Higuchi Bound in Massive Gravity and Bigravity Andrew J Tolley Case Western Reserve University Based on work with Matteo Fasiello 1206.3852 (Massive Gravity) 1207.???? (Bigravity) Overview • Why Massive Gravity? • Problems with FRW • FRW on de Sitter Massive gravity/bigravity • Higuchi and Vainshtein • Resolution - Inhomogenities? Bigravity Why Massive Gravity? Massive Gravity Theories are a remarkably a constrained modification of general relativity at large distance scales - graviton is assumed to acquire a mass In present talk I shall only be concerned with models where this occurs without breaking Lorentz or de Sitter symmetries They are interesting in that as in GR, there are a finite number of consistent allowed terms in the Lagrangian that do not give rise to ghosts By Massive Gravity we mean a nonlinear completion of Fierz-Pauli coupled to matter Markus Fierz and Wolfgang Pauli, 1939 hµ + · · · = m (hµ 2 1 µ 5 = 2s + 1 Fierz-Pauli mass term guarantees 5 rather than 6 propagating degrees of freedom Massless spin-two in Minkowski makes sense! h) Why Massive Gravity? Adding a mass to gravity weakens the strength e mr VY ukawa ⇠ of gravity at large (cosmological) distances r But thats not all! Screening mechanism Self-acceleration? Degravitation mechanism? Why Massive Gravity? Self-acceleration? Gravitons can condense to form a condensate whose energy density sources self-acceleration ⇢matter ⇠ 0 H ⇠ m 6= 0 Analogous to well-known mechanism in DvaliGabadadze-Porrati model (DGP), however here it seems possible to remove the DGP ghost?? Deffayet 2000 Koyama 2005 Charmousis 2006 Why Massive Gravity? Gravitons can condense to form a condensate whose energy density compensates the cosmological constant Screening mechanism - The Cosmological Constant can be LARGE with the cosmic acceleration SMALL In a Massive Theory - the c.c. is a `redundant’ operator Why Massive Gravity? mass term Gµ⌫ @L M 2 +m = @gµ⌫ ⇤gµ⌫ Graviton condensate: Spacetime is Minkowski in presence of an arbitrary large gµ⌫ = ✓ 1+f ✓ ⇤ m2 ◆◆ ⌘µ⌫ Gµ⌫ = 0 2 @LM m @gµ⌫ = ⇤ ⇤gµ⌫ Equivalent Statement: The cosmological constant can be reabsorbed into a redefinition of the metric and coupling constants - and is hence a redundant operator Why Massive Gravity? Screening Degravitation One strong motivation for considering Massive Gravity is as a toy model of higher dimensional gravity models (eg Cascading Gravity) that potentially exhibit degravitation de Rham et al 2007 Degravitation = Dynamical Evolution to a Screened Solution from generic initial conditions Dvali, Hofmann, Khoury 2007 so far it is safe to say that this idea has not YET been fully realized Why Massive Gravity? Departure from GR is governed by essentially a single parameter - Graviton Mass Vainshtein Screening mechanism ensures recovery of GR in limit m ! 0 This ensures massive gravity can be easily made to be consistent with most tests of GR (effectively placing an upper bound on m) without spoiling its role as an IR modification Why Massive Gravity? Massive Gravity is a natural Infrared Completion of Galileon Theories Galileon: Nicolis, Rattazzi, Trincherini 2010 Decoupling limit of Massive Gravity on Minkowski is a Galileon Theory de Rham and Gabadadze 2010 Decoupling limit of Massive Gravity on de Sitter is a Galileon Theory (with slightly different coefficients) de Rham and Renaux-Petel 2012 The allowed Galileon Interactions are in direct correspondence with the allowed MG interactions Why Massive Gravity? Massive Gravity models share many nice features in common with extra dimensional models such as DGP and Cascading Gravity ..... e.g. Vainshtein mechanism, Galileon limit, self-acceleration, possible screening .... however without the difficulty of having to solve fundamentally higher dimensional equations nomials The of the eigenvalues ofpotential the tensorU that has no ghosts [20] is bu most general generalization of the theory proposed in [20]. The Lagrangia ! " # ! 