• Tidak ada hasil yang ditemukan

Gauge invariance and holographic renormalization

N/A
N/A
Protected

Academic year: 2024

Membagikan "Gauge invariance and holographic renormalization"

Copied!
7
0
0

Teks penuh

(1)

Contents lists available atScienceDirect

Physics Letters B

www.elsevier.com/locate/physletb

Gauge invariance and holographic renormalization

Keun-Young Kim

a

, Kyung Kiu Kim

a

, Yunseok Seo

b

, Sang-Jin Sin

c,d,

aSchoolofPhysicsandChemistry,GwangjuInstituteofScienceandTechnology,Gwangju500-712,RepublicofKorea bResearchInstituteforNaturalScience,HanyangUniversity,Seoul133-791,RepublicofKorea

cDepartmentofPhysics,HanyangUniversity,Seoul133-791,RepublicofKorea dSchoolofPhysics,KoreaInstituteforAdvancedStudy,Seoul130-722,RepublicofKorea

a r t i c l e i n f o a b s t ra c t

Articlehistory:

Received5March2015

Receivedinrevisedform20July2015 Accepted22July2015

Availableonline28July2015 Editor:M.Cvetiˇc

Keywords:

Gauge/gravityduality Holographicrenormalization Gaugeinvariance

We study the gauge invariance of physical observables in holographic theories under the local diffeomorphism.Wefindthatgaugeinvarianceisintimatelyrelatedtotheholographicrenormalization:

thelocalcountertermsdefinedintheboundarycancelmostofgaugedependencesoftheon-shellaction aswellasthedivergences.Thereisamismatchinthedegreesoffreedombetweenthebulktheoryand theboundaryone.Weresolvethisproblembynoticingthatthereisaresidualgaugesymmetry(RGS).By extendingtheRGSsuchthatitsatisfiesinfallingboundaryconditionatthehorizon,wecanunderstand theprobleminthecontextofgeneralholographicembeddingofaglobalsymmetryattheboundaryinto thelocalgaugesymmetryinthebulk.

©2015TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense (http://creativecommons.org/licenses/by/4.0/).FundedbySCOAP3.

1. Introduction

AccordingtoAdS/CFT correspondence, anyglobalsymmetry at theboundarytheoryisliftedtoalocalsymmetryinthebulk[1,2].

Thegaugesymmetry isessential toreducethedegreeoffreedom whichisenlargedbygoingintoonehigherdimension.Thephysical goalinholographyistheboundaryquantitieswhichdo notknow the presence of higher dimension or gauge degrees of freedom, while we use the tools in the bulk theory. Therefore the gauge invarianceofaphysicalquantityisacriticalissueforthevalidityof theAdS/CFT.Alsotracingthegaugeinvariancegivesmuchintuition onthewayhowholographyactuallyworks, especiallyhowglobal symmetryisencodedinthelocalgaugesymmetry.

One can find gauge invariant combinations of the fields, and expressthephysicalquantities intermsofsuch mastervariables, however,itisnotalwayseasytofindsuchgaugeinvariantcombi- nation.Eveninthecasetheyareavailable,itisnotveryconvenient to usesuch fields,especially if manyfields are coupled,because thephysical quantitiesare definedinterms ofthefield variables whichareformallygaugedependent.Forexample[2],energymo- mentum tensor and chemical potential are defined in terms of metric/gaugefieldwhichis notgaugeinvariant.Similarly,heatcur- rentscan be relatedtothe metricperturbationdefinedonly ina

*

Correspondingauthor.

E-mailaddresses:[email protected](K.-Y. Kim),[email protected] (K.K. Kim),[email protected](Y. Seo),[email protected](S.-J. Sin).

specific gauge where time periodhas definite relationwithtem- perature.

In recentworks[3,4],based on[5,6], wedeveloped asystem- aticmethodtonumericallycalculatetheGreen’sfunctionsandall AC transportsquantities simultaneouslyforthe casewheremany fields are coupled and there are constraints due to gauge sym- metry. Althoughwe have testedthe validityof the procedureby showingtheagreementofzerofrequencylimitsofACconductivi- tieswiththeknownanalyticDCconductivities[7–9]westillthink thatwe needtoprovethegaugeinvarianceofourprocedureasa matterofprinciple.Wefoundthatthebulkgaugeinvarianceisin- timately relatedto theholographic renormalization. Althoughthe local counter termswere introduced to kill thedivergences, they alsokill mostofgaugedependence.

