ScienceDirect
Nuclear Physics B 903 (2016) 262–303
www.elsevier.com/locate/nuclphysb
Scale
invariance
of
the
η
-deformed
AdS
5
×
S
5
superstring,
T-duality
and
modified
type II
equations
G. Arutyunov
a,b,1S. Frolov
c,1,
B. Hoare
d,∗,
R. Roiban
e,
A.A. Tseytlin
f,2 aII. Institut für Theoretische Physik, Universität Hamburg, Luruper Chaussee 149, 22761 Hamburg, GermanybZentrum für Mathematische Physik, Universität Hamburg, Bundesstrasse 55, 20146 Hamburg, Germany cHamilton Mathematics Institute and School of Mathematics, Trinity College, Dublin 2, Ireland dInstitut für Theoretische Physik, ETH Zürich, Wolfgang-Pauli-Strasse 27, 8093 Zürich, Switzerland
eDepartment of Physics, The Pennsylvania State University, University Park, PA 16802, USA fThe Blackett Laboratory, Imperial College, London SW7 2AZ, UK
Received 30November2015;accepted 21December2015
Availableonline 23December2015
Editor: StephanStieberger
Abstract
Weconsider the ABF background underlying the η-deformedAdS5×S5 sigma model. This back-ground fails to satisfy the standard IIBsupergravity equationswhich indicatesthat the corresponding sigmamodelisnotWeylinvariant,i.e.doesnotdefineacriticalstringtheoryintheusualsense.Weargue thattheABFbackgroundshouldstilldefineaUVfinitetheoryonaflat2dworld-sheetimplyingthatthe η-deformedmodelisscaleinvariant.ThispropertyfollowsfromtheformalrelationviaT-dualitybetween theη-deformedmodelandtheonedefinedbyanexacttype IIBsupergravitysolutionthathas6isometries albeitbrokenbyalineardilaton.WefindthattheABFbackgroundsatisfiescandidatetype IIBscale in-varianceconditionswhichfortheR–Rfieldstrengthsareofthesecondorderinderivatives.Surprisingly, wealsofindthattheABFbackgroundobeysaninterestingmodificationofthestandardIIBsupergravity equationsthatarefirstorderinderivativesofR–Rfields.Thesemodifiedequationsexplicitlydependon KillingvectorsoftheABFbackgroundand,althoughnotuniversal,theyimplytheuniversalscale invari-anceconditions.Moreover,weshowthatitispreciselythenon-isometricdilatonoftheT-dualsolutionthat
* Correspondingauthor.
E-mail addresses:[email protected](G. Arutyunov),[email protected](S. Frolov),[email protected]
(B. Hoare),[email protected](R. Roiban),[email protected](A.A. Tseytlin).
1 CorrespondentfellowatSteklovMathematicalInstitute,Moscow. 2 AlsoatLebedevInstitute,Moscow.
http://dx.doi.org/10.1016/j.nuclphysb.2015.12.012
leads,afterT-duality,tomodificationoftype IIequationsfromtheirstandardform.Weconjecturethatthe modifiedequationsshouldfollowfromκ-symmetryoftheη-deformedmodel.Allourobservationsapply alsotoη-deformationsofAdS3×S3×T4andAdS2×S2×T6models.
©2015TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense (http://creativecommons.org/licenses/by/4.0/).FundedbySCOAP3.
1. Introduction
Thestudy of integrabledeformationsof theAdS5×S5 superstring sigmamodelis an im-portantdirection inthe searchfor newsolvable examplesof AdS/CFTduality. Aninteresting one-parameter integrablegeneralisationof theclassical AdS5×S5 Green–Schwarz action re-latedtothequantumdeformationoftheunderlyingsupergroupsymmetrywasfoundin [1].Just fromtheconstructionofthis“η-model”(basedonaparticular current–currentdeformationof thesupercosetaction [2]generalisingthebosonicmodelof [3])thereisnoapriorireasonwhyit shoulddefineascaleinvariant(UVfinite)2dtheoryand,moreover,whyitshouldpreservethe conformal(Weyl)invarianceandhencestillcorrespondtoaconsistentsuperstringtheoryasthe undeformedAdS5×S5modeldoes.3
Theonlyindicationinthisdirectionisthattheη-modelaction,liketheoriginalAdS5×S5
action,isinvariantunderaversionoffermionicκ-symmetry [1],whichreducesthenumberof fermionsbyhalf.However,the usualclaimthat κ-symmetryimpliesthecorrespondingaction canbeinterpretedasthatofaGSsuperstringpropagatinginabackgroundthatisaconsistent type IIsupergravitysolution(andhencedefinesaconsistentcriticalsuperstringtheory)assumes thattheκ-symmetryisofthestandardGS“projector”form [5].Thisismostprobablynotthe casefor theη-modelathigherordersinfermions.Indeed,itwasfoundin [6,7]thatthetarget spacebackgroundcorrespondingtotheη-modelaction [1],interpretedasaGSaction,doesnot representatype IIBsupergravitysolution.
StartingwiththeGSLagrangianwritteninsuperspaceform(ZM=(xm,θα))
L=(√hhabEMr ENsηrs−abBMN)∂aZM∂bZN , (1.1)
onecansolvethestandardtypeIIsuperspaceconstraintsandBianchiidentitiesforE(Z),B(Z)
(whichimplythesupergravityequations)inordertoexpresstheGSactionintermsofcomponent fields.Onethenobservesthatthe dilatonφ (whichispartof thedilatonsuperfield (Z)that isintroducedintheprocessofsolving theconstraints)entersthe world-sheetaction(i)inthe combinationF=eφF withtheR–Rfieldstrengthsstartingatorderθ2 and(ii)viaderivatives
∂mφ startingat order θ4 (see [8] and the references therein). Thisactionhas classical Weyl invarianceandκ-symmetry,whichwillbebroken,ingeneral,byquantum corrections.Asfor thebosonicstring [9],tocancelthe2dstresstensortraceanomalyrequiresaddingthefamiliar 1-loopdilatoncounterterm∼d2z√hR(2) (Z)(see [10,11]andthereferences therein).4
Thecaserelevanttoourdiscussionbelowisaspecialisometrictype IIsolutionforwhichthe metricGmn,B-fieldBmnandR–Rfields Fm1...mn areinvariantwhileφislinearintheisometric
3 Thisisincontrast,e.g.,totheintegrabledeformation[4]basedonTsTdualitytransformations,whichpreserve
conformality.Inparticular,theTsTdeformedbackgroundisasolutionoftype IIBsupergravity.
4 Thisadditionaltermiscertainlyrequiredtoreproducethestandard1-loopWeyl-invarianceconditionsfortheGand
directions.InthiscasetheGSactionwilldependontheisometriccoordinatesonlythroughtheir 2dderivativesandcanthusbeT-dualised.Asweshallsee,inthiscasetheT-dualmodelwillbe scaleinvariantbutmaynotbeWeylinvariant(onemaynotbeabletocanceltheWeylanomaly byalocalcounterterm),i.e.maynotcorrespondtoatype IIsupergravitysolution.
The ABF background[6,7]includes the10d metric G,the B-fieldandthe R–R fields Fn (n=1,3,5)thatareextractedfromthequadraticfermionicpartoftheactionof [1]putintothe usualGSform,
A= −T
d2z
1
2(η abG
mn−abBmn)∂axm∂bxn
+iθ ¯ mDθ ∂xm+ ¯θ mF· nθ ∂xm∂xn+. . .
. (1.2)
For the standard GS actionin atype IIB supergravity background Fn are interpreted as the productsofthedilatonandtheR–Rfieldstrengths Fn=eφFnbutintheη-modelcasethereis noindependentinformationaboutthe dilaton,andthereexistsnodilatonfieldthatcompletes
G,B,FnoftheABFbackgroundtoatype IIBsolution [7].
