Masters Thesis Caputo
Masters Thesis Caputo
Masters Thesis Caputo
Tesi di Laurea
Relatore Laureanda
Prof. Sabino Matarrese Claudia Caputo
Correlatore
Dr. Daniele Bertacca
Introduction 4
2 Gravitational slip 14
2.1 Probing gravitational slip with large-scale structures . . . . . . . . . . . . . . . . 15
2.1.1 Growth of structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.2 CMB temperature anisotropies and CMB polarizations . . . . . . . . . . . 16
2.1.3 CMB and LSS cross-correlations . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2 Gravitational slip as signature of non-standard propagation of tensor modes . . 19
3 The Effective Field Theory approach: unifying single-field models of Dark Energy
and Modified Gravity 21
3.1 The most general EFT parametrization . . . . . . . . . . . . . . . . . . . . . . . . 21
3.1.1 Background evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.1.2 Perturbed equations: Poisson equation and anisotropy constraint . . . . . 23
3.1.3 Phenomenological functions in QSA . . . . . . . . . . . . . . . . . . . . . 24
3.2 EFT of Horndeski gravity and beyond Horndeski . . . . . . . . . . . . . . . . . . 25
3.2.1 The α-parametrization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2.2 Stability conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2.3 Horndeski gravity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2.4 Beyond Horndeski models . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2.5 GWs propagation in scalar-tensor theories . . . . . . . . . . . . . . . . . . 36
1
Contents
Conclusions 64
A Relations among EFT operators, α property functions and Horndeski- beyond Horn-
deski’s free functions 66
A.0.1 Horndeski gravity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
A.0.2 Beyond Horndeski gravity . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
D Fisher information 77
D.0.1 Covariance matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
References 79
2
Introduction
The universe is undergoing a phase of accelerated expansion and this is a fact supported by a
vast amount of observational data collected over the years. The first probe of such an accel-
eration came from Supernova Type Ia (SNIa) measurements in 1998 [1]- [2] and it was later
confirmed by independent cosmological observational data from baryon acoustic oscillations
(BAO), cosmic microwave background (CMB) and large scale structures (LSS) [3]. However, a
satisfactory explanation of the cosmological phenomena, dubbed ”dark energy” (DE), remains
until today completely elusive.
In the years that followed the discovery of the expansion of the universe a landscape of theo-
ries were proposed, the simplest of them is the one that modifies the r.h.s of the Einstein’s field
equations introducing the so called cosmological constant Λ. It works as a negative pressure
against gravity in such a way to produce an accelerated expansion. This theory corresponds to
the cosmological concordance model ΛCDM (Lambda Cold Dark Matter) that consists in a flat
universe in which the gravity is described by general relativity (GR) and the evolution history
of the universe is the one predicted by the standard cosmological Hot Big Bang model in which
the energy content of the universe is roughly subdivided in this way: 25% of cold dark matter,
5% baryons and 70% dark energy. This simple scenario is in impressive agreement with obser-
vations: GR has been shown to be consistent with all current observations, including tests in
the solar system and binary pulsars; moreover data analysis of Planck 2018 combined with SN
Ia and BAO data constrain the value of the equation of state (EOS) (wDE = pDE /ρDE ) to be close
to ” − 1”. Thus, up to now the standard six-parameters ΛCDM scenario represents the cur-
rent cosmological concordance model that is in general agreement with observed phenomena.
However, this model suffers some theoretical issues, among them there is the difficulty to con-
ciliate the vacuum energy density theoretically predicted by particle physics (ρvac ' 1074 GeV4 )
with the observed value of dark energy (ρΛ ' 10−47 GeV4 ) [4]- [106]. Furthermore, beside these
theoretical problems, it was recently disclosed that there is a substantial discordance inferred
by ΛCDM between observations of large-scale structures (e.g. weak lensing measurements
from KiDS (Kilo Degree Survey) ) and CMB measurements made by Planck .
These theoretical and observational problems have motivated the searching for alternative so-
lutions. In particular, they can be divided in two categories: dark energy (DE) and modified
gravity (MG) models. The former are models in which the r.h.s of the Einstein’s field equations
is modified, i.e. the energy-momentum tensor T µν ; the latest are models in which the geomet-
rical part of the Einstein’s field equations is modified (l.h.s).
In order to be in agreement with observational data a theory of gravity should have an ex-
pansion history that is indistinguishable from that of ΛCDM and thus it should give rise to a
sufficiently long radiation and matter eras, as well as a transition to a stable the Sitter era [58];
it should also satisfy local gravity constraints. However, such a theory can still differs, with
respect to ΛCDM, in the evolution of the cosmological perturbations and thus predicts differ-
ent growth of structures and also a non-standard relation between the scalar potentials that
characterize the scalar-mode perturbations in the Newtonian gauge, i.e. Ψ , Φ
In this dissertation we consider MG models that predict such a non-standard relation between
the scalar potentials, i.e theories in which there is gravitational slip.
The thesis is organized as follows:
3
Contents
Chapter 1 We write down the closed system of equations which describe the evolution of cosmo-
logical perturbations at linear order in GR. Then, in order to measure the deviation of the phe-
nomenology of LSS from GR in a model-independent way, we introduce the functions µ(a, k),
η(a, k) and Σ(a, k). Departures from GR can be fully characterized by using only two functions
of scale and time. In the last part of this chapter we briefly list other popular parametrizations
commonly used in the literature to measure deviations from GR.
Chapter 2 We focus on gravitational slip making a distinction between the anisotropic stress
sourced by matter and the gravitational slip itself. Then we quickly mention how LSS can be
used to probe gravitational slip and how, in a certain class of theories, it is related with the
non-standard propagation of the tensor modes.
Chapter 3 We introduce the effective field theory (EFT) formalism by considering two differ-
ent parametrizations: the first is the most general EFT approach which consists in an unified
description of DE and MG models; the second, the α-parametrization, is specifically designed
to study the phenomenological aspects of Horndeski theories and, by adding new parameters,
of its extensions. Using these formalisms we write down the modified Poisson equation, the
modified anisotropy constraint and, consequently, we obtain the analytic expressions for the
phenomenological functions µ, η and Σ in the quasi-static approximation. Then we focus on
Horndeski gravity and beyond Horndeski models specifying the above-mentioned functions
for these two classes of theories.
Chapter 4 We account for the recent binary neutron-star merger that put a very stringent bound
on the tensor speed at low redshifts. In particular, we classify the survived Horndeski theories
in terms of the α-parameters finding for which values of these property functions the theo-
ries exhibit gravitational slip. We also consider the expressions of µ, η and Σ, for both Horn-
deski and beyond Horndeski theories, in two limiting cases (sub-Compton and super-Compton
regimes) and we discuss how, using these expressions, cosmological observations can be used
to rule out subclasses of these theories.
Chapter 5 This chapter is devoted to Horndeski gravity. In particular, we summarize what we
find in the previous chapter and following [13] we discuss the conjecture (µ − 1)(Σ − 1) ≥ 0
which characterizes this class of models. We finally reports the results of two recent analysis
intended to constrain these theories using observational data and the publicly available Boltz-
mann codes EFTCMB and EFTCosmoMC.
Chapter 6 We shortly list possible sources of degeneracies that one has to take into account
when testing for departures from GR and that can significantly limit our ability in constrain-
ing cosmological parameters (and in particular MG parameters).
Chapter 7 In this final chapter we report the results of some forecasting analysis in order to have
an overview of how future surveys will improve our ability in constraining MG parameters and
thus, testing departures from GR.
In this dissertation we adopt the metric signature (−, +, +, +) and we use the natural unit such
that the speed of light is c = 1, the reduced Planck’s constant is ~ = 1 and the Boltzmann’s
constant is also kB = 1.
4
Chapter 1
In this chapter we see how deviations from GR can be characterized using a model-independent
approach. In order to do so we introduce some of the possible parametrizations used in the lit-
erature and we analyse the behaviour of these deviations from GR in the sub-horizon regime
and in the super-horizon regime. We thus introduce the gravitational slip and its most popular
parametrizations.
To derive the evolution equations for the linear cosmological perturbations one needs to per-
turb the FLRW metric and the energy-momentum tensor Tµν
(0) (0)
gµν = gµν + δgµν , Tµν = Tµν + δTµν (1.1)
(0) (0)
where gµν and Tµν are the background terms and δgµν , δTµν are their small perturbations.
In particular, we specialize to the conformal Newtonian (or longitudinal) gauge because it is
the most convenient to our purpose, so that the perturbed energy-momentum tensor is
where ρ and p are the density and the pressure of the unperturbed perfect fluid, vi the velocity
field and Πij ≡ Tji − Tkk δji /3 denotes the traceless component of the energy-momentum tensor.
The line element of the perturbed spatially-flat FRW metric, in the Newtonian gauge, is
where τ is the conformal time, a(τ) the scale factor and Ψ and Φ are respectively the Newtonian
potential and the curvature potential. In the Newtonian limit the potential Ψ is that which
enters in the matter equation of motion x~¨ = −∇Ψ
~ , while Φ enters in the Poisson equation ∇2 Φ =
2
−4πGa δρ; from this limit one can understand that the first determines the geodesic motion
5
1. Parametrizing deviations from GR in the evolution of cosmological
perturbations
of matter, the second is the potential generated by matter sources and, in general, they do not
coincide.
The evolution of the perturbations can be obtained by linearising the Einstein’s field equations
and subtracting the background evolution. The equations that characterize the evolution of
linear perturbations are the Poisson equation and the anisotropy constraint which are obtained
respectively by combining the time-time and time-space component and by the traceless part
of the space-space component. The former relates the over-density to the curvature potential1 ,
while, the second, describes the relationship among the two scalar potentials and the shear
stress.
In GR they can be written, in the Fourier space, as
X
k 2 Φ = −4πGa2 ρ̄i ∆i (1.6)
i
X
k 2 (Ψ − Φ) = −12πGa2 ρ̄i (1 + wi )σi (1.7)
i
where the index i denotes the sum over different matter species. Then ρi is the density of the
i-species, wi (ρ, S) = Pi (ρ, S)/ρi , ∆i and σi are respectively the comoving density perturbation
and the dimensionless shear stress perturbation2 which are defined by:
H j
∆ = δ+3 (1 + w)θ, (ρ + p)σ ≡ −(k̂i k̂ j − δi /3)Πij (1.8)
k2
here H = (da/dτ)a−1 = aH, δ = (ρ − ρ̄)/ ρ̄ is the density contrast, ik j δTj0 = (ρ + p)θ 3 .
µν
Then one needs the energy-momentum conservation equation to close the system4 : T;µ = 0
!
0 0 δp
δ = −(1 + w)(θ − 3Φ ) + 3Hδ w − (1.9)
δρ
w0
" # " #
0 2 δp/δρ
θ = − (1 − 3w)H + θ+k δ+Ψ −σ (1.10)
1+w 1+w
where we considered the averaged quantities on the i-species and the prime denotes the deriva-
tion with respect to the conformal time. Using these two equations one finds the evolution
equation for ∆
θ
∆0 = 3(1 + w)(Φ 0 + HΨ − Hσ ) + 3Hw∆ − [k 2 + 3(H2 − H0 )] (1 + w) (1.11)
k2
The motion of relativistic particles is determined by the equation for the Weyl potential ΦW =
(Φ + Ψ )/2 which is obtained by combining (1.6) and (1.7)
X
2 2
k (Ψ + Φ) = −8πGa ρ̄i ∆i + 3ρ̄i (1 + wi )σi /2 (1.12)
i
Departures from GR can be characterized using two approaches: a) one writes down the La-
grangian and derives a set of predictions which can be tested to constrain the parameters of the
1 there is an analogous equation for the Newtonian potential.
2 This quantity is related to the anisotropic stress perturbation Π defined by Kodama & Sasaki (1984) through
3
2 (ρm + pm )σ = pm Π.
3 θ is the divergence of the peculiar velocity v and their relation in the Fourier space is ik i v = θ.
i i
4 The semicolon stands for the covariant derivative, thus the energy-momentum tensor is covariantly conserved.
6
1.2. Parametrized evolution of linear cosmological perturbations in a
general theory of gravity
theory; b) one parametrizes the deviations from GR using model-independent functions which
can be reconstructed by observations, then one can obtain their specific form for each model
and can use them to rule out the theories which are not consistent with observations.
Modified gravity models, in general, affect both the background and the perturbation evolu-
tion. However, many of these theories of gravity, can mimic the same background evolution
history but they can still differ in the evolution of perturbations and thus predict measur-
able differences in the growth rate of large-scale structures. In particular, the observational
constraints on H(t), reduce the number of viable models to those in which the background evo-
lution is close to that predicts by the cosmological concordance model, ΛCDM, which fit very
well the observational data.
Using the approach b) the deviations from GR can be completely characterized by two phe-
nomenological functions which depend on scale and time. A possible choice is the following
X
k 2 Ψ = −4πGµ(a, k)a2 ρ̄i ∆i (1.13)
i
Φ
= η(a, k) (1.14)
Ψ
Here Gµ ≡ Gmatter is the effective gravitational coupling felt by non relativistic particles; η is
the gravitational slip parameter and measure the mismatch between the two scalar potentials;
in the following we refer to gravitational slip as that determined only by modifications of the
l.h.s. of Einstein’s equations5 .
Sometimes, however, can be useful to introduce the function
(1 + η)
Σ(a, k) = µ (1.15)
2
It enters in the equation for the Weyl potential and characterizes the effective gravitational
coupling felt by relativistic particles (Glight ≡ GΣ)
X
k 2 (Φ + Ψ ) = −8πGΣ(a, k)a2 ρ̄i ∆i (1.16)
i
1 Q(1 + R)
Q = µη, R= , =Σ (1.19)
η 2
All these functions, thus, are not independent, indeed the deviations from GR can by fully
characterized by choosing two of them. The couples {µ, η} and {µ, Σ} have been widely adopted
5 The free-streaming of relativistic particles also determines a difference in these two potentials, but this is a
non-linear effect at late time.
7
1. Parametrizing deviations from GR in the evolution of cosmological
perturbations
in the literature.
In [16] is pointed out that in a general theory of gravity the dynamics of linear perturbations
may not match that one obtains by averaging over small scale fluctuations. In particular, N-
body simulations, used to study the dynamics in the non-linear regime, in scalar-tensor theo-
ries of chameleon type reduced to the linear one on large-scale. But this fact is not general and
thus the relation between Ψ and ∆ could be non-linear leading to non-linear phenomenologi-
cal functions.
In what follows we consider, for simplicity, matter of type ”dust”, i.e w = 0, in particular we
only consider CDM, for which the sound velocity is negligible at least at linear order6 . The
equations (1.10), (1.9) and (1.11) become
δ0 = −θ + 3Φ 0 (1.20)
0 2
θ = −Hθ + k Ψ (1.21)
θ
∆0 = 3(Φ 0 + HΨ ) − [k 2 + 3(H2 − H0 )] (1.22)
k2
One can recast (1.13) in the following way
a2 ρ
Ψ =− µ∆ (1.23)
2Mp2 k 2
a2 ρ 0
Ψ0 =− (µ − µH)∆ + µ∆0 , Φ 0 = η 0 Ψ + ηΨ 0
2 2
(1.24)
2Mp k
By solving these equations one finds ∆ and θ, then the solutions for the potentials follow from
(1.13) and (1.14).
In [38] it is shown that in any theory of gravity in which the energy-momentum tensor is
covariantly conserved, if the entropy perturbations can be neglected and the long wavelength
curvature perturbations have a well defined behaviour in the infrared limit7 , the following
consistency relation must hold [63]:
!0
1 a2 Φ
0
1 a
2
ζ + O(k ) = 2 +Ψ −Φ (1.27)
a H a H
where ζ is the curvature perturbation on hypersurfaces of uniform density [79] and k (in this
section) denotes the comoving wave number. On super-horizon scales (k << aH) it is indepen-
dent by time if one neglects terms of order O(k 2 ζ), thus one can use this expression to define ζ
6 It is also negligible for baryons at late time.
7 We are assuming also flat background
8
1.2. Parametrized evolution of linear cosmological perturbations in a
general theory of gravity
in this way8
Φ 0 + HΨ
ζ=Φ+ , (k = 0) (1.28)
H(1 − H0 /H2 )
The conservation of this quantity on super-horizon scales yields a second order differential
equation for Φ
!
Ḧ Ḣ Ḧ
ζ̇ = Φ̈ − Φ̇ + Ψ̇ + − Ψ = 0, (k = 0) (1.29)
Ḣ H Ḣ
where here the overdot denotes derivative with respect to lna 9 . Thus, once the relation be-
tween the two gravitational potentials is specified, i.e. once the gravitational slip is known,
this equation solves the evolution of linear perturbations on super-horizon scales. Then, if all
the assumptions we made hold, the time and scale dependence in this regime must factorize
and thus one obtains
One can show that this equation can be equivalently obtained by combining (1.10), (1.11),
(1.13) and (1.14) in the limit k 2 /(a2 H 2 µη) → 0 and assuming that η and µ are close to their
GR values10 . The equation (1.32) shows that in this regime the scalar potential evolution is
independent from µ, this means that on super-horizon scales the only degree of freedom, for a
generic theory of gravity, is η.
This regime corresponds to k >> aH and in this limit the equation for θ and ∆ are the following:
∆0 = −θ (1.33)
a2 ρ
θ 0 = −Hθ − µ∆ (1.34)
2Mp2
a2 ρ
=⇒ ∆00 + H∆0 − µ∆ = 0 (1.35)
2Mp2
This is a second order differential equation in ∆ where the only phenomenological function
that appears is µ; it means that on sub-horizon scales the growth of matter perturbations is
affected only by µ.
8 This approximation is valid if the Jeans length is smaller than the cosmological scale we are considering; this is
true for z < 30 [38].
9 In terms of derivatives of ”lna” and H = da/dt (with t the standard cosmological time) the curvature perturba-
tion is ζ = Φ − (Φ̇ + Ψ )H/ Ḣ, where ˙≡ d/dlna.
10 Such an assumption is indeed in well accordance with observations.
9
1. Parametrizing deviations from GR in the evolution of cosmological
perturbations
1.2.3 The ΛCDM limit
In this limit µ and η are equal to one so that the equations (1.25) and (1.26) become
2 −1
2 3a ρ
h i
0 2 2 0
∆ = −θ k + 3(H − H ) k + (1.36)
2Mp2
a2 ρ
θ 0 = −Hθ − ∆ (1.37)
2Mp2
if the radiation component can be neglected then we have 3(H2 − H0 ) = 3a2 ρ/2Mp2 and thus the
equation (1.36) reduces to
∆0 = −θ (1.38)
this last equation combined with (1.37) gives the second order differential equation for ∆
a2 ρ
∆00 + H∆0 − ∆=0 (1.39)
2Mp2
from the above expression we can see that it looks like the equation obtained in the sub-horizon
regime except that, in the first, there is a factor µ in the expression. This shows that in this case
the time evolution of ∆ is scale-independent.
The parametrized post-Friedmann (PPF) formalism has been introduced with the purpose of
parametrizing in a phenomenological and model-independent way the new degrees of freedom
arising from modification of general relativity at cosmological scales; it was the analogous of
the parametrized post-Newtonian (PPN) formalism which was introduced in the 1970s to test
alternative gravity theories in the solar system and also in binary systems11 . Indeed, PPN and
PPF, can be thought of as complementary parametrizations that cover different gravitational
regimes.
There are different formulations of the PPF parametrization, among them we find the one pro-
posed by Hu and Sawicki in [5]. In this paper they consider three different regimes of modified
gravity: a) the super-horizon scales in which the evolution of cosmological perturbations has
to be compatible with the background expansion; b) the intermediate scales, i.e. quasi-static
regime, in which the Poisson equation is modified; c)the non-linear regime in which the addi-
tional degrees of freedom must be suppressed in order to satisfy local gravity constraints. Thus,
they develop a PPF framework that includes all these three regimes and which gives rise to ac-
celerated expansion without dark energy. Their parametrization consists in three functions
which characterize deviations from GR and one parameter that controls the transition between
the first two regimes. In particular, in the super-horizon regime and in the quasi-static regime,
they define the metric ratio function that in the Newtonian gauge is
Φ −Ψ
g(a, k) = (1.40)
Φ +Ψ
this function specifies the relation between the two scalar potential and determines the evolu-
tion of the metric fluctuations. In particular, in [5], is pointed out that even if the definition
11 The PPN formalism consist in a parametrization of the metric tensor that satisfies a set of reasonable assump-
tions in the slow-motion, weak-gravitational field limit.
10
1.3. Other parametrizations
of the function holds in the two regimes, the metric evolution is not the same, indeed in the
quasi-static regime one can have variations of the effective Newton’s constant. In terms of the
gravitational slip parameter η this function is
η −1
g= (1.41)
η +1
so that a non zero value implies deviation from GR. Then using the Weyl potential ΦW = (Φ +
Ψ )/2 one has that
Φ = (1 + g)ΦW , Ψ = (1 − g)ΦW (1.42)
This means that under the assumptions made in section 1.2.1, on super-horizon scales, the
scalar metric perturbations can be completely specified by the metric ratio ”g” and ”H”. In
GR this function is completely determined by the ratio of anisotropic stress to energy density;
but since the matter anisotropic stress is negligible at linear order at late time, this quantity is
determined by the anisotropic stress of dark energy; however, in dark energy models based on
scalar fields it is also negligible at linear order giving g = 0.
The other two functions used by Hu and Sawicki are fζ (a) and fG (a). The former expresses
the super-horizon relationship between the metric and density, in particular, the choice of this
function is not important from an observational point of view. The latest is introduced for the
quasi-static regime to allow modification of the effective gravitational coupling in the following
way
4πG 2
k 2 ΦW = a ρ∆ (1.43)
1 + fG
The parameter introduced in order to make a bridge between the two linear regimes is cΓ (for
details see [5].
Finally for the non-linear regime, in order to satisfy local gravity constraints and thus to sup-
press additional degrees of freedom, they develop a non-linear ansatz based on the halo model
of non-linear clustering that gives a qualitative description of the main features of this regime
in modified gravity. In particular, they decompose the non-linear matter power spectrum in
the sum of two pieces, one accounts for the correlations between DM halos and the other for
the correlations within DM halos. Thus it can be written as
where PL is the linear power spectrum of the density fluctuations and the expressions for Ii (k)
(i = 1, 2) can be found in [5]).
An other formulation of the PPF formalism has been introduced in [6]. The idea is to construct
a parametrization that allows for modifications of the Einstein’s field equations without speci-
fying any precise form. In particular, these modifications can introduce new scalar degrees of
freedom and also they can modify the degrees of freedom already present in GR.
