410\,dx\,dt\\ & = \int_{-\infty}^0\int_{{\mathbb{r}}^n}[\psi v\delta p+v\nabla\psi\cdot\nabla p - p\nabla v\cdot\nabla\psi]\,dx\,dt\\ & = \int_{-\infty}^0\int_{{\mathbb{r}}^n}\left[v\delta p+\frac{1}{2 t } v(x\cdot\nabla p)-\frac{1}{2t}p(x\cdot\nabla v)\right]\psi\,dx\,dt.\end{aligned}\ ] ] we now use the identities @xmath1411 , @xmath1412 , @xmath1413 to arrive at @xmath1414\frac{\psi}{2t}\,dx\,dt\\ & = \int_{-\infty}^0\int_{{\mathbb{r}}^n}\partial_t v\psi p\,dx\,dt\\ & = \langle\partial_t v,\psi p\rangle.\end{aligned}\ ] ] therefore , @xmath1415 .
since @xmath474 is a nonpositive measure , this implies that actually @xmath1400 and the proof is complete .
[ def : pk ] throughout the rest of the paper we denote by @xmath1362 , @xmath76 , @xmath1416 , the class of @xmath61-homogeneous harmonic polynomials described in proposition [ prop : char - sing - point](ii ) . in the rest of this section we state our main results concerning the singular set : @xmath61-differentiability at singular points ( theorem [ thm : k - diff - sing - p ] ) and a structural theorem on the singular set ( theorem [ thm : sing - points - nonzero ] )
. the proofs will require additional technical tools ( two families of monotonicity formulas ) that we develop in the next section .
the proofs themselves will be given in section [ sec : struct - sing - set ] .
[ thm : k - diff - sing - p ] let @xmath479 with @xmath1417 in @xmath110 , @xmath1357 in @xmath1188 , @xmath1118 , and @xmath1418 for @xmath1419 , @xmath77 .
then , there exists a nonzero @xmath1420 such that @xmath1421 moreover , if @xmath1422 with @xmath308 , @xmath1423 and @xmath1424 is obtained as in proposition [ prop : uk - def ] for @xmath1425 , then in the taylor expansion @xmath1426 the mapping @xmath1427 from @xmath1428 to @xmath1362 is continuous .
[ rem : pk - norm]note that since @xmath1362 is a convex subset of a finite - dimensional vector space , namely the space of all @xmath61-homogeneous polynomials , its topology can be induced from any norm on that space .
for instance , the topology can be induced from the embedding of @xmath1362 into @xmath1429 , or even into @xmath1430 , since the elements of @xmath1362 can be uniquely recovered from their restriction to the thin space . [
rem : taylor - cntrex ] we want to emphasize here that the asymptotic development , as stated in theorem [ thm : k - diff - sing - p ] , does not generally hold for @xmath1431 .
indeed , consider the following example .
let @xmath1432 be a continuous function such that * @xmath1433 for @xmath904 and @xmath1434 .
* in @xmath1435 , @xmath94 solves the dirichlet problem @xmath1436 * in @xmath1437 , we extend the function by even symmetry in @xmath248 : @xmath1438 it is easy to see that @xmath94 solves the parabolic signorini problem with zero obstacle and zero right - hand side in all of @xmath116 . moreover , @xmath94 is homogeneous of degree two and clearly @xmath1439 .
now , if @xmath1440 , then @xmath1441 and we have the following equalities : @xmath1442 so for @xmath1443 the difference @xmath1444 is not @xmath1445 , despite being zero for @xmath1434 .
we next want to state a structural theorem for the singular set , similar to the ones in @xcite for the classical obstacle problem and @xcite for the thin obstacle problem . in order to do so
, we define the spatial dimension @xmath1446 of @xmath1428 at a given point @xmath78 based on the polynomial @xmath1447 .
[ def : dim - sing - point ] for a singular point @xmath1423 we define @xmath1448 which we call the _ spatial dimension _ of @xmath1428 at @xmath78 .
clearly , @xmath1449 is an integer between @xmath1385 and @xmath1450 .
then , for any @xmath1451 define @xmath1452 the case @xmath1453 deserves a special attention .
[ lem : timelike - blowup ] let @xmath1454 , @xmath1419 .
then @xmath1455 for some positive constant @xmath919 .
in other words , @xmath1456 depends only on @xmath248 and @xmath55 and is unique up to a multiplicative factor .
we call such singular points _ time - like_. the condition @xmath1457 is equivalent to the following property of @xmath1458 : @xmath1459 for any multi - index @xmath1460 and @xmath1201 such that @xmath1461 .
it is easy to see that this is equivalent to vanishing of @xmath1462 on @xmath1094 for @xmath1463 . on the other hand
, @xmath1462 is caloric in @xmath775 and is also even symmetric in @xmath248 variable , implying that @xmath1464 on @xmath1094 .
then , by holmgren s uniqueness theorem @xmath1462 is identically @xmath1385 in @xmath775 , implying that @xmath1465 depends only on @xmath248 and @xmath55 .
the homogeneity of @xmath1360 implies that we can write it in the form @xmath1466 the rest of the proof is then elementary .
we say that a @xmath1467-dimensional manifold @xmath1468 , @xmath1469 , is _ space - like _ of class @xmath89 , if locally , after a rotation of coordinate axes in @xmath67 one can represent it as a graph @xmath91 where @xmath25 is of class @xmath89 , i.e. , @xmath25 and @xmath71 , @xmath92 are continuous .
we say that @xmath1470-dimensional manifold @xmath1468 is _ time - like _ of class @xmath87 if it can be represented locally as @xmath86 where @xmath25 is of class @xmath87 .
( 150,150 ) ( 0,0)$ ] for the solution @xmath19 with @xmath1471 , @xmath1434 with zero thin obstacle .
note that the points on @xmath1472 and @xmath1473 are space - like , and the points on @xmath1474 are time - like.,title="fig : " ] ( 120,90)@xmath1474 ( 60,110 ) @xmath1474 ( 72,40)@xmath1473 ( 75,50)(0,2)30 ( 125,30)@xmath1472 ( 25,50)@xmath1472 ( 162,40)@xmath1475 ( 180,105)@xmath346 [ thm : sing - points - nonzero ] let @xmath307 with @xmath272 , @xmath1476 . then , for any @xmath1419 , @xmath77 , we have @xmath1477 . moreover , for every @xmath1478 , the set @xmath1479 is contained in a countable union of @xmath1467-dimensional space - like @xmath89 manifolds and @xmath1480 is contained in a countable union of @xmath1470-dimensional time - like @xmath87 manifolds . for a small illustration ,
see fig .
[ fig : sing - set - ex ] .
in this section we construct two families of monotonicity formulas that will play a crucial role in the study of the singular set .
they generalize the corresponding formulas in @xcite in the study of the elliptic thin obstacle problem .
the first family of monotonicity formulas goes back to the work of weiss @xcite in the elliptic case and @xcite in the parabolic case ; see also @xcite .
[ thm : weiss ] let @xmath479 with @xmath1417 in @xmath110 , @xmath57 .
for any @xmath1481 , define the weiss energy functional @xmath1482 then , for any @xmath1483 there exists @xmath483 depending only on @xmath484 , @xmath1052 , @xmath301 , and @xmath254 , such that @xmath1484 in particular , the function @xmath1485 is monotonically nondecreasing for @xmath335 .
the proof is by direct computation using proposition [ prop : diff - form - signor ] .
we have @xmath1486 hence , using the integral estimate on @xmath108 as at the beginning of section [ sec : class - free - bound ] , we obtain @xmath1487 which yields the desired conclusion .
note that in theorem [ thm : weiss ] , we do not require @xmath1488 .
however , if we do so , then we will have the following fact .
[ lem : w0 + ] let @xmath94 be as in theorem [ thm : weiss ] , and assume additionally that @xmath1489 , @xmath60
. then , @xmath1490 the proof will require the following growth estimate , which we will use a few more times in the remaining part of the paper .
[ lem : humr2kappa ] let @xmath265 with @xmath1417 , @xmath1491 , and @xmath1492 with @xmath60 and let @xmath1483 .
then @xmath1493 with @xmath919 depending only on @xmath484 , @xmath1052 , @xmath254 .
take @xmath1494 and proceed as in the proof of lemma [ lem : humr3 ] .
we omit the details .
we can now prove lemma [ lem : w0 + ] .
since @xmath1495 , by lemma [ lem : hu - mu ] we have @xmath1496 further , by lemma [ lem : humr2kappa ] above , we will have @xmath1497 hence , we obtain @xmath1498 the next monotonicity formula is specifically tailored for singular points .
it goes back to the paper of monneau @xcite for the classical obstacle problem .
the theorem below is the parabolic counterpart of the monneau - type monotonicity formula in @xcite .
[ thm : monneau ] let @xmath479 with @xmath1417 in @xmath110 , @xmath1357 in @xmath1188 , @xmath1118 .
suppose that @xmath1499 with @xmath1419 , @xmath77 .
further , let @xmath1360 be any parabolically @xmath61-homogeneous caloric polynomial from class @xmath1362 as in definition [ def : pk ]
. for any such @xmath1360 , define monneau s functional as @xmath1500 then , for any @xmath1483 there exists a constant @xmath919 , depending only on @xmath484 , @xmath1052 , @xmath301 , and @xmath254 , such that @xmath1501 in particular , the function @xmath1502 is monotonically nondecreasing for @xmath493 for a constant @xmath919 depending @xmath484 , @xmath1052 , @xmath301 , @xmath254 , @xmath1503 , and @xmath1504 . first note that @xmath1505 is constant in @xmath565 , which follows easily from the homogeneity of @xmath1360 .
then , since also @xmath1506 , by lemma [ lem : w0 + ] we have @xmath1507 , implying that @xmath1508 therefore , integrating by parts , we have @xmath1509 next , we want to compute the derivative of @xmath1510 . with this objective in mind , we remark that we can compute the derivative of @xmath1511 by a formula similar to that in lemma [ lem : differentiation - formulae - smooth ] : @xmath1512 for a.e .
indeed , we first note that lemma [ lem : differentiation - formulae - smooth ] holds for smooth functions with a polynomial growth at infinity , since the same spatial integration by parts used to derive lemma [ lem : differentiation - formulae - smooth ] is still valid .
then , as in proposition [ prop : diff - form - signor ] approximate @xmath94 by the solutions @xmath176 of the penalized problem , apply lemma [ lem : differentiation - formulae - smooth ] for @xmath1513 , where @xmath1514 , and then pass to the limit as @xmath190 and @xmath391 as in proposition [ prop : diff - form - signor ] .
we thus arrive at the formula for @xmath1515 given above .
we therefore can write @xmath1516 we now estimate each term on the right hand side .
to estimate the first term we use that , in view of theorem [ thm : weiss ] , for an appropriately chosen @xmath919 , the function @xmath1517 is nondecreasing , and therefore @xmath1518 where in the last equality we have used lemma [ lem : w0 + ] .
the second term can be estimated by the cauchy - schwarz inequality , the integral estimate on @xmath108 at the beginning of section [ sec : class - free - bound ] , and lemma [ lem : humr2kappa ] : @xmath1519 for the last term , just notice that on @xmath1520 , @xmath1521 and thus , @xmath1522 combining all these estimates , we obtain , @xmath1523 which is the desired conclusion .
in this section , we prove our main results on the singular set , stated at the end of section [ sec : struct - sing - set ] as theorems [ thm : k - diff - sing - p ] and [ thm : sing - points - nonzero ] .
we start by remarking that in the following proofs it will be more convenient to work with a slightly different type of rescalings and blowups , than the ones used up to now .
namely , we will work with the following @xmath61-_homogeneous rescalings _ @xmath1524 and their limits as @xmath1525 .
the next lemma shows the viability of this approach .
[ lem : nondeg - sing - nonzero ] let @xmath479 with @xmath1356 in @xmath110 , @xmath1526 in @xmath1188 , @xmath1118 , and @xmath1527 for @xmath60 .
then , there exists @xmath1528 such that @xmath1529 assume the contrary .
then for a sequence @xmath653 one has @xmath1530 since @xmath1385 is a singular point , by proposition [ prop : char - sing - point ] , we have that , over a subsequence , @xmath1531 as described in theorem [ thm : exist - homogen - blowups ] , for some nonzero @xmath1532 .
now , for such @xmath80 we apply theorem [ thm : monneau ] to @xmath1533 . from the assumption on the growth of @xmath94
is is easy to recognize that @xmath1534 therefore , using the monotonicity of @xmath1535 ( see theorem [ thm : monneau ] ) for an appropriately chosen @xmath483 , we will have that @xmath1536 or equivalently @xmath1537 after rescaling , we obtain @xmath1538 which can be rewritten as @xmath1539 now from the arguments in the proof of lemma [ lem : hu - est ] , we have @xmath1540 , for any @xmath1541 , for sufficiently small @xmath565 .
hence , choosing @xmath1542 , we will have that @xmath1543 .
thus , passing to the limit over @xmath653 , we will arrive at @xmath1544 which is a contradiction , since @xmath1545 .
one consequence of the nondegeneracy at singular points is that the singular set has a topological type @xmath1546 ; this will be important in our application of whitney s extension theorem in the proof of theorem [ thm : sing - points - nonzero ] .
[ lem : sing - set - f - sigma - nonzero ] for any @xmath1547 with @xmath272 , the set @xmath1428 with @xmath1419 , @xmath1416 , is of topological type @xmath1546 , i.e. , it is a union of countably many closed sets . for @xmath1548 , @xmath1549 , let @xmath1550 be the set of points
@xmath1551 satisfying @xmath1552 by lemmas [ lem : humr2kappa ] and [ lem : nondeg - sing - nonzero ] , we have that @xmath1553 so to complete the proof , it is enough to show that @xmath1550 are closed .
indeed , if @xmath1554 , then it readily satisfies .
that , in turn , implies that @xmath1555 and by proposition [ prop : char - sing - point ] that @xmath1556 .
hence @xmath1557 .
this completes the proof .
we next show the existence and uniqueness of blowups with respect to @xmath61-homogeneous rescalings .
[ thm : uniq - blowup - sing - nonzero ] let @xmath479 with @xmath1417 in @xmath110 , @xmath1558 in @xmath1188 , @xmath1118 , and @xmath1418 with @xmath60 .
then there exists a unique nonzero @xmath1420 such that @xmath1559 by lemmas [ lem : humr2kappa ] and [ lem : nondeg - sing - nonzero ] , there are positive constants @xmath1560 and @xmath919 such that @xmath1561 this implies that any limit @xmath581 of the @xmath61-homogeneous rescalings @xmath1562 over any sequence @xmath535 is just a positive multiple of a limit of regular rescalings @xmath1563 , since over a subsequence @xmath1564 converges to a positive number . since the limits of rescalings @xmath533 are polynomials of class @xmath1362 , we obtain also that @xmath1565 . to show the uniqueness of @xmath581 , we apply theorem [ thm : monneau ] with @xmath1566 .
this result implies that the limit @xmath1567 exists and can be computed by @xmath1568 the latter equality is a consequence of theorem [ thm : exist - homogen - blowups ] , and of the argument at the beginning of this proof .
in particular , we obtain that @xmath1569 as @xmath1525 ( not just over @xmath535 ! ) .
thus , if @xmath1570 is a limit of @xmath1562 over another sequence @xmath1571 , we conclude that @xmath1572 this implies that @xmath1573 and completes the proof of the theorem .
[ thm : cont - dep - blowup - nonzero ] let @xmath307 with @xmath272 , @xmath1118 , and @xmath1419 , @xmath1416 .
for @xmath1423 , let @xmath1574 be as constructed in proposition [ prop : uk - def ] for the function @xmath1575 , and denote by @xmath1576 the @xmath61-homogeneous blowup of @xmath1424 at @xmath78 as in theorem [ thm : uniq - blowup - sing - nonzero ] , so that @xmath1577 then , the mapping @xmath1578 from @xmath1428 to @xmath1362 is continuous .
moreover , for any compact subset @xmath1249 of @xmath1428 there exists a modulus of continuity @xmath1579 , @xmath1580 such that @xmath1581 for any @xmath1582 .
given @xmath1583 and @xmath174 fix @xmath1584 such that @xmath1585 then , there exists @xmath1586 such that if @xmath1587 and @xmath1588 one has @xmath1589 this follows from the continuous dependence of @xmath1424 on @xmath1590 in @xmath1591 norm , which is a consequence of @xmath56 regularity of the thin obstacle @xmath18 .
then , from theorem [ thm : monneau ] , we will have that @xmath1592 for a constant @xmath1593 depending on @xmath1594 norms of @xmath1595 and @xmath1447 , which can be made uniform for @xmath1596 in a small neighborhood of @xmath78 .
letting @xmath1597 we will therefore obtain @xmath1598 this shows the first part of the theorem ( see remark [ rem : pk - norm ] ) . to show the second part
, we notice that we have @xmath1599 for @xmath1588 , @xmath998 , or equivalently @xmath1600 where @xmath1601 is the homogeneous rescaling of @xmath1595 . making a finite cover of the compact @xmath1249 with full parabolic cylinders @xmath1602 for some @xmath1603 , @xmath1604
, we see that is satisfied for all @xmath1605 , @xmath1606 and @xmath1607 .
now note that @xmath1608 solves the parabolic signorini problem in @xmath109 with zero thin obstacle and the right - hand side @xmath1609 besides , @xmath1610 also solves the parabolic signorini problem with zero thin obstacle , and zero right - hand side .
this implies @xmath1611 and therefore using the @xmath869 bounds as in the proof of theorem [ thm : exist - homogen - blowups](iii ) , we obtain that @xmath1612 for all @xmath1605 , @xmath1613 and @xmath1614 as @xmath190 .
it is now easy to see that this implies the second part of the theorem .
we are now ready to prove theorems [ thm : k - diff - sing - p ] and [ thm : sing - points - nonzero ] . simply combine theorems [ thm : uniq - blowup - sing - nonzero ] and [ thm : cont - dep - blowup - nonzero ] .
_ step 1 : parabolic whitney s extension . _ for any @xmath1423 let the polynomial @xmath1615 be as in theorem [ thm : cont - dep - blowup - nonzero ] . write it in the expanded form @xmath1616 then , the coefficients @xmath1617 are continuous on @xmath1618 .
next , for any multi - index @xmath1619 and integer @xmath1620 , let @xmath1621 then , we have the following compatibility lemma .
[ lem : whitney - compat - nonzero ] let @xmath1622 for some @xmath1548 , as in lemma [ lem : sing - set - f - sigma - nonzero ] .
then for any @xmath1623 @xmath1624 with @xmath1625 where @xmath1626 is a certain modulus of continuity .
\1 ) in the case @xmath1627 we have @xmath1628 and the statement follows from the continuity of @xmath1617 on @xmath1428 .
\2 ) for @xmath1629 we have @xmath1630 suppose now that there exists no modulus of continuity @xmath1631 such that is satisfied for all @xmath1632
. then , there exists @xmath1194 and a sequence @xmath1633 , with @xmath1634 such that @xmath1635 consider the rescalings @xmath1636 without loss of generality we may assume that @xmath1637 and @xmath1638 .
then , by theorem [ thm : cont - dep - blowup - nonzero ] , for any @xmath149 and large @xmath1639 we have for a modulus of continuity @xmath1579 @xmath1640 and therefore @xmath1641 note that we do not necessarily have the above convergence in the full parabolic cylinder @xmath1642 . next , consider the rescalings at @xmath1643 instead of @xmath1644 @xmath1645 then , by the same argument as above @xmath1646 we then claim that the @xmath56 regularity of the thin obstacle @xmath18 implies that @xmath1647 for any @xmath149 , or equivalently @xmath1648 indeed , if @xmath1649 ( as usual ) denotes the @xmath276-th parabolic taylor polynomial of the thin obstacle @xmath1650 at @xmath78 , then are the same ] @xmath1651 and this implies the convergence , if we write the explicit definition of @xmath1652 using the construction in proposition [ prop : uk - def ] . to proceed further , we consider two different cases : \1 ) there are infinitely many indexes @xmath1639 such that @xmath1653 .
\2 ) there are infinitely many indexes @xmath1639 such that @xmath1654 . in both cases
, passing to subsequences we may assume that @xmath1655 @xmath1656 for all indexes @xmath1639 . in case 1 ) we proceed as follows .
if we take any @xmath1657 , because of the nonpositivity of @xmath1658 we have @xmath1659 . passing to the limit in , , and ,
we thus obtain @xmath1660 because of the real analyticity of polynomials , it follows that holds in fact for all @xmath1661 .
but then , we also obtain @xmath1662 in particular , this implies that @xmath1663 on the other hand , dividing both sides of by @xmath1664 and passing to the limit , we obtain @xmath1665 a contradiction . in case 2 ) , when there are infinitely many indexes @xmath1639 so that @xmath1654 , passing to the limit in , , and , we obtain for @xmath1657 .
again by real analyticity , we have the same conclusion for all @xmath1661 .
then , we complete the proof arguing as in case 1 ) .
so in all cases , the compatibility conditions are satisfied and we can apply the parabolic whitney s extension theorem that we have proved in appendix [ sec : parab - whitn - extens ] , see theorem [ thm : parab - whitney ] . thus , there exists a function @xmath1666 such that @xmath1667 for any @xmath1668 . _
step 2 : implicit function theorem .
_ suppose now @xmath1669 .
we will consider two subcases : @xmath1670 and @xmath1453 .
\1 ) @xmath1671 . then , there are multi - indexes @xmath1672 with @xmath1673 , for @xmath1674 such that @xmath1675 are linearly independent .
on the other hand , @xmath1676 therefore , in view of the implicit function theorem , the linear independence of @xmath1677 implies that @xmath1678 is contained in a @xmath88-dimensional space - like @xmath89 manifold in a neighborhood of @xmath78 . finally , since @xmath1679
this implies the statement of the theorem in the case @xmath1680 .
\2 ) suppose now @xmath1453 . in this case , by lemma [ lem : timelike - blowup
] , we have @xmath1681 on the other hand , @xmath1682 thus , by the implicit function theorem we obtain that @xmath1678 in a neighborhood of @xmath78 is contained in a time - like @xmath1470-dimensional @xmath87 manifold , as required .
this completes the proof of the theorem .
in this section we give the proofs of the estimates stated in section [ sec : estimates - w2 - 1_2s_1 + ] . [
step:1-w21-gauss](@xmath362 ) for the given @xmath1683 , choose @xmath1684 . then , note that without loss of generality we may assume that @xmath406 , by multiplying @xmath94 with a smooth cutoff function @xmath400 such that @xmath401 on @xmath1685 $ ] and @xmath403 near @xmath794 .
next , fix a cutoff function @xmath1686 and let @xmath1687 be so large that @xmath1688 vanishes outside @xmath1689 and @xmath94 vanishes outside @xmath1690 $ ] . then , approximate @xmath94 with the solutions of the penalized problem @xmath408,\\ \partial_{x_n}u^{\varepsilon}=\beta_{\varepsilon}(u^{\varepsilon})&\quad\text{on } b_r'\times(-1,0],\\ u^{\varepsilon}=0&\quad\text{on } ( \partial b_r)^+\times(-1,0],\\ u^{\varepsilon}(\cdot , -1)=0&\quad\text{on } b_r^+ , \end{aligned}\ ] ] where @xmath175 is a mollification of @xmath108 .
let now @xmath1691 $ ] be arbitrary .
then , for any small @xmath424 and @xmath1692)$ ] , vanishing on @xmath1693 $ ] , we will have @xmath1694 dxdt=-\int_{s_r'\setminus s_\delta'}\beta_{\varepsilon}(u^{\varepsilon})\eta dx'dt.\ ] ] now , for the cutoff function @xmath1688 as above , define the family of homogeneous functions in @xmath328 by letting @xmath1695 then , choosing @xmath1696 , we will have @xmath1697 where we have used that @xmath1698 .
next , recalling that @xmath1699 and that @xmath1700 , we can rewrite the above inequality as @xmath1701 we then can use the standard arguments in the proof of energy inequalities , except that we need to handle the term involving @xmath1702 . making the change of variables
@xmath455 , @xmath1703 and using the identities @xmath1704 we obtain @xmath1705\hat\zeta_0 ^ 2(y ) g(y,-1)dyd\lambda\\ & \qquad= -\int_{{\mathbb{r}}^n_+ } [ r^2u^{\varepsilon}(ry
,- r^2)^2-\delta^2 u^{\varepsilon}(\delta y,-\delta^2)^2]\hat\zeta_0 ^ 2(y ) g(y,-1)dy\\ & \qquad\qquad+\int_\delta^r\int_{{\mathbb{r}}^n_+ } 2\lambda u^{\varepsilon}(\lambda y,-\lambda^2)^2\hat\zeta_0 ^ 2 g(y,-1)dyd\lambda\\ & \qquad \geq -r^2\int_{{\mathbb{r}}^n_+ } u^{\varepsilon}(ry ,- r^2)^2\hat\zeta_0 ^ 2(y ) g(y,-1)dy\\ & \qquad =- r^2\int_{{\mathbb{r}}^n_+ } u^{\varepsilon}(\cdot ,-
r^2)\zeta_0 ^ 2g(\cdot ,- r^2)dy,\end{aligned}\ ] ] where we have used integration by parts is @xmath372 variable .
thus , using young s inequality , we conclude that @xmath1706 g\big).\end{gathered}\ ] ] now , integrating over @xmath1707 $ ] , we obtain @xmath1708 g.\ ] ] letting first @xmath1709 , then letting the support of the cutoff function @xmath1688 sweep @xmath1 , @xmath391 , and replacing the integral over @xmath1710 by @xmath110 , we conclude that @xmath1711 [ step:2-w21-gauss](@xmath362 ) let @xmath1688 , @xmath1687 , homogeneous functions @xmath1712 , the approximations @xmath176 , @xmath1707 $ ] , and @xmath463 be as above .
then , choosing as a test function @xmath1713 , @xmath1714 , with @xmath1715 vanishing in @xmath1716 $ ] and integrating by parts , we obtain @xmath1717dxdt=-\int_{s_r'\setminus s_\delta'}\beta_{\varepsilon}'(u^{\varepsilon})u^{\varepsilon}_{x_i}\eta dx'dt.\ ] ] then , plugging @xmath1718 in the above identity , we will have @xmath1719 then , again using the identity @xmath1699 , we may rewrite the above inequality as @xmath1720 arguing as in step ( [ step:1-w21-gauss]@xmath362 ) above , we have @xmath1721\hat\zeta_0(y)^2 g(y,-1 ) \\ & \qquad+4\int_\delta^r\int_{{\mathbb{r}}^n_+}\lambda^3(u_{x_i}^{\varepsilon})^2\hat\zeta_0 ^ 2 g\\&\geq -r^4\int_{{\mathbb{r}}^n_+ } u_{x_i}^{\varepsilon}(ry ,- r^2)^2\hat\zeta_0(y)^2 g(y,-1)dy \\&= -r^2\int_{{\mathbb{r}}^n_+ } u_{x_i}^{\varepsilon}(\cdot ,-
r^2)^2\zeta_1 ^ 2 g(\cdot ,- r^2).\end{aligned}\ ] ] we further estimate , by the appropriate young inequalities , @xmath1722g,\end{aligned}\ ] ] with a small constant @xmath1723 , where in the last step we have used that the following claim .
[ clm : v2x2 t ] for any @xmath1724 and @xmath1725 we have @xmath1726 using that @xmath1699 , and then integrating by parts , we obtain @xmath1727 which implies the desired estimate . combining the estimates above
, we obtain @xmath1728 g\big)\\ & \qquad\qquad+c_{n,\rho}\int_{s_r^+\setminus s_\delta^+ } |d^2 u^{\varepsilon}|^2\zeta_2 ^ 2 g.\end{aligned}\ ] ] [ step:3-w21-gauss](@xmath362 ) using the notations of the previous step , taking a test function @xmath1729 and integrating by parts , we will obtain @xmath1730 dxdt=-\int_{s_r'\setminus s_\delta ' } [ u_t^{\varepsilon}\eta+\nabla'u^{\varepsilon}\nabla'\eta]dx'dt.\ ] ] plugging @xmath1731 , we will have @xmath1732+|\nabla ' u^{\varepsilon}_{x_i}|^2\beta'_{\varepsilon}(u^{\varepsilon})\zeta_2 ^ 2 g\\ & \qquad\qquad-\int_{s_r'\setminus s_\delta'}2u^{\varepsilon}_{x_n}\nabla'u^{\varepsilon}\nabla'\zeta_2\zeta_2 g.\end{aligned}\ ] ] we therefore have @xmath1733 to estimate @xmath426 we argue as before , however we now take into account that the spatial dimension is less by one : @xmath1734\hat\zeta_0(y')^2 g(y',-1)dy'\\ & \qquad -\frac32\int_\delta^r\int_{{\mathbb{r}}^{n-1}}\lambda^2\mathcal{b}_{\varepsilon}(u^{\varepsilon}(\lambda y',-\lambda^2))\hat\zeta_0(y')^2g(y'-1)dy'd\lambda\\ & \leq c_{n,\rho}{\varepsilon},\end{aligned}\ ] ] since @xmath1735 for any @xmath1736 and we have used that @xmath1737 on @xmath1738 $ ] . for more details
see the proof of proposition [ prop : diff - form - signor ] , step ( [ step:3.3.2-diff - form]@xmath362 ) . in order to estimate @xmath432 we write it as a solid integral @xmath1739 the terms in @xmath1740 are estimated in a standard way by the appropriate young inequalities .
the integral @xmath1741 is estimated by using claim [ clm : v2x2 t ] : @xmath1742,\end{aligned}\ ] ] for a small constant @xmath1723 .
we further treat the term @xmath1743 analogously to the similar term with @xmath1744 in ( [ step:2-w21-gauss]@xmath362 ) . collecting all estimates in this step ,
combined with appropriate young inequalities , we obtain @xmath1745 g\\ & \qquad+c_{n,\rho}\int_{s_r^+\setminus s_\delta^+ } |d^2 u^{\varepsilon}|^2\zeta_2 ^ 2 g + c_{n,\rho}{\varepsilon}.\end{aligned}\ ] ] [ step:4-w21-gauss](@xmath362 ) now combining the estimates in ( [ step:2-w21-gauss]@xmath362 ) and ( [ step:3-w21-gauss]@xmath362 ) above and integrating over @xmath1707 $ ] we obtain @xmath1746 g\\ & \qquad+c_{n,\rho}{\varepsilon}.\end{aligned}\ ] ] as before , passing to the limit as @xmath1747 , increasing the support of @xmath1688 , and then letting @xmath391 , we conclude that @xmath1748 finally noticing that @xmath1749 , we obtain the desired integral estimate for @xmath1750 as well .
the proof is complete .
the proof is very similar to part ( [ step:1-w21-gauss]@xmath362 ) of the proof of lemma [ lem : w212 ] . indeed , for approximations @xmath1751 , @xmath331
, we have the integral identities @xmath1752 dxdt=-\int_{s_r'\setminus s_\delta'}\beta_{\varepsilon}(u_i^{\varepsilon})\eta dx'dt.\ ] ] taking the difference , choosing @xmath1753 , and using the inequality @xmath1754(u_1^{\varepsilon}-u_2^{\varepsilon})\geq 0,\ ] ] we complete the proof as in step ( [ step:1-w21-gauss]@xmath362 ) of the proof of lemma [ lem : w212 ] .
let @xmath154 be a compact subset of @xmath116 and @xmath1755 a certain continuous function . here
we want to establish a theorem of whitney type ( see @xcite ) that will allow the extension of the function @xmath108 to a function of class @xmath1756 , @xmath77 .
in fact , for that we need to have a family of functions @xmath1757 , where @xmath1619 and @xmath1201 is a nonnegative integer .
[ thm : parab - whitney ] let @xmath1757 be a family of functions on @xmath154 , with @xmath1758 , satisfying the following compatibility conditions : there exists a family of moduli of continuity @xmath1759 , such that @xmath1760 and @xmath1761 then , there exists a function @xmath1762 such that @xmath1763 on @xmath154 and moreover @xmath1764 on @xmath154 , for @xmath129 . _
step 1 : _ parabolic whitney cube decomposition of @xmath1765 .
we say that @xmath1766 is a parabolic ( @xmath276-)dyadic cube if it has a form @xmath1767\times\cdots\times [ a_n 2^{-k } , ( a_n+1)2^{-k}]\times[b2^{-2k},(b+1)2^{-2k}],\ ] ] where @xmath1768 .
we will call @xmath1769 the size of @xmath1766 and denote it by @xmath1770 .
we will also call @xmath1771 $ ] the center of @xmath1766 .
the proof of the following lemma is very similar to its euclidean counterpart and is therefore omitted .
( a slightly different version of this lemma can be found in @xcite*lemma 1.67 , for more general homogeneous spaces . ) for every @xmath1778 let @xmath1779 be the center and @xmath1780 the size of the parabolic cube @xmath1778 .
then let @xmath1781+(x_i , t_i)$ ] , where @xmath1782 is the parabolic dilation .
clearly , the family of @xmath1783 is no longer disjoint , however , we every point in @xmath1784 has a small neighborhood that intersects at most @xmath1785 cubes @xmath1786 , provided @xmath1787 is small . then define @xmath1788 where @xmath1789 such that @xmath1790 where @xmath1791\times\cdots\times[-\lambda/2,/2]\times[-\lambda^2/2,\lambda^2/2].\ ] ] we also observe that @xmath1792 for some constants @xmath1793
next , we define a partition of unity @xmath1794 subordinate to @xmath1783 as follows .
let @xmath1795 note that in the definition of @xmath1796 , the sum is locally finite and therefore @xmath1797 in @xmath1784 , and also satisfies @xmath1798 .
then @xmath1799 are also @xmath1797 in @xmath1784 and we have @xmath1800 where again the sum is locally finite .
moreover , it is easy to see that , similarly to @xmath1801 , we have the estimates @xmath1802 _ step 2 : _ for every @xmath1803 let @xmath1804 in addition to @xmath1403 , it is convenient to introduce @xmath1805 for @xmath129 . note that in fact @xmath1806 .
then by definition @xmath1807 for any @xmath1808 . we will prove the latter formula .
it is enough to check that the partial derivatives @xmath1812 of both sides equal to each other for @xmath1813 , as both sides are polynomials of parabolic degree @xmath1814 .
we have @xmath1815 which implies the desired equality .
_ step 3 : _ we are now ready to define the extension function @xmath1816 . for
every @xmath1778 let @xmath1817 be such that @xmath1818 .
note that @xmath1819 is not necessarily unique . then define @xmath1820 from the local finiteness of the partition of unity , it is clear that @xmath1816 is @xmath1797 in @xmath1784
. then we can define @xmath1821 for @xmath129 .
note that for @xmath1825 , the estimates follows from the compatibility assumptions . for @xmath1826
, we have @xmath1827{\varphi}_i^*(x , t)\big|\\ & \qquad\leq\sum_{i } \sum_{|\alpha|+2j\leq 2m}|r_{\alpha , j}(y_i , s_i;x_0,t_0 ) |\frac{|x - y_i|^\alpha}{\alpha!}\frac{|t - s_i|^j}{j!}{\varphi}_i^*(x , t)\\ & \qquad\leq \sum_{i , |\alpha|+2j\leq 2m}\omega_{\alpha , j}(c_n\|(x - y_i , t_0-s_i)\|)\|(x - y_i , t - s_i)\|^{2m-|\alpha|-2j+|\alpha|+2j}{\varphi}_i^*(x , t)\\ & \qquad\leq \tilde\omega(\|(x - x_0,t - t_0)\|)\|(x - x_0,t - t_0)\|^{2 m } , \end{aligned}\ ] ] using that @xmath1828 for @xmath1829 . the second estimate in the lemma is obtained in a similar way .
indeed , we can write @xmath1830|\\ & \qquad=|\partial_x^\alpha\partial_t^j\sum_i [ p(x , t;y_i , s_i)-p(x , t;x_0,t_0)]{\varphi}_i^*(x , t)|\\ & \qquad=\sum_{i,\beta\leq \alpha , k\leq j } c^{\alpha , j}_{\beta , k}[p_{\beta , k}(x , t;y_i , s_i)-p_{\beta , k}(x , t;x_0,t_0)]\partial_x^{\alpha-\beta}\partial_t^{j - k}{\varphi}_i^*(x , t)\end{aligned}\ ] ] and then we argue as above by using lemma [ lem : we - diff ] and the estimates . note
that lemma [ lem : we - f - p ] implies that @xmath1831 at every @xmath1825 .
the same equalities hold also in @xmath1784 , by the definition of @xmath1832 . thus , arguing by induction in the order of the derivative @xmath129 and using lemma [ lem : we - f - p ] , we prove that everywhere in @xmath116 @xmath1833 we also note that @xmath1832 are continuous by lemma [ lem : we - f - p ] .
the proof is complete .
|
we give a comprehensive treatment of the parabolic signorini problem based on a generalization of almgren s monotonicity of the frequency .
this includes the proof of the optimal regularity of solutions , classification of free boundary points , the regularity of the regular set and the structure of the singular set .
|
modern optical magnetometers@xcite enable the detection of magnetic fields with a sensitivity reaching or even exceeding 1 ft / hz@xmath1 .
yet , the magnetometers are not sensitive to magnetic field _ per se _ but rather to its effect on energies of light - coupled magnetic sublevels . with this respect ,
optical magnetometers are intrinsically sensitive to any spin couplings , including exotic ones .
this underlies the foundations for searches for physics beyond the standard model and formed the basis for many experiments over the last several decades ( see , for example , ref . and reference therein ) .
despite the high sensitivities of the experiments searching for physics beyond the standard model , all the experiments have provided null results.@xcite this has falsified several theoretical models , leading to their redefinition . however , to reach the precision , long - time averaging of the detected signals is required .
this brings an important question : _ what if the sought couplings ( signals ) are of time dependent ( oscillatory or transient ) character ?
_ obviously , in such a case , the averaging would lead to deterioration of the signal of interest rather than increase of a signal - to - noise ratio . over the last years
, several theoretical models postulating the existence of time - dependent exotic spin couplings have been proposed .
such couplings would either lead to oscillating signals , e.g. , due to oscillating electric dipole moment@xcite or dark - matter couplings,@xcite or to transient signals , e.g. , due to a passage through a topological defect ( e.g. , domain wall ) of an exotic field@xcite or interaction with a jet of exotic particles.@xcite detection of oscillating couplings requires new experiments , which are currently under construction in mainz , germany and boston , usa ( cosmic axion spin precession experiment)@xcite and stanford , usa ( dark - matter radio).@xcite similarly , the transient signals will be investigated with a global network of optical magnetometers for exotic physics ( gnome),@xcite described here .
gnome is a network of synchronized optical magnetometers separated by a distance of hundreds or even thousands of kilometers .
each magnetometer is characterized with a femtotesla sensitivity and operates in a magnetically shielded , precisely controlled magnetic environment .
in such a measurement , the magnetometer readouts are principally static , but the nature of the measurements burdens the signals with noise .
the spectral characteristics of the noise may be quite complex , with frequency - independent ( e.g. , due to photon shot noise ) , @xmath2 ( e.g. , due to vibrations ) , and band ( e.g. , due to ac lines ) components .
additionally , infrequent spikes ( in the time domain ) can be observed in the signals .
the nature of this noise is particularly hard to identify and eliminate , making untriggered transient signals difficult to investigate . despite the complexity of the magnetometer readouts ,
the observed noise is typically local and hence the readouts of even same - type devices are characterized with different dependence .
particularly , it is unlikely to observe correlated transient signals in two or more magnetometers triggered by uncorrelated ( local ) events .
in such a case , correlation measurements provide information about global transient disturbances .
the source of these disturbances may be magnetic ( due to the earth - magnetic - field change and imperfections of magnetic shieldings ) but such cases may be falsified with proper vetoing techniques ( e.g. , via correlating the signals with readouts of outside - of - the - shield sensors ) , leaving the system sensitive to global disturbances of nonmagnetic origins . from the point of view of the network operation , precise timing of the signal acquisition
is of crucial importance . in gnome ,
the timing is provided by the global positioning system ( gps ) , which offers a time precision better than 100 ns . at the same time
, the global spread of the magnetometers loosens requirements for sampling rate as even light - speed - propagating perturbations require several to several tens of milliseconds to transit between different stations . to meet these requirements ,
the gnome collaboration developed a data - acquisition system , offering the ability to multi - channel data acquisition with up to 1 ks / s sampling rate and sub - microsecond time accuracy .
the systems are now installed in all gnome stations to acquire synchronously the magnetometer readouts . from the point of view of foreseen measurements , correlation of signals with the same spectral characteristics
is essential . to meet this requirement
, the magnetometer signals need to be standardized .
this is achieved by the so - called whitening procedure , which consists of normalizing the signals by an exponentially weighted signal history.@xcite this procedure produces a gaussian distributed , zero - mean and unity variance data stream , which can be reliably correlated ( each magnetometer contributes at the same level to the correlation signals ) .
the current gnome consists of 5 stations situated in 4 countries ( germany , poland , switzerland , and the us ) .
several other stations are under construction . in the existing arrangement ,
the network performed a 24-hour trial run in march 2016 .
the signals from all magnetometers were stored and transmitted to the main server ( mainz , germany ) , where they were processed .
the spectrogram of a sample time series detected with the krakw magnetometer is shown in fig .
[ fig : data](a ) ( top panel ) .
similarly to signals of all other magnetometers ( not shown ) , the data reveals strong @xmath2-noise but also two strong modulation components at 1 hz ( slowly drifting @xmath3 hz ) and 12 hz ( stable over time ) . as shown in bottom panel of fig .
[ fig : data](a ) , @xmath2-noise and the modulation components are strongly suppressed with the whitening procedure ; the whole noise amplitude span was reduced by about 5 orders of magnitude leaving a detectable , yet strongly reduced trace only at 1 hz .
whitening procedure enables the identification of transient events , manifesting as local maxima in the time - frequency spectrogram .
however , from the point of view of gnome operation , correlation measurements are of crucial importance .
thus , the correlations of two hour - long time series of krakw and berkeley magnetometers [ fig .
[ fig : data](b ) ] are presented . to perform reliable analysis
, we first estimated the background by correlating time series separated by several tens of minutes , where there are no correlations .
this enables identification of a false positive - event rate ( solid line ) .
then , we correlated the signals with no time offset , i.e. , a specific case for which the correlations could be observed ( gray points ) .
for such a case , several tens of events were identified , all consistent with the false - positive background .
this result presents the ability to identify correlation between the stations but also demonstrates the necessity of improving performance of the magnetometers and developing more sophisticated ( multi - station ) correlation algorithms . ]
the results of the trail run demonstrated the ability to identify transient events in signals of two magnetometers . currently , the network magnetometers are being optimized to reach sensitivity of roughly 10 ft / hz@xmath1 and bandwidth of @xmath4 hz and 24/7 operation .
the start of a multi - month run is planned at the end of 2016 .
this will enable a search for axion - like - field domain structure , exotic field bursts , and exotic coherent oscillations .
however , the universality of the network makes it a unique tool for testing theories beyond the standard model .
the author acknowledges support from the polish ministry of science and higher education within the iuventus plus program ( grant 0390/ip3/2015/73 ) .
d. f. jackson kimball and s. k. lamoreux and t. e. chupp _ tests of fundamental physics with optical magnetometers _ in _ optical magnetometry _ , d. budker and d. f. jackson kimball eds . cambridge university press , cambridge , 2013 .
|
in this letter , we describe a novel scheme for searching for physics beyond the standard model .
the idea is based on correlation of time - synchronized readouts of distant ( @xmath0100 km ) optical magnetometers .
such an approach limits hard - to - identify local transient noise , providing the system with unique capabilities of identification of global transient events .
careful analysis of the signal can reveal the nature of the events ( e.g. , its nonmagnetic origin ) , which opens avenues for new class of exotic - physics searches ( searches for global transient exotic spin couplings ) and tests of yet unverified theoretical models .
|
there are metals in the intergalactic medium ( igm ) even in low density regions such as lyman @xmath8 forest ( e.g. , @xcite ) .
since metal and dust are relating each other , it is sure that dust grains also exist in the igm .
however , the amount of the intergalactic ( ig ) dust is still quite uncertain although it seems to be not so abundant .
we try to put a new constraint on the amount of the ig dust by using the igm thermal history suggested by the recent observations of lyman @xmath8 forest ( e.g. , @xcite ) .
after the early attempts to estimate the ig extinction @xcite , @xcite have obtained an upper limit on the amount of the ig dust in terms of the density parameter as @xmath9 , by comparing the observed spectral energy distribution of distant ( but @xmath10 ) giant elliptical galaxies with an average spectrum of similar nearby galaxies ( see also @xcite ) .
observations of the redshift evolution of the qsos spectral slope provide us with another approach to the ig dust @xcite .
although cheng et al . ( 1991 ) found no evidence for an appreciable ig reddening up to @xmath113 , this may not be so strict constraint because it is difficult to measure the uv slope of qsos enough precisely owing to the broad and complex iron band superposing on the continua .
recently , we obtain further constraint on the ig dust from observations of high - z supernovae ( sne ) @xcite
. an upper limit of the colour excess by the ig reddening is @xmath12 mag . by the way
, @xcite suggests that the ig dust can be `` gray '' by a selection rule in the transfer of grains from the host galaxies to the igm .
this occurs when small ( @xmath13 ) grains are destroyed selectively by the thermal sputtering in the hot gas halo of the host galaxies before they reach the igm .
then , the extinction property becomes `` gray '' since the relatively large dust survives .
importantly , such `` gray '' ig dust can avoid the detection by the ig reddening survey like above . even if the ig dust is really `` gray '' , its evidence should be imprinted in the cosmic microwave background ( cmb ) and infrared background because the dust emits thermal radiation in the wave - band from the far - infrared to submillimetre ( submm ) @xcite . according to the current status , _ cobe _ data
provides us with a rough upper limit on the ig dust @xcite . on the other hand
, the submm background radiation may give a more strict constraint on the ig dust .
the emission from the ig dust expected from the modern cosmic star formation history with the grain transfer mechanism of @xcite contributes to a substantial fraction ( @xmath14 75% ) of the measured background radiation in the submm @xcite .
it is interesting to develop the approach of @xcite to match a strong constraint by scuba whose result of the number count accounts @xmath15 90% of the submm background light ( @xcite and references therein ) . in any current status ,
the amount of the ig dust is still highly uncertain .
hence , it is worth trying to find a new constraint on the amount of the ig dust . in this letter
, it is demonstrated that we can obtain such a new constraint of the ig dust by comparing the observational igm temperature with a theoretical one if we take into account the dust photoelectric heating in the theoretical model .
this is possible because the dust heating is an efficient mechanism in the igm @xcite .
interestingly , the effect of the dust photoelectric heating depends on not only the amount but also the size of dust .
therefore , we can put a limit on the amount of the ig dust as a function of the typical size of the ig grains .
to assess photoelectric effect , we must specify the charge on grain , @xmath16 ( in the electron charge unit ) , which is given by @xcite : @xmath17 where we have assumed a charge equilibrium because a typical charging time - scale is very short ( @xmath18^{-1}$ ] yr under the uv background radiation dominated by qso light given in 3 ) . in the equation , @xmath19 is the collisional charging rate by @xmath20-th charged particle , and @xmath21 is the photoelectric charging rate .
we assume spherical grains for simplicity .
the collisional charging rate is @xmath22 where @xmath23 is the charge in the electron charge unit , @xmath24 is the sticking coefficient , @xmath25 is the number density , @xmath26 is the grain radius , @xmath27 is the velocity , @xmath28 is the minimum velocity required to collide with a grain , @xmath29 is the dimensionless collisional cross section depending on both charges and @xmath27 , and @xmath30 is the velocity distribution function .
we simply assume @xmath24 is always unity . in the above equation
, we neglected the effect of the secondary emission .
we can neglect the effect of the `` image potential '' @xcite on the collisional cross section because the obtained charges are enough large . if we assume the maxwellian velocity distribution for the particle , the integral in the r.h.s . of equation ( [ eq2 ] )
is reduced to @xmath31 and @xmath32 for @xmath33 or @xmath34 for @xmath35 , where @xmath36 is the boltzmann s constant , @xmath37 is the gas temperature , @xmath38 is the charged particle s mass , and @xmath39 . the photoelectric charging rate is given by @xmath40 where @xmath41 is the absorption coefficient of grains , @xmath42 is the photoelectric yield , @xmath43 is the intensity ( averaged for solid angle ) of the incident radiation at a frequency @xmath44 , @xmath45 is the plank constant , and @xmath46 is the maximum frequency of the incident radiation .
the minimum frequency , @xmath47 , is the threshold photon frequency required for an electron to escape from a grain .
that is , @xmath48 is equal to the ionization potential , @xmath49 , of a grain : @xmath50 , where @xmath51 is the work function . here
we neglect quantum effects on @xmath49 because it is very small for grains with @xmath52 ( @xcite , hereafter wd01 ) . according to wd01
, we adopt @xmath53 ev for graphites and @xmath54 ev for silicates . for @xmath41
, we adopt the values of `` graphite '' and `` smoothed uv astronomical silicate '' calculated by @xcite .
the value of @xmath42 is estimated , based on the way constructed by wd01 , which includes approximately the effect of the energy distribution of photoelectrons and the geometrical enhancement for small grains , and reproduces recent results of the laboratory experiments even for very small grains .
the both equilibrium charges obtained for graphites and silicates are similar when their radii are equal .
this is consistent with the result of nath et al .
( 1999 ) . in the following discussion
, we consider mainly silicate case .
the photoelectric heating rate by a grain with radius @xmath26 is expressed by @xmath55 and the cooling rate of the electron capture ( recombination ) by the grain is @xmath56 , where we have assumed the maxwellian distribution for the gaseous electrons .
we also have defined @xmath57 , the mean kinetic energy of photoelectrons that are emitted from a grain with radius @xmath26 and charge @xmath16 when a photon with energy @xmath58 is absorbed .
this is determined from the energy distribution function of the photoelectrons , @xmath59 , which is assumed to be a parabolic function introduced by wd01 , is equal to @xmath60 in wd01 . ] and is consistent with a typical energy distribution of photoelectrons in the laboratory experiments .
hence , we estimate @xmath61 where @xmath62 is the maximum energy of the photoelectrons and @xmath63 is the minimum energy of the photoelectrons appearing on the grain surface ( however falling back into the grain ) . for the parabolic function adopted here , a typical kinetic energy of the photoelectrons is about half of @xmath64 .
to estimate the total photoelectric heating rate , we specify a characteristic grain size in the igm instead of the size distribution of the ig grains .
this is because the size distribution in the igm is quite unknown .
here we consider three cases for the typical grain size ; 0.001 , 0.01 , and 0.1 .
the smallest case of 0.001 is based on the size of the primary grains produced in the ejecta of sne ii @xcite .
it is interesting to obtain an information about the size of the ig dust as well as to put limits on the amount . from the analysis of the case of very small ig dust
, we can assess the selection rule of the ig dust suggested by @xcite .
the grain number density is given as @xmath65 , where @xmath66 is the mass of a grain with the radius @xmath26 , @xmath67 @xmath68 is the grain material density , and @xmath69 is the volume grain mass density , which is given by @xmath70 , where @xmath71 is the mean atomic weight , @xmath72 is the proton mass , @xmath73 is the baryon number density , and @xmath0 is the dust - to - gas mass ratio . we introduce a relative dust - to - gas ratio , @xmath74 , defined as @xmath75 , where @xmath76 is the galactic dust - to - gas ratio @xcite , for convenience below .
therefore , the total photoelectric heating rate and electron capture cooling rate per unit volume are @xmath77 and @xmath78 , respectively .
the total heating rate is approximately proportional to @xmath79 under the situation considered in the next section .
this is because the heating rate per a grain roughly has a dependence of @xmath80 , i.e. , the grain cross section plus an additional @xmath26 dependence caused by @xmath81 , whereas the grain number density is proportional to @xmath82 .
therefore , for a fixed dust mass , the photoelectric effect becomes more important as the grain size is smaller .
the cooling rate by the electron capture is always about an order of magnitude smaller than the photoelectric heating rate .
suppose an ideal fluid element with the mean density of the igm .
the time evolution of its temperature is governed by ( e.g. , @xcite ) @xmath83 where @xmath84 is the hubble constant , @xmath73 is the cosmic mean number density of baryon , @xmath85 and @xmath86 are the total heating and cooling rates per unit volume , respectively , @xmath87 is the number ratio of the total gaseous particles to the baryon particles , i.e. , @xmath88 , where @xmath89 is the number density of the @xmath20-th gaseous species and we consider h i , h ii , he i , he ii , he iii , and electron .
then , we solve equation ( [ eqt ] ) coupled with the non - equilibrium rate equations for these gaseous particles . for the rate coefficients , we use those compiled by @xcite .
the adopted initial conditions are @xmath90 , h i / h(total)=0.99 , he i / he(total)=0.99 , he ii / he(total)=0.0099 at redshift @xmath91 , where @xmath92 k is the temperature of the cmb radiation at @xmath93 .
the result is insensitive to the choice of the initial condition if calculation is started at an epoch before h i reionization .
we also adopt @xmath94 km s@xmath95 mpc@xmath95 , @xmath96 , @xmath97 , @xmath98 , and he mass abundance @xmath99 .
to add to the photoelectric heating and the electron capture cooling by grains , we consider the following heating / cooling mechanisms : for cooling , we use the recombination cooling , collisional ionization / excitation cooling , bremsstrahlung cooling , and compton cooling compiled by @xcite . for heating rates , to take account of the radiative transfer effect @xcite , we multiply usual photoionization heating rates by a correction factor of @xmath100 , where @xmath101 is the fractional abundance of h i , he i , and he ii relative to total number of h and he .
this factor mimics the situation that atoms are ionized by higher energy photons when the optical depth is large ( i.e. , @xmath101 is nearly unity ) relative to the optically thin limit .
the coefficient @xmath102 is a parameter determined by solving the cosmological radiative transfer .
the metal line cooling is negligible because we consider only the case that metallicity is less than 1/100 solar value and temperature is less than a few @xmath103 k ( see e.g. , @xcite ) .
we need to give the uv background radiation and the cosmic reionization history to obtain the temperature evolution .
we adopt , for uv background , @xmath104 , where @xmath105 is the mean intensity at the lyman limit of h i in unit of @xmath106 erg s@xmath95 @xmath107 sr@xmath95 hz@xmath95 and its evolution is given by equation ( 3 ) in @xcite .
we assume the spectral index to be @xmath108 as a qso dominated background radiation .
for the reionization history , we assume , for simplicity , a sudden reionization of h i ( and he i ) at @xmath109 ( @xcite ) h i reionization , our constraint of the ig dust dose not change because it is determined mainly from the igm temperature after the he ii reionization . ] ; there is no uv photon before this redshift . to mimic the he ii reionization at @xmath110 @xcite
, we set the maximum photon energy of the uv background to be 54.4 ev ( he ii lyman limit ) for @xmath111 and to be 1.24 kev for @xmath112 ( i.e. , a sudden he ii reionization ) .
other choice of @xmath8 will be discussed later .
we compare our theoretical thermal history at the mean density of the igm with the observational one obtained by @xcite .
they observe the ly@xmath8 forest clouds with the column density of @xmath113 @xmath107 ( i.e. , slightly over density regions ) , and then they convert the temperature at this density range estimated from the minimum @xmath114 ( doppler ) parameters into that at the mean density of the igm by using the observed equation of state of the igm .
thus , we can compare both thermal histories directly . in figure 1
, we show the igm thermal history obtained for no dust cases and dusty cases of @xmath115 ( silicate grains ) . for no dust cases ( three dotted curves ) ,
the calculated thermal histories are consistent with the data points of @xcite if @xmath1167.0 . for dusty cases ( solid [ @xmath117 0.1 ] ,
dash - dotted [ @xmath117 0.01 ] , and dashed curves [ @xmath117 0.001 ] ) , we observe , especially after the he ii reionization , temperature enhancement by the dust photoelectric heating . in table 1 , we summarize the temperature enhancement factors at some redshifts for some sets of dust size and dust - to - gas ratio .
all calculations do not include the evolution of the dust - to - gas ratio , for simplicity . and
he ii at @xmath118 .
data points with errorbars are taken from schaye et al .
three dotted curves are no dust cases ; bottom : optically thin case , middle : @xmath119 case , top : @xmath120 case , where @xmath102 is the correction factor of the radiative transfer effect . the dashed , dash - dotted , and solid curves are cases of grain radius @xmath121 , 0.01 , and 0.1 , respectively . for these dusty cases ,
the dust - to - gas mass ratio is set to be 1/100 galactic value , @xmath119 is assumed , and the grain type is silicate .
the middle dotted curve in @xmath111 is almost superposed on the solid curve.,title="fig:",height=207 ] , dust - to - gas mass ratio of 1/100 galactic value , and radiative transfer correction @xmath119 .
solid curve is the photoelectric heating rate by grains .
dash , dotted , and dash - dotted curves are the photoionization heating rates by h i , he i , and he ii , respectively .
two spikes at @xmath109 and 3.4 are produced by h i and he i reionization , and he ii reionization , respectively .
two thin solid curves are the photoelectric heating rate of the cases of @xmath121 ( top ) and @xmath122 ( bottom ) .
grain type is silicate.,title="fig:",height=207 ] we also present the time evolution of heating rates per a unit volume for the case of @xmath123 and @xmath115 in figure 2 .
the solid , dashed , dotted , and dash - dotted curves are the dust photoelectric , h i , he i , and he ii photoionization heating rates , respectively .
these heating rates show sudden change at the he ii reionization epoch ( @xmath118 ) .
this is because the maximum photon energy increases from 54.4 ev to 1.24 kev suddenly at that time .
we note here that the heating rate by grains is saturated even when the maximum photon energy increases more than about 300 ev in the case of the spectral index of the incident radiation of @xmath108 .
thus , our results are robust for the choice of the maximum energy after the he ii reionization as long as we adopt the energy larger than 300 ev . in addition
, we show the cases of @xmath122 and 0.001 ( and @xmath115 ) by thin curves in figure 2 .
even if we adopt other sets of @xmath26 and @xmath74 , the other curves of the photoionizations do not change so significantly .
let us discuss each case in more detail .
for the cases of @xmath1230.1 , before the he ii reionization , the photoelectric heating by grains is minor heating mechanism relative to the h i photoionization heating .
thus , the effect of the photoelectric heating on the igm temperature is very small , i.e. , temperature enhancement is less than a few % .
after the he ii reionization , the photoelectric heating dominates other heating mechanisms ( but by a factor of less than 3 ) when @xmath115 .
however , the temperature enhancement factor is still less than 1.5 ( solid and dash - dotted curve in figure 1 ) .
thus , the thermal histories of these cases are consistent with that obtained by @xcite when @xmath119 is adopted . for the case of very small ( @xmath124 ) grains ,
the photoelectric heating is always dominant when @xmath125 .
especially , this heating exceeds the he ii photoionization heating by a factor of 10 after the he ii reionization .
then the temperature enhancement factor becomes about 2 at @xmath126 ( dashed curve in figure 1 ) .
thus , the thermal history in this case may be inconsistent with the data from @xcite .
this means that if the typical grain size in the igm is @xmath124 , the dust - to - gas mass ratio in the igm should be much less than 1/100 galactic value .
.temperature enhancement by ig dust . [ cols="^,^,^,^,^",options="header " , ] in table 2 , we summarize the obtained upper limits of the amount of the ig dust at @xmath127 .
let us compare these limits with the ig dust amount estimated from a possible cosmic star formation history ( e.g. , @xcite ) .
once we assume a cosmic star formation history , the cosmic metal amount can be estimated .
roughly , we find @xmath128 a few @xmath129 at @xmath130 ( e.g. , @xcite ) . if we postulate @xmath15 0.5 as the metal depletion to dust grains , which number is suitable for the milky way , we obtain @xmath131 as the total cosmic dust amount at the redshift .
we show in table 2 the upper limits in terms of the density parameter converted from the dust - to - gas ratios by @xmath132 . for the larger grain models , we obtain an upper limit , @xmath133 which is comparable to @xmath134 .
this means that our upper limit may not deny the possibility that most of the dust grains produced in galaxies before @xmath135 escape to the igm as an extreme case .
for very small ( 0.001 ) grain case , we obtain @xmath136 .
thus , a small fraction ( @xmath137 10% ) of such very small grains can escape from the host galaxies to the igm because of @xmath138 .
the latter conclusion is consistent with the suggestion by @xcite , who proposes that the igm is very difficult to be polluted by the very small grains .
the obtained values of the heating rate by the dust photoelectric effect are consistent with those in nath et al .
it indicates that our result is robust against uncertainties of the adopted efficiency factor @xmath41 , photoelectric yield @xmath42 , and energy distribution function of the photoelectrons , etc . , because these are somewhat different between our model and nath et al .
if we consider only graphite grains , the photoelectric heating rate is found to be about 70% of that of only silicate grains , i.e. , silicate grains can be more efficient heating source .
the temperature in the graphite case is lower than that in the silicate case , while the decrement is @xmath137 10% .
these results are also consistent with nath et al .
( 1999 ) .
we shall comment some uncertainties of our analysis .
first , the effect of the spectral index of @xmath8 is examined . as shown in zheng et al .
( 1997 ) , for example , a softer spectrum than @xmath108 is still compatible with observations . if @xmath139 , the heating rates by dust at @xmath140 decrease by a factor of 0.25 ( @xmath117 0.1 @xmath71m)0.5 ( @xmath117 0.001 @xmath71 m ) relatively to our case of @xmath108 .
if @xmath141 is used , the heating rates at @xmath140 decrease by a factor of 0.1 ( 0.1 @xmath71m)0.25 ( 0.001 @xmath71 m ) . in these calculations , @xmath115 and @xmath119 are assumed . as the dust heating becomes less efficient ,
a more abundant ig dust is allowed .
then the upper limits for the ig dust amount increase by a factor of 2 ( 0.001 @xmath71m)4 ( 0.1 @xmath71 m ) for @xmath139 and by a factor of 4 ( 0.001 @xmath71m)10 ( 0.1 @xmath71 m ) for @xmath141 relative to the @xmath108 case , because the dust heating rate is proportional to the dust - to - gas ratio linearly .
to summarize , a softer spectral index constrains more loosely the amount of the ig dust by a factor . it might be possible to take very large @xmath8 .
for example , we consider a background radiation field dominated by only galaxies ( e.g. , @xmath142 ) . in this case , the dust heating rates at @xmath140 become 1/100 ( 0.1 @xmath71m)1/10 ( 0.001 @xmath71 m ) of those in @xmath108 .
however , the thermal histories with @xmath142 never reproduce a steep rise of the observed temperatures at the he ii reionization . in addition
, the transfer correction should be smaller as the spectrum is softer @xcite .
therefore , the background spectrum is unlikely to be as soft as @xmath142 after the he ii reionization .
we should assume a harder index .
a caveat against our results is in the correction factor of the transfer effect .
indeed , the effects of the photoelectric heating and the transfer correction can be offset .
thus , we should check this point by solving the cosmological radiative transfer in the future work .
the current work may give a hint for the origin of the large @xmath114 region of the observed igm clouds .
there is a possibility that the large @xmath114 regions are localized .
if the very small grains are located at the restricted areas , a systematic heating owing to the dust photoelectrons may explain the large-@xmath114 at the localized region .
the principal role of the photoelectric heating in the igm temperature also indicates that the theoretical equation of state in the igm should be reconstructed by including the photoelectric heating effect .
all of these works are being developed by the authors .
we would appreciate valuable comments by the referee , dr .
a. aguiree .
we are grateful to r. hirata for continuous encouragement , and also thank h. hirashita and s. bianchi for very useful discussion .
aki is supported by the research fellowships of the japan society for the promotion of science for young scientists .
abel , t. , & haehnelt , m. g. , 1999 , apj , 520 , l13 aguirre , a. , 1999 , apj , 525 , 583 aguirre , a. , & haiman , z. , 2000 , apj , 532 , 28 becker , r. h. , et al .
, 2001 , aj , 122 , 2850 calzetti , d. , 2001 , pasp , 113 , 1449 cen , r. , 1992 , apjs , 78 , 341 cheng , f. h. , gaskell , c. m. , & koratkar , a. p. , 1991 ,
apj , 370 , 487 cowie l. l , songaila a. , kim t. s. , & hu e. m. , 1995 , aj , 109 , 1522 crane , p. , & hoffman , a. , 1973 , apj , 186 , 787 draine , b. t. , & lee , h. m. , 1984 , apj , 285 , 89 draine , b. t. , & sutin , b. , 1987 , apj , 320 , 803 eigenson , m. s. , 1949 , soviet astr . , 26 , 278 ferrara , a. , nath , b. , sethi , s. k. , & shchekinov , y. , 1999 , mnras , 303 , 301 hui , l. , & gnedin , n. y. , 1997 , mnras , 292 , 27 humason , m. l. , mayall , n. u. , & sandage , a. r. , 1956 , aj , 61 , 97 kitayama , t. , susa , h. , umemura , m. , & ikeuchi , s. , 2001 , mnras , 326 , 1353 kogut , a. , et al .
2003 , apj , submitted ( astro - ph/0302213 ) laor , a. , & draine , b. t. , 1993 , apj , 402 , 441 loeb , a. , & haiman , z. , 1997 , apj , 490 , 571 madau , p. , pozzetti , l. , & dickinson , m. , 1998 , apj , 498 , 106 nath , b. b. , sethi , s. k. , & shchekinov , y. , 1999 , mnras , 303 , 1 nickerson , b. g. , & partridge , r. b. , 1971 , apj , 169 , 203 perlmutter , s. , et al .
, 1999 , apj , 517 , 565 riess , a. g. , et al .
, 1998 , aj , 116 , 1009 rowan - robinson , m. , negroponte , j. , & silk , j. , 1979 , nature , 281 , 635 schaye , j. , theuns , t. , rauch , m. , efstathiou , g. , & sargent , w. l. w. , 2000 , mnras , 318 , 817 sutherland , r. s. , & dopita , m. a. , 1993 , apjs , 88 , 253 spitzer , l. , 1941 , apj , 93 , 369 spitzer , l. , 1978 , physical processes in the interstellar medium , wiley , new york takase , b. 1972 , pasj , 24 , 295 telfer , r. c. , kriss , g. a. , zheng , w. , davidsen , a. f. , & tytler , d. , 2002 , apj , 579 , 500 theuns , t. , et al . , 2002 ,
apj , 574 , l111 todini , p. , & ferrara , a. , 2001 , mnras , 325 , 726 weingartner , j. c. , & draine , b. t. , 2001a , apjs , 134 , 263 ( wd01 ) weingartner , j. c. , & draine , b. t. , 2001b , apj , 548 , 296 wright , e. l. , 1981 , apj , 250 , 1 zheng , w. , kriss , g. a. , telfer , r. c. , grimes , j. p. , & davidsen , a. f. 1997 , apj , 475 , 469
|
this letter investigates the amount of dust in the intergalactic medium ( igm ) . the dust photoelectric heating can be the most efficient heating mechanism in the igm where the density is very small and there are a lot of hard ultraviolet photons . comparing the observational thermal history of igm with a theoretical one taking into account the dust photoelectric heating
, we can put an upper limit on the dust - to - gas ratio , @xmath0 , in the igm . since the rate of the dust photoelectric heating depends on the size of dust , we find the following results : if the grain size is @xmath1 , @xmath0 at @xmath2 is @xmath3 galactic value corresponding to @xmath4 . on the other hand , if the grain size is as small as @xmath5 , @xmath0 is @xmath6 galactic value corresponding to @xmath7 .
[ firstpage ] cosmology : theory dust , extinction intergalactic medium quasars : absorption lines
|
one of the central issues currently under intense investigation in both nuclear physics and astrophysics is the equation of state ( eos ) of neutron - rich nuclear matter @xcite .
for cold nuclear matter of isospin asymmetry @xmath0 at density @xmath1 , the energy per nucleon @xmath2 can be expressed as an even series of @xmath3 that respects the charge symmetry of strong interactions , namely , @xmath4 where @xmath5 is the so - called symmetry energy of the _ _
i__th order @xcite and @xmath6 is the eos of symmetric nuclear matter .
the quadratic term @xmath7 is most important and its value at normal nuclear matter density @xmath8 is known to be around 30 mev from analyzing nuclear masses within liquid - drop models .
essentially , all microscopic many - body calculations have indicated that the higher - order terms are usually negligible around @xmath8 , leading to the so - called empirical parabolic law of eos even for @xmath3 approaching unity for pure neutron matter .
the @xmath7 is then generally regarded as the symmetry energy .
for instance , the value of the quartic term has been estimated to be less than 1 mev at @xmath8 @xcite .
however , the presence of higher - order terms at supra - saturation densities can significantly modify the proton fraction in neutron stars at @xmath9-equilibrium and thus the cooling mechanism of proto - neutron stars @xcite .
it was also found that a tiny quartic term can cause a big change in the calculated core - crust transition density in neutron stars @xcite .
therefore , precise evaluations of the quartic symmetry energy in neutron - rich matter are useful .
although much information about the eos of symmetric nuclear matter @xmath6 has been accumulated over the past four decades , our knowledge about the density dependence of @xmath5 is unfortunately still very poor .
it has been generally recognized that the @xmath5 , especially the quadratic and quartic terms , is critical for understanding not only the structure of rare isotopes and the reaction mechanism of heavy - ion collisions , but also many interesting issues in astrophysics @xcite .
therefore , to determine the @xmath5 in neutron - rich matter has recently become a major goal in both nuclear physics and astrophysics . while significant progress has been made recently in constraining the @xmath7 especially around and below the saturation density , see , e.g. , @xcite , much more work needs to be done to constrain more tightly the @xmath5 at supra - saturation densities where model predictions are rather diverse @xcite .
as dedicated experiments using advanced new detectors have now been planned to investigate the high density behavior of @xmath7 at many radioactive beam facilities around the world , it has become an urgent task to investigate theoretically more deeply the fundamental origin of the extremely uncertain high density behavior of @xmath7 .
it is also of great interest to evaluate possible corrections due to the @xmath10 term to the equation of state of asymmetric nuclear matter . among existing proposals for extracting information about @xmath5 using terrestrial laboratory experiments ,
transport model simulations have shown that many observables in heavy - ion reactions are particularly useful for studying @xmath5 in a broad density range . in these transport model simulations of heavy - ion reactions ,
the eos enters the reaction dynamics and affects the final observables through the single - nucleon potential @xmath11 where @xmath12 is the nucleon momentum .
except in situations where statistical equilibrium is established and thus many observables are directly related to the binding energy @xmath2 after correcting for finite - size effects , what is being directly probed in heavy - ion reactions is the single - nucleon potential @xmath11 .
the latter is , however , directly related to the symmetry energy @xmath7 through the underlying nuclear effective interaction as first pointed out by brueckner , dabrowski and haensel @xcite using k - matrices within the brueckner theory .
they showed that if one expands @xmath11 to the leading order in @xmath3 as in the well - known lane potential @xcite , i.e. , @xmath13 where @xmath14 and @xmath15 are , respectively , the nucleon isoscalar and isovector ( symmetry ) potentials , the quadratic symmetry energy is then @xcite @xmath16 where @xmath17 is the nucleon kinetic energy at the fermi momentum @xmath18 in symmetric nuclear matter of density @xmath1 .
the above equation indicates that the symmetry energy @xmath7 depends only on the single - particle kinetic and potential energies at the fermi momentum @xmath19 .
this is not surprising since the microscopic origin of the symmetry energy is the difference in the fermi surfaces of neutrons and protons .
the first term @xmath20 is the trivial kinetic contribution due to the different fermi momenta of neutrons and protons ; the second term @xmath21 is due to the momentum dependence of the isoscalar potential and also the fact that neutrons and protons have different fermi momenta ; while the term @xmath22 is due to the explicit isospin dependence of the nuclear strong interaction . for the isoscalar potential @xmath14
, reliable information about its density and momentum dependence has already been obtained from high energy heavy - ion collisions , see , e.g. , ref .
@xcite , albeit there are still some rooms for further improvements , particularly at high momenta / densities . on the contrary ,
the isovector potential @xmath15 is still not very well determined , especially at high densities and momenta , and has been identified as the key quantity responsible for the uncertain high density behavior of the symmetry energy as stressed in ref.@xcite . in the present work ,
we first show , using both the differential and integral formulations of the hugenholtz - van hove ( hvh ) theorem @xcite , that the relation in eq.([esym ] ) is valid in general .
we then derive an expression for the quartic symmetry energy @xmath10 in terms of the single - nucleon potential by keeping higher - order terms in the expansion of both the eos and the single - nucleon potential . applying the hvh formalism to two model single - nucleon potentials , namely ,
the bombaci - gale - bertsch - das gupta ( bgbd ) potential @xcite and a modified gogny momentum - dependent - interaction ( mdi)@xcite , which are among the most widely used ones in studying isospin physics based on transport model simulations of heavy - ion reactions @xcite , we examine the relative contributions from the kinetic and various potential terms to @xmath7 and @xmath10 .
we put the emphasis on identifying those terms that dominate the high density behaviors of @xmath7 .
finally , we evaluate the relative importance of the @xmath10 term by studying the @xmath23 ratio as a function of density . the paper is organized as follows . in section [ symmetry ] ,
based on the hvh theorem we derive general expressions for the higher - order symmetry energy terms @xmath7 and @xmath10 in terms of the single - nucleon isoscalar and isovector potentials . the derivation is carried out in section [ hvh ] using the differential form of the hvh theorem by starting from the neutron and proton chemical potentials and in section [ fermi ] using the integral form of the hvh theorem by starting from the total energy density of the system .
numerical results and discussions for both the bgbd and mdi interactions are given in sections [ bgbd ] and [ mdi ] , respectively .
finally , we give a summary in section [ summary ] .
in this section , we present two alternative approaches to derive expressions for the quadratic and quartic symmetry energy terms @xmath7 and @xmath10 .
both are based on the fermi gas model of interacting nucleons and satisfy the hvh theorem that was first derived in ref .
there are , however , some technical differences between the two approaches in that a taylor - series expansion is made on the single - nucleon energy in one approach but on the total energy of the system in the other .
the hugenholtz - van hove theorem @xcite describes a fundamental relation among the fermi energy @xmath24 , the average energy per particle @xmath25 and the pressure of the system @xmath26 at the absolute temperature of zero . for a one - component system , in terms of the energy density @xmath27
, the general hvh theorem can be written as @xcite @xmath28 the above relation has been strictly proven to be valid for any interacting self - bound infinite fermi system . it does not depend upon the precise nature of the interaction . in the special case of nuclear matter at saturation density where the pressure @xmath26 vanishes ,
the average energy per nucleon becomes equal to the fermi energy , i.e. , @xmath29 .
it is worthwhile to stress that the general hvh theorem of eq.([hvh ] ) is valid at any arbitrary density as long as the temperature remains zero @xcite .
in fact , a successful theory for nuclear matter is required not only to describe satisfactorily all saturation properties of nuclear matter but also to fulfill the general hvh theorem at any density . in the following , we use the general hvh theorem to derive the relation between the nuclear symmetry energy and the single - nucleon potential . according to the hvh theorem , the chemical potentials of neutrons and protons in isospin asymmetric nuclear matter of energy density @xmath30 are
, respectively @xcite , @xmath31 where @xmath32 is the kinetic energy and @xmath33 is the neutron / proton single - particle potential .
the fermi momenta of neutrons and protons are @xmath34 and @xmath35 , respectively . subtracting eq.([chemup ] ) from eq.([chemun ] ) gives @xcite @xmath36+[u_n(\rho,\delta , k_f^n)-u_p(\rho,\delta , k_f^p)]=\frac{\partial \xi } { \partial \rho_n}-\frac{\partial \xi } { \partial \rho_p}.\end{aligned}\ ] ] the nucleon single - particle potentials can be expanded as a power series of @xmath3 while respecting the charge symmetry of nuclear interactions under the exchange of neutrons and protons , @xmath37 @xmath38 if one neglects the higher - order terms ( @xmath39 , @xmath40 , ... ) , eq.([unup1 ] ) and eq.([unup2 ] ) reduce to the lane potential in eq.([lane ] ) .
expanding both the kinetic and potential energies around the fermi momentum @xmath19 , the left side of eq.([unminusup ] ) can be further written as @xmath41+[u_n(\rho,\delta , k_f^n)-u_p(\rho,\delta , k_f^p ) ] \nonumber\\ = & & \sum_{i=1,2,3 ... }\frac{1}{i!}\frac{\partial^i [ t(k)+u_0(\rho , k)]}{\partial k^i}|_{k_f } k_f^i \nonumber\\ \times&&[(\sum\limits_{j=1,2,3 .. }f(j)\delta^j)^i-(\sum\limits_{j=1,2,3 .. }f(j)(-\delta)^j)^i ] \nonumber\\ + & & \sum_{l=1,2,3 ... }u_{sym , l}(\rho , k_f)[\delta^{l}-(-\delta)^{l } ] \nonumber\\ + & & \sum_{l=1,2,3 ...
