Contents lists available atScienceDirect
Physics
Letters
B
www.elsevier.com/locate/physletb
Massive
mimetic
cosmology
Adam
R. Solomon
a,
∗
,
Valeri Vardanyan
b,
c,
Yashar Akrami
d,
baDepartmentofPhysics&McWilliamsCenterforCosmology,CarnegieMellonUniversity,Pittsburgh,PA 15213,USA bLorentzInstituteforTheoreticalPhysics,LeidenUniversity,P.O.Box9506,2300RALeiden,theNetherlands cLeidenObservatory,LeidenUniversity,P.O.Box9513,2300RALeiden,theNetherlands
dDépartementdePhysique,ÉcoleNormaleSupérieure,PSLResearchUniversity,CNRS,24rueLhomond,75005Paris,France
a
r
t
i
c
l
e
i
n
f
o
a
b
s
t
r
a
c
t
Articlehistory: Received14March2019
Receivedinrevisedform24May2019 Accepted28May2019
Availableonline30May2019 Editor:H.Peiris
Westudythefirstcosmologicalimplicationsofthemimetictheoryofmassivegravityrecentlyproposed byChamseddineandMukhanov.Thisisanoveltheoryofghost-freemassivegravitywhichadditionally containsamimeticdarkmattercomponent.Inanechoofothermodifiedgravitytheories,thereare self-acceleratingsolutionswhichcontainaghostinstability.Intheghost-freeregionofparameterspace,the effectofthegravitonmassonthecosmicexpansionhistoryamountstoaneffectivenegativecosmological constant,aradiationcomponent,andanegativecurvatureterm.Thisallowsustoplaceconstraintsonthe model parameters—thegravitonmassandthe Stückelbergvacuumexpectationvalue—byinsisting that theeffective radiationandcurvaturetermsbewithinobservationalbounds.Thelate-timeacceleration must be accountedfor by aseparatepositive cosmological constantorother darkenergy sector. We imposefurtherconstraintsatthelevelofperturbationsbydemandinglinearstability.Wecommenton the possibilityof distinguishingthistheory fromCDM withcurrentand future large-scale structure surveys.
©2019TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense (http://creativecommons.org/licenses/by/4.0/).FundedbySCOAP3.
1. Introduction
Chamseddine and Mukhanov have recently proposed [1,2] a novel ghost-free theory of massive gravity in which one of the fourStückelbergscalars isconstrainedinthesamewayasinthe mimetictheoryofdarkmatter [3],spontaneouslybreakingLorentz invariance.InthisLetter,we studytheimmediateimplicationsof thismimeticmassivegravityforcosmologicaltheoryand observa-tion.
From a field-theoretic perspective, general relativity is the uniquetheory(in fourspacetimedimensions)ofamassless spin-2particle,orgraviton. Itis thereforenaturaltoask whetheritis possibleto endow thegraviton witha non-zero mass, andwhat sort of theoretical structures would result [4]. A closely related lineofinquiryaskswhetheritispossiblefortwo ormore gravi-tonsto interact[5]. Most nonlinear realizations of such theories sufferfromtheso-calledBoulware-Deserghostinstability[6].The pastdecadehasseentheconstructionofmodelswhichavoidthis instability, allowing forthe construction ofghost-free theoriesof
*
Correspondingauthor.E-mailaddresses:adamsolo@andrew.cmu.edu(A.R. Solomon), vardanyan@lorentz.leidenuniv.nl(V. Vardanyan),akrami@ens.fr(Y. Akrami).
massive gravity [7–13] andbimetric andmultimetric gravity [12,
14,15].Wereferthereadertothereviews[16,17] onmassive grav-ityand[18,19] onbimetricgravity.Thetheoryofmimeticmassive gravityproposedinRefs. [1,2] takesanewandalternativepathto aghost-freenonlineartheoryofmassivegravity.
A generic theory ofmassive gravity propagatessix degreesof freedom,whichshould bethoughtofasthefivehelicitystatesof a massive graviton plus an additional, ghostly scalar. The easiest waytounderstandthedegrees-of-freedom countingisto observe that a graviton mass breaks diffeomorphism invariance.This isa gauge symmetry andso can be restored by the addition of four Stückelberg scalars
A, which propagate in addition to the two (nowpotentiallymassive)tensormodesofgeneralrelativity.
As an illustration, considera Lorentz-invariant theory of mas-sivegravity.Inordertoconstructnon-trivial, non-derivative inter-actions for the metric, one requires a second “reference” metric. The simplestchoiceforthis metricisthat offlat space,
η
μν , buttheadditionofthispriorgeometrybreaksdiffeomorphism invari-ance;forinstance,therearepreferredcoordinatesystemsinwhich
η
μν=
diag(−
1,1,1,1). But diffeomorphism invariance is simplya redundancy in description, and can be restored by the addi-tion ofredundantvariables, i.e., replacing
η
μν→
η
A B∂
μA
∂
νB,
where
η
A B=
diag(−
1,1,1,1)andthefourfieldsA transformas https://doi.org/10.1016/j.physletb.2019.05.045
spacetimescalars.Onecanalways,bymeansofadiffeomorphism, choosethe unitary gauge inwhich
A
=
xA,andwe recoverthe originaldescriptionofthetheoryintermsofasymmetry-breaking referencemetric. Generic interaction terms for the graviton, e.g., genericfunctionsof gμαη
A B∂
αA
∂
νB,willleadtodynamicsfor
eachofthesefourscalars,inadditiontothetwomodesofgeneral relativity,foratotalofsixdegreesoffreedom.
At thelinear level,i.e., linearizing themetric aboutflat space inunitarygauge, gμν
=
η
μν+
hμν andA
=
xA,wefindthatoneof the sixdegrees of freedom leads to a ghost instability unless we specifically arrange the mass term into the Fierz-Pauli form,
L
mass∼
h2μν−
h2,inwhichcasethedynamicsoftheghostlymodetake the form of a total derivative. Continuing thisprocedure at higher orders in perturbation theory—i.e., continually packaging ghostlyoperatorsintototalderivativestructures—leadsuniquelyto thenon-linearmassivegravitytheoryofdeRham,Gabadadze,and Tolley(dRGT)[8,9].