The starting point is to add to the linearised field equations a piece arising from these modifi-
cations
δGµν = 8πGδTµν + δUµν
metric d.o.f . matter
(1.45)
δUµν = δUµν + δUµν + δUµν
Thus they decompose the additional term δUµν in three parts among which the third can be
eliminated in favour of the first two. Therefore the additional terms due to modified gravity are
those arising from scalar perturbation of the metric and of the new degrees of freedom. Then,
they choose, as derivative order of the parametrization, the second order so that the Ostrograd-
ski’s instability is avoided. Moreover they require gauge-invariance of the parametrization and
that the coefficients of the parametrization must be functions of the zero order background
quantities, for FRW universe this means that they must be functions of scale and time.
11
1. Parametrizing deviations from GR in the evolution of cosmological
perturbations
metric and δU d.o.f .
They expand δUµν µν in terms of gauge-invariant variables, in particular, for the
first, they choose to use the Bardeen potential Φ̂ and a linear combination of the two Bardeen
potentials Γ̂ = (Φ̂˙ + H Ψ̂ )/k; for the latest they introduce the gauge-invariant variable χ̂, that
represents a gauge-invariant parametrization of the additional scalar degree of freedom. The
coefficients of this expansion, which are function of time and scale, are twenty-two and are not
all independent; moreover their number can be reduced by imposing restrictions on the types
of theories of gravity one is considering. This parametrization covers scales at which the linear
perturbation theory holds, a detailed analysis can be found in [5].
Thus this parameter measures the difference between the two scalar potentials or equivalently,
the scalar shear fluctuations in a dark energy component. It can be thought as the similar of
the PPN parameter γ 12 in a cosmological context. Its relation with this parameter is $ ' 1 − γ
in the limit of weak departure from GR. On scales of tens kiloparsec the constraint on the
PPN parameter translates in a constraint on the Caldwell’s parameter so that $ = 0.02 ± 0.07
(68% CL); while on few hundred kiloparsec scales comparison between weak lensing and X-ray
based masses gives the value $ = 0.03 ± 0.10 (68% CL) [62].
A non-zero value of the Caldwell’s gravitational slip affect the anisotropy constraint (eq. (1.7))
in the following way
k 2 (Ψ − Φ) = −12πGa2 ρ̄σ |γ,ν + $k 2 Φ (1.47)
where the first contribution is that due to the free-streaming of the relativistic particles and it
is negligible at late time. Thus, a non vanishing value of this parameter, can be interpreted as
a non-standard relation between the two scalar potentials (i.e. Φ , Ψ ) due to modifications of
the gravity; the specific form of $ can be obtained for a given theory of gravity.
In [62] the authors start from a universe described by the ΛCDM model and introduce the
Caldwell’s gravitational slip as an extension of this simple model that accounts for the acceler-
ated expansion of the late time universe when the dark energy component becomes important;
the resulting model was denoted as $ΛCDM in [39]. Since in this model the gravitational slip
appears only at late time they use the following phenomenological parametrization
ρDE (a, k)
$(a, k̄) = $0 (a, k) (1.48)
ρm (a, k)
where the parameter $0 has to be |$0 | . 10 to satisfy local gravity constraints and |$0 | . 100
on the scale of a galaxy halo.
Then one can expand to first order in perturbations around the FRW background obtaining
ρ̄DE (a) Ω
$(a, k) ' $0 = $0 DE (1 + z)−3 = $(a) (1.49)
ρ̄m (a) Ωm
where ρ̄DE and ρ̄m are the background values of respectively the DE density and the matter
density.
12 It is the PPN parameter which measures the amount of spacetime curvature per unit mass; it was strongly
constrained to be close to one (that is the GR value of this parameter) in the solar system by the Cassini mission [70]
which measured γ −1 = (2.1±2.3)·10−5 . On galactic size scales (kiloparsec scales) its value has been measured to be
0.98 ± 0.07 (68% CL) by comparing, in two different ways, the mass of 15 elliptical lensing galaxies from the Sloan
Lens ACS Survey [71].
12
1.3. Other parametrizations
From (1.49) one can see that, at the lowest order in perturbations, the scale dependence disap-
pears, so the Caldwell’s gravitational slip depends only by time (or redshift), while the gravi-
tational slip parameter η introduced in section (1.2) is function of time and scale. If the scale
dependence can be neglected, the relation between the two gravitational slip parameters is the
following
$(a) = η(a)−1 − 1 (1.50)
The EG parameter is a useful quantity because its determination combines three different
probes of LSS [7]- [8]: galaxy-galaxy lensing, galaxy clustering and galaxy redshift distortions.
This combination ensures that this parameter is insensitive to the galaxy bias but very sensitive
to modification of gravity that generates gravitational slip or a difference in the rate of growth
of structures with respect to GR. Indeed, the galaxy-galaxy lensing, is sensitive to the Weyl
potential ΦW = (Ψ + Φ)/2, while the galaxy clustering depends on the Newtonian potential Ψ
and, therefore, they can be used to probe any non-standard relation between the two scalar
potentials. On the other hand, the galaxy redshift distortions is sensitive to the growth of the
structures and thus can be used to discriminate among different theories of gravity that predict
different growth rates.
Following Zhang et al. (2007) [7], we define EG as the ratio of the Laplacian of the Weyl poten-
tial over the peculiar velocity divergence
∇2 (Ψ + Φ)
" #
EG (k) = (1.51)
3H02 (1 + z)θ(k) z̄,k=`/ χ̄
where χ̄ is the mean angular comoving distance at the redshift z̄ and θ = ∇~v /H(z) is the velocity
field divergence. This latter can be written in term of the matter perturbation and the growth
factor (f (z)) using the equation (1.20) in the sub-horizon limit as θ = f (z)δm . In the ΛCDM
model EG is scale-independent and is given by
GR Ω0m
EG = (1.52)
f (z)
where Ω0m is the relative matter density today. The EG parameter is scale-independent also in
models with large sound speed and negligible anisotropic stress, but potentially, in a general
theory of gravity, this parameter can exhibit scale-dependence.
The quantity defined above corresponds to the expectation value of the estimator ÊG , i.e.
hÊG i = EG . The estimator ÊG is obtained by the ratio of the crossing galaxy-lensing power
spectrum over the velocity galaxy cross power spectrum
here Ckg (`, ∆`) is the band power of the crossing galaxy-lensing power spectrum centred in `
with band width ∆`; P α is the band power of the galaxy velocity cross power spectrum Pgθ
between kα and kα+1 and fα is the corresponding weighting function.
In order to obtain this estimator they use the relation between the matter overdensity and
the peculiar velocity field (equation 1.20 on sub-horizon scale); in this way, by measuring the
velocity field at a given redshift, one is able to extract the corresponding matter overdensity.
The velocity field can be obtained by the redshift space distortions (RSD) of the galaxy power
spectrum; while the lensing signal at the corresponding redshift is measured using the cross
correlation of the galaxies and the lensing maps reconstructed from background galaxies.
13
Chapter 2
Gravitational slip
The gravitational slip is the mismatch between the two scalar potentials appearing in the per-
turbed FRW metric, in the Poisson gauge (or in the Newtonian gauge), due to modifications of
the geometrical part of the Einstein’s field equations. It is a general signature of non-minimal
gravitational coupling and thus, it is a key quantity in order to characterize the nature of the
dark energy.
The traceless part of the (ij)-component of the linearized Einstein’s equations, i.e. the anisotropy
constraint (or shear equation), in a general theory of gravity can be recast in the following
form [32]- [33]
Ψ (a, k) − Φ(a, k) = σ (a)Π + πm (2.1)
where the quantity Π is a function of the parameters which characterize the theory, while σ (a)
is a background function and it depends only by the parameters of the theory. Finally πm is
the anisotropic stress sourced by the matter sector and it is non vanishing for imperfect fluids.
In GR we have seen that the anisotropy constraint equation is given by equation (1.7) and
specifying the anisotropic stress contribution one gets
one can see that the only source of anisotropic stress comes from the quadrupole moments
of relativistic particles, i.e. photons and neutrinos, that are generated by the free-streaming
of these particles. However, the contribution of photons is very small and that of neutrinos
becomes negligible at late time1 , thus, at linear order in perturbations, it disappears and the
relation between the two gravitational potentials becomes trivial Φ = Ψ , this means that η = 1.
At second order in perturbations, the contribution of the anisotropic stress is not vanishing but
it is constrained in the late universe by the bound |η − 1| . 10−3 .
However, if an imperfect fluid drove the cosmic acceleration, its contribution to the anisotropic
stress might be not negligible at linear order.
In [33]- [32] two fundamental point are highlighted:
• There is a subtle distinction between anisotropic stress sourced by imperfect-fluid and
gravitational slip, the former is a property of matter, the latest is a purely geometrical
effect.They both give rise to a difference between the two scalar potentials but, as we will
see later, only the second leads to a non-standard propagation of the tensor modes (GWs)
in many theories of modified gravity.
• The gravitational slip parameter η(a, k) = Φ/Ψ can depend in principle by the reference
frame; however, one can remove this ambiguity by adopting an operational definition
1 The contribution of neutrinos to the anisotropic stress in the ΛCDM scenario is significant only after decou-
ρ +p
pling during the radiation-dominated era when η − 1 = 25 ρ̄+ ν ν
p̄ ; while during the matter-dominated era this
−1
contribution becomes negligible since it scales as η − 1 ∝ a .
14
2.1. Probing gravitational slip with large-scale structures
In section 1.2.1 we have seen that on large scale the only degree of freedom necessary to fully
specify a general theory of gravity is η. In [63] Bertschinger and Zukin distinguish between
two classes of theories of gravity:
• Scale-independent modified gravity models: are theories in which the terms of order O(ζk 2 )
can be neglected also at sub-horizon scales which are higher than Jeans length scale.
Then in this case the gravitational slip completely characterizes deviations from GR and
following [63], it can be parametrized as follows
1 + α1 k 2 as 1 + β1 k 2 as
µ(a, k) = η(a, k) = (2.4)
1 + α2 k 2 as 1 + β2 k 2 as
where s > 0 is a constant and αi , βi (i = 1, 2) are arbitrary constants with the units of
length squared. Moreover, to avoid divergence in the above expressions, one must require
α2 , β2 > 0. Finally, in order to have a theory of gravity which is attractive, one needs
α1 > 0. In this parametrization there are at least three physical scales; in particular,
considering theories in which the αi and βi are all comparable, we have: the Hubble scale
√
(1/H), the transition scale α1 = a1+s/2 and the non -linear length scale for LSS formation
(∼ 10 today).
At sub-horizon scales and at late time, the evolution of growth of density perturbation2 is given
by
4πGa2 ρδ
scale-independent case
δ00 + Hδ0 = −k 2 Ψ = (2.5)
2
4πGµ(k, a)a ρδ scale-dependent case
where we have used (1.20)-(1.22). This equation has two solutions: a decaying mode and a
growing mode. We are interesting in the latter which is the solution that leads to the structures
2 We are considering non relativistic pressureless perturbations or CDM.
15
2. Gravitational slip
formation. In GR the linear growth factor of perturbations, denoted as D+ (a), is the quantity
that relates the overdensity δ(a) at a given ”a” to that at some initial ”ai ”
D+ (a)
δ(a) = δ(a ) (2.6)
D+ (ai ) i
where D+ (ai ) and δ(ai ) are fixed by the initial conditions. From this quantity one can define
an other quantity relevant from an observational point of view, this is the growth rate and is
defined as follows
d lnD
f (a) = (2.7)
d lna
Then one can define the theoretical matter power spectrum as
(3)
hδ(z, k)δ(z, k0 )i = (2π)3 P (z, k)δD (k − k0 ) (2.8)
P (z, k) = As k ns T 2 (k)G2 (z) (2.9)
(3)
where δD is the delta of Dirac, As k ns is the primordial fluctuations power spectrum (with ns =
0.9652±0.0062 from observation); T (k) is the transfer function which describes the evolution of
perturbations through the horizon crossing and the radiation-matter transition. Finally G(a) ≡
D+ (a)/a is the scale-independent growth at late time.
Then, to connect this theoretical prediction with observations coming from galaxy surveys, one
has to account for the galaxy bias b(z, k) = δg (z, k)/δm (z, k) and the redshift space distortions
(RSD) that introduces a factor f (z)µ2RSD (µRSD is the cosine of the angle to the line of sight.
Thus the galaxy power spectrum is
h i2
s
Pgg (z, k, µRSD ) = As k ns T 2 (k)G2 (z) b(z, k) + f (z)µ2RSD (2.10)
On small scales, the power spectrum must be modified to include non-linear effects.
Departures from GR can affect T (k), G(z) and also the growth rate f (z). As a consequence, the
galaxy power spectrum can be modified in shape and amplitude.
In the scale-independent class of models, as we just said above, the deviations from GR are
characterized by η(a), therefore, in these theories, the matter power spectrum can be affected
only through an enhancement or diminution of the Newtonian potential (with respect to GR
case) that correspond respectively to η < 1 and η > 1.
Instead, in the scale-dependent case, a new degree of freedom is introduced in order to char-
acterize modifications of gravity (in the above parametrization µ(a, k)) and, as we have seen in
section 1.2.3, in this case the modification of the growth rate depends only by µ(a, k) which is
independent by η(a, k).
The observations of CMB (Cosmic Microwave Background) are fundamental in order to under-
stand the early universe and its evolution. In particular, its temperature anisotropies and its
polarization maps provide a powerful tool in testing modified gravity.
The CMB temperature anisotropies were measured for the first time by the COBE satellite in
1992 [77], then other experiments gave more precise measures of these temperature fluctua-
tions. The CMB polarization maps can be decomposed in curl-free E-modes and gradient-free
B-modes; these latter are generated by gravitational lensing of CMB by LSS. B-modes have been
observed by two independent experiments [21]- [22] and are very interesting because they give
us information about primordial gravitational waves 3 as well as about the lensing potential.
3 By measuring primordial tensor perturbations, i.e GWs, one obtains information about the energy scale of
inflation.
16
2.1. Probing gravitational slip with large-scale structures
The CMB power spectrum is affected by modifications of gravity in several ways, as an example
it is modified at low multipoles (large angular scales) by the late ISW (Integrated-Sachs-Wolfe)
effect caused by the time variation of the gravitational potentials. Moreover, modified gravity
theories that predict gravitational slip, affect the B-mode spectra in two different ways: 1) the
first effect is the modification of the lensing potential as well as of the TT, EE and BB spectra;
2) the second effect is a possible shift of the position of the primordial B-mode peak which is
determined by the tensor speed cT , thus a theory which predicts cT , 1 predicts also a change
in the position of this peak.
The ISW effect is a secondary anisotropy of the CMB temperature fluctuations due to the fact
that photons from the last scattering surface reach us crossing potential wells and voids. At
late time (z < 2), this effect can be induced only by departures from GR: the difference in the
energy that photons gain by falling down to the potential wells and that they loose by climbing
out of these wells can differ from zero because of modification of gravity or the presence of
a dark energy component; while the photons which cross voids reduce their energy and thus
their temperature.
The ISW effect is given by the following integral
Z τ∗
∆T ∂(Ψ + Φ)
(n̂) = − dτ (2.11)
T τ0 ∂τ
here T is the CMB temperature, τ∗ the conformal time at the last scattering surface, τ0 is the
conformal time of the observer and n̂ the photon’s direction.
This effect modifies the CMB temperature power spectrum at larger angular scales (` < 10).
The presence of gravitational slip, as well as a different evolution in the scalar potentials,
modifies the CMB temperature power spectrum. In particular, in [63], they show that for
scale-independent models with 0.2 < η < 1 this effect reduces the low multipoles part of this
spectrum because of the destructive interference with the primary anisotropy contribution.
As photons travel from the last scattering surface to us, their paths are gravitationally deflected
by the LSS. These deviations induce very small distortions in the CMB temperature and polar-
ization maps that, in turn, lead to the generation of 3 and 4 point correlation functions and to
the conversion of the E-mode polarization into B-modes. These non-Gaussian B-mode signals
can be used to constrain different cosmological parameters, among these the sum of neutrino’s
masses.
In order to calculate the weak lensing effect on the CMB anisotropies, we define the lensing
potential as [78] Z χ∗
f (χ − χ)
ψ(n̂) = −2 dχ K ∗ Φ (χn̂; τ0 − χ) (2.12)
0 fK (χ∗ )fK (χ) W
where ΦW = (Ψ + Φ)/2 is the Weyl gravitational potential; χ∗ is the conformal distance to the
surface of last scattering, τ0 − χ is the conformal time at which the photon is at the position
χn̂. The deflection angle, i.e the angle formed by the deflected direction and the one that the
photon would have in an unperturbed universe, is denoted as α and it is related, in the first
order approximation, with the lensing potential by α = ∇ψ. Then, the lensed CMB temperature
in a given direction n̂, is denoted with Te(n̂) and it is equal to the unlensed temperature, T, in
the deflected direction n̂0 = n̂ + α, that is Te(n̂) = T (n̂ + α).
Since we are considering a spatially-flat universe we can set in the above definition fK (χ) = χ.
17
2. Gravitational slip
P
Then, one can expand the lensing potential in spherical harmonics ψ = lm ψlm Ylm .
The power spectrum of the lensing potential in a given direction is
ψ
hψlm ψl∗0 m0 i = δll 0 δmm0 Cl (2.13)
Z "Z χ #2
ψ dk ∗
χ∗ − χ
Cl = 16π P (k) TΦW (k; τ0 − χ)jl (kχ ) (2.14)
k R 0 χ∗ χ
here TΦW is the transfer function and in a linear theory it relates the primordial comoving
curvature perturbation to the Weyl potential t through ΦW (k, τ) = TΦW (k; τ)R(k); while j` (r) =
√
π/2rJl+1/2 ()r with J` (r) are the Bessel functions of first kind. The Weyl potential can be ex-
pressed in terms of the gravitational slip and of the Newtonian potential, therefore
(1 + η) (1 + η)2
ΦW = Ψ =⇒ PΦW = PΨ (2.15)
2 4
one finds that if η , 1 the power spectrum of the Weyl potential acquires a factor (1 + η)2 /2
inside the integral of (2.14). Moreover, in [78], it is shown that the lensed B-mode power
spectrum for ` << 1000 and at lowest order can be approximated as follows
Z
B 1 ψ
C` '
e d` 0 ` 0 C`0 CE`0 (2.16)
4
where CE` is the unlensed E-mode spectrum. Thus if η , 1, the presence of a gravitational slip
ψ
affects the power spectrum of the lensing potential (C` ) by a factor (1+η)2 /4 and since it enters
linearly in (2.16), also the B-mode power spectrum gets enhanced, at large scales, by the same
factor.
The accelerated expansion of the universe leads to the decay of the gravitational potentials and
this decay in turn induces the ISW that, as we have just mentioned, is due to the time variation
of the scalar potentials and it is responsible for secondary anisotropies of the CMB tempera-
ture fluctuations. A consequence of this fact is that, if there is a cluster of galaxies in a given
direction of the sky, it is very likely to observe a correlation in the temperature anisotropies of
the CMB in the same direction if the photons of the CMB have crossed that region during the
accelerated expansion. The ΛCDM model predicts a positive correlation signal, but departures
from this model can potentially change the sign of this correlation. In order to identify this sig-
nal, one can define the two-point angular cross correlation function between the surveys of the
large scale structures and the CMB temperature anisotropy as follows [9]- [39]
where θ = |n̂1 − n̂2 |, ∆ISW (n̂) is the integral (2.11) and δLSS (n̂) is the density constrast of a clump
of luminous matter observed by a given survey in the direction n̂2 and it is given by
Z
δLSS (n̂2 ) = b dzφ(z)δm (n̂2 , z) (2.18)
here b is the bias of the galaxy, φ(z) the selected function of the survey and δm is the matter
density fluctuation.
Then one can expand C X (θ) into a Legendre series
∞
X
X 2l + 1
C (θ) = ClX Pl (cosθ) (2.19)
4π
l=2
18
2.2. Gravitational slip as signature of non-standard propagation of
tensor modes
where Pl (cosθ) are the Legendre polynomials and ClX is the power spectrum of the cross corre-
lation which is given by the following expression
Z
X 9 dk 2 ISW
Cl = 4π ∆ I (k)IlLSS (k) (2.20)
25 k R l
where ∆2R is the primordial power spectrum; IlISW (k) and IlLSS (k) are given by
Z
d((1 + η)Ψk )
IlISW (k) = − dze−κ(z) jl [kr(z)]
dz
Z (2.21)
IlLSS (k) = b dzφ(z)δk (z)jl [kr(z)]
where Ψk and δk are the Fourier components respectively of the Newtonian potential Ψ and
matter perturbation;
Rτ jl [kr(z)] are the spherical Bessel functions, r(z) the comoving distance at
”z” and κ(τ) = τ 0 dτ κ̇(τ) is the total optical depth.
Thus, from (2.20) and (2.21), one can see that a non-vanishing gravitational slip can affect not
only the value of IlISW (k) but in principle it could affect also its sign with respect to IlLSS (k).
As like stated previously, the propagation of the scalar tensor modes are affected by modifica-
tions of gravity, in particular the presence of gravitational slip implies a non-standard propa-
gation of these modes in a number of MG theories.
To understand what we mean by ”non-standard” propagation let us consider the perturbed
line element in a spatially flat FRW universe, neglecting the tensor and scalar perturbations, it
can be written in the following way
h i
ds2 = −a(τ)2 −dτ 2 + (δij + hij )dxi dxj (2.22)
where hij is a symmetric, traceless and divergence-free tensor which represents the tensor per-
turbation of the metric, i.e. the GWs.
The equation of motion for the tensor modes can be obtained by linearising the Einstein’s field
equation. In GR one gets the following equation (in the Fourier space)
where cT is the tensor speed and, in GR, it is equal to the speed of light, i.e ct = 1; the source
term ΠTij = (Tij − Tkk δij /3)(T T ) is the transverse-traceless projection of the anisotropic matter
stress tensor that is zero for perfect fluids. The prime stands for the derivation with respect to
conformal time. This equation describes the standard propagation of the GWs.
However, in a general theory of gravity, any additional degrees of freedom can alter this form,
thus, in such a theory, the propagation equation for the tensor modes can be written as follows
where
• ν ≡ H −1 dM∗2 /dt: this parameter characterize the running of the effective Planck mass,
M∗ , so that a theory of gravity that predicts an evolution in time of the Plank mass also
predicts a non-standard propagation of the GWs;
19
2. Gravitational slip
• cT : it is the tensor speed and, in general, it can differ from the speed of light4 , thus, if
cT , 1, the dispersion relation changes and also the propagation of the tensor modes;
• m2g : it is the squared mass of the graviton and it appears in massive bi-gravity;
• Γ γij : here γij is a transverse-traceless tensor which represent the source term; in partic-
ular, in the bimetric massive gravity, it is the tensor perturbation of the second metric;
while if matter has anisotropic stress, γij contains this contribution.