Whilenotsolvingthestandardtype IIBequationsdirectlythisABF backgroundstill turns outtobeveryspecial:itisrelatedbyT-dualitytoanexacttype IIBsupergravitysolution [12, 13].ThelatterHTbackgroundinvolvesanon-diagonalmetricGˆ,animaginary5-formFˆ5and
the dilatonφˆ,andtheT-dualityappliedinall6 isometricdirections actsonlyonthefields Gˆ
andFˆ5=eφˆFˆ5 enteringthe correspondingGSaction (1.2)onaflat 2dbackground. TheGS
actionforanytype IIsolution(andthusfortheHTbackground)shouldbeWeylinvariantand,in particular,scaleinvariant.AstheT-dualityappliedtotheGSaction [14]isasimplepathintegral transformation,theT-dualityrelationbetweenthe ABFandHTbackgroundsimplies [12]that theη-modelactionshoulddefineascaleinvariant2dtheoryatleastto1-looporder.
However,theremaybeaproblemwithWeylinvariancefortheη-modelactiononacurved 2dbackground.TheHTdilatonφˆ hasatermlinearlydependingontheisometricdirectionsof
ˆ
GandFˆ5andthusonecannotdirectlyapplythestandardT-dualitytransformationrules [15]to
thefullHTbackgroundtogetafullT-dualsupergravitysolution,andthustheWeylinvariance oftheT-dualsigmamodelrequiresfurtherinvestigation.5Thisisofcourseconsistentwiththe observation [7]thatthe ABFbackgrounddoesnotsatisfythefullsetof type IIBsupergravity equations.
The aim of the present paper istofurther clarifyandextend theseobservations.Weshall demonstratethattherelationbyformalT-dualitybetweentheABFandHTbackgroundsimplies thattheformer,whilenotasupergravitysolution,shouldsatisfythefollowingtwogeneralisations or“modifications”ofthetype IIsupergravityequations:
(i)thescaleinvarianceconditionsforthetype IIsuperstringsigmamodel(withequationson theR–RfieldsFbeingof2ndorderinderivatives)
5 The1-loopWeylinvarianceconditionsoftheNSRorGStype IIsuperstringsigmamodelarebelievedtobeequivalent
tothefieldequationsoftype IIsupergravity.Whilethisisawell-establishedfactintheNS–NSsector[9,16]thiswas neverdemonstrateddirectlywiththeR–Rcouplingsincluded(forsomerelatedwork,mostlyfortheheteroticstring,see
(ii)asetofequations thatarestructurallysimilartothoseoftype IIsupergravity(with 1st-orderequationsfortheR–Rfields F)butinvolving,insteadofderivativesofthedilaton,acertain co-vectorZmplayingnowtheroleofthedilatonone-formandaKillingvectorImresponsible forthe“modification”oftheequationsfromtheirstandardform.6
Whilethe scaleinvarianceconditions areuniversal,the secondsetof equations(whichwe shallrefertoas“I-modified”type IIequations)onlyapplytoparticularbackgroundswith iso-metricG,Band F-fields,whicharerelatedbyformalT-dualitytoatype IIsolution(G,ˆ B,ˆ Fˆ,φˆ) withthe dilatonφˆ containing aterm linearintheisometriccoordinates.Suchadilaton back-ground,breakingisometriesbyalineartermonly,isspecial.Asthetype IIsupergravityequations writtenintermsoftheF-fieldsonlydependonthedilatonthroughitsderivatives,theyremain independentoftheisometricdirections.Asaresult,thestandardtype IIsupergravityequations fortheT-dualsolution(G,ˆ B,ˆ Fˆ, ˆφ)canbere-interpretedascertainmodifiedtype IIequations fortheoriginalfields(G,B,F),alsodependingonthevectorsZ andI.TheKillingvectorIm
dependenceisfixedbythetermlinearinisometriccoordinatesinthedilaton,whilethevector
ZmisdeterminedbyapplyingthestandardT-dualityrulestothepartofthedilatonindependent oftheisometriccoordinates.7
ItispossibletoexpressthemodifiedequationsfortheNS–NSfieldsintermsofjustonesingle vectorXm=Zm+Im,whichisthevectorthatappearsinthescaleinvarianceconditions.The su-perstringscaleinvarianceconditionsgeneralisethefamiliarone-loopscaleinvarianceconditions forthebosonicsigmamodelwithcouplingsGmnandBmn(cf. (1.2))
βmnG ≡Rmn−14HmklHnkl= −DmXn−DnXm, (1.3)
βmnB ≡12DkHkmn=XkHkmn+∂mYn−∂mYn. (1.4)
Here the terms involving Xm [17] andYm do not contribute toon-shell UV divergencesor, equivalently, reflect the freedom of renormalisation by reparametrisations andB-fieldgauge transformations.The Xm terms drop outof the actionif thesigmamodel field xm issubject to the classical equations, or, equivalently, they can be absorbed in a field renormalisation,
xm→xm+Xmlog.TheoriginoftheXkHkmntermcanbeunderstoodeitherbystartingwith acountertermproportionalto(DmXn+DnXm)∂axn∂bxn+. . .,integratingbypartsandusing theequations ofmotion for xm,orbyobservingthat Bmn transformsunder acombinationof reparametrisationsandgaugetransformationsasXk∂kBmn+∂mXkBkn−∂nXkBkm+∂mYn−
∂nYm =XkHkmn+∂mYn−∂nYm whereYm or Ym drop out of the sigmamodel actionupon integrationbyparts.
TheWeylinvarianceconditionsareequivalenttothevanishingof thetraceofthe2dstress tensoroperatoronacurved2dbackground.FortheNSRtype IIsuperstringsigmamodeltheycan besatisfiedprovidedoneaddsthedilatonterm ∼R(2)φ(x)[9,19–21,16]:theyareastrongerform ofthescaleinvarianceconditions (1.3), (1.4)withXmandYmnolongerarbitrary,butgivenby
Xm=∂mφ , Ym=0. (1.5)
TheWeylinvarianceequations (1.3), (1.4), (1.5)implythe“centralcharge”identity [9,22]
∂mβ¯φ=0, β¯φ≡R−121Hmnk2 +4D2φ−4∂mφ∂mφ , (1.6)
6 Inwhatfollowsweshallnotdistinguishbetweenco-vectorsandvectors,referringtobothXmandXmasvectors. 7 LetusstressthatFˆandφˆexplicitlydependontheisometriccoordinates.ItisG,ˆ B,ˆ Fˆ,dφˆ,andG,B,F,Zthatare
i.e.thattheeffectivedilaton“β-function”isaconstant(whichshouldbezeroincriticalstring theory). The fullsetof Weylinvariance equations for G,B andφ follows fromthe effective actionwiththesameformastheNS–NSsectorofthetype IIsupergravityaction(F≡eφF)
S=
ddx√Ge−2φβ¯φ+F F+. . .
=
ddx√G e−2φβ¯φ+FF+. . ., (1.7)
wherewehaveindicatedthepresenceoftheR–Rfieldstrengthtermsforfuturereference. The generalisationof thescaleinvariance conditionstothepresenceof R–Rfieldsisgiven by (1.3), (1.4)withextraFF terms,togetherwithasetofsecond-derivativeequationsforthe R–R fields F that directlyenter the GSaction (1.2), 12D2F+. . .=X∂F +F∂X. Herethe r.h.s.standsforreparametrisation(Liederivative)termswiththesameX-vectorasin (1.3), (1.4) anddotsindicatenon-linearterms.InthespecialcasewhenXm=∂mφtheseequationsarethe consequenceof the type IIB equations or Weylinvariance conditions, whichare 1st orderin
F =e−φF,i.e.d F+. . .=0 anddF+. . .=0.8Theseuniversalscaleinvarianceconditions willbesatisfiedbytheABFbackgroundforaparticularchoiceofthevectorsXmandYm.