The recent proposal of Chamseddine and Mukhanov takes a novel alternative approach to eliminating the dangerous ghostly mode [1,2]. Noticing that the ghost can be associated to the
0 Stückelberg mode, they propose imposing the constraint gμν
∂
μ0
∂
ν0
= −
1. This is motivatedby a similar constructionknown as mimetic gravity [3], in which the constrained scalar winds up behaving like dark matter.1 Mimetic massive gravity takesthisconstrainedscalartobeoneoftheStückelbergmodesof amassivegraviton,eliminatingtheghost.Theyproposethe follow-ing action,designedto ensurestability at thelinear level(notice thatthemasstermisnotoftheFierz-Pauliform),
S
=
d4x√
−
g M2 Pl 2 R+
m2M2 Pl 8 1 2h¯
2− ¯
h2 A B+ λ(
X+
1)
+
Smatter,
(1) withX≡
gμν∂
μ0
∂
ν0,and
¯
hA B≡
gμν∂
μA
∂
νB
−
η
A B.
(2)Internal indices (given by capital Roman letters) are raised and loweredwiththeMinkowskimetric.Thefieldequationsare2
Gμν
=
1 M2 Pl Tμν−
2λ
M2 Pl∂
μ0
∂
ν0
+
m2 2¯
hA B−
1 2h¯
η
A B∂
μA
∂
νB
−
1 4h¯
A Bg μν,
(3) 0= ∇
μ 2λ
M2Pl∂
μ0
δ
0 A−
m2 2¯
hA B−
1 2h¯
η
A B∂
μB
,
(4) X= −
1.
(5)The last of thesealigns
˙
0 withthe lapse of gμν . An upshot ofthisconstructionisthattheconstrainedmode behavesasa pres-sureless fluid, i.e., this theory provides a (mimetic) dark matter candidate[1,2].3
1 Foranearlierconstructioninwhichaconstrainedscalarmimicsdarkmatter
anddarkenergy,andwhichcontainsmimeticdarkmatterasasubset,seeRef. [20].
2 Notethesigndifferencesbetweentheright-handsideoftheEinsteinequations
andthe correspondingequationinRef. [1],whichisdue tothe mostlypositive metricconventionweemploy.
3 Oneshouldnotethatthephenomenologyofmimeticdarkmatterisstillinthe
earlystagesofdevelopmentcomparedtotraditionalparticle darkmattermodels suchasweaklyinteractingmassiveparticles(WIMPs)oraxions,anditispremature toconsidermimetic gravity asaserious alternativetothosemodels. For
exam-We endthissection bymakinga connectionwiththeexisting literature on Lorentz-violatingmassivegravity anddemonstrating the absence of certain well-known features of Lorentz-invariant massive gravity, namely the van Dam-Veltman-Zakharov (vDVZ) discontinuity [22,23] andthe Higuchi bound [24]. The vDVZ dis-continuityreferstothefailureoflinearizedLorentz-invariant mas-sive gravity to reduce to general relativityin the massless limit; thisrequiresnonlineareffectsinordertorestoregeneralrelativity in the Newtonian limit [25,26]. The Higuchi bound is a stabil-itybound formassive gravityon de Sitterspace, placing a lower bound on thegraviton mass,m2
≥
2H2,with H the Hubble rate. Itiswell-knownthat breakingLorentzinvariancechangesbothof theseconclusionsdramatically[27,28].At theleveloflinearperturbations aroundflatspace, the gen-eralSO(3)-invariantmassterminunitarygauge(A
=
xA) canbe writtenas[27]L
mass=
1 8M 2 Plm20h200
+
2m21h20i−
m22h2i j+
m23hii2−
2m24h00hii.
(6)Thelinearizedmasstermineq. (1) inunitarygaugeis(treating
λ
asfirst-order)L
mass=
m2MPl2 8−
1 2h 2 00+
2h20i−
h2i j+
1 2h 2 ii−
h00hii+ λ
h00.
(7)The
λ
equation ofmotion setsh00=
0, whichwe can impose in theaction4tofindm20
=
m24=
0,
m12=
m22=
2m23=
1.
(8)This allows us to easily make contact with the existing litera-tureonLorentz-violatingmassivegravity.TheanalysisofRef. [27] shows that for these mi parameters, the Newtonian limit is the usualone,whilethevDVZdiscontinuityisabsent.Theanalogueof theHiguchiboundinLorentz-violatingmassivegravitywasderived in Ref. [28], andfor ourvalues of the mi parameters, it reduces simplytoH2
>
0,whichistriviallysatisfied.2. Flat-spaceperturbations
In this section, we briefly review the behavior of perturba-tions aboutflatspaceinmimeticmassivegravity, asdiscussedin Refs. [1,2].Thiswillplacestabilityconditionsonthetheorywhich willberelevantwhenwemovetocosmologicalsolutions.
Theequationsofmotion(3)–(5) invacuumaresolvedby5
gμν
=
η
μν,
A
=
xA,
λ
=
0.
(9)ple,sincethemimeticdarkmatteronlyinteractsgravitationallywiththeStandard Model,wedonotexpecttohaveathermalproductionmechanism,incontrastto manytraditionaldarkmatterscenariossuchasWIMPs.Indeed,whenthetheoryis shift-symmetricin0,theenergydensityofthiscomponentissetentirelybyan
integrationconstantandsoisdeterminedbyinitialconditions.Itmayalsobe nec-essarytotunetheparametersofthemodelinordertoobtaintherightvaluesof thedarkmatterdensityovertheentirecosmichistory,andhigher-derivative effec-tivefieldtheorycorrectionsplayanimportantrole[21].Wereferthereaderto,e.g., Ref. [21] fordiscussionsoftheconstraintsthatearly-Universeconsiderationsplace onthepropertiesandevolutionofmimeticdarkmatterthroughoutcosmichistory.
4 Thisisjustifiedbecause,onshell,theh
00equationofmotionsimplysetsthe
valueofλ,whileh00 dropsoutofthehi j equationsofmotion.Thedynamicsare
thereforeequivalent.
