Phenomenological Gravitational Phase Transition: Early and Late Modifications

Nima Khosravi1, ∗ and Marzieh Farhang1, †

1
Department of Physics, Shahid Beheshti University, 1983969411, Tehran, Iran
(Dated: September 23, 2021)
In this work we generalize the idea of a gravitational phase transition or GPT at late times [1] to allow for a
modified gravity scenario in the early universe as well. The original GPT was shown to simultaneously relax
the H0 and σ8 tensions and Planck internal inconsistencies. However, the primary results from GPT predic-
tions implied disagreement with the data from the baryonic acoustic oscillations (BAO). Here we investigate
whether the generalized GPT scenario could address the Hubble tension in the presence of the BAO data as
well. We find that, despite the vastly enlarged gravitational parameter space, the BAO data strongly prefer the
ΛCDM paradigm with a low inferred Hubble constant, 68.03 ± 0.76 km/s/Mpc, in tension with the local H0
arXiv:2109.10725v1 [astro-ph.CO] 22 Sep 2021

measurements from [2], H0 = 74.03 ± 1.42 km/s/Mpc. This failure of the generalized GPT scenario to host
the P18-BAO-R19 trio is of significance and noticeably shrinks the space of possible gravitational solutions to
the Hubble tension.

I. INTRODUCTION AND MOTIVATIONS
The recent cosmological tensions between early and late
time observations in the context of the standard model of
cosmology, the so-called ΛCDM, have been extensively ad-
dressed by many cosmologists in the last few years. However,
a truly convincing theoretical model for their explanation is
still missing. The main tension is the well-known H0 -tension
which is the ∼ 4.2σ discrepancy between the local measure-
ments of the Hubble constant [2, 3, 4] and its inferred value
from the cosmic microwave background (CMB) data in the
ΛCDM framework [5]1 . There have also been many attempts
for H0 estimation from other independent observations, with
higher H0 values [8, 9, 10, 11, 12]. Beside the Hubble tension,
there is the milder σ8 tension with its local measurements im-
pling less clumpiness in matter distribution compared to the
CMB-based inferences [5]. Certain internal inconsistencies
are also reported in Planck results, specifically in the CMB
lensing amplitude and the low- and high-` discrepancy. If
not due to systematics, and if considered statistically signif-
icant, theoretical models beyond ΛCDM are required to ex-
plain them.
Many different models are proposed to address these cos-
mological tensions [see 13, for a comprehensive review]. The
models could in general be categorized into late and early,
based on the era of the modification. The motivation behind
the late time modifications is clear and there are many at-
tempts [see, e.g., 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24,
25, 26, 27, 28, 29, 30, 31]. The early modifications try to
reduce the sound horizon at the last scattering surface which
results in a higher value for H0 [32, 33, 34, 35]. There are
many other different ideas to approach the H0 tension, such
as dark matter [36], primordial magnetic fields [37] and etc.
Neither of these models are consistent with all the datasets, in
particular when data from the baryonic acoustic oscillations
(BAO) are included [38].
∗ Electronic address: n-khosravi@sbu.ac.ir
† Electronicaddress: m farhang@sbu.ac.ir
1 H (z) = H0 Ωr (1 + z) + Ωm (1 + z) + ΩΛ (z)
For some controversy on the significance of the Hubble tension see [6, 7].
In a recent work [1] the authors proposed a phenomeno-
logical phase transition in the gravity sector by assuming that
the early standard ΛCDM transits to a modified gravity model
both at the background and perturbation levels. We verified
that the model could resolve the CMB-lensing, low- and high-
`’s, σ8 and H0 tensions simultaneously (at different levels).
The BAO datasets were not included in the main analysis and
parameter estimation of [1]. However, the predictions of GPT
for BAO observables were shown to disagree with some BAO
data points. In this work we generalize the model to allow for
transition between two general early and late gravity models
at some intermediate redshift. We will study this model in the
presence of BAO datasets. These standard rulers provide us
with independent observations at low redshifts, where a tran-
sition in the gravity sector was shown to relax the tensions
between CMB and local Hubble measurements. On one hand,
BAO data are expected to break the parameter degeneracies
and improve parameter measurements. On the other hand, the
enhanced space of gravitational degrees of freedom at early
times may have the potential to accommodate both BAO data
and local H0 values. This is the driving motivation of this pa-
per: to explore the huge subset of gravity models effectively
modelled by a transition at some redshift, with possible si-
multaneous deviations from GR in early and late times, and
search for a possible sweet spot where CMB, BAO and local
H0 measurements can all co-exist in peace.
II. THE GPT MODEL
The gravitational phase transition, or GPT model, intro-
duced in [1], suggests a transition in the gravitational theory,
phenomenologically modelled by a "tanh" function. It was
assumed that the early phase of gravitation is governed by
the standard Einstein-Hilbert model, referred to as GR, which
forms the gravity model in the ΛCDM scenario. In this work
we relax this assumption and study the consequences.
We assume our Universe is isotropic, homogeneous and
flat, where the background evolution is given by
 
