1511 05962 PDF

Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Searching for scalar gravitational interactions in current and future cosmological data

Alireza Hojjati1,2 , Aaron Plahn2 , Alex Zucca2 , Levon Pogosian2 ,


Philippe Brax3 , Anne-Christine Davis4 , Gong-Bo Zhao5,6
1

arXiv:1511.05962v1 [astro-ph.CO] 18 Nov 2015

Department of Physics and Astronomy, University of British Columbia, Vancouver, BC, V6T 1Z1, Canada
2
Department of Physics, Simon Fraser University, Burnaby, BC, V5A 1S6, Canada
3
Institut de Physique Theorique, CEA, IPhT, CNRS, URA 2306, F-91191Gif/Yvette Cedex, France
4
DAMTP, Centre for Mathematical Sciences, University of Cambridge, Wilberforce Road, Cambridge CB3 0WA, U.K.
5
National Astronomy Observatories, Chinese Academy of Science,
Beijing, 100012, Peoples Republic of China and
6
Institute of Cosmology and Gravitation, University of Portsmouth, Portsmouth, PO1 3FX, United Kingdom
Modified gravity theories often contain a scalar field of gravitational strength which interacts with
matter. We examine constraints on the range and the coupling strength of a scalar gravitational
degree of freedom using a subset of current data that can be safely analyzed within the linear
perturbation theory. Using a model-independent implementation of scalar-tensor theories in MGCAMB
in terms of two functions of the scale factor describing the mass and the coupling of the scalar degree
of freedom, we derive constraints on the f (R), generalized chameleon, Symmetron and Dilaton
models. Since most of the large scale structure data available today is from relatively low redshifts,
only a limited range of observed scales is in the linear regime, leading to relatively weak constraints.
We then perform a forecast for a future large scale structure survey, such as Large Synoptic Survey
Telescope (LSST), which will map a significant volume at higher redshifts, and show that it will
produce much stronger constraints on scalar interactions in specific models. We also perform a
principal component analysis and find that future surveys should be able to provide tight constraints
on several eigenmodes of the scalar mass evolution.

I.

INTRODUCTION

A non-vanishing cosmological constant, , is the simplest and the most common explanation of the observed
cosmic acceleration [1, 2]. Because, gravitationally, is
equivalent to the large vacuum energy predicted in particle physics, its value requires a technically unnatural
fine-tuning [3, 4] in order to be consistent with observations. The cosmological constant could be embedded
in a larger class of dark energy models, where dynamics
dictate the value of the vacuum energy. Because of the
absence of apparent violation of Lorentz invariance in the
Universe, dark energy is commonly described by the field
theory of a scalar. Usually, some degree of fine-tuning of
the parameters of the model must be introduced.
Another explanation could be provided by a modification of the laws of gravity on large scales. Such modifications generically involve a scalar degree of freedom which
can lead to dynamical dark energy when the range of
the scalar interaction is cosmological. As a result, scalartensor models with couplings to matter represent a wellmotivated and versatile class of dark energy. Theories
describing the behaviour of the scalar field involve conformal [5] and disformal couplings to matter [6, 7]. It turns
out that the disformal coupling is severely constrained by
local experiments and cosmological observations [8, 9].
On the other hand, the conformal couplings, albeit large
on cosmological scales, can be screened in the local environment where none of their effects, such as deviations
from Newtons law, have been uncovered.
In this paper, we will focus on scalar-tensor models
with screening mechanisms that are broadly classified
to be of chameleon type [10, 11], i.e. where either the

mass of the scalar and/or its coupling to matter has a


dependence of the local matter density. Specifically, we
will consider three types of models with the chameleon
property: the f (R) theories [12, 13], the environmentally
dependent Dilatons [14] and the Symmetron [15]. The
latter two models use the DamourPolyakov mechanism
for screening [16]. We will take advantage of the fact
that these three very different types of models can be
described using the same formalism defined in terms of
two dynamical functions m(a) and (a), where a is the
scale factor [17, 18]. The first one represents the mass
of the scalar in the cosmological background at the redshift 1 + z = a1 and the second one is the coupling
of the scalar to matter. The growth of cosmological perturbations in these models in the linear regime and on
sub-horizon scales can be entirely described using a single function, (k, a) = 2 2 (a)/[1 + m2 (a)a2 /k 2 ], which
appears in the modification of Newtons constant and in
the modified relation between the curvature and the gravitational potential. While in this paper we shall restrict
ourselves to observables which are sensitive to the linear
regime only, we note that, given m(a) and (a), one can
also reconstruct the full non-linear dynamics of the models. Namely, using a known evolution of the background
matter density, (a), one can express the mass and the
coupling as functions of local matter density: m((x, t))
and ((x, t)) and use them to perform N-body simulations of these models or to analyse local gravitational
tests.
Modified gravity (MG) and its comparison with dark
energy has been investigated using various cosmological
probes in the last ten years [1944]. Some models of modified gravity, such as the f (R) theories, have been strongly

2
constrained by observations both cosmological and astrophysical. The strongest bound on the range of the scalar
interaction, expressed in terms of the parameter fR0 , is
at the level of 107 and comes from astrophysical tests
of modified gravity using the period of cepheids or the
gas dynamics of dwarf galaxies [4547]. The cosmological bounds are less effective, at the level of 105 [43, 48].
On the other hand, dilatons and symmetrons have not
been constrained as systematically as f (R) on cosmological scales. Only a few tests have been performed using the [m(a), (a)] parameterisation [42]. The strongest
bounds on dilatons and symmetrons still spring from local gravitational tests [49]. Local tests of gravity for the
chameleon-type models of modified gravity imply that
the range of the scalar interaction cannot exceed 1 Mpc,
implying that linear analyses are limited to probing only
some of the features of the chameleon screening mechanisms. On the other hand, studying effects of modified
gravity on shorter scales requires the use of either semianalytical methods suited to the quasi-linear regime of
cosmological perturbations or N-body simulations, both
of which are model-specific. Here, in order to keep our
analysis as model-independent as possible, we shall restrict ourselves to observables which are sensitive to the
linear regime only.
The range of scales that are safely in the linear regime
at low redshifts is quite limited. Most of the large scale
structure data available today is from relatively low redshifts and provides only weak constraints on scalar-tensor
models unless one considers information from non-linear
scales. The only way to do so is to run N-body simulations
for specific models. On the other hand, future surveys,
such as LSST [50] and Euclid [51], will provide a high volume of data from higher redshifts at which the range of
linear scales is significantly larger, allowing one to deduce
stronger constraints on scalar interactions not only for
specific models but in a more general model-independent
way. In this paper, we start by deriving constraints on
f (R), Symmetron and Dilatons models from the subset
of todays data that can be safely considered to be in
the linear regime. Then we perform a Fisher forecast for
the same models assuming data from a future LSST-like
survey in combination with other types of data expected
over the next 5-10 years to show that they will be significantly tighter. Finally, we perform a principal component
analysis (PCA) forecast of m(a) for the same future data,
assuming that (a) is a slowly varying function that can
be taken to be a O(1) constant over the range of redshifts
relevant to LSST.

II.

THE MODEL

where g is the Einstein frame metric, are the matter


fields that follow geodesics of A2 ()g , and L is the
scalar field Lagrangian given by
L =

()2
V () .
2

The action in Eq. (1) is a Generalized Brans-Dicke


(GBD) theory [5] that includes a potential for the scalar
field. In all GBD, the scalar field mediates an additional
gravitational interaction between massive particles. The
net force on a test mass is given by
~ d ln A()
~ ,
f~ =
d

(3)

where is the Newtonian potential. Since solar system


and laboratory tests severely constrain the presence of
the scalar force, GBD can only be viable if either the
coupling of the scalar field to matter is always negligible,
or if there is a dynamical screening mechanism that suppresses the force in dense environments. The latter can be
accomplished with appropriately chosen functional forms
of A() and V (). Because of its coupling to matter, the
scalar field dynamics are determined by an effective potential which takes into account the presence of the conserved matter density of the environment
Veff () = V () + (A() 1).

