The relation between general relativity and a class of Hořava gravity theories

Nicola Franchini, Mario Herrero-Valea, and Enrico Barausse

SISSA, Via Bonomea 265, 34136 Trieste, Italy and INFN Sezione di Trieste and
IFPU - Institute for Fundamental Physics of the Universe, Via Beirut 2, 34014 Trieste, Italy

Violations of Lorentz (and specifically boost) invariance can make gravity renormalizable in the
ultraviolet, as initially noted by Hořava, but are increasingly constrained in the infrared. At low
energies, Hořava gravity is characterized by three dimensionless couplings, α, β and λ, which vanish
in the general relativistic limit. Solar system and gravitational wave experiments bound two of
these couplings (α and β) to tiny values, but the third remains relatively unconstrained (0 ≤ λ .
0.01 − 0.1). Moreover, demanding that (slowly moving) black-hole solutions are regular away from
the central singularity requires α and β to vanish exactly. Although a canonical constraint analysis
shows that the class of khronometric theories resulting from these constraints (α = β = 0 and λ 6= 0)
cannot be equivalent to general relativity, even in vacuum, previous calculations of the dynamics
of the solar system, binary pulsars and gravitational-wave generation show perfect agreement with
general relativity. Here, we analyze spherical collapse and compute black-hole quasinormal modes,
and find again that they behave exactly as in general relativity, as far as observational predictions are
arXiv:2103.00929v2 [gr-qc] 26 Mar 2021

concerned. Nevertheless, we find that spherical collapse leads to the formation of a regular universal
horizon, i.e. a causal boundary for signals of arbitrary propagation speeds, inside the usual event
horizon for matter and tensor gravitons. Our analysis also confirms that the additional scalar degree
of freedom present alongside the spin-2 graviton of general relativity remains strongly coupled at
low energies, even on curved backgrounds. These puzzling results suggest that any further bounds
on Hořava gravity will probably come from cosmology.

I. INTRODUCTION The action is written in terms of a preferred spacetime

foliation described by T = const, and the metric has been
Lorentz symmetry is one of the cornerstone of our un- decomposed in the 3+1 form1
derstanding of theoretical physics, and has been tested
ds2 = N 2 dT 2 − γij dxi + N i dT dxj + N j dT , (3)



to exquisite precision in particle physics experiments [1–
4]. Bounds on Lorentz violations (LVs) in gravity are
however much weaker [5–7]. This is particularly inter- where we recognise a lapse function N , a shift three-
esting because violations of boost symmetry in gravity vector Ni and the spatial three-metric γij . Also defined
may allow for constructing a theory of quantum gravity in terms of this decomposition are the other quantities
that is power counting (or even perturbatively) renor- appearing in the action, e.g. the determinant of the three-
malizable in the ultraviolet [8, 9]. This proposal, ini- metric γ; the extrinsic curvature of the foliation,
tially put forward by Hořava [8], may still pass particle
1
physics tests of Lorentz symmetry if a mechanism is in- Kij = − (∂T γij − Di Nj − Dj Ni ) , (4)
cluded to prevent “percolation” of large LVs from gravity 2N
to matter. Among such putative mechanisms are renor- where the covariant derivative Di is defined with re-
malization group flows (whereby Lorentz invariance may spect to γij ; the three-dimensional Ricci scalar (3) R;
be recovered, at least in matter, in the infrared) [10–13], K = Kij γij ; and ai = ∂i ln N . With Ψ we refer here
accidental symmetries allowing for different degrees of to standard matter fields, which couple to the full four-
LVs in gravity and matter [14], or the suppression of LVs dimensional metric gµν . This action can be obtained
in matter via a large energy scale [15]. from that of non-projectable Hořava Gravity [16] by ne-
The infrared limit of Hořava gravity, also known as glecting operators with more than two derivatives, rel-
khronometric theory, is characterized by three dimen- evant only at high energies. Note that by introduc-
sionless coupling parameters α, β and λ, in terms of ing a preferred foliation, Lorentz symmetry is broken at
which the theory’s action is [8, 16, 17] the local level. The action is invariant under foliation-
1−β √  1+λ 2 preserving diffeomorphisms (T → T̃ (T ), xi → x̃i (x, T ))
Z
S= dT d3 x N γ Kij Kij − K but not under full four-dimensional diffeomorphisms.
16πG 1−β
(1) The same action can be recast in covariant form by pro-
1 (3) α 
+ R+ ai ai + Smatter [gµν , Ψ] , moting the coordinate T to a (timelike) scalar field (the
1−β 1−β
“khronon”) and defining a unit-norm, timelike “æther”
in units where c = 1 (used throughout this article), and
where the bare gravitational constant G is related to the
one measured on Earth and in the solar system (GN ) by
G 1 From now on, Latin indices will run only over space directions,
GN = . (2) while Greek indices will also include time.
1 − α/2
 2

vector field orthogonal to the hypersurfaces of T = const, which parametrize violations of the strong equivalence
principle at the leading post-Newtonian (PN) order [27,
∇µ T 28, 31] – can be shown to vanish exactly in mHG [25,
uµ = √ α , (5)
∇ T ∇α T 30]. Therefore, no dipole GW emission from binaries of
compact objects is expected in mHG [25, 30], unlike for
where we assume a + − −− metric signature (as in the
generic α, β (where this effect was used to test the theory
following). This allows for writing the action as [18]
with binary pulsars [27, 28]).
1
Z
√ h A possible caveat regarding these experimental bounds
S=− d4 x −g R + λ (∇µ uµ )2 is that the khronon becomes strongly coupled around
16πG
i the Minkowski and Robertson-Walker geometries in the
+ β∇µ uν ∇ν uµ + α aµ aµ + Smatter [gµν , Ψ] , (6) mHG limit, since the scalar field T becomes non-
propagating (i.e. its speed diverges) when α, β → 0 and
where aµ ≡ uν ∇ν uµ . Here, LVs are made apparent by λ 6= 0 [23, 32]. We stress that strong coupling does not
the fact that the vector field u is timelike, i.e., according mean that the theory is not viable, but simply that the
to the definition (5), linearized calculations on the simple backgrounds men-
tioned above may provide incorrect results. However,
uµ uµ = 1. (7) since the strong coupling affects the khronon and not the
Although still weaker than in matter, LVs in grav- tensor sector, the linear calculation of the speed of GWs
ity are becoming increasingly constrained, especially by (used to compare to the GW170817 observations) is ex-
gravitational wave (GW) experiments. Bounds on the pected to provide trustworthy results.
propagation speed of GWs from GW170817 constrain As for the PN calculations of the solar system dynam-
|β| . 10−15 [19, 20], which paired with bounds from so- ics and GW generation, it should be noted that (i) the
lar system experiments also allows for constraining |α| . PN scheme is an expansion in powers of 1/c, and it thus
10−7 (with λ left unconstrained), or |α| . 0.25×10−4 and includes non-linear terms; and (ii) the Newtonian/PN
λ ≈ α/(1 − 2α) [21–25]. Measurements of the abundance dynamics is strongly coupled in GR as well, and yet it
of primordial elements produced by Big Bang Nucleosyn- gives meaningful results. Indeed, at leading (Newtonian)
thesis (BBN) constrain λ . 0.1 [26–28], with λ ≥ 0 re- order the gravitational field does not propagate in GR
quired to ensure absence of ghosts [16, 20]. These bounds (i.e. the equation describing it is elliptic), and propa-
therefore seem to suggest that α and β should be tiny, gation only appears at higher PN orders [21, 33]. We
while λ could still be sizeable. Indeed, an additional therefore expect the results from a PN expansion of the
theoretical constraint – namely that black holes moving field equations to remain valid also in the mHG limit.
slowly relative to the preferred foliation remain regular Given this wealth of (non-cosmological) observables for
except for their central singularity – would require α and which mHG provides the same predictions as GR, it is
β to vanish exactly [25]. natural to wonder whether mHG and GR may be equiv-
We will refer to the theory with α = β = 0 and alent, at least in some regimes. Obviously, a full equiv-
λ 6= 0 as minimal Hořava gravity (mHG) in the follow- alence between GR and mHG can be excluded, since
ing. Remarkably, all non-cosmological observables that the cosmological expansion history is different in the two
have been computed in Hořava gravity reduce to their theories (a fact that is used to constrain λ with BBN
GR counterparts in the mHG case. For instance, the dy- data [27, 28]), but it may hold in more specific settings.
namics in the solar system (i.e. at first post-Newtonian For instance Refs. [34, 35], based on a constraint analysis
order) exactly matches that of GR [23, 24]. GWs also of mHG, claimed that the theory may be equivalent to
propagate exactly at the speed of light [23]. Moreover, GR in vacuum and under asymptotically flat boundary
static spherically symmetric black holes are described by conditions. While suggestive in the light of the “coinci-
the Schwarzschild metric [29], and so are those moving dences” presented above, this conclusion disagrees with
slowly relative to the preferred foliation [25]. The same that of Ref. [36], which solved the (tertiary) constraint
applies to stars, for which both static spherically symmet- equation of mHG and showed explicitly that the theory
ric solutions and ones describing slowly moving bodies are cannot be equivalent to GR unless N = 0 (in which case
characterized by the same (GR) geometry [30]. Note that the metric is degenerate).
for both stars and black holes the khronon configuration In this work, we will therefore attempt to identify
is non-trivial, but does not backreact on the geometry in (non-cosmological) astrophysical observables for which
mHG. This is quite surprising – because objects at rest mHG may differ from GR, focusing on fully non-
and in motion are expected to be described by the same perturbative calculations, or on ones that involve pertur-
metric only in a Lorentz-symmetric theory such as GR, bations over backgrounds different from the Minkowski
and not (a priori) in a theory with LVs – and has im- and Robertson-Walker geometries (on which the khronon
plications also for the dynamics of binaries of compact is strongly coupled). In more detail, in Sec. II we will
objects and for GW generation. review linear perturbations of mHG on flat space. We
Indeed, since the geometry of slowly moving stars and will then study the non-linear dynamics of spherically
black holes is the same as in GR, the “sensitivities” – symmetric collapse (in Sec. III), showing that a uni-
 3