Toexplaintheoriginofthesecond“I-modified”setofequationsletusfirstignoretheR–R fieldsandassumethatthereexiststhefollowingmetric-dilatonbackgroundthatsolvestheWeyl invarianceequations(i.e.Rmn+2DmDnφ=0,β¯φ=const)
ˆ
ds2=e2a(x)ˆ [dyˆ+ ˆAμ(x)dxμ]2+gμν(x)dxμdxν , φˆ= −cyˆ+f (x) . (1.8)
Herethemetrichasanisometrywhichisbroken bythelinearterminthedilaton(c=const). Examplesofsuchnon-trivialsolutions9canbefoundbytakingspeciallimitsofgaugedWZW backgrounds [13].T-dualisingthismetric,wefindadiagonalmetricGandB-field,i.e.
ds2=e2a(x)dy2+gμν(x)dxμdxν, B= ˆAμ(x) dy∧dxμ, a= −ˆa . (1.9)
Forc=0 (i.e.whenφˆ isisometric)thesefieldstogetherwiththeT-dualitytransformeddilaton
φ= ˆφ− ˆa would solve the standardWeylinvariance equations (1.3), (1.4)with Xm=∂mφ,
Ym=0.Fornon-zero ctheequationRˆmn+2DˆmDˆnφˆ=0 (fortheoriginalbackground (1.8)) expressedintermsofthedualfieldsG,Bwillcontainadditionalc-dependenttermsobstructing (fornon-constanta(x))theintroductionofanewdilatonscalar.Still,theycanbeputinamore generalformRmn+DmXn+DnXm=0 withaspecialvectorXgivenby10
Xmdxm≡Imdxm+Zmdxm=c e−2ady+ ∂μ(φˆ− ˆa)+cAˆμ
dxμ. (1.10)
Thedilatonequationβ¯φ=0 fortheoriginalbackground (1.8)alsocanberewrittenasthe fol-lowinggeneralisedequation(cf. (1.6))11
8 Therelationbetweenthe1st-orderand2nd-orderequationsonFhasthesamespiritastherelationbetweentheDirac
andtheKlein–Gordon(squaredDirac)equationsforspinorfields.
9 Itisimportantthatdilatonhasalineartermina“warped”isometricdirectionofthemetric,i.e.a(x),Aμ(x)are
non-constant,otherwisetheeffectofaddingthelineardilatonwouldbetrivial.
10 TheneedtointroducethevectorXm,whichis notsimplyagradientofascalar,isthereforedirectlyrelatedtothe
feature∂yˆφˆ= −c =0.
11 Notethatthisequationisnotpresentinthelistofscaleinvarianceconditions,andWeylinvarianceconditionsrequire
¯
βX≡R−121Hmnk2 +4DmXm−4XmXm=0, (1.11)
thatissatisfiedfortheT-dualbackground.
TheT-dualbackground(G,B)definesasigmamodelthatisscaleinvariantonaflat2d back-ground(satisfyingequations (1.3), (1.4)withYm=Xm)butwhichisnot Weylinvariant.The traceof stress tensorT =βmnG ∂axm∂axn+βmnB ab∂axm∂bxn is atotalderivative T = ∇aNa,
Na=2(Xm∂axm+abYm∂bxm)(uptotermsproportionaltothexmequationsofmotion).This cannotbecancelledbyalocalcounterterm(theclassicaldilatonterm)unlessXm=∂mφ,Ym=0
[19,20],whichisnotthecasefortheABFbackground.Thesigmamodelsbasedon (1.9)(with explicitbackgroundsgivenbelow)thusrepresentparticularexamplesof2dscaleinvariant theo-riesthatavoidtheZamolodchikov–Polchinskitheorem [23]duetotheirnon-compactness(and/or non-unitarity related tothe presence of time-like directions).It thus remains unclear if such backgroundsrelatedbyformalT-dualitytoWeylinvariantmodels (1.8)canalsobeassociated somehowwithaconsistentcriticalstringtheory.
Asweshallseebelow,asimilargeneralisationofthefullsetofthebosonictype IIsupergravity equationsalsoexistsinthepresenceofR–Rfields Fnthathavethesameisometriesasthemetric (i.e.when (1.8)isextendedtoananalogoftheHTsolution [12]).Thusingeneral,givenatype II solution withnon-isometric lineardilatonthere will bean associated (“T-dual” or ABF-like) backgroundsolvingsuchamodifiedsetoftype IIequations.
Therestofthispaperisorganisedasfollows.Insection2weshallpresentthegeneralscale invarianceconditionsforthecouplingsG,Bofthesigmamodel (1.2)thatgeneralise (1.3), (1.4) tothepresenceoftheR–Rfields FandshowthatthereexistsuchvectorsX=Z+I andY that theseequationsaresatisfiedbytheABFbackground.Insection3weshallderiveamodification ofthe standard1st-orderIIBsupergravityequations ofthe R–Rfluxesthatis “driven”bythe specialisometryvectorI andwhicharesatisfiedbytheABFbackground.Insection4weshall showthatcombiningthese1st-orderequationsonecanfind2nd-orderequationsforFthathave therightstructure(whengeneralisedtoarbitraryvectorX)tobeinterpretedasscaleinvariance conditionsontheR–Rcouplings.Insection5weexplainhowthestandardtype IIsupergravity equationsforasolutionwiththedilatonlinearalongtheisometricdirectionsismappedtothe modifiedequationsforT-dualsolution.
Ournotationandsomeusefulrelationsaresummarisedin Appendix A.In Appendix Bwe present the explicit form of the ABF backgroundand the T-dual type IIB HT solution. Ap-pendix Ccontainsthederivationoftheidentity∂mβ¯X=0 fromthemodifiedtype IIequations whichiscloselyrelatedtotheon-shellconservationofR–Rstresstensor.In Appendix D,starting withthemodifiedtype IIequations,wederivethe2nd-orderequationsfortheR–Rfieldsthatare candidatesforthecorrespondingscaleinvariance conditions.In Appendix Ewe remarkonan alternativederivationoftherelation (2.13)forthevectorZ,whichplaystheroleofthedilaton one-forminthemodifiedequations.In Appendix FwesummarisetheanalogsoftheABFand HTbackgrounds intheAdS2×S2×T6andAdS3×S3×T4casesandgivethecorresponding expressionsforthevectorsX,Y andI thatsolvethescaleinvarianceandmodifiedtype II equa-tions.In Appendix Gweexplainhowthe2nd-orderequationsfortheR–RcouplingsFemerge astheone-loopconditionsofscaleinvariancefortheGSsigmamodel (1.2).
2. Scale invariance conditions and modified type II equations: NS–NS sector
Appendix G).Theθ¯Fθ ∂x∂xtermsintheGSaction (1.2)shouldleadtoone-loopdiagrams(with onebosonicandonefermionicline)contributinglogarithmicUVdivergences ∼FF∂x∂x.These termswillproduceextra FFtermsintheβ-functionsin (1.3)and (1.4).Inparticular,theanalog of theEinsteinequation (1.3)shouldpickuptheR–RstresstensortermandtheB-field equa-tion (1.4),theFFtermasintheIIsupergravityequations.12ThisisexpectedasforXm=∂mφ,
Ym=0 theresultingequationsaretheWeylinvarianceequationsthatshouldbeequivalenttothe type IIsupergravityequations.
ThescaleinvarianceequationsfortheF-fields(tobediscussedinsection4)willnot,however, havethefamiliarsupergravityformof1st-orderequationsforF (theseshouldfollowfromthe Weylinvarianceconditions).Insteadtheywillbeof2ndorder,D2F+. . .=X-dependentterms, andforXm=∂mφwillbeaconsequenceofthe1st-ordersupergravityequations.