2 (4) 2 the µeigenvalues polynomials of of the tensor (4) µ µα M − gravity g R + 2mmatter U(g, )plus LaM .− Gravity nstein with a scalar fun Ghost-free Massive g fανthat is(2.1) Kν f(g, f+ )= δνpotential Pl ! sl U that has no!ghosts "[20] is build Kout µ µ µα f of characteristic g (g, f ) = δ − # αν ν ν 2 (4) 2 hat (4)g L = M − R + 2m U(g, f ) + LM . Pl ues of the tensor U(g, H) = U2 + α3 U3 + α4 U4 , so that ! ral potential U that has no ghosts [20] is build out of chara µ µ µα re the αn f are U(g, H) = U2 + α3 U (2.2) Kν (g, ) =free δν parameters, − g fαν and 3 + α4 U4 , the eigenvalues "of the tensor # 2 2 where αn are free and U2 the = [K] − [K ] ,parameters, ! "µ 3 # µ2 µα 3 " # g f f ) = δ − αν [K] − 3[K][K ] + 2[K ] , ν3(g, ν U(g, H) =UU32=+K α U + α U , (2.3) 2 2 3 4 4 − [K3 ] , " 4 U2 =2 [K] # 2 2 2 4 " + U4 = [K] − 6[K ][K] 8[K ][K] + 3[K ] 3−#6[K ] , 3 2 meters, and U3 = [K] − 3[K][K ] + 2[K ] , " 4 PRL,2 106,2 2311013 (2011) de Rham, Gabadadze, Tolley, 2 Th 2 # re [. . .]2 represents the H) trace of a tensor with respect to the metric g . U(g, = U + α U + α U , U4 =2 [K]3 −3 6[K ][K] µν ] − 4 4 + 8[K ][K] + 3[K − [K ] , (2.4) Proven fully ghost free in ADM formalism: Hassan and Rosen host for this theory for# a Minkowski background metric was shown in the d 2 3 2011 ofthe re free parameters, + 2[K ,and [. fully . .] ]represents the beyond trace a tensor with (2.5) respect to the t−in3[K][K [17, where 18,] 20], non-linearly decoupling limit in [21, 22 # Result reconfirmed in Stueckelberg decomposition: Result reconfirmed in helicity decomposition: 2 of 2 3this 2for 2 a Minkowski 4 [23, n−the Stückelberg and helicity languages in 24]. " # ghost for theory background metric wa 6[K de ][K] + 8[K ][K] + 3[K ] − 6[K ] , (2.6) 2 Gabadadze, 2 Tolley 2011 Rham, de Rham, Gabadadze, Tolley 2011 U = [K] − [K ] , 2 Hassan, Varying with respect to Strauss the metric gµν we find the equations of motion limit in [17, 18, 20], fully beyond the decoupling Schmidt-May, von 2012 non-linearly " 3 Kluson 2012 2 # 3 metric g . The absence ace of a tensor with respect to the in the Stückelberg and2]helicity languages in [23, 24]. U3 = as[K] − 3[K][K ] + 2[K , µν −2 G + m X = M T " # Now several other proofs: Mehrdad Mirbaryi 2011, AJT to appear µν µν µν pl Minkowski background metric was in the 4Varying2with 2 respect 3toshown 2 2 gdecoupling 4find the equa the metric we µν U = [K] − 6[K ][K] + 8[K ][K] + 3[K ] − 6[K ] , dRGT model: allowed mass terms de Rham, Gabadadze, Tolley 2011 Build out of unique combination Mass terms are characteristic polynomials K µ = gµ f µ U (g, f ) = i Ui (K) i n=d det( µ + K µ )= n Un (K) n=0 Finite number of allowed interactions in any dimension Interactions protected by a Nonrenormalization theorem Generalized to arbitrary (dynamical - bigravity) reference metrics by Hassan, Rosen 2011 3]. Away from the decoupling limit this is related to the constraint that was en we see the constraint for the FRW metric to all orders, by Refs. [1, 4, 5]. Here we see the constraint for the FRW metric to all orde 4) w.r.t. f : of (14) w.r.t. f : taking variation A No-Go 2 3 2 2 3 2 m ∂ (a − a ) = 0 . Gravity in Minkowski) m ∂0 (a − a(dRGT ) = 0model . 0 - Massive (15) The simplest model not support spatially flat (or closed) solutions Thisdoes constraint makes time evolution of the scale cosmological factor impossible. As we have time evolution of the scale factor impossible. As we have noted 3 4 whichthe areKFRW meaning homogeneous and isotropic above, keeping and K terms in (12) can only modify the polynomial fu 4 and Kwhich terms in (12) can only modify the polynomial function of a on ∂ acts in (15). Therefore, there are no nontrivial 0 where overdot denotes the time derivative ∂homogeneou 0 . We em in (15). Therefore, no homogeneous Argument isinsimple: as inare GR wenontrivial have equation sotropic solutions thethere theory of massive GR,Friedman defined by (12). andand appears in the Lagrangian