qi
versal horizon (i.e. a boundary for signals of arbitrary h0i = Zi + s4 , (13)
speeds) [37, 38] naturally forms inside the usual horizon q
for tensor gravitons and matter. Nevertheless, the col- where Xi , Yi and Zi are divergenceless vectors – i.e.
lapse is completely indistinguishable from GR as far as Di X i = Di Y i = Di Z i = 0; si , v0 and h00 are scalars;
observable quantities are concerned. In Sec. IV we will and ζij is a transverse-traceless tensor – thus satisfying
then derive the equations for linear metric perturbations Di ζ ij = Dj ζ ij = ζii = 0.
over static spherically symmetric black holes, and show The two constraints (7) and (8) kill three of these de-
that they also coincide with the GR ones, when focusing grees of freedom, as previously mentioned. At the linear
on the tensor modes. The scalar mode remains instead level, they impose
strongly coupled (like in flat space), but decouples from
the tensor sector. Our conclusions are finally presented h00 + 2v0 = 0, ǫijk q j Y k = 0, (14)
in Sec. V.
where the second of these conditions is satisfied by set-
ting Y k = 0. Once these conditions are enforced, the
II. KHRONOMETRIC THEORY AROUND momentum space Lagrangian for the perturbations, re-
FLAT SPACE taining only quadratic terms, becomes

The dynamics of the action (6) is described in terms L = − 81 (λ + β)ω 2 s21 − 21 (1 + λ)ω 2 s21 − λqωs1 s3
of the metric gµν and the æther vector u. The latter is + 41 −q 2 − (1 + β + 2λ)ω 2 s21 − 12 (λ + β)qωs1 s3


constrained to be unit-norm and timelike [cf. Eq. (5)]
+ 21 −(λ + β)q 2 − αω 2 s23 + qωs1 s4 − αω 2 s3 s4 (15)


and hypersurface orthogonal, i.e. it must have, from the
− 21 αω 2 s24 + 81 q 2 + (1 − β)ω 2 ζab ζ ab − 12 qωX a Za


Fröbenius theorem, zero vorticity
+ 41 (1 − β)ω 2 Xa X a − αqωs4 v0 − 12 αq 2 v02 − αqωs3 v0
u[µ ∇ν uσ] = 0 . (8)
+ 41 (1 − β)q 2 − 2αω 2 Za Z a + q 2 s1 v0 ,


(16)
Since the theory breaks boost invariance at the local
level, it should propagate additional degrees of freedom where we have omitted a global factor of G.
besides the usual spin-2 graviton field hµν of GR. In- We are left with the task of choosing a suitable gauge.
deed, a generic four-dimensional vector u contains four Since the action (6) from which we started is covari-
degrees of freedom – which can be arranged into a three- ant (being related to the “unitary gauge” action (1) by
dimensional divergence-less vector and two scalars. How- a Stuckelberg transformation), we need to choose four
ever, the unit norm (7) and vorticity (8) conditions elim- gauge conditions. These can be given as the requirement
inate three of these degrees of freedom, leaving a single that two scalars and one of the three-dimensional diver-
scalar behind (corresponding obviously to the khronon genceless vectors vanish. We choose s1 = v0 = X a = 0,
scalar field T defining the preferred foliation). and replacing these conditions (as well as those that fol-
This can be seen directly at the level of the action by low from the equations of motion of these fields) back in
perturbing both the metric and æther around flat space: the action, we obtain

gµν = ηµν + hµν , uµ = (1, 0) + v µ . (9) 1−β

ζab ω 2 − c22 q 2 ζ ab


L=
8
Replacing this into the action (6), we first go to mo- (β − 1)(λ + β)(2 + 2λ + β)
mentum space – where ∂t ≡ iω and Di ≡ iqi , with ω + s̃(ω 2 − c20 q 2 )s̃ , (17)
4(1 + λ)2
and qi the frequency and three-momentum respectively
– and perform a 3 + 1 decomposition adapted to the foli- where we have also rescaled the remaining scalar as s̃ =
q
ation orthogonal to the background æther (1, 0), i.e. we ω s3 . This is the Lagrangian of two modes propagating
decompose the metric perturbation and æther as with speeds [23, 39]
1
 
h00 h0i c22 = , (18)
hµν =   , vµ = (v0 , vi ). (10) 1−β
h0i hij (λ + β)(2 − α)
c20 = . (19)
α(1 − β)(2 + 3λ + β)
We then split the various quantities in modes that trans-
form as scalars, vectors and tensors under rotations As previously mentioned, we find an extra propagating
  scalar field with velocity c0 , besides the usual transverse-
qi qj qi qj qi qj traceless graviton with velocity c2 . Both propagation
hij = ζij + Xj + Xi + 2 s1 + δij − 2 s2 ,
q q q q velocities can be different from the speed of light, al-
(11) though the coincident observation of GW170817 and
qi GRB 170817A constrains c2 to match c to within about
vi = Yi + s3 , (12) 10−15 (which in turns bounds |β| . 10−15 ). As for the
q
 4

scalar mode, cosmic ray observations require c0 & 1, be- present spin-0, spin-2 and matter horizons at (generally)
cause otherwise ultrahigh energy particles would lose en- distinct locations.
ergy to the khronon in a Cherenkov-like cascade [40]. Even more worryingly, when terms of higher order in
Superluminality is of course not surprising, since the the (spatial) derivatives are included in the infrared ac-
theory is not boost-invariant, and thus c = 1 is not a uni- tion (1), Hořava gravity predicts that the dispersion re-
versal maximum speed. However, although in general the lations (for both the gravitons and matter) will take the
scalar velocity (19) is finite, it diverges in the mHG limit, form ω 2 = c2i q 2 +Aq 4 +Bq 6 , with A and B constant coef-
if λ 6= 0. This is a signal that the linearized expansion ficients and ci the species infrared phase velocity. There-
breaks down for the dynamics of the scalar field, which fore, the group velocity dω/dq of all species will diverge
is then out of reach of perturbative techniques, while the in the ultraviolet limit, which questions whether it makes
tensor mode remains healthy. The same conclusion is sense to talk about event horizons at all. The problem is
achieved by performing an identical expansion around even more evident in mHG, where the spin-0 propagation
FRW space-times or around any maximally symmetric speed diverges already in the infrared limit (cf. Sec. II).
spacetime [41]. Note also that this potentially problem- However, an unavoidable requirement for any physical
atic behaviour of the scalar field only appears at low en- modes is that they propagate in the future, as defined
ergies. At higher energies the action must be extended by by the preferred foliation. Therefore, the topology of the
operators with higher number of derivatives, which de- hypersurfaces of constant khronon plays a crucial role in
form the dispersion relations and lead to a healthy prop- defining the spacetime’s causal structure. In the infrared
agating scalar mode. black hole solutions of [37, 38], there exists indeed a a
In light of this “strong-coupling problem” for the hypersurface of T = const that is also a hypersurface of
khronon on flat space, we pursue in the following two dis- constant radius. Once inside this hypersurface – which
tinct calculations in mHG, namely spherically symmet- was called the “universal horizon” in [37, 38] – no modes
ric gravitational collapse and linear perturbations over can escape, even if they propagate at infinite speeds, sim-
spherically symmetric static black hole spacetimes. We ply because they need to move in the future direction de-
will aim to assess whether the khronon dynamics remains fined by the background preferred foliation. Note that in
strongly coupled when non-linearities are included in the the special case of mHG, this universal horizon coincides
equations of motion, or when spacetimes more general with the spin-0 horizon, since the khronon propagation
than Minkowski space are considered speed diverges already in the infrared limit.
Despite their attractive features, as mentioned above,
black holes are not the only static and spherically sym-
III. SPHERICAL COLLAPSE metric solutions of khronometric theory. Indeed, generic
values of the æther tilt parameter yield solutions that
are singular at the spin-0 horizon. In particular, if the
Unlike in GR, Birkhoff’s theorem does not hold in
tilt parameter is such that the æther does not present any
khronometric theories, and vacuum spherically symmet-
radial component at spatial infinity, that component van-
ric solutions (even when one imposes that they are static
ishes throughout the entire spacetime (i.e. the æther is
and asymptotically flat) are not unique [37, 38, 42–44]. In
always parallel to the timelike Killing vector), and the re-
more detail, in a given khronometric theory, there exists
sulting solutions describe the exterior spacetime of static
a two-parameter family of static, spherically symmetric
spherically symmetric stars (whose matter “covers” the
and asymptotically flat vacuum solutions. One of the
singularity at the spin-0 horizon) [43, 44].
parameters characterizing these solutions is (like in GR)
For concreteness, let us examine the special case of
their mass, while the second parameter regulates the ra-
mHG, where these static and spherically symmetric vac-
dial tilt of the æther near spatial infinity [37, 38, 42–44].
uum solutions can be obtained analytically and read [29]
In particular, for a given mass, only a specific value of
this second parameter yields solutions that are regular B(r)2 2
everywhere except for the central r = 0 curvature sin- ds2 = f (r)dt2 − dr − r2 dΩ2 , (20)
f (r)
gularity. These are the solutions that are expected to
1 + f (r)A(r)2
 