}\sum_{i=1,2,3 ... }\frac{1}{i!}\frac{\partial^iu_{sym , l}(\rho , k)}{\partial k^i}|_{k_f } k_f^i \nonumber \\ \times & & [ ( \sum\limits_{j=1,2,3 .. }f(j)\delta^j)^i \delta^{l}-(\sum\limits_{j=1,2,3 .. }f(j)(-\delta)^j)^i ( -\delta)^{l } ] \nonumber \\ = & & [ \frac{2}{3 } \frac{\partial [ t(k)+u_0(\rho , k)]}{\partial k}|_{k_f}k_f + 2 u_{sym,1}(\rho , k_f)]\delta + ... , \end{aligned}\ ] ] where we have introduced the function @xmath42 $ ] . for the right side of eq.([unminusup ] ) , expanding in powers of @xmath3 gives @xmath43 comparing the coefficient of each @xmath44 term in eq.([leftside ] ) with that in eq.([rightside ] ) then gives the symmetry energy of any order .
for instance , the quadratic term @xmath45}{\partial k}|_{k_f}k_f + \frac{1}{2 } u_{sym,1}(\rho , k_f)\nonumber \\ & = & \frac{1}{3 } t(k_f ) + \frac{1}{6 } \frac{\partial u_0}{\partial k}\mid _ { k_f}\cdot k_f + \frac{1}{2}u_{sym,1}(\rho , k_f)\end{aligned}\ ] ] is identical to that in eq.([esym ] ) , while the quartic term can be written as @xmath46}{\partial k}|_{k_f } k_f \right .
\nonumber \\ & & \left .
- \frac{1}{108 } \frac{\partial^2 [ t(k)+u_0(\rho , k)]}{\partial k^2}|_{k_f } k_f^2 + \frac{1}{648 } \frac{\partial^3 [ t(k)+u_0(\rho , k)]}{\partial k^3}|_{k_f } k_f^3 \right.\nonumber \\ & & \left . - \frac{1}{36 } \frac { \partial u_{sym,1}(\rho , k)}{\partial k}|_{k_f } k_f + \frac{1}{72 } \frac { \partial^2 u_{sym,1}(\rho , k)}{\partial k^2}|_{k_f } k_f^2 \right . \nonumber \\ & & \left .
+ \frac{1}{12 } \frac{\partial u_{sym,2}(\rho , k)}{\partial k}|_{k_f } k_f + \frac{1}{4 } u_{sym,3}(\rho , k_f)\right].\end{aligned}\ ] ] the symmetry energy of any order obtained in the previous subsection can also be derived within the interacting fermi gas model @xcite .
although the derivation is more tedious , it is physically interesting and mathematically instructional .
there are different ways to calculate the total energy of a fermi system at a given density @xmath1 . as explained in detail by bertsch and das gupta @xcite
, one could determine the total energy starting with empty space , adding particles until the desired density is reached .
each added particle would contribute an energy @xmath47 , where @xmath12 is the fermi momentum corresponding to the density @xmath48 of particles already added to the system .
as stressed by bertsch and das gupta , the single - particle potential @xmath49 is not the same as the potential energy per particle as one might at first guess . in this way ,
the total energy density of the asymmetric nuclear matter written in coordinate space is @xmath50 d\rho_x + \int_0^{\rho^p } [ k(\rho_x)^2/2m+u_p(\rho_x , k(\rho_x))]d\rho_x,\ ] ] where @xmath51 .
the single particle potential @xmath49 for the neutron or proton in the first or second integral in eq.([erspace ] ) can be rewritten as @xmath52 where @xmath53 in terms of the local isospin asymmetry @xmath54 ( to be discussed in detail in the following ) .
the single particle potential can be further expanded as a series of @xmath54 @xmath55 similar to the interpretation of eq.([erspace ] ) , one can raise the momenta of all particles in the system from zero to the fermi momentum corresponding to the density @xmath1 . in either the coordinate or momentum space , as one increases the density or momentum for both neutrons and protons to build up the desired fermi system , the local isospin asymmetry @xmath54 changes continuously as a function of density or momentum , i.e. , @xmath56\rightarrow \delta^*=\left[\frac{k^3}{k_f^3}-1\right]\end{aligned}\ ] ] where @xmath48 is either the neutron or proton density used in the two terms of eq.([erspace ] ) , and @xmath12 is the neutron or proton momentum used in the two terms of eq.([ekspace ] ) below .
at the respective fermi surfaces of neutrons and protons , the local isospin asymmetry @xmath54 reduce to the global one @xmath57=\left[\frac{(k_f^n)^3}{k_f^3}-1\right]= \delta , \nonumber\\ & & \delta^*(k_f^p)=\left[\frac{2\rho^p}{\rho}-1\right]=\left[\frac{(k_f^p)^3}{k_f^3}-1\right]= ( -\delta).\end{aligned}\ ] ] in order to make use of the same technique in expanding both the kinetic and potential terms in @xmath3 series , it is more convenient to express the energy density in momentum space @xmath58 k^2dk + \int_0^{k_f^p } [ t(k)+u_p(\rho,\delta^*,k ) ] k^2dk\right].\ ] ] before we proceed to derive expressions for @xmath7 and @xmath10 , it is critical to examine whether the energy density given in eq.([erspace ] ) or eq.([ekspace ] ) satisfies the hvh theorem . noticing that @xmath59}{\partial k_f^n}\right ) = \frac{\partial \xi } { \partial k_f^n } \left[\frac{\pi^2}{(k_f^n)^2}\right],\end{aligned}\ ] ]
then it is straightforward to show using eq.([ekspace ] ) that @xmath60 ( k_f^n)^2 \right\ } \left[\frac{\pi^2}{(k_f^n)^2}\right ] \nonumber\\ & = & t(k_f^n)+u_n(\rho,\delta , k_f^n).\end{aligned}\ ] ] similarly , one can show that @xmath61 thus , the hvh theorem is indeed satisfied by the energy density expressed in both eq.([erspace ] ) and eq.([ekspace ] ) . to obtain expressions for @xmath7 and @xmath10
, we make a taylor - series expansion of the energy density in eq.([ekspace ] ) .
to proceed , it is useful to first recall that for any continuous function @xmath62 the taylor - series expansion of the integral @xmath63 around @xmath19 leads to @xcite @xmath64\cdot(k_f^i - k_f)^2+ ... ,\end{aligned}\ ] ] moreover , it is easy to show that @xmath65k_f \nonumber \\k_f^p - k_f & = & \left[-\frac{1}{3}\delta-\frac{1}{9}\delta^2-\frac{5}{81}\delta^3-\frac{10}{243}\delta^4 + ... \right]k_f.\end{aligned}\ ] ] the kinetic energy per nucleon can then be expanded as @xmath66 \nonumber \\ & + & \frac{\alpha}{2 ! } \frac{\partial [ t(k ) k^2]}{\partial k}|_{k_f}[(k_f^n - k_f)^2+(k_f^p - k_f)^2 ] \nonumber \\ & + & \frac{\alpha}{3 ! } \frac{\partial^2 [ t(k ) k^2]}{\partial k^2}|_{k_f } [ ( k_f^n - k_f)^3+(k_f^p - k_f)^3 ] \nonumber \\ & + & \frac{\alpha}{4 ! } \frac{\partial^3 [ t(k ) k^2]}{\partial k^3}|_{k_f}[(k_f^n - k_f)^4+(k_f^p - k_f)^4 ] \nonumber \\ & = & 2 \alpha \int_0^{k_f } t(k ) k^2dk + \frac{1}{6 } \frac{\partial t(k)}{\partial k}\mid_{k_f}k_f \delta^2 \nonumber \\
& + & \left[\frac{5}{324}\frac{\partial t(k)}{\partial k}\mid_{k_f}k_f - \frac{1}{108}\frac{\partial^2 t(k)}{\partial k^2}\mid_{k_f}k_f^2 + \frac{1}{648}\frac{\partial^3 t(k)}{\partial k^3}\mid_{k_f}k_f^3 \right]\delta^4 + ... , \nonumber\\\end{aligned}\ ] ] where @xmath67 .
thus , the kinetic contribution to the @xmath7 is @xmath68 and its contribution to @xmath10 is @xmath69 for the potential energy per nucleon written in momentum space , we first expand the single - nucleon potential @xmath33 in local isospin asymmetry @xmath70 @xmath71 \nonumber \\&+ & \sum_{i=1,2,3 ... }\left[\alpha \int_0^{k_f^n } u_{sym , i}(\rho , k)(\delta^*)^i k^2dk + \alpha
\int_0^{k_f^p } u_{sym , i}(\rho , k)(\delta^*)^i k^2dk\right].\nonumber\\\end{aligned}\ ] ] we then expand each integral in the above equation around the fermi momentum @xmath19 using eq .
( [ ts ] ) .
the @xmath72 contribution to @xmath73 is @xmath74 \delta^4+ .... \nonumber\\\end{aligned}\ ] ] contributions from the higher - order potential terms in @xmath3 can be obtained similarly . for the second order - symmetry energy @xmath7 ,
only the @xmath75 term contributes @xmath76 k^2dk + \alpha \int_0^{k_f^p } u_{sym,1}(\rho , k ) [ ( \frac{k}{k_f})^3 - 1 ] k^2dk\nonumber \\ & = & 2\alpha \int_0^{k_f } u_{sym,1}(\rho , k ) [ ( \frac{k}{k_f})^3 - 1 ] k^2dk \nonumber \\&+ & \frac{1}{2 } u_{sym,1}(\rho , k_f ) \delta^2 + \left [ - \frac{1}{36 } \frac { \partial u_{sym,1}(\rho , k)}{\partial k}|_{k_f } k_f + \frac{1}{72 } \frac { \partial^2 u_{sym,1}(\rho , k)}{\partial k^2}|_{k_f } k_f^2 \right ] \delta^4+ .... \nonumber\\\end{aligned}\ ] ] to obtain the fourth - order symmetry energy @xmath10 , all @xmath75 , @xmath77 , and @xmath78 terms are needed , i.e. @xmath79 ^ 2 k^2dk + \alpha \int_0^{k_f^p } u_{sym,2}(\rho , k ) [ ( \frac{k}{k_f})^3 - 1]^2 k^2dk\nonumber \\ & = & 2\alpha \int_0^{k_f } u_{sym,2}(\rho , k ) \left[\left(\frac{k}{k_f}\right)^3 - 1\right]^2 k^2dk + \frac{1}{12 } \frac{\partial u_{sym,2}(\rho , k)}{\partial k}|_{k_f } k_f
\delta^4+ ... ,\nonumber\\\end{aligned}\ ] ] and @xmath80 ^ 3 k^2dk + \alpha \int_0^{k_f^p } u_{sym,3}(\rho , k ) [ ( \frac{k}{k_f})^3 - 1]^3 k^2dk\nonumber \\ & = & 2\alpha \int_0^{k_f } u_{sym,3}(\rho , k ) \left[\left(\frac{k}{k_f}\right)^3 - 1\right]^3 k^2dk + \frac{1}{4 } u_{sym,3}(\rho , k_f ) \delta^4+ .... \nonumber\\\end{aligned}\ ] ] combining all coefficients of the @xmath39 and @xmath81 terms in both the kinetic and potential parts given above , one sees that the @xmath7 and @xmath10 obtained here are exactly the same as those given in using directly the hvh theorem . moreover , they agree with earlier results obtained by brueckner _
et al_. @xcite .
as shown in the previous section , the symmetry energy can be explicitly separated into the kinetic energy term @xmath82 and the potential terms @xmath72 and @xmath83 at the fermi momentum @xmath19 . to evaluate their relative contributions to the symmetry energies , especially for the second - order and fourth - order terms @xmath7 and @xmath10
, we consider in this section two typical single - nucleon potentials that have been widely used in tansport model simulations of heavy - ion reactions . as a first example , we use the phenomenological potential of bombaci - gale - bertsch - das gupta @xcite @xmath84\delta\nonumber\\ & & + \frac{4}{5\rho_0}\left[{\frac{1}{2 } ( 3c-4z_1 ) \mathcal{i_{\tau } } + ( c+2z_1)\mathcal{{i_{\tau^{\prime}}}}}\right]+ \left({c \pm \frac{c-8z_1}{5}\delta}\right)u\cdot g(k),\end{aligned}\ ] ] where @xmath85 is the reduced density and @xmath86 is for neutrons / protons . in the above , we have @xmath87\int d^3k f_{\tau}(k)g(k)$ ] with @xmath88 $ ] being a momentum regulator and @xmath89 being the phase space distribution function .
the parameter @xmath90 has the value @xmath91 , where @xmath92 is the nucleon fermi wave number in symmetric nuclear matter at @xmath8 . with a=-144 mev , b=203.3 mev , c=-75 mev and @xmath93
, the bgbd potential reproduces all ground state properties including an incompressibility @xmath94=210 mev for symmetric nuclear matter @xcite .
the three parameters @xmath95 and @xmath96 can be adjusted to give different symmetry energy @xmath7 and the neutron - proton effective mass splitting @xmath97 @xcite .
for example , the parameter set @xmath98 mev , @xmath99 and @xmath100 leads to @xmath101 while the one with @xmath102 mev , @xmath103 and @xmath104 leads to @xmath105 at all non - zero densities and isospin asymmetries .
on expanding the bgbd potential in @xmath3 , the coefficients of the first few terms are @xmath106 + c u\cdot g(k),\end{aligned}\ ] ] @xmath107\nonumber\\ & + & \frac{c-8z_1}{5}u\cdot g(k_f ) + \frac{c-8z_1}{5}u\cdot g(k),\end{aligned}\ ] ] @xmath108 @xmath109 thus , the second - order symmetry energy @xmath7 is given by @xmath110 \nonumber\\&+ & \frac{c-8z_1}{5}u\cdot g(k_f),\end{aligned}\ ] ] and the fourth - order symmetry energy @xmath10 is @xmath111 \nonumber \\ & = & \frac{\hbar ^2}{162 m } ( \frac{3\pi^2 \rho}{2})^{\frac{2}{3}}- \frac{c}{81 } u \frac{k_f^2}{\lambda^2}(\frac{10k_f^4}{\lambda^4}+\frac{5k_f^2}{\lambda^2}+1 ) g(k_f)^4 + \frac{c-8z_1}{135}u \frac{k_f^2}{\lambda^2}(\frac{5k_f^2}{\lambda^2}+1 ) g(k_f)^3.\nonumber\\\end{aligned}\ ] ] ) and the isovector potential part ( @xmath75 ) of the symmetry energy @xmath112 from the bgbd potential with @xmath101 ( left ) and for @xmath113 ( right).,width=453 ] in fig .
[ bgbde2 ] we compare @xmath7 and its three components in the two cases of @xmath101 and @xmath105 .
it is seen that the kinetic and isoscalar contributions are the same in both cases . however , they have significantly different isovector potentials @xmath75 , leading thus to different @xmath7 especially at supra - saturation densities . with the bgbd potential for @xmath101 ( left ) and for @xmath113 ( right).,width=453 ] various contributions to the fourth - order symmetry energies
@xmath10 in the two cases are compared in fig.[bgbde4 ] .
similar to @xmath7 , the contributions of the @xmath82 and @xmath72 terms to @xmath10 are positive and they are the same in both cases .
interestingly , the @xmath75 term also plays the most important role in determining the high - density behavior of @xmath10 .
it is positive in the case of @xmath101 but negative in the case of @xmath105 , resulting in very different behaviors of @xmath10 at supra - saturation densities .
moreover , it is interesting to note that @xmath10 receives no contribution from the @xmath77 term .
this is not surprising because the @xmath77 term in the bgbd interaction is momentum independent and its contribution to @xmath10 is actually @xmath114=0 . on the contrary
, the @xmath78 term still contributes to @xmath10 via @xmath115 although it is momentum independent too . in the two cases considered here , the contributions from the @xmath78 term also have opposite sign . over @xmath112 with the bgbd potential for @xmath101 ( left ) and for @xmath113 ( right).,width=453 ] to compare the fourth - order term @xmath10 with the second - order term @xmath7 more clearly
, we show in fig.[bgbdr ] their ratio @xmath116 as a function of the reduced density @xmath117 .
obviously , the relative value of @xmath10 is generally small .
however , it can reach up to about @xmath118 at high densities for both cases of @xmath119 and @xmath113 .
it may thus lead to an appreciable modification in the proton fraction and therefore the properties of neutron stars at @xmath9-equilibrium . in this subsection
, we discuss the symmetry energy obtained from the mdi interaction @xcite , which is derived from the hartree - fock approximation using a modified gongy effective interaction @xcite @xmath120 in the above , @xmath121 ( @xmath122 ) for neutrons ( protons ) and @xmath123 ; @xmath124 is the density - dependence parameter ; @xmath125 is the phase space distribution function at coordinate @xmath126 and momentum @xmath127 .
the parameters @xmath128 and @xmath90 are obtained by fitting the nuclear matter saturation properties @xcite .
the momentum dependence of the symmetry potential stems from the different interaction strength parameters @xmath129 and @xmath130 for a nucleon of isospin @xmath131 interacting , respectively , with unlike and like nucleons in the background fields .
more specifically , @xmath132 mev while @xmath133 mev .
the quantities @xmath134 and @xmath135 are parameters . the parameters @xmath136 and @xmath137 in the mdi single - particle potential are related to the @xmath138 and @xmath139 in the gogny effective interaction via @xmath140 and @xmath141 @xcite . the parameter @xmath142 is related to the spin(isospin)-dependence parameter @xmath143 via @xmath144 @xcite . on expanding the single - nucleon potential in @xmath3 ,
the first four terms are @xmath145,\nonumber\\\end{aligned}\ ] ] @xmath146 @xmath147[(p - p_f)^2+\lambda^2 ] } - \ln \frac{(p+p_{f})^{2 } + \lambda ^{2}}{(p - p_{f})^{2}+\lambda ^{2 } } \right],\nonumber\\\end{aligned}\ ] ] @xmath148 ^ 2[(p - p_f)^2+\lambda^2]^2 } \right.\nonumber \\ & & \left.- \ln \frac{(p+p_{f})^{2 } + \lambda ^{2}}{(p - p_{f})^{2}+\lambda ^{2}}\right].\end{aligned}\ ] ] according to eq.([esym2 ] ) , the second - order symmetry energy @xmath7 is @xmath149 \nonumber \\+ & & \frac{(a_l - a_u)}{4 } \frac { \rho } { \rho_0 } - x\frac{b}{\sigma+1}\frac{\rho^ { \sigma}}{\rho_0^{\sigma } } \nonumber \\+ & & \frac{(c_{\tau,\tau}-c_{\tau,\tau'})}{3\rho_0 } \frac{\pi\lambda^2}{h^3}2p_f\textrm{ln}\left(\frac{4p_f^2+\lambda^2}{\lambda^2}\right),\end{aligned}\ ] ] and according to eq.([esym4 ] ) the fourth - order symmetry energy @xmath10 is @xmath150 \nonumber \\&= & \frac{\hbar ^2}{162 m } ( \frac{3\pi^2 \rho}{2})^{\frac{2}{3 } } - \frac{c_{\tau,\tau}}{3^{5}\rho _ { 0}\rho } \left ( \frac{4\pi } { h^{3}}\right ) ^{2}\lambda ^{2}\left [ 7\lambda ^{2}p_{f}^{2}\ln \frac{4p_{f}^{2}+\lambda ^{2}}{\lambda ^{2}}-\frac{4(7\lambda
^{4}p_{f}^{4}+42\lambda ^{2}p_{f}^{6}+40p_{f}^{8}}{(4p_{f}^{2}+\lambda ^{2})^{2}}\right ] \nonumber \\&-&\frac{c_{\tau,\tau'}}{3^{5}\rho _ { 0}\rho } \left ( \frac{4\pi } { h^{3}}\right ) ^{2}\lambda ^{2}\left [ ( 7\lambda ^{2}p_{f}^{2}+16p_{f}^{4})\ln \frac{4p_{f}^{2}+\lambda ^{2}}{\lambda ^{2}}-28p_{f}^{4}-\frac{8p_{f}^{6}}{\lambda ^{2}}\right].\nonumber\\\end{aligned}\ ] ] as one expects , the above expressions are identical to those derived directly from the exact mdi eos using @xcite @xmath151 ) and the isovector potential part ( @xmath75 ) of the symmetry energy @xmath112 from the mdi interaction with @xmath152 , 0 and -1.,width=453 ] in fig .
[ mdie2 ] , we show the kinetic ( t ) , isoscalar ( @xmath72 ) and isovector ( @xmath75 ) potential contributions to @xmath112 for the three different spin ( isospin)-dependence parameter @xmath152 , 0 , and -1 .
we notice that the kinetic ( @xmath82 ) and the isoscalar potential ( @xmath72 ) contributions are the same for the three different @xmath142 values . as pointed out in ref .
@xcite , it is the isovector potential @xmath75 that is causing the different density dependence of @xmath112 .
for instance , with @xmath152 the @xmath75 term decreases very quickly with increasing density and thus results in a super - soft symmetry energy at supra - saturation densities . on the contrary ,
the symmetry energy @xmath112 at supra - saturation densities is very stiff for both @xmath153 and @xmath1541 as the contribution of the @xmath75 term becomes very positive with smaller values of @xmath142 . from the mdi interaction.,width=453 ] unlike the second - order term @xmath112 , the fourth - order symmetry energy @xmath155 is independent of the spin ( isospin)-dependence parameter @xmath142 .
shown in fig .
[ mdie4 ] are the various contributions to the fourth - order symmetry energy @xmath155 . comparing these with the results obtained using the bgbd in fig .
[ bgbde4 ] , we find that the @xmath82 and @xmath72 terms from these two interactions are almost identical
. however , there exists some differences for other terms . for the mdi interaction ,
the @xmath77 term is negative and becomes very important for determining @xmath155 .
one the contrary , the contributions from the @xmath75 and @xmath78 terms are positive and they are relatively small as compared to @xmath77 .
generally , the behavior of @xmath155 from the mdi interaction is very similar to that from the bgbd interaction for the case of @xmath101 . over @xmath112 obtained from the mdi interaction as a function of reduced density @xmath117 for @xmath152 , 0 , and -1.,width=453 ] to compare more directly @xmath155 with @xmath112 , their ratio @xmath156
is plotted in fig .
[ mdir ] as a function of reduced density for @xmath152 , 0 , and -1 .
it is seen that with @xmath152 there is a sharp break in the curve around @xmath157 .
this is because the second - order symmetry energy @xmath112 changes from positive to negative around @xmath157 in this case .
however , this is not the case for both @xmath158 and @xmath153 where @xmath112 remains positive at all densities . in all cases
, @xmath155 is very small compared to @xmath112 . for both bgbd and mdi interactions , the small values of @xmath155 up to several times
the normal density clearly shows that the parabolic approximation of the eos is well justified for most purposes .
however , cares have to be taken in evaluating the core - crust transition density where the energy curvatures are involved @xcite .
in summary , using the hugenholtz - van hove theorem we have derived general expressions for the quadratic and quartic symmetry energies in terms of single - particle potentials in isospin asymmetric nuclear matter .
identical results are obtained by using two approaches , i.e. , one based on the single - particle potential and the other based on the total energy , that are physically identical although mathematically different . by using the derived analytical formulas ,
the symmetry energies are explicitly separated into the kinetic and several potential parts .
the formalism is applied to two typical single - nucleon potentials , namely the bombaci - gale - bertsch - das gupta ( bgbd ) potential and the modified gogny momentum - dependent - interaction ( mdi ) , that are widely used in transport model simulations of heavy - ion reactions .
we find that for both interactions the isovector potential is responsible for the uncertain high density behavior of the quadratic symmetry energy .
also , the magnitude of the quartic symmetry energy in both cases is found to be significantly smaller than that of the quadratic symmetry energy .
we expect that the analytical formulas for the nuclear symmetry energies derived in the present study will be useful in extracting reliable information about the eos of neutron - rich nuclear matter from heavy - ion reactions .
this work is supported in part by the us national science foundation grants phy-0757839 and phy-0758115 , the research corporation under grant no.7123 , the welch foundation under grant no .
a-1358 , the texas coordinating board of higher education grant no.003565 - 0004 - 2007 , the national natural science foundation of china grants 10735010 , 10775068 , 10805026 , and 10975097 , shanghai rising - star program under grant no .
06qa14024 , the national basic research program of china ( 973 program ) under contract no .
2007cb815004 and 2010cb833000 .
|
using the hugenholtz - van hove theorem , we derive general expressions for the quadratic and quartic symmetry energies in terms of single - nucleon potentials in isospin asymmetric nuclear matter .
these analytical relations are useful for gaining deeper insights into the microscopic origins of the uncertainties in our knowledge on nuclear symmetry energies especially at supra - saturation densities . as examples ,
the formalism is applied to two model single - nucleon potentials that are widely used in transport model simulations of heavy - ion reactions . , , , symmetry energy , nuclear potential , heavy - ion collision , transport model
|
the increase of wireless internet - of - things ( iot ) requires a large volume of vacant frequency bands that the current dedicated spectrum policy can not cope with . to handle the spectrum shortage ,
the devices need to detect and access an unoccupied spectrum in an opportunistic manner @xcite , @xcite . however , most wireless iot devices are hard to perform precise spectrum sensing by themselves due to their limited hardware capability and less cost @xcite .
spectrum sensors are essential as a part of the infrastructure for the sole purpose of interference monitoring @xcite . the locational difference of an iot device and the corresponding sensor causes an observation error of spectrum occupancy .
a mathematical model reflecting this spatial relationship is thus required . to this end
, we derive a conditional interference distribution ( cid ) at an iot device for a given measured interference from the sensor using stochastic geometry ( sg ) .
we find that its shape is skewed to the left of the sensor measurement and it has a long tail to the right side .
this asymmetric tendency becomes intensified with increasing the measured interference level at the sensor . in other words ,
the iot devices may experience less interference than the measured value with high probability . from the perspective of an opportunistic spectrum access
, the iot device would has more chances to exploit the band while guaranteeing the quality of services ( qos ) of primary communications .
it is worth mentioning that the existing interference distributions based on sg @xcite@xcite can not explain the above asymmetric spatial correlation between the sensors and the iot devices , which is a key to design cognitive radio iot networks .
we propose a novel cognitive random access algorithm to adjust its transmission probability in a distributed manner according to the measured interference .
based on the cid , the proposed algorithm improves an area spectral efficiency ( ase ) in return while satisfying the requirement of primary users .
analytic and numerical results show that our algorithm outperforms conventional aloha @xcite and threshold based random access protocol with a hard decision , where an iot device can access the medium only when a measured sensor value is lower than the predetermined threshold .
the rest of the paper is organized as follows . in section
ii , we describe the system model and analyze the cid . in section iii
, we investigate our cognitive random access scheme and verify the performance through simulations .
section iv concludes the paper .
in cognitive iot networks with spectrum sensors , primary and iot services try to access a shared spectrum band . the primary transmitter ( pt ) has a license to access the spectrum .
the iot transmitter , denoted as secondary transmitter ( st ) , may acquire opportunistic access to the spectrum by exploiting the sensor measurement as shown in fig .
[ scenario ] . consider a pair of st and a sensor , where the sensor is located at the center .
the distance between the sensor and st is @xmath0 .
the sensor and the st are surrounded by pts , whose locations follow a poisson point process ( ppp ) @xmath1 of density @xmath2 .
the locations of sts follow another independent ppp @xmath3 of density @xmath4 . pts and sts in the network use transmission powers
@xmath5 and @xmath6 , respectively .
we consider distance - dependent path loss , where @xmath7 .
parameter @xmath8 is a path - loss exponent .
fading is modeled as an independent and identical random variable @xmath9 .
transmissions made by pus impose an aggregate interference @xmath10 at the sensor as follows : @xmath11 in this paper , we assume that spectrum sensors use energy detection @xcite and measure the aggregate interference @xmath10 without an error .
we focus on an interference distribution at the point at a distance @xmath0 from the sensor , when the sensor measurement @xmath12 is given .
we specify the conditional interference distribution ( cid ) function as @xmath13 to derive the cid , we consider geometric situations , where pts impose an aggregate interference to a sensor and its adjacent st as depicted in fig .
let @xmath14 denote the distance between a sensor and its nearest pt .
the variable @xmath15 is the distance between the pt and the corresponding st .
two lines from the sensor to the pt and the st form an angle @xmath16 in radian unit .
* proposition 1 . *
_ when a measured interference @xmath17 , the cid function @xmath18 is given as : _
@xmath19 where @xmath20 , and @xmath21 is a solution of the following equation : @xmath22 . _
proof : _ appendix . in proposition 1 , we consider that the nearest pt has a dominant effect on the cid and approximate the distribution in terms of @xmath21 denoting the estimated distance between the sensor and its nearest pt . when the pathloss exponent @xmath23 is 4 , then the estimated distance @xmath21 is equal to @xmath24 .
the approximation of the cid is tight when the pt density @xmath2 is up to 0.003 , or 3000 pts / km@xmath25 .
it implies that the approximation would be effective to the case that the primary network is cellular network .
[ simul_cid1 ] and fig .
[ simul_cid2 ] show two cids with respect to the different sensor measurement @xmath12 .
the shape of cid is skewed to the left of the sensor measurement and characterized by having a long tail .
variance and skewness of the cid both increase with @xmath26 it implies that the actual received interference at st @xmath27 may be lower than the sensor measurement @xmath12 with a considerable probability .
in other words , there would be more transmission opportunities for sts .
this phenomenon gives us an insight that sts can access the medium more aggressively while not degrading the primary communication qualities .
mw , @xmath28 , @xmath29 , and @xmath30 , or 3000 pts / km@xmath25.,scaledwidth=27.0% ] mw , @xmath28 , @xmath29 , and @xmath30 , or 3000 pts / km@xmath25.,scaledwidth=27.0% ]
fig . [ simul_topo ] shows the snapshot of the network topology where pts and sts are randomly located with the density of @xmath31 , and @xmath32 .
although the same density is applied to the network , we can observe the regional variance of the population , which makes different local interference conditions .
the sts in subarea a are located in relatively sparse environment with low population of pts , where the sts could access the channel without interruption . on the other hand ,
the sts in subarea b are located in relatively high region interference imposed by pts , and the transmission attempt of sts in b would be obstructed . to deal with these regional differences ,
we propose a cognitive random access by tuning each user s transmission probability based on the cid ( 3 ) .
it is worth noting that the cid gives us the probability that the aggregate interference at a st is lower than an arbitrary threshold . unlike the conventional aloha ,
sts have the different transmission probabilities with respect to the interference measured by the sensors .
we assume that the time is slotted and synchronized in the networks .
we focus on a snapshot of the communication process , where the network topology does not change during each time slot .
each transmitter in the network always has enough data to transmit .
let us assume that the sts know the corresponding sensor location .
the time delay that the st receives interference measurement from its sensor is negligible .
is 0.001 and the st density @xmath4 is 0.002 .
, scaledwidth=25.0% ] -9pt our purpose is to improve the area spectral efficiency ( ase ) @xmath33 , the sum of data rates per unit bandwidth in the unit area @xcite , while protecting primary networks . under a constraint of satisfying a required outage probability of primary transmissions ,
we determine transmission probabilities of sts @xmath34 in order to maximize @xmath33 as follows : @xmath35 p_{s } \log(1+\beta ) \\ & & \hspace{-40pt}\quad\quad\quad\text{subject to}\text { } \quad\quad \text{pr}\{sir_p \leq \beta \ } \leq \tau ,
\label{eq : p1c1}\\ & & \hspace{-40pt}\text { } \quad\quad\quad\quad\quad\qquad\qquad 0\leq p_i \leq1 \quad\forall i , \label{eq : p1c2}\end{aligned}\ ] ] where @xmath36 is a target sir threshold , and @xmath37 is transmission success probability of secondary user .
we neglect noise in our analytical calculations . in the perspective of a typical st ,
interferer density is @xmath38 $ ] . from @xcite ,
the probability @xmath37 can be calculated as follows : @xmath39=e^{-(\lambda_p+\lambda_s\text{e}_i [ p_i ] ) r_s^2 \left(\frac{p_p\beta}{p_s}\right)^{2/\alpha}c(\alpha)}.\ ] ] where @xmath40 and @xmath41 .
the parameter @xmath42 is a distance between secondary transmitter and receiver . here
, we assume that the distance @xmath42 is same for all sts and their corresponding receivers .
the constraint assures that the outage probability of the primary communications can not exceed the target value .
* proposition 2 . *
_ the optimal transmission probability @xmath43 should satisfy _
@xmath44\!=\!\max\!\left[0,\min\!\left\{\!1,\frac{1}{\lambda_s}\left(\frac{\ln\left(1/(1-\tau)\right)}{r_p^2\left(\frac{p_s\beta}{p_p}\right)^{2/\alpha}{\frac{2}{\alpha}}c(\alpha)}-\lambda_p\ ! \right)\!\right\}\!\right]\!.\ ] ] _ proof _ : from the outage probability in @xcite , we can represent the constraint as follows : @xmath45)r_p^2 \left(\frac{p_s\beta}{p_p}\right)^{2/\alpha}c(\alpha)},\ ] ] @xmath46\leq\frac{1}{\lambda_s}\left\{\frac{\ln\left(1/(1-\tau)\right)}{r_p^2 \left(\frac{p_s\beta}{p_p}\right)^{2/\alpha}c(\alpha)}-\lambda_p \right\}.\ ] ] we can rewritten the expectation term @xmath47 $ ] of as @xmath48 without loss of generality , where @xmath49 is the number of sts .
then , we can relax constraint and obtain the following lagrangian function : @xmath50 p_{s } \log(1+\beta ) \nonumber\\ & + \mu \left \{\frac{n}{\lambda_s}\left(\frac{\ln\left(1/(1-\tau)\right)}{r_p^2 \beta^{2/\alpha}c(\alpha)}-\lambda_p \right ) -\sum_i^np_i \right\},\end{aligned}\ ] ] where @xmath51 is a nonnegative lagrangian multiplier .
the karush - kuhn - tucker ( kkt ) conditions of the problem * p1 * are necessary for optimality . the kkt conditions are given as follows : = -0.5mu=-0.5mu=-0.5mu@xmath52 @xmath53 = -0.5mu=-0.5mu=-0.5mu@xmath54 @xmath55 @xmath56 the variable @xmath51 should be positive .
it can be proved as follows .
when @xmath57 , the variable @xmath58 should be zero for all @xmath27 since the term ( @xmath59 ) is always positive .
then , the term ( @xmath60 ) always has a positive value , and makes the partial derivative @xmath61 positive .
it does not satisfy the condtion .
therefore , @xmath62 and the variable @xmath51 is positive and equal to the term ( @xmath59 ) . @xmath63 @xmath2 , @xmath4 , and @xmath64 @xmath65 : measure the aggregate interference @xmath66 @xmath67 : get @xmath66 from the spectrum sensor compute a probability @xmath68 for @xmath66 using @xmath69 caculate an expectation @xmath70 $ ] + @xmath70\leftarrow \max\!\left[0,\min\!\left\{\!1,\frac{1}{\lambda_s}\left(\frac{\ln\left(1/(1-\tau)\right)}{r_p^2\left(\frac{p_s\beta}{p_p}\right)^{2/\alpha}{\frac{2}{\alpha}}c(\alpha)}-\lambda_p\ !
\right)\!\right\}\!\right]\!$ ] @xmath71}e[p_i^*]\right]$ ] go line 2 the proposition 2 determines only the expectation @xmath70 $ ] of the optimal transmission probability @xmath72 in the network , but it can not find the optimal @xmath72 for each st @xmath27 .
instead , we can combine in proposition 1 and in proposition 2 , in order to make an algorithm finding a suboptimal @xmath73 .
we propose a simple algorithm to find @xmath73 that satisfies the necessary condition .
let @xmath74 be a probability weight factor for st @xmath27 .
the transmission probability @xmath73 is determined as follows : @xmath75}e[p_i^*]\right],\ ] ] here , the cumulative cid @xmath76 determines the factor @xmath74 in the following manner .
@xmath77 where @xmath66 is a measured interference by the sensor adjacent to the st @xmath27 , and @xmath64 is a given interference threshold .
the weight @xmath74 means the probability that the st @xmath27 receives an aggregate interference lower than the threshold @xmath64 . in , the normalized probability @xmath78 $ ] adjusts a chance of spectrum access .
for example , as shown in fig [ simul_topo ] , the sts in sparse environment like area a may have a high weight @xmath74 .
the whole process of the proposed cognitive random access is described in algorithm 1 . ) .
, scaledwidth=26.0% ] ( @xmath79 .
, scaledwidth=27.0% ] we evaluate the proposed random access scheme through 1,000,000 simulations . at every simulation ,
pts and sts are independent and identically distributed according to a homogeneous ppp with intensity @xmath2 and@xmath4 , respectively , in a 100 m @xmath80 100 m area .
spectrum sensor is located at a distance of @xmath0 from the corresponding st .
the communication distances of primary and secondary pairs is 3 m. the distance between the st and its adjacent sensor is set to 1 m. the interference threshold @xmath64 is set to 2 dbm .
the primary and secondary transmission powers are 23 dbm and 5 dbm , respectively .
we set the noise power as -70 dbm and consider rayleigh fading .
[ fig ase_s ] shows ase performance as a function of the st density @xmath4 .
the proposed scheme surpasses the conventional aloha scheme . also , we conducted the comparison with adaptive transmission that deterministically attempts to send data when the measurement of an adjacent sensor is lower than an interference threshold .
the proposed scheme shows better ase performance than the other schemes for high density of st @xmath4 .
this ase difference comes from the probabilistic transmission based on the cid , which gives us a probability that actual received interference at st may be lower than the measurement . with this information , the st attempts to access the spectrum more aggressively , producing higher ase .
also , we observe ase performance of the proposed scheme with respect to the pt density @xmath2 as shown in fig .
[ fig ase_p ] .
when the primary outage probability constraint @xmath81 is small , ase is sensitive to @xmath2 .
this paper focuses on cognitive radio ( cr ) based iot networks where multiple sensors are deployed to monitor interference temperature in the area .
an inherent characteristic of this cr network induces the different spectrum occupancy at the locations of the cr iot devices and the sensors . to compensate this difference
, we derive a conditional interference distribution ( cid ) at the cr device for a given measured interference at the sensor .
we find that the shape of the cid is a left - skewed distribution .
this statistic property implies that an actual received interference at cr device may be lower than the sensor measurement with a considerable probability . reflecting this phenomenon
, we devise a cognitive random access scheme which adaptively adjusts transmission probability with respect to the cid .
our scheme improves area spectral efficiency compared to conventional aloha and threshold based random access protocol , where an iot device can access the medium only when a measured sensor value is lower than the predetermined threshold .
the measured aggregate interference @xmath12 at the sensor can be decomposed as follows : @xmath82 where @xmath83 , @xmath84 is the nearest pt from the sensor , and @xmath14 is the distance between the sensor and pt @xmath84 .
since spectrum sensors measure interference for a enough time duration , the fading effect can be averaged out in the measurement .
we assume that sensors are close to sts enough to consider pt @xmath84 as a common dominant interferer for both the sensor and st @xmath27 .
now then , we investigate how these two sets of interferers have an influence on the st @xmath27 .
first , pt @xmath84 imposes an interference @xmath85 to st @xmath27 as @xmath86 . for a fixed interference @xmath87 ,
the angle @xmath16 determines interference strength @xmath85 . as @xmath16 increases , the interference @xmath85 decreases by @xmath88 . using this geometric property
, we can transit the cumulative cid function @xmath89 to the probability with respect to @xmath16 as follows : @xmath90 here , we use an approximation that interference @xmath91 at st @xmath27 from pts in @xmath92 is equal to mean interference @xmath93=\lambda_p
\int_0^{2\pi}\!\!\!\int_{r_1}^\infty pl(o , r ) rdrd\phi=2p\pi \lambda_p\frac{r_1^{2-\alpha}}{\alpha-2 } $ ] .
let @xmath21 denote the estimated distance from the nearest pt .
for a given @xmath12 equal to @xmath94 in , we can find the estimated distance @xmath21 which is the solution of the following equation : @xmath95 . now
, then we can specify the cumulative cid @xmath96 , of which derivative is the cid function @xmath18 .
@xmath97 where @xmath98 .
|
this paper focuses on cognitive radio ( cr ) internet - of - things ( iot ) networks where spectrum sensors are deployed for iot cr devices , which do not have enough hardware capability to identify an unoccupied spectrum by themselves . in this sensor - enabled iot cr network , the cr devices and
the sensors are separated .
it induces that spectrum occupancies at locations of cr devices and sensors could be different . to handle this difference
, we investigate a conditional interference distribution ( cid ) at the cr device for a given measured interference at the sensor .
we can observe a spatial correlation of the aggregate interference distribution through the cid . reflecting the cid ,
we devise a cognitive random access scheme which adaptively adjusts transmission probability with respect to the interference measurement of the sensor .
our scheme improves area spectral efficiency ( ase ) compared to conventional aloha and an adaptive transmission scheme which attempts to send data when the sensor measurement is lower than an interference threshold . cognitive random access , dynamic spectrum access , adaptive transmission probability , internet - of - things , conditional interference distribution
|
t tau has been extensively studied from radio to ultraviolet wavelengths because it has long been considered the prototype of a class of pre - main sequence stars . in recent years
it has become clear that t tau is in reality a very complex system and that it differs from other t tauri stars .
it is in fact known to be a binary system ( dyck et al .
1982 ) , containing an optical stellar component ( t tau n ) and an infrared companion 0.7 arcsec to the south ( t tau s ) , corresponding to 100 au at the 140 pc distance of the taurus auriga dark cloud ( wichmann et al . 1998 ) .
t tau s dominates the bolometric luminosity of the system ( ghez et al .
1991 ) , but it has no detectable optical counterpart to a limiting magnitude of v=19.6 , suggesting an optical extinction greater than 7 mag ( stapelfeldt et al .
1998 ) .
since herbig s ( 1950 ) optical spectra of burnham s nebula , surrounding t tau several arcsecond across , resembling a so - called hh object , it was clear that the interaction of stellar winds with the surrounding molecular medium is at work .
molecular outflow activity was first mapped by edwards & snell ( 1982 ) in @xmath8co j=1 - 0 and j=2 - 1 , who found that 95% of the high velocity molecular gas is associated with blueshifted material .
the direction of the detected outflow is roughly parallel to the line of sight , but the emission also shows a region extending 2 arcmin to the south and east of t tau n with a secondary peak in the blueshifted wing .
higher resolution maps of the @xmath8co j=3 - 2 , j=6 - 5 and c@xmath9o j=1 - 0 , j=2 - 1 and hco@xmath10 emission later showed ( schuster et al .
1993 , momose et al .
1996 , schuster et al .
1997 , hogerheijde et al .
1998 ) a complex outflow system that could originate from the different components of the binary system .
the fast stellar winds observed through forbidden optical line emission ( bhm & solf 1994 ) , revealed five distinct kinematic components that suggest that both the primary star and the companion may drive separate bipolar outflows . a giant herbig - haro flow was recently discovered ( reipurth et al .
1997 ) around t tau and is interpreted as originating several thousand years ago from t tau s. strong and extended @xmath11 ro - vibrational emission was found quite early around t tau ( beckwith et al .
recently , high resolution @xmath11 imaging ( herbst et al . 1996 , 1997 ) indicated that the extended molecular hydrogen emission arises from the impact on the ambient cloud of two outflow systems oriented nw - se and e - w .
these originate from the two stars , each with its circumstellar disk , and the emission is distributed equally over t tau n and t tau s. infrared adaptive optics observations in @xmath11 show instead that the emission is concentrated on t tau s and is interpreted in terms of shocks occurring as matter accretes onto the circumstellar disk of t tau s ( quirrenbach & zinnecker 1997 ) .
no firm conclusion is therefore reached on this problem .
t tau has associated a substantial amount of mass of dust , it is therefore luminous in the millimeter continuum ( adams et al .
1990 , beckwith et al . 1990 ) .
a circumstellar disk has been detected both with co interferometry and infrared scattered light ( weintraub et al .
1989 , momose et al .
later , millimeter continuum interferometry at 0.9 and 3 mm ( hogerheijde et al .
1997 , akeson et al .
1998 , respectively ) was used to derive a total mass of 0.04 @xmath12 for the circumstellar disk around t tau n and at least 10 times smaller for that associated to t tau s. a circumbinary envelope would also be required to fit the continuum energy distribution .
far - infrared spectroscopy provides powerful diagnostic lines from abundant molecular species like co , h@xmath0o and oh , that can be used to clarify the physical processes at work in the complex t tau system . in this paper we present the far - infrared spectrum observed from the long wavelength spectrometer ( lws , clegg et al .
1996 ) onboard the infrared space observatory ( iso , kessler et al .
additional data from the short wavelength spectrometer ( sws , de graauw et al .
1996 ) are also used for discussing the molecular emission properties .
the main results of the sws are presented by van den ancker et al.(1999 ) .
the continuum far - infrared spectrum of t tau will be discussed in a forthcoming paper ( pezzuto et al . in preparation ) .
t tau has been observed with iso ( infrared space observatory ) using the lws ( long wavelength spectrometer , clegg et al .
1996 ) . a full low resolution
( r @xmath2 200 ) spectrum of the source from 45 to 197 @xmath1 m was obtained during revolution 680 , corresponding to september , 25 , 1997 .
the beamsize is on average 80 arcsec , depending on the wavelength .
the spectrum was made up of 23 full grating scans oversampled at 1/4 of a resolution element , with each spectral sample integrated for 11.5 sec , with a total integration time of 4265 sec . besides the observation on - source , full grating spectra
were also collected at four off - source positions . in table 1
we present the journal of the lws observations , which includes source and off - source positions and total observing time ( ott ) .
the raw data were reduced and calibrated using version 7 of the lws pipeline , which achieves an absolute accuracy of about 30 % ( swinyard et al .
post - pipeline processing was carried out with the isap package and included removal of spurious signals due to cosmic ray impacts and averaging the grating scans of each detector .
besides the data measured by the lws , we also discuss in this paper the detection of h@xmath0o and oh emission lines observed by the sws .
the details of these observations are reported by van den ancker et al.(1999 ) .
.journal of the iso - lws observations of t tau [ cols="<,^,^,^ " , ] notes : @xmath13 : wavelength was fixed for deblending .
the detection of [ cii]158@xmath1 m in the four off - source positions around t tau ( see table 5 ) clearly shows that most of the ionized carbon emission ( @xmath14 ) is from an extended region and not originated in the vicinity of t tau .
the intrinsic emission in the lws beam centered on t tau is about 2 @xmath15 10@xmath16 w @xmath17 .
this implies a ratio [ oi]63@xmath1m/[cii]158@xmath1 m of 115 , greatly in excess of that expected from photodissociation region models ( kaufman et al.1999 ; burton et al .
1990 ) .
finally the ratio of [ oi]63@xmath1m/145@xmath1m=28.3 is such that no oxygen self - absorption should occur , as it often appears to be the case towards pre - main sequence sources ( saraceno et al .
this suggests that there is no cold gas in front of the source , in accordance with the geometry of the outflow directed towards the observer .
as outlined in van den ancker et al .
( 1999 ) , based on a larger set of fine - structure lines detected at shorter wavelengths , we argue that the atomic line emission observed is consistent with the presence of j - type dissociative shocks . on the other hand , the excess [ cii]158@xmath1 m emission on - source could also be due to a local photodissociation region ( pdr ) , possibly originated from the far - uv field of t tau ( see sect .
4 ) . table 6 summarises the physical quantities derived from the observed molecular spectra of co , h@xmath0o and oh , adopting the two models considered : the column densities and the total cooling luminosities are given for each molecular species .
we also give the observed values of the cooling derived from the sum of all the detected fluxes . for deriving l@xmath18
, we used the line fluxes reported in van den ancker et al .
( 1999 ) and the line @xmath11 1 - 0 s(1 ) at 2.12@xmath1 m , given in carr ( 1990 ) .
a comparison between the observed and modeled cooling shows that the observations of co , h@xmath0o and oh can account for most of the modeled cooling .
the underestimate of the oh cooling by the 900 k model confirms that this latter is probably inadeguate to explain the observed oh emission .
the total radiated cooling observed from these species , including [ oi ] , sums up to about 0.04 l@xmath7 , and has to be considered as a lower limit .
taking the outflow parameters from the literature ( levreault 1988 ; mundt 1984 ) we derive a total mechanical luminosity of about 0.05 l@xmath7 ( we have taken an average velocity of 10 @xmath19 , a total outflow mass of 0.22 m@xmath7 , and a dynamical timescale of 40,000 years ) . the radiative luminosity observed in the
far - infrared is therefore comparable to the outflow mechanical luminosity .
this is expected if the stellar winds from the stars are driving the outflows and the shocks , traced by the far - infrared lines , accelerate the ambient medium into the molecular outflow ( davis & eislffel 1995 ) .
lccc + & observed@xmath20 & `` lower t '' & `` higher t '' + + temperature t(k ) & & 300 & 900 + density @xmath21 & & 4@xmath1510@xmath22 & 2@xmath1510@xmath23 + n@xmath24 ( @xmath17 ) & & @xmath25 & @xmath26 + n@xmath27 ( @xmath17 ) & & @xmath28 & @xmath29 + n@xmath30 ( @xmath17 ) & & 4@xmath1510@xmath31 & @xmath32 + l@xmath24 ( l@xmath7 ) & 0.0067 & 0.0096 & 0.010 + l@xmath27
( l@xmath7 ) & 0.008 & 0.014 & 0.009 + l@xmath30 ( l@xmath7 ) & 0.012 & 0.019 & 0.009 + l@xmath33}$ ] ( l@xmath7 ) & 0.014 & & + l@xmath18 ( l@xmath7 ) & 0.005 & & + + notes : @xmath20 : the observed cooling is computed by summing all the detected lines .
once the physical conditions of the emitting gas in the vicinity of t tau have been established , we can now proceed to compare the results with existing models of shock excitation .
the far - infrared line emission of young stellar objects is mainly originated from two physical processes : the excitation from photoionized and photodissociated ( pdr ) regions ( tielens & hollenbach 1985 ) and the shock excitation produced by the interaction of supersonic winds with the ambient medium .
depending on wind velocity , magnetic field and ion density , two kinds of shocks with different far - infrared spectra are predicted from models : * high velocity dissociative j shocks ( e.g. hollenbach & mckee 1989 ) , in which temperature , density and velocity have a discontinuous jump ( j ) on the shock front , molecules are dissociated and atomic lines are the dominant coolants ; * low velocity non - dissociative c shocks ( e.g. kaufman & neufeld 1996 , draine et al . 1983 ) in which the ion alfvn velocity is larger than the shock velocity and the magnetic field transmits energy faster than the shock velocity ; in this case temperature , density and velocity have a continuous ( c ) variation and molecules are the dominant coolants . instead of using intensities of many tens of different lines , we can obtain a better comparison between our data and the shock model predictions using the total cooling from a single species .
in fig.7 we show the water cooling as a function of the [ oi]63@xmath1 m cooling , both normalized to the high - j co cooling .
the c - type shock models of kaufman & neufeld ( 1996 ) are considered .
the j - type shock models ( hollenbach & mckee 1989 ) are also indicated ( see figure caption for details ) . together with the position of t tau , we also show in this figure the positions of l1448 ( nisini et al .
1999b ) , ic1396 ( saraceno et al .
1999 ) , iras16293 - 2422 ( ceccarelli et al .
1998 ) , and the herbig - haro objects hh54 ( liseau et al . 1996 ) , hh25 and hh26 ( benedettini et al . 1998 ) .
t tau is in a central position , showing that both c - type and j - type shock models could explain the observations .
it has to be noted , however , that according to kaufman & neufeld models , its position implies a shock velocity between 10 and @xmath34 , in a range where water production is triggered but it is not at its maximum efficiency .
we can see in the figure that other pre - main sequence sources also cluster in the same region of t tau , indicating that these shock conditions are fairly common in the environment of young stellar objects ( nisini et al .
1998 ) .
8 shows the water cooling as a function of the oh cooling , both normalized to the high - j co cooling . as before
, we consider both j - type and c - type shocks .
the position of t tau in this plot appears to be consistent with j - type shocks .
the main reason why c - shocks fail to reproduce the observed cooling is the overabundance of oh molecules , by at least a factor 10 .
only a model which includes the effects of the presence of a high uv field from t tau itself can reproduce the observed values . fig .
9 shows the water cooling as a function of the @xmath11 cooling , both normalized to the high - j co cooling .
t tau lies exactly in the center of the region of c - type shocks and can not be explained by j - type shocks .
as expected , the strong h@xmath0 emission can not be accounted for by j - type shocks .
we therefore rule out the hypothesis that the major responsible of the observed excitation are j - shocks .
because photodissociation of water by an uv field , that is not included in shock models , can convert water to oh , we suppose that the overabundance of the oh molecule is due only to the strong far - uv radiation field associated to t tau ( herbig & goodrich , 1986 ) . the photodissociation cross sect . of water at the @xmath35 frequency is in fact ten times larger than the one of oh ( van dishoeck & dalgarno 1984 ) . a similar situation has been found in supernova remnants , where oh 1720 mhz emission is explained as originated from c - type shocks , allowing that the action of an uv field creates sufficient oh from water dissociation ( wardle et al .
1998 ; lockett et al .
1999 ) .
if this is the case , the far - infrared molecular emission spectrum of t tau is primarily due to c - type shocks . to reconcile the oh observations with c - type shock models we need that the oh abundance , and thus its total cooling , were reduced by a factor 10 in favor of water cooling . from table 6
, if we reduce by a factor 10 the oh column density and we increase of the corresponding amount that one of the water , passing from @xmath28 to @xmath36 , we will increase the total water cooling by a factor of about 3 .
increasing the water cooling by this factor and decreasing the oh cooling by a factor 10 would move the position of t tau in the three diagrams of fig.7 , 8 and 9 in positions fully consistent with c - type shock models ( see the open triangles in the figures ) , at pre - shock densities of about 10@xmath37 @xmath38 and shock velocities of @xmath39 . comparing the pre - shock densities of these models with the densities derived from our lvg models , we found that a moderate compression factor ( @xmath40 10 ) would be required . in support of the fact that the molecular emission is due to c - type shocks
is the evidence of the presence of strong magnetic fields in the outflow region associated with t tau s ( ray et al . 1997 ) .
our conclusion that the observed far - infrared molecular emission from t tau can be explained by c - type shocks and that the atomic emission is probably originated in j - type shocks is in agreement with the findings of van den ancker et al .
their models , based on the near to mid - infrared @xmath11 emission imply two temperature components at 440 k and 1500 k , which again are roughly in agreement with our finding of two components : one ranging from 300 k to 900 k and another at an higher temperature traced by the higher @xmath41 co transitions . as to the origin of the c - type shocks , responsible for the observed far - infrared molecular emission , we know from the lvg models that the emission region has a size of only few hundreds au , assuming a spherical geometry .
this implies that the shocks occur in a very compact region , presumably very close to the binary system .
there are at least three mechanisms not mutually exclusive to explain the origin of the shocks : 1 . from the interaction region of winds coming from the two stars ; 2 . from disk accretion on the youngest component of the binary system ; 3 . from the interaction of the stellar wind with the molecular material in the circumstellar envelope .
the first possibility seems the best one for originating fast ( @xmath42 ) dissociative j - type shocks , because the wind interaction would occur close to the stars where velocities are supposed to be high and the stellar field strong to dissociate molecules .
the second possibility has already been suggested by quirrenbach & zinnecker ( 1997 ) to explain the near - infrared @xmath11 extended emission .
the third one is probably at work in any case , because it is needed to explain the strong emission from co , h@xmath0o , oh , as well as that from @xmath11 ( van den ancker et al .
to summarise our results , we list the main findings of this study : 1 .
the far - infrared spectrum associated to the binary system of t tau shows strong emission from co , h@xmath0o and oh molecules . 2 .
optically thin co emission lines from high - j transitions are used to constrain the physical regimes of the gas : t@xmath2300 - 900 k and n@xmath310@xmath4@xmath5 .
the detection of co lines with @xmath41 of 24 and 25 seems to indicate that a warmer component is also needed .
h@xmath0o and oh emission is consistent with such conditions ,
however the higher excitation lines at the shorter wavelenghs are not well fitted by these models , indicating that higher temperature gas should also be present , in agreement with the co emission .
4 . from the assumption that all the far - infrared molecular emission observed originate from the same region a very compact size of 300 - 400 au of diameter is derived . 5 .
the detection of [ cii]158@xmath1 m off - source at large distances from t tau shows that most of the ionised carbon emission ( @xmath14 ) is from an extended region and not originated in the vicinity of t tau .
the intrinsic emission in the lws beam centered on t tau implies a ratio [ oi]63@xmath1m/[cii]158@xmath1 m of 115 , much in excess of what is expected from photodissociation region models .
this emission is probably due to j - type shocks , however it is not ruled out the possibility of a contamination by a pdr .
the total cooling observed in the various species is comparable to the mechanical luminosity of the outflow , indicating that the stellar winds could be the ultimate responsible of the line excitation through shocks .
the study of the cooling from the main radiative components ( [ oi ] , co , h@xmath0o , oh and @xmath11 ) is used to show that our data are consistent with c - type shock models , only if we assume an overabundance of the oh molecule ( by a factor 10 ) probably due to the intense far - ultraviolet radiation field .
we finally consider the origin of the shocks responsible for the far - infrared emission : colliding winds from the components of the binary system , disk accretion in the more embedded star and wind interaction with the molecular ambient cloud . due to the large beam size of the lws instrument , we are unable to specify which of the processes is the dominant one .
we wish to thank the lws consortium and the iso staff at vilspa ( esa , villafranca , spain ) for having operated the lws instrument and the iso satellite .
we thank riccardo cesaroni for running his oh code to evaluate the effects of a local radiation field .
mvda acknowledges financial support from nwo grant 614.41.003 and through a nwo _ pionier _ grant to l.b.f.m .
waters .
adams f.c . , emerson j.p . , fuller g.a .
1990 , apj 357 , 606 .
akeson r.l .
, koerner d.w . , jensen e.l.n . 1998 , apj 505 , 358 .
beckwith s.v.w .
, gatley i. , matthews k. , neugebauer , g. 1978 , apj 223 , l41 .
beckwith s.v.w .
, sargent a. , chini r. guesten , r. 1990 , aj 99 , 924 .
benedettini m. , giannini t. , nisini b. et al .
1998 , proc . of the universe as seen by iso , paris , 20 - 23 october 1998 , in press .
bergin e.a . ,
melnick g.j . ,
neufeld d.a .
1998 apj 499 , 777 .
bhm k .- h .
, solf j. 1994 , apj 430 , 277 .
burton m.g , hollenbach d.j . ,
tielens a.g.g .
1990 , apj 365 , 620 .
carr j.s . , 1990 , aj 100 , 1244 .
ceccarelli c. , caux e. , white g.j .
1998 , a&a 331 , l17 .
cesaroni r. , wamsley c.m .
, 1991 , a&a 241 , 537 .
chackerian c. , tipping r.h .
1983 , j. of molecular spectroscopy 99,431 clegg p.e . , ade p.a.r .
, armand c. et al .
1996 , a&a 315 , l38 .
cohen m. , hollenbach d.j . ,
haas m.r . , erickson e.f .
1988 , apj 329 , 863 .
davis c.j . , eislffel j. 1995 , a&a 300 , 851 . de graauw t. , haser l.n .
, beintema d.a .
1996 , a&a 315 , l49 .
draine b.t .
, roberge w.g . and dalgarno a. 1983 , apj 264 , 485 .
dyck h. m. , simon t. , zuckerman b. 1982 , apj 255 , 103 .
edwards s. , snell r.l .
1982 , apj 261 , 151 .
ghez a.m. , neugebauer g. , gorham p.w .
1991 , aj 102 , 2066 .
giannini t. , lorenzetti d. , tommasi e. et al .
1999 , a&a 346 , 617 . gonzlez - alfonso e. , cernicharo j. , van dishoeck e. , wright c.m . , heras , a. 1998 , apj 502 , l169 .
green s. , maluendes s. , mclean a.d .
1993 , apjs 85,181 harwit m. , neufeld d.a . ,
melnick g.j .
, kaufman m.j .
1998 , apj 497 , l105 .
herbig g.h . , 1950 ,
apj 111 , 11 .
herbig g.h . ,
goodrich r.w .
1986 , apj 309 , 294 .
herbst t.m . ,
beckwith s.v.w .
, glindemann a. et al .
1996 , aj 111 , 2403 .
herbst t.m . , robberto m , and beckwith s.v.w .
1997 , aj 114 , 774 .
hogerheijde m.r .
, van langevelde h.j . ,
mundy l.g . ,
blake g.a . ,
van dishoeck , e.f .
1997 , apj 490 , l99 .
hogerheijde m.r . ,
van dishoeck e.v . , blake g.a . ,
van langevelde h.j .
1998 , apj 502 , 315 .
hollenbach d. , mckee c.f .
1989 , apj 342,306 .
kaufman m.j .
, neufeld d.a .
1996 , apj 456 , 250 .
kaufman m.j .
, wolfire m.g . ,
hollenbach d.j . ,
luhman m.l . 1999 ,
apj in press .
kessler m.f .
, steinz j.a .
, anderegg m.e .
et al . , 1996 ,
a&a 315 , l27 kobayashi n. , nagata t. , hodapp k. , hora , j.l .
1994 , pasj 46 , l183 .
levreault r.m .
1988 , apj 330 , 897 .
liseau r. , ceccarelli c. , larsson b. et al .
1996 , a&a 315 , l181 .
lockett p. , gauthier e. , elitzur m. , 1999 , apj 511 , 235 .
mckee c.f .
, storey j.w.v . , watson d.m .
, green s. 1982 apj 259 , 647 .
momose m. , ohashi n. , kawabe r. hayashi m. , nakano t. 1996 , apj 470 , 1001 .
mundt r. 1984 , apj 280 , 749 .
nisini b. , giannini , t. , molinari , s. et al .
1998 in star formation with the infrared space observatory , yun j.l . &
liseau r. ( eds . ) , asp conference series , vol .
132 , p.233 nisini b. , benedettini , m. , giannini , t. et al .
1999a , a&a , 343 , 266 nisini b. , benedettini , m. , giannini , t. et al .
1999b , a&a , in press offer a.r . , van dishoeck e.f .
1992 , mnras 257 , 377 .
quirrenbach a. , zinnecker h. , 1997 , eso messenger 87 , 36 .
ray t.p . ,
muxlow t.w.b .
, axon d.j .
1997 , nat 385 , 415 .
reipurth b. , bally j. , and levine d. 1997 , aj 114,2708 .
rothman l.s . ,
gamache r.r .
, goldman a. et al .
1987 , appl.opt . , 26 , 4078 .
schuster k .- f . ,
harris a.i . ,
anderson n.l . ,
russel a.p.g .
1993 , apj 412 , l67 .
schuster k .- f . ,
harris a.i . ,
russell a.p.g .
1997 , a&a 321 , 568 .
saraceno p. , nisini b. , benedettini m. , et al .
, 1998 , in star formation with the infrared space observatory , yun j.l .
, liseau r. ( eds . ) , asp conference series , vol . 132 , p.233 stapelfeldt k.r . , burrows c.j .
, krist j.e .
1998 , apj 508 , 736 .
swinyard b.m . ,
burgdorf m.j .
, clegg p.e .
1998 , proc .
spie , vol.3354 , p.888 , infrared astronomical instrumentation , albert m. fowler ( ed . ) , 1998 .
tielens a.g.g.m .
, hollenbach d. , 1985 , apj 291 , 722 .
van den ancker m.e .
, wesselius p.r .
, tielens a.g.g.m . , van dishoeck e.f .
, spinoglio , l. 1999 , a&a in press .
van dishoeck e.f .
, dalgarno a. , 1984 , icarus 59,305 .
wardle m. , yusef - zadeh f. , geballe t.r .
1998 , the central parsecs , galactic center workshop 1998 , tucson , sep .
7 - 11 , 1998 : to be published in the asp conf .
weintraub d.a . , zuckerman b. , masson , c.r .
1989 , apj 344 , 915 .
wichmann r. , bastian u. , krautter j. , jankovics i. , rucinski s. m. 1998 , mnras 301 , l39 .
|
the first complete far - infrared spectrum of t tau has been obtained with the lws spectrometer on - board the infrared space observatory ( iso ) , which detected strong emission from high - j ( j=14 - 25 ) co , para- and ortho - h@xmath0o and oh transitions over the wavelength range from 40 to 190 @xmath1 m .
in addition the [ oi]63@xmath1 m , [ oi]145@xmath1 m and [ cii]158@xmath1 m atomic lines were also detected . most of the observed molecular emission can be explained by a single emission region at t@xmath2300 - 900 k and n@xmath310@xmath4@xmath5 , with a diameter of about 2 - 3 arcsec .
this corresponds to a very compact region of 300 - 400 au at the distance of 140 pc .
an higher temperature component seems to be needed to explain the highest excitation co and h@xmath0o lines .
we derive a water abundance of 1 - 7 @xmath6 and an oh abundance of @xmath2 3@xmath6 with respect to molecular hydrogen , implying h@xmath0o and oh enhancements by more than a factor of 10 with respect to the expected ambient gas abundance .
the observed cooling in the various species amounts to 0.04 l@xmath7 , comparable to the mechanical luminosity of the outflow , indicating that the stellar winds could be responsible of the line excitation through shocks . in order to explain the observed molecular cooling in t tau in terms of c - type shock models , we hypothesise that the strong far - ultraviolet radiation field photodissociates water in favour of oh .
this would explain the large overabundance of oh observed .
the estimated relatively high density and compactness of the observed emission suggest that it originates from the shocks taking place at the base of the molecular outflow emission , in the region where the action of the stellar winds from the two stars of the binary system is important .
|
entropy is a crucial concept in thermal physics . a solid understanding of what entropy is and how it works is the key to understanding a broad range of physical phenomena , and enhances one s understanding of an even wider range of phenomena .
the idea of entropy and its ( nearly ) inevitable increase are thus core concepts that we should hope physics students at all levels would study carefully and learn to apply successfully.@xcite unfortunately , the approach to teaching entropy found in virtually all introductory and many upper - level textbooks is more abstract than it needs to be .
such texts@xcite introduce entropy in the context of macroscopic thermodynamics , using heat engines and carnot cycles to motivate the discussion , ultimately defining entropy to be the quantity that increases by @xmath1 when energy @xmath2 is transferred by heating to a system at temperature @xmath3 . in spite of the way that this approach mirrors the historical development of the idea of entropy
, our students have usually found the intricate logic and abstraction involved in getting to this definition too subtle to give them any intuitive sense of what entropy is and what it measures ; the concept ultimately remains for them simply a formula to memorize.@xcite most introductory texts attempt to deal with this problem by including a section about how entropy quantifies `` disorder '' . the accuracy and utility of this idea , however , depends on exactly how students visualize `` disorder '' , and in many situations this idea is unhelpful or even actively misleading.@xcite the macroscopic approach therefore generally does _ not _ help students develop the kind of accurate and useful intuitive understanding of this very important concept that we would like them to have .
the _ statistical _ interpretation of entropy ( as the logarithm of the the number of quantum microstates consistent with a system s macrostate ) is , comparatively , quite concrete and straightforward .
the problem here is that it is not trivial to show from this definition that the entropy of a macroscopic system of interacting objects will inevitably increase.@xcite this paper describes our attempt to link the statistical definition of entropy to the second law of thermodynamics in a way that is as accessible to students as possible.@xcite our primary motivation was to create a more contemporary approach to entropy to serve as the heart of one of the six modules for _ six ideas that shaped physics _ , a new calculus - based introductory course developed with the support and assistance of the introductory university physics project ( iupp ) .
however , we have found this approach to be equally useful in upper - level undergraduate courses .
the key to our approach is the use of a _
computer _ to count the numbers of microstates associated with various macrostates of a simple model system .
while the processes involved in the calculation are simple and easy to understand , the computer makes it practical to do the calculation for systems large enough to display the irreversible behavior of macroscopic objects .
this vividly displays the link between the statistical definition of entropy and the second law without using any sophisticated mathematics .
our approach has proved to be very helpful to upper - level students and is the basis of one of the most successful sections of the _ six ideas _
introductory course.@xcite depending on the level of the course and the time available , the computer - based approach can be used either as the primary tool or as a springboard to the more powerful analytic methods presented in statistical mechanics texts .
a number of choices are available for the specific computational environment : a standard spreadsheet program such as microsoft _ excel _ or borland _ quattropro _ ; a stand - alone program specially written for the purpose ; or even a graphing calculator such as the ti-82 .
we will use spreadsheets in our examples in the main body of this paper , and briefly describe the other alternatives in section vi .
after some preliminary definitions in section ii , we present the core of our approach , the consideration of two interacting `` einstein solids '' , in section iii . in the remaining sections we derive the relation between entropy , heat , and temperature , then go on to describe a number of other applications of spreadsheet calculations in statistical mechanics .
the aim of statistical mechanics is to explain the behavior of systems comprised of a very large number of particles .
we can characterize the physical _ state _ of such a system in either of two ways , depending on whether we focus on its characteristics at the macroscopic or molecular level .
the _ macrostate _ of a system is its state as viewed at a macroscopic level .
for instance , to describe the macrostate of a tank of helium gas , we could specify the number of molecules @xmath4 it contains , the total volume @xmath5 that they occupy , and their total internal energy @xmath6 . as long as the gas is in internal thermal equilibrium
, these three parameters suffice to determine its macrostate .
( other thermodynamic variables , such as the temperature @xmath3 and the pressure @xmath7 of the gas , can then be calculated using various `` equations of state '' such as the ideal gas law @xmath8 and the relation @xmath9 , where @xmath10 is boltzmann s constant . ) in general , the macrostate of a system is determined by the various conservation laws and physical constraints that it is subject to .
the _ microstate _ of a system is its state as viewed at the molecular level . to describe the microstate of our tank of helium gas
, we would have to specify the exact position and velocity ( or quantum mechanical wavefunction ) of each individual molecule .
note that if we know a system s microstate , we also know its macrostate : the total energy @xmath6 , for instance , is just the sum of the energies of all the molecules . on the other hand ,
knowing a system s macrostate does not mean we re anywhere close to knowing its microstate . the number of microstates that are consistent with a given macrostate is called the _ multiplicity _ of that macrostate , denoted @xmath11 .
equivalently , @xmath11 is the number of _ accessible _ microstates , limited only by the constraints that determine the macrostate . for our tank of helium gas
, the multiplicity would be a function of three macroscopic variables , for instance , @xmath12 . in most cases
the multiplicity will be a _ very _ large number , since there will be an enormous number of different ways to distribute a given amount of energy among the system s molecules ( and , for fluids , to distribute the molecules within the available volume ) .
the _ fundamental assumption _ of statistical mechanics is that an isolated system in a given macrostate is equally likely to be in any of its accessible microstates .
this means that the probability of finding the system in any given microstate is @xmath13 , since there are @xmath11 microstates , all equally probable . in the next section
we will see how this assumption leads to the conclusion that macroscopic objects must exhibit _ irreversible _ behavior , such as the spontaneous flow of energy from a hot object to a cold one . in order to make this argument concrete , though
, we will focus our attention on a specific model for a thermodynamic system , since the exact way that a macrostate s multiplicity depends on @xmath4 , @xmath6 , and @xmath5 depends on what physical system we are considering . once we understand how the argument works for our specific model , we will be able to understand ( at least qualitatively ) how it would apply to other physical systems .
there are only a few realistic systems whose multiplicities can be calculated using elementary methods .
even an ideal gas is too complicated in this regard a direct derivation of its multiplicity function requires some fairly sophisticated tools.@xcite a much simpler system would be one whose particles are all fixed in place ( so the volume is irrelevant ) , storing energy in units that are all the same size .
this idealized model is treated in many statistical mechanics texts.@xcite physically , it corresponds to a collection of identical quantum harmonic oscillators . in 1907 , einstein proposed such a model to describe the thermal properties of a simple crystalline solid , treating each atom as an independent three - dimensional harmonic oscillator .
we will therefore call this model an _ einstein solid_. we will use @xmath4 to denote the number of one - dimensional _ oscillators _ in an einstein solid .
( so the number of atoms is @xmath14 . ) if we neglect the static zero - point energy ( which plays no role in the thermal behavior ) , then the energy of any oscillator can be written @xmath15 where @xmath16 is a nonnegative integer and @xmath17 is a constant ( @xmath18 , where @xmath19 is the oscillator s natural frequency ) . in other words ,
each oscillator can store any number of units of energy , all of the same size .
the total energy of the entire solid can thus be written @xmath20 where @xmath21 is another nonnegative integer .
the multiplicity of an einstein solid is some function of @xmath4 and @xmath21 : the number of ways of arranging @xmath21 units of energy among @xmath4 oscillators . for a very small einstein solid one can simply _ count _ all the ways of arranging the energy among the oscillators . for example , if @xmath22 , the following 10 distinct distributions are possible ( where the first number in each triplet specifies the number of energy units in the first oscillator , the second the number of energy units in the second oscillator , etc . ) : 300 , 030 , 003 , 210 , 120 , 201 , 102 , 021 , 012 , 111 .
thus the multiplicity of this macrostate is @xmath23 . by directly counting microstates in this manner
, one can not only vividly illustrate the concepts of macrostates and microstates to first - year students , but also show how the multiplicity increases very sharply with both @xmath4 and @xmath21 , without using any difficult mathematics .
the general formula for the multiplicity of an einstein solid is @xmath24 an elegant derivation of this formula can be found in the textbook of callen.@xcite in brief , the trick is to represent any microstate as a sequence of dots ( representing energy units ) and lines ( representing partitions between one oscillator and the next ) .
a total of @xmath21 dots and @xmath25 lines are required , for a total of @xmath26 symbols .
the number of possible arrangements is just the number of ways of choosing @xmath21 of these symbols to be dots .
whether or not one derives formula ( [ esolidmult ] ) in class , students have little trouble believing it after they have verified a few simple cases by brute - force counting . in an introductory course , this is all we usually have them do .
to understand irreversible processes and the second law , we now move on to consider a system of _ two _ einstein solids , @xmath27 and @xmath28 , which are free to exchange energy with each other but isolated from the rest of the universe .
first we should be clear about what is meant by the `` macrostate '' of such a composite system . for simplicity
, we assume that our two solids are _ weakly coupled _ , so that the exchange of energy between them is much slower than the exchange of energy among atoms within each solid .
then the individual energies of the two solids , @xmath29 and @xmath30 , will change only slowly ; over sufficiently short time scales they are essentially fixed .
we will use the word `` macrostate '' to refer to the state of the combined system , as specified by the ( temporarily ) constrained values of @xmath29 and @xmath30 . for any such macrostate we can compute the multiplicity , as we shall soon see .
however , on longer time scales the values of @xmath29 and @xmath30 _ will _ change , so we will also talk about the _
multiplicity of the combined system , considering all microstates to be `` accessible '' with only @xmath31 held fixed . in any case
, we must also specify the number of oscillators in each solid , @xmath32 and @xmath33 , in order to fully describe the macrostate of the system .
let us start with a very small system , in which each of the `` solids '' contains only three oscillators and the solids share a total of six units of energy:@xcite @xmath34 given these parameters , we must still specify the individual value of @xmath35 or @xmath36 to describe the macrostate of the system
. there are seven possible macrostates , with @xmath37 , 1 , @xmath38 6 . for each of these macrostates
, we can compute the multiplicity of the combined system as the product of the multiplicities of the individual systems : @xmath39 this works because the systems are independent of each other : for each of the @xmath40 microstates accessible to solid @xmath27 , there are @xmath41 microstates accessible to solid @xmath28 .
a table of the various possible values of @xmath40 , @xmath41 , and @xmath42 is shown in fig . 1 .
students can easily construct this table for themselves , computing the multiplicities by brute - force counting if necessary . from the table
we see that the fourth macrostate , in which each solid has three units of energy , has the highest multiplicity ( 100 ) .
now let us invoke the fundamental assumption : on long time scales , _ all _ of the 462 microstates are accessible to the system , and therefore , by assumption , are equally probable .
the probability of finding the system in any given _ macro_state is therefore proportional to that macrostate s multiplicity .
for instance , in this example the fourth macrostate is the most probable , and we would expect to find the system in this macrostate 100/462 of the time .
the macrostates in which one solid or the other has all six units of energy are less probable , but not overwhelmingly so .
another way to construct the table of fig .