The first two quantities are defined in the matter Jordan frame. By comparing this equation
with (2.23) one can see that departures from GR affect the homogeneous part of the propaga-
tion equation and thus, the non-standard behaviour of the tensor modes, is due to these kind
of modifications. In particular, in scalar-tensor theories, where mg = 0, one can have ν , 0 and
cT , 1, so that the propagation of the GWs is modified; in what follows we will see that these
parameters are strictly related to the gravitational slip and in Horndeski gravity it implies a
non-standard propagation of GWs, while in beyond Horndeski theories it is not necessarily
true. Also in Einstein-Aether models and in bimetric gravity a non-vanishing gravitational slip
implies that the propagation of the GWs is modified.
4 However, there are very stringent constraints on this parameter, in particular, the recent event GW100817,
forces cT to be very close to one at low redshifts.
20
Chapter 3
In this chapter we introduce the Effective Field Theory (EFT) of DE which allows the unifi-
cation of the description of single scalar-field models of dark energy and modified gravity.
The advantage of this approach is that it is a very efficient way to describe a wide number of
existing theories in the same language and in a formalism that has a clear connection to cos-
mological observations. In the following, we present two different EFT parametrizations : the
first is the most general EFT approach to dark energy while, the second, the α-parametrization,
is specifically designed to study the phenomenological aspects of Horndeski theories and, by
adding new parameters, also of its extensions, among which we consider the ”beyond Horn-
deski models”. Using these formalisms, we write down the modified Poisson equation and
the modified anisotropy constraint and we derive the analytic expressions for the phenomeno-
logical functions µ, η and Σ in the quasi-static approximation. In particular, using the second
parametrization, we analyse for what values of the α-parameters there is a non-vanishing grav-
itational slip.
The Effective Field Theory (EFT) approach was first applied in a cosmological context to study
inflation [41]- [42], then, it was extended to the DE sector [40]- [43] with the aim of unifying
the description of single-field models of DE and MG and also in order to have a theory that
was readily testable by observations.
The idea is to start by a perturbed FRW universe in which there are gravity, a single scalar
field and a matter sector obeying to the weak equivalence principle (WEP), i.e. it is assumed
the existence of a conformal frame, the Jordan frame, in which all matter fields are minimally
coupled to the metric. Then, it is possible to construct the most general action that is invariant
under time-dependent spatial diffeomorphism by choosing the unitary gauge. In this gauge
the dynamic of the scalar field perturbation is ”eaten” by the metric, i.e. δφ vanishes and
the time coordinate is function of the scalar field t = t(φ); thus the scalar field φ defines a
preferred slicing of the spacetime in space-like hypersurfaces where φ = const. Finally, the
time diffeomorphism invariance, can be restored via ”Stückelberg trick”, which consists in
performing an infinitesimal time diffeomorphism t → t + π(xµ ), xi → xi after which the scalar
degree of freedom reappears as a new field (π(x)) in the action.
21
3. The Effective Field Theory approach: unifying single-field models of
Dark Energy and Modified Gravity
The EFT action is given by [43]
Z m2
√
S = d 4 x −g 0 Ω(t)R + Λ(t) − c(t)δg 00
2
M 4 (t) M 4 (t)
+ 2 (δg 00 )2 + 3 (δg 00 )3 + . . .
2 2
3
M̄ (t) M̄ 2 (t) M̄ 2 (t) µ
− 1 δg 00 δK − 2 (δK)2 − 3 δKν δKµν + . . .
2 2 2
2
+ λ1 (t)(δR) + λ2 (t)δRµν δK + γ1 (t)Cµνσ λ C µνσ λ + γ2 (t)µνσ λ Cµν αβ Cσ λαβ + . . .
µν
2 µ ν 00 00 2 µν µ ν 00 00
+ m1 (t)n n ∂µ g ∂ν g + m2 (t)(g + n n )∂µ g ∂ν g + . . . + Sm [gµν , ψi ]
(3.1)
p
where hµν ≡ gµν + nµ nν is the induced spatial metric, nµ ≡ −∂µ φ/ −(∂φ2 ) is the normal to the
ρ
hypersurfaces of constant time and Kµν ≡ hµ ∇ρ nν is the extrinsic curvature; then δg 00 , δR,
(0)
δRµν , δK µν = Kµν − Kµν = Kµν + 3H(g µν + nµ nν ) and δK are the perturbations respectively to the
time-time component of the metric, the Ricci scalar, the Ricci tensor, the extrinsic curvature
tensor and to its trace; C µνσ λ is the Weyl tensor.
Moreover, we have m20 = (8πG)−1 (= Mpl 2
) (the bare squared Planck mass) and the operators Ω,
Λ, c, Mi , λi , γi and mi which are functions of time; the subset {Mi , λi , γi , mi } only affects the
behaviour of the perturbations.
However, one finds that only a reduced set of these operators are required to fully characterize
the linear perturbation theory. Thus, the EFT action describing the most general single-scalar
field model of dark energy, up to quadratic order in perturbations, is
Z m2
4 √
S = d x −g 0 Ω(t)R + Λ(t) − c(t)δg 00
2
M24 (t) 3
00 2 M̄1 (t) 00 M̄22 (t) 2
2 M̄3 (t) j M̂ 2 (t) 00 (3)
+ (δg ) − δg δK − (δK) − δKji δKi + δg δR
2 2 2 2 2
+ m22 (t)(g µν + nµ nν )∂µ g 00 ∂ν g 00 + Sm [gµν , ψi ]
(3.2)
here the term δR(3) is the perturbation of the three dimensional spatial Ricci scalar of constant-
time hypersurfaces.
This action encompasses a broad number of models, among which: DGP braneworld models,
Galileons models, ghost condensate and in particular it contains the Horndeski gravity and
beyond Horndeski theories.
From this action one can see that a theory belonging to this class can be specified by the fol-
lowing set of functions of time: {Ω(t), M13 (t), M24 (t), M32 (t), M24 (t), M̂ 2 (t), m22 (t)}.
22
3.1. The most general EFT parametrization
By varying the action with respect to the metric and, assuming a spatially flat universe, one
gets the background evolution equations
Ω̇
3m20 Ω H 2 + H = ρ̄m − Λ + 2c (3.4)
Ω
Ω̈ Ω̇
2 2
m0 Ω 3H + 2Ḣ + + 2H = −Λ − p̄m (3.5)
Ω Ω
Then, using these expressions, one finds that two of the first three functions can be expressed
in terms of the others parameters, so only one function of the set {Ω(t), Λ(t), c(t)} is needed to
characterize the theory at background level.
ρ̄m + p̄m Ω̈ H Ω̇
2
c(t) = − − m0 Ω Ḣ + − (3.6)
2 2Ω 2 Ω
Ω̈ Ω̇
Λ(t) = −p̄m − m20 Ω 3H 2 + 2Ḣ + + 2H (3.7)
Ω Ω
Finally, one can recast the equations (3.4)-(3.5) introducing the DE energy density and the DE
pressure.
1
H2 = (ρ̄m + ρDE ) (3.8)
3Ωm20
1
Ḣ = − (p̄m + pDE + ρ̄m + ρDE ) (3.9)
2Ωm20
where
ρDE = −Λ − 2c − 3m20 H Ω̇, pDE = Λ + m20 Ω̈ + 2m20 H Ω̇ (3.10)
By combining the background equations and the matter continuity equation one obtains the
continuity equation for dark energy
At this stage we can restore the time diffeomorphism invariance by using the above-mentioned
”Stückelberg trick” so that the scalar degree of freedom explicitly appears in the action as π,
which is the perturbed part of the scalar field.
The perturbed field equations can be obtained by varying the action (3.2) with respect to the
metric and then by subtracting the background contribution (3.4)- (3.5).
The resulting equations can be written in this form
µ µ (m) µ (DE)
m20 ΩδGν = δTν + δTν (3.12)
From them, it is straightforward to obtain the Poisson equation and the anisotropy constraint
k2
2m20 Ω Φ = −ρ̄m ∆ + ∆P (3.13)
a2
k2
m20 Ω (Φ − Ψ ) = p̄m Π + ∆S (3.14)
a2
23
3. The Effective Field Theory approach: unifying single-field models of
Dark Energy and Modified Gravity
with
k2 k2
∆P = − 2c(π̇ − Ψ ) + m20 Ω̇ − 3ḢΦ + 2 π − 3(Φ̇ + HΨ ) − 4M24 (π̇ − Ψ ) − M¯13 3Φ̇ + 3HΨ + 3Ḣπ − 2 π
a a
k 2 k 2 k 2
− 2H M¯32 2 π − 4M̂ 2 2 (Φ + Hπ) + 8m22 2 (Ψ − π̇)
a a a
(3.15)
k2 2k
2 2M̄˙ 3 2k
2
∆S =m20 Ω̇ π − M̄3 π̇ + H + π + 2M̂ (Ψ − π̇)
a2 a2 M̄3 a2
(3.16)
In order to derive analytic expression for the functions η(a, k), µ(a, k) and Σ(a, k) one needs the
equations (3.13)-(3.14) plus the equation of motion for π (obtained by varying the action with
respect to this field) in the quasi-static limit.
The quasi-static approximation consists in neglecting the time variation of all parameters and
perturbations of the metric which are assumed to be small with respect to the Hubble time
where X here stands for a generic perturbation of the metric or the scalar field. In particular,
in [14], it is shown that this approximation breaks down outside of the sound horizon rather
than outside the Hubble horizon.
In this limit the dominant contributions to the perturbation equations are those containing the
terms δm and k 2 /a2 (i.e. the terms involving spatial derivative of the fields)
k2
2
|X| >> H 2 |X| (3.18)
a
This approximation works well for sub-horizon perturbation when, the mass of the scalar field,
is at most of the order of the Hubble parameter (H) and thus, the oscillating mode of the
scalar field perturbation, is suppressed relative to the matter-induced mode; while, when this
mass is larger than H, the oscillating mode of the scalar field cannot be neglected and, this
approximation, does not hold anymore.
Then, if this approximation can be applied, one obtains the following system of equations in
the Fourier space
k2 k2 k2
A1 Φ + A2 π + A 3 Ψ ' −ρ̄m ∆ (3.19)
a2 a2 a2
B1 Ψ + B2 Φ + B3 π ' 0 (3.20)
k2 k2 k2
C1 2 Φ + C2 2 Ψ + Cπ + C3 2 π ' 0 (3.21)
a a a
24
3.2. EFT of Horndeski gravity and beyond Horndeski
a2
B 2 C 3 − B 3 C 1 + k 2 B2 C π
µ(a, k) = 2m20 2
a
A1 (B3 C2 − B1 C3 ) + A2 (B1 C1 − B2 C2 ) + A3 (B2 C3 − B3 C1 ) − k 2 (A1 B1 − A3 B2 )Cπ
(3.22)
Analogously, one finds the analytic expression for the gravitational slip parameter
2
B3 C2 − B1 C3 − ka2 B1 Cπ
η(a, k) = 2
(3.23)
B2 C3 − B3 C1 + ka2 B2 Cπ
(1+η)
Σ can be obtained by using the relation Σ = µ 2 .
a2
B 3 C2 − B1 C 3 + B2 C3 − B3 C1 + k 2 (B2 − B1 )C π
Σ(a, k) = m20 2
A1 (B3 C2 − B1 C3 ) + A2 (B1 C1 − B2 C2 ) + A3 (B2 C3 − B3 C1 ) − ka2 (A1 B1 − A3 B2 )Cπ
(3.24)
In [45]- [46] an alternative parametrization of the EFT theory of dark energy, specifically de-
signed for Horndeski models, was introduced; it can be extended to beyond Horndeski models
and others models by adding new parameters.
In this approach, the evolution of linear perturbations, can be fully specified by the back-
ground evolution H(t), the constant ρm0 , i.e. the matter density today, plus a set of four inde-
pendent functions of time {αM (t), αB (t), αT (t), αK (t)} (and αH (t) for beyond Horndeski). The
α-functions are arbitrary and independent of the first two, thus they characterize the physical
properties of the dark energy model. This formulation completely separates the background
description from the perturbations, contrary to the original EFT approach in which the opera-
tors {Ω(t), Λ(t), c(t)} enter both the background and the perturbations evolution.
The starting point is a general unitary gauge action expressed in the Arnowitt-Deser-Misner
(ADM) coordinates. The ADM coordinate are particularly useful when the gradient of the
scalar field is timelike and thus, one has a preferred slicing of the spacetime in spacelike hyper-
surfaces of constant φ which, in these coordinates, coincide with constant time hypersurfaces.
The line element in this metric is the following
here Ni is the shift function and N the lapse function; hij is the three-dimensional spatial
metric.
In these coordinates one has that
φ̇2 (t)
X = g 00 = − (3.26)
N2
∇µ φ
nµ = − √ = (n0 = −N , ni = 0) (3.27)
−X
1
Kµν = (gµσ + nσ nµ )∇σ nν =⇒ Kij = (ḣ − Di Nj − Dj Ni ) (3.28)
2N ij
25
3. The Effective Field Theory approach: unifying single-field models of
Dark Energy and Modified Gravity
The most general unitary gauge action in the ADM formalism is [46]- [47]
Z
√
Sg = d 4 x −gL(N , Kij , Rij , hij , Di , t) (3.29)
where Di is the covariant derivative associated to the three-dimensional spatial metric (hij ) and
the Lagrangian for Horndeski and beyond Horndeski model is of the following form
5
X
L= Li (3.30)
i=2
with
L2 = A2 (t, N ), (3.31)
L3 = A3 (t, N )K, (3.32)
L4 = A4 (t, N )(K 2 − Kij K ij ) + B4 (t, N )R, (3.33)
1
j
L5 = A5 (t, N )(K 3 − 3KKij K ij + 2Kij K ik Kk ) + B5 (t, N )K ij Rij − hij R (3.34)
2
In particular, Horndeski theories can be obtained by imposing
X
A4 = −B4 + 2XB4,X , A5 = − B5,X (3.35)
3
Then one can derive the evolution of the background by varying the action (3.29) with respect
to the lapse function and the scale factor. The evolution of linear perturbations can be studied
by expanding the Lagrangian in the action (3.29) up to quadratic order. In [46] it is shown that
the EFT quadratic action for linear perturbations which give rise to propagation equation with
no more than two space derivatives, can be written as follows
Z 2 √
3 M∗ δ h
(2) 3 ij 2
S = d xdta δKij δK − δK + (1 + αT ) R 3 + δ2 R
2 a (3.36)
2 2 (2)
+ αK H δN − 2αB HδKδN + (1 + αH )RδN + Sm [gµν , ψi ]
(2)
this is equivalent to the standard EFT action in which M̄22 = −M̄32 and m22 = 0. Here Sm is
the perturbed matter action in the Jordan frame and M∗ , is the effective squared Planck mass,
which, in general, is function of time. Thus, it is convenient to introduce the dimensionless
parameter:
1 d
αM ≡ lnM∗2 (3.37)
H dt
It is defined by (3.37) and measures the running of the Planck mass. This parameter, along
with αT (and αH in beyond Horndeski models), controls the existence of the gravitational slip
and it is related with the non-standard propagation of tensor modes [32]- [33]. In general,
when αM , 0 one also has αB , 0, this happens for all known models. Moreover, αM modifies
the evolution of the vector modes and affects the lensing potential as well as the amplitude of
the primordial polarization peak in the B-mode power spectrum.
26
3.2. EFT of Horndeski gravity and beyond Horndeski
It quantifies the deviation of the speed of gravitational waves from that of the light: αT = cT −1.
A non-zero value implies higher order derivative coupling of the scalar field to the metric
and this non-linearity gives rise to gravitational slip. Furthermore, this parameter affects the
position of the primordial peak in the B-mode power spectrum.
Kinetic braiding: αB
The non-vanishing of this function means a kinetic mixing between gravitational and scalar
degrees of freedom, i.e. a non-zero contribution of the term δKδN in the quadratic action
that implies a coupling between the metric and the scalar field. Furthermore, if αB , 0 this
parameter and αK determine a transition scale in the dynamics of the gravitational potentials
called ”braiding scale”, kB . In particular, αB controls whether dark energy clusters at all.
Kineticity: αK
This function measures the independent kinetic energy of the scalar degree of freedom, that
is the contribution deriving from the term δN δN . Thus, one has αK , 0 in minimal coupling
models of dark energy. Large values of this parameter lead to a suppression of the sound speed
of the scalar modes. As we just mentioned, it is related to the braiding scale and specifically it
determines the scale at which dark energy begins to cluster.
This parameter has been introduced to extend the α-parametrization to beyond Horndeski
models and thus, it is zero for Horndeski gravity. It characterizes theories in which there is a
kinetic coupling between the matter fields and the additional scalar degree of freedom. Fur-
thermore, it modifies the matter sound speed and contribute to gravitational slip. In [53], the
authors show that this parameter leads to a damping of the matter power spectrum on both
large and small scale. In the same paper, they show that it also affects the temperature CMB
power spectrum at low multipoles leading to an enhancement of this latter as well as to a de-
creasing of the lensing potential.
In conclusion these property functions represent the maximum information about the nature
of the dark energy that one can obtain from the evolution of linear cosmological perturbations.
In particular, we are interesting in theories in which there is gravitational slip, i.e. a non-
vanishing ∆S in (3.14). We can rewrite the anisotropy constraint (3.14) with (3.16) in terms of
the α functions
k2 2k
2
M∗2 [Φ − Ψ ] = p̄ m Π + M∗ 2 [HαM π − αT (Φ + Hπ) + αH (Ψ − π̇)] (3.38)
a2 a
It is easy to see that, since at late time the anisotropic stress sourced by matter is negligible,
i.e Π ∼ 0, the set {αT , αM , αH } determines the relation between the two scalar potentials, thus
it controls the existence of gravitational slip. In particular, in Horndeski gravity, it is different
from zero if M∗2 αT = −M̄22 = M̄32 = 2X[2G4,X − 2G5,φ − (φ̈ − φ̇H)G5,X ] , 0 or if we have a the-
ory with non-minimal coupling with gravity (Ω̇(t) , 0) (expression (3.16)), this means that to
have gravitational slip at least one of the parameters αT , αM must be non-vanishing. Indeed,
combining the expression in table A.2 of the appendix A, one finds M∗2 (1 + αT ) = m20 Ω; this
27
3. The Effective Field Theory approach: unifying single-field models of
Dark Energy and Modified Gravity
relation along with (3.37) and (3.14) shows that if αT = 0, the only way to have gravitational
slip is by a non-zero value of αM , i.e dM∗2 /dt , 0; this case is equivalent to have Ω̇ , 0, that is a
non-minimal coupling with gravity. In what follows we will see that there is a deep connection
among non-standard propagation of GWs, non-minimal coupling with gravity and gravita-
tional slip. In beyond Horndeski theories the situation is complicated by the presence of the
property function αH ; this last gives rise to additional contribution to the gravitational slip in
such a way that it is possible to have Φ , Ψ even if with αM = 0 = αT .
There are some stability conditions that a theory of gravity must satisfy in order to be a viable
model [18]- [19]:
a) no-ghost conditions: this kind of instabilities arises whenever in the high-k limit the
kinetic matrix has negative eigenvalues that destabilise the high energy vacuum state;
b) positive squared speeds of propagation: it is required in order to avoid gradient instabil-
ities at high values of k;
c) no-tachyonic instabilities: they are related with the presence of negative mass squared
terms.
These conditions impose constraints on the free parameters of the theory. In particular, the
first two stability conditions, in terms of the α-parametrization, are for tensor and scalar modes
respectively [45]- [46]
M∗2
QT = >0 (3.39)
8
cT2 = 1 + αT > 0 (3.40)
2M∗2 α
QS = >0 (3.41)
(2 − αB )2
(2 − αB )2
" ! !#
2 1 + αH Ḣ 2 d 1 + αH
cS = − 1 + αT − 2 1 + αM − 2 − −
2α 2 − αB H H dt 2 − αB
(3.42)
(1 + αH )2 (ρ̄m + p̄m )
− >0
α M∗2 H 2
where we have defined α ≡ αK + 32 αB2
From the above expressions one can see that if QT > 0 then (3.41) is satisfied if α > 0. In the
following we assume that these stability conditions hold.
In particular, for Horndeski theories, the expression (3.42) reduces to
(2 − αB )[Ḣ − 12 H 2 αB (1 + αT ) − H 2 (αM − αT )] − H α̇B + (ρ̄m + p̄m )/M∗2
cS2 = − >0 (3.43)
H 2α
Thus one finds that the only difference in the stability conditions between Horndeski and be-
yond Horndeski model is in the gradient of the scalar sector.
In 1974 Horndeski [65] derived the most general class of four-dimensional scalar-tensor the-
ories whose Lagrangian leads to second-order equations of motion; this property ensures that
this class of theories are free of Ostrogradski’s instabilities1 . In 2011, Deffayet et al. showed
1 This kind of instabilities are those that arise when there are ghost-like degrees of freedom and are usually
related to higher order time-derivative.
28
3.2. EFT of Horndeski gravity and beyond Horndeski
that the original Horndeski action is equivalent to the action of the so called generalized
Galileon theories [66]- [67].
The Horndeski action can be cast in the following form
Z 5
√ X
SH [gµν , φ] = d 4 x −g Li (3.44)
i=2
where φ is the scalar field and X = − 21 ∇µ φ∇µ φ is its kinetic term; then K(φ, X) and Gi (φ, X)(i =
3, 4, 5) are functions of the first two and we use the notations Gi,X ≡ ∂Gi /∂X, Gi,φ ≡ ∂Gi /∂φ, =
g µν ∇µ ∇ν . Finally, g is the determinant of the metric, R the Ricci scalar and Gµν the Einstein’s
tensor.
The Horndeski action includes a wide number of single scalar field models: to different choices
of the functions K and Gi correspond different theories. In the table 3.1 the functions K and
Gi are specified for some of the most popular theories belonging to this class.
Table 3.1: Explicit expressions of the Horndeski functions K and Gi (i = 3, 4, 5) for some of the well-
known models belonging to this class. In this table we adopt the notations of [23] except that we use K
instead of G2 and −G3 instead of G3 .
here V (φ) is the potential of the scalar field, Mpl is the bare Planck mass while m and c are constants;
we consider covariant Galileon models without field potential where ci (i = 1, 3, 4, 5) are dimensionless
coefficients and M is a constant having the dimension of a mass.
ωBD is the Brans-Dicke parameter; it can be shown that for ωBD = 0 this theory is equivalent to f (R)
gravity in the metric formalism with V (φ) = 21 Mpl
2
(RF −f ) [30]; while, for ωBD → ∞, we recover GR with
a quintessence scalar field.
F(R) ≡ ∂f /∂R corresponds to the propagating degree of freedom of the metric f (R) gravity. ξ (n) (φ) ≡
∂ξ n /∂φn where ξ(φ) is the coupling between the scalar field and the Gauss-Bonnet curvature invariant
defined by G = R2 − 4Rαβ Rαβ + Rαβγδ Rαβγδ .