(4)

For some forms of V () and A(), the effective potential can have a density dependent minimum, (). The
scalar force will be screened if either the mass of the
field happens to be extremely large or the coupling happens to be negligibly small at the minimum of Veff ().
Such models can be broadly classified as Generalized
Chameleons (GC), and include the original chameleon
model [10], f (R), dilatons [14] and symmetrons [15].
We note that the GC scalar-tensor theories considered
in this work are viable only if the field stays at the minimum of the effective potential Veff () [18]. In this case,
the effective dark energy equation of state is indistinguishable from 1 and the expansion history practically
the same as in the CDM model. Furthermore, as long
as the scalar field is at its density dependent minimum,
(), the theory can be described parametrically from
the sole knowledge of the mass function m() and the
coupling () at the minimum of the potential [17, 18]
Z c
() c
1
()
= 2
d 2 ,
(5)
mPl
mPl
m ()
where we have identified the mass as the second derivative
m2 () =

We consider scalar-tensor theories defined by the action




Z
R
4
2
S = d x g
+ L + Lm [, A ()g ] , (1)
16G

(2)

d2 Veff
|=()
d2

(6)

and the coupling


() = mPl

d ln A
|=() .
d

(7)

3
It is often simpler to characterize the functions m() and
() using the time evolution of the matter density of the
Universe
0
(8)
(a) = 3
a
where a is the scale factor whose value now is a0 = 1.
This allows one to describe characteristic models in a
simple way and the full dynamics can be recovered from
the time evolution of the mass and coupling functions,
m(a), (a).
A.

Evolution of linear perturbations

While the scalar-tensor theories considered in this work


predict the same expansion history as CDM, the existence of the additional scalar interaction gives them
distinguishing features in the evolution of linear matter
and metric perturbations. More specifically, the attractive force mediated by the scalar enhances the overall
growth of inhomogeneities. In addition, the relation between the curvature perturbation and the Newtonian
potential is modified [52]. Both of these effects can be
captured in terms of two phenomenological functions employed in MGCAMB1 [25, 34, 54], parametrizing effective
modifications to the Poisson and the anisotropy Einstein
equations in Fourier space. Namely, one defines (a, k)
and (a, k), such that
k 2 = 4Ga2 (k, a),

= (k, a)

(9)
(10)

where is the comoving matter density contrast2 . In the


quasi-static approximation, whose validity is discussed
below, functions (k, a) and (k, a) can be expressed in
terms of m(a) and (a) as [18].
(a, k) = A2 ()(1 + (k, a)),
1 (k, a)
(a, k) =
,
1 + (k, a)

(11)

In the quasi-static approximation, the equation governing the evolution of matter density contrast reads
3
00 + H 0 m H2 (k, a) = 0
2

(14)

where 0 is the derivative with respect to conformal time


and H = a0 /a. Two regimes can be distinguished. When
the mode k is outside the Compton wavelength of the
scalar field, i.e. k  am(a),   1 and the growth is not
modified. Inside the Compton wavelength, k  am(a),
gravity is enhanced by 1+2 2 (a), implying more growth.
In addition, in the Symmetron and Dilatons models, the
coupling (a) depends on the matter density and controls
the transition to the enhanced growth.
Since  is a manifestly non-negative number, the
growth is generically enhanced. Also, generically, < 1
in these models. At the same time, the relation between
the lensing potential + and the matter density is
effectively unchanged. Namely, if one defines (k, a) as
k 2 ( + ) = 8Ga2 (k, a),

(15)

= A2 ()

(16)

then

and is effectively unity for all viable modes. Thus, a clear


detection of 6= 1 would not only signal a breakdown of
CDM but would rule out the entire class of GBD models. We note that, even though is constrained to be very
close to unity in viable GBD models, its time derivative,
can, in principle, be non-negligible and affect the ob,
servables via the Integrated Sachs-Wolfe (ISW) effect.
When functions m(a) and (a) are regular, which is the
case for chameleon models such as f (R), and for dilatons,
the error introduced by working in the quasi-static approximation scales as H/k [55]. For models such as the
Symmetron, in which the functions m(a) and (a) vanish
with a power n < 1 for a > a? and are zero for a < a?
(and thus have a diverging derivative at a? ), the accuracy
is reduced to (H/k)n [55, 56].

(12)
B.

Functions m(a) and (a) in f(R)

where
(k, a) =

2 2 (a)
.
1 + m2 (a)a2 /k 2

(13)

The conformal factor A2 () that appears in Eq. (11) is


indistinguishable from unity for viable models within the
class of scalar-tensor theories considered in this paper,
and can be safely ignored. CDM is recovered when 
0 and = = 1.

1
2

MGCAMB is a publicly available patch to CAMB [53].


These equations are valid at late times, when the contribution
of relativistic species can be neglected. Equations used in MGCAMB are more general and are valid at all times [35].

In what follows, we briefly motivate specific functional


forms of (m(a) and (a) adopted for the analysis in Sections III and IV. Given the forms of m(a) and (a),
the predictions for the observables can be calculated
using MGCAMB [54]. Plots of the CMB temperature
anisotropy and the matter power spectra for a few representative models are shown in Fig. 1.
Among theories exhibiting chameleon screening are the
f (R) class of models [57, 58] described by the action


Z
f (R)
4
S = d x g
+ Lm [, g ]
(17)
16G
where the function f (R) is designed to depart from the
Einstein-Hibert form at smaller values of the curvature

6000
104

4000

P ( k)

(l +1)ClTT (2)1

5000

3000
2000

103
102

0
0.000
0.001
0.002
0.003
0.004
0.005
0.006

101

102

103

101

P(k)/PCDM(k)

ClTT /ClTTCDM

1000

101

102

CDM

103

Symmetron

2.0

10-3

10-2

10-1

100

10-3

10-2

10-1

100

1.5
1.0
0.5
0.0

Dilaton

k [h/Mpc]

Hu-Sawicki

FIG. 1. Plots of CMB temperature anisotropy ClT T (left) and the matter power spectrum P (k) (right) for the models studied
in this paper. The parameters used for the Symmetron model are: a? = 0.25, ? = 1 and ? = 103 . The parameters used for
the Dilatons model are: 0 = 3 and 0 = 6 103 . The parameters used for Hu-Sawicki f (R) model are fR0 = 104 and n = 1.
The yellow shaded region shows the scales that are not taken into account in the data analysis.

R. As a specific example, we take the form proposed by


Hu and Sawicki (HS) [12],
f (R) = R 2 +

fR0 R0n+1
,
n Rn

(18)

where is the cosmological constant term, R0 is the value


of the curvature today and fR0 (1 df /dR)R=R0 . As
argued in [12, 59, 60], all viable f (R) models should be
of such disappearing cosmological constant type [60],
and models similar to HS were proposed
in [59, 60].
For all f (R) models, (a) = 1/ 6, while the mass
function is model dependent. In the HS model, we have

m(a) = m0

4 + m a3
4 + m

(n+2)/2
(19)

where and m are the dark energy and matter density fractions today, and m0 is a mass scale that can be
expressed in terms of fR0 as [18]
s
4 + m
m 0 = H0
.
(20)
(n + 1)fR0
6
Local tests of gravity require fR0 <
10 [49], while astrophysical constraints from dwarf galaxies imply that

7
fR0 <
10 [47]. These bounds depend on accurate modelling of non-linear physics. In what follows, we will derive
the constraint on fR0 from current cosmological data using only information from linear scales, and also forecast
constraints expected from future surveys like LSST.

Representative CMB and matter power spectra for


f (R) are shown in Fig. 1. A notable effect on the CMB
spectrum is the suppression of power at small multipoles, which is due to the reduced Integrated Sachs-Wolfe
(ISW) effect. The magnitude of the ISW effect is proportional to the net change in the gravitational potential
along the line of sight. In CDM, the change in the potential is a reduction caused by the onset of cosmic acceleration. In f (R), the additional scalar force enhances the
potential which, combined with the decay due to acceleration, leads to a smaller net change and, thus, a smaller
ISW effect. The other notable impact of f (R) on the
CMB spectrum is the enhanced lensing, which has the effect of slightly dumping the peaks. The enhanced growth
is more evident in the plot of P (k). Qualitatively, these
features are common to all GBD models.