5 Thisistheonlysolutionthatismanifestlyinvariantunderrotations,i.e.,with gμν=diag(−1,1,1,1)andA=ϕ(t), βxi.Apriori itmaybepossibletohave
flatsolutionswithinhomogeneousStückelbergsA,orequivalentlysolutionswith A=xAandg
Weexpandtheaction(1) toquadraticorderaroundtheMinkowski solution,focusingonscalarmodes,
g00
= −(
1+
2φ),
(10) g0i= ∂
iB,
(11) gi j= (
1−
2ψ )δ
i j+
2∂
i∂
jE,
(12)A
=
xA+
π
0, ∂
iπ
,
(13)λ
= δλ.
(14)Threeof thesefields—φ, B,and
δλ—are
auxiliary,asthey appear withouttimederivativesintheaction,andsocanbeintegratedout usingtheirequationsofmotion.Notethattheauxiliarystructureis preciselythesameasingeneralrelativity,sincethemasstermand Lagrangemultiplierdonotintroduceanyderivativesofthemetric. Wecanusediffeomorphisminvariancetoremoveafurthertwo modes.Whengaugefixingattheleveloftheaction,onemusttake care to only eliminate variables whose equations of motion are containedin theequations ofmotion ofthe remaining variables, otherwisewe willloseinformationafterpicking agauge. Follow-ingtheprocedureofRefs. [29,30],weseethat wecansafelytakeπ
0andoneof(
E,
π
)
tovanish.Pickingunitarygauge,π
0=
π
=
0,weobtaintheflat-spacequadraticaction(inFourierspace),
δ
2S=
dt M2Pl
− ˙
X
TK ˙
X
+
X
Tk2
G
+
m2M
X
,
(15)where
X ≡ (ψ,
k2E
)
andthematricesK
,G
,andM
aregivenbyK
=
3+
4k2 m2 1 1 0,
(16)G
=
1 0 0 0,
(17)M
=
1 4 3 1 1−
1.
(18)As described in Ref. [2], this system can be diagonalized by replacing
ψ
withtheLagrange multiplierδλ,
which we had pre-viouslyintegratedoutusingδλ
=
M 2 Pl 4(
4k2+
3m2)ψ
+
k2m2E,
(19) tofindδ
2S=
dt 1 4k2+
3m2 k4m2M2Pl˙
E2− (
k2+
m2)
E2−
1 MPl2 16 m2δλ
˙
2−
4δλ
2.
(20)Ifwetakem2
>
0,wecancanonicallynormalize,δλ
c≡
4 mMPl 2k2+
3 2m2δλ,
(21) Ec≡
mMPlk2 2k2+
3 2m2 E,
(22)toobtainthefinalaction,
δ
2S=
dt 1 2E˙
2 c−
1 2(
k 2+
m2)
E2 c−
1 2δλ
˙
2 c+
1 8m 2δλ
2 c.
(23)The only dynamical degree of freedom here is Ec, which is healthyandhasmassm.Thefield
δλ
c hasthewrongsignonbothitskineticandmassterms,butdoesnotpropagateduetothe ab-senceofagradientterm;itsequationofmotion,
¨
δλ
c+
m24
δλ
c=
0,
(24)leads toa dispersionrelation
ω
2=
m2/4 and
issolved simplyby[2]
δλ
c=
C(
x
)
sin mt 2+
D(
x
)
cos mt 2,
(25)whereC andD are space-dependentconstantsofintegration.The authorsofRef. [2] identifythismodewiththemimeticdark mat-ter.6
When we discusscosmology inthenext section, we will find ourselvestemptedbythepossibilityoftakingm2
<
0.Apriori thisis merely a parameter choice, but the flat-space analysis shows whythiswouldbe apoordecision.Bylookingattheaction(20), we see that, fornegative m2,the overall signin front ofthe ac-tion flips depending onwhetherk2
>
3|
m2|/
4 or k2<
3|
m2|/
4,a signofpathologicalbehavior.Inparticular,forscalesk2>
3|
m2|/
4,uponcanonicallynormalizingwefindtheaction(23) withan over-allminussign,sothatthedynamicalmode Ec isaghost.
3. Cosmologicalsolutions
In this section we investigate Friedmann-Lemaître-Robertson-Walker(FLRW) cosmologicalsolutionsofmimeticmassivegravity. Considerthehomogeneousandisotropicansatz
gμν
=
diag(
−
1,
a(
t)
2δ
i j),
(26)A
=
ϕ
(
t), β
xi.
(27)Inprincipleonecouldallow
β
todependontime,butthisbreaks homogeneityandisotropyasitinducesx-dependenttermsinthe stress-energy tensorof the Stückelberg fields.Note that on-shell, theLagrangemultiplierenforces
ϕ
=
t (uptoaconstant). Wewill include a general matter sector with densityρ
and pressure p.Wewillfindthissector needstocontainacosmologicalconstant, muchlikeingeneralrelativity,butdoesnot needtoincludedark matter, as this role can be played by the mimetic dark matter (whichisanexactlypressurelessperfectfluid).
TheEinsteinandscalarequationsofmotionare
3H2
=
ρ
M2Pl−
2λ
M2Pl−
3m2 16β
4 a4−
6β
2 a2+
5,
(28) 2H˙
+
3H2= −
p M2 Pl−
m2 16 3−
β
4 a4−
2β
2 a2,
(29) 0=
d dt a3 3m2 4 1−
β
2 a2+
2λ
M2 Pl.
(30)Wecansolvefor
λ
byintegratingthe0 equationofmotion(30), finding
6 SeeRef. [2] foranargumentforwhythismodeisnotaghost,despitehaving
anoverallwrong-signaction.Inprinciple,onemightworrythatwhenquantizingor consideringnonlinearities,acouplingwillbeinducedbetweenδλcandotherfields
whichwillleadtoanOstrogradskiinstability.Ontheotherhand,duetothelack ofagradienttermthismodeisnotapropagatingdegreeoffreedomintheusual sense.Wewillremainagnosticaboutthisquestionandlimitourselvesto consid-erationsofclassical,linearstability,whichthissystemclearlysatisfiesform2>0.
−
2λ
M2Pl=
C
a3+
3m2 4 1−
β
2 a2,
(31)where
C
is an integration constant. Pluggingthis into the Fried-mannequation(28),weobtain3H2
=
ρ
M2 Pl+
C
a3−
3m2 16 1−
β
2 a2 2.