Explicitly,thescaleinvarianceconditions (1.3)and (1.4)generaliseto
βmnG ≡Rmn−14HmklHnkl−Tmn= −DmXn−DnXm, (2.1)
βmnB ≡12DkHkmn+Kmn=XkHkmn+∂mYn−∂nYm, (2.2)
Tmn≡12FmFn+14FmpqFnpq+4×14!FmpqrsFnpqrs−21Gmn(12FkFk
+ 1 12FkpqF
kpq) ,
(2.3)
Kmn≡12FkFkmn+121FmnklpFklp. (2.4)
HereFm,Fmnk,FmnklpareR–Rfieldsoftype IIBsupergravity(fornotationsee Appendix A). ForXm=∂mφ,Ym=0 theseequationsfollowfromtype IIBsupergravityaction (1.7).Tmnis thefamiliarstresstensorthatfollowsfromthetype IIBaction (1.7)uponvariationoverGmn.13
Aswas notedin [12],the existence ofthe HTsolution relatedtotheABF backgroundby T-duality,suggeststhattheGSsigmamodelforthelatterdefinedonaflat2dbackgroundshould be scaleinvariant(atleasttoleading,1-loop,order).Our keyobservationisthat indeedthere existvectorsXm andYm such thateqs.(2.1) and(2.2)aresatisfied for theABF background
(B.1).ThevectorXmrequiredtosatisfy (2.1)turnsouttobe(see Appendix Bfornotation)
X≡Xmdxm=c0
1+ρ2
1−2ρ2dt+c1ρ 2
sin2ζ dψ2+c2
ρ2cos2ζ
1+2ρ4sin2ζdψ1
+c3
1−r2
1+2r2dϕ+c4r
2sin2ξ dφ 2+c5
r2cos2ξ
1+2r4sin2ξdφ1
+ 2ρ4sin 2ζ
2(1+2ρ4sin2ζ )dζ+
1
ρ
1− 3
1−2ρ2+
2 1+2ρ4sin2ζ
dρ
+ 2r4sin 2ξ
2(1+2r4sin2ξ )dξ+
1
r
1− 3
1+2r2+
2 1+2r4sin2ξ
dr . (2.5)
Xmcanbesplitinthefollowingway
Xm=Im+Zm, DmIn+DnIm=0, DmIm=0, (2.6)
12 ForanargumentsupportingthisintheNSRformalismsee[18].
13 Notethatinthefirst(NS–NS)termof(1.7)onedoesnotneedtovarythe√Gfactorasitscontributionvanishesafter
whereIm=6i=1ci(I(i))m. Theindex i labelsthe 6 isometricdirections yi =(t,ψ2,ψ1,ϕ, φ2,φ1)ofthe 10dABF metricandci arearbitrary constantcoefficients.(I(i))m arethe6 in-dependent commuting Killing vectors of the ABF background: the Lie derivatives of the G,
B and F-fieldsin [6]along Imallvanish.Ifwe splitthecoordinatesas xm=(yi,xμ)where
μ =1,2,3,4 labelsthenon-isometricdirectionsxμ=(ζ,ρ,ξ,r),then
Im=
6
i=1
δmi ciGii(xμ) , Im=δimci=const, Zm=δmμZμ(xν) . (2.7)
ThevectorYmrequiredtosatisfy (2.2)ontheABFbackgroundisfoundtobe14
Y≡Ymdxm=4
1+ρ2
1−2ρ2dt+2
ρ2cos2ζ
1+2ρ4sin2ζdψ1
+4 1−r 2
1+2r2dϕ−2
r2cos2ξ
1+2r4sin2ξdφ1
+ 2ρ4sin 2ζ
2(1+2ρ4sin2ζ )dζ+
1
ρ
1− 3
1−2ρ2+
2(−1c 2−1)
1+2ρ4sin2ζ
dρ
+ 2r4sin 2ξ
2(1+2r4sin2ξ )dξ+
1
r
1− 3
1+2r2−
2(−1c5+1)
1+2r4sin2ξ
dr . (2.8)
Weobservethatifwefixciin (2.5)tothefollowingspecificvalues
c0=c3=4 , c1=c4=0, c2= −c5=2 , (2.9)
thenYmandXmcoincide
Ym=Xm. (2.10)
Thenextsurprisingobservationisthatforthesespeciallychosenvaluesofci in (2.9)thevector
Xmsatisfiesalsoadirectgeneralisation (1.11)ofthedilatonequation (1.6)(∂mφ→Xm)15:
¯
βX≡R− 1 12H
2
mnk+4DkXk−4XkXk=0. (2.11)
Asweshallshowin Appendix Cthisβ¯Xsatisfiesthegeneralisationofthedilatonidentity (1.6)
∂mβ¯X=0. (2.12)
Thereasonforthisparticularchoiceofciin (2.9)canbetracedtotheformofthelinearterms inthedilatonφˆ of theT-dual HTsolution (B.3).That isthepresenceof theI-terminXm in
(2.6)reflectsthe presenceofthenon-isometriclineartermsinφˆ.Therefore,theseterms drive themodificationof theequationssatisfiedbytheABFbackgroundfromtheirstandardtype II form.InthissensetheZmpartofXmmaybeinterpretedastheanalogof∂mφinthemodified equations.Indeed,onecancheckthat forIm in (2.7)withci chosenasin (2.9)thefollowing relationissatisfied
∂mZn−∂nZm+IkHkmn=0. (2.13)
14 Yisofcoursedefinedmoduloatotalderivative. 15 SinceDnXn=DμZμ, XmXm=Gijc
Thismaybeinterpretedasamodified“dilatonBianchiidentity”:ifImisformallysettozerothen
Zmbecomesaderivativeofascalar,∂mφ.Ingeneral,assumingthatImrepresentsanisometryof theB-field,i.e.theLiederivative(LIB)mn=Ik∂kBmn+Bkn∂mIk−Bkm∂nIkvanishes(modulo agaugetransformationterm∂mUn−∂nUm),wecansolve (2.13)as16
Zm=∂mφ+BkmIk, (2.14)
where∂mφ termrepresentsthetrivial “zero-mode”solution.In theparticularcaseof theABF backgroundwithZmandImgivenby (2.5), (2.6), (2.7)andci fixedasin (2.9)wefind
Xm=Ym=Im+Zm=∂mφ+(Gkm+Bkm)Ik, (2.15)
φ=12log (1−κ
2ρ2)3(1+κ2r2)3
(1+κ2ρ4sin2ζ )(1+κ2r4sin2ξ ). (2.16)
Thescalarφin (2.16)ispreciselytheonethatisfound [12]byapplyingthestandardT-duality transformationruletotheisometricpartofthedilatonφˆoftheHTsolutionin (B.3)(cf. (1.10)).
3. Modified type II equations: first-order equations for R–R couplings
Let us now explorewhat modificationof the type IIBequations for the R–R couplings is satisfiedbytheABFbackground.
Thestandardequationsoftype IIBsupergravity [28]intheR–Rsectorwrittenintermsofthe rescaled F=eφF fieldstrengthsarepairsof dynamicalequations andBianchiidentities(see Appendix Afornotation)17
DmFm−ZmFm−16HmnpFmnp=0, dF1−Z∧F1=0, (3.1)
DpFpmn−ZpFpmn−16HpqrFmnpqr=0, dF3−Z∧F3+H3∧F1=0,
(3.2)
DrFrmnpq−ZrFmnpq+361εmnpqrstuvwHrstFuvw=0, dF5−Z∧F5+H3∧F3=0.
(3.3)
HereZ=Zmdxm=dφisthedilatonone-form.Thefive-form F5isalsorequiredtosatisfythe
self-dualityequationF5=F5 whichimpliestheequivalenceofthefirstandsecondequation
in(3.3).
An apriori surprisingobservation isthat thereexist directgeneralisationsof the 1st-order equations (3.1)–(3.3)involvingZ=ZmdxmandI=Imdxmin (2.5), (2.6),withfixedvaluesof thecoefficientsci asgivenin (2.9),whicharesolvedbytheABFbackground (B.1).Explicitly, theequationsfortheone-formF1in (B.1)are
DmFm−ZmFm−16HmnpFmnp=0, ImFm=0, (3.4)
(dF1−Z∧F1)mn−IpFmnp=0. (3.5)
16 Ingeneral,wefindZm=∂mφ+BkmIk−Um.UndergaugetransformationsofBthevectorUmtransformssothat φmaybeassumedtobeinvariant.IntheparticularcaseoftheABFbackground(B.1)withtheB-fieldchoseninthe manifestlysymmetricformwehaveUm=0.