B(r) 1
form in gravitational collapse [45] and which are usually uα dxα = dt + − A(r)2 dr ,
referred to as “black holes” in the literature [37, 38, 42]. 2A(r) 2A(r) f (r)
Although the geometry of these black holes is similar (21)
to that of the Schwarzschild solution of GR (with which where
it actually coincides exactly in the mHG limit), the ex-
istence of the khronon mode has profound implications 2GN M
f (r) = 1 − , B(r) = 1 , (22)
for their causal structure. As shown in Sec. II, at low r !
r
energies Hořava gravity propagates both spin-2 and spin- 2 4
1 ræ ræ
0 gravitons, whose speeds are generally different from c A(r) = − 2 + f+ 4 (23)
f r r
(i.e. the limiting speed for matter modes). As a result,
different causal boundaries exist for spin-0, spin-2 and The two parameters characterizing each solution are the
matter modes, i.e. black holes in khronometric theories mass M and the “radial tilt” ræ . These solutions are
 5

singular at the universal horizon (which in mHG coin- (where p, ρ and U are the fluid’s pressure, energy density
cides with the spin-0 horizon, as mentioned above) unless and four-velocity) is covariantly conserved (∇µ T µν = 0),
ræ = 33/4 GN M/2 [29]. The latter value describes instead since in Eqs. (1) and (6) matter couples only to the four-
a black hole with a regular universal horizon (located at dimensional metric gµν . The explicit form of H, Hi and
areal radius rU = 3GN M/2), while ræ = 0 describes a E ij is given in the Appendix A (see also [24]) for generic
static æther u ∝ ∂t . khronometric theories, from which the mHG equations
Note that Eqs. (20)–(23), whatever the value of ræ , can be obtained by setting α = β = 0.
yield ∇µ uµ = 0. If we now express the class of solutions
given by Eqs. (20)–(23) in the unitary gauge (where the To simplify the algebra, let us choose spatial coordi-
khronon is used as the time coordinate T ), the unit-norm, nates on the preferred slices such that N i = 0. Unlike in
future directed æther vector u becomes orthogonal to the GR, however, the lapse is not a gauge field, i.e. we have
preferred foliation T = const, which therefore presents already chosen our time coordinate to be the khronon
K = ∇µ uµ = 0. Therefore, Eqs. (20)–(23) yield the when writing the action (1), and thus no further condi-
Schwarzschild geometry foliated in maximal (preferred) tions can be imposed on N . The most generic ansatz
slices K = 0, which have long been studied in the context that we can write in spherical symmetry is therefore
of numerical relativity [46–51].
In order to ascertain which solution, in the class de-
scribed by Eqs. (20)–(23), is produced as the end-point
of gravitational collapse, let us consider the equations γij dxi dxj = A(T, R)dR2 + R2 B(T, R)dΩ2 , (25)
of motion for time dependent configurations, which can N = Z(T, R). (26)
be obtained by varying the action (1). Variation with
respect to the shift yields the momentum constraint
Hi = 0. Variation with respect to the lapse yields an
equation H = 0 that reduces to the GR energy con- For the matter, we assume that ρ and p are also functions
straint when α, β, λ → 0, but which is not a priori a of T and R alone, and that the three-velocity is radial, i.e.
constraint equation in khronometric theory. In fact, for only U R (T, R) and U T (T, R) are non-zero (and related
generic α, β, λ the resulting equation is not a constraint, by the normalization condition Uα U α = 1). Note that
but corresponds to the khronon’s evolution equation in the class of metrics given by Eqs. (20)–(23) can be easily
the covariant formalism of action (6). Finally, by varying put in the form of Eq. (25) by performing first a coordi-
with respect to γij one obtains the evolution equations nate transformation t = T + H(r) (to go to the unitary
E ij = 0. We describe matter by a perfect fluid, whose gauge) and then a further (time dependent) coordinate
stress-energy tensor transformation r = RB(T, R)1/2 to eliminate the shift
and render the metric diagonal. More explicitly, at large
T µν = (ρ + p)U µ U ν − pg µν , (24) radii the solution given by Eqs. (20)–(23) yields

−2k12 + 2k1 + 2k2 − 1 G3N M 3

GN M (2k1 − 1) G2N M 2
Z =1− + + +
R 2R2 2R3
2 4  5
+ G4N M 4 8k13 − 12k12 + 8k2 − 4k1 (4k2 − 3) − 5
 
−8GN M T ræ + 4ræ 1
+ + O , (27)
8R4 R
 4
2 (k1 + k2 − 2) G2N M 2 2
+ G3N M 3 2k12 − 8k1 − 6k2 + 8
 

2GN M 4T ræ 1
A =1+ − + +O , (28)
R R2 R3 R
 4
k 2 + 2k2 G2N M 2


2k1 GN M 2k1 k2 G3N M 3 − 2T ræ
2
1
B =1+ + 1 + + O , (29)
R R2 R3 R

where k1 and k2 are free parameters entering in the co- and B are time dependent, the dependence on time ap-
ordinate transformation.2 Note that even though Z, A pears at sub-leading order in 1/R. Moreover, the trace of

2 Indeed, our ansatz (25) does not complete fix the gauge, as it persurfaces, but merely ensures that once a set of spatial coordi-
is invariant under a time-independent redefinition of the radius. nates is chosen on some initial T = constant hypersurface, then
This residual gauge freedom arises because choosing Ni = 0 does the spatial coordinates are fixed in the whole spacetime. The
not completely fix the spatial coordinates on T = constant hy- choice of coordinate on the initial slice, however, is arbitrary.
 6
√
the extrinsic curvature vanishes, as it did in the original and γ ≈ 1 + Aγ exp[iωγ (T − R)]/R (with AN , Aγ , ωN ,
foliation given by Eqs. (20)–(23), since K = ∇µ uµ is a and ωγ free coefficients), then at large R we find
scalar under four-dimensional diffeomorphisms.
In order to have only first order equations for our sys- Aγ ωγ2
∂R K ≃ − exp[iωγ (T − R)]. (32)
tem, let us then introduce KA ≡ KR R = −∂T A/2AZ, R
KB ≡ Kθ θ = Kϕ ϕ = −∂T B/2BZ, DZ ≡ ∂R log Z, Requiring that ∂R K = 0 implies Aγ = 0 and thus K = 0.
DA ≡ ∂R log A and DB ≡ ∂R log B. With these vari- Finally, let us note that if K = 0 at all times, the
√
ables, K = −∂T ln γ/N = KA + 2KB . spherical collapse equations and the constraints become
With our ansatz, the non-trivial field equations are identical to the GR ones, written in the maximal slicing
H = HR = E RR = E θθ = 0. As mentioned above, H = 0 gauge K = 0 [and in our zero-shift ansatz (25)]. Since
becomes the energy constraint in the GR limit, but is K = ∇µ uµ (with u the unit-norm future-directed vec-
not generically a constraint in khronometric theory. To tor orthogonal to the foliation) is a scalar under four-
check whether H = 0 is a constraint in mHG, let us con- dimensional diffeomorphisms, one can then transform the
sider the time derivative of H. By using the equations of spherical collapse equations of mHG into those of GR
motions to simplify the expressions, one obtains with maximal time-slicing K = 0, but more general spa-
tial coordinates (i.e. ones yielding general non-vanishing
λ  2