Furthermore,thereisaresidualgaugesymmetry(RGS)evenaf- terwefixtheaxialgauge grx=0.While equationsofmotioncan be written in terms of the gauge invariant master fields Ph,Pχ (3.8),itturnsoutthatthequadraticon-shellaction,thegenerating functionfortwopointretardedGreen’sfunctions,cannot bewrit- tenassuch.However,weprovethattheGreen’sfunctionsarestill invariantundersuchasymmetry.

Thereisamismatchinthedegreesoffreedominthebulkand thoseattheboundary:thereareonlytwoindependentbulksolu- tions satisfyingthe in-fallingboundaryconditions whilewe need threesolutionsattheboundarysincetherearethreeindependent sourcefields.TheRGSistheonethatresolvestheproblem:sinceit cannotsatisfyaproperboundarycondition,itisnotapropergauge http://dx.doi.org/10.1016/j.physletb.2015.07.058

0370-2693/©2015TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense(http://creativecommons.org/licenses/by/4.0/).Fundedby SCOAP3.

(2)

symmetrybuta‘solutiongeneratingsymmetry’.Itgeneratethede- siredsolutionattheboundaryandthereforeweshouldaccept its bulkcounterpartasanewphysicaldegreeoffreedomaswellal- thoughit cannot satisfytheinfalling boundarycondition (BC).By extendingtheRGSsuch that itsatisfies infallingboundary condi- tionatthehorizon,wecanmakethebulksolutionmorenaturalin thesense thatit satisfiestheinfalling BC. Withsuchsolution we canalsounderstand theprobleminthe contextof generalstruc- tureofholography, namely thecorrespondence betweena global symmetry atthe boundaryandthe local gauge symmetryin the bulk.

2. Actionandbackgroundsolution

Letusfirstbrieflyreviewthesystemwewilldiscuss,whichhas beenanalysedin detailin[3,7,10].The holographicallyrenormal- ized action(Sren)isgivenby

Sren

=

SEM

+

+

Sc

,

(2.1)

where SEM

=

M

d4x

g

R

2

1 4F2

2

M

d3x

− γ

K

,

(2.2)

is the usual action for charged black hole in AdS space (<0) withtheGibbons–Hawkingtermand

Sψ

=

M

d4x

g

1 2

2 I=1

(∂ψ

I

)

2

,

(2.3)

istheactionfortwofreemasslessscalarsaddedforamomentum relaxationeffect.Sc isthecounterterm

Sc

= η

c

M

dx3

− γ −

4

R

[ γ ] +

1

2

2 I=1

γ

μν

μ

ψ

I

ν

ψ

I

,

(2.4)

whichisincludedtocancelthedivergencein SEM+Sψ.Here we introduced

η

c to keeptrackof theeffectof thecounter term. At theendofthecomputationwewillset

η

c=1.

Theaction(2.1)yieldsgeneralequationsofmotion1 RMN

=

1

2gMN R

2

1 4F2

1

2

2 I=1

(∂ψ

I

)

2

+

1 2

I

M

ψ

I

N

ψ

I

+

1

2FMPFNP

,

(2.5)

MFMN

=

0

, ∇

2

ψ

I

=

0

,

(2.6)

whichadmitthefollowingsolutions ds2

=

GMNdxMdxN

= −

f

(

r

)

dt2

+

dr2

f

(

r

) +

r2

δ

i jdxidxj

,

(2.7) f

(

r

) =

r2

− β

2

2

m0 r

+ μ

2

4 r20 r2

,

m0

=

r30 1

+ μ

2

4r20

− β

2 2r02

,

(2.8)

A

= μ

1

r0 r

dt

,

(2.9)

ψ

I

= β

Iixi

= βδ

Iixi

.

(2.10)

1 Indexconvention:M,N,· · · =0,1,2,r,andμ,ν,· · · =0,1,2,andi,j,· · · =1,2.

These are reduced to AdS–Reissner–Nordstrom (AdS–RN) black branesolutionswhenβ=0.HerewehavetakenspecialβIi,which satisfies 122

I=1βI· βI=β2 forgeneralcases.

The solutions (2.7)–(2.10) are characterized by three parame- ters:r0,

μ

,andβ.r0istheblackbranehorizonposition(f(r0)=0) andcanbereplacedbytemperatureT forthedualfieldtheory:

T

=

f

(

r0

)

4

π =

1 4

π

3r0

− μ

2

+

2

β

2 4r0

.