2 2 4 3
(1)
 2

and linear scalar perturbations are described by

P18
ds2 = a2 (τ )[ − (1 + 2Ψ(τ, ~x))dτ 2 P18+R19
+ (1 − 2Φ(τ, ~x))d~x2 ] (2) P18+BAO
in the Newtonian gauge. Here Ψ and Φ characterize the scalar
perturbations and a(τ ) is the scale factor. In Fourier space the
linear Einstein equations can be modified phenomenologically

P
as
k 2 Ψ = −µ(z) 4πGa2 [ρ∆ + 3(ρ + P )σ] (3)
k 2 [Φ − γ(z)Ψ] = µ(z) 12πGa2 (ρ + P )σ, (4)
where µ(z) and γ(z) are responsible for any general redshift-
dependent modifications to GR. Equations (3) and (4) can also
be written as 50 60 70 80 90 100
k 2 (Φ + Ψ) = 8πGa2 Σ(z) ρ ∆ (5)
H0
which is particularly useful for studying the gravitational lens- FIG. 1: The 1-D marginalized likelihood for the Hubble constant
ing. In the GPT scenario, we assume a transition in the gravity H0 is shown for different datasets. The P18 case, represented by
sector and model the transition through "tanh" function. At the solid green curve, shows no inconsistencies with R19. Adding
the background level we therefore have R19 (the dashed red curve) results in a shift to higher H0 values as
  expected. In P18+BAO case, however, the preferred H0 value, repre-
early 1 + tanh α (zt − z) sented by dotted blue curve, becomes much smaller in comparison to
ΩΛ (z) = ΩΛ + ∆Λ . (6) R19. This shows the power of BAO in putting very tight constraints
2
on our model parameters.
Here zt is the transition redshift and α is the inverse of the
transition’s width. Imposing the flatness condition gives Ωr +
Ωm + ΩΛ = 1, where ΩΛ ≡ ΩΛ (z = 0) = Ωearly + ∆Λ [1 + dataset consists of 6DFGS [39], SDSS MGS [40] and BOSS
tanh(αzt )]/2. Similarly, the modifications to the perturbation DR12 [41] and we refer to them in general as BAO.
equations are characterized by transitions in µ(z) and γ(z), We use CosmoMC3 , the Cosmological Monte
  Carlo code [42], to estimate the gGPT parameters
early 1 + tanh α (zt − z) (∆Λ , µearly , ∆µ , γ early , ∆γ , α, zt ) as well as the stan-
µ(z) = µ + ∆µ ,
2  dard cosmological parameters (Ωb h2 , Ωdm h2 , θ, τ, ns , As )
1 + tanh α (zt − z) with the above datasets. We assume uniform prior on all of
γ(z) = γ early + ∆γ . (7) these parameters. For parameter estimation we use P18+BAO
2
and P18+R19. We know the consistency of these data
where we do not impose any restricting assumptions on µearly , combinations from the standard ΛCDM and the late GPT
γ early and Σearly . This is in contrast to the late GPT model model respectively. To be more precise, there is no tension
where we had µearly = γ early = 1 and Σearly = −1. This between P18 and BAO in ΛCDM and between P19 and R19
generalization allows for transitions happening between two in late GPT. Therefore there should be no tension in the
general gravity models and not necessarily starting from GR extended encompassing parameter space of gGPT neither.
in the past. We refer to this generalized model as gGPT.