only linearly. The same rem Ittheory is also instructive to show absence FRW 1st solutions the unitary equation - thethe 2nd follows from by diffininvariance theRaychaudhuri of massive GR, defined byof(12). α4no- fitfield is just theinspecial structure of theInterm a α3a and µ for which φ = δ x , and appears the action to begin with. this µ e to show the absence of FRW solutions in the unitary gauge, ˙ that ensures that f enters only linearly! This is a conse theBut most general homogeneous and isotropic ansatz involves the lapse function in MG diff invariance is broken and so 2nd does not follow from nd no f fielddecoupling appears inlimit the action to begin with. In this gauge, the equations of motion of this theory 1st - consistencyds of2 two imposes condition on scale factor 2 2 2 2 = −N (t)dt + a (t)d% x , ogeneous and isotropic ansatz involves the lapse function N(t), derivatives acting on the helicity-0 field in particular [3]. from and the2 Lagrangian (12)2 with = αdecoupling 2 Away 2α3the 24 = 0 reads limit this is related to t ds = −N (t)dt + a (t)d%x , (16) ! [1,2 4, 5]. Here we see the constraint" for the in Refs. aȧ 2 2 3 2 2 3 2 L = 3M − − m (a − a ) + m N(2a − 3a + a) . taking variation of (14) w.r.t. f : Pl 12) with α3 = α4 = 0Nreads " 3 2 2 m ∂ (a − a ) = 0 . As a can verified, the condition (15) in this case arises D’Amico et al 2011 0 ȧ be straightforwardly 2 3 2 2 3 2 − − mof (a − a ) +ofmtheN(2a − 3a + a) for . the two fields, (17)a a requirement consistency equations of motion ! 2 eeping the K3 and K4 terms in (12) can only modify the polynom which ∂0 acts in (15). Therefore, there are no nontrivial homo c solutions in the theory of massive GR, defined by (12). also instructive to show the absence of FRW solutions in the un a a µ h φIt = δ x , and no f field appears in the action to begin with. I µ is possible to find exact solutions in which the metric takes t general homogeneous andthe isotropic ansatz involves the lapse fu form ... A No-Go? 2 2 2 2 2 ds = −N (t)dt + a (t)d%x , Lagrangianin(12) with 0 reads type equation which a(t)αsatisfies a Friedman 3 = α4 = ! " D’Amico et al 2011 2 a ȧ 2 2 3 2 2 3 2 Volkov 2011 L = 3MPl − − m (a − a ) + m N(2a − 3a + a) . Koyama et al 2011 N Gratia et al 2012 Kobayashi 2012case a be straightforwardly verified, the condition (15)etinalthis But is achievedofbythe introducing fields carry ment ofthis consistency equationsStuckelberg of motion for which the two fiel the specifically, inhomogeneities meaning that(15) these are difference not truly More one can obtain bysolutions taking the i.e. the for perturbations inhomogeneous ivative of FRW, the e.o.m. N and thearee.o.m. for a. Technicall Two paths Accept inhomogeneities: D’Amico, de Rham, Dubovsky, Gabadadze, Pirtskhalava, Tolley `Massive Cosmologies’ 2011 Not as bad as it sounds! Vainshtein mechanism should guarantee inhomogeneities unobservable before late times Inhomogenities only appear on scale set by inverse graviton mass Two paths Or modify assumptions to allow FRW: Open Universe solutions: Gumrukcuogli et al 2011 Anisotropic solutions: Gumrukcuogli et al 2012 Felice et al 2012 * Make reference metric de Sitter - AJT and Fasiello - 1206.3852 (for decoupling limit see de Rham, Renaux-Petel 2012) * Make reference metric dynamical - Bigravity/Bimetric von Strauss et al 2011 Comelli et al 2011 Crisostomi et al 2012 de Sitter MG and bigravity Qualitatively for the present discussion there is no great distinction between bigravity and de Sitter Massive gravity This is because the second metric may not directly couple to our observable matter (absence of ghosts) other than having its own cosmological constant Thus for suitably low energies bigravity looks like MG on de Sitter (or Minkowski/AdS) However, we will see later that quantitatively there is a different for the Higuchi bound Crux of problem Although we can obtain FRW like solutions, number of issues ... The `mass’ of a graviton gets dressed by the background Generically the mass grows with increasing H Thus the Vainshtein mechanism is more subtle!! We must send m ! 