shift).3
∂T H = − (2 + R D)∂R K + r ∂R K , (30)
R AZ The maximal time-slicing gauge has been extensively
used in GR to study gravitational collapse, as it allows
where D = 2DZ −DA /2+DB . Eq. (30) vanishes either in
for penetrating the black hole horizon [46–51]. One can
the GR limit λ = 0, or when the quantity within brackets
therefore utilize the results of GR simulations (either per-
is zero.
formed in the maximal slicing gauge, or transformed to
Barring the case λ = 0, one therefore has to solve
that gauge a posteriori) to gain insight on spherical col-
∂T H = 0 (which follows from the original field equation
lapse in mHG.
H = 0) at each time T . Actually, the generic solution to
Indeed, GR collapse simulations in the maximal time-
∂T H = 0 is simply
slicing found that there exists a “limiting slice”, i.e. a
A1/2 limiting hypersurface that the maximal slices approach at
∂R K = C(T )2 , (31) late times [46–51]. In more detail, the slicing that arises
R2 BZ 2 in these simulations outside the collapsing sphere turns
where C(T ) is an integration constant. In a gravitational out to be described by the unit-norm future-directed vec-
collapse, e.g. of a star, one requires regularity at the cen- tor u given by Eq. (21), with the parameter ræ asymp-
ter of the coordinates to obtain a physically meaningful totically approaching the critical value 33/4 GN M/2. The
solution [50, 51]. Necessary conditions for regularity are limiting slice is therefore defined by areal radius r =
that A, Z, B are finite (and non-vanishing) at R = 0, and 3GN M/2 [46–51].
that K and its radial derivative also remain finite at the While in GR the foliation of the spacetime in time
center. From Eq. (31), it is therefore clear that the only slices has no physical meaning (as it is merely a coordi-
way to impose regularity at r = 0 is to set C(T ) = 0 for nate effect), the slicing has instead an important physical
any T , i.e. the extrinsic curvature K must be constant meaning in mHG, since we are using the unitary gauge,
on any given spatial foliation, i.e. K(T, R) = k(T ). Note where the time coordinate coincides with the khronon
that a spatially constant trace was also expected from scalar field. Indeed, the appearance of the limiting slice
the Hamiltonian analysis of [35, 52]. r = 3GN M/2 in the GR maximal-slicing collapse simu-
If k(T ) 6= 0, by exploiting time-reparametrization in- lations corresponds to the formation of a universal hori-
variance one can set k(T ) = 1. The evolution equations zon in mHG.4 This can be understood because in spher-
E RR = E θθ = 0 and the momentum constraint HR = 0
then take the same form as in GR, whereas the equa-
tion H = 0 [which is now a bona fide Hamiltonian con- 3 Note that we are changing the spatial coordinates to reinstate
straint since C(T ) = 0] contains a term proportional to the shift, but not the time coordinate, which still coincides with
λ. However, as noted e.g. by [35], if asymptotically flat the khronon field (unitary gauge).
4 The correspondence between the appearance of a limiting folia-
boundary conditions are assumed, at spatial infinity one
tion in GR in the maximal time slicing gauge and the formation
must necessarily have K = k(T ) = 0 at all times [cf. of a universal horizon was also noticed in Ref. [54] for the case of
also Eqs. (27)–(29)]. Boundary conditions at spatial in- Cuscuton theories. The latter are indeed equivalent to mHG if
finity that are not necessarily flat, but which are time- their scalar potential is quadratic [55, 56] (even though the equiv-
independent, will yield also K = k(T ) = 0 at all times alence is subtle when it comes to the hypersurface orthogonality
condition (8), as a result of which suitable boundary conditions
(see e.g. [53] for an example of one such GR collapse
are required to obtain an exact equivalence between the two the-
solution). ories [55]). However, Ref. [54] worked in the decoupling limit
Similarly, outgoing boundary conditions at infinity also (i.e. neglecting the backreaction of the Lorenz-violating field on
imply K = k(T ) = 0 at all times. This can be seen by the geometry) and with zero potential (in which case the theory
noting that if one imposes N ≈ 1+AN exp[iωN (T −R)]/R is not equivalent to mHG).
 7

kh
ical symmetry the universal horizon is, by definition, the where Gµν = Rµν − gµν R/2 and Tµν contains the con-
outermost hypersurface r = const that is also orthog- tribution from the khronon:
onal to u (or equivalently, the outermost hypersurface
kh 1
r = const that is also a hypersurface of constant khronon Tµν = uµ uν uσ ∇σ ∇ρ uρ − gµν ∇ρ uρ + gµν (∇σ uσ )2
T = const). We can therefore conclude that spherical col-   2
ρ ρ σ
lapse in mHG produces “regular” black holes, i.e. ones + 2 g(µ uν) − u(µ uν) u ∇ρ ∇σ u . (34)
described by Eqs. (20)–(23). In particular, no singularity
forms at the spin-0/universal horizon. Variation of the khronon field T yields a scalar equation
We stress, however, that an analysis of the principal kh
that is equivalent to the covariant conservation of Tµν ,
part of the fully non-linear spherical collapse equations already a consequence of (33) [18]. As such, it does not
in generic khronometric theories, which we present in need to be independently enforced if all of the ten com-
Appendix B, shows that the characteristic speed of the ponents of Eq. (33) are satisfied. However, we show it
scalar mode diverges in the mHG limit. This suggests here for completeness:
that the effect of the khronon on spherical collapse in  µν
(g − uµ uν ) (∇ν ∇σ uσ )

mHG may vanish simply because it satisfies an elliptic
equation. It is therefore unclear if gravitational collapse κ ≡ λ∇µ √ = 0. (35)
∇α T ∇α T
will be the same as in GR when the assumption of spheri-
cal symmetry is relaxed. To partially tackle this problem, Let us now perturb the metric and khronon fields
as well as to assess if moving away from flat space can fix around a curved background geometry, characterized by
the strong coupling of the khronon reviewed in Sec. II, the pair {gµν , uµ }:
in the next section we will consider linear, but otherwise
generic, perturbations of black holes in mHG. gµν = g µν + ǫ hµν + O(ǫ2 ),
uµ = uµ + ǫ vµ + O(ǫ2 ) , (36)
IV. QUASI-NORMAL MODES
where ǫ is a perturbative parameter (which sets the
amplitude of the perturbations of the metric and
Linear gravitational perturbations of black hole space- æther/khronon, which needs to be small for the linear
times in GR have been studied for decades, since the theory to be a good approximation). In the following,
seminal work by Regge, Wheeler and Zerilli for the to keep the analysis more general, we will simply assume
Schwarzschild geometry [57, 58] and by Teukolsky for the a spherically symmetric and static background, given by
Kerr one [59]. The frequency spectrum of these pertur- the ansatz of Eqs. (20)–(21). To restrict to a black hole
bations, once ingoing/outgoing boundary conditions are background, one can then simply assume the validity of
imposed at the event horizon/far from the black hole, is Eqs. (22)–(23), with ræ = 33/4 GN M/2.
discrete and consists of complex frequencies. Since the Inserting Eq. (36) into Eq. (33) and expanding to lin-
imaginary part of the latter is such that the spectrum ear order in ǫ, we obtain the equations of motion for the
modes are exponentially damped (thus pointing, in par- perturbations in covariant form,
ticular, to linear stability of the Schwarzschild and Kerr
solutions, at least for non-extremal spins), these modes E µν + ǫ δEµν + O(ǫ2 ) = 0, (37)
are usually referred to as quasi-normal modes (QNMs).
Interestingly, because the Kerr geometry can only de- where E µν = 0 is automatic from the choice of back-
pend on two “hairs” [60–63] (mass and spin5 ), the QNMs ground. From now on we will drop the O(ǫ2 ) symbol
frequencies are found to only depend on the same two everywhere for notational clarity. Note that the æther
quantities. This observation has long prompted sug- field enters Eq. (34) both with upper and lower indices.
gestions to use QNM observations to test the no-hair This implies that even if we set vµ = 0, we do not triv-
theorem and thus GR [65, 66], a proposal that the ially recover the same equations for the perturbation as
LIGO/Virgo collaboration is starting to tentatively ap- in GR, since there are still non-negligible contributions
ply to real data [67–70], even though really constraining to δEµν coming from uµ ≈ uµ + ǫv µ − ǫuν hµν . Note that
tests will probably have to wait for future detectors [71]. this signals that the gravitational perturbations “feel”
In order to compute QNM frequencies in mHG, let us the presence of a background violating Lorentz invari-
start from the equations of motion in vacuum derived ance through the presence of the preferred foliation.
from the covariant action (6). From variations of the Since the background {gµν , uµ } is spherically symmet-
metric, one obtains ric, it is convenient to expand the perturbations in spin-
kh weighted spherical harmonics. Using the standard Regge-
Eµν = Gµν − λTµν = 0, (33)
Wheeler gauge [57] for the metric perturbations and per-
forming a Fourier transform in the time coordinate (ex-
ploiting the staticity of the background), we obtain
5 The electric charge is believed to be zero or extremely small for
hµν = e−iωt heven odd


astrophysical black holes [64]. µν + hµν sin θ ∂θ Pℓ (cos θ), (38)
 8

where Pℓ (x) is the ℓ-th Legendre polynomial, with ℓ the where Λ = ℓ(ℓ + 1). Replacing the black hole metric
angular momentum eigenvalue, and [Eq. (22)], this reduces exactly to the potential found in
  GR for the same Schwarzschild solution. We thus con-
f (r)H0ℓ (r) H1ℓ (r) 0 0 clude that no differences from GR arise in the equations

 H ℓ (r) H2ℓ (r)
 for odd perturbations, nor in the QNM frequencies in this
1 f (r) 0 0 
heven
µν = 

,
 parity sector.
2 ℓ

 0 0 r K (r) 0 
 In the even sector, the manipulation of the equations
2 ℓ 2 gets more complicated as they now involve the khronon
0 0 0 r K (r) sin θ
perturbations as well. The full derivation of the equations
(39) presented below is shown in Appendix C, as it is rather
lengthy and not particularly enlightening. In summary,
  the system is reduced to two second-order equations, for
0 0 0 hℓ0 (r)
  φ3 (r) and for an additional variable Ψ(r) defined in Ap-
 0 0 0 hℓ1 (r) pendix C. These two modes represent the perturbations
hodd = . (40)
 
µν
 0 of the khronon and metric, respectively. The equation
0 0 0 
  for Ψ decouples and reads
hℓ0 (r) hℓ1 (r) 0 0
d2 Ψ  2 
+ ω − Veven (r) Ψ = 0, (44)
Here, without loss of generality (thanks to spherical dr∗2
symmetry), we have set the azimuthal number m =
0. The functions H0 (r), H1 (r), H2 (r), K(r), h0 (r) and with the potential
h1 (r), where we have dropped the index ℓ to keep the 
notation compact, characterize the radial profile of the f
Veven = 2 (1 + Λ) (Λ(Λ − 2) + 3)
degrees of freedom of the metric perturbations. The per- r (1 + Λ − 3f )2
turbation of the æther vµ depends on that of the khronon