(32)Notethatthecontributionfrom
λ
exactlycancelsoutthatfromthe lasttermoftheEinsteinequation(3),sotheverysimpleformforρ
ϕ≡ −
3m2MPl2(1
− β
2
/
a2)
2/16 is
entirelyduetothetermpropor-tionaltogμν inthestresstensor.Theintegrationconstantprovides adust-likecontributiontotheFriedmannequation,whichistobe expectedasthisisatheoryofmimeticdarkmatter.
We can geta better sense of the physicalpicture by expand-ing out the Friedmannequation andabsorbing the mimetic dark matter
C
intoρ
,finding3H2
=
ρ
M2 Pl−
3m2 16β
4 a4−
2β
2 a2+
1.
(33)Form2
>
0 (m2<
0),we seethatthe masstermgenerates an ef-fectivenegative (positive)cosmologicalconstant,an effective neg-ative (positive) curvature, and an effective radiation component withnegative (positive)energydensity.Notethat theseaddonto anycosmologicalconstant,radiation,andcurvaturealreadypresent cosmologically;forexample,whilewehaveassumedaflat cosmol-ogyasouransatz, observational boundsonspatial curvaturewill constrainthesumofanypre-existingcurvatureandthe curvature-liketermgeneratedbythegravitonmass.Note that for m2
<
0 we have late-time acceleration, witheff
=
3|
m2|/
16.However,asdiscussedintheprevioussection,weneedm2
>
0 inorder to avoida ghost around flat space. Thisis reminiscentofthesituationintheDvali-Gabadadze-Porrati(DGP) model [33], where one branch of solutions has self-accelerating cosmologicalexpansion[34,35] butisplaguedby aghost[36,37], whilethe other branch is healthybut cannot account forcosmic acceleration.Letus assume that the energydensity
ρ
ineq. (33) contains dust(includingthe mimeticdarkmatter), radiation,anddark en-ergycomponents.Then,intermsofthedensityparameters,i,0
=
ρ
i,03MPl2H02
,
(34)thecomponentsoftheFriedmannequationwhicharemodifiedby mimeticmassivegravityare
,0
= ¯
,0−
m2 16H20 (35)K,0
=
m2 8H20β
2,
(36)r,0
= ¯
r,0−
m2 16H20β
4,
(37)where
¯
,0 and¯
r,0 arethe densities associatedto darkenergyandStandardModelradiation.Usingobservationalboundsonthe curvatureandradiationdensities,wecanplaceconstraintsonthe model parameters m2 and
β.
We will not consider any boundscomingfrom thepresence ofthe effectivecosmologicalconstant, even though it contributes a negative and potentially large (if
m2
H0) amount to
,0.Particle physics also predicts a large
(and potentially negative) vacuum energy, and since we are not worrying about that, it seems inconsistent to worry about the
contributionfrommimeticmassivegravity.Onemightexpectthat whateversolvestheformerproblemwillalsosolvethelatter.7
We will use observational constraints on
K,0 and
r,0 to
boundourtwofreeparameters,m2 and
β.
Planck 2018constrainsK,0
=
0.0007±
0.0019, which we parametrize as|
K,0|
< δ
K, withδ
K∼
0.003 [39].Wewilltake thistobea constraintonthe contributionfrommimeticmassivegravityalone,m2
8H02
β
2
< δ
K
.
(38)We remind the reader that what we are really bounding is the sum ofthe mimetic massive gravity contribution andany“bare” curvature,butunlessthereissignificanttuningbetweenthesetwo, we can simply take thisas a constrainton the mimetic massive gravitypiecealone.
Toboundthemimeticcontributiontotheradiationdensity,we will use constraintsfrom big bangnucleosynthesis (BBN). At the time of BBN,radiation dominates.The exact value ofthe Hubble rate at the time of nucleosynthesis, which dependson the radi-ation density, determines the freeze-out abundance of neutrons andthereforetheprimordialabundanceofhelium-4,whichis sub-jecttotightobservationalbounds.Theconstraintsareconveniently phrasedintermsofthe“speed-upfactor”
ζ
≡
H/ ¯
H ,where H and¯
H are theHubblerateandits expectedvalue, respectively,atthe time of BBN.The difference betweenthe observedandpredicted helium-4 abundance,
|
YP|
, isrelated to the speed-upfactor by [40]YP
=
0.
08(ζ
2−
1).
(39) Currentobservationalboundsimply[41]|
YP|
0.
01.
(40)Comparing the Friedmann equation (33) with and without the mimetic radiation contribution, and focusing on radiation domi-nation,wefind
ζ
2−
1= −
m2
β
416
¯
r,0H20,
(41)where the value for the present-day radiationdensity associated to photonsandneutrinos,
¯
r,0∼
10−4,is determinedentirely bytheCMBtemperatureandtheeffectivenumberofneutrinospecies and is therefore not dependent on our modification of gravity.8 Combiningthiswitheq. (40) wearriveattheconstraint
m2 16H20
β
4< δ
r,
(42) whereδ
r≡
max(
|Y
P|) ¯
r,0 0.
08≈
O
(
10 −5).
(43)Wecanrewriteourconstraints(38) and(42) asinequalitiesfor
m
/
H0 andβ
aloneintwodifferentrégimes,m H0
<
⎧
⎪
⎨
⎪
⎩
√ 8δK β, β <
2δr δK 4√δr β2,
β >
2δr δK.
(44)TheseareplottedinFig.1.
7 SeeRef. [38] foraproposedsolutiontothecosmologicalconstantproblemin
thecontextofLorentz-violatingmassivegravity,whichiscloselyrelatedtomimetic massivegravity.
Fig. 1. Upper limits on m/H0andβfor(δK, δr)= (0.003,10−5). Finally,wenotethatthestrong-couplingscaleforthistheoryis oforder
2
=
√
mMPl[2].Ifm isoforderthepresent-dayHubblescale,m
∼
10−33eV,thenthestrongcouplingscaleis2
∼
meV,i.e., the theory breaks down slightly below the millimeter scale. Aswe seefrom eq. (44), forsufficiently small
β
,m could poten-tiallybemuchlargerthan H0,leadingtoacorrespondinglylarger strong-couplingscale.4. Cosmologicalperturbations
As we have seen, atthe background level, cosmological solu-tions in mimetic massive gravity do not differ appreciably from
CDM.