5
C.

scalar force. For cosmological densities, the transition occurs at

Functions m(a) and (a) for dilatons

Another relevant example is the environmentally dependent Dilaton [14], where the screening mechanism is
of the Damour-Polyakov type [16]. This model, inspired
by string theory in the large string coupling limit, has an
exponentially runaway potential
V () = V0 e/mPl ,

(21)

with the value of V0 set to generate the current acceleration of the Universe, while the coupling function is
A() = 1 +

A2
( ? )2 .
2m2Pl

(22)

In dense environments, the minimum of the effective potential approaches = ? , and the coupling function
(a) vanishes. The coefficient A2 has to be large to satisfy
6
local tests of gravity; typically A2 >
10 . These models
can be described by a mass function given by
m2 (a) = 3A2 H 2 (a)

? =

and
r
(a) = ?

? =

where 0 = /m 2.7 is related to V0 , and is determined by requiring that plays the role of dark energy.
We will present our constraints on the mass in terms of
a scalar-force range parameter 0 , defined as
H0
1
0 =
=
,
c m0
3A2

(25)

 a 3
?

H0 1
.
c m?

(30)

(31)

Representative CMB and matter power spectra for this


model are shown in Fig. 1.

E.

Generalized Chameleon models

In our forecasts, we will also consider generalized models of chameleon type [61] defined by
m(a) = m0 ar , (a) = 0 as .

where m0 = m(a = 1). We show representative CMB and


matter power spectra for the Dilaton model in Fig. 1,
with parameter values being large on purpose to exaggerate the qualitative features of the model.

D.

while (a) = 0 for a < a? . As in the case of dilatons, we


represent our bounds in terms of a range parameter ? ,
defined as

(23)

(24)

(28)

where m is the matter density today. Thus, one can


work with a? , along with m? and ? , as the three free
parameters of the theory. At a > a? , the model can be
described by
r
 a 3
?
m(a) = m? 1
(29)
a

and, assuming matter domination, a coupling function


(a) = 0 a3 ,

m
mpl m2? 2?
=
,
3
a?
2?

(32)

In practically all viable chameleon models, the coupling


function is expected to vary extremely slowly at redshifts
probed by large scale structure surveys. Thus, for all
practical purposes, it can be taken to be a constant of
order unity.

Functions m(a) and (a) for symmetrons

Another example of a GBD model with the DamourPolyakov screening mechanims is the Symmetron [15],
where the scalar field has a quartic potential,
"

2
 4 #
m2? 2?
1

1
V () = V0 +

+
(26)
2
2 2?
4 ?
and a coupling function,
A() = 1 +

? 2
.
2?

(27)

When matter density is large, the effective potential has


a minimum at = 0 and A() 1, thus decoupling
the scalar from matter. At lower densities, the effective
potential acquires a non-zero minimum, activating the

F.

Binned Model

As discussed so far, for any of the aforementioned models, each with its own theoretical motivation, one can determine the functional forms of m(a) and (a). This effectively reduces the two free functions m(a) and (a) to a
handful of parameters. However, one might be interested
in knowing how well the two functions are constrained in
general, without regard for any specific model. One can
then proceed by discretizing either of the two functions
in bins of redshift space and treating the amplitude in
each bin as a free parameter to be constrained.
Varying both, the coupling and the mass functions,
simultaneously would be redundant, since their effect is
largely degenerate. Since it is the mass parameter that affects the shape of the matter power spectrum, we fix (a)

6
to a constant value of order unity and bin m(a) in redshift. If a non-zero m1 (a) were detected, it would signal
the presence of a scalar interaction and further investigation would be required to determine if the variation
occurs in (a), m(a) or both.
While a binning scheme gives a model independent
(rather a far less model dependent) treatment of m(a),
the larger number of parameters (values of m in each bin)
results in weaker constraints on the individual parameters. To extract useful information, we apply the Principal Components Analysis (PCA) technique (reviewed
in Section IV G). The resulting Principal Components
(PCs) are linear combinations of the original bin values
and the propagated uncertainty (from original errors on
the bins) in their values can inform us about those PCs
that are best constrained by data and the number of degrees of freedom the can potentially be constrained.

III.

CONSTRAINTS FROM CURRENT DATA

In this Section, we use a combination of currently


available CMB, lensing and Baryonic Acoustic Oscillation (BAO) data, as well as measurements of the matter
power spectrum, to derive constraints on the GBD parameters. To compute the observables, we implemented
the parametrizations described in the previous Section in
MGCAMB. We then use it with an appropriately modified version of CosmoMC [62] to obtain the posterior distributions for the model parameters. Since current data
is unable to simultaneously constrain multiple GBD parameters, we will only consider models from the previous
Section for which meaningful constraints are possible.

A.

The datasets used in the analysis

We use the measurements of CMB temperature


anisotropy from the second data release of the Planck
survey [63] in the form of the full Planck TT high-` likelihood (30 < ` < 2500) along with the low-` polarization
(` < 30). We refer to the above datasets as PLC. We also
consider the Planck 2015 lensing potential spectrum [64]
extracted from mode-coupling correlations, and refer to
this dataset as CMBLens.
In addition to inducing higher order correlations, lensing by large scale structures affects the TT spectrum at
higher `, slightly damping the oscillatory features. In [65],
and subsequently in [63], the lensing contribution to TT
was quantified via an amplitude AL multiplying the lensing power spectrum in the calculation of the theoretical
prediction for TT. The parameter AL was used to quantify the significance of detection of the lensing contribution to TT. However, instead of measuring the expected
value of AL = 1, since the lensing contribution to TT is
calculated from the same model as the rest of the spectrum, the best fit value obtained for LCDM from the
PLC dataset in [63] was AL = 1.22 0.10, or two stan-

P
P
Fixed
m
Varying
m
P
Data sets
f R0
fR0
m (eV)
PLC+BAO
0.05 (0.14)
0.08 (0.23)
0.24(0.35)
+CMBLens 3 (8) 103 0.6 (1.6) 102 0.22(0.31)
+MPK
0.6(1.6) 104 0.7 (1.7) 104 0.24(0.34)
+WL
3 (7) 105
4 (9) 105
0.23(0.33)
TABLE I. The 68% (95%) CL upper limits of fR0 and the
sum of neutrino masses using different combinations of data
sets shown in the table.

dard deviations away from the expectation. As discussed


in [65] (see also [66]) this is due to an apparent tension
between the higher-` and lower-` data when trying to fit
LCDM to Planck TT data. To negate the effect of this
tension, the parameter AL was sometimes co-varied with
other parameters when deriving constraints on LCDM in
[63]. In what follows, we take the view that AL is not a
physical parameter and should be held fixed to 1 when
deriving constraints on cosmological models. However, we
also investigate and discuss the effect of co-varying AL
in the case of f (R).
For BAO measurements, we used data from the 6dF
survey [67] and from SDSS, specifically the MGS [68]
and BOSS data releases (LOWZ and CMASS) [69].
We also use the matter power spectrum (referred to
as MPK) from SDSS LRG DR4 [70], but only on linear
scales, k 0.1 h/Mpc. We are aware of the fact that nonlinear corrections can play a role even at k <
0.1Mpc
and that a proper treatment of the bias and the redshift
space distortions (RSD) must take them into account.
This was studied at length in [71] for the SDSS DR9
power spectrum and it was found that the differences in
the upper bounds on neutrino masses obtained using four
different RSD models were under 20%. Based on this, we
expect that bounds on the GBD parameters (such as fR0 )
obtained from MPK are accurate to within 30%, which
is sufficient given that constraints form current data are
relatively weak.
Finally, we consider the weak lensing data from
the Canada France Hawaii Telescope Lensing Survey
2DCFHTLenS [72], referred to as WL. To avoid dealing with non-linear scales, we adopt a conservative cut
and exclude < 300 from the measurements of the correlation function , which corresponds to k < 0.1 h/Mpc
scales.

B.

