Cosmological Tests of Quintessence in Quantum Gravity: A, B, B, C, B, D, B

Cosmological tests of quintessence
in quantum gravity
Sukannya Bhattacharyaa,1 , Giulia Borghettob, 2 , Ameek Malhotrab, 3 ,

arXiv:2410.21243v1 [astro-ph.CO] 28 Oct 2024
Susha Parameswaranc,4 , Gianmassimo Tasinatob,d,5 , Ivonne Zavalab,6

a
Instituto de Fı́sica Téorica UAM/CSIC, Calle Nicolás Cabrera 13-15, Cantoblanco,
28049, Madrid, Spain
b
Physics Department, Swansea University, SA2 8PP, UK
c
Department of Mathematical Sciences, University of Liverpool, Liverpool, L69 7ZL, UK
d
Dipartimento di Fisica e Astronomia, Università di Bologna and
INFN, Sezione di Bologna, I.S. FLAG, viale B. Pichat 6/2, 40127 Bologna, Italy
We use a suite of the most recent cosmological observations to test models of dynamical
dark energy motivated by quantum gravity. Specifically, we focus on hilltop quintessence
scenarios, able to satisfy theoretical constraints from quantum gravity. We discuss their
realisation based on axions, their supersymmetric partners, and Higgs-like string construc-
tions. We also examine a specific parameterisation for dynamical dark energy suitable for
hilltop quintessence. We then perform an analysis based on Markov Chain Monte-Carlo
to assess their predictions against CMB, galaxy surveys, and supernova data. We show to
what extent current data can distinguish amongst different hilltop set-ups, providing model
parameter constraints that are complementary to and synergetic with theoretical bounds
from quantum gravity conjectures, as well as model comparisons across the main dark
energy candidates in the literature. However, all these constraints are sensitive to priors
based on theoretical assumptions about viable regions of parameter space. Consequently,
we discuss theoretical challenges in refining these priors, with the aim of maximizing the
informative power of current and forthcoming cosmological datasets for testing dark energy
scenarios in quantum gravity.
1
[email protected]
2
[email protected]
3
[email protected]
4
[email protected]
5
[email protected]
6
[email protected]
Contents
1 Introduction 1
2 Quintessence in string theory 4

2.1 dS minima, plateaus, runaways, maxima and saddles vs the Swampland 4
2.2 String models of hilltop quintessence 5
2.3 Cosmological equations for hilltop quintessence 9
3 Parameterisation of the equation of state for hilltop models 10

3.1 Derivation of the Dutta-Scherrer(-Chiba) (DSCh) parameterisation 11
3.2 Testing the DS parameterisation for hilltop quintessence models 14
3.3 Comparison: testing DSCh parameterisation in exponential quintessence 19
3.4 Bounds on – and from – the initial conditions 20
4 Cosmological analysis 22
4.1 Axion hilltop 23
4.2 Saxion hilltop 25
4.3 Higgs-like hilltop 28
4.4 DS parameterisation analysis 29
4.5 Model comparison 32
5 Outlook and Future Challenges 33
A Saxion-axion stringy hilltops 36

A.1 Saxion hilltops 36
A.2 Subleading corrections to the axion 37
B Constraints for all the parameters of our models 39

B.1 Axion hilltop 39
B.2 Saxion hilltop 41
B.3 Higgs-like hilltop 43
B.4 Dutta-Scherrer parameterisation 45
C Dutta-Scherrer-Chiba parameterisation, including curvature 48
References 50
1 Introduction
One of the most significant and challenging problems in contemporary fundamental physics
is to understand the microscopic nature of the Dark Energy (DE) that dominates our
1
Universe today, driving its current accelerated expansion. From one perspective, DE may
appear a low-energy problem, since the observed DE scale is small, lying around the milli-
eV. However, the fact that vacuum energy – an ultraviolet sensitive quantum phenomenon
– behaves as DE once included in Einstein’s General Relativity, frames the problem as a
high-energy one. This makes it all the more exciting that cosmological observations are
probing the behaviour of DE with an ever-increasing degree of precision, possibly opening
a path to connect quantum gravity to observations. In this paper, we consider classes of
quintessence models for DE that are currently allowed by quantum gravity considerations,
and we test them against the most recent cosmological data. At the same time, we identify
trends that current cosmological results indicate for DE model building in quantum gravity.
String theory provides an excellent framework for the DE problem (for reviews, see
e.g. [1, 2]). The simplest candidate for DE has long been considered to be a positive vacuum
energy (corresponding to a de Sitter vacuum), which, however, must be fine-tuned to a level
of one part in 10120 – the so-called cosmological constant problem. The supposed String
Landscape of exponentially large numbers of finely-spaced metastable de Sitter vacua,
together with eternal inflation to populate them, lended itself to an anthropic explanation
of this fine-tuning (see e.g. [3] for a review). However, despite impressive technical progress,
after two decades of effort there is still no consensus on a single example of an explicit,
well-controlled de Sitter vacuum in string theory. Instead, a number of obstructions are
invariably met in the hunt for metastable de Sitter string vacua, including challenges in
satisfying global and local constraints, tachyonic instabilities, and a lack of parametric or
even numerical control in the perturbative expansions used.
At the same time, there has been a growing focus on the expectation that, in string
theory and quantum gravity, not everything goes: not every effective field theory (EFT)
can be ultraviolet completed into a theory of quantum gravity and those EFTs that are
not consistent with quantum gravity are deemed to be in the Swampland. In mapping out
which EFTs lie in the Swampland, and which are safely in the Landscape, a number of
Swampland Conjectures have been put forward (for recent reviews see e.g. [4–6]). Among
them, the de Sitter Swampland Conjecture proposes that (meta)stable de Sitter vacua are
inconsistent with quantum gravity. In terms of a string compactification’s low energy EFT
ingredients, it is supposed that the scalar potential of the string moduli (describing sizes,
shapes and positions in the extra dimensions, and the string coupling) should satisfy [7, 8]:
√
∇a V ∇a V c min(∇a ∇b V ) c′
≥ or ≤− 2 , (1.1)
V MPl V MPl
where “min()” denotes the minimal eigenvalue and c and c′ O(1) positive constants. Whilst
there exist physical arguments for these inequalities to hold in asymptotic regions of the
moduli space [8] – where large moduli correspond to weak couplings in the corresponding
perturbative expansions – the conjecture speculates that it holds everywhere in moduli
2
space. This is not uncontroversial, but it is based on the empirical evidence previously
discussed, together with conceptual issues with observers in de Sitter space, such as how
to define an S-matrix in this context [9–11].
Conjecture (1.1) rules out a metastable de Sitter vacuum as the explanation for DE,
as we cannot have simultaneously ∇a V = 0, V > 0, and eigenvalues(∇a ∇b V ) > 0. The
main alternative model for DE is slow-roll quintessence, and a priori it would be natural
to expect that quintessence candidates are found amongst the string moduli. Although
the conjecture is in tension with the simplest realisations of slow-roll inflation for the early
Universe – the left-hand-sides of (1.1) corresponding directly to the potential slow-roll
parameters, ϵV and ηV , which need to be small in single-field slow-roll inflation (but see
e.g. [12] for potential counter-examples) – it leaves some room to play within the context
of late-time quintessence, depending on the values of c and c′ . Indeed, whereas around
60 e-folds of inflation in the early Universe are required to solve the horizon problem, the
late-time accelerated expansion has been occurring for less than one e-fold of expansion,
suggesting a viable window in (1.1).
Three distinct classes of simple, smooth slow-roll quintessence potentials then come to
mind: plateaus, runaways, and hilltops. Plateaus, corresponding to both first and second
derivatives of the potential being small, are in tension with (1.1); moreover, they are as
difficult to obtain from string theory as metastable dS vacua. Runaway potentials are
ubiquitous in string theory, and have both parametric control and a suppressed vacuum
energy emerging at the asymptotics of field space. However, generally these potentials
turn out to be too steep to source a slow-roll7 accelerated expansion [13, 14], consistently
with the expectation from (1.1) that the slow-roll parameters should be large. Moreover,
the small window that could be consistent with (1.1) and slow-roll quintessence does not
agree with observations. In fact, string compactifications generically lead to asymptotic
√
exponential potentials, V (ϕ) = V0 e−λϕ , which satisfy the conjecture with λ ≳ 2. Such
potentials can source a transient late-time accelerated expansion that follows epochs of ra-
√
diation and matter domination, provided that λ ≲ 3 [15], but this parameter space turns
out to be ruled out by the most recent cosmological data [16–18]. It is worth emphasising
that these results demonstrate a powerful synergy between quantum gravity considerations
and cosmological observations: by themselves quantum gravity would allow exponential
√ √
quintessence with 2 ≲ λ ≲ 3 and observations would allow exponential quintessence
with λ ≲ 0.537 [16]; taken together, exponential runaway quintessence is currently dis-
favoured. It remains to consider the option of hilltop potentials: this is our aim in this
work.
We focus on hilltop quintessence scenarios and explore the interplay between quantum
gravity constraints and the most recent cosmological datasets. The models we consider
can be made consistent with Conjecture (1.1) as well as other swampland conjectures;
7
Any potential can have some amount of accelerated expansion simply by tuning initial conditions such
that the scalar starts by rolling up its potential.
3
moreover, they are expected to arise naturally within string theory. In particular, we
consider axion hilltops, hilltops for a saxion within 4D N = 1 supergravity, and a Higgs-
like potential that, close to the hilltop, plays the role of a generic quadratic hilltop. After
presenting these general models, motivating their initial conditions, and discussing their
connections with string theory and particle physics in Section 2, in Section 3 we review
how they can be analysed in a unified way using an appropriate parameterisation of the
quintessence equation of state, put forward by Dutta and Scherrer [19] (considering also its
generalisation for more general thawing quintessence models by Chiba [20]). We test this
general parameterisation against our concrete scenarios and also show that it provides a
useful analytical understanding of the degree of fine-tuning of initial conditions necessary
to be consistent with observations.
We are then ready, in Section 4, to test our string-motivated hilltop models – together
with general hilltops using the Dutta-Scherrer parameterisation – against a suite of re-
cent cosmological data, from CMB observations, galaxy surveys, and Type IA supernovae
data. We find the best-fit and mean values and bounds for the fundamental parameters in
our models and investigate how these observational constraints stand against independent
bounds from quantum gravity considerations. Furthermore, we identify which model is
preferred by the data, comparing also to the fits of the ΛCDM model, the exponential
runaway model, and the w0 wa CDM model – the latter corresponding to the alternative
Chevakkier-Polarski-Linder (CPL) parameterisation [21, 22], assuming that the equation of
state parameter evolves linearly with the scale factor. We summarise our results in Section
5, where we also discuss the most important model-building challenges to be addressed in
order to extract the maximal information on quantum gravity scenarios from current and
forthcoming data sets. A number of technical appendices follows.
2 Quintessence in string theory

In this section we introduce the quintessence models whose cosmologies we study. Im-
portantly, these models are consistent with string theory expectations, but have different
interpretations within string theory, and thus distinct associated microscopic parameters
with specific interplays with particle physics. After motivating and presenting the models,
we analyze the associated cosmological equations that will be used in the following sections
to confront them with data8 .
2.1 dS minima, plateaus, runaways, maxima and saddles vs the Swampland

As discussed in the introduction, the swampland conjectures on the properties of scalar
potentials in string theory (1.1) (and the difficulties in constructing controlled metastable
8
In the recent paper [23], model-independent cosmological constraint on c are explored, finding larger-
than-one values for this quantity. However, the analysis of [23] does not include hilltops, and the parameter
c′ is not considered individually. Specifically, [23] focuses on the combined quantity Γ = −c′ /c2 in the
range Γ > 1, which excludes hilltop scenarios (and exponentials, which have Γ = 1).
4
dS string vacua that themselves motivated the conjectures) suggest that the current ac-
celeration of the universe cannot be attributed to a small positive cosmological constant.
The same swampland constraints (1.1) are also in tension with a slow-roll quintessence
characterised by a scalar potential with a plateau. We have also reviewed how the con-
jecture (1.1) suggests that runaways are typically too steep to source slow-roll accelerated
expansion, and the small window that is allowed is ruled out by observations.
It is then interesting to note that dS maxima and saddles generically satisfy the swamp-
land constraints in (1.1) and, moreover, seem to be easier9 to find in string constructions
compared to dS minima, plateaus and slow-roll runaways. In the following, we will use
the term dS “hilltops” to refer to both dS maxima and dS saddles. If one considers a
modulus close to a dS hilltop, it can remain frozen by Hubble friction during the epochs
of radiation and matter domination, acting as a small cosmological constant, until it only
recently starts to slowly roll down its potential sourcing dynamical dark energy. This pro-
vides a string-motivated, (thawing10 ) dynamical dark energy candidate, towards which –
tantalizingly – recent data seem to hint [27–29]11 .
2.2 String models of hilltop quintessence

We will consider three classes of string theory hilltop models, with the quintessence field
descending from string moduli or matter fields. Within string theory, two types of moduli
– both amongst the most generic predictions of the field content of string scenarios – can
be identified as hilltop quintessence candidates12 :
1. Axion hilltops. Axions descend from the dimensional reduction of higher dimen-
sional p-forms, giving rise to the so-called string axiverse13 [36]. Their scalar poten-
tial can be generated by non-perturbative effects such as instantons, with the leading
contribution taking the form:

θ
V (θ) = V0 1 − cos , (2.1)
f
where f is the axion decay constant, or shift-symmetry breaking scale, and V0 has an
exponential suppression in the instanton action, V0 ∼ M 4 e−Sinst , with M the scale of
9
It should, however, be acknowledged that so far all explicit top-down constructions of dS maxima
and saddles involve some size moduli that are smaller than the string-length, implying that they might be
spoilt by large α′ corrections; see e.g. [24–26]. On the other hand, as we will discuss below, one can make
rather general arguments for their existence.
10
Hilltop quintessence is an example of thawing quintessence where wϕ starts close to -1 and then
increases, as opposed to freezing quintessence where wϕ starts above -1 and decreases towards it.
11
Of course, we have to wait for future more precise data to have a conclusive answer regarding the
dynamical behaviour of dark energy.
12
Generally, more moduli could be rolling over their potential – however we focus on the simplest case
of a single dynamical field.
13
For early work on axions as quintessence in field theory see e.g. [30–35].
5
the instanton physics. Both 1/f and Sinst typically go as the saxionic superpartner
for the axion, leading to f Sinst ∼ xMPl with x an order number [33]. Also, we have
assumed (some resolution to the cosmological constant problem and) a Minkowski
minimum at θmin = 0, whilst there is a dS maximum at θmax = πf .
Depending on the value of the decay constant, axion dark energy can occur either
(i) close to the minimum of the potential at θ = 0, and up to its inflection point
at θ = πf /2, which requires f > MPl , or (ii) near the hilltop at θmax = πf , which
allows for f ≲ MPl . We are interested in the latter case, since it allows for values of
the decay constant that are usually found in string theory constructions.
Indeed, f has been argued to be always f ≲ O(1)MPl by the so-called weak gravity
conjecture [37], which for axions implies that there must exist an instanton whose
action satisfies14 :
MPl
Sinst ≲ , (2.2)
f
In the strong version of the WGC this instanton must be the one with smallest action
i.e. the leading effect. To keep control of the instanton expansion assumed in (2.1),
we require Sinst > 1 and thus f < MPl . In the case that Sinst ≫ 1, the exponential
suppression in V0 naturally realises the necessary hierarchy between the dark energy
potential and the leading-order potential that fixes the volume moduli.
As pseudo-scalars, axions can evade stringent fifth force constraints even if they
are extremely light. Furthermore, their approximate shift symmetries restrict their
allowed couplings and protect the axion mass and potential energy density, which
are otherwise UV sensitive quantities. An important open question is why initial
conditions would be fine-tuned close to the hilltop. One possible mechanism to
achieve this is if in the early Universe, the leading non-perturbative effects that are
active stabilise the axion in a Minkowsi or adS minimum, and at some later time a
further non-perturbative effect dynamically comes into play, turning this minimum
into a (nearby) dS maximum (see Appendix A.2 for a working example).
2. Saxion hilltops. Geometric string moduli, corresponding to sizes and shapes of

the extra dimensions and positions within them, as well as the dilaton, also arise
generically in string compactifications and are often associated with some small ex-
pansion parameter in the low-energy effective field theory description assumed. They
14
There is a large literature that attempts to obtain super-Planckian decay constants, usually involving
multiple axions for which there exists a generalised mutifield weak gravity conjecture (WGC). The axion
alignment mechnism [38] invokes two (or more) axions and a fine-tuning between their axion decay con-
stants that produces a large effective axion decay constant. In N -flation [39], a large√number N of axions,
each with axion decay constant f , lead to an effective axion decay constant feff = N f . However, both
axion alignment and N -flation require some extra model building to satisfy the multifield WGC, and in
any case violate the strong WGC. See e.g. [4] for a further discussion of the literature, including top-down
model building attempts, where the challenges encountered may well seem consistent with a quantum
gravity censorship of large axion decay constants.
6
also have potentials that include dS maxima and saddles; indeed, one can formulate
rather general arguments for the existence of such hilltops. Consider, for example,
a compactification that stabilises all moduli in a regime of parametric or numerical
control to a supersymmetric AdS vacuum, with one modulus lighter than the others.
If the leading correction to the potential at asymptotic values of the light modulus is
positive, then a dS maximum must exist to the right of the AdS minimum, and thus
also under control [40]. The setup becomes a priori more complex if there is more
than one light modulus, since then the minimum may not be accompanied by an
extremum in all directions. It is therefore interesting to note that the recent explicit
constructions [41] of supersymmetric AdS minima in type IIB flux compactifications
with many moduli – which are under numerical control15 – have been found to be
accompanied by dS maxima.
For concreteness, in the following, we consider a specific, simple saxion hilltop model
that is well-motivated from supergravity and was studied recently in [42]. This
model starts with a supersymmetric Minkowski setup with one flat direction, which
is lifted by a leading-order supersymmetry-breaking non-perturbative contribution,
with details given in Appendix A. The corresponding scalar potential, expressed in
terms of the canonical normalised field, is given by:
√ √ √ √
− 2ϕ −2αe 2ϕ 2 2 2ϕ
V (ϕ) = V0 e e −2 + 4 α e + 4 α e 2ϕ , (2.3)
with α a constant that depends on the type of non-perturbative effect in play. For
example, for gaugino condensation in a hidden SU (N ) gauge group (from wrapped
D7-branes) in type IIB string models, we have α = 2π/N . The scalar potential (2.3)
has a maximum at:
1 1
ϕmax = √ log √ , (2.4)
2 2α
which lies in a “weak-coupling” regime, say ϕmax ≳ 0.4 for the canonically normalised
field16 , for around α ≲ 0.4, or N ≳ 16. Note that geometrical and topological
constraints imply that N cannot be arbitrarily large17 , but we can safely take, say,
N ≲ O(100). On the other hand, as discussed in Appendix A, we can in any case
expect at best numerical control of the expansion in the non-perturbative effects at
15
Note that – even though the solutions in [41] include some two-cycle volumes that are small (c.f.
footnote 9) – it has been checked explicitly that e.g. the worldsheet instanton expansions are under
control. Moreover, whilst control of the dS minima (a.k.a. “KKLT vacua”) in [41] is under question
because the concrete examples have gs M ≲ 1, this regime is only of concern in the presence of a warped
throat, in which case the supergravity expansion breaks down. The warped throat is a necessary ingredient
for the KKLT dS minimum, but not for the precursor supersymmetric AdS vacuum of interest here.
16
This corresponds to the original field φmax ≳ 1.8 (see (A.5)).
17
E.g. for gaugino condensation from wrapped D7-branes in type IIB, [43] (see also [44]) found that
N ≲ O(10)h1,1 with h1,1 related to the number of size moduli of the compactification (more explicitly,
h1,1 is the Hodge number of the Calabi-Yau manifold counting for the number of Kähler moduli).
7
the hilltop.
Although the exponential suppression in the saxion potential energy turns out to
cancel at ϕmax , its scale can match the observed dark energy by being multiply
exponentially suppressed in the vevs of the supersymmetrically stabilised moduli
[24]. Saxions do not enjoy a shift symmetry like the axions, but constraints from
time variation of fundamental constants, fifth forces and radiative corrections can
potentially be avoided if the quintessence couples only indirectly, via gravity and with
some further geometric suppression, to the Standard Model and the supersymmetry
breaking sector. The fine-tuning of initial conditions could be explained e.g. via
high temperature effects [45] or some other dynamics [46] turning the maximum into
a transient minimum analogously to symmetry restoration in the Higgs potential;
alternatively, anthropic arguments might be relevant, since without fine-tuning to
the hilltop, the saxion would runaway to decompactification or decoupling and an
unviable universe18 .
In what follows, we explore hilltop quintessence using the two concrete moduli examples
above – the axion hilltop and the supergravity saxion hilltop – together with a more generic
Higgs-like quadratic hilltop, which can approximate any quadratic hilltop near the top, and
which might descend from a stringy saxion modulus, stringy axion modulus, or a stringy
matter field:
3. Higgs-like hilltops. The dynamics from any quadratic hilltop potential can be
approximated by using a Higgs-like potential
2 ! 2
ϕ
V (ϕ) = V0 1− . (2.5)
ϕ0
This potential is bounded from below by having a quartic term compared to the
more standard quadratic field theory hilltops. This – and any other terms that
might appear in the Taylor expansion of the hilltop potential about its maximum
– will not affect the dynamics for as long as the field stays sufficiently close to its
hilltop.
Should the field ϕ explore a significant part of its field-range, the swampland dis-
tance conjecture [48] would limit ϕ0 ≲ O(1)MPl . In hilltop quintessence scenarios,
ϕ remains frozen for most of the cosmological history, allowing this constraint to be
relaxed.
18
These arguments do not work in the same straightforward way for axions. E.g. the axion’s remnant
discrete shift symmetry implies that axion-matter couplings are such that finite temperatures, and other
dynamical effects, typically only change the effective axion-mass and not the position of the minima in
axion potentials (see e.g. [45, 47]). Also, relaxing initial conditions away from the hilltop would lead to the
axion rolling down to a minimum whose vacuum energy is of similar magnitude to that at the maximum
and hence equally anthropically viable.
8
We focus on the cosmological aspects of the hilltop quintessence models, from the cos-
mological evolution and comparison with cosmological observations in the CMB, galaxy
surveys and type IA supernovae catalogues, to a discussion on the initial conditions and
implications for inflation and reheating. We aim to ascertain to what extent theoretical
and observational constraints might favour a specific model of hilltop quintessence, and
what insights observational constraints give into the microscopic parameters of stringy
hilltop quintessence models.
2.3 Cosmological equations for hilltop quintessence

We now set up the equations of motion that describe the background cosmological evolution
for the models of interest. We consider a universe whose dark energy (DE) component is
described microscopically by one of the hilltop quintessence fields introduced above, with
a canonical kinetic term and a scalar potential functional V (ϕ), minimally coupled to
gravity. We further include radiation and (dark) matter and – though we do not consider
the detailed string theory model building required to achieve it – we assume that they are
decoupled from the quintessence field. Given theoretical expectations and observational
prospects, we also allow for non-zero curvature of the 3D space slices at this stage19 .
The 4D FLRW metric with arbitrary curvature given is given by:
dr2

2 2 2 2
2 2 2

ds = −dt + a (t) + r dθ + sin θdφ , (2.6)
1 − kr2
where k = 0, ±1 denotes the curvature of the 3D slices. The energy momentum tensor
is described by a set of perfect fluids describing the radiation, matter, quintessence, and
effective “curvature fluid” components. The energy density and pressure, ρi , pi , for these
components are related by their equation of state parameter, wi , as:
pi = wi ρi , (2.7)
1
where i = r, m, ϕ, k runs over radiation, matter, quintessence, and curvature; wr = 3
and
wm = 0, whilst for the scalar:
ϕ̇2 ϕ̇2 pϕ
ρϕ = + V (ϕ) , pϕ = − V (ϕ) , wϕ = , (2.8)
2 2 ρϕ
and for the curvature component:
3k k 1
ρk = − , pk = , wk = − . (2.9)
a2 a2 3
We can now write down the cosmological equations of motion for this system, which are
19
For completeness, in Appendix C we collect the cosmological evolution including curvature.
9
given by (we set κ = 8πGN = 1 for now):
ρeff
H2 = , (2.10a)
3
ä ρeff
=− (1 + 3weff ) , (2.10b)
a 6
ϕ̈ = −3H ϕ̇ − Vϕ . (2.10c)
In the last equation, Vϕ ≡ ∂ϕ V . Moreover, we defined

X X
ρeff = ρi , peff = pi , peff = weff ρeff . (2.11)
i i
From this definition it is clear that

X
weff = w i Ωi , (2.12)
i
where
ρi
.
Ωi = (2.13)
3H 2
Moreover, from (2.10b), we learn that acceleration requires weff < −1/3.
3 Parameterisation of the equation of state for hilltop models

In all hilltop quintessence models, the quintessence field is initially frozen by Hubble friction
close to the hilltop and starts to slowly roll, as the Universe expands and Hubble friction
falls, in recent times. All hilltop quintessence potentials are therefore well-approximated
for the full cosmological history by their behaviour close to the hilltop, i.e. their Taylor
expansion around the maximum up to second order. It is therefore not surprising that
all hilltop models can be described in a universal way. In fact, Dutta and Scherrer have
derived in [19] a parameterisation of the equation of state parameter for hilltop quintessence
models, and Chiba has shown in [20] that this parameterisation can actually be extended
to more general thawing quintessence models.
In this section, we first outline the derivation of the Dutta-Scherrer(-Chiba) (DS(Ch))
parameterisation and then test how it fairs for the specific hilltop models in our focus,
showing that it performs much better than the more commonly used CPL parametersation
[21, 22]. For comparison, we also test the DSCh parameterisation against a quintessence
model without hilltop, specifically with an exponential potential. Finally, we use the DS
parameterisation to obtain a bound on the initial displacement from the hilltop, which,
as expected, depends on the curvature at the hilltop. As we will show, this subsequently
leads to a bound on the scale of inflation or reheating.
10
3.1 Derivation of the Dutta-Scherrer(-Chiba) (DSCh) parameterisation
Dutta and Scherrer obtained their parameterisation of the equation of state parameter for
hilltop quintessence by computing the general solution to the scalar field in a flat universe
whose dynamics is dominated by the scalar and matter. We now outline the calculation
of DS [19], pointing out along the way how the derivation also applies in the presence of
curvature, so long as the curvature is subdominant, as it is in our observed universe. As
already mentioned, the beauty of the DS parameterisation is that it is analytically justified
for generic hilltop potentials, which can all be approximated by the same Taylor expansion
around the maximum as:
1
V (ϕ) ≈ V (ϕmax ) + V ′′ (ϕmax )(ϕ − ϕmax )2 . (3.1)
2
To begin, we consider the scalar field equation, (2.10c), in the background of matter and
the hilltop potential energy, which corresponds to an effective cosmological constant Λ =
V (ϕmax ). That is, we assume that the rolling of the scalar field away from its hilltop hardly
affects the overall background expansion of the Universe, and moreover, that radiation and
curvature are negligible during the epochs of interest. Note that the energy density from the
curvature grows more slowly than that from radiation and matter as one tracks backwards
in time, so if the curvature is subdominant today – as it is – then it was subdominant
throughout the history of the universe. Given our assumed background cosmology, which
is effectively a flat ΛCDM (i.e. neglecting the subdominant radiation, curvature and time-
varying quintessence), the expansion as a function of time is given by:
1/3
1 − ΩΛ,0

a(t)
= sinh2/3 (t/tΛ ) , (3.2)
a0 ΩΛ,0
where a0 is the scale-factor today, we denoted the present-day density parameter from the
hilltop potential energy as ΩΛ,0 , and defined
2 2
tΛ ≡ p =√ , (3.3)
3H0 ΩΛ,0 3Vmax
with Vmax ≡ V (ϕmax ). Let us now define a new variable u(t) as follows [19]:
u(t) = (ϕ − ϕmax )a3/2 (t) . (3.4)
In terms of u, the equation for ϕ (2.10c) becomes