0 in away that compensates growth with H Generalized Higuchi bound implies 2 mdressed (H) > 2H Successful Vainshtein mechanism (recovery of GR at large H) and Higuchi bound are incompatible for FRW solutions Tolley and Fasiello (to appear tomorrow) 2 Generalized Higuchi bound Fasiello and AJT - 1206.3852 Previous Work: 2 Higuchi 1987, Deser and Waldron 2001 (de Sitter) m Grisa, Sorbo 2009 Generalized to FRW Berkhahn et al 2010 (Similar results to above) Grisa and Sorbo obtain: m2 > 2(H 2 + Ḣ) seemingly no problem in deccelerating universe ?!?! However! These authors assumed the equivalence of the background FRW metric and reference metric - this is inconsistent with known behaviour of dRGT and de Sitter/ Bigravity generalization Necessary to use correct nonlinear theory to obtain result! 2H 2 dRGT trics Massive Gravity ! " (4) # 2 2 (4)g L = M − Rreference +reference 2m U(g, f )fµν + L . heory of massive gravity defined on an arbitrary metric [53] isMjust Pl avity defined on an arbitrary metric f is just a µν fghtforward massive gravity defined on an arbitrary reference metric fµν generalization of the theory proposed in [20]. The Lagrangian takes ation of potential thegravity theory proposed in Lagrangian takes general Uof that has no [20]. ghosts is function build ofLagrang charact m ofgeneralization Einstein with matter a potential thatThe is [20] a scalar d theplus theory proposed in [20]. ofout The o metrics with matter plus a potential that is a scalar function of sity of the eigenvalues of the tensor 2 ! " (4) # a potential that is a scalar fu MPl with insteinStarting gravity matter plus 2 (4) point L= − g R + 2m U(g, f ) + LM . (2.1) 2 ! ! " # cs µ 2 " µ 2general potential (4)U that ! µα (4) # ost has no ghosts is build out of characteristic polyno- (2.1) g f K (g, f ) = δ − MPl − g R + 2m U(g, f ) + L . αν ν ν M 2 (4) 2 (4) Sketch of argument of the eigenvalues L = ofMthePltensor − g R + 2m U(g, f ) + LM . ! ial U that has no µ ghosts µ [20]µαis build out of characteristic (2.2) Kν (g, f ) = δν − g fαν , neral potential U that has no ghosts [20] is build out of char alues of the tensor U(g, H) = U2 + α3 U3 + α4 U4 , tf the eigenvalues of the tensor ! U(g, H) = U2 + α3 U3 + α4 U4 , (2.3) µ µ µα αn areKνfree parameters, and g f (2.2) (g, f ) = δ − ! αν ν de Sitter spacetime metric the αn are free parameters, and µ µ fµ⌫ µα " 2 Kν 2(g, # f ) = δν − g fαν # ] , 2 − [K 2 UU22 = = 1 "[K] [K] − [K ] , (2.4) " # 2! 3 2 # 3 "[K] 1= U(g, U + α U + α U , (2.3) UU3 H) = − 3[K][K ] + 2[K ] , 3 32 2 3 34 4 (2.5) # 3 = " [K] − 3[K][K ] + 2[K ] , 3! 4 U(g, H) 2 =2U + α3 U + α U2 ,2 4 " # U = [K] − 6[K ][K] + 8[K ][K] + 3[K ] − 6[K ] , 2 3 3 4 4 1 4 and ameters, 4 2 2 3 2 2 4 U4 = [K] − 6[K ][K] + 8[K ][K] + 3[K ] − 6[K ] , (2.6) 4! #parameters, and are free 2 2 represents the trace of a tensor with respect to the metric(2.4) gµν . The ab − [K ] , For experts U1 is removed by tadpole condition and U0 is a c.c. # " # this theory for a Minkowski background metric was shown in the deco 3 2 3 2 2 which can be absorbed –, 3 – into definition of matter −U3[K][K ] + 2[K ] , (2.5) = [K] − [K ] 2 , 18, 20], fully non-linearly beyond the decoupling limit in [21, 22], a # 4 consider FRW solut Let us de Sitter. For simplicity I willasassume k =Sitter. 0. We denote theI d de For simplicity he a scale factor is properly calculated e Gravity " !# on ✓ ◆ 2 2 2 2 2 example in two dimensions Deriving Friedman equation ˙0 1 d reference d ȧon an arbitrary ḃ ds = fµνNis just dt + x ravity 1defined metric a a(t) d~ = (1.5) ˙0 ation of the theory proposed The Lagrangian takes N dt N N ˙0 dt in [20]. and theofreference metric as W and solutions in reference massive on a fixed FRW such as vity withthe matter plusgravity a potential that is a geometry scalar function metric as Nice approach is with Stuckelberg fields icity I will assume k = 0. We denote the dynamical metric as brackets is the acceleration of the reference metric which for de ! " # 2 (4) 2 2 (2.1) be positive. Thus if the dynamics the a MPl − (4) g R + 2m U(g, L . 