To complete our description we add to (3.44) the action for the matter fields assuming a matter
perfect fluid minimally coupled to gravity. Then, the complete action is
29
3. The Effective Field Theory approach: unifying single-field models of
Dark Energy and Modified Gravity
Z 5
√
X
4
S= d x −g Li + Lm (gµν , ψi ) (3.49)
i=2
One can show that the EFT action (3.2) for the Horndeski theories corresponds to [44]
Z m2 M24 (t)
4 √ 0
S= d x −g Ω(t)R + Λ(t) − c(t)δg 00 + (δg 00 )2
2 2
(3.50)
M̄13 (t) 00 M̄22 (t) 2 i j 1 00 (3)
− δg δK − δK − δKj δKi − δg δR + Sm [gµν , ψi ]
2 2 2
thus we have m22 = 0 and 2M̂ 2 = M̄22 = −M̄32 ; in this case all the coefficients in table A.1 are
k-independent.
In [44] Bloomfield finds via software analysis the correspondences among the EFT operators
and the Horndeski free functions K(φ, X) and Gi (φ, X).
After introducing the α-functions, we can rewrite the expressions (3.22)-(3.23)-(3.24) special-
izing to Horndeski gravity (2M̂ 2 = M̄22 = −M̄32 and m22 = 0)
−1
m20 a2 MC2
#
a2 2 f3
"
µ(a, k) = 2 1 + 2 MC + (3.51)
2M∗2 f1 k 2 1 + αT
M∗ k
#−1
f5 a2 MC2
"
a2 2
η(a, k) = + 2
1 + 2 MC
(3.52)
f1 k 1 + αT k
−1
m20 a2 MC2
!#
f 5 a2 2
"
1 f3
Σ(a, k) = 1 + + 2 MC 1 + + (3.53)
2M∗2 f1 k 1 + αT 2M∗2 f1 k 2 1 + αT
here the quantities fi are functions of time and their definitions can be found in table A.3; in
particular, for the Horndeski gravity, their explicit expression are given by (A.4)-(A.6). The
new term MC2 ≡ Cf π is a transition scale associated to the ”Compton wavelength” of the scalar
1
field (∝ 1/MC ) which is a fundamental quantity in the study of LSS. It is the scale inside which
DE begins to cluster (if αB , 0) so that the growth of large-scale structures can differ from GR.
We consider these functions in two limiting cases:
1 f i−1
η0 = (1 + αT )−1 = η∞ = 5 = 1 + βB βξ /2 1 + αT + βξ2
h ih
(3.55) (3.58)
cT2 f1
m20 m2 h i
Σ0 = (2 + αT ) (3.56) Σ∞ = 02 2 + αT + βξ2 + βB βξ /2 (3.59)
2M∗2 2M∗
In the super-Compton limit it easy to see that any deviation from GR comes from the modifi-
cation of the tensor sector as well as from the running of the Planck mass. In particular, as we
will see in the next section, the value of the effective gravitational coupling in this limit at a = 1
must coincide with the value of the Newtonian constant measured in terrestrial experiments,
where the fifth force is hidden by screening mechanisms. This fact implies µ0 (a = 1) = 1 and
30
3.2. EFT of Horndeski gravity and beyond Horndeski
Σ0 (a = 1) = 12 (2 + αT )(1 + αT )−1 , but allows variations in the past. Then if one observes µ0 < 12 ,
in principle, if αT , 0, one can have Σ0 > 1. However, it would require a fine-tuning of the
functions involved in the above expressions, thus it is very likely that Σ0 < 1.
In the sub-Compton limit the quantities βξ and βB are defined in the appendix A ( respectively
(A.9) and (A.10)). They represent the contribution due to the propagation of the fifth force.
Since βξ2 > 0 it follows that, in general, µ∞ > µ0 and the condition µ∞ , µ0 or Σ∞ , Σ0 are
signature of a fifth force. Thus, if a scale dependence is detected, so that one can distinguish
between the two regimes, then a measurement of µ∞ < µ0 would disqualify Horndeski gravity.
In particular the presence of the fifth force term makes possible to have µ0 < 1 and µ∞ > 1. As
in the case k << aMC , since Σ and µ are controlled by the same functions one should expect
that if one measures µ∞ > 1(< 1) then it is likely to have Σ∞ > 1(< 1).
These two cases are the most interesting because the observational window it is very likely
to fall in one of these two regimes, indeed, for most of the well-known models, it is usually
completely inside the Compton wavelength or completely outside it3 . In particular if a scale
dependence is observed, these kind of models would be ruled out4 . Otherwise, if no scale-
dependence is detected, one has to test independently these two regimes.
Modified gravity can give rise to modifications of the effective gravitational couplings Gmatter ,
Glight and thus, can affect both the growth of large-scale structures and weak lensing. In gen-
eral, in accordance with what we have found in the previous section (µ∞ > µ0 ), this class of
models predicts enhanced growth within the Compton wavelength; this enhancement is due
to the fifth force induced by the scalar-matter interaction which is attractive for theoretically
consistent Horndeski theories. However, in order to satisfy the constraints imposed by solar
system tests of gravity and thus to suppress the propagation of the fifth force on small scale,
screening mechanisms [75] are required and should be environmental dependent. Among
these mechanisms, we find the chameleon mechanism [68]- [69] and the Vainshtein mecha-
nism [74].
The former is based on the following idea: there is an effective scalar potential which is the
sum of two terms, one of which is density-dependent; this potential gives rise to a mass term
which depends on the local matter density in such a way that on cosmological scale the mass
of the scalar field can be very small (∼ H0 ) meanwhile, in high-density regions, the scalar field
begins very massive and the the fifth force is hidden.
The Vainshtein mechanism is used in many alternative theories of gravity5 to hide the propaga-
tion of the additional degrees of freedom via non-linear effects. In particular, in the context of
Horndeski theories, this mechanism, also called ”k-mouflage”, is performed through the non-
linear self-interaction terms6 of the scalar field that lead to the decoupling of the field from
matter within the Vainshtein radius which depends on the surrounding density.
Therefore, if such a screening mechanism is at work, the fifth force is suppressed around local
source and the screened gravitational coupling is
1 1
Gsc (a) = = (3.60)
16πG4 (φ(a)) 8πM∗2 (a)
2 This can happens in self-accelerating models, in which the evolution of M 2 (a) can lead to a value of µ less then
∗ 0
one.
3 In general, for self-accelerating models it is of the order of the Hubble radius (λ ∼ 1/H), while in chameleon-
C
type models it is very small (λC < 1Mpc)
4 We may take as an example self-accelerating models where in general the mass of the scalar field is very small
, then it follows that the observations fall in the limit ka >> MC where the scale dependence disappears.
5 Fierz-Pauli massive gravity, Dvali–Gabadadze– Porrati (DGP), f(G) gravity
6 A well-known example is the term Xφ appearing in the Lagrangian density of the cubic Galileon theories.
31
3. The Effective Field Theory approach: unifying single-field models of
Dark Energy and Modified Gravity
where we have assumed that cT = 1; indeed such an assumption, after the aforementioned
GW170817 event, has proved to be a very good assumption.
Moreover, this screened value of the gravitational coupling at a = 1 (≡ a0 ) and in the limit k → 0,
has to match the current value of the effective gravitational coupling measured in Cavendish-
type experiments on Earth7
1
Gef f (a0 , k = 0) = Gµ(a0 , k = 0)) = (3.61)
8πM∗2 (a0 )
The relation (3.60) suggests the normalization µ0 (a0 ) = 1 so that Gsc (a0 , k = 0) = G and M∗2 (a0 ) =
(8πG)−1 . Thus one has that (in the limit k → 0)
2 2
M∗2 (a0 ) M∗ (a) < M∗ (a0 ) =⇒ µ0 (a) > 1
µ0 (a) = =⇒ (3.62)
M∗2 (a) M∗2 (a) > M∗2 (a0 ) =⇒ µ0 (a) < 1
So if in the past M∗2 (a) < M∗2 (a0 ) , we can potentially have Gef f < G, but it is not a sufficient
condition because in the full expression of µ , the effective Planck mass and the positive term
related with the propagation of the fifth force (βξ2 ) can combine in such a way that one can still
have Gef f > G.
Moreover, the screened gravitational coupling must satisfies the constraints imposed by Lunar
Laser Ranging (LLR) experiments [56]- [57]:
G˙sc −12 −1
< 1.3 · 10 yr = 0.02H0 =⇒ |α0M | < 0.02 (3.63)
Gsc
where Ġsc (a0 ) = −H0 αM /(8πM∗2 (a0 )). Thus, the bound on the variation of the gravitational
coupling implies a bound on the value of the running Planck mass parameter.
Additional bounds on the variation of the gravitational coupling come from observations of the
light elements’ abundances in the Universe [72], binary pulsars [73] and CMB. In particular,
this variation, must be consistent with the standard primordial nucleosynthesis scenario (BBN)
and this requires that the gravitational coupling at the BBN time should not differ by more than
10% from the value measured on Earth.
As just mentioned previously, a viable theory of gravity must satisfies a set of stability con-
ditions, among which there is the absence of ghost-instabilities, in particular according to
the Ostrogradski’s theorem, this kind of instability arises in theories with non-degenerate La-
grangian with higher time derivatives. For many years the Horndeski gravity has been con-
sidered the most general class of scalar-tensor theories which does not suffer from Ostrograd-
ski’s instabilities; recently, an extension of these theories has been introduced in [48]- [49]:
they are a new class of scalar-tensor theories, dubbed ”beyond Horndeski” models or GPLV
(Gleyzes-Langlois-Piazza-Vernizzi) theories; they are the minimum extension of Horndeski
gravity. Gleyzes et al. showed, via Hamiltonian analysis, that even if they have higher-order
derivative equations of motion, the beyond Horndeski theories are free from Ostrogradski’s
instabilities, indeed the true propagating degrees of freedom, which are three as like as in
Horndeski gravity, obey to second-order equations of motion.
The action of the beyond Horndeski models is that of the Horndeski theories plus two addi-
tional terms that modify L4 and L5
Z 5
√ X BH
SBH [gµν , φ] = d 4 x −g Li (3.64)
i=2
7 we live in a screened environment
32
3.2. EFT of Horndeski gravity and beyond Horndeski
LBH
2 = K(φ, X), (3.65)
LBH
3 = −G3 (φ, X)φ, (3.66)
LBH
4 = G4 (φ, X)R + G4,X (φ, X)[(φ)2 − (∇µ ∇ν φ)(∇µ ∇ν φ)]+
0 0 0 (3.67)
+ F4 (φ, X)µνρ σ µ ν ρ σ ∇µ φ∇µ0 φ∇ν ∇ν 0 φ∇ρ ∇ρ0 φ
LBH µ ν
5 = G5 (φ, X)Gµν (∇ ∇ φ)−
1
− G5,X (φ, X)[(φ)3 − 3(φ)(∇µ ∇ν φ)(∇µ ∇ν φ) + 2(∇µ ∇ρ φ)(∇ρ ∇σ φ)(∇σ ∇µ φ)]+ (3.68)
6
0 0 0
+ F5 (φ, X)µνρσ µ ν ρ σ ∇µ φ∇µ0 φ∇ν ∇ν 0 φ∇ρ ∇ρ0 φ∇σ ∇σ 0 φ
here the notation is the same we use in the previous section and µνρσ is the totally antisym-
metric four-dimensional Levi-Civita tensor. This Lagrangian corresponds to the Horndeski
Lagrangian (3.44)-(3.68) when the free functions F4 (φ, X) and F5 (φ, X) are setted to zero; in-
deed, from the table A.2, one can see that in this case αH = 0. In [48]- [49] it is also shown that
one can use disformal transformations to relate the subclasses of these theories that one ob-
tains by setting L4 = 0 or L5 = 0 (not at the same time) to theories with manifest second-order
equations of motions, i.e. the Horndeski theories.
Here again one has to add the matter fields action under the same assumptions made previ-
ously. Then, the complete action is the following
Z 5
√
X
S= d 4 x −g LBH
i + L (g ,
m µν iψ ) (3.69)
i=2
The standard EFT action (3.2) for beyond Horndeski theories corresponds to [44]
Z m2 M 4 (t)
√
S= d 4 x −g 0 Ω(t)R + Λ(t) − c(t)δg 00 + 2 (δg 00 )2
2 2
(3.70)
3
M̄1 (t) 00 M̂ 2 (t) 00 (3) M̄22
2 i j
− δg δK + δg δR − δK − δKj δKi + Sm [gµν , ψi ]
2 2 2
Disformal transformations
where Ω2 and Γ are respectively the conformal and disformal factors; the WEP is preserved un-
der disformal transformations if one assumes that Ω and Γ are the same for all matter species.
In particular, the theories whose action is invariant under this kind of transformation can be
subdivided in three classes:
a) Horndeski theories where the conformal and the disformal factors are independent by X:
Ω(φ), Γ (φ);
b) Beyond Horndeski theories where the disformal factor depends also by X: Ω(φ), Γ (φ, X);
33
3. The Effective Field Theory approach: unifying single-field models of
Dark Energy and Modified Gravity
c) Degenerate Higher Order Scalar-Tensor (DHOST) which are the most general theories in-
variant under disformal transformations: Ω(φ, X), Γ (φ, X).
When the dependence of the disformal factor by X is allowed, then the theory contains higher
order derivatives; this is the case of b) and c).
If one chooses the scalar field as the time coordinate, ∂µ φ = δµ0 , in the ADM formalism the
components of the metric tensor transform as follows
In [48], the authors showed that using this class of transformations, in particular the class
a), one can map the subclasses of beyond Horndeski theories involving respectively L4 ( with
L5 = 0 ) and L5 (with L4 = 0) to a theory that has second-order equations of motions, i.e a
theory that belongs to Horndeski models. In particular, in the former case, starting from a
theory in which the coefficients Ā4 (φX̄) and B̄4 (φX̄) satisfy the first condition of (3.35), one
obtains a theory in which the the new coefficients A4 (φX), B4 (φX) and Γ are related by
A4 + B4 − 2XB4,X
Γ4,X = (3.74)
X 2 A4
3A5 + XB5,X
Γ5,X = (3.75)
3X 2 A5
so that, a Γ which is independent by X, gives a theory in which the new coefficients satisfy
(3.35) (with A4 = B4 = 0).
In the same paper is pointed out that one cannot map an arbitrary theory with both L4 , 0 and
L5 , 0 to an Horndeski theory because it is not possible to satisfy (3.74) and (3.75) at the same
time.
Coupling to matter
In the previous section we said that the Lagrangian of beyond Horndeski models is preserved
under disformal transformations whose disformal coefficient depends by the scalar field and
its gradient; this last dependence leads to the so called Kinetic Matter Mixing (KMM). In [52]
the authors define the KMM in a frame-independent way. They assume the existence of a frame
in which the matter is minimally-coupled to gravity, i.e the Jordan frame, so that the matter
action is, as usual, Z
Sm = d 4 x −ḡLm (ḡµν , ψi )
p
(3.76)
where ḡµν is the Jordan-frame metric which is disformally related to the metric gµν through
(3.71). In the frame of gµν , the KMM manifest itself with the appearing of a kinetic coupling
between the matter fields and the new scalar degree of freedom which is parametrized by
introducing the additional function of time
X 2 ∂Γm
αX,m = (3.77)
Ωm ∂X
34
3.2. EFT of Horndeski gravity and beyond Horndeski
which is valued on the background. In the Jordan frame αX,m = 0 and the information about
the kinetic mixing are transferred to the beyond Horndeski function αH , which now encodes
the KMM. Thus, since it is a physical effect, the authors in order to characterize the kinetic
matter mixing in a way which is independent from the reference frame, define a new parameter
λ2 ∝ (αH − αX,m )2 . It measures the degree of this mixing between matter and the scalar filed
and, as it is shown in the reference, it does not depend on the frame. Furthermore, they show
that if λ , 0, the KMM is present and the dispersion relation of the scalar modes is modified as
follow
(ω2 − cS2 k 2 )(ω2 − cm
2
k) = λ2 cS2 ω2 k 2 (3.78)
thus, the propagation modes are mixed states of matter and scalar fields; this leads to several
observational effects. In particular, in [52], the authors analyse the observational signatures of
KMM finding that it affects both the matter power spectrum and the CMB.
As like as Horndeski gravity, assuming that the oscillating mode of the scalar filed fluctuation
can be neglecting, we can rewrite the expressions (3.22)-(3.23)-(3.24) for beyond Horndeski
models (2M̂ 2 , M̄22 = −M̄32 and m22 = 0)
#−1
m20 ∆f1 a2 2 f3BH a2 2 (1 + αH )2
" #"
µ(a, k) = 2 1 + H + 2 MC + MC (3.79)
M∗ f1 k 2M∗2 f1H k 2 1 + αT
" BH −1
a2 a2
# " #
f 1 + αH ∆f
η(a, k) = 5 H + 2 MC2 1 + H1 + 2 MC2 (3.80)
f1 k 1 + αT f1 k
#−1
m20 ∆f1 f5BH a2 2 f3BH a2 2 (1 + αH )2
" !# "
1 + αH
Σ(a, k) = 1 + H + H + 2 MC 1 + + MC (3.81)
2M∗2 f1 f1 k 1 + αT 2M∗2 f1H k 2 1 + αT
here the quantities fi are defined in table A.3; MC2 ≡ Cπ /f1H where f1H is defined in (A.4). We
write the functions fi , for beyond Horndeski models as a sum of an Horndeski term (fiH ) plus
an additional term (∆fi ) arising from the contribution of the αH : fiBH ≡ fiH + ∆fi . The explicit
expressions of these functions can be found in A.0.2.
As for the Horndeski case, we consider these functions in the sub-Compton and super-Compton
regimes:
m20 1 + αT
µ0 = (3.82)
M∗2 (1 + αH )2
1 + αH
η0 = (3.83)
1 + αT
m20 (2 + αH + αT )
Σ0 = (3.84)
2M∗2 (1 + αH )2
From the above expressions, one can see that in this case one can have gravitational slip even
if the tensor speed excess is vanishing because of the presence of the property function αH ;
moreover, also in Σ0 and µ0 there is more freedom than in the Horndeski’s case and thus, one
can have potentially (Σ − 1)(µ − 1) < 0.
35
3. The Effective Field Theory approach: unifying single-field models of
Dark Energy and Modified Gravity
Sub-Compton: k >> aMC
H 2 2
m20 2M∗2 f1BH m20 α(cS ) (1 + αT + βξ ) + 2β1H
µ∞ = 2 = 2 (3.85)
M∗ f3BH M∗ α(cSH )2 + β3H
H 2
f BH α(cS ) 1 + βB βξ /2 + 2β5H
η∞ = 5BH = (3.86)
f1 α(cSH )2 (1 + αT + βξ2 ) + 2β1H
H 2 2
m20 α(cS ) 2 + αT + βξ + βB βξ /2 + 2(β1H + β5H )
Σ∞ = (3.87)
2M∗2 α(cSH )2 + β3H
where the quantities β1H , β3H and β5H are defined in A.0.2 ((A.34)-(A.37)).
The sub-Compton limit is much more complicated than the previous one because we have
several additional terms, so it is not simple to analyse without any assumptions that simplify
the expressions. The novelty here is that, thanks to the new terms β1H , β3H and β5H , which
can be negatives, in beyond Horndeski models the fifth force can be repulsive leading to a
weakening of gravity and a suppression of clustering, contrary to what happens in Horndeski
gravity where it is always attractive; this effect is due to the presence of the kinetic matter
mixing (KMM).
Let us consider the perturbed line element for tensor modes in a spatially-flat FRW universe:
By varying this action with respect to hij and using (3.37)-(3.42) one finds the evolution equa-
tion for tensor modes which, in the Fourier space, is
!(T T )
k2 2 δij k
ḧij + (3 + αM )H ḣij + (1 + αT ) 2 hij = 2 Tij − T (3.90)
a M∗ 3 k
here the term in brackets on the r.h.s is the transverse-traceless projection of the anisotropic
matter stress tensor that vanishes for perfect fluids.
From this equation one immediately sees that if both αM and αT are zero, then the propagation
of GWs is not affected by modifications of gravity. On the contrary, a non-zero value of αM ,
introduces a modification in the friction term that affects the amplitude of GWs and moreover
the same parameter, along with αT , modifies the frequency of the observed GW. Thus, if at
least one of these functions are different from zero, we have a non-standard propagation of
tensor modes and since the same property functions (together with αH ) are responsible for
η , 1, then, it becomes clear that there is a deep connection between these two effects.
In addition, as we mentioned in section 2.1.2, the tensor speed cT determines the horizon cross-
ing of the tensor modes and thus, a value of αT different from zero, shifts the position of the
first peak of the B-mode power spectrum. Furthermore, the theoretical BB-spectrum predicts
another peak at ` ∼ 5 and one finds that also the position of this second peak depends on the
tensor speed and thus it would be shifted if cT , 1.
In [33]- [32] the authors pointed out that the existence of the gravitational slip is controlled by
36
3.2. EFT of Horndeski gravity and beyond Horndeski
the property functions αM , αT and αH and that the first two are the same property functions
responsable for the non-standard propagations of GWs. Thus, in many modified gravity the-
ories, the presence of gravitational slip is related with the modifications of the tensor modes
propagation. In Horndeski models, in principle, one can construct a theory in which there is no
gravitational slip even if one of αM , αT are different from zero. However, this requires a very
tuned choice of the α-functions, i.e. αT = 0 and αM = −αB /2, thus in general, in Horndeski
gravity, the modification of the propagation of the GWs waves is associated to the presence of
the gravitational slip. In beyond Horndeski models the situation is different and the presence
of gravitational slip does not imply the non-standard propagation of the tensor modes, indeed
one can have a theory with αM = 0 = αT and thus a standard propagation of GWs but a non
vanishing gravitational slip arising from the contribution of αH , 0.
37
Chapter 4
On the 17th August of 2017 the LIGO-VIRGO collaborations detected the first binary neutron-
star merger, the event GW170817; exactly (1.74 ± 0.05)s later, the Fermi and the Interna-
tional Gamma-Ray Astrophysics Laboratory detected its electromagnetic counterpart, the short
gamma-ray burst GRB170817A [25]- [24].
This event puts a very strong constraint on the propagation speed of the tensor modes (GWs):
Before this event there were bounds on cT coming from the non-observation of gravitational
Cherenkov radiations and the variation of the orbital period of binaty pulsar. In particular,
the former imposes a lower bound on cT : for the cosmic rays which have a galactic origin this
bound gives 1 − cT < 2 · 10−15 , if the cosmic rays have an extragalactic origin the bound is of
order 1 − cT < 2 · 10−19 [28].
These constraints in terms of the tensor speed excess translate in: |αT | ≤ 10−15 .
As a result, many alternative theories of gravity have been ruled out. In the next sections
we consider the consequences for two classes of theories: Horndeski and beyond Horndeski
models.