1 is to use a computer spreadsheet program .
both microsoft _ excel _ and borland _ quattropro _ include built - in functions for computing binomial coefficients : @xmath43 the formulas from the _ excel _ spreadsheet that produced fig . 1
are shown in fig . 2 .
the advantage in using a computer is that it is now easy to change the parameters to consider larger systems .
changing the numbers of oscillators in the two solids can be accomplished by merely typing new numbers in cells b2 and d2 . changing the number of energy units requires also adding more lines to the spreadsheet .
the spreadsheet becomes rather cumbersome with much more than a few hundred energy units , while overflow errors can occur if there are more than a few thousand oscillators in each solid .
fortunately , most of the important qualitative features of thermally interacting systems can be observed long before these limits are reached .
figure 3 shows a spreadsheet - generated table , and a plot of @xmath42 vs. @xmath35 , for the parameters @xmath44 we see immediately that the multiplicities are now quite large . but more importantly
, the multiplicities of the various macrostates now differ by _
factors _ that can be very large
more than @xmath45 in the most extreme case .
if we again make the fundamental assumption , this result implies that the most likely macrostates are @xmath45 times more probable than the least likely macrostates . if this system is initially in one of the less probable macrostates , and we then put the two solids in thermal contact and wait a while , we are almost _ certain _ to find it fairly near the most probable macrostate of @xmath46 when we check again later . in other words ,
this system will exhibit irreversible behavior , with net energy flowing spontaneously from one solid to the other , until it is near the most likely macrostate .
this conclusion is just a statement of the second law of thermodynamics .
there is one important limitation in working with systems of only a few hundred particles . according to the graph of fig . 3 ,
there are actually quite a few macrostates with multiplicities nearly as large as that of the most likely macrostate . in relative terms , however , the width of this peak will decrease as the numbers of oscillators and energy units increase .
this is already apparent in comparing figs . 1 and 3 .
if the numbers were increased further , we would eventually find that the relative width of the peak ( compared to the entire horizontal scale of the graph ) is completely negligible .
unfortunately , one can not verify this statement directly with the spreadsheets .
students in an introductory class have little trouble believing this result , when reminded of the fact that macroscopic systems contain on order of @xmath47 particles and energy units .
in an upper - division course , one can prove using stirling s approximation that the relative width of the peak is of order @xmath48 or @xmath49 , whichever is larger .
thus , statistical fluctuations away from the most likely macrostate will be only about one part in @xmath50 for macroscopic systems.@xcite using this result , we can state the second law more strongly : aside from fluctuations that are normally much too small to measure , any isolated macroscopic system will inevitably evolve toward its _ most _ likely accessible macrostate .
although we have proved this result only for a system of two weakly coupled einstein solids , similar results apply to all other macroscopic systems .
the details of the counting become more complex when the energy units can come in different sizes , but the essential properties of the combinatorics of large numbers do not depend on these details . if the subsystems are not weakly coupled , the division of the whole system into subsystems becomes a mere mental exercise with no real physical basis ; nevertheless , the same counting arguments apply , and the system will still exhibit irreversible behavior if it is initially in a macrostate that is in some way `` inhomogeneous '' .
instead of always working with multiplicities , it is convenient to define the _ entropy _ of a system as @xmath51 the factor of boltzmann s constant is purely conventional , but the logarithm is important : it means that , for any given macrostate , the total entropy of two weakly interacting@xcite systems is the _ sum _ of their individual entropies : @xmath52 since entropy is a strictly increasing function of multiplicity , we can restate the second law as follows : aside from fluctuations that are normally much too small to measure , any isolated macroscopic system will inevitably evolve toward whatever ( accessible ) macrostate has the largest entropy .
it is a simple matter to insert columns for @xmath53 , @xmath54 , and @xmath55 into the spreadsheet in fig .
a graph of these three quantities is shown in fig . 4 .
notice that @xmath55 is _ not _ a strongly peaked function .
nevertheless , the analysis of the previous section tells us that , when this system is in equilibrium , it will almost certainly be found in a macrostate near @xmath56 . , @xmath54 , and @xmath55 for the system of two einstein solids represented in fig .
3 . at equilibrium ,
the slopes of the individual entropy graphs are equal in magnitude .
away from equilibrium , the solid whose slope @xmath57 is _ steeper _ tends to gain energy , while the solid whose slope @xmath57 is _ shallower _ tends to lose energy.,width=302 ] now let us ask how entropy is related to temperature .
the most fundamental way to define temperature is in terms of energy flow and thermal equilibrium : two objects in thermal contact are said to be at the same temperature if they are in thermal equilibrium , that is , if there is no spontaneous net flow of energy between them . if energy _ does _ flow spontaneously from one to the other
, then we say the one that loses energy has the higher temperature while the one that gains energy has the lower temperature .
referring again to fig .
4 , it is not hard to identify a quantity that is the same for solids @xmath27 and @xmath28 when they are in equilibrium .
equilibrium occurs when @xmath55 reaches a maximum , that is , when the slope @xmath58 . but
this slope is the sum of the slopes of the individual entropy graphs , so their slopes must be equal in magnitude : @xmath59 if we were to plot @xmath54 vs. @xmath36 instead of @xmath35 , its slope at the equilbrium point would be equal , in sign and magnitude , to that of @xmath60 .
so the quantity that is the same for both solids when they are in equilibrium is the slope @xmath57 .
temperature must be some function of this quantity.@xcite to identify the precise relation between temperature and the slope of the entropy vs. energy graph , look now to a point in fig .
4 where @xmath35 is larger than its equilibrium value . here
the entropy graph for solid @xmath28 is steeper than that for solid @xmath27 , meaning that if a bit of energy were to pass from @xmath27 to @xmath28 , solid @xmath28 would gain more entropy than solid @xmath27 loses .
since the _ total _ entropy would increase , the second law tells us that this process will happen spontaneously . in general , the steeper an object s entropy vs. energy graph , the more it `` wants '' to gain energy ( in order to obey the second law ) , while the shallower an object s entropy vs. energy graph , the less it `` minds '' losing a bit of energy .
we therefore conclude that temperature is _ inversely _ related to the slope @xmath61 .
in fact , the reciprocal @xmath62 has precisely the units of temperature , so we might guess simply @xmath63 the factor of boltzmann s constant in the definition ( [ sdef ] ) of entropy eliminates the need for any further constants in this formula . to confirm that this relation gives temperature in ordinary kelvin units
one must check a particular example , as we will do in the following section .
our logical development of the concepts of entropy and temperature from statistical considerations is now complete . in the following section
, we will apply these ideas to the calculation of heat capacities and other thermal properties of some simple systems .
however , at this point in the development , many instructors will wish to first generalize the discussion to include systems ( such as gases ) whose volume can change .
the derivatives in eqs .
( [ slopesequal ] ) and ( [ tempdef ] ) then become partial derivatives at fixed volume , and one can not simply rearrange eq .
( [ tempdef ] ) to obtain @xmath64 .
there are at least two ways to get to the correct relation between entropy and temperature , and we now outline these briefly . the standard approach in statistical mechanics texts@xcite is to first consider two interacting systems separated by a movable partition , invoking the second law to identify pressure by the relation @xmath65 . combining this relation with eq .
( [ tempdef ] ) one obtains the thermodynamic identity @xmath66 . for quasistatic processes ( in which the change in volume is sufficiently gradual ) , the second term is the ( infinitesimal ) work @xmath67 done on the system.@xcite but the first law says that for any process , @xmath68 , where @xmath2 is the ( infinitesimal ) energy added by heating .
therefore , for quasistatic processes only , we can identify @xmath69 or @xmath70 from this relation one can go on to treat a variety of applications , most notably engines and refrigerators . in an introductory course , however , there may not be time for the general discussion outlined in the previous paragraph .
fortunately , if one does not need the thermodynamic identity , some of the steps can be bypassed ( at some cost in rigor ) .
we have already seen that , when the volume of a system does not change , its entropy changes by @xmath71 now consider a system whose volume does change , but that is otherwise isolated from its environment . if the change in volume is quasistatic ( i.e. , gradual ) , then the entropy of the system can not change .
theoretically , because while the energies of the system s quantum states ( and of the particles in those states ) will be shifted , the process is not violent enough to knock a particle from one quantum state to another ; thus the multiplicity of the system remains fixed .
alternatively , we know from experiment that by reversing the direction of an adiabatic , quasistatic volume change we can return the system to a state arbitrarily close to its initial state ; if its entropy had increased , this would not be possible even in the quasistatic limit . by either argument ,
the work done on a gas as it is compressed should not be included in the change in energy @xmath72 that contributes to the change in entropy .
the natural generalization of eq .
( [ dsatconstv ] ) to non - constant - volume quasistatic processes is therefore to replace @xmath72 with @xmath73 , the energy added to the system by heating _
alone_. thus we again arrive at eq .
( [ sqrelation ] ) .
so far in this paper we have used spreadsheet calculations only to gain a qualitative understanding of the second law and the relation between entropy and temperature . now
that we have developed these concepts , however , we can use the same tools to make quantitative predictions about the thermal behavior of various systems .
figure 5 shows a spreadsheet calculation of the entropy of a _ single _ einstein solid of @xmath74 oscillators with anywhere from 0 to 100 units of energy .
then , in the fourth column , we have calculated the temperature @xmath75 using a pair of closely spaced entropy and energy values .
we have actually used a centered - difference approximation , so the cell d4 , for example , contains the formula @xmath76 .
since we have omitted the constants @xmath10 and @xmath17 from the formulas , the temperatures are expressed in units of @xmath77 .
we now know the temperature for every energy , or vice versa . in practice , however , one usually compares such predictions to experimental measurements of the _ heat capacity _ , defined ( for systems at constant volume ) as @xmath78 this quantity , normalized to the number of oscillators , is computed in the fifth column of the table , again using a centered - difference approximation ( for instance , cell e5 contains the formula @xmath79 ) .
since we are still neglecting dimensionful constants , the entries in the column are actually for the quantity @xmath80 . to the right of the table
is a spreadsheet - generated graph of the heat capacity vs. temperature for this system .
this theoretical prediction can now be compared to experimental data , or to the graphs often given in textbooks.@xcite the quantitative agreement with experiment verifies that the formula @xmath81 is probably correct.@xcite at high temperatures we find the constant value @xmath82 , as expected from the equipartition theorem ( with each oscillator contributing two degrees of freedom ) . at temperatures below @xmath77 , however , the heat capacity drops off sharply .
this is just the qualitative behavior that einstein was trying to explain with his model in 1907 . to investigate the behavior of this system at lower temperatures
, one can simply increase the number of oscillators in cell e1 . with 45000 oscillators
the temperature goes down to about @xmath83 , corresponding to a heat capacity per oscillator of about @xmath84 . at this point
it is apparent that the curve becomes _ extremely _ flat at very low temperatures .
experimental measurements on real solids confirm that the heat capacity goes to zero at low temperature , but do not confirm the detailed shape of this curve .
a more accurate treatment of the heat capacities of solids requires more sophisticated tools , but is given in standard textbooks.@xcite a collection of identical harmonic oscillators is actually a more accurate model of a completely different system : the vibrational degrees of freedom of an ideal diatomic gas . here
again , experiments confirm that the equipartition theorem is satisfied at high temperatures but the vibration ( and its associated heat capacity ) `` freezes out '' when @xmath85 .
graphs showing this behavior are included in many textbooks.@xcite perhaps the most useful tool of statistical mechanics is the _ boltzmann factor _ , @xmath86 , which gives the relative probability of a system being in a microstate with energy @xmath87 , when it is in thermal contact with a much larger system ( or `` reservoir '' ) at temperature @xmath3
. a fairly straightforward derivation of this result can be found in most textbooks on statistical mechanics.@xcite a less general derivation , based on the exponentially decreasing density of an isothermal atmosphere , is given in serway s introductory text.@xcite although it is no substitute for an analytic treatment , a spreadsheet calculation can easily display this exponential dependence for the case of a single one - dimensional oscillator in thermal contact with a large einstein solid.@xcite figure 6 shows such a calculation , for the parameters @xmath88 , @xmath89 , and @xmath90 . here
solid @xmath27 is the `` system '' , while solid @xmath28 , which is much larger , is the `` reservoir '' .
the graph shows the natural logarithm of the total multiplicity , plotted as a function of @xmath35 .
according to the fundamental assumption , the total multiplicity is proportional to the probability of finding the small system in each given microstate .
note that the graph is a straight line with a negative slope : this means that the probability does indeed decay exponentially ( at least approximately ) as the energy @xmath91 of the small system increases . according to the boltzmann factor
, the slope of the graph of multiplicity ( or probability ) vs. @xmath92 should equal @xmath93 .
since the small system always has unit multiplicity and zero entropy , the entropy of the combined system is the same as the entropy of the reservoir : @xmath94 .
once they know this , students can check either numerically or analytically that the slope of the graph is @xmath95 where the derivatives and the temperature @xmath96 are all evaluated at @xmath97 .
we have used the einstein solid model in our examples for several reasons : it is fairly realistic , its behavior is intuitive , and , most importantly , it is mathematically simple .
other systems generally lack at least one of these qualities . to carry out our spreadsheet calculations , however , all we require is mathematical simplicity .
the simplest possible system would consist of `` atoms '' that have only two allowed energy levels .
a common example is a paramagnetic material , with one unpaired electron per atom , immersed in a uniform magnetic field @xmath28 .
let us say that @xmath28 points in the @xmath98 direction , and quantize the spin ( hence the magnetic moment ) of the atom along this axis .
when the magnetic moment points up , the atom is in the ground state , while when the magnetic moment points down , the atom is in the excited state .
the energy levels of these two states are @xmath99 where @xmath100 is the magnitude of the atom s magnetic moment . in a collection of @xmath4 such atoms , the total energy will be @xmath101 where @xmath102 and @xmath103 are the numbers of atoms in the up and down states .
these numbers must add up to @xmath4 : @xmath104 the total magnetization of the sample is given by @xmath105 to describe a macrostate of this system we would have to specify the values of @xmath4 and @xmath6 , or equivalently , @xmath4 and @xmath102 ( or @xmath4 and @xmath103 ) . the multiplicity of any such macrostate is the number of ways of choosing @xmath102 atoms from a collection of @xmath4 : @xmath106 from this simple formula we can calculate the entropy , temperature , and heat capacity .
figure 7 shows a spreadsheet calculation of these quantities for a two - state paramagnet of 100 `` atoms '' .
the rows of the table are given in order of increasing energy , which means decreasing @xmath102 and decreasing magnetization . alongside the table
is a graph of entropy vs. energy .
since the multiplicity of this system is largest when @xmath102 and @xmath103 are equal , the entropy also reaches a maximum at this point ( @xmath107 ) and then falls off as more energy is added .
this behavior is unusual and counter - intuitive , but is a straightforward consequence of the fact that each `` atom '' can store only a limited amount of energy . to appreciate the consequences of this behavior ,
imagine that the system starts out with @xmath108 and that we gradually add more energy .
at first it behaves like an einstein solid : the slope of its entropy vs. energy graph decreases , so its tendency to suck in energy ( in order to maximize its entropy in accord with the second law ) decreases . in other words ,
the system becomes hotter .
as @xmath6 approaches zero , the system s desire to acquire more energy disappears .
its temperature is now infinite , meaning that it will willingly give up a bit of energy to any other system whose temperature is finite ( since it loses no entropy in the process ) .
when @xmath6 increases further and becomes positive , our system actually _ wants _ to give up energy .
intuitively , we would say that its temperature is higher than infinity .
but this is mathematical nonsense , so it is better to think about the reciprocal of the temperature , @xmath109 , which is now lower than zero , in other words , negative . in these terms
, our intuition even agrees with the precise relation ( [ tempdef ] ) .
the paramagnetic system can be an invaluable pedagogical tool , since it forces students to abandon the naive notion that temperature is merely a measure of the average energy in a system , and to embrace instead the more general relation ( [ tempdef ] ) , with all its connotations .
one can even imagine using the paramagnet in place of the einstein solid as the main paradigm for the lessons of sections iii and iv.@xcite in particular , it is not hard to set up a spreadsheet model of two interacting paramagnets , to look at how energy can be shared between them . however , this system exhibits behavior not commonly seen in other systems or in students prior experience .
we therefore prefer not to use it as a first example , but rather to save it for an upper - level course in which students have plenty of time to retrain their intuition .
putting these fundamental issues aside , we can go on to calculate both the heat capacity and the magnetization of our two - state paramagnet as a function of temperature .
plots of these quantities are shown in fig .
since the heat capacity goes to zero at high temperature , the equipartition theorem fails utterly to describe this system ( as it must , since it applies only to degrees of freedom for which the energy is a quadratic function of a coordinate or momentum ) .
the magnetization is complete at @xmath3 just greater than zero , but drops off as the temperature is increased .
the behavior in the high - temperature limit is known as curie s law , as described in many textbooks.@xcite
the ideas presented in this paper have been tested , in various forms , in calculus - based introductory courses at pomona college , grinnell college , smith college , amherst college , and st .
lawrence university , and in an upper - division thermodynamics course at weber state university .
we have found that students at either level are quite able to understand the fundamental concepts of statistical mechanics that we are trying to teach . in particular , the fact that the statistical approach to the second law is more `` modern '' does not , in our opinion , make it unsuitable for first - year students . on the contrary
, we have had far more success in teaching the statistical approach than in teaching the concept of entropy from the older classical viewpoint used in most texts .
the use of computers in our approach does not seem to pose a barrier to today s students .
many of them have had previous experience with spreadsheet programs , and the rest are able to learn very quickly with the help of a one - page instruction sheet that we have prepared.@xcite we generally present all of the material of section iii in class , and hand out reproductions of the figures from that section . for homework , we ask students to reproduce fig . 1 , then change the parameters to look at some larger systems .
here is a typical homework assignment , written for students who will be using _ excel _ : 1 .
list all the possible microstates for an einstein solid containing three oscillators and five units of energy .
verify that @xmath110 .
2 . set up an _ excel _ spreadsheet identical ( initially ) to the one shown on the handouts : two einstein solids , each containing three harmonic oscillators , with a total of six units of energy .
you ll need to type in all the labels and formulas down to row 4 , as well as cell a5 .
after that , you can use the `` fill down '' command to fill in the rest ( except for cell e11 ) . use the `` chart wizard tool '' ( the second icon from the right ) to create the graph . check that your spreadsheet yields all the same results as on the handout . then modify your spreadsheet to show the case where one einstein solid contains six harmonic oscillators and the other contains four harmonic oscillators ( with the total number of energy units still equal to six ) . turn in a printout of the modified spreadsheet .
assuming that all microstates are equally likely , what is the most probable macrostate , and what is its probability ? what is the least probable macrostate , and what is its probability ?
modify your spreadsheet from the previous problem to show the case where solid @xmath27 contains 100 oscillators , solid @xmath28 contains 200 oscillators , and there are 100 units of energy in total .
it s easiest to delete the old graph and make a new one , and this time it s best to make it an `` xy ( scatter ) '' graph instead of a column graph .
turn in a printout of the first several rows of the spreadsheet as well as the graph .
what is the most probable macrostate , and what is its probability ?
what is the least probable macrostate , and what is its probability ?
calculate the entropies of these two macrostates ( expressed as unitless numbers , neglecting the factor of boltzmann s constant ) .
4 . starting with your spreadsheet from the previous problem , add columns for the entropy of solid @xmath27 , the entropy of solid @xmath28 , and the total entropy .
print a graph showing all three entropies , plotted vs. the energy of solid @xmath27 .
use the formula @xmath111 to calculate the temperatures of both solids ( in units of @xmath77 ) at the equilbrium point and also at the point @xmath56 .
do your results make sense ?
explain .
we have also given computer - based homework assignments on the material of section v. in upper - division courses we have asked students to generate fig . 5 before seeing it in class .
the treatment of this material can vary greatly depending on how the class is taught , however , so we will not reproduce our homework exercises here .
spreadsheet programs provide an almost ideal computational environment for the calculations described in this paper : they are versatile , easy to learn , and readily available on most campuses .
other computing options may be more suitable in some contexts , however . before we discovered spreadsheets , one of us ( t.a.m . ) wrote a stand - alone program for the macintosh to produce tables and graphs such as those shown in figs . 1 and 3 .
the stand - alone program is even easier to use than a spreadsheet , and can conveniently handle somewhat larger systems and values of @xmath21 .
it also allows one to completely omit formula ( [ esolidmult ] ) from the course if desired
students can simply be asked to verify the computer s computations in a few simple cases .
this program is sufficient to teach the core of our approach , as outlined in section iii .
a copy of the program is available on request.@xcite a much cheaper , but less convenient , option is to perform the calculations on a modern graphing calculator .
we do not particularly recommend this ; in our experience , graphing calculators are _ much _ harder to use than spreadsheet programs
. however , a motivated student who already knows how to use the calculator should be able to reproduce all the major results in this paper without too much difficulty . as an example
, we describe here an appropriate set of instructions for the ti-82.@xcite to set up tables like those shown in section iii , one can use the `` table '' feature of the ti-82 , which displays seven rows at a time , calculating each row as it is displayed .
the first column of the table , labeled x , should be used for @xmath35 .
the following formulas in the `` y= '' screen will then generate the remaining four columns : here we are using variables a , b , and q to hold the constants @xmath32 , @xmath33 , and @xmath116 , respectively . before plotting a graph , it is best to set xmin to zero and xmax to 94 ; this insures that each pixel corresponds to an integer value of x. to calculate temperatures and heat capacities on the ti-82 , one must use `` lists '' instead of `` tables '' .
the following sequence of instructions will reproduce most of fig . 5 , storing @xmath21 , @xmath11 , @xmath117 , @xmath118 , and @xmath119 in lists l@xmath120 through l@xmath121 : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ @xmath122 + @xmath123 + @xmath124 + @xmath125 + @xmath126 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ again we assume that the number of oscillators is stored in variable a. lists 4 and 5 will be offset upward by one and two places respectively , but this can be fixed by entering `` stat - edit '' mode and inserting zeros .
one should also add zeros at the end to make all lists the same length .
it is then a simple matter to plot the heat capacity vs. the temperature , as shown in fig .
5 . unfortunately
, the lists will not update automatically when the value of a is changed .
for instance , david halliday , robert resnick , and jearl walker , _ fundamentals of physics _ ( wiley , new york , 1993 ) , 4th ed . ;
raymond a. serway , _ physics for scientists and engineers _ ( saunders college publishing , philadelphia , 1996 ) , 4th ed . ; francis w. sears and gerhard l. salinger , _
thermodynamics , kinetic theory , and statistical thermodynamics _ ( addison - wesley , reading , ma , 1975 ) , 3rd ed .
rudolf clausius himself remarked that
the second law of thermodynamics is much harder for the mind to grasp than the first , " as quoted in abraham pais , _ subtle is the lord _ ( oxford university press , oxford , 1982 ) , p. 60 .
for example , a deep understanding of entropy and the behavior of self - organizing systems makes it clear that biological life is fully consistent with ( and is indeed an expression of ) the increase of entropy , even though life apparently involves an _ increase _ of order in the universe .
similarly , it is not obvious that a disorderly pile of ice chips has less entropy than a simple and seemingly orderly cup of the same amount of water .
boltzmann himself had a great deal of trouble convincing the scientific community of the validity of this statistical approach to entropy .
see pais , ref . , pp . 6068 , 8283 .
arthur beiser comments on page 315 in his _ concepts of modern physics _ text ( 4th ed . , mcgraw - hill , new york , 1987 ) that `` battles with nonbelievers left him severely depressed '' and contributed to his eventual suicide in 1906 , just as experimental evidence in favor of the atomic hypothesis and boltzmann s approach to thermal physics began to become overwhelming . for another attempt to address the same issue , see ralph baierlein , `` entropy and the second law : a pedagogical alternative , '' am .
j. phys . *
62*(1 ) , 1526 ( 1994 ) .
baierlein s approach is less quantitative than ours , but full of physical insight .
a complete text for a three - week introductory - level unit on thermal physics based on this approach is available . [
update : this text is now published as _
six ideas that shaped physics : unit t _ , 2nd edition ( mcgraw - hill , 2009 ) . ] a partial derivation can be found in keith stowe , _ introduction to statistical mechanics and thermodynamics _
( wiley , new york , 1984 ) , pp .
a very nice indirect approach to the multiplicity of an ideal gas has been given by david k. nartonis , `` using the ideal gas to illustrate the connection between the clausius and boltzmann definitions of entropy , '' am .
j. phys . *
44*(10 ) , 10081009 ( 1976 ) . see , for example , henry a. bent , _ the second law _ ( oxford university press , new york , 1965 ) , pp .
144162 ; herbert b. callen , _ thermodynamics and an introduction to thermostatistics _
( wiley , new york , 1985 ) , 2nd ed . , p. 334 ; w. g. v. rosser , _ an introduction to statistical physics _
( ellis horwood limited , chichester , 1982 ) , pp .
3141 ; charles a. whitney , _
random processes in physical systems _
( wiley , new york , 1990 ) , pp . 97139 .
an even simpler model is a two - state system , as treated , for example , in charles kittel and herbert kroemer , _ thermal physics _
( freeman , san francisco , 1980 ) , 2nd ed . , pp .
we treat this model in section v.c below , and also discuss there why we do nt use it as our primary example .
a nice calculation of the width of the peak for a simpler system has been given by ralph baierlein , `` teaching the approach to thermodynamic equilibrium : some pictures that help , '' am .
j. phys . *
46*(10 ) , 10421045 ( 1978 ) . the definition of entropy , like that of macrostate or multiplicity , depends on the time scale under consideration .
here we are interested in time scales that are long compared to the relaxation time of each individual system , but short compared to the relaxation time of the composite system . for a strongly coupled system this clean separation of time scales would not be possible .
note that on very long time scales , even our weakly coupled system would have a slightly larger entropy , equal to @xmath127 where @xmath11 is the total multiplicity for _ all _ allowed values of @xmath29 and @xmath30 . for macroscopic systems
this number is practically indistinguishable from the quantity @xmath128 computed for the most likely macrostate .
a careful discussion of the role of time scales in defining entropy is given in l. d. landau and e. m. lifshitz , _ statistical physics _ ,
j. b. sykes and m. j. kearsley ( pergamon , oxford , 1980 ) , 3rd edition , part 1 , p. 27 .
notice that the preceding argument depends crucially on the additivity of @xmath129 for the composite system .
thus the logarithm in the definition of @xmath130 is essential , if @xmath130 and @xmath3 are to be related in a simple way .
the standard notation is @xmath131 , but it is too tempting to read this incorrectly as the `` change in work '' ( a meaningless phrase ) , rather than correctly as an infinitesimal `` amount of work '' .
we therefore call it simply @xmath67 , with the understanding that in this context it is an infinitesimal quantity
. similarly , we use @xmath2 instead of @xmath132 .
note also that we take @xmath67 to be positive when work is done _ on _ the system .
most textbooks define the temperature scale operationally in terms of a gas thermometer .
the most direct verification of @xmath81 would therefore be to apply this formula directly to an ideal gas .
but the gas thermometer can be used to measure the heat capacity of a solid , and this result , in turn , can be compared to our theoretical prediction .
note , in particular , that the factor of boltzmann s constant in the definition of @xmath130 conveniently makes the temperature come out in ordinary kelvin units , with no further constants needed .
the two - state model system is used to introduce statistical concepts in the popular text of kittel and kroemer , _ op .
cit._. this text includes a nice calculation of the sharpness of the multiplicity function for a system of two interacting paramagnets .
this instruction sheet , which is written for users of _ excel _ on the macintosh , is available on request from daniel schroeder .
[ update : the instructions are now available at < http://physics.weber.edu/thermal/computer.html>[<http://physics.weber.edu/thermal/computer.html > ] . ]
please send requests to thomas moore .
[ update : versions of this computer program for mac osx , windows , and linux are now available at < http://www.physics.pomona.edu/sixideas/sicpr.html>[<http://www.physics.pomona.edu/sixideas/sicpr.html > ] , and a web - based version is at < http://physapps.pomona.edu/>[<http://physapps.pomona.edu/ > ] . ]
|
the basic notions of statistical mechanics ( microstates , multiplicities ) are quite simple , but understanding how the second law arises from these ideas requires working with cumbersomely large numbers . to avoid getting bogged down in mathematics
, one can compute multiplicities numerically for a simple model system such as an einstein solid a collection of identical quantum harmonic oscillators . a computer spreadsheet program or comparable software
can compute the required combinatoric functions for systems containing a few hundred oscillators and units of energy .
when two such systems can exchange energy , one immediately sees that some configurations are overwhelmingly more probable than others .
graphs of entropy vs. energy for the two systems can be used to motivate the theoretical definition of temperature , @xmath0 thus bridging the gap between the classical and statistical approaches to entropy .
further spreadsheet exercises can be used to compute the heat capacity of an einstein solid , study the boltzmann distribution , and explore the properties of a two - state paramagnetic system .
_ 1997 american association of physics teachers . _ published in am .
j. phys .
* 65 * ( 1 ) , 2636 ( 1997 ) , < http://dx.doi.org/10.1119/1.18490>[<http://dx.doi.org/10.1119/1.18490 > ] .
|
-coupled low - density parity - check ( sc - ldpc ) codes have attracted many attentions in recent years due to their ability to achieve the shannon limit of binary memoryless symmetric ( bms ) channels under iterative decoding @xcite .
the research on the sc - ldpc codes was started with the advent of convolutional ldpc codes by felstrom and zigangirov @xcite .
there have been lots of research results on construction and analysis of the convolutional ldpc codes @xcite@xcite . especially , sridharan _ et al . _
@xcite derived the belief propagation ( bp ) threshold which is a theoretical performance limit of the iterative decoding for the convolutional ldpc codes and lentmaier _ et al . _
@xcite empirically showed that the bp thresholds of convolutional ldpc codes are close to the map thresholds of the underlying regular block ldpc codes .
this remarkable improvement of the bp thresholds was named as _ threshold saturation _ which was analytically proved in @xcite , @xcite .
recently , proving the threshold saturation effect has been done through many different approaches such as a potential function @xcite , @xcite and an one - dimensional continuous coupled system @xcite .
these approaches utilize the unique decoding process of sc - ldpc codes ; the reliable messages are propagated from the both ends to the center on the tanner graph in a _ wave - like manner_. sc - ldpc codes are constructed by coupling @xmath0 disjoint ldpc codes with the boundary condition which causes the wave - like propagation . however , at the cost of the wave - like propagation , the boundary condition reduces the overall code rate compared with that of underlying codes , where the rate loss converges to zero at a speed @xmath1 .
thus , we can obtain a capacity approaching code ensemble for large chain length because the rate loss is vanished while the wave - like decoding process is still successful even for extremely large chain lengths , provided that the iteration number increases with the chain length . in a practical situation , however
, sc - ldpc codes with too large @xmath0 imply a large number of iterations for the wave - like propagation in the bp decoding @xcite .
furthermore , an in - depth analysis regarding a scaling law of sc - ldpc ensembles shows that large @xmath0 is not good for the finite - length performance @xcite .
on the other hand , sc - ldpc ensembles with short @xmath0 have good decoding properties such as high bp thresholds , low required numbers of iterations for the wave - like propagation , and good finite - length performances but it show large rate loss . some modifications of sc - ldpc ensembles have been studied to achieve performance improvement in practical situations . among them , we focus on the modification by properly connecting multiple sc - ldpc ensembles .
the idea of coupling is not limited to connection of several ldpc ensembles into a chain ( an sc - ldpc ensemble ) but also extended to connection of several chains to overcome some problems that a single chain can not solve .
for example , the multi - dimensional sc - ldpc codes @xcite were considered for their robustness against burst erasures and the parallelly connected sc - ldpc chains @xcite were proposed to obtain asymptotically good code ensembles with wide rate range .
especially , in @xcite@xcite , a loop ensemble obtained by connecting two sc - ldpc chains was proposed with improved bp thresholds and finite - length performances . at the connecting points in the loop ensemble ,
consecutive positions consisting of high degree variable nodes , so called protected region , are created and reliable messages are emitted from the protected region . due to these reliable messages ,
the loop ensemble is decoded as two independent sc - ldpc chains with short chain length and achieves performance improvements .
however , in @xcite , it is shown that the loop ensemble shows worse finite - length performance than the sc - ldpc ensemble for long chain length .
the proposed coupled ensemble , called overlapped circular sc - ldpc ( oc - ldpc ) ensemble , is obtained based on the idea to connect multiple chains , but it is different from the loop ensemble in the sense that the oc - ldpc ensemble is constructed from two circular sc - ldpc ensembles rather than two sc - ldpc ensembles . since there is no reliable position initiating decoding waves in the symmetric structure of the circular sc - ldpc ensemble @xcite , the performance improvement by coupling does not appear in the circular sc - ldpc ensemble . to make a protected region from which decoding waves start to propagate in the circular ensemble ,
our proposed approach is overlapping some variable nodes in two circular sc - ldpc ensembles , which introduces high - degree variable nodes . since high - degree variable nodes in ldpc codes
are successfully decoded with high probability , a protected region is formed by the introduced high - degree variable nodes like the loop ensemble .
similar to the loop ensemble , the oc - ldpc ensemble behaves as two _ split _ short sc - ldpc chains which are separately decoded with improved decoding performance and small number of iterations in the decoding process . since the performance improvement of the loop and oc - ldpc ensembles comes from almost the same effect , these ensembles look similar .
however , in terms of degree distribution of the protected region , we will show that the protected region of the oc - ldpc ensemble is more protected compared to the protected region of the loop ensemble , and thus the oc - ldpc ensemble behaves as two split short sc - ldpc chains regardless of the chain length unlike the loop ensemble .
contrary to the randomly defined ensemble , codes from protograph - based ensembles @xcite are well suited for designing practical codes with several advantages such as good finite - length performance and an implementation - friendly structure .
also , protograph - based sc - ldpc ensembles have good minimum distance properties as well as capacity approaching bp thresholds @xcite .
thus , we design finite - length oc - ldpc codes based on protograph and numerical results show that the proposed codes provide better finite - length performance than the conventional sc - ldpc codes for both short and long chain lengths .
furthermore , it is also verified that a construction method called precoding @xcite is applicable to the proposed ensemble to further improve the performance .
the remainder of the paper is organized as follows .
section [ sec : construction ] gives construction methods of coupled ensembles such as sc - ldpc , loop , and oc - ldpc ensembles . in section [ sec : analysis ] , the behavior of the oc - ldpc ensemble in the decoding process is analyzed by the density evolution ( de ) and the expected graph evolution . in section
[ sec : protograph ] , the protograph - based ensembles for the proposed coupled ensembles are presented .
finally , the conclusion is given in section [ sec : conclusion ] .
let @xmath2 and @xmath3 denote variable and check node degrees of regular ldpc ensembles , respectively .
we consider the sc - ldpc ensemble in @xcite denoted by the @xmath4 ensemble , where @xmath0 and @xmath5 are called the chain length and the coupling length , respectively .
the @xmath4 ensemble consists of @xmath6 variable nodes of degree @xmath2 located at each @xmath0 variable node positions with @xmath7 check nodes of degree @xmath3 located at each @xmath8 check node positions .
each of the @xmath2 edges of a variable node at position @xmath9 is uniformly at random and independently connected to check nodes at the positions @xmath10 for @xmath11 . in @xcite , the connectivity between the positions in the graph is represented by a @xmath12 connectivity matrix @xmath13 where @xmath14 and @xmath15 are the number of check and variable node positions , respectively .
@xmath16_{u , v}$ ] equals to @xmath17 if a variable node at position @xmath18 is connected to some check nodes at position @xmath19 .
then the connectivity matrix @xmath20 of the @xmath4 ensemble is represented as @xmath21_{u , v}=\begin{cases } 1 \textrm { , if $ u\in [ v , v+w-1]$ } & \\ 0 \textrm { , otherwise } \end{cases}\label{eq : t_sc}\end{aligned}\ ] ] and @xmath22 , @xmath23 .
the total number of variable nodes @xmath24 and check nodes @xmath25 in the graph are @xmath26 and @xmath27 , respectively .
however , some check nodes at the leftmost @xmath28 and rightmost @xmath28 boundary positions can not be connected to any variable node in the ensemble by the randomness of the connectivity .
the expected number of check nodes @xmath29 with at least one connection to the variable nodes in the graph is @xcite @xmath30 which is less than @xmath31 .
thus , the design rate @xmath32 of the @xmath33@xmath34 ensemble is given as ( * ? ? ?
* lemma 3 ) @xmath35 where the first term is the design rate of the underlying regular ldpc ensemble , the second term represents the rate loss by coupling , and the third term represents the rate gain by the unconnected check nodes .
since the third term is small in comparison with the second term , the design rate of the sc - ldpc ensemble is reduced from the first term , but it converges to the first term as @xmath0 increases .
let @xmath36 denote the loop ensemble @xcite@xcite .
as in @xcite@xcite , we mainly consider the loop ensemble for the trivial case when @xmath37 and @xmath38 .
the @xmath39 ensemble is constructed by connecting two independent @xmath40 ensembles .
the connectivity matrix of the @xmath39 ensemble is @xmath41 { \bf l}_1 & { \bf t}_{\rm sc}(l,3 ) \end{array } \right]\ ] ] where the @xmath42 matrix @xmath43 has ones at the entries @xmath44 , @xmath45 , and @xmath46 and the @xmath42 matrix @xmath47 has ones at the entries @xmath48 , @xmath49 , and @xmath50 @xcite .
then , the size of @xmath51 becomes @xmath52 which is two times the size of @xmath53 .
the number of variable nodes @xmath24 and check nodes @xmath25 in the graph are @xmath54 and @xmath55 , respectively . since all check nodes in the @xmath39 ensemble are connected to the variable nodes contrary to the @xmath4 ensemble , the design rate of @xmath56 of the @xmath39 ensemble is given as @xmath57 before introducing construction method of oc - ldpc ensembles , we first consider the circular sc - ldpc ensemble @xmath58 presented in @xcite .
let @xmath59 and @xmath60 $ ] be a set of integers between @xmath9 and @xmath61 for integers @xmath9 , @xmath61 . in the @xmath58 ensemble , both of variable and
check nodes are placed at the positions in @xmath62 $ ] with @xmath6 variable nodes and @xmath7 check nodes at each position .
then , each of @xmath2 edges of a variable node at position @xmath9 is uniformly at random and independently connected to check nodes at the positions @xmath63 for @xmath64 $ ] .
the connectivity matrix of the @xmath58 ensemble becomes @xmath65_{u , v}=\begin{cases } 1 \textrm { , if $ u\in \{\left < v , l+w-1\right>,\ldots,\left < v+w-1,l+w-1\right>\}$ } & \\ 0 \textrm { , otherwise}. \end{cases}\label{eq : t_c}\end{aligned}\ ] ] let @xmath66 denote the oc - ldpc ensemble . to construct the @xmath66 ensemble ,
we first construct two independent @xmath67 ensembles for positive integer @xmath68 such that @xmath69 .
next , divide the variable node positions of each @xmath67 ensembles into two parts ; the first part consists of variable node positions from @xmath17 to @xmath68 and the second part consists of the remaining variable node positions .
then , we can also separate the connectivity matrix @xmath70 into two parts such as @xmath71\label{eq : t_c}\ ] ] where the matrix @xmath72 given in ( [ eq : t_sc ] ) corresponds to the first part and the @xmath73 matrix @xmath74 corresponds to the second part . then , connect two circular sc - ldpc ensembles by overlapping each pair of variable nodes located in the second part of each circular sc - ldpc ensembles .
more precisely , connect the check nodes , which are connected with the @xmath75th variable node at position @xmath9 in the second @xmath67 ensemble , to the @xmath75th variable node at position @xmath9 in the first @xmath67 ensemble for @xmath76 $ ] and @xmath77 $ ] while removing the last @xmath78 variable node positions in the second @xmath67 ensemble .
then @xmath79 variable nodes of degree @xmath80 are generated at the positions from @xmath81 to @xmath82 in the first @xmath67 ensemble .
the order of the positions is rearranged from the positions of the first circular @xmath67 ensemble to the positions of the second @xmath67 ensemble . by following the above procedures , the connectivity matrix @xmath83 of the @xmath66 ensemble is obtained as @xmath84 & { \bf t}_{\rm c , r}(l_s , w ) & { \bf t}_{\rm sc}(l_s , w ) \end{array } \right].\label{eq : t_oc}\ ] ] for example , @xmath85 is represented as @xmath86 $ 0 $ & $ 0 $ & $ 0 $ & $ 1 $ & $ 1 $ & $ 1 $ & $ 0 $ & $ 0 $ \\ $ 0 $ & $ 0 $ & $ 0 $ & $ 0 $ & $ 1 $ & $ 1 $ & $ 1 $ & $ 0 $ \\ $ 0 $ & $ 0 $ & $ 0 $ & $ 0 $ & $ 0 $ & $ 1 $ & $ 1 $ & $ 1 $ \\ $ 0 $ & $ 0 $ & $ 0 $ & $ 1 $ & $ 0 $ & $ 0 $ & $ 1 $ & $ 1 $ \\ $ 0 $ & $ 0 $ & $ 0 $ & $ 1 $ & $ 1 $ & $ 0 $ & $ 0 $ & $ 1 $ \end{tabular}\right].\ ] ] clearly
, the size of @xmath83 is @xmath87 which is equal to the size of @xmath20 in ( [ eq : t_sc ] ) .
since the numbers of variable and check nodes are @xmath26 and @xmath88 , the design rate of @xmath66 ensemble is given as @xmath89 as an example , we include the graphical representations of @xmath90 , @xmath91 , and @xmath85 in fig .
[ fig : graph ] . compared to the design rate of the @xmath4 ensemble in ( [ eq : sc_design_rate ] ) ,
the design rate of the @xmath66 ensemble in ( [ eq : oc_design_rate ] ) is slightly lower by the rate gain from the unconnected check nodes for given @xmath0 . on the other hand ,
the design rates of the @xmath39 and @xmath66 ensembles are the exactly the same for the trivial case @xmath92 . in terms of degree of variable node at each position ,
the three coupled ensembles ( sc - ldpc , loop , oc - ldpc ) have different degree distributions .
let @xmath93 be a degree distribution vector of an ensemble @xmath94 where @xmath95_{i}$ ] is the degree of variable node at position @xmath9 . for the @xmath4 ensemble ,
all variable nodes in the ensemble have @xmath2 edges , i.e. , @xmath96_i = d_l$ ] for all @xmath9 . on the contrary
, the @xmath39 ensemble has the irregular degree distribution @xmath97_i=\begin{cases } 4 , & \textrm{if $ i\in { \mathcal
r}_{\rm l , 1}\cup{\mathcal r}_{\rm l , 2}$ } \\ 3,&\textrm{otherwise } \end{cases}\ ] ] where @xmath98 and @xmath99 which indicate the positions where high degree variable nodes exist .
similarly , the degree distribution of the variable nodes in the @xmath66 ensemble is irregular such as @xmath100_i=\begin{cases } 2d_l,&\textrm{if $ i\in { \mathcal r}_{\rm oc}$ } \\
d_l,&\textrm{otherwise } \end{cases}\label{eq : degree_oc}\ ] ] where @xmath101 . as discussed in @xcite ,
the @xmath4 ensemble has two protected regions at each boundary position that are protected by the connected low degree check nodes . for the @xmath39 ensemble , there are four protected regions ; two of them at the boundary positions are formed by the connected low degree check nodes like the @xmath4 ensemble and the other two protected regions are @xmath102 and @xmath103 . on the other hand ,
the @xmath66 ensemble has one protected region @xmath104 .
although the protected regions @xmath102 , @xmath103 , and @xmath104 are similar because they are formed by high - degree variable nodes , the degree of variable node in the protected region @xmath104 is higher than that in @xmath102 or @xmath103 for @xmath105 , which implies that @xmath104 is more protected .
in this paper , the channel is assumed to be the binary erasure channel ( bec ) with erasure probability @xmath106 .
the de is often used as an analytical tool for the decoding performance of ldpc codes in the asymptotic setting such as infinite codeword length and infinite iteration numbers . for the coupled ensembles
, the de keeps track of the average erasure probability of messages at each variable node position .
let @xmath107 denote the average erasure probability of messages emitted from the variable nodes at position @xmath9 at the iteration @xmath108 .
set the initial conditions as @xmath109 for all @xmath9 . given a @xmath12 connectivity matrix @xmath13 and degree distribution vector @xmath110 , the evolution of @xmath111 can be expressed in vector form as @xmath112 where @xmath113 denotes the transpose and @xmath114_i&=\frac{(1-(1-{x_i})^{(d_r-1)})}{\sum\limits_{t=1}^{q}[{\bf t}]_{i , t } } , { \rm~for~ } i\in[1,p]\\ [ { \bf f(x)}]_i&=\frac{\epsilon{x_i}^{[{\bf d}]_i-1}}{\sum\limits_{t=1}^{p}[{\bf t}]_{t , i } } , { \rm~for~ } i\in[1,q].\end{aligned}\ ] ] with the de equation , we can obtain the bp threshold as a measure of decoding performance which is defined as below . the bp threshold @xmath115 of a coupled ensemble @xmath94 is defined as @xmath116 an asymptotically good ensemble with good bp decoding performances has a large bp threshold .
the bp threshold , however , assumes an infinite number of iterations which is infeasible in a practical situation .
thus , we include the required number of iterations as another measure of performance to design a practically good ensemble that works properly with the limited number of iterations . the required number of iterations @xmath117 of a coupled ensemble @xmath94 is defined as @xmath118 denote the bp and the map thresholds of the @xmath119-regular ldpc ensemble by @xmath120 and @xmath121 , respectively . for the bec with @xmath122 , where the variable nodes at each position of @xmath4 are decoded without the wave - like propagation , @xmath123 is independent of @xmath0 .
we focus on the required number of iterations when @xmath124 . in that range of erasure probability
, the required number of iterations means the minimum decoding time such that the variable nodes at all positions are successfully decoded via the propagation of reliable messages .
thus , the required number of iterations @xmath123 increases proportionally to @xmath0 . ensembles . ] in fig .
[ fig : sc_d_r_r ] , we observe how the measures of the @xmath40 ensemble behave as the chain length @xmath0 increases which is summarized as follows a. the design rate increases and approaches to the design rate of the underlying regular ldpc ensemble .
b. the bp threshold decreases for a while and then approaches to the map threshold of the underlying regular ldpc ensemble .
c. the required number of iterations for @xmath125 gradually increases almost linearly with @xmath0 . as the chain length @xmath0 increases , we obtain capacity approaching sc - ldpc ensembles due to the steadily increasing design rates and the saturated thresholds . however , the sc - ldpc ensemble with short chain length are expected to show good decoding performance because they have higher bp thresholds and lower required number of iterations . in @xcite@xcite
, it is shown that the asymptotic performance of the @xmath39 ensemble is comparable to that of the sc - ldpc ensemble with short chain length @xmath126 .
that is because the positions from @xmath17 to the first position in @xmath127 and from the last position in @xmath128 to @xmath129 , called the outer segment in @xcite , are almost surely decoded over the bp threshold of the short sc - ldpc chain with @xmath130 .
similar to the @xmath39 ensemble , the @xmath66 ensemble shows asymptotic performance comparable to the sc - ldpc ensemble with short chain length @xmath68 for @xmath69 .
recall that the de equation in ( [ eq : de ] ) is formulated if a connectivity matrix @xmath13 and a degree distribution vector @xmath110 are given .
consider the de equation of the @xmath66 ensemble with @xmath83 and @xmath131 in ( [ eq : degree_oc ] ) . from the symmetry of @xmath83 and @xmath131
, it is obvious that @xmath132 for @xmath133 $ ] and thus it is sufficient to keep track of the evolution of @xmath134 for @xmath135 $ ] .
the evolution of @xmath136 is expressed using only the upper half part of @xmath83 in ( [ eq : t_oc ] ) , which is equal to @xmath70 in ( [ eq : t_c ] ) , with the degree distribution given @xmath137_i=\begin{cases } 2d_l,&\textrm{if $ i\in [ l_s+1,l_s+w-1]$ } \\
d_l,&\textrm{if $ i\in [ 1,l_s]$}. \end{cases}\label{eq : degree_oc2}\ ] ] on the other hand , the evolution of @xmath138 of the @xmath139 ensemble can also be obtained from the segment of the evolution of @xmath136 which is derived by the de equation with the connectivity matrix @xmath70 and the degree distribution such as @xmath140_i=\begin{cases } \infty , & \textrm{if $ i\in [ l_s+1,l_s+w-1]$}\\ d_l , & \textrm{if $ i\in [ 1,l_s]$}. \end{cases}\label{eq : degree_sc2}\ ] ] the degree @xmath141 of the positions in @xmath142
$ ] imposes the boundary condition on @xmath136 by setting @xmath143 for @xmath144 $ ] at each iteration . to sum up , in terms of the de equation ,
the @xmath66 and @xmath139 ensembles are quite similar except for the variable node operation at the last @xmath28 positions , where known variable nodes exist for the @xmath139 ensemble while high - degree variable nodes exist for the @xmath66 ensemble .
the following lemmas show this similarity .
the solution of the de equation of the @xmath139 ensemble satisfies the following properties @xcite , @xcite .
a. @xmath145 [ lemma : sc_1 ] b. @xmath146 .
[ lemma : sc_2 ] [ lemma : sc ] the solution of the de equation of the @xmath66 ensemble satisfies the following properties : a. @xmath145 [ lemma : oc_1 ] b. @xmath147 [ lemma : oc_2 ] c. @xmath148 [ lemma : oc_3 ] d. @xmath149 [ lemma : oc_4 ] e. @xmath150 [ lemma : oc_5 ] [ lemma : oc ] we prove this lemma by a similar method used for the proof of lemma [ lemma : sc ] in @xcite .
all the properties are satisfied for @xmath151 by the initial condition .
suppose that the properties are satisfied for the inductive step @xmath152 .
consider the de equation with @xmath153 $ ] and @xmath154 in ( [ eq : degree_oc2 ] ) .
note that @xmath155_i=(1-(1-x_i))^{(d_r-1)}/w$ ] and @xmath156_i=\epsilon x_i^{[{\bf d}_{\rm oc , 2}]_i-1}/w$ ] are monotonic functions operating on each element @xmath157 .
define an @xmath158 permutation matrix @xmath159 such as @xmath160_{i , j}=1 $ ] if @xmath161 .
considering the definition of @xmath162 $ ] in ( [ eq : t_c ] ) , @xmath163 and @xmath164 are symmetric under simultaneous row - column reversal ( i.e. , @xmath165_{i , j}=[{\bf t}_{\rm sc}]_{l_s+w - i , l_s+1-j}$ ] and @xmath166_{i , j}=[{\bf t}_{\rm c , r}]_{l_s+w - i , w - j})$ ] and this implies that @xmath167 and @xmath168 .
define @xmath169 as @xmath170 & { \bf m}_{w-1 } \end{array } \right].\ ] ] then , the properties i ) and ii ) for step @xmath171 imply @xmath172 and @xmath173\left [ \begin{array}{c;{3pt/3pt}c } { \bf m}_{l_s } & \\\hdashline[3pt/3pt ] & { \bf m}_{w-1 } \end{array } \right ] \\ & = \left[\begin{array}{c;{3pt/3pt}c } { \bf t}_{\rm sc}{\bf m}_{l_s } & { \bf t}_{\rm c , r } { \bf m}_{w-1 } \end{array}\right]\\ & = \left[\begin{array}{c;{3pt/3pt}c } { \bf m}_{l_s+w-1}{\bf t}_{\rm sc } & { \bf m}_{l_s+w-1}{\bf t}_{\rm c , r } \end{array}\right]\\ & = { \bf m}_{l_s+w-1}{\bf t}_{\rm c}.\end{aligned}\ ] ] then , the inductive step for @xmath174 follows from @xmath175 where ( a ) follows from @xmath176 and the fact that @xmath154 of @xmath156_i=\epsilon x_i^{[{\bf d}_{\rm oc , 2}]_i-1}/w$ ] is separated into two parts . due to the symmetry ,
the properties iii ) , iv ) , and v ) for the @xmath171th inductive step can be expressed as @xmath177 for @xmath178 , i.e. , @xmath179 is monotonically decreasing for @xmath178 . for @xmath180_{i}$ ] , we have @xmath181_{i}-[{\bf t}_{\rm c}{\bf x}^{(k)}]_{i+1}&=\sum_{j=0}^{w-1}x^{(k)}_{\left < i - j , l_s+w-1\right>}-\sum_{j=0}^{w-1}x^{(k)}_{\left < i+1-j , l_s+w-1\right>}\\ & = ( x^{(k)}_{i-(w-1)}-x^{(k)}_{i+1})\ge 0\end{aligned}\ ] ] if @xmath182 or @xmath183 and this is equivalent to @xmath184 .
thus , @xmath180_{i}$ ] is decreasing for @xmath184 .
since @xmath185 does not affect the monotonicity , @xmath186_i$ ] is also decreasing over the same interval .
we can also show that @xmath187_i$ ] is decreasing for @xmath178 in a similar manner with @xmath180_{i}$ ] .
lastly , considering the definition of @xmath154 , @xmath188_{i}$ ] also decreasing for @xmath178 and thus the properties iii ) , iv ) , and v ) for step @xmath174 are satisfied . from the properties
i ) and iii ) in lemma [ lemma : oc ] , we can see that a vector @xmath189 of the @xmath66 ensemble is unimodal , which is the same as that of the @xmath139 ensemble shown in lemma [ lemma : sc ] . on the other hand ,
for the last @xmath28 positions of the @xmath66 ensemble , there exists another unimodal vector @xmath190 corresponding to @xmath104 . if the high - degree variable nodes of the @xmath66 ensemble are decoded , i.e. , @xmath191 for @xmath192 , the de equations of the @xmath66 and @xmath139 ensembles are exactly identical and then two ensembles have the same bp threshold and required number of iterations . however , these two performance measures of the @xmath66 ensemble are bounded by those of the @xmath139 ensemble as described in the following theorem . for the @xmath139 and the @xmath66 ensembles for @xmath193 , we have a. @xmath194 b. @xmath195 for @xmath196 , @xmath197 for the de equation of the @xmath66 ensemble while @xmath191 for the de equation of the @xmath139 ensemble .
thus , the solutions of the de equations for the @xmath66 ensemble are lower bounded by the solutions of the de equations for the @xmath198 ensemble , which implies the theorem . [
theorem : bounds ] the bounds in theorem [ theorem : bounds ] are achieved when @xmath191 for @xmath192 , but erasure probabilities are always larger than zero before the evolution of the de equation is successfully completed .
however , the erasure probabilities of @xmath104 are less than or equal to the minimum erasure probability of variable nodes in the remaining positions from the property v ) in lemma [ lemma : oc ] . also , through extensive numerical experiments , we observe that erasure probabilities of @xmath104 rapidly converge to zero after some iterations and the bounds in theorem [ theorem : bounds ] are almost achieved .
table [ table : small l ] shows the design rate @xmath199 , bp threshold @xmath200 , and the required number of iterations @xmath201 for each of @xmath40 , @xmath39 and @xmath202 ensembles for various values of @xmath0 up to @xmath203 . as discussed before ,
the loop and oc - ldpc ensembles suffer from additional rate loss compared to the sc - ldpc ensemble for the same @xmath0 .
however , the bp thresholds of the loop and oc - ldpc ensembles are improved compared to the corresponding sc - ldpc ensemble . regarding the required number of iterations ,
the loop and oc - ldpc ensembles significantly outperform the sc - ldpc ensembles .
such improvement comes from the property that asymptotic behaviors of the loop and oc - ldpc ensembles with chain length @xmath0 converge to that of the short sc - ldpc chain with chain length @xmath126 and @xmath204 , respectively .
for example , @xmath200 and @xmath201 of the @xmath205 and @xmath206 ensembles are comparable to those of the @xmath207 and @xmath208 ensembles , respectively . [ cols="^,^,^,^,^,^,^,^,^,^,^",options="header " , ] [ table : comparison_proto ] fig .
[ fig : simulation_fer ] compares the block error rate of particular code instances lifted from the @xmath209 , @xmath210 , and @xmath211 ensembles for @xmath212 $ ] with lifting factor @xmath213 and the loop ensemble in @xcite@xcite with lifting factor @xmath214 .
all codes have an equal number of transmitted bits as @xmath215 for @xmath216 and @xmath217 for @xmath218 .
in addition , we include the simulation result of two independent sc - ldpc codes with @xmath219 ( denoted by @xmath220sc @xmath219 ) as the performance bounds of the oc - ldpc and ocp - ldpc codes with @xmath69 .
the decoder uses the bp algorithm and runs the algorithm until the state of all variable nodes stays the same .
in other words , we restrict the maximum number of iterations @xmath221 to @xmath141 . for a short chain length @xmath216 , it is shown in fig .
[ fig : simulation_fer_l18 ] that the oc - ldpc code shows improved block error rate compared with the sc - ldpc and loop codes . in addition , the ocp - ldpc code shows further improved performance which is close to the performance of two independent sc - ldpc codes with @xmath222 . for a long chain length @xmath218 in fig .
[ fig : simulation_fer_l50 ] , the performance improvement has also been observed as the case of @xmath223 although the sc - ldpc ensemble has the same bp threshold as the oc - ldpc and ocp - ldpc ensembles for @xmath218 .
this is predicted by the expected graph evolution of the sc - ldpc and the oc - ldpc ensembles analyzed in section [ sec : analysis ] .
meanwhile , the ocp - ldpc code also shows the same performance as the oc - ldpc code because the oc - ldpc code already achieves the performance bound , which means that the precoding technique is not required for large @xmath0 .
furthermore , when the maximum number of iterations is restricted to @xmath224 , the block error rate of the codes for @xmath218 is compared in fig . [
fig : simulation_fer_l50_iter100 ] , which shows a noticeable improvement of the oc - ldpc code .
the additional performance improvement comes from the low required number of iterations @xmath225 discussed in section [ sec : analysis ] .
in this paper , a new coupled ensemble constructed by overlapping two circluar sc - ldpc ensembles is proposed .
a key property of the proposed ensemble called oc - ldpc ensemble is that it behaves as two sc - ldpc ensembles with shorter chain lengths during the decoding process . since the sc - ldpc ensemble is more easily decoded as the chain length decreases , the oc - ldpc ensemble gives more improved bp thresholds compared with the sc - ldpc ensemble . as a measurement of the decoding complexity
, we use the required number of iterations which is an important measure to design practically usable codes .
the numerical results show that the oc - ldpc ensemble reduces the required number of iterations compared with the sc - ldpc ensemble .
the construction method of random - based ensembles can be extended to protograph - based ensembles . from the numerical analysis
, we also verify that the performance improvement of the oc - ldpc codes can be achieved not only for asymptotic performances but also for finite - length performances .
10 s. kudekar , t. j. richardson , and r. l. urbanke , `` spatially coupled ensembles universally achieve capacity under belief propagation , '' _ ieee trans .
inf . theory _ ,
77617813 , dec . 2013 .
a. j. felstrom and k. s. zigangirov , `` time - varying periodic convolutional codes with low - density parity - check matrix , '' _ ieee trans .
inf . theory _ ,
45 , no . 6 , pp
. 21812190 , sep . 1999 .
a. sridharan , m. lentmaier , d. j. costello jr . , and
k. s. zigangirov , `` convergence analysis of a class of ldpc convolutional codes for the erasure channel , '' _ in proc .
allerton conf .
control comput .
_ , monticello , il , oct . 2004 , pp . 953962 .
r. m. tanner , d. sridhara , a. sridharan , t. e. fuja , and d. j. costello , jr .
, `` ldpc block and convolutional codes based on circulant matrices , '' _ ieee trans .
inf . theory _ ,
29662984 , dec . 2004 .
a. pusane , a. j. feltstrom , a. sridharan , m. lentmaier , k. s. zigangirov , and d. j. costello , jr .
, `` implementation aspects of ldpc convolutional codes , '' _ ieee trans .
commun_. , vol .
56 , no . 7 , pp
. 10601069 , jul .
2008 m. lentmaier , a. sridharan , d. j. costello , jr . , and k. sh .
zigangirov , `` iterative decoding threshold analysis for ldpc convolutional codes , '' _ ieee trans .
inf . theory _ ,
52745289 , oct .
a. e. pusane , r. smarandache , p. o. vontobel , and d. j. costello , jr .
, `` deriving good ldpc convolutional codes from ldpc block codes , '' _ ieee trans .
inf . theory _ ,
2 , pp . 835857 , feb . 2011 .
d. g. m. mitchell , a. e. pusane , and d. j. costello , jr .
, `` minimum distance and trapping set analysis of protograph - based ldpc convolutional codes , '' _ ieee trans .
inf . theory _ , vol .
1 , pp . 254281 , jan .
s. kudekar , t. j. richardson , and r. l. urbanke , `` threshold saturation via spatial coupling : why convolutional ldpc ensembles perform so well over the bec , '' _ ieee trans .
inf . theory _
2 , pp . 803834 , feb . 2011 .
a. yedla , y .- y .
jian , p. s. nguyen , and h. d. pfister , `` a simple proof of threshold saturation for coupled scalar recursions , '' in _ proc .
turbo codes iterative inf . process .
( istc ) _ , gothenburg , sweden , aug .
2012 , pp . 5155 .
, `` a simple proof of maxwell saturation for coupled scalar recursions , '' _ ieee trans .
inf . theory _ ,
60 , no.11 , pp . 69436965 ,
s. kudekar , t. j. richardson , and r. l. urbanke , `` wave - like solutions of general 1-d spatially coupled systems , '' _ ieee trans .
inf . theory _ ,
41174157 , aug .
p. m. olmos and r. urbanke , `` a scaling law to predict the finite - length performance of spatially coupled ldpc codes , '' _ ieee trans .
inf . theory _ , vol .
6 , pp . 31643184 , jun . 2015 .
r. ohashi , k. kasai , and k. takeuchi , `` multi - dimensional spatially coupled codes , '' in _ proc .
theory ( isit ) _ , istanbul , turkey , jul .
2013 , pp . 24482452 .
y. liu , y. li , and y. chi , `` spatially coupled ldpc codes constructed by parallelly connecting multiple chains , '' _ ieee commun .
d. truhachev , d. g. m. mitchell , m. lentmaier , and d. j. costello , jr , `` connecting spatially coupled ldpc code chains , '' in _ proc .
conf commun ( icc ) _ , ottawa , canada , jun .
2012 , pp . 21762180 .
, `` new codes on graphs constructed by connecting spatially coupled chains , '' in _ proc .
ieee ita _ , 2012 , pp .
, `` new codes on graphs constructed by connecting spatially coupled chains , '' [ online ] .
available : http://arxiv.org/abs/1312.3368 p.
m. olmos , d. g. m. mitchell , d. truhachev , and d. j. costello , jr .
, `` improving the finite - length performance of spatially coupled ldpc codes by connecting multiple code chains , '' [ online ] .
available : http://arxiv.org/abs/1402.7170 , `` improving the finite - length performance of long sc - ldpc code chains by connecting consecutive chains , '' in _ proc .
ieee istc _ ,
2014 , pp .
7276 . s. kudekar , c. measson , t. j. richardson , and r. l. urbanke , `` threshold saturation on bms channels via spatial coupling , '' in _ proc .
turbo codes iterative inf . process .
( istc ) _ , brest , france , sep .
2010 , pp . 309313 .
j. thorpe , `` low density parity check ( ldpc ) codes constructed from protographs , '' jpl , california inst .
( caltech ) , ca , usa , tech .
rep . 42154 , 2003 .
d. g. m. mitchell , m. lentmaier , and d. j. costello , jr .
, `` spatially coupled ldpc codes constructed from protographs , '' _ ieee trans .
9 , pp . 48664889 , sep . 2015 .
d. divsalar , s. dolinar , c. r. jones , and k. andrews , `` capacity approaching protograph codes , '' _ ieee j. sel .
areas commun .
27 , no . 6 , pp . 876888 , aug .
t. j. richardson and r. urbanke , _ modern coding theory_. cambridge , u.k .
: cambridege univ . press , 2008 .
v. aref , l. schmalen , and s. ten brink , `` on the convergence speed of spatially coupled ldpc ensembles , '' in _ proc .
51st annu .
allerton conf .
commun . , control , comput .
( allerton ) _ , monticello ,
il , usa , oct .
2013 , pp . 342349
. m. stinner and p. m. olmos , `` analyzing finite - length protograph - based spatially coupled ldpc codes , '' in _ proc .
inf . theory ( isit ) _ , honolulu , hi , usa , jun./jul .
2014 , pp . 891895 .
, `` on the waterfall performance of finite - length sc - ldpc codes constructed from protographs , '' in _ ieee j. sel .
areas commun .
_ , vol .
34 , no . 2 ,
pp . 345361 , feb .
m. b. tavares , k. s. zigangirov , and g. p. fettweis , `` tail - biting ldpc convolutional codes , '' in _ proc .
inf . theory ( isit ) _ , nice , france , jun .
2007 , pp . 23412345
|
in this paper , we propose new coupled codes constructed by overlapping circular spatially - coupled low - density parity - check ( sc - ldpc ) codes , which show better asymptotic and finite - length decoding performance compared to the conventional sc - ldpc codes .
the performance improvement comes from the property that the proposed codes effectively split into two separated sc - ldpc codes with shorter chain length during the decoding process .
we verify that the property of the proposed codes is valid in asymptotic setting via analysis tools such as the density evolution and the expected graph evolution .
experimental results show that the proposed codes also outperform the conventional sc - ldpc codes in terms of the finite - length performance under belief propagation decoding .
circular spatially - coupled low - density parity - check codes , density evolution , expected graph evolution , finite - length performance , low - density parity - check ( ldpc ) codes , spatially - coupled ldpc ( sc - ldpc ) codes .
|
as predicted by early band theory studies,@xcite charge carriers in graphene are chiral quasiparticles that have a linear dirac - like dispersion relation resulting in fascinating electronic @xcite and transport @xcite properties . as a result of this linear energy - momentum relationship , the valence and conduction bands intersect at the dirac points located at the brillouin zone corners . for neutral graphene ,
the chemical potential lies exactly at the dirac points .
use of a gate potential can shift the chemical potential away from the dirac point and thus tune the charge density of the graphene device over a large range ( as compared to other electronic materials ) .
the high degree of control over the charge density and the linear dispersion relation are features that distinguish graphene from the two - dimensional electron gas in semiconductor heterostructures and other metallic conductors .
understanding the mechanism for charge carrier transport in graphene in the low carrier density regime ( the linear dispersion regime ) and in the presence of disorder has received considerable attention.@xcite the wide variety of fabrication techniques @xcite has resulted in graphene devices that operate in different disorder regimes . the strength of the disorder charge carriers encounter can be quantified via the electron mobility .
the standard fabrication method of mechanical exfoliation and deposition of graphene flakes on sio@xmath1 substrates @xcite can result in good quality devices with mobilities approaching a few tens of thousands @xmath2 , values comparable to those found in conventional semiconductor devices .
yet , the substrate provides a source of impurities @xcite and corrugations @xcite for the electrons to scatter , limiting the mean free path @xmath3 to about a hundred nanometers .
etching away the substrate is routinely done to produce suspended graphene devices , removing some sources of scattering .
etching followed by current annealing increases the carrier mobility in suspended samples to values exceeding 200,000 @xmath2 , resulting in mean free paths comparable to the size of the device.@xcite similar sample quality ( 80,000 @xmath2 carrier mobility ) is obtained in devices where the substrate is single - crystal hexagonal boron nitride.@xcite it is apparent that replacing oxide based substrates by high - quality crystal substrates or suspended graphene results in devices with improved mobility and reduced electron scattering events.@xcite for transport near or at the dirac point , where the density of states vanishes , the conductivity has a minimum value ( @xmath4 ) , even for ideal pristine graphene.@xcite experimentally , the phase coherent transport in graphene devices with short - and - wide geometries has been characterized as ballistic @xcite indicating evanescent wave transport . in an ideal ( no impurities or defects ) graphene device , the mean free path is longer than the size of the device resulting in ballistic conductivity .
thus , transport in graphene qualitatively changes from _ metallic diffusive _ in sio@xmath1 substrates to what has been characterized as _ ballistic _ in suspended systems and when using crystalline boron nitride substrates .
our paper is organized as follows . in sec .
ii we discuss the usual transport measures , emphasizing the null results obtained from the fano shot noise , and present two alternative methods that can also estimate the transport mechanism . in sec .
iii we present the computational method used to model transport in two - terminal graphene devices . in sec .
iv we report results of extracting the transport mechanisms from modeled transport measurements .
we present a summary and conclusion in sec .
the possibility of several transport mechanisms makes it essential to develop measures that clearly distinguish and classify the different transport regimes . in principle ,
fano shot noise is the metric of choice to assess the mechanism for electronic transport in mesoscopic devices .
however , the fano shot noise is of limited applicability , with the same value for the fano factor found in disordered or pristine devices .
shot noise measurements can determine the statistics relevant for transport in mesoscopic conductors .
@xcite shot noise is a consequence of charge quantization and can be assessed by the fano factor defined as the ratio of noise power and mean current , @xmath5 from the above definition , the fano factor should be zero for conventional ballistic systems , _
i.e. _ if perfect transmission is present .
it has also been established that for poisson processes @xmath6 and for diffusive metallic conductors @xmath7.@xcite however , the fano factor calculated for pristine graphene,@xcite disordered systems,@xcite samples with substrate roughness,@xcite and for transport along an n - p junction @xcite all have similar magnitude ( @xmath0 ) to that found in diffusive metallic conductors , making the fano shot noise a cumbersome measure of the transport mechanism in graphene devices.@xcite calculations in pristine graphene predict that the fano factor at the dirac point is identically 1/3 when ( 1 ) a wide device ( where the aspect ratio
w / l is above 3 ) is contacted by doped graphitic leads , @xcite ( 2 ) when the leads are quantum wires creating an effective - contact model simulating the metallic lead / graphene interface , @xcite and ( 3 ) for graphene junctions with realistic metal contacts .
@xcite away from the dirac point , the fano factor also indicates diffusive metallic transport .
@xcite a slightly smaller yet universal value ( @xmath8 ) was found numerically for disordered systems .
@xcite when disorder originates mainly from roughness in the substrate , the fano factor lies slightly above the 1/3 value and has been shown numerically to increase slightly with disorder .
@xcite transport along an n - p ( electron rich / hole rich ) junction is selective of those quasiparticles that approach the n - p interface almost perpendicularly and results in a fano factor of @xmath9 .
@xcite again , in all of these different situations , the fano factor measurement results in 1/3 or a value numerically close to this number .
experiments have found similar results .
the fano factor measured in clean devices @xcite is close to the value of 1/3 found analytically for ballistic samples .
@xcite measurements in disordered devices @xcite follow the numerical trends of ref . .
these similar shot noise values warrant developing an alternative quantitative understanding of the carrier dynamics crucial for testing the transport regime of graphene devices .
our quest for a better measure to assess transport mechanisms begins with a simple question : how does the conductance change with respect to the vertical displacement of one of the leads ? by calculating the conductance between two metallic contacts , we model transport in two - terminal devices .
we extract the conductance profile from the dependence of the conductance on the position of one of the leads , and use this to distinguish the mechanism of transport .
in particular , we study the transport properties of clean and disordered graphene devices contacted by narrow metallic leads where one of the two leads can be moved along the edge of the device .
moreover , we simulate scanning probe microscopy ( spm ) @xcite measurements for the same devices , which sheds light on the charge flow _ inside _ the device
. @xcite diffusive transport in disordered systems is based on the charge carriers scattering multiple times off impurities or boundaries as they traverse the system .
this mechanism can be described classically by a random walk . in ballistic transport , the charge carriers traverse the system with minimal scattering .
ballistic transport is expected in an ideal graphene strip , given that the crystal lattice has no defects and no impurities are present .
yet for ideal graphene , the dynamics of the electrons produce the same shot noise as that found in classical diffusion . @xcite determining the transport mechanism in graphene devices is important given that transport experiments are possible in the quasi - ballistic limit , that is , where the mean free path is of the order of the size of the system . here , we argue that the transition from diffusive to ballistic transport in graphene , along with the limiting cases , can be quantified by alternative methods based on measuring the flow of charge in the system . first , the conductance profile , that is , the conductance as a function of the vertical displacement of one of the leads , can give a measure of the flow inside the device . for the case of ballistic transport , scattering does not impede the flow of charge carriers and the conductance profiles are well described by a linear ( triangular ) fit due to the convolution of the square windows created by the two leads . in the case of diffusive transport ,
the charge carriers experience several scattering events and the transport is then well described by brownian motion with a drift . thus , charge flow will be a gaussian function characteristic of diffusion ; for this reason , the conductance profile can be fitted to a gaussian function .
we use the nonequilibrium green s function ( negf ) formalism @xcite to calculate the transport properties of a normal - conductor / graphene / normal - conductor junction , given that in experiments the electronic contacts are usually metallic .
the metallic lead / graphene interface we have investigated follows that of robinson and schomerus .
@xcite a schematic of the geometry of the devices investigated is shown in fig .
[ fig1 ] , where rectangular ( width w and length l ) graphene flakes have armchair boundaries along their longitudinal direction and narrow metallic leads connected along the zigzag boundary .
the two metallic terminals of width w@xmath10 are modeled by semi - infinite square lattice regions that have a quadratic dispersion relation .
( color online ) schematic setup for the transport calculation .
a rectangular graphene flake , depicted by red and blue atomic sites arranged in a honeycomb lattice , is contacted between two narrow metallic contacts , depicted by the green atomic sites arranged in a square lattice , forming a normal - conductor / graphene / normal - conductor junction .
this schematic is not to scale ; simulated devices were considerably bigger with @xmath11 carbon atoms along the zigzag edge . ] the tight - binding model hamiltonian of the device is given by @xmath12 where @xmath13 is the on - site potential of the lattice site @xmath14 , @xmath15 is the annihilation ( creation ) operator acting on site @xmath14 , the second sum is over nearest neighbors @xmath16 , and @xmath17 denotes the hopping matrix elements . the on - site potential changes due to contributions from impurities and the gate voltage applied to the device , @xmath18 .
the graphene section consists of atomic sites placed in the hexagonal lattice with lattice constant @xmath19 . as outlined in ref .
, metallic contacts are modeled as a region with a square lattice arrangement with lattice constant @xmath20 , matching the a(b)-atom in the zigzag interface at the left(right ) of the graphene flake .
disorder is introduced in two different ways . to generate edge - disordered samples ,
we randomly remove 30% of carbon atoms from sites located in the three outer atomic layers of the device .
@xcite this type of disordered edge without disorder puddles in the bulk of the system is a reasonable model for suspended graphene devices .
bulk disordered in graphene devices is linked to the presence of charged impurities in the substrate .
@xcite to generate such disorder potentials , @xmath21 lattice sites are selected at random from the total number of atomic sites ( @xmath22 ) in the device .
the position @xmath23 of each of the @xmath21 lattice sites has an on - site potential amplitude @xmath24 chosen randomly from the interval @xmath25 and smoothed out over a range @xmath26 by convolution with a gaussian , @xmath27 the range of the convolution is important given that in the extreme case of @xmath28 the atomic scale disorder potential would break the a - b symmetry caused by having two atoms in the unit cell . for our calculations
we assume that @xmath29 resulting in a short - range potential ( when compared to other length scales in the system ) that varies smoothly on the atomic scale and suppresses the effect of intervalley scattering .
the parameters @xmath21 and @xmath30 determine the mean free path @xmath3 of the disordered system . using the born approximation , the mean free path can be quantified as @xcite @xmath31 where @xmath32 is the fermi wave vector and @xmath33 is the dimensionless correlator given by @xmath34 for the type of bulk disordered systems considered here ( @xmath35 ) , rycersz _ et al . _
@xcite have found that @xmath36 where @xmath37 the negf formalism @xcite is a sophisticated framework for obtaining the transmission and other quantities in realistic devices .
the retarded green s function in the atomic orbital basis is given by @xmath38^{-1},\ ] ] where the non - hermitian self - energy matrix @xmath39 introduces the effect of attaching leads to the left and right ends of the device .
the energy @xmath40 tunes the fermi level from the dirac point ( @xmath41 ) to any charge density induced by the gate voltage in the device .
the self - energies determine the escape rates of electrons from the device into the semi - infinite ideal leads . using the landauer formula @xcite
, it is possible to obtain the transmission function @xmath42,\ ] ] where @xmath43 is the advanced green s function .
the matrices @xmath44\ ] ] introduce a level broadening due to the coupling of the leads and a source - to - drain voltage given by @xmath45 .
the fano factor [ eq . ( [ eq : fano ] ) ] for a particular energy and @xmath45 is calculated using the following expression , @xmath46}{t}.\ ] ] finally , the current in two - terminal devices can be obtained from the landauer formula @xmath47\ ] ] where the energy window is defined from the difference of the fermi functions of the macroscopic reservoirs where the leads terminate . in our calculations
we assume the linear response limit ( @xmath48 ) , where the relationship between conductance @xmath49 and current is given by @xmath50 .
we investigate rectangular graphene devices with a two - terminal geometry where the source and drain leads are assumed to be perfect ballistic and metallic conductors .
this geometry has been investigated previously in the context of quantum scars in graphene @xcite and for the geometry - dependent conductance fluctuations in graphene quantum dots .