IV. RESULTS AND DISCUSSION

III. ANALYSIS AND DATASETS

In our previous work [1] we studied the H0 and σ8 tensions

In this work the goal is to explore whether a transition in the
(referred to as external Planck tensions) and the two internal
gravity sector can address the tensions between CMB, BAO
Planck inconsistencies. We showed that the P18, R19 and
and local data of Hubble constant measurements. We use the
DES datasets can be used consistently in the GPT framework.
Planck measurements of the CMB temperature and polariza-
However, including BAO to the analysis would push the pre-
tion anisotropies and lensing [5] and refer to it as P18. For
ferred scenario toward ΛCDM with low H0 and thus restoring
the local H0 data we use [2], labeled here as R192 . Our BAO
the tension with R19. Here we explore the possibility that the
new degrees of freedom in the early gravity sector of the gGPT
can host all the independent datasets without any tensions.
2 We are aware of concerns regarding the absolute magnitude datasets and
using R19 data. However, our main interest here is the impact of BAO’s on
parameter measurement and believe our main conclusions and results will
not be affected by this concern. 3 https://cosmologist.info/cosmomc/
 3

Ω b h2 Ω c h2 H0 τ log(1010 As ) ns
P18+R19 0.0227 ± 0.0002 0.1182 ± 0.0016 73.69 ± 1.47 0.049 ± 0.008 3.03 ± 0.02 0.968 ± 0.005
P18+BAO 0.0226 ± 0.0002 0.1183 ± 0.0012 68.03 ± 0.76 0.050 ± 0.008 3.03 ± 0.02 0.967 ± 0.004

TABLE I: Standard cosmological parameters as measured by the data combinations, P18+R19 and P18+BAO, in the gGPT framework.

In Fig. 1 we have plotted the 1D-likelihood curves for P18

the Hubble constant with various data combinations of P18, 6 P18+R19
P18+BAO
P18+R19 and P18+BAO in gGPT. Clearly, P18 gives a wide 4
likelihood, with no inconsistencies with R19. We expected
this result as the original GPT, with fewer free parameters, was 2

∆H (z)
found to be consistent with R19 [1]. However, we find that
0
adding BAO to P18 data would tighten the likelihood around
the lower value of H0 = 68.03 ± 0.76, and thus disagree −2
with R19 measurements. We should emphasize that these re-
−4
sults are in the presence of the early-, as well as late-time,
degrees of freedom, allowing for the gravity theory to devi- −6
ate from GR even before transitioning to a possibly modified
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
late gravity model. The other standard parameters, measured
z
with P18+BAO and P18+R19, are found to be consistent, as
presented in Table I. P18
P18+R19
In the following we investigate how different data combi- 1.0 P18+BAO
nations would constrain the evolution of the main free param-
eters in the model.
Background evolution: Fig. 2 shows the evolution of H 0.8
ΩΛ(z)
and ΩΛ , calculated for various data combinations. For better
visual comparison, the P18 best-fit measurement of H(z) is 0.6