2
 
field T . If we make this explicit in the equations, the ex- − 3f (1 + Λ) + 3f (f − 1 − Λ) . (45)
pressions quickly become very cumbersome. Instead, and
equivalently, we choose to write a generic æther vector This is a wave equation (in Fourier space) and again it
perturbation agrees exactly with the GR result [58], once specialized to
  the black hole background [Eq. (22)]. Therefore, the even
vµ = φℓ1 (r), φℓ2 (r), 2ut φℓ3 (r)∂θ , 0 Pℓ (cos θ)e−iωt , (41) QNM frequencies for the metric perturbations coincide
with their GR counterparts.
The equation for φ3 (r), however, remains coupled to
where the factor of ut is chosen for convenience, since
Ψ(r), which enters as a source
it makes the resulting equations simpler. Imposing here
the unit-norm and hypersurface-orthogonality conditions φ′′3 (r) + W1 (r)φ′3 (r) + W0 (r)φ3 (r) = j(r), (46)
[Eqs. (7) and (8)] expanded at linear order in ǫ allows one
to eliminate two of the three free functions appearing in with
Eq. (41).
Focusing first on the odd part of Eq. (37), we find that j(r) = U1 (r)Ψ′ (r) + U0 (r)Ψ(r) (47)
there are only three potentially independent equations,
corresponding to δEtθ , δErθ and δEθφ . Notice that the and
perturbation of the æther, arising from the perturbation −4A4 f 2 + 2A2 (5f + 3) − 4 ω 2i − 2iA2 f


of a scalar field, has no odd contribution and therefore W1 (r) = 2 + ,
r (A2 f + 1) A2 f 2 + f
does not appear in the odd sector.
The function h0 (r) can be algebraically solved from the (48)

3
2
system and, after defining Q(r) ≡ f (r)h1 (r)/r, we find i(3f + 1)ω A2 f − 1 2 2
ω A f −1
that one of the remaining equations implies the other. W0 (r) = −
We are thus left with a single independent equation of f 2 r (A2 f + 1)3 f 2 (A2 f + 1)2
the Regge-Wheeler form [57] 4A2 Λ
− 2 . (49)
r2 (A2 f + 1)
d2 Q  2 
2
+ ω − Vodd (r) Q = 0, (42)
dr∗ The Ui (r) are (very complicated) functions of the ge-
ometry, the frequency ω and the angular momentum ℓ,
where we have introduced the tortoise coordinate in the
and explicit expressions for them are given in the Sup-
usual way, i.e. dr/dr∗ = f (r). The effective potential
plemental Material as Mathematica [72] files. We have
Vodd (r) reads
confirmed that this is a general result by looking at the
Λf (r) 2f (r)[f (r) − 1] f ′ (r)f (r) eigensystem of the generalized linear problem, when all
Vodd = + − , (43) equations are taken together: There is no (linear) change
r2 r2 r
 9

of variables which decouples the system into two indepen- V. CONCLUSIONS

dent differential equations.
Taking a closer look at our result, it may seem that the In this work, we have explored the possibility of find-
khronon field, which was strongly coupled around maxi- ing astrophysical signatures of theories of gravity that vi-
mally symmetric spaces (cf. Sec. II), is now propagating, olate Lorentz symmetry, focusing on the case of Hořava
since Eq. (46) has a potential W0 including ω 2 , which gravity. Because of existing experimental and theoretical
seems to indicate a finite propagating speed. However, constraints, we have enforced α = β = 0 in the action (1)
this is just an illusion due to a poor choice of variables, that describes the low energy limit of Hořava gravity, ob-
since the equation also contains terms proportional to taining a theory (for which we coined the name mini-
φ′3 (r). Performing a change of variables φ3 (r) = g(r)φ(r) mal Hořava gravity) depending on only one dimension-
and choosing g(r) to cancel all terms proportional to less coupling parameter λ, on which experimental bounds
φ′ (r), we get are relatively loose. We have focused on two phenomena
Rr
g(r) = C1 e 1
dz l(z)
, (50) that explore both the linear (around curved spacetime)
2 4 2 2
and non-linear dynamics of mHG, i.e., gravitational col-
iω A f − 1 2 + 2A f − A (3 + 5f ) lapse of spherically symmetric matter configurations and
l(r) = + , (51)
f A2 f + 1 r(A2 f + 1)2 the dynamics of the QNMs of the black holes produced
where C1 is an integration constant. The equation thus by the collapse.
becomes We have found that spherical collapse proceeds exactly
as in GR as far as the spacetime metric is concerned, but
φ′′ (r) − Vφ (r)φ(r) = j(r). (52)
that the khronon field undergoes a non-trivial dynamical
The new potential Vφ (r) has no term proportional to ω 2 , evolution. In more detail, we have found that the hyper-
whose contribution has been cancelled by that coming surfaces of constant khronon follow the same evolution as
from g(r). This corresponds to a field propagating with the maximal time slices K = 0 of GR, which are known
infinite speed (so that c−2φ = 0, with cφ the propaga- to asymptote to a limiting slice (corresponding to areal
tion speed). This is analog to the situation in flat space. radius r = 3GN M/2) in spherical collapse [46–51]. Al-
Therefore, we conclude that the khronon field remains though in the case of GR the appearance of this limiting
strongly coupled also around black hole geometries (and slice is just a coordinate effect (since the choice of folia-
actually around any static spherically symmetric solution tion has only practical but not physical meaning), the fo-
of the class described by Eqs. (22)–(23)). liation has a physical meaning in mHG (where diffeomor-
This result can also be confirmed by looking at the phism invariance is broken). We interpret the appearance
position of the spin-0 horizon for radial khronon modes of the aforementioned limiting slice as the formation of
(ℓ = 0). In more detail, if one restores time derivatives in a universal horizon (i.e. a causal boundary for signals
Eq. (46) by replacing ω → i∂t and takes the eikonal limit of arbitray speed). This is a hallmark of black holes in
– thus keeping only the highest radial and time deriva- Hořava gravity [37, 38], and we thus conclude that col-
tives of φ3 (r) – we find that Eq. (46) can be rewritten lapse in mHG leads to the formation of a black hole (as
as opposed to other vacuum solutions of the theory).
We have also studied QNMs around this family of
gAB ∂A ∂B φ3 (r) ≈ 0, (53) Lorentz-violating black holes, which are described by the
where the indices A, B run on {t, r} and the (inverse) Schwarzschild metric, but which present a non-trivial
effective metric g is given by khronon configuration. By using the standard Regge-
  Wheeler gauge, we have shown that the metric perturba-
2 2 2
1 − A f f (A f − 1)(1 + A f ) tions, both in the even and odd sectors, satisfy the same
gAB =  .
f (A2 f − 1)(1 + A2 f ) f 2 (1 + A2 f )2 (linearized) equations as in GR. We have also found that
the extra scalar mode of the theory, i.e. the perturba-
(54) tion of the khronon field, remains strongly coupled also
By computing the null cones of this effective met- around static and spherically symmetric spacetimes.
ric, one finds that the apparent horizon for the khronon We therefore conclude that no (classical) observable
modes – which coincides with the event horizon, since the deviations from GR arise in either spherical collapse or
metric is static – lies at the outermost radius at which in the spectrum of black hole QNMs, at least if the
grr = 0 (or equivalently = gtt = 0). Since that location khronon (which undergoes a non-trivial dynamics) does
corresponds to 1 + A2 (r)f (r) = 0, one can conclude that not couple directly to matter. While direct coupling of
the spin-0 horizon coincides with the universal horizon, the khronon to matter is certainly possible, this would
which we recall is the causal boundary for signals of in- produce large violations of Lorentz invariance in the mat-
finite speed. This again signals that the khronon field ter sector, which are tightly constrained by experiments.
remains strongly coupled even on spherically symmetric Another possibility to further test mHG may be provided
and static black hole spacetimes. We stress however that by cosmology (which already places mild constraints on
the khronon does not couple directly to matter (and in mHG via e.g. BBN) and in general by spacetimes which
particular to GW detectors) at tree level. are not asymptotically flat.
 10

ACKNOWLEDGMENTS finally, variation with respect to the metric γij gives

1 (3) ij
G ij ≡ (3) Rij − Rγ
We are grateful to M. Bezares, L. Lehner, E. Lim and 2
C. Palenzuela for helpful discussions about spherical col- 1
DT (1 − β)Kij − (1 + λ)γ ij K
 
−
lapse, and to D. Blas and S. M. Sibiryakov for numer- N
ous conversations about Lorentz-violating gravity. We 2 h  i
+ Dk N (i (1 − β)K j)k − (1 + λ)Kγ j)k
acknowledge financial support provided under the Euro- N
pean Union’s H2020 ERC Consolidator Grant “GRavity 1 (A3)
− γ ij (1 − β)Kkl Kkl + (1 + λ)K2
 
from Astrophysical to Microscopic Scales” grant agree- 2
ment no. GRAMS-815673. 1
+ 2(1 − β)Kik Kj k − Di Dj N − γ ij Dk Dk N