@xcite for all results presented here and as illustrated in fig .
[ fig1 ] , the edges of the graphene along the transport direction are in the armchair configuration .
we use eq .
[ eq : fano2 ] to calculate the fano factor for pristine , edge disordered , and bulk disordered graphene devices . as presented in fig .
[ fig : fano ] , our results are similar to those found in the literature.@xcite the fano factors found are near the theoretical value @xmath51 that applies to pristine graphene as @xmath52 near the dirac point limit . as towrzydlo _
_ reported,@xcite the @xmath51 value is a theoretical maximum for armchair edge graphene devices , increasing the charge density reduces the value of the fano factor .
for pristine devices , the fano factor does show a trend as a function of the system size .
the longer device calculated ( 83 nm ) presented the lowest value of @xmath53 while the largest value corresponds to the 44 nm graphene sample .
likewise , the fano factor for the graphene devices with disordered edges do not exhibit a trend in the sizes studied .
however , the fluctuations in @xmath53 from the mean value , illustrated with the standard deviation ( error bars ) , show that as the devices get longer , the deviation from the mean increases .
in contrast , for diffusive graphene , we find that the value of the fano factor does show a monotonic dependence as a function of the @xmath54 ratio , approaching the @xmath51 value in the larger devices .
yet , given that the shot noise metric is not significantly different from the value found in classical diffusive transport , even in the narrow - leads geometry , the fano factor is an inadequate metric for predicting the transport mechanism .
( color online ) fano factor at a fermi energy @xmath55 ( with hopping matrix element @xmath56 ) . the limit @xmath51 as @xmath52 is shown as a dotted line .
the data points of the disordered systems have error bars showing the variance of the value over several realizations .
each of the data points corresponds to different system lengths ( l = 11 , 23 , 44 , and 85 nm ) while the widths of the graphene device and of the leads are kept constant at w = 74 nm and w@xmath10 = 10 nm , respectively .
the mean free path is @xmath57 nm in the system with bulk disorder.,scaledwidth=55.0% ] in order to visualize the charge flow at the edge of the sample , we calculate the profile of the conductance @xmath58 , that is , the conductance as a function of the position of the drain lead with respect to the source lead that remains fixed . in fig .
[ fig2 ] , we present the conductivity as a function of the displacement of the drain lead @xmath59 for several system lengths ( l=11 , 21 , 43 nm ) and for the three models considered here .
both the width of the device and the width of the leads are kept constant at w = 74 nm and w@xmath10 = 10 nm , respectively .
thus , the schematic shown in fig .
[ fig1 ] is not representative of the size of the devices considered in this work as our calculation are considerably larger with about 700 atomic sites along the zigzag edge ( width of system ) .
numerical calculations of these system sizes require the use of efficient recursive green s function methods .
( color online ) profile of the conductance for pristine , edge disordered , and bulk disordered systems .
the conductance versus displacement of one of the leads from the center of the device @xmath58 is calculated at a fermi energy @xmath55 ( with hopping matrix element @xmath56 ) .
each of the curves corresponds to different system lengths ( l = 11 , 21 , and 43 nm ) while the width of the graphene device and of the leads are kept constant at w = 74 nm and w@xmath10 = 10 nm , respectively .
( a ) clean pristine system .
( b ) system with edge disorder .
( c ) system with bulk disorder where the mean free path is @xmath57 nm.,scaledwidth=37.0% ] while the displacement of one of the leads from the center of the device is a theoretical construct that can be easily investigated numerically , the experimental implementation is not trivial . our proposal of moving the leads " could be realized by fabricating several samples where one of the leads is attached at a different location in each device .
another possibility is to replace one of the metallic leads by the tip of a scanning tunneling microscope . when the tip is brought close enough to the edge , the tunneling current could be measured as a function of the position of the scanning tunneling microscope .
figure [ fig2](a ) corresponds to pristine graphene devices , fig .
[ fig2](b ) corresponds to realizations with edge disorder , and fig .
[ fig2](c ) to systems where the presence of bulk disorder would form electron - rich and hole - rich puddles @xcite and the mean free path is @xmath57 nm , as estimated from the born approximation . figure [ fig2](a ) shows that the maximum conductance is associated with the two leads being collinear ( @xmath60 ) .
for a short device the maximum conductance occurs at @xmath61 and as the length of the system in increased there are local maxima near @xmath61 .
increasing the length of the device reduces the conductance peak due to the increased number of reflections at the boundary of the sample . yet
, we find an envelope of maximum conductance when the vertical positions of the two leads overlap with each other .
similar features are seen in the conductance profiles for the disordered samples .
we find that for all three systems the width of the central peak is approximately @xmath62 , corroborating the idea that the peak in the displacement conductance is due to the overlap of the two leads .
although our interest is to extract the transport mechanisms from the conductance profiles , there are certain features visible in the conductance curves of fig .
[ fig2 ] that are worth explaining such as the fluctuations near the edges of the conductance curves .
we carefully checked that these fluctuations are not only present for individual conductance curves and remain after averaging over an energy window .
thus , the fluctuations are consistent with universal conductance fluctuations ( ucfs ) , @xcite which we therefore review . in a disordered mesoscopic conductor where the system is of comparable size or smaller than the phase coherence length of the charge carriers but large compared to the average impurity spacing
, the transmission of carriers is affected by interference of many different paths through the system . as these paths are typically long compared to the wavelength of the charge carriers , the accumulated phase along the paths changes randomly in response to variation in an external parameter ( e.g. , the magnetic field or gate voltage ) .
this results in a random interference pattern and reproducible fluctuations in the conductance of a universal magnitude on the order of @xmath63 .
@xcite ucfs can also be created by the displacement of a single scatterer .
@xcite but how can ucfs occur in a ballistic device ?
the role of disorder in providing a distribution of random phases can as well be taken by chaos .
thus , ballistic mesoscopic cavities like quantum dots in high mobility two - dimensional electron gases that form chaotic billiards show the same universal fluctuations .
@xcite remarkably , experiments in graphene quantum dots found strong indications of chaos in billiard systems .
@xcite in fig .
[ fig3 ] we presents our conductance profile analysis for the 11 nm devices .
a striking feature of these conductance profiles is the apparent fit of the peaks in the displacement conductance curves to either triangular or gaussian functions .
the conductance profile for a pristine device of length @xmath64 nm is shown in fig .
[ fig3](a ) and the peak is presented in fig .
[ fig3](b ) . the envelope of the peak in the conductance is well described by a triangular fit ( red line ) . in the case of conventional ballistic transport , the classical expectation for the charge flow at the edge is a triangular shape of width @xmath62 .
this triangular shape is the result of the convolution of two square windows , where the window sizes are given by the lead width @xmath65 .
gaussian fits for the conductance profile ( not shown ) differ significantly particularly near the cusp of the curve . diffusive transport is based on the multiple scattering paths taken by charge carriers as they transverse a device . in fig .
[ fig3](c ) we present the conductance profile for an edge disordered system of length @xmath64 nm and a close - up of the peak with a fit shown in fig .
[ fig3](d ) . in this case , and in contrast with the wide and ballistic device , the data for the edge - disordered device is best fitted by a gaussian curve , where the envelope of the conductance profile is the result of the spatial overlap of the leads .
similarly , a gaussian curve describes well the conductance profile of a diffusive system , as seen in figs .
[ fig3](e ) and [ fig3](f ) .
we note that barthelemy _ et al . _
@xcite used a similar approach to distinguish normal diffusion ( reflected by a gaussian profile ) from the anomalous transport associated with lvy transport .
while we have been able to account for the features of the envelope of the central peak of width @xmath62 in a quasi - classical way , the conductance profile curves exhibit several quantum interference effects that need a different interpretation . along with ucfs ,
fabry - prot conductance interferences show up as regularly spaced fluctuations seen in the conductance profiles of the pristine systems . scanning probe microscopy ( spm ) @xcite
has recently been used to image mesoscopic transport effects such as universal conductance fluctuations @xcite and weak localization @xcite in graphene devices .
spm has also been used to image current transport in a graphene quantum dot connected to leads through two small constrictions , imaging conductance resonances of the quantum dot and observing localized states .
@xcite the tip of the cryogenic spm capacitively couples to the graphene device inducing a movable scatterer .
conductance maps created via spm have been proposed as a way to probe the chiral nature of the charge carriers in graphene .
@xcite the work carried out by berezovsky _ et al . _
@xcite and schnez _ et al . _
@xcite probed the regime of coherent diffusive transport in graphene . in light of the above ,
it is interesting to consider the possibility of using spm to gain insight into the transport mechanism by imaging the flow of charge carriers particularly , the ballistic to diffusive crossover regime in graphene devices .
we simulate the effect of the capacitively - coupled tip of the spm on the graphene device as a point charge @xmath66 located a height @xmath19 above the substrate .
the charge induces a local charge density perturbation given by @xmath67 where @xmath68 is the in - plane radial coordinate away from the position of the point charge .
the charge @xmath66 is chosen to yield an rms charge density @xmath69 , in agreement with the observed charge puddles in graphene .
the tip height @xmath19 controls the width of the induced charge density ; we have chosen @xmath70 nm , a routinely used tip - to - substrate distance in spm experiments which results in a half - width for the induced density puddle of about 20 nm . during spm measurements , a transport measurement is carried out while the tip of the microscope scans the over the device .
the resulting conductance maps are generated by calculating the conductance as a function of spm tip position @xmath71 and rastering the tip position in a plane above the device .
the effect of the spm tip is included by adding to the on - site potential of each of the lattice sites the energy due to the charge density perturbation created by the tip .
the total on - site energy is obtained from the contributions of the perturbation from the tip , the effect from the gate voltage , and the impurities present , @xmath72 . in the previous section we extracted conductance profiles that allowed the classification of transport as ballistic or diffusive . to explain these observations and to obtain an intuitive picture of the transport mechanisms
, we explore charge flow as a function of the position of one of the leads . as in the conductance profile calculations ,
the injection lead remains fixed in the left edge of the sample .
the two terminals are identical . as seen in the panels of fig .
[ deltay - scan ] , the spm conductance maps exhibit a significant change in conductance when the charge flow between the leads is obstructed or enhanced by the presence of the perturbation induced by the scanning probe . the conductance map shown in fig .
[ deltay - scan](a ) was obtained when the right lead is positioned near the bottom edge of the device . as the scanning probe rasters along the top half of the device ,
the change in conductance is minimal as reflected by the mostly uniform conductance in that region of the image .
however , as the tip scans the bottom half of the device , the conductance map shows a region of lower conductance that connects the left lead to the right lead , as indicated by the large blue feature across the bottom of the image . in this case , the trans - conductance is reduced when the local perturbation induced by the tip of the spm is over a region of considerable charge flow . in fig .
[ deltay - scan](b ) , the right lead contacts the device at a slightly higher position .
consequently , the blue feature corresponding to the region of charge flow is now at a higher position , following " the position of the right lead .
furthermore , there are two regions of higher conductance in the image , corresponding to an increase of the conductance due to the presence of the local perturbation .
the higher conductance ( red ) bands are a consequence of the redirection of the charge flow that without the tip being present would not have contributed to the conductance .
thus , the tip redirects charge into the drain lead that otherwise would not have exited the device . similarly , placing the right lead closer to the central region of the device ( fig .
[ deltay - scan](c ) ) , we find that the position of conductance features are correlated with the path starting at the left lead and ending in the right lead . in all images , the effect of narrow lead constrictions
is seen from the size of the band of near constant conductance .
surprisingly , the feature does not become significantly wider as the charge flows away from the source lead .
all of the panels present conductance fluctuations of order @xmath73 . the conductance map presented in fig .
[ deltay - scan](d ) corresponds to a device where the leads are collinear ( @xmath61 ) . here , the conductance is higher than in the ones to the left .
along with the central band of conductance drop , fig .
[ deltay - scan](d ) also shows regions of constructive interference along bands parallel to the transport .
the limit to the size of the features present in the conductance maps is proportional to the fermi wavelength of the system @xcite , which for graphene is inversely proportional to the fermi energy given by @xmath74 ( color online ) conductance maps for two fermi energies , ( left ) @xmath75 and ( right ) @xmath76 , corresponding to fermi wavelengths of 13 nm and 1.5 nm , respectively .
, scaledwidth=45.0% ] the fermi wavelength in the system presented in fig .
[ deltay - scan ] is @xmath77 nm . figure [ fig : fermiwavelength ] presents the simulated conductance maps for the same configuration of the leads for @xmath78 and @xmath79 , corresponding to fermi wavelengths of 13 nm and 1.5 nm .
the resolution resolved in the images , i.e. , the size of the pixels in each of the maps , is also 1.5 nm .
previous analytical results of braun _ et al . _
@xcite consider ballistic trajectories between two constrictions in the presence of the tip scattering potential .
to first order , as treated in ref . , there are only two trajectories that interfere with each other : the source - to - drain trajectory and the source - tip potential - drain trajectory , revealing an interference pattern in the conductance maps .
in particular , the first order calculation predicts that the interference pattern depends on the fermi wavelength of the system . comparing the conductance maps for three different energies ( fig .
[ deltay - scan](d ) and those in fig .
[ fig : fermiwavelength ] ) , the interference pattern are very similar . increasing the fermi energy of the system does not reveal the fermi wavelength of the particles but rather results in better resolution of the interference patterns .
in contrast to analytical results , @xcite our numerical calculation treats all possible trajectories including reflections from the edge of the device and multiple scattering events .
we believe that the pattern found is due to the redirection of the charge carriers as the tip scans the device and does not dependent on the wavelength but rather on the geometry of the system and the interference of several trajectories the charge carriers can take as they traverse the device . in what follows we study the effect of edge and bulk disorder on the simulated conductance maps . in fig .
[ clean_edge_bulk ] , we present conductance maps for graphene devices with edge disorder ( left panel ) and bulk disorder ( middle and right panels ) .
as seen in the left panel , the interference pattern is present in edge - disordered graphene devices .
the effect of disorder at the edges of the device induces random scattering , modifying previously interfering trajectories and thus reducing the strength of the pattern .
this is more noticeable near the edge of the device .
still , the result for edge - disordered systems is a comparable pattern found for pristine graphene ( fig .
[ deltay - scan](d ) ) .
our calculations show an important difference for bulk - disordered systems . surprisingly
, the spm simulations indicate diffusive rather than ballistic transport ( in the mesoscopic systems considered ) even on the regime where the mean free path ( estimated from the born approximation ) is longer than the system size . this result does not question the validity of the born approximation ; rather , it shows the sensitivity of the spm method for the geometry studied .
even when the size of the system is smaller than @xmath3 , for this geometry , the oath traveled by the charge carrier from initially entering the device to exiting at the drain lead can be significantly longer .
our calculations take into account all possible trajectories that the charge carriers can take , including the possibility of several reflections from the leads and multiple scattering events from the walls of the device . the presence of random disorder due to charged impurities in the substrate results in conductance maps with no interference patterns , ( fig .
[ clean_edge_bulk ] , middle and right panels ) .
the middle panel of fig .
[ clean_edge_bulk ] is the conductance map for a system where the mean free path ( @xmath80 nm ) is larger than the size of the device ( @xmath81 nm ) .
this value of @xmath3 is attainable in graphene devices with a crystalline bn - substrate , resulting in relatively high mobility @xcite .
the mean free path of the system pictured in the right panel is @xmath82 nm , comparable to high quality graphene devices in a silicon oxide substrate .
ballistic transport is expected when the mean free path is larger than the size of the system .
in contrast , in graphene devices with weak bulk disorder such that @xmath83 , we find diffusive transport signatures .
this indicates that for the system and sizes considered , multiple reflections of the dirac electrons from the edges dominate .
this mesoscopic effect needs to be carefully considered in both theoretical and experimental studies on mesoscopic graphene flakes . as a result
, spm imaging could be used to explore the ballistic to diffusive transition in graphene devices .
the visible interference pattern in the conductance maps would be robust against the presence of edge disorder , inevitable in most experimental realizations , but is sensitive to the presence of disorder due to bulk disorder in the system .
finally , we would like to comment on possible experimental realizations of spm measurements in graphene .
several spm measurements have been carried out for graphene devices on a sio@xmath1 substrate @xcite and can be readily carried out in suspended graphene , where the ballistic transport regime can be reached , in graphene membranes , @xcite and in devices with crystalline substrate , where the ballistic to diffusive cross - over should be observable .
we have proposed a method to study the mechanisms of electronic transport based on displacement conductance , that is , the conductance as a function of the position of the drain lead in a two - terminal device .
the method extracts the conductance profile of the charge flow at the edge of the device and it can be applied to discern the mechanism of transport from ballistic to diffusive .
it is worth noting that this technique might be capable of uncovering signatures of anomalous transport in mesoscopic systems .
several quantum interference effects ( fabry - prot resonances and ucfs ) are also found in the resulting conductance curves .
the method that we applied to graphene is general , resolves the transport mechanism in graphene which can not be accomplished via fano shot noise , and can be used as well to study other systems such as semiconductor heterostructures . in this article
, we simulated spm measurements for the same devices considered in the conductance profile study .
probe imaging is important as we are able to visualize the flow of charge and do not have to rely on transport metrics . for the graphene devices considered , the spm simulations suggest diffusive rather than ballistic transport even in the clean " regime where the mean free path is larger than the systems size .
our numerical results suggest that in the process of escaping the device , charge carriers perform multiple reflections . from our simulations
, we expect that spm measurements are well suited to study the crossover between ballistic and diffusive transport in graphene devices .
we are grateful to s. barraza - lopez , s. bhandari , j. berezovsky , and r. m. westervelt for valuable discussions .
m.f.b . and e.j.h .
were supported by the department of energy , office of basic science ( grant no . de - fg02 - 08er46513 ) , and h.h .
acknowledges funding through the german research foundation ( grant no .
he 6312/1 - 1 ) .
while most of the calculations were run on the odyssey cluster supported by the fas at harvard university , the rest were performed on the vaquero cluster supported by oklahoma state university and outfitted with a tesla k20 gpu donated by the nvidia corporation .
b. k. nicolic , l. p. zarbo , and s. souma , spin currents in semiconductor nanostructures : a nonequilibrium green - function approach , in a. v. narlikar and y. y. fu ( eds . ) , _ the oxford handbook on nanoscience and technology frontiers and advances _ , ( oxford university press , oxford , 2010 ) .
|
we investigate the transport of electrons in disordered and pristine graphene devices .
fano shot noise , a standard metric to assess the mechanism for electronic transport in mesoscopic devices , has been shown to produce almost the same magnitude ( @xmath0 ) in ballistic and diffusive graphene devices and is therefore of limited applicability .
we consider a two - terminal geometry where the graphene flake is contacted by narrow metallic leads .
we propose that the dependence of the conductance on the position of one of the leads , a conductance profile , can give us insight into the charge flow , which can in turn be used to analyze the transport mechanism .
moreover , we simulate scanning probe microscopy ( spm ) measurements for the same devices , which can visualize the flow of charge inside the device , thus complementing the transport calculations . from our simulations
, we find that both the conductance profile and spm measurements are excellent tools to assess the transport mechanism differentiating ballistic and diffusive graphene systems .
|
the observation of surface cooling of accreting neutron stars on timescales of days to years after they go into quiescence has opened up a new window on the physics of neutron star crusts .
six neutron stars in low mass x - ray binaries have now been observed to cool after extended accretion outbursts that were long enough to heat the crust significantly out of thermal equilibrium with the core @xcite .
the subsequent thermal relaxation of the crust depends on its physical properties such as its thickness , thermal conductivity , and heat capacity , with deeper regions being probed at successively later times in the cooling light curve ( * ? ? ?
* ; * ? ? ?
* hereafter bc09 ) .
@xcite and bc09 showed that the observed cooling curves of ks 1731260 and mxb 165929 imply that the inner crust has a thermal conductivity corresponding to an impurity parameter of order unity .
@xcite find that the cooling curve of xte j1701462 is compatible with similar crust microphysics to ks 1731260 and mxb 165929 .
chemical separation occurs when , as a material freezes , the equilibrium compositions of the liquid and solid phases are different .
this important process can drive sedimentation or mixing in white dwarf @xcite and giant planet ( e.g. , * ? ? ?
* ) interiors , and earth s core @xcite .
@xcite carried out molecular dynamics simulations of the freezing of the multicomponent plasma expected in the outer layers of accreting neutron stars , and found chemical separation occurred , with lighter nuclear species being preferentially retained in the liquid phase . in a previous paper ( * ? ? ? * hereafter mc11 ) we considered the effect of chemical separation while accretion is ongoing . in that case , as matter is driven to higher pressure and crosses the freezing depth , the light elements released into the neutron star ocean will drive convection , mixing them throughout the ocean and raising the light element fraction . at the same time , because the convection is driven by composition gradients in an otherwise thermally stable layer , the convective heat flux is inward , potentially heating the layers deep in the ocean . in this paper ,
we consider the cooling of the outer layers of the star in quiescence , during which the liquid layers freeze into a solid and chemical separation occurs , enriching the ocean further in light elements .
we show that the resulting compositionally driven convection can significantly modify the expected cooling curves of accreting transients .
we first discuss the expected size of the compositionally driven heat flux and show that it is easily comparable to the cooling flux in the ocean ( section [ sec : theory ] ) .
we then present simulations of ocean cooling including heat transport by convection , showing that the light curve is significantly modified ( section [ sec : simulation ] ) .
we conclude with a discussion of the theoretical uncertainties and implications for observed sources ( section [ sec : discuss ] ) .
the outermost layers of an accreting neutron star form a gaseous atmosphere and liquid ocean @xcite , in regions where the coulomb coupling parameter @xmath0 @xcite , where @xmath1 is the number average over the mixture of nuclei with charge @xmath2 , @xmath3 is the electron sphere radius , and @xmath4 is the temperature . a useful measure of the depth is the column depth @xmath5 , related to the pressure by @xmath6 where @xmath7 is the gravity .
the base of the ocean lies at y_b,14 = 1.8 ( ) ^4 ( ) ^-4()^-1 , [ eq : yb ] where @xmath8 , @xmath9 k , @xmath10 , and the subscript ` @xmath11 ' signifies that the quantities are taken at the base of the ocean .
note that in eq .
( [ eq : yb ] ) we assume @xmath12 , appropriate for an oxygen - enriched ocean ; a heavy - element rich can be much shallower , @xmath13 .
bc09 showed that when accretion ends , the cooling of the star proceeds by the temperature profile relaxing from the outside - in . at
any given time , lower density layers have cooled and adopted a constant flux temperature profile , whereas deeper layers have yet to thermally relax .
the transition occurs at a depth given by setting the thermal timescale = 19 days y_14()^-1()^2()^-1 [ eq : tthermal ] ( henyey & lecuyer 1969 ; equation 7 of bc09 ) equal to the current time . in eq .
( [ eq : tthermal ] ) , @xmath14 is the mass density , @xmath15 is the specific heat , @xmath16 is the thermal conductivity , and @xmath17 is the pressure scale height . in the ocean , the pressure is dominated by the relativistic degenerate electrons with @xmath18 , giving the scalings @xmath19 , @xmath20 , and @xmath21 and @xmath4 constant ( e.g. , mc11 and references therein ) . equations ( [ eq : yb ] ) and ( [ eq : tthermal ] ) show that tens of days after the onset of quiescence , the base of the ocean will start to cool and solidify . as new crust is formed , the bottom of the ocean becomes enriched in light elements because of chemical separation @xcite . in mc11 , we estimated the timescales for particle nucleation , growth , and sedimentation of solid particles , finding that these `` microscopic '' timescales were much shorter than `` macroscopic '' timescales such as the time to accrete the ocean , or more relevant here , the time for the ocean to cool .
the picture then is that fluid elements with a light composition are deposited at the base of the ocean and will rise upwards , driving convective mixing .
moreover , the convection occurs in a medium that is otherwise thermally stratified .
therefore the convective heat flux , which is proportional to the excess of the temperature gradient @xmath22 in the star compared to the adiabatic gradient @xmath23 , or @xmath24 , is negative ; the compositionally driven convection transports heat inwards .
we can estimate the expected size of @xmath25 by comparing the heat and composition fluxes . in mixing length theory , the heat flux is @xmath26 , where @xmath27 is the convective velocity ; whereas the composition flux is @xmath28 , where @xmath29 is the composition gradient in the star @xmath30 ( mc11 ) , and @xmath31 is the light element mass fraction .
the convective velocity is given by the superadiabaticity , v_conv^2 g l^2 , where @xmath32 is the mixing length , @xmath33 , and @xmath34 .
in the ocean , the convection is extremely efficient and therefore close to marginal stability , @xmath35 .
the heat and composition fluxes are then related by f_conv = - c_pt ( ) .
[ eq : fconvfx ] during cooling , the composition of the ocean is changing with time as the ocean is enriched in light elements . for efficient convection ,
the light element fraction is fairly constant over the ocean ( @xmath36 ) , and so @xmath37 is roughly constant throughout the ocean .
the composition flux is then @xmath38 , and the heat flux is f_conv = -y c_pt .
[ eq : fconvx ] taking @xmath39 and @xmath40 ( mc11 ) , we find a flux that increases steeply inwards , @xmath41 .
if cooling is unaffected by compositionally driven convection , we estimate the timescale on which @xmath31 is changing , @xmath42 , to be a typical thermal time at the ocean floor : in the standard cooling model the bulk of the ocean freezes on a thermal time ; as it does it releases nearly all of its light elements into the remaining ocean and approximately doubles the light element content there .
equation ( [ eq : fconvx ] ) becomes f_conv-10 ^ 25 erg cm^-2 s^-1 y_14 ^ 5/4 ( ) ^-1 [ eq : fconv ] where we have assumed @xmath43 and @xmath44 [ and @xmath45 and @xmath46 in eq .
( [ eq : tthermal ] ) ] . the flux from compositionally driven convection in eq . ( [ eq : fconv ] ) easily outweighs the cooling flux , @xmath47 .
this suggests that the cooling of an ocean with chemical separation included should be significantly different than without , and motivates our numerical calculations that will be presented in the next section .
as we shall see below , the freezing of the ocean is strongly regulated by the convective heat flux , and in turn keeps the heat flux at a much lower level than is suggested by eq .
( [ eq : fconv ] ) .
we solve for the thermal relaxation of the neutron star ocean and crust by solving the thermal diffusion equation following bc09 , but including convective heat fluxes modeled using mixing length theory and assuming the convection is efficient as described in mc11 , and in addition by following the composition profile as chemical separation and mixing occur .
for the examples shown here , we take the ocean composition to be a mixture of oxygen and selenium , for which the phase diagram is shown in the lower panel of figure 1 in mc11 .
a full description of our numerical code will be presented elsewhere @xcite ; here we focus on the resulting light curves and describe the influence of compositionally driven convection on the evolution .
is the time from the end of the accretion outburst and @xmath48 is the effective temperature ( @xmath49 ) as seen by an observer at infinity .
the curves were generated with the initial temperature profile shown in fig .
[ fig : tcool ] ; @xmath50 for each run is as labeled .
the labels that appear above the graph denote the duration of the stages of convection for the @xmath51 case ( see text ) .
the spikes and wiggles in the convection curves are the result of the oscillations of @xmath52 around @xmath53 as described in the text , coupled with our finite numerical resolution . ]
value and , for the convective models , the stage is given in parentheses .
for each convection curve , the temperature at the ocean - crust boundary is marked with an open circle .
the curves were generated with @xmath51 ( as in the solid and dotted curves of fig .
[ fig : lightcurve ] . ) ] ) .
the base of the ocean is located at the right - most extent of each curve ; while the top of the convection zone is located at the depth where @xmath31 ( the light element fraction ) reaches the burning layer level of 0.02 ( paper i ) or , for @xmath54 days , at the top of the ocean . ]
figure [ fig : lightcurve ] shows an example light curve with and without compositionally driven convection included .
figures [ fig : tcool ] and [ fig : xenrich ] show the temperature and composition profiles , respectively , at different times as the ocean and crust cool .
the temperature profile at the end of the accretion outburst is similar to that assumed by bc09 , with an inward directed heat flux , and ( for figs .
[ fig : tcool ] and [ fig : xenrich ] ) an initial base composition @xmath55 , corresponding to the steady state ( mc11 ) .
the kink in the temperature profile at @xmath56 is due to the jump in composition between the ocean and overlying h / he layer ( cf .
paper i ) .
the neutron star mass and radius are @xmath57 and @xmath58 , giving a redshift factor of @xmath59 .
we find that the evolution proceeds in four stages . during stage 1 ,
the base of the ocean has not yet started to cool and so the evolution is the same with or without convection included .
the light curve is a power - law with slope given by equation 8 of bc09 , with the modification that @xmath60 in the ocean rather than 3/4 as in the outer crust . in stage 2
( @xmath61 in figs .
[ fig : lightcurve][fig : xenrich ] ) , the cooling wave has reached the bottom of the ocean , and new crust begins to form , driving convection . because convection transports heat inward from the ocean to the ocean - crust boundary , cooling at the boundary is delayed at the expense of more rapid cooling in the ocean . as a result ,
the temperature profile in the ocean becomes very steep and the light curve drops faster than without convection . during this stage ,
the convective heat flux is strong and the boundary cools very slowly ; the transition depth and temperature @xmath62 and @xmath63 remain close to their values at the onset of quiescence . in this way
, the convection acts analogously to the latent heat .
the cusp in the temperature profile is due to the jump in conductive flux that must occur to balance the large inward convective flux in the ocean .
the temperature profile can not steepen indefinitely , because eventually the temperature gradient at the base of the ocean @xmath52 approaches @xmath64 , the liquidus temperature gradient ( i.e. , how the melting temperature varies with pressure ; see mc11 ) .
when @xmath65 , multiple depths at the bottom of the ocean freeze simultaneously .
this rapid freezing quickly suppresses itself , however , as strong compositionally driven convection heats the base of the ocean and melts the top of the crust , mixing a heavy - element fluid into the bottom of the ocean and thereby stabilizing the ocean against further convection .
cooling resumes , but convection remains off , such that the steep temperature profile at the ocean base that had been supported by convection quickly flattens due to heat conduction .
once the heavy - element - enriched fluid at the base of the ocean completely solidifies , compositionally driven convection resumes and the temperature profile steepens again . in this manner @xmath52 oscillates around @xmath53 .
we refer to this phase as stage 3 ( @xmath66 in figs .
[ fig : lightcurve][fig : xenrich ] ) . in this stage convection is sporadic in the ocean and can no longer prevent the ocean base from cooling . as a result , the ocean boundary moves outward and the cooling wave then continues its inward motion through the crust .
we find generally that @xmath67 ; i.e. , that the ocean - crust boundary cools rapidly but moves outward slowly .
this is due to the compensating effect of light element enrichment on the freezing depth . during stage 3 ,
the surface temperature @xmath49 still drops with time , but at a rate similar to without convection ; and now @xmath65 is constant while @xmath63 and @xmath62 drop and @xmath68 increases . during stage 4 ( @xmath69 in figs .
[ fig : lightcurve][fig : xenrich ] ) , the crust is thermally relaxed , the ocean cools too slowly for convection to support the steep gradient @xmath65 , and the light curve returns to the shape it would have if there was no ocean convection .
note , however , that due to light element enrichment the asymptotic value of @xmath49 is slightly higher in the case with convection than without ( the ocean thermal conductivity @xmath70 , and so for a given base temperature the outwards flux is greater for a lower @xmath71 ) .
alternatively , the ocean experiences an abrupt transition to stage 4 when the base is saturated with light elements ( @xmath72 ) and chemical separation halts ; this happens at @xmath73 days in the @xmath74 case of fig .
[ fig : lightcurve ] .
previous calculations of the thermal relaxation of accreting neutron stars in quiescence have assumed that the ocean cools passively . in contrast , we have shown here that mixing in the ocean driven by chemical separation at the base leads to a significantly different evolution , changing the expected cooling curve . the early time ( 1100 days )
cooling curve of quiescent neutron stars potentially offers a remarkable new probe of the freezing and chemical separation of a classical plasma in `` real time '' .
the timing of the rapid drop in flux is sensitive to the composition and temperature of the ocean at the end of the outburst .
the magnitude of the effect depends on the composition of the ocean , in particular the fraction of light elements and contrast in the atomic mass of the light and heavy nuclei . and a core temperature of @xmath75 k , and use an impurity parameter @xmath76 and @xmath55 ( solid and dotted curves ) and @xmath77 and @xmath78 ( dashed curve ) . for igr
j174802446 we set @xmath79 and @xmath80 , and use @xmath81 k and @xmath82 ( solid and dotted curves ) and @xmath83 k and @xmath84 ( dashed curve ) . ]
detecting the signature of convection will require better sampling of the early phase of the cooling curve .
we can generally fit currently available light curves equally well with and without convection .
these fits will be presented in a companion paper @xcite ; two examples are shown in fig .
[ fig : both ] .
xte j1701462 @xcite is interesting because its high temperature means that the ocean remains liquid for hundreds of days .
@xcite fit the light curve with a standard cooling model ( neglecting the two data points marked in fig .
[ fig : both ] that are argued to be contaminated with residual accretion ) . including convection we find that we can match the drop in luminosity at @xmath85@xmath86 days .
a general result is that convection lessens , but does not remove , the need for a shallow heat source in the ocean during accretion ( bc09 ; paper i ; * ? ? ?
* ) , because light element enrichment increases the thermal conductivity and reduces the temperature gradient , making the ocean hotter .
convection could help with recent observations of two classical transients ( with shorter 23 month outbursts ) . in igr
j174802446 @xcite , the flux remains elevated above the value observed before the outburst .
convection allows @xmath50 to change from one accretion episode to the next , modifying the ocean conductivity and thereby changing the late time temperature ( compare the solid curve with the hatched region in fig .
[ fig : both ] ) .
xte j1709267 showed a rapid decrease in temperature during a single 8 hour xmm observation @xcite .
if due to crust cooling , @xcite suggested that a strong heat source must be operating at low densities within the ocean during the outburst , necessary for significant thermal relaxation after a short outburst of only 23 months .
we can reproduce the rapid drop in temperature with convection if stage 2 of the light curve ( section [ sec : simulation ] ) occurs during the observation .
there remains much to be explored theoretically .
we have included only two species in our calculations , oxygen and selenium , which approximates the rp - process ashes used by @xcite .
the phase diagram for multicomponent mixtures is complex but can be calculated @xcite and should be included .
we have assumed that solid particles form at a single depth . however
, electron capture reactions may occur in the ocean ( for example , @xmath87fe captures at a density of @xmath88 , @xcite ) , lowering the @xmath71 at that depth , and potentially leading to formation of solid particles pre - electron capture above the post - electron capture liquid layers
. it will be important to include carbon burning in the models .
enrichment of the ocean with carbon remains a major issue for superburst models @xcite .
chemical separation during the cooling phase will significantly enrich the ocean in light elements immediately following an outburst , much more efficiently than gravitational sedimentation during quiescence .
this may have implications for the puzzling superburst observed immediately before the onset of an accretion outburst in exo 1745248 @xcite .
we thank chuck horowitz , nathalie degenaar , and chris fontes for useful discussions .
is supported by an nserc discovery grant and is an associate member of the cifar cosmology and gravity program .
we are grateful for the support of an international team on neutron star crusts by issi in bern .
z.m . was supported by a lanl director s postdoctoral fellowship .
this research was carried out in part under the auspices of the national nuclear security administration of the u.s .
department of energy at los alamos national laboratory and supported by contract no .
de - ac52 - 06na25396 .
|
we show that convection driven by chemical separation can significantly affect the cooling light curves of accreting neutron stars after they go into quiescence .
we calculate the thermal relaxation of the neutron star ocean and crust including the thermal and compositional fluxes due to convection .
after the inward propagating cooling wave reaches the base of the neutron star ocean , the ocean begins to freeze , driving chemical separation .
the resulting convection transports heat inward , giving much faster cooling of the surface layers than found assuming the ocean cools passively .
the light curves including convection show a rapid drop in temperature weeks after outburst . identifying this signature in observed cooling curves would constrain the temperature and composition of the ocean as well as offer a real time probe of the freezing of a classical multicomponent plasma .
|
the study of information theoretic secrecy was initiated by shannon in @xcite . following shannon s work ,
significant contributions were made by wyner @xcite who established the rate - equivocation region of a degraded broadcast channel .
wyner s result was generalized to the case of a general broadcast channel by csiszar and korner @xcite .
recently , there has been a resurgence of activity in studying multi - terminal and vector extensions of @xcite , @xcite . in this paper
, we investigate a secure transmission problem from a source coding perspective .
in particular , we first consider a simple setup consisting of four terminals .
terminal @xmath10 ( say alice ) observes an i.i.d .
source @xmath0 which it intends to transmit losslessly to terminal @xmath11 ( say bob ) .
a malicious but passive user ( say eve ) gets to observe the coded output of alice . in other words ,
the communication link between alice and bob is public , i.e. , insecure .
it is clear that since the malicious user gets the same information as the legitimate user , there can not be any positive secret rate of transmission .
on the other hand , if there is a helper , say helen , who observes an i.i.d .
source @xmath2 which is correlated with the source @xmath0 and transmits information over a secure rate - limited link to bob , then one can aim for creating uncertainty at the eavesdropper ( see figure @xmath10 , @xmath11 and @xmath12 , secure links are shown by * bold * lines . ] ) . for the model shown in figure @xmath10 , we completely characterize the rate - equivocation region . from our result , we observe that the classical achievablity scheme of ahlswede and korner @xcite and wyner @xcite for source coding with rate - limited side information is robust in the presence of a passive eavesdropper .
secondly , we consider the model where alice also has access to the coded output of helen and completely characterize the rate - equivocation region .
we will call this model the two - sided helper model ( see figure @xmath11 ) . from our result
, we observe that the availability of additional coded side information at alice allows her to increase uncertainty of the source at eve even though the rate needed by alice to transmit the source losslessly to bob remains the same .
this observation is in contrast with the case of insecure source coding with side information where providing coded side information to alice is of no value @xcite .
we next generalize the setup of figure @xmath11 to the case when there are both secure and insecure rate - limited links from helen and there is additional side information @xmath7 at bob and additional side information @xmath8 at eve . in particular , there is a secure link of capacity @xmath13 , whose output is available at alice and bob and an insecure link , of capacity @xmath14 , whose output is available at all three terminals , i.e. , at alice , bob and eve ( see figure @xmath12 ) .
the presence of both secure and insecure links from helen can be interpreted as a source - coding analogue of a degraded broadcast channel from helen where alice and bob receive both secure and insecure streams @xmath15 and @xmath16 , whereas , eve only receives the insecure stream @xmath16 .
we completely characterize the rate - equivocation region for this model when @xmath17 forms a markov chain .
we explicitly compute the rate - equivocation region for the cases of one - sided helper and two - sided helper for a pair of binary symmetric sources .
we show that having access to helen s coded output at alice yields a strictly larger equivocation than the case of one - sided helper .
the secure source coding setup shown in figure @xmath10 was considered in @xcite where it was also assumed that eve has access to additional correlated side information @xmath8 .
inner and outer bounds for the rate - equivocation region were provided for this setup , which do not match in general .
the rate - equivocation region was completely characterized in @xcite for the case when bob has complete uncoded side information @xmath2 and eve has additional side information @xmath8 .
this result also follows from @xcite where a similar three terminal setup was studied and the maximum uncertainty at eve was characterized under the assumption of no rate constraint in the lossless transmission of the source to bob .
a similar model was also studied in @xcite where bob intends to reconstruct both @xmath0 and @xmath2 losslessly .
it was shown that slepian - wolf type coding suffices for characterizing the rate - equivocation region when the eavesdropper does not have additional correlated side information .
this setup was generalized in @xcite to the case when the eavesdropper has additional side information @xmath8 , and inner and outer bounds were provided , which do not match in general . in @xcite ,
a multi - receiver secure broadcasting problem was studied , where alice intends to transmit a source @xmath0 to @xmath18 legitimate users .
the @xmath19th user has access to a correlated source @xmath20 , where @xmath21 , for @xmath22 , and the eavesdropper has access to @xmath8 , where @xmath23 , and the noise sequences @xmath24 are mutually independent and also independent of the source @xmath0 .
furthermore , it was assumed that alice also has access to @xmath25 . for sources with such modulo - additive structure
, it was shown that to maximize the uncertainty at the eavesdropper , alice can not do any better than describing the error sequences @xmath26 to the legitimate users .
this model is related to the two - sided helper model shown in figure @xmath11 ; see section @xmath27 for details . for the model shown in figure @xmath12 ,
when we set @xmath28 , i.e. , in the absence of helen , we recover the result obtained in @xcite .
therefore , our result can also be viewed as a generalization of the result obtained in @xcite . by setting @xmath29 , i.e. , in the absence of the secure rate - limited link ,
the resulting model is related to the model considered in @xcite where the aim is to generate a secret key between two terminals via an insecure rate - limited two - sided helper . in the model studied in this work
, the aim is to securely transmit the source @xmath0 to bob .
note that , when @xmath29 and @xmath30 , both the secret key generation capacity @xcite and secure transmission rate are zero since eve has access to both @xmath16 and @xmath6 along with @xmath8 . on the other hand , in the presence of a secure link ,
i.e. , when @xmath31 , even when @xmath30 , we can still create uncertainty at the eavesdropper .
this is possible since helen can choose not to transmit any information on the insecure link and transmit only a coded description of @xmath2 by using the secure link at the rate @xmath13 , which plays the role of a correlated key .
furthermore , being correlated with the source @xmath0 , the coded description of @xmath2 also permits alice to lower the rate of transmission when compared to the case of using an uncorrelated secret key , where alice transmits at a rate @xmath32 .
in section @xmath33 , we present the rate - equivocation region for the case of one - sided helper . we show that slepian - wolf type coding alone at alice is optimal for this case .
we present the rate - equivocation region for the case of two - sided helper in section @xmath27 . for the case of two - sided helper
, alice uses a conditional rate - distortion code to create an auxiliary @xmath34 from the source @xmath35 and the coded output @xmath36 received from helen .
this code construction is reminiscent of lossy - source coding with two - sided helper @xcite,@xcite , @xcite . for the case of lossy source coding , a conditional rate - distortion code
is used where @xmath34 is selected to satisfy the distortion criterion . on the other hand ,
the purpose of @xmath34 in secure lossless source coding is to confuse the eavesdropper . from this result , we demonstrate the insufficiency of slepian - wolf type coding at alice by explicitly utilizing the side information at alice .
this observation is further highlighted in section @xmath37 where we compare the rate - equivocation regions of two - sided helper and one - sided helper cases for a pair of binary symmetric sources .
for this example , we show that for all @xmath38 , the information leakage to the eavesdropper for the two - sided helper is strictly less than the case of one - sided helper .
we finally generalize the result of two - sided helper to the case when there are both secure and insecure rate - limited links from the two - sided helper and additional side informations @xmath39 and @xmath40 , at bob and eve , respectively .
the presence of secure and insecure rate - limited links from helen can be viewed as a source - coding analogue of a degraded broadcast channel from helen to ( alice , bob ) and eve .
we characterize the rate - equivocation region for this model when the sources satisfy the condition @xmath41 .
we consider the following source coding problem .
alice observes an @xmath42-length source sequence @xmath0 , which is intended to be transmitted losslessly to bob .
the coded output of alice can be observed by the malicious user eve .
moreover , helen observes a correlated source @xmath2 and there exists a noiseless rate - limited channel from helen to bob .
we assume that the link from helen to bob is a secure link and the coded output of helen is not observed by eve ( see figure @xmath10 ) .
the sources @xmath43 are generated i.i.d . according to @xmath44 where @xmath44 is defined over the finite product alphabet @xmath45 .
the aim of alice is to create maximum uncertainty at eve regarding the source @xmath0 while losslessly transmitting the source to bob .
an @xmath46 code for this model consists of an encoding function at alice , @xmath47 , an encoding function at helen , @xmath48 , and a decoding function at bob , @xmath49 .
the uncertainty about the source @xmath0 at eve is measured by @xmath50 .
the probability of error in the reconstruction of @xmath0 at bob is defined as @xmath51 .
a triple @xmath52 is achievable if for any @xmath53 , there exists a @xmath54 code such that @xmath55 and @xmath56 .
we denote the set of all achievable @xmath52 rate triples as @xmath57 .
the main result is given in the following theorem .
[ theorem1 ] the set of achievable rate triples @xmath57 for secure source coding with one - sided helper is given as @xmath58 where the joint distribution of the involved random variables is as follows , @xmath59 and it suffices to consider such distributions for which @xmath60 .
the proof of theorem @xmath10 is given in the appendix .
we note that inner and outer bounds for source coding model considered in this section can be obtained from the results presented in @xcite although these bounds do not match in general .
these bounds match when bob has complete uncoded side information @xmath2 , i.e. , when @xmath61 .
the achievability scheme which yields the rate region described in theorem @xmath10 is summarized as follows : 1 .
helen uses a rate - distortion code to describe the correlated source @xmath62 to bob .
alice performs slepian - wolf binning of the source @xmath35 with respect to the coded side information at bob .
therefore , our result shows that the achievable scheme of ahlswede , korner @xcite and wyner @xcite is optimal in the presence of an eavesdropper .
moreover , on dropping the security constraint , theorem @xmath10 yields the result of @xcite,@xcite .
we next consider the following generalization of the model considered in section @xmath33 . in this model , alice also has access to the coded output of helen besides the source sequence @xmath0 ( see figure @xmath11 ) .
an @xmath46 code for this model consists of an encoding function at alice , @xmath63 , an encoding function at helen , @xmath48 , and a decoding function at bob , @xmath49 .
the uncertainty about the source @xmath0 at eve is measured by @xmath50 .
the probability of error in the reconstruction of @xmath0 at bob is defined as @xmath64 .
a triple @xmath52 is achievable if for any @xmath53 , there exists a @xmath54 code such that @xmath55 and @xmath56 .
we denote the set of all achievable @xmath52 rate triples as @xmath65 .
the main result is given in the following theorem .
[ theorem2 ] the set of achievable rate triples @xmath65 for secure source coding with two - sided helper is given as @xmath66 where the joint distribution of the involved random variables is as follows , @xmath67 and it suffices to consider distributions such that @xmath68 .
the proof of theorem @xmath11 is given in the appendix .
we remark here that the proof of converse for theorem @xmath11 is closely related to the proof of the converse of the rate - distortion function with a two - sided helper @xcite,@xcite,@xcite .
the achievability scheme which yields the rate region described in theorem @xmath11 is summarized as follows : 1 .
helen uses a rate - distortion code to describe the correlated source @xmath62 to both bob and alice through a coded output @xmath36 .
2 . using the coded output @xmath36 and the source @xmath35
, alice jointly quantizes @xmath69 to an auxiliary random variable @xmath34 .
she subsequently performs wyner - ziv coding , i.e. , bins the @xmath34 sequences at the rate @xmath70 such that bob can decode @xmath34 by using the side information @xmath36 from helen .
alice also bins the source @xmath35 at a rate @xmath71 so that having access to @xmath72 , bob can correctly decode the source @xmath35 .
the total rate used by alice is @xmath73 .
therefore , the main difference between the achievability schemes for theorems @xmath10 and @xmath11 is at the encoding at alice and decoding at bob .
also note that selecting a constant @xmath34 in theorem @xmath11 corresponds to slepian - wolf type coding at alice which resulted in an equivocation of @xmath74 in theorem @xmath10 .
we will show in the next section through an example that the uncertainty about the source at eve for the case of two - sided helper can be strictly larger than the case of one - sided helper and selecting @xmath34 as a constant is clearly suboptimal .
besides reflecting the fact that the uncertainty at eve can be strictly larger than the case of a one - sided helper , theorem @xmath11 has another interesting interpretation .
if alice and helen can use sufficiently large rates to securely transmit the source @xmath0 to bob , then the helper can simply transmit a secret key of entropy @xmath32 to both alice and bob .
alice can then use this secret key to losslessly transmit the source to bob in perfect secrecy by using a one - time pad @xcite .
in other words , when @xmath1 and @xmath3 are larger than @xmath32 , one can immediately obtain this result from theorem @xmath11 by selecting @xmath36 to be independent of @xmath75 and uniformly distributed on @xmath76 .
finally , selecting @xmath77 , we observe that @xmath78 , yielding perfect secrecy .
we note that , here , @xmath36 plays the role of a secret key .
now consider the model where the side information @xmath2 is of the form @xmath79 , where @xmath80 , and @xmath81 is independent of @xmath0 .
moreover , assume that the side information @xmath2 is available to both alice and bob in an uncoded manner .
for this model , it follows from @xcite that , to maximize the uncertainty at the eavesdropper , alice can not do any better than describing the error sequence @xmath81 to bob .
note that our two - sided helper model differs from this model in two aspects : first , in our case , the common side information available to alice and bob is coded and rate - limited , secondly , the sources in our model do not have to be in modulo - additive form .
our encoding scheme at alice for the case of two - sided helper comprises of two steps : ( a ) using the coded side information @xmath36 and the source @xmath35 , alice creates @xmath34 and transmits the bin index of @xmath34 at a rate @xmath70 so that bob can estimate @xmath34 using @xmath36 , and , ( b ) alice bins the source @xmath35 at a rate @xmath71 and transmits the bin index of the source @xmath35 . note that if for a pair of sources @xmath75 , the optimal @xmath36 is of the form @xmath82 , where @xmath80 and @xmath83 is independent of @xmath35 , then it suffices to choose @xmath84 , so that @xmath85 and @xmath86 . in other words , for such sources ,
step ( b ) in our achievability scheme is not necessary , which is similar to the achievability for the case of modulo - additive sources in @xcite . on the other hand , for a general pair of sources @xmath75
, the optimal @xmath36 may not be of the form @xmath82 and the optimal @xmath34 may not always satisfy @xmath86 .
therefore , we need the step ( b ) for our achievability scheme .
this differentiates our achievable scheme from that of @xcite , which holds for modulo - additive sources with uncoded side information .
also note that if @xmath2 is not of the form @xmath87 , and if @xmath88 , then helen can transmit a secret key which will enable perfectly secure transmission of @xmath0 by using a one - time pad @xcite .
this phenomenon does not always occur when the side information @xmath2 is available to both alice and bob in an uncoded fashion @xcite .
before proceeding to further generalizations of theorems @xmath10 and @xmath11 , we explicitly evaluate theorems @xmath10 and @xmath11 for a pair of binary sources .
let @xmath35 and @xmath62 be binary sources with @xmath89 and @xmath90 , where @xmath91 . for this pair of sources ,
the region described in theorem @xmath10 can be completely characterized as , @xmath92 and the region in theorem @xmath11 can be completely characterized as , @xmath93 where @xmath94 is the binary entropy function , and @xmath95 .
we start with the derivation of ( [ regn1 ] ) . without loss of generality , we assume that @xmath96
. achievability follows by selecting @xmath97 , where @xmath98 , where @xmath99 substituting , we obtain @xmath100 which completes the achievability .
converse follows by simple application of mrs .
lemma @xcite as follows .
let us be given @xmath101 .
we have @xmath102 which implies @xmath103 .
gerber s lemma states that for @xmath90 , with @xmath91 , if @xmath104 , then @xmath105 . we therefore have , @xmath106 and @xmath107 this completes the converse . for the case of two - sided helper ,
we compute the equivocation @xmath4 as follows .
we choose @xmath36 as @xmath108 where @xmath98 as in the case of one - sided helper .
we choose @xmath34 as , @xmath109 we next compute the term @xmath110 as follows , @xmath111 and therefore @xmath112 for the converse part , we also have that @xmath113 the rate from alice , @xmath1 and the equivocation @xmath4 for the cases of one - sided and two - sided helper are shown in figure @xmath33 for the case when @xmath114 . for the one - sided helper
, we can observe a trade - off in the amount of information alice needs to send versus the uncertainty at eve .
for small values of @xmath3 , alice needs to send more information thereby leaking out more information to eve .
the amount of information leaked is exactly the information sent by alice . on the other hand , for the case of two - sided helper , the uncertainty at the eavesdropper is always strictly larger than the uncertainty in the one - sided case .
also note that for this pair of sources , perfect secrecy is possible for the case of two - sided helper when @xmath115 which is not possible for the case of one - sided helper .
in this section , we consider a generalization of the model considered in section @xmath27 .
we consider the case when there are two links from helen ( see figure @xmath12 ) .
one link of rate @xmath13 is secure , i.e. , the output of this link is available to only alice and bob .
the second link of rate @xmath14 is public and the output of this link is available to alice , bob and eve .
the sources @xmath116 are generated i.i.d . according to @xmath117 where @xmath118 is defined over the finite product alphabet @xmath119 .
a @xmath120 code for this model consists of an encoding function at helen , @xmath121 , where @xmath122 , @xmath123 , an encoding function at alice , @xmath124 , and a decoding function at bob , @xmath125 .
the uncertainty about the source @xmath0 at eve is measured by @xmath126 .
the probability of error in the reconstruction of @xmath0 at bob is defined as @xmath127 .
a quadruple @xmath128 is achievable if for any @xmath53 , there exists a @xmath120 code such that @xmath55 and @xmath129 .
we denote the set of all achievable @xmath130 rate quadruples as @xmath131 . the main result is given in the following theorem .
[ theorem3 ] the set of achievable rate quadruples @xmath132 for secure source coding with secure and insecure links from a two - sided helper , additional side information @xmath39 at bob and @xmath40 at eve is given as @xmath133 where the joint distribution of the involved random variables is as follows , @xmath134 and it suffices to consider such distributions for which @xmath135 , @xmath136 .
the proof of theorem @xmath12 is given in the appendix . the achievability scheme which yields the rate region described in theorem @xmath12 is summarized as follows : 1 .
helen first uses the secure link to describe the source @xmath62 to both alice and bob at a rate @xmath137 , where @xmath39 plays the role of side information .
subsequently , the insecure link is used to provide another description of the correlated source @xmath62 at a rate @xmath138 , where @xmath139 plays the role of side information .
therefore , the key idea is to first use the secure link to build common randomness at alice and bob and subsequently use this common randomness to send information over the insecure link at a lower rate .
2 . having access to the coded outputs @xmath140 from helen and the source @xmath35 , alice jointly quantizes @xmath141 to an auxiliary random variable @xmath34 .
she subsequently performs wyner - ziv coding , i.e. , bins the @xmath34 sequences at the rate @xmath142 such that bob can decode @xmath34 by using @xmath39 and the coded outputs @xmath140 from the helper .
she also bins the source @xmath35 at a rate @xmath143 so that having access to @xmath144 , bob can correctly decode the source @xmath35 .
the total rate used by alice is @xmath145 .
we note here that using theorem @xmath12 , we can recover theorem @xmath11 by setting @xmath146 , and selecting @xmath147 . on setting @xmath29 ,
the resulting model is similar to the one considered in @xcite although the aim in @xcite is to generate a secret key to be shared by alice and bob , while we are interested in the secure transmission of the source @xmath35 .
if @xmath28 , then we recover the result of @xcite as a special case by setting @xmath148 . therefore , theorem @xmath12 can also be viewed as a generalization of the result of @xcite where no rate constraint is imposed on the transmission of alice and the goal is to maximize the uncertainty at eve while losslessly transmitting the source to bob .
in this paper , we considered several secure source coding problems .
we first provided the characterization of the rate - equivocation region for a secure source coding problem with coded side information at the legitimate user .
we next generalized this result for two different models with increasing complexity .
we characterized the rate - equivocation region for the case of two - sided helper .
the value of two - sided coded side information is emphasized by comparing the respective equivocations for a pair of binary sources .
it is shown for this example that slepian - wolf type coding alone is insufficient and using our achievable scheme , one attains strictly larger uncertainty at the eavesdropper than the case of one - sided helper .
we next considered the case when there are both secure and insecure rate - limited links from the helper and characterized the rate - equivocation region .
1 . codebook generation at helen : from the conditional probability distribution @xmath150 compute @xmath151 .
generate @xmath152 codewords @xmath153 independently according to @xmath154 , where @xmath155 .
2 . codebook generation at alice : randomly bin the @xmath156 sequences into @xmath157 bins and index these bins as @xmath158 , where @xmath159 .
3 . encoding at helen : on observing the sequence @xmath160 , helen tries to find a sequence @xmath153 such that @xmath161 are jointly typical . from rate - distortion theory , we know that there exists one such sequence as long as @xmath162 .
helen sends the index @xmath163 of the sequence @xmath153 .
4 . encoding at alice : on observing the sequence @xmath156 , alice finds the bin index @xmath164 in which the sequence @xmath156 falls and transmits the bin index @xmath164 .
decoding at bob : on receiving @xmath163 and the bin index @xmath164 , bob tries to find a unique @xmath156 sequence in bin @xmath164 such that @xmath165 are jointly typical .
this is possible since the number of @xmath156 sequences in each bin is roughly @xmath166 which is @xmath167 and the existence of a unique @xmath156 such that @xmath165 are jointly typical is guaranteed by the markov lemma @xcite .
equivocation : @xmath168 therefore , @xmath169 is achievable .
this completes the achievability part .
let the output of the helper be @xmath170 , and the output of alice be @xmath6 , i.e. , @xmath171 first note that , for noiseless reconstruction of the sequence @xmath0 at the legitimate decoder , we have by fano s inequality @xmath172 we start by obtaining a lower bound on @xmath1 , the rate of alice , as follows @xmath173 where ( [ e1 ] ) follows by ( [ fano ] ) . in ( [ e2 ] ) , we have defined @xmath174 in ( [ etemp2 ] ) , we have defined , @xmath175 where @xmath176 is uniformly distributed on @xmath177 and is independent of all other random variables .
we now have the main step , i.e. , an upper bound on the equivocation rate of the eavesdropper , @xmath182 where ( [ e4 ] ) follows from the markov chain @xmath183 and ( [ e5 ] ) follows from ( [ fano ] ) .
this implies @xmath169 also note that the following is a markov chain , @xmath184 therefore , the joint distribution of the involved random variables is @xmath59 from support lemma @xcite , it can be shown that it suffices to consider such joint distributions for which @xmath60 . 1 . codebook generation at helen : from the conditional probability distribution @xmath150 compute @xmath151 .
generate @xmath152 codewords @xmath153 independently according to @xmath154 , where @xmath155 .
2 . codebook generation at alice : from the distribution @xmath186 , compute @xmath187 . generate @xmath188 sequences @xmath189 independently according to @xmath190 , where @xmath191 .
next , bin these sequences uniformly into @xmath192 bins .
+ also , randomly bin the @xmath156 sequences into @xmath193 bins and index these bins as @xmath194 .
3 . encoding at helen : on observing the sequence @xmath160 , helen tries to find a sequence @xmath153 such that @xmath161 are jointly typical . from rate - distortion theory , we know that there exists one such sequence .
helen sends the index @xmath163 of the sequence @xmath153 .
encoding at alice : the key difference from the one - sided helper case is in the encoding at alice . on observing the sequence @xmath156
, alice first finds the bin index @xmath164 in which the sequence @xmath156 falls .
alice also has the sequence @xmath153 received from helen .
alice next finds a sequence @xmath195 such that @xmath196 are jointly typical .
let the bin index of this resulting @xmath195 sequence be @xmath197 .
+ alice transmits the pair @xmath198 which is received by bob and eve .
the total rate used by alice is @xmath199 .
decoding at bob : on receiving the pair @xmath198 from alice and the index @xmath163 from helen , bob first searches the bin @xmath197 for a sequence @xmath200 such that @xmath201 are jointly typical .
this is possible since the number of @xmath195 sequences in each auxiliary bin is approximately @xmath202 which is @xmath203 and therefore with high probability , bob will be able to obtain the correct @xmath195 sequence . + using the estimate @xmath200 and @xmath153 , bob searches for a unique @xmath156 sequence in the bin @xmath164 such that @xmath204 are jointly typical .
this is possible since the number of @xmath156 sequences in each bin is approximately @xmath205 which is @xmath206 .
equivocation : @xmath207 where ( [ equi1 ] ) follows from the fact that @xmath164 is the bin index of the sequence @xmath0 , ( [ equi2 ] ) follows from the fact that conditioning reduces entropy , ( [ equi3 ] ) follows from the fact that @xmath208 , ( [ equi4 ] ) follows from the fact that @xmath197 is the bin index of the sequence @xmath209 , i.e. , @xmath210 forms a markov chain and subsequently using the data processing inequality . finally , ( [ equi6 ] ) follows from the fact that @xmath211 .
we therefore have @xmath212 is achievable .
this completes the achievability part for the case of two - sided helper .
the only difference in the converse part for the case of two - sided helper is when deriving an upper bound on the equivocation rate of the eavesdropper : @xmath213 where we have defined @xmath214 and @xmath215 we also have , @xmath216 this implies @xmath212 the joint distribution of the involved random variables is as follows , @xmath217 note that in the achievability proof of theorem @xmath11 , joint distributions of the following form are permitted , @xmath218 i.e , we have the markov chain , @xmath219 . with the definition of @xmath36 and @xmath34 as in ( [ defv])-([defu ] ) , these random variables do not satisfy this markov chain .
this implies that what we have shown so far is the following , @xmath220 where @xmath221 where the joint distribution of the involved random variables is as given in ( [ outerr ] ) . however , we observe that the term @xmath110 depends only on the marginal @xmath222 .
similarly , the terms @xmath223 and @xmath74 depend only on the marginal @xmath224 .
we use these observations to show that the region @xmath225 is the same when it is evaluated using the joint distributions of the form given in ( [ innerr ] ) .
this is clear by noting that once we are given a distribution of the form given in ( [ outerr ] ) , we can construct a new distribution of the form given in ( [ innerr ] ) with the same rate expressions .
consider any distribution @xmath226 of the form given in ( [ outerr ] ) .
using @xmath226 , compute the marginal @xmath222 as , @xmath227 we now construct a distribution @xmath228 as , @xmath229 such that the terms @xmath110 , @xmath223 and @xmath74 are the same whether they are evaluated according to @xmath226 or according to @xmath230 .
therefore , we conclude that it suffices to consider input distributions satisfying the markov chain @xmath231 when evaluating @xmath225 and hence @xmath232 .
this completes the converse part .
we remark here that this observation was useful in obtaining the converse for the rate - distortion function with common coded side information @xcite,@xcite,@xcite . 1 . codebook generation at helen : from the conditional probability distribution @xmath234 compute @xmath235 and @xmath236 .
generate @xmath237 sequences @xmath238 independently according to @xmath239 , where @xmath240 .
bin these sequences uniformly and independently in @xmath241 bins .
denote the bin index of the sequence @xmath238 as @xmath242 .
+ next generate @xmath243 sequences @xmath244 independently according to @xmath245 , where @xmath246 .
bin these sequences uniformly and independently in @xmath247 bins .
denote the bin index of the sequence @xmath244 as @xmath248 .
2 . codebook generation at alice : from the distribution @xmath249 , compute @xmath187 . generate @xmath250 sequences @xmath189 independently according to @xmath190 , where @xmath251 .
next , bin these sequences uniformly into @xmath252 bins .
+ also , randomly bin the @xmath156 sequences into @xmath253 bins and index these bins as @xmath254 .
3 . encoding at helen : on observing the sequence @xmath160 , helen tries to find a sequence @xmath238 such that @xmath255 are jointly typical . from rate - distortion theory , we know that there exists one such sequence .
helen sends the bin index @xmath242 of the sequence @xmath238 on the secure link which is received by alice and bob .
+ helen also finds a sequence @xmath244 such that @xmath256 are jointly typical . from rate - distortion theory
, we know that there exists one such sequence .
helen sends the bin index @xmath248 of the sequence @xmath244 on the insecure link which is received by alice , bob and eve .
4 . encoding at alice : on observing the sequence @xmath156 , alice first finds the bin index @xmath164 in which the sequence @xmath156 falls .
alice also receives the bin indices @xmath242 and @xmath248 from helen .
she first looks for a unique @xmath257 in the bin @xmath242 such that @xmath258 are jointly typical .
alice can estimate the correct sequence @xmath238 as long as the number of sequences in each bin is less than @xmath259 .
note from the codebook generation step at helen , that the number of @xmath260 sequences in each bin is approximately @xmath261 .
since @xmath262 forms a markov chain , we have @xmath263 and therefore the number of @xmath260 sequences in each bin is less than @xmath259 and consequently alice can estimate the correct sequence @xmath238 .
+ using @xmath156 and @xmath257 , alice looks for a unique @xmath264 in the bin @xmath248 such that @xmath265 are jointly typical .
alice can estimate the correct sequence @xmath244 as long as the number of sequences in each bin is less than @xmath266 .
the number of @xmath244 sequences in each bin is @xmath267 . from the markov chain @xmath268
, we have @xmath269 and therefore alice can correctly estimate the sequence @xmath244 . +
she next finds a sequence @xmath195 such that @xmath270 are jointly typical .
let the bin index of this resulting @xmath195 sequence be @xmath197 .
+ alice transmits the pair @xmath198 which is received by bob and eve .
the total rate used by alice is @xmath271 .
decoding at bob : on receiving the pair @xmath198 from alice and the bin indices @xmath242 and @xmath248 from helen , bob looks for a unique @xmath257 in the bin @xmath272 such that @xmath273 are jointly typical .
he can estimate the correct sequence @xmath238 with high probability since the number of @xmath260 sequences in each bin is approximately @xmath274 . using the estimate @xmath257 and @xmath275
, he looks for a unique @xmath264 in the bin @xmath248 such that @xmath276 are jointly typical . with high probability
, the correct sequence @xmath244 can be decoded by bob since the number of @xmath277 sequences in each bin is approximately @xmath278 . +
he next searches the bin @xmath197 for a sequence @xmath200 such that @xmath279 are jointly typical .
since the number of @xmath195 sequences in each auxiliary bin is approximately @xmath280 which is @xmath281 , with high probability , bob will be able to obtain the correct @xmath195 sequence .
+ using the estimates @xmath200 , @xmath257 , and @xmath264 , bob searches for a unique @xmath156 sequence in the bin @xmath164 such that @xmath282 are jointly typical .
this is possible since the number of @xmath156 sequences in each bin is approximately @xmath283 , i.e. , @xmath284 .
therefore , bob can correctly decode the @xmath156 sequence with high probability .
equivocation : @xmath285 @xmath286 where ( [ e12l ] ) follows from the fact that @xmath164 is the bin index of the sequence @xmath0 , ( [ e12ltem1 ] ) and ( [ e12ltem2 ] ) follow from the fact that conditioning reduces entropy , ( [ e22l ] ) follows from the fact that @xmath287 , ( [ e32l ] ) follows from the fact that @xmath197 is the bin index of the sequence @xmath209 and @xmath288 is the bin index of the sequence @xmath289 .
this completes the achievability part .
let the coded outputs of the helper be denoted as @xmath290 , where @xmath15 denotes the coded output of the secure link , @xmath16 denotes the coded output of the insecure link , and the output of alice be denoted as @xmath6 , i.e. , @xmath291 first note that , for noiseless reconstruction of the sequence @xmath0 at bob , we have by fano s inequality @xmath292 we start by obtaining a lower bound on @xmath1 , the rate of alice , as follows , @xmath293 @xmath294 where ( [ ex1 ] ) follows by ( [ fano2l ] ) and ( [ exx1 ] ) follows from the following markov chain , @xmath295 in ( [ ex2 ] ) , we have defined @xmath296 next , we provide a lower bound on @xmath14 , the rate of the insecure link , @xmath299 where ( [ 2l2 ] ) follows from the csiszar s sum lemma @xcite , and ( [ 2l3 ] ) follows from the following markov chain , @xmath300 we now have the main step , i.e. , an upper bound on the equivocation rate of the eavesdropper , @xmath301 where ( [ 2l4 ] ) follows from the markov chain @xmath302 and ( [ 2l5 ] ) follows by ( [ fano2l ] ) .
now , using csiszar s sum lemma @xcite , we have @xmath303 and we also have , @xmath304 where ( [ 2l7a ] ) follows from the memorylessness of the sources and ( [ 2l7b ] ) follows from the markov chain @xmath17 .
now , defining , @xmath305 and @xmath306 we have from ( [ 2l6 ] ) and ( [ 2l7b ] ) , @xmath307 now consider , @xmath308 we also have @xmath309 therefore , we have @xmath310 finally , on substituting ( [ 2l8 ] ) , ( [ 2l9 ] ) and ( [ 2l10 ] ) in ( [ 2l5 ] ) , we arrive at @xmath311 this implies @xmath312 also note that the following is a markov chain , @xmath313 therefore , the joint distribution of the involved random variables is @xmath314 on the other hand , the joint distribution of the involved random variables in the achievability proof of theorem @xmath12 is in the following form , @xmath315 i.e. , they satisfy the markov chain @xmath316 . now using the observation that @xmath317 depends on the marginal @xmath318 and @xmath319 depends on the marginal @xmath320 and using similar arguments used in the converse proof of theorem @xmath11
, it can be shown that it suffices to consider distributions of the form given in ( [ innner2s ] ) when evaluating our outer bound .
this completes the proof of the converse part .
|
we consider a secure lossless source coding problem with a rate - limited helper . in particular , alice observes an i.i.d .
source @xmath0 and wishes to transmit this source losslessly to bob at a rate @xmath1 .
a helper , say helen , observes a correlated source @xmath2 and transmits at a rate @xmath3 to bob .
a passive eavesdropper can observe the coded output of alice .
the equivocation @xmath4 is measured by the conditional entropy @xmath5 , where @xmath6 is the coded output of alice .
we first completely characterize the rate - equivocation region for this secure source coding model , where we show that slepian - wolf type coding is optimal .
we next study two generalizations of this model and provide single - letter characterizations for the respective rate - equivocation regions .
in particular , we first consider the case of a two - sided helper where alice also has access to the coded output of helen .
we show that for this case , slepian - wolf type coding is suboptimal and one can further decrease the information leakage to the eavesdropper by utilizing the side information at alice .
we finally generalize this result to the case when there are both secure and insecure rate - limited links from helen and additional uncoded side informations @xmath7 and @xmath8 are available at bob and eve , respectively . for this model
, we provide a complete characterization of the rate - equivocation region when @xmath9 forms a markov chain .
|
let @xmath2 denote a complete separable metric space , and let @xmath3 denote a borel probability measure on @xmath2 .
one way to measure the interplay between the metric @xmath4 and the measure @xmath3 is by means of an isoperimetric inequality . recall that minkowski s ( exterior ) boundary measure of a borel set @xmath5 , which we denote here by @xmath6 , is defined as @xmath7 , where @xmath8 denotes the @xmath9 extension of @xmath10 with respect to the metric @xmath4 .
the isoperimetric profile @xmath11 is defined as the pointwise maximal function @xmath12 \rightarrow { \mathbb{r}}_+$ ] , so that @xmath13 , for all borel sets @xmath5 .
an isoperimetric inequality measures the relation between the boundary measure and the measure of a set , by providing a lower bound on @xmath14 by some function @xmath15 \rightarrow { \mathbb{r}}_+$ ] which is not identically @xmath16 . since @xmath10 and @xmath17 will typically have the same boundary measure , it will be convenient to also define @xmath18 \rightarrow { \mathbb{r}}_+$ ] as @xmath19 . another way to measure the relation between @xmath4 and @xmath3
is given by concentration inequalities .
the log - concentration profile @xmath20 is defined as the pointwise maximal function @xmath21 such that @xmath22 for all borel sets @xmath5 with @xmath23 .
note that @xmath24 for all @xmath25 .
concentration inequalities measure how tightly the measure @xmath3 is concentrated around sets having measure @xmath26 as a function of the distance @xmath27 away from these sets , by providing a lower bound on @xmath28 by some non - decreasing function @xmath29 , so that @xmath30 tends to infinity .
the two main differences between isoperimetric and concentration inequalities are that the latter ones only measure the concentration around sets having measure @xmath26 , and do not provide any information for small distances @xmath27 ( smaller than @xmath31 ) .
we refer to @xcite for a wider exposition on these and related topics and for various applications .
it is known and easy to see that an isoperimetric inequality always implies a concentration inequality , simply by `` integrating '' along the isoperimetric differential inequality ( see section [ sec : pre ] ) .
in fact , it will be useful to also consider other _
intermediate _ levels between these two extremes , such as functional inequalities ( e.g. poincar , sobolev and log - sobolev inequalities ) and transport - entropy inequalities ( e.g. talagrand s @xmath32 inequality and its various variants ) . these will be introduced later on , but for now , let us just mention that it is known that these types of inequalities typically follow from appropriate isoperimetric inequalities , and imply appropriate concentration inequalities ( see @xcite or section [ sec : pre ] and the references therein ) .
schematically , this can represented in the following hierarchical diagram : @xmath33 all of the converse statements to the implications above are in general known to be false , due to the possible existence of narrow `` necks '' in the geometry of the space @xmath2 or the measures @xmath3 .
however , when such necks are ruled out by imposing some semi - convexity assumptions on the geometry and measure in the riemannian - manifold - with - density setting ( defined below ) , it was shown in our previous work @xcite that isoperimetric and concentration inequalities are in fact equivalent , with quantitative estimates which _ do not depend _ on the dimension of the underlying manifold ( see section [ sec : pre ] for a precise formulation ) . the main purpose of this work is to obtain several new applications of this equivalence between isoperimetry and concentration ( and hence of all the intermediate levels as well ) , which seem to have been previously inaccessible .
we will henceforth assume that @xmath34 is a smooth complete oriented connected @xmath35-dimensional ( @xmath36 ) riemannian manifold @xmath37 , that @xmath4 is the induced geodesic distance , and that @xmath3 is an absolutely continuous measure with respect to the riemannian volume form @xmath38 on @xmath39 .
[ def : ca ] we will say that our _ smooth @xmath40-semi - convexity assumptions _ are satisfied @xmath41 if @xmath3 is supported on a geodesically convex set @xmath42 , on which @xmath43 with @xmath44 , and as tensor fields on @xmath45 : @xmath46 we will say that our _ @xmath40-semi - convexity assumptions _ are satisfied if @xmath3 can be approximated in total - variation by measures @xmath47 so that each @xmath48 satisfies our smooth @xmath40-semi - convexity assumptions .
+ when @xmath49 , we will say in either case that our _ ( smooth ) convexity assumptions _ are satisfied .
here @xmath50 denotes the ricci curvature tensor of @xmath37 and @xmath51 denotes the second covariant derivative .
@xmath52 is the well - known bakry mery curvature tensor , introduced in @xcite ( in the more abstract framework of diffusion generators ) , which incorporates the curvature from both the geometry of @xmath37 and the measure @xmath3 .
when @xmath53 is sufficiently smooth and @xmath54 , our @xmath40-semi - convexity assumption is then precisely the curvature - dimension condition @xmath55 ( see @xcite ) .
an important example to keep in mind is that of euclidean space @xmath56 equipped with a probability measure @xmath57 with @xmath58
. one central theme in this work will be in deducing new stability results of isoperimetric and functional inequalities with respect to perturbations of the measure @xmath3 , in the presence of our semi - convexity assumptions . more precisely ,
if @xmath59 are two probability measures on @xmath37 such that @xmath60 is close to @xmath61 with respect to some ( not necessarily symmetric ) distance , we will show that under appropriate semi - convexity assumptions , @xmath62 inherits from @xmath63 a quantitatively comparable isoperimetric or functional inequality .
a - priori , it seems very difficult to analyze the stability of these questions directly ( at least for non - trivial distances ) , possibly due to the fact that in general no stability is possible and that some further weak convexity conditions need to be imposed .
our approach for obtaining stability results in such cases , is to decouple the stability question from the convexity assumptions .
we first pass from the isoperimetric or functional inequality to an appropriate concentration inequality ( this is always possible without any further assumptions ) ; the stability question on the level of concentration turns out to be elementary , and it is easy to obtain a concentration inequality for the perturbed measure ; lastly , we utilize our semi - convexity assumptions and pass back to the isoperimetric or functional level , by employing the equivalence between concentration and isoperimetry .
our results apply to several different notions of distance between @xmath61 ( the original measure ) and @xmath60 ( the perturbed measure ) : 1 .