subtracted from the H(z) measurements, and the upper panel

represents ∆H = H(z) − HP18 (z). The Hubble tension is 0.4
most vivid, with ∼ 4σ, at z = 0 in the ∆H plot, between
P18+BAO and P18+R19 measurements.
0.2
The trajectories of the time evolution of these parameters
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
and their tensions are quite interesting. First, BAO is found z
to be quite strong in confining the trajectories to the ΛCMD4
best-fit. The tight constraints imposed by BAO data and their
FIG. 2: The 1σ trajectories of the background evolution in the gen-
strong preference for ΛCDM, in the whole z range, lead to the
eralized GPT model, presented by the expansion history H(z) (top)
significant discrepancy with R19. As is seen from the plots, and ΩΛ (z) (bottom). For better visual comparison the expansion
the tension in H(z) goes beyond z = 0 to z ∼ 0.5, with de- history is plotted compared to Planck measurements of H(z), i.e.,
creasing significance and reminds us of the H(z) tension re- ∆H = H(z) − HP18 (z).
ported in [15, 43]. The discrepancy between the cosmological
implications of BAO and R19 in the gGPT framework is also
present in ΩΛ (z), in z . 3 with lower significance. P18+BAO observation: BAO systematically prefers µ < 1 and γ > 1
prefers a constant dark energy (similar to the cosmological and just touches µ = γ = 1 in the edge of its 1σ contours.
constant) while P18+R19 requires a dynamical dark energy. This result, together with the observed anti-correlation of µ
At higher z, with little dependence of the expansion history and γ as shown in Fig. 4, imply that if more constraining data
on dark energy, we do not expect any tension. push one of the two parameters away from GR, the other one
Perturbation parameters: As is obvious from Fig. 3, the would also be driven away from GR. BAO data is therefore
constraints on the perturbations’ degrees of freedom µ and γ expected to play a crucial role in discriminating between the-
are not very tight. All datasets are compatible with each other oretical gravitational models.
and with ΛCDM, i.e., µ = γ = 1. There is also a remarkable

V. FINAL WORDS

4 It should be noted that, although the x-axis in the ∆H (z) plot, by defini-
tion, corresponds to the best-fit gGPT with P18 data, its agreement with
the best-fit ΛCDM is quite well (as it is obvious from solid green line in
ΩΛ (z) plot). However, the 1σ spans of the trajectories around this com-
mon best-fit differ by a large amount.
In this work we generalized our previous GPT model [1]
to allow the gravity sector go beyond GR both in early and
late times. In [1] we showed that a late-only phase of de-
viation from GR can address H0 and σ8 tension, as well as
 4

1.30
P18 the internal inconsistencies of Planck. However, there were
P18+R19
P18+BAO
hints for tensions with some BAO observations based on GPT
1.25
predictions for BAO distances. In the current work we in-
1.20 cluded BAO in our study of the gGPT scenario to fully study
1.15
their impact on parameter estimation. In brief, even in this en-
hanced framework, BAO did not allow for deviations from GR
µ(z)

1.10
in the background level. Therefore, our gGPT model could
1.05 not resolve the H0 tension among the P18-R19-BAO trio. At
1.00 the perturbation level, on the other hand, the data are found
to be consistent. The ∼ 1σ deviation of µ and γ from GR
0.95
predictions with BAO, although not statistically significant, is
0.90 persuading for future investigation of these perturbation pa-
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
rameters with more informative data and hints to the power
z
of these parameters in differentiating between various scenar-
P18 ios and better understanding of the origin of the cosmological
P18+R19
1.4
P18+BAO
tensions including H0 and σ8 .
The failure of this relatively general and effective extension
1.2 of GR, both in the background and in perturbations, should
be taken seriously. That is because the proposed gGPT sce-
γ(z)

