Dark Energy versus Modified Gravity

Martin Kunz∗ and Domenico Sapone†

Département de Physique Théorique, Université de Genève,
24 quai Ernest Ansermet, CH–1211 Genève 4, Switzerland
(Dated: December 17, 2006)
There is now strong observational evidence that the expansion of the universe is accelerating. The
standard explanation invokes an unknown “dark energy” component. But such scenarios are faced
with serious theoretical problems, which has led to increased interest in models where instead General
Relativity is modified in a way that leads to the observed accelerated expansion. The question then
arises whether the two scenarios can be distinguished. Here we show that this may not be so easy,
demonstrating explicitely that a generalised dark energy model can match the growth rate of the
DGP model and reproduce the 3+1 dimensional metric perturbations. Cosmological observations
are then unable to distinguish the two cases.
arXiv:astro-ph/0612452v1 17 Dec 2006

PACS numbers: 95.36.+x; 04.50.+h; 98.80.-k

INTRODUCTION conserved on the brane, ρm satisfies the usual conserva-

tion equation. Comparing this to the normal Friedmann
The observed accelerated expansion of the late-time equation with an additional dark energy component, we
universe, as evidenced by a host of cosmological data like see that we can move the crossover term to the right hand
type Ia supernovae (SN-Ia) [1], the cosmic microwave side and think of it as a dark energy contribution with
background radiation [2] and large scale structure [3] ρDE = 3H/(8πGrc ). Looking at the conservation equa-
came as a great surprise to cosmologists. Although it tion we find that it is solved if the effective dark energy
is straightforward to explain the effect within the frame- has an equation of state given by
work of Friedmann-Robertson-Walker cosmology by in- Ḣ
troducing a cosmological constant or a more general (dy- 1 + wDE = − . (2)
3H 2
namical) dark energy component, all such explanations
give rise to severe coincidence and fine-tuning problems. Consequently, it is impossible to rule out “dark energy”
An alternative approach postulates that General Rel- based on measurements of the cosmic expansion history
ativity is only accurate on small scales and has to be (e.g. SN-Ia data).
modified on cosmological distances. This in turn leads to Recently there have been claims that it is instead pos-
the observed late-time acceleration of the expansion of sible to use the growth rate of structures for this purpose
the universe [4, 5, 6, 7]. One of the best-studied exam- [6, 10, 11, 12] (but see also [13] for cautionary remarks).
ples is the Dvali-Gabadadze-Porrati (DGP) brane-world This is based on the observation that we can fix the equa-
model [8], in which the gravity leaks off the 4-dimensional tion of state parameter w of the dark energy from back-
Minkowski brane into the 5-dimensional bulk Minkowski ground data and then predict the evolution of the dark
space-time. On small scales the gravity is bound to the matter perturbations in a standard cosmological model
4-dimensional brane and the Newtonian gravity is recov- with dark energy. If the observed growth rate is different
ered to a good approximation. from the predictions, then general relativity with dark
One important question is whether such a scenario can energy would be ruled out.
be distinguished from one invoking an invisible dark en- However, in this paper we will show that this conclu-
ergy component. It is well known that any expansion sions makes additional, very strong assumptions about
history (as parametrised by the Hubble parameter H(t)) the nature of the dark energy, and that in general the
can be generated by choosing a suitable equation of state growth rate of structure is not sufficient to distinguish
for the dark energy (parameterised by the equation of between dark energy models and modifications of grav-
state parameter w = p/ρ of the dark energy). This can ity. We will show how the dark energy perturbations
for example be seen from Eq. (1) of [9]. Let us illustrate influence the dark matter and the metric perturbations,
this explicitely for the DGP model, for which the Hubble and provide an explicit example of a dark energy model
parameter evolves as which reproduces the 3+1 dimensional metric perturba-
tions of the DGP scenario.
H 8πG
H2 − = ρm (1)
rc 3
SETTING THE STAGE
where rc , the crossover scale, separates the 5D and the
4D regimes. It has to be of the order of 1/H0 in or- We start by discussing the fluid perturbations in stan-
der to generate late-time acceleration. Since matter is dard 3+1 dimensional cosmologies. The perturbations in
 2

the energy density are given by δ = δρ/ρ and to repre-

sent the fluid velocity we use V = ikj T0j /ρ. Working in
the Newtonian (longitudinal) gauge, the metric can be
written as

ds2 = − (1 + 2ψ) dt2 + a2 (1 − 2φ) dxi dxi (3)

with two scalar potentials φ and ψ describing the per-

turbations in the metric. Perturbations in cosmic fluids
evolve according to [14]
 