(2.11)

Non-vanishing components of energy–momentum tensor and chargedensityread

Ttt

=

2m0

,

Txx

=

Ty y

=

m0

,

Jt

= μ

r0

.

(2.12) Ttt=2Txx implies that charge carriers are still of massless character.Fromherewesetr0=1 nottoclutter.

3. Gaugefixingandresidualgaugetransformation

Tostudyelectric,thermoelectric,andthermalconductivitieswe introducesmallfluctuationsaroundthebackground(2.7)–(2.10)

δ

Ax

(

t

,

r

) =

−∞

d

ω

2

π

e

iωtax

( ω ,

r

) ,

(3.1)

δ

gtx

(

t

,

r

) =

−∞

d

ω

2

π

e

iωtr2htx

( ω ,

r

) ,

(3.2)

δ

grx

(

t

,

r

) =

−∞

d

ω

2

π

e

iωtr2hrx

( ω ,

r

) ,

(3.3)

δψ

1

(

t

,

r

) =

−∞

d

ω

2

π

e

iωt

χ ( ω ,

r

) .

(3.4)

The fluctuationsare chosen to be independent of x and y. This isallowed since all thebackground fieldsappearingin theequa- tionsofmotionturnouttobeindependentofxand y.Thegauge fieldfluctuation(δAx(t,r))sourcesmetric(δgtx(t,r),δgrx(t,r))and scalarfield(δψ1(t,r))fluctuationandviceversaandalltheother fluctuationsaredecoupled.Wewillworkinmomentumspaceand htx(

ω

,r) andhrx(

ω

,r)is definedso that it goesto constant asr goestoinfinity.

By linearizing the full equation of motion, we get four equa- tions. However oneof themcan be obtainedby theothers.Thus wemayconsiderfollowingthreeequations:

( χ

− β

hrx

) −

i

μω

ax

β

r2f

(

r

) −

ir2

ω (

htx

+

i

ω

hrx

)

β

f

(

r

) =

0

,

(3.5) ax

(

r

) +

ax

(

r

)

f

(

r

)

f

(

r

) + ω

2ax

(

r

)

f

(

r

)

2

+ μ (

htx

+

i

ω

hrx

)

f

(

r

) =

0

,

(3.6) f

(

r

)

f

(

r

)( χ

(

r

) − β

hrx

) +

f

(

r

)

2

( χ

− β

hrx

)

+

2f

(

r

)

2

( χ

− β

hrx

)

r

+ ω

2

χ (

r

) −

i

β ω

htx

(

r

) =

0

.

(3.7) Ifwe differentiate thethird equation withrespect tor,all equa- tions can be written interms of three variables, Pχ,Ph, andax, where

P

χ

≡ χ

− β

hrx

, P

h

htx

+

i

ω

hrx

.

(3.8) Therefore, hrx is a non-dynamical degree of freedom. Indeed, Pχ,Ph,andax areinvariantunderadiffeomorphismgeneratedby
(3)

ξμ=(0,ζ (r)eiωt,0,0),underwhichthefieldsaretransformedas follows:

δ

hrx

=

1

r2

( ∇

r

ξ

x

+ ∇

x

ξ

r

) = ζ

(

r

)

eiωt

,

(3.9)

δ

htx

=

1

r2

(∇

t

ξ

x

+ ∇

x

ξ

t

) = −

i

ω ζ (

r

)

eiωt

,

(3.10)

δ χ = βζ (

r

)

eiωt

,

(3.11)

δ

ax

=

0

.

(3.12)

Usingthisgaugedegreeoffreedom,onemaysethrx=0,whichis so-called theaxialgauge.Thenumericalcalculationin[3]hasbeen performedin thisgauge. Aquestion arises whethertheresulting physicalquantitiesareindependentofsuchgaugefixingcondition.