 N
1
+ α ai aj − γ ij a2 − (1 + β + 2λ)Kij K
2
− 4πGN (2 − α)(T ij − N i N j T 00 ) = 0,

where Di is the covariant derivative compatible with γij

and DT ≡ ∂T − Nk Dk .
Appendix A: Equations of motion in the unitary
gauge
Appendix B: Characteristic speeds of khronometric
theory in spherical symmetry
The variation of the action (1) with respect to the lapse
N yields In this Appendix, we write the evolution equations for
the metric and khronon [Eqs. (A3) and (A1)] for the
ansatz (25) in generic khronometric theories and in spher-
ical symmetry, and compute their characteristic speeds.
H ≡(3) R − (1 − β)Kij Kij + (1 + λ)K2 We also refer the reader to [45, 73] for more details on
(A1)
−αai ai − 2αDi ai − 8πGN (2 − α)N 2 T 00 = 0; spherical collapse in generic khronometric theories.
By combining with the momentum constraint √ (A2)
and√ by introducing√the variables X √ ≡ ∂T A/Z, Y ≡
∂T B/Z, AR ≡ ∂R A and BR ≡ ∂R B, Eqs. (A3) and
on the other hand, a variation with respect to the shift (A1) can be put in the first order form
Ni gives
∂T u + M · ∂R u = S (B1)
2
√
  ∂R Z = SZ (B2)
i ij 1 + λ ij
H ≡Dj K − γ K where u = (X, Y, BR , AR ), M is the characteristic ma-
1−β
(A2) trix
2 − α 0i
+ 4πGN N (T + N i T 00 ) = 0;
1−β

 √ √ 2(α−2)(λ+1)Z

(β + λ) ZABk1 2(λ + 1) ZBAk1 α(β−1)(β+3λ+2) √
AB
0
 √ √
 (β + λ) AZBk 2(λ + 1) ZBAk − (α−2)(β+λ)Z

2 2 α(β−1)(β+3λ+2)A 0
M =
 


 0 −Z 0 0
−Z 0 0 0

√  √ √ 
while S and SZ are complicated source terms that de- Ṙ = ABZ k1 (β + λ) B + 2(λ + 1)k2 A (B4)
pend on Z, ∂R Z, A, B, X, Y , BR , AR and the matter
variables. Z
Ṙ = ±c0 √ (B5)
The characteristic matrix has four eigenvalues A

Ṙ = 0 (B3) where k1 and k2 are functions of T and R, which can

 11

be chosen arbitrarily (as they regulate how the momen- Appendix C: The linearized field equations for the
tum constraint is linearly combined with the evolution even-parity sector
equations), and c0 is the propagation speed for the spin-
0 modes in Minkowski space [cf. Eq. (19)], This means Let us start from the trace-reversed system
that the sub-system X, Y, BR , AR is strongly hyperbolic
if k1 = k2 6= 0 and if c0 is real and finite, while Eq. (B2) 1
Ẽµν ≡ Eµν − gµν Eαβ g αβ = 0 , (C1)
can be solved as an ordinary differential equation at each 2
time step (provided that suitable boundary conditions
are imposed on it). Note however, as stressed in the and use it to compute the linearized equations δ Ẽµν . In
main text, that c0 diverges in the mHG limit, signaling order to simplify them, we make use of the background
a strong-coupling problem. field equations, and of the unit-norm and hypersurface-
orthogonality constraints (7)-(8), in order to get rid of
φ1 , φ2 and their derivatives.
In more detail, the seven non-trivial linearized equa-
tions have the following structure:

"
tt tt tt tt
δ Ẽtt ∝ λ CK K + CH 0
H0 + CH 1
H1 + CH 2
tt
H2 + Cφtt3 φ3 + CK ′ tt ′ tt ′ tt ′ tt ′
′ K + CH ′ H0 + CH ′ H1 + CH ′ H2 + Cφ′ φ3
0 1 2 3

#
tt tt tt tt (3) (3f + 1)iωH1 ω 2 H2 (C2a)
+ CK ′′ K
′′
+ CH ′′
′′ H0 + CH ′′
′′ H1 + CH ′′
′′ H2 + Cφtt′′3 φ′′3 + Cφtt(3) φ3 + −
0 1 2 3 2r 2
1 (f − 1)K ′ f (f + 3)H0′ f (f − 1)H2′ f ΛH0 f
+ H0′′ f 2 + + + iωH1′ f − − − Kω 2 = 0,
2 2r 4r 4r 2r2
"
tr tr tr tr
δ Ẽtr ∝ λ CK K + CH 0
H0 + CH 1
H1 + CH 2
tr
H2 + Cφtr3 φ3 + CK ′ tr ′ tr ′ tr ′ tr ′
′ K + CH ′ H0 + CH ′ H1 + CH ′ H2 + Cφ′ φ3
0 1 2 3

#
+ tr
CK ′′ K
′′
+ tr
CH ′′
′′ H0 + tr
CH ′′
′′ H1 + tr
CH ′′
′′ H2 + Cφtr′′3 φ′′3 +
(3)
Cφtr(3) φ3 (C2b)
0 1 2 3

i(3f − 1)Kω H1 Λ iH2 ω

− − + − iωK ′ = 0,
2f r 2r2 r

"
rr rr rr rr
δ Ẽrr ∝ λ CK K + CH0
H0 + CH1
H1 + CH2
rr
H2 + Cφrr3 φ3 + CK ′ rr ′ rr ′ rr ′ rr ′
′ K + CH ′ H0 + CH ′ H1 + CH ′ H2 + Cφ′ φ3
0 1 2 3

#
rr rr rr rr (3)
+ CK ′′ K
′′
+ CH0
′′
′′ H0 + CH1
′′
′′ H1 + CH2
′′
′′ H2 + Cφrr′′3 φ′′3 + Cφrr(3) φ3
3
(C2c)
2
 
ω Λ i(f − 1)H1 ω 3(f − 1)H0′ (3f + 1)H2′
H2 − + + −
2f 2 2f r2 2f 2 r 4f r 4f r
iωH1′ (3f + 1)K ′ H0′′
− + − + K ′′ = 0,
f 2f r 2

"
δ Ẽφφ θθ θθ θθ θθ
δ Ẽθθ + ∝ λ CK K + CH H0 + CH H1 + CH θθ
H2 + Cφθθ3 φ3 + CK ′ θθ ′ θθ ′
′ K + CH ′ H0 + CH ′ H1
sin2 θ 0 1 2 0 1

#
θθ θθ ′ θθ θθ θθ θθ θθ ′′ (3)
+ CH2
′
′ H2 + Cφ′ φ3 + CK ′′ K
3
′′
+ CH0
′′ ′′ ′′
′′ H0 + CH ′′ H1 + CH ′′ H2 + Cφ′′ φ3 +
1 2 3
Cφθθ(3) φ3
3
(C2d)
2 2
 
r ω
− f rH0′ − f rH2′ + f r2 K ′′ + (3f r + r)K ′ + K −Λ+2
f
 
H0 Λ Λ
+ + H2 − − 2 − 2iH1 rω = 0,
2 2
 12
"
3
2 !
(3f + 1) A2 f + 1 A2 f − 1 iω A4 f 2 − 1


δ Ẽtθ ∝ λ H0 4 2
+
32A f r 32A4 f 2

2
3 !
A2 f + 1 3A4 f 3 + A2 − 12 A2 f 2 + 3 − 4A2 f + 1 iω A2 f + 1 A2 f − 1





+ H1 +
16A4 f 2 r 16A4 f 2

3
2 4 2 !
(3f + 1) A2 f + 1 A2 f − 1 iω A2 f + 1 A f − 6A2 f + 1



+ H2 +
32A4 f 2 r 32A4 f 2
−A8 f 4 − 6A6 f 3 + 6A2 f + 1 H0′ A4 f 2 − 1 K ′



+ +
32A4 f 4A2 (C2e)
3
2
2 !
i(3f + 1)ω A2 f + 1 A2 f − 1 ω 2 A4 f 2 − 1 Λ A2 f + 1



+ φ3 − −
8A4 f 2 r 8A4 f 2 2A2 r2
4 2 2

2
2 2

2
3 !
2A f − A (5f + 3) + 2 A f + 1 iω A f − 1 A f + 1
+ φ′3 − −
4A4 r 4A4 f
4 3
2
4 #
A2 f + 1 H1′ A2 f − 1 A2 f + 1 H2′ iKω A2 f + 1 A2 f + 1 φ′′3




− − − +
16A4 f 32A4 f 4A2 f 8A4
(f − 1)H1 f H1′ 1 1
+ − − iH2 ω − iKω = 0,
2r 2 2 2

"
4
3 !
(3f + 1) A2 f − 1 iω A2 f + 1 A2 f − 1


δ Ẽrθ ∝ λ H0 − −
32A4 f 3 r 32A4 f 3

4 !
−3A8 f 5 − A2 − 12 A6 f 4 + 4A6 f 3 − 12A2 f 2 + 3 − 4A2 f + 1 iω A2 f − 1



+ H1 −
16A4 f 3 r 16A4 f 3

4 !
2
(3f + 1) A2 f − 1 iω A8 f 4 − 6A6 f 3 + 6A2 f − 1 A4 f 2 − 1 H2′


+ H2 − − +
32A4 f 3 r 32A4 f 3 32A4 f 2

3
2 4 2
A2 f − 1 A2 f + 1 H1′ A2 f − 1 A f + 6A2 f + 1 H0′ iKω A4 f 2 − 1




+ + + (C2f)
16A4 f 2 32A4 f 2 4A2 f 2
4 3
!
Λ A4 f 2 − 1 i(3f + 1)ω A2 f − 1 ω 2 A2 f + 1 A2 f − 1