However, one must take into account the fact that, in principle, one can have a cT , 1 at hight
redshift and, as pointed out in [24], its value may depend on the frequency at which it is
measured, cT = cT (k). In particular the event GW170817 has been detected at energy scale
2 2 1/3
close to the cut-off scale (Λ3 = (Mpl H0 ) ) at which the EFT of dark energy breaks down.
They also underline that an EFT may have a tensor speed different from the speed of light, but
one must have cT = 1 at high energy, specifically they show that it can happen for Horndeski
theories.
this relation must hold for any value of φ, its derivative and H, so if we exclude a cancellation
among G4,X , G5,φ , G5,X that requires a tuning among the functions appearing in the above
38
4.1. Horndeski gravity after GW170817
Then, thanks to the Bianchi identity the term Gµν (∇µ ∇ν φ) in (3.68) vanishes and the La-
grangian density for Horndeski models is restricted to be of the following form
This Lagrangian represents the subclass of the Horndeski models survived the event GW170817.
In particular, it includes quintessence and k-essence models, BD theory and metric f(R) grav-
ity. In these last models the evolution of the scalar field on cosmological scale is negligible and
screening mechanisms are required in order to satisfy solar system constraints , thus, these
theories, do not self-accelerate cosmological expansion. An other class of models which is con-
sistent with the bound (4.1) includes kinetic braidings and their extensions; these theories, in
general, exhibit self-acceleration.
In this subsection we consider the analytic expressions for the phenomenological functions
(3.79)-(3.81) setting cT = 1 and considering each subclass of surviving Horndeski theories. The
general expressions, after imposing the bound (4.1), become
−1
m20 m20
#
a2 2 1 a2 2
" " #
∆1 (a, k)
µ(a, k) = 2 1 + 2 MC + M = 2 1+ (4.5)
M∗ k 1 + βξ2 k 2 C M∗ 1 − ∆1 (a, k)
" #−1
1 + βB βξ /2 a2 2 a2
2
η(a, k) = + 2 MC 1 + 2 MC = 1 + ∆2 (a, k) (4.6)
1 + βξ2 k k
2 −1
m20 2 + βξ + βB βξ /2 a2 m20
1 a 2
2
2
Σ(a, k) = + 2M + M = [1 + ∆3 (a, k)] (4.7)
C
2M∗2 1 + βξ2 k2 1 + βξ2 k 2 C M∗2
with the parameters appearing in these expressions defined in (A.8)-(A.12) but now βξ2 =
1
[α + 2αM ]2 .
2αcs2 B
In particular, assuming that the stability conditions (3.39)-(3.42) hold, from the expression of
∆2 , ∆1 and ∆3 1 one can see that
• ∆2 , 0 if
2
βB βξ /2 − βξ2 = − αM (2αM + αB ) , 0 =⇒ η,1 (4.8)
αcS2
This means that to have gravitational slip one must have αM , 0 and it must be αM ,
−αB /2.
In terms of Horndeski free functions this implies one needs dM∗2 /dt = dG4 /dt , 0 =⇒
G4 = G4 (φ), i.e a conformal coupling to gravity.
• ∆1 , 0 if
1 m20
βξ2 = (α + 2αM )2 , 0
2 B
=⇒ µ> 2 (4.9)
2αcs M∗
39
4. The day after GW170817: implications for scalar-tensor theories
• ∆3 , 0 if
1 m20
βB βξ /2 + βξ2 = αM (2αM + αB )(αM + αB ) , 0 =⇒ Σ, (4.10)
αcS2 M∗2
Thus one finds Σ = m20 /M∗2 for αM = 0 = αB , αB = −2αM and αB = −αM , then, according to
the sign of the product (2αM + αB )(αM + αB ), one has that Σ ≷ m20 /M∗2 .
The two limiting cases, setting αT = 0, yield
m2 m20 h i
µ0 = 02 (4.11) µ∞ = 2 1 + βξ2 (4.14)
M∗ M∗
h ih i−1
η0 = 1 (4.12) η∞ = 1 + βB βξ /2 1 + βξ2 (4.15)
m20 m20 h i
Σ0 = (4.13) Σ∞ = 2 + βξ2 + βB βξ /2 (4.16)
M∗2 2M∗2
This subclass corresponds to K = K(φ, X), G3 = 0, G4 = m20 /2, thus the action is
" 2 #
m0
Z
4 √
S= d x −g R + K(φ, X) + Sm [gµν , ψi ] (4.17)
2
In particular, the quintessence models, are characterized by a scalar field with canonical kinetic
term and a slowly varying potential (K(φ, X) = X − V (φ)); while in K-essence models the scalar
field has non-canonical kinetic term. K-essence models include: ghost condensate models,
tachyon field and Dirac-Born-Infeld (DBI) theories. In this class of theories the DE equation of
state is time-dependent and the phenomenological functions are trivial
µ = 1, η = 1, Σ = 1, (4.18)
This result holds at any scale and any time, therefore the equations (1.13)-(1.16) are not mod-
ified and neither gravitational slip nor modification of Gmatter Glight are expected for these
theories. Thus, any deviation of these functions from the unity, as well as any detection of a
scale-dependence or a time-dependence of the same, would rule out these models.
φ̇XG3,X
Kinetic braiding models and its extensions: αB + αM = H G4
In these models the Horndeski free functions are K = K(φ, X), G3 = G3 (φ, X), G4 = G4 (φ). This
class includes two special cases: 1) αB = −2αM , and 2) αM = 0, αB , 0. In particular in the
2 These are more stringent constraints with respect to those we found in section (3.2.3) where we did not impose
cT = 1.
40
4.1. Horndeski gravity after GW170817
former βξ = 0, in the latest βξ = βB /4; the result, in both cases, is that Σ = µ and η = 1.
Kinetic braiding theories correspond to the particular choice
This class is also a subclass of the covariant Galileons, which in [11] is dubbed H3 and which
in particular contains the cubic Galileons3 . The expressions (4.5)-(4.7) in this case are
−1
a2 2 2αcS2
#
a2 2
"
µ(a, k) = Σ(a, k) = 1 + 2 MC + M , η=1 (4.19)
k 2αcS2 + αB2 k 2 C
1
where we have used βξ2 = α2
2αcS2 B
= βB2 /4 and M∗2 = m20 .
We have that for these models the two scalar potentials are equally enhanced by the cubic
derivative coupling G3 so that, also in this case, there is no gravitational slip.
1
µ∞ = Σ∞ = 1 + αB2 (4.22)
µ 0 = Σ0 = 1 (4.20) 2αcS2
η0 = 1 (4.21) η∞ = 1 (4.23)
Since they are self-accelerating models the scalar field mass is small and λC ∼ 1/H, conse-
quently the observational window, in this case, falls in the small scale limit if the QSA holds;
this means that if a scale dependence in µ and Σ were to prove to be true it would rule out this
class of models. Moreover also an observation of Σ , µ or η , 1 would disqualify these theories.
√ m2 h
Z i
SGBD = d 4 x −g 0 F(φ)R − Z(φ)g µν ∇µ φ∇ν φ − 2U (φ) + Sm [gµν , ψi ] (4.24)
2
the variation of the action with respect to the metric gives the modified Einstein’s equations
h i 1 (m)
FGµν − Z(φ)∇µ φ∇ν φ + gµν Z∇ρ φ∇ρ φ + 2U − ∇µ ∇ν F + gµν F = 2 Tµν (4.25)
m0
1 1 k(m)
(2ZF + 3F 02 )φ + (2ZF + 3F 02 )∇µ φ∇µ φ = 2 F 0 Tk − 4F 0 U + 2U 0 F (4.26)
2 m0
3 Appendix C.2.
4 an equivalent form of this action can be found in appendix D.
5 in particular the choice Z(φ) = 1 (1 − 6Q2 )F(φ), corresponds to Brans-Dicke theory with scalar potential if one
2
redefine the scalar field in this way F(φ) = φ0 = e−2Qφ with 3 + 2ωBD = 1/(2Q2 ); one recovers GR for Q −→ 0
(ωBD −→ ∞)
41
4. The day after GW170817: implications for scalar-tensor theories
where the prime denotes the derivative with respect to the scalar field.
From the above equation we obtain the modified Friedmann equation
1 1
3FH 2 = ρ̄ + 3ḞH + Z φ̇2 + U
2 m
(4.27)
m0 2
Finally, by perturbing at linear order the field equations and by subtracting the background
equations, one gets the Poisson equation and the anisotropy constraint, that in QSA are
1
Fk 2 Ψ = −4πGa2 ∆ − k 2 δF (4.28)
2
Fk 2 (Φ − Ψ ) = k 2 δF (4.29)
2
8πGa2
k 2 (Φ + Ψ ) = − ρ∆ (4.30)
F
Therefore, we can write the phenomenological parameter Σ(a) = F(φ)−1 . This expression shows
that Σ is inversely proportional to the background value of the conformal factor and it is k-
independent. Then, one can use the relation (3.60) with the normalization µ(a = 1, k = 0) = 1
so that Σ(a = 1) = F(φ(a = 1))−1 = 1. Furthermore, the variation of the conformal factor
is constrained by the presence of screening mechanisms in such a way that one must have
|F(z = 1) − F(z = 0)| /F(z = 0) . 10−6 [76]6 .
In terms of the α-parametrization the phenomenological functions are
−1
m20 a2 2 2αcS2 m20
#
a2 2
"
µ(a, k) = 2 1 + 2 MC + M , Σ(a) =
M∗ k 2αcS2 + αB2 k 2 C M∗2
#−1 (4.31)
2αcS2 − αB2 a2 2
"
a2 2
η(a, k) = + M 1 + 2 MC
2αcS2 + αB2 k 2 C k
1
where we have used βξ2 = 2αcS2
(−αB )2 = βB2 /4.
m20 m20
1 2
µ∞ = 2 1 + αB , Σ∞ = (4.34)
m20 M∗ 2αcS2 M∗2
µ 0 = Σ0 = 2 (4.32)
M∗ 2αcS2 − αB2
η∞ = (4.35)
η0 = 1 (4.33) 2αcS2 + αB2
This subclass of the survived Horndeski theories predicts gravitational slip. Furthermore, from
the above expressions, one can see that the parameter Σ depends only by time and it is related
with the effective value of the Planck mass. Additionally, since α and cS2 in a consistent theory
of gravity are positive quantities, one has µ ≥ m20 /M∗2 , Σ = m20 /M∗2 and η ≤ 1. One can conclude
that any measurement of these functions which are not consistent with these last conditions
would rule out the GBD theories.
6 For non-universally coupled models this constraints are less stringent and one can obtain value of Σ slightly
different from one.
42
4.2. Beyond Horndeski models after GW170817
We said that the cut-off of the scalar-tensor EFT of dark energy, where the interactions between
the scalar field fluctuations and the gravitons become important, lies near the scale at which
the GW170817 has been detected. In [55] the authors show that, at this scale, the decay of
gravitons becomes very efficient for theories with αH , 0, consequently, the observation of
GWs would imply the ruling out of this kind of theories. However, they also show that one
can avoid the decay of GWs by imposing that the propagation speed of scalar modes is equal
to that of the light, i.e. cS2 = 1, it can be used like a constraint that allows the elimination of one
of the α property functions.
If we impose the bound (4.1) and thus cT = 1, i.e. αT = 0 we have
this relation must hold for any value of φ, its derivative and H, thus one has to require that
in [26] they show that the Lagrangian that satisfies these constraints is of the following form
where B4 (φ, X) = (G4 − XG5,φ )/2 and B4,X = −2XF4 . In [26] the authors point out that the
functions G5,X and F5 must vanish separately in order to avoid the pathology M∗2 = 0.
#−1
m20 f1BH a2 2 f3BH a2 2 m20
" #"
2
µ(a, k) = 2 + M + M (1 + αH ) = [1 + ∆BH
1 (a, k)] (4.39)
M∗ f1H k 2 C 2M∗2 f1H k 2 C M∗2
" BH # " BH #−1
f5 a2 2 f1 a2 2
η(a, k) = + 2 MC (1 + αH ) + 2 MC = 1 + ∆BH
2 (a, k) (4.40)
f1H k f1H k
#−1
m20 f1BH f5BH a2 2 f3BH a2 2
" #"
2
Σ(a, k) = + H + 2 MC (2 + αH ) + MC (1 + αH ) =
2M∗2 f1H f1 k 2M∗2 f1H k 2
(4.41)
m2 h i
= 02 1 + ∆BH 3
M∗
The quantities ∆BHi (a, k) (i = 1, 2) are defined in A.0.2. From these expressions one can see that,
in general, the new property function αH gives rise to gravitational slip and modifies the grav-
itational interaction felt by both relativistic and non-relativistic particles. Indeed, unlike the
Horndeski case, in which there are same combinations of the property functions α for which
the additional contributions ∆H i (A.12)-(A.14) to the phenomenological functions cancel out
for each scale and each time, here the expressions of ∆BH i (A.40)-(A.41) involve combinations
of the α and background quantities, thus it is extremely unlikely to have a cancellation of these
terms. In particular
43
4. The day after GW170817: implications for scalar-tensor theories
• ∆BH
2 , 0 (A.41)
It is given by αH (t) plus a contribution that depends on time and scale ; this means that
in general, if αH , 0, one has gravitational slip, also when αM = 0 or αM = −αB /2, i.e.
βξ βB /2 − βξ2 = 0, that correspond to the cases in which, in the Horndeski gravity, the
gravitational slip disappears.
• ∆BH
1 , 0 (A.40)
2
It is given by −(2αH + αH )/(1 + αH )2 plus a contribution that depends on time and scale;
thus, also in this case, in general ∆BH2 , 0 and the gravitational interaction with matter is
modified with respect to GR.
• ∆BH
3 , 0 (A.42)
The phenomenological function Σ can be obtained by the first two through Σ = µ(1+η)/2,
thus one finds ∆3 = ∆1 + ∆2 /2 + ∆1 ∆2 /2 and, in general, ∆BH
3 , 0, because cancellations
among these terms are unlikely. This means that also effective gravitational coupling
Glight = ΣG is modified in beyond Horndeski models.
The expressions of the three functions in the super-Compton and sub-Compton regimes are
the following:
m20 1
µ0 = (4.42)
M∗ (1 + αH )2
2
η0 = 1 + αH (4.43)
m20 (2 + αH )
Σ0 = (4.44)
2M∗2 (1 + αH )2
In this regime the phenomenological functions depends only by αH and the effective squared
Planck mass. We see that the situation is different with respect to the Horndeski theories; here
there is always gravitational slip, i. e. η0 , 1, because of the presence of the αH functions.
This means that if a scale dependence would be detected, then a measurement of η0 = 1 would
disqualify this class of theories. Furthermore an αH , 0 leads to a different gravitational inter-
action with matter with respect to GR, in particular there will be an enhancement µ0 > m20 /M∗2
(or a decreasing µ0 < m20 /M∗2 ) of this latter if αH < 0 (αH > 0) as one can see from (4.42). In this
limit, also the effective gravitational interaction felt by relativistic particles, is affected by the
presence of a non-zero αH .
The expressions of the phenomenological functions in the sub-Compton limit involves not
only αH but also the others property functions and the background functions, therefore, the
discussion of this case, is not so simple as the super-Compton limit and, as just commented for
44
4.2. Beyond Horndeski models after GW170817
the general case, one has modifications with respect to GR; indeed a cancellations among the
additional contributions would require a tuning among the αs, H and Ḣ.
45
Chapter 5
In this chapter we summarize what we have found in the previous one about Horndeski gravity
and, in particular, we analyse the conjecture (µ−1)(Σ−1) ≥ 0. Finally, we briefly report some of
the results obtained in the literature using the Boltzmann codes EFTCMB and EFTCosmoMC;
in particular, we consider the following papers: Espejo (2018) [12], Peirone (2017) [13] and the
Planck Collaboration 2015 [37].
5.1 Summary
Horndeski gravity, unless there is an highly unlikely fine-tuning among M∗2 , αT , βB and βξ2 ,
predicts
(µ − 1)(Σ − 1) ≥ 0 (5.1)
in principle, a different sign in the two factors, is not impossible to obtain in Horndeski gravity,
however, for anything which has been said, such a result would disfavour these models.
Moreover if one
• measures µ ≤ m20 /M∗2 or Σ , m20 /M∗2 =⇒ GBD theories would be ruled out;
• measures η , 1 =⇒ theories with αM = 0 or αM = −αB /2 would be ruled out;
• detects a k-dependence =⇒ the Covariant Galileons would be ruled out and furthermore
i) µ∞ < µ0 would rule out all the Horndeski theories;
ii) µ∞ ≥ µ0 Horndeski gravity would be not disqualified; in particular µ∞ > µ0 would
imply a non vanishing βξ2 , i.e. it is evidence of a fifth force.
After GW170817
The bound (4.1) for z ≤ 0.009 allows us to set αT = 0, this in turn implies G4 = G4 (φ) and
G5 = const. The survived Horndeski theories can be classified in terms of the α parameters as
follows:
A) Quintessence and K-essence: αB = 0 = αM ;
46
5.2. Analysing the conjecture (µ − 1)(Σ − 1) ≥ 0
φ̇XG3,X
B) Kinetic braiding and its extensions: αB + αM = HG4 . In particular the case αM = 0
corresponds to kinetic braiding;
Following [13], we consider the limit k << aMC and k >> aMC and analyse the conditions under
which this constraint can be violated. It is clear, from what we have seen in section 4.1.1, that
the theories belonging to class A) and the subclass of B) for which αM = 0 or αM = −αB /2 cannot
violate this conjecture in any case since they predict Σ = µ.
In this regime the phenomenological functions are given by (3.54)-(3.56), thus, in order to
have gravitational slip (η0 , 1), one has to require αT , 0. We know that at low redshifts this
is impossible to achieve because of the bound (4.1); however, there are no constraints on cT at
high redshifts. A zero tensor speed excess, as we saw, implies also µ0 = Σ0 and therefore, in
this case, the constraint cannot be violated by Horndeski theories. If one allows for αT , 0 in
principle there is any chance of breaking the conjecture:
1 2 1 αT
a) µ0 = Ω (1 + αT ) >1 Σ0 = Ω (1 + αT )(1 + 2 )<1
αT
⇐⇒ (1 + αT )(1 + ) < Ω < (1 + αT )2 =⇒ αT > 0 =⇒ Ω>1 (5.2)
2
1 2 1 αT
b) µ0 = Ω (1 + αT ) <1 Σ0 = Ω (1 + αT )(1 + 2 )>1
αT
⇐⇒ (1 + αT )2 < Ω < (1 + αT )(1 + ) =⇒ αT < 0 =⇒ Ω<1 (5.3)
2
where to obtain these inequalities we have used (3.54)-(3.56) and M∗2 (1 + αT ) = m20 Ω1 .
From the above expressions it is easy to see that a violation of (5.1), in this regime, would imply
αT , 0 that, in turn, would constrain the conformal coupling to be Ω , 1. But, as we mentioned
before, these quantities are subjected to other strong constraints2 and so the conclusion is that
it would be very hard to obtain and also to observe a violation of the conjecture (5.1).
1 Appendix A, table A.2
2 In particular we have the following bounds: 1) |Ω(z = 0) − 1| < 0.1 coming from the non-detection of the fifth
force on Earth; 2)|Ω(z = 1100) − 1| < 0.1 coming from the necessity to be consistent with BBN and CMB [13]- [76].
47
5. Constraining Horndeski gravity with µ, η and Σ
In this limit the phenomenological functions are given by (3.57)-(3.59). Unlike the previous
case, one can have gravitational slip even if αT = 0, thanks to the presence of the fifth force
term. Here again, we have two possibilities to violate the condition above:
β 2 +β β /2
a) µ∞ = Ω1
(1 + αT )(1 + αT + βξ2 ) > 1 1
Σ∞ = Ω (1 + αT ) 1 + α2T + ξ 2ξ B < 1
2
αT βξ + βξ βB /2 Ω
⇐⇒ 1+ + < < 1 + αT + βξ2 (5.4)
2 2 1 + αT
βξ2 +βξ βB /2
1 1 αT
b) µ∞ = Ω (1 + αT )(1 + αT + βξ2 ) < 1 Σ∞ = Ω (1 + αT ) 1+ 2 + 2 >1
2
Ω α βξ + βξ βB /2
⇐⇒ 1 + αT + βξ2 < < 1+ T + (5.5)
1 + αT 2 2
Thus, also in this case, the breaking of the conjecture (5.1) would require very specific condi-
tions. We can then consider the case in which αT = 0, so that the above disequalities simplify.
In order to satisfy (5.4) and (5.5) one must require respectively 3
βξ βB 2
αB
αM > 0, αM > −αB /2
a) < βξ =⇒ αM αM + > 0 =⇒ (5.6)
2 2 αM < 0, αM < −αB /2
βξ βB
α
αM > 0,
αM < −αB /2
> βξ2 αM + B < 0
b) =⇒ αM =⇒ (5.7)
2 2
αM < 0,
αM > −αB /2
5.3.1 Espejo(2018)-Peirone(2017)
In [12] and in [13] the authors perform numerical Monte Carlo simulations with the publicly
available Boltzmann codes EFTCMB and EFTCosmoMC5 in order to verify the consistency of
this conjecture (5.1), to check if there is any correlation between the values of µ and Σ and
also to test the validity of the QSA. They start to solve for the background solution requiring
that it is consistent with the concordance model ΛCDM6 and that the conformal coupling Ω(t)
satisfies the constraints coming from CMB and BBN. Then, they check the stability conditions
3 are the conditions (35) and (36) of [13]
4 these expressions contains the pre factor 1/(αc2 ) that is positive if the stability conditions (3.39)-(3.42) hold.
S
5 They are patches of the Einstein-Boltzmann solver CAMB.
6 They impose theoretical prior in order to constrain the parameters’ space in such a way to be consistent with
observations; for example they impose a Gaussian prior on H(t).
48
5.3. Horndeski gravity with EFTCAMB and EFTCosmoMC
of the model and, if these conditions hold, they evolve the liner perturbations finding the exact
solution of Φ, Ψ and ∆; then, for a given model, they reconstruct the exact phenomenological
functions. They make simulations by parametrizing the EFT functions using a Padé expansion
and by considering three class of models: the full class of Horndeski gravity (Hor) with the
bound (4.1), the class of models for which cT = 1 (HS ) and GBD theories, which have a standard
kinetic term.
In particular, in the former review, they sample the (a, k)-plane at the following values: a ∈
{0.25, 0.575, 0.9}, k ∈ {0.001, 0.05, 0.1}. The ensuing results are:
• The QSA breaks down for modes k . 0.001 h/Mpc even though they are inside the scalar
field’s sound horizon;
• The conjecture (5.1) is not violate for GBD models; it holds also for the other two class
of models, except for about 10% of these theories that violate the conjecture at k =
0.001h/Mpc giving µ < 1 and Σ > 1. Furthermore, they show that the stability condi-
tions (3.39)-(3.42) work against the breaking of this conjecture disfavouring solutions
with µ < 1 and Σ > 1.