Furthermore,evenafterwefixhrx=0,onecanstillfindaresid- ualgauge transformation whichis givenby constant ζ [11]. This residualdiffeomorphismdoesn’tchangethegaugefixingcondition hrx=0 and generates constant shift on htx and

χ

, because the equationsofmotioncontainonlyderivativesofhtx and

χ

andthe linear combinationof them,

ωχ

(r)iβhtx(r), which is invariant under

htx

htx

+

h0

,

and

χ → χ +

i

β

ω

h0

,

(3.13)

whereh0isaconstant.Thusthereisoneparameterconstantsolu- tiongivenby

ax

=

0

,

htx

=

h0

, χ =

i

β

ω

h0

,

(3.14)

whichdoesnotsatisfy in-fallingboundaryconditionsoitisnota physicaldegreeoffreedom.2Wecallittheresidualgaugesymme- try(RGS)becauseitisgeneratedbythezeromodeofadiffeomor- phismgenerator.Thiskindofsolutionwasfirstintroducedin[12].

Whyshouldtherebesucharesidualdegreeoffreedom?Itcan betracedtothedifferenceofthedifferentialequationnearhorizon andthosenearboundary.Neartheblackholehorizon(r1)the solutionsareexpandedas

htx

= (

r

1

)

ν±+1

htx(I)

+

h(txI I)

(

r

1

) + · · · ,

ax

= (

r

1

)

ν±

a(xI)

+

a(xI I)

(

r

1

) + · · · , χ = (

r

1

)

ν±

χ

(I)

+ χ

(I I)

(

r

1

) + · · ·

,

(3.15)

where

ν

±= ±i4

ω

/(12+2β2+

μ

2)= ∓i

ω

/(4

π

T)andtheincom- ingboundarycondition correspondsto

ν

=

ν

+.Byinsertingthese to theequationsof motion,one caneasily find a linearrelations betweenthezero-thmodes:

( ν +

1

)

htx(I)

+ μ

a(xI)

+ β χ

(I)

=

0

.

(3.16) Noticethat allother modesaregeneratedby these. Thusthereis a well defined constraintequation which reducesthe degrees of freedom.

Ontheotherhand,byinsertingtheexpansionnearthebound- ary(r→ ∞)

htx

=

htx(0)

+

1

r2h(tx2)

+

1

r3h(tx3)

+ · · · ,

ax

=

a(x0)

+

1

ra(x1)

+ · · · , χ = χ

(0)

+

1

r2

χ

(2)

+

1

r3

χ

(3)

+ · · · ,

(3.17)

2 Itisaregularsolutionatfuturehorizon.

to the equations of motion,we cannot get any relationbetween the zero-thmodesa(x0),h(tx0),and

χ

(0),allofwhichare relatedto thehighermodes.Moreexplicitly,

ω ( ωχ

(0)

i

β

h(tx0)

) −

2

χ

(2)

=

0

,

i

β( ωχ

(0)

i

β

h(tx0)

) −

2htx(2)

=

0

,

(3.18) which are evolution equations in r-direction. Therefore, there is no constraintequation. Then thereis a crisis ofmismatchofde- greesoffreedomandthiscrisisisresolvedbytheeffectiveresidual degree of freedom described above. However, thisresidualgauge degree of freedom raises another issue of invariance of physics underthissymmetry.WewilladdressthisissueattheendofSec- tion5.

4. Holographicrenormalizationandgaugeinvariance

Nowwecomebacktothequestionwhetherphysicalquantities are independent of the choice of the gauge condition hrx(r)=0.

Wewillshowthisbyprovingthatthegeneratingfunctionofphys- ical quantities, the on-shell action, is invariant even in the case withhrx(r)=0.

Theon-shellrenormalized actiontoquadraticorderinfluctua- tionfields,S(ren2),is

Sren(2)

=

lim

r→∞

d3x

δψ

1

1

2

β

f

δ

grx

1 2f r2

δψ

1

+

2

r

δ

gtx2

1

2f

δ

Ax

δ

Ax

− δ

gtx

1

2

δ

g

˙

rx

1 2r2

( δ

gtx

r2

)

+ μ

2r2

δ

Ax

+ η

c

δψ

1

r2

δψ ¨

1

2

f

− β δ ˙

gtx 2

f

+ β δψ ˙

1

δ

gtx

2

f

2

f

δ

g2tx

,

(4.1)

where f(r)=r2β22mr0 +4rμ22.Wedroppedtheboundarycon- tributionfromthehorizonasaprescriptionfortheretardedGreen function[13].3Nearboundaryr→ ∞,thefluctuationfieldsinmo- mentumspace,(3.1)–(3.4),maybeexpandedas

htx

( ω ,

r

) =

n=0

h(txn)

( ω )

rn

,

hrx

( ω ,

r

) =

n=0

h(rxn)