V 3 δp
δ ′ = 3(1 + w)φ′ − − − wδ (4)
Ha2 a ρ
k2
 
V δp
V ′ = −(1 − 3w) + + (1 + w)(ψ − σ) (5)
a Ha2 ρ
where a prime denotes a derivative with respect to the
scale factor a. The physical properties of the fluid are
given by the anisotropic stress σ and the pressure pertur-
bation δp (in general both can be functions of k). The
latter is often parametrised in terms of the rest-frame
sound speed c2s , FIG. 1: This figure shows how the growth of the matter
perturbations depends on the clustering properties of the
V dark energy. From the top downward the sound speed is
δp = c2s δρ + 3Ha(c2s − c2a )ρ (6) c2s = −2 × 10−4 (cyan dash-dotted line), c2s = −10−4 (ma-
k2
genta long dashed line), c2s = 0 (blue dotted line) and c2s = 1
where c2a = ṗ/ρ̇ is the adiabatic sound speed. Collision- (red dashed line). For comparison we also plot the growth
less cold dark matter has zero pressure (wm = 0), van- factor of the DGP model (black solid line).
ishing sound speed (c2s,m = 0) and no anisotropic stress
(σm = 0). For the dark energy all these quantities are a
priori unknown functions and have to be measured. For THE IMPORTANCE OF DARK ENERGY
the special case of dark energy due to a minimally cou- PERTURBATIONS
pled scalar field we have a variable w (corresponding to
the choice of the scalar field potential, and fixed by the We start by noticing that the growth factor is not
expansion history of the universe), c2s,DE = 1 and σ = 0 uniquely determined by the expansion history of the uni-
(see e.g. [15, 16]). verse (and hence wDE ). Although the main effect of the
The perturbations in different fluids are linked via the dark energy is to change H, leading to g < 1 at late times,
perturbations in the metric φ and ψ. Introducing the there is an additional link through the gravitational po-
comoving density perturbation ∆ ≡ δ + 3HaV /k 2 , their tential ψ. Different dark energy perturbations will lead
evolution in the standard cosmology is given by to a different evolution of ψ, which can modify the be-
X haviour of g. Conventionally one assumes that the dark
k 2 φ = −4πGa2 ρi ∆i (7)
energy perturbations are unimportant, e.g. [17]. This
i
2 2
X is a good assumption for scalar field dark energy where
k (φ − ψ) = 12πGa (1 + wi ) ρi σi (8) the high sound speed prevents clustering on basically all
i scales. However, a small sound speed c2s,DE ≈ 0 is not ex-
where the sum runs over matter and dark energy in our cluded. Indeed, it could even be negative, leading to very
case. rapid growth of the dark energy perturbations. It could
The quantity of interest to us is the growth factor also vary in time. We show in Fig. 1 how the growth fac-
g ≡ ∆m /a which parameterises the growth of structure tor of the dark matter changes in response to large dark
in the dark matter. The growth factor is normalised so energy perturbations [23].
that g = 1 for a ≪ 1 (using that ∆m ∝ a during matter What happens is that, as we decrease the sound speed,
domination and on sub-horizon scales). For definiteness the dark energy is able to cluster more and more. The in-
we fix k = 200/H0 for the numerical results. We assume creased dark energy perturbations lead to enhanced met-
that g is an observable quantity (even though of course ric perturbations. The dark matter in turn falls into the
large scale structure surveys observe luminous baryonic potential wells created by the dark energy, leading to an
matter, not dark matter, adding yet another layer of com- increase of the growth factor. Although clearly g is not
plications). uniquely determined by wDE , we notice that it always
 3

increases as we decrease c2s,DE (at least as long as the lin-

earised theory is applicable, see also [18]). Looking at the
evolution equations (4) and (5) for σ = 0 (⇔ φ = ψ) we
see that the response of the fluids to the metric perturba-
tions is governed by the sign of 1+w. Non-phantom dark
energy (as required to mimic the DGP expansion history)
clusters therefore in fundamentally the same way as the
dark matter and can only increase the growth of matter
relative to the case of negligible dark energy perturba-
tions (excluding highly fine-tuned initial conditions).
So although the dark energy perturbations can influ-
ence the growth factor of the dark matter, they only seem
capable of enhancing it. But Fig. 1 also shows the predic-
tion for the growth factor in the DGP model from [19],
and it is smaller than the one of a smooth dark energy
component. We therefore need to change something else
if we want to mimic DGP with dark energy. For this we
need to take a closer look at the DGP model.