@xmath64 - we show in section [ sec : stab1 ] that when @xmath60 satisfies our semi - convexity assumptions , and @xmath61 satisfies a strong enough isoperimetric or functional inequality , the latter property is inherited by @xmath60 .
we formulate some results specifically for the log - sobolev inequality , since it lies precisely on the border of our method , and since it is very useful in applications .
our results extend beyond the classical stability result for the log - sobolev inequality of holley and stroock @xcite ; in addition , under our convexity assumptions , we improve the classical quantitative dependence on @xmath65 from linear to logarithmic .
@xmath66 - we introduce a new metric @xmath67 between exponentially integrable probability measures , which we call the @xmath68-lipschitz metric .
we show in section [ sec : stab2 ] analogous stability results for this distance as for the first one , when @xmath60 satisfies our semi - convexity assumptions .
the main advantage of using this distance over the first one is that this result may be applicable even when the measures @xmath59 are mutually singular .
@xmath69 - using the usual @xmath1-wasserstein distance @xmath70 , we show in section [ sec : stab2 ] that under our convexity assumptions ( @xmath71 case ) , linear ( cheeger type ) isoperimetric inequalities or poincar inequalities are easily inherited .
this also applies when having control over the relative entropies : @xmath72 or @xmath73 . the importance of obtaining dimension - free log - sobolev and poincar inequalities in the context of statistical mechanics has been clarified in the works of stroock
zegarlinski @xcite , yoshida @xcite and bodineau helffer @xcite , who showed ( roughly speaking ) that these are equivalent to the decay of spin - spin correlations ( implying in particular the uniqueness of the corresponding gibbs measure in the thermodynamic limit ) .
the stability results described above are therefore relevant in understanding the effects of perturbing the hamiltonian potential in this context .
another application pertains to transport - entropy inequalities , first introduced by marton @xcite and developed by talagrand @xcite ( see section [ sec : pre ] ) . in general ,
given two convex functions @xmath74 , one may define a @xmath75 transport - entropy inequality as the following statement : @xmath76 where @xmath77 denotes the cost - function @xmath78 , @xmath79 is the wasserstein distance with cost - function @xmath80 , and @xmath81 denotes the relative entropy ( see section [ sec : pre ] for definitions ) . by jensen s inequality ,
it is immediate that @xmath82 implies the following ( in general , strictly ) weaker statement : @xmath83 it was shown by bobkov and gtze @xcite that the latter inequality is in fact equivalent to a variant of a concentration inequality .
we clarify this equivalence in proposition [ prop : con - te ] , which seems new and may be of independent interest , removing the inherent dimension dependence in previous results by djellout guillin
wu @xcite , bolley villani @xcite , and gozlan
leonard @xcite . using this characterization ,
we show that under our semi - convexity assumptions , it is possible in many cases to obtain a converse to jensen s inequality , up to dimension - independent constants , passing back from ( [ eq : intro - te - weak ] ) to ( [ eq : intro - te ] )
. it would be very hard to describe our results in more detail in any reasonably sized introduction , without first defining many necessary notions .
these are given in section [ sec : pre ] , which serves as an extended introduction to this work .
we describe the hierarchy ( [ eq : intro - hierarchy ] ) in detail , and formulate our main result from @xcite , asserting the equivalence of concentration and isoperimetric inequalities under the semi - convexity assumptions , which permits reversing the hierarchy and completing the equivalence .
the stability results with respect to perturbation of the measure are given in sections [ sec : stab1 ] ( for the distance @xmath84 ) and [ sec : stab2 ] ( for the @xmath68-lipschitz , wasserstein and relative entropy distances ) . in section
[ sec : inter ] , we set the ground for properly relating between transport - entropy and concentration inequalities ; parts of it may be of independent interest . in section [ sec : te - equiv ] , we briefly state the results on the equivalence of transport - entropy inequalities with different cost - functions . * acknowledgments .
* i would like to thank cedric villani , alexander kolesnikov and franck barthe for their interest and remarks on this work .
i also gratefully acknowledge the support of the institute of advanced study , where this work was initiated , and of the university of toronto ( and in particular , of robert mccann ) , where this work was concluded .
final thanks go out to the referee , for valuable and effective comments .
we reserve the use of @xmath85 etc . to indicate universal numeric constants , independent of all other parameters ( and in particular of the dimension of any underlying manifold ) ,
whose values may change from one occurrence to the next .
we also use the notation @xmath86 to signify that there exist constants @xmath87 so that @xmath88 , and that these constants do not depend on any other parameter , unless explicitly stated otherwise . when @xmath89 depend on some additional set of parameters @xmath45 , we may also use the notation @xmath90 .
let us start by recalling some of the notions mentioned in the introduction .
let @xmath91 denote the space of functions which are lipschitz on every ball in @xmath2 , and let @xmath92 .
functional inequalities compare between some type of expression measuring the @xmath3-averaged oscillation of @xmath93 , and an expression measuring the @xmath3-averaged magnitude of the gradient @xmath94 . in the general metric - space setting
, one may define @xmath95 as the following borel function : @xmath96 ( and we define it as 0 if @xmath97 is an isolated point - see @xcite for more details ) .
some well known examples of functional inequalities include the poincar and sobolev - gagliardo - nirenberg inequalities , but the one which will be of most interest to us in this work is the _ log - sobolev inequality _ , introduced by gross @xcite in the study of the gaussian measure , which corresponds to the case @xmath98 below .
the extension to the range @xmath99 $ ] is due to bobkov and zegarlinski @xcite .
@xmath100 satisfies a @xmath101-log - sobolev inequality ( @xmath99 $ ] ) if : @xmath102 where @xmath103 denotes the entropy of a non - negative function @xmath104 : @xmath105 the best possible constant @xmath65 above is denoted by @xmath106 .
another way to measure the interplay between the metric @xmath4 and the measure @xmath3 is given by transport - entropy ( or te ) inequalities , first introduced by marton @xcite , and significantly developed by talagrand @xcite .
these compare between the cost of optimally transporting between @xmath3 and a second probability measure @xmath107 ( with respect to some cost function @xmath108 ) , and the relative entropy of @xmath107 with respect to @xmath3 .
the transportation cost , or wasserstein distance , is defined as : @xmath109 where the infimum runs over the set @xmath110 of all probability measures @xmath111 on the product space @xmath112 with marginals @xmath107 and @xmath3 .
we reserve the notation @xmath113 to denote @xmath114 ( @xmath115 ) , which is known ( e.g. ( * ? ? ?
* theorem 6.9 ) ) to metrize the appropriate weak topology on the space of borel probability measures @xmath3 on @xmath2 having a finite @xmath116-th moment : @xmath117 . here
@xmath118 is some ( equivalently , any ) fixed point in @xmath34 .
the relative entropy , or kullback leibler divergence , is defined for @xmath119 as : @xmath120 and @xmath121 otherwise .
an important example of a transport - entropy inequality is given by talagrand s @xmath32 inequality , corresponding to the case @xmath122 below : @xmath100 satisfies a @xmath123 transport - entropy inequality ( @xmath124 ) if : @xmath125 the best possible constant @xmath65 above is denoted by @xmath126 .
the restriction to @xmath99 $ ] and @xmath127 above is necessary .
indeed , setting @xmath128 in ( [ eq : ls - inq - def ] ) and letting @xmath129 , one checks that the left - hand - side behaves like @xmath130 whereas the right - hand - side behaves like @xmath9 .
similarly , setting @xmath131 with @xmath132 in ( [ eq : te - inq - def ] ) , the right - hand - side behaves like @xmath133 whereas the left - hand - side behaves like @xmath9 .
it is however possible to extend these definitions to the range @xmath134 and @xmath135 $ ] by using appropriate modified log - sobolev and te inequalities with cost function modified to be quadratic for small distances , in the spirit of talagrand @xcite . to describe these variants ,
let us introduce the convex function @xmath136 , which is given by : @xmath137 \\ \frac{x^p}{p } + \frac{1}{2 } - \frac{1}{p } & x \in ( 1,\infty ) \end{cases } \;\;\ ; \text { if } p \in [ 1,2 ] ~.\ ] ] denoting by @xmath138 the legendre transform of @xmath139 , where @xmath140 , one checks that : @xmath141 \;\;\ ; \text{and } \;\;\ ; \varphi_{*,q}(\lambda ) : = \begin{cases } \frac{\lambda^2}{2 } & \lambda \in [ 0,1 ] \\
\frac{\lambda^q}{q } + \frac{1}{2 } - \frac{1}{q } & \lambda \in ( 1,\infty ) \end{cases } \;\;\ ; \text { if } q \in [ 2,\infty ] ~.\ ] ] @xmath100 satisfies a @xmath101-modified - log - sobolev inequality ( @xmath142 $ ] ) if : @xmath143 the best possible constant @xmath65 above is denoted by @xmath144 .
here @xmath145 denotes the class of functions @xmath93 such that @xmath146 , and @xmath147 should be understood as @xmath148 .
note that substituting @xmath149 above when @xmath99 $ ] , we see that @xmath101-modified and @xmath101-log - sobolev inequalities coincide , with @xmath150 .
the case @xmath151 ( @xmath152 ) was first introduced by bobkov and ledoux @xcite and further studied by bobkov gentil ledoux @xcite .
the extension to the entire range @xmath134 is due to gentil
guillin miclo @xcite .
@xmath100 satisfies a @xmath153 transport - entropy inequality ( @xmath115 ) if : @xmath154 where @xmath155 denotes the cost function @xmath156 .
the best possible constant @xmath65 above is denoted by @xmath157 .
again , in the case @xmath127 , these are just obviously identical to the usual @xmath158 transport - entropy inequalities with @xmath159 , so the novelty lies in the extension to the range @xmath135 $ ] .
the case @xmath152 was first introduced by talagrand @xcite in his study of the exponential measure on @xmath160 , and further characterized in @xcite ( see below ) .
the entire range @xmath135 $ ] has been subsequently considered by various authors ( see ( * ? ? ?
* chapter 22 ) and the references therein ) , and in particular by gentil guillin
miclo @xcite , who connected them to modified @xmath101-log - sobolev inequalities .
due to their non - homogeneous nature , various variants of @xmath153 te and @xmath101-modified - log - sobolev inequalities have been used in the literature , with corresponding constants @xmath161 and @xmath162 appearing in front of different terms .
we stand behind our convention , since all of the constants @xmath65 above and throughout this work scale linearly with the metric , i.e. @xmath163 . in any case , it is easy to modify these variants into our form , by using the following easy to verify properties of @xmath139 : @xmath164 & \rightarrow & c \varphi_p(x ) \geq \varphi_p(\min(c,1 ) x ) \;\;\ ; \forall x \geq 0 ~ ; \\ \label{eq : varphi*_q - prop } q \in [ 2,\infty ] & \rightarrow & c \varphi_{*,q}(\lambda ) \leq \varphi_{*,q}(\max(\sqrt{c},1 ) \lambda ) \;\;\ ; \forall \lambda \geq 0 ~.\end{aligned}\ ] ] before proceeding , we mention a useful characterization of the case @xmath152 ( @xmath151 ) obtained by bobkov gentil
ledoux @xcite : the @xmath165 te inequality is equivalent to the @xmath166-modified - log - sobolev inequality , which in turn is equivalent ( @xcite ) to the well - known poincar inequality : @xmath167 denoting the best possible constant @xmath65 above by @xmath168 , the equivalence is in the sense that : @xmath169 we refer to @xcite for a more precise statement .
as mentioned in the introduction , it is known that functional and transport - entropy inequalities may be used to interpolate between isoperimetric and concentration inequalities . to dispense of unneeded generality ,
let us illustrate this hierarchy in an important family of examples .
@xmath100 is said to satisfy a _
@xmath116-exponential isoperimetric inequality _
, @xmath170 , if : @xmath171 where @xmath172 denotes the probability measure on @xmath160 with density @xmath173 ( and @xmath174 is a normalization factor ) .
we denote the best constant @xmath65 above by @xmath175 .
the case @xmath176 corresponds to the standard gaussian measure on @xmath160 , and is called a _ gaussian isoperimetric inequality_. @xmath100 is said to satisfy a _
@xmath116-exponential concentration inequality _ ,
if : @xmath177 we denote the best constant @xmath65 above by @xmath178 .
the purpose of the @xmath179 above is to emphasize that this only provides information on the behaviour of @xmath28 in the large , and could be replaced by any constant strictly smaller than @xmath180 .
[ rem : iso_p ] it is known ( see @xcite,@xcite ) that given @xmath170 , @xmath181 uniformly on @xmath182 $ ] , with @xmath183 ( the lower bound on @xmath184 is universal , but the upper bound will depend on @xmath116 ) .
the space @xmath185 is a prototype for all of the inequalities mentioned above , and it is known that it satisfies @xmath186 for all @xmath170 .
as already mentioned , it is known and easy to see ( e.g. ( * ? ? ?
* proposition 1.7 ) ) that an isoperimetric inequality always implies a concentration inequality , simply by `` integrating '' along the isoperimetric differential inequality .
namely , if @xmath187 is a continuous function , then : @xmath188 \\ \downarrow \\ { \mathcal{k}}(r ) \geq \alpha(r ) \;\;\ ; \forall r \geq 0 \;\;\ ; \textrm{where } \;\ ;
\alpha^{-1}(x ) = \int_{\log 2}^x \frac{dy}{\gamma(y ) } ~. \end{array}\ ] ] it is immediate to check that ( [ eq : i - c ] ) yields the following implication ( @xmath189 ) : @xmath190 using the same notation as above , the following known series of implications clearly interpolates between these two extremes ( for @xmath191 and @xmath183 ) : @xmath192 more precisely , given @xmath193 , there exist constants @xmath194 so that : @xmath195 it is also possible to extend the above hierarchy to the range @xmath196 , but this is slightly less known and requires further explanation : @xmath197 more precisely , there exist constants @xmath194 so that for any @xmath135 $ ] ( and setting @xmath183 ) : @xmath198 the third implication in ( [ eq : hierarchy ] ) or ( [ eq : hierarchy-2 ] ) for general @xmath115 is due to marton @xcite . the fact that a @xmath101-modified - log - sobolev inequality implies @xmath116-exponential concentration follows from the herbst argument " ( see @xcite ) .
the second implication in ( [ eq : hierarchy ] ) or ( [ eq : hierarchy-2 ] ) for @xmath199 is due to otto and villani @xcite in the manifold setting .
this was subsequently given a different proof in the euclidean setting using the hamilton - jacobi equation by bobkov
ledoux @xcite , who also established the implication ( along with its converse ) for @xmath152 .
the implication was then extended to the range @xmath135 $ ] and @xmath191 in @xcite , @xcite , respectively .
see also ( * ? ? ?
* theorem 22.28),@xcite for generalizations , extensions to more general measure - metric spaces , and further techniques .
the first implication in ( [ eq : hierarchy ] ) or ( [ eq : hierarchy-2 ] ) for @xmath199 is due to m. ledoux @xcite , later refined by beckner ( see @xcite ) . in the general @xmath127 case ,
it is due to bobkov and zegarlinski @xcite ( see also @xcite ) . for @xmath152
, it follows from the characterization ( [ eq : bgl - equiv ] ) and cheeger s inequality @xcite , which implies that @xmath200 for some universal constant @xmath201 . in the general @xmath135 $ ] case ,
this implication is due to a. kolesnikov ( * ? ? ?
* theorem 1.1 ) , but this requires some further explanation
. kolesnikov showed ( in particular ) that given @xmath202 $ ] , if @xmath203 satisfies a @xmath116-exponential isoperimetric inequality ( with constant @xmath204 ) , then : @xmath205 with @xmath206 , where the constant @xmath65 depends on @xmath204 , @xmath116 and a lower bound on a variant of the poincar constant @xmath207 ( which is equivalent to it up to constants , see e.g. ( * ? ? ?
* lemma 2.1 ) ) . by cheeger s inequality as above
, we have @xmath208 for some universal constants @xmath209 , thus removing @xmath207 from the list of parameters .
a further careful inspection of the proof reveals that the estimate on @xmath65 does not depend on any integrability properties of the measure @xmath3 , and hence dimension independent .
the most delicate part is to notice that the estimate is actually uniform in @xmath202 $ ] , and hence extends to the case @xmath152 , if one is willing to use a different universal constant @xmath210 in ( [ eq : kol ] ) ; this may also be seen from the simpler tightening procedure suggested by barthe and kolesnikov ( * ? ? ?
* theorem 2.4 ) , using @xmath211 .
it follows that if @xmath212 , then ( [ eq : kol ] ) holds for some universal @xmath213 uniformly in @xmath135 $ ] . using ( [ eq : varphi*_q - prop ] )
, it follows that if @xmath212 then @xmath214 , for all @xmath135 $ ] .
but since our constants scale linearly in the metric , it must follow that @xmath215 , concluding our claim .
we also note that kolesnikov s method is not particular to euclidean space , and extends to the riemannian - manifold - with - density setting .
in general , it is known that it is not possible to reverse any of the implications in ( [ eq : hierarchy ] ) and ( [ eq : hierarchy-2 ] ) , at least not for general @xmath116 in the corresponding range .
that the first implication can not be reversed follows for instance from known criteria for hardy - type inequalities on @xmath216 @xcite . as for the second implication ,
this was settled by cattiaux and guillin @xcite in the case @xmath199 ; the extension to the case @xmath217 $ ] may be obtained by combining the results of barthe
roberto @xcite and gozlan @xcite , which respectively characterize in that range @xmath101-modified - log - sobolev inequalities and @xmath153 te inequalities on the real line . as already mentioned , in the case @xmath152 , @xmath166-modified - log - sobolev and @xmath165 te inequalities are actually known to be equivalent @xcite .
the third implication is certainly false in general .
indeed , note that any compactly supported measure always satisfies a concentration inequality , since @xmath218 for all @xmath219 , where @xmath220 denotes the diameter of the support . on the other hand ,
as remarked in @xcite , since @xmath153 te inequalities for @xmath135 $ ] imply a poincar inequality ( see section [ sec : te - equiv ] ) , it follows that when the support is in addition disconnected , the third implication can not be reversed ; a similar argument works for @xmath221 .
some partial reversal results have been obtained under some additional assumptions , typically involving convexity . under our convexity assumptions ,
it was shown by ledoux @xcite ( extending buser @xcite ) that a poincar inequality implies back a @xmath1-exponential ( or cheeger type ) isoperimetric inequality , up to universal constants . the semi - group method developed by ledoux and bakry
ledoux @xcite allowed reversing general functional inequalities with a @xmath222 term under our semi - convexity assumptions , and in particular applies to the log - sobolev inequality .
this method was extended to handle general @xmath223 terms in @xcite , and in particular applies to @xmath101-log - sobolev inequalities . under our semi - convexity assumptions ,
it was shown by otto
villani @xcite ( see also @xcite for a semi - group proof ) that a strong - enough @xmath224 te inequality implies back a log - sobolev inequality , with dimension independent estimates .
under our convexity assumptions , this was extended to the general @xmath135 $ ] case by gentil guillin miclo @xcite ; these authors comment that it would be possible to extend their result to handle the @xmath40-semi - convexity assumptions ( @xmath225 ) , but in view of the spirit of our results , we are not certain this is so .
this was also extended to the @xmath221 case by wang @xcite in the riemannian setting ( but with dimension dependent bounds ) , and by balogh et al .
@xcite in a more general one ( at least in the @xmath49 case , but their proof should generalize to arbitrary @xmath225 ) .
we do not know whether previous attempts have been considered to deduce isoperimetric , functional or te inequalities from concentration inequalities .
a weaker variant has been considered by many authors , including wang @xcite , chen
wang @xcite , bobkov @xcite , barthe @xcite , barthe kolesnikov @xcite , where an integrability condition of the measure , together with @xmath40-semi - convexity assumptions , was shown to guarantee an appropriate isoperimetric or functional inequality .
unfortunately , as explained in @xcite , these types of results will unavoidably always yield dimension - dependent estimates .
however , it was shown in our previous work @xcite using tools from riemannian geometry ( see also ledoux @xcite for an alternative approach ) , that under our semi - convexity assumptions , general concentration inequalities imply back their isoperimetric counterparts , with quantitative estimates which _ do not depend _ on the dimension of the underlying manifold .
this implies in particular that all of the tiers in our hierarchies are equivalent ( up to universal constants ) in this case .
the precise formulation is as follows : [ thm : main - equiv ] let @xmath226 and let @xmath29 denote an increasing continuous function so that : @xmath227 then under our @xmath40-semi - convexity assumptions , the concentration inequality : @xmath228 implies the following isoperimetric inequality : @xmath229 \;\;\ ; , \;\;\ ; \textrm{where } \;\ ; \gamma(x ) = \frac{x}{\alpha^{-1}(x ) } ~,\ ] ] and @xmath230 are constants depending solely on their arguments . moreover ,
if @xmath71 , we may take @xmath231 and @xmath232 for some universal constant @xmath201 . if @xmath225 , the dependence of @xmath233 on @xmath30 may be expressed only via @xmath234 and @xmath235 .
in this work , we focus on some of the consequences of theorem [ thm : main - equiv ] to the study of isoperimetric , functional and transport - entropy inequalities .
the three main motives appearing in this work are : * under our convexity assumptions , both hierarchies ( [ eq : hierarchy ] ) and ( [ eq : hierarchy-2 ] ) may be reversed . * under our @xmath40-semi - convexity assumptions , the hierarchy ( [ eq : hierarchy ] ) for @xmath236 may be reversed .
* under our @xmath40-semi - convexity assumptions , the hierarchy ( [ eq : hierarchy ] ) for @xmath176 may be reversed , if a strong - enough concentration inequality is satisfied . by going up and down the hierarchies
, we deduce the various announced results .
we start by deducing several stability results with respect to a rather restrictive notion of proximity of @xmath60 to @xmath61 , given by @xmath237 ( assuming that @xmath238 ) .
the main observation behind the contents of this section is the following elementary : [ lem : conc - going - down ] let @xmath61,@xmath60 denote two probability measures on a common metric space @xmath2 , and assume that : @xmath239 let @xmath240 denote the corresponding log - concentration profiles , and assume that @xmath241 where @xmath242 is strictly increasing and continuous .
then : @xmath243 recall that @xmath241 just means that : @xmath244 which is easily seen to be equivalent to : @xmath245 now let @xmath10 denote a borel subset of @xmath34 with @xmath246 .
the condition ( [ eq : going - down ] ) implies that @xmath247 .
it follows from ( [ eq : conc - reversed ] ) and the definition of @xmath248 that @xmath249 . applying ( [ eq :
going - down ] ) and ( [ eq : conc ] ) once again , we deduce that : @xmath250 recalling the definition of @xmath251 , the desired assertion follows .
[ thm : iso - stability ] let @xmath61,@xmath60 denote two probability measures on a common riemannian manifold @xmath37 so that @xmath238 , and assume that ( [ eq : going - down ] ) holds .
assume that our @xmath40-semi - convexity assumptions are satisfied for @xmath62 ( @xmath226 ) , and let @xmath252 denote a continuous positive function so that : @xmath253 let @xmath254 denote the corresponding isoperimetric profiles . if @xmath63 satisfies the isoperimetric inequality : @xmath255 ~,\ ] ] then @xmath62 satisfies the following isoperimetric inequality : @xmath256 ~,\ ] ] where @xmath257 is defined as : @xmath258 and @xmath259 depend solely on their arguments .
it is obvious that the assertion of the theorem is completely false without the semi - convexity assumption on @xmath62 ( e.g. consider the case that the support of @xmath60 is disconnected ) . using the same notations as in lemma [ lem : conc - going - down ] ,
if follows from ( [ eq : i - c ] ) that : @xmath260 the growth condition ( [ eq : gamma - growth ] ) on @xmath261 ensures that @xmath262 satisfies the following growth condition : @xmath263 applying lemma [ lem : conc - going - down ] , we deduce from ( [ eq : going - down ] ) that : @xmath264 where : @xmath265 clearly @xmath266 inherits the growth condition ( [ eq : alpha_1-growth ] ) from @xmath262 : @xmath267 since our @xmath40-semi - convexity assumptions are satisfied for @xmath62 and since @xmath268 , we may apply theorem [ thm : main - equiv ] and deduce the isoperimetric inequality : @xmath269 ~,\ ] ] where : @xmath270 since @xmath271 for @xmath272 , the proof is complete .
[ cor : iso - stability ] under measure perturbations of the form @xmath273 and our @xmath40-semi - convexity assumptions on @xmath62 ( @xmath226 ) : 1 . if @xmath49 , then @xmath116-exponential isoperimetric inequalities for @xmath170 are stable under perturbation : @xmath274 2 . if @xmath225 , then @xmath116-exponential isoperimetric inequalities for @xmath275 are stable under perturbation : @xmath276 3 . if @xmath225 , then a strong - enough gaussian isoperimetric inequality ( @xmath176 case ) is also stable under perturbation : @xmath277 here @xmath278 is a constant depending solely on its arguments , and @xmath279 are functions depending solely on their arguments , which in addition are strictly positive as soon as @xmath280 is .
recalling remark [ rem : iso_p ] , by definition of a @xmath116-exponential isoperimetric inequality we have : @xmath281 ~,\ ] ] for some universal constant @xmath201 and @xmath183 .
the first two claims then easily follow from theorem [ thm : iso - stability ] applied to @xmath282 , after noting that : @xmath283 note that in the first case ( @xmath49 ) , the dependence on @xmath284 may be shown to be linear , due to the remarks at the end of the formulation of theorem [ thm : main - equiv ] .
the third claim follows similarly , once it is checked ( see e.g. @xcite ) that when @xmath176 : @xmath285 where @xmath286 denotes the standard gaussian measure on @xmath160 .
this implies that if @xmath287 , then setting @xmath288 , the condition ( [ eq : gamma - growth ] ) is satisfied for some big enough @xmath118 .
this completes the proof .
the case @xmath49 and @xmath152 of corollary [ cor : iso - stability ] was also deduced in our previous work @xcite . in that work
, it was shown that one may use : @xmath289 and that up to universal constants , this result is sharp .
indeed , ( [ eq : c_1,d ] ) may also be seen from ( [ eq : good - estimate ] ) in the proof of theorem [ thm : iso - stability ] .
in fact , this can be extended to the following estimate : @xmath290 it is also possible to derive an analogue of lemma [ lem : conc - going - down ] for the case when the roles of @xmath59 are interchanged , so that condition ( [ eq : going - down ] ) is replaced by @xmath291 , when @xmath292 . unfortunately , in this case , we can only deduce a lower bound on @xmath251 which will always be smaller than @xmath293 , and in particular , we can not deduce that @xmath251 increases to infinity . repeating the arguments of theorem [ thm : iso - stability ] , this would only allow us to deduce an isoperimetric inequality of the form @xmath294 in the range @xmath295 $ ] , for some @xmath296 sufficiently close to @xmath26 .
this is not very useful in general , except in the case that our convexity assumptions are satisfied ( @xmath49 ) , where this is already enough to imply a _ linear _ isoperimetric inequality - see @xcite and subsection [ subsec : stab2-w1 ] . although it is possible to formulate an analogue to theorem [ thm : iso - stability ] in the language of more general functional and transport - entropy inequalities , we prefer to restrict ourselves in this subsection to log - sobolev inequalities , since these lie on the border of our method and are the most interesting in applications . before stating our result ,
let us first recall the following well - known stability result due to holley and stroock @xcite : let @xmath297 ( @xmath298 ) denote two probability measures on a common riemannian manifold @xmath37 , so that : @xmath299 if @xmath63 satisfies a log - sobolev inequality : @xmath300 then @xmath62 also satisfies a log - sobolev inequality : @xmath301 this result was obtained in the context of statistical mechanics , where @xmath302 represent some hamiltonian potentials , and the condition ( [ eq : hs - two - sided ] ) is interpreted as the assumption that @xmath303 is a bounded perturbation of @xmath304 .
although it is quite useful in various situations in this context , the condition ( [ eq : hs - two - sided ] ) is unavoidably restrictive , since it is easy to check that there can be no stability in general without assuming both the upper and lower bounds on the perturbation . the following result , on the other hand , permits to dispose of the upper bound in ( [ eq : hs - two - sided ] ) , at the expense of an additional semi - convexity assumption ; in addition , under our convexity assumptions , the quantitative dependence on the perturbation parameter is improved from exponential to linear : [ thm : log - sob - stability ] let @xmath297 ( @xmath298 ) denote two probability measures on a common riemannian manifold @xmath37 .
assume that : @xmath305 with some @xmath226 . if @xmath63 satisfies a strong - enough log - sobolev inequality : @xmath306 then @xmath62 satisfies a log - sobolev inequality : @xmath307 where @xmath308 depends solely on its arguments .
moreover , when @xmath71 , one may use : @xmath309 where @xmath201 is a universal constant . as usual , we use the same notations as in lemma [ lem : conc - going - down ] . by the herbst argument ( see @xcite ) , which was already mentioned in section [ sec : pre ] , it is known that the log - sobolev inequality ( [ eq : ls - assump ] ) implies the following laplace - functional inequality : @xmath310 it is easy to check ( see e.g. lemma [ lem : conc - laplace ] ) that this implies the following concentration inequality on @xmath63 : @xmath311 using lemma [ lem : conc - going - down ] we deduce that : @xmath312 since @xmath313 , @xmath251 clearly satisfies the growth condition required to apply theorem [ thm : main - equiv ] : @xmath314 consequently , theorem [ thm : main - equiv ] implies that the following isoperimetric inequality is satisfied : @xmath315 ~.\end{aligned}\ ] ] this means that @xmath62 satisfies a gaussian ( or 2-exponential ) isoperimetric inequality in the notation of section [ sec : pre ] . as described there , it is known that this implies the log - sobolev inequality ( [ eq : ls - concl ] ) , concluding the proof . note that when @xmath71 , the remarks at the end of the formulation of theorem [ thm : main - equiv ] imply that one may use @xmath316 above , with @xmath201 a universal constant , implying the last assertion of the theorem .
a natural situation where only the lower bound in ( [ eq : hs - two - sided ] ) is available , is when @xmath60 is the restriction of @xmath61 onto some `` event '' having positive probability .
we state this explicitly as an immediate corollary of theorem [ thm : log - sob - stability ] . to further elucidate this scenario ,
we restrict ourselves to the case @xmath49 in the euclidean setting , although it is of course possible to formulate the following more generally : [ cor : log - sob ] let @xmath317 denote a probability measure on @xmath318 .
let @xmath319 be such that : @xmath320 set @xmath321 . if @xmath322 satisfies a log - sobolev inequality : @xmath300 then @xmath323 satisfies a log - sobolev inequality : @xmath324 where @xmath325 is a universal constant .
analogous stability results may be obtained for single - sided perturbations of other functional or te inequalities , under our semi - convexity assumptions .
we mention here an analogue of the holley stroock two - sided perturbation lemma , only recently obtained by gozlan , roberto and samson @xcite , for talagrand s @xmath224 transport - entropy inequality ( and more general ones ) , which yields the same exponential dependence on @xmath326 and @xmath327 . as for the log - sobolev inequality
, this may be substantially improved under our semi - convexity assumptions .
in this section , we set the ground for the next sections , which deal with transport - entropy inequalities . besides recalling known results , we show a complete equivalence between concentration and certain transport - entropy inequalities , which may be of independent interest .
we will say that @xmath100 satisfies a ( weak ) laplace - functional inequality if there exists @xmath328 , @xmath329 and an increasing convex function @xmath330 with @xmath331 , so that : @xmath332 where @xmath333 denotes the legendre transform of @xmath334
. recall that by the monge - kantorovich - rubinstein dual characterization of @xmath70 ( e.g. @xcite ) , we have that : @xmath335 this characterization is the key ingredient in the following ( mild adaptation of a ) theorem of bobkov and gtze @xcite ( see also @xcite ) : [ thm : bg ] let @xmath330 denote an increasing convex function so that @xmath331 . then for any @xmath336 , the following weak transport - entropy inequality : @xmath337 is equivalent to the weak laplace - functional inequality ( [ eq : trunc - laplace ] ) . by the dual characterization ( [ eq : mkr - duality ] ) of @xmath70 , the definitions of @xmath338 and @xmath81 , and denoting @xmath339 , ( [ eq : lem - trunc - tc ] ) is equivalent to the statement that : @xmath340 for any @xmath1-lipschitz function @xmath93 and non - negative @xmath3-integrable function @xmath341 so that @xmath342 . denoting : @xmath343 we see that @xmath344 , for all @xmath341 as above .
this is well known to be equivalent to @xmath345 , which is equivalent to ( [ eq : trunc - laplace ] ) .
this gives rise to the following : we will say that @xmath100 satisfies a weak @xmath346 transport - entropy inequality ( @xmath115 ) if : @xmath347 the best constant @xmath65 above will be denoted by @xmath348 .
it is well known that the weak laplace - functional inequality ( [ eq : trunc - laplace ] ) is equivalent to our usual notion of concentration inequality .
this is made precise in the following : [ lem : conc - laplace ] let @xmath330 denote an increasing convex function so that @xmath331 . 1 .
if @xmath349 then ( [ eq : trunc - laplace ] ) implies : @xmath350 with @xmath351 , @xmath352 and @xmath353 .
2 . if @xmath354 and @xmath355 then for any @xmath356 , ( [ eq : trunc - conc ] ) implies ( [ eq : trunc - laplace ] ) with @xmath357 , @xmath358 and @xmath359 . we will show statement ( 2 ) , statement ( 1 ) is simpler and follows along the same lines .
it is immediate that ( [ eq : trunc - conc ] ) is equivalent to : @xmath360 where @xmath361 denotes a median of @xmath93 with respect to @xmath3 , i.e. a value so that @xmath362 and @xmath363
. now let @xmath93 denote a 1-lipschitz function with @xmath364 . using ( [ eq : med - bound ] ) to evaluate : @xmath365 with @xmath366 , one checks using ( [ eq : med - bound ] ) again that : @xmath367 where @xmath368 . integrating by parts again and using ( [ eq : exp - bound ] ) ,
we evaluate : @xmath369 using that @xmath370 for all @xmath371 , it follows that : @xmath372 from which point it is immediate to verify the claim .
[ rem : conc - laplace ] note that since ( [ eq : trunc - conc ] ) is only meaningful whenever the right - hand side exceeds @xmath180 , we can always change the values of @xmath373 to @xmath374 so that @xmath375 and @xmath376 are arbitrary , as long as one of these inequalities is strict . since @xmath377 , the same applies to ( [ eq : lem - trunc - tc ] ) and changing the values of @xmath378 to @xmath379 so that @xmath380 and @xmath381 are arbitrary , as long as one of these inequalities is strict . in either case , @xmath382 or @xmath383
are determined in a manner depending on all the other parameters and in addition on @xmath384 .
combining theorem [ thm : bg ] with the equivalence between weak laplace - functional and concentration inequalities given by lemma [ lem : conc - laplace ] and remark [ rem : conc - laplace ] , it is easy to check that weak @xmath346 transport - entropy and @xmath116-exponential concentration inequalities are precisely equivalent : [ cor : conc - weak - te ] @xmath385 uniformly in @xmath115 .
we will also need to use a tight " form of our weak @xmath346 transport - entropy inequalities for some of the results in the next sections .
this is summarized in the following proposition , which extends beyond the previously mentioned results in this section , and which may be of independent interest : [ prop : con - te ] the following inequalities are equivalent : 1 .
the @xmath116-exponential concentration inequality : @xmath386 2 .
the weak @xmath346 transport - entropy inequality : @xmath387 3 .
the @xmath388 transport - entropy inequality : @xmath389 where , recall , @xmath139 is given by ( [ eq : def - varphi ] ) .
the equivalence is in the sense that the best constants above satisfy @xmath390 uniformly in @xmath115 . by corollary [ cor :
conc - weak - te ] , @xmath385 uniformly in @xmath115 , so it remains to prove that @xmath391 uniformly .
it will be convenient to slightly change our normalization , so we remark that @xmath392 is clearly the best constant in the following inequality : @xmath393 since @xmath394 , it is immediate that @xmath395 .
the other direction is the tricky part .
we will assume that @xmath396 , the case @xmath152 follows by approximation . by theorem [ thm : bg ] ,
( [ eq : lem - wte-p ] ) is equivalent to the statement that : @xmath397 to conclude , we will need to deduce from this that for some universal constant @xmath201 : @xmath398 which will imply that @xmath399 by theorem [ thm : bg ] ( recall that @xmath400 is given by ( [ eq : def - varphi * ] ) ) . by taking @xmath201 smaller than some universal constant @xmath401 ,
it is easy to check that ( [ eq : prop - assumption ] ) implies ( [ eq : prop - conclusion ] ) for @xmath402 , so it remains to check ( [ eq : prop - conclusion ] ) in the range @xmath403 $ ] .
fix any @xmath1-lipschitz function @xmath93 so that @xmath364 .
we proceed by denoting @xmath404 , the logarithm of the laplace transform .
note that @xmath405 by jensen s inequality . further denoting @xmath406
, it is immediate to check that : @xmath407 by the cauchy - schwarz inequality , it follows that : @xmath408 hence , using ( [ eq : prop - assumption ] ) with @xmath409 and a standard application of the markov chebyshev inequality and integration by parts , we conclude that @xmath410 whenever @xmath411 , for some universal constant @xmath412 . since @xmath413 and @xmath414
, we conclude that @xmath415 for @xmath416 $ ] . denoting @xmath417
, this implies that : @xmath418 ~.\ ] ] this confirms the validity of ( [ eq : prop - conclusion ] ) in the range @xmath403 $ ] and concludes the proof .
a previous characterization of @xmath419 transport - entropy inequalities was obtained by djellout , guillin and wu @xcite , and strengthened by bolley and villani @xcite .
this was generalized to more general @xmath420 te inequalities by gozlan and leonard @xcite .
all of these characterizations were in terms of an integrability condition of the form @xmath421 for some ( equivalently , all ) @xmath422 .
the problem with these criteria is that they all inevitably result in bad quantitative dependence when trying to estimate @xmath423 via @xmath424 and @xmath425 ; in particular , when the underlying space is an @xmath35-dimensional manifold , they will all result in dimension dependent bounds ( see @xcite ) .
proposition [ prop : con - te ] ( which clearly extends to more general @xmath420 te inequalities ) demonstrates that the right characterization is via concentration inequalities ( as opposed to integrability criteria ) .
in addition , all of the above mentioned criteria may be easily recovered from it , since it is easy to check ( see e.g. ( * ? ? ?
* section 7.2 ) ) that : @xmath426 [ rem : no - tight ] by checking what happens for measures with compact yet disconnected support , it is not difficult to realize that the equivalence between the weak and tight transport - entropy inequalities is rather special to the @xmath70 distance , and that analogous results can not hold in general for @xmath113 , @xmath427 .
however , under our semi - convexity assumptions , it is in fact possible to tighten these weak te inequalities , by passing through the appropriate concentration inequality and employing theorem [ thm : main - equiv ] .
a drawback of using a distance of the form @xmath84 ( or @xmath428 ) to measure the extent of a perturbation when analyzing the stability of various inequalities , as in section [ sec : stab1 ] , is that the estimates become meaningless when the measures @xmath59 have disjoint supports , or more generally , are mutually singular . in this section ,
we analyze the stability with respect to several new distances , which provide further flexibility and generalize some of the previous results : * the wasserstein distance @xmath429 and consequently the relative entropies @xmath430 and @xmath431 . * a new distance @xmath432 which we introduce , called the @xmath68-lipschitz metric .
let @xmath433 denote the space of probability measures @xmath3 satisfying that @xmath434 for any @xmath435 and some ( any ) @xmath422 . on this space
, we introduce the following distance : @xmath436 it is clear that @xmath67 satisfies the triangle inequality and that it is symmetric .
it is also easy to see that @xmath437 if and only if @xmath438 , for instance by using the hahn - banach theorem together with the stone - weierstrass theorem ( in its lattice version ) and the fact that functions of the form @xmath439 as above separate points in a metric space .
we consequently verify that @xmath67 is a metric on @xmath433 , which we call the @xmath68-lipschitz metric .
another immediate property using the monge - kantorovich - rubinstein dual characterization ( [ eq : mkr - duality ] ) of @xmath70 , is that : @xmath440 indeed , this is seen by testing in ( [ eq : def - wpsi1 ] ) functions @xmath104 of the form @xmath441 , where @xmath93 is an arbitrary 1-lipschitz function , and taking the limit as @xmath9 tends to @xmath16 .
we comment that , in analogy to the dual characterization ( [ eq : mkr - duality ] ) of @xmath70 , this metric may have some relation to the more standard @xmath68-wasserstein metric : @xmath442 but we have not been able to make this relation precise .
the key reason to use the @xmath67 metric is the following lemma , which asserts that ( weak ) laplace - functional inequalities are stable under perturbations in this metric : let @xmath59 denote two borel probability measures on a common metric space @xmath2 .
assume that @xmath443 satisfies the following ( weak ) laplace - functional inequality : @xmath444 where @xmath330 is an arbitrary function .
then @xmath445 satisfies the following weak laplace - functional inequality : @xmath446 by definition , we have that : @xmath447 if @xmath448 for a @xmath1-lipschitz function @xmath93 , we know by ( [ eq : mkr - duality ] ) and ( [ eq:1-psi_1 ] ) that @xmath449 ,
so we can bound the first term on the right hand side above by : @xmath450 this completes the proof .
since weak laplace - functional inequalities are equivalent to concentration inequalities by lemma [ lem : conc - laplace ] , we deduce that concentration inequalities are also stable under @xmath67 perturbation . hence , as in the previous section , this stability may be transferred to the level of isoperimetric and functional inequalities , once our semi - convexity assumptions on @xmath445 are satisfied .
the formulations are completely analogous to those of theorem [ thm : iso - stability ] , corollary [ cor : iso - stability ] and theorem [ thm : log - sob - stability ] , so we do not explicitly state them here , and leave this to the interested reader . for the results of this subsection , let us recall some further notation and results from @xcite . given a measure - metric space @xmath100 , we denote by @xmath451 the best constant @xmath65 in the following _ first - moment inequality _ : @xmath452 where as usual , @xmath361 denotes a median of @xmath93 with respect to @xmath3 .
recall from section [ sec : pre ] that @xmath453 is the _
@xmath116-exponential concentration constant _ , i.e. the best constant @xmath65 so that : @xmath454 recall also that @xmath455 denotes the poincar constant , and that @xmath456 denotes the _ exponential isoperimetric constant _ , which by remark [ rem : iso_p ] is equivalent ( up to universal constants ) to the _ cheeger constant _ , i.e. the best constant @xmath65 in the following _ linear isoperimetric inequality _ : @xmath457 ~.\ ] ] the following theorem was proved in @xcite ( see also @xcite for a slightly stronger statement and simplified proof ) : [ thm : main - roc ] if @xmath100 satisfies our convexity assumptions then : @xmath458 where @xmath459 are some numeric constants .
[ rem : main - roc - trivial ] all of the above inequalities except for the first hold without any additional convexity assumptions : the second is trivial and the others follow from ( [ eq : hierarchy-2 ] ) and the subsequent comments .
the inequality @xmath460 under our convexity assumptions also follows from theorem [ thm : main - equiv ] .
theorem [ thm : main - roc ] was used in @xcite to obtain stability results for @xmath461 when @xmath3 undergoes a perturbation so that our convexity assumptions are preserved .
the control on the perturbation was measured in terms of control over some notion of distance between the original measure @xmath61 and the perturbed measure @xmath60 .
three distances were analyzed : @xmath462 , @xmath463 and the total variation distance @xmath464 .
although some essentially sharp estimates were obtained in @xcite , there is a certain drawback in using any of the above distances , which was already mentioned in the beginning of this section . in this subsection
, we analyze the stability with respect to two new distances : the wasserstein distance @xmath429 and the relative entropies @xmath430 and @xmath431 , which provide further flexibility over the previous distances .
the idea is based on the following elementary : [ lem : w1-fm ] let @xmath61,@xmath60 denote two probability measures on a common metric space @xmath2 .
then : @xmath465 let @xmath466 denote a @xmath1-lipschitz function . then : @xmath467 where we have used the dual characterization ( [ eq : mkr - duality ] ) of @xmath70 in the last inequality .
taking supremum on @xmath93 as above , and exchanging the roles of @xmath61 and @xmath60 , the assertion immediately follows .
using theorem [ thm : main - roc ] ( and remark [ rem : main - roc - trivial ] ) , we immediately deduce the following stability result with respect to the @xmath1-wasserstein distance : [ thm : w1-stability ] if @xmath445 satisfies our convexity assumptions then : @xmath468 where @xmath469 are some numeric constants . exchanging the roles of @xmath61 and @xmath60 ,
we conclude : if @xmath443 and @xmath445 satisfy our convexity assumptions and @xmath470 , then @xmath471 , where the constants implied by @xmath472 depend linearly on @xmath473 . in practice ,
it is convenient to estimate @xmath429 by using the relative entropies @xmath431 or @xmath430 .
these two possibilities turn out to be rather different . by corollary [ cor :
conc - weak - te ] , there exists a universal constant @xmath201 so that for every @xmath115 : @xmath474 plugging this into the estimate ( [ eq : con-1 ] ) of theorem [ thm : w1-stability ] , and using the obvious fact that @xmath475 for @xmath115 , we obtain : [ thm : going - down - h ] if @xmath445 satisfies our convexity assumptions then : @xmath476 where @xmath477 are some numeric constants .
moreover , for any @xmath189 : @xmath478 a different estimate is obtained from lemma [ lem : w1-fm ] and theorem [ thm : main - roc ] by proceeding as above but reversing the roles of @xmath61 and @xmath60 .
this time , we can not use the weak @xmath479 transport - entropy inequality given by ( [ eq : wtc - p ] ) as in the proof of theorem [ thm : going - down - h ] ( the reader may want to check this ) , so we employ its tight equivalent form given by proposition [ prop : con - te ] : @xmath480 where , recall , @xmath481 is given by ( [ eq : def - varphi ] ) .
[ thm : going - up - h ] there exists a universal constant @xmath201 so that if @xmath482 and @xmath445 satisfies our convexity assumptions , then : @xmath483 where @xmath484 are some other universal constants .
the first and last inequalities follow from theorem [ thm : main - roc ] and remark [ rem : main - roc - trivial ] .
to deduce the middle inequality , we use lemma [ lem : w1-fm ] and ( [ eq : tc - ql ] ) : @xmath485 multiplying by @xmath486 and using theorem [ thm : main - roc ] , we conclude that : @xmath487 for some universal constant @xmath488 .
the assertion now clearly follows . a corollary which summarizes the resulting stability is : if @xmath443 and @xmath445 satisfy our convexity assumptions then : @xmath489 where @xmath490 are numeric constants .
in particular : * if @xmath491 then @xmath492 . * if @xmath493 then @xmath494 . note that when @xmath495 , the results of theorems [ thm : going - down - h ] and [ thm : going - up - h ] may be recovered from our previous results from @xcite , where the total - variation distance @xmath496 was employed .
this follows from the well - known pinsker - csizsar - kullback inequality ( * ? ? ?
* chapter 6 ) : @xmath497 analogous stability results to those in this subsection were shown in @xcite when @xmath498 , for @xmath499 .
since always @xmath500 , ( [ eq : pinsker ] ) suggests that when @xmath501 , we can not formally obtain the results in this subsection from the previous ones in @xcite .
all the results in this subsection regarding @xmath502 also apply to the poincar constant @xmath503 , since these two are equivalent under our convexity assumptions ( see theorem [ thm : main - roc ] or subsection [ subsec : pre - reverse ] ) .
denote by @xmath504 the best possible constant in the following @xmath75 te inequality : @xmath505 where as usual @xmath77 denotes the cost - function @xmath78 . when @xmath425 ( @xmath53 ) is the function @xmath506
, we simply write @xmath507 ( @xmath508 ) to be consistent with previous notation . using jensen s inequality and the fact that @xmath509 is convex when @xmath510 , it is immediate to check that : @xmath511 in view of the equivalence @xmath512 stated in ( [ eq : bgl - equiv ] ) and proposition [ prop : con - te ] , the inequality between first and last terms in ( [ eq : te - jensen ] ) for @xmath152 should be interpreted as a trivial proof of the well - known implication , due to gromov and v. milman @xcite , that a poincar inequality always implies exponential concentration . in general
, this shows that a @xmath153 transport - entropy inequality implies @xmath116-exponential concentration ( without relying on marton s method ) .
similarly , when @xmath513 , ( [ eq : te - jensen ] ) should be interpreted as a trivial proof of the known implication that a @xmath153 transport - entropy inequality ( @xmath202 $ ] ) implies the poincar inequality .
this was shown in the case @xmath176 by otto
villani @xcite , for @xmath514 by bobkov
gentil ledoux @xcite , and for @xmath135 $ ] by gentil
guillin miclo @xcite ; for further generalizations , see ( * ? ? ?
* theorem 22.28 ) .
we conclude that transport - entropy inequalities provide a certain framework for deducing concentration and other inequalities , just by employing elementary tools such as jensen s inequality . as a further application of theorem [ thm : main - equiv ] ,
we demonstrate in this section that under our various semi - convexity assumptions , reverse inequalities may be obtained , implying that te inequalities with different cost - functions are in fact equivalent ( up to universal constants ) under these assumptions .
our procedure should by now be self - evident , so we omit the proofs .
[ thm : te - equiv-2 ] assume that our convexity assumptions are satisfied ( @xmath49 ) for @xmath515 . then there exists a universal constant @xmath201 , so that for any @xmath135 $ ] , all of the constants in ( [ eq : te - jensen ] ) are equivalent for all @xmath516 $ ] : @xmath517 1 .
if @xmath49 , then @xmath123 transport - entropy inequalities are equivalent for @xmath516 $ ] : @xmath519 2 . if @xmath225 and @xmath236 , then @xmath123 transport - entropy inequalities are equivalent for @xmath516 $ ] : @xmath520 3 . if @xmath225 , then a strong - enough @xmath419 transport - entropy inequality implies a @xmath224 transport - entropy inequality : @xmath521 t. bodineau and b. helffer .
correlations , spectral gap and log - sobolev inequalities for unbounded spins systems . in _
differential equations and mathematical physics ( birmingham , al , 1999 ) _ , volume 16 of _ ams / ip stud . adv . math .
_ , pages 5166 .
soc . , providence , ri , 2000 .
j. cheeger .
a lower bound for the smallest eigenvalue of the laplacian . in _ problems in analysis ( papers dedicated to salomon bochner , 1969 ) _ , pages 195199 .
princeton univ . press , princeton ,
n. j. , 1970 .
e. milman .
a converse to the mazya inequality for capacities under curvature lower bound . in _ around the research of vladimir mazya i. function spaces ( a. laptev editor ) _ ,
volume 11 of _ international mathematical series _ , pages 321348 .
springer and tamara rozhkovskaya publisher , 2010 .
|
various properties of isoperimetric , functional , transport - entropy and concentration inequalities are studied on a riemannian manifold equipped with a measure , whose generalized ricci curvature is bounded from below .
first , stability of these inequalities with respect to perturbation of the measure is obtained .
the extent of the perturbation is measured using several different distances between perturbed and original measure , such as a one - sided @xmath0 bound on the ratio between their densities , wasserstein distances , and kullback leibler divergence .
in particular , an extension of the holley stroock perturbation lemma for the log - sobolev inequality is obtained , and the dependence on the perturbation parameter is improved from linear to logarithmic .
second , the equivalence of transport - entropy inequalities with different cost - functions is verified , by obtaining a reverse jensen type inequality . the main tool used is a previous precise result on the equivalence between concentration and isoperimetric inequalities in the described setting . of independent interest is a new dimension independent characterization of transport - entropy inequalities with respect to the @xmath1-wasserstein distance , which does not assume any curvature lower bound .
|
over the past decade , many exotic close binary systems with supergiant components have been discovered .
systems like ngc 300 x-1 a wolf - rayet / black hole x - ray binary ( ( * ? ? ?
* crowther 2010 ) ) and sn2010da a sgb[e]/neutron star x - ray binary ( ( * ? ? ?
* villar 2016 ) ) are two such examples of a coupling between a massive star in a short - lived evolutionary phase and a compact stellar remnant .
interestingly , no x - ray binaries with confirmed red supergiant ( rsg ) counterparts have been discovered ( rsgs have been proposed as the candidate donor star for a few ultraluminous x - ray sources , see ( * ? ? ?
* heida 2016 ) ) .
this may be partially explained by the rarity of rsgs ; however , though rare , rsgs are both longer - lived and more common ( due to the smaller 10 - 25 @xmath0 initial masses of their zero age main sequence progenitors ) than most other evolved massive stars .
rsg x - ray binaries , if they exist , offer a view into an interesting edge case of accretion ; their extended envelopes and strong winds ( @xmath1 yr@xmath2 , ( * ? ? ?
* van loon 2005 ) ) could allow for accretion from both the wind and roche - lobe overflow in an environment continually enriched with dust produced by the rsg .
rsg x - ray binaries are also the immediate progenitors of thorne - ytkow objects stars with embedded neutron star cores ( ( * ? ? ?
* thorne & ytkow 1975 ) ) assuming the neutron star plunges into the rsg as it expands ( ( * ? ? ?
* taam 1978 ) ) . to search for rsg x - ray binaries
, we used the photometry of the local group galaxy survey ( lggs , ( * ? ? ?
? * ; * ? ? ?
* massey & olsen 2003 , massey 2006 , 2007 ) ) , which covers m31 , m33 , the magellanic clouds and 7 dwarf galaxies in the local group .
following @xcite to find rsgs among the nearly - identical foreground dwarfs , we cross - referenced the positions of the lggs rsgs with the _ chandra _ source catalog ( csc , ( * ? ? ?
* evans 2010 ) ) , and found one rsg coincident with an x - ray source .
lggs j004527.30 + 413254.3 ( j0045 + 41 hereafter ) is a bright ( @xmath3 ) object of previously - unknown nature in the disk of m31 . @xcite
classify j0045 + 41 as an eclipsing binary with a period of @xmath4 days . j0045 + 41
was also observed with the palomar transient factory ( ptf ) ; the @xmath5-band lightcurve shows evidence for a @xmath6 day period . on the other hand @xcite
identify j0045 + 41 as a globular cluster , and it has been included in catalogs of m31 globular clusters as recently as 2014 ( ( * ? ? ?
* wang 2014 ) ) .
the lggs photometry is consistent with the color and brightness of a rsg .
indeed , following @xcite , we found that , as an rsg , j0045 + 41 would have an effective temperature of @xmath73500 k and bolometric magnitude of -6.67 , consistent with a 12 - 15 m@xmath8 rsg .
however , a complete sed fit to photometry from the panchromatic hubble andromeda treasury ( phat , ( * ? ? ?
* dalcanton 2012 ) ) using the bayesian extinction and stellar tool ( beast , ( * ? ? ?
* gordon 2016 ) ) yields an unphysical result of 300 m@xmath8 , 10@xmath9 k star , extincted by @xmath104 magnitudes .
furthermore , the object appears extended in the phat images ( though its radial profile appears similar to that of other nearby stars ) .
j0045 + 41 is separated by @xmath11 from an x - ray source .
the source , cxo j004527.3 + + 413255 , is bright ( @xmath12 erg s@xmath2 @xmath13 ) and hard ; fitting a spectrum obtained by williams et al .
( in prep ) yields a power law with @xmath14 .
the best - fit neutral hydrogen column density is @xmath15 @xmath13 , which corresponds to @xmath16 .
the apparent periodicity of j0045 + 41 and its apparent association with a hard and unabsorbed x - ray source prompted us to obtain follow - up spectroscopic observations to determine the true nature of this object .
we obtained a longslit spectrum of j0045 + 41 using the gemini multi - object spectrograph ( gmos ) on gemini north .
four 875 second exposures were taken 2016 july 5 using the ` b600 ` grating centered on 5000 , and four 600 second exposures were taken 2016 july 9 using the ` r400 ` grating centered on 7000 , with a blocking filter to remove 2@xmath17-order diffraction . due to the gaps between gmos s three ccds ,
two of each set of exposures were offset by + 50 .
the data were reduced using the standard ` gemini ` ` iraf ` package .
the final reduced spectrum has continuous signal from @xmath74000 to @xmath79100 at a resolution of @xmath18 .
line is telluric.,scaledwidth=90.0% ] the spectrum ( figure [ fig : spectrum ] ) shows that j0045 + 41 is a quasar at @xmath19 ( measured with h@xmath20 , h@xmath21 , [ oiii]@xmath22 and ca ii h and k ) . while a false positive , this quasar is quite interesting on its own .
the ( low - significance ) detections of periodicity on short timescales by multiple sources are difficult to explain .
furthermore , mistaking a blue quasar for a red star would imply a high reddening along the line of sight , which is consistent with a sightline through m31 ; however , the low h column density implied by the x - ray counterpart s spectrum and our inability to achieve a satisfactory fit to the optical spectrum by reddening the ( redshifted ) quasar template spectrum from @xcite indicate that j0045 + 41 belongs to a small and intriguing class of intrinsically red quasars , observed through a low - extinction region of the ism in m31 ( ( * ? ? ?
* richards 2003 ) ) .
if a quasar can be misidentified as a rsg , are there red stars especially x - ray bright stars in already - existing quasar catalogs ? to answer this question
, we turned to the quasar catalog of the sloan digital sky survey ( sdss , ( * ? ? ?
* york 2000 ) ) .
we selected all sdss objects automatically tagged as quasars that were within 0.2 magnitudes of j0045 + 41 in @xmath23 vs. @xmath24 vs. @xmath25 color - space j0045 + 41 is too faint in @xmath26 to utilize @xmath27 and ignored any warning flags to avoid throwing out interesting objects that were not easily identified by the sdss algorithm .
1098 objects in this sample had associated spectroscopic observations .
interestingly , many of the spectroscopically - determined redshifts were unbelievably small or even negative , implying that these objects are in a regime of color - space where classification algorithms may fail .
indeed , on visual inspection of these spectra , many of them are stellar .
we used ` emcee ` , a python implementation of markov chain monte carlo ( mcmc ) by @xcite , to fit gaussian profiles to the ca ii triplet ( @xmath28 ) , which we use to identify stars .
the posterior distributions of the parameters allow us to determine if the triplet is well fit , and estimate errors for each parameter . because the relative centroids and strengths of the lines are fixed , a good fit guarantees the lines are actually detected , while simultaneously measuring with accurate errors the radial velocity and equivalent width ( @xmath29 ) of the triplet . after a follow - up inspection by eye of spectra that were noisy or missing data , we find 344 confirmed cool stars , representing @xmath731% of the total sample .
figure [ fig : ew_hist ] shows the distribution of @xmath29 for @xmath30 .
we follow @xcite to estimate luminosity from @xmath29 , and find that most stars are dwarfs ( @xmath31 ) but @xmath740 stars have larger equivalent widths indicating they are likely giants or supergiants ( the relationship is dependent on effective temperature and metallicity , so these labels are approximate ) .
this result demonstrates that , when looking for rare objects like rsg x - ray binaries , it is important to look in unlikely places ; e.g. , a red and x - ray bright star may be confused for a quasar if the classification algorithm mistakes the continuum between the tio bands in a rsg spectrum for an emission line . the fact that some of the color - space containing m - dwarfs by far the most common type of star is a regime where classification algorithms fail underlines the importance of improving on these algorithms until they perform as well as the human eye .
indeed , many of these stars were previously identified ( see ( * ? ? ?
* west 2011 ) ) , but are still listed as quasars on the sdss online data portal . future work will focus on improving our star - finding algorithm to use alternate spectral features when the ca ii triplet is missing or obscured by noise , and on finding which areas of color - space contain significant numbers of these misidentified stars , with the goal of finding rsg x - ray binaries as well as improving our knowledge of where classification algorithms fail .
|
thanks to incredible advances in instrumentation , surveys like the sloan digital sky survey have been able to find and catalog billions of objects , ranging from local m dwarfs to distant quasars .
machine learning algorithms have greatly aided in the effort to classify these objects ; however , there are regimes where these algorithms fail , where interesting oddities may be found
. we present here an x - ray bright quasar misidentified as a red supergiant / x - ray binary , and a subsequent search of the sdss quasar catalog for x - ray bright stars misidentified as quasars .
|
it is well established nowadays that qcd undergoes a chiral restoration transition , as confirmed by different lattice simulations @xcite .
the transition is presumably a crossover for 2 + 1 massive light flavors , which means that the critical point could be different for different observables . in the lattice ,
the main order parameters are the quark condensate and the light scalar susceptibility .
nevertheless , there have been several efforts in the literature to try to find signals of chiral symmetry restorations in physical observables which could be eventually measured in heavy ion or nuclear matter experiments .
the most promising signals are related to the behaviour of resonances in hot and dense matter , such as the @xmath2 degeneracy @xcite or the role of the @xmath1 resonance in the dilepton spectrum @xcite . in the latter case
, a significant dropping of the @xmath1 mass would be interpreted as a signal of chiral restoration according to the brown - rho scaling @xcite or qcd sum rules @xcite scenarios . however
, this does not seem to be the case experimentally in heavy ion collisions , where the alternative explanation of broadening dominance @xcite is equally valid to explain the data and it is actually favored in recent experiments , such as na60 with dimuons @xcite . at rhic energies ( phenix ) , even the broadening picture is insufficient to explain the observed excess in the region of low invariant mass @xcite .
the situation is less clear in cold nuclear matter experiments .
the e325-kek collaboration @xcite has reported a measurable shift in the masses of vector mesons compatible with theoretical predictions based on brown - rho scaling and qcd sum rules . on the other hand ,
the jlab - clas experiment @xcite has obtained results compatible with vanishing mass shift , as predicted by most in - medium hadronic many - body analysis , where broadening is the dominant effect @xcite .
the @xmath3 meson is also a suitable candidate to study signals of chiral restoration through its in - medium behaviour , since it has the quantum numbers of the vacuum .
a simple @xmath4 model description would suggest that the dropping of @xmath5 driven by chiral restoration should imply a significant reduction of the @xmath0 mass , at least in the chiral limit .
such a reduction could produce experimentally an enhancement of @xmath6 scattering and cross section near the point where @xmath7 and the phase space shrinks .
this is the so called threshold enhancement effect @xcite , whose signals are observed in cold nuclear matter reactions @xmath8 @xcite and @xmath9 @xcite .
our approach is to consider the temperature and density modifications of the lightest meson resonances in the unitarized chiral lagrangian framework @xcite .
we rely mostly on chiral perturbation theory ( chpt ) , which provides a model - independent low - energy expansion for a given @xmath6 scattering partial wave with total isospin @xmath10 and angular momentum @xmath11 as @xmath12 with @xmath13 the center of mass energy squared and @xmath14 the temperature .
according to the usual chpt scheme , @xmath15 is the tree level contribution from the lowest order lagrangian @xmath16 , while @xmath17 includes both the tree level from @xmath18 and the one - loop diagrams from @xmath16 .
the latter contain the dependence with the temperature @xcite , as well as the imaginary part demanded by unitarity , which at finite temperature reads @xmath19 , where : @xmath20\label{thps}\ ] ] is the two - pion thermal phase space , with @xmath21^{-1}$ ] the bose - einstein distribution function and @xmath22 .
the chpt series can not reproduce a resonant behaviour , which can be recovered by using a unitarization method , which amounts to find a unitary amplitude @xmath23 satisfying exactly @xmath24 and matching the chpt series when expanded at low energies or temperatures .
we follow the inverse amplitude method ( iam ) which reads : @xmath25 where @xmath26 is a known function constructed from @xmath15 and @xmath17 ensuring that the zeros of the unitarized amplitude coincide with those of the perturbative one and which can be alternatively obtained using dispersion relations @xcite .
it is irrelevant for poles far from the real axis but should be included in medium analysis where , as commented above , chiral symmetry may induce a vanishing imaginary part and in fact the @xmath27 function in that case prevents the appearance of spurious poles @xcite .
once we have a unitarized @xmath6 amplitude , the @xmath0 and @xmath1 are generated as poles in the second riemann sheet dynamically , i.e. , without introducing them explicitly in the lagrangian nor assuming anything about their nature .
the results with the finite-@xmath14 iam are showed in fig.[fig : rhosigmat ] .
the so - called low - energy constants of @xmath18 are fixed to yield the mass and width of the @xmath1 at their physical values at @xmath28 .
the main conclusions of our analysis are : i ) the main effect for the @xmath1 is thermal broadening , in agreement with dilepton data , the mass being reduced only slightly for temperatures below chiral restoration , ii ) such broadening does not only come from phase space increasing @xmath29 but also from the increasing of the effective @xmath30 vertex , extracted from the residue of the amplitude at the pole position , iii ) in the @xmath0 channel , mass reduction dominates , presumably driven by chiral restoration , and at some point overcomes the thermal width phase space increasing so that the imaginary part ( width ) reaches a maximum value and then starts decreasing , iv ) near chiral restoration , the width of the @xmath0 remains sizable ( its spectral function is not peaked around the mass as in a breit - wigner resonance ) so that no threshold enhancement is expected in a heavy ion environment , v ) the mass dropping pattern in both the @xmath0 and @xmath1 cases does not scale with the condensate as expected from brown - rho scaling or simple @xmath4 model arguments for the @xmath0 , not even near the chiral limit .
@xmath31{rhowidth } & \includegraphics[height=.21\textheight]{mrho}\\ \includegraphics[height=.21\textheight]{msigvscond140 } & \includegraphics[height=.21\textheight]{msigvscond10 } \end{array}$ ] the next question is how to incorporate the effect of the nuclear density @xmath1 in our approach . at @xmath28 , a simple procedure @xcite which encodes most of the relevant corrections is to scale the pion decay constant @xmath32 according to @xcite : @xmath33 where @xmath1 is the nuclear density , @xmath34 45 mev is the pion - nucleon sigma term and @xmath35 0.17 @xmath36 is the normal or saturation nuclear matter density . in this way we are actually examining the scaling properties of the resonances when we scale @xmath37 , or equivalently the quark condensate
actually , in this simple approach there is no broadening source , so that only the masses change .
the results of the mass scaling in our iam approach are showed in fig.[fig : scalingfpi ] . in the @xmath38 channel
, the scaling corresponds to the @xmath0 of the @xmath4 model .
when the pole approaches the real axis , threshold enhancement is clearly observed and when it reaches it a pole - doubling occurs : one of the poles remains near threshold and eventually jumps into the first riemann sheet becoming a @xmath6 bound state , while the other one ( plotted in fig.[fig : scalingfpi ] ) tends to degenerate with the pion as its chiral partner .
although this phenomenon takes place at rather high @xmath39 for this approach to be fully trusted , it has been obtained also in @xcite and reveals the molecular - like character of the @xmath40 state @xcite .
the chiral degeneracy between the dynamically generated @xmath0 and the pion for @xmath41 had been also noticed in @xcite . in the @xmath1 channel ,
the @xmath32 scaling induces a linear @xmath42 scaling in agreement with the experimental value of @xcite and the theoretical predictions of @xcite and @xcite , which however must be taken with a pinch of salt because all the broadening effects coming from many - body interactions are neglected .
scaling of the mass in the iam , compared with that of @xmath43 . in the @xmath38 channel ,
the masses displayed correspond to the lowest masses of the second - sheet poles.,title="fig : " ] scaling of the mass in the iam , compared with that of @xmath43 . in the @xmath38 channel , the masses displayed correspond to the lowest masses of the second - sheet poles.,title="fig : " ] the picture that emerges from our analysis is then that when broadening effects are ignored , chiral restoration in the @xmath0 channel takes place more or less along the expected @xmath4 pattern .
however , the @xmath44 study shows that thermal broadening distorts the picture of a @xmath45-like narrow state and , despite experiencing a large mass reduction , the state is still wide near the transition .
this is confirmed by the detailed many - body calculation performed in @xcite for the @xmath0 channel , including both nuclear density and temperature effects .
the unitarization scheme used is the bethe - salpeter one , more suitable for nuclear many - body analysis , and takes into account the pion self - energy in the nuclear medium plus all the other relevant diagrams compatible with chiral symmetry at the same order in density .
the results in figure [ fig : tempdens ] for the imaginary part of the amplitude show some distinctive features of this full approach : the abrupt change with respect to the vacuum or thermal case , originated by physical in - medium excitations such as particle - hole or @xmath46-hole , which produce also strength below threshold , the sizable threshold enhancement , even though the pole is still rather away from the axis like in the thermal case , and finally the amplification of the threshold strength by the combined effect of temperature and density , which in fact accelerates the migration of the pole towards threshold by sudden mass decrease
. imaginary part of the @xmath6 amplitude in the @xmath38 channel at finite temperature and nuclear density in the many - body bs approach.,title="fig : " ] imaginary part of the @xmath6 amplitude in the @xmath38 channel at finite temperature and nuclear density in the many - body bs approach.,title="fig : " ]
in the hadron gas formed in a relativistic heavy ion collision , it is phenomenologically well justified to consider a phase where thermal equilibrium still prevails but inelastic collisions involving particle number change are negligible , so that chemical equilibrium is lost .
for instance , according to the estimates in @xcite , at @xmath47 the relaxation time of elastic @xmath6 collision is @xmath48 , whereas that of the process @xmath49 is @xmath50 , the hadronic phase lifetime being about @xmath51 . in other words , there is a temperature window between the chemical freeze - out temperature @xmath52 180 mev , below which approximately only elastic collisions remain , and the thermal freeze - out one @xmath53 100 - 120 mev of hadron decoupling , where the chemical potential associated to pion number conservation @xmath54 is not zero @xcite .
actually , in this phase @xmath54 is a function of @xmath14 so that @xmath55 . on the other hand ,
different phenomenological fits of particle yields and ratios at sps and rhic energies predict @xmath56 70 - 100 mev @xcite . in a recent work @xcite , we have considered a chpt description of this phase in the pion gas , within a quantum field theory treatment of chemical nonequilibrium for boson fields .
a crucial point is that the total pion number is only approximately conserved , unlike for instance the electric charge or the third isospin component .
this makes the usual path - integral description in terms of field states inconvenient for this case .
we have used instead an holomorphic representation , which allows to develop the formalism in terms of creation and annihilation operators , in terms of which one can easily express the free particle number .
another technical complication is that there are no kms - like periodicity conditions , which prevents the imaginary - time thermal formalism to be used . as in other nonequilibrium formulations ,
the appropriate choice is a real - time contour in complex time .
this implies the doubling of the pion degrees of freedom , so that the propagators have a two - dimensional matrix structure .
one of the two types of fields is unphysical and only appears in internal lines . after a detailed analysis
, we obtain the generating functional , which can be read from the usual one at @xmath57 with the following replacements for the free propagators and for the free partition function : @xmath58=d_{21}(-p_0,e_p)\nonumber\\ \log z_\beta^0&=&-v\int\frac{\mathrm{d}^3\vec{p}}{(2\pi)^3}\ \left[\frac{\beta e_p}{2}+\log\left(1-\mathrm{e}^{-\beta(e_p-\mu)}\right)\right ] \label{rtpropmom}\end{aligned}\ ] ] with @xmath59 the system volume and @xmath60 . with the above ingredients
, we can calculate the different thermodynamical variables of the pion gas in this regime , by evaluating the corresponding closed diagrams , whose topology is the same as the @xmath57 case @xcite .
we have extended our analysis up to @xmath61 . at that order ,
particle - changing processes can be seen to become already important , but the perturbative scheme remains consistent since an additional physical condition has to be imposed in order to regulate the plasma dynamical evolution from thermal to chemical equilibrium , namely the @xmath62 function . a simple way to model
this is to demand that the ratio of entropy density to pion particle number density @xmath63 remains constant @xcite .
this isentropic condition is nothing but the combination of entropy conservation without dissipation and approximate particle number conservation in this phase .
the crucial point is that this ratio is a decreasing function of both @xmath14 and @xmath54 , so that as the systems cools down , the chemical potential grows in order to keep it constant .
the precise value of @xmath63 can be fixed either by the chemical freeze - out temperature at which @xmath64 or by the value @xmath65 at thermal freeze - out .
we show in figure [ fig : mut ] our results for different chiral orders , comparing also with the virial expansion approach followed in @xcite .
we see that one of the consequences of including the pion interactions is to lower the value of @xmath66 by about 25 mev with respect to the ideal gas .
left : the function @xmath62 with the isentropic condition for different approximations .
right : the pion mass dependence with @xmath14 and @xmath54.,title="fig : " ] left : the function @xmath62 with the isentropic condition for different approximations .
right : the pion mass dependence with @xmath14 and @xmath54.,title="fig : " ] our diagrammatic approach allows also to calculate the pion self - energy corrections in @xmath14 and @xmath54 . the detailed analysis and results
are given in @xcite .
to leading order in the pion density ( dilute gas regime ) the self - energy obeys a luscher - like relation : @xmath67+{{\cal o}}(n^2 ) \label{gamma}\end{aligned}\ ] ] where @xmath68 is the forward @xmath6 scattering amplitude .
the @xmath69 and unitarized ( iam ) amplitudes produce a decreasing pion mass with both @xmath14 and @xmath54 . actually , at a given temperature
, the mass can reach the chemical potential , as showed in figure [ fig : mut ] .
this implies the interesting possibility that pion bose - einstein condensation could be reached dynamically , i.e. , driven by interactions .
the corresponding @xmath70 where this condition is met is above but not far from the isentropic curves for realistic chemical freeze - out conditions . on the other hand , from the pion width @xmath71
, we can estimate the elastic mean collision time @xmath72 and the thermal freeze - out temperature where @xmath73 10 fm / c .
taking into account the @xmath54 dependence of @xmath72 through the isentropic @xmath62 yields a reduction @xmath74 -20 mev with respect to neglecting chemical nonequilibrium .
transport coefficients measure the linear response of the system to a deviation from equilibrium induced by an external source .
their knowledge is essential to describe many phenomenological aspects regarding dissipation effects in heavy ion collisions .
recently , we have developed an extensive programme for the calculation of transport coefficients within diagrammatic chpt , with direct application to the mesonic phase of the plasma expansion @xcite .
the diagrammatic treatment is technically nontrivial because transport coefficients involve the zero external frequency and momentum limit of retarded correlators .
this implies the appearance of the so called pinching poles , proportional to @xmath75 with @xmath76 the collisional width of the particles in the medium .
this generates nonperturbative contributions , since @xmath76 is proportional to the collision amplitude ( see e.g. eq.([gamma ] ) ) and therefore perturbatively small .
thus , the usual chiral power counting of chpt has to modified to account properly for these contributions @xcite .
the viscosity coefficients are particularly interesting for phenomenology .
the shear viscosity to entropy density ratio @xmath78 can be determined through elliptic flow analysis , the present rhic data pointing to a small value @xmath79 @xcite , an almost perfect fluid .
our chpt result is greatly influenced by a correct unitarized description of the scattering amplitudes entering the pion width .
in fact , without unitarization , at @xmath69 , @xmath78 is a monotonically decreasing function of @xmath14 , violating the so called kss bound @xmath80 @xcite , as showed in figure [ fig : visco ] .
the unitarized result satisfies that bound , gives phenomenologically compatible values at the relevant temperatures and is compatible with the existence of a minimum for this ratio near the transition @xcite .
the case of the bulk viscosity @xmath81 is also important because it is directly sensitive to variations of conformal symmetry .
thus , it should have a similar behaviour @xcite to the trace anomaly : @xmath82 where @xmath83 is the energy - momentum tensor and @xmath84 the thermodynamic pressure .
the trace anomaly has a maximum at the qcd transition , originated mostly from the anomalous glue contribution to the beta function and observed in the lattice @xcite .
therefore , significantly large values for @xmath81 could be observed near the critical point .
our results in figure [ fig : visco ] show a clear correlation between these quantities , both developing a two - peak structure .
the first maximum comes from the explicit conformal breaking due to the quark mass and thus it disappears in the chiral limit .
actually , in that temperature range @xmath85 @xcite .
the second maximum is the critical one and is almost unchanged in the chiral limit , not being related to the quark condensate or to the quark mass , but to deconfinement effects encoded in the gluon condensate anomalous conformal breaking .
we see again that incorporating properly unitarization is crucial to describe correctly the high temperature region .
it is also essential to consider the speed of sound squared @xmath86 to @xmath61 in order to capture its minimum at the transition arising from anomalous conformal breaking .
our analysis does not rely on assumptions on the behaviour of the @xmath87 spectral function , as it is the case of the original proposal in @xcite , which has been recently corrected @xcite .
it is therefore unclear from general considerations how the correlation between the trace anomaly and the bulk viscosity should actually hold .
in fact , it does not hold in simple models mimicking qcd @xcite .
we believe then that our pion gas description can be useful to understand this and other issues regarding the behaviour of transport coefficients below the critical temperature .