Wethereforeproceedtostudycosmologicalperturbations aroundthe FLRWbackground.This willtell ushowcosmological large-scalestructure(LSS)evolvesinthistheory incomparisontoCDM.
Since mimetic massive gravity differs from general rela-tivity,we wouldexpectmodifications tothe gravitationalPoisson equationandthe sliprelation,whichcould inprinciple allow for observational tests of this alternative model againstCDM
and distinguishthetwousingthecurrentandfutureLSSsurveys. How-ever,aswewillsee,stabilityofcosmologicalperturbationsandthe bounds(44) placestrongconstraintsonthe modelwhichsuggest that thistheory should be observationally indistinguishablefrom GRinthelinearrégime.4.1.Stabilitybound
We begin by studying the stability of cosmological pertur-bations using the second-order action formalism. Since, as dis-cussedin section 3,this theory doesnot possessghost-free self-accelerating solutions, we include a cosmological constant, al-thoughitwillnotaffectanyoftheresultsinthissection.Sincethe theory alreadycontains a pressurelessfluid, namely themimetic darkmatter,weneednotintroduceanadditionalmatterfield.Our analysisisthereforevalidforalltimesaftermatter-radiation equal-ity.
We define the linearized metric, Stückelberg fields, and La-grangemultiplieras ds2
= −(
1+
2φ)
dt2+
2a∂
iBdtdxi+
a2(
1−
2ψ )δ
i j+
2∂
i∂
jE dxidxj,
(45)0
=
t+
π
0,
(46)i
= β
xi
+ ∂
iπ
,
(47)λ
= ¯λ + δλ,
(48)wherewe are restrictingourselves toscalar perturbations, and
¯λ
isthe background value given ineq. (31). The calculation of thequadratic action proceeds analogously to the flat-space case dis-cussedinsection2.Expandingtheaction(1) (withacosmological constant)toquadraticorderinperturbations,wefindthatthe vari-ables
φ,
B,andδλ
areauxiliary—thatis,theyappearwithouttime derivatives—andcanthereforebe integratedout usingtheir equa-tions ofmotion. To safelyfix a gauge at the level of the action, weagainfollowtheprocedureofRefs. [29,30],findingthatwecan eliminateoneeachof(ψ,π
0)and(E,π
).Wewillchoosetoworkinunitarygauge,
π
0=
π
=
0,sothatA
= (
t,
β
xi)
isunperturbed. Thefinalaction,inFourierspaceandafterintegrationsbyparts,isδ
2S=
dt MPl2a3− ˙
X
TK ˙
X
+
X
T k2 a2G
+
m 2M
X
,
(49)where
X ≡ (ψ,
k2E
)
andthematricesK
,G
,andM
aregivenbyK
=
3−
8a2 β2−3a2 k 2 m2β2 1 1 0 (50)G
=
1 0 0 0 (51)M
=
1 8β
2 a2 1+
β
2 a23 1 1
−
1 (52)Sinceweareinterestedintheimplicationsofmimeticmassive gravityforthegrowthandpropertiesoflarge-scalestructureinthe late Universe, let usfocus on subhorizon scales (i.e., k2
a2H2) and assume the quasi-static (QS)approximation. In order to use thisapproximation,wefirstneedtoensurethatfluctuationsinthis régime are stable. Ignoring time variation in a(
t),
which will be subdominant in the limit k2a2H2, and assuming solutions of
theform
X =
X
0eiωt,theequationsofmotionfollowingfromthe action(49) are−
ω
2K
+
k 2 a2G
+
m 2M
X
=
0.
(53)We canthen derive stability conditionsfromthe dispersion rela-tions,obtainedbysolving
0
=
det−
ω
2K
+
k 2 a2G
+
m 2M
=
ω
4 a2+ β
2−
ω
2k2 a2(
3a2− β
2)
−
5ω
2m2β
2 8a4+
k2m2β
2 8a6+
m4β
4(
a2+ β
2)
16a8 (54) forω
2.The dispersionrelations arisingfromeq. (54) arecomplicated, butsimplifysignificantly in thelimit k
aH whenwe take into account theconstraints(44) onm/
H0,whichwe obtainedby re-quiringthattheradiationandcurvaturedensitiesgeneratedbythe masstermnotexceedobservationalbounds.Consider replacingmand
β
ineq. (54) withthefollowingtwoparameters,91
≡
mβ
k 2,
2
≡
mβ
2 ka 2.
(55)Weproceedtoshowthatthebounds(44) implythateachofthese ismuchsmallerthanunityonsubhorizonscalesforalltimesafter matter-radiationequality.
9 Todothisreplacement, firstreplacem→√
1β/k,and thenreplaceany
Forboth
1 and
2 wecan putupperboundson the
numera-torsandlower boundsonthedenominators.Letusstartwiththe numerators. For
1, multiply each side of eq. (44) by
β
.We seethereisastrictupperboundonthecombinationm
β
,m
β
≤
8δ
KH0≈
0.
15H0 (56)wherewehavetaken
δ
K∼
0.003 asarepresentativevalue.Wecan similarlyfindaboundonthenumeratorof2 bymultiplyingboth
sidesofeq. (44) by
β
2,findingm
β
2≤
4δ
rH0≈
10−2H0 (57)for
δ
r∼
10−5.Now we move on to the denominators. The subhorizon limit isgivenbyk
aH .Forthesake ofargumentlet usbe conserva-tiveandassumethatk isonlyslightlysubhorizon,k/
a≈
O(1)
H .10At any giventime frommatter-radiation equality to the present, wherewe can trustour analysis, theHubblerate H is relatedto itspresent-dayvalue H0 byH
=
H0,0
+
m,0a−3.Puttingthistogether with the boundswe have derived on m
β
andmβ
2, we find1 0
.
02,0a2
+
m,0a−11
,
(58)2 10−4
,0a4
+
m,0a1 for z3000
.
(59)Note that while the upper bound on
1 is always much smaller
thanunityfor0
<
a≤
1,the upperboundon2 infactgrows as a−1 atearly times.However, itgrows slowly and hasa factorof 10−4tocompetewith,sothat max(
2
)
doesnotreachunityuntil z∼
3000,rightaroundmatter-radiationequality.Thereforein prin-cipletheremightbeahandfulofmodes—rightaroundthehorizon scaleandattheearliestmomentsofmatterdomination—forwhich termsgoingas2 affectthesubhorizondispersionrelation,if m
β
2takes the largest value allowed by the constraints. We will con-tinuetotake
2
1,withtheunderstandingthatifthisparticular
situation is realized,then atthose very early times we are only consideringmodeswithk
10aH ,forwhich2iscertainlysmaller
thanunity.