Constraints on f (R)

The Hu-Sawicki f (R) model has two parameters, fR0


and n. In what follows, we fix n = 1 because that is a
common choice in the literature, and also because the
two parameters are highly correlated and the current
data cannot simultaneously constrain both. We chose a
flat prior on log10 fR0 within the [7, 0] range. We have
checked that changing the range of the flat prior does not

7
P
P
f (R)+AL , fixed
m
f (R)+AL , varying
m
P
fR0
AL
fR0
AL
m
+0.07
(0.12)
+0.10
(0.16)
5
4
3 (8) 10
1.080.05 (0.13) 0.4 (1.0) 10
1.110.06 (0.15) 0.30 (0.38)
TABLE II. 68% (95%) CL bounds on fR0 , AL and

m using all the data sets: PLC+BAO+CMBlens+MPK+WL

1.0

1
2

log10fR0

P/Pmax

0.8
0.6
0.4

4
5
6

0.2
0.0 -7
10

7
0.0

10-6
PLC+BAO

10-5

10-4

fR0

10-3

+CMBLens

10-2

10-1

100

+CMBLens+MPK+WL

FIG. 2. The marginalized posterior distribution for the fR0


parameter in the Hu-Sawicki model (n = 1) for different combinations of datasets. The solid lines show
P the PDF in case of
massive neutrinos with a fixed mass
m = 0.06 eV, while
the dashed lines show the PDF for the case when the neutrinoP
mass was varying. Due to the degeneracy between fR0
and
m , we see that the constraint on fR0 become weaker
when the neutrino mass is varied. The datasets are labeled according to the notation introduced in Sec. III A. The symbol
+ means that we add data on top of the PLC+BAO dataset.
For example, +lensing means PLC+BAO+lensing.

affect our results.


Fig. 2 shows constraints on fR0 for different combinations of datasets described in Sec. III A, after marginalizing over all the other cosmological parameters. We considered
P the case in which the total neutrino mass is fixed
at
m = 0.06 eV (solid
P lines), and the case where it
can vary within 0
m 1 eV (dashed lines). The
results from Fig. 2 are summarized in Table I.
We can see that the combination of PLC and
BAO datasets (blue lines) only weakly constrains the
model. Modified gravity affects the CMB temperature
anisotropy spectrum in two ways: it affects the low-`
power spectrum through the ISW effect and enhances the
damping at high-` due to the enhancement in clustering
and, as a consequence, the lensing potential. Thus, the
observed lack of power at low-` multipoles and the apparent preference of enhanced lensing in CMB TT, when
compared to the CDM prediction, can be reconciled
by a non-zero fR0 . This is the reason for the peak in

0.1

PLC+BAO

0.2

0.3

0.4

m (eV)

+CMBLens

0.5

0.6

0.7

+CMBLens+MPK+WL

P
m in the Hu-Sawicki
FIG. 3. Joint contours for fR0 and
model (n = 1) after marginalizing over all other cosmological parameters. The darker and lighter shades correspond
respectively to the 68% C.L. and the 95% C.L. . Datasets are
described in text and also in the caption of Fig. 2

the PLC+BAO likelihood. Adding the CMBLens data


(red lines) tightens the constraint substantially. The enhancement of growth due to the extra scalar interaction
affects the lensing potential measured by Planck, which
is known to be in excellent agreement with the LCDM
prediction [64]. Thus, the weak preference for larger fR0
coming from PLC+BAO is overwhelmed by the stronger
CMBLens data that is consistent with fR0 = 0. The constraint becomes even tighter after adding the MPK and
WL datasets (green lines).
The dashed lines in Fig. 2 show the impact
of coP
varying the combined mass of neutrinos,
m , along
with fR0 . Massive neutrinos suppress the growth and can
partially compensate for the enhanced clustering in f (R),
slightly weakening the bounds on fR0 . The extent of the
degeneracy can be inferred from Fig. 3 which shows the
joint confidence contours for the two parameters. We see
that, although the constraint on fR0 becomes
P tighter as
we add the LSS data, the constraint on
m remains
roughly the same. This is because we are restricting our
analysis to linear scales, while the effect of massive neutrinos becomes more relevant on smaller scales and, hence,
causes only a small degradation of fR0 constraints.
Up to this point, we kept the unphysical lensing amplitude parameter AL fixed at its expected value of 1.
However, one may wonder if the discrepancy in AL observed in the CDM model also persists in f (R), and

PLC+BAO
+CMBLens+MPK+WL

0.8

P/Pmax

log10fR0

4
5
6
7

fixed m
X
varying m

1.0

0.6
0.4
0.2

0.6

0.8

1.0

AL

1.2

1.4

0.0 -6
10

FIG. 4. Joint contours for fR0 and AL in the Hu-Sawicki


model. The darker and lighter shades corresponf respectively
to the 68% C.L. and the 95% C.L. Using PLC+BAO data
sets only it is possible to detect high values of fR0 that can
cure the tension in lensing amplitude AL . However such high
values are ruled out once we add lensing and LSS data sets.

what effect co-varying AL has on the bounds on fR0 .


The results for two different combinations of data are
shown in Fig. 4. Although it seems that, in the case
of PLC+BAO, the lensing amplitude tension has been
reconciled, we argue that this is not due to a genuine
signal of modified gravity. As discussed previously, the
PLC+BAO data yields a peak in the likelihood of fR0
because the preference for enhanced lensing and the lack
of power at low ` in C`T T can be reconciled with a nonzero fR0 . The enhanced lensing appears to cure the AL
problem and this is depicted in Fig. 4, where we see that
there is a strong degeneracy between AL and fR0 for large
values of the latter (blue contours). However such large
values of fR0 are ruled out once we add the datasets that
probe clustrering (green contours). Still, the value of AL
when co-fit with fR0 is in better disagreement with the
prediction. For the combination of all data we find
AL = 1.11+0.20
0.14

68 % C.L., all datasets.

10-5

10-4

10-3

10-2

10-1

100

FIG. 5. The marginalized posterior distribution for ? in Symmetron model


P with ? = 1 and a? = 0.25 considering neutrinos with
mP
= 0.06 eV (red solid line) and marginalizing
over a varying
m (blue dashed line). The data sets used
in this analysis are PLC+BAO+lensing+MPK+WL as described in section III A.

solid
P line) as well as after marginalizing over a varying
m (blue dashed line). We find an upper bound of
? < 1.5 103 at 95 % C.L, which corresponds to a
Compton wavelength of a few Mpc. Our bounds are
summarized in Table III.
As mentioned above, current data is unable to simultaneously constrain all the model parameters because they
are highly correlated. We also note that one cannot derive meaningful constraints for smaller values of coupling
constant ? as the modification of growth is relatively
small for the scales and redshifts currently probed. Further, since a? sets the onset of modified growth, we would
see tighter constraints on ? for smaller a? values. Nevertheless, as we will show in Sec. IV, future surveys with
larger sky and deeper redshift coverage will be able to
constrain ? along with the other two parameters.

(33)
D.

Constraints on the Dilaton model

The results of the analysis with varying AL are summarized in Table II.
C.

Constraints on the Symmetron model

In this Subsection we derive constraints on the inverse


mass parameter, ? , defined in Eq. (31), which represents
the Compton wavelength of the scalar interaction. We fix
the other two Symmetron parameters, taking a? = 0.25
and ? = 1, since current data is unable to constrain
them simultaneously with ? .
Fig. 5 shows the posterior probability
distribution for
P
the ? parameter with a fixed
m = 0.06 eV (red

P
P
Fixed
m
Varying
m
P
?
?
m
Symmetron
0.8 (1.5) 103 0.9 (1.8) 103 0.16 (0.27)
P
0
0
m
Dilaton
2.1 (3) 103
2.3 (3) 103 0.15 (0.25)
TABLE III. Summary of the 95% CL upper limits of the MG
parameters and the sum of neutrino masses (in unit of eV)
derived from current observations described in Sec. III A.