• There are correlations among the background parameters and the phenomenological
functions µ and Σ and these correlations should be taken into account when searching
for signatures of MG.
In [13] the authors focusing on the reconstruction of wDE , Σ and µ via PCA technique7 . They
derive many statistical properties of these functions among which the mean values and the
distribution functions as well as their joint covariances and the functional forms of their corre-
lations functions, independently for the three class of models. The results can be summarized
as follows:
• The mean values of Σ and µ do not have significant variations with redshift, in particular
for HS and Hor these values remain very close to 1 (∼ σ ); in GBD the means values are
close to one within two standard deviations but tend to values below one because of the
pre factor m20 /M∗2 . The mean value of wDE is close to one at lower redshifts and tends to
zero at higher redshifts8 .
• All classes of models show strong correlation between Σ and µ and it must be taken into
account when these functions are constrained by data. The correlation between wDE and
Σ/µ decreases as the number of parameters used to characterize a given class of models
increases9 . Finally, the cross correlations among wDE , Σ and µ vary significantly from
one class of models to another: it is very strong for GBD, it is barely visible for HS and
negligible for Hor.
• They use the generalized CPZ parametrization and fit it to their numerical results to
obtain the functional forms of the correlation functions. They found the same result for
the three class of models: the correlation functions for µ and Σ scale with ”a” while for
wDE scale with ”lna”.
7 The Principal Component Analysis is a method that allows the reconstruction of these functions in a non-
parametrically way by binning them in redshift and in scale. The advantage of this technique is that it compresses
all information about modes coming from observations and uses them to derive constraints on the parameters of a
specific model.
8 This difference is due to the fact that at lower redshifts the mean values are strongly influenced by SN data while
those at higher redshifts are determined by the dominant density component because of the tracking behaviour of
the effective DE fluid which is induced by the non-minimal coupling of the scalar field.
9 In particular they use two parameters to specify GBD models (Σ, Λ) and add two parameters for H and three
S
for Hor
49
5. Constraining Horndeski gravity with µ, η and Σ
Linear model
αM0 (95%CL) < 0.052 < 0.072 < 0.057 < 0.074 < 0.050 < 0.043
Exponential model
αM0 (95%CL) < 0.063 < 0.092 < 0.066 < 0.097 < 0.054 < 0.062
β 0.87+0.57
−0.27 0.91+0.54
−0.26 0.88+0.56
−0.28 0.92+0.53
−0.25 0.90+0.55
−0.26 0.92+0.53
−0.24
Table 5.1: In this table are shown the marginalized mean value at 68% confidence intervals for the
EFT parameters obtained via MCMC (Monte Carlo Markov chain) methods using the Boltzmann code
EFTCAMB.
In [37] they use the data coming from the ”Planck Collaboration I 2015” in combination with
other data sets to analyse the implications for DE and MG models. Also in this analysis, it is
used EFTCAMB to fit observational data sets via MC methods in order to constrain the param-
eters of the EFT of dark energy. In particular, they focusing on Horndeski models (αH = 0) and
thus impose the conditions m22 = 0 and 2M̂ 2 = M̄22 = −M̄32 by setting these parameters to zero
(M̄22 = −M̄32 = 0), consequently also αT is setted to zero. Furthermore, they impose the condi-
tion αM = −αB (GBD theories) and fix the background to be that of ΛCDM. With these choices
the only free parameters of the theory is αM and it is related with the conformal coupling of
the EFT by10
a dΩ
αM = (5.8)
Ω da
Constraints on αM
The Planck Collaboration constrains the αM parameter making the following ansatz: αM =
αM0 aβ , where αM0 is the actual value of the parameter and β ∈ (0, 3] is a parameter that quan-
tifies how fast GR is recovered in the past. The EFTCAMB allows us to choose among different
models for the function Ω(a), Plank Collaboration fits data using the exponential and the linear
models with Ω0 = αM0 /β ∈ [0, 1]
!
αM0 β αM0
Ω(a) = exp a − 1, Ω(a) = a (5.9)
β β
Therefore, for αM0 = 0, one recovers ΛCDM and for small Ω0 the exponential model reduces
to the linear one. Moreover, as the scaling parameter goes to zero (β → 0), then αM = const
and it cannot vanishes in the past. The results are shown in table 5.3.2 and in Fig.5.1. In
particular, the viability conditions required by EFTCAMB lead to a sharp cut-off-off for β ∼
1.5 and strictly reduce the number of viable models; in [37] is pointed out that the filtering
conditions implemented in EFTCAMP may be too restrictive and as a consequence may exclude
more models than the necessary.
10 table A.2.
50
5.3. Horndeski gravity with EFTCAMB and EFTCosmoMC
Figure 5.1: Marginalized posterior distributions for the parameters αM0 and β of the exponential model
(at 68% and 95% C.L., on the left) and for the background parameter Ω of the linear model (on the
right); here Planck refers to Planck TT+lowP. In particular, at perturbation level, in both models, for
Ω0 = 0 the ΛCDM model is recovered. In the figure on the left one can see that the inclusion of WL data
enlarges the contours of the marginalized posterior distribution; this fact is due to a moderately tension
between WL data and the others.
Constraints on η, µ and Σ
In order to characterize deviations from GR, they choose the phenomenological functions η
and µ parametrizing them in the following way
1 + c1 (λH/k)2
µ(a, k) = 1 + f1 (a) (5.10)
1 + (λH/k)2
1 + c2 (λH/k)2
η(a, k) = 1 + f2 (a) (5.11)
1 + (λH/k)2
where the fi are functions of time which quantify the extent of the deviations from GR, while
the parameters ci and λ are constants which give information about the scale dependence.
Then, they implement the model using two different parametrizations of the functions fi (a)
(i = 1, 2)
The two parametrizations are complementary, in particular in the former the time evolution
is based on the effective dark energy density ΩDE (a) and the deviations from GR are not al-
lowed at early time; on the contrary, the last parametrization has two parameters: Ei1 that
characterizes deviations at low redshifts and Ei2 that allows for deviations at high redshifts,
i.e. at early time. The authors make clear that they are not using the QSA, but only a minimal
parametrization that allow for a scale dependence of the phenomenological functions which
is required because of their data cover a wide range of scales. Then, for each parametrization,
they consider the scale-dependent case and the scale independent case by setting for this last
case c1 = 1 = c2 . The results of their analysis are reported in table 5.2 and can be summarized
as follows:
• The scale dependence in any case does not play a significant role, indeed they find that
the chi-quadro distributions are approximately the same in both scale dependent and
independent case; thus they focusing on this last case, which is the simplest to analyse.
51
5. Constraining Horndeski gravity with µ, η and Σ
• The Eij values are obtained by constraining data; from them the actual values of µ and
η, denoted with the index zero11 , are reconstructed using the above parametrization; the
today’s value of Σ is obtained using (1.15). The plots of the marginalized distributions in
the plane (µ0 − 1, η0 − 1) (for both DE-related and time-related parametrizations) and that
in the plane (µ0 − 1, Σ0 − 1) (for the DE-related case) are shown respectively in Fig.5.3 and
Fig.5.2.. In these plots the dashed lines represent the values of these functions predicted
by the ΛCDM scenario (µ = η = Σ = 1) and, in general, the results are compatible with
this scenario, even if some tension is visible. In particular, they found that for the case a)
the deviations, for each data set, is at least of 2σ (Plank TT+lowP); the tension increases
when BAO+RSD are included reaching the maximal value of 3σ when they consider the
full set of data (Plank TT+lowP+BAO+RSD+WL). Including also CMB lensing the tension
decreases with a maximal value of 1.7σ , obtained when all data sets are combined. In
the second parametrization (time-related) the tension is lower going from 1σ for PLanck
TT+lowP to a maximum of 2.1σ for the combination Plank TT+lowP+BAO+RSD+WL.
The higher tension observed in the DE-relates parametrization is probably due, as the
authors comment in the paper, to the fact that the data set Planck TT+lowP , slightly
prefers models which have a lower CMB temperature power spectra at low multipoles
(ISW effect) and higher CMB lensing potential with respect to ΛCDM12 .
Parameter Planck TT+ Planck TT+ Planck TT+ Planck TT+lowP+ Planck TT+lowP Planck TT,TE,EE+
+lowP +lowP+BSH +lowP+WL +BAO/RSD +WL+BAO/RSD +lowP+BSH
E11 0.099+0.34
−0.73 0.06+0.32
−0.69 −0.20+0.19
−0.47 −0.24+0.19
−0.33 −0.30+0.18
−0.30 0.08+0.33
−0.69
Table 5.2: Marginalized mean values (68% C.L.) of the phenomenological parameters obtained by
Planck Collaboration for the DE-related scale-independent case.
Figure 5.2: Contour plot for marginalized posterior distribution for 68% and 95% C.L. in the planes
(µ0 − 1, Σ0 − 1) for the DE-related parametrization with no scale-dependence.
11 In this paper µ , η and Σ refers to today’s values of these functions, while in the previous sections we use the
0 0 0
same notations for the super-Compton limit of the functions.
12 Their best fit power spectrum Planck TT+lowP in the DE-related parametrization is obtained for the following
values of the parameters: E11 = −0.3, E22 = 2.2. These values are close to those of the model with E11 = 0, E22 = 1 in
Fig.5.4.
52
5.3. Horndeski gravity with EFTCAMB and EFTCosmoMC
Figure 5.3: Contour plots for marginalized posterior distributions in the planes (µ0 − 1, η0 − 1) at 68%
and 95% C.L. for the DE-related and time-related parametrizations with no scale-dependence. Here
Plancks stands for PlanckTT+TEB. These plots show a similar behaviour for both parametrization.
Figure 5.4: In these pictures it is shown how deviations from GR affect the power spectra of CMB tem-
perature (left panel) and that of the lensing potential (right panel). From (5.10)-(5.11) one can see that
these deviations, in the parametrization adopted in this paper, are characterized by the parameters Eii ;
in particular for E22 , 0 gravitational slip appears and in any case the power spectra of CMB tempera-
ture is modifies at low multipoles, while the power spectra of the lensing potential increases if E22 > 0
and decreases for E22 < 0.
53
Chapter 6
The neutrino flavours oscillations is evidence of the fact that neutrinos are massive particles.
Since they are the most abundant particles after photons, they play an important role in the
history of the universe and in the structure formation and thus, the evidence of massive neu-
trinos, has important implications in cosmology. In particular, in [80], it is shown that in a
cosmological model with massive neutrinos the free-streaming of these particles leads to a
modification of the evolution of both matter and metric perturbations at scale below their free-
streaming length which results into a scale-dependent suppression of the growth of structures
at late time. It is due to the combination of two effects: 1) the matter-radiation equality take
place a bit later than in a model with massless neutrinos; 2) the gravitational potentials slowly
decays during the matter-dominated era on scales below the free-streaming scale. Thus, from
observations of the CMB and LSS, one can obtain information about neutrinos, such as a bound
on the sum of the neutrino masses.
However, it is also well known that many theories of modified gravity predict an enhancement
of the growth of large-scale structures, due to the presence of an attractive fifth force, that
could cancel out the effects of massive neutrinos and thus, from an observational point of view,
54
6.1. Cosmic degeneracy between modified gravity and massive neutrinos
such a theory would be indistinguishable from GR with very light neutrinos. In the literature
there are many studies on this topic and on the possibility of breaking such a degeneracy, called
cosmic degeneracy, and in all these works, they use as representative models of modified grav-
ity, the well known metric f(R) model.
In particular, in some of these works, the cosmic degeneracy is analysed using the DUSTGRAIN-
pathfinder simulations1 , which is a suite of cosmological N-body simulations specifically de-
signed to study cosmological models in which gravity is modified and neutrinos are massive
particles. The modified gravity model implemented by the DUSTGRAIN project is the f(R)
gravity model proposed by Hu and Sawicki in [86], for which it is possible to vary the values
of the propagating degree of freedom fR (z = 0) = df (R)/dR|z=0 and the mass of neutrinos in a
certain range. Among these papers there are
Peel et al. (2018) [81]
In this paper the authors propose to use non-Gaussian information from weak-lensing
statistic of order higher than second to break the cosmic degeneracy: the aperture mass
skewness (third order), the kurtosis (fourth order ) and the peak counts statistic as a
function of map filtering scale and source galaxy redshift. They confirm the fact that at
linear order and also at non-linear order, second-order statistics is not able to break the
cosmic degeneracy; moreover, the main result they obtain is that the only higher-order
observable, among those they have considered, which is potentially capable to distin-
guish efficiently between standard cosmology and modified gravity models with massive
neutrinos, is the peak counts and its efficiency depends on redshift and angular scale.
Merten et al. (2018) [83] In this paper the authors explore the possibility to break cosmic
degeneracy making use of three types of machine learning techniques. In particular, they
use the DUSTGRAIN-pathfinder simulations to generate the mass map which constitute
their main data set. Then, they apply to these data set different methods of characteriza-
tion and classification. In order to characterize the mass maps they use, among others, a
Convolutional Neural Network (CNN) to extract the characterising features directly from
data. The conclusion is that these techniques are promising but further investigation is
needed.
Hagstotz et al. (2019) [85]
It was already pointed out in previous papers the redshift dependence of this degeneracy;
in this paper, the authors suggest to break the cosmic degeneracy exploiting kinematic
information related to either the growth rate on large-scales or the virial velocities inside
of collapsed structures. They use the DUSTGRAIN-pathfinder simulations finding that
the suppression of the growth due to neutrino’s mass dominates in the past, while the
enhancement due to the fifth force will dominate in the future evolution. They also
study kinematic inside clusters, finding that the velocity dispersion is not affected by
neutrinos and that, in general, kinematic information is a very powerful observable in
order to detect a potential fifth force.
Finally we mention a recent works in which the authors propose to break this degeneracy with
redshift space distortions (RSD).
55
6. Degeneracies with modifications to GR
the cosmological models that, at a given redshift have similar matter power spectrum,
have different growth rate and thus, one can distinguish between them by exploiting the
velocities information encoded in the RSD. They find that the information that one can
obtain from the quadrupole of the redshift space power spectrum can in principle break
the cosmic degeneracy, but further investigation is required. They also study the redshift
evolution of the degeneracy finding that, in general, if it is present at a given redshift, it
is likely that the same disappears at another redshift.
Systematic effects in cosmological tests can be degenerate with the modifications of gravity,
therefore in order to distinguish between genuine signatures of modified gravity and to im-
prove the precision in constraining cosmological parameters it is necessary to identify and to
control them.
Weak gravitational lensing, in particular that due to LSS, i.e cosmic shear, represents a very
powerful tool in testing gravity; but weak lensing measurements are subject to a number of
systematic effects that can reduce the precision that one can reach in cosmological tests and
thus, the capacity in distinguishing different cosmological models [87]- [89]. In what follows,
we consider the three principal systematic effects that affect this kind of measurements: bary-
onic effects, photometric redshift uncertainties and galaxy intrinsic alignments.
Baryonic effects
It is well know that the physics of structure formation impacts on weak lensing observables and
many studies demonstrate that on small scales the presence of baryonic matter cannot be ig-
nored [88], otherwise one would introduce bias in the estimation of cosmological parameters;
in particular, they affect the matter power spectrum. This is because baryons have different
physical properties with respect to CDM and one of them is that baryons have non-zero pres-
sure and this fact prevents clustering on small scales. Other effects are radiative cooling, star
formation and supernova (SN) feedback [88]- [90]. In the literature there are many studies that
investigate these systematic effect using hydrodynamical simulations.
The spectroscopic distance measures of an object on cosmological scale, i.e the determination
of the redshift, requires long time observations and a very sensitive instrumentation and thus,
the estimation of the distances of all cosmological objects with high precision, is a very hard
task to achieve in this way. For this reason, in order to determine redshift distances, alternative
methods have been introduced, among them there are those based on photometric measure-
ments2 . This technique, also called photo-z’s, is based on imaging data3 which allows inferring
2 The photometric redshift was first introduced by Baum in 1962, which used multi band photometry to estimate
galaxy redshifts; then this technique was adopted by Pushell, Owen and Laing which in their paper used for the
first time the term ”photometric redshift”.
3 This means that the only information used in photometric redshift is the galaxy magnitude and colours through
a few broad filters.
56
6.2. Systematic effects in cosmological tests and degeneracies with
modifications of gravity
redshifts of a large number of objects; however, in photometric redshift techniques there are
many factors which can limit the accuracy in redshift determination. In the literature can be
found various papers on this topic and in particular, on the uncertainties associated with this
kind of measurements, e.g. [91]- [92].
The intrinsic alignment of galaxies is one of the most relevant systematic effects in weak lens-
ing measurements and thus, many studies has been made to understand and to mitigate this
effect, e.g. [93]- [94]. For intrinsic alignment one means the fact that the shape and orientation
of the galaxies are not random but are statistically determined by a number of factors (such as
formation environment, history and galaxy type) that produce such an alignment that, when
observed, can be mistaken for cosmic shear. It follows the necessity to identify these effects in
order to isolate the weak lensing signals.
As an example of how these systematic effects can impact on the precision in constraining MG
parameters we mention the paper [95] by Ferté et al. (2017). In this study, the authors use cos-
mic microwave background anisotropy measurements from the Planck satellite, cosmic shear
from CFHTLenS and DES-SV, and redshift-space distortions from BOSS data release 12 and the
6dF galaxy survey to constrain the phenomenological functions µ and Σ. They also investigate
the impact of systematics on the estimation of these functions by performing a forecast analy-
sis through Fisher matrix technique4 . Thus, they compare the constraints on µ and Σ obtained
when the systematic effects are neglected, to the case in which they are not neglected, finding
that: photometric redshift effects and the shear calibration do not affect the constraints on these
functions; on the contrary, the intrinsic alignment, shifts the constraints on Σ to higher values.
Therefore, they pay particular attention to this last effect showing that the presence of IA in
the data leads to an underestimation of the uncertainties in the phenomenological functions.
Figure 6.1: [95] This picture shows the forecasted 68% and 95% confidence contour plots on Σ0 and µ0
for a future DES-Y5 like survey when IA are considered (blue) and when it is neglected (green).
4 Appendix D.
57
6. Degeneracies with modifications to GR
The degeneracies among cosmological parameters and, in particular among those that charac-
terize deviations from GR, arise because the observables can depend by combinations of them;
however, one can try to break such degeneracies by combining different cosmological probes.
One way to investigate degeneracies is based on the Fisher information formalism5 . This ap-
proach allows us to forecast the precision under which one would be able to constrain the
parameters of a given cosmological model and to check for correlations among these parame-
ters.
In particular the phenomenological functions µ, η and Σ have been shown to be correlated each
other and also with background parameters in many theories of MG gravity (e.g. Horndeski
models, see section 5.3.1 ). To this respect, we mention two papers in which the authors study
degeneracies in modified gravity theories.
Casas et al. (2017) [101] This work is a forecasting analysis on modified gravity for
forthcoming surveys in which authors choose to characterize deviations from GR with η
and µ, considering both linear and non-linear scales. They use different time-dependent
parametrizations of these functions neglecting the scale dependence and performing a
redshift binning. The authors use the Fisher matrix approach for their analysis and in
particular, they study the correlations between the two phenomenological functions us-
ing a prior covariance matrix derived from the CMB Planck observations obtained by per-
forming a MCMC analysis. They compute the Fisher matrix on a set of 15 independent
parameters {Ωm , Ωb , ln1010 As , ns , µi , ηi }, where ηi and µi are the values of the functions
at five bin-intervals corresponding to the following redshift z = {0 − 0.5, 0.5 − 1.0, 1.0 −
1.5, 1.5 − 2.0, 2.0 − 2.5}. They also perform a Zerophase Component Analysis (ZCA) in
order to further decorrelate MG parameters and to identify which combinations of the
MG parameter amplitudes, in different redshift bins, are best constrained by future sur-
veys. They find that, in general η and µ, for different redshift, are strongly correlated but,
these correlations, are reduced or disappear when adding non-linear scales. They also
5 Appendix D.
6 CMB temperature and polarization (T and E), SNe observations, weak lensing (WL) shear of distant galaxies,
galaxy number counts (GC) and their cross-correlation.
7 Indeed these kind of observations measures δ = bδ which can be combined with the equations (1.20)-(1.21) in
g m
the sub-horizon limit giving a second order equation in δg from which it is easy to see the origin of such degeneracy.
58
6.3. Degeneracies among MG parameters
show that the higher are correlations the lower is the power in constraining MG parame-
ters. Furthermore, their study shows that there is a strong anti-correlations between the
primordial amplitude of the density fluctuation As and µi : this degeneracy can be under-
stood by considering that a large initial fluctuation amplitude leads to the same physical
effects of a larger growth of structures; but also this degeneracy is mitigate by includ-
ing non-linear scales. Moreover, they quantify the effects of adding non-linear scales on
the correlations comparing the FoC (D.5, appendix D) obtained using only linear scales
and the one obtained by including non-linear scales: the former gives FoC' 62, the lat-
est FoC' 35. Thus, this analysis, point out that by including non-linear scales one can
significantly decrease the degree of degeneracy among MG parameters.
59
Chapter 7
We are in the era of precision cosmology in which very significant steps forward have been
made in the understanding of the universe thanks to a wide number of cosmological probes
such as the CMB, galaxy clustering, weak lensing, and supernovae. The future cosmological
surveys promise to further improve this understanding by providing an enormous amount of
data with high precision. Among them we find: the Euclid mission, whose launch is planned
for 2022, with the main aim of investigating the nature and the properties of dark energy,
dark matter and gravity; DESI (Dark Energy Spectroscopic Instrument) which will start in the
present year with the purpose of measuring the effect of dark energy on the expansion of the
universe; the eBOSS (Extended Baryon Oscillation Spectroscopic Survey) which has the aims
to improve constraints on the nature of dark energy and to measure with high precision most
of the expansion history of the Universe.
In this chapter we briefly touch upon some of the recent forecasting analyses which allow an
overview of how such future surveys will improve our ability in constraining MG parameters
and thus in testing departures from GR.
Ferté et al. (2017) [95]
In this paper the authors use a wide range of observational data to constrain the phe-
nomenological functions Σ0 and µ0 1 finding no evidence for modified gravity. Moreover,
the authors, using the Fisher formalism, perform a forecasting analysis for the full five-
years DES survey and an LSST-like survey showing that these forthcoming experiments
will significantly improve the constraints on MG parameters Σ and µ. As we already
mentioned in the previous chapter, they also investigate the impact of systematic effects
on such constraints. In particular, the LSST survey, is expected to give standard devia-
tions of an order of magnitude smaller than DES-Y5 survey. The projected uncertainties
they obtained are the following
Casas et al. (2017) [101] We already mention this paper in regard of degeneracies among
MG parameters. In this chapter we come back to it to give more details on their forecasts
about the precision under which future surveys will be able to constrain MG parame-
ters. Their forecasting analysis includes a wide number of forthcoming surveys (Euclid,
DESI-ELG, SKA1 and SKA2) and make use of the Fisher information techniques which
1 They use different definitions with respect those we have adopted in this dissertation: their {Σ, µ} = our {Σ, µ}−1.
60
is applied to weak lensing and galaxy cluster using as prior the Planck data from CMB.