( ω )

rn

,

ax

( ω ,

r

) =

n=0

a(xn)

( ω )

rn

, χ ( ω ,

r

) =

n=0

χ

(n)

( ω )

rn

,

(4.2)

andusingtheequationsofmotion,wecan obtainaquadraticac- tionasfollows

3 Infact,thecontributionoftheincomingsolutionatthehorizoniszeroin(4.1), whichisreal.However,forageneratingfunctionofretardedGreen’sfunctions,we willtakeonlypartof(4.1)asexplainedbelow(4.3),whichiscomplex.Inthiscase, it turnsoutthatthecontributionfromthehorizonispureimaginary.Fromthis perspective,weshoulddropthecontributionfromthehorizon.

(4)

S(ren2)

=

V2 2

0

d

ω

2

π

− μ

a

¯

(x0)h(tx0)

− μ

h

¯

tx(0)a(x0)

2m0h

¯

(tx0)h(tx0)

+ ¯

a(x0)a(x1)

+

¯ χ

(0)

+

i

β

ω

h

¯

(0)

tx 3

χ

(3)

+ β

hrx(4)

+ ( η

c

1

)

3

4h

¯

tx(0)h(tx0)

2

4h

¯

(tx1)htx(0)

+

4ih

¯

(tx0)h(rx2)

ω +

i

β

h

¯

tx(0)

χ

(0)

ω −

2ih

¯

(tx0)h(rx3)

ω + β

2h

¯

(tx0)h(tx0)

+

4ih

¯

(tx1)h(rx2)

ω −

4h

¯

(tx2)h(tx0)

+

i

β χ ¯

(0)h(tx0)

ω

− ¯ χ

(0)

χ

(0)

ω

2

2m0h

¯

(tx0)htx(0)

4h

¯

(tx0)h(tx3)

2i

ω

h

¯

tx(1)h(rx3)

+ β

2h

¯

(tx1)h(tx0)

+

i

β ω

h

¯

(tx1)

χ

(0)

4i

ω

h

¯

(tx2)h(rx2)

4h

¯

(tx3)htx(0)

+

i

β ω χ ¯

(1)h(tx0)

− ω

2

χ ¯

(1)

χ

(0)

+ [

c.c

],

(4.3)

wheretheargumentofthefields4is

ω

.V2 denotesvolumeinxy spaceand[c.c]means the complexconjugated terms.From here, we will drop the [c.c] term since we want to compute retarded Green’sfunctions[13].

The second line is proportional to a gauge invariant combi- nation under (3.13). Furthermore,one ofthe equation ofmotion includingh(rx4)is

h(rx4)

1

β

2

− ω

2

3i

ω

htx(3)

i

μω

a(x0)

3

β χ

(3)

=

0

.

(4.4) Onecanshowthat(4.4) isequivalenttoaWardidentity

μ

Tμν

+

Fλν

Jλ

− O

I

ν

ψ

I

=

0

,

(4.5) by using the boundary metric andthe other fields in the linear approximationgivenasfollows:

ds2

= η

μνdxμdxν

+

2h(tx0)eiωtdtdx

,

Tμν

=

T(0)μν

+

T(1)μν

F

= −

i

ω

a(x0)eiωtdt

dx

,

Jμ

=

J(0)μ

+

J(1)μ

= ( μ ,

0

,

0

) +

0

,

a(x1)

− μ

h(tx0)

,

0

eiωt

ψ

I

= (β

x

, β

y

) ,

O

I

= O

(1)I

=

3

χ

(3)

+ β

h(rx4)

,

0

eiωt

,

(4.6) where

T(0)μν

=

m0

2 0 00 1 0 0 0 1

⎠ ,

T(1)μν

=

2m0h(tx0)

3htx(3)

+

i

ω

h(rx4)

0 1 01 0 0 0 0 0

eiωt

.

(4.7) One may ask why Ward identity of the boundary theory is in- cluded in the bulk equation of motion. It is not accidental: The

4 a¯(x0)(ω)a(x0)(ω)=a(x0)(ω)bytherealityconditionofδAx.Thesamenota- tionandrealityconditionapplytoalltheotherfields.

translation,xx+ξ0attheboundarytheoryisimbeddedintothe bulk diffeomorphismxx+ξ(x),whichinduces thefield trans- formation+δξ,whichinturnisaspecialcaseofgeneral variation, +δ. Now the equation of motion is coming fromtheinvarianceofbulkactionδSB=0 underthegeneralvari- ation,whiletheWardidentityistherequirementoftheboundary action under the translation δξ0Sb=0. Because AdS/CFT request SB=Sb attheonshell,thelatteriscontainedinthehugetowerof equationofmotionasatinypiece.