Droppingtermssubdominantin
1 and
2,thedispersion
rela-tion(54) becomes 0
≈
ω
4 a2+ β
2−
ω
2k2 a2(
3a2− β
2)
+
k2m2β
2 8a6.
(60)Solving for
ω
2, and again dropping terms subleading in1
=
(
mβ/
k)
2,wefindthedispersionrelationsforourtwomodes,ω
2≈
k 2 a2 a2+ β
2 3a2− β
2,
(61)ω
2≈
m2β
2 8a2 3−
β
2 a2.
(62)Eachoftheseimpliesthesamestabilitycondition,
β
2a2
<
3.
(63)Thistellsusthatnomatterwhatthevalueof
β
is,our cosmologi-calsolutionsareunstable atsufficientlyhighredshifts,10 Ofcourse,the deeperinthesubhorizon régimek is,the smaller 1 and2
become.
z
>
√
3β
−1−
1.
(64)This early time instability can howeverbe safely pushed back to unobservably early times by taking the parameter
β
to be suffi-ciently small.11 Becauseweareassuming matteranddarkenergy domination, we can trust our stability condition as far back as matter-radiation equalityat zeq≈
3400.Demandingstabilityfromzeq onward,wefindaconstrainton
β
,12β
5×
10−4.
(65)4.2. Cosmologicaltensormass
Another possible cosmological bound on the parameters m
and
β
comes from constraints on the graviton mass. The tight-estboundscurrentlycomefromLIGO,mT≤
7.7×
10−23eV[44].13To compute the mass of tensor fluctuations on a cosmological background,we linearizethe Einsteinequation (3) around gμν
=
¯
gμν
+
hμν ,withgμν¯
=
diag(−
1,a2δ
i j
),
h00=
0,andhi j transverse andtraceless,i.e.,hii= ∂
ihi j=
0.TheEinsteinequationis¨
hi j+
3Hh˙
i j−
∇
2 a2hi j+
m 2 Thi j=
0 (66)withthetensormassgivenby
m2T
≡
m 2 2β
2 a2 1+
β
2 a2 (67)The structureoftheEinstein equation issuchthat m2T
/
m2 has to be a (quadratic) polynomial inβ
2/
a2. What is non-trivial is that thedegree-zeroterminthat polynomialcancelsout, i.e.,the expressionform2T
/
m2 startsatorderβ
2/
a2.Thismeansthatgrav-itational waves propagating over cosmological distances (at low redshift,i.e.,a
∼
O(
1))donotdependonm alone;insteadthey in-volvethecombinationsmβ
andmβ
2 which,aswehaveseen,are stronglyconstrainedbythecosmologicalbackground.Inparticular, recalling that m2β
210−2H2
0 andm2
β
410−4H 20,we see that mT atthepresenteraisatmostoforder10−1H0
∼
10−34eV,far belowtheLIGObounds.Moreover,ourstabilitycondition (65) has no bearingon mT. No matter howtinyβ
is,the constraints(44) placeaconstantupperboundonmβ
,sothatthesmallerβ
is,the largerm isallowedtobe,leavingmT≈
mβ/(
√
2a)fixed.Itis inter-esting to note that, without demanding that thismodelprovides cosmic acceleration, the tensormass isnevertheless forcedtobe smallerthan theHubblescale.Finally,we notethat arounda flat background,thetensormassissimplym, solocaltestsofgravity mightbeabletoplaceconstraintsonm thatarenotpossiblewith gravitationalwavesthatpropagateovercosmologicaldistances.4.3. Quasistaticlimit
Finally, let uscomment on the testability ofmimetic massive gravity using near-future LSSsurveys. We willfind it convenient
11 Thisissimilartomassivebimetricgravity,whichpossessesanearly-time
in-stabilitythatcanberenderedsafeinthelimitwheretheratioofthetwoPlanck massesbecomessmall [43].
12 Itisplausiblethattheresult(63) holds,atleastonanorder-of-magnitudebasis,
throughradiationdominationaswell(see,again,theexampleofbigravity[43]).In thiscase,weshoulddemandthattheinstabilitybepushedbacktobeforebigbang nucleosynthesis,withzBBN≈3×108,whichwouldimplyastrongerconstraintof
β10−8.Wedonothavemuchobservationalhandleonthepresumably
radiation-dominatederabeforeBBN,andthereforeshouldnotdemandthattheinstabilitybe absentthen;indeed,amildenoughinstabilitymighthaveinterestingconsequences, suchastheformationofprimordialblackholes.
13 SeeRef. [45] forahelpfulsummaryofboundsonthe gravitonmassfroma
towork in Newtonian gauge, B
=
E=
0.Linearizing the Einstein equations (3), and leaving in a generic stress-energy tensor Tμνforcompleteness,weobtain
6H2
φ
−
2 a2∂
i∂
iψ
+
6H˙ψ
=
1 MPl2δ
T 0 0+
2δλ
M2Pl−
m2 4β
2 a2 3−
β
2 a23
ψ
+ ∂
i∂
iπ
,
(68)−
2∂
i˙
ψ
+
Hφ
=
1 MPl2δ
T 0 i+
2¯λ
MPl2∂
iπ
0+
m2 4 3−
β
2 a2∂
iπ
0− β
2∂
iπ
˙
,
(69) 6¨ψ +
3H˙ψ +
H˙ + (
3H2+
2H˙
)φ
+
2 a2∂
i∂
i(φ
− ψ)
=
1 M2 Plδ
Tii−
m 2 4β
2 a2 1+
β
2 a23
ψ
+ ∂
i∂
iπ
,
(70) 1 a2∂
i∂
j(ψ
− φ) =
1 M2Plδ
T i j+
m2 2β
2 a2 1+
β
2 a2∂
i∂
jπ
,
i=
j.