17 Notethatallequationsincluding(2.2)areinvariantunderthesimultaneouschangeofsignofH
WehaveaddedtheconditionImFm=0 asanindependentequationon F1.18
Similarly,theequationsthatgeneralise (3.2)andaresatisfiedforthethree-form F3in (B.1)
arefoundtobe
DpFpmn−ZpFpmn−16HpqrFmnpqr−(I∧F1)mn=0, (3.6)
(dF3−Z∧F3+H3∧F1)mnpq−IrFmnpqr=0. (3.7)
TheequationssatisfiedbyF5oftheABFbackgroundarefoundtobe
DrFrmnpq−ZrFrmnpq+361εmnpqrstuvwHrstFuvw−(I∧F3)mnpq=0, (3.8)
(dF5−Z∧F5+H3∧F3)mnpqrs+16εmnpqrstuvwItFuvw=0. (3.9)
Thesetwoareequivalentinviewoftheself-dualityofF5.
Thesemodified equations (3.4)–(3.9)reduceback to (3.1), (3.2), (3.3)ifwedrop allterms withImandassumethatdZ=0,i.e.ifweset
Zm→∂mφ , Im→0. (3.10)
Thestructureof (3.4)–(3.9)supportstheinterpretationofZasageneralised“dilatonone-form”, whiletheisometryvectorI effectivelydrivesthedeformationofthestandardtype IIBequations. Aninterestingobservationisthatthereexistcertaincombinationsoftheequations (3.4)–(3.9) thatdependonZ andI onlythroughthecombinationX=Z+I,whichenteredtheNS–NS equationsoftheprevioussection.Thesearefoundbyaddingtogetherequationsofequalform degree,forexample,theequationofmotionfortheR–Rthree-formandtheBianchiidentityfor theR–Rone-form.TheresultingX-dependentequationsaregivenby
DmFm−XmFm−16HmnpFmnp=0, (3.11)
DpFpmn−XpFpmn−16HpqrFmnpqr+(dF1−X∧F1)mn=0, (3.12)
DrFrmnpq−XrFrmnpq+361εmnpqrstuvwHrstFuvw+(dF3−X∧F3+H3∧F1)mnpq
=0. (3.13)
Usingtheself-dualityofF5thelastequationcanbealsowrittenas
(dF5−X∧F5+H3∧F3)pqrlmn−16εpqrlmnvst u(DvFstu−XvFstu−FvHst u)=0.
(3.14)
Aswillbediscussedbelow,thesethreeequationsarealreadysufficientforderivingcandidates forthescaleinvarianceequationsfortheF-fields,whichare2ndorderinderivatives.
Itisusefultorewrite (3.1)–(3.3)inthenotationofforms(see Appendix Aforconventions). Todosoweintroducethedualformsdefinedby
F1=F9, F3= −F7, F5=F5, F7= −F3, F9=F1. (3.15)
Thenthecompletesetofthetype IIsupergravityequationsforR–RstrengthsandBianchi iden-tities (3.1)–(3.3)isgivenby19
18 Alternatively,onecanderivethisequationfromtheBianchiequation(3.5),theinvarianceofF
1undertheisometry,
theorthogonalityofI andZ,andtheconditionthatZisnotanexactone-form.Indeed,multiplying(3.5)byImone finds∂n(ImFm)−ZnImFm=0.Thus,ifImFm =0 thenZ=dln(ImFm).Wefind,however,itmoreconvenientto
addImFm=0 asanindependentequation,andinferfromittheorthogonalityofIandZ. 19 WeassumethatF
dF2n+1−Z∧F2n+1+H3∧F2n−1=0, n=0,1, . . . ,
d F2n+1−Z∧F2n+1−H3∧F2n+3=0, n=0,1, . . . , (3.16)
whereZ=dφ.
The“I-modified”equations (3.4)–(3.9)aregivenby20
dF2n+1−Z∧F2n+1+H3∧F2n−1−(I∧F2n+3)=0, n= −1,0, . . . ,
d F2n+1−Z∧F2n+1−H3∧F2n+3+(I∧F2n−1)=0, n=0,1, . . . .
(3.17)
Dueto (3.15)thetwoequationsin (3.16)areequivalentandthesameistruefor (3.17).
LetusnotethatthedeformedR–Requations (3.17)togetherwiththerelation (2.13)ordZ+ ιIH3=0 implythefollowingrelation
LIF2n+1=(I·Z)F2n+1. (3.18)
ThustheconditionthattheF-fieldsareinvariantundertheisometryI isequivalenttothe con-ditionI ·Z=0, whichis clearlysatisfiedfor the ABF backgroundas is evidentfrom (2.5), (2.7).
4. Second-order equations for R–R couplings as scale invariance conditions
LetusreturntothediscussionofthescaleinvarianceconditionsforthecouplingsoftheGS sigmamodel (1.2)insection2andconsidertheequationsfortheR–RcouplingsF thatshould followfromtherequirementof(1-loop)UVfinitenessofthe2dmodel.Onecanarguethatthe conditionsanalogoustoeqs.(2.1), (2.2)fortheGandB-fieldcouplingsshouldhavetheform
βkF
1...ks ≡ 1 2D
2F
k1...ks +. . .=X m∂
mFk1...ks +
i
Fk1...m...ks∂kiX
m, (4.1)
wherewehaveomittedpossiblenon-lineartermssuchas RF+DHF+. . . onthel.h.s.The
X-dependentLiederivativetermonther.h.s.reflects,asin (2.1), (2.2),thereparametrisation(or off-shell xm-renormalisation) freedom.Forexample, startingwiththe linearised RGequation
dFn(x)
dt =βnF=12∂ 2F
n(x),t=loganddoingthecoordinateredefinitionxm→xm+t Xm,one endsupwith dFndt(x)=12∂2Fn(x) −Xm∂mFn−Fm∂nXm.
Weshall discussthe computationof1-loop logarithmicUVdivergencesforthe GSaction (1.2)in Appendix GclarifyingthestructureofβF.
ForXm=∂mφtheequations (4.1)shouldbeaconsequenceofstrongerWeylinvariance con-ditions,21whichshouldbeequivalenttothetype IIsupergravityequations (3.1)–(3.3)or (3.16) whereZ=X=dφ.Indeed,combining(“squaring”)thefamiliardF+. . .=0,d F+. . .=0
20 Notethathereweincluden= −1 asinthedeformedtheoryitisnolongertrivial:itgivesthesecondequationin (3.4),i.e.(I∧F1)=ImFm=0.
21 Forexample,usingtheNSRapproachonaflatbackgroundwemayconsidertheR–Rvertexoperatorsbuiltoutof
equationsleadstod d F+d dF+. . .=0 orD2F+. . .=0,wheretheleadingtermisthe Hodge–deRhamoperator.
Moreover, the same equations should follow also from the modified type II equations (3.4)–(3.9)or (3.17)(as,e.g., theABF backgroundthat solvesthe modifiedequations should alsobeasolutionofthescaleinvarianceconditions).Thisshouldprovideanon-trivial consis-tencycheck:afterproperly“squaring” (3.4)–(3.9)thedependenceontheZandI vectorsinany candidatescaleinvarianceequationsshouldappearonlythroughtheirsumX=Z+I asin (2.1), (2.2).
StartingfromthemodifiedtypeIIequations (3.4)–(3.9)(whichincludethestandardtype IIB supergravityequationsasaspecialcase (3.10),Im=0),letusoutlinethederivationofthe2nd orderequationsfortheR–Rcouplingsthatshouldbeequivalenttothescaleinvarianceconditions for FnoftheGSsigmamodel (1.2).Tobeacandidateforthescaleinvarianceconditionsthese equationsshouldhavethefollowingproperties:
(i)vanishonthesupergravityequations (2.1), (2.2), (2.11), (3.1)–(3.3)withX=dφ,Y=0 (ii)dependonZandI throughX=Z+I
(iii)dependonXthroughLiederivatives.22
Startingwiththemodifiedequations (3.17)andactingwithdonthefirstequationandd
onthesecondandthenusingthemodifiedequations(asdescribedin Appendix D)wearriveat thefollowingequation,whichsatisfiestheaboveproperties
d d F2n+1+d dF2n+1+14R∧F2n+1−18 (H3∧H3)∧F2n+1
−H3∧(H3∧F2n+1)−(H3∧(H3∧F2n+1))
−d (H3∧F2n+3)−(H3∧dF2n+3)+d (H3∧F2n−1)+H3∧d F2n−1
=LXF2n+1+LXF2n+1−(d X)∧F2n+1+βB∧F2n−1−(βB∧F2n+3) .