Analogously to the Symmetron model, we constrain

9
X

fixed m
X
varying m

1.0

P/Pmax

0.8
0.6
0.4
0.2
0.0 -5
10

10-4

10-3

10-2

10-1

FIG. 6. Marginalized posterior distribution for 0 in the Dilatons model with 0 = 5. The datasets used in the analysis are PLC+BAO+lensing+MPK+WL as described in section III A. The red solid line shows
the case with massive neuP
trinos with a fixed mass of
m = 0.06eV, while the blue
solid
P lines shows the PDF after marginalizing over a varying
m .

the inverse mass parameter 0 defined by Eq. (25), and fix


0 to a constant. Fig. 6 shows the posterior distribution
for 0 with the current value of the coupling parameter
fixed at 0 = 5. We find an upper bound of 0 < 3
103 (95 % C.L.). As for symmetrons, the sensitivity to
the coupling is weak due to the lack of data on linear
scales. However, as we will see in Sec. IV, constraints will
improve significantly with future surveys. Our results for
the Dilaton model are summarized in Table III

IV.

order unity coupling , to see how well future datasets


can constrain an evolving mass parameter.

A.

The data assumed in the forecast

The data we consider in our forecast include CMB


temperature anisotropy (T) and polarization (E) power
spectra with characteristics of the Planck survey, weak
lensing shear (WL) and galaxy number count (GC) from
an LSST-like survey [50], with the survey parameters
adopted from [73], and their cross-correlations. In some
cases, we compare this to constraints expected from the
Dark Energy Survey (DES) [74].
Theoretical power spectra are calculated assuming the
LSST (DES) GC data is partitioned into 10 (4) tomographic redshift bins, while the WL shear field is split
into 6 (4) tomographic redshift bins. In addition, we assume a flat FRW geometry and vary h, c h2 , b h2 , ,
ns , w and As , together with the modified gravity parameters. The fiducial values of the cosmological parameters
are taken to be the Planck 2015 best fit results. To calculate the WL and GC auto- and cross-correlation spectra
in our scalar-tensor models, we have applied the MGCAMB patch to CAMBSources [75]. The details of the
implementation are described in [25, 35].

B.

Fisher analysis

For a given model, one can calculate the Fisher matrix


[76] to determine how well future surveys can constrain
its parameters. The inverse of the Fisher matrix provides
a lower bound on the covariance matrix of the model
parameters via the Cramer-Rao inequality, C F1 .
For zero-mean Gaussian-distributed observables, such as
the angular correlations C`XY , the Fisher matrix is given
by

FORECASTS

Constraints on scalar gravitational interactions derived


in the previous Section, using current information available on linear scales, are relatively weak when compared
to bounds available from astrophysical tests. With improved redshift resolution, depth and sky coverage that
future surveys will provide, the number of modes in the
linear regime will dramatically increase. Thus, it is interesting to know if future constraints from linear scales can
become compatible with astrophysical bounds.
In what follows, we perform a series of Fisher forecasts
for the model parameters described in the previous Section, using, where possible, the current bounds on model
parameters as fiducial values in the forecast. Where there
was no upper bound, we use fiducial values motivated
by a combination of theoretical considerations and existing constraints from non-linear scales. We also perform a
principal component analysis (PCA) of m(a) for a fixed

Fab = fsky

`X
max
`=`min

2` + 1
Tr
2

C` 1 C` 1
C
C
pa ` pb `


, (34)

` is the
where pa is the ath parameter of our model and C
observed covariance matrix with elements C`XY that
include contributions from noise:
C`XY = C`XY + N`XY .

(35)

Eq. (34) assumes that all fields X(


n) are measured over
contiguous regions covering a fraction fsky of the sky.
The value of the lowest multipole can be approximately
inferred from `min /(2fsky ). The noise matrix N`XY
includes the statistical noise as well as the expected systematic errors. We refer the reader to [25, 35] for the
details of the Fisher matrix calculations for the individual experiments considered in our analysis.

10

1E-4

GC
+WL
Full

6E-5

fR0

fR0

8E-5

LSST
DES
LSST, fixed n

6E-5

4E-5

4E-5
2E-5

2E-5
1.0

1.5

2.0

FIG. 7. Expected 1 bounds on the parameters of the HuSawicki model. The assumed fiducial model is marked with
a star. The importance of using theFull set of observables
(WL, GC and their cross-correlation) is clearly demonstrated.
The Planck CMB data is included in all cases and is important
for constraining the standard cosmological parameters.

C.

The Symmetron forecast

Fig. 9 shows the bounds on the parameters of the Symmetron parameters expected from LSST+. As a fiducial
model, we assume ? = 1 and a mass scale of ? = 103 ,
which corresponds to a range of a few Mpc. Current data
is unable to constrain ? if a? = 0.5 or larger. For this
reason, the bound on ? in Sec. III was derived for a

0.05

0.10
m [eV]

0.15

FIG. 8. Comparison of the uncertainties expected from


LSST+ vs those from DES+ for the fR0 paramater of the
n = 1 Hu-Sawicki model and the total mass of neutrinos. The
assumed fiducial model is marked with a star. The effect of
fixing n, as opposed to marginalizing over it, is also shown.

The f(R) forecast

In Fig. 7 we show 1 constraints on parameters of the


Hu-Sawicki f (R) model, as expected from LSST+ (LSST
WL + LSST GC + Planck CMB). Recall that current
data is unable to constrain fR0 unless one assumes a fixed
value for n, since the two parameters are highly degenerate. Thus, the forecast in Fig. 7 depends strongly on the
assumed fiducial value, indicated with a ? on the plot.
What we see is that for n 1 or smaller, future data will
be able to constrain both parameters simultaneously.
Fig. 7 also shows the importance of including the crosscorrelation between WL and GC. The information from
GC alone is largely diluted by the unknown galaxy bias.
Weak lensing, while not sensitive to the bias, is plagued
by degeneracies coming from projection effects. Combining them helps determine the bias and break the degeneracies coming from projections.
Fig. 8 compares joint 1 constraints
on fR0 and the
P
combined mass of neutrinos,
m , as expected from
LSST+ vs those expected from DES+. We see that
LSST+ can reduce uncertainties in both parameters by
a factor of 3. The plot shows the effect of marginalizing
over n, however the outcome depends on the assumed
fiducial value of n (which is n = 1).

D.

2.5

a =0.5
a =0.25

1.05

0.5

1.00
0.95

0.504

6E-4

8E-4

1E-3

1.2E-3

1.4E-3

6E-4

8E-4

1E-3

1.2E-3

1.4E-3

0.502

0.500
0.498
0.496

FIG. 9. Expected 1 constrains from LSST+ on the parameters of the Symmetron model. The assumed fiducial models
are marked with a star. Unlike current data, LSST+ can simultaneously constrain ? and a? to a few percent level, and
will improve the current bounds on ? . See Table IV for a
quantitative comparison.

fixed a? = 0.25. We perform a forecast using two different fiducial values: a? = 0.25 and 0.5. In the former
case, LSST+ clearly improves on the bound in Sec. III,
even after marginalizing over a? and ? . It will also be
able to provide a non-trivial bound on ? for a? = 0.5,
which is the value assumed in much of the previous literature. The current and expected bounds are summarized
in Table IV.
It is interesting to examine the possible degeneracy be-

11

2E-03
1E-03

1.1

1E-03

1.2

a Marg.
a Fixed
a , Fixed

1E-03

0.9

8E-04

0.8

0.05

0.1

0.15 0.2
m[eV]

FIG. 10. Expected 1 bounds on the ? parameter


of the
P
Symmetron model and the mass of neutrinos,
m . The assumed fiducial model is marked with a star. Fixing the other
MG parameters in this model, as opposed to marginalizing
over them, does not change the degree of degeneracy, neither
it improves the constraints.

tween the Symmetron parameters and the total mass of


neutrinos. Fig. 10 shows the joint uncertainties in ? and
P
m expected from LSST+ assuming P
a fiducial model
with ? = 1, ? = 103 , a? = 0.5 and
m = 0.06eV.
It is clear from the figure that there is practically no degeneracy between ? and m which is because they affect
the growth on different scales. Fixing the other MG parameters in this model, as opposed to marginalizing over
them, does not change the degree of degeneracy, neither
it improves the constraints.

E.