2metric switched 2 from 2 0 2 2 2 2 2 f) 2 + of M 0 ˙ ds = N dt + a(t) d~x (1.1) ds = dt + b ( )d~ x 0 ˙ ration, this must be accompanied by a change of sign of which ial U that hasthere no ghosts [20] is build out ofwecharacteristic 0 is an ambiguity in how take the square root. Fortuna where is the time-Stuce etric as the lapse of the reference metric must vanish. galues that 0 0 of the tensor p 0 H b µ µ isisresolved in theegvierbein formulation of massive gravity, which wherebecause the time-Stuceklberg field. If the fixed metric is µ! b( ) = e with H cons in de Sitter problem when we evaluate the tensor K = g ⌘ 2 b !⌫ ⌫ ⌫ 2 2! 2 0 2 0 ˙ 0 p ds = dt + b ( )d~ x (1.2) 0µ µH Stückelberg fields. This tells us that the correct way to takethat the squ b µ! µ µα b( Kν)(g,=Q e with H constant. o evaluate = g ⌘ . Now since The constraint co !⌫ b (2.2) f )⌫ = δν − g fαν ˙ is of the sign of ! e-Stuceklberg If the fixedthat metric is de Sitter thenthe we have a branch (the normal bran ! comes Thefield. constraint from variation of the actio 2 ˙ 0 ˙0 p 0j Hb constant. 1 0 1 N j 2 N g f = 0) a branch (the normal branch) of solutions for which g f = (1.6) 0 has ˙ 0 2 b( that comes from the variation of the action with respect to b( ) U(g, H) = U2 + α30U3 + α4 U4 , (2.3) 0i a ij i ij 2 a al branch) of solutions for which ameters, and With this choice the action for FRWaH solutions of massive gravity i = bH b or (1.3) aH = bH b We2 #mustsince choose sign of square root to correlate it is constructed out of characteristic polynomials of K it is li 2 ] − [K ] , (2.4) 0 ˙ ˙ 0 problem in having pass though zero. Implicity what we #1 arises with sign of – – 3 or 2 3 p ] ḃ, p ] − 3[K][K ] +a 2[K (2.5) 0 ˙ ˙0 < ȧ + =for the zeroth eigenvalue for > 0 and the for # (1.4) ḃ ȧ 4 2 2 3 ˙0 2 2 4 N ][K] +this ] − 6[K ][K] formulation + 8[K 3[K choice ] − 6[K ] , have The (2.6) acceleration of now the aam s would been= automatic and 0 ˙ N arisen. the a scale factor is properly calculated as Deriving Friedman equation Since mass terms are characteristic polynomials of K - linear in ˙0 ⇣ Lmass = a3 N A( 0 , a) + ˙0 B( 0 , a) ⌘ It is clear than we can integrate by parts to remove ˙0 dependence to give a nondynamical equation for 0 in terms of a and H Constraint equation Consistency of Friedman and Raychauduri equation (or equation for zero Stuckelberg field) impies 1 + 2(1 + ↵3 ) + (↵3 + ↵4 ) 2 ✓ b a H H0 ◆ Normal branch of solutions is b b = a H = a H0 1 Equation fixes dynamics of Stuckelberg field Perturbations subtlety If metric transits from acceleration to decceleration we need ˙0 to change sign At thisp point one of the eigenvalues of g 1 f vanishes How do we define this perturbatively? Vierbein formulation a The vierbein formulation is analytic in the Mass term is Det ⇥ a eµ + a b ⇤b f c @ µ reference vierbein ⇤ c As long as it is possible to solve the equation for the a T ⇤ ⇤⌘⇤ = ⌘ Lorentz Stuckelberg fields b e µa b c ⇤c f d @ µ d e µb b c ⇤c fd @ µ d =0 6 equations for 6 unknown Lorentz transformations Even when ˙0 = 0 we can solve for a ⇤b = ...@ c H is the Hubble rate of the dynamical metric and H0 that of the refer cal) metric, and α3 and α4 are the two free parameters in the dR his should be compared with the associated Friedmann equation Friedman equation 2 m 1 2 H 2 ρ − (6 + 4α3 + α4 ) + (3 + 3α3 + α4 )m − H = 2 3MPl 3 H0 2 2 3 m H 2H (1 + 2α3 + α4 )m 2 + (α3 + α4 ) . 