ANISOTROPIC STRESS AND MODIFIED FIG. 2: In this figure we show how the anisotropic stress of
GRAVITY MODELS the dark energy affects the growth of the dark matter pertur-
bations. The red dashed line corresponds to scalar field dark
An important aspect of DGP and other brane-world energy with c2s = 1 and σ = 0. The dotted blue line shows
models is that the dark matter does not see the higher- how the dark matter growth factor decreases for a constant
σDE = −0.1. The long-dashed magenta line uses the theoreti-
dimensional aspects of the theory as it is bound to the
cal anisotropic stress of Eq. (12) with c2s = 1, which suppresses
three-dimensional brane. Its evolution is then the same the growth of the matter perturbations too much. Finally, the
as in the standard model. The modifications appear only dash-dotted cyan line (nearly on top of black solid DGP line)
in the gravitational sector, represented by the metric per- uses the same σDE but sets the pressure perturbation of the
turbations. dark energy to δp = (1 + w)ρσ in its rest frame.
The metric perturbation in DGP can be written as
[19, 20]
Fig. 2 again the growth factor for scalar field dark energy
and the DGP model, but now also a family of dark en-
 
2 2 1
k φ = −4πG a 1 − ρm ∆m (9) ergy models with non-vanishing anisotropic stress σ. We
3β
  notice that these models can easily suppress the growth
1
k 2 ψ = −4πG a2 1 + ρm ∆m (10) of perturbations in the dark matter for σ < 0 and mimic
3β
the behaviour of the DGP model.
where the parameter β is defined as: Formally we can recover the DGP metric perturbations
by choosing
!
Ḣ 2 ρm
β = 1 − 2rc H 1 + = 1 + 2rc HwDE (11) σDE = ∆m . (12)
3H 2 9β(1 + wDE ) ρDE

The dark matter does not care if the metric perturba- for the anisotropic stress of the dark energy, if we can
tions are generated (in addition to its own contribution) also generate dark energy perturbations with
by a modification of gravity or by an additional dark en- 1
ergy fluid. Its response to them is identical. Or to put it ρDE ∆DE = − ρm ∆m . (13)
3β
differently, if the dark energy and dark matter together
can create the φ and ψ of Eqs. (9) and (10) then the We notice that these are very large dark energy pertur-
growth factor (and indeed all other cosmological observ- bations. Indeed, if we keep c2s = 1 and set σ to the
ables) will be the same as in the DGP scenario. expression (12) we suppress the growth of the matter
We see immediately that in order to generate these perturbations too much, see Fig. 2. Since β < 0 the
metric perturbations we will need to introduce an large dark energy perturbations of Eq. (13) then increase
anisotropic stress since φ 6= ψ. This seems to be a very the matter clustering back to the DGP value.
generic property of modified gravity that is also present in The required size of the dark energy perturbations in
f (R) models [21] and has been noticed before. We plot in itself is no problem, as we can lower the sound speed
 4