3 ä 1 2
ü − u + H + a3/2 Vϕ = 0 . (3.5)
2 a 2
11
Using (2.10a) and (2.10b), this becomes
3
ü + u peff + a3/2 Vϕ = 0 , (3.6)
4
where remember that we are assuming that peff includes only the contribution from the
hilltop maximum (with wΛ = −1) and matter (with wm = 0) and therefore:
peff ≃ −Vmax . (3.7)
′′
Recalling the expansion (3.1), and writing Vmax ≡ V ′′ (ϕmax ), we have Vϕ = Vmax
′′
(ϕ − ϕmax ).
Therefore, (3.6) can be reduced to:

′′ 3
ü + u Vmax − Vmax = 0 . (3.8)
4
Defining r
3 ′′
kV ≡ Vmax − Vmax , (3.9)
4
the general solution to (3.8) is given by
u(t) = A sinh(kV t) + B cosh(kV t) . (3.10)
To fix the integration constants, consider an initial, finite value for ϕ at t = 0:
ϕ(0) ≡ ϕi ≡ ϕmax + ∆ϕi . (3.11)
Then (3.4) together with (3.2) requires B = 0 in (3.10), and also fixes the value of A in
terms of ∆ϕi , giving the final solution for ϕ = ϕmax + u/a3/2 :
∆ϕi sinh(kV t)
ϕ = ϕmax + , (3.12)
K sinh(t/tΛ )
where we defined
K ≡ kV tΛ . (3.13)
The next step is to find an expression for the equation of state parameter associated to
ϕ as a function of a, wϕ (a). First notice that
ϕ̇2 ϕ̇2
1 + wϕ = ≃ , (3.14)
ρϕ Vmax
where we used the approximation that ρϕ ≃ Vmax . Normalising this expression to the
present-day value of wϕ , denoted by w0 , and using the solutions for ϕ and the scale factor
12
a, eqs. (3.12) and (3.2), one arrives at the final expression as in [19]:
3(K−1) 2
(K − F (a))(1 + F (a))K + (K + F (a))(F (a) − 1)K

1 + wϕ (a) a
= , (3.15)
1 + w0 a0 (K − F0 )(1 + F0 )K + (K + F0 )(F0 − 1)K
where, approximating ΩΛ,0 ≃ Ωϕ,0 :

r
′′
4 Vmax
K = 1− , (3.16a)
3 Vmax
s −3
1 − Ωϕ,0

a
F (a) ≡ 1 + , F0 = F (a0 ) . (3.16b)
a0 Ωϕ,0
We see that the parameterisation involves two free parameters: K, which depends on the
curvature of the potential around the hilltop as in (3.16a); and w0 , which depends on K
and the initial displacement of the scalar field from its maximum, ∆ϕi , as:
 p !K p !K 2
3 ∆ϕ2i (1 − Ωϕ,0 )  1 + Ωϕ,0 1 − Ωϕ,0
1 + w0 = (K − F0 ) p + (K + F0 ) p  .
16 K 2 Ωϕ,0 1 − Ωϕ,0 1 − Ωϕ,0
(3.17)
Chiba’s generalization of (3.15) [20] purports to extend the validity of the parameter-
isation beyond hilltops to general thawing quintessence models by allowing the field to
start at some arbitrary initial value, ϕi , and keeping all terms in the Taylor expansion of
the potential around V (ϕi ) up to second order (c.f. (3.1)). Formally, this parameterisation
ends up coinciding with (3.15), but with K in (3.16a) defined via the initial values of Vi′′
and Vi , rather than their values at the maximum. On the other hand, it is not obvious
why – if both first and second order terms in the Taylor expansion are significant – third
order terms and beyond can be neglected.
It is interesting to compare the DSCh parameterisation with the commonly used linear
CPL parameterisation [21, 22], with parameters w0 and wa :
wϕ (a) = w0 + (1 − a)wa , (3.18)
which – though it lacks the analytical justification of the DS parameterisation for hill-
tops – is based on a Taylor expansion of the equation of state parameter itself, where
the leading term is the linear one. In the following, we test the DS(Ch) (3.15) and CPL
(3.18) parameterisations against both hilltop models and the exponential runaway po-
tential, using the modification by Chiba [20] for the latter. A comparison of different
phenomenological parameterisations was performed in [49] for a particular hilltop model
with V (ϕ) = V0 e−cϕ (1+αϕ), while the DSCh parameterisation was analysed against recent
13
data in [50].
3.2 Testing the DS parameterisation for hilltop quintessence models

In this section, we test the DS [19] parameterisation against the explicit hilltop models
presented in Section 2. For comparison, in the next subsection 3.3, we test the parameteri-
sation also against the exponential dark energy model, using the Chiba [20] generalisation.
We start by checking how the parameter K in the DS parameterisation is related to the
more fundamental parameters in the different models that we are interested in analysing:
1. Axion hilltops. The scalar potential is given by (2.1) and for hilltop quintessence,
we are interested in the case f ≲ 1, which is also consistent with typical values found
in string theory and suggested by swampland constraints [37]. For this model, the
parameter K is given by: r
2
Kax = 1 + 2 . (3.19)
3f
p
Thus for f ≲ 1, Kax ≳ 5/3 ∼ 1.3. In Figure 1 we compare the true evolution of
the equation of state parameter obtained using CAMB with the DS and CPL param-
eterisations, for the best-fit value of the decay constant using DESI year one data
plus Union3 supernova data, f = 0.15 (see Table 6 in Appendix B), and for f = 0.5.
As we commented before, since the spatial curvature k is subdominant throughout
the cosmological evolution, adding a small non-zero k does not change the results.
We collect the evolution with non-zero spatial curvature in Appendix C.
As we can see from the Figure 1, the DS parameterisation works very well through
the full cosmological evolution, well beyond the reach of current and near-future Dark
Energy surveys; indeed, the derivation of the DSCh parameterisation suggests that
it should work as soon as radiation is negligible (and recall that zeq ≈ 3400). In the
figures, the evolution starts in the matter domination epoch from zm = 3000. For
the CPL parameterisation we fit the linear behaviour to obtain suitable values of
w0 , wa . It is very clear that the linear parameterisation (3.18) is not appropriate for
axion hilltop quintessence throughout the evolution; rather it works only for small
red-shifts as f decreases. We will see the same pattern also for the other hilltop
potentials below.
Finally, in Figure 2, we compare the evolution of the dark energy equation of state
for different initial conditions as indicated in the plot.
14
θi = (π − 0.01)f
−0.99997
f = 0.15 f = 0.5
DS: Ωφ0 = 0.680, w0 = −0.95 DS: Ωφ0 = 0.680, w0 = −0.99997
−0.96 CPL: w0 = −0.95, wa = −0.34 CPL: w0 = −0.99997, wa = −0.00007
wDE −0.99998
−0.98 −0.99999
−1.00 −1.00000
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
a a
Figure 1. Evolution of the equation of state parameter for the axion hilltop for different values
of f , and its comparison to the DS (3.15) and CPL (3.18) parameterisations. The initial value
for θ is given at the top of the plots and for the DS parameterisation we used the values of Ωϕ0
and w0 as indicated to the right, as obtained from the evolution with CAMB. The CPL parameters
are obtained by fitting the linear behaviour between a = 0.9 and a = 1.
f = 0.15
−0.95 θi = (π − 0.01)f
−0.96 θi = (π − 0.005)f
θi = (π − 0.0001)f
wDE
−0.97
−0.98
−0.99
−1.00
0.0 0.2 0.4 0.6 0.8 1.0
a
Figure 2. Evolution of the equation of state parameter for the axion hilltop for different initial
conditions θi .
2. Saxion hilltop. The potential of this model is given by (2.3) [42], where typical
values for α are 2πN
with, recall, N an integer that we assume to be N ≲ O(100).
Interestingly, for this model, K is independent of the potential parameters! Indeed,
s √
19 + 8 2
Ksugra = ≃ 3.179 , (3.20)
3
and thus the curvature at the maximum is independent of α and the evolution will
mostly be dependent on ϕi . To better understand the constraints on α in this case,
15
it is useful to write the potential as follows:
√ " 2 3 #
− 2

e ϕ ϕ
V = V0 α a0 − a2 κ2 ϕ2max − 1 + a3 κ3 ϕ3max − 1 + ... ,
3 ϕmax ϕmax
(3.21)
where we restored Planck units κ = 1/MPl , an are numerical constants independent
√
of α given by a0 = 12, a2 = 12(2 + 2) and a3 = 32, and recall that ϕmax is given
in terms of α by (2.4). Note that ϕmax can be positive or negative, depending on the
√
value of α (see eq. (2.4)). In particular for α ≥ 1/ 2, ϕmax ≤ 0. On the other hand,
ϕmax becomes super-Planckian for α ≲ 0.17 ∼ 2π 37
. As we mentioned before, α is also
constrained by ensuring theoretical control to α ≲ 0.4. Therefore in the next section
we will focus on α ∈ 2π 2π

,
32 9
.
From the expansion around the maximum (3.21), we also see that contrary to the
axion and field theory Higgs-like model, there is a non-zero cubic contribution. Thus,
for ϕ/ϕmax < 1, the cubic term changes sign, as the potential becomes steeper (and
unbounded) on the left hand side. (On the other hand, the axion and Higgs potentials
have the same curvature to either side). Due to the cubic contribution, we expect
the DS parameterisation to be a little less accurate in this case, compared to the
axion and Higgs.
In Figure 3 we compare the evolution of the equation of state with the DS and
CPL parameterisations, using the best-fit value for the parameter α obtained from
DESI year one data plus Union3 supernova data i.e. α = 0.37 ∼ 2π 17
(see Table 7
in Appendix B). From the figure we see that the DS parameterisation does slightly
worse here compared to the Axion (see Figure 1) and Higgs (see Figure 5) cases,
through the full cosmological evolution, whereas the linear parameterisation breaks
down rather quickly.
In Figure 4, we compare the evolution of the dark energy equation of state for the
saxion model, for different initial conditions.
16
φi = 0.5
−0.95
α = 0.37
−0.96 DS: Ωφ0 = 0.669, w0 = −0.95
CPL: w0 = −0.95, wa = −0.19
−0.97
wDE
−0.98
−0.99
−1.00
0.0 0.2 0.4 0.6 0.8 1.0
a
Figure 3. Evolution of the equation of state parameter for the supergravity hilltop and its
comparison to the DS (3.15) and CPL (3.18) parameterisations. The initial value for the saxion
is given at the top of the plot. For the DS parameterisation we used the values of Ωϕ0 , w0
obtained from the evolution with CAMB as indicated to the right, while we fit the linear behaviour
between a = 0.9 and a = 1 to obtain (w0 , wa ).
α = 0.37
φi = 1.001 φmax
φi = 1.05 φmax
−0.96 φi = 1.08 φmax
wDE
−0.98
−1.00
0.0 0.2 0.4 0.6 0.8 1.0
a
Figure 4. Evolution of the equation of state parameter for the supergravity hilltop for different
initial conditions ϕi .
3. Higgs-like hilltop. The potential for this model is given by (2.5), where recall that
in order to avoid issues with the numerical and cosmological analysis, we completed
the quadratic hilltop to a Higgs-like potential. For this potential, K is given by
s
16
KHiggs = 1 + 2 . (3.22)
3ϕ0
If we expect ϕ0 to be p
less than or at most one (recall we are using Planck units),
ϕ0 ≲ 1, then KHiggs ≳ 19/3 ∼ 2.5.
17
In Figure 5 we compare the evolution of the equation of state with the DS and CPL
parameterisations, for the best-fit value of ϕ0 = 0.69 using from DESI year one data
plus Union3 supernova data (see Table 8 in Appendix B) and for ϕ0 = 1.3. Again,
we learn that the DS parameterisation works very well through the full cosmological
evolution also in this example, whereas the CLP parameterisation works only for
smaller redshifts.
Finally in Figure 6 we compare the evolution of the equation of state for the field
theory model for different values of ϕi .
φi = 0.036φ0
−0.996
−0.92 φ0 = 0.69 φ0 = 1.3
DS: Ωφ0 = 0.671, w0 = −0.91 DS: Ωφ0 = 0.671, w0 = −0.996
CPL: w0 = −0.92, wa = −0.39 CPL: w0 = −0.996, wa = −0.011 −0.997
−0.94
wDE
−0.96 −0.998
−0.98 −0.999
−1.00 −1.000
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
a a
Figure 5. Evolution of the equation of state parameter for the Higgs-like hilltop (2.5) and its
comparison to the DS (3.15) and (3.18) parameterisations. For DS we used Ωϕ0 , w0 as obtained
from the evolution with CAMB, while we fitted the linear behaviour between a = 0.9 and a = 1
for the CPL to obtain (w0 , wa ) as indicated to the right. The initial value for the scalar field is
indicated at the top of the plot.
φ0 = 1.0
−0.90
φi = 0.001
−0.92 φi = 0.05
φi = 0.1
−0.94
wDE
−0.96
−0.98
−1.00
0.0 0.2 0.4 0.6 0.8 1.0
a
Figure 6. Evolution of the equation of state parameter for the Higgs-like hilltop for different
initial conditions, ϕi .
18
3.3 Comparison: testing DSCh parameterisation in exponential quintessence
For comparison, having verified the success of the DS parameterisation for various explicit
hilltop models, we now test how its generalisation by Chiba fares with a potential that
does not have a maximum. We do this using an exponential potential, whose cosmology
has recently been studied in [16, 17, 51]:
V = V0 e−λϕ . (3.23)
′′
Recall that the DSCh parameterisation corresponds to (3.15) where Vmax , Vmax in K are
′′ ′′
replaced by Vi = V (ϕi ), Vi = V (ϕi ). Though in general K will depend on ϕi , for the
exponential case, it is independent of it, becoming purely dependent on λ:
r
4λ2
Kexp = 1 − . (3.24)
3
2
Note that Kexp can be negative, giving rise to oscillatory behaviour
p in the parameterisa-
tion [20]. In this case, one should replace K → iK̃ with K̃ = 4Vi′′ /3Vi − 1 in (3.15).
2
√
Interestingly, Kexp < 0 for λ > 3/2 ∼ 0.866. In Figure 7 we compare the evolution of the
equation of state with the DSCh and CPL parameterisations for different values of λ. As
2
we can see from the comparison, the DS parameterisation works rather well for Kexp > 0,
2 2
but it does not do well for Kexp < 0. For Kexp > 0, the DSCh parameterisation works
2
out to larger redshifts than the linear CPL parameterisation, but for Kexp < 0 the CPL
parameterisation does better.
19
φi = 0
λ = 0.25 λ = 0.5
−0.9925 DSCh: Ωφ0 = 0.680, w0 = −0.9909 DSCh: Ωφ0 = 0.680, w0 = −0.96 −0.97
wDE CPL: w0 = −0.9909, wa = −0.0142 CPL: w0 = −0.96, wa = −0.06
−0.9950 −0.98
−0.9975 −0.99
−1.0000 −1.00
√ √
λ = 3/2 λ= 2 −0.7
−0.900
DSCh: Ωφ0 = 0.680, w0 = −0.889 DSCh: Ωφ0 = 0.680, w0 = −0.7
−0.925 CPL: w0 = −0.888, wa = −0.169 CPL: w0 = −0.7, wa = −0.4
−0.8
wDE
−0.950
−0.9
−0.975
−1.000 −1.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
a a
Figure 7. Evolution of the equation of state parameter for the exponential potential (3.1) and
its comparison to the DSCh and CPL parameterisations. For the latter we used Ωϕ0 , w0 obtained
from the evolution with CAMB, while for CPL we fitted the linear behaviour between a = 0.5 and
a = 1 to obtain the parameters, indicated to the right.
3.4 Bounds on – and from – the initial conditions