99 y. aoki _ et al .
_ , jhep * 0906 * , 088 ( 2009 ) .
m. cheng _ et al . _ ,
d * 81 * , 054504 ( 2010 ) .
r. rapp and j. wambach , adv .
phys . * 25 * , ( 2000 ) 1 .
d. cabrera , d. jido , r. rapp and l. roca , prog .
phys . * 123 * , 719 ( 2010 ) .
g. david , r. rapp and z. xu , phys .
* 462 * , 176 ( 2008 ) .
g. e. brown and m. rho , phys . rev .
* 66 * , ( 1991 ) 2720 ; phys .
rept . * 363 * , 85 ( 2002 ) .
t. hatsuda , s. h. lee and h. shiomi , phys .
c * 52 * ( 1995 ) 3364 .
r. arnaldi _ et al . _
[ na60 collaboration ] , phys .
lett . * 96 * , ( 2006 ) 162302 .
a. adare _ et al .
_ [ phenix collaboration ] , phys .
b * 670 * , 313 ( 2009 ) .
a. drees , nucl .
a * 830 * , 435c ( 2009 ) .
m. naruki _ et al .
_ , phys .
* 96 * ( 2006 ) 092301 .
r. nasseripour _ et al . _ [ clas collaboration ] , phys .
lett . *
99 * , ( 2007 ) 262302 .
m. urban , m. buballa , r. rapp and j. wambach , nucl .
a * 641 * ( 1998 ) 433 .
d. cabrera , e. oset and m. j. vicente vacas , nucl .
a * 705 * , ( 2002 ) 90 .
t. hatsuda and t. kunihiro , phys .
lett . * 55 * , ( 1985 ) 158
. f. bonutti _ et al .
_ [ chaos collaboration ] , nucl .
a * 677 * , ( 2000 ) 213 .
a. starostin
_ et al . _
[ crystal ball collaboration ] , phys .
lett . * 85 * , ( 2000 ) 5539 .
j. g. messchendorp _ et al . _ , phys .
lett . * 89 * , ( 2002 ) 222302 .
a. gomez nicola , f. j. llanes - estrada and j. r. pelaez , phys .
b * 550 * , 55 ( 2002 ) .
a. dobado , a. gomez nicola , f. j. llanes - estrada and j. r. pelaez , phys .
c * 66 * , 055201 ( 2002 ) .
d. fernandez - fraile , a. gomez nicola and e. t. herruzo , phys .
d * 76 * , 085020 ( 2007 ) .
d. cabrera , d. fernandez - fraile and a. gomez nicola , eur .
j. c * 61 * , 879 ( 2009 ) .
a. gomez nicola , j. r. pelaez and g. rios , phys .
d * 77 * ( 2008 ) 056006 .
v. thorsson and a. wirzba , nucl .
phys . * a589 * , ( 1995 ) 633 .
u. g. meissner , j. a. oller and a. wirzba , annals .
phys . * 297 * , ( 2002 ) 27 .
a. patkos , z. szep and p. szepfalusy , phys .
* d68 * , ( 2003 ) 047701 .
j. a. oller , arxiv : hep - ph/0007349 . c. song and v. koch , phys .
c * 55 * , 3026 ( 1997 ) .
h. bebie , p. gerber , j. l. goity and h. leutwyler , nucl .
b * 378 * , 95 ( 1992 ) . c. m. hung and e. v. shuryak , phys .
c * 57 * , 1891 ( 1998 ) .
p. f. kolb and r. rapp , phys .
c * 67 * , 044903 ( 2003 ) .
d. fernandez - fraile and a. gomez nicola , phys .
d * 80 * , 056003 ( 2009 ) .
p.gerber and h.leutwyler , nucl .
b*321 * , 387 ( 1989 ) .
a. dobado and j. r. pelaez , phys .
d * 59 * , 034004 ( 1999 ) .
d. fernandez - fraile and a. gomez nicola , phys .
d * 73 * , 045025 ( 2006 ) .
d. fernandez - fraile and a. g. nicola , phys .
lett . * 102 * , 121601 ( 2009 ) .
d. fernandez - fraile and a. gomez nicola , eur .
j. c * 62 * , 37 ( 2009 ) .
k. dusling and d. teaney , phys .
c * 77 * , 034905 ( 2008 ) .
p. kovtun , d. t. son and a. o. starinets , phys .
lett . * 94 * , 111601 ( 2005 ) .
a. dobado , f. j. llanes - estrada and j. m. torres - rincon , phys .
d * 79 * , 014002 ( 2009 ) .
f.karsch , d.kharzeev and k.tuchin , phys.lett.b*663 * , 217 ( 2008 ) .
p. romatschke and d. t. son , phys .
d , 2009 , * 80 * : 065021 .
d. fernandez - fraile , arxiv:1009.2741 [ hep - ph ] .
|
we review recent results on properties of the meson gas relevant for heavy ion collision and nuclear matter experiments , within the framework of chiral lagrangians . in particular , we describe the temperature and density evolution of the @xmath0 and @xmath1 poles and its connection with chiral symmetry restoration , as well as the chemical nonequilibrated phase and transport coefficients . address = departamento de fsica terica ii .
univ . complutense .
28040 madrid .
spain .
address = nuclear theory group , department of physics , brookhaven national laboratory , upton , ny-11973 , usa .
|
the main aim of this paper is to discuss a limiting relation between two classes of generalized schrdinger operators in @xmath1 .
one of them are measure - type perturbations of the laplacian , in particular , we will be interested in potentials which are a negative multiple of the dirac measure supported by a finite graph @xmath4 , in other words , they are formally given by the expression @xmath5 with some @xmath6 .
we are going to show that such operators can be approximated in the strong resolvent sense by families of point - interaction hamiltonians @xcite with two - dimensional @xmath3 potentials suitably arranged .
stated in this way the problem is an interesting mathematical question ; the indicated result will represent an extension the approximation theorem derived in @xcite to the two - dimensional situation , which is not straightforward because properties of the @xmath3 potentials are dimension - dependent .
in addition , we have a strong physical motivation coming from the fact that hamiltonians of the type ( [ formal ] ) are good models of various graph - type nanostructures which in distinction to the usual description @xcite take quantum tunneling into account see @xcite for a bibliography to this problem . to investigate such models one has to find spectral properties of the operator ( [ formal ] ) for various shapes of @xmath2 .
this is in general a complicated task even if the original pde problems is reformulated by means of the generalized birman - schwinger principle @xcite into solution of an appropriate integral equation . on the other hand , spectral problem for point - interaction hamiltonians
is reduced in a standard way to solution of an algebraic equation with coefficients containing values of the free green function .
hence an approximation of the mentioned type would be of practical importance ; our second aim is to illustrate this aspect of the problem on examples .
let us describe briefly contents of the paper .
in the next section we collect the needed preliminaries stated in a way suitable for the further argument . in section [ main result ]
we formulate and prove our main result , which is the approximation indicated above . in the following two sections we discuss examples of two simple graph classes .
the first are graphs of the form of a ring , full or open . in this case the spectrum of the operator ( [ formal ] ) can be found by mode matching @xcite which enables us to compare the approximation with the `` exact '' result , in particular , to assess the rate of its convergence .
in contrast , in section [ stargraph ] we discuss star - shaped graphs . here
an alternative method for numerical solution of the spectral problem is missing , however , we can derive several conclusions for ( an infinite ) star analytically and show how does the approximation results conform with them .
computations of this type were performed already in @xcite but the number of points used there was too small to provide a reasonable approximation . in the last section we address another question which is more difficult and no analytical results are available presently .
it concerns the scattering problem on infinite leaky graphs with asymptotically straight `` leads '' . on a heuristic level
, it is natural to expect that the states from the negative part of the continuous spectrum can propagate being transversally confined to the graph edges , hence a nontrivial geometry should yield an @xmath7 on - shell scattering matrix , where @xmath8 is the number of the leads .
this remains to be proved , however , and even more difficult it will be to compute the s - matrix mentioned above in terms of the graph geometry . for simplicity we will restrict ourselves to the simplest case when @xmath2 is an infinite asymptotically straight curve , i.e. @xmath9 , without self - intersections .
one natural conjecture is that if the distance between the curve points is small somewhere so that @xmath2 is close to a loop shape locally , the system can exhibit resonances due to quantum tunneling . on the other hand ,
simple bends are unlikely to produce distinguished resonances ; the reason is that the transverse @xmath3 coupling in ( [ formal ] ) has a single bound state , hence there is no analog here to higher thresholds which give rise to resonances in bent hard - wall tubes @xcite . to support these conjectures ,
we have let us inspire by the so - called @xmath10 approach to resonances , which was in the one - dimensional case set on a rigorous ground in @xcite .
specifically , we study the spectrum of a finite segment of @xmath2 as a function of the cut - off position . in the first case we find in this dependence
avoided eigenvalue crossings , the more narrow the smaller the gap to tunnel is , while no such effect is seen in the second case . on the other hand ,
a tunneling between arcwise distant points of the graphs is not the only source of resonances . to illustrate this claim , we analyze in our last example a z - shaped graph with two sharp bends separated by a line segment .
we find again avoided crossings the widths of which vary widely as functions of the bending angle .
this provides an indirect but convincing indication that the above described on - shell s - matrix for a bent curve in the shape of a broken line is nontrivial .
we start with the definition of schrdinger operators of the form @xmath11 in @xmath12 , where @xmath13 is a finite positive measure on the borel @xmath14-algebra of @xmath2 , which is assumed to be a non - empty closed subset of @xmath15 and also the support of the measure @xmath13 .
furthermore , @xmath16 is a bounded and continuous function , acting from @xmath2 to @xmath17 .
we suppose that the measure @xmath13 belongs to the generalized kato class , in two - dimensional case it means that the following condition holds , @xmath18 centered at @xmath19 . due to the fact that the measure @xmath13 belongs to kato class , for each @xmath20 there exists @xmath21 such that any @xmath22 from the schwartz space @xmath23 satisfies the inequality @xmath24 as it was proven in the paper @xcite .
since @xmath23 is dense in @xmath25 we can define a bounded linear transformation @xmath26 using this transformation , the inequality ( [ m estimate ] ) can be extended to the whole @xmath25 with the function @xmath22 on l.h.s .
replaced by @xmath27 . by employing the klmn theorem ,
see ( * ? ? ?
x.17 ) , we conclude that the quadratic form @xmath28 given by @xmath29 is lower semi - bounded and closed in @xmath12 .
hence there exists a unique self - adjoint operator @xmath30 associated with the form @xmath28 .
the described definition applies to more general class of measures than we need here .
if @xmath2 is a graph consisting of a locally finite number of smooth edges meeting at nonzero angles , i.e. having no cusps , there is another way to define @xmath30 , namely via boundary conditions imposed on the wavefunctions . to this aim ,
consider first the operator @xmath31 acting as @xmath32 for any @xmath22 of the domain consisting of functions which belong to @xmath33 , are continuous at @xmath2 with the normal derivatives having there a jump , @xmath34 then it is straightforward to check that @xmath31 is e.s.a . and
by green s formula it reproduces the form @xmath35 on its core , so its closure may be identified with @xmath36 defined above .
an important tool to analyze spectra of such operators is the generalized birman - schwinger method . if @xmath37 belongs to the resolvent set of @xmath30 we put @xmath38 .
the free resolvent @xmath39 is defined for @xmath40 as an integral operator with the kernel @xmath41 now we shall use @xmath39 to define three other operators . for the sake of generality ,
suppose that @xmath42 are positive radon measures on @xmath15 with @xmath43 for any @xmath44
. in our case they will be the measure @xmath13 on @xmath2 and the lebesgue measure @xmath45 on @xmath15 in different combinations . by @xmath46
we denote the integral operator from @xmath47 to @xmath48 with the kernel @xmath49 , i.e. @xmath50 holds @xmath51-a.e . for all @xmath52 . with this notation
one can express the resolvent @xmath53 as follows @xcite : [ gen krein ] let @xmath40 .
suppose that @xmath54 is invertible and the operator @xmath55^{-1 } r^k_{m,{{\rm d}}x}\ ] ] from @xmath12 to @xmath12 is everywhere defined .
then @xmath37 belongs to @xmath56 and @xmath57 .
the invertibility hypothesis is satisfied for all sufficiently large negative @xmath37 because for such a @xmath37 the operator norm of @xmath58 acting in @xmath59 is less than 1 , see again @xcite , and similar result can be proven for the operator norm in @xmath60 following @xcite . thus from now on we consider @xmath61 such that both these norms are less than one . for later considerations it is useful we rewrite operator @xmath62 in the form @xmath63 , where we have introduced @xmath64 since function @xmath16 acquires only non - negative values and @xmath13 is positive measure , @xmath65 is a positive number .
the resolvent reads @xmath66 it can be rewritten alternatively as @xmath67 where @xmath68 and @xmath69 represents the unique solution to the equation @xmath70 since @xmath71 belongs to @xmath72 , using sobolev s embedding theorem we conclude that it has a bounded and continuous version , and therefore @xmath14 also has a representative which is bounded and continuous on the set @xmath73 .
consider a discrete and finite subset @xmath74 and the positive constant @xmath65 defined above .
as it is well known we can not regard the operator @xmath75 with the interaction supported by @xmath76 as before , i.e. as ` @xmath77 + measure ' .
instead , we define this operator via its domain : each function @xmath78 behaves in the vicinity of a point @xmath79 as follows @xmath80 where the generalized boundary values @xmath81 and @xmath82 satisfy @xmath83 for any @xmath79 with @xmath84 ; for a justification of this definition and further properties of the operators @xmath75 see , e.g. , ( * ? ? ?
ii.4 ) . the form ( [ bc ] ) which we have chosen
is adapted for the indicated use of these operator in the approximation . in particular
, the coupling parameter @xmath85 depends linearly on the cardinality of the set @xmath76 .
however , it is important to stress that our problem differs substantially form its three - dimensional analogue considered in @xcite where no sign of @xmath65 played a preferred role . in
the two - dimensional setting the coupling parameter tends to @xmath86 in the limit of weak and strong coupling , respectively .
consequently , we will be able to find an approximation for operators ( [ formal ] ) with _ attractive interactions _ only .
the krein s formula for the resolvent @xmath87 is the basic ingredient in the spectral analysis of the point - interaction hamiltonians .
let @xmath61 and @xmath88 be the matrix @xmath89 given by @xmath90 \delta_{xy } \nonumber \\ & & - g_k(x - y)(1-\delta_{xy})\ , , \end{aligned}\ ] ] where @xmath91 is the euler constant .
suppose that this matrix is invertible .
then @xmath92 and we have @xmath93^{-1 } ( y , y')\ , g_k ( x - y ) \nonumber \\ & & \phantom{aaaaa } \times r_0^k \psi \,(y ' ) \end{aligned}\ ] ] for any @xmath68 .
one can see easily that the matrix @xmath88 is invertible for sufficiently large negative @xmath94 .
indeed , the diagonal part is dominated by the term @xmath95 , while the non - diagonal elements vanish as @xmath96 , see the asymptotic formula @xcite for the hankel function @xmath97 . in view of our special choice of the coupling , an alternative way how to make the matrix @xmath98 invertible is to take a sufficiently large set @xmath76 .
[ lemma : lambda ] let @xmath99 and @xmath100 be a sequence of non - empty finite subsets of @xmath101 such that @xmath102 as @xmath103 and the following inequality holds @xmath104 then there exists a positive @xmath105 and @xmath106 such that the matrix @xmath107 is invertible and @xmath108 holds for all @xmath109 . here @xmath110 means the norm of the map from @xmath111 to @xmath112 ; the specification is superfluous here but it will be useful in the following . _
proof : _ let us first decompose the matrix @xmath113 into the diagonal and non - diagonal parts , @xmath114 and @xmath115 , respectively . for @xmath116 being large enough
the diagonal matrix @xmath114 is invertible and its operator norm in @xmath117 converges to @xmath65 as @xmath118 . due to the strict inequality in the hypothesis there is an @xmath119 such that the inequality ( [ schur test ] ) holds with @xmath65 replaced by @xmath120 .
then the schur - holmgren bound ( * ? ? ?
c ) yields @xmath121 , which in turn implies that the matrix sum @xmath122 is invertible for a sufficiently large @xmath116 . in analogy with the expression ( [ sing krein alt ] ) it is possible to rewrite krein s formula ( [ point krein ] ) in the form @xmath123 where @xmath124 solve the following system of equations , @xmath125 q_y - \sum_{y ' \in y , y'\neq y } g_k ( y - y ' ) q_{y ' } = r_0^k \psi \ , ( y)\,,\ ] ] for all @xmath126 .
with the above preliminaries we can proceed to the main goal we will prove that for a chosen generalized schrdinger operator with an attractive interaction one can find an approximating sequence of point - potential schrdinger operators under requirements which will be specified below .
the assumption about positions of the point potentials is obvious loosely speaking , as the sites of potentials are getting denser in the set @xmath2 , they must copy the measure @xmath127 . then we have to specify the coupling - constant behaviour in the approximating operators . we have already mentioned that in the analogous situation in dimension three the coupling constants scale by @xcite linearly with the number @xmath128 of point potentials , and suggested the same behaviour here .
this requires an explanation , because it is well known that the coupling constants are manifested differently in dimension three and two ; just consider a pair of point potentials and let their distance vary . to get a hint that the scaling behavior for the approximation remains nevertheless the same , consider an infinite straight polymer as in (
* iii.4 ) , denote the coupling constant by @xmath65 and the period by @xmath129 .
the threshold of the continuous spectrum is given as the unique solution @xmath130 to the implicit equation @xmath131 now let the number @xmath116 increase . if we want to keep the solution @xmath132 preserved as @xmath133 , then @xmath65 must grow linearly with @xmath116 ; recall that for @xmath134 this means that the individual point interactions are becoming _
weaker_. this motivates the choice of the coupling constants in the form @xmath135 which we made in the boundary condition ( [ bc ] ) .
now we can prove the announced approximation result .
let @xmath2 be a closed and non - empty subset of @xmath15 and let @xmath13 be a finite positive measure on borel @xmath14-algebra of @xmath2 with @xmath136 , which belongs to the kato class .
let @xmath137 be a nontrivial bounded continuous function .
choose @xmath138 with @xmath40 such that the equation ( [ eq sigma ] ) has a unique solution @xmath14 which has a bounded and continuous version on @xmath139 .
finally , suppose that there exists a sequence @xmath140 of non - empty finite subsets of @xmath101 such that @xmath102 and the following relations hold @xmath141 for any bounded continuous function @xmath142 , @xmath143 for @xmath144 .
the operators @xmath145 and @xmath36 defined in section [ prelim ] then satisfy the relation @xmath146 in the strong resolvent sense as @xmath133 .
_ proof : _ since both hamiltonians @xmath145 and @xmath30 are self - adjoint it is sufficient to prove the weak convergence , i.e. to check that @xmath147 holds for arbitrary @xmath148 . using the formulae ( [ sing krein alt ] ) and ( [ point krein alt ] ) for the resolvents we arrive at @xmath149 \\ & & + { 1\over |y_n| } \sum_{y ' \in y_n }
( r_0^k \bar{\phi})(y')\ , \sigma(y ' ) - \int ( r^k_{\mu,{{\rm d}}x } \bar{\phi})(y ) \,\sigma(y ) \mu ( d y ) .
\end{aligned}\ ] ] the sum of the last two terms tends to zero as @xmath144 , which follows from the hypothesis ( [ hypothesis 1 ] ) .
since the function @xmath150 is continuous and bounded , as we have already mentioned , it is enough to prove the following claim , @xmath151 for the @xmath152 norm , where @xmath153 , @xmath154 . to this end
, we employ the equations ( [ eq sigma ] ) and ( [ eq q ] ) and we substitute the term @xmath155 in one equation from the other .
the equation ( [ eq sigma ] ) holds @xmath127-a.e .
, so we must consider continuous representatives of the functions involved here . in this way
we get @xmath156 q_y \nonumber \\ & & - \sum_{y'\in y_n , y'\neq y } g_k(y - y ' ) q_{y'}. \end{aligned}\ ] ] by adding two extra terms to both sides of the equation we arrive at @xmath157 = - { \log ( i k/2)+c_e \over 2\pi |y_n| } \,\sigma(y ) \nonumber \\ + { 1 \over |y_n| } \sum_{y'\in y_n , y'\neq y } g_k(y - y ' ) \sigma(y ' ) -\int g_k(y - y ' ) \sigma(y ' ) \mu(d y ' ) .
\nonumber \end{aligned}\ ] ] we denote the vector on the r.h.s by @xmath158 , then the previous formula reads @xmath159 . since we assume that the inequality ( [ hypothesis 2 ] ) holds , the lemma [ lemma : lambda ] is applicable here .
therefore there exists @xmath106 such that the matrix @xmath160 is invertible for all @xmath161
. then we can write @xmath162 from the lemma [ lemma : lambda ] and the relation @xmath163 , @xmath164 is a operator acting on @xmath165 , we conclude that the operator norm in the inequality above is bounded by @xmath166 for some @xmath167 .
finally , let us look on the norm @xmath168 : the first term of @xmath169 converges to zero uniformly w.r.t @xmath170 as @xmath144 ( recall that @xmath14 is bounded ) and the hypothesis ( [ hypothesis 3 ] ) yields that remaining two terms go to zero as well .
let us now pass to examples . the first class of graphs to which we apply the approximation developed in the section [ main result ]
are rings , both full or open .
since the spectral properties of hamiltonians with interaction supported by these ring graphs were already explored in the paper @xcite see also the three - dimensional analogue discussed earlier in @xcite this gives us an opportunity to compare the approximation with the `` exact '' results , in particular , to asses the convergence rate of the approximation .
consider a circle @xmath171 with the radius @xmath172 and let @xmath16 be a function from @xmath2 to @xmath173 such that @xmath174}(\varphi)$ ] , where @xmath175 , @xmath6 and @xmath176 .
the schrdinger operator with the @xmath3 interaction supported by @xmath2 given formally by ( [ formal ] ) will be denoted by @xmath177 ; it can be given meaning in either of the two ways described in section [ prelim ] .
the spectrum and eigenfunctions are found easily in the full ring case , @xmath178 , when @xmath177 is reduced by the angular momentum subspaces .
every eigenstate except the ground state is twice degenerate .
in contrast , for a cut ring , @xmath179 , the spectrum is simple ; the problem can be solved numerically by the mode - matching method @xcite .
we start the presentation of the numerical results with the full ring .
first , we put @xmath180 and @xmath181 .
the discrete spectrum of @xmath177 consists of three eigenvalues , the ground state corresponds to the angular momentum @xmath182 , the other two correspond to @xmath183 , respectively , and they are twice degenerate ; the number of eigenvalues is given be the inequality @xmath184 .
the choice of the approximating point - potential operators @xmath145 is obvious
@xmath8 point potentials are spread periodically all over the circle and the coupling constant @xmath65 equals @xmath185 .
the task of finding the eigenvalues @xmath186 of @xmath145 means to solve the implicit equation @xmath187 we plot the eigenvalues of @xmath75 as @xmath188 increases in fig .
[ fig : conv1 ] .
on the number of point potentials @xmath8 for @xmath181 and the ring graph with @xmath180 .
the dotted lines are the exact eigenvalues @xmath189 , @xmath190 , and @xmath191.,width=340,height=226 ] on the number of point potentials @xmath8 for @xmath192 and the ring graph with @xmath180 .
the dotted lines are the exact eigenvalues @xmath193 , @xmath194 , @xmath195 , @xmath196 , and @xmath197.,width=340,height=226 ] in case of a stronger interaction , @xmath192 , one gets a similar picture , just the number of levels rises to five , see fig .
[ fig : conv2 ] .
note that the convergence of eigenvalues is slower than it is for the previous system . to estimate the rate of convergence , we calculate the difference between the exact eigenvalue and the eigenvalue computed using the approximation .
the result is shown in fig .
[ fig : convdif ] , we observe that the aforementioned difference decays roughly as @xmath198 with @xmath120 being less than 1 . in the logarithmic scale .
dotted line corresponds to @xmath180 and @xmath181 and the solid line corresponds to @xmath180 and @xmath192.,width=340,height=226 ] the approximation by point - potential schrdinger operators allows us to find easily the eigenfunctions . by , they can be written as a linear combination of the free green functions : @xmath199 where @xmath200 is the eigenvalue , i.e. @xmath201 , and @xmath202 is the solution to @xmath203 .
the eigenfunctions obtained in this way behave as one expects : they decrease exponentially if moving transversally away of the circle , and for @xmath204 they copy a sine function if moving along the circle
. a closer inspection of the eigenfunctions ( [ point eigf ] ) shows , of course , a logarithmic peak at the site of each point potential , as pictures fig .
[ fig : cir - eigf1 ] and fig .
[ fig : cir - eigf2 ] demonstrate . to our opinion
the contributions to energy coming from these spikes are responsible for the slow convergence of the approximation . ) for @xmath180 , @xmath205 and @xmath206.,width=340,height=226 ] ) for @xmath180 , @xmath205 and @xmath206.,width=340,height=226 ] next , we consider an open ring , @xmath207 , i.e. one sixth of the perimeter is missing .
for example , the fifth excited state for @xmath180 and @xmath192 has the energy @xmath208 , the corresponding eigenfunction is shown in fig .
[ fig : cir - e5 ] .
the approximation by 1000 point potentials yields the energy @xmath209 and the corresponding eigenfunction shown in fig .
[ fig : cir - app5 ] . for @xmath180 , @xmath192 and @xmath207 .
the solid line represents the ring @xmath2.,width=188,height=188 ] for @xmath210 , which approximates the fifth excited state of @xmath177 from fig .
[ fig : cir - e5 ] .
the solid lines represent the nodal lines , the dotted line represents the graph @xmath2.,width=226,height=188 ]
another class of leaky - graph systems to which we will apply the approximation by point - interaction hamiltonians are the star - graph hamiltonians where @xmath2 is a collection of segments coupled at a point . in distinction
to the previous section no `` direct '' method to solve the problem is available in this case , and thus the approximation represents the only way how to obtain a numerical description of the eigenvalues and eigenfunctions .
first we want to make a remark about finiteness of such graphs .
our interest concerns primarily infinite stars in which the arms are halflines , in particular , since they support localized states despite the fact that the graph geometry would allow escape to infinity . on the other hand
, the above approximation result applies to finite graphs only because it requires @xmath211 to be finite .
nevertheless , the result can be used , due to the fact that the infinite star hamiltonian is approximated , again in the strong resolvent sense , by a family of operators with cut - off stars . to justify this claim ,
it is sufficient to realize that the corresponding family of quadratic forms is by ( [ form ] ) monotonous and bounded from below , so that thm .
viii.3.11 of @xcite applies .
let us thus discuss in the beginning what can be derived analytically about infinite star graphs .
given an integer @xmath212 , consider an @xmath213-tuple @xmath214 of positive numbers such that @xmath215 denote @xmath216 , where we put conventionally @xmath217 .
let @xmath218 be the radial halfline , @xmath219 , which is naturally parametrized by its arc length @xmath220 .
the support of the interaction is given by @xmath221 .
the star - graph hamiltonians @xmath222 are defined formally by ( [ formal ] ) as schrdinger operators with an attractive potential supported by the graph @xmath2 and with the coupling constant @xmath16 , and the proper meaning is given to this operator in the way described in section [ prelim ] .
properties of the operator @xmath222 certainly depend on the order of the angles in @xmath223 .
however , the operators obtained from each other by a _ cyclic _
permutation are obviously unitarily equivalent by an appropriate rotation of the plane .
let us mention two trivial cases : \(a ) @xmath224 corresponding to a straight line can be written as @xmath225 , where @xmath226 is the one - center point - interaction hamiltonian @xcite on @xmath227 .
consequently , its spectrum is purely a.c . and equal to @xmath228 .
+ [ 1 mm ] ( b ) @xmath229 with @xmath230 again allows separation of variables being @xmath231 .
hence the a.c .
part of @xmath232 is the same as above , and in addition , there is a single isolated eigenvalue @xmath233 corresponding to the eigenfunction @xmath234 .
first we notice that the essential spectrum of @xmath222 does not extend below that of @xmath235 corresponding to a straight line .
@xmath236 holds for any @xmath8 and @xmath223 .
_ proof : _ by neumann bracketing .
we decompose the plane into a finite union @xmath237 where @xmath238 is a wedge of angle @xmath239 , @xmath240 is a halfstrip centered at @xmath218 which is obtained by a euclidean transformation of @xmath241 $ ] , and @xmath242 is the remaining polygon containing the vertex of @xmath2 .
imposing neumann boundary conditions at the common boundaries , we get a lower bound to @xmath222 by an operator which is a direct sum corresponding to the decomposition ( [ neub ] ) .
the wedge parts have an a.c .
spectrum in @xmath243 while the polygon has a purely discrete spectrum .
finally , the halfstrip spectrum is a.c .
again and consists of the interval @xmath244 , where @xmath245 is the lowest eigenvalues of @xmath246 on @xmath247)$ ] .
it is straightforward to see that to any @xmath248 there is @xmath249 such that @xmath250 , and since the decomposition ( [ neub ] ) can be chosen with @xmath249 arbitrarily large , the proof is finished . in fact , the essential spectrum is exactly that of a straight line .
@xmath251 holds for any @xmath8 and @xmath223 .
_ proof : _ in view of the previous proposition , it is sufficient to check that @xmath252 . given a function @xmath253 with @xmath254 and @xmath255 in the vicinity of @xmath256 , we construct @xmath257 with @xmath258 and @xmath259 , where the points @xmath260 can be chosen , e.g. , as @xmath261 .
it is easy to see that the vectors @xmath262 weakly as @xmath133 and that they form a weyl sequence of @xmath222 referring to the value @xmath263 .
this yields the desired result .
the first question concerns naturally the existence of isolated eigenvalues .
it follows from two observations of which one is rather simple .
[ addleg ] @xmath264 holds for any @xmath8 and angle sequence @xmath265 with @xmath266 .
_ proof _ follows directly from the definition by the quadratic form ( [ form ] ) .
on the other hand , the second one is rather nontrivial .
it has been proven in @xcite for a wide class of piecewise continuous non - straight curves which includes , in particular , a broken line .
[ exist - pr ] @xmath267 is nonempty unless @xmath268 . combining these two results with the minimax principle ( recall that @xmath222 is below bounded ) we arrive at the following conclusion .
[ exist - thm ] @xmath269 is nonempty except if @xmath9 and @xmath268 .
next one has to ask how many bound states does a star graph support .
the answer depends on its geometry .
there are situations , however , where their number can be large .
[ card - thm ] fix @xmath8 and a positive integer @xmath116 . if at least one of the angles @xmath239 is small enough , @xmath270 . _
proof : _ in view of proposition [ addleg ] it is again sufficient to check the claim for the operator @xmath271 . choose the coordinate system in such a way that the two `` arms '' correspond to @xmath272 .
we employ the following family of trial functions @xmath273 supported in the strip @xmath274 , with @xmath275 satisfying @xmath276 , and @xmath277 with @xmath278 .
let us ask under which conditions the value of the shifted energy form @xmath279 : = \|\nabla\phi\|^2 - { 2\gamma\over \cos(\beta/2 ) } \|f\|^2 + { \gamma^2\over 4 } \|\phi\|^2\ ] ] is negative .
since @xmath280 and @xmath281 , this is equivalent to @xmath282 by minimax principle the system there will be at least @xmath116 isolated eigenvalues provided @xmath283 where @xmath284 means an @xmath116-dimensional subspace in @xmath285)$ ] , i.e. , if @xmath286 now one should optimize the r.h.s . w.r.t .
@xmath287 , but for a rough estimate it is sufficient to take a particular value , say @xmath288 which yields @xmath289 it is obvious that the last inequality is for any fixed @xmath116 satisfied if @xmath223 is chosen small enough .
[ card - cor ] independently on @xmath223 , @xmath290 exceeds any fixed integer for @xmath8 large enough . the estimate used in the proof also shows that the number of the bound states for a sharply broken line is roughly proportional to the inverse angle , @xmath291 as @xmath292 .
this is the expected result , since the number is given by the length of the effective potential well which exists in the region where the two lines are so close that they roughly double the depth of the transverse well .
now we are going to show how the spectral problem for the operators @xmath222 can be reformulated in terms of suitable integral operators .
we will employ the resolvent formula for measure perturbations of the laplacian derived in @xcite .
notice that this technique was used in @xcite to derive a result which implies our proposition [ exist - pr ] . since the operators @xmath222 are defined by the quadratic form ( [ form ] ) they satisfy the generalized birman - schwinger principle . if @xmath37 belongs to the resolvent set of @xmath222 we put @xmath293 .
we already know the krein - like formula for the resolvent from thm [ gen krein ] , here it has the form @xmath294^{-1 } r^k_{m,{{\rm d}}x } , \ ] ] with @xmath13 denoting the dirac measure on @xmath2 .
one can express the generalized bs principle as follows @xcite : [ bs ] @xmath295 for any @xmath138 with @xmath40 .
consequently , the original spectral problem is in this way equivalent to finding solutions to the equation @xmath296 in @xmath297 , where @xmath298 .
furthermore , in analogy with ( * ? ? ?
ii.1 ) one expects that a non - normalized one corresponding to a solution @xmath299 of the above equation is @xmath300 a proof of this claim can be found in the sequel to the paper @xcite .
let us look now how the equation ( [ speceq ] ) looks like for graphs of the particular form considered here .
define @xmath301 with @xmath302 , in particular , @xmath303 . by @xmath304
we denote the operator @xmath305 with the kernel @xmath306 then ( [ speceq ] ) is equivalent to the matrix integral - operator equation @xmath307 on @xmath308 . notice that the above kernel has a monotonicity property , @xmath309 if @xmath310 .
this has the following easy consequence : [ mono2 ] each isolated eigenvalue @xmath311 of @xmath271 is an increasing function of @xmath223 in @xmath312 .
_ proof : _ notice first that the eigenfunction related to @xmath311 is even with respect to the interchange of the two halflines . without loss of generality we may assume that @xmath313 .
the odd part of @xmath271 then corresponds to the dirichlet condition at @xmath314 .
the spectrum of this operator remains the same if we change @xmath315 to @xmath316 .
removing then the dirichlet condition , we get the operator @xmath235 with @xmath317 , so by minimax principle no eigenfunction of @xmath271 can be odd . on the symmetric subspace the diagonal matrix element of the matrix integral operator @xmath318 equals @xmath319 with the first term at the r.h.s .
independent of @xmath223 .
next we we notice that @xmath320 is a type ( a ) analytic family around any @xmath321 ; the derivative @xmath322 is a bounded operator or the form @xmath323 where @xmath324 has the kernel @xmath325 at the same time @xmath320 is a type ( a ) analytic family w.r.t . @xmath132 around any @xmath326 and the corresponding derivative is a bounded operator @xmath327 with the kernel @xmath328 let @xmath329 be a normalized eigenvector of @xmath330 corresponding to the eigenvalue @xmath331 . by feynman - hellmann theorem
we find @xmath332 and similarly @xmath333 .
the solution @xmath334 of the implicit equation @xmath335 thus satisfies @xmath336 and consequently , the eigenvalue @xmath337 is increasing w.r.t .
@xmath223 .
having explained analytically how does the discrete spectrum of @xmath222 depend on the number of arms and the angles @xmath223 , we employ now the approximation by point - potential schrdinger operators to obtain the numerical results for cut - off stars which would illustrate the above conclusions .
consider a two - arms star graph @xmath338 with both arms having the same length , @xmath339 , and put @xmath340 .
as we already know , the threshold of the continuous spectrum of @xmath271 is @xmath341 .
we approximate @xmath271 by point - potential schrdinger operator @xmath75 that has 1 potential on the center of the star graph and @xmath342 equidistant point potentials on each arm .
only the lower part of the discrete spectrum of @xmath75 approximates the discrete spectrum of the star - graph operator , while the upper part corresponds to the interval @xmath343
\subset \sigma_{{\rm ess}}(h_2(\beta))$ ] .
therefore only the states with energy below the threshold may be indeed understood as the approximation of the bound states of @xmath271 . on the angle @xmath223 for the symmetric two - arms star @xmath338 .
the dotted line represents the threshold @xmath344.,width=340,height=226 ] on the angle @xmath223 for the non - symmetric two - arms star @xmath345 .
the dotted line represents the threshold @xmath344.,width=340,height=226 ] the main result of the analytic argument presented above was the dependence of the eigenvalues on the angle @xmath223 : if @xmath223 decreases , the eigenvalues decrease and their number grows .
the discrete spectrum of @xmath75 for @xmath223 varying is in a good agreement with this fact as fig . [
fig : star - spec1 ] illustrates .
the eigenvalue crossings we observe are actual , which is a consequence of the symmetry of the graph @xmath338 .
for a non - symmetric graph the crossings become avoided . to demonstrate it in fig . [
fig : star - spec2 ] , we slightly change the length of one arm , @xmath346 , while the other parameters are preserved .
also the setting of the point potentials approximating the new graph @xmath345 stays the same , up to extra four potentials added on the longer arm . , @xmath347 , which approximates the ground state of @xmath348 with @xmath223 being the 5-tuple of @xmath349 and @xmath192 .
the contours correspond to logarithmically scaled horizontal cuts , the dotted lines represents the graph @xmath2.,width=188,height=188 ] , @xmath350 .
it approximates the third excited state of @xmath351 with @xmath223 being the 9-tuple of @xmath352 and @xmath192 .
the convention is the same as above , the solid line is the nodal line.,width=226,height=188 ] the expression ( [ point eigf ] ) again yields the eigenfunctions in the form of a sum of the free green functions .
we limit ourselves to a pair of examples : the ground state of @xmath348 in fig .
[ fig : star - eigf0 ] and the third excited state of @xmath351 in fig .
[ fig : star - eigf3 ] .
the length of the cut arms is 30 , in the former case the approximating operator @xmath75 has 601 point potentials and in the latter case the number of point potentials is 1001 .
we see , in particular , that for @xmath8 large enough @xmath353 may have closed nodal lines ; it is an interesting question what is the minimal @xmath8 for which this happens .
our last example , as indicated in the introduction , concerns the situation when @xmath2 is a single infinite curve ; we want to see whether resonances in the scattering of states propagating along @xmath2 may be detected by inspecting the spectrum of the cut - off problem with the curve of a finite length which is a parameter to be varied .
consider a curve @xmath2 from fig .
[ fig : res - graph ] : its central part consists of three segments of a circle with the radius @xmath180 and it has two infinite `` legs '' .
the distance between two closest points of the curve , i.e. the bottleneck width , is denoted by @xmath358 .
the coupling constant @xmath16 equals 1 .
we cut the `` legs '' of the graph to a finite length @xmath355 and we plot the eigenvalues computed using the approximation for @xmath355 varying .
the number of the point potentials involved in @xmath75 is chosen so that the distance between every two adjacent points equals @xmath359 . for small values @xmath354 and @xmath356 the tunneling effect occurs and we can see the _ plateaux _ in figs . [
fig : res - cir1 ] , [ fig : res - cir2 ] , which indicate the existence of resonances ; the width of the avoided crossings increase with @xmath358 as expected . on the other hand , for more open curve , @xmath357
, we get a different picture , where the _ plateaux _ are absent , see fig . [
fig : res - cir3 ] .
the second graph type we are interested in are simple bends , in terms of section [ stargraph ] it is a two - arms star @xmath360 . as we have already mentioned , the avoided eigenvalue crossings are not expected here because the transport along the graph arm involves a single transverse mode this is confirmed in fig .
[ fig : res - star ] , which shows the results of the cut - off method for @xmath361 .
finally to illustrate that resonances may also result from multiple reflections rather than from a tunneling , we apply the cut - off method to graphs which are of a stair - type , or z - shaped .
they consists of three line segments : the central one has a finite length @xmath362 , i.e. the `` heigth '' of the stair , the other two are parallel ( cut - off ) halflines .
if the angle @xmath363 between the segments is less than @xmath364 the term z - shaped is more appropriate .
the results for @xmath180 and @xmath205 are plotted in figs .
[ fig : res - stair1 ] and [ fig : res - stair2 ] . in the former case
the stair is `` skewed '' to @xmath365 , in the latter we have @xmath366 ; the distance between the point potentials of the approximating operator @xmath75 equals @xmath367 .
we find avoided crossings , however , they are very narrow for @xmath365 and barely visible in the right - angle case ; this observation can be naturally understood in terms of the reflection probability through a single bend
compare with the angle dependence of the spectrum on fig .
[ fig : star - spec1 ] .
p. exner : spectral properties of schrdinger operators with a strongly attractive @xmath3 interaction supported by a surface , in _ proceedings of the nsf summer research conference ( mt . holyoke 2002 ) _ ; ams contemporary mathematics " series , providence , r.i .
, 2003 ; to appear .
|
we prove an approximation result showing how operators of the type @xmath0 in @xmath1 , where @xmath2 is a graph , can be modeled in the strong resolvent sense by point - interaction hamiltonians with an appropriate arrangement of the @xmath3 potentials .
the result is illustrated on finding the spectral properties in cases when @xmath2 is a ring or a star .
furthermore , we use this method to indicate that scattering on an infinite curve @xmath2 which is locally close to a loop shape or has multiple bends may exhibit resonances due to quantum tunneling or repeated reflections . _
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 1 nuclear physics institute , academy of sciences , 25068 e + near prague , czechia + 2 doppler institute , czech technical university , behov 7 , + 11519 prague , czechia + 3 institute of theoretical physics , fmp , charles university , + v holeovikch 2 , 18000 prague , czechia _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
|
the hydration thermodynamics of analogs of amino acid side - chains or of amino acid side - chains in small model peptides have often been used to understand protein folding and protein - protein association .
examples of such approaches abound in the biochemical literature . an admittedly biased list of a very small number of the many pioneering investigations where this approach has been used include work identifying hydrophobicity as a dominant force in protein folding @xcite , investigations of protein denaturation @xcite , interpretation of calorimetric data on protein unfolding @xcite , and more recently , investigations on the role of osmolytes in protein folding @xcite . drawing upon insights attributed to cohn and edsall , tanford @xcite formulated this approach into a quantitative , predictive framework . in his approach ,
the free energy of unfolding is given as a sum of the free energy of transfer of `` the small component groups of the molecule , from the environment they have in the native form , to the environment they have in the unfolded form '' @xcite .
accounting for subsequent refinements that included corrections for the solvent - exposure of the `` small component groups '' ( cf .
ref . ) , the generic equation of the unfolding free energy ( @xmath1 ) can be written as @xmath2 where @xmath3 is the free energy of transferring the group @xmath4 from some reference phase to liquid water , and @xmath5 is a factor that corrects for the change in solvent - exposure of the group @xmath4 in protein unfolding . in interpretation of calorimetric data @xcite ,
for example , a gas - phase reference is natural , as is also the case in this article .
the same form of equation can also be used to describe the effect of osmolytes on protein unfolding : identifying @xmath3 with the free energy for transferring the group between water and the aqueous osmolyte solution and @xmath5 with the change in the solvent exposure upon unfolding in the osmolyte solution relative to water , the above equation directly gives the so - called @xmath6-value for 1 m osmolyte @xcite . at a time when theory , simulations , and experimental techniques were much less developed than they are now , the group - additive approach a pragmatic first step to understand the hydration thermodynamics of a complicated macromolecule .
but it is also important to assess its limitations , and probe if the _ physical conclusions _ based on this approach are valid .
for example , the group - additive model appears to capture the effect of osmolytes rather well @xcite : the predicted and experimentally determined @xmath6-values agree to within a couple kcal / mol for proteins with about 100 residues . on this basis
it has been suggested that osmolytes largely act by tuning the solubility of the peptide backbone @xcite .
our recent simulation - based scrutiny of the transfer for a peptide , however , suggests that the good agreement in @xmath6-values likely arises because solvent - mediated correlations between two groups largely cancel in the water - to - osmolyte transfer @xcite .
further , in the vapor - to - liquid transfer , a situation where solvent - mediated correlations are preserved , the identified group - additive contribution of the peptide group depends on whether blocked - gly@xmath7 or cyclic - diglycine was used as a model : that is the group - contribution was not context independent as often assumed .
building on the work noted above @xcite , and other recent studies that have exposed limitations of the additive model @xcite , here we study how well a model of the side - chain describes the hydration of the side - chain and , most importantly , whether the physical conclusions about hydrophobicity of a model are applicable to the side - chain .
the problem we consider is the vapor - to - water transfer ( hydration ) of an isoleucine side - chain in the context of model extended and helical peptides .
butane in the _ gauche _ conformation , matching exactly the side - chain conformation of the isoleucine in extended peptides , is used as a small - molecule analog of the isoleucine side - chain .
as before , we parse the hydration free energy into hydrophobic and hydrophilic contributions .
we find that the net hydration free energy of the target peptide igggg or ggigg is predicted to within 2 kcal / mol ( in a net free energy of about 30 kcal / mol ) using the hydration free energy of ggggg and @xmath0-butane , appropriately adjusting for solvent exposure of ggggg and isoleucine side - chain in the target peptide .
but , the underlying components convey a different picture .
using the @xmath0-butane model will cause one to overestimate substantially the hydrophobicity of the isoleucine side - chain in the peptide models .
since the @xmath0-butane model also underestimates the hydrophilic contributions , the net chemical potential works out reasonably well .
importantly , using group - additivity the magnitude of the excess enthalpy and entropy of hydration are also incorrectly predicted , but these also balance one - another .
similar trends are also found when we use the conditional hydration free energy of isoleucine in ggigg to predict the hydration thermodynamics of model helical deca - peptides .
our work suggests that while the group - additive approach may describe the net free energy reasonably well , caution is necessary in drawing physical conclusions based on this observed agreement .
the excess chemical potential @xmath8 and its enthalpic ( @xmath9 ) and entropic ( @xmath10 ) components are of primer concern in this work .
the excess chemical potential is that part of the free energy that would vanish if all intermolecular interactions were neglected .
formally , @xmath11 , where it is understood that @xmath8 is referenced relative to the ideal gas at the same density and temperature .
the averaging @xmath12 is over the solute - solvent binding energy distribution @xmath13 and @xmath14 , with @xmath15 the temperature and @xmath16 the boltzmann constant .
the direct approach to calculate @xmath8 by characterizing @xmath13 usually fails since the high - energy ( high-@xmath17 ) regions of @xmath13 are never well - sampled in simulations .
the usual approach then is to calculate @xmath8 by accumulating the work done in changing the strength of the solute - solvent interaction from a non - interacting solute to the fully - interacting solute .
here we approach the problem differently , with a view to better dissecting the physics of hydration .
imagine an inner - shell ( or hydration shell ) , defined by a length parameter @xmath18 , around the solute . at equilibrium ,
the solute samples a distribution of coordination states @xmath19 , where @xmath20 is the number of water molecules within the inner shell .
the association of the @xmath20-water molecules with the solute can be described in terms of an equilibrium constant @xmath21 .
a basic mass - balance then gives the probability @xmath22 that the inner - shell is empty of solvent by @xmath23 , where @xmath24 is the bulk density of water @xcite .
the free energy to allow water molecules to populate a formerly empty inner - shell is just @xmath25 . for a solute with an empty inner - shell the solute - solvent interactions are necessarily of a longer - range than for the solute without this restriction .
but these long - range interactions are better behaved , such that for a sufficiently large inner - shell ( typically up to the first hydration shell ) , @xmath26 , the probability density of the solute - solvent binding energy with the restriction of an empty inner - shell ( denoted by @xmath27 ) , is gaussian .
thus the specification of the inner - shell helps separate strong , short - range solute - water interactions from weaker , non - specific , longer - range interactions .
we can consider an analogous process of solvent clustering within the same defined inner - shell but in the absence of the solute ; @xmath28 , the probability to find a cavity in the solvent , has an expansion similar to @xmath22 .
the quantity @xmath29 is the work done to create a cavity in the liquid and is of principal interest in understanding hydrophobic hydration @xcite . in terms of the short - range chemical , long - range nonspecific , and cavity ( packing ) contributions , we then have @xcite @xmath30}_{\rm local\ ; chemistry } \underbrace{- \ln p_0[\phi_\lambda]}_{\rm packing } + \underbrace{\beta\mu^{\rm ex}[p(\varepsilon|\phi_\lambda)]}_{\rm long - range\ ; or\ ; outer } \ ; . \label{eq : qc}\end{aligned}\ ] ] figure [ fg : cycle ] provides a schematic of the decomposition of @xmath8 according to eq .
[ eq : qc ] .
the constraint @xmath27 is essentially an a field of range @xmath18 that is used to regularize the problem of calculating @xmath8 from @xmath13 . ) .
schematic reproduced from ref . with permission from elsevier.,width=312 ] in simulations , we grow the external field @xmath27 ( instead of changing the underlying system hamiltonian ) to calculate @xmath22 and @xmath28 @xcite . in practice , the field has a soft - boundary at @xmath18 @xcite , unlike the sharp - demarcation suggested in the schematic .
it is straightforward to correct for this effect @xcite , but we do not pursue that here . for @xmath31 a gaussian , we have @xmath32 = \langle \varepsilon | \phi_\lambda\rangle + \frac{\beta}{2 } \langle \delta\varepsilon^2 | \phi_\lambda\rangle \ , .
\label{eq : gaussian}\end{aligned}\ ] ] the excess entropy of hydration of the solute is given by @xmath33 where @xmath34 , the excess energy of hydration , is essentially equal to the excess enthalpy ( @xmath9 ) of hydration .
the above equation ignores small corrections due to a finite excess volume of hydration and corrections arising from the thermal expansion coefficient and the isothermal compressibility of the pure liquid .
the above framework is particularly apposite to investigate the additive description of @xmath8 . to this end , for simplicity consider the hydration of a dimer aa . within the additive model
the packing contribution of aa is obtained from the packing contribution of a alone .
given that a cavity is already formed around one center , the work required to create the cavity around the second center will necessarily be less than the work required to form the same cavity in isolation .
this is because for the cavity in isolation more water molecules are pushed away and a larger interface is created than when another cavity is in proximity .
thus a naive additive approximation of the cavity contribution for the pair aa will be more positive than dictated by the molecular reality . for the same reasons , less work is needed to empty the inner - shell around the pair aa than given by the sum of the chemistry contribution for each a. that is , the pair additive description of the chemistry contribution will always be more negative than dictated by molecular reality
. such correlation effects are also present in the long - range contribution @xcite . in this case , however , the correlated fluctuation of binding energies can either raise or lower the free energy .
for aa , we expect the favorable interaction of the solvent with one center to come at the expense of favorable interaction with the other center . for such anti - correlated binding energies ,
the free energy of the pair is lowered , as has been discussed earlier @xcite .
even simple glycyl - peptides show context - dependent positive and negative correlations . in practice ,
the @xmath8 of a is usually scaled by extent of solvent exposure in aa to model the @xmath8 of the dimer ( see eq .
[ eq : areascaling ] below ) . in terms of the quasichemical decomposition
, we can appreciate that the area - scaling tries to correct for the overestimation of the packing contribution and the underestimation of the chemistry contribution .
how a a local area - based scaling can correct long - range interaction contributions is difficult to appreciate , but this is precisely what is done in experimental analysis .
the pentapeptides ggggg , ggigg , and igggg are modeled in the extended configuration with the long axis aligned with the diagonal of the simulation cell and the center of the peptide placed at the center of the simulation cell .
helical deca - glycine ( g@xmath35 g ) and a helical peptide with nine alanine and one glycine residues ( a@xmath35 g ) served as the reference for the helical peptides , g@xmath35i and a@xmath35i , respectively . in the helical peptides ,
isoleucine was substituted at position 6 , roughly in the center of the peptide .
butane in the _ gauche _ conformation was built using the isoleucine conformation in the extended pentapeptide .
the g@xmath35 g and a@xmath35 g peptides were built from a helical deca - alanine .
these structures were extensively energy minimized with weak restraints on heavy atoms to prevent fraying of the helix .
the g@xmath35i and a@xmath35i helices were built by appropriately grafting the conformation of isoleucine in the ggigg system onto position 6 .
( thus we ensure that the internal conformation of the isoleucine is the same in both @xmath0-butane , ggigg , and the helical peptides . )
all peptides had an acetylated ( ace ) n - terminus and an n - methyl - amide ( nme ) c - terminus .
after energy minimization , the peptide atoms are held fixed throughout the subsequent simulation .
the solvent was modeled by the tip3p @xcite model and the charmm @xcite forcefield with correction terms for dihedral angles @xcite was used for the peptide .
a total of 2006 tip3p molecules solvated the pentapeptide ; 3500 water molecules were used for studies with helical peptides .
the conditional hydration free energy of isoleucine in ggigg relative to the model ggggg is defined by @xmath36 = \mu^{\rm ex}[{\rm ggigg } ] - \mu^{\rm ex}[{\rm ggggg}]$ ] @xcite .
similar definitions apply to the excess enthalpy and entropy of hydration and to isoleucine in context of other reference systems . in experiments
it is often the case that the hydration thermodynamics of the gxg tripeptide is compared with the ggg peptide to ascertain the contribution of x alone .
the zwitterionic amino acid x alone is not used primarily to minimize end - effects . by this yardstick
, we should expect the ggigg peptide constrained to be in the extended conformation to be an even more conservative model to ascertain the contribution of the isoleucine side - chain . capping the ends with ace and nme
further serves to dampen end - effects arising from long - range electrostatic interactions .
the free energy calculations and error analysis exactly follow the procedure described earlier @xcite .
briefly , the free energy to apply the field is obtained by growing the field between 0 and 5 .
we use gauss - legendre quadratures with five gauss - points per unit ngstrom to integrate the force exerted by the field on the solvent molecules versus the distance @xmath18 . at each gauss - point , the system was equilibrated for 0.5 ns and data collected over the subsequent 0.5 ns .
the long - range contribution was obtained by performing particle insertion calculations in the appropriate molecular - shaped cavity ( cf .
[ fg : cycle ] ) to calculate @xmath31 ( @xmath37 ) .
water with the appropriate cavity was simulated for 1 ns and 1250 frames from the last 0.625 ns used for analysis . confirming the gaussian - distribution of binding energies , particle extraction calculations agree with the particle insertion procedure ( data not shown ) . the excess energy was obtained by adapting the shell - wise calculation procedure used earlier for studying the hydration of methane @xcite . for the peptides , water molecules in the first two hydration shells
were considered , where the hydration shell is defined by the union of shells of radius @xmath18 centered on the heavy atoms . for ease ,
@xmath38 defined the first shell , @xmath39 defined the second shell , and @xmath40 defined the third shell .
we find that the excess energy contributions from the third shell is close to zero , justifying our use of just the first two shells for calculating @xmath9 . for calculating the excess energy we equilibrated the solvated peptide system an additional 1.5 ns ( beyond what was used in the free energy calculation ) , and the propagated the trajectory for an additional 3 ns , collecting data every 500 ps for a total of 6000 frames .
entropies obtained using eq . [ eq : entropy ] agree with entropy calculated using @xmath41 ( data not shown ) .
further details on the entropy calculations will be described separately . in experiments ,
conditional or small molecule thermodynamic quantities are used in reconstructing the free energy of a macromolecule by scaling these quantities by the solvent - exposure in the macromolecule .
we follow this procedure and find solvent accessible surface area ( sasa ) using a standard code @xcite and bondi radii @xcite .
the area - scaled group - additive estimate of the free energy of the solute @xmath42 is given by @xmath43 where @xmath44 refers to the side - chain model ( either @xmath0-butane or isoleucine in ggigg ) , @xmath45 is the hydration free energy of the reference for the remaining peptide , and @xmath46 and @xmath47 are , respectively , the fractional solvent exposure of the side - chain and of the reference in @xmath42 . as an example of this procedure , to find the group - additive free energy of the g@xmath35i helix using the conditional hydration of isoleucine in the ggigg peptide , we compute the sasa of g@xmath35i , @xmath48 , and
identify the contribution due to the side - chain alone , @xmath49 .
the contribution of the rest of the peptide is then @xmath50 .
we likewise find the sasa of isoleucine in the reference ggigg , @xmath51 , and of the g@xmath35 g peptide , @xmath52 .
then the group - additive free energy of g@xmath35i is given by @xmath53 = \alpha_{sc } \cdot \mu^{\rm ex } [ { \rm i } | { \rm ggigg } ] + \alpha_{back } \cdot \mu^{\rm ex } [ { \rm g_9g}]$ ] , with @xmath54 and @xmath55 .
for an additive model to be satisfactory in predicting the thermodynamics of folding , dill@xcite has suggested that the allowable error per group for a modest 100 amino - acid protein is about 10 cal / mol / group ( or about 0.02 @xmath56/group at 298 k ) .
this order of magnitude estimate sets the scale against which we can compare the success or failure of the additive approach . beyond the numerical prediction of thermodynamic quantities ,
our aim is also to question if the physical conclusions based on translating small - molecule hydration data can be used to understand the hydration of a macromolecule .
table [ tb : penta ] summarizes the results on the conditional hydration of isoleucine in the ggigg and ggggg pentapeptides . for reference
we also show the value expected from @xmath0-butane . [
cols="^,>,>,>,>,>,>",options="header " , ] it is evident that the reconstructed free energy of the helices deviate from the simulated values by at least 1.5 kcal / mole , and the small net deviation results from large compensating deviations in the predicted chemical , packing , and long - range contribution .
for both g@xmath35i and a@xmath35i , the magnitude of the minimum error is set by the errors in the long - range contribution to hydration . as suggested in sec .
[ sec : qual ] , trying to model the long - range contributions using a scaling based on surface areas will have limitations : the surface area scaling is effectively a local construct , whereas the long - range interactions , as the name indicates , involves interaction of the group with the entire solvent bath .
the above analysis shows that one of the most often used premises to model and/or interpret protein hydration , namely using data from gxg after scaling for solvent - exposure , is inadequate in predicting the hydration of the same residue in a somewhat different context .
therefore , all results based on such constructions must be considered with caution .
there are two main conclusions that emerge from this work .
first , the conditional solvation of an amino acid in a peptide model is rather different from the solvation of the corresponding amino acid side - chain in bulk solvent .
this point is has been made before , most notably by ben - naim .
what we show here is that assuming a non - polar solute ( @xmath0-butane ) models the side - chain of isoleucine one would ascribe to the isoleucine residue a greater hydrophobicity than is warranted .
we anticipate that even the temperature dependencies of the conditional packing contribution to be rather different from the packing contribution for the nonpolar solute . on this basis concerns
are raised about drawing conclusions about the presumed dominance of hydrophobic hydration in protein folding based on small molecule hydration .
the second allied conclusion is that even if an additive description of free energy predicts the numerical values in reasonable accord with experiments , it is likely that this success arises primarily due to compensating failures in describing the underlying physics .
we thank dr .
gillian lynch for providing the structure of a deca - alanine in the helical conformation .
we thank arieh ben - naim for helpful comments on the manuscript .
bmp gratefully acknowledge the financial support of the national institutes of health ( gm 037657 ) , the national science foundation ( che-1152876 ) and the robert a. welch foundation ( h-0037 ) .
this research used resources of the national energy research scientific computing center , which is supported by the office of science of the u.s .
department of energy under contract no .
de- ac02 - 05ch11231 .
tomar , d. s. , v. weber , and d. asthagiri , 2013 .
solvation free energy of the peptide group : its model dependence and implications for the additive transfer free energy model .
_ biophys .
j. _ 105:14821490 .
pratt , l. r. , and d. asthagiri , 2007 .
potential distribution methods and free energy models of molecular solutions . _ in _ c. chipot , and a. pohorille , editors , free energy calculations : theory and applications in chemistry and biology , springer , berlin , de , volume 86 of _ springer series in chemical physics _ , chapter 9 , 323351 .
weber , v. , s. merchant , and d. asthagiri , 2011 .
regularizing binding energy distributions and thermodynamics of hydration : theory and application to water modeled with classical and _ ab initio _ simulations .
_ j. chem .
_ 135:181101 .
, a. d. , m. feig , and c. l. brooks iii , 2004 . extending the treatment of backbone energetics in protein force fields : limitations of gas - phase quantum mechanics in reproducing protein conformational distributions in molecular dynamics simulations .
_ j. comp .
_ 25:14001415 .
|
the hydration thermodynamics of the gxg tripeptide relative to the reference ggg is often used to define the _ conditional _ hydration contribution of x. this quantity or the hydration thermodynamics of a small molecule analog of the side - chain or some combination of such estimates , usually including factors to account for the solvent - exposure of the side - chain in the protein , have anchored the interpretation of seminal experimental studies in understanding protein stability and folding . using simulations
we show that such procedures to model protein hydration thermodynamics have significant limitations .
we study the conditional hydration thermodynamics of the isoleucine side - chain in an extended pentapeptide and in helical deca - peptides using corresponding extended penta - glycine or helical deca - peptides as reference .
hydration of butane in the _ gauche _ conformation provides a small molecule reference for the side - chain .
we use the quasichemical approach to parse the hydration thermodynamics into chemical , packing , and long - range interaction contributions .
the chemical contribution of @xmath0-butane , reflecting the contribution of solvent clustering within the defined inner - shell of the solute , is substantially more negative than the conditional chemical contribution of isoleucine .
the packing contribution gives the work required to create a cavity in the solvent , a quantity of central interest in understanding hydrophobic hydration .
the packing contribution for @xmath0-butane substantially overestimates the conditional packing of isoleucine .
the net of such compensating contributions disagrees with the conditional free energy of isoleucine but by a lesser magnitude .
the excess enthalpy and entropy of hydration of @xmath0-butane model are also more negative than the corresponding conditional quantities for the side - chain .
the conditional solvation of isoleucine in ggigg also proves unsatisfactory in describing the conditional solvation of isoleucine in the helical peptides .
|
elements up to the iron peak in the solar abundance distribution can be made in hydrostatic stellar burning processes , whereas heavier elements require extremer conditions , such as explosions resulting from thermonuclear burning or rapid ejection and decompression of gravitationally condensed and heated material .
the former are connected to he - shell flashes in stars with less than 8 solar masses and to type ia supernovae , the latter are realized in core - collapse supernovae and neutron star mergers .
although nuclear processes under explosive conditions depending on the specific phenomenon may also produce nuclei around and below the fe - ni region , heavier nuclides can not be produced in conditions encountered in hydrostatic burning . due to the different temperature and density ranges of explosive burning with respect to hydrostatic burning and due to the fact that heavier nuclei are involved , also different aspects of the nuclear processes have to be studied and the behavior of reaction sequences
is often governed by different considerations than those for lighter nuclei .
this is especially important for experimental studies which not only may encounter additional problems when studying heavy nuclei but also have to be adapted in order to extract the data actually important for constraining nucleosynthesis .
nuclear theory also faces different challenges in heavy nuclei than in light ones and has to focus on the prediction of the actually relevant nuclear properties when applied to nucleosynthesis .
finally , a 3-d hydrodynamical simulation of exploding dense matter not only probes the limits of our nuclear models but also is a considerable computational challenge . the combined astrophysical and nuclear uncertainties lead to generally less well constrained conditions for nucleosynthesis in such phenomena and thus to considerable leeway in the interpretation and consolidation of astrophysical models .
in this section , i want to highlight the special considerations that apply to the nuclear aspects of nucleosynthesis of intermediate and heavy nuclei . some of the connected astrophysical uncertainties will be summarized in later sections .
the connection between the astrophysical and the nuclear physics aspects of nucleosynthesis , in general , is made in the stellar reaction rate @xmath4 , which for any reaction @xmath5 is derived from the product of effective reaction area @xmath6 and the flux of incoming projectiles @xmath7 , where @xmath8 , @xmath9 are the number densities of target nuclei and projectile nuclei , respectively , in the stellar plasma with temperature @xmath10 , and @xmath11 is the relative velocity between the reaction partners @xmath12 and @xmath13 .
the stellar reaction cross section @xmath14 also depends on @xmath10 because of the possible presence of nuclei in excited states and not only in the ground state ( g.s . ) due to thermal excitation in the plasma .
the implications of this will be discussed in later sections . to obtain the number of reactions per volume and per second , an integration over all relative velocities has to be performed , where the probability to find particles at a velocity @xmath15 is given by the maxwell - boltzmann distribution ( mbd ) @xmath16 and thus @xmath17 using center - of - mass ( c.m . )
coordinates , with @xmath18 being the reduced mass , and inserting the mbd explicitly , this can also be rewritten as integration over the c.m .
energy @xmath19 when realizing that @xmath20 and @xmath21 , @xmath22 which allows to use the more common form of an energy - dependent reaction cross section @xmath23 ( see also eqs . [ eq : effcs ] , [ eq : stellcs]).@xcite instead of number densities @xmath24 , astrophysical investigations often use abundances @xmath25 , which implicitly contain the plasma matter density @xmath26 and avogadro s number @xmath27 . in explosive burning
, however , it is not always necessary to determine @xmath4 at all temperatures . as introduced above
, explosive burning involves high temperatures and sometimes also high densities .
typical temperatures and densities depend on the actual site and process but relevant temperatures , at which reaction rates have to be known , do not exceed 3.5 gk because forward and reverse rates are in equilibrium when the plasma is hotter .
this means that the timescales for both forward and reverse reaction are short enough compared to the process timescale ( the time during which temperature and density remain at values allowing the reactions to significantly modify the abundances ) to permit the abundances of all involved nuclei to reach their equilibrium values , given by @xcite @xmath28 for a reaction @xmath29 and @xmath30 for a reaction @xmath31 ( see also eq .
[ eq : revphoto ] ) , where @xmath32 are the g.s .
spin factors , @xmath33 the normalized partition functions for a nuclide @xmath34 , and @xmath35 the reaction @xmath36 value when starting with @xmath37 . note that this does not imply that the abundances remain constant , they still depend on temperature @xmath10 which may vary with time as well as on @xmath38 and @xmath39 , which also depend on the plasma density @xmath26 .
the fact that individual rates do not appear anymore implies that the rates have only to be determined up to about 3.5 gk while the reaction @xmath36 values ( derived from nuclear masses ) determine abundances at higher temperature .
it should further be noted that not all reactions may be in equilibrium for given conditions .
if all reactions were in equilibrium , this would lead to full _ nuclear statistical equilibrium _ for which all abundances can be computed from coupled equations without the use of rates.@xcite on the other hand , any subset of reactions may be in equilibrium ( e.g. , neutron or proton captures and their reverse reactions ) and the above equations can then be used to calculate the abundances affected by these reactions ( which will be the fastest reactions occuring and thus be dominating ) . the lower temperature limit is not defined as clearly as it depends much more on reaction cross sections , process duration , and the temperature and density profiles obtained in the astrophysical models . generally , charged particle reactions freeze out at higher temperatures than reactions with neutrons due to the coulomb barrier and are mostly too slow to change abundances considerably at @xmath40 gk .
equilibrium considerations limit the temperature range in which astrophysical rates have to be known . since the relevant energy window is determined by the energies giving non - negligible contributions to the integral in eq .
[ eq : stellrate ] for a chosen temperature @xmath10 , the thermal energies of the interacting nuclei are low by nuclear physics standards .
these energies are below 10 mev for charged particles and up to a few 100 kev for reactions with neutrons.@xcite especially for charged particle reactions on trans - iron nuclei , this implies tiny reaction cross sections which pose a significant challenges to their experimental determination . of great importance in the nucleosynthesis of trans - iron nuclides is the fact that heavier nuclei have a higher nuclear level density ( ld ) and thus smaller level spacing than light species .
this intrinsic ld is important in three ways .
firstly , the ld at the formation energy of the compound nucleus ( which is the sum of incident projectile energy and separation energy of the projectile in the compound nucleus , @xmath41 ) determines the dominant reaction mechanism . in the absence of levels to be populated close to the given energy ,
_ direct reactions _ to low - lying final states dominate the reaction cross section.@xcite when there is a small number of well separated excited states close to @xmath42 in the compound nucleus @xmath37 , this gives rise to resonances in the reaction cross sections , which can be described by the breit - wigner formula or by the r - matrix method.@xcite the challenge therein is to determine the properties of the resonances contributing to the reaction rate integral in eq .
( [ eq : stellrate]).@xcite it is experimentally challenging to perform the required measurements for unstable nuclei and/or at low energy and theoretical _ ab initio _ methods can not be applied to heavy nuclei , yet .
an extreme case of resonant reactions appears when the ld is high , leading to a large number of unresolved resonances which can be described by averaged resonance parameters .
this is called the hauser - feshbach approach.@xcite the vast majority of reactions in nucleosynthesis of intermediate and heavy nuclei can be described in this model.@xcite despite of the increased number of nuclear transitions to be included in the model , this facilitates the predictions somewhat , as averaged quantities can be used .
calculations of reaction rates from smooth hauser - feshbach cross sections are also somewhat more `` forgiving '' to fluctuations around the `` true '' cross section value because of the integration over the projectile energy distribution ( see eq .
[ eq : stellrate ] ) .
different quantities , though , have to be known as input to the calculations than for direct or resonant reactions , such as optical potentials ( related to the effective interaction in a many - nucleon system ) and level densities ( but not , e.g. , spectroscopic factors of isolated levels).@xcite exceptions are reactions with very low or negative @xmath43 because then @xmath42 is shifted to very low excitation energies at which the ld may be too low to apply the hauser - feshbach model even for heavier nuclei .
typically this occurs close to the driplines . due to the lower ld at magic nucleon numbers ,
the hauser - feshbach model may not be applicable for magic target nuclei and low plasma temperatures .
this applies mainly for neutron captures , however , because the projectile energy range relevant for the calculation of the reaction rate is shifted to higher compound excitation energies when there are charged particles in entrance or exit channel.@xcite the second consequence of a high ld is the availability of a larger number of states at low excitation energy which can be thermally populated in the stellar plasma .
temperatures in explosive burning are higher than in hydrostatic burning and forming nuclei at higher excitation energies . combining the higher ld with higher @xmath10 has the consequence that the calculation of the _ stellar _ rate includes a larger number of reactions on excited states of the target nucleus than would be the case for lighter nuclei , for which excited state contributions are often negligible .
in fact , most reactions in trans - iron nucleosynthesis processes require to include more nuclear transitions in the calculation of reaction rates than just the ones originating from the g.s . of the target nucleus .
the contribution of the rate @xmath44 obtained for a state @xmath45 with spin @xmath46 and excitation energy @xmath47 to the stellar rate @xmath4 can be quantified as @xcite @xmath48 the statistical weights @xmath49 are given by @xcite @xmath50 the reaction cross section @xmath51 of state @xmath45 is defined as @xmath52 and the effective cross section as @xmath53 where @xmath54 is the partial cross section from target level @xmath45 to final level @xmath55 ( cross sections at zero or negative energies are zero).@xcite the modified weights @xmath56 are specified in eq .
( [ eq : new ] ) . for the g.s . , eq .
( [ eq : xfactor ] ) simplifies to @xcite @xmath57 it is very important to note that this is different from the simple ratio @xmath58 of g.s . and stellar rates , respectively.@xcite ( the ratio @xmath59 is called the _ stellar enhancement factor _ and in the past was mistakenly assumed to quantify the excited state contributions . ) using the above equation , the total excited state contribution @xmath60 can be defined , using a sum over discrete excited states and an integration over a ld @xmath61 above the last included excited state at energy @xmath62 , with generalized @xmath63 depending on excitation energy @xmath64 , spin @xmath65 and parity @xmath66 , analogous to eq .
( [ eq : xfactor ] ) . in light nuclei , where the energy of the first excited state is larger than @xmath67 , @xmath68 and thus @xmath69 .
then @xmath70 .
on the other hand , @xmath71 and @xmath72 for heavy nuclei already at low plasma temperatures . a straightforward measurement of a reaction with the target nucleus being in the g.s .
will therefore only constrain a small fraction of the stellar rate .
even when experimentally determining several transitions , one has to be aware that many more transitions contribute to rates involving heavier nuclei than for light nuclides .
this is especially pronounced for charged particle reactions but it was shown that already neutron capture rates on nuclei in the rare - earth region exhibit sizeable excited state contributions even at @xmath73 kev.@xcite finally , a high ld alters the determination of resonance widths for resonant reactions and averaged widths in the hauser - feshbach model . in light nuclei , only few nuclear transitions @xmath74 between a small number of states determine the reaction cross sections . even when the cross sections can not be measured directly , often indirect methods allow to determine the few dominating transitions and thus to derive the cross sections and rates .
the situation is different for nuclei with high ld , as many more transitions contribute .
thus , resonance widths comprising all these transitions have to be measured or calculated .
the utility of indirect methods or of studying reverse reactions is much smaller .
for example , ( d , p ) and ( d , n ) reactions frequently used to study neutron and proton states , respectively , and their spectroscopic factors in light nuclei are of limited use here , as the number of levels at @xmath42 is too large when the hauser - feshbach model is applicable .
also , the ( d , p ) and ( d , n ) cross sections themselves do not provide further information on the astrophysical rates unless they are applied to systems with low ld , for which the direct reaction mechanism dominates .
they can be used , however , to identify low - lying isolated states which are of significance in the calculation of particle widths .
similarly , coulomb dissociation is a helpful tool in studying light nuclei and can also be applied at radioactive ion beam facilities@xcite but for heavier nuclei only a small subset of the relevant transitions can be accessed by this method .
the utility of indirect methods to study rates involving heavier nuclei is further decreased by the larger excited state contributions @xmath75 to the stellar rate , implying even more contributing transitions , as discussed above . on the other hand ,
theoretical predictions are simpler for averaged quantities entering the prediction of widths , such as the compound nucleus ld in the calculation of @xmath2 widths , and do not require _ ab initio _ models but only effective theories .
since systems with high ld exhibit different dependences of the cross sections on nuclear properties , an important quantity to consider is the sensitivity @xmath0 of cross sections or rates @xmath76 to a change in an input quantity @xmath77 , @xcite @xmath78 it is a measure of a change by a factor of @xmath79 in @xmath76 as the result of a change in the quantity @xmath77 by the factor @xmath80 , with @xmath81 when no change occurs and @xmath82 when the final result changes by the same factor as used in the variation of @xmath77 , i.e. , @xmath82 implies @xmath83 .
this is equivalent to writing @xmath84 with @xmath85 and @xmath86 , as used in standard sensitivity analysis . comparing the sensitivities of laboratory cross sections ( for nuclei in their g.s . ) and stellar reaction rates allows an improved judgement of the utility of a given experiment to constrain the rate or some relevant input.@xcite since @xmath0 is strongly energy dependent , a measurement outside the energy range mainly contributing to the rate may not provide the desired , astrophysically relevant information .
different sensitivities of laboratory cross sections and rates also reflect the contribution of reactions on excited states to the stellar rate .
such reactions on excited states proceed at smaller relative energy and thus have a different sensitivity than those on the g.s . of a nucleus . in this context
it should be noted that most determinations of relevant energies at which cross sections contribute significantly to the reaction rate , the _ gamow window _ , only provide them for the g.s .
cross sections @xmath87 but when @xmath71 the relevant energies are much lower for the excited state contributions dominating the rate .
the sensitivity @xmath0 is also central to the assessment of the remaining uncertainties in a rate after the inclusion of experimental data .
this will be further discussed in section [ sec : sproc ] below .
in order to determine the actual rate uncertainties it is important to not only know the experimental error ( or uncertainty factor ) but also to have an initial estimate of the theoretical uncertainty involved . would we know the uncertainty in the prediction of , say , the averaged widths , the sensitivities defined above could be directly used to determine the resulting uncertainty in the stellar rate ( or reaction cross section ) .
unfortunately , this poses a fundamental problem because theoretical uncertainties are fundamentally different from experimental errors , at least when talking about statistical errors in data .
several factors contribute to the theoretical uncertainty .
first , an appropriate model has to be chosen , for example , by choosing which reaction mechanism to treat and in which way .
obviously , it is _ a priori _ impossible to exactly determine the correctness of a model without being able to falsify it through comparison with data .
this impossibility poses the fundamental problem of quantifying uncertainties of the theory approach.@xcite it is loosely related to systematic errors in experiments due to instruments and external effects , which are also hard to estimate . assuming that a `` correct '' model is used
, the next step in the uncertainty analysis would be to check the reliability of the input required to perform calculations with the model
. some of the input may be data bearing some measurement error .
it is possible to propagate this error into the final model output , either analytically or , more likely , by variation studies , for example by a monte carlo variation of such input values within their errors .
part of the input , however , may again be in the form of a model ( e.g. , the ld or an energy - dependent @xmath2-strength function ) . in this case
the initial problem of determining the correctness of a model returns .
even when several choices for model descriptions of a quantity are available , the uncertainty can not be determined by switching from one description to the other and observing the change in the result because this can not be considered drawing a random sample from a statistically distributed set .
the different treatments may even yield similar results at one energy but vastly differ at another . without further knowledge of the underlying model assumptions it is even impossible to sensibly compare different approaches .
therefore it is fundamentally impossible to specify an error bar for a model , especially not a statistically distributed one .
theory uncertainties have to be estimated in comparison to available data but can not be as strictly defined as statistical errors in experimental data .
it is in this sense that the term `` theory uncertainty '' is used below .
monte carlo variation of input can therefore only determine part of the full uncertainty , specifically the uncertainty stemming from input data ( including natural constants ) .
in addition to uncertainty analysis , the sensitivity can also be used to directly infer the impact of a ( experimentally or theoretically ) newly determined quantity @xmath77 ( for example , an averaged width in a hauser - feshbach calculation ) on the cross section or reaction rate . with @xmath88 being the previous value of the cross section or rate of interest , the new value @xmath89 is simply given by @xmath90 when @xmath91 is the factor by which the newly determined @xmath77 differs from its previous value used to calculate @xmath88 .
in the following i want to discuss relevant nuclear and astrophysical uncertainties in selected nucleosynthesis scenarios .
the list of topics and challenges is not exhaustive but rather is supposed to provide examples which enable the reader to generalize and apply the considerations when dealing with further details and other nucleosynthesis processes .
the astrophysical slow neutron - capture process ( @xmath0 process ) produces about half of the elemental abundances between fe and bi.@xcite the @xmath0 process is attributed to environments of neutron number densities @xmath92 of typically @xmath93 @xmath94.@xcite when an unstable nucleus is produced by neutron capture , @xmath95 decays are usually faster than subsequent neutron captures , so the reaction path follows the valley of stability .
the @xmath0 process takes place in different stellar sites .
in particular , the @xmath0-process abundances in the solar system are made by contributions from different generations of stars , providing three different @xmath0-process components , a main , a weak and a strong component.@xcite the main component dominates the @xmath0 contributions to the region between zr and pb .
it is mainly associated with thermally pulsing asymptotic giant branch ( agb ) stars of @xmath96 @xmath97 with an initial composition close to solar.@xcite during the agb phase , he burning takes place in a shell surrounding the c / o core of the star .
thermal pulses are caused by he - shell flashes which occur because he burning can not sustain hydrostatic equilibrium within a thin shell . as a consequence of the mixing processes during the thermonuclear flashes , protons can be mixed downwards from the h - shell into the he - shell where they are used to produce @xmath98c from @xmath99c via @xmath100n in the interpulse phase .
the he - shell contracts and heats up before igniting another he - shell flash .
this allows neutrons to be released through @xmath98c(@xmath101,n ) before and in such a shell flash.@xcite during the temperature peak neutrons are further released in @xmath102ne(@xmath101,n ) reactions .
the strong component also originates in agb stars but old ones , with metallicities much lower than solar.@xcite it is responsible for about half of the solar @xmath103pb abundance and for the @xmath0-process abundance of bi .
the astrophysical challenge in the investigation of the @xmath0 process in agb stars is in the detailed modelling of the downmixing of protons from the h - shell into the he - shell and the formation of a @xmath98c `` pocket '' .
the convection and mixing can only be understood as a 3d effect but a fully 3d stellar model including nucleosynthesis is computationally not yet feasible .
therefore all agb @xmath0-process nucleosynthesis models include _ ad hoc _ assumptions and phenomenological parameters describing the @xmath98c pocket and it is not clear how these evolve for stars with different masses and compositions .
the weak @xmath0 process takes place in massive stars ( with more than 8 @xmath97 ) which later explode as a core - collapse supernova , and is producing most of the @xmath0 abundances in the mass region between fe and zr.@xcite in these stars , neutrons are mostly produced at the end of core he - burning and during the later convective c - shell burning phase via @xmath102ne(@xmath101,n ) .
typical neutron energies are @xmath104 and 0.03 mev for interpulse and flash burning and around 0.09 mev in massive stars .