Thetermsproportionalto(

η

c1)in(4.3)includethedivergent terms with,a regularization parameter, andfinite termswith- out .Aremarkablefactisthat withthecounter termofweight

η

c=1,notonlythedivergenttermsarecanceled,butalsoallthe hrxdependentfinitetermsdisappearsfromtheon-shellaction,as weclaimedinthebeginningofthissection.

5. Gaugeinvarianceundertheresidualgaugetransformation Ourstartingpointistheaction5

S(ren2)

=

V2 2

0

d

ω

2

π

− μ

a

¯

(x0)h(tx0)

2m0h

¯

tx(0)h(tx0)

+ ¯

a(x0)a(x1)

3h

¯

tx(0)h(tx3)

+

3

χ ¯

(0)

χ

(3)

+

β χ ¯

(0)

+

i

ω

h

¯

(tx0)

hrx(4)

+

c.c

,

(5.1)

which is still dependent on residual gauge (3.13) even after we set hrx=0. Since it is just a constant shift of the solution , its effects are only shifts of zero-th modesand (r) andall of its modes,especially (a(x1),h(tx3),

χ

(3)):=a are intact. Noticethat the recurrence relations derived fromequations of motionrelate highermodeswiththezero-thmodes Ja=(a(x0),h(tx0),

χ

(0)).How- ever,alldependencesofhighermodesonzerothmodesisthrough thegaugeinvariant combination

ωχ

(0)iβh(tx0).See,forexample, (3.18).Thusallhighermodesaregaugeinvariant,whichmakesthe gaugeinvarianceofthe(r)intactinspiteofthecomplicatedde- pendenceofhighermodesonthezerothmodes.

The residual gauge dependence of(5.1) can be understood as follows. The full on shell action should be invariant under the residualgaugetransformation.However,whatwearelookingatis thequadraticpartoftheaction S(ren2),whichgeneratesthe2-point function,intheexpansionof

Sren

[δ] =

S(ren0)

+

S(ren1)

[δ] +

S(ren2)

[δ] + · · ·,

(5.2) whereδ=μνμI) collectivelydenotes thesources of the dualfield theory,whichare boundaryvalues of r12δgμν,δAμ andδψI.S(ren1)[δ]andS(ren2)[δ]aregivenasfollows:

S(ren1)

[δ] =

d3x

1

2

δ

μν

T(0)μν

+ δ

μ

J(0)μ

+ δ

I

O

(0)I

,

(5.3)

S(ren2)

[δ] =

d3x

1

2

δ

μν

T(1)μν

+ δ

μ

J(1)μ

+ δ

I

O

(1)I

.

(5.4)

5 Itcomesfrom(4.1)beforewegetEq.(4.3),forwhichwehavetousetheequa- tionsofmotion.

(5)

Under the residual gauge transformation6 with h0= −i

ω

ζ0, the variationsoftheseactionsare

δ

S(ren1)

[δ] =

V2

d

ω

2

π ζ ¯

0

i

ωμ

a(x0)

+

2i

ω

m0h(tx0)

+

c.c

,

(5.5)

δ

S(ren2)

[δ] = −δ

S(ren1)

[δ]

+

V2

d

ω

2

π ζ ¯

0

3

β χ

(3)

3i

ω

h(tx3)

+

i

ωμ

a(x0)

+

β

2

− ω

2

hrx(4)

+

c.c

.

(5.6)

ThusthetotalvariationisproportionaltotheWardidentity(4.4).

Noticethat SrenisgaugeinvariantbutS(ren2),whichisstartingpoint toderivetheGreenfunction,isnotinvariantbyitself.Nevertheless physical observablesderived from S(ren2) are invariant becausethe Greenfunctionsare second derivatives ofthefull on shellaction atthezerosourcelimit.

At thispoint one can discussa puzzlein countingdegreesof freedom.Thereareonlytwoindependentbulksolutionssatisfying thein-fallingboundaryconditions,7whileweneedthreesolutions atthe boundary since there are three independent source fields.