As we have anticipated, a problem particular to hilltop quintessence is the fine-tuning of
initial conditions close to the dS maximum. Intuitively, the higher the curvature of the
hilltop, the closer to the hilltop ϕ needs to start in order to drive cosmic acceleration.
Even if such initial conditions can be selected via some physical mechanism in the early
universe before big bang nucleosynthesis – e.g. dynamically or anthropically – they need to
subsequently survive quantum diffusion effects. The DS parameterisation provides us with
an analytical expression for the initial displacement from the top of the potential, ∆ϕi , in
terms of the curvature of the potential at the hilltop, K, the equation of state parameter
today, w0 , and the density parameter for quintessence today, Ωϕ,0 :
K−1
r
(1 + w0 ) (1 − Ωϕ,0 ) 2
∆ϕi = 4KΩϕ,0 K K .
3
p p p p
K Ωϕ,0 − 1 1 + Ωϕ,0 + K Ωϕ,0 + 1 1 − Ωϕ,0
(3.25)
Note that as w0 increases from −1 to 1, ∆ϕi also grows, and as Ωϕ,0 increases from 0 to 1,
∆ϕi decreases. Although the overall dependence on K is complex, it is clear that the larger
the value of K (function of the hilltop curvature in (3.16a)), the larger the curvature, and
therefore, the smaller initial displacement from the hilltop (smaller |∆ϕi |) is expected. We
20
make manifest such behaviour with Figure 8, where we plot the K-dependence of |∆ϕi | us-
ing the best-fit values for w0 and Ωϕ0 (see Table 9) from the cosmological analysis described
in Subsection 4.4, together with the derived 1σ and 2σ limits on the derived parameter
|∆ϕi |. In the same plot we include the predictions for different hilltop quintessence models.
0.06
0.04
|∆φi |
0.02
0.00
2 4 6 8 10 12
K
Figure 8. Analytic results for ∆ϕi obtained from eq. (3.25) for the hilltop quintessence models,
and posterior contours furnished by a MCMC analysis in the K-|∆ϕi | plane. The analytic results
are represented in dark grey line using best-fit values for Ωϕ,0 and w0 from the data combination
with Union3. Dark blue shapes indicate the points corresponding to the best-fit values (see Tables
6-8) for model parameters (ϕ0 and f ): circle for axion model, star for sugra model and triangle
for the field theory model. In the same figure, we show with blue contours the 1σ and 2σ bounds
in the K − |∆ϕi | plane from the constraints on the DS parameterisation for the data combination
with Union3. See Figure 19 in the appendix for the analogous figure including constraints from
all the data combinations.
The 2σ tail extending towards larger |∆ϕi | in Figure 8 is expected, since very small
curvature is still allowed when ϕi is far away from the hilltop, and, in fact, this tail follows
the analytical curve. This asymmetric part in 2σ eventually leads to the skewed posterior
distribution for small K values shown in Figure 12 in next section.
Once the best-fit value of the displacement from the hilltop inferred from the cosmolog-
ical data has been found, this in turn sets an upper bound on the Hubble scale at the end
of reheating, if we assume that the initial conditions were set up by this time. Indeed, for
quantum diffusion at around the time of reheating – and thus any time thereafter – not to
kick ϕ too far away from the value inferred from observations, we require:
Hrh ≪ 2π∆ϕi . (3.26)
21
For example, choosing the mean values for the DS parameterisation in Table 4 for the
data combination with Union3, we obtain an upper bound on the reheating scale, Hrh ≪
0.06MPl . We will obtain analog bounds in our model-by-model analyses below.
4 Cosmological analysis
We modify the cosmological Boltzmann code CAMB to implement the three hilltop mod-
els as well as the DS parameterisation described above. For each model, we perform
a Markov Chain Monte-Carlo (MCMC) analysis of the parameter space, varying two
model specific parameters (described below) alongside the baseline cosmological param-
eters {Ωb h2 , Ωc h2 , H0 , τ, As , ns } for which we adopt wide uniform priors. We make use of
the following datasets:
1. CMB from Planck 2018:
- Planck 2018 low-ℓ temperature and polarisation likelihood [52].

- Planck high-ℓ CamSpec TTTEEE temperature and polarization likelihood using
NPIPE (Planck PR4) data [53].
- Planck 2018 lensing likelihood [54].
Hereafter, we collectively refer to all the Planck CMB likelihoods as ‘CMB’.
2. BAO likelihoods from DESI DR1 [28, 55, 56]
3. Pantheon+ [57], Union3 [58] and DES-Y5 [27] type Ia supernovae likelihoods.
We sample the likelihoods using the MCMC sampler [59, 60], provided in Cobaya [61]. Our
convergence criteria for the MCMC chains is reached at the value R − 1 = 0.02 for the
Gelman-Rubin diagnostic. The constraints and posterior distribution plots for each model
are generated using the GetDist package [62]. We also run the Py-BOBYQA [63, 64] mini-
mizer via Cobaya to obtain the maximum likelihood point and the corresponding χ2 values.
Recent results from the DESI BAO analysis [28], alone as well as when combined with
supernovae data from Pantheon+ [57], Union3 [58] and DESY5 [27], exhibit a preference
for dynamical dark energy with a fairly rapid evolution in the recent past [28, 65, 66].
The significance of this deviation from ΛCDM ranges from 2 − 4σ, depending upon the
supernovae dataset chosen.20 When it comes to the quintessence models considered here,
these datasets allow us to provide constraints on the underlying model parameters as well
as test whether these models can provide a better fit to the data compared to ΛCDM, or
to the CPL parameterisation.
20
Different interpretations of these results as well as their various cosmological implications are also
discussed in [67–84].
22
4.1 Axion hilltop
For the axion model discussed in Section 2.2 we sample the axion decay constant f and the
initial field value θi rescaled by f , i.e. θi /f , for f < 2. The results are plotted in Figure
9 and 68% limits summarised in Table 1 for the parameters21 {f, θi /f, Ωb h2 , Ωc h2 , H0 }.
Focusing on the 1D marginalised constraints for the parameter f , we notice a preference
for larger (f ≳ 1) values in the Pantheon+ dataset, which decreases progressively as we
change the supernovae dataset to Union3 or DESY5. In addition, from the constraints in
the (f -θi /f ) plane using the DESY5 SN dataset, we learn that for smaller f , the allowed
values of θi /f are squeezed to a small region around π, while for larger f the region around
θi /f = π is excluded. This happens because for larger f (smaller slope), one has to start
farther away from the hilltop (maxima) to obtain dynamical dark energy at the present
epoch. These effects are much less pronounced for the other SN datasets, reflecting the fact
that these do not deviate from ΛCDM as much as DESY5. In other words, for the DESY5
dataset, the preferred field evolution in the axion model requires either the field starting
far away from the maximum, if f is large, or the field starting close to the maximum, if f
is small.
Table 1 indicates that the combined data sets give a lower bound on f at around f ≳ 0.7
(68% C.L), and including the DESY5 data gives a mean value f = 0.88+0.24 −0.54 . Note that the
lower-limits and means presented here derive from a Bayesian analysis of the model against
the cosmological data, thus they are highly prior-dependent. As we can see in both Figure
9 and Table 1, the data are not particularly constraining when it comes to the parameters
f, θi /f and – as long as this is the case – the prior dependence of the limits is expected
to remain. On the other hand, refinement of the theory priors on f (or even θi ) will lead
to tighter constraints on these parameters. If evidence for dynamical dark energy persists,
the parameter region f > 1 with θi close to the maximum will be strongly disfavoured.
Motivated by the swampland constraints discussed in Section 2, which suggest f ≲
O(1)MPl , we have assumed the prior that f < 2MPl , restoring Planck units. It would
be important to refine the order one constants that appear in the swampland constraints.
Nevertheless, the values of f favoured by our analysis are rather large from the string
theory point of view and they could only be pushed further up by extending the priors
to allow larger values for f , until the data is sufficiently constraining to make the fits
prior-independent. In particular, for such large values of22 f ≳ 0.7MPl , the weak gravity
conjecture (2.2) implies that the instanton that generates the scalar potential has action
Sinst ≲ 1.4. This means that an additional source of exponential suppression is needed to
21
For each model studied in this section, the results for the full set of parameters including the best-fit
parameter combinations are presented in Appendix B. In this section, we focus mainly on the dark energy
model specific and the cosmological background parameters as, in any case, the constraints on the other
cosmological parameters do not differ significantly across the different models or the different datasets.
22
Large values of f also imply large values for θi in Planck units. The distance conjecture [48] generally
puts super-Planckian field ranges in the swampland, but, in our hilltop scenario, the field range actually
explored would be small due to Hubble friction.
23
achieve the hierarchically small scale of Dark Energy – e.g. in the form of polyinstantons
as discussed in [85, 86]. Recalling also that control of the instanton expansion requires
Sinst ≳ 1, theoretical and observational constraints combine to give a very narrow window
of possibilities that would have to rely on numerical control.
Finally, it is interesting to consider the implications of the observational constraints on
the initial conditions for the inflationary or reheating scale, by demanding that the initial
conditions are safe from quantum diffusion at the time of reheating. Following Section 3.4,
and using the mean values for θi and f from the CMB+DESI+DESY5 data in Table 1 for
illustration, we find, restoring MPl :
Hrh ≪ 2π(πf − θi ) = 4.1MPl , (4.1)
that is, there is no effective constraint on the reheating scale.

Note that the axion model has been analysed previously in [87], using the cosmological
data available at that time. Our results with the new datasets are in agreement with their
results, with the slight differences in the constraints on the axion parameters mainly driven
by the new datasets.
24
CMB+DESI+Pantheon+
CMB+DESI+Union3
CMB+DESI+DESY5
3.0
2.6
θi /f
2.2
0.0225
Ωb h 2
0.0220
0.120
Ωc h 2
0.118
0.116
68
67
H0
66
65
0.5 1.0 1.5 2.2 2.6 3.0 0.0220 0.0225 0.117 0.120 65 66 67 68
f θi /f Ωb h 2 Ωc h 2 H0
Figure 9. Parameter constraints on the Axion model, eq. (2.1) (68% and 95% contours).
Parameter +Pantheon+ +Union3 +DESY5

f > 0.946 > 0.779 0.88+0.24
−0.54
θi /f — — < 2.62
Ωc h2 0.11842 ± 0.00081 0.11842 ± 0.00083 0.11847 ± 0.00086
Ωb h2 0.02227 ± 0.00012 0.02227 ± 0.00013 0.02227 ± 0.00013
H0 67.49+0.51
−0.37 67.23+0.81
−0.40 66.79+0.74
−0.62
θi 3.1+1.1
−1.4 2.73+0.93
−1.6 2.11+0.40
−1.2
Table 1. Axion model: parameter means and 68% limits for the addition of the different
supernovae datasets to the CMB+DESI combination.
4.2 Saxion hilltop

For the saxion model we sample the parameter α ∈ 2π 2π

,
32 9
, consistent with our theory
discussion in Section 2, and the initial field value ϕi , rescaled by ϕmax , i.e. ϕi /ϕmax . The
25
results are plotted in Figure 10 and the 68% limits summarised in Table 2 for the parameters
{α, ϕi /ϕmax , Ωb h2 , Ωc h2 , H0 }.
We interpret these parameter constraints in terms of the series expansion of the Saxion
potential in (3.21), where we learn that α effectively plays the role of the potential’s
curvature through the term ϕmax . Since the field value at the potential maximum is a
function of α (2.4), the potential has a higher curvature for smaller α and lower curvature
for larger α. Thus, for smaller α the parameter ϕi /ϕmax is constrained to be close to
one. This manifests as a somewhat curved degeneracy in the α–ϕi /ϕmax plane in Figure
10. Focusing in particular on the DESY5 contours (green), whilst ϕi /ϕmax moves towards
smaller values as α decreases, it cannot be too close to 1 as otherwise one would not obtain
dynamical dark energy at the present epoch, as is preferred by the DESY5 data. Although
current data do not put any tight constraint on the fundamental parameter α, if e.g. ϕi
is forced to be closer to the hilltop and upcoming surveys continue to see evidence for
dynamical dark energy, the preferred region for α would be driven towards lower values.
Finally, we note the constraints on the reheating scale, by demanding that the initial
conditions are safe from quantum diffusion at that time. Following Section 3.4, and using
the mean values for ϕi and α from the CMB+DESI+DESY5 data in Table 2 for illustration,
we find, restoring MPl :

1 1
Hrh ≪ 2π ϕi − √ log √ = 0.38MPl . (4.2)
2 2α
26
CMB+DESI+Pantheon+
CMB+DESI+Union3
CMB+DESI+DESY5
1.4
1.3
φi /φmax
1.2
1.1
0.0226
0.0224
Ωb h 2
0.0222
0.0220
0.120
Ωc h 2
0.118
0.116
68
67
H0
66
0.3 0.4 0.5 0.6 1.1 1.2 1.3 1.4 0.0220 0.0225 0.117 0.120 66 67 68
α φi /φmax Ωb h 2 Ωc h 2 H0
Figure 10. Parameter constraints (68% and 95% contours) for the Saxion model (2.3).