( note that the relevant plasma temperature in gk ( @xmath105 ) is then given by @xmath106 when @xmath67 is in mev . ) on the nuclear physics side , the challenges lie in the accurate determination of the neutron producing and consuming rates . in the last decades
, considerable effort has been put into the measurement of precise and accurate neutron capture cross sections at @xmath0-process energies.@xcite it has been understood that the uncertainties in the present neutron capture rates are dominated by the experimental errors which are currently below 5% for most @xmath0-process captures.@xcite recently it has been shown , however , that even at the comparatively low @xmath0-process temperatures excited state contributions @xmath75 are not negligible , especially for deformed nuclei.@xcite this leads to larger remaining uncertainties because the laboratory cross section only constrains a fraction of the stellar rate .
the situation is even worse when the experimental `` rate '' is inferred not from neutron capture but its reverse reaction , ( @xmath2,n ) .
ground - state contributions @xmath107 are @xmath108 orders of magnitude smaller in ( @xmath2,n ) rates than in ( n,@xmath2 ) ones at the same temperature .
this will be further discussed in section [ sec : gamma ] . here , we take a look at an important example in the @xmath0 process , @xmath109w(n,@xmath2 ) .
the nucleus @xmath109w is unstable and the location of a branching in the @xmath0-process path where neutron capture and @xmath95 decay are competing .
the current version of the neutron capture compilation for the @xmath0 process ( kadonis , v0.3 ) recommends an experimental value with a 9% error at @xmath110 kev.@xcite even when assuming that the uncertainties in the predictions of the transitions originating on the g.s . and
the excited states are uncorrelated , this would suggest a strong experimental constraint on the rate because for @xmath109w(n,@xmath2 ) , @xmath111 at @xmath112 kev and @xmath113 at @xmath110 kev .
the new uncertainty factor @xmath114 of the stellar rate is constructed from the original theory uncertainty factor @xmath115 and the experimental uncertainty factor @xmath116 by using @xcite @xmath117 the 9% experimental error would translate into @xmath118 . due to the large @xmath107 ,
even a sizeable theory uncertainty of a factor of 2 or more would be almost completely removed .
a closer inspection of the source of the quoted rate , however , reveals that the recommended value is obtained by averaging the rates given by two photodisintegration experiments@xcite , measuring @xmath119w(@xmath2,n)@xmath109w cross sections .
to arrive at a rate , the ( @xmath2,n ) cross section of the g.s . of @xmath119w
was compared to a hauser - feshbach prediction.@xcite it was found that the prediction had to be scaled to reproduce the data .
consequently , the ( n,@xmath2 ) rate predicted by the same model was scaled by the same factor and this value was used as `` experimental '' rate .
the quoted error actually applies to the ( @xmath2,n ) cross section .
therefore it has to be interpreted differently when asking for the remaining uncertainty in the neutron capture rate . of the stellar @xmath109w(n,@xmath2 ) rate to variations in the neutron and @xmath2 widths as function of plasma temperature.[fig :
w185],scaledwidth=110.0% ]
if we could constrain the reverse _ stellar _ rate well , the forward rate could be computed equally reliably because of the reciprocity relation for _ stellar _ rates .
based on the detailed balance theorem for nuclear reactions , it relates the rates of the forward and reverse reactions and reads @xmath120 for a photodisintegration reaction @xmath121 ( see also eqs . [
eq : stellrate ] , [ eq : equilcap]).@xcite equation ( [ eq : revphoto ] ) only applies to _ stellar _ rates in both the forward and reverse channel , i.e. , it can only be used with a rate or cross section of a nucleus in the g.s . as usually encountered in laboratory measurements when @xmath122 ( as defined in eq .
[ eq : exccontrib ] ) , and thus @xmath123 and @xmath124 . for @xmath119w(@xmath2,n ) , however , @xmath125 at @xmath126 and @xmath127 at @xmath128 , for instance.@xcite it immediately follows that the measured contribution to the stellar rate is actually negligible .
does the measurement at least provide information on a transition that is important for the neutron capture cross section ?
the @xmath107 for the photodisintegration are closely related to the relevance of the measured @xmath2 transition to / from the g.s . of @xmath119w . among the @xmath2s emitted in the capture , the ones directly to the g.s .
are only a tiny fraction of all .
the majority are emitted with energies of @xmath129 mev,@xcite about half - way between the g.s . and the neutron separation energy ( see also section [ sec : gamma ] ) .
thus , the ( @xmath2,n ) experiment probes the @xmath2-strength function at an energy which is important neither for the capture nor for the stellar ( @xmath2,n ) rate . and
finally , even if the @xmath2 width was constrained , this would still not fully constrain the ( n,@xmath2 ) rate . as shown in fig .
[ fig : w185 ] , the rate is equally sensitive to uncertainties in the @xmath2- as well as the neutron width at low temperatures . from the above considerations it follows that the @xmath109w(n,@xmath2 ) rate is experimentally still unconstrained at @xmath0-process temperatures and its measurement remains a challenge due to the half - life of @xmath109w .
( the limitations of photodisintegration experiments are more generally discussed in sec .
[ sec : gamma ] . ) from @xmath0-process models it is found that the solar @xmath119os/@xmath130sm abundance ratio , which is determined by the branching at @xmath109w , is best reproduced by a value which is 1.81 times larger than the averaged value of the two photodisintegration experiments.@xcite this is larger by a factor of 1.5 than the previously recommended value which is based on a hauser - feshbach prediction and is a reasonable theory uncertainty for such a neutron capture rate.@xcite the above does not only apply to neutron captures in the @xmath0 process but similar investigations of contributions and sensitivities have to be performed for any reaction in any nucleosynthesis process , preferrably before the measurement is undertaken . in order to explain the solar abundance pattern beyond fe , another neutron capture process had to be postulated , the @xmath1 process.@xcite the abundance peaks at magic neutron numbers at stability are accounted for in the @xmath0 process , due to diminished neutron capture cross sections of neutron - magic nuclei .
the existence of additional peaks shifted to lower mass number was explained by neutron captures far off stability , where neutron - magic nuclei are encountered at lower mass number.@xcite this requires neutron number densities larger than 10@xmath131 @xmath94 and temperatures around 1 gk.@xcite at such conditions , ( n,@xmath2)-(@xmath2,n ) equilibrium is established within each isotopic chain . applying eq .
( [ eq : equilcap ] ) to neutron captures , it can be shown that the abundances within such a chain vary quickly with neutron separation energy @xmath132 and only one or two isotopes receive non - negligible abundances . because @xmath95-decays are slower than the equilibrated reactions , even far from stability , the isotope with the maximal abundance in each chain is called _ waiting point_. the next isotopic chain
is reached only by decay of the waiting point nucleus.@xcite since the waiting points in each chain are disconnected from each other and equilibrium is reached almost instantaneously ( on a timescale on the order of reaction timescales of 10@xmath133 s or less ) , there is no well - defined reaction `` path '' as in the @xmath0 process . an @xmath1-process path is sometimes shown , nevertheless , as an imaginary line connecting the waiting points of all isotopic chains . but also the fact that the @xmath1-process peaks are wider than the @xmath0-process ones shows that even this type of path , i.e. , the waiting points , is not as well defined as the @xmath0-process path .
the waiting points are located far from stability and therefore their half - lives are short , a few 100 ms .
the total processing time to go from @xmath134 to @xmath135 and beyond is on the order of seconds .
this short timescale along with the high neutron densities and high temperature indicates an explosive environment .
so far , the actual site of the @xmath1 process is unknown .
the long time favorite have been the innermost , still ejected layers of a massive star in its core - collapse supernova ( ccsn ) explosion.@xcite these layers cool down from very high temperatures while moving in a neutrino wind emitted from the nascent neutron star .
the neutrino interactions render the matter neutron rich , so that an @xmath1 process can ensue once the matter has cooled sufficiently from nuclear statistical equilibrium to allow neutron captures.@xcite recent multi - d models of core - collapse explosions , however , have been unable to provide the entropies required to explain the full solar @xmath1-process pattern.@xcite additionally , it would be problematic to avoid overproduction at mass numbers below 80 when producing all three @xmath1-process peaks in a single core - collapse event.@xcite a promising alternative site are neutron star mergers which may allow nucleosynthesis in ( n,@xmath2)-(@xmath2,n ) equilibrium in even neutron - richer matter than in core - collapse ejecta and thus require a smaller entropy range to produce solar abundances.@xcite since such mergers are rare events which occur late in the galaxy , they can not explain the @xmath1-process patterns found in very old stars , so - called ultra - metal poor stars , which formed early in the galaxy.@xcite these patterns , however , suggest that there may be two @xmath1-process components , a weak component contributing to nuclides below ba and a stronger , robust component contributing to abundances of ba and above.@xcite independently , two components have been suggested based on meteoritic data.@xcite further alternatives for possible sites are neutron - rich material in magnetically driven jets during ccsn , which circumvent the problems in reproducing solar abundances in neutrino - driven outflows.@xcite a `` cold '' @xmath1-process neutrino - wind scenario was also postulated , in which ( n,@xmath2)-(@xmath2,n ) equilibrium is not achieved.@xcite the astrophysical uncertainties , i.e. , the knowledge of mechanisms and conditions , are considerable , especially in neutrino - wind models .
these are strongly coupled to the lack of our knowledge regarding the ccsn explosion mechanism .
further insights can only be brought by 3d hydrodynamics calculations with realistic neutrino transport to constrain the actual conditions , such as neutrino - wind properties ( determining the evolution of the neutron - richness of the ejected material ) as well as the behavior of the ejecta regarding fallback and wind - termination shock , which can alter the time evolution of the nucleosynthesis conditions considerably.@xcite the stellar reaction rates entering the ( post - processing ) nucleosynthesis calculations also bear considerable uncertainties . due to the involvement of very neutron - rich nuclides far off stability , theory is probed at the edge of our knowledge .
experiments are still unfeasible , except for studies of masses and half - lives somewhat closer to stability , but future radioactive ion beam facilities may allow to gather further information on these nuclei and their reactions .
they will have to overcome severe limitations on what kind of nuclear properties can be studied with respect to astrophysical applications . due to the uncertainties in astrophysical sites and conditions , the range of nuclei to be studied is also large , much larger than , e.g. , in the @xmath0 process with its well - determined path .
contrary to the @xmath0 process , neutron capture rates , however , do not have to be known as long as ( n,@xmath2)-(@xmath2,n ) equilibrium is valid.@xcite the equilibrium abundances are solely determined by @xmath132 which , in turn , is derived from nuclear mass differences .
individual neutron captures become important in the freeze - out from equilibrium ( or in a cold @xmath1 process ) , when they compete with @xmath95-decays at @xmath40 gk .
depending on the freeze - out timescale , this may or may not impact the final abundances.@xcite while the lower @xmath1-process peaks remain almost unaffected by the freeze - out ( because it is fast for its progenitor nuclei ) , the third peak and the rare - earth abundances in between the second and third peak have been found to show some sensitivity to freeze - out.@xcite the relevance of these sensitivities , however , is not yet clear because considerable contributions to the rare - earth region are expected from fission fragments of nuclei produced above bi and these alone may robustly explain the rare - earth peak.@xcite individual neutron capture rates also determine how long ( n,@xmath2)-(@xmath2,n ) equilibrium can be upheld while density and temperature are dropping . predictions and experiments are likewise complicated by the fact that during freeze - out the abundances move from nuclei with low @xmath132 to such with higher @xmath132 and therefore several reaction mechanisms may contribute , depending on the ld ( see sec . [
sec : nld ] ) .
close to the neutron dripline and at magic neutron numbers far off stability , direct capture may be non - negligible although single resonances may still dominate the cross sections , depending on the level structure.@xcite since the nuclei are heavy and their level structure is unknown , it is conceivable that isolated resonances already play a role even for the neutron - richest species appearing during @xmath1-process nucleosynthesis .
resonant capture definitely becomes an issue in later freeze - out phases and in a cold @xmath1 process .
resonant rates are extremely sensitive to the resonance energy , as it only contributes within the relevant energy range of the rate integral ( eq .
[ eq : stellrate ] ) , but it can not be predicted with the necessary accuracy .
closer to stability , the hauser - feshbach model may already yield a good approximation to the rates , even at comparatively low ld , as long as the averaged resonance parameters used in this model are close to the average of the resonances within the astrophysically relevant neutron energy window ( see also sec .
[ sec : nld ] ) .
some attempts have been undertaken to combine predictions of direct capture rates with hauser - feshbach rates but the uncertainties are still considerable.@xcite for experiments , resonant reactions pose a challenge because indirect methods ( such as coulomb dissociation but also photodisintegration , see sec . [
sec : gamma ] ) rely on the fact that only few transitions are important for the calculation of the rate and this is only the case for direct reactions .
another problem are the elevated temperatures encountered in the @xmath1 process which lead to considerable population of excited states with all the consequences of additional transitions discussed before in secs .
[ sec : nld ] , [ sec : sproc ] .
further required information includes rates for @xmath95-decay , @xmath95-delayed neutron emission , @xmath95-delayed and neutron - induced fission , and the fission fragment distributions .
the temperature dependence for these rates ( similarly to what was discussed before for neutron rates ) has also to be taken into account .
this has not even been consistently treated in theoretical rate predictions , yet .
nucleosynthesis of elements heavier than iron is dominated by neutron capture processes . however , there are 35 proton - rich isotopes between sr and hg , which can not be synthesized by the neutron - induced capture reactions of the _ s _ and _ r _ processes.@xcite their abundances are typically factors of @xmath136 smaller than @xmath0 and @xmath1 abundances .
the bulk ( 32 nuclides ) of these so - called _ p _ nuclei are believed to be produced in a series of photodisintegration reactions , referred to as the @xmath2 process , @xcite which is found in the outer layers of massive stars during their explosion in a core - collapse supernova @xcite or in explosions of white dwarfs as type ia supernovae ( snia).@xcite at temperatures between 2 and 3 gk , a heavy seed distribution is first altered by ( @xmath2,n ) reactions and thus proton - rich nuclei are created .
after several neutron emissions , ( @xmath2,p ) and ( @xmath2,@xmath101 ) reactions start to compete with neutron emission , deflecting the reaction flux .
the whole process lasts only a few seconds , before the environment becomes too cool to permit photodisintegration and @xmath95-decay chains finally populate the next stable isobar encountered for a given mass number .
the seed nuclei either are @xmath0- and @xmath1-nuclei initially present ( inherited by the massive star from its proto - stellar cloud or accreted from a binary companion in the case of white dwarfs ) or made in the object during its evolution ( @xmath0 processing in the accreted layer on the white dwarf or during the accretion ) . a longstanding problem is posed by the comparatively large abundances , relative to the other @xmath137 nuclei , of @xmath138mo and @xmath139ru .
they are not produced in massive stars as there is not enough @xmath0- and @xmath1-process material present to act as seed to produce these nuclides in sufficient quantities .
recently , it has been found that they can be produced in thermonuclear explosions of white dwarfs which accreted highly @xmath0-process enriched matter from a companion agb star.@xcite although this @xmath0-process enrichment is crucial for the success of the model , the details are uncertain and current @xmath2-process studies in snia make _ ad hoc _ assumptions because self - consistent simulations of @xmath0 processing in the accreted layer are difficult and unavailable to date .
another open question is in the details of the snia explosion , requiring multi - d hydrodynamical models which also provide the conditions for the nucleosynthesis in the explosion and the white dwarf fragments .
finally , it remains unclear what fraction of snia is actually caused by mass accretion from a companion agb star ( single degenerate scenario ) or by collisions of two white dwarfs ( double degenerate scenario ) .
a recent review discusses in detail the astrophysical and nuclear uncertainties when studying the origin of @xmath137 nuclei.@xcite on the nuclear side , detailed reaction network calculations,@xcite have to consider several hundred reactions for which the reaction rates under the given conditions have to be known .
measurements are challenging or even impossible for several reasons : the cross sections at astrophysically relevant energies are very small , the g.s .
contributions @xmath107 are tiny , and many potentially important reactions involve unstable target nuclei.@xcite therefore experiments are limited to testing model predictions of certain transitions which are also appearing in the calculation of the astrophysical rate . nevertheless , the experimental database of relevant reactions and nuclear properties for the @xmath2 process is far from complete.@xcite the majority of reaction rates has to be predicted by theory making use of the hauser - feshbach approach ( see sec .
[ sec : nld ] ) .
a crucial input parameter of the hauser - feshbach model are optical model potentials , describing the interaction between nuclei and light nuclear particles and required to determine particle transitions .
particularly large discrepancies have been found between predicted and measured @xmath101-induced cross section data in the energy region of relevance.@xcite .
this also has consequences for understanding isotope ratios in meteorites containing matter preserved from the early solar system.@xcite therefore many experimental studies relevant to the production of @xmath137 nuclei have focused on better constraining the @xmath101+nucleus potential .
it is interesting to note that so far it was not possible to find a global optical potential which can consistently describe all the known ( @xmath101,@xmath2 ) and ( @xmath101,n ) data at low energy , especially when including the @xmath140sm(@xmath101,@xmath2)@xmath141gd measurement below 12 mev.@xcite this may require a complicated energy dependence of the optical potential parameters , fine - tuned to each nucleus separately .
recently , a different solution to this @xmath101-potential puzzle was proposed.@xcite it was suggested that an additional reaction channel may contribute to the total reaction cross section , which is not yet considered in the present hauser - feshbach models . instead of modifying the energy - dependence of the optical potential , leading to a different total cross section
, the new idea is that this total cross section is actually described well by the standard potential but that rather it has to be distributed differently between the possible reaction channels at low energy .
such an additional channel acting at subcoulomb energy ( but being negligible at intermediate energies ) is coulomb excitation of target states .
test calculations have shown that this can consistently explain the @xmath140sm(@xmath101,@xmath2 ) measurement and other low - energy data for nuclei with @xmath142.@xcite remaining uncertainties include the @xmath143 values ( especially for odd nuclei ) required for the calculation of the coulomb excitation cross section as well as the knowledge of excited states contributing , especially for unstable nuclei .
the fact that photodisintegrations play a role in certain astrophysical processes , specifically in the @xmath2 process , often gives rise to the misconception that a photodisintegration measurement will directly help with determining the astrophysical reaction rate .
some of the problems when using photodisintegration reactions to determine stellar rates have already been addressed in section [ sec : sproc ] . here
i intend to provide further details .
full reviews of all relevant effects @xcite and photodisintegration in the @xmath2-process @xcite are given elsewhere .
tables of sensitivities of cross sections and stellar rates to various nuclear properties and all g.s .
contributions @xmath107 to the stellar rate are also available.@xcite the main reason for the limited usefulness of photodisintegration experiments is that , even when exciting the nucleus to an astrophysically relevant energy ( and this is important because at higher energies so many additional effects are coming into play that it becomes difficult to impossible to extract the ones relevant in the astrophysical energy range ) , the photon - induced transition from the g.s . with subsequent particle emissions
is only a tiny fraction of all transitions appearing in the astrophysical case ( see also secs . [
sec : nld ] , [ sec : sproc]).@xcite the majority of the relevant @xmath2-transitions have lower @xmath2-energy ( about @xmath129 mev ) and start at an already excited state of the nucleus ( and therefore will probably also select different spin and parity quantum numbers).@xcite the g.s .
contribution @xmath107 to the stellar photodisintegration rate is always very small , often less than a percent.@xcite ( a typical example is given in sec .
[ sec : sproc ] for @xmath119w(@xmath2,n ) but the situation is worse for the @xmath2-process because the higher plasma temperatures lead to even smaller @xmath107 . ) therefore , it is always recommended to measure the capture because its g.s .
contribution is larger and thus the rate can be better constrained by experiment ( although some theory may still be necessary).@xcite ( the situation may be different for nuclei with very low particle separation energy ( or reaction @xmath36 value ) but these nuclei are far from stability and not yet accessible by experiments.@xcite ) three components act together in making photodisintegrations less favorable : @xcite 1 .
the principle of detailed balance in nuclear reactions states that there is a simple relation between forward and reverse reaction between each discrete initial state @xmath45 and final state @xmath55 .
the cross sections @xmath54 are related by a simple phase space factor .
2 . in astrophysics , most nuclei are in thermal equilibrium with the surrounding plasma and therefore are present also as excited nuclei , not just nuclei in their g.s .
the fraction of nuclei in a specific excited state is given by a boltzmann population factor depending on the energy of the excited state @xmath47 and the plasma temperature @xmath10 : @xmath144 .
this leads to the probabilities @xmath145 introduced in equation ( [ eq : statweight ] ) .
the projectile ( or @xmath2 ) energies are also thermally distributed in a plasma .
obviously , the same energy distribution must act on a nucleus in the g.s . and one in an excited state .
this implies that the respective energy distributions ( for example , mbd or planck distribution ) are shifted relative to the integration energy @xmath19 relative to the g.s . and
extending to higher excitation energy @xmath146 when acting on an excited state.@xcite it is important to realize that the above applies to all initial and final nuclei simultaneously in a reaction in an astrophysical environment .
this is also why it is possible to derive the reciprocity relation between forward and reverse _ stellar _ rates , as shown in eq .
( [ eq : revphoto ] ) . already from the notion that a capture reaction ( with positive @xmath36 value ) releases @xmath2s to a wider range of final states than particles ( e.g. , neutrons )
are released in its reverse ( photodisintegration ) reaction it can be seen that a larger number of @xmath2 transitions are relevant than particle transitions ( see also sec . [ sec : sproc]).@xcite from the reciprocity relation in eq .
( [ eq : revphoto ] ) it is easy to show that the number of participating transitions is larger in the exit channel of a reaction with positive @xmath36 value also in stellar rates.@xcite often it is confusing to only consider the boltzmann population factors @xmath147 of the excited states as given above .
they seem to give similar weights to low - lying excited states , regardless of which reaction direction they are applied to .
it is incorrect , however , to take only these factors as a measure of the importance of initial states .
rather , folding these factors with the shifted energy distributions of the projectiles ( or @xmath2s ) and expressing everything in an energy scale relative to the ground state leads to a transformed weight @xcite @xmath148 this can be seen in the full expression for the stellar cross section , @xcite @xmath149 to be used in eq .
( [ eq : stellrate ] ) to calculate the stellar reaction rate .
this means that the partial cross sections for reactions on excited states are evaluated at @xmath150 instead of the usual @xmath19 and their contributions weighted by @xmath56 .
cross sections for individual transitions @xmath151 are zero for negative energies.@xcite equation ( [ eq : stellcs ] ) applies to the forward as well as the reverse reaction .
this has two implications .
firstly , the @xmath56 are linear , more slowly declining than the exponentially declining boltzmann factor . secondly , the weight is relative to the integration energy @xmath19 in eq .
( [ eq : stellrate ] ) and thus energies in the gamow window will contribute most .
the gamow window is at higher excitation energy of the target nucleus in a reaction with negative @xmath36 value.@xcite ( when its energy is @xmath152 in the capture , then it is @xmath153 in the photodisintegration . ) therefore , the weights decline more slowly and reach higher excitation energy in the case of photodisintegration than in the case of capture ( see also fig . 9 of ref .
@xcite ) .
finally , folding the actual astrophysical weights @xmath154 with the relevant @xmath2 transition strengths ( which are - close to stability - roughly about halfway between the ground state and the particle separation energy @xcite ) shows that the ground state contribution to astrophysical photodisintegration rates is small.@xcite from the above , it is found that photodisintegration ( with real photons or by coulomb fragmentation ) can not directly constrain the astrophysical reaction rate for systems with large inherent ld and at high temperature , as found in explosive nucleosynthesis above fe .
capture reactions are much better suited for this as their g.s .
contribution to the stellar rate is much larger than for photodisintegrations . scrutinizing the complete tables@xcite of g.s .
contributions across the nuclear chart , it can be seen that _ stellar _ photodisintegration rates almost always have extremely tiny g.s
. contributions ( with very few exceptions @xcite ) .
therefore a measurement provides a value closer to the rate required in astrophysics .
nevertheless , further corrections from theoretical calculations may still be needed
. can photodisintegration experiments help at all ? from the above , it can be seen that a ( @xmath2,n ) experiment tests the @xmath2-ray strength function ( or @xmath2 width ) at much larger @xmath2-ray energy than the one of the @xmath2-rays actually contributing mostly to the _ stellar _ photodisintegration rate.@xcite in the past , a few ( @xmath2,n ) cross sections were measured@xcite and an astrophysical rate was derived by renormalizing the predicted rate by the same factor as found when comparing the measured result with a prediction in the same model as used to calculate the rate ( as discussed in sec .
[ sec : sproc ] ) .
this implicitly assumes that any discrepancy found between experiment and theory for the much larger @xmath2 energy of the g.s .
photodisintegration equally applies also to the actually relevant transitions with smaller relative energy ( or are due to the particle transitions in the exit channel ) .
usually , this is not a good assumption because the photon strength function behaves differently at low energy and the allowed partial waves in the particle transitions may have different relative angular momentum than the ones appearing in the astrophysical rate .
therefore the experiment does not constrain the astrophysical rate . studying the behavior of the @xmath2-strength function itself
aids the prediction of astrophysical rates , both for captures and photodisintegrations . as mentioned above , the relevant @xmath2 energies are of the order of @xmath129 mev.@xcite changes in the strength function within this energy range directly affect astrophysical capture and photodisintegration rates , changes outside this range are of smaller importance.@xcite unfortunately , this @xmath2-energy range can only be accessed indirectly , requiring a combination of experiment and theory .
for example , ( @xmath2,@xmath155 ) data can help to constrain the low - energy @xmath2 strength or -width .
the ( @xmath2,n ) data can not be used for this by themselves .
recent experiments@xcite realized this and fit experimental ( @xmath2,@xmath155 ) data with theoretical models to use such models for the calculation of the stellar photodisintegration rate .
note that in such experiments also any ( @xmath2,n ) cross section , if measured , does not contribute significantly to the actual constraint of the stellar rate .
because of the relevance of reactions proceeding on thermally excited target states in an astrophysical plasma , it is important to study ( particle ) transitions from these excited states
. this can be done by studying the reverse transitions to such excited states in a particle exit channel .
photodisintegrations can achieve this , as can inelastic particle scattering ( e.g. , ( n , n)).@xcite one has to be careful in the interpretation of such experiments , however , because starting from a specific ground state introduces a selection of possible quantum numbers ( relative angular momenta ) which may be different from the astrophysically relevant ones .
nevertheless , this may allow testing the predictions of particle transitions to / from excited states by theoretical models and also the predicted ratios to the ground state transitions ( e.g. , ( @xmath2,n@xmath156)/(@xmath2,n@xmath157 ) , ( @xmath2,n@xmath158)/(@xmath2,n@xmath157 ) , ; ( @xmath2,@xmath159)/(@xmath2,@xmath160 ) , ( @xmath2,@xmath161)/(@xmath2,@xmath160 ) , ) .
@xcite ( additional information required in predictions of captures and photodisintegrations are low - lying discrete states and nuclear level densities above those discrete states . )
what is essential in all types of experiments is to really measure within the relevant energy region .
the @xmath2 strength function has to be known around @xmath129 mev.@xcite the compound formation energy is given by the gamow window .
@xcite for neutron captures , the upper end of the relevant energy window is at most at 0.2 mev , even at the high plasma temperatures encountered in explosive burning environments .
this translates into an excitation energy of the compound nucleus of @xmath162 mev ( with the neutron separation energy @xmath132 ) .
the energy window of charged particles is shifted to slightly higher energy due to the coulomb barrier but does not exceed several mev.@xcite measuring at much larger energy than the astrophysical one does not yield much relevant information in most cases .
this is because at higher energy the cross sections show a different sensitivity ( see sec .
[ sec : sensi ] ) to photon- and particle - strengths than at astrophysical energies and higher partial waves also play a role.@xcite furthermore , additional reaction mechanisms may occur which are not appearing in the astrophysical energy range.@xcite at high energy , so many additional effects are coming into play that it becomes difficult to impossible to extract and compare the ones relevant in the astrophysical energy range .
much can be learned from photodisintegration experiments ( subsuming experiments with real photons and coulomb dissociation ) . however , it should be clear that there are severe limitations in these methods regarding astrophysical rates for heavier nuclei , despite their success in light nuclear systems . without further information
( such as ( @xmath2,@xmath155 ) data ) , a ( @xmath2,n ) measurement can not be used to constrain the astrophysical rate or to even only test the reliability of model predictions of such rates .
data showing the relative particle emission to g.s . and excited states , though , may test to a certain extent the prediction of thermal modification of the stellar capture reaction and thus of @xmath75 .
also the energy - dependence of optical potentials can be studied at low energy because of the decreasing relative transition energies for reactions leading to excited states in the final nucleus .
the situation is similar in photodisintegrations emitting charged particles .
if measurements could study , however , charged particle emission below the coulomb barrier , this would constrain the rate because at low relative energy the cross section is determined by the charged particle width ( as can be verified by looking at the sensitivities @xmath0).@xcite unfortunately , such cross sections are very small and outside the reach of current methods , especially for unstable nuclei .
the above applies also to indirect studies of reactions in other nucleosynthesis processes , such as described in secs .
[ sec : sproc]@xmath163[sec : nup ] .
the fact that the proton dripline is closer to the line of stability than the neutron dripline may not be taken as an indication that fewer problems are encountered on the proton - rich side .
in general , the effective energies contributing to the reaction rate integral in eq .
( [ eq : stellrate ] ) are higher for charged particle reactions than when dealing with neutron captures on neutron - rich nuclei.@xcite moreover , temperatures in nucleosynthesis processes on the proton - rich side are higher than on the neutron rich side , typically @xmath164 gk , because otherwise the rates would be too slow for short , explosive timescales .
this somewhat alleviates the problem of the reaction mechanisms as discussed in sec .
[ sec : nld ] .
the hauser - feshbach model seems valid up to close to the proton dripline , even for proton captures.@xcite as pointed out before , this is a challenge for experimental studies using indirect methods ( except at the proton dripline ) but eases a theoretical treatment .
nucleosynthesis beyond fe on the proton - rich side not only requires higher temperatures but also a high proton number density .
an important process at such conditions is the rapid proton capture process or @xmath165 process .
a certain type of regular x - ray bursts ( type i ) is explained by an @xmath165 process in proton - rich matter accreted on the surface of a neutron star in a binary system.@xcite nuclear burning on the surface of neutron stars is discussed elsewhere in this volume.@xcite therefore i focus on another proton - rich process in the following , the neutrino - induced ( or neutrino - aided ) rapid proton capture or @xmath3 process .
hydrodynamical simulations indicate that part of the innermost , still ejected layers of ccsn may become proton - rich due to the interaction with the intense @xmath166 wind ( and @xmath167 wind ) within which they move outward .
the matter quickly cools from the initially high temperature , assembling nucleons mainly to @xmath168ni and @xmath101-particles in a nuclear statistical equilibrium , leaving a large number of free protons . at sufficiently low temperature ( @xmath169 gk ) ,
rapid proton captures ensue on @xmath168ni .
production of heavier nuclei beyond @xmath170ge would not be possible without a tiny fraction of free neutrons which is created by @xmath167 captures on the free protons .
the supply of free neutrons in the neutrino - wind allows ( n , p ) reactions bypassing slow electron captures and @xmath171 decays .
this provides the possibility that sequences of proton captures and ( n , p ) reactions produce nuclei with larger and larger @xmath172 and @xmath173.@xcite charged particle reactions then freeze out quickly below 1.5 gk , leaving only ( n , p ) and ( n,@xmath2 ) acting at late time which push the matter back to stability . the @xmath3 process : equilibrium abundance in each isotonic chain ( color - shaded ) , dominating reactions ( arrows ) , and nuclear uncertainties ( circles and diamonds )
. stable nuclei are marked as white squares and the @xmath174 line is shown.,scaledwidth=110.0% ] the location of the effective @xmath175p - process path remains remarkably unaffected by variations of the astrophysical parameters , such as entropy , expansion timescale , and details of the reverse shock @xcite . also
systematic variations of reaction rates show only small effects , if any , regarding the path location @xcite .
all these variations , however , determine how far up the path is followed or whether it is terminated already at low @xmath172 .
thus , also the achieved abundances within the path are determined by these conditions .
the reason for this behavior can easily be understood when realizing that ( p,@xmath2)-(@xmath2,p ) equilibrium is upheld until late times . in consequence ,
the abundance maximum in each isotonic chain is given by the equilibrium abundance ( eq . [ eq : equilcap ] ) , color - shaded for each nucleus in fig .
[ fig : equiabuns ] . the corresponding fig .
[ fig : flux ] shows the same rate field but with the lifetimes of the nuclei .
these _ rate field plots _ are extremely helpful to examine a nucleosynthesis process .
the shade of each nucleus in the plot is either its relative abundance within an isotopic chain ( fig .
[ fig : equiabuns ] ) or related to its lifetime with respect to the fastest reaction destroying it .
the arrows give the direction of the fastest _ net rate per target nucleus _
@xmath176 ( see also eq . [
eq : stellrategeneral ] ) .
the net rate is obtained from the difference of a rate per target nucleus and its reverse rate .
these are not the actual rates in a reaction network as these would also depend on the abundance of the target nucleus but they allow to quickly gather which reaction dominates on a nucleus and which direction the synthesis path would take , were the nucleus actually produced .
same as fig .
[ fig : equiabuns ] but the color shade of each nucleus gives its lifetime against the dominating destruction reaction.,scaledwidth=110.0% ] same as fig .
[ fig : flux ] but at very late times ; neutron captures drive the path even further to stability.,scaledwidth=110.0% ] due to the ( p,@xmath2)-(@xmath2,p ) equilibrium each proton - rich nucleus in an isotonic chain is rapidly ( compared to the expansion timescale ) converted into the nucleus favored by the equilibrium conditions .
therefore , the maximum within a chain is characterized by a low ( p,@xmath2 ) reaction @xmath36 value because the relative speed of forward and reverse rate depends exponentially on the @xmath36 value ( eq .
[ eq : revphoto ] ) .
the largest flux into the next isotonic chain occurs at these nuclei , which are also called waiting points . as can be seen in fig .
[ fig : flux ] , these nuclei indeed exhibit longer lifetimes but they may still be overcome if the neutron abundance is sufficiently high .
this implies that a variation of the neutron density or the ( n , p ) rate on these waiting points will mostly affect how fast nuclei with larger @xmath172 can be reached within the timescale given by the expansion .
another nuclear - structure determined feature can be seen immediately in figs .
[ fig : equiabuns ] and [ fig : flux ] : while the waiting points follow the @xmath177 line up to mo , they are extending further and further to neutron - richer isotopes between mo and sn , pushing the path gradually away from the @xmath177 line .
the path is pushed strongly towards stability at the sn isotopes and above , providing a strong barrier for the efficient production of any elements beyond sn .
two effects are acting here : the location of the waiting - points at neutron - richer nuclides with longer lifetimes with respect to both decays and ( n , p ) reactions , and the fact that also the proton captures become slower due to the higher coulomb barriers at larger @xmath172 .
the latter leads to a dominance of ( n,@xmath2 ) reactions over ( p,@xmath2 ) and ( @xmath2,p ) as can be seen in figs .
[ fig : equiabuns ] and [ fig : flux ] at high @xmath172 , and especially in fig .
[ fig : lateflux ] showing the situation at late times .
the neutron captures push the nucleosynthesis strongly towards stability and even beyond and prevent the production of appreciable amounts of matter above the sn region .
the above general conclusions hold for any trajectory yielding an appreciable @xmath175p - process .
moreover , the range of conditions permitting such a process are also set by nuclear properties . without neutrons ,
the waiting points can not not be passed but a large neutron density ( caused , e.g. , by a higher @xmath167 flux ) would make the ( n , p ) reactions faster than ( p,@xmath2 ) on _ any _ proton - rich nuclide above @xmath168ni .
on the other hand , too high a temperature prevents the outbreak from the @xmath168ni seed as ( @xmath2,p ) and ( @xmath2,@xmath101 ) reactions are faster than proton captures and always establish qse with @xmath168ni .
finally , in a freeze - out charged - particle captures are suppressed according to the coulomb barriers and a further increase in @xmath172 is more and more hampered .
the exact values determining the window of favorable conditions can thus be estimated from nuclear properties in the spirit of the classical b@xmath178fh paper.@xcite the astrophysical hydrodynamics just determines how long the ejected matter is subjected to those favorable conditions and thus how far up in @xmath172 and @xmath173 @xmath175p nucleosynthesis can proceed .
the uncertainties in the astrophysical conditions are similar to those for the @xmath1 process in ccsn ( sec .
[ sec : rproc ] ) because they stem from the same problems in the supernova modelling .
the explosion mechanism and the neutrino interactions not only determine how much matter is ejected at high entropy from the innermost layers and the ejection speed but also influence the time evolution of the neutron - to - proton ratio .
the details of the hydrodynamics also determine the wind termination when the supersonic inner layers run into the slower outer layers of the exploding star and abruptly decelerate.@xcite another possible problem concerning the @xmath166 and @xmath167 spectra close above the neutrino spheres was pointed out recently.@xcite the difference in mean field potentials between neutrons and protons was incompletely considered in previous studies . including an improved treatment seems to render the ejected layers less proton - rich .
the actual impact on the @xmath3 process is pending consistent nucleosynthesis studies . even when using the improved mean field effects
, however , the remaining hydrodynamical uncertainties may be too large to arrive at a definitive conclusion regarding the significance of the @xmath3 process . on the nuclear reaction side ,
the results are mostly insensitive to proton captures due to the prevailing ( p,@xmath2)-(@xmath2,p ) equilibrium . only at late freeze - out times
, this equilibrium is left and gives rise to some sensitivity to a variation of rates.@xcite the equilibrium abundances , in turn , are determined by the proton separation energies @xmath179 and thus by nuclear masses . in figs .
[ fig : equiabuns][fig : lateflux ] circles mark nuclei where remaining experimental uncertainties in the @xmath36 values ( mass differences ) affect the @xmath175p - process results .
it should be noted that the waiting points at lowest @xmath179 within an isotonic chain are somewhat different from waiting points in the @xmath165 process .
the proton number density is much lower in the @xmath3 process than in the @xmath165 process and therefore there are no cases of uncertain waiting points which may be overcome by proton captures when the masses are changed within uncertainties .
therefore also @xmath170ge still is a waiting point , although it is not anymore in the @xmath165 process after a recent mass measurement.@xcite changing the masses of the nuclei marked in the figures , nevertheless , affects the abundance pattern along the @xmath3 path which is obtained after decay to stability and may also impact up to which mass significant abundances can be created .
another difference to @xmath165 waiting points is the fact that ( n , p ) reactions are much faster than @xmath95-decays and electron captures and therefore there is no actual `` waiting '' in the reaction flow , contrary to @xmath165- and @xmath1 process waiting points . the flow to heavier nuclei is determined by ( n , p ) reactions on the waiting point nuclei ( figs . [
fig : equiabuns][fig : lateflux ] ) and thus a knowledge of these is essential.@xcite neutron captures on proton - rich nuclei are of some relevance at large @xmath172 and/or at late times ( fig .
[ fig : lateflux ] ) .
because of the large @xmath132 of proton - rich , unstable nuclei , ( n , p ) reaction @xmath36 values are large and thus @xmath42 is large and the hauser - feshbach model will be applicable . as pointed out in sec .
[ sec : nld ] , this is bad news for experiments probing direct transitions or relying on the fact that only few transitions contribute . on the other hand , even though the hauser - feshbach model is well established , its input has to be determined reliably , especially for the cases where it predicts competitions between rates . in figs .
[ fig : equiabuns][fig : lateflux ] diamonds mark targets for which ( p,@xmath2 ) , ( n , p ) , or ( n,@xmath2 ) rates are close and theoretical uncertainties in the predicted rates may be important .
these occur only at large @xmath172 , reached at late time if at all . a special class of reactions
are those which govern the onset of the @xmath3-process at high temperature . since the details depend on competitions between reactions , they are mostly trajectory - independent and it is possible to identify a set of reactions for which it would be desireable to know the rates ( or ratios of rates of possibly competing reaction types ) with good accuracy .
when freezing out from nuclear statistical equilibrium at high temperature , the @xmath3 process is delayed by several issues . at high temperature , ( @xmath2,p )
reactions are fast and the equilibrium abundances are always located around @xmath168ni . since the main abundance is concentrated in @xmath168ni , further processing is halted until the @xmath168ni waiting point can be bridged effectively and the ( p,@xmath2)-(@xmath2,p ) equilibrium abundance maxima in the subsequent isotonic chains are moved to higher @xmath172.@xcite this depends on the competing rates of ( @xmath2,p ) , ( n,@xmath2 ) , and ( n , p ) on @xmath168ni and occurs at @xmath180 gk .
rate field plot showing an example of a possible limiting cycle at zn in the @xmath3 process , the main path is drawn to guide the eye . ]
rate field plot showing an example of a possible limiting cycle at ge in the @xmath3 process , the main path is drawn to guide the eye . ] whether further processing occurs already at this temperature , however , depends on the relative speeds of ( @xmath2,@xmath101 ) , ( p,@xmath101 ) , and ( n,@xmath101 ) reactions on waiting point isotopes of zn and ge compared to the ( n , p ) , ( n,@xmath2 ) , or ( p,@xmath2 ) reactions required to commence with nucleosynthesis to heavier elements .
we find many of these rates to be comparable within the theoretical uncertainties and therefore they have to be considered as important competition points at early times . to illustrate the importance of accurately knowing the relative strengths of these reactions , two examples are presented here .
the reactions @xmath181zn(@xmath2,@xmath101)@xmath182ni , @xmath181zn(p,@xmath101)@xmath183cu , and @xmath184ge(n,@xmath101)@xmath185zn are close to the competing ( n , p ) , ( n,@xmath2 ) , or ( p,@xmath2 ) reactions . the flow beyond ge
would be considerably impeded were these reactions dominating because the reaction cycles shown in figs .
[ fig : breakout1 ] and [ fig : breakout2 ] would form . initially
the first cycle would be due to the @xmath181zn(@xmath2,@xmath101)@xmath182ni reaction being faster than other reactions on @xmath181zn , hindering any production beyond zn . while cooling down , the situation would change to the one shown in fig .
[ fig : breakout1 ] where @xmath181zn(@xmath2,@xmath101 ) has become slow but there still would be a cycle caused by the reaction @xmath181zn(p,@xmath101)@xmath183cu . to know whether this cycle occurs , the relative strengths of @xmath181zn(p,@xmath101 ) , @xmath181zn(n,@xmath2 ) , @xmath181zn(n , p ) , @xmath181zn(n,@xmath101 ) have to be known . in the current rate
set , @xmath181zn(p,@xmath101 ) , @xmath181zn(n,@xmath2 ) , and @xmath181zn(n,@xmath101 ) are quite comparable down to @xmath186gk .
since a ( p,@xmath101 ) reaction is competing with neutron - induced reactions in this possible zn cycle , one might wonder whether an increase in @xmath187 would also help to avoid the cycle .
this can only be achieved by a larger @xmath188 because a simple scaling of the density @xmath26 equally affects the ( p,@xmath101 ) reaction .
there are two problems with this , however .
firstly , both ( n,@xmath2 ) and ( n,@xmath101 ) would be increased and the competiton between these reactions within their uncertainties would not be lifted . secondly , to make the neutron - induced reactions faster than ( p,@xmath101 ) , the @xmath188 has to be increased so much that ( n,@xmath101 ) or ( n , p ) reactions become faster than ( p,@xmath2 ) for all nuclei in the subsequent isotonic chains , thereby dissolving the ( p,@xmath2)-(@xmath2,p ) equilibrium and pushing nucleosynthesis towards stability without the possibility to further move up in @xmath172 .
nevertheless , if @xmath188 were increased moderately , this avoided a dominating ( p,@xmath101 ) reaction but required some fine - tuning . in the second example
, a cycle would form at ge if the reaction @xmath184ge(n,@xmath101)@xmath185zn was dominating , as shown in fig .
[ fig : breakout2 ] . using the current rate set ,
the reaction @xmath184ge(n,@xmath101)@xmath185zn is found to be in strong competition with @xmath184ge(n,@xmath2 ) but a re - evaluation of the reaction rates may change this picture .
this cycle would be robust with respect to changes in @xmath188 .
another cycle , close to @xmath168ni and acting down to @xmath189 gk was previously found:@xcite the reaction @xmath183cu(p,@xmath101)@xmath168ni cycles matter back to @xmath168ni .
this cycle is also affected by possible competitions with other reactions . this nicely illustrates the uncertainties in @xmath3 nucleosynthesis due to reaction rates and underlines the importance of studying competing reactions . to better constrain the efficiency of the @xmath3 process from the nuclear physics side
, it is necessary to establish tight bounds on the relative strengths of the strongly competing ( p,@xmath2 ) , ( p,@xmath101 ) , ( n,@xmath2 ) , ( n , p ) , and ( n,@xmath101 ) reactions on targets from ni to ge and with @xmath190 .
these conclusions concerning the important rate competitions do not strongly depend on the chosen trajectory .
the ratio of proton- to neutron - induced rates is independent of density changes .
modifying the density at a given temperature affects the ratio of particle - induced to photodisintegration reactions and thus leads to altered break - out temperatures from @xmath168ni or from ( p,@xmath2)-(@xmath2,p ) equilibrium . since the temperature dependence of the reactivities , however , is exponential and the density dependence only linear , the break - out temperatures are shifted only slightly .
the same is true when varying @xmath191 alone .
due to the densities encountered in the ejected layers , @xmath192 always stays well below the values of an @xmath165 process , where ( p,@xmath2 ) would dominate .
moreover , a realistic range of entropies in ejected proton - rich layers implies only a narrow range of temperatures at a given density .
the only strong astrophysical dependence is on @xmath188 created by the @xmath167 flux present at a given temperature .
however , this does not change the ratio between ( n,@xmath2 ) , ( n ,
p ) , and ( n,@xmath101 ) reactions , the latter being a hindrance to the flow up to heavier nuclei . furthermore ,
as pointed out above , there is not much room for modification of @xmath188 without destroying the possibility for @xmath3 nucleosynthesis .
a low @xmath188 prevents @xmath3 processing because of slow ( n , p ) and ( n,@xmath2 ) reactions but a much higher @xmath188 also does not help the break - out as discussed above . nevertheless ,
although the rate competitions are robust with respect to changes in the hydrodynamic evolution , the impact of , for example , cycles as discussed above depends on the trajectory .
the time evolution of the conditions within a trajectory determines how long the favorable conditions for a cycle ( if existing ) are upheld .
if the relevant temperature - density-@xmath188 regime for a possible cycle is crossed quickly whereas the later evolution proceeds slower , leaving enough time for further @xmath3 processing , then the relevance of the cycle is diminished .
for example , although the cycle formed by the @xmath183cu(p,@xmath101)@xmath168ni reaction is appearing at a given temperature , it has only moderate impact on the final abundances when chosing a regular freeze - out without significant wind termination shock .
using a trajectory in which the wind termination shock keeps the temperature at slightly above 3 gk for an extended time will yield a large impact , however.@xcite in conclusion , although the astrophysical constraints seem to be similarly weak for the @xmath3 process as for the @xmath1 process , it is better constrained by nuclear physics and exhibits smaller uncertainties therein , at least in the dominating rates .
nevertheless , an experimental verification of the predicted rates will be difficult , not only because of the short - lived , intermediate and heavy nuclei involved but also due to the high plasma temperatures , giving rise to considerable thermal excitation and thus small g.s .
contributions @xmath107 to the stellar rate .
nucleosynthesis beyond fe poses additional challenges not encountered when studying astrophysical processes involving light nuclei .
it requires different approaches , both in theory and experiment .
the main considerations were presented for a selection of nucleosynthesis processes and reactions .
the presentation should not be viewed as confining the discussed problems to the specific processes .
the intention was to generally introduce the concepts and possible pitfalls along with some examples .
similar problems may apply to further astrophysical processes involving nuclei from the fe region upward and/or at high plasma temperatures .
the framework and strategies presented here shall aid the conception of experimental and theoretical approaches to further improve our understanding of the origin of trans - iron nuclei .
this work was supported by the swiss nsf , the european research council , and the thexo collaboration within the 7@xmath193 framework program ensar of the eu .
t. rauscher , astrophys .
j. lett .
* 755 * , l10 ( 2012 ) ( note : although this applies the derived formalism to s - process neutron capture , the same considerations with respect to the g.s .
contribution to the stellar rate and resulting rate uncertainties can be applied to any other reaction , including reactions with charged particles and stellar photodisintegration reactions . )
|
nucleosynthesis beyond fe poses additional challenges not encountered when studying astrophysical processes involving light nuclei .
astrophysical sites and conditions are not well known for some of the processes involved .
on the nuclear physics side , different approaches are required , both in theory and experiment .
the main differences and most important considerations are presented for a selection of nucleosynthesis processes and reactions , specifically the @xmath0- , @xmath1- , @xmath2- , and @xmath3-processes . among the discussed issues
are uncertainties in sites and production conditions , the difference between laboratory and stellar rates , reaction mechanisms , important transitions , thermal population of excited states , and uncertainty estimates for stellar rates .
the utility and limitations of indirect experimental approaches are also addressed .
the presentation should not be viewed as confining the discussed problems to the specific processes .
the intention is to generally introduce the concepts and possible pitfalls along with some examples .
similar problems may apply to further astrophysical processes involving nuclei from the fe region upward and/or at high plasma temperatures .
the framework and strategies presented here are intended to aid the conception of future experimental and theoretical approaches . _ the only way of discovering the limits of the possible + is to venture a little way past them into the impossible .
+ arthur c. clarke _
|
the meridional overturning circulation in the atlantic is a central challenge to our understanding of global climate dynamics . _
gnanadesikan _ [ 1999 ] ( _ g99 _ hereafter ) presented a model for the deep meridional circulation in terms of the pycnocline depth ( pd ) .
this idealised model has been under intense investigation as a possible paradigm for the meridional overturning circulation [ _ gnanadesikan and hallberg _ , 2000 ; _ saenko and weaver _ , 2002 ; _ gnanadesikan et al . _ ,
furthermore it has been used to investigate the qualitative importance of different physical feedbacks on the oceanic circulation [ _ klinger et al . _ , 2003 , _
gnanadesikan et al . _ , 2003 ; _ kamenkovich and sarachik _ ,
a qualitative feature of the deep meridional overturning circulation is the scaling relation between the volume transport and the meridional density difference in the atlantic [ _ bryan _ 1987 ] . picking up bryan s scaling arguments but assuming a constant pd in the atlantic _ rahmstorf _ [ 1996 ] proposed a linear relation which he demonstrated in the oceanic general circulation model ( gcm ) mom-2 . _ park _ [ 1999 ] and
_ scott et al . _ [ 1999 ] derived the same scaling in a stommel - type box model .
gcm simulations of the ocean suggest that this linear relation carries over from the density to the pressure difference [ _ hughes and weaver _ , 1994 ; _ thorpe et al . _ ,
the linear scaling relation between pressure difference and maximum overturning strength has since been demonstrated to be a robust feature in oceanic gcm simulations ( _ griesel _ , pers .
comm . , 2004 ) .
in sec . 4 we present simulations with the coupled climate model climber-3@xmath0 further supporting these findings .
g99 _ model contains four physical processes which influence the pd in the ocean .
the balance of the pressure gradient in the north atlantic and the frictional forces within the boundary currents leads to an equation for the northward volume transport @xmath1 the pressure gradient is parameterised through the density difference in the na @xmath2 , the north - south distance @xmath3 over which the gradient occurs and the pd @xmath4 .
@xmath5 the constant @xmath6 combines @xmath3 with @xmath7 , @xmath8 and @xmath9 ( the meridional derivative of the coriolis parameter @xmath10 , the density and a proportionality constant of order one ) .
@xmath11 is the gravity constant .
the quadratic dependence on @xmath4 occurs due to the vertical integration in order to obtain a volume transport . in
the so the model includes the drake passage effect through a wind - driven upwelling which does not explicitly depend on the pd @xmath12 . @xmath13 and @xmath14 are the wind stress in the so and the circumference around earth at the latitude of drake passage . additionally _
g99 _ includes an eddy induced return flow @xmath15 where @xmath16 is the transport velocity which _ g99 _ parameterised following _ gent and mcwilliams _ [ 1990 ] while we focus here on its dependence on the pd .
the fourth term in the model is associated with low - latitudinal upwelling described by an advection - diffusion balance @xmath17 in the tropics which yields @xmath18 where @xmath19 and @xmath20 are the diapycnal diffusivity and the horizontal area of upwelling , respectively .
all non - negative constants @xmath21 have been introduced for convenience .
note that the underlying assumption of the model is that these four process can be described using the same value @xmath4 for the pd throughout the atlantic .
( [ equ_tn ] ) requires futhermore that the vertical extension of the northward volume flow is also given by @xmath4 . accepting these assumptions ,
the conservation of volume then results in the governing equation of the model @xmath22 it can be shown that for all parameter settings the model has at most one solution with non - negative pd . in sec .
[ sec_solution ] we give this solution analytically in terms of the volume transport @xmath23 as a function of the pressure difference @xmath24 and discuss , in sec . [ sec_scaling ] , its scaling with @xmath24 . in sec .
4 we compare the results with simulations with the coupled climate model climber-3@xmath0 .
in order to obtain an analytic solution of the model we rewrite equ .
( [ equ_tn ] ) to get an expression of the pd as a function of volume transport @xmath23 and pressure difference @xmath24 @xmath25 in the most interesting case of non - zero volume transport , @xmath26 , we can insert the equality ( [ equ_d_tn_dp ] ) into the volume conservation equ .
( [ equ_governing ] ) to get @xmath27 multiplying by @xmath23 yields a quadratic equation in @xmath23 with two solutions of which only one is non - negative @xmath28 note that despite the fact that the governing equ .
( [ equ_governing ] ) is cubic in @xmath4 , the model does have at most one physical solution given by equ .
( [ equ_solution ] ) .
the model does therefore not bear the possibility of multiple stable modes of the deep meridional overturning circulation as suggested by simulations with climate models of different complexity [ _ stommel _ , 1961 ; _ manabe and stouffer _ , 1988 ; _ rahmstorf _ , 1995 , 1996 ; _ ganopolski et.al._ , 2001 ; _ prange et al .
_ , 2003 ] .
this is to be expected given that the model does not include a salt - advection feedback as proposed by _
[ fig_analytic_solution ] shows the solution for different diapycnal mixing coefficients @xmath19 .
the results were obtained using the numerical values given by _
g99_. note that the solution ( [ equ_solution ] ) depends continuously on the diapycnal mixing coefficient @xmath29 .
no change in the quality of the solution ( [ equ_solution ] ) occurs in the absence of the low - latitudinal upwelling , where @xmath30 in contrast to the behaviour for vanishing @xmath19 , the elimination of the so processes changes the quality of the solution as can be seen from equ .
( [ equ_solution ] ) and will be discussed in the next section .
the analytic solution of the conceptual model for the volume transport @xmath23 as a function of the meridional pressure difference @xmath24 for different values of the diapycnal diffusivity @xmath19 .
the dots represent simulations with the coupled model climber-3@xmath0 which contains an oceanic gcm .
the simulations exhibit a linear scaling in contrast to the conceptual model . ]
next , let us discuss the scaling of the volume transport @xmath23 with the meridional pressure difference @xmath24 .
first , consider the situation without the so processes , i.e. @xmath31 . the scaling can be obtained from the general solution in equ .
( [ equ_solution ] ) with @xmath32 .
more illustrative is the derivation from the original equations for the volume transport ( [ equ_tn ] ) and ( [ equ_tu ] ) .
the fact that the northern downwelling has to be balanced by the low - latitude upwelling @xmath33 implies that @xmath34 i.e. the pd decreases with increasing pressure difference in the na . using this expression to replace @xmath4 in the parameterisation of @xmath23 in equ .
( [ equ_tn ] ) yields @xmath35 in connection with equ .
( [ equ_dp ] ) we get the scaling @xmath36 which was derived first by _ bryan _ [ 1987 ] .
+ next let us add the so winds , but neglect the eddy - induced return flow , i.e. @xmath37 . the solution ( [ equ_solution ] ) then becomes @xmath38 , which goes to a constant @xmath39 in the wind - driven limit , i.e. for small vertical diffusivity @xmath40 . as expected no scaling between pressure difference and volume transport
is observed in this case . in order to discuss the scaling behaviour in the presence of both so processes in the model
, we plot the solutions shown in fig .
[ fig_analytic_solution ] in double logarithmic scale in fig .
[ fig_analytic_solution_log ] . for small @xmath24
all solutions which include the so processes have slope one which corresponds to a linear scaling of the volume transport with the pressure difference . for comparison the solution without the so processes from equ .
( [ equ_noso_scaling ] ) has been included as the solid curve in fig .
[ fig_analytic_solution_log ] showing the one - half slope .
this result can be understood from the general solution in equation ( [ equ_solution ] ) which also sets the scale @xmath41 for which the linear relation holds .
for @xmath42 the solution ( [ equ_solution ] ) becomes @xmath43 i.e. @xmath23 is linear in the meridional pressure difference @xmath24 . using the numerical values given by _
, we obtain an estimate for the pressure scale @xmath44 which is consistent with the scaling seen in fig .
[ fig_analytic_solution_log ] .
simulations with the oceanic general circulation model mom-3 show a linear scaling of @xmath23 with @xmath24 for a variety of parameter settings , including the case of zero diapycnal mixing ( _ griesel _ , ( pers .
comm . , 2004 ) .
the pressure scale in these simulations is of the order of @xmath45 which is in good agreement with the above estimate .
the physical meaning of the scaling regime is seen when multiplying @xmath41 with @xmath4 which gives @xmath46 .
which means that the scaling occurs only when the circulation is completely dominated by the so processes , i.e. when the eddy - induced return flow in the southern ocean is much stronger than the downwelling in the na .
this situation is not consistent with the underlying physical assumption of the model of an interhemispheric meridional overturning circulation and it does not describe the observed circulation in the ocean . from equ .
( [ equ_d_tn_dp ] ) and ( [ equ_small_dp ] ) we can see that in the linear scaling regime the pycnocline depth does not vary with the pressure difference , in contrast to the situation without so processes ( equ .
( [ equ_d_noso ] ) ) where @xmath4 decreases with @xmath24 . from equ .
( [ equ_dp ] ) we see that for constant @xmath4 the pressure difference scales in the same way as the density difference @xmath47 making the linear scaling a simple consequence of the initial assumption that @xmath48 ( equ .
[ equ_tn ] ) .
the same curves as in fig .
[ fig_analytic_solution ] in double logarithmic scale .
the dashed line corresponds to the solution without the so processes and shows a one - half scaling @xmath49 .
in contrast to this the solutions which include the so processes show a linear relationship @xmath50 . ]
the linear relationship between the maximum overturning strength and the density difference @xmath2 which was observed by _ rahmstorf _ [ 1996 ] in a oceanic gcm is reflected in the parameterization of the northern downwelling in equ .
( [ equ_tn ] ) . in the conceptual model ,
however , it does not carry over to the pressure difference , as was shown in the previous section . in order to check this scaling we carried out simulations with the coupled climate model climber-3@xmath0 .
the model contains an atmosphere and a sea - ice module as well as the oceanic general circulation model mom-3 .
the effect of baroclinic eddies was included through a parameterization following _ gent and mcwilliams _ [ 1990 ] with a coefficient of @xmath51 . for a full description of the model see _
montoya et al . _
starting from the present day equilibrium simulation with a maximum overturning strength of 12 sv , we apply a negative salinity forcing of different strength to the na convection sites ( between @xmath52n and @xmath53n ) as described in _
levermann et al . _
this leads to a decrease in the meridional pressure difference in the na and therefore a weakening of the meridional overturning .
a positive salinity forcing strengthens the overturning and increases the pressure difference .
[ fig_analytic_solution ] shows the simulations as black dots .
the pressure was taken at a depth of 1500 m corresponding to the center of the overturning cell in the simulations .
the differences were taken between the zonal average between @xmath52n and @xmath53n and the zonal average between @xmath54n and @xmath55n .
this corresponds with the meridional pressure difference in the na that enters equ .
( [ equ_tn ] ) . as seen in fig .
[ fig_analytic_solution ] the maximum meridional overturning in the atlantic scales linearly with the pressure difference in the na in the simulations .
the vertical diffusivity in coupled model was kept constant at @xmath56 .
thus the simulations correspond to the dashed solution curve in fig .
[ fig_analytic_solution ] .
simulations and conceptual model do neither agree quantitatively using the values suggested by _
g99 _ nor is the qualitative behaviour of the two main quantities ( pressure and volume transport ) reproduced in the conceptual model .
these results are supported by recent findings by _ griesel _ ( pers .
comm . , 2004 ) with an oceanic gcm .
their work shows that the linear scaling between pressure and overturning strength is a robust feature .
it is independent of changes to various parameters including the gent and mcwilliams diffusivity coefficients . in order to emphasize that fact that the linear scaling @xmath50 corresponds to constant @xmath4 we plot in fig .
[ fig_pycnocline ] the pd as defined in _
g99 _ for our simulations .
in contrast to _ g99 _ in an ogcm we find in our coupled model no significant variation of the pd for varying pressure difference .
pycnocline depth for the simulations shown in fig .
[ fig_analytic_solution ] as a function of the meridional pressure difference @xmath24 in the na . definition and
displayed depth range were taken as in _ g99_. ]
by giving an analytic expression for the meridional overturning strength @xmath23 as a function of the meridional pressure difference @xmath24 , we discuss the scaling of the two main quantities of the conceptual model introduced by _
g99_. the model exhibits two scaling regimes which both correspond to unphysical situations .
linear scaling occurs in a situation where the eddy - induced return flow is much stronger than the northern downwelling .
this corresponds to a circulation which is localized entirely in the so and in which all downward volume transport is due to the eddy - induced return flow .
this situation is inconsistent with the physical assumption of an interhemispheric overturning cell underlying the model and the isopycnal nature of the return flow .
the second scaling regime corresponds to a purely northern cell where the upwelling takes place entirely in low latitudes , described by an advection - diffusion balance . in this case
the overturning is proportional to the square root of the pressure as reported by _
bryan _ [ 1987 ] .
the scaling was checked using the coupled climate model climber-3@xmath0 with a parameter setup comparable to the conceptual model , i.e. including effects of baroclinic eddies following _ gent and mcwilliams _ [ 1990 ] and a vertical diffusivity of @xmath57 .
the simulations exhibit a linear scaling and therefore support previous studies [ _ hughes and weaver _ , 1994 ; _ rahmstorf _ , 1996 ; _ thorpe et al .
_ , 2001 ] with comprehensive climate models .
the pd does not vary significantly as a function of the pressure difference in our simulations .
we conclude that the conceptual model of the pd can not reproduce the scaling between its central variables , the pressure and the volume transport .
besides possible criticism regarding the specific parameterizations of the four physical processes contained in the model , the assumption of a universal @xmath4 for all these processes seems questionable .
a.l . was funded by the gary comer foundation .
a.g . were funded through the james s. mcdonnell foundation centennial fellowship .
the authors are grateful to two anonoumous referees for enriching comments .
klinger , b.a .
, s. drijhout , j. marotzke , and j.r .
scott , sensitivity of basin - wide meridional overturning to diapycnal diffusion and remote wind forcing in an idealized atlantic - southern ocean geometry .
, 33:249 , 2003 .
montoya , m. , a. griesel , a. levermann , j. mignot , m. hofmann , a. ganopolski , and s. rahmstorf , the earth system model of intermediate complexity climber-3@xmath0 : performance for present day conditions and sensitivity .
, in preparation , 2004 .
thorpe , r.b .
gregory , t.c .
johns , r.a .
wood , and j.f.b .
mitchell , mechanisms determining the atlantic thermohaline circulation response to greenhouse gas forcing in a non - flux - adjusted coupled climate model .
, 14:3102 , 2001 .
|
we present an analysis of the model by _ gnanadesikan _ [ 1999 ] for the pycnocline depth in the ocean .
an analytic solution for the overturning strength as a function of the meridional pressure difference is derived and used to discuss their mutual scaling .
we show that scaling occurs only in two unphysical regimes of the model . in the absence of the southern ocean ( so ) processes , i.e. for a northern overturning cell , the volume transport is proportional to the square root of the pressure difference .
linear scaling is seen when the overturning is restricted entirely to the so , i.e. when no northern downwelling exists . for comparison ,
we present simulations with the coupled climate model climber-3@xmath0 which show linear scaling over a large regime of pressure differences in the north atlantic ( na ) .
we conclude that the pycnocline model is not able to reproduce the linear scaling between its two central variables , pressure and volume transport .
|
the outer components of nuclear forces are dominated by pion - exchanges and involve just a few basic subamplitudes , describing pion interactions with either nucleons or other pions .
the simplest process @xmath0 , corresponding to the emission or absorption of a single pion by a nucleon , is rather well understood and gives rise to the one - pion exchange @xmath1 potential ( @xmath2 ) .
the scattering reaction @xmath3 comes next and determines both the very important two - pion exchange term in the @xmath1 force and the leading three - body interaction , as shown in fig.[f1 ] .
amplitude ( a ) and two - pion exchange two - body ( b ) and three - body ( c ) potentials . ] the theoretical understanding of the @xmath4 amplitude proved to be very challenging and a suitable description was only produced by means of chiral symmetry .
this framework provides a natural explanation for the observed smallness of @xmath4 scattering lengths and plays a fundamental role in nuclear physics .
nowadays , the use of chiral symmetry in low - energy pion - interactions is justified by @xmath5 .
the small masses of the quarks @xmath6 and @xmath7 , treated as perturbations in a chiral symmetric lagrangian , give rise to a well defined chiral perturbation theory ( chpt ) .
hadronic amplitudes are then expanded in terms of a typical scale @xmath8 , set by either pion four - momenta or nucleon three - momenta , such that @xmath9 gev .
this procedure is rigorous and many results have the status of _ theorems_. in general , these theorems are written as power series in the scale @xmath8 and involve both _
leading order terms _ and _ chiral corrections_. the former can usually be derived from tree diagrams , whereas the latter require the inclusion of pion loops and are the main object of chpt . at each order ,
predictions for a given process must be unique and the inclusion of corrections can not change already existing leading terms .
the relationship between chiral expansions of the @xmath4 amplitude and of two - pion exchange @xmath10 nuclear forces is discussed in the sequence . for the @xmath4 amplitude ,
tree diagrams yield @xmath11 terms and corrections up to @xmath12 have already been evaluated , by means of both covariant@xcite(cf ) and heavy baryon@xcite(hbf ) formalisms . in the case of the @xmath1 potential ,
the leading term is @xmath13 and given by the @xmath2 .
the tree - level @xmath4 amplitude yields @xmath14 contributions at @xmath15 and corrections at @xmath12 are available , based on both hbf@xcite and cf@xcite .
tree - level @xmath4 results also determine the leading @xmath16 three - body force and partial corrections at @xmath12 begin to be derived @xcite . as this discussion suggests ,
@xmath12 corrections to both two- and three - nucleon forces require just the @xmath16 @xmath4 amplitude .
the full empirical content of the @xmath4 amplitude can not be predicted by chiral symmetry alone .
experimental information at low energies is usually encoded into the subthreshold coefficients introduced by hhler and collaborators@xcite which can , if needed , be translated into the low - energy contants ( lecs ) of chiral lagrangians
. therefore , in order to construct a @xmath16 @xmath4 amplitude , one uses chiral symmetry supplemented by subthreshold information , as indicated in fig .
the first two diagrams correspond to the nucleon pole , whereas the other ones represent a smooth background .
the third graph reproduces the weinberg - tomozawa contact interaction , the fourth one summarizes lec contributions and the last two describe medium range pion - cloud effects .
amplitude at @xmath16 . ]
with the purpose of discussing the problem of predicted @xmath17 observed chiral hierarchies , in this section we review briefly results obtained by our goup@xcite for the @xmath14-@xmath1 potential at @xmath12 .
this component is determined by the three families of diagrams shown in fig .
family @xmath18 begins at @xmath19 and implements the minimal realization of chiral symmetry@xcite , whereas family @xmath20 depends on @xmath21 correlations and is @xmath12 .
they involve only the constants @xmath22 and @xmath23 and all dependence on the lecs is concentrated in family @xmath24 , which begins at @xmath16 . as far as chiral orders of magnitude are concerned , on finds that the various components of the force begin as follows@xcite : @xmath25 and @xmath26 , where the superscripts @xmath27 and @xmath28 refer to terms proportional to either the identity or @xmath29 in isospin space .
an interesting feature of these results is that the role played by family @xmath20 is completely irrelevant .
on the other hand , family @xmath18 dominates almost completely the components @xmath30 , @xmath31 , @xmath32 and @xmath33 , whereas family @xmath24 does the same for @xmath34 , @xmath35and @xmath36 . and
@xmath37 contributions to spin - spin ( left ) and tensor ( right ) nuisovector components.,title="fig : " ] and @xmath37 contributions to spin - spin ( left ) and tensor ( right ) nuisovector components.,title="fig : " ] the relationship between the @xmath38 $ ] and @xmath39 $ ] contributions to the @xmath36 and @xmath35 profile functions is shown in fig .
[ f4 ] , where it is possible to see that the chiral hierarchy is respected . in fig .
[ f5 ] , the two central components @xmath40 $ ] and @xmath41 $ ] are displayed side by side and two features are to be noted .
the first one concerns the favorable comparison with the empirical argonne@xcite potentials in both cases .
the second one is that @xmath42 in regions of physical interest , defying strongly the predicted chiral hierarchy .
this problem will be further discussed in the sequence .
the profile function @xmath43 together with @xmath30 , are displayed in fig .
drift corrections begin at @xmath12 and , in principle , should be smaller than the spin - orbit terms , which begin at @xmath16 .
however , in this channel , the hierarchy is again not respected .
the leading term in the three - nucleon potential , known as @xmath14-@xmath44 , has long range and corresponds to the process shown in fig.1c , in which a pion is emitted by one of the nucleons , scattered by a second one , and absorbed by the last nucleon .
is this case , the intermediate @xmath4 amplitude , which is @xmath45 for free pions , becomes @xmath19 and the three - body force begins at @xmath16 .
the first modern version of this component of the force was produced by fujita and miyazawa@xcite , its chiral structure has bee much debated since the seventies@xcite and , nowadays , a sort o consensus has been reached about its form@xcite .
the leading @xmath14-@xmath44 has a generic structure given by & & v_l(123 ) = - _ ab a - b ^3 _ 12 _ 23 + d ^3 i_bac _ c^(2 ) i ^(2 ) _ 12 _ 23 + & & ( g_a /2 f_)_a^(1 ) ^(1 ) _ 12 ( g_a /2 f_)_b^(3 ) ^(3 ) _ 23 y(x_12 ) y(x_23 ) , where @xmath46 is the pion mass and @xmath47 , @xmath48 and @xmath7 are strength parameters , determined by either lecs or subhtreshold coefficients .
the evaluation of @xmath12 corrections requires the inclusion of single loop effects and is associated with a large number of diagrams , which are being calculated by epelbaum and collaborators@xcite . in order to produce a feeling for the structure of these corrections , we discuss a particular set of processes belonging to the @xmath14-@xmath44 class , considered recently@xcite .
full results involve expressions which are too long and cumbersome to be displayed here .
however , their main qualitative features can be summarized in the structure @xmath49 $ ] , where @xmath50 is the leading term shown above and the factors within square brackets are chpt corrections .
the function @xmath51 can be obtained directly from @xmath50 , by replacing @xmath52 , where the @xmath53 indicate changes smaller than @xmath54 .
this part of the chpt correction corresponds just to shifts in the parameters of the leading component .
the term @xmath55 , on the other hand , represents effects associated with new mathematical functions involving both non - local operators and complicated propagators containg loop integrals , in place of the yukawa functions .
the strengths of these new functions are determined by a new set of parameters @xmath56 , which are also typically about @xmath54 of the leading ones . in summary
, chpt gives rise both to small changes in already existing coefficients and to the appearance of many new mathematical structures .
the latter are the most interesting ones , since they may be instrumental in explaining effects such as the @xmath57 puzzle .
chiral symmetry has already been applied to about 20 components of nuclear forces , allowing a comprehensive picture to be assessed . according to chpt
, the various effects begin to appear at different orders and the predicted hierarchy is displayed in the table below . & @xmath2 & @xmath37 & @xmath37 + @xmath13 & @xmath58 & & + @xmath19 & @xmath59 & @xmath60 & + @xmath16 & & @xmath61 & @xmath62 + @xmath12 & & @xmath63 & @xmath64 + in ref . , the relative importance of @xmath65 , @xmath66 and @xmath67 terms in each component of the @xmath37-@xmath68 has been studied . in general , convergence at distances of physical interest is satisfactory , except for @xmath34 , where the ratio between @xmath12 and @xmath16 contributions is larger than @xmath69 for distances smaller than @xmath70 fm .
as far as the relative sizes of the various dynamical effects are concerned , one finds strong violations of the predicted hierarchy when one compares @xmath34 with @xmath33 and @xmath43 with @xmath30 , as discussed above .
it is interesting to note that , in both cases , the unexpected enhancements occur in the isoscalar sector .
the numerical explanation for this behavior is that some of the lecs used in the calculation are large and generated dynamically by delta intermediate states .
however , it is also possible that perturbation theory may not apply to isoscalar interactions at intermediate distances .
this aspect of the problem is explored in the next section .
the structure of @xmath34 was scrutinized in ref . and found to be heavily dominated by a term of the form v_c^+(r ) ~- ( 4/f_^2)(c_3 - 2c_1 ) - c_3 ^2/2 _ n_n(r ) , where the @xmath71 are lecs and @xmath72 is the leading contribution from the pion cloud to the nucleon scalar form factor .
this close relationship between @xmath72 and @xmath34 indicates that the study of the former can shed light into the properties of the latter .
the nucleon scalar form factor is defined as n(p ) | _ sb | n(p ) = _ n(t ) ( p ) u(p ) , where @xmath73 is the symmetry breaking lagrangian .
it has already been expanded@xcite up to @xmath12 and receives its leading @xmath19 contribution from a tree diagram associated with the lec @xmath74 .
corrections at @xmath16 and @xmath12 are produced by two triangle diagrams , involving nucleon and delta intermediate states . in configuration space@xcite , the scalar form factor
is denoted by @xmath75 and one writes _
n ( ) = - 4 c_1 ^2 ^3 ( ) + _ n_n(r ) + _ n_(r ) , where @xmath72 and @xmath76 are the finite - range triangle contributions .
the symmetry breaking lagrangian can be expressed in terms of the chiral angle @xmath77 as @xmath78 .
the ratio @xmath79 describes the density of the @xmath80 condensate around the nucleon and is displayed in fig .
one notes that it vanishes at large distances and increases monotonically as one approaches the center .
this means that the function @xmath81 becomes meaningless beyond a critical radius @xmath82 , corresponding to @xmath83 , since the physical interpretation of the quark condensate requires the condition @xmath84 . in ref . , the condensate was assumed to no longer exist in the region @xmath85 and the @xmath4 sigma - term was evaluated using the expression _
n = r^3 f_^2 ^2 + 4_r^drr^2_n ( ) . this procedure yields 43 mev@xmath8649 mev , depending on the value adopted for the @xmath87 coupling constant , in agreement with the empirical value 45@xmath88 mev .
this picture of the nucleon scalar form factor is sound and can be used to gain insight about @xmath34 . inspecting fig .