+ φ3 − +
2A2 f r2 8A4 f 3 r 8A4 f 3

2 !
2
′ 2A8 f 4 − A6 f 2 (5f + 3) + A2 (5f + 3) − 2 iω A4 f 2 − 1 A2 f − 1 K ′
+ φ3 + −
4A4 f r 4A4 f 2 4A2 f

3 #
A2 f − 1 A2 f + 1 φ′′3


(3f − 1)H0 (f + 1)H2 iH1 ω H0′ K′
− + − − − + = 0,
8A4 f 4f r 4f r 2f 2 2

δ Ẽθφ ∝ H0 − H2 = 0 . (C2g)

The explicit expressions for the coefficients Ckij are given covariant gravitational action (6) (without the matter
in the Supplemental Material as Mathematica [72] files. contribution) one obtains the generalized Bianchi iden-
Let us notice that these seven equations contain only tity [25]
five independent variables H0 , H1 , H2 , K and φ3 . This κp
seems to imply that the system may be over-determined. ∇µ E µν = − ∇α T ∇α T uν . (C3)
This turns out not to be the case, since some of these 2
equations are redundant due to the Bianchi identity. Taking linear combinations to cancel out the explicit de-
In more detail, from diffeomorphism invariance of the pendence on T and κ and performing trivial manipula-
 13

tions, one can then write the identity which take the schematic form

∇ν (E µν uα − E αν uµ ) = E µν ∇ν uα − E αν ∇ν uµ , (C4)
F0 ≡ H0′′ − d1 H0′ + d2 K ′ + d3 H0 + d4 K = 0, (C6)
which can be used to show that two of the seven equa-
tions can be eliminated from the system without loss of FK ≡ K ′′ − d5 H0′ + d6 K ′ + d7 H0 + d8 K = 0, (C7)
generality. This can also be seen by direct manipulation H1 − d9 H0′ + d10 K ′ + d11 H0 + d12 K = 0, (C8)
of the equations of motion, as we will now show.
From δ Ẽθφ = 0, we obtain
where the di are functions of r, ω and Λ. We have
H2 (r) = H0 (r), (C5)
checked that the derivative of Eq. (C8) coincides with
which allows us to get rid of H2 completely. The struc- the previous analytic solution that we had found for H1′ ,
ture of the remaining equations is then the following: on so Eq. (C8) is redundant. We are thus left with three
the one hand, the equations δ Ẽtt , δ Ẽtr , δ Ẽrr , δ Ẽθθ con- independent equations, corresponding to Fφ = 0, F0 = 0
tain up to second derivatives of the metric perturbations and FK = 0, which depend only on three variables φ3 , H0
and up to third derivatives of φ3 (r)6 ; on the other hand, and K, with the scalar field present only in Fφ . Moreover,
both F0 and Fk are independent of λ, and Fφ contains
in δ Ẽtθ and δ Ẽrθ one can find up to first derivatives of
only first derivatives of H0 and K.
the metric perturbations and up to second derivatives of
φ3 (r). Thus, from δ Ẽtθ = 0 and δ Ẽrθ = 0 we can solve In the GR limit, λ → 0, the dependence on φ3 also
algebraically for H1′ (r) and φ′′3 (r). This defines an equa- disappears from Fφ . In that case, compatibility of the
tion for the scalar field, which we denote by Fφ = 0. The system would require that one of the equations is redun-
next step is to use the expressions for H1′ (r) and φ′′3 (r) dant. Note that this must be the case since we know
(3)
(and their derivatives) to eliminate H1′ , H1′′ , φ′′3 and φ3 that in the GR limit the energy constraint is re-instated,
from the rest of the equations. By doing so, we obtain cf. Eq. (30). Since Fφ reduces to a first order equation
Ẽrr ∝ Ẽtt . in the GR limit, it can be used, upon substitution into
We then solve δEtt = 0, δErt = 0 and δEθθ = 0 and the other equations, to reduce the whole system to two
get algebraic expressions for H0′′ (r), K ′′ (r) and H1 (r), first-order equations relating H0 and K:

f 2 (−Λ) + f Λ2 − 2Λ + 14r2 ω 2 + Λ − 2Λr2 ω 2 − 6r2 ω 2

− H0
f r (f Λ − Λ + 4r2 ω 2 )
(C9a)
f 2 Λ2 − 2Λ + 9r2 ω 2 − 2f (2Λ + 1)r2 ω 2 + 4r4 ω 4 + r2 ω 2



+ K + H0′ = 0,
f 2 r (f Λ − Λ + 4r2 ω 2 )

H0 −2f Λ + Λ2 + 4r2 ω 2 K f Λ2 − 2Λ + 6r2 ω 2 − 2(Λ + 1)r2 ω 2

+ + K ′ = 0. (C9b)
−f Λr − 4r3 ω 2 + Λr f r (f Λ − Λ + 4r2 ω 2 )

Since Eqs. (C6)-(C7) can be shown to be independent After performing this transformation, Eq. (C9) reduces
of λ, Eq. (C9) also holds in the general case, as can be to the simple equation
checked by direct substitution.
d2 Ψ  2
A last simplification occurs by introducing the same

2
+ ω − Veven (r) Ψ = 0, (C12)
variable transformation as in [58, 74], given by dr∗

Λ(1 + Λ) − 3(2 + Λ)f + 6f 2 with potential

K= Ψ + f Ψ′ , (C10)
2r (1 + Λ − 3f ) f

1 + Λ − 3Λf + 3f 2 ′ Veven = 2 (1 + Λ) (Λ(Λ − 2) + 3)
r (1 + Λ − 3f )2
H0 = − Ψ
2(1 + Λ − 3f ) 
(C11) 2
 
− 3f (1 + Λ) + 3f (f − 1 − Λ) . (C13)
(Λ − 2)2 (1 + Λ) rω 2
 
1 + Λ − 3f
+ + − Ψ.
6r 3r(1 + Λ − 3f )2 f However, in the general case of non-vanishing λ, the
third equation Fφ = 0 remains independent and serves
6 as the equation of motion for the scalar field:
Third radial derivatives appear after imposing the hypersurface-
orthogonality condition, Eq. (8). φ′′3 (r) + W1 (r)φ′3 (r) + W0 (r)φ3 (r) = j(r), (C14)
 14

with

j(r) = U1 (r)Ψ′ (r) + U0 (r)Ψ(r). (C15)

The explicit forms of the functions Wi (r) are

ω 2i − 2iA2 f


−4A4 f 2 + 2A2 (5f + 3) − 4
W1 (r) = 2 + , (C16)
r (A2 f + 1) A2 f 2 + f

3
2
i(3f + 1)ω A2 f − 1 ω 2 A2 f − 1 4A2 Λ
W0 (r) = 3 − 2 − 2, (C17)
f 2 r (A2 f + 1) f 2 (A2 f + 1) r2 (A2 f + 1)

while U0 (r) and U1 (r) are included in the Supplemental Material as Mathematica [72] files.

[1] V. A. Kostelecky, Phys. Rev. D69, 105009 (2004), arXiv:1010.5249 [hep-th].