3 H0 3 H0 H = H0 1 ⇢dark energy –2– H0 is Hubble parameter of reference metric (e.g. dRGT has 2 free parameters) and even if we employ at full the freedom reference metric, once observations are taken into account the bound remain stringent. We hint to a possible resolution in the Conclusions. Our final bound by ! " 2 H H 2 2 2 H (3 + 3α3 + α4 ) − 2(1 + 2α3 + α4 ) + (α3 + α4 ) 2 ≥ 2 m̃ (H) (H) = m mdressed H0 H0 H0 Dressed Mass and Higuchi where H is the Hubble rate of the dynamical metric and H0 that of the referenc 2 parameters in the 2 dRGT dynamical) metric, and αbound are the twomfree 3 and α4 is Generalized Higuchi (H) > 2H dressed [20]. This should be compared with the associated Friedmann equation H = H0 arises 2 m 1 2 H 21 ρ − (6 + 4α3 + α4 ) + (3 + 3α3 + α4 )m − H = 2 3MPl 3 H0 2 2 3 H m H from coefficient of kinetic term for helicity zero mode 2 (1 + 2α3 + α4 )m 2 + (α3 + α4 ) . 3 H0 3 H0 Lhelicity zero / 2 2 mdressed (mdressed –2– 2 2H )(@⇡) This is a similar polynomial to what arises in the Friedman equation 2 Partially Massless Gravity Coefficient of kinetic term in general is proportional to ✓ m2dressed H 2H = m H0 2 2 (3 + 3↵3 + ↵4 ) H H 2(1 + 2↵3 + ↵4 ) + (↵3 + ↵4 ) 2 H0 H0 If we make the special choice ↵3 = 3/2 ↵4 = 3/2 2 mdressed and so if we choose 2 m = 2 2H0 2 ◆ 2H 2 Precisely Claudia’s values!! (N.B. my conventions are different) de Rham and Renaux-Petel 2012 2 H 2 2 2H = 2 (m H0 2 2H0 ) m2dressed = 2H 2 Kinetic term vanishes regardless of matter source!!! Higuchi versus Vainshtein 2 mdressed (H) 2 = m (1 + ) (1 2 ⇢dark energy = 3m ( Higuchi 2 (2 + ↵3 ( 2 ) + m ↵3 (3 2 2) 3 H = H0 ↵4 )) ) + ↵4 m 2 3 Vainshtein d d 2 2 2 ⇢ ⌧ H dark energy mdressed (H) > 2H dt dt 2 m 3 2 mdressed ⇠ ⇢dark energy ⇠ 3 H If ↵3 + ↵4 6= 0 H0 2 m 2 2 ↵3 + ↵4 = 0 (generic) mdressed ⇠ ⇢dark energy ⇠ 2 H H0 2 m ↵4 = ↵3 = 1 m2dressed ⇠ ⇢dark energy ⇠ H H0 1 an FRW reference metric, which has revealed several, interesting things. First of all once stability and observations are taken into account, the Higuchi bound is still quite stringent on the value of m, the parameter deforming GR. In summary the bound states that ! " 2 H H 2 2 H (3 + 3α3 + α4 ) − 2(1 + 2α3 + α4 ) + (α3 + α4 ) 2 ≥ 2H 2 . m̃ (H) = m H0 H0 H0 Higuchi versus Vainshtein where H is the Hubble rate of the dynamical metric and H0 that of the reference metric. Remarkably Ḣ drops our of generalized bound!!!! This bound differs significantly from previous conditions, in particular [39] (see also [41–43]) in that it is independent of Ḣ and the precise form of matter. Indeed A direct consequence of the ghost-free form This is a different, better theory of MG, where the freedom on fµν is also fully employed. W onlybranch firstwhich derivatives coefficient have(action indeed seenexpressible how there is now with an additional does relax the- Higuchi bound. of helicity zero mode kinetic term is just a function of first derivatives of metric in Stueckelberg analysis) 10 – 17 – so the window found by Grisa and Sorbo for deccelerating solutions m2 > 2(H 2 + Ḣ) is not present an FRW reference metric, which has revealed several, interesting things. First of all once stability and observations are taken into account, the Higuchi bound is still quite stringent on the value of m, the parameter deforming GR. In summary the bound states that ! " 2 H H 2 2 H (3 + 3α3 + α4 ) − 2(1 + 2α3 + α4 ) + (α3 + α4 ) 2 ≥ 2H 2 . m̃ (H) = m H0 H0 H0 Higuchi versus Vainshtein where H is the Hubble rate of the dynamical metric and H0 that of the reference the qualitative form of these results goes through in metric. H0 conditions, case ofsignificantly bigravity from where is dynamical - but [39] (see This the bound differs previous in particular also [41–43]) in that it is independent Ḣ and the precise form of matter. Indeed with a oftwist (later) 10 This is a different, better theory of MG, where the freedom on fµν is also fully employed. W have indeed seen how there is now an additional branch which does relax the Higuchi bound. Their is no regime for the de Sitter MG spatially flat cosmologies which is simultaneously obervationally – 17 – acceptable and ghost-free as long as the helicity zero mode is present Partia$y massless case is not included in this statement Resolution? One resolution