[ f7 ] ( right ) , one learns that the hierarchy predicted by chpt is subverted for distances smaller than 1.5 fm , since the @xmath12 delta becomes more important than the @xmath16 nucleon .
on the other hand , the good prediction obtained for the nucleon @xmath89-term ( and also for the @xmath90 @xmath89-term@xcite ) indicates that the functions @xmath91 and @xmath92 can be trusted up to the critical radius @xmath93 0.6 fm . just outside this radius ,
the chiral angle is close to @xmath94 , indicating that the pion cloud is non - perturbative in that region .
this picture is supported by fig .
[ f5 ] ( right ) since , at least up to 1 fm , the prediction for @xmath34 agrees well with the argonne phenomenological potentials .
this leads to our main conclusion , namely that the range of validity of calculations based on nucleon and delta intermediate states is wider that that predicted by chpt .
it was a great pleasure participating in the chiral 2007 meeting and i would like to thank the organizers for the very nice conference , for the warm and friendly hospitality , and for supporting my stay in osaka .
n. kaiser , r. brockman and w. weise , nucl .
phys . a * 625 * , 758 ( 1997 ) ; n. kaiser ,
c*64 * , 057001 ( 2001 ) ; phys .
c*65 * , 017001 ( 2001 ) ; e. epelbaum , w. glckle and u - g .
meissner , nucl .
a * 637 * , 107 ( 1998 ) ; _ ibid . _ a * 671 * , 295 ( 2000 ) ; d.r .
entem and r. machleidt , phys .
c * 66 * , 014002 ( 2002 ) .
g. hhler , group i , vol.9 , subvol.b , part 2 of landlt - bornstein numerical data and functional relationships in science and technology , ed .
h.schopper , 1983 ; g.hhler , h.p.jacob and r.strauss , nucl.phys .
b * 39 * , 273 ( 1972 ) .
yang , phys .
c * 10 * , 2067 ( 1974 ) ; s.a .
coon , m.d .
scadron , p.c .
mcnamee , b.r .
barrett , d.w.e .
blatt and b.h.j .
mckellar , nucl .
a * 317 * , 242 ( 1979 ) ; s.a .
coon and w. glckle , phys .
c * 23 * , 1790 ( 1981 ) , h. t. coelho , t. k. das , and m. r. robilotta , phys .
c * 28 * , 1812 ( 1983 ) ; m.r .
robilotta and h.t .
coelho , nucl .
phys . a * 460 * , 645 ( 1986 ) ; s.a .
coon and h.k .
han , few body syst . * 30 * , 131 ( 2001 ) .
|
chiral expansions of the two - pion exchange components of both two- and three - nucleon forces are reviewed and a discussion is made of the predicted pattern of hierarchies .
the strength of the scalar - isoscalar central potential is found to be too large and to defy expectations from the symmetry .
the causes of this effect can be understood by studying the nucleon scalar form factor .
c
|
the recent indication of the possible existence of a new resonance of mass around 750 gev decaying into two photons @xcite , reported by the atlas and cms experiments at the large hadron collider , has been enthusiastically received by the community and triggered a great interest in its possible origin ( see _ e.g. _ @xcite ) .
one of the possibilities is based on a simple observation that a particle decaying into two photons can be also produced in the photon photon fusion . since a photon can couple to the proton elastically , without breaking it apart ,
it is possible to consider the following process ( @xmath1 is the hypothetical 750 gev @xmath2 resonance ) : @xmath3 such a process provides an interesting possibility of the exclusive measurement , where all final - state particles are registered .
this can be achieved with dedicated forward proton detectors , which can register protons scattered at very small angles into the accelerator beam pipe .
the atlas and cms / totem experiments at the lhc are already equipped with such systems : the atlas forward proton ( afp ) detectors @xcite and the cms - totem proton precision spectrometer ( ct - pps ) @xcite .
if the existence of the 750 gev resonance is confirmed , an exclusive measurement would provide additional information about its nature .
in particular , such a measurement would ensure the two - photon production mechanism of the resonance .
this is because the gluon - mediated exclusive processes at high masses are suppressed with respect to the photon - mediated ones @xcite .
the exclusive measurement with two photons in the final state was already discussed in the context of general searches for new physics via anomalous @xmath4 couplings @xcite and also in the context of the hypothetical @xmath5 resonance @xcite . in this letter , for the first time , a possibility of a semi - exclusive measurement of the light - by - light scattering is discussed .
a signal and background simulation is performed and the effects of ( semi-)exclusivity selection are studied .
the analysis is performed for the example case of the 750 gev resonance .
however , the presented method is more general and can be used for studies of @xmath0 processes in a wide range of invariant masses .
the state - of - the - art calculation of the cross sections for the two - photon production can be found in @xcite .
the authors report 15% 20% uncertainties of their results .
the present study does not attempt to approach such a high precision ; the goal is rather to discuss the measurement feasibility with a relatively simple and straightforward model .
nevertheless , the following calculations are expected to provide characteristics of the events that are sufficiently close to the reality .
the model is based on the equivalent photon approximation , in which the cross section for a given two - photon process factorises into the cross section for the photon photon interaction and the photon
photon luminosity : @xmath6 the photon
photon luminosity is a product of the fluxes of both interacting photons : @xmath7 for the following study it is natural to single out two possible types of photon emission .
first , the photon can be emitted coherently by the proton , which then remains intact after the interaction .
second , the photon can be emitted inelastically , _
e.g. _ when it couples directly to a single quark , which leads to the dissociation of the proton . in this work
the elastic photon flux is taken from the approximate parameterisation @xcite : @xmath8 \\
\left [ \log a - \frac{11}{6 } + \frac{3}{a } - \frac{3}{2a^2 } + \frac{1}{3a^2 } \right ] , \end{gathered}\ ] ] where @xmath9 is the proton momentum fraction carried by the photon , @xmath10 and @xmath11 , \end{gathered}\ ] ] with @xmath12 being the proton mass and @xmath13 the centre - of - mass energy squared .
the inclusive photon flux is taken from the mrst2004 qed parton distribution set @xcite .
the inelastic flux is taken as the difference between the inclusive and the elastic flux .
+ + since in the two - photon process the photons must be emitted by both interacting protons , one can distinguish three production modes : * fully elastic process : both photons are emitted leaving the protons intact , * single dissociation : one photon is emitted elastically and the other one inelastically , * double dissociation : both photons are emitted inelastically .
the photon
photon luminosities for different mechanism and experimental signatures are shown in fig .
[ fig : lumi_m ] . for the mass of 750 gev
the total effective luminosity equals 1.33 pb , where the double - dissociation , single - dissociation and elastic contributions are 0.69 , 0.54 and 0.11 pb , respectively .
the single - dissociation contribution takes into account the factor of two because of the symmetry of the initial state ( each of the two protons can dissociate ) .
in addition , the elastically emitted protons may or may not be registered in the forward proton detectors .
this leads to three possible event signatures : * double tag ( exclusive measurement ) : protons are registered in the detectors on both sides , * single tag ( semi - exclusive measurement ) : a proton is registered in the detectors on one of the sides , * no tag : no protons are registered . using an approximate acceptance of the forward proton detectors based solely on the energy of the proton ( neglecting its transverse momentum ) @xcite , it is possible to estimate how a requirement of a given proton - tag signature affects the photon
photon luminosity . here
, it is assumed that the detectors can register protons that have lost between 5% and 10% of their energy .
this range is quite conservative and it has been chosen to illustrate the advantage of the semi - exclusive measurement over the fully exclusive one .
[ fig : lumi_acc_m ] shows the effective luminosities for different measurement modes .
obviously , the double - tag signature can be obtained only for the fully elastic process , while the single - tag signature can be obtained for the elastic and single dissociative ones .
one can see that the conservative acceptance to a large extent compromises the possibilities of a double - tag measurement at mass of 750 gev .
on the other hand , the acceptance for the semi - exclusive measurement extends over the full mass range .
in addition , the fact that the single dissociation production mode contributes to the semi - exclusive measurement greatly enhances the resulting photon photon luminosity .
in the following analysis it is assumed that the 750 gev resonance is produced in two - photon processes , as described in the previous section .
the model contains three free parameters : the cross section for the @xmath0 production , the mass and the width of the resonance .
the mass was chosen as 750 gev and the width as 45 gev , following @xcite .
the value of the cross section does not affect the majority of the presented results and is chosen arbitrary , as discussed later on .
the transverse momenta of the virtual photons are neglected .
it is also assumed that the resonance decays only to photons and that the decay is isotropic . for the fully exclusive measurement
the dominant background was identified as photon pairs produced in hard non - diffractive events overlaid with independent pile - up events of soft single diffraction @xcite .
it is reasonable to expect the same situation also for the semi - exclusive measurement discussed here .
therefore , the background events are simulated with pythia 8.215 @xcite configured for @xmath14 and @xmath15 parton level processes at @xmath16 tev .
the simulation of pile - up is performed for a given value of pile - up multiplicity @xmath17 .
the actual number of soft diffractive protons in a given event is generated from a poissonian distribution , assuming the inelastic @xmath18 cross section of 77 mb and the single diffractive cross section of 5.6 mb .
the transverse momenta of the diffractive protons are neglected , while the distribution of @xmath19 , the relative momentum loss of the proton , is assumed to be of @xmath20 shape between @xmath21 and 1 .
the @xmath22 value corresponds to the threshold for the @xmath23 process . in order to make the analysis realistic , pile - up protons
are added both to the background and to the signal events . in the analysis
the following experimental effects were taken into account : * acceptance of the forward proton detectors : an event is accepted if a proton with energy loss between 5% and 10% and @xmath24 was present ( only detectors on one side of the interaction point are considered ) , * forward detector energy resolution @xcite , * electromagnetic calorimeter energy resolution @xcite , * longitudinal spread of the vertex ( affects the photon pseudo - rapidity reconstruction , since in two - photon events the vertex reconstruction may not be possible ) .
for an exclusive measurement a typical procedure is to use the information provided by the forward proton detectors to calculate the properties of the centrally produced state and to compare them with the direct measurement .
this approach is not possible for the proposed semi - exclusive measurement , since the single - proton measurement is not sufficient to calculate the momentum of the central state .
however , the reversed procedure is possible : from the central state measurement one can estimate the energy of the forward proton , @xmath25 , which can than be compared with the measured value : @xmath26 .
the relative energy of the forward proton can be calculated as : @xmath27 where it is assumed that the photon
photon system has no transverse momentum . .
] it was checked that the signal events exhibit a strong correlation between @xmath28 and @xmath26 , while for the background the @xmath28 vs @xmath26 distribution is approximately flat .
the shape of the correlation suggests a choice of @xmath29 as the optimal discriminating variable . the @xmath30 distribution for the discussed processes , assuming the pile - up multiplicity of @xmath31 , is shown in fig .
[ fig : discrimination ] . for the elastic and single - dissociation signal events the distributions exhibit very strong peaks around zero . on the other hand , flat distributions
are observed for the non - diffractive production , inelastic two - photon production and single - diffractive production with the intact proton on the wrong side .
.background rejection factors for different values of the pile - up multiplicity . [ cols=">,^,^,^,^ " , ] [ tab : rejection ] fig .
[ fig : discrimination ] suggests the signal selection defined by @xmath32 , with @xmath33 .
however , since the present analysis is relatively simple , a more conservative value of @xmath34 is taken , which partially takes into account effects that are not explicitly included in these considerations . with this cut value ,
the signal efficiency is around 25% and its dependence on the pile - up multiplicity is very small . on the other hand
, background rejection depends very strongly on the @xmath17-value , as shown in tab .
[ tab : rejection ] . in order to illustrate a possible result of the semi - exclusive measurement , fig .
[ fig : mass ] shows the photon
photon mass distributions for the signal and background processes for the pile - up multiplicity of @xmath31 .
contrary to all previous results , here the absolute normalisation of the signal processes is needed .
it has been chosen so that the relative normalisation of the background and double - dissociation signal resembles the atlas results on the 750 gev excess event , hence the possible contributions from elastic and single - diffraction processes should be suppressed . ] .
one can observe that the proposed measurement method greatly improves the signal - to - background ratio .
in addition , the relative contributions of different two - photon production mechanisms in the final sample are different than before the selection ( fig . [
fig : lumi_m ] ) .
the inclusive production is dominated by double dissociation . after applying the exclusivity selection the dominant contribution
is due to single dissociation with the proton @xmath24 .
the next contribution is the fully elastic one , then the fully inelastic one and finally the single dissociation with the proton @xmath35 .
after the ( semi-)exclusivity selection . ]
the above results show that a semi - exclusive measurement of the @xmath0 processes , in particular mediated via the hypothetical 750 gev resonance , is a promising possibility .
the photon
photon luminosity is significantly higher in the single - dissociation mode than in the fully elastic one , which will translate into higher statistics of the collected events .
in addition , the acceptance of the forward proton detectors is not as crucial as in the case of the double - tag measurements .
the proposed ( semi-)exclusive selection provides a significant reduction of the background .
the improvement of the signal - to - background ratio strongly depends on the pile - up multiplicity : for @xmath36 it is a factor of 65 , while for @xmath37 it is around 8 . in the presented analysis only the dominant background was considered and the source of the pile - up protons was treated quite roughly . a future , more detailed analysis
should also take into account additional background processes : ( semi-)exclusive photon
photon production via quark and @xmath38 boson loops , which is predicted by the standard model , diffractive photon
photon production @xcite and backgrounds due to the photon misidentification .
taking into account all these processes , and possibly a more sophisticated simulation of the experimental effects , could allow more stringent exclusivity selection and thus a better background rejection and signal - to - background ration optimisation .
a proper treatment of the pile - up effects can have a two - fold consequence .
first , for the proton energy loss values of the order of 10% processes other than single diffraction can contribute to pile - up effects . therefore , the background contribution will be increased .
on the other hand , taking into account the transverse momenta of the protons and the appropriate resolutions of the forward detectors will provide an additional constraint for the two - photon events selection .
atlas collaboration , arxiv:1606.03833 [ hep - ex ] .
cms collaboration [ cms collaboration ] , cms - pas - exo-15 - 004 .
r. franceschini _ et al .
_ , jhep * 1603 * , 144 ( 2016 )
. j. ellis , s. a. r. ellis , j. quevillon , v. sanz and t. you , jhep * 1603 * , 176 ( 2016 ) .
cms - totem collaboration , cern - lhcc-2014 - 021 .
v. a. khoze , a. d. martin and m. g. ryskin , eur .
j. c * 23 * , 311 ( 2002 )
. s. fichet , g. von gersdorff , b. lenzi , c. royon and m. saimpert , jhep * 1502 * ( 2015 ) 165 .
s. fichet , g. von gersdorff and c. royon , phys .
lett . * 116 * , no .
23 , 231801 ( 2016 ) . l. a. harland - lang , v. a. khoze and m. g. ryskin , jhep * 1603 * ( 2016 ) 182 .
m. drees and d. zeppenfeld , phys .
d * 39 * , 2536 ( 1989 ) .
a. d. martin , r. g. roberts , w. j. stirling and r. s. thorne , eur .
j. c * 39 * , 155 ( 2005 ) .
a. buckley , j. ferrando , s. lloyd , k. nordstrm , b. page , m. rfenacht , m. schnherr and g. watt , eur .
j. c * 75 * , 132 ( 2015 ) .
m. trzebiski , proc .
spie int .
eng . * 9290 * , 929026 ( 2014 ) .
t. sjstrand _ et al .
_ , comput .
commun . * 191 * , 159 ( 2015 ) .
atlas collaboration , jinst * 3 * , s08003 ( 2008 ) . c. brenner mariotto and v. p. goncalves , phys .
d * 88 * , no . 7 , 074023 ( 2013 ) .
|
the two - photon production of photon pairs , _
i.e. _ the @xmath0 process , is studied .
different production modes regarding the elastic or inelastic coupling of the intermediate - state photons to the protons are considered .
the semi - exclusive measurement , where one intact proton is registered by a dedicated forward proton detector , is discussed . as an example , the signal and background simulations are performed for the @xmath0 process mediated by the hypothetical 750 gev resonance .
|
studies of @xmath2 and @xmath3 meson decays shed light on many important aspects of particle physics , such as the extraction of the cabibbo - kobayashi - maskawa ( ckm ) matrix elements , cp violation , rare decays , and hints of physics beyond the standard model .
such decays involve not only the electroweak currents , but also strong interaction dynamics in the form of hadronic matrix elements .
calculation of these matrix elements is crucial for high precision tests of the standard model .
a very promising approach to determining hadronic matrix elements from first principles in qcd is provided by numerical simulations of hadrons using lattice gauge theory .
studies of leptonic and semi - leptonic decays using lattice simulations have already been done using various techniques to handle heavy quarks on the lattice , such as the static approach@xcite , nonrelativistic qcd ( nrqcd)@xcite , and a reformulation of conventional wilson and clover lattice fermions@xcite . in each approach , the electroweak currents must be constructed in terms of appropriate operators comprised from the gluon and quark fields of the lattice theory by calculating the renormalization factors which relate the hadronic matrix elements in the lattice - regulated theory to those in the @xmath6 scheme .
these factors are crucial for both decay constant and semi - leptonic form factor calculations . in this paper
, we construct the time component of the heavy - light axial vector current in terms of operators suitable for simulations@xcite in which the heavy quark is treated using lattice nrqcd and the light quark propagates according to the @xmath4-improved clover action@xcite , where @xmath5 is the lattice spacing .
the standard wilson action is used for the gluons .
this construction is carried out to one - loop order in perturbation theory and to order @xmath0 , where @xmath1 is the mass of the heavy quark .
the expansion of the heavy - light axial vector current in terms of appropriate lattice operators is achieved by matching relevant scattering amplitudes in perturbation theory .
consider the heavy - light pseudoscalar meson decay constant @xmath7 defined in some continuum renormalization scheme , such as @xmath6 , by , @xmath8 where @xmath9 is the heavy - light axial vector current and @xmath10 is a pseudoscalar meson state of momentum @xmath11 .
as will be seen , the axial vector current operator can be written as a sum of operators @xmath12 in the lattice - regulated theory .
for instance , one finds that the temporal component , @xmath13 , is a sum of three operators which contribute through @xmath14 , where @xmath15 is the qcd coupling : @xmath16 since the role of the dimensionless coefficients @xmath17 is to compensate for low - energy effects in the current operator from the loss of short - wavelength qcd modes in the lattice theory , one expects that , due to asymptotic freedom , they may be computed to a good approximation in perturbation theory .
they are fixed by requiring that scattering amplitudes involving these current operators agree in continuum and lattice qcd to a given order in @xmath15 , @xmath5 , and @xmath0 . in a perturbative determination , a process involving quarks and gluons as asymptotic states may be used for this matching ; one need not consider hadron scatterings which are much more complicated to evaluate .
note , however , that the @xmath17 coefficients are independent of the process chosen .
a one - loop evaluation of the @xmath17 coefficients then involves the following steps : ( 1 ) select a quark - gluon scattering process induced by the heavy - light current and calculate the one - loop amplitude for this process in continuum qcd ; ( 2 ) expand the amplitude in powers of @xmath0 ; ( 3 ) identify operators in the lattice theory that reproduce the terms in this expansion ; ( 4 ) calculate the one - loop mixing matrix of these operators in the lattice theory ; ( 5 ) adjust the @xmath17 coefficients to produce a linear combination of lattice current operators whose one - loop scattering amplitude agrees with that from the @xmath6 current to a given order in @xmath0 and @xmath5 .
note that the lattice currents are matched directly to the continuum qcd currents ; we do not need to use a two - step matching procedure in which the lattice currents are first matched to those in a continuum heavy - quark effective theory@xcite , which are then matched to the full qcd currents@xcite .
this paper is organized as follows .
first , in sec . [ sec : continuum ] , we discuss various issues relevant to our matching procedure , such as the choice of expansion parameter in the perturbation series , the definition of the heavy quark mass , the regularization of infrared divergences , and the selection of a renormalization scheme .
next , a scattering process appropriate for constructing the heavy - light axial vector current in the lattice nrqcd - clover theory is chosen .
we then discuss the continuum qcd calculations , and the necessary current operators in the lattice regularization scheme are identified .
[ sec : lattice ] describes the mixing matrix calculation in the lattice theory .
the scattering amplitudes in the continuum and lattice schemes are matched in sec .
[ sec : results ] , and our main results , the values of the coefficients @xmath17 for various heavy quark mass values , are presented .
renormalon ambiguities are discussed , and subtleties with infrared divergences for wilson light quarks are pointed out .
our goal here is the expansion of the temporal component of the ( renormalized ) axial vector current in terms of ( bare ) operators suitable for use in lattice simulations of @xmath2 and @xmath3 mesons .
note that because the axial vector current is partially conserved in qcd , its operator does not actually require renormalization .
this expansion is determined by matching suitable matrix elements evaluated both in continuum qcd and in the lattice theory . since this expansion is an operator relation , the expansion coefficients are independent of the external states of the matrix elements chosen . of course
, the normalizations of the external states used in the continuum and lattice matrix elements must match in order to expose the required operator relation .
since the external states in our perturbative calculations are free quark states , this can be done at tree level by matching the normalizations of the continuum and lattice dirac spinors ( to the appropriate order in @xmath5 ) .
the use of _ on - shell _ wave function renormalization for both the heavy and light quark fields is then the simplest way to ensure this matching beyond tree level . in matching the @xmath0 expansions of heavy - light matrix elements in the continuum and lattice theories , it is important to establish the relations between the quark mass parameters ( both heavy and light ) in each .
equality of the lattice and continuum mass parameters can be ensured in perturbation theory using the pole mass definition .
the pole mass is gauge - invariant and , within the context of perturbation theory , is a physical observable .
hence , in our calculations , we use the pole mass for @xmath1 and also for the light quark mass @xmath18 . in using the pole mass definition for our quarks , we compute the continuum and lattice matrix elements using _ on - shell _ quark mass renormalization .
note that since the pole mass is not defined outside of perturbation theory , it is not a useful quantity for lattice simulations ; the bare mass @xmath19 appearing in the lattice action is much more suitable .
thus , once we determine the coefficients @xmath20 in terms of @xmath1 , we re - express our results in terms of @xmath21 using the perturbative relation between @xmath1 and @xmath19 .
we must also know the relationship between the expansion parameters used in the continuum and lattice perturbative expressions in order to carry out the operator matching .
for example , we could use @xmath22 in the continuum calculations and a coupling @xmath23 defined in terms of the momentum - space static quark potential @xmath24 in the lattice computations .
the two couplings are related by @xmath25 .
our one loop calculations do not make use of the @xmath26 correction in the relation between the lattice and continuum expansion parameters ; hence , as long as we use couplings which match at leading order , we can simply use @xmath15 to refer to both the lattice and continuum couplings .
of course , when numerical values for the coefficients @xmath20 are needed , we must specify a renormalization scheme for @xmath15 .
the coefficients @xmath20 depend on the ultraviolet regulator but contain no infrared divergences .
however , our choice of on - shell mass and wave function renormalization will lead to infrared divergences at intermediate stages in our matrix element calculations . in order to demonstrate the cancellation of these divergences in the @xmath17 coefficients ,
the same infrared regulator must be used in the evaluation of both the continuum and lattice matrix elements . since the triple gluon vertex plays no role in any of our feynman diagrams ,
we introduce a gluon mass @xmath27 for this purpose .
errors in our operator matching are measured in terms of powers of the lattice spacing @xmath5 , the qcd coupling @xmath15 , and the inverse heavy quark mass @xmath0 .
however , nrqcd is a non - renormalizable effective field theory in which the lattice spacing can not be taken to zero ; rather , @xmath28 must be imposed .
hence , our error determinations must take into account the fact that @xmath29 is of order unity .
since we wish to construct a current operator which is correct through @xmath30 , we must remove all @xmath4 errors .
we also remove all @xmath31 errors since it is easy and desirable to do so .
the @xmath32 errors can be eliminated at tree - level in perturbation theory .
the one - loop calculations are then used to remove the @xmath33 and @xmath34 errors .
the remaining @xmath35 errors are much smaller than @xmath4 and @xmath33 and are expected to be comparable to @xmath36 .
a scattering process which is suitable for our study of the heavy - light axial vector current is depicted in fig .
[ fig : feynabc ] . in this process , an incoming heavy quark @xmath37 with momentum @xmath11 scatters off a heavy - light current @xmath38 into an outgoing light quark @xmath39 with momentum @xmath40 .
the axial - vector current operator is given by @xmath41 in terms of the light quark field @xmath42 and the heavy quark field @xmath43 .
the first step in our study is the calculation of the amplitude for this continuum - qcd process to one - loop order in perturbation theory . using the on - shell mass and wave function renormalization scheme in feynman gauge and expanding in @xmath0 ( except for the dirac spinor of the heavy quark )
, one finds @xmath44 \quad \ ; + \quad a_2 \,\left[\frac{p_0}{m}\;\bar{u}_q(p^{\prime } ) \hat\gamma_5 u_h(p)\right ] \nonumber \\ & & + \;a_3\,\left [ \frac{p \cdot p^{\prime}}{m^2}\ ; \bar{u}_q(p^{\prime } ) \hat\gamma_5 \hat\gamma_0 u_h(p)\right ] \ ; + \ ; a_4 \,\left[\frac{p_0^{\prime}}{m}\;\bar{u}_q(p^{\prime } ) \hat\gamma_5 u_h(p ) \right ] \nonumber \\ & & + \;a_5\,\left [ \frac{p \cdot p^{\prime}}{m^2 } \frac{p_0}{m } \ ; \bar{u}_q(p^{\prime } ) \hat\gamma_5 u_h(p)\right ] \quad + \quad o(1/m^2),\end{aligned}\ ] ] with @xmath45 , \nonumber \\
a_2 & = & \qquad\ \frac{\alpha_s}{3 \pi } \ 2 , \nonumber \\
a_3 & = & \qquad\ \frac{\alpha_s}{3 \pi } \left [ \;\;6\ , \ln\frac{m}{\lambda } \,-\,\frac{8\pi}{3 } \,\frac{m}{\lambda}\,+ \,\frac{1}{2 } \right],\nonumber\\ a_4 & = & \qquad\ \frac{\alpha_s}{3 \pi } \left[- 2\ , \ln\frac{m}{\lambda } \,+\ , \frac{1}{2 } \right ] , \nonumber \\ a_5 & = & \qquad\ \frac{\alpha_s}{3 \pi } \left [ -4\ , \ln\frac{m}{\lambda } \,+\ , 5 \right ] , \end{aligned}\ ] ] where @xmath46 and @xmath47 are the standard spinors for the heavy and light quarks , respectively , which satisfy the dirac equation . the light quark mass is set equal to zero .
our conventions for the dirac @xmath48-matrices in minkowski space are given in the appendix .
ultraviolet divergences are regulated using dimensional regularization and fully anti - commuting @xmath49 matrices are used .
as previously mentioned , we use a gluon mass @xmath27 in order to regulate infrared divergences .
the three diagrams which contribute to this amplitude at one - loop order are shown in fig .
[ fig : feynabc ] . in lattice nrqcd
, the heavy quark is described in terms of a two - component ( in spin space ) field @xmath50 .
the dirac field @xmath43 is related to @xmath51 ( and the antiquark field @xmath52 by a unitary foldy - wouthuysen transformation@xcite , @xmath53 this transformation decouples the upper and lower components of the dirac field , thereby separating the quark field from the antiquark field . to facilitate the identification of lattice nrqcd operators capable of matching eq .
[ fqcdterms ] , we similarly transform the external state spinor @xmath46 ( with normalization @xmath54 ) into a nonrelativistic pauli spinor : @xmath55 \,u_q(p ) \quad + \quad o(1/m^2),\ ] ] where @xmath56 and @xmath57 is a two - component external state spinor depending only on the spin of the heavy quark . using eq .
[ fwu ] , the relation @xmath58 , and the dirac equation for the light quark @xmath59 , eq .
[ fqcdterms ] may be written @xmath60 where @xmath61 , and @xmath62 the coefficients in eq .
[ fqcdtermsb ] may be written @xmath63 where @xmath64 , \nonumber \\ \tilde{b}_1 & = & \frac{1}{3 \pi } \left [ 3\ , \ln\frac{m}{\lambda } - \frac{19}{4 } \right ] , \nonumber \\
\tilde{b}_2 & = & \frac{1}{3 \pi } \left[12 - \frac{16 \pi}{3 } \frac{m}{\lambda } \right].\end{aligned}\ ] ] having obtained the @xmath0 expansion of the above scattering amplitude in continuum qcd , the next step is to identify operators in the lattice theory which can reproduce the terms in this expansion .
an inspection of eq .
[ fqcdtermsb ] suggests immediately that matrix elements of the following three lattice operators should be considered : @xmath65 where @xmath42 is now the light quark field in the lattice theory , @xmath19 is the bare heavy quark mass , and @xmath66 is related to the heavy quark field @xmath50 in lattice nrqcd by @xmath67 we use the bare quark mass @xmath19 in the above definitions since it is the natural mass to use in lattice simulations and because the pole mass @xmath1 is not well defined beyond perturbation theory .
the covariant finite difference operators @xmath68 and @xmath69 are defined as usual in terms of the link variables @xmath70 which are the parallel transport operators from sites @xmath71 to neighboring sites @xmath72 in the gauge field .
the definitions of these operators are given below , along with other derivative operators which will be needed later : @xmath73 , \\
o(x)\ a\!\overleftarrow{\nabla}_\mu & = & \frac{1}{2u_0}\left [ o(x\!+\!a_\mu ) u^\dagger_\mu(x)-o(x\!-\!a_\mu )
u_\mu(x\!-\!a_\mu ) \right ] , \\
a^2\delta^{(2 ) } o(x ) & = & \sum_{k=1}^{3 } \left ( u_0^{-1}\left [ u_k(x)o(x\!+\!a_k)+u^\dagger_k(x\!-\!a_k )
o(x\!-\!a_k)\right]-2o(x ) \right ) , \\ a^2\nabla^{(2)}o(x ) & = & \sum_{\mu=0}^{3 } \left ( u_0^{-1}\left [ u_\mu(x)o(x\!+\!a_\mu)+u^\dagger_\mu(x\!-\!a_\mu ) o(x\!-\!a_\mu)\right ] -2o(x ) \right),\end{aligned}\ ] ] where @xmath74 is an operator defined at lattice site @xmath71 with appropriate color structure , and @xmath75 is the mean link parameter introduced by the tadpole improvement procedure@xcite . note that these lattice operators are defined in euclidean space ; our euclidean space conventions are outlined in the appendix .
in this section , we describe the one - loop calculation in the lattice theory of the mixing matrix @xmath76 defined by @xmath77 where @xmath78 are the euclidean - space counterparts of the @xmath79 defined in eqs .
[ omega0]-[omega2 ] .
first , the lattice actions used in these calculations are specified .
the necessary feynman diagrams are then presented , and their evaluation in lattice perturbation theory is outlined . for the heavy quark ,
we use the following nrqcd action density@xcite : @xmath80 where @xmath81 the positive integer @xmath82 is introduced to stabilize the highest momentum modes in the heavy quark propagator@xcite ; the condition , @xmath83 , has proven to be a reliable guide .
note that the above nrqcd action does not include the heavy quark mass term .
the qcd coupling @xmath84 is related to @xmath15 in the usual manner , @xmath85 , and @xmath86 are the standard pauli spin matrices . at tree level , @xmath87 ; the one - loop contribution to @xmath88 is an @xmath26 effect in our mixing matrix calculation and hence can be ignored here .
the chromomagnetic field is given by @xmath89 , where the hermitian and traceless field strength tensor @xmath90 is defined at the sites of the lattice in terms of clover - leaf operators : @xmath91 with @xmath92 for @xmath93 .
for the light quarks , we use the clover action@xcite , @xmath94 where @xmath95 , @xmath96 is the bare light quark mass , @xmath97 $ ] , and we set the wilson parameter @xmath98 .
the one - loop correction to the clover coefficient is an @xmath26 effect in our matching calculation and can be neglected .
lattice perturbation theory calculations are much more laborious than those in continuum perturbation theory .
not only do more diagrams contribute to a given process , but the complicated functional forms of the propagators and vertex functions ( presented in the appendix ) necessitate the use of numerical methods in evaluating the integrals over internal loop momenta . for the scattering process considered here , there are five additional one - loop feynman diagrams in the lattice theory : three are vertex corrections , shown in fig . [
fig : feyndef ] , and two are extra external leg corrections , shown in fig . [
fig : feyngh ] .
the feynman rules are determined by expanding the total lattice action in terms of @xmath84 using @xmath99,\ ] ] and @xmath100 , then fourier transforming into momentum space .
@xmath101 is the lattice gluon field defined at the midpoints of the links connecting neighboring sites .
we calculate the amplitudes corresponding to the feynman diagrams of figs .
[ fig : feynabc]-[fig : feyngh ] in two very different ways and verify that the results agree . in the first method , all spin matrix manipulations and derivatives with respect to external momenta are done by hand .
the resulting integrals are also simplified by hand , including the subtraction of terms to remove infrared divergences .
the remaining infrared - finite integrals are then done numerically using monte carlo and adaptive gaussian quadrature techniques . in the second method
, nearly all aspects of the calculation are automated .
first , the propagators and vertex functions are expressed as functions in c++ whose arguments are the appropriate four momenta implemented using a class fourvec .
spin algebra is accomplished using explicit matrix representations ; this is done by defining new class structures in c++ , such as pauli and dirac , and overloading all of the necessary arithmetic operators for ease of use .
derivatives with respect to external momenta are taken using automatic differentiation@xcite .
this is implemented by defining in c++ a class tcomplex which carries out the multivariate taylor series expansions .
a variable of type tcomplex is treated by the end user just as if it were a regular complex scalar variable since all appropriate arithmetic operators and mathematical functions are overloaded .
however , a tcomplex variable is actually an array containing the function value and all of its derivatives up to some order ; an indexing member function is used to return a specific term in the taylor series expansion .
all derivatives are taken automatically using analytical techniques ; the end user never needs to write subroutines for taking these derivatives .
the above tools allow one to easily write c++ functions to compute the integrands for all of our one - loop feynman diagrams .
the integrals corresponding to the graphs in figs .
[ fig : feyndef ] and [ fig : feyngh ] are easily evaluated using monte carlo techniques .
we use the integration routine vegas@xcite . note that , in lattice perturbation theory , each internal loop momentum @xmath102 is restricted to the first brillouin zone @xmath103 .
however , due to infrared divergences , the other integrals can not be directly calculated .
we evaluate these integrals in a sequence of steps .
first , we subtract from the lattice integrand @xmath104 , where @xmath102 is the loop momentum , the analogous integrand from the continuum theory , @xmath105 .
the integral of this difference over the first brillouin zone is infrared finite and can be evaluated using the monte carlo method .
we then multiply the continuum integrand by a factor @xmath106 to render the integral insensitive to the shape of the first brillouin zone .
this factor must also be chosen such that the integral of @xmath107\ i_{\rm con}(k)$ ] is infrared finite .
for example , @xmath108 where @xmath109 is often a good choice .
we then use the monte carlo method to evaluate the integral of @xmath107\ i_{\rm con}(k)$ ] .
the infrared divergence has then been isolated in the integral of @xmath110 . since
@xmath106 is chosen to remove any appreciable sensitivity of the integral to the edge of the brillouin zone , we can change to hyperspherical coordinates and integrate over the interior of an infinitely large sphere to a good approximation .
this simplifies the calculation and allows us to easily identify and analytically manipulate the infrared divergent pieces of the integrand .
using these two methods , we obtain results for the mixing matrix @xmath76 in eq .
[ latterms ] , again using the on - shell renormalization scheme in feynman gauge .
note that the nonrelativistic external - state spinors @xmath111 are identical in the lattice and continuum theories . for the clover action
, the external light - quark spinors @xmath47 differ only at @xmath112 . at one - loop
, the mixing matrix elements may be written @xmath113,\ ] ] where @xmath114 and @xmath115 are the contributions from the light- and heavy - quark external leg corrections ( that is , from wave function renormalization factors ) , and @xmath116 denote the contributions from the vertex corrections .
our use of an on - shell renormalization scheme with lattice operators defined in terms of the bare mass @xmath19 is responsible for the term proportional to @xmath117 , where @xmath118 \ ,
m_0 + o(\alpha_s^2 ) .
\label{massrenorm}\ ] ] note that although the current operators @xmath12 are defined using @xmath19 , the pole mass @xmath1 must appear in @xmath78 .
the factors in eq .
[ zij ] may be further decomposed : @xmath119 where @xmath120 , @xmath121 , @xmath122 , and @xmath123 are infrared finite and independent of the tadpole improvement factor @xmath75 , and @xmath124 and @xmath125 contain the infrared divergences and tadpole improvement contributions , respectively , from the vertex corrections .
contributions to @xmath114 and @xmath117 from the tadpole improvement counterterms are denoted by @xmath126 and @xmath127 , respectively .
to complete the operator matching , we transform eq .
[ fqcdtermsb ] from minkowski to euclidean space , use eqs .
[ latterms ] and [ zij ] , then peel off the dependence on the external states .
this yields the following operator relation : @xmath128\right ) \
j_{a,{\rm lat}}^{(0 ) } \nonumber \\ & & + \left ( 1 + \alpha_s \left [ \tilde{b}_1 - \frac{1}{2}(\tilde{c}_q+\tilde{c}_q ) -\tilde{c}_m
-\tilde{\zeta}_{01 } - \tilde{\zeta}_{11}\right]\right ) \
j_{a,{\rm lat}}^{(1 ) } \nonumber \\ & & + \qquad \;\ ; \alpha_s \left [ \tilde{b}_2 - \tilde{\zeta}_{02 } - \tilde{\zeta}_{12}\;\right ] \
j_{a,{\rm lat}}^{(2 ) } \quad + \quad o(\alpha_s^2,a^2,1/m^2,\alpha_s a / m).\end{aligned}\ ] ] as expected , we find that the infrared divergences from the various terms in the expansion coefficients cancel .
we now have the desired expansion coefficients @xmath17 of eq . [ a0 ] in terms of the pole mass @xmath1 .
as previously mentioned , the bare mass @xmath19 is a much more convenient mass parameter .
our results are easily expressed in terms of @xmath21 using eq .
[ massrenorm ] , which , in this case , simply amounts to replacing @xmath1 by @xmath19 : @xmath129\right )
\ j_{a,{\rm lat}}^{(0 ) } \nonumber \\ & & + \left ( 1 + \alpha_s \left [ b_1 - \frac{1}{2}(c_q+c_q^{\rm ti}+ c_q ) -c_m - c_m^{\rm ti } -\tau_1 - \tau_1^{\rm ti}\right]\right ) \
j_{a,{\rm lat}}^{(1 ) } \nonumber \\ & & + \qquad \;\ ; \alpha_s \left [ b_2 - \tau_2 \;\right ] \
j_{a,{\rm lat}}^{(2 ) } \quad + \quad o(\alpha_s^2,a^2,1/m^2,\alpha_s a / m),\end{aligned}\ ] ] where @xmath130 , @xmath131 , @xmath132 , @xmath133 , @xmath134 , and @xmath135 .
the factors arising from tadpole improvement counterterms are given by @xmath136 for the usual plaquette definition @xmath137 in the wilson gluonic action , @xmath138 .
for massless clover quarks , @xmath139 .
results for @xmath140 , @xmath141 , @xmath142 , @xmath121 and @xmath122 for various values of @xmath21 are listed in table [ tab : one ] .
explicit calculation reveals that @xmath142 behaves as @xmath143 as @xmath19 becomes large , where @xmath144 is found numerically to be @xmath145 .
this factor can be viewed as arising from a discretization correction @xmath146 to @xmath147 , @xmath148 analogous to the @xmath4 correction @xmath149 , where @xmath150 is the pseudoscalar density , to the axial current in light - light quark systems@xcite .
we can then define an _ improved _ current operator @xmath151 where we write @xmath152 this decomposition is not unique , and hence , we leave @xmath153 as a free parameter . different choices of @xmath153 lead to different @xmath26 contributions being included in eq .
[ ajalph2 ] . taking this into consideration , eq . [ ajalph2 ]
can be written @xmath154 so that the @xmath17 coefficients of eq .
[ a0 ] become @xmath155 , @xmath156 , and @xmath157 .
this equation is our final result .
note that at @xmath158 , the @xmath153 dependence in @xmath159 cancels that in @xmath160 .
numerical results for @xmath161 , @xmath162 , and @xmath163 are given in table [ tab : two ] . in ref .
@xcite , @xmath2 meson decay constants were computed in simulations in which the heavy quark propagation was described by an nrqcd action different from eq . [ nrqcdact ] ; in particular , higher order corrections were included .
the values of @xmath161 , @xmath162 , and @xmath163 for the action used in these simulations are given in table [ tab : three ] . in order to use eq .
[ finalresult ] in a simulation , a value for @xmath15 must be specified . to do this
, one must first choose a renormalization scheme for @xmath15 ; in the case of a running coupling , a means of setting the scale must then be devised ; lastly , the value of the coupling at some reference scale must be determined .
a coupling @xmath164 defined in terms of the short - distance static potential with a scale - setting prescription based on the mean value theorem is advocated in ref .
@xcite ; the value of @xmath165 can be obtained from measurements of the average plaquette .
unfortunately , we have not computed the @xmath166 scales for the @xmath161 , @xmath162 , and @xmath163 coefficients .
however , based on findings in refs .
@xcite and @xcite , we expect that @xmath167 .
an alternative choice for the expansion parameter is a non - running boosted coupling @xmath168 ; the value of this coupling is typically comparable to @xmath169 . for the range of @xmath21 values considered here ,
our results show no evidence that perturbation theory is failing . from tables
[ tab : two ] and [ tab : three ] , one sees that the one - loop corrections to @xmath170 are very small .
values for @xmath161 are typically near @xmath171 for all @xmath21 considered ; multiplying by @xmath172 yields one - loop corrections to @xmath170 which are approximately @xmath173 of the tree - level contribution . for the large range of @xmath21 values studied , @xmath162 varies between @xmath174 and @xmath175 ; hence , one - loop corrections to @xmath176 are never larger than about @xmath177 .
the magnitudes of the values for @xmath163 are significantly larger than those for @xmath161 and @xmath162 .
if we set @xmath178 , then @xmath179 and @xmath180 which , using @xmath172 , varies between @xmath181 and @xmath182 for the range of @xmath21 values studied here .
since @xmath183 has no tree - level contribution , at least a two - loop result is needed to check the convergence of its perturbative expansion . as in all applications of perturbative qcd , nonperturbative contributions to @xmath17 are possible .
for example , the perturbative expansions of the @xmath17 coefficients contain renormalon ambiguities arising from a slight mismatch between the infrared physics of the lattice theory and that of continuum qcd .
this mismatch is caused by discretization artifacts and the truncation of the @xmath0 expansion . in our calculation , contributions coming from momenta small relative to a given infrared scale @xmath184 are all suppressed by @xmath185 or @xmath186 since the infrared structure of our lattice theory is designed to agree with continuum qcd through @xmath187 .
this infrared suppression is not true diagram - by - diagram ( some diagrams are even infrared divergent ) , but the infrared contribution is suppressed by @xmath188 or @xmath189 when all diagrams are summed .
for example , in some analyses@xcite , coefficients like @xmath170 are separated into the contribution @xmath190 from the leading operator ( here @xmath191 ) and contributions @xmath192 from mixings with higher - dimension operators ( here @xmath193 and @xmath194 ) . in our analysis
, such a separation is artificial and would be a mistake since it induces @xmath187 ambiguities in the separate pieces @xmath190 and @xmath192 .
these ambiguities cancel when the pieces are recombined , leaving contributions to @xmath170 suppressed by @xmath185 or @xmath186 .
thus , any renormalon ambiguities are suppressed by the same factors and are at worst comparable to the other truncation errors in the analysis .
our calculation differs from conventional lattice calculations in that the @xmath188 truncation errors , perturbative and nonperturbative , can not be made arbitrarily small by reducing the lattice spacing@xcite .
this is because nrqcd has nonrenormalizable interactions whose couplings do not vanish as @xmath195 .
the perturbative expansions for @xmath17 have power - law terms of the form @xmath196 which ruin the convergence of perturbation theory if @xmath5 is taken to zero . in practice
, this problem is avoided by ensuring that @xmath197 is not large . our inability to take @xmath198 in nrqcd
fundamentally limits the precision of our @xmath187 accurate formalism .
if improved precision is needed , we must reduce the truncation errors by using more accurate discretizations of the lattice currents and action . in this way truncation errors , renormalon ambiguities , etc .
are pushed off to @xmath199 or higher .
this approach differs dramatically from that used in the static - quark expansion ; in the latter approach , problems with power divergences arise when the limit @xmath198 is attempted in a brute - force manner .
the role of the power - law terms in the @xmath17 coefficients is to cancel unphysical power - divergent contributions in the lattice nrqcd current matrix elements , thereby ensuring the correct matching to full continuum qcd through the relevant order in @xmath15 and @xmath0 .
if the unphysical terms behaving as @xmath196 in the @xmath17 coefficients and the current matrix elements become comparable to or larger than the physical terms of @xmath200 , this cancellation becomes a very delicate issue and the need for nonperturbative subtraction methods may arise .
such is the case in the static - quark expansion ; taking the brute - force limit @xmath198 obscures the physically relevant contributions by amplifying the power - divergent unphysical terms . here , this problem is circumvented by keeping the lattice spacing large enough that @xmath197 never becomes large .
the success of low - order perturbation theory in cancelling the unphysical terms depends , of course , on the relative size of the physical contributions .
recent numerical simulations@xcite find no evidence of large power - law contamination in the current matrix elements for the range of @xmath21 appropriate for @xmath201-quark physics , suggesting that power - divergent nonperturbative and higher - order perturbative contributions to the matching coefficients are also not large .
this finding applies to the temporal component of the heavy - light axial vector current . for other operators ,
the situation could be different and each case must be examined separately . the results in table [ tab : two ] are suitable for values of @xmath21 appropriate for current simulations of heavy - light systems using nrqcd
however , terms proportional to @xmath202 cause the @xmath158 contributions to @xmath170 and @xmath176 to become large as @xmath21 becomes large . in such cases
, the renormalization group should be used to improve upon the estimates from one - loop perturbation theory . since the left - hand side of eq .
[ a0 ] is independent of the lattice spacing , it follows that , neglecting @xmath112 terms , @xmath203 = 0.\ ] ] using eq .
[ latterms ] and @xmath204 , a renormalization group equation for the @xmath17 coefficients , collected into a vector @xmath205 , can be obtained which describes the change in @xmath17 as the lattice spacing is varied : @xmath206 where the anomalous dimension matrix is given by @xmath207 in the limit of large @xmath21 , we find @xmath208 and eq .
[ rg1 ] can be easily solved .
first , express the @xmath17 coefficients as a function of @xmath209 and @xmath29 instead of @xmath209 and @xmath210 since @xmath1 is fixed and does not run with @xmath5 .
if the lattice theory is renormalized in such a way that the renormalization group @xmath211-function @xmath212 is independent of @xmath19 and @xmath96 , then @xmath213 where @xmath214 and @xmath215 , for @xmath216 light quark flavors . for large , fixed @xmath1 , eq .
[ rg1 ] then tells us that the change in @xmath170 and @xmath176 in going from an initial lattice spacing @xmath217 to a smaller lattice spacing @xmath218 is given , in leading logarithmic approximation , by @xmath219^{1/(2\beta_0\pi ) } , \label{rg3}\ ] ] for @xmath220 , where @xmath221 is the familiar qcd running coupling ; its two - loop form is @xmath222^{-1}.\ ] ] let @xmath223 and @xmath224 , then @xmath225 hence , to the order at which we are working , eq .
[ rg3 ] is equivalent to @xmath226^{-1/(2\beta_0\pi ) } , \label{rg4}\ ] ] for large @xmath21 and @xmath227 .
when @xmath21 is not large , the dependence of @xmath228 on @xmath21 can not be neglected and eq .
[ rg4 ] must be suitably modified .
as an aside , we mention the following fact .
if one wishes to use the wilson action instead of the clover action for the light quark , one can not include the discretization correction term @xmath146 .
if one includes @xmath146 while using wilson light quarks , one finds an uncancelled logarithmic infrared divergence upon attempting to match the continuum and lattice scattering amplitudes .
this divergence is removed by including contributions from the @xmath4-correction term in the clover action .
in this paper , the temporal component of the heavy - light axial vector current @xmath229 was expanded in terms of lattice operators suitable for use in simulations of @xmath2 and @xmath3 mesons .
the expansion was carried out to @xmath30 by matching relevant scattering amplitudes to one - loop order in perturbation theory .
the ( massless ) light quark was described in the lattice theory using the @xmath4-improved clover action of eq .
[ cloveract ] , and the nrqcd action of eq .
[ nrqcdact ] was used to treat the heavy quark .
the standard wilson action was used for the lattice gluons .
the expansion of the heavy - light current was found to be @xmath230 where the lattice operators are defined in eqs .
[ jop0 ] , [ jop1 ] , [ jop2 ] , and [ impop ] , and @xmath155 , @xmath156 , @xmath157 and @xmath231 $ ] .
values for @xmath161 , @xmath162 , and @xmath163 are listed in table [ tab : two ] for various bare heavy quark masses .
@xmath153 remains as a free parameter ; different choices for the value of @xmath153 lead to different @xmath26 contributions being included in eq .
the one - loop corrections for @xmath170 and @xmath176 were shown to be small relative to the tree - level contributions for the range of @xmath21 studied ; no evidence of a breakdown in perturbation theory was found .
since @xmath183 has no tree - level contribution , at least a two - loop calculation would be needed to check the behavior of its perturbative expansion .
renormalon ambiguities were argued to be at worst of the same order as our other systematic errors .
our results have already been used in @xmath2-meson simulations@xcite measuring @xmath232 .
more recently , @xmath2 meson decay constants were computed@xcite in simulations using an nrqcd action with higher - order interactions not included in eq .
[ nrqcdact ] ; the values of @xmath161 , @xmath162 , and @xmath163 for the action used in these simulations are given in table [ tab : three ] . in the future ,
we plan to apply the methods described in this paper to the expansion of other currents , such as the vector current , in terms of appropriate lattice operators .
we would like to thank peter lepage , arifa ali khan , tanmoy bhattacharya , and geoff bodwin for helpful discussions .
we would also like to thank christine davies for checking some of our continuum qcd calculations .
this work was supported by the u.s .
doe , grants no .
de - fg03 - 90er40546 and de - fg02 - 91er40690 , and by nato grant crg 941259 .
the nrqcd and light quark actions are given in eqs .
[ nrqcdact ] and [ cloveract ] in sec .
[ sec : lattice ] .
the feynman rules of perturbation theory can be derived from these actions using , for instance , the methods of ref .
most of the rules relevant for the calculations of this article are collected in this appendix .
various conventions used in our computations are also outlined . to simplify notation , we set the lattice spacing @xmath233 in this appendix .
minkowski - space quantities are indicated either by a caret , such as @xmath234 , or by a subscript or superscript @xmath235 , such as @xmath236 .
the metric tensor in minkowski space is taken to be @xmath237 and the dirac matrices satisfy @xmath238 .
we use the dirac - pauli representation : @xmath239 where @xmath86 are the standard pauli spin matrices .
also , we define @xmath240 euclidean - space four - vectors are defined in terms of minkowski - space four - vectors using @xmath241 and @xmath242 , for @xmath243 . for the derivative operator , @xmath244 and @xmath245 .
note that the gauge field @xmath246 wick - rotates into euclidean space as a covariant vector , as does the gauge - covariant derivative @xmath247 .
the euclidean - space dirac matrices satisfy @xmath248 and are related to their minkowski - space counterparts by @xmath249 , @xmath250 , and @xmath251 .
also , @xmath252 $ ] . to be consistent with our conventions for euclidean - space @xmath253 matrices , we define euclidean - space quark - bilinear axial and vector currents by @xmath254 , @xmath255 , @xmath256 , and @xmath257 .
the chromoelectric and chromomagnetic fields are defined in terms of the field strength tensor by @xmath258 and @xmath259 , where @xmath260 is the fully antisymmetric levi - civita tensor and the field strength tensor is given on the lattice by eq .
[ fieldstrength ] .
the heavy quark propagator is diagonal in both spin and color , and is given in momentum space by @xmath261 the light quark propagator is diagonal in color : @xmath262 in this paper , we work only with the @xmath263 and @xmath264 case .
the gluon propagator is diagonal both in color and the lorentz indices , and is given in feynman gauge by @xmath265 we now list the vertex factors associated with the interaction of a single gluon with a heavy - quark line .
let @xmath266 be the outgoing heavy - quark momentum , @xmath102 be the incoming quark momentum , and @xmath267 be the polarization index of the emitted gluon .
these vertex factor all have a color factor of @xmath268 , where @xmath5 is the color index of the gluon , and @xmath201 and @xmath269 are the color indices of the outgoing and incoming quarks , respectively . for @xmath270 ,
the vertex factor is @xmath271 for @xmath272 , the vertex factor from the @xmath273 term is @xmath274 \ , \left [ e^{-ik^{\prime}_0 } f^n(k^{\prime})+ e^{-ik_0 } f^n(k)\right ]
\;s_{n}(k^{\prime},k ) , & \\ & s_{n}(k^{\prime},k ) = \sum_{l=0}^{n-1 } f^{l}(k^{\prime})\ , f^{n - l-1}(k).&\end{aligned}\ ] ] for @xmath275 from the @xmath276 term , the vertex factor is given by @xmath277 \ ; \left [ e^{-ik^{\prime}_0 } f^{2n}(k^{\prime})+ e^{-ik_0 } f^{2n}(k)\right].\ ] ] the single - gluon vertex factor for the light quark ( clover action ) is @xmath278 + r \sin[{\textstyle\frac{1}{2 } } ( k^{\prime } + k)_\mu ] + \frac{r}{2 } \sum_\nu \sigma_{\nu \mu } \sin(k^{\prime } - k)_\nu \cos[{\textstyle\frac{1}{2 } } ( k^{\prime } - k)_\mu]\right\},\ ] ] for all gluon polarizations @xmath267 , where @xmath279 is the outgoing ( incoming ) momentum of the light quark .
next , the vertex factors associated with the interaction of two gluons with a heavy - quark line are given .
let @xmath266 be the outgoing heavy - quark momentum , @xmath102 be the incoming quark momentum , and @xmath280 , @xmath281 be the outgoing momenta of the emitted gluons having polarization indices @xmath282 , @xmath283 , respectively . only the factors for @xmath284 ( and @xmath285 )
are given , which is all that is required for the one - loop tadpole graphs .
the color factor for these cases is @xmath286 , where @xmath5 , @xmath287 are the color indices of the gluons , and @xmath201 and @xmath269 are the color indices of the outgoing and incoming quarks , respectively . for @xmath288 ,
the vertex factor is @xmath289 for @xmath290 , the vertex factor from the @xmath273 term in the nrqcd action is @xmath291 \;\biggl(\displaystyle{\frac{-1}{4nm}}\ ; \cos[{\textstyle\frac{1}{2 } } ( k^{\prime}\!+\!k)_j ] \ ;
s_{n}(k^{\prime},k ) & \\ & + { \displaystyle \frac{1}{(2nm)^2}}\;\sin[{\textstyle\frac{1}{2 } } ( k^{\prime}\!+\!k\!-\!q_2)_j ] \ ; \sin[{\textstyle\frac{1}{2}}(k^{\prime}\!+\!k\!+\!q_1)_j ] \ ; s_{g , n}(k^{\prime},k^{\prime}+q_1,k)\biggr ) & \\ & + g^2 \ ; { \displaystyle\frac{1}{(2nm)^2 } } \;\sin[{\textstyle\frac{1}{2}}(k^{\prime}\!+\!k\!-\!q_2)_j ] \ ; \sin[{\textstyle\frac{1}{2}}(k^{\prime}\!+\!k\!+\!q_1)_j ] \ ; s_{n}(k^{\prime},k - q_2 ) \ ; s_{n}(k^{\prime}+q_1 , k ) \ ; e^{-i(k^{\prime } + q_1)_0 } , \end{aligned}\ ] ] where @xmath292 the vertex factor for @xmath290 from the @xmath276 term is @xmath293 \ ; \cos[{\textstyle\frac{1}{2}}(k^{\prime}-k+q_1)_j ] & \\ & + \mbox{\rm terms that vanish for $ q_1+q_2 = 0$}.\end{aligned}\ ] ] the heavy - light currents , @xmath12 , are not part of the action and we list their feynman rules separately . again
, let @xmath266 and @xmath102 be the outgoing and incoming quark momenta , respectively , and let @xmath294 and @xmath295 denote the momenta and polarizations , respectively , of the emitted gluons . also , let @xmath296 be the momentum carried off by the external heavy - light current , where @xmath297 .
we use the notation @xmath298^{(n)}$ ] to indicate the vertex factor for @xmath82-gluon emission from operator @xmath299 . at tree level , @xmath300^{(0 ) } & = & - \frac{i}{2 m } \ ; \sum_j \sin[{\textstyle\frac{1}{2}}(k^{\prime } + k + q_{ext})_j ] \ ; \left(\gamma \ , \gamma_j \right),\\ \left [ j^{(2)}_{a,{\rm lat } } \right]^{(0 ) } & = & - \frac{i}{2 m } \ ; \sum_j \sin[{\textstyle\frac{1}{2}}(k^{\prime } + k - q_{ext})_j ] \ ; \left ( \gamma_j \ , \gamma \right).\end{aligned}\ ] ] for one gluon emission ( @xmath301 , color factor = @xmath268 ) : @xmath302^{(1 ) } & = & - \;g \;\frac{i}{2 m } \ ; \cos[{\textstyle\frac{1}{2}}(k^{\prime } + k + q_{ext})_j ] \ ; \left(\gamma \ , \gamma_j \right),\\ \left [ j^{(2)}_{a,{\rm lat } } \right]^{(1 ) } & = & - \;g \;\frac{i}{2 m } \ ; \cos[{\textstyle\frac{1}{2}}(k^{\prime } + k - q_{ext})_j ] \ ; \left ( \gamma_j \ , \gamma \right).\end{aligned}\ ] ] for two gluon emission ( @xmath290 , color factor = @xmath303 ) : @xmath302^{(2 ) } & = & g^2 \ ; \frac{i}{4 m } \ ; \sin[{\textstyle\frac{1}{2}}(k^{\prime } + k + q_{ext})_j ] \ ; \left(\gamma \ , \gamma_j \right),\\ \left [ j^{(2)}_{a,{\rm lat } } \right]^{(2 ) } & = & g^2 \ ; \frac{i}{4 m } \ ; \sin[{\textstyle\frac{1}{2}}(k^{\prime } + k - q_{ext})_j ] \ ; \left ( \gamma_j \ , \gamma \right).\end{aligned}\ ] ] e. eichten , nucl . phys . b ( proc . suppl . ) * 4 * , 170 ( 1988 ) . b.a . thacker and g.p .
lepage , phys .
d*43 * 196 ( 1991 ) .
lepage _ et al .
, _ phys .
d*46 * 4052 , ( 1992 ) .
el - khadra , a.s .
kronfeld , p.b .
mackenzie , phys .
d*55 * , 3933 ( 1997 ) .
a. ali khan _ et al . , _ phys . rev .
d * 56 * , 7012 ( 1997 ) .
a. ali khan _ et al . , _ hep - lat/9801038 , submitted to phys .
b. b. sheikholeslami and r. wohlert , nucl .
b*259 * , 572 ( 1985 ) . o. hernandez and b. hill , phys .
d*50 * , 495 ( 1994 ) .
p. bouchaud , c. lin , and o. pene , phys .
d*40 * , 1529 ( 1989 ) ; d*41 * , 3541(e ) ( 1990 ) .
p. boucaud _ et al .
, _ phys .
d*47 * , 1206 ( 1993 ) .
a. borrelli and c. pittori , nucl .
b*385 * 502 , ( 1992 ) .
m. golden and b. hill , phys .
b*254 * , 225 ( 1991 ) .
foldy and s.a .
wouthuysen , phys .
rev . * 78 * , 29 ( 1950 ) .
lepage and p.b .
mackenzie , phys .
d*48 * , 2250 ( 1993 ) .
see , for example , c. bendtsen and o. stauning , technical report , dept . of mathematical modelling , technical university of denmark , 1997 ( unpublished ) , and
references contained therein .
lepage , j. comput .
* 27 * , 192 ( 1978 ) .
k. jansen _ et al .
, _ phys .
b*372 * , 275 ( 1996 ) . c. morningstar , phys .
d*50 * , 5902 ( 1994 ) .
g. martinelli and c. sachrajda , nucl .
b*478 * , 660 ( 1996 ) .
j. hein _ et al .
, _ in preparation . c. morningstar , phys .
d*48 * , 2265 ( 1993 ) .
.values of the coefficients @xmath140 , @xmath141 , @xmath142 , @xmath121 , and @xmath122 appearing in eq .
[ ajalph2 ] for various values of the bare heavy - quark mass @xmath21 and nrqcd stability parameter @xmath82 .
uncertainties in the determinations of these parameters due to the use of monte carlo integration are included .
[ tab : one ] [ cols= " > , > , > , > , > , > , > " , ]
|
the temporal component of the heavy - light axial vector current is constructed to one - loop order in perturbation theory and to order @xmath0 , where @xmath1 is the heavy quark mass , in terms of operators suitable for use in lattice simulations of @xmath2 and @xmath3 mesons .
the @xmath4-improved clover action is used for the massless light quark , where @xmath5 is the lattice spacing , and propagation of the heavy quark is described by a nonrelativistic lattice action .
|
coronal mass ejections ( cmes ) are energetic expulsions of plasma from the solar corona that are driven by the release of magnetic energy typically in the range of @xmath0 ergs .
the majority of cmes originate from the eruption of pre - existing large - scale helmet streamers @xcite .
less common fast cmes typically come from smaller , more concentrated locations of magnetic flux referred to as active regions . in this case , the cmes often occur shortly after the flux has emerged at the photosphere , but can also happen even as the active region is decaying . while cmes occur in a wide range of circumstances , all cmes originate above photospheric magnetic polarity inversion lines ( neutral lines ) , which exhibit strong magnetic shear .
shear implies that the magnetic field has a strong component parallel to the photospheric line that separates magnetic flux of opposite sign , and in this configuration , the field possesses significant free energy .
in contrast , a potential field runs perpendicular to the inversion line and has no free energy .
there is enormous evidence for the existence of highly sheared magnetic fields associated with cmes and large flares . at the photosphere
, magnetic shear is measured directly with vector magnetographs ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
higher in the atmosphere , the magnetic field is difficult to measure directly , but its direction may be inferred from plasma structures formed within the field . seen in chromospheric h@xmath1 absorption , filaments form only over photospheric inversion lines @xcite along which the magnetic field is nearly parallel @xcite . fibrils and h@xmath1 loops that overlay photospheric bipolar active regions are also indicative of magnetic shear @xcite .
comparisons between vector magnetograms and h@xmath1 images show that the direction of the sheared photospheric magnetic field coincides with the orientation of such fibril structures @xcite .
higher in the corona , evidence of magnetic shear is found in loops visible in the extreme ultraviolet @xcite and x - ray sigmoids @xcite .
these structures run nearly parallel to the photospheric magnetic inversion line prior to cmes , and are followed by the reformation of closed bright loops that are much more potential in structure .
finally , observations by the transition region and coronal explorer ( trace ) show that 86 percent of two - ribbon flares show a strong - to - weak shear change of the ribbon footpoints that indicates the eruption of a sheared core of flux @xcite
. sheared magnetic fields are at the epicenter of solar eruptive behavior .
large flares are preferentially found to occur along the most highly sheared portions of magnetic inversion lines @xcite .
more recent analysis by @xcite found that shear flows associated with flux emergence drove enhanced flaring .
similarly , active region cme productivity is also strongly correlated with magnetic shear as shown by @xcite .
it is not coincidental that large flares and cmes are strongly associated with filaments , which are known to form only along sheared magnetic inversion lines @xcite .
the buildup of magnetic shear is essential for energetic eruptions , and for this reason , it is of fundamental importance to understanding solar activity .
currently the majority of cme initiation models rely on the application of artificially imposed shear flows
. examples of such models include @xcite . until recently , there was no theory to account for these large scale shear flows . in this paper , we discuss a series of simulations that illustrate a physical process by which these shear flows are self - organized in emerging magnetic fields . we will give a close comparison of these simulations in the context of new observations that make a more complete and compelling picture of a fundamental cause of eruptive solar magnetic activity .
shearing motions and magnetic field alignment with the polarity inversion line , so frequently observed in active regions , are readily explained as a response to the lorentz force that arises when magnetic flux emerges in a gravitationally stratified atmosphere .
the motions take the form of large - amplitude shear alfvn waves in which the magnetic tension force drives horizontal flows in opposite directions across the polarity inversion line .
the physical process was first shown by @xcite and found in later simulations by @xcite . as will be discussed in greater detail ,
what ultimately produces the shearing lorentz force is the nonuniform expansion of the emerging magnetic field .
the shearing process was first simulated by @xcite , with a two - and - a - half - dimensional ( 2.5d ) simulation of flux emergence from a horizontal magnetic layer placed 2 - 3 pressure scale heights below the photosphere .
the crucial aspect of this simulation is that the magnetic field is initially in a _ sheared _ configuration oriented at a 45 degree angle to the plane of variation . in this case , for spatial and temporal variations of the instability described by @xmath2 , @xmath3 is oblique to the magnetic field ( @xmath4 ) .
this distinguishes the mode of instability from a parker mode @xcite in which @xmath3 is parallel to @xmath4 or an interchange mode in which @xmath3 is perpendicular to @xmath4 . with @xmath3 oblique to @xmath4 , the instability can be thought of as a mixed mode , ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . in this case , there will be a component of the lorentz force out of the plane of variation any where the magnetic component out of the plane ( @xmath5 in the chosen coordinates ) is not constant along field lines . in the simulation by @xcite
, it was shown that the departure of @xmath5 from constant values along field lines produced strong shear flows in the emerging flux . when the magnetic flux rises forming an arcade as shown in figure [ layer ] , expansion causes the magnitude of @xmath5 to decrease in the arcade .
the gradient in @xmath5 along field lines results in the lorentz force that drives the shear flows . examining the expression for the @xmath6 component of the lorentz ( tension ) force , @xmath7
, we see the reason for the shear .
the gradient in @xmath5 is negative moving up the arcade in the direction of @xmath8 on one side of the arcade , while on the opposite side , the gradient of @xmath5 along @xmath8 is positive .
these shear flows are illustrated in color in the top row of figure [ layer ] .
here , we see the persistence of shear flows as the magnetic field ( drawn with black lines ) expands in the atmosphere .
the magnitude of the shear flow velocity reaches a peak of 15 km / s high in the arcade , 3 - 4 km / s at the photosphere and approximately 1 km / s below the photosphere .
the maximum shear speed typically reaches half the value of the local alfvn speed .
the bottom row of figure [ layer ] shows the time evolution of magnetic field lines integrated in three - dimensional space , and shown from above projected onto the @xmath9 plane .
footpoints of the field lines are shown with black dots .
these panels show the shear displacement of the footpoints and how the magnetic field evolves to be parallel to the polarity inversion line .
note that the field is most highly sheared in close proximity to the polarity inversion line , and grows more nearly perpendicular to the line with greater distance from it .
this same shearing process is found in more complex simulations such as three - dimensional emerging flux ropes @xcite .
panel ( a ) of figure [ rope ] shows in greater detail than before the structure of the magnetic field and current systems in the emerging flux rope of @xcite that leads to the lorentz force , which drives the shear flows . here
, the axial field strength of the rope is shown in color , while the direction of the magnetic field and current density ( confined to the plane ) are shown with white and black lines respectively on the vertical plane that is at the central cross section of the flux rope . the large expansion of the upper part of the flux rope produces a vertical gradient in the axial component of the magnetic field @xmath10 that results in a horizontal current @xmath11 .
the magnetic field crosses the current in opposite direction on opposite sides of the flux rope .
this produces the lorentz force out of the plane that reverses direction across the flux rope , and drives the horizontal shear flow .
panel ( b ) shows the magnitude of these shear flows on the same central cross section of the flux rope .
the shear flow is greatest precisely where the field is most nearly perpendicular to the current reaching a magnitude of approximately 30 km / s . at this time
, the flux rope has risen twice as high as the flux that emerged from the magnetic layer and similarly the shear flow in the rope is also twice as fast . in the case of @xcite
the flux rope erupted , which was not found in earlier simulations such as @xcite .
this eruption resulted from high speed shear flows close to the neutral line that formed a highly sheared core that lifted off , and was followed by magnetic reconnection .
the bottom panels show the magnetic and current structure at the time of the eruption .
panel ( c ) gives a view of the system that shows the vertical magnetic field in color and the horizontal field with arrows at the photosphere .
a current sheet is illustrated with an isosurface of current density .
the sigmoid shape of the sheet reflects the shape of both the coronal field lines passing close to the sheet as well as the highly sheared field at the photosphere .
this model strongly suggests that x - ray sigmoids seen prior to and during cmes are indicative of a highly sheared magnetic field geometry .
panel ( d ) shows that at this current sheet , reconnection is taking place that separates the upper part of the rope that is erupting from the lower part that has v - shaped field lines full of dense plasma that remains attached to the photosphere .
such flux separations in cmes have been suggested by @xcite based on observations of filament eruptions .
finally we discuss a numerical simulation of an arcade eruption .
this simulation is essentially identical to the simulation of @xcite , with the exception that it is carried out in fully three - dimensional space with the bats - r - us mhd code developed at the university of michigan .
the left panel of figure [ arcade ] shows the initial state of the simulation with field lines drawn in black showing the sheared magnetic arcade . on the boundaries of the computational domain ,
the plasma mass density is shown in color along with the computational grid shown with white lines . the center and right panels show the shear and vertical velocities respectively in color on the central plane of the simulation . the magnetic field confined to the plane ( ignoring the component out of the plane )
is drawn with white lines in both panels . in these panels , we find that the arcade erupts very violently with a shear flow that reach a magnitude of 200 km / s
. the arcade rises at a peak velocity of 150 km / s , and drives a shock ahead of it in the corona . in this simulation ,
the eruption is caused by shear flows driven by the lorentz force .
a shearing catastrophe occurs when the @xmath5 component of the magnetic field can not be equilibrated along field lines .
this simulation is different from the pervious two discussed in that it does not treat flux emergence through the photosphere , but only models the coronal plasma at a temperature of one million degrees .
not needing to resolve the photospheric pressure scale hight allows much larger cells and a computational domain that is 25 times larger than that of the previous two simulations .
the result of this coronal arcade simulation is an eruption that is 5 times faster than the flux rope eruption , and extends out to half a solar radius above the surface .
this progression of eruption velocity with the size of the flux system being treating offers compelling evidence that shear flows driven by the lorentz are capable of producing fast cmes from large active regions .
the build up of magnetic energy in active regions is essential to the onset of cmes and flares .
the magnetic stress must pass from the convection zone into the corona in the form of non - potential fields , and effectively couple layers of the atmosphere .
observations show that such non - potential fields occurs along magnetic polarity inversion lines where the magnetic field is highly sheared .
the evidence that magnetic shear is essential to cmes and flares is provided by the very strong correlation between photospheric shear flows , flux emergence and the onset of cmes and large flares @xcite .
these shear flows are found to be strongest along the magnetic inversion line precisely where flares are found @xcite .
this complements earlier evidence that in the case of two - ribbon flares @xcite , prior to the eruption , h@xmath1 arches over the inversion line are highly sheared , afterward , the arches are nearly perpendicular to the inversion line .
@xcite found a similar pattern of magnetic shear loss in the apparent motion of footpoints in two - ribbon flares .
there are even now observations of subphotospheric shear flows with a magnitude of 1 - 2 km / s 4 - 6 mm below the photosphere that occur during flux emergence @xcite , which simulations show is fully consistent with the lorentz force driving mechanism @xcite .
these ubiquitous shear flows and sheared magnetic fields so strongly associated with cmes are readily explained by the lorentz force that occurs when flux emerges and expands in a gravitationally stratified atmosphere .
this physical process explains and synthesizes many observations of active regions and gives them meaning in a larger context .
this shearing mechanism explains ( 1 ) the coincidence of the magnetic neutral line with the velocity neutral line , ( 2 ) the impulsive nature of shearing in newly emerged flux , ( 3 ) the magnitude of the shear velocity in different layers of the atmosphere , ( 4 ) the large scale pattern of magnetic shear in active regions , ( 5 ) the transport of magnetic flux , and energy from the convection zone into the corona , ( 6 ) eruptions such as cmes and flares . with so much explained ,
it still remains a numerical challenge to model an active region with sufficient resolution to produce a large scale cme by this shearing mechanism .
the rope emergence simulation discussed here only produces a flux concentration that is one tenth the size of an active region , which at this scale simply can not produce an eruption the size of a cme .
however , this simulation illustrates the basic process by which cmes and flares must occur , and current simulations already show a very favorable scaling of eruption size . with increases in computer power ,
simulations of flux emergence should soon be producing cme size eruptions by shear flows driven by the lorentz force .
|
shear flows have been prescribed in numerical models of coronal mass ejections and flares for decades as a way of energizing magnetic fields to erupt .
while such shear flows have long been observed in the solar atmosphere , until recently , there was no compelling physical explanation for them .
this paper will discuss the discovery that such shear flows are readily explained as a response to the lorentz force that naturally occurs as bipolar magnetic fields emerge and expand in a gravitationally stratified atmosphere .
it will be shown that shearing motions transport axial flux , and magnetic energy from the submerged portion of the field to the expanding portion , strongly coupling the solar interior to the corona .
this physical process explains active region shear flows and why the magnetic field is found to be nearly parallel to photospheric polarity inversion lines where prominences form .
finally , shear flows driven by the lorentz force are shown to produce a loss of equilibrium and eruption in magnetic arcades and flux ropes offering a convincing explanation for cmes and flares .
|
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.