α > 0.521 0.537+0.16
−0.045 0.49 ± 0.11
ϕi /ϕmax < 1.23 1.201+0.071
−0.17 1.205+0.058
−0.14
Ωc h2 0.11838 ± 0.00084 0.11842 ± 0.00083 0.11846 ± 0.00085

Ωb h2 0.02228 ± 0.00013 0.02228 ± 0.00013 0.02228 ± 0.00012
H0 67.44+0.49
−0.43 67.21+0.62
−0.55 66.86 ± 0.52
ϕi 0.209+0.064
−0.18 0.241+0.083
−0.21 0.32+0.13
−0.22
Table 2. Saxion model: parameter means and 68% limits for the addition of the different
supernovae datasets to the CMB+DESI combination. The full set of constraints can be found in
Table 7 and corresponding plots in Figure 16.
27
4.3 Higgs-like hilltop
For the Higgs-like hilltop23 , we vary the parameter ϕ0 and the initial relative field value
ϕi /ϕ0 . In fact these two quantities play a role similar to f , θi /f in the Axion model,
representing the steepness of the potential and the initial displacement from the maximum.
The resulting parameter posterior distributions can also be interpreted similarly and are
plotted in Figure 11, with the 68% limits summarised in Table 3. In particular, we notice
the same squeezing of the allowed ϕi when the potential is steep, i.e. for smaller ϕ0 , as
found in the axion model of Section 4.1. The limits from the current cosmological data are
strongly prior dependent and once again, a more precise understanding of the theoretical
priors on ϕ0 and ϕi will be crucial to constraining the parameter space of this model.
Finally, we estimate the constraints on the reheating scale from quantum diffusion and
the degree of fine-tuning in the initial conditions that the data indicates. Following Section
3.4, and using the mean value for ϕi from the CMB+DESI+DESY5 data in Table 3 for
illustration, we find, restoring MPl :
Hrh ≪ 2πϕi = 1.6MPl , (4.3)
that is, no effective constraint on the reheating scale.

23
See [88] for a cosmological analysis of a pure quadratic hilltop.
28
CMB+DESI+Pantheon+
CMB+DESI+Union3
CMB+DESI+DESY5
0.3
φi /φ0 0.2
0.1
0.0226
0.0224
Ωb h 2
0.0222
0.0220
0.120
Ωc h 2
0.118
0.116
68
H0
66
64
0.5 1.0 1.5 0.1 0.2 0.3 0.0220 0.0225 0.117 0.120 64 66 68
φ0 φi /φ0 Ωb h 2 Ωc h 2 H0
Figure 11. Constraints on the Higgs-like hilltop model (2.5) (68% and 95% contours).

ϕ0 > 1.29 > 1.24 > 1.17
ϕi /ϕ0 < 0.142 0.151+0.073
−0.12 0.169 ± 0.081
Ωc h2 0.11838 ± 0.00082 0.11835 ± 0.00084 0.11829 ± 0.00084
Ωb h2 0.02228 ± 0.00012 0.02228 ± 0.00013 0.02228 ± 0.00013
H0 67.29+0.59
−0.45 66.7+1.0
−0.70 66.44 ± 0.64
ϕi 0.174+0.071
−0.17 0.235+0.088
−0.23 0.26+0.10
−0.24
Table 3. Higgs-like hilltop model: parameter means and 68% limits for the addition of the
different supernovae datasets to the CMB+DESI combination.
4.4 DS parameterisation analysis

For the DS parameterisation, we vary the parameters K and w0 along with the other six
CDM parameters as in the previous sections. The results for the posterior probability
29
distributions are plotted in Figure 12 and the corresponding 68% limits are presented in
the Table 4. The full set of posterior distributions and corresponding limits are reported
in Figure 18 and Table 9, respectively. The 1D posterior distributions of K for all data
combinations are skewed towards smaller values – as expected since smaller curvature is
required when the initial field value is far from the hilltop (see also the discussion around
Figure 8). For what concerns the combination with Pantheon+ results, the large K tail of
the 1D posterior is a consequence of the condition w0 ≃ −1.0 being preferred when this
data set is added24 .
Although a one-to-one match among the posteriors of f and ϕ0 with that of K is
not expected due to the non-linear relation between these parameters (3.19), and (3.22)
– as well as the complicated mapping between ϕi and w0 – the values of K obtained
directly from f and ϕ0 can nevertheless be compared with the posterior of K from DS
parameterisation. Using the mean values for f and ϕ0 from Tables 1 and 3 (for +Union3)
mean mean
leads to Kax < 1.45, and KHiggs < 2.11, while Ksugra = 3.179 is independent of model
25
parameters . However, K in DS parameterisation is not very tightly constrained for any
data combination, as compared to its prior range, allowing for the values of K from the
individual model constraints to lie within 1σ – 3σ of their DS counterparts, depending on
the data combination under consideration. This can also be clearly seen in Figure 8 in the
relative position of the dots (for model predictions) with respect to the shaded 1σ and 2σ
regions in the K-|∆ϕi | plane.
A similar analysis for the DS parameterisation can be found in [50], where a number
of parameters including K have been varied, while CDM parameters were kept fixed at
CMB values. This analysis obtained even larger values for K, albeit with large error. Our
results and those in [50] are consistent within 2σ error.
Finally, we note that w0 itself is also poorly constrained. Due to the weak constraints
on the DS parameters w0 , K, the derived constraints on |∆ϕi | shown in Figure 8 also allow
for a large range of values. Since wϕ is best constrained at around z ≈ 0.4 – as manifest
also from the DESI reconstruction plotted in Figure 13 below – we also present the derived
constraints on wϕ (z = 0.4) in Figure 20 in Appendix B as 2D posterior plots from the
analysis of DS parameterisation. In particular, the w0 -wϕ (z = 0.4) plot demonstrates that
the constraints on wϕ (z = 0.4) are improved by at least an order of magnitude as compared
to the constraints on w0 only.
24
Our prior range for w0 excludes the w0 < −1.0 region.
25 best−fit
Using best-fit values for f and ϕ0 from Tables 6 and 8 respectively (for +Union3), viz, Kax =
best−fit
5.3685, and KHiggs = 3.4765.
30
K 7.6+2.5
−2.1 8.2+2.7
−1.4 8.4+2.4
−1.1
w0 < −0.709 −0.11+0.39

−0.61 −0.35+0.24
−0.33
Ωc h2 0.11830 ± 0.00081 0.11828 ± 0.00086 0.11825 ± 0.00082

Ωb h2 0.02226 ± 0.00013 0.02227 ± 0.00013 0.02227 ± 0.00013
H0 67.53 ± 0.39 66.73 ± 0.59 67.04 ± 0.42
Table 4. DS parameterisation: parameter means and 68% limits for the addition of the different
CMB+DESI+Pantheon+
CMB+DESI+Union3
CMB+DESI+DESY5
0.5
w0
0.0
−0.5
0.120
Ωc h 2
0.118
0.116
68
67
H0
66
65
0.0226
0.0224
Ωb h 2
0.0222
0.0220
2 6 10 14 −0.6 0.0 0.6 0.117 0.120 65 66 67 68 0.0220 0.0225

K w0 Ωc h 2 H0 Ωb h 2
Figure 12. DS parameterisation: parameter means and limits for the addition of the different
31
4.5 Model comparison
To compare the quality of the fit to the data provided by the different models, we perform
a model comparison based on the Akaike information criterion (AIC) [89, 90], which also
takes into account the number of free parameters of the model. The AIC value is defined
as follows
AIC = 2n − 2 ln Lmax , (4.4)
where Lmax stands for the maximum likelihood value in the model and n the number of
free parameters. Models with smaller AIC are favoured by the data with the best model
having the lowest AIC value. To compare between models, one looks at the difference in
AIC values with ∆AIC12 ≡ AIC1 − AIC2 ≲ 2 indicating no preference between Model 1
and Model 2, whilst ∆AIC12 ≳ 5 indicates a strong preference for Model 2 over Model 1.
The AIC values for the different models we have studied are provided in Table 5,
where we also consider ΛCDM, the CPL parameterisation and the exponential runaway
quintessence model. We learn that the preference for the hilltop quintessence models
over ΛCDM is strongest for the dataset combination CMB+DESI+DESY5 and weakest
for CMB+DESI+Pantheon+. Out of all the different models studied here, the CPL pa-
rameterisation remains the most favoured, irrespective of the dataset combination chosen.
However, its improvement with respect to the DS parameterisation is data dependent,
with best improvement for +Union3 (∆AICDS,CPL ≃ 8.3) and only mild improvement for
+DESY5 (∆AICDS,CPL ≃ 3.4). We attribute this improvement to the more rapid evolution
of the dark energy equation of state in this model, as well as the phantom-like behaviour in
the past [28], which matches very well the DESI reconstruction of background quantities
w(z) and h(z) ≡ H(z)/H0 , shown in Figure 13. As is clear from Figure 13, these two
features cannot be produced in the hilltop quintessence models. On the other hand, the
redshift evolution seen here for the hilltop quintessence models is much closer to the DESI
reconstruction than what can be produced in the exponential potential model (compare
Figure 10 of [16]). At the same time, Table 5 tells us that the overall improvement in
the fit compared to the exponential model is not significant enough to compensate for the
additional parameter introduced by the hilltop models.
We also notice that the DS parameterisation fares (marginally) better than the concrete
hilltop quintessence models we considered, with the greatest improvement for +DESY5
with ∆AICmodel,DS ≃ 4, and the lowest improvement for +Pantheon+ with ∆AICmodel,DS ≲
0.6. One can then also adopt the approach to start from the DS parameterisation, compare
it with data, and infer the preferred values for the curvature parameter K. We can then
use these results to deduce a preference for specific hilltop scenarios, given that each
model favours its own preferred range of K, depending on each model parameters. The
evolution of the background parameters plotted in Figure 13 shows clearly that the DS
parameterisation tracks the predictions from hilltop quintessence models well until very
32
low redshift, z ≳ 0.1, beyond which it leads to a faster evolution of w(z).
AIC Axion Sugra Higgs DS ΛCDM CPL Exp

CMB+DESI+Pantheon+ 12409.55 12409.40 12409.07 12408.9 12406.04 12401.70 12407.19
CMB+DESI+Union3 11030.07 11029.49 11030.38 11027.9 11028.69 11019.62 11029.00
CMB+DESI+DESY5 12644.67 12645.65 12644.89 12641.2 12649.01 12637.79 12644.73
Table 5. Comparison between the different models considered in this section.
0.0 1.10
Sugra Higgs
Axion DS
−0.5 1.05
h(z)/hLCDM
w(z)
−1.0
1.00
−1.5
0.95
−2.0
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
z z
Figure 13. The quantities wϕ (z), and h(z) ≡ H(z)/H0 are plotted for the best-fit hilltop models
as indicated, and compared to the DESI reconstruction (blue line) using CMB+DESI+Union3
data [65]. The CPL parameterisation matches the DESI reconstruction. The shaded regions
represent the 95% confidence regions. At z = 0, w0 = −0.65 for the reconstruction, while the
values for the hilltop models are as follows w0 = −0.77 for the axion model, w0 = −0.83, for
the saxion and w0 = −0.79 for the Higgs-like hilltop. For comparison, w0exp = −0.89 for the
exponential model (see Figure 10 of [16]). For the DS parameterisation, the best-fit value is
w0DS = −0.60. The best-fit parameter values for each model can be found in Appendix B.
5 Outlook and Future Challenges