arXiv:hep-th/0312310 [hep-th]. [16] D. Blas, O. Pujolas, and S. Sibiryakov,
[2] V. A. Kostelecky and N. Rus- Phys. Rev. Lett. 104, 181302 (2010),
sell, Rev. Mod. Phys. 83, 11 (2011), arXiv:0909.3525 [hep-th].
arXiv:0801.0287 [hep-ph]. [17] D. Blas, O. Pujolas, and S. Sibiryakov,
[3] D. Mattingly, Living Rev. Rel. 8, 5 (2005), JHEP 04, 018 (2011), arXiv:1007.3503 [hep-th].
arXiv:gr-qc/0502097 [gr-qc]. [18] T. Jacobson, Phys. Rev. D81, 101502 (2010), [Erratum:
[4] T. Jacobson, S. Liberati, and D. Mat- Phys. Rev.D82,129901(2010)], arXiv:1001.4823 [hep-th].
tingly, Annals Phys. 321, 150 (2006), [19] B. P. Abbott et al. (Virgo, Fermi-GBM, INTE-
arXiv:astro-ph/0505267 [astro-ph]. GRAL, LIGO Scientific), Astrophys. J. 848, L13 (2017),
[5] T. Jacobson, Proceedings, Workshop on From quantum arXiv:1710.05834 [astro-ph.HE].
to emergent gravity: Theory and phenomenology (QG- [20] A. Emir Gumrukcuoglu, M. Saravani, and
Ph): Trieste, Italy, June 11-15, 2007, PoS QG-PH, 020 T. P. Sotiriou, Phys. Rev. D97, 024032 (2018),
(2007), arXiv:0801.1547 [gr-qc]. arXiv:1711.08845 [gr-qc].
[6] S. Liberati, Class.Quant.Grav. 30, 133001 (2013), [21] C. M. Will, Theory and Experiment in Gravitational Physics
arXiv:1304.5795 [gr-qc]. (Cambridge University Press, 1993).
[7] A. V. Kostelecky and J. D. Tas- [22] C. M. Will, Living Rev. Rel. 17, 4 (2014),
son, Phys. Rev. D83, 016013 (2011), arXiv:1403.7377 [gr-qc].
arXiv:1006.4106 [gr-qc]. [23] D. Blas and H. Sanctuary,
[8] P. Hořava, Phys. Rev. D79, 084008 (2009), Phys. Rev. D84, 064004 (2011),
arXiv:0901.3775 [hep-th]. arXiv:1105.5149 [gr-qc].
[9] A. O. Barvinsky, D. Blas, M. Herrero- [24] M. Bonetti and E. Barausse,
Valea, S. M. Sibiryakov, and C. F. Phys. Rev. D91, 084053 (2015), [Erratum: Phys.
Steinwachs, Phys. Rev. D93, 064022 (2016), Rev.D93,029901(2016)], arXiv:1502.05554 [gr-qc].
arXiv:1512.02250 [hep-th]. [25] O. Ramos and E. Barausse,
[10] S. Chadha and H. B. Nielsen, Phys. Rev. D99, 024034 (2019),
Nucl. Phys. B217, 125 (1983). arXiv:1811.07786 [gr-qc].
[11] G. Bednik, O. Pujolas, and S. Sibiryakov, [26] S. M. Carroll and E. A. Lim,
JHEP 11, 064 (2013), arXiv:1305.0011 [hep-th]. Phys. Rev. D70, 123525 (2004),
[12] A. O. Barvinsky, D. Blas, M. Herrero- arXiv:hep-th/0407149 [hep-th].
Valea, S. M. Sibiryakov, and C. F. [27] K. Yagi, D. Blas, E. Barausse, and N. Yunes,
Steinwachs, Phys. Rev. Lett. 119, 211301 (2017), Phys. Rev. D89, 084067 (2014), [Erratum: Phys.
arXiv:1706.06809 [hep-th]. Rev.D90,no.6,069901(2014)], arXiv:1311.7144 [gr-qc].
[13] A. O. Barvinsky, M. Herrero-Valea, and S. M. [28] K. Yagi, D. Blas, N. Yunes, and E. Ba-
Sibiryakov, Phys. Rev. D 100, 026012 (2019), rausse, Phys. Rev. Lett. 112, 161101 (2014),
arXiv:1905.03798 [hep-th]. arXiv:1307.6219 [gr-qc].
[14] S. Groot Nibbelink and M. Pospelov, [29] P. Berglund, J. Bhattacharyya, and
Phys. Rev. Lett. 94, 081601 (2005), D. Mattingly, Phys. Rev. D85, 124019 (2012),
arXiv:hep-ph/0404271 [hep-ph]. arXiv:1202.4497 [hep-th].
[15] M. Pospelov and Y. Shang, [30] E. Barausse, Phys. Rev. D 100, 084053 (2019),
Phys. Rev. D85, 105001 (2012), arXiv:1907.05958 [gr-qc].
 15

[31] C. M. Will, Class. Quant. Grav. 35, 085001 (2018), Numerical Relativity: Solving Einstein’s Equations on the Compu
arXiv:1801.08999 [gr-qc]. (Cambridge University Press, 2010).
[32] T. Kobayashi, Y. Urakawa, and M. Yamaguchi, [52] R. Loll and L. Pires, Phys. Rev. D 90, 124050 (2014),
JCAP 04, 025 (2010), arXiv:1002.3101 [hep-th]. arXiv:1407.1259 [hep-th].
[33] E. Barausse and L. Lehner, [53] D. Markovic and S. Shapiro,
Phys. Rev. D 88, 024029 (2013), Phys. Rev. D 61, 084029 (2000), arXiv:gr-qc/9912066.
arXiv:1306.5564 [gr-qc]. [54] M. Saravani, N. Afshordi, and R. B.
[34] R. Loll and L. Pires, ArXiv e-prints (2014), Mann, Phys. Rev. D 89, 084029 (2014),
arXiv:1407.1259 [hep-th]. arXiv:1310.4143 [gr-qc].
[35] J. Bellorin and A. Restuccia, [55] N. Afshordi, Phys. Rev. D80, 081502 (2009),
Int. J. Mod. Phys. D21, 1250029 (2012), arXiv:0907.5201 [hep-th].
arXiv:1004.0055 [hep-th]. [56] J. Bhattacharyya, A. Coates, M. Colombo,
[36] M. Henneaux, A. Kleinschmidt, and G. Lu- A. E. Gumrukcuoglu, and T. P.
cena Gómez, Phys. Rev. D 81, 064002 (2010), Sotiriou, Phys. Rev. D 97, 064020 (2018),
arXiv:0912.0399 [hep-th]. arXiv:1612.01824 [hep-th].
[37] E. Barausse, T. Jacobson, and T. P. [57] T. Regge and J. A. Wheeler,
Sotiriou, Phys. Rev. D83, 124043 (2011), Phys. Rev. 108, 1063 (1957).
arXiv:1104.2889 [gr-qc]. [58] F. J. Zerilli, Phys. Rev. Lett. 24, 737 (1970).
[38] D. Blas and S. Sibiryakov, [59] S. A. Teukolsky, Astrophys. J. 185, 635 (1973).
Phys. Rev. D84, 124043 (2011), [60] W. Israel, Phys. Rev. 164, 1776 (1967).
arXiv:1110.2195 [hep-th]. [61] S. W. Hawking, Commun. Math. Phys. 25, 152 (1972).
[39] T. Jacobson and D. Mattingly, [62] B. Carter, Phys. Rev. Lett. 26, 331 (1971).
Phys. Rev. D70, 024003 (2004), [63] D. Robinson, Phys. Rev. Lett. 34, 905 (1975).
arXiv:gr-qc/0402005 [gr-qc]. [64] E. Barausse, V. Cardoso, and
[40] J. W. Elliott, G. D. Moore, and H. Stoica, P. Pani, Phys. Rev. D89, 104059 (2014),
JHEP 08, 066 (2005), arXiv:hep-ph/0505211 [hep-ph]. arXiv:1404.7149 [gr-qc].
[41] B. Audren, D. Blas, J. Lesgourgues, and S. Sibiryakov, [65] O. Dreyer, B. J. Kelly, B. Krishnan,
JCAP 1308, 039 (2013), arXiv:1305.0009 [astro-ph.CO]. L. S. Finn, D. Garrison, and R. Lopez-
[42] C. Eling and T. Jacobson, Aleman, Class. Quant. Grav. 21, 787 (2004),
Class. Quant. Grav. 23, 5643 (2006), [Erra- arXiv:gr-qc/0309007.
tum: Class. Quant. Grav.27,049802(2010)], [66] E. Berti, V. Cardoso, and A. O.
arXiv:gr-qc/0604088 [gr-qc]. Starinets, Class. Quant. Grav. 26, 163001 (2009),
[43] C. Eling and T. Jacobson, arXiv:0905.2975 [gr-qc].
Class. Quant. Grav. 23, 5625 (2006), [Erra- [67] B. P. Abbott and et al. (LIGO Scientific and Virgo Col-
tum: Class.Quant.Grav. 27, 049801 (2010)], laborations), Phys. Rev. Lett. 116, 221101 (2016).
arXiv:gr-qc/0603058. [68] B. Abbott et al. (LIGO Scientific,
[44] C. Eling, T. Jacobson, and M. Coleman Miller, Virgo), Phys. Rev. D 100, 104036 (2019),
Phys. Rev. D76, 042003 (2007), [Erratum: Phys. arXiv:1903.04467 [gr-qc].
Rev.D80,129906(2009)], arXiv:0705.1565 [gr-qc]. [69] M. Isi, M. Giesler, W. M. Farr, M. A. Scheel, and
[45] D. Garfinkle, C. Eling, and T. Ja- S. A. Teukolsky, Phys. Rev. Lett. 123, 111102 (2019),
cobson, Phys. Rev. D76, 024003 (2007), arXiv:1905.00869 [gr-qc].
arXiv:gr-qc/0703093 [GR-QC]. [70] R. Abbott et al. (LIGO Scientific, Virgo), (2020),
[46] F. Estabrook, H. Wahlquist, S. Christensen, B. DeWitt, arXiv:2010.14529 [gr-qc].
L. Smarr, and E. Tsiang, Phys. Rev. D 7, 2814 (1973). [71] E. Berti, A. Sesana, E. Barausse, V. Cardoso,
[47] L. I. Petrich, S. L. Shapiro, and S. A. Teukolsky, and K. Belczynski, Phys. Rev. Lett. 117, 101102 (2016),
Phys. Rev. D 31, 2459 (1985). arXiv:1605.09286 [gr-qc].
[48] S. L. Shapiro and S. A. Teukolsky, [72] W. R. Inc., “Mathematica, Version 12.2,” .
Astrophys. J. 298, 34 (1985). [73] J. Bhattacharyya, A. Coates, M. Colombo, and
[49] R. Beig and N. O’Murchadha, T. P. Sotiriou, Phys. Rev. D 93, 064056 (2016),
Phys. Rev. D 57, 4728 (1998), arXiv:gr-qc/9706046. arXiv:1512.04899 [gr-qc].
[50] M. Alcubierre, Introduction to 3+1 numerical relativity , [74] N. Sago, H. Nakano, and M. Sasaki,
International series of monographs on physics (Oxford Phys. Rev. D 67, 104017 (2003), arXiv:gr-qc/0208060.
Univ. Press, Oxford, 2008).
[51] T. W. Baumgarte and S. L. Shapiro,