and even make it negative. However, while for σ = 0 we tween the two possibilities.
were not able to decrease ∆m with the help of the dark Although the construction of a matching dark energy
energy perturbations, we find that with a large, nega-
model for the DGP case may seem very fine tuned, we
tive anisotropic stress we are unable to increase it. The are here more concerned with the question to what de-
required anisotropic stress is far larger than the gravi-
gree this is possible at all. Just measuring a growth factor
tational potential ψ, and it starts to be the dominant
that does not agree with scalar field dark energy is not
source of dark energy clustering in Eq. (5). As it enters sufficient to rule out “dark energy” and General Relativ-
with the opposite sign it now leads to anti-clustering of
ity. But clearly, if the expansion history and the growth
the dark energy with respect to the dark matter which of matter perturbations were to match those predicted
feels only ψ (ie. dark matter overdensities are dark en- from a physically motivated and self-consistent modified
ergy voids). There is still enough freedom in the choice
gravity model, a statistical analysis would rule out a fine
of σ to match the growth factor very precisely, but if we tuned dark energy model. However, we should not for-
could measure both φ and ψ separately then we could
get that as observations seem to indicate wDE ≈ −1 it is
detect the differences between the two models.
rather the modified gravity models that are about to be
Is it really not possible to match both ψ and φ of ruled out [22] or look increasingly fine tuned.
the DGP model within a generalised fluid dark energy
model? Yes, it is: The metric perturbations have two M.K. and D.S. are supported by the Swiss NSF. It is a
degrees of freedom, and we do have two degrees of free- pleasure to thank Chiara Caprini and Ruth Durrer for in-
dom of the dark energy to adjust, σ and δp. As it turns teresting discussions, and Eric Linder and Roy Maartens
out, the parametrisation in terms of the rest-frame sound for comments on the draft.
speed is too restrictive. This can happen for example if
the dark energy is not composed of a single fluid, see
e.g. the discussion in [16]. Allowing free use of the pres-
sure perturbations, we can choose them for example to ∗
Electronic address: [Link]@[Link]
cancel the direct effect of σ onto the dark energy pertur- †
Electronic address: [Link]@[Link]
bation in Eq. (5) by setting δp = (1+w)ρσ. This reverses [1] A.G. Riess et al, Astron. J. 116, 1009 (1998); S. Perl-
the sign of ∆DE , and minor adjustments to the pressure mutter et al., Astrophys. J. 517, 565 (1999).
perturbations can then provide the required match to [2] D.N. Spergel et al., ApJS 148, 175 (2003).
∆m . For the cyan dash-dotted curve in Fig. 2 we set [3] M. Tegmark et al., Phys. Rev. D 69, 103501 (2004); M.
δp = (1 + w)ρσ + 3Hac2a ρV /k 2 , ie. we cancelled the con- Tegmark et al., Astrophys. J. 606, 702 (2004).
[4] C. Deffayet, Phys. Lett. B 502, 199 (2001)
tribution of the anisotropic stress in the dark energy rest
[5] P. Binetruy, C. Deffayet, U. Ellewanger and D. Langlois;
frame. This provides a very good solution to Eqs. (12) Phys. Lett. B 477 285 (2000).
and (13) during matter domination. It is easy to improve [6] R. Maartens; astro-ph/0602415.
the solution to the point where it is impossible to distin- [7] S. Capozziello, V.F. Cardone, S. Carloni and A. Troisi,
guish observationally between the DGP scenario and this Int. J. Mod. Phys. D 12, 1969 (2003).
generalised dark energy model. [8] G.R. Dvali, G. Gabadadze and M. Porrati, Phys. Lett. B
Is linear perturbation theory still valid with such a 484, 112 (2000).
[9] C. Bonvin, R. Durrer and M. Kunz, Phys. Rev. Lett. 96,
large anisotropic stress? Using Eq. (13) we can rewrite
191302 (2006).
Eq. (12) as σDE = −2/(3(1+wDE ))∆DE . The anisotropic [10] E.V. Linder, Phys. Rev. D 72 043529 (2005).
stress is therefore comparable in size to ∆m and ∆DE , and [11] D. Huterer and E.V. Linder, astro-ph/0608681.
at high redshift DGP approaches GR. It is thus safe to [12] D. Polarski, astro-ph/0605532.
study the dark energy with linear perturbation theory as [13] E.V. Linder, astro-ph/0610173.
long as the dark matter perturbations stay in the linear [14] C.P. Ma and E. Bertschinger, Astrophys. J. 455, 7
regime, even in the presence of the anisotropic stresses. (1995).
[15] W. Hu, astro-ph/0402060.
[16] M. Kunz and D. Sapone, Phys. Rev. D 74, 123503 (2006).
[17] W.J. Percival, A&A 443, 819 (2005).
CONCLUSIONS
[18] S. Dutta and I. Maor, astro-ph/0612027.
[19] K. Koyama, R. Maartens, JCAP 0601, 016 (2006).
We have shown in this letter that the growth factor [20] K. Koyama, JCAP 0603, 017 (2006).
is not sufficient to distinguish between modified gravity [21] V.F. Mukhanov, H.A. Feldman and R.H. Brandenberger,
and generalised dark energy, even if the expansion his- Phys. Rep. 215, 206 (1992).
[22] R. Maartens and E. Majerotto, Phys. Rev. D 74, 023004
tory (and so the effective equation of state of the dark
(2006).
energy) has been fixed by observations. We have also [23] We emphasize that we discuss only how the dark energy
demonstrated that in some cases (notably DGP) the dark perturbations can modify the behaviour of the dark mat-
energy can match the metric perturbations completely ter, without taking into account limits from observations.
so that cosmological observations cannot distinguish be-