(4.2)
HereβBisthe2-formanalogof (2.2),i.e.
βB≡12 d H3+K=(X∧H3)+dY . (4.3)
ThisisthenacandidateforthescaleinvarianceequationfortheR–Rform F2n+1.
Usingtheidentity
LXF2n+1=LXF2n+1+(d X)∧F2n+1+βG·F2n+1,
βG·F2n+1≡
i
βmG
inFm1...mi−1
n
mi+1...m2n+1, (4.4)
whereβmnG isdefinedin (2.1),wefindthat (4.2)becomes
d d F2n+1+d dF2n+1+14R∧F2n+1−18 (H3∧H3)∧F2n+1
−H3∧(H3∧F2n+1)−(H3∧(H3∧F2n+1))
−d (H3∧F2n+3)−(H3∧dF2n+3)+d (H3∧F2n−1)+H3∧d F2n−1
=2LXF2n+1+βG·F2n+1+βB∧F2n−1−(βB∧F2n+3) . (4.5)
22 Moreover,sincetheR–RfieldsFareinvariantundertheisometriesgeneratedbyI,theirLiederivativesalongI
ThedependenceoftheseequationsonXratherthanseparatelyonZandIcanberelatedtotheir closeconnectiontotheparticularX-dependentcombinationsofthemodifiedequationsin (3.11), (3.12), (3.13),i.e.to(heren ∈Zasin (3.17))
2n≡dF2n−1−X∧F2n−1+H3∧F2n−3
+(−1)n (dF9−2n−X∧F9−2n+H3∧F7−2n)=0. (4.6) Wealsodefineasin (1.11), (2.2)
¯
βB≡12 d H3+K−(X∧H3)−dX=0,
¯
βX≡R−12 (H3∧H3)+4 d X−4 (X∧X)=0. (4.7)
Deconstructingthederivationin Appendix D,wefindthatthe2nd-orderequationfortheR–R fluxes (4.2)canalsobewrittenas
d2n−X∧2n+H3∧2n−2−F2n−1∧ ¯βB
+(−1)n (d8−2n−X∧8−2n+H3∧6−2n−F7−2n∧ ¯βB
+1
4F2n+1∧ ¯β
X
=0. (4.8)
Finally,letuspresenttheexplicitformofeq.(4.5)incomponents.For F1wefind
D2Fm−RmnFn+14(R−34H2)Fm
+1 2H
pnkH
mpnFk−16DmHpnkFpnk−12HpnkDpFnkm
=2(XpDpFm+DmXpFp)+βmnGF
n−1
2β
B nkF
nk
m. (4.9)
Using the identity D[mHnpk] = 0 the term 61DmHpnkFpnk in (4.9) can be replaced by
1
2DpHmnkF
pnk.TheequationforF
3maybewrittenas
D2Fnkm−Ra[nFakm]+Rab[nkFabm]+
1 4(R−
3
4H
2)F
nkm
+1 2H
abcH
ab[nFkm]c−12HabcHa[nkFm]bc
+DaHa[nkFm]+Ha[nkDaFm]−FaDaHnkm
−1 6D[nH
abcF
km]abc−12HabcDaFbcnkm
=2(XaDaFnkm+D[nXaFkm]a)+βaG[nFakm]+β[BnkFm]−12βabBFabnkm, (4.10) whiletheequationfor F5canbeputintotheform
D2Fij klm−Ra[iFaj klm]+Rab[ijFabklm]+
1 4(R−
3 4H
2)F
ij klm
+1 2H
abcH
ab[iFj klm]c−12HabcHa[ijFklm]bc
+DaHa[ijFklm]+Ha[ijDaFklm]−Fa[ijDaHklm]
+ 1
12εij klmbdef(DaH
abcFdef+HabcD
aFdef−FabcDaHdef)=
=2(XaDaFij klm+D[iXaFj klm]a)+βaG[iF a
j klm]+β[BijFklm]
+ 1
12εij klmabcde(β
B)abFcde.
(4.11)
Thisexpressionisconsistentwiththeself-dualityof F5(inparticular,thethirdandforthlines
These2nd-orderequationsfor F1, F3and F5exhibitobviousstructuralsimilarities.In
par-ticular,theycontaintheexpectedHodge–deRhamoperatortermsandthevectorXonlyenters throughthereparametrisationterms as in(4.1).The βG andβB terms intheseequationsare definedasin (2.1), (2.2)butcanalsobereplacedbyexpressionsonther.h.s.of (2.1), (2.2).
Asweshall discussin Appendix G,similarequations comeout ofthe computationof the one-loopbeta-functionsfortheR–RcouplingsintheGSsigmamodel (1.2).
5. Origin of modified equations: T-duality relation to type II equations for backgrounds with non-isometric linear dilaton
Givenascaleinvariantsigmamodelinflat2dspaceT-dualityinanisometricdirectionshould alsoproduceascaleinvariantsigmamodel.Similarly,givenaWeylinvariantsigmamodelon curved2dspacewithallcouplingsincludingthedilatonbeingisometrictheT-dualbackground shouldalso beWeyl invariant(provided thedilatontransforms intheusualway [32,33]).As discussedintheintroduction,ingeneralthisisnotsoifthe dilatonisnotisometric:T-duality will still preserve scale invariance but not Weyl invariance. Thusgiven asolution of type II supergravityequations whichhas linearnon-isometric termin thedilatonits T-duality image will no longer solve the standardtype II equations but will satisfy instead amodified setof type IIequationsasdiscussedabove.
5.1. Simpleexamples
Hereweshallmaketheoriginofthemodifiedequationsexplicitbyshowingthatthey repre-senttheoriginaltype IIequationsforasolutionwithnon-isometriclineardilaton,rewrittenin termsofthefieldsoftheT-dualbackground.Toexplainhowthishappensinsimpletermsletus firststartwithabosonicbackground (1.8)withAμ=0,i.e.
ˆ
ds2=e−2a(x)dyˆ2+gμν(x)dxμdxν, φˆ= −cyˆ+ϕ(x)−12a(x) . (5.1)
ThenthecorrespondingWeylanomalycoefficients
¯
βmnG = ˆRmn+2DˆmDˆnφ ,ˆ β¯φ= ˆR+4Dˆ2φˆ−4Gˆmn∂mφ∂ˆ nφ ,ˆ (5.2)
havethefollowingcomponentsunderthexˆm=(y,ˆ xμ)splittingofcoordinates23 ¯
βμνG =Rμν−∂μa∂νa+2DμDνϕ , β¯yGˆyˆ=e−2a(DμDμa−2∂μa∂μϕ) , (5.3)
¯
βyμGˆ = −2c ∂μa , β¯φ=R−∂μa∂μa+4D2ϕ−4∂μϕ∂μϕ−4c2e2a. (5.4)
Weseethatifc=0,i.e.thedilatonisisometric,thentheWeylinvarianceconditionsβ¯μνG =0, ¯
βφ=0 areinvariantunderT-dualityiny,i.e.underaˆ= −a→a,φˆ→ ˆφ+aor ϕ→ϕ.The
c= −∂yˆφˆ dependenttermsin (5.4)thusrepresentobstructionstomappingoneWeylinvariant modeltoanother.TheT-dualmetricthensolvesweaker,modified,equations
Rmn+DmXn+DnXm=0, β¯X=R+4DmXm−4XmXm=0, (5.5)
withXmbeing(cf. (1.10))
23 HereR
Xμ=Zμ=∂μφ=∂μ(ϕ+12a) , Xy=Iy= −Gyy∂yˆφˆ=ce2a. (5.6)
Similar conclusions are reached in the casewe havea non-diagonalmetric in (1.8)(see the generaldiscussionbelow).Thepresenceofnon-zeroAˆμisinfactnecessarytohaveasolution oftheWeylinvarianceconditionswhenc =0 (cf. (5.4))andthetargetspaceshouldthusbeat least3-dimensional.Anexample ofsuchasolutionwasfoundin [13].Itrepresentsalimitof thebackgroundassociatedwiththeSO(4)/SO(3)gWZWmodel,whichhasthefollowingmetric anddilaton [34]
ˆ
ds2=dt2+tan
2t dp2+cot2t dq2
1−p2−q2
=dt2+cot2t (dθ+tanψcotθ dψ )2+tan 2t
sin2θdψ 2,
(5.7)
ˆ
φ= −lnb2−p2−q2sin 2t= −lnsinθ cosψ sin 2t, (5.8)
wherep=sinψ,q=cosθcosψandt,ψ,θareanglesofthecosetparametrisationoftheSO(4)
groupelement.Thisbackground(whichsolvestheWeylinvarianceconditionβ¯G=0 withβ¯φ=
const)has noisometries.Oneoption togeneratean isometryistosett =iz andthenshiftz
byaninfiniteconstant.Doingsowegetlineardilatoninz,butthezdirectiondecouplesinthe metric.Anon-trivialalternativeistosetψ=iyˆ andtoshiftyˆ byan infiniteconstant(which correspondstoinfiniterescalingofp,q generatingascalingisometryinthemetric (5.7)).The resultingbackground(wedropaninfiniteconstantinthedilaton)
ˆ
ds2=dt2−tan
2t dp2+cot2t dq2
p2+q2 =dt
2+cot2t (dθ+cotθ dy)ˆ 2−tan2t
sin2θdyˆ
2, (5.9)
ˆ
φ= −lnp2+q2sin 2t= − ˆy−lnsinθ sin 2t, (5.10)
isthereforeofthesametypeasin (1.8)anddefinesaconformalsigmamodel.24Similarhigher dimensionalbackgrounds canbe constructedstarting fromSO(n)/SO(n−1)gWZWmodels withn>4[13].