Therefore,thereisa crisisofmismatchofdegreesoffreedombe- tweenthebulkandboundary.WhatsolvestheproblemistheRGS (3.14). We call it RGS because it is generated by the zero mode ofadiffeomorphismgenerator.Ontheotherhand,tobeaproper gaugedegreeoffreedominthebulk,thediffeomorphismgenerator shouldsatisfytheproperboundaryconditions:infallingathorizon andDirichlet at boundary. The residual gauge symmetry genera- torisaglobalshiftandthereforeitcansatisfyneitherofthem.So suchashiftbythediffeomorphismzeromodeisnotatruegauge symmetry,whileitisasymmetryofthebulkequationsofmotion.

Inotherwords,theRGSisa“solutiongeneratingsymmetry”rather thanagaugesymmetry.Therefore,thegaugeorbitofRGScanpro- videusthenecessarydegree offreedom (d.o.f)nearboundary. To matchthed.o.f,weneedtoacceptitsbulkorbitasphysicalconfig- urationinspiteofthefactthattheresultingbulksolutiondoesnot satisfytheinfallingBC.8Onecangiveamorenaturalbulksolution byextendingRGStoadiffeomorphismwhichsatisfiestheinfalling boundaryconditionanditisreducedtoourpreviousRGSnearthe boundary.Itisgeneratedbyξμ=(0,ζ (r)eiωt,0,0),with9

ζ (

r

) = (

f

(

r

)/

r2

)

iω/(4πT)

,

(5.7) where f isthemetricfactorgiveninEq.(2.8)and isaconstant parameter.NoticethattheRGSisthecasewhereζ (r)isconstant.

We willcall this“boundaryshifting diffeomorphism”(BSD). Now we can understand the degree of freedom mismatchas follows:

Since it is not satisfyingthe Dirichlet bc, it is still not a proper gaugetransformation.Noticealsothatunder(5.7),thegaugeslice is shiftedand some ofthe gauge fieldsbecome singular. Forthe discussiononthetreatingtheseissues,wereferthereadertop. 24 ofRef. [9].10This isthereasonwhytheBSD cangeneratea new

6 Thistransformation changesthe sourcesoftheaction, δμνμI. One shouldnotethattherearenon-vanishingtransformationsforδ00andδ0.

7 Wehavetwosecondorderdifferentialequationsand onefirstorderonein threevariables:ax,htx.Therefore,thereare5boundaryconditionstofix.Ifwe fixthein-fallingboundaryconditionsforallthreevariables,weareleftwithtwo degreesoffreedom.Werecallequations(3.15)and(3.16).

8 SofarwediscussedthedegreeoffreedommismatchusingtheRGS,sinceour formalismin[3]tocalculatetheconductivityisbasedonit.

9 Wethanktheanonymousrefereeforsuggestingtoconsiderthis.

10 ItisverytemptingtoconsiderBSDasagaugetransformationatleastfrombulk pointofview.Ifwedoit,wegettotheproblem:Itsorbitintheboundarygenerate physicalconfigurationwhileitdoesnotinthebulk,sothatcrisisofd.o.fbecomes real!

solution in theboundary. It is preciselythe same logic whyRGS generate new solution.11 Since RGS and BSD shift the boundary values of fields,they generate the Ward identity forthe transla- tion invariance.Thisisatypical examplehowa globalsymmetry isencodedinalocalgauge transformationandhowtheapparent paradox ofthedegree offreedomcan beresolved becauseofthe holographiccorrespondence.12

6. Basisindependence

In[3],weconstructed aformalism toperformtheACconduc- tivities for the case where multiple fields are coupled together.

We had to choose a basis of initial conditions and one can ask whetherdifferentchoicesofbasisgivethesameresult.Answering this question will also provide an alternative reasoning ofgauge invariance.Toprovidethesetup,let usconsider N fieldsa(x,r), (a=1,2,· · ·,N),

a

(

x

,

r

) =

ddk

(

2

π )

de

ikxrp

a

(

k

,

r

) ,

(6.1)

where the index a may include components of higher spin fields. For convenience, rp is multiplied such that the solution a(k,r)goestoconstantatboundary.Inourcase, (1,2,3)= (ax,htx,

χ

)andp=0 for1,3 andp=2 for2.