nario is quite general and its many degrees of freedom can

1.0
effectively encompass various modifications to GR as long as
their cosmological implications for CMB and BAO are con-
0.8 cerned. The fact that CMB is only sensitive to the expan-
sion history in projection leads to large degeneracies among
cosmological models with differing post-recombination his-
0.6
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
tories. Due to this degeneracy the detailed features of the
z history do not matter and one expects effective parameteriza-
tions suffice to represent different models. It should however
FIG. 3: The 1σ trajectories of the modifications to the perturbed Ein- be noted that BAO data partially break this degeneracy, espe-
stein equations in the generalized GPT model, parametrzied by µ(z) cially at late times. We therefore interpret this failure as the
(top) and γ(z) (bottom). insufficiency of quite general (scale-independent) early and
late modifications to gravity, in the background and perturba-
tions to solve the Hubble tension. This conclusion can break if
P18
P18+R19 the proposed modification differs vastly and observably from
−0.82
P18+BAO
the gGPT scenario, even effectively and in projection. As
−0.84 other ways around this conclusion one could consider pre-
−0.86
recombination modifications5 , transitions at different transi-
tion redshifts for background an perturbations, and exploring
ρ(γ, µ)

−0.88 k-dependent modifications. Possibilities beyond-gGPT will

−0.90
be explored in future works.

−0.92

−0.94

Acknowledgement
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
z
The numerical calculations of this work were carried out on
FIG. 4: The correlation between µ(z) and γ(z), for the three data the computing cluster of the Canadian Institute for Theoretical
combinations used in this work. Astrophysics (CITA), University of Toronto.

[1] M. Farhang and N. Khosravi, Phys. Rev. D 103, 083523 (2021),
2011.08050.
[2] A. G. Riess, S. Casertano, W. Yuan, L. M. Macri, and D. Scol-
nic, Astrophys. J. 876, 85 (2019), 1903.07603.
[3] A. G. Riess et al., Astrophys. J. 826, 56 (2016), 1604.01424.
[4] A. G. Riess et al., Astrophys. J. 861, 126 (2018), 1804.10655.
[5] Planck Collaboration, N. Aghanim, Y. Akrami, M. Ashdown,
J. Aumont, C. Baccigalupi, M. Ballardini, A. J. Banday, R. B.
Barreiro, N. Bartolo, et al., Astron. Astrophys. 641, A6 (2020),
1807.06209.