T-dualisingthemetric (5.9)alongyˆwegeta(G,B)backgroundthatwillsolvethemodified
(G,B)equations (2.1), (2.2)withnon-trivialXm=Im+Zm,whereIy= −∂yˆφˆandZmisgiven by (2.14),withφ= ˆφ−12logGyˆyˆ.ThesemodifiedequationswillbetheoriginalWeylinvariance conditionsrewrittenintermsofthedualGandB-fields.
Given the2d CFT in(5.9)with2d stress-tensordefinedtaking into accountthe dilatonin (5.10),onemayformallycompactifyyˆ andaskifthisCFT hasT-dualityasasymmetryofits spectrum.Theanswerappearstobenoasthe2dstress-tensorwillnotbeinvariantunderT-duality (i.e.mapping momentumintowindingmodes).25 Thisiscompatiblewithourexpectationthat formallyT-dualisingthemetric (5.9)willnotleadtoaconsistentCFT.
24 Thecentralchargeforthisd=3 conformalmodelisgivenbyc=d−3 2α(R−
1
12H2+4D2φ−4DmφDmφ)+. . .=
3−32α×12+. . ..Herethescaleofthespacewassettoone,sothatαisthentheinverseoftheWZWlevelk.Thisisin agreementwiththeusualcountofthecentralchargeforthe SO(4)/SO(3)gWZWmodelc=6k/(k+4)−3k/(k+2)=
3−18/k+. . .,whichshouldbeunchangedinthecoordinatelimitleadingfrom(5.7)to(5.9).
25 GivenafreecompactscalarCFTL=r2(∂φ)2withφ≡φ+2πthespectrumofdimensionsofprimaryoperators
ThesameconclusionsarereachedfortypeIIsolutionswithalineardilatonandnon-zeroR–R fluxes(withisometricG,Band Fn=eφFn),suchastheHTsolutiondualtoABFbackgroundin theAdS5×S5caseanditscounterpartsintheAdS2×S2×T6andAdS3×S3×T4casesdiscussed in Appendix F.Explicitly,inthecaseofasolution(G,ˆ B,ˆ Fˆn,φ)ˆ withseveralisometriesbroken onlybythelineardilatonterm
ˆ
φ=φ0−ciyˆi+f (xμ) , (5.11)
wewill getageneralisationof thetypeIIsupergravityequations,dependingonthefollowing twovectorsZandI (cf. (2.6), (2.7), (2.14))
I=ciGyiyidy
i, Z=dφ+ι
IB , φ=f−
1 2
i
logGˆyˆiyˆi . (5.12)
HereGandBarethebackgroundfieldsoftheT-dualbackground.
Belowweshallillustratetheserelationsinthegeneralcasewithoneisometry.
5.2. NS–NSsector
Weconsiderthefollowingtwod-dimensionalbackgrounds(hereweuseKμinsteadofAˆμin
(1.8))
ds2=e2a(dy+Aμdxμ)2+gμνdxμdxν,
B=Kν(dy+12Aμdxμ)∧dxν+12bμνdxμ∧dxν,
φ= −c y+ϕ+12a , Iy= ˆc , Iμ=0, (5.13)
ˆ
ds2=e−2a(dyˆ+Kμdxμ)2+gμνdxμdxν,
ˆ
B=Aν(dyˆ+12Kμdxμ)∧dxν+12bμνdxμ∧dxν ,
ˆ
φ= −ˆcyˆ+ϕ−12a , Iˆyˆ=c , Iˆμ=0. (5.14)
Here y and yˆ are the directions that are assumed tobe (shift) isometries of their respective metricsandB-fields.Weusetheindicesμ,ν,. . .=1,. . . ,d−1 andm,n,. . .=1,. . . ,d.For
c= ˆc=0 (5.13)and (5.14)arerelatedbystandardT-duality, suchthat ϕ istheanalogofthe duality-invariantdilatonfield.Letusalsodefine
Z=dφ+ιIB= −c dy+dϕ+12da+ ˆc Kμdxμ,
X=Z+I=(−c+ ˆc e2a)dy+dϕ+12da+(c Kˆ μ+ ˆc e2aAμ)dxμ,
ˆ
Z=dφˆ+ιIˆBˆ= −ˆc dyˆ+dϕ−12da+c Aμdxμ,
ˆ
X= ˆZ+ ˆI=(−ˆc+c e−2a)dyˆ+dϕ−12da+(c Aμ+c e−2aKμ)dxμ, (5.15)
where
Z·I= ˆZ· ˆI = −cc .ˆ (5.16)
restoredinthe“doubled”formulationifthelineardilatontermweregivenbyqφ+ ˜qφ˜whereφ˜isthedualfield(with
1 √
rq↔ r
√
Thetwo(G,B)backgroundsin (5.13)and (5.14)areT-dualtoeachotherandthusforc= ˆc=0 solve theequivalentWeylinvariance equations(see,e.g., [37]andthereferences therein).We willnowshowhowthisrelationalsoextendstothemoregeneralcasewithlineardilatons.
Letusfirstconsiderthegeneraliseddilatonequation
R−121H2+4DmXm−4XmXm=0. (5.17)
The questionwe wanttoaddress is:if thebackground (5.13) satisfies (5.17) does that imply that (5.14)satisfies (5.17).Asthetwobackgrounds (5.13)and (5.14)arerelatedbytheobvious symmetry
a→ −a , Aμ↔Kμ, c↔ ˆc , y↔ ˆy , (5.18)
itissufficienttocomputetheleft-handsideof (5.17)for (5.13)andcheckthatitisinvariant(or atleastcovariant)under (5.18).