Nearhorizon(r=1),solutionscanbeexpandedas

ai

(

k

,

r

) = (

r

1

)

νa±

ϕ

ia

+ ˜ ϕ

ai

(

r

1

) + · · ·

,

(6.2)

whereanewsubscripti isintroducedtodenotethesolutionscor- responding toa specificindependent setofinitial conditions.For example,

ϕ

ai maybechosenas

ϕ

1a

=

1

, −( μ ˜ + ˜ β)/(

1

+ ν ),

1

, ϕ

2a

=

1

, −( μ ˜ − ˜ β)/(

1

+ ν ), −

1

,

(6.3)

whereweused(3.16)and

ν

= −i

ω

/(4

π

T)asshownbelow(3.15) forincomingboundaryconditiontocomputetheretardedGreen’s function[13].Duetoincomingboundarycondition,

ϕ

ai determines

˜

ϕ

ai throughhorizon-regularityconditionsothatwecandetermine thesolutioncompletely.Eachinitialvalue vector

ϕ

i yieldsasolu- tion,denotedbyi(r),whichisexpandedas

ai

(

k

,

r

) → S

ai

+ · · · + O

ai

rδa

+ · · · (

near boundary

) ,

(6.4) where Sai are the sources (leading terms) of i-th solution and Oai are the operatorexpectation values corresponding to sources (δa1).

Notice thatwehaveonlytwosolutions whilewehaveathree dimensional vector spaceJof boundaryvalues Ja,a=1,2,3.To

11 Thisargumentisfurtherjustifiedifweconsiderthenumericalcalculationstart- ingfromtheboundaryinsteadfromhorizon.Afterchoosing3fields’svalues,wecan adjusttwo“expectationvalues”suchthatwecangetinfallingboundaryvaluesat thehorizon.Itiseasytoshowthatonlywhenwestartfromasubspaceofcodimen- sion 1,wegetthreeinfallingsolutionnearthehorizon.Ifwestartfromapointoff thisplane,wegetoneinfallingandtwofieldswhicharemixtureofinfallinganda constant.Inthiscalculationthegaugeconditionhrx=0 isintact.Thisdemonstrates thatwecannot imposeinfallingbcforallfieldsathands.Ifwedothesamenumer- icalexperimentforBSD,thepictureisfollowing.TheBSDgeneratetheorbitandit alsomovethegaugeslice.Nowinthiscaseeveninthecasewestartfromtheoff theplane,wecangetthreeinfallingfieldsatthehorizon.Weneedtocalculatethe r-evolutionateach‘gaugefixing’planewhichpassthroughtheinitialdata.

12 Theapparent‘mismatch’isduetothedifferenceinviewingthegaugeorbitof BSD(orRGS)betweenthebulkandboundary.Inthebulk,onecouldviewitas gaugeorbit.Ontheotherhand,fromtheboundarytheorypointofview,thereisno gaugestructureandtheorbitoftranslationsymmetryisphysicaldegreeoffreedom.

(6)

fixsuchmismatchofdegreeoffreedom, weintroduce aconstant solution0(r)= S0=(0,1,iβ/

ω

)alongthegauge-orbitdirection oftheresidualgaugetransformationsothatSa1,Sa2,Sa0 formabasis ofJ.NowSandOaregenericregularmatricesoforder 3.

Thegeneralsolutionisalinearcombinationofthem:let

a

(

k

,

r

) =

ai

(

k

,

r

)

ci

,

(6.5) withrealconstantsci’s.Wecanchooseci suchthatthecombined sourcetermmatchestheboundaryvalue Ja:

Ja

= S

aici

,

(6.6)

whichyields

a

(

k

,

r

) =

ai

(

k

,

r

)

ci

Ja

+ · · · +

a rδa

+ · · · ,

(

near boundary

)

(6.7)

where,with(6.4)and(6.6),

a

= O

aici

= O

ai

(S

1

)

ibJb

=:

CabJb

.

(6.8) NoticethatbothaandCabareinvariantunderthetransformation JbJb+

Sb0becauseCbaSb0=Oai(S1)ibSb0=Oa0=0,whereOa0= 0 sinceitisthesub-leadingtermoftheconstantsolutions.

Ageneralon-shellquadraticactioninmomentumspacehasthe formof

S(ren2)

=

1 2

ddk

(

2

π )

d

¯

Ja

A

ab

(

k

)

Jb

+ ¯

Ja

B

ab

(

k

)

b

,

(6.9)

whereAandBareregularmatricesoforderN. ¯Ja means

Referensi

Dokumen terkait