Embedding models of dark energy in quantum gravity and string theory is a notoriously
difficult task, given stringent theoretical constraints on model building associated with
quantum gravity conjectures. At the same time, recent and forthcoming cosmological
results are going to probe the behaviour of dark energy with an ever-increasing degree
of precision, offering the concrete possibility to test our understanding of dark energy in
quantum gravity with cosmological data.
With the aim of exploiting such opportunities, in this work we considered a class of
dark energy models – hilltop quintessence – which is able to satisfy theoretical bounds
from de Sitter and other quantum gravity conjectures. We examined various realisations of
quintessence hilltops based on axions, their supersymmetric partners, and Higgs-like string
embeddings. Axion hilltops are widely considered and particularly well-motivated, as their
33
shift symmetry evades problems with UV sensitivities, fifth-forces and time-variation of
fundamental constants, and a priori – at the hilltop – they can be consistent with the weak
gravity conjecture. Moreover, we have presented a dynamical mechanism by which hilltop
initial conditions could be set up.
We studied the cosmological consequences of our three string-motivated hilltop mod-
els and discussed a convenient parameterisation of their associated equation of state. We
then tested their predictions by means of a Bayesian MCMC analysis with recent CMB,
galaxy surveys, and supernova data. We showed to what extent current data can distin-
guish amongst hilltop models and impose constraints on their parameters. Interestingly,
such experimental results are complementary to theoretical bounds from quantum gravity
conjectures, and we discussed the consequences of these features for our current under-
standing of dark energy in string theory. Notably, observational constraints on the axion
decay constant for axion hilltop quintessence are in tension with the weak gravity conjec-
ture, illustrating how synergies between constraints from observations and from quantum
gravity can rule out otherwise well-motivated models. So far, model comparisons favour
the CPL parameterisation over any of our hilltops, ΛCDM, and exponential runaways.26
However CPL is only mildly improved with respect to the DS parametrisation, especially
for the dataset CMB+DESI+DESY5. At the same time, the limited constraining power
of current data means that the model parameter constraints and comparisons that we
obtained are sensitive to our priors, which are based on theoretical assumptions about vi-
able regions of parameter space. Consequently, given that our inference from cosmological
data strongly depends on theoretical assumptions, it is imperative to refine our theoreti-
cal understanding of the priors, so as to maximize the informative power of current and
forthcoming cosmological datasets for testing dark energy scenarios in quantum gravity.
Assumptions entering into theoretical priors constrain the size and location of the al-
lowed region in parameter space associated with a given model. This includes the possible
values of parameters entering in the model Lagrangian and the allowed initial conditions
or field ranges associated with the dynamics of the quintessence scalar. Such information
can be theoretically refined by better specifying the ‘order one’ constants27 entering the
de Sitter conjecture in the inequalities (1.1) and the weak gravity conjecture in (2.2), or
by embedding hilltop quintessence into more complete early universe models, able to accu-
rately specify their initial conditions. Also, the range of allowed priors can be limited by
enriching the hilltop models to include additional Standard Model matter fields: then, one
should take into account further constraints on the parameter space from limits on fifth
forces and the time variation of fundamental constants. All these theoretical questions
can be addressed by developments of the current theoretical tools at our disposal. We be-
26
See e.g. [91] for possible maps of the CPL parameterisation to physical quintessence or barotropic
dark energy models, though no quintessence potential will give rise to the phantom behaviour seen in
Figure 13.
27
See e.g. [13, 92] for work towards fixing these order one constants using dimensional reduction.
34
lieve that such questions are very timely, and addressing them will allow us to exploit the
synergy between theory and observations offered by current and forthcoming cosmological
probes. We look forward to continuing this analysis in forthcoming publications.
Acknowledgements
The work of GB, AM, GT and IZ is partially funded by STFC grant ST/X000648/1
and the work of SP is partially funded by STFC grant ST/X000699/1. SB acknowl-
edges the “Consolidación Investigadora” grant CNS2022-135590 and her work is partially
supported by the Spanish Research Agency (Agencia Estatal de Investigación) through
the Grant IFT Centro de Excelencia Severo Ochoa No CEX2020-001007-S, funded by
MCIN/AEI/10.13039/501100011033. We also acknowledge the support of the Supercom-
puting Wales project, which is part-funded by the European Regional Development Fund
(ERDF) via Welsh Government. For the purpose of open access, the authors have applied a
Creative Commons Attribution licence to any Author Accepted Manuscript version arising.
Research Data Access Statement: No new data were generated for this manuscript.
35
A Saxion-axion stringy hilltops
In this appendix we summarise the saxionic hilltop potential we consider in the main
text (2.3). More details can be found in [42] (see also [14]). We also provide a concrete
example where the minima of stringy axions can become maxima upon turning on further
subleading corrections, thus providing a possible dynamical mechanism that tunes the
initial conditions of the axion to the hilltop.
A.1 Saxion hilltops

The model considers some inflationary early Universe scenario that ends in a supersym-
metric Minkowski minimum, where most of the string moduli are stabilised and heavy,
except for a flat direction corresponding to a chiral superfield Φ. The Kähler potential for
such flat direction takes the form:

K = −n ln Φ + Φ̄ , (A.1)
where n takes different values depending on the type of modulus. As we will see, we
are interested in n = 1, which may correspond to some (non-overall) volume modulus,
or a complex structure modulus, or the dilaton. The superpotential is given by non-
perturbative effects which lift the flat direction at some scale before BBN. The leading
term in the non-perturbative superpotential is then given by
Wnp = Ae−α Φ . (A.2)
Here α is a constant that can arise from different instanton types or gaugino condensation
and can e.g. take values α = 2π/N , with N = 1 for an Euclidean D3-brane instanton and
N ≥ 2 for gaugino condensation with condensing group rank SU (N ). The scalar potential
can then be computed using the supergravity formula
h i
K ij̄ 2
V = e K Dj W Dj̄ W̄ − 3|W | , (A.3)
where Dj W ≡ ∂j W + Kj W . Writing the complex scalar field component of the chiral

superfield as Φ = φ + iθ, we obtain:
A2 −2 α φ −n 2 2 2

V = e φ n + n(−3 + 4αφ) + 4α φ , (A.4)
2n n
which, for n = 1, has a dS maximum at:
1
φmax = √ . (A.5)
2α
Notice that at the leading order considered, the axion θ remains a flat direction, but
36
it will be lifted by subleading non-perturbative terms [42]. Indeed, adding a subleading
contribution
Wnp sub = Be−βΦ (A.6)
to the leading contribution (A.2), will generate a minimum for the axion, whilst preserving
the (slightly shifted) maximum in the saxion direction, with |m2axion | < |m2saxion |. For
2mπ
example for n = 1, β = 2α, B = −A/20, the axion has a minimum at θ = β−α ,m∈Z
[42]. Note that the exponential suppression of the subleading non-perturbative term is only
√
by a factor 2 at the hilltop, so control of the expansion in non-perturbative effects can
be at best numerical there. It is useful to express the potential in terms of the canonically
normalised field, which for n = 1, is
r
1
ϕ = MPl log φ . (A.7)
2
The final potential becomes (2.3).
A.2 Subleading corrections to the axion

Let us focus on the axion in the model considered above. As we saw, the axion is lifted
after adding subleading correction (A.6). Now further subleading corrections of the form
Wsub sub = Ce−γ Φ , (A.8)
may turn the minimum of the axion potential at C = 0 into a maximum when C ̸= 0.
This will happen for suitable values of the parameters α, β, γ, and A, B, C. Schematically,
the potential (A.3) including the two subleading corrections, (A.6), (A.8) takes the form
V = g(φ0 )+f1 (φ0 )AB cos ((β − α)θ)+f2 (φ0 )AC cos ((γ − α)θ)+f3 (φ0 )BC cos ((γ − β)θ) ,
(A.9)
where g(φ0 ), fi (φ0 ) are functions of the saxion (as well as α, β, γ) evaluated at its extremum.
For the case C = 0 and B < 0, as discussed above, the potential has a minimum at
2mπ
θmin = β−α . When C is turned on, this minimum can become a maximum for suitable
values of the parameters. For example, for n = 1 as above, α = 2π/16, β = 2α, γ = 3α,
B = −A/20, C = A/35 (A = 1 for concreteness), the minima for the axion at C = 0
become maxima for C ̸= 0. It is also possible that some minima stay minima, while only
some become maxima. Of course, more complex modulated structures can arise.
37
C=0
C=A/35
V(θ)
0 10 20 30
θ
Figure 14. Subleading corrections to axion potential (A.9) (in arbitrary units) as described in
the text for (A, B, C) = (1, −A/20, A/35), (α, β, γ) = (2π/16, 2α, 3α). The minima become a
maxima when the subleading correction is turned on.
38
B Constraints for all the parameters of our models
In this section we present the plots and tables with the full set of parameters for the
three hilltop models we considered in the main text, as well as the Dutta-Scherrer (DS)
parameterisation.
B.1 Axion hilltop
CMB+DESI+Pantheon+
CMB+DESI+Union3
CMB+DESI+DESY5
3.0
2.6
θi/f
2.2
4
θi
0.0225
Ωb h2
0.0220
0.120
Ωc h2
0.118
0.116
68
67
H0
66
65
3.08
log(1010As)
3.04
3.00
0.97
ns
0.96
0.07
0.06
τreio
0.05
0.04
0.5 1.0 1.5 2.2 2.6 3.0 2 4 6 0.022 0.117 0.120 65 66 67 68 3.04 3.08 0.96 0.97 0.04 0.06
f θi/f θi Ωb h2 Ωc h2 H0 log(1010As) ns τreio
Figure 15. The complete parameter constraints for the Axion model.
39
f > 0.946 (1.22) > 0.779 (0.15) 0.88+0.24
−0.54 (0.19)
θi /f — (2.25) — (3.12) < 2.62 (3.07)

Ωc h2 0.11842 ± 0.00081 (0.11184) 0.11842 ± 0.00083 (0.118) 0.11847 ± 0.00086 (0.118)
log(1010 As ) 3.045 ± 0.014 (3.046) 3.044 ± 0.015 (3.048) 3.045 ± 0.014 (3.053)
ns 0.9664 ± 0.0036 (0.9647) 0.9664 ± 0.0037 (0.9624) 0.9663 ± 0.0037 (0.967)
H0 67.49+0.51
−0.37 (67.35) 67.23+0.81
−0.40 (66.24) 66.79+0.74
−0.62 (65.95)
Ωb h2 0.02227 ± 0.00012 (0.02223) 0.02227 ± 0.00013 (0.0221) 0.02227 ± 0.00013 (0.0226)

τreio 0.0573+0.0066
−0.0074 (0.0592) 0.0568+0.0067
−0.0076 (0.0586) 0.0573 ± 0.0071 (0.061)
θi 3.1+1.1
−1.4 (2.75) 2.73+0.93
−1.6 (0.483) 2.11+0.40
−1.2 (0.591)
Table 6. Axion model: full parameter means and 68% limits for the addition of the different
supernovae datasets to the CMB+DESI combination. The values in parentheses denote the best-
fit parameters for this model.
40
B.2 Saxion hilltop
CMB+DESI+Pantheon+
CMB+DESI+Union3
CMB+DESI+DESY5
1.4
1.3
φi/φmax
1.2
1.1
0.8
0.4
φi
0.0
0.0225
Ωb h2
0.0220
0.120
Ωc h2
0.118
0.116
68
67
H0
66
3.08
log(1010As)
3.04
3.00
0.97
ns
0.96
0.06
τreio
0.04
0.4 0.6 1.2 1.4 0.0 0.4 0.8 0.0223 0.116 0.120 66 68 3.04 3.08 0.96 0.97 0.04 0.06
α φi/φmax φi Ωb h2 Ωc h2 H0 log(1010As) ns τreio
Figure 16. Full parameter constraints for Saxion model.
41
α > 0.521 (0.66) 0.537+0.16
−0.045 (0.366) 0.49 ± 0.11 (0.311)
ϕi /ϕmax < 1.23 (1.43) 1.201+0.071
−0.17 (1.14)
+0.058
1.205−0.14 (1.10)
Ωc h2 0.11836 ± 0.00084 (0.1184) 0.11841 ± 0.00083 (0.1181) 0.11841 ± 0.00085 (0.1175)

Ωb h2 0.02228 ± 0.00013 (0.02228) 0.02228 ± 0.00012 (0.02223) 0.02228 ± 0.00013 (0.02231)
log(1010 As ) 3.045 ± 0.014 (3.040) 3.044 ± 0.014 (3.048) 3.044 ± 0.014 (3.045)
ns 0.9665 ± 0.0037 (0.9650) 0.9664 ± 0.0037 (0.9644) 0.9664 ± 0.0037 (0.9665)
H0 67.44+0.49
−0.43 (67.61) 67.21+0.62
−0.55 (66.16) 66.86 ± 0.52 (66.31)
τreio 0.0573 ± 0.0070 (0.0552) 0.0568+0.0067
−0.0075 (0.0574) 0.0572+0.0066
−0.0076 (0.0572)
ϕi 0.209+0.064
−0.18 (0.07) 0.241+0.083
−0.21 (0.53)
+0.13
0.32−0.22 (0.64)
Table 7. Saxion model: full parameter means and 68% limits for the addition of the different
42
B.3 Higgs-like hilltop
CMB+DESI+Pantheon+
CMB+DESI+Union3
CMB+DESI+DESY5
0.3
φi/φ0
0.2
0.1
0.6
0.4
φi
0.2
0.0225
Ωb h2
0.0220
0.120
Ωc h2
0.118
0.116
68
H0
66
64
3.08
log(1010As)
3.04
3.00
0.97
ns
0.96
0.06
τreio
0.04
0.5 1.0 1.5 0.1 0.2 0.3 0.1 0.4 0.7 0.0223 0.117 0.120 64 66 68 3.04 3.08 0.96 0.97 0.04 0.06
φ0 φi/φ0 φi Ωb h2 Ωc h2 H0 log(1010As) ns τreio
Figure 17. Full parameter constraints for the Higgs-like model.
43
ϕ0 > 1.29 (1.23) > 1.24 (0.69) > 1.17 (0.62)
ϕi /ϕ0 < 0.142 (0.106) 0.151+0.073
−0.12 (0.051) 0.169 ± 0.081 (0.034)
Ωc h2 0.11838 ± 0.00082 (0.1188) 0.11835 ± 0.00084 (0.1176) 0.11829 ± 0.00084 (0.1189)
Ωb h2 0.02228 ± 0.00012 (0.02221) 0.02228 ± 0.00013 (0.00223) 0.02228 ± 0.00013 (0.02223)
log(1010 As ) 3.044 ± 0.014 (3.041) 3.046 ± 0.014 (3.044) 3.045 ± 0.014 (3.033)
ns 0.9666 ± 0.0036 (0.9641) 0.9667 ± 0.0037 (0.9694) 0.9669 ± 0.0036 (0.9645)
H0 67.29+0.59
−0.45 (66.98) 66.7+1.0
−0.70 (66.30) 66.44 ± 0.64 (66.06)
τreio 0.0567 ± 0.0071 (0.0555) 0.0576 ± 0.0071 (0.056) 0.0576 ± 0.0072 (0.0514)
ϕi 0.174+0.071
−0.17 (0.1313) 0.235+0.088
−0.23 (0.0035) 0.26+0.10
−0.24 (0.0021)
Table 8. Higgs-like model: full parameter means and 68% limits for the addition of the different
44
B.4 Dutta-Scherrer parameterisation
CMB+DESI+Pantheon+
CMB+DESI+Union3
CMB+DESI+DESY5
0.0
w0
−0.4
−0.8
0.0225
Ωb h2
0.0220
0.120
Ωc h2
0.118
0.116
68
H0
66
3.08
log(1010As)
3.04
3.00
0.97
ns
0.96
0.06
τreio
0.04
5 10 −0.5 0.0 0.0220 0.0225 0.117 0.120 66 68 3.01 3.05 0.96 0.97 0.04 0.06
K w0 Ωb h2 Ωc h2 H0 log(1010As) ns τreio
Figure 18. Full parameter plots for the DS parameterisation.
45
K 7.6+2.5
−2.1 (7.49) 8.2+2.7
−1.4 (6.16) 8.4+2.4
−1.1 (9.52)
w0 < −0.709 (-0.79) −0.11+0.39