For (5.13)wehave
Xy= −c+ ˆc e2a, Xμ=∂μϕ+12∂μa+ ˆc Kμ+ ˆc e2aAμ. (5.19)
Itwillalsobeusefultodefinethefollowingobjects
Fμν≡∂μAν−∂νAμ, Hμν≡∂μKν−∂νKμ,
hμνρ≡(db+12A∧dK+12K∧dA)μνρ , (5.20)
whereweobservethathisinvariantunder (5.18).Nowusingthedimensionalreductionformulae in Appendix Awefind
R−121H2+4DmXm−4XmXm
=R−∂μa∂μa−121h2−14e2aFμνFμν−14e−2aHμνHμν+4∇μ∂μϕ−4∂μϕ∂μϕ
+4c∇μAμ+4cˆ∇μKμ−8(c Aμ+ ˆc Kμ)∂μϕ
−4c2(AμAμ+e−2a)−4cˆ2(KμKμ+e2a)+8cc (ˆ 1−AμKμ) , (5.21) whichisindeedinvariantunder (5.18).Therefore,if (5.17)issatisfiedforbackground (5.13)itis satisfiedforbackground (5.14)andviceversa.26
LetusnowturntothemodifiedmetricandB-fieldequationstoshowthatthetwo combina-tionsappearingin (1.3)and (1.4)
Rmn−14HmpqHnpq+DmXn+DnXm, (5.22)
1
2D
pH
mnp−XpHmnp−DmXn+DnXm, (5.23)
are covariantunderthesymmetry (5.18).27 Theniftheyvanishforthebackground (5.13)this impliesthattheyvanishfor (5.14)andviceversa.As (5.22)issymmetricand (5.23)is antisym-metric,wemayjustconsidertheirdifference
26 Thisgeneralisestheusual(c= ˆc=0)discussionoftheT-dualityinvarianceofthestringeffectiveaction(1.7)with
√
G e−2φ= √g e−2ϕ.
27 HereweassumeYmin(1.4),(2.2)isequaltoXmin(1.3),(2.1)asisthecasefortheI-modifiedequations
sat-isfiedbytheABFbackground.Moregenerally,givenascaleinvariantsigmamodelwithanisometryandtheGand
B-fieldcouplingssatisfying(1.3),(1.4),itsT-dualcounterpartwillalsosatisfy(1.3),(1.4)withtherolesofXmandYm
Cmn≡Rmn−14HmpqHnpq−12DpHmnp+2DmXn+XpHmnp, (5.24)
whichwecandecomposeintoapartindependentofcandcˆ(C(mn0))andapartlinearincandcˆ (Cmn(1))as
Cmn=Cmn(0)+2Cmn(1), (5.25)
Cmn(0)=Rmn−41HmpqHnpq−12DpHmnp+2DmXn(0)+X(0)pHmnp, (5.26)
Cmn(1)=DmX(n1)+12X
(1)pH
mnp. (5.27)
Herewehaveusedthefactthatallthecand ˆcdependenceiscontainedinX=X(0)+X(1),where
X(0)isthec- and ˆc-independentandX(1)isthec- andcˆ-dependentpart.Usingthespecificform ofXforthebackground (5.13),asgivenin (5.19),wehave
Xy(0)=0, X(μ0)=∂μϕ+12∂μa , Xy(1)= −c+ ˆc e2a, Xμ(1)= ˆc Kμ+ ˆc e2aAμ. (5.28)
Usingtheformulaein Appendix AwefindthefollowingrelationsforCmn(0) andCmn(1)evaluated onthebackground (5.13)
Cyy(0)=2e2a∂μϕ∂μa−e2a∇μ∂μa+14e4aFμνFμν−14HμνHμν ,
Cyμ(0)−Gyμ Gyy
Cyy(0)=e2a(12∇ν+∂νa−∂νϕ)Fμν+14hμνρHνρ
+(12∇ν−∂νa−∂νϕ)Hμν+14e2ahμνρFνρ,
Cμy(0)−Gμy Gyy
Cyy(0)=e2a(12∇ν+∂νa−∂νϕ)Fμν+14hμνρHνρ
−(12∇ν−∂νa−∂νϕ)Hμν−14e2ahμνρFνρ ,
Cμν(0)−Gμy Gyy
Cyν(0)−Gyν Gyy
Cμy(0)+Gμy Gyy
Gyν
Gyy
Cyy(0)
=Rμν−∂μa∂νa+2∇μ∂νϕ−12e2aFμνFνρ−21e−2aHμρHνρ
−1 2∇
ρh
μνρ−14hμρσhνρσ+hμνρ∂ρϕ , (5.29)
Cyy(1)=e2a(cAμ+ ˆc Kμ)∂μa ,
Cyμ(1)−Gyμ Gyy
Cyy(1)= −12(e2aFμν+Hμν)(cAν+ ˆc Kν)+(c− ˆc e2a)∂μa ,
Cμy(1)−Gμy Gyy
Cyy(1)= −12(e2aFμν−Hμν)(cAνc Kˆ ν)+(c+ ˆc e2a)∂μa ,
Cμν(1)−Gμy Gyy
Cyν(1)−Gyν Gyy
Cμy(1)+Gμy Gyy
Gyν
Gyy
Cyy(1)
=1
2c(∇μAν+ ∇νAμ)+ 1 2c eˆ
2a(∇
μAν− ∇νAμ)
+1
2c(ˆ ∇μKν+ ∇νKμ)+ 1 2ce−
2a(∇
Thenusing themap (5.18)betweenthe backgrounds (5.13)and (5.14),we findthe following relationsforCmn
Cyy Gyy = −
ˆ
Cyˆyˆ
ˆ
Gyˆyˆ
, 1
Gyy Cyμ−
Gyμ GyyCyy
=1
ˆ
Gyˆyˆ ˆ
Cyμˆ −Gˆyμˆ
ˆ
Gyˆyˆ ˆ Cyˆyˆ,
1
Gyy
Cμy−GGμy yyCyy
= −1
ˆ
Gyˆyˆ ˆ
Cμyˆ−Gˆμyˆ
ˆ
Gyˆyˆ ˆ Cyˆyˆ
,
ˆ
Cμν−GGμy yyCˆyν−
Gyν GyyCˆμy+
Gμy Gyy
Gyν
GyyCˆyy= ˆCμν−
ˆ
Gμyˆ
ˆ
GyˆyˆCˆyνˆ −
ˆ
Gyνˆ
ˆ
GyˆyˆCˆμyˆ+
ˆ
Gμyˆ
ˆ
Gyˆyˆ
ˆ
Gyνˆ
ˆ
GyˆyˆCˆyˆyˆ, (5.31)
wherethe left-handside isevaluatedon (5.13)andthe right-handside on (5.14).From these equalities it follows that the vanishingof the tensors (5.22), (5.23)on thebackground (5.13) implies their vanishingalso on the background(5.14), andin thissense are covariantunder T-duality.
LetusbrieflycommentonthegeneralisationwhenIμ= ˆIμ =0 inthebackgrounds (5.13) and (5.14) (i.e.,when thereareextraisometries inxμ directions).Runningthroughthe same analysiswefindthattheresultstillholdsonlyifIμsatisfiescertainproperties.Inparticular,the T-dualityrelationbetweentheequationsfor (5.13)and (5.14)stillholdsif
IμAμ=IμKμ=0. (5.32)
ThisrequirementisalsosufficientfortheT-dualityofthemodifiedequationsofmotionforthe R–RfieldsdiscussedinthefollowingsectiontocontinuewhenIμ= ˆIμ =0.
OnecancheckthattheserelationsarevalidateachstageinthesequenceofT-dualities re-quiredtotransformfromthesupergravityHTsolutionsof [12]totheABFbackground (B.1)and itsAdS3×S3andAdS2×S2counterparts (F.4), (F.13).
5.3. R–Rsector
Letusnow considerthecaseofnon-zeroisometricR–Rfields Fn.Thecontributionofthe R–RfieldstothemetricandB-fieldequationsappearsintheusualunmodifiedformandhence wecanfocusourattentiononthemodifiedequationsofmotionfortheR–Rfields (3.17).Written in terms of the forms fk≡e−a/2Fk these take the followingform (dropping the distinction betweenyandyˆ)
Ek≡dfk−Z∧fk+H3∧fk−2− ˆc ιyfk+2=0, (5.33)
ˆ
Ek≡dfˆk− ˆZ∧ ˆfk+ ˆH3∧ ˆfk−2−c ιyfˆk+2=0, (5.34)
wherewehaveintroduced(K=Kμdxμ,A=Aμdxμ)
Z=Z−1
2da= −c dy+dϕ+ ˆc K , Zˆ
≡ ˆZ−1
2daˆ= −ˆc dy+dϕ+c A , (5.35) whicharerelatedtoeachotherunderT-dualityas
ˆ
Z=Z+c(dy+A)− ˆc(dy+K)≡Z+δZ . (5.36)
RecallthattheinvarianceoftheR–Rformsundertheisometryalongyrequires