(71)MovingtoFourierspace, specializingtoa pressurelessfluid with-outanisotropic stress,andtakingthequasistaticlimit, X
¨
∼
HX˙
∼
H2Xk2X for any perturbation X , eqs. (68), (70) and (71)
be-come 2k2 a2
ψ
=
1 MPl2(
2δλ
− ¯
ρ
δ)
−
m2 4β
2 a2 3−
β
2 a23
ψ
−
k2π
,
(72) 2k2 a2(φ
− ψ) =
m2 4β
2 a2 1+
β
2 a23
ψ
−
k2π
,
(73) 1 a2(φ
− ψ) = −
m2 2β
2 a2 1+
β
2 a2π
,
(74)where
ρ
¯
andδ
arethebackgrounddensityandoverdensity ofthe dustcomponent.Notethatthesearedegeneratewiththemimetic darkmatter,asexpected.Combining these equations, we obtain the modified Poisson equationandthesliprelation,
−
k2ψ
=
4π
Gμ
(
a,
k)
a2(δ
ρ
−
2δλ),
(75)ψ
=
η
(
a,
k)φ,
(76)wherethemodified-gravityparameters
μ
andη
aregivenbyμ
(
a,
k)
=
1 1+
12m2β2 k23
−
β2 a2,
(77)η
(
a,
k)
=
1 1+
12m2β2 k21
+
β2 a2.
(78)Theseparametrizeobservabledeviationsfromgeneralrelativity,in which
μ
=
η
=
1.Theconstraintswehavealreadyderived onm and
β
precludeμ
andη
fromdeviating fromunity ata levelaccessible to near-futureobservations.Thestabilityconstraint(63) requirestheterms inparenthesesto beO(
1),while the backgroundconstraint(56) setsm2β
20.02H2 0,sothat
μ
−
1∼
η
−
1∼
O
m2β
2 k2 10−2 H0 k 2.
(79)It istherefore highly unlikely that cosmologicalobservations will be able to test this model against
CDM
in the linear and sub-horizonrégime.5. Conclusions
In this Letter we have studied the first cosmological implica-tions of the recently-proposed theory of mimetic massive grav-ity. We find that the theory is unable to self-accelerate without introducinga ghost.Itseffects on Friedmann-Lemaître-Robertson-Walker cosmological backgrounds are tointroduce effective radi-ation, curvature, and cosmological constant terms, as well as a dust-like mimetic dark matter component. We place constraints (44) on the theory parameters by demanding that the effective radiationandcurvatureterms bewithin observational bounds.In theghost-freeregionofparameterspace,m2
>
0,theeffectivecos-mologicalconstant isnegative-definite, soa separate darkenergy sector,whichwetaketobeapositivecosmologicalconstant,is re-quiredtoexplainthelate-timeaccelerationoftheUniverse.
Wefurther studiedthebehaviorofcosmologicalperturbations inthe subhorizon,quasistaticlimit.The modelgenerically suffers fromaninstabilityatearlytimes.However,sinceouranalysisonly includeda pressurelessdustcomponent (inaddition toa cosmo-logicalconstant),thecalculationcanonlybetrustedasfarbackas matter-radiation equality.This allowedusto place afurther con-straintonthetheoryparametersbyinsistingthattheinstabilitybe absentthroughoutmatterdomination.Withtheseconstraints,the deviationsfrom
CDM
inthelinear,subhorizonrégimeare likely toosmalltobeobservable.Notsurprisingly,sincethisisatheoryofmassivegravity,it pre-dicts massive tensormodes. We have calculatedthe tensormass around cosmologicalbackgrounds andfoundthat, takinginto ac-count the constraints imposed by the cosmological background, thismassmustbeatleastanorderofmagnitudebelowthe Hub-ble scale, far outside the currently-available constraints on the graviton mass. Unlikeother theoriesof massivegravity, in which the graviton mass is comparableto the Hubblescale in order to providelate-timeacceleration,thisbound onthegraviton massis solelyduetotherequirementthattheeffectiveradiationand cur-vaturetermsintheFriedmannequationnotbetoolarge.
What are the remaining prospects for cosmological tests of mimetic massive gravity? We emphasize that our analysis does notapply intwoimportantrégimes:horizon-sizescalesand non-linearscales. Oneor bothofthesemaypossesssignatures which couldbeusedtodistinguishmimeticmassivegravityfrom
CDM,
or otherwise to rule out additional regions of parameter space. OneexpectsthatnonlinearscaleswillrequireN-bodysimulations, while at horizon-size scales we cannot apply the quasistatic ap-proximation andwouldneed to solvethe perturbation equations numerically,asinother theoriesofmodified gravity[46]. Forthe latter, we note that themass scales appearing in theaction (49) forcosmologicalperturbationsarenotsimplym,whichcanbe ar-bitrarily large (in the limit of smallβ
), butrather mβ
andmβ
2, which we haveshownmust bothbe at leastan order of magni-tudesmallerthantheHubblescale.Itthereforemightbe difficult forthistheory toproduceeffectsathorizonscalesthatarelarger than cosmic variance. Note that scales k∼
mβ
and k∼
mβ
2 are super-horizonandthereforenotobservable.Acknowledgements
helpful discussions. We thank the referee for useful comments on the draft. A.R.S. is supported by DOE HEP grants DOE DE-FG02-04ER41338 and FG02-06ER41449 and by the McWilliams Center for Cosmology, Carnegie Mellon University. V.V. is sup-ported by a de Sitter PhD fellowship of the Netherlands Organi-zation for Scientific Research (NWO). Y.A. is supported by LabEx ENS-ICFP:ANR-10-LABX-0010/ANR-10-IDEX-0001-02PSL*.Y.A.also acknowledges support fromthe NWO and the Dutch Ministryof Education, Culture and Science (OCW), and also from the D-ITP consortium,aprogramoftheNWOthatisfundedbytheOCW.
References
[1]A.H.Chamseddine,V.Mukhanov,J.HighEnergy Phys.06(2018)060,arXiv: 1805.06283 [hep-th].
[2]A.H.Chamseddine,V.Mukhanov,J.HighEnergy Phys.06(2018)062,arXiv: 1805.06598 [hep-th].
[3]A.H.Chamseddine,V.Mukhanov,J.HighEnergy Phys.11(2013)135,arXiv: 1308.5410 [astro-ph.CO].