They considers two regimes, the linear one and the midly non-liner one (up to k ' 0.5
h/Mpc), for which they obtain the power spectrum using respectively a Boltzmann code
and two different methods for the latter (Halofit and a a semi-analytic prescription in-
cluding screening mechanisms). As we said previously, they choose η(a) and µ(a) to spec-
ify modifications of gravity and use two different approach for their analysis:
1) They bin the time dependence of the two phenomenological functions neglecting
the scale dependence without specifying any parametrized evolution
N −1 " !#
X p(zi+1 ) − p(zi ) z − zi+1
p(z) = p(z1 ) + 1 + tanh s (7.3)
2 zi+1 − zi
i=1
where z1 = 0.5 is the redshift value of the first bin, p = µ, η, N is the number of
binned value and s = 10 a smoothing parameter.
2) They use the same two parametrizations adopted by Planck Collaboration 2015
5.3.2 calling ”late-time” parametrization and ”early-time” parametrization what in
[37] are called respectively DE-related (5.12) and time-related (5.13) parametriza-
tion.
We summarize their predictions on the accuracy that future survey will reach in con-
straining µ, η and Σ by reporting their results in the table 7.1.
Table 7.1: The table shows the forecasted results for the phenomenological functions from Casas et al.
(2017) for the early-time and late-time parametrizations: 1σ fully marginalized errors on the cosmo-
logical parameters {Ωm , Ωb , ln1010 As , ns , µi , ηi } comparing different surveys in the linear and non-linear
case. For a detailed discussion of the results see [101].
61
7. Future cosmological constraints on GR and MG parameter forecasts
We finally mention a recent paper in which the authors perform a forecasting study on the
potential of future galaxy-cluster surveys in constraining the gravitational slip parameter η
using mass profiles.
Pizzuti et al. (2019) [103]
They combine simulated information of galaxy cluster mass profiles inferred from galaxy
cluster kinematics and lensing observations to reconstruct η in a model-independent way
without make any assumptions on its functional form. In particular, the dynamical mass
profile is obtained using the MAMPOSSt method, while the lensing one is obtained us-
ing previous information from the analysis of the cluster MACS 1206. They assume: 1)
that the density distribution of the cluster is dominated by dark matter for which they
assume a Navarro-Frenk-White (NFW) profile; 2)that clusters are in thermal and hydro-
static equilibrium (i.e are relaxed) and have a spherical symmetry; 3) flat ΛCDM back-
ground. By using these assumptions they arrive to an expression of the gravitational slip
parameter in terms of the lensing and dynamical masses of the cluster, respectively Mlens
and Mdyn
Rr
ds
s2
[2Mlens (s) − Mdyn (s)]
η(r) = Rr
ds
(7.4)
s2
Mlens (s)
in GR one has Mlens = Mdyn , thus η = 1.
They fix redshift at z = 0 and consider two cases: a) η scale-independent; b)η scale-
dependent. Then, for each case, they reconstruct η from synthetic sample of 15 and 75
clusters taking as fiducial value η = 1. Their results can be summarized as follows:
• The forecasting constraining power of η, as was to be expected, is higher in the
scale-independent case. The accuracies found in their analysis are reported in table
7.2;
• They investigate on the dependence of the constraints on cluster masses and num-
ber of galaxy members in the clusters. They find that the former dependence is
moderate; while the number of tracers in the clusters has a negligible impact on
the constraints. In order to deepen this effects, they consider increasing number
of tracers (Ng ) used in the dynamics fit for one single cluster finding that, for the
scale-independent case , the constraining power does not change in an apprecia-
ble way passing from 27% uncertainties at 1σ for Ng = 100 to 21% for Ng = 2000
and for already Ng & 500 the accuracy remains approximately the same. Thus, they
conclude that, with a moderate number of clusters for which a hundred measured
spectroscopic redshifts are available, combined with an honest lensing mass profile,
one can reach an high precision in constraining the gravitational slip parameter.
• Finally, they point out that their analysis is purely statistical and several possible
systematic effects have been neglected and thus, further investigation is required.
This means that, once these effects are properly taken in account, future surveys can
determine η down to the percent level of accuracy.
Table 7.2: In the table are shown the forecasting accuracy in constraining η for the scale-dependent (
evaluated at the reference scale r = 1.5 Mpc ) and for the scale-independent cases from Pizzuti et al.
(2019)
.
The overall picture emerging from these forecasting analysis is encouraging: the huge amount
62
of data coming from different cosmological probes, combined with the development of appro-
priate and differentiate analysis techniques to deal with them, lead us to believe that in the not
too distant future, we would be able to reach unprecedented level of accuracy in testing gravity
at cosmological scales.
63
Conclusions
In this thesis we studied how cosmological observations can be used to constrain modified
gravity models reducing their parameters’ space and eventually ruling out some of them. In
this work we focus on the most general scalar-tensor theories with second order equations of
motion, i.e. Horndeski gravity, and on its minimal extension, i.e beyond Horndeski models,
paying particular attention to the subclasses of these theories which predict non-zero gravita-
tional slip.
In the first chapter we wrote down the full set of the equations that describe the linear evolution
of the cosmological perturbations, parametrizing the deviations from GR with µ(a, k), η(a, k)
and Σ(a, k). The scale dependence of these functions allows for modifications of gravity on cos-
mological scales, while local gravity constraints, can still be satisfied. Following Bertschinger
and Zukin [63], we distinguished between two classes of theories of gravity: scale-independent
and scale-dependent models. We saw that on super-horizon scales the evolution of the pertur-
bations, in a generic theory of gravity, can be completely specified by the gravitational slip
parameter, η(a); this is also true, on sub-horizon scales bigger than the Jeans wavelength, for
scale-independent modified gravity models. On the contrary, the scale-dependent models,
need two independent functions to fully characterize the departures from GR; in particular,
we chose η(a, k) and µ(a, k) and we found that, this last, is the only that affects the growth of
matter perturbations on sub-horizon scale.
In the second chapter we clarified our definition of gravitational slip [33]: it is the mismatch
between the two scalar potentials appearing in the perturbed line element of the FRW met-
ric, in Newtonian gauge, due to modification of the geometrical part of the Einstein’s field
equations. Moreover, we saw that in Horndeski theories a non-standard propagation of GWs
in general, implies gravitational slip, while in beyond Horndeski models this is no longer
true. Then, we summarized very quickly the effects of a non-zero gravitational slip on LSS:
it can induce changes in the matter power spectrum and it can also lead to distortions in the
CMB temperature anisotropies and in the CMB polarization maps. Furthermore, when there
is a non-vanishing gravitational slip, the lensing potential power spectrum acquires a factor
(1 + η)2 /4; as a consequence, the B-mode power spectrum, is affected by the same factor for
` << 1000. Finally, the gravitational slip can also induce a change in the sign of CMB and LSS
cross-correlations, which in the ΛCDM is predicted to be positive.
In the third chapter we introduced the EFT formalism and wrote down the modified Pois-
son equation and the modified anisotropy constraint in terms of the EFT operators. We ob-
tained the analytical expressions for η(a, k), µ(a, k) and Σ(a, k) in the quasi-static approximation
and we introduced the α- parametrization that allowed us to characterize very efficiently the
phenomenology of the Horndeski and beyond Horndeski models. In particular, the modified
anisotropy constraint, expressed in terms of this parametrization (3.38), showed that the exis-
tence of the gravitational slip is controlled by the functions αM (running of the Planck mass),
αT (tensor speed excess) and αH (kinetic matter mixing). We then considered the analytical
expressions for the phenomenological functions in two limiting cases: super-Compton limit
(k << aMC ) and sub-Compton limit (k >> aMC ) where MC defines a transition scale between
the two extreme cases and it is related with the Compton wavelength (MC ∝ λ−1 C ). These two
limits are the most interesting cases because it is very probably that, the observational window
64
of a given theory, falls in one of these two regimes.
In the fourth chapter we accounted for the gravitational waves event GW170817 which con-
strained the GWs speed cT to be very very close to the speed of light at low redshifts. We
analysed the implications for scalar-tensor theories:
- Horndeski Lagrangian reduced to (4.4). Then we found that, if αT = 0, the conditions neces-
sary to have gravitational slip depend only by the parameter αM , which must be αM , 0 and
αM , −αB /2. Following [23], we classified the survived Horndeski theories in terms of the α
functions:
A) quintessence and K-essence: αB = 0 = αM ;
B) kinetic braiding and its extensions: αB + αM = (φ̇XG3,X )/(HG4 );
C) GBD theories: αB = −αM .
We saw that, only GBD theories and the extensions of the kinetic braiding with αM , 0 and
αM , −αB /22 , can exhibit gravitational slip, indeed, it is strictly related to the non-minimal
coupling with gravity, i.e the Horndeski function G4 (φ), which cannot be G4 = const in order
to have η , 1.
- Beyond Horndeski Lagrangian reduced to (4.38). In this case, we found that the analytical
expressions for the phenomenological functions are much more cumbersome than the Horn-
deski case and, in general, one always has gravitational slip (because αH , 0). Moreover, in
beyond Horndeski models, the fifth force can be repulsive leading to a weakening of gravity
and a suppression of clustering, contrary to what happens in Horndeski gravity where it is
always attractive.
Furthermore, we analysed, for these two classes of models, how measurements of µ(a, k), η(a, k)
and Σ(a, k) can be used to rule out subclasses of modified gravity theories.
In the fifth chapter we collected the result of the previous chapter about Horndeski gravity
and, following [13], we verified that it is very likely that the functions µ and Σ satisfy the con-
dition (µ − 1)(Σ − 1) ≥ 0, this means that a measurement of (µ − 1)(Σ − 1) < 0 would disfavour
this class of models. We also reported the results obtained in the following papers: [12], [13]
and in [37].
In the sixth chapter we briefly considered the most important sources of degeneracies one must
take into account in cosmological tests of modified gravity. In particular, we considered the cos-
mic degeneracies, i.e. the enhancement of the growth of structures predicted in some modified
gravity theories can be cancelled out by the effect of massive neutrinos and thus, such a model,
can be degenerate with GR with very light neutrinos. We also considered degeneracies arising
from systematic effects (e.g. baryonic effect, photometric redshift uncertainties and intrinsic
alignments) and degeneracies among MG parameters.
Finally, in the seventh chapter, we reported some recent forecasting analyses which allow an
overview of how future surveys will improve our ability in constraining MG parameters and
thus in discriminating between different theories of gravity.
2 This special case was analysed in [36] and it was dubbed ”no slip gravity”.
65
Appendix A
The expression for the term Cπ appearing in the equation of motion for the scalar field π, in
the QSA, is
m20 3 9 3
Cπ = Ω̇Ṙ(0) − 3Ḣc + [Ḣ(3H + ∂t ) + Ḧ]M̄13 + Ḣ 2 M̄22 + Ḣ 2 M̄32 (A.1)
4 2 2 2
In the following table one can find the expressions for the coefficients Ai , Bi , Ci (i = 1, 2, 3)
Index Ai Bi Ci
m20 1 3 3H
2 −m20 Ω̇ − M̄13 + 2H M̄32 + 4H M̂ 2 1 − 2 H 2
2 Ω̇ − 2 M̄1 − 2 M̄2 − 2 M̄3 + 2H M̂
2
2M̄˙ 3
1 k2
3 −8m22 − Ω̇
Ω + m2 Ω
H + M̄
M̄32 c − 12 (H + ∂t )M̄13 + 2a 2 − 3 Ḣ M̄22
0 3
k2 2 2 2
+ 2a 2 − Ḣ M̄3 + 2(H + Ḣ + H∂t )M̂
Table A.1: In this table are shown the coefficients of the system of equations (3.19)-(3.21). One can notice
that the only term that depends on the scale is C3 , in particular, in Horndeski gravity, this dependence
vanishes.
In the following table can be found the correspondences among the Horndeski free functions
K(φ, X) ,Gi (φ, X) (and F4,5 (φ, X) for beyond Horndeski models), the EFT operators and the α-
functions. We adopt the notations of [45], here the definition of αB differs from that in [46]
and [23]; the relation between the two definitions is αB ≡ αBB = −2αBG
66
Parameters Horndeski-Beyond Horndeski EFT operators
M∗2 H αM d 2
dt M∗ m20 Ω̇ + M̄˙ 22
Table A.2: Here the relations among the operators of the EFT actions (3.70) and those appearing in the
actions (3.44) and (3.64). The terms in bold in the middle column are those that one has to add with
respect to the Horndeski case to extend this class to beyond Horndeski models; indeed, these terms
contain the free function F4 (φ, X) and F5 (φ, X) that in Horndeski theory are zero.
f1 f2 f3 f4 f5
Table A.3: The table shows the fi functions appearing in the quasi-static approximation of the phe-
nomenological functions µ(a, k), η(a, k) and Σ(a, k)
67
A. Relations among EFT operators, α property functions and Horndeski-
beyond Horndeski’s free functions
A.0.1 Horndeski gravity
In particular, for Horndeski gravity one finds, using tab.A.2, the following relations
2M̂ 2 = M̄22 = −M̄32 , m22 = 0 =⇒ M∗2 = m20 Ω + 2M̂ 2 , M∗2 αT = −2M̂ 2 =⇒ M∗2 (1 + αT ) = m20 Ω
M̄˙ 22 = 4M̂˙ M̂ = −2M̄3 M̄˙ 3 , −M̄˙ 13 = ḢM∗2 αB + H 2 M∗2 αM αB + HM∗2 α˙B + m20 Ω̈
Now using the tables A.1 and A.2, along with the expressions (3.6)-(3.37)-(3.42), one obtains
the following coefficients
Index Ai Bi Ci
HM∗2
2 HM∗2 αB 1 2 αB
H 2 M∗2 αB2
3 0 −H(αM − αT )(1 + αT )−1 2 αcS2 + 2 (1 + αT ) + 2αB (αM − αT )
Table A.4: In this table are reported the coefficients of the system of equations (3.19)-(3.21) specialized
to the case of Horndeski gravity and in terms of the α-parametrization
" !#
3 2 ρ̄m + p̄m 2 2 ḦH
Cπ = M∗ Ḣ + (2 − αB )Ḣ − H α˙B − αB 3H + H αM + (A.3)
2 M∗2 Ḣ
M∗2 H 2 αcS2 h i
f1 = 1 + αT + βξ2 (A.4)
2(1 + αT )
M 4H 2 2
f3 = ∗ αc (A.5)
1 + αT S
M∗2 H 2 αcS2 h i
f5 = 1 + βB βξ /2 (A.6)
2(1 + αT )
2M∗2 f1 f5 1 + βB βξ /2
=⇒ = 1 + αT + βξ2 , = (A.7)
f3 f1 1 + αT + βξ2
68
where we have defined the following quantities
3
α ≡ αK + αB (A.8)
2
2
2 αB
2
βξ ≡ 2 (1 + αT ) + (αM − αT )
αcS 2
(A.9)
2 2
βB2 ≡ αB (A.10)
αcS2
" !#
3Ḣ(1 + αT ) ρ̄m + p̄m ḦH
MC2 ≡ 2 2 2
+ (2 − αB )Ḣ − H α˙B − αB 3H + H αM + (A.11)
αcS H 2 (1 + αT + βξ2 ) M∗2 Ḣ
2 " −1
βξ a2 2
#
∆1 (a, k) ≡ 1 + M (A.12)
1 + βξ2 k2 C
βB βξ /2 − βξ2
#−1
a2 2
"
∆2 (a, k) ≡ 1 + 2 MC (A.13)
1 + βξ2 k
−1
βB βξ /2 + βξ2 1
a 2
+ 2 MC2
∆3 (a, k) ≡ 2
2
(A.14)
2(1 + βξ ) 1 + βξ k
The relations between the free functions appearing in (3.64) and those appearing in the ADM
action (3.31)-(3.34) are the following
√ G3,φ
Z
A2 = K2 + −X dX √ , (A.15)
2 −X
Z √ √
A3 = dX −XG3,X − 2 −XG4,φ , (A.16)
X
A4 = −G4 + 2XG4,X + G5,φ − X 2 F4 , (A.17)
2
√ Z G5,φ
B4 = G4 + −X dX √ , (A.18)
4 −X
1√ 3 √
A5 = − −X G5,X + −X 5 F5 , (A.19)
3
Z √
B5 = − dX −XG5,X (A.20)
Using the tables A.1 and A.2, along with the expressions (3.6)-(3.37)-(3.42), one finds the fol-
lowing coefficients
69
A. Relations among EFT operators, α property functions and Horndeski-
beyond Horndeski’s free functions
Index Ai Bi Ci
HM∗2
2 HM∗2 (αB + 2αH ) 1 2 (αB + 2αH )
M∗2 Ḣ M∗2 H
3 0 −H(αM − αT )(1 + αT )−1 2 [αB + 2αH − 2(1 + 2αT )] + 2 [α̇B + 2α̇H ]+
Table A.5: In this table are reported the coefficients of the system of equations (3.19)-(3.21) specialized
to the case of beyond Horndeski theories and in terms of the α-parametrization
By comparing the table A.5 with the table A.4, one can see that, the coefficient Ai , Bi and Ci ,
for beyond Horndeski theories, can be written as a sum of the ones of the Horndeski gravity
plus a new term which contains the parameter αH
2
ABH H
1 = A1 + 2M∗ αH (A.21)
2
ABH H
2 = A2 + 2HM∗ αH (A.22)
ABH H
3 = A3 = 0 (A.23)
BBH H
1 = B1 − αH (1 + αT )
−1
(A.24)
BBH H
2 = B2 = 1 (A.25)
BBH
3 = B3
H
(A.26)
C1BH = C1H + HM∗2 αH (1 + αM ) + M∗2 α̇H (A.27)
C2BH = C2H + HM∗2 αH (A.28)
C3BH = C3H + ḢM∗2 αH + H 2 M∗2 αH (1 + αM ) + M∗2 H α̇H (A.29)
From the above expressions, one can obtain the fi functions for beyond Horndeski model
i α − αT
f1BH = f1H + H 2 M∗2 ∆C + HM∗2 αH (1 + αM ) + M∗2 α̇H H M
h
1 + αT
4 2
" #
M∗ H Ḣ
f3BH = f3H + (2 − αB )(1 + αH )∆C + αH (αB + 2αH ) (1 + αH ) − (1 + αM )
(1 + αT ) H
" #
βξ βB βξ βB
+ αH (2 + αH )(1 + )− α(cSH )2
2 αB
M∗2 H 2
( " #)
BH H αH Ḣ 1 (ρ̄m + p̄m )
f5 = f5 + (1 + αH )∆C + (α − 4αT − 2) + α̇B + αB (1 + αM ) −
(1 + αT ) 2 H2 B H M∗2 H 2
Thus
where cSH is the sound speed of the scalar perturbations for the Horndeski gravity (αH = 0)
which is given in (3.43). The other new parameters appearing in the above expressions are
70
defined as follows
Ḣ 1
∆C ≡ 2
αH + (1 + αM )αH + α̇H (A.33)
H H
β1H ≡ ∆C (1 + αM ) − (αM − αT )αH (A.34)
" #
Ḣ
β3H ≡ (2 − αB )(1 + αH )∆C + αH (αB + 2αH ) (1 + αH ) − (1 + αM ) (A.35)
H
" #
βξ βB βξ βB
+ αH (2 + αH )(1 + )− α(cSH )2 (A.36)
2 αB
α(cSH )2
!
βB βξ α
β5H ≡ 1+ αH + ∆C (1 + αH ) − H (2 − αB )(αM − αT ) (A.37)
2 2 2
One can see that, in beyond Horndeski models, the expressions of the phenomenological func-
tions are much more cumbersome because one has additional contributions coming from αH ;
if one set this last parameter to zero then, the above expressions, reduce to the ones we found
for the Horndeski gravity, indeed, in this case, ∆C = 0 and β1H = β3H = β5H = 0.
We have thus
2 #−1
2M∗2 f1BH (1 + αH )2 − f3BH f3BH a2 2
" #"
(2αH + αH ) 2
∆BH
1 (a, k) = − + + MC (1 + αH )
(1 + αH )2 f1H 2M∗2 f1H k 2
α(cSH )2 [(1 + βξ2 )(1 + αH )2 − 1] + β1H (1 + αH )2 − β3H
2
(2αH + αH ) 2
=− + M ·
∗
(1 + αH )2 H 2 2
α(cS ) (1 + βξ )
(A.40)
−1
α(cSH )2 + β3H
a 2
2 2
· H 2 2
+ 2 MC (1 + αH )
α(cS ) (1 + βξ ) k
#−1
f5BH − f1BH f1BHa2 2
" #"
∆BH
2 (a, k) = αH + + 2 MC
f1H f1H k
−1
α(cSH )2 βB βξ /2 − βξ2 + 2(β5H − β1H )
2β 1H a2 (A.41)
2
= αH + 1 + + M
C
α(cSH )2 (1 + βξ2 ) α(cSH )2 (1 + βξ2 ) k 2
1 BH 1 BH
∆BH BH BH
3 (a, k) = ∆1 (a, k) + ∆2 (a, k) + ∆1 (a, k)∆2 (a, k) (A.42)
2 2
71
Appendix B
In this appendix we report the Einstein’s field equations, in the Newtonian gauge, obtained
in [43] and where the background contribution has been subtracted, in the case of flat universe.
• 00-component: energy constraint
" 2 #
2 k
m0 Ω 2 2 Φ + 6H(Φ̇ + HΨ ) = −δρ − ρ̇DE π − 2c(π̇ − Ψ )+
a
# (B.1)
k2
"
2 2
+ m0 Ω̇ 3π H − Ḣ + 3H(π̇ − Ψ ) + 2 π − 3 Φ̇ + HΨ
a
B.0.1 α−parametrization
In [45]- [46] the authors find the following modified field equations in terms of the α property
functions
• 00-component: energy constraint
k2
3(2 − αB )H Φ̇ + (6 − αK − 6αB )H 2 Ψ + 2(1 + αH ) 2 Φ − (αK + 3αB )H 2 v̇X
" 2 !# a (B.5)
k ρ̄m + p̄m ρ̄m δ
− 2 (αB + 2αH ) − 3ḢαB + 3 2Ḣ + 2
HvX = − 2
a M∗ M∗
72
• 0i-component: momentum constraint
!