0.08

0.25

The Dilaton forecast

Fig. 11 shows expected bounds on the Dilaton model


parameters, with 0 = 1 and 0 = 103 as the fiducial
values. Similar to the Symmetron case, we find that an
LSST-like survey can constrain the inverse mass parameter 0 to a percent level accuracy which is a significant
improvement over current constraints. Constraints on the
coupling constant 0 , however, are not as tight as those
on ? in the Symmetron case. This is due to a lesser impact of the Dilaton on the linear matter power spectrum.
One can see from Fig. 1 that for the chosen fiducial values, P (k) would deviate from the LCDM prediction far
less in the Dilaton case compared to the Symmetron. The
bottom panel in Fig. 11 shows the expected joint constraints on the neutrino masses, which are tighter than
those for the Symmetron. Again, this is because Dilatons
have a much lesser impact on the growth on linear scales.

The Generalized Chameleon model

Forecasts for the Generalized Chameleon provide a


general estimate of how well one could constrain the

m[eV]

6E-04
0

F.

1.0

0.07
0.06
0.05
9.6E-4

9.8E-4

1E-3

1.02E-3

1.04E-3

FIG. 11. 1 bounds on the neutrino masses and parameters of


the Dilaton model expected from LSST+. The fiducial values
are marked with stars.

scalar gravitational interactions with a next generation


WL survey such as LSST. In Fig. 12 we show forecasted uncertainties on the parameters of the generalized
Chameleon model for two fiducial values of r, assuming
that the coupling is constant (s = 0). For a slower evolution with time (r = 1), the scalaron mass decreases slower
and modification to growth extends back to larger redshifts, leading to significantly tighter constraints. Thus,
while LSST+ can constrain the coupling, the mass and
the time-variation of the scalaron mass simultaneously,
the strength of the bounds depends strongly on the assumed fiducial model.

G.

Principal Component Analysis of m(a)

In addition to considering particular functional forms


of (a) and m(a) as motivated by the scalar-tensor models mentioned in preceding sections, it is also interesting
to treat the coupling and the mass as two general functions and ask what features of these two functions can
be constrained by the future data. In principle, one could
discretize the functions (a) and m(a) into N bins in a
and treat the bins values as free parameters. However,
we find that even future data will not be able to simultaneously constrain m(a) and (a) in a completely modelindependent way, since the two parameters are largely
degenerate in their effect on the observables on linear
scales, as they appear together in (a, k) (see Eq. (13)).
For this reason, we fix at a constant value of order unity
and discretize m(a) into bins with m(ai ), i = 1, ..., N .
As with earlier forecasts, we can calculate the Fisher

12

9.6E-4

1E-3

1.04E-3 0.97

1.03

0.506
0.504

0.5
0.497

0.494
0.8
0.7
0.6
0.5
0.4
0.3
0.2

5E-4 1.E-3 1.5E-3

FIG. 12. Expected 1 constrains on the the Generalized


Chameleon parameters for a fiducial model with r = 3 (top)
and r = 1 (bottom) as a fiducial model. In each case, the
value of r is varied and marginalized over.

matrix, and invert it to find the covariance matrix,


Cij h(pi pi )(pj pj )i ,

(36)

where pi are the fiducial values, and parameters include


the bins m(ai ), as well as the rest of cosmological parameters. We then isolate the N N block of the matrix, C m
corresponding to the covariance of m(ai ) after marginalization over other parameters. Since the individual bins
of m(ai ) bins are highly correlated, the covariance matrix for these parameters will be non-diagonal, and the
value of m in any particular bin will be practically unconstrained. The Principal Component Analysis (PCA)
[27, 35, 36, 77, 78] is a way to decorrelate the parameters and find their linear combinations that are best constrained by data. Namely, we solve an eigenvalue problem
to find a matrix W m that diagonalizes C m :
C m = (W m )T W m ; ij = i ij ,

(37)

where Wijm ei (aj ) are the eigenvectors (or eigenmodes)


and i s are the eigenvalues. In the limit of large N , one
can write an arbitrary m(a) as an expansion into ei (a):

It is customary to order the eigenmodes from the best


constrained to the worst. Then the ith eigenmode is referred to as the ith principal component (PC). Typically,
one finds that only the first few modes are well constrained by the data, while most of them are practically
unconstrained.
For our forecast, we partition m(a) into 11 bins, with
10 of them evenly spaced in redshift within z [0, 3],
and the 11th bin ranging from z = 3 to z = 30. The
last bin can be taken to be wide because the observables
we work with are weakly sensitive to modifications at
high redshifts. In what follows, we marginalize over the
11th bin, since it is largely degenerate with some of the
cosmological parameters, most prominently with m . We
take the fiducial model to be = 0.4 and m(ai ) = m0
for all i, with m0 = H0 /c and = 103 , corresponding
to m0 = 0.2 h/Mpc.
The left panel in Fig. 13 shows the the forecasted uncertainties in the measurement of the eigenvectors for two
cases: when is fixed, and when is marginalized over.
In both cases, we marginalize over all cosmological parameters and the 11th m-bin. The right panel in Fig. 13
presents the first four best constrained eigenvectors of
m(a) after marginalizing over . One can interpret the
best constrained mode (PC1) as that corresponding to
a weighted average value of m(a). The second best constrained mode (PC2) has a single node and corresponds
to the difference between the high-z and low-z values of
m(a). The third best mode (PC3) has 3 nodes, PC4 has
4 nodes, and so on.
The eigenvalue plot demonstrates that marginalizing
over affects the first eigenmode of m(a), but not the
others. This is because the main effect of a constant
is an overall rescaling of the strength of the 5th force. It
is largely degenerate with the average value of m(a), but
has no impact on the detectability of time-variation of
m(a). After marginalizing over , LSST+ can measure
one mass parameters (the average m(a)) with an accuracy that is better than 0.01 h/Mpc, or about 5% of the
fiducial m0 , and another 3 parameters, describing more
rapid evolution of the mass with time, with accuracy better than 0.02 h/Mpc, or 10% of the fiducial value.
The extrema of the eigenmodes indicate the sweet
spots in redshift, or epochs at which variations in m(a)
are best constrained with LSST+. It is evident from the
right panel in Fig. 13, for instance the shape of PC2, that
LSST+ is more sensitive to time-variations at z > 1.5.
This is because at higher redshifts there is a larger number of Fourier modes that are still in the linear regime.

V.

m(a) m(a)

N
X

i ei (a)

(38)

i=1

in which case i can be interpreted as the variance of i ,


i = 2 i .

(39)

DISCUSSION AND CONCLUSIONS

Modifications of gravity on cosmological scales can potentially explain the origin of cosmic acceleration. The
Generalized Brans-Dicke theory, in which there is an additional scalar degree of freedom that mediates a fifth
force, is one of the viable MG models that are able to fit

13

101
10-1

ei (z)

i (z)

100

Marg.
Fixed

10-2
10-3

6
PC#

10

0.8
0.6
0.4
0.2
0.0
0.2
0.4

PC #1
PC #2
PC #3
PC #4

0.5 1.0 1.5 2.0 2.5 3.0

FIG. 13. Left: The uncertainties (square roots of eigenvalues) associated with the eigenmodes of m(a) for the case when the
coupling is fixed at = 0.4 (solid line), and when it is marginalized over (dashed line). Right: The first four best constrained
eigenmodes of m(a) after marginalizing over .

Hu-Sawicki f (R)
Symmetron
Dilaton
Parameters
f R0
n=1
?
? = 1 a? = 0.25 (0.5)
0
0 = 1 (5)
Current 1 4 105
103
unconstrained (2.3 103 )
LSST+ 1 6 106 0.3
2 (2.9) 104 0.05 (0.07) 0.001 (0.005)
2.7 105
2.3 101
TABLE IV. The current 68 % C.L. uncertainties and those expected from LSST+. The blocks with mean the parameter
was fixed at its fiducial value. The values in parenthesis indicate those obtained for an alternative fiducial value.

observations after the required tuning of model parameters.