to realise something like out universe in Massive Gravity models is to return to the inhomogenous solutions D’Amico et al 2011 Higuchi constraint is implied by representation theory of de Sitter group. Introducing inhomogenity in the metric breaks this relation Known exact solutions are self-accelerating type and sit in different branches than the generic solution - as yet the general solution - the one with all 5 degrees of freedom propagating which is continuously connected with the normal Minkowski vacuum is not known tion, it leads to the usual ΛCDM - like cosmological expansion of the background in the sub-horizon approximation used here. This is clear form the fact that the stress-tensor (24) gives rise to a de Sitter background as was shown in the previous subsection. Hence, in the comoving coordinate system – which differs from the one used above – the invariant de Sitter space will be the self-accelerating solution. All this can be reiterated by performing an explicit coordinate transformation see the presence of will thebe FRW toWe the can comoving coordinates. This done solutions in two steps.inInthe thefamous so-called Fermi 3 2 M ! 1 ⇤3be=locally m Mwritten normal coordinates, space and for all decoupling limitthe FRW P metric can P heldin fixed times, as a small perturbation over Minkowski space-time: de Rham et al 2010 ! " # $ µ ν 1 2 2 2 2 2 2 2 FRW ds = −[1 − (Ḣ + H )x ]dt + 1 − H x dx = ηµν + hµν dx dx , (25) 2 Reasons to be hopeful? where the corrections to theform abovefor expression are suppressed by higher powers of The generic solution the helicity zero mode near x=0 H 2 x2 . The Fermi normal coordinates, on the other hand, are related to those used in which is isisotropic in this deformation limit is of Minkowski space(14) (in which the FRW metric a small conformal 2 time), by an infinitesimal gauge transformation [15]. The latter does not change the ⇡ ⇠ A(t) + B(t)x π expression (24), since Tµν is invariant under infinitesimal gauge transformations in the decoupling limit. fix OnA the the Fermi normal coordinates can be Equations of motion and other B - forhand, example for pure cc source B=constant transformed into the standard comoving coordinates (tc , xc2) as follows [15] A = Bt ! " 1 1 2 x 2 tc = t − H(t)x , xc = 1 + H (t)x2 . (26) 2 a(t) 4 direct calculations to the 6 order show that the helicity-1 fluctua degravitating branch of positive solutions, and this being by any instability kinetic termwithout as long as κ/(qplagued − 1) > 0. This suggests the p 0 at least in the decoupling limit. attractor solution for degravitation. The special case a3 late-time separately below. Reasons to be hopeful? 4.2 Degravitation in ghostless massive gravity 4.2.2 de Sitter branch Forde convenience we start by recalling the decoupling limit Lagrangian (9) coupled Rham, Gabadadze, Heisenberg, Pirtzkhalava 2010 - ofdecoupling Incloser the presence a cosmological constant, the fieldand equations (4 This solution to the of usual GR deonly Sitter configuration only exis to usual GR de is Sitter configuration and exists if to the an external source limit 2 admit a second branch of solutions; these connect with the self-acc 3this ≥ 3a1 a3 .can The stability of# solution can by be analyzed as previously by loo ity of thisa2solution be analyzed as previously looking 1 µν αβ 1 them a3, n and(n) µν µν as the de Sitter solutio presented in section we refer to at fluctuations around this background configuration, L = − X [Π] + h E h + h h Tµν . (35) αβ µν this background 2configuration, 3(n−1) µν MPl eters for these solutions should satisfy Λ3 n=1 1 3 2 2 1 q Λ + φ , π = a1 + 3a q dS2a23qx 3 2 dS 3 dS = 0 , qdS Λ3 x + φ , (51) π = 13 2 3 ! " λ 2Λ 2 3 21 2 3 2 2 + H =dS a q + a2 qdS + a3 qdS . h = − H x dS 2ηµν + χµν , 1 dS µν 1 2 2 3MPl MPl 2 hµν = − HdS x ηµν + χµν , (52) Tµν = −ληµν + τµν . 