[6] G. Efstathiou, Mon. Not. Roy. Astron. Soc. 505, 3866 (2021),
2103.08723.
[7] D. Camarena and V. Marra, Mon. Not. Roy. Astron. Soc. 504,
5164 (2021), 2101.08641.
[8] W. L. Freedman et al. (2019), 1907.05922.
[9] C. D. Huang, A. G. Riess, W. Yuan, L. M. Macri, N. L. Za-
kamska, S. Casertano, P. A. Whitelock, S. L. Hoffmann, A. V.
Filippenko, and D. Scolnic (2019), 1908.10883.
[10] K. C. Wong et al., Mon. Not. Roy. Astron. Soc. 498, 1420
(2020), 1907.04869.
[11] N. Khetan et al. (2020), 2008.07754.
[12] D. Pesce et al., Astrophys. J. Lett. 891, L1 (2020), 2001.09213.
[13] E. Di Valentino, O. Mena, S. Pan, L. Visinelli, W. Yang, A. Mel-
chiorri, D. F. Mota, A. G. Riess, and J. Silk, Class. Quant. Grav.
38, 153001 (2021), 2103.01183.
[14] E. Di Valentino, A. Melchiorri, and O. Mena, Phys. Rev. D 96,
043503 (2017), 1704.08342.
[15] G.-B. Zhao et al., Nature Astron. 1, 627 (2017), 1701.08165.
[16] N. Khosravi, S. Baghram, N. Afshordi, and N. Altamirano,
Phys. Rev. D 99, 103526 (2019), 1710.09366.
[17] W. Yang, S. Pan, E. Di Valentino, R. C. Nunes, S. Vagnozzi,
and D. F. Mota, JCAP 09, 019 (2018), 1805.08252.
[18] W. Yang, A. Mukherjee, E. Di Valentino, and S. Pan, Phys. Rev.
D 98, 123527 (2018), 1809.06883.
[19] A. Banihashemi, N. Khosravi, and A. H. Shirazi, Phys. Rev. D
99, 083509 (2019), 1810.11007.
[20] X. Li and A. Shafieloo, Astrophys. J. Lett. 883, L3 (2019),
1906.08275.
[21] R. E. Keeley, S. Joudaki, M. Kaplinghat, and D. Kirkby, JCAP
12, 035 (2019), 1905.10198.
[22] M. Raveri, Phys. Rev. D 101, 083524 (2020), 1902.01366.
[23] S.-F. Yan, P. Zhang, J.-W. Chen, X.-Z. Zhang, Y.-F. Cai, and
E. N. Saridakis, Phys. Rev. D 101, 121301 (2020), 1909.06388.
[24] E. Di Valentino, A. Mukherjee, and A. A. Sen (2020),
2005.12587.
[25] M. Braglia, M. Ballardini, W. T. Emond, F. Finelli, A. E. Gum-
rukcuoglu, K. Koyama, and D. Paoletti, Phys. Rev. D 102,
023529 (2020), 2004.11161.
5
[26] A. Gómez-Valent, V. Pettorino, and L. Amendola, Phys. Rev. D
101, 123513 (2020), 2004.00610.
[27] M. Lucca and D. C. Hooper (2020), 2002.06127.
[28] E. Di Valentino, A. Melchiorri, O. Mena, and S. Vagnozzi,
Phys. Dark Univ. 30, 100666 (2020), 1908.04281.
[29] E. Di Valentino, A. Melchiorri, O. Mena, and S. Vagnozzi,
Phys. Rev. D 101, 063502 (2020), 1910.09853.
[30] A. Banihashemi, N. Khosravi, and A. H. Shirazi, Phys. Rev. D
101, 123521 (2020), 1808.02472.
[31] A. Banihashemi, N. Khosravi, and A. Shafieloo, JCAP 06, 003
(2021), 2012.01407.
[32] V. Poulin, T. L. Smith, T. Karwal, and M. Kamionkowski, Phys.
Rev. Lett. 122, 221301 (2019), 1811.04083.
[33] L. Knox and M. Millea, Phys. Rev. D 101, 043533 (2020),
1908.03663.
[34] R. E. Keeley, A. Shafieloo, D. K. Hazra, and T. Souradeep,
JCAP 09, 055 (2020), 2006.12710.
[35] S. Vagnozzi (2021), 2105.10425.
[36] K. Vattis, S. M. Koushiappas, and A. Loeb, Phys. Rev. D 99,
121302 (2019), 1903.06220.
[37] K. Jedamzik and L. Pogosian, Phys. Rev. Lett. 125, 181302
(2020), 2004.09487.
[38] K. Jedamzik, L. Pogosian, and G.-B. Zhao, Commun. in Phys.
4, 123 (2021), 2010.04158.
[39] F. Beutler, C. Blake, M. Colless, D. H. Jones, L. Staveley-
Smith, L. Campbell, Q. Parker, W. Saunders, and F. Watson,
Mon. Not. Roy. Astron. Soc. 416, 3017 (2011), 1106.3366.
[40] A. J. Ross, L. Samushia, C. Howlett, W. J. Percival, A. Burden,
and M. Manera, Mon. Not. Roy. Astron. Soc. 449, 835 (2015),
1409.3242.
[41] S. Alam et al. (BOSS), Mon. Not. Roy. Astron. Soc. 470, 2617
(2017), 1607.03155.
[42] A. Lewis and S. Bridle, Phys. Rev. D 66, 103511 (2002), astro-
ph/0205436.
[43] L. Pogosian, M. Raveri, K. Koyama, M. Martinelli, A. Silvestri,
and G.-B. Zhao (2021), 2107.12992.