In this work, we have investigated the observational
constraints on three MG models within the general
framework of the GBD theory, namely, the f (R), the
Symmetron and the Dilaton models, using latest observations of CMB, BAO, weak lensing and galaxy clustering.
In all cases, we used observables on linear scales to avoid
the complexities of the modelling of nonlinearities and
redshift-space distortions.
We find that the CDM model is consistent with all
observations. Specifically, we find the constraint on fR0 ,
the model parameter in the Hu-Sawicki f (R) model, to
be fR0 < 8 105 (95% CL) when the sum of neutrino
masses is fixed to be 0.06 eV. Since both massive neutrinos and MG models studied in this paper can alter
the structure growth in a scale-dependent way, a degeneracy is expected. Therefore we perform another analysis with the neutrino mass varying, and we find that
the constraint is diluted to fR0 < 1.0 104 (95% CL).
For the Symmetron model, the 95% CL upper limit is
? < 1.8 103 with ? and a? fixed at 1 and 0.25, respectively. For the Dilaton model, we find 0 < 3 103
at 95% CL when 0 = 5. Tables II and III summarize the
current bounds.
We have also performed a forecast for ongoing and upcoming imaging surveys including DES and LSST, and
present the results in Sec. IV. A comparison between the
current and future constraints on model parameters is
shown in Table IV. As one can see, the improvement

is significant and, despite the high level of degeneracy,


more than one parameter can be constrained simultaneously. In the Hu-Sawicki model, the upper limit of fR0
is reduced by a factor of 6.7 and n can be constrained
with 25% accuracy for n = 1. For the Dilaton model,
current data is unable to constrain 0 if 0 = 1. However, we find that LSST+ can simultaneously constrain
0 at few 105 and measure 0 1 with 20%
accuracy. In the Symmetron model, the constraint on ?
is improved by a factor of 3, while simultaneously constraining a? and ? within a few percent. This is compatible with current bounds derived from astrophysical tests,
such as the cluster profile [79], galactic dynamics and so
on, which requires high-resolution N-body or hydrodynamical simulations [80, 81] of the MG models. Additionally, to demonstrate the capabilities of an LSST-like
survey, we have presented constraints on the Generalized
Chameleon model in Fig. 12.
Given the power of future surveys, a modelindependent analysis will become possible. In this work,
we performed a PCA study of m(a), to forecast the maximum number of parameters of the scalaron mass function
that can be well determined. We find that an LSST-like
survey will be able to measure the average mass parameter with an accuracy of 0.01 h/Mpc and another 3 parameters quantifying the time-variation of m(a) with an
accuracy that is better than 0.02 h/Mpc. Finally, we note
that future spectroscopic and HI surveys, such as eBOSS
and SKA [82, 83], will also provide powerful constraints
on MG parameters that will be highly complementary to

14
those from a photometric survey like LSST [84].

ACKNOWLEDGMENTS

The research of AH, AP, AZ and LP is supported by


the Natural Sciences and Engineering Research Council
of Canada (NSERC). AP was sponsored in part by the
Robert Frazier Memorial Fellowship. AZ is supported in
part by the Bert Henry Memorial Entrance Scholarship

[1]
[2]
[3]
[4]
[5]
[6]
[7]

[8]
[9]
[10]
[11]

[12]
[13]

[14]

[15]
[16]
[17]
[18]

[19]
[20]
[21]
[22]
[23]
[24]

A. G. Riess, et al., Astron. J. 116, 1009 (1998).


S. Perlmutter, et al., Astrophys. J. 517, 565 (1999).
C. P. Burgess, arXiv:1309.4133 [hep-th].
S. Weinberg, Rev. Mod. Phys. 61, 1 (1989).
C. Brans and R. H. Dicke, Phys. Rev. 124, 925 (1961).
J. D. Bekenstein, Phys. Rev. D 48, 3641 (1993) [grqc/9211017].
M. Zumalacarregui, T. S. Koivisto, D. F. Mota
and P. Ruiz-Lapuente, JCAP 1005, 038 (2010)
[arXiv:1004.2684 [astro-ph.CO]].
H. Y. Ip, J. Sakstein and F. Schmidt, arXiv:1507.00568
[gr-qc].
J. Sakstein and S. Verner, arXiv:1509.05679 [gr-qc].
J. Khoury and A. Weltman, Phys. Rev. D 69 (2004)
044026 [astro-ph/0309411].
P. Brax, C. van de Bruck, A. C. Davis, J. Khoury and
A. Weltman, Phys. Rev. D 70 (2004) 123518 [astroph/0408415].
W. Hu and I. Sawicki, Phys. Rev. D 76 (2007) 064004
[arXiv:0705.1158 [astro-ph]].
P. Brax, C. van de Bruck, A. C. Davis and D. J. Shaw,
Phys. Rev. D 78 (2008) 104021 [arXiv:0806.3415 [astroph]].
P. Brax, C. van de Bruck, A. C. Davis and D. Shaw,
Phys. Rev. D 82 (2010) 063519 [arXiv:1005.3735 [astroph.CO]].
K. Hinterbichler and J. Khoury, Phys. Rev. Lett. 104
(2010) 231301 [arXiv:1001.4525 [hep-th]].
T. Damour and A. M. Polyakov, Nucl. Phys. B 423
(1994) 532 [hep-th/9401069].
P. Brax, A. -C. Davis and B. Li, Phys. Lett. B 715, 38
(2012) [arXiv:1111.6613 [astro-ph.CO]].
P. Brax, A. -C. Davis, B. Li and H. A. Winther,
Phys. Rev. D 86, 044015 (2012) [arXiv:1203.4812 [astroph.CO]].
M. Ishak, A. Upadhye and D. N. Spergel, Phys. Rev. D
74 (2006) 043513 [astro-ph/0507184].
E. V. Linder, Phys. Rev. D 72 (2005) 043529 [astroph/0507263].
M. Kunz and D. Sapone, Phys. Rev. Lett. 98 (2007)
121301 [astro-ph/0612452].
P. Zhang, M. Liguori, R. Bean and S. Dodelson, Phys.
Rev. Lett. 99 (2007) 141302 [arXiv:0704.1932 [astro-ph]].
L. Amendola, M. Kunz and D. Sapone, JCAP 0804
(2008) 013 [arXiv:0704.2421 [astro-ph]].
Y. S. Song and K. Koyama, JCAP 0901 (2009) 048
[arXiv:0802.3897 [astro-ph]].

at SFU. GBZ is supported by the Strategic Priority Research Program The Emergence of Cosmological Structures of the Chinese Academy of Sciences Grant No.
XDB09000000. This research was enabled in part by support provided by WestGrid [85] and Compute Canada
[86]. P.B. acknowledges partial support from the European Union FP7 ITN INVISIBLES (Marie Curie Actions,
PITN- GA-2011- 289442) and from the Agence Nationale
de la Recherche under contract ANR 2010 BLANC 0413
01. ACD acknowledges partial support from STFC under
grants ST/L000385/1 and ST/L000636/1.