ρ̄m + p̄m ρ̄m + p̄m
2Φ̇ + (2 − αB )HΨ − αB H v̇X − 2Ḣ + 2
vX = − vm (B.6)
M∗ M∗2
3 (ρ̄m + p̄m )
(1 + αH )Ψ − (1 + αT )Φ − (αM − αT )HvX + αH v̇X = − σm (B.7)
2 M∗2
the relation with the mass term defined in (A.3) is H 2 Mπ2 = 2Cπ /M∗2
The scalar velocity potential is defined as vX ≡ −δφ/ φ̇ and represents the perturbations of
the scalar field. This quantity is introduced because the value of the scalar field, φ, is not
an observable, then one can always redefine the field φ̃ ≡ φ̃(φ) without changing √ observable
quantities. The scalar field gradient can be thought of as a four velocity uµ ≡ ∂µ / 2X if one
requires that ∂µ φ is timelike. Consequently, φ, can be interpreted as a time variable and its
gradient is the time direction for an observer which is at rest in this frame; thus vx is the scalar
velocity potential and it is invariant under redefinition of φ.
73
Appendix C
Then, by linearising the field equations and subtracting the background solutions, in QSA, one
finds the Poisson equation and the anisotropy constraint
1
φ0 k 2 Ψ = −4πGa2 ρ∆ − k 2 δφ (C.7)
2
φ0 k 2 (Φ − Ψ ) = k 2 δφ (C.8)
74
C.2. Covariant Galileons
where the scalar field can be written as φ(a, x) = φ0 (a) + δφ(a, x), with φ0 the background value
of the scalar field.
The equation for the Weyl potential follows by the first two
8πGa 2
k 2 (Φ + Ψ ) = − ρ∆ (C.9)
φ0
In particular, in Brans-Dicke theories, i.e. ω(φ) = ωBD and Λ(φ) = 0, the parameter ω has been
constrained by several solar system experiments. For instance, the Cassini mission [70] put
strong constraints on the post-Newtonian deviation from the Einstein’s gravity: ω > 4.3 · 104 at
a 2σ level. The effective gravitational coupling measured by Cavendish-type experiments and
the PPN (parametrized post-Newtonian) parameters γ, in this case, are given by
G 2ω + 4 1+ω
Gef f = , γ= (C.10)
φ0 2ω + 3 2+ω
3ξ 4 ξ5
" #
c4 2 c5 2
M∗2 = m20 − 6 2
X − 4φ̇H 9 X = m0 1 − c4 4 − c5 4 (C.12)
M6 M 2E E
1 Ḣ 3
M∗2 HαM = 4 4 c4 ξ 4 + c5 ξ 5 (C.13)
E H 2
m20
" !#
2 4 5 Ḣ
M∗ αT = 4 2c4 ξ + c5 ξ 1 + 2 (C.14)
E H
where E ≡ H(t)/H0 .
In the cubic Galileons αT = 0 = αM , therefore this theory predicts no gravitational slip. More-
over, the stringent bound on cT (4.1) in turn implies strong constraints on c4 and c5 [25]:
4 5
αT −17 2 αT −17 2
|c4 | < 4 ∼ 2.8 · 10 , |c5 | < ∼ 3.8 · 10 (C.15)
2ξ ξ 0.75ξ 5 ξ
1 This solution must be reached before the scalar field gives a non negligible contribute to the energy density of
the universe.
75
C. Some details on GBD and covariant Galileons theories
thus, if we exclude fine-tuning among the parameters of these models, the quintic and quartic
covariant Galileons are strongly disfavoured after GW 170817. On the other hand, the cubic
Galileon models, even if satisfy the bound (4.1), are in tension with ISW data.
76
Appendix D
Fisher information
Fisher information formalism (Fisher, 1935) [97] is a statistical method widely used in cosmol-
ogy to forecast the accuracy under which one is able to estimate the parameters of a models
from a given data set in upcoming experiments and thus, the ability in constraining combined
set of cosmological parameters. Indeed, the Fisher information is encoded in the Fisher matrix
which, under some specific assumptions1 , turn out to be the inverse of of the covariance matrix
which contains information about the uncertainties of the model’s parameters and also about
potential correlations (i.e. degeneracies) among them. Therefore, this approach allows to have
an estimation of the parameter’s errors before. It is a less expensive and quicker alternative to
Markov Chain Monte Carlo (MCMC) [98].
Let us suppose to have a data set depending on n random real number {x1 , . . . , xn } (that can
be represented by the vector x~) whose probability distributions depends on a set of m model
parameters {θ1 , . . . , θm } then the Fisher matrix is defined as follows [99]- [98]:
* 2 +
∂ lnL
Fij = − (D.1)
∂θi ∂θj
where ∂2 lnL/∂θi ∂θj is the Hessian matrix that one obtains by expanding the log likelihood
in Taylor series around its maximum, under the assumptions that it is locally a multi-variate
Gaussian surface. Thus, the Fisher matrix, is the ensemble average of this Hessian matrix over
observational data. A useful property of the Fisher matrix is that it is additive for independent
data sets, this is consequence of the fact that the likelihood, for independent data sets, is the
product of the likelihoods.
Assuming that the likelihood is a Gaussian around its maximum, Fisher matrix is related to the
covariance matrix by
F = C −1 (D.2)
where C is the covariance matrix.
C = h∆θ∆θ T i (D.3)
where ∆θ = θ − hθi. The diagonal part of the matrix, contains the square of errors, i.e the
variances, for each parameters; the off-diagonal elements represent the correlations among the
1 In particular, one has to require that the likelihood assumes a multivariate Gaussian form in the model param-
eters, see for example [99], [100].
77
D. Fisher information
parameters, this means that if C is not diagonal there is degeneracy among the parameters
which specify the cosmological model we are considering. Therefore, once C is known, one
can compute the correlation matrix P as
Cij
Pij = p (D.4)
Cii Cjj
the correlations can be easily visualized using the marginalized ellipsoidal contour plot in the
parameter hyperspace. The volume of the error √ ellipsoid plot is proportional to the square-root
of the determinant of the covariance matrix, detC 2 . Then, it is possible to define respectively
the ”Figure of Merit (FoM)” and the ”Figure of Correlations (FoC)” as follows
1 1
FoM = − ln[detC], FoC = − ln[detP ] (D.5)
2 2
The FoM quantifies how well model parameters are constrained by data sets, this means that
more stringent constraints correspond to higher FoM; it is a very useful tool because can
be used to compare how well the parameters are constrained in different experiments. The
FoC measures the degree of correlation among parameters, the higher is the correlation (off-
diagonal terms) the lower is the logarithm of the determinant3 , i.e. higher FoC. [101].
2 If one diagonalize the covariance matrix then, it easy to see that, the determinant of this matrix, represents the
logarithm of an error volume [101].
3 With increasing correlations among parameters the determinant tend to zero and thus the logarithm tend to
−∞, the result is that the FoC tend to +∞.
78
Acknowledgements
79
Bibliography
[1] A. G. Riess et al. [Supernova Search Team], “Observational evidence from supernovae for
an accelerating universe and a cosmological constant,” Astron. J. 116 (1998) 1009 [astro-
ph/9805201].
[4] J. Martin, “Everything You Always Wanted To Know About The Cosmological Con-
stant Problem (But Were Afraid To Ask),” Comptes Rendus Physique 13 (2012) 566
[arXiv:1205.3365 [astro-ph.CO]].
[6] T. Baker, P. G. Ferreira and C. Skordis, “The Parameterized Post-Friedmann framework for
theories of modified gravity: concepts, formalism and examples,” Phys. Rev. D 87 (2013)
no.2, 024015 [arXiv:1209.2117 [astro-ph.CO]].
[7] P. Zhang, M. Liguori, R. Bean and S. Dodelson, “Probing Gravity at Cosmological Scales
by Measurements which Test the Relationship between Gravitational Lensing and Matter
Overdensity,” Phys. Rev. Lett. 99 (2007) 141302 [arXiv:0704.1932 [astro-ph]].
[9] P. S. Corasaniti, T. Giannantonio and A. Melchiorri, “Constraining dark energy with cross-
correlated CMB and large scale structure data,” Phys. Rev. D 71 (2005) 123521 [astro-
ph/0504115].
[10] M. Motta, I. Sawicki, I. D. Saltas, L. Amendola and M. Kunz, “Probing Dark Energy
through Scale Dependence,” Phys. Rev. D 88 (2013) no.12, 124035 [arXiv:1305.0008
[astro-ph.CO]].
[11] L. Pogosian and A. Silvestri, “What can cosmology tell us about gravity? Constraining
Horndeski gravity with Σ and µ,” Phys. Rev. D 94 (2016) no.10, 104014 [arXiv:1606.05339
[astro-ph.CO]].
80
Bibliography
81
Bibliography
[30] A. De Felice and S. Tsujikawa, “f(R) theories,” Living Rev. Rel. 13 (2010) 3
[arXiv:1002.4928 [gr-qc]].
[31] L. Amendola, S. Fogli, A. Guarnizo, M. Kunz and A. Vollmer, “Model-independent con-
straints on the cosmological anisotropic stress,” Phys. Rev. D 89 (2014) no.6, 063538
[arXiv:1311.4765 [astro-ph.CO]].
[32] I. D. Saltas, I. Sawicki, L. Amendola and M. Kunz, “Anisotropic Stress as a Signature
of Nonstandard Propagation of Gravitational Waves,” Phys. Rev. Lett. 113 (2014) no.19,
191101 [arXiv:1406.7139 [astro-ph.CO]].
[33] I. Sawicki, I. D. Saltas, M. Motta, L. Amendola and M. Kunz, “Nonstandard gravitational
waves imply gravitational slip: On the difficulty of partially hiding new gravitational de-
grees of freedom,” Phys. Rev. D 95 (2017) no.8, 083520 [arXiv:1612.02002 [astro-ph.CO]].
[34] L. Amendola, G. Ballesteros and V. Pettorino, “Effects of modified gravity on B-mode
polarization,” Phys. Rev. D 90 (2014) 043009 [arXiv:1405.7004 [astro-ph.CO]].
[35] L. Amendola, M. Kunz, I. D. Saltas and I. Sawicki, “Fate of Large-Scale Structure in Mod-
ified Gravity After GW170817 and GRB170817A,” Phys. Rev. Lett. 120 (2018) no.13,
131101 [arXiv:1711.04825 [astro-ph.CO]].
[36] E. V. Linder, “No Slip Gravity,” JCAP 1803 (2018) no.03, 005 [arXiv:1801.01503 [astro-
ph.CO]].
[37] P. A. R. Ade et al. [Planck Collaboration], “Planck 2015 results. XIV. Dark energy and
modified gravity,” Astron. Astrophys. 594 (2016) A14 [arXiv:1502.01590 [astro-ph.CO]].
[38] E. Bertschinger, “On the Growth of Perturbations as a Test of Dark Energy,” Astrophys. J.
648 (2006) 797 [astro-ph/0604485].
[39] S. F. Daniel, R. R. Caldwell, A. Cooray and A. Melchiorri, “Large Scale Structure as a Probe
of Gravitational Slip,” Phys. Rev. D 77 (2008) 103513 [arXiv:0802.1068 [astro-ph]].
[40] G. Gubitosi, F. Piazza and F. Vernizzi, “The Effective Field Theory of Dark Energy,” JCAP
1302 (2013) 032 [JCAP 1302 (2013) 032] [arXiv:1210.0201 [hep-th]].
[41] S. Weinberg, “Effective Field Theory for Inflation,” Phys. Rev. D 77 (2008) 123541
[arXiv:0804.4291 [hep-th]].
[42] P. Creminelli, M. A. Luty, A. Nicolis and L. Senatore, “Starting the Universe: Stable Vio-
lation of the Null Energy Condition and Non-standard Cosmologies,” JHEP 0612 (2006)
080 [hep-th/0606090].
[43] J. K. Bloomfield, É. É. Flanagan, M. Park and S. Watson, “Dark energy or modified grav-
ity? An effective field theory approach,” JCAP 1308 (2013) 010 [arXiv:1211.7054 [astro-
ph.CO]].
[44] J. Bloomfield, “A Simplified Approach to General Scalar-Tensor Theories,” JCAP 1312
(2013) 044 [arXiv:1304.6712 [astro-ph.CO]].
[45] E. Bellini and I. Sawicki, “Maximal freedom at minimum cost: linear large-scale struc-
ture in general modifications of gravity,” JCAP 1407 (2014) 050 [arXiv:1404.3713 [astro-
ph.CO]].
[46] J. Gleyzes, D. Langlois and F. Vernizzi, “A unifying description of dark energy,” Int. J.
Mod. Phys. D 23 (2015) no.13, 1443010 [arXiv:1411.3712 [hep-th]].
[47] J. Gleyzes, D. Langlois, F. Piazza and F. Vernizzi, “Essential Building Blocks of Dark En-
ergy,” JCAP 1308 (2013) 025 [arXiv:1304.4840 [hep-th]].
82
Bibliography
[48] J. Gleyzes, D. Langlois, F. Piazza and F. Vernizzi, “Healthy theories beyond Horndeski,”
Phys. Rev. Lett. 114 (2015) no.21, 211101 [arXiv:1404.6495 [hep-th]].
[49] J. Gleyzes, D. Langlois, F. Piazza and F. Vernizzi, “Exploring gravitational theories beyond
Horndeski,” JCAP 1502 (2015) 018 [arXiv:1408.1952 [astro-ph.CO]].
[50] M. Mancarella, “Consistency tests of the Universe and cosmic relics,” Cosmology and
Extra-Galactic Astrophysics [astro-ph.CO]. Université Paris-Saclay (2017).
[51] A. De Felice, K. Koyama and S. Tsujikawa, “Observational signatures of the theories be-
yond Horndeski,” JCAP 1505 (2015) no.05, 058 [arXiv:1503.06539 [gr-qc]].
[52] G. D’Amico, Z. Huang, M. Mancarella and F. Vernizzi, “Weakening Gravity on Redshift-
Survey Scales with Kinetic Matter Mixing,” JCAP 1702 (2017) 014 [arXiv:1609.01272
[astro-ph.CO]].
[53] D. Traykova, E. Bellini and P. G. Ferreira, “The phenomenology of beyond Horndeski
gravity,” arXiv:1902.10687 [astro-ph.CO].
[54] Scott F. Daniel, “Effects of gravitational slip on the higher-orde moments of the matter
distribution,” [arXiv:0909.0532v1 [hep-th]].
[55] P. Creminelli, M. Lewandowski, G. Tambalo and F. Vernizzi, “Gravitational Wave Decay
into Dark Energy,” JCAP 1812 (2018) no.12, 025 [arXiv:1809.03484 [astro-ph.CO]].
[56] J. G. Williams, S. G. Turyshev and D. H. Boggs, “Progress in lunar laser ranging tests of
relativistic gravity,” Phys. Rev. Lett. 93 (2004) 261101 [gr-qc/0411113].
[57] E. Babichev, C. Deffayet and G. Esposito-Farese, “Constraints on Shift-Symmetric Scalar-
Tensor Theories with a Vainshtein Mechanism from Bounds on the Time Variation of G,”
Phys. Rev. Lett. 107 (2011) 251102 [arXiv:1107.1569 [gr-qc]].
[58] I. D. Saltas and M. Kunz, “Anisotropic stress and stability in modified gravity models,”
Phys. Rev. D 83 (2011) 064042 [arXiv:1012.3171 [gr-qc]].
[59] A. Hojjati, L. Pogosian and G. B. Zhao, “Testing gravity with CAMB and CosmoMC,” JCAP
1108 (2011) 005 [arXiv:1106.4543 [astro-ph.CO]].
[60] G. B. Zhao, T. Giannantonio, L. Pogosian, A. Silvestri, D. J. Bacon, K. Koyama, R. C. Nichol
and Y. S. Song, “Probing modifications of General Relativity using current cosmological
observations,” Phys. Rev. D 81 (2010) 103510 [arXiv:1003.0001 [astro-ph.CO]].
[61] E. Bertschinger, “One Gravitational Potential or Two? Forecasts and Tests,” Phil. Trans. R.
Soc. A, 369 (2011), 4947-4961 [arXiv:1111.4659 [astro-ph.CO]].
[62] R. Caldwell, A. Cooray and A. Melchiorri, “Constraints on a New Post-General Relativity
Cosmological Parameter,” Phys. Rev. D 76 (2007) 023507 [astro-ph/0703375 [ASTRO-
PH]].
[63] E. Bertschinger and P. Zukin, “Distinguishing Modified Gravity from Dark Energy,” Phys.
Rev. D 78 (2008) 024015 [arXiv:0801.2431 [astro-ph]].
[64] A. Joyce, L. Lombriser and F. Schmidt, “Dark Energy Versus Modified Gravity,” Ann. Rev.
Nucl. Part. Sci. 66 (2016) 95 [arXiv:1601.06133 [astro-ph.CO]].
[65] G. W. Horndeski, “Second-order scalar-tensor field equations in a four-dimensional
space,” Int. J. Theor. Phys. 10 (1974) 363.
[66] C. Deffayet, X. Gao, D. A. Steer and G. Zahariade, “From k-essence to generalised
Galileons,” Phys. Rev. D 84 (2011) 064039 [arXiv:1103.3260 [hep-th]].
83
Bibliography
[68] J. Khoury and A. Weltman, “Chameleon fields: Awaiting surprises for tests of gravity in
space,” Phys. Rev. Lett. 93 (2004) 171104 [astro-ph/0309300].
[69] J. Khoury and A. Weltman, “Chameleon cosmology,” Phys. Rev. D 69 (2004) 044026
[astro-ph/0309411].
[70] B. Bertotti, L. Iess and P. Tortora, “A test of general relativity using radio links with the
Cassini spacecraft,” Nature 425 (2003) 374.
[72] T. Clifton, J. D. Barrow and R. J. Scherrer, “Constraints on the variation of G from primor-
dial nucleosynthesis,” Phys. Rev. D 71 (2005) 123526 [astro-ph/0504418].
[74] A. I. Vainshtein, “To the problem of nonvanishing gravitation mass,” Phys. Lett. 39B
(1972) 393.
[75] A. Joyce, B. Jain, J. Khoury and M. Trodden, “Beyond the Cosmological Standard Model,”
Phys. Rept. 568 (2015) 1 [arXiv:1407.0059 [astro-ph.CO]].
[76] J. Wang, L. Hui and J. Khoury, “No-Go Theorems for Generalized Chameleon Field Theo-
ries,” Phys. Rev. Lett. 109 (2012) 241301 [arXiv:1208.4612 [astro-ph.CO]].
[77] G. F. Smoot et al. [COBE Collaboration], “Structure in the COBE differential microwave
radiometer first year maps,” Astrophys. J. 396 (1992) L1.
[78] A. Lewis and A. Challinor, “Weak gravitational lensing of the CMB,” Phys. Rept. 429
(2006) 1 [astro-ph/0601594].
[80] J. Lesgourgues and S. Pastor, “Massive neutrinos and cosmology,” Phys. Rept. 429 (2006)
307 [astro-ph/0603494].
[81] A. Peel, V. Pettorino, C. Giocoli, J. L. Starck and M. Baldi, “Breaking degeneracies in mod-
ified gravity with higher (than 2nd) order weak-lensing statistics,” Astron. Astrophys. 619
(2018) A38 [arXiv:1805.05146 [astro-ph.CO]].
[82] C. Giocoli, M. Baldi and L. Moscardini, “Weak Lensing Light-Cones in Modified Gravity
simulations with and without Massive Neutrinos,” Mon. Not. Roy. Astron. Soc. 481 (2018)
2813 [arXiv:1806.04681 [astro-ph.CO]].
84
Bibliography
[85] S. Hagstotz, M. Gronke, D. Mota and M. Baldi, “Breaking cosmic degeneracies: Disen-
tangling neutrinos and modified gravity with kinematic information,” arXiv:1902.01868
[astro-ph.CO].
[86] W. Hu and I. Sawicki, “Models of f(R) Cosmic Acceleration that Evade Solar-System Tests,”
Phys. Rev. D 76 (2007) 064004 [arXiv:0705.1158 [astro-ph]].
[87] H. Hoekstra and B. Jain, “Weak Gravitational Lensing and its Cosmological Applications,”
Ann. Rev. Nucl. Part. Sci. 58 (2008) 99 [arXiv:0805.0139 [astro-ph]].
[88] Van Daalen, M. P., Schaye, J., Booth, C. M., et al., “The effects of galaxy formation on
the matter power spectrum: A challenge for precision cosmology,” [arXiv:1104.1174v2
[astro-ph.CO]]
[89] R. Mandelbaum, “Weak lensing for precision cosmology,” Ann. Rev. Astron. Astrophys.
56 (2018) 393 [arXiv:1710.03235 [astro-ph.CO]].
[90] I. Mohammed and U. Seljak, “Analytic model for the matter power spectrum, its covari-
ance matrix, and baryonic effects,” Mon. Not. Roy. Astron. Soc. 445 (2014) no.4, 3382
[arXiv:1407.0060 [astro-ph.CO]].
[91] H. Hildebrandt, S. Arnouts, P. Capak, et al. 2010, “ PHAT: PHoto-z Accuracy Testing,” [
arXiv:1008.0658v1 [astro-ph.CO]]
[93] M. A. Troxel and M. Ishak, “The Intrinsic Alignment of Galaxies and its Impact on
Weak Gravitational Lensing in an Era of Precision Cosmology,” Phys. Rept. 558 (2014)
1 [arXiv:1407.6990 [astro-ph.CO]].
[94] D. Kirk et al., “Galaxy alignments: Observations and impact on cosmology,” Space Sci.
Rev. 193 (2015) no.1-4, 139 [arXiv:1504.05465 [astro-ph.GA]].
[95] A. Ferté, D. Kirk, A. R. Liddle and J. Zuntz, “Testing gravity on cosmological scales
with cosmic shear, cosmic microwave background anisotropies, and redshift-space dis-
tortions,” [arXiv:1712.01846 [astro-ph.CO]].
[96] A. Hojjati, “Degeneracies in parametrized modified gravity models,” JCAP 1301 (2013)
009 [arXiv:1210.3903 [astro-ph.CO]].
[97] R. A. Fisher, ”The Logic of Inductive Inference.” Journal of the Royal Statistical Society
98, no. 1 (1935)
[99] L. Verde, “A practical guide to Basic Statistical Techniques for Data Analysis in Cosmol-
ogy,” (2007)[arXiv:0712.3028 [astro-ph]].
[100] D. Coe, “ Fisher Matrices and Confidence Ellipses: A Quick-Start Guide and Software,”
[arXiv:0906.4123v1 [astro-ph.IM]]
[101] S. Casas, M. Kunz, M. Martinelli and V. Pettorino, “Linear and non-linear Modified Grav-
ity forecasts with future surveys,” Phys. Dark Univ. 18 (2017) 73 [arXiv:1703.01271 [astro-
ph.CO]].
85
Bibliography
86