−0.61 (-0.60) −0.35+0.24
−0.33 (-0.21)
Ωc h2 0.11830 ± 0.00081 (0.11782) 0.11828 ± 0.000856 (0.11815) 0.11825 ± 0.00082 (0.11757)
Ωb h2 0.02226 ± 0.00013 (0.02228) 0.02227 ± 0.00013 (0.02229) 0.02227 ± 0.00013 (0.02230)
H0 67.53 ± 0.39 (67.76) 66.73 ± 0.59 (65.92) 67.04 ± 0.42 (67.23)
log(1010 As ) 3.045 ± 0.014 (3.046) 3.044 ± 0.014 (3.045) 3.045 ± 0.014 (3.043)
ns 0.9659 ± 0.0037 (0.9687) 0.9661 ± 0.0037 (0.9676) 0.9661 ± 0.0036 (0.9665)
τreio 0.0574 ± 0.0071 (0.0588) 0.0573 ± 0.0070 (0.0580) 0.0576 ± 0.0069 (0.0565)
Table 9. DS parameterisation: full parameter means and 68% limits for the addition of the
different supernovae datasets to the CMB+DESI combination. The values in parentheses denote
the best-fit parameters for this model.
+Union3 +DESY5 +Pantheon+

0.15
0.10
|∆φi |
0.05
0.00
2 4 6 8 10 12
K
Figure 19. Analytical result from eq. (3.25), results for the hilltop quintessence models and
posterior contours from MCMC analysis in the K-|∆ϕi | plane. The analytical result is in grey
lines, where three different patterns use best-fit values for Ωϕ,0 and w0 from different data com-
binations (solid: +Pantheon+, dashed: +Union3, dot-dashed: +DESY5). In the same figure,
we show the 1σ and 2σ bounds from the constraints on the DS parameterisation for all data
combinations. Values in the K-|∆ϕi | plane corresponding to the hilltop quintessence models are
indicated by coloured shapes: circle for axion model, star for sugra model and triangle for the
Higgs-like model, using best-fit values with Union3/Pantheon+/DESY5 for model parameters
(ϕ0 and f ) denoted by dark blue/dark red/dark green shapes. All the best-fit values can be
found in Tables 6-9.
46
Union3 DESY5 Pantheon+
w(z = 0.4)
−0.97
−0.98
−0.99
−0.5 0.0 0.5 2 4 6 8 10 12 14

w0 K
Figure 20. Derived bounds on wϕ (z = 0.4) from the MCMC analysis for the DS parameterisa-
tion.
47
C Dutta-Scherrer-Chiba parameterisation, including curvature
In this appendix we collect the evolution of the equation of state including non-zero cur-
vature, Ωk and its comparison with the DS parameterisation, for all the models discussed
in the main text. In all plots we take Ωk = 0.005.
θi = (π − 0.01)f
f = 0.15 f = 0.5
DS: Ωφ0 = 0.675, w0 = −0.95 DS: Ωφ0 = 0.675, w0 = −0.99997
−0.96
−0.99998
wDE
−0.98 −0.99999
−1.00 −1.00000
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
a a
Figure 21. Evolution of wDE for the Axion model including non-zero curvature and its compar-
ison with the DS parameterisation.
φi = 0.5
α = 0.37
−0.96 DS: Ωφ0 = 0.664, w0 = −0.95
wDE
−0.98
−1.00
0.0 0.2 0.4 0.6 0.8 1.0
a
Figure 22. Evolution of wDE for the Saxion model including non-zero curvature and its com-
parison with the DS parameterisation.
48
φi = 0.036φ0
−0.996
−0.92 φ0 = 0.69 φ0 = 1.3
DS: Ωφ0 = 0.668, w0 = −0.92 DS: Ωφ0 = 0.668, w0 = −0.996
−0.997
−0.94
wDE
−0.96 −0.998
−0.98 −0.999
−1.00 −1.000
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
a a
Figure 23. Evolution of wDE for the Higgs-like model including non-zero curvature and its
comparison with the DS parameterisation.
φi = 0
λ = 0.25 λ = 0.5
−0.992
DSCh: Ωφ0 = 0.675, w0 = −0.991 DSCh: Ωφ0 = 0.675, w0 = −0.96 −0.97
−0.994
wDE
−0.98
−0.996
−0.99
−0.998
−1.000 −1.00
√ √ −0.7
−0.900 λ = 3/2 λ= 2
DSCh: Ωφ0 = 0.675, w0 = −0.890 DSCh: Ωφ0 = 0.675, w0 = −0.7
−0.925 −0.8
wDE
−0.950
−0.9
−0.975
−1.000 −1.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
a a
Figure 24. Evolution of wDE for the Exponential model including non-zero curvature and its
comparison with the DSCh parameterisation.
49
References
[1] M. Cicoli, S. De Alwis, A. Maharana, F. Muia, and F. Quevedo, “De Sitter vs
Quintessence in String Theory,” Fortsch. Phys. 67 no. 1-2, (2019) 1800079,
arXiv:1808.08967 [hep-th].
[2] M. Cicoli, J. P. Conlon, A. Maharana, S. Parameswaran, F. Quevedo, and I. Zavala,
“String cosmology: From the early universe to today,” Phys. Rept. 1059 (2024) 1–155,
arXiv:2303.04819 [hep-th].
[3] J. Polchinski, “The Cosmological Constant and the String Landscape,” in 23rd Solvay
Conference in Physics: The Quantum Structure of Space and Time, pp. 216–236. 3, 2006.
arXiv:hep-th/0603249.
[4] E. Palti, “The Swampland: Introduction and Review,” Fortsch. Phys. 67 no. 6, (2019)
1900037, arXiv:1903.06239 [hep-th].
[5] M. van Beest, J. Calderón-Infante, D. Mirfendereski, and I. Valenzuela, “Lectures on the
Swampland Program in String Compactifications,” Phys. Rept. 989 (2022) 1–50,
arXiv:2102.01111 [hep-th].
[6] M. Graña and A. Herráez, “The Swampland Conjectures: A Bridge from Quantum
Gravity to Particle Physics,” Universe 7 no. 8, (2021) 273, arXiv:2107.00087 [hep-th].
[7] S. K. Garg and C. Krishnan, “Bounds on Slow Roll and the de Sitter Swampland,” JHEP
11 (2019) 075, arXiv:1807.05193 [hep-th].
[8] H. Ooguri, E. Palti, G. Shiu, and C. Vafa, “Distance and de Sitter Conjectures on the
Swampland,” Phys. Lett. B 788 (2019) 180–184, arXiv:1810.05506 [hep-th].
[9] E. Witten, “Quantum gravity in de Sitter space,” in Strings 2001: International
Conference. 6, 2001. arXiv:hep-th/0106109.
[10] T. Banks, “The Top 10500 Reasons Not to Believe in the Landscape,” arXiv:1208.5715
[hep-th].
[11] G. Dvali, C. Gomez, and S. Zell, “Quantum Breaking Bound on de Sitter and
Swampland,” Fortsch. Phys. 67 no. 1-2, (2019) 1800094, arXiv:1810.11002 [hep-th].
[12] M. Cicoli, M. Licheri, P. Piantadosi, F. Quevedo, and P. Shukla, “Higher derivative
corrections to string inflation,” JHEP 02 (2024) 115, arXiv:2309.11697 [hep-th].
[13] T. Rudelius, “Dimensional reduction and (Anti) de Sitter bounds,” JHEP 08 (2021) 041,
arXiv:2101.11617 [hep-th].
[14] B. Valeixo Bento, D. Chakraborty, S. L. Parameswaran, and I. Zavala, “Dark Energy in
String Theory,” PoS CORFU2019 (2020) 123, arXiv:2005.10168 [hep-th].
[15] D. Andriot, S. Parameswaran, D. Tsimpis, T. Wrase, and I. Zavala, “Exponential
quintessence: curved, steep and stringy?,” JHEP 08 (2024) 117, arXiv:2405.09323
[hep-th].
[16] S. Bhattacharya, G. Borghetto, A. Malhotra, S. Parameswaran, G. Tasinato, and I. Zavala,
50
“Cosmological constraints on curved quintessence,” JCAP 09 (2024) 073,
arXiv:2405.17396 [astro-ph.CO].
[17] O. F. Ramadan, J. Sakstein, and D. Rubin, “DESI constraints on exponential
quintessence,” Phys. Rev. D 110 no. 4, (2024) L041303, arXiv:2405.18747
[astro-ph.CO].
[18] G. Alestas, M. Delgado, I. Ruiz, Y. Akrami, M. Montero, and S. Nesseris, “To curve, or
not to curve: Is curvature-assisted quintessence observationally viable?,”
arXiv:2406.09212 [hep-th].
[19] S. Dutta and R. J. Scherrer, “Hilltop Quintessence,” Phys. Rev. D 78 (2008) 123525,
arXiv:0809.4441 [astro-ph].
[20] T. Chiba, “Slow-Roll Thawing Quintessence,” Phys. Rev. D 79 (2009) 083517,
arXiv:0902.4037 [astro-ph.CO]. [Erratum: Phys.Rev.D 80, 109902 (2009)].
[21] M. Chevallier and D. Polarski, “Accelerating universes with scaling dark matter,” Int. J.
Mod. Phys. D 10 (2001) 213–224, arXiv:gr-qc/0009008.
[22] E. V. Linder, “Exploring the expansion history of the universe,” Phys. Rev. Lett. 90
(2003) 091301, arXiv:astro-ph/0208512.
[23] R. Arjona and S. Nesseris, “A swampland conjecture DESIderátum?,” arXiv:2409.14990
[astro-ph.CO].
[24] S. L. Parameswaran, S. Ramos-Sanchez, and I. Zavala, “On Moduli Stabilisation and de
Sitter Vacua in MSSM Heterotic Orbifolds,” JHEP 01 (2011) 071, arXiv:1009.3931
[hep-th].
[25] D. Andriot, “Open problems on classical de Sitter solutions,” Fortsch. Phys. 67 no. 7,
(2019) 1900026, arXiv:1902.10093 [hep-th].
[26] S. Parameswaran and M. Serra, “On (A)dS Solutions from Scherk-Schwarz Orbifolds,”
arXiv:2407.16781 [hep-th].
[27] DES Collaboration, T. M. C. Abbott et al., “The Dark Energy Survey: Cosmology
Results With ˜1500 New High-redshift Type Ia Supernovae Using The Full 5-year
Dataset,” arXiv:2401.02929 [astro-ph.CO].
[28] DESI Collaboration, A. G. Adame et al., “DESI 2024 VI: Cosmological Constraints from
the Measurements of Baryon Acoustic Oscillations,” arXiv:2404.03002 [astro-ph.CO].
[29] DESI Collaboration, K. Lodha et al., “DESI 2024: Constraints on Physics-Focused
Aspects of Dark Energy using DESI DR1 BAO Data,” arXiv:2405.13588
[astro-ph.CO].
[30] J. A. Frieman, C. T. Hill, A. Stebbins, and I. Waga, “Cosmology with ultralight pseudo
Nambu-Goldstone bosons,” Phys. Rev. Lett. 75 (1995) 2077–2080,
arXiv:astro-ph/9505060.
[31] K. Choi, “String or M theory axion as a quintessence,” Phys. Rev. D 62 (2000) 043509,
51
arXiv:hep-ph/9902292.
[32] J. E. Kim and H. P. Nilles, “A Quintessential axion,” Phys. Lett. B 553 (2003) 1–6,
arXiv:hep-ph/0210402.
[33] P. Svrcek, “Cosmological Constant and Axions in String Theory,” arXiv:hep-th/0607086.
[34] N. Kaloper and L. Sorbo, “Where in the String Landscape is Quintessence,” Phys. Rev. D
79 (2009) 043528, arXiv:0810.5346 [hep-th].
[35] S. Panda, Y. Sumitomo, and S. P. Trivedi, “Axions as Quintessence in String Theory,”
Phys. Rev. D 83 (2011) 083506, arXiv:1011.5877 [hep-th].
[36] A. Arvanitaki, S. Dimopoulos, S. Dubovsky, N. Kaloper, and J. March-Russell, “String
Axiverse,” Phys. Rev. D 81 (2010) 123530, arXiv:0905.4720 [hep-th].
[37] N. Arkani-Hamed, L. Motl, A. Nicolis, and C. Vafa, “The String landscape, black holes
and gravity as the weakest force,” JHEP 06 (2007) 060, arXiv:hep-th/0601001.
[38] J. E. Kim, H. P. Nilles, and M. Peloso, “Completing natural inflation,” JCAP 01 (2005)
005, arXiv:hep-ph/0409138.
[39] S. Dimopoulos, S. Kachru, J. McGreevy, and J. G. Wacker, “N-flation,” JCAP 08 (2008)
003, arXiv:hep-th/0507205.
[40] J. P. Conlon, “The de Sitter swampland conjecture and supersymmetric AdS vacua,” Int.
J. Mod. Phys. A 33 no. 29, (2018) 1850178, arXiv:1808.05040 [hep-th].
[41] L. McAllister, J. Moritz, R. Nally, and A. Schachner, “Candidate de Sitter Vacua,”
arXiv:2406.13751 [hep-th].
[42] Y. Olguin-Trejo, S. L. Parameswaran, G. Tasinato, and I. Zavala, “Runaway Quintessence,
Out of the Swampland,” JCAP 01 (2019) 031, arXiv:1810.08634 [hep-th].
[43] J. Louis, M. Rummel, R. Valandro, and A. Westphal, “Building an explicit de Sitter,”
JHEP 10 (2012) 163, arXiv:1208.3208 [hep-th].
[44] F. Carta, J. Moritz, and A. Westphal, “Gaugino condensation and small uplifts in KKLT,”
JHEP 08 (2019) 141, arXiv:1902.01412 [hep-th].
[45] E. Hardy and S. Parameswaran, “Thermal Dark Energy,” Phys. Rev. D 101 no. 2, (2020)
023503, arXiv:1907.10141 [hep-th].
[46] J. M. Gomes, E. Hardy, and S. Parameswaran, “Dark energy with the help of interacting
dark sectors,” Phys. Rev. D 110 no. 2, (2024) 023533, arXiv:2311.08888 [hep-ph].
[47] P. Sikivie, “Axion Cosmology,” Lect. Notes Phys. 741 (2008) 19–50,
arXiv:astro-ph/0610440.
[48] H. Ooguri and C. Vafa, “On the Geometry of the String Landscape and the Swampland,”
Nucl. Phys. B 766 (2007) 21–33, arXiv:hep-th/0605264.
[49] G. Pantazis, S. Nesseris, and L. Perivolaropoulos, “Comparison of thawing and freezing
dark energy parametrizations,” Phys. Rev. D 93 no. 10, (2016) 103503, arXiv:1603.02164
52
[astro-ph.CO].
[50] I. D. Gialamas, G. Hütsi, K. Kannike, A. Racioppi, M. Raidal, M. Vasar, and H. Veermäe,
“Interpreting DESI 2024 BAO: late-time dynamical dark energy or a local effect?,”
[51] Y. Akrami, R. Kallosh, A. Linde, and V. Vardanyan, “The Landscape, the Swampland and
the Era of Precision Cosmology,” Fortsch. Phys. 67 no. 1-2, (2019) 1800075,
arXiv:1808.09440 [hep-th].
[52] Planck Collaboration, N. Aghanim et al., “Planck 2018 results. V. CMB power spectra
and likelihoods,” Astron. Astrophys. 641 (2020) A5, arXiv:1907.12875 [astro-ph.CO].
[53] E. Rosenberg, S. Gratton, and G. Efstathiou, “CMB power spectra and cosmological
parameters from Planck PR4 with CamSpec,” Mon. Not. Roy. Astron. Soc. 517 no. 3,
(2022) 4620–4636, arXiv:2205.10869 [astro-ph.CO].
[54] Planck Collaboration, N. Aghanim et al., “Planck 2018 results. VIII. Gravitational
lensing,” Astron. Astrophys. 641 (2020) A8, arXiv:1807.06210 [astro-ph.CO].
[55] DESI Collaboration, A. G. Adame et al., “DESI 2024 IV: Baryon Acoustic Oscillations
from the Lyman Alpha Forest,” arXiv:2404.03001 [astro-ph.CO].
[56] DESI Collaboration, A. G. Adame et al., “DESI 2024 III: Baryon Acoustic Oscillations
from Galaxies and Quasars,” arXiv:2404.03000 [astro-ph.CO].
[57] D. Brout et al., “The Pantheon+ Analysis: Cosmological Constraints,” Astrophys. J. 938
no. 2, (2022) 110, arXiv:2202.04077 [astro-ph.CO].
[58] D. Rubin et al., “Union Through UNITY: Cosmology with 2,000 SNe Using a Unified
Bayesian Framework,” arXiv:2311.12098 [astro-ph.CO].
[59] A. Lewis and S. Bridle, “Cosmological parameters from CMB and other data: A Monte
Carlo approach,” Phys. Rev. D66 (2002) 103511, arXiv:astro-ph/0205436 [astro-ph].
https://arxiv.org/abs/astro-ph/0205436.
[60] A. Lewis, “Efficient sampling of fast and slow cosmological parameters,” Phys. Rev. D87
no. 10, (2013) 103529, arXiv:1304.4473 [astro-ph.CO].
https://arxiv.org/abs/1304.4473.
[61] J. Torrado and A. Lewis, “Cobaya: Code for Bayesian Analysis of hierarchical physical
models,” JCAP 05 (2021) 057, arXiv:2005.05290 [astro-ph.IM].
[62] A. Lewis, “GetDist: a Python package for analysing Monte Carlo samples,”
arXiv:1910.13970 [astro-ph.IM].
[63] C. Cartis, L. Roberts, and O. Sheridan-Methven, “Escaping local minima with
derivative-free methods: a numerical investigation,” arXiv e-prints (Dec., 2018)
arXiv:1812.11343, arXiv:1812.11343 [math.OC].
[64] C. Cartis, J. Fiala, B. Marteau, and L. Roberts, “Improving the Flexibility and
Robustness of Model-Based Derivative-Free Optimization Solvers,” arXiv e-prints (Mar.,
53
2018) arXiv:1804.00154, arXiv:1804.00154 [math.OC].
[65] R. Calderon et al., “DESI 2024: Reconstructing Dark Energy using Crossing Statistics
with DESI DR1 BAO data,” arXiv:2405.04216 [astro-ph.CO].
[66] K. Lodha et al., “DESI 2024: Constraints on Physics-Focused Aspects of Dark Energy
using DESI DR1 BAO Data,” arXiv:2405.13588 [astro-ph.CO].
[67] E. O. Colgáin, M. G. Dainotti, S. Capozziello, S. Pourojaghi, M. M. Sheikh-Jabbari, and
D. Stojkovic, “Does DESI 2024 Confirm ΛCDM?,” arXiv:2404.08633 [astro-ph.CO].
[68] Y. Carloni, O. Luongo, and M. Muccino, “Does dark energy really revive using DESI 2024
data?,” arXiv:2404.12068 [astro-ph.CO].
[69] C.-G. Park, J. de Cruz Perez, and B. Ratra, “Using non-DESI data to confirm and
strengthen the DESI 2024 spatially-flat w0 wa CDM cosmological parameterization result,”
[70] D. Wang, “The Self-Consistency of DESI Analysis and Comment on ”Does DESI 2024
Confirm ΛCDM?”,” arXiv:2404.13833 [astro-ph.CO].
[71] M. Cortês and A. R. Liddle, “Interpreting DESI’s evidence for evolving dark energy,”
[72] Z. Wang, S. Lin, Z. Ding, and B. Hu, “The role of LRG1 and LRG2’s monopole in
inferring the DESI 2024 BAO cosmology,” arXiv:2405.02168 [astro-ph.CO].
[73] B. R. Dinda, “A new diagnostic for the null test of dynamical dark energy in light of DESI
2024 and other BAO data,” arXiv:2405.06618 [astro-ph.CO].
[74] K. S. Croker, G. Tarlé, S. P. Ahlen, B. G. Cartwright, D. Farrah, N. Fernandez, and R. A.
Windhorst, “DESI Dark Energy Time Evolution is Recovered by Cosmologically Coupled
Black Holes,” arXiv:2405.12282 [astro-ph.CO].
[75] D. Wang, “Constraining Cosmological Physics with DESI BAO Observations,”
[76] O. Luongo and M. Muccino, “Model independent cosmographic constraints from DESI
2024,” arXiv:2404.07070 [astro-ph.CO].
[77] P. Mukherjee and A. A. Sen, “Model-independent cosmological inference post DESI DR1
BAO measurements,” arXiv:2405.19178 [astro-ph.CO].
[78] H. Wang and Y.-S. Piao, “Dark energy in light of recent DESI BAO and Hubble tension,”
[79] G. Efstathiou, “Evolving Dark Energy or Supernovae Systematics?,” arXiv:2408.07175
[astro-ph.CO].
[80] Y. Tada and T. Terada, “Quintessential interpretation of the evolving dark energy in light
of DESI,” arXiv:2404.05722 [astro-ph.CO].
[81] W. Yin, “Cosmic Clues: DESI, Dark Energy, and the Cosmological Constant Problem,”
arXiv:2404.06444 [hep-ph].
54
[82] K. V. Berghaus, J. A. Kable, and V. Miranda, “Quantifying Scalar Field Dynamics with
DESI 2024 Y1 BAO measurements,” arXiv:2404.14341 [astro-ph.CO].
[83] D. Shlivko and P. Steinhardt, “Assessing observational constraints on dark energy,”
[84] G. Alestas, M. Caldarola, S. Kuroyanagi, and S. Nesseris, “DESI constraints on
α-attractor inflationary models,” arXiv:2410.00827 [astro-ph.CO].
[85] M. Cicoli, F. Cunillera, A. Padilla, and F. G. Pedro, “From Inflation to Quintessence: a
History of the Universe in String Theory,” arXiv:2407.03405 [hep-th].
[86] M. Cicoli, F. G. Pedro, and G. Tasinato, “Natural Quintessence in String Theory,” JCAP
07 (2012) 044, arXiv:1203.6655 [hep-th].
[87] N. Schöneberg, L. Vacher, J. D. F. Dias, M. M. C. D. Carvalho, and C. J. A. P. Martins,
“News from the Swampland — constraining string theory with astrophysics and
cosmology,” JCAP 10 (2023) 039, arXiv:2307.15060 [astro-ph.CO].
[88] W. J. Wolf, C. Garcı́a-Garcı́a, D. J. Bartlett, and P. G. Ferreira, “Scant evidence for
thawing quintessence,” arXiv:2408.17318 [astro-ph.CO].
[89] H. Akaike, “A New Look at the Statistical Model Identification,” IEEE Transactions on
Automatic Control 19 (Jan., 1974) 716–723.
[90] A. R. Liddle, “How many cosmological parameters?,” Mon. Not. Roy. Astron. Soc. 351
(2004) L49–L53, arXiv:astro-ph/0401198.
[91] R. J. Scherrer, “Mapping the Chevallier-Polarski-Linder parametrization onto Physical
Dark Energy Models,” Phys. Rev. D 92 no. 4, (2015) 043001, arXiv:1505.05781
[astro-ph.CO].
[92] B. Heidenreich, M. Reece, and T. Rudelius, “Sharpening the Weak Gravity Conjecture
with Dimensional Reduction,” JHEP 02 (2016) 140, arXiv:1509.06374 [hep-th].
55