[25] G. B. Zhao, L. Pogosian, A. Silvestri and J. Zylberberg,


Phys. Rev. D 79, 083513 (2009) [arXiv:0809.3791 [astroph]].
[26] Y. S. Song and O. Dore, JCAP 0903 (2009) 025
[arXiv:0812.0002 [astro-ph]].
[27] G. B. Zhao, L. Pogosian, A. Silvestri and J. Zylberberg,
Phys. Rev. Lett. 103, 241301 (2009) [arXiv:0905.1326
[astro-ph.CO]].
[28] S. F. Daniel, E. V. Linder, T. L. Smith, R. R. Caldwell,
A. Cooray, A. Leauthaud and L. Lombriser, Phys. Rev.
D 81 (2010) 123508 [arXiv:1002.1962 [astro-ph.CO]].
[29] G. B. Zhao, T. Giannantonio, L. Pogosian, A. Silvestri,
D. J. Bacon, K. Koyama, R. C. Nichol and Y. S. Song,
Phys. Rev. D 81 (2010) 103510 [arXiv:1003.0001 [astroph.CO]].
[30] L. Pogosian, A. Silvestri, K. Koyama and G. -B. Zhao,
Phys. Rev. D 81, 104023 (2010) [arXiv:1002.2382 [astroph.CO]].
[31] C. Shapiro, S. Dodelson, B. Hoyle, L. Samushia
and B. Flaugher, Phys. Rev. D 82 (2010) 043520
[arXiv:1004.4810 [astro-ph.CO]].
[32] S. F. Daniel and E. V. Linder, Phys. Rev. D 82 (2010)
103523 [arXiv:1008.0397 [astro-ph.CO]].
[33] Y. S. Song, G. B. Zhao, D. Bacon, K. Koyama,
R. C. Nichol and L. Pogosian, Phys. Rev. D 84 (2011)
083523 [arXiv:1011.2106 [astro-ph.CO]].
[34] A. Hojjati, L. Pogosian and G. -B. Zhao, JCAP 1108,
005 (2011) [arXiv:1106.4543 [astro-ph.CO]].
[35] A. Hojjati, G. -B. Zhao, L. Pogosian, A. Silvestri, R. Crittenden and K. Koyama, Phys. Rev. D 85, 043508 (2012)
[arXiv:1111.3960 [astro-ph.CO]].
[36] A. Hojjati, JCAP 1301, 009 (2013) [arXiv:1210.3903
[astro-ph.CO]].
[37] A. Hojjati, L. Pogosian, A. Silvestri and S. Talbot,
Phys. Rev. D 86, 123503 (2012) [arXiv:1210.6880 [astroph.CO]].
[38] S. F. Daniel and E. V. Linder, JCAP 1302 (2013) 007
[arXiv:1212.0009 [astro-ph.CO]].
[39] F. Simpson et al., Mon. Not. Roy. Astron. Soc. 429
(2013) 2249 [arXiv:1212.3339 [astro-ph.CO]].
[40] A. Silvestri, L. Pogosian and R. V. Buniy, Phys. Rev.
D 87, no. 10, 104015 (2013) [arXiv:1302.1193 [astroph.CO]].
[41] A. Hojjati, L. Pogosian, A. Silvestri and G. B. Zhao,
Phys. Rev. D 89, no. 8, 083505 (2014) [arXiv:1312.5309
[astro-ph.CO]].
[42] B. Hu, M. Liguori, N. Bartolo and S. Matarrese, Phys.
Rev. D 88 (2013) 12, 123514 [arXiv:1307.5276 [astro-

15
ph.CO]].
[43] J. Dossett, B. Hu and D. Parkinson, JCAP 1403 (2014)
046 [arXiv:1401.3980 [astro-ph.CO]].
[44] P. A. R. Ade et al. [Planck Collaboration], Planck
2015 results. XIV. Dark energy and modified gravity,
arXiv:1502.01590 [astro-ph.CO].
[45] B. Jain, V. Vikram and J. Sakstein, Astrophys. J. 779
(2013) 39 [arXiv:1204.6044 [astro-ph.CO]].
[46] V. Vikram, A. Cabr, B. Jain and J. T. VanderPlas, JCAP
1308 (2013) 020 [arXiv:1303.0295 [astro-ph.CO]].
[47] J. Sakstein, B. Jain and V. Vikram, Int. J. Mod. Phys. D
23 (2014) 12, 1442002 [arXiv:1409.3708 [astro-ph.CO]].
[48] J. Bel, P. Brax, C. Marinoni and P. Valageas, Phys. Rev.
D 91 (2015) 10, 103503 [arXiv:1406.3347 [astro-ph.CO]].
[49] P. Brax, Phys. Rev. D 90 (2014) 2, 023505
[arXiv:1310.2562 [astro-ph.CO]].
[50] http://www.lsst.org
[51] http://www.euclid-ec.org
[52] C. Schimd, J. P. Uzan and A. Riazuelo, Phys. Rev. D 71,
083512 (2005) [astro-ph/0412120].
[53] http://www.camb.info
[54] Http://www.sfu.ca/~aha25/MGCAMB.html
[55] P. Brax and P. Valageas, Phys. Rev. D 88, no. 2, 023527
(2013) [arXiv:1305.5647 [astro-ph.CO]].
[56] C. Llinares and D. Mota, Phys. Rev. Lett. 110, no. 16,
161101 (2013) [arXiv:1302.1774 [astro-ph.CO]].
[57] S. Capozziello, S. Carloni and A. Troisi, Recent Res. Dev.
Astron. Astrophys. 1, 625 (2003) [astro-ph/0303041].
[58] S. M. Carroll, V. Duvvuri, M. Trodden and M. S. Turner,
Phys. Rev. D 70, 043528 (2004) [astro-ph/0306438].
[59] S. A. Appleby and R. A. Battye, Phys. Lett. B 654, 7
(2007) [arXiv:0705.3199 [astro-ph]].
[60] A. A. Starobinsky, JETP Lett. 86, 157 (2007)
[arXiv:0706.2041 [astro-ph]].
[61] P. Brax, A. C. Davis, B. Li, H. A. Winther and
G. B. Zhao, JCAP 1304 (2013) 029 [arXiv:1303.0007
[astro-ph.CO]].
[62] http://cosmologist.info/cosmomc/
[63] P. A. R. Ade et al. [Planck Collaboration], Planck
2015
results.
XIII.
Cosmological
parameters,
arXiv:1502.01589 [astro-ph.CO].
[64] P. A. R. Ade et al. [Planck Collaboration],
arXiv:1502.01591 [astro-ph.CO].
[65] P. A. R. Ade et al. [Planck Collaboration], Astron. Astrophys. 571, A16 (2014) [arXiv:1303.5076 [astro-ph.CO]].

[66] G. E. Addison, Y. Huang, D. J. Watts, C. L. Bennett, M. Halpern, G. Hinshaw and J. L. Weiland,


arXiv:1511.00055 [astro-ph.CO].
[67] F. Beutler, C. Blake, M. Colless, D. H. Jones, and
L. Staveley-Smith, L. Campbell, L. Q. Parker, W. Saunders and F. Watson. Mon. Not. Roy. Astron. Soc. 416,
3017 (2011) [arXiv:1106.3366 [astro-ph.CO]]
[68] A. J. Ross, L. Samushia, C. Howlett, W. J. Percival,
A. Burden and M. Manera, Mon. Not. Roy. Astron. Soc.
449, no. 1, 835 (2015) [arXiv:1409.3242 [astro-ph.CO]].
[69] L. Anderson et al. [BOSS Collaboration], Mon. Not.
Roy. Astron. Soc. 441, no. 1, 24 (2014) [arXiv:1312.4877
[astro-ph.CO]].
[70] M. Tegmark et al. [SDSS Collaboration], Phys. Rev. D
74, 123507 (2006) [astro-ph/0608632].
[71] G. B. Zhao et al., Mon. Not. Roy. Astron. Soc. 436, 2038
(2013) [arXiv:1211.3741 [astro-ph.CO]].
[72] Kilbinger, M., Fu, L., Heymans, C., et al., MNRAS 430,
2200 (2013) [arXiv:1212.3338 [astro-ph.CO]]
[73] Z. Ivezic, J. A. Tyson, et al for the LSST Collaboration,
arXiv:0805.2366 [astro-ph].
[74] http://www.darkenergysurvey.org/
[75] http://www.camb.info/sources/
[76] M. Tegmark, A. Taylor and A. Heavens, Astrophys. J.
480, 22 (1997) [arXiv:astro-ph/9603021].
[77] D. Huterer and G. Starkman, Phys. Rev. Lett. 90, 031301
(2003) [astro-ph/0207517].
[78] R. G. Crittenden, L. Pogosian and G. B. Zhao, JCAP
0912, 025 (2009) [astro-ph/0510293].
[79] H. Wilcox et al., Mon. Not. Roy. Astron. Soc. 452, no.
2, 1171 (2015) [arXiv:1504.03937 [astro-ph.CO]].
[80] P. Brax, A. C. Davis, B. Li, H. A. Winther and
G. B. Zhao, JCAP 1210 (2012) 002 [arXiv:1206.3568
[astro-ph.CO]].
[81] P. Brax, A. C. Davis, B. Li, H. A. Winther and
G. B. Zhao, JCAP 1304, 029 (2013) [arXiv:1303.0007
[astro-ph.CO]].
[82] G. B. Zhao et al., arXiv:1510.08216 [astro-ph.CO].
[83] G. B. Zhao, D. Bacon, R. Maartens, M. Santos and
A. Raccanelli, arXiv:1501.03840 [astro-ph.CO].
[84] D. Bacon et al., PoS AASKA 14, 145 (2015)
[arXiv:1501.03977 [astro-ph.CO]].
[85] http://www.westgrid.ca
[86] http://www.computecanada.ca

You might also like