2 background plus 15 (53) To second order in fluctuations, the resulting action is thensplit of the form perturbations Tµν = −ληµν + τµν . tuations, the resulting action is then6H of2the form 1 M (2) µν αβ dS Pl αβ ν χαβ L + = − χ Eµν χαβ + 2 2 6HdS MPl (a2 + 3 It is Λ3interesting 3a3 qdS )φ!φ + 3 Λ 13 1 µν (a2 + 3a3 qdS )φ!φ + χ τµν . MPl χµν τµν . (54) to point out again that the helicity-0 fluctuation φ then decou MPl coefficient helicity zero order simple from matterof sources at quadratic (however the coupling reappears at the c nt out function again thatStability the helicity-0 φ then decouples order). is therefore ensured if the parameters satisfy o ↵of3 this↵solution of 4fluctuation ˜ hµν Tµν 1 2 2 = − HdS x ηµν + χµν , 2 = −ληµν + τµν . (52) (53) Reasons to be hopeful? To second order in fluctuations, the resulting action is then of the form L(2) 2 1 µν 1 µν αβ 6HdS MPl = − χ Eµν χαβ + (a2 + 3a3 qdS )φ!φ + χ τµν . 3 2 Λ3 MPl (54) ItDecoupling is interesting tolimit point implies out again that the helicity-0 fluctuation φ then decouples existence of inhomogenous from matter sources at quadratic order (however the coupling reappears at the cubic cosmological solutions for massive gravity in Minkowski order). Stability of this solution is therefore ensured if the parameters satisfy one of (dRGT) for suitable of parameters of free the followingwhich three constrains, (settingrange a1 = −1/2 and λ̃ > 0) from Higuchi bound a2 < 0 and Remarkable or 1 − 3a2 λ̃ − (1 − 2a2 λ̃)3/2 2a22 ≤ a3 < − , 2 3 3λ̃ (55) helicity zero does not couple to matter perts no vDVZ discontinuity 1 − 3a λ̃ + (1 − 2a λ̃)3/2 1 a2 < Absence or 2λ̃ and a3 > 2 2 3λ̃2 , of Higuchi bound frees up possibility for background Vainshtein1effect - consistency with known 2 2 a2 ≥ and a3 > − a2 . 3 cosmology 2λ̃ (56) (57) ce metric, once observations are taken into account the bound rem nt. We hint to a possible resolution in the Conclusions. Our final bo Or .... Bigravity ! " 2 H H 2 H + 3αmetrics 2(1 + 2α + α ) + (α + α ) H) Now = m make(3both 3 + α4 ) − 3 4 3 4 dynamical, H0 H0 H02 meaning add EH term for f metric 2p 2p M H is the Hubble rate of the dynamical metric and H that of the refer MP 0 f 2 L = g R(g) + 2m U (g, f ) + f R(f ) ical) metric, 2and α3 and α4 are the two free parameters in the dR 2 hisFriedman should beunchanged compared with the associated Friedmann equation 2 m 1 2 H ρ − (6 + 4α3 + α4 ) + (3 + 3α3 + α4 )m − H = 2 3MPl 3 H0 2 2 3 H m H (1 + 2α3 + α4 )m2 2 + (α3 + α4 ) . 3 H0 3 H0 2 H = H0 1 we still obtain b H = a H0 s the Hubble rate of the dynamical metric and H0 that of the referenc ) metric, and α3 and α4 are the two free parameters in the dRGT Bigravity HiguchiFriedmann bound equation should be compared with the- associated 2 m 1 2 H 2 ρ − (6 + 4α3 + α4 ) + (3 + 3α3 + α4 )m − H = 2 3MPl 3 H0 2 2 3 H m H 2 (1 + 2α3 + α4 )m 2 + (α3 + α4 ) . 3 H0 3 H0 b H = a H0 H = H0 1 – 2 – Higuchi bound is now 2 mdressed (H) 2 H + 2 2 H 0 MP Mf2 ! 2H 4 Bigravity - Higuchi bound Higuchi bound is now 2 mdressed (H) suppose 2 H + 2 2 H 0 MP Mf2 ! 2H 4 H ⌧ H0 then the f-metric Friedman equation gives suppose 3H02 Mf2 ⇡ (3 + 3↵3 + 2 mdressed 3 2 2 H0 ↵4 )m MP 3 H 2 ⇡ (3 + 3↵3 + ↵4 )m 2 p H Mf H0 ⇠ 3 mdressed MP 2 MP H H0 Bigravity - Higuchi bound Higuchi bound is now 2 mdressed (H) H0 ⇠ p 2 H + 2 2 H 0 MP Mf2 ! 2H 4 H 2 Mf 3 mdressed MP Higuchi bound is approximately 4 4 ⇠ 3H 2H but since 3 > 2 bound is automatically satisfied!!!!! (as long as H ⌧ H0 ) Note that in the MG decoupling limit Mf ! 1 we recover a problem Summary • FRW (fully homogeneous and isotropic) solutions are a problem in Massive Gravity • For Partially Massless Gravity - Higuchi bound is automatica$y satisfied for any choice of matter • For Massive Gravity on a fixed reference metric, bound is in conflict with Vainshtein mechanism • For Bigravity, bound is almost always satisfied regardless of the choice of matter as long as H ⌧ H0 • Generalized Higuchi bound is insensitive to equation of state for matter i.e. Ḣ making it more stringent than previously expected