Cosmological Tests of Quintessence in Quantum Gravity: A, B, B, C, B, D, B

Uploaded by

Copyright:

Available Formats

Cosmological Tests of Quintessence in Quantum Gravity: A, B, B, C, B, D, B

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cosmological Tests of Quintessence in Quantum Gravity: A, B, B, C, B, D, B

Uploaded by

Copyright:

Available Formats

Cosmological tests of quintessence

Sukannya Bhattacharyaa,1 , Giulia Borghettob, 2 , Ameek Malhotrab, 3 ,

Susha Parameswaranc,4 , Gianmassimo Tasinatob,d,5 , Ivonne Zavalab,6

2 Quintessence in string theory 4

3 Parameterisation of the equation of state for hilltop models 10

5 Outlook and Future Challenges 33

A Saxion-axion stringy hilltops 36

B Constraints for all the parameters of our models 39

C Dutta-Scherrer-Chiba parameterisation, including curvature 48

2 Quintessence in string theory

2.1 dS minima, plateaus, runaways, maxima and saddles vs the Swampland

2.2 String models of hilltop quintessence

2. Saxion hilltops. Geometric string moduli, corresponding to sizes and shapes of

2.3 Cosmological equations for hilltop quintessence

and for the curvature component:

In the last equation, Vϕ ≡ ∂ϕ V . Moreover, we defined

From this definition it is clear that

3 Parameterisation of the equation of state for hilltop models

u(t) = (ϕ − ϕmax )a3/2 (t) . (3.4)

In terms of u, the equation for ϕ (2.10c) becomes

peff ≃ −Vmax . (3.7)

u(t) = A sinh(kV t) + B cosh(kV t) . (3.10)

To fix the integration constants, consider an initial, finite value for ϕ at t = 0:

ϕ(0) ≡ ϕi ≡ ϕmax + ∆ϕi . (3.11)

where, approximating ΩΛ,0 ≃ Ωϕ,0 :

wϕ (a) = w0 + (1 − a)wa , (3.18)

3.2 Testing the DS parameterisation for hilltop quintessence models

3.4 Bounds on – and from – the initial conditions

Hrh ≪ 2π∆ϕi . (3.26)

1. CMB from Planck 2018:

- Planck 2018 low-ℓ temperature and polarisation likelihood [52].

Hereafter, we collectively refer to all the Planck CMB likelihoods as ‘CMB’.

2. BAO likelihoods from DESI DR1 [28, 55, 56]

Hrh ≪ 2π(πf − θi ) = 4.1MPl , (4.1)

that is, there is no effective constraint on the reheating scale.

Parameter +Pantheon+ +Union3 +DESY5

4.2 Saxion hilltop

Parameter +Pantheon+ +Union3 +DESY5

Ωc h2 0.11838 ± 0.00084 0.11842 ± 0.00083 0.11846 ± 0.00085

Hrh ≪ 2πϕi = 1.6MPl , (4.3)

that is, no effective constraint on the reheating scale.

Parameter +Pantheon+ +Union3 +DESY5

4.4 DS parameterisation analysis

w0 < −0.709 −0.11+0.39

Ωc h2 0.11830 ± 0.00081 0.11828 ± 0.00086 0.11825 ± 0.00082

2 6 10 14 −0.6 0.0 0.6 0.117 0.120 65 66 67 68 0.0220 0.0225

AIC = 2n − 2 ln Lmax , (4.4)

AIC Axion Sugra Higgs DS ΛCDM CPL Exp

Table 5. Comparison between the different models considered in this section.

5 Outlook and Future Challenges

A.1 Saxion hilltops

Wnp = Ae−α Φ . (A.2)

where Dj W ≡ ∂j W + Kj W . Writing the complex scalar field component of the chiral

The final potential becomes (2.3).

A.2 Subleading corrections to the axion

Wsub sub = Ce−γ Φ , (A.8)

B.1 Axion hilltop

θi /f — (2.25) — (3.12) < 2.62 (3.07)

Ωb h2 0.02227 ± 0.00012 (0.02223) 0.02227 ± 0.00013 (0.0221) 0.02227 ± 0.00013 (0.0226)