[4]M.Fierz,W.Pauli,Proc.R.Soc.Lond.A173(1939)211. [5]C.Isham,A.Salam,J.Strathdee,Phys.Rev.D3(1971)867. [6]D.Boulware,S.Deser,Phys.Rev.D6(1972)3368.
[7]P.Creminelli,A.Nicolis,M.Papucci,E.Trincherini,J.HighEnergyPhys.0509 (2005)003,arXiv:hep-th/0505147 [hep-th].
[8]C.deRham,G.Gabadadze, Phys.Rev.D82(2010) 044020,arXiv:1007.0443 [hep-th].
[9]C.deRham,G.Gabadadze,A.J.Tolley,Phys.Rev.Lett.106(2011)231101,arXiv: 1011.1232 [hep-th].
[10]S. Hassan, R.A. Rosen, Phys. Rev.Lett. 108(2012) 041101, arXiv:1106.3344 [hep-th].
[11]S.Hassan,R.A.Rosen,A.Schmidt-May,J.HighEnergyPhys.1202(2012)026, arXiv:1109.3230 [hep-th].
[12]S.Hassan,R.A.Rosen,J.HighEnergyPhys.1204(2012)123,arXiv:1111.2070 [hep-th].
[13]S.Hassan,A.Schmidt-May,M.vonStrauss,Phys.Lett.B715(2012)335,arXiv: 1203.5283 [hep-th].
[14]S.Hassan,R.A.Rosen,J.HighEnergyPhys.1202(2012)126,arXiv:1109.3515 [hep-th].
[15]K.Hinterbichler, R.A.Rosen,J.HighEnergyPhys.07(2012)047,arXiv:1203. 5783 [hep-th].
[16]K.Hinterbichler,Rev.Mod.Phys.84(2012)671,arXiv:1105.3735 [hep-th]. [17]C.deRham,LivingRev.Relativ.17(2014)7,arXiv:1401.4173 [hep-th]. [18]A.R.Solomon,CosmologyBeyondEinstein,Ph.D.thesis,CambridgeU.,Cham,
2015,arXiv:1508.06859 [gr-qc].
[19]A.Schmidt-May,M.vonStrauss,J.Phys.A49(2016)183001,arXiv:1512.00021 [hep-th].
[20]E.A.Lim,I.Sawicki,A.Vikman,J.Cosmol.Astropart.Phys.1005(2010)012, arXiv:1003.5751 [astro-ph.CO].
[21]L.Mirzagholi,A.Vikman,J.Cosmol.Astropart.Phys.1506(2015)028,arXiv: 1412.7136 [gr-qc].
[22]H.vanDam,M.J.G.Veltman,Nucl.Phys.B22(1970)397.
[23]V.I.Zakharov,JETPLett.12(1970)312;Pis’maZh.Eksp.Teor.Fiz.12(1970) 447.
[24]A.Higuchi,Nucl.Phys.B282(1987)397. [25]A.Vainshtein,Phys.Lett.B39(1972)393.
[26]E.Babichev,C.Deffayet,Class.QuantumGravity30(2013)184001,arXiv:1304. 7240 [gr-qc].
[27]S.L.Dubovsky,J.HighEnergyPhys.10(2004)076,arXiv:hep-th/0409124 [hep -th].
[28]D.Blas,D.Comelli,F.Nesti,L.Pilo,Phys.Rev.D80(2009)044025,arXiv:0905. 1699 [hep-th].
[29]M.Lagos,M.Bañados,P.G.Ferreira,S.García-Sáenz,Phys.Rev.D89(2014) 024034,arXiv:1311.3828 [gr-qc].
[30]H.Motohashi,T.Suyama,K.Takahashi,Phys.Rev.D94(2016)124021,arXiv: 1608.00071 [gr-qc].
[31]S.Dubovsky,T.Gregoire,A.Nicolis,R.Rattazzi,J.HighEnergyPhys.03(2006) 025,arXiv:hep-th/0512260 [hep-th].
[32]S.Endlich,A.Nicolis,R.Rattazzi,J.Wang,J.HighEnergyPhys.04(2011)102, arXiv:1011.6396 [hep-th].
[33]G.R.Dvali,G.Gabadadze,M.Porrati,Phys.Lett.B485(2000)208,arXiv:hep -th/0005016 [hep-th].
[34]C.Deffayet,Phys.Lett.B502(2001)199,arXiv:hep-th/0010186 [hep-th]. [35]C.Deffayet,G.R.Dvali,G.Gabadadze, Phys.Rev.D65(2002)044023,arXiv:
astro-ph/0105068 [astro-ph].
[36]D.Gorbunov,K.Koyama,S.Sibiryakov,Phys.Rev.D73(2006)044016,arXiv: hep-th/0512097 [hep-th].
[37]C.Charmousis,R.Gregory,N.Kaloper,A.Padilla,J.HighEnergyPhys.10(2006) 066,arXiv:hep-th/0604086 [hep-th].
[38]J.Khoury,J.Sakstein,A.R.Solomon,J.Cosmol.Astropart.Phys.1808(2018)024, arXiv:1805.05937 [hep-th].
[39]N.Aghanim,etal.,Planck,arXiv:1807.06209 [astro-ph.CO],2018.
[40]X.-l.Chen,R.J.Scherrer,G.Steigman,Phys.Rev.D63(2001)123504, arXiv: astro-ph/0011531 [astro-ph].
[41]R.H.Cyburt,B.D.Fields,K.A.Olive,T.-H.Yeh,Rev.Mod.Phys.88(2016)015004, arXiv:1505.01076 [astro-ph.CO].
[42]D.J.Fixsen,Astrophys.J.707(2009)916,arXiv:0911.1955 [astro-ph.CO]. [43]Y.Akrami,S.F.Hassan,F.Könnig,A.Schmidt-May,A.R.Solomon,Phys.Lett.B
748(2015)37,arXiv:1503.07521 [gr-qc].
[44]B.P.Abbott,etal.,LIGOScientific,VIRGO,Phys.Rev.Lett.118(2017)221101, arXiv:1706.01812 [gr-qc];Erratum:Phys.Rev.Lett.121 (12)(2018)129901. [45]C.de Rham,J.T. Deskins, A.J.Tolley, S.-Y.Zhou,Rev.Mod. Phys. 89 (2017)
025004,arXiv:1606.08462 [astro-ph.CO].