PHYSICAL REVIEW D 111, 023532 (2025)

DESI 2024: Constraints on physics-focused aspects

of dark energy using DESI DR1 BAO data
K. Lodha ,1,2 A. Shafieloo ,1,2 R. Calderon ,1 E. Linder ,3,4,5 W. Sohn,1 J. L. Cervantes-Cota ,6 A. de Mattia,7
J. García-Bellido ,8 M. Ishak ,9 W. Matthewson,1 J. Aguilar,3 S. Ahlen ,10 D. Brooks,11 T. Claybaugh,3
A. de la Macorra ,12 A. Dey ,13 B. Dey ,14 P. Doel,11 J. E. Forero-Romero ,15,16 E. Gaztañaga,17,18,19
S. Gontcho A. Gontcho ,3 C. Howlett ,20 S. Juneau,13 S. Kent ,21,22 T. Kisner ,3 A. Lambert,3 M. Landriau ,3
L. Le Guillou ,23 P. Martini ,24,25,26 A. Meisner ,13 R. Miquel,27,28 J. Moustakas ,29 J. A. Newman ,14 G. Niz ,30,31
N. Palanque-Delabrouille ,7,3 W. J. Percival ,32,33,34 C. Poppett,3,4,5 F. Prada ,35 G. Rossi,36 V. Ruhlmann-Kleider ,7
E. Sanchez ,37 E. F. Schlafly ,38 D. Schlegel,3 M. Schubnell,39,40 H. Seo ,41 D. Sprayberry,13 G. Tarlé ,40
B. A. Weaver,13 and H. Zou 42

(DESI Collaboration)
1
Korea Astronomy and Space Science Institute,
776, Daedeokdae-ro, Yuseong-gu, Daejeon 34055, Republic of Korea
2
University of Science and Technology, 217 Gajeong-ro, Yuseong-gu, Daejeon 34113, Republic of Korea
3
Lawrence Berkeley National Laboratory, 1 Cyclotron Road, Berkeley, California 94720, USA
4
Space Sciences Laboratory, University of California,
Berkeley, 7 Gauss Way, Berkeley, California 94720, USA
5
University of California, Berkeley, 110 Sproul Hall No. 5800 Berkeley, Calfornia 94720, USA
6
Departamento de Física, Instituto Nacional de Investigaciones Nucleares,
Carreterra México-Toluca S/N, La Marquesa, Ocoyoacac, Estado de México C.P. 52750, México
7
IRFU, CEA, Université Paris-Saclay, F-91191 Gif-sur-Yvette, France
8
Instituto de Física Teórica (IFT) UAM/CSIC, Universidad Autónoma de Madrid,
Cantoblanco, E-28049, Madrid, Spain
9
Department of Physics, The University of Texas at Dallas, Richardson, Texas 75080, USA
10
Physics Department, Boston University, 590 Commonwealth Avenue,
Boston, Massachusetts 02215, USA
11
Department of Physics and Astronomy, University College London,
Gower Street, London, WC1E 6BT, United Kingdom
12
Instituto de Física, Universidad Nacional Autónoma de México, Ciudad de México C.P. 04510, México
13
NSF NOIRLab, 950 North Cherry Avenue, Tucson, Arizona 85719, USA
14
Department of Physics and Astronomy and Pittsburgh Particle Physics, Astrophysics, and Cosmology
Center (PITT PACC), University of Pittsburgh, 3941 O’Hara Street, Pittsburgh, Pennsylvania 15260, USA
15
Departamento de Física, Universidad de los Andes,
Cra. 1 No. 18A-10, Edificio Ip, CP 111711, Bogotá, Colombia
16
Observatorio Astronómico, Universidad de los Andes,
Cra. 1 No. 18A-10, Edificio H, CP 111711 Bogotá, Colombia
17
Institut d’Estudis Espacials de Catalunya (IEEC), 08034 Barcelona, Spain
18
Institute of Cosmology and Gravitation, University of Portsmouth,
Dennis Sciama Building, Portsmouth, PO1 3FX, United Kingdom
19
Institute of Space Sciences, ICE-CSIC, Campus UAB, Carrer de Can Magrans s/n,
08913 Bellaterra, Barcelona, Spain
20
School of Mathematics and Physics, University of Queensland, Brisbane 4072, Australia
21
Department of Astronomy and Astrophysics, University of Chicago,
5640 South Ellis Avenue, Chicago, Illinois 60637, USA
22
Fermi National Accelerator Laboratory, PO Box 500, Batavia, Illinois 60510, USA
23
Sorbonne Université, CNRS/IN2P3, Laboratoire de Physique Nucléaire et de Hautes Energies (LPNHE),
FR-75005 Paris, France
24
Center for Cosmology and AstroParticle Physics, The Ohio State University,
191 West Woodruff Avenue, Columbus, Ohio 43210, USA
25
Department of Astronomy, The Ohio State University,
4055 McPherson Laboratory, 140 W 18th Avenue, Columbus, Ohio 43210, USA
26
The Ohio State University, Columbus, 43210 Ohio, USA

2470-0010=2025=111(2)=023532(13) 023532-1 © 2025 American Physical Society

K. LODHA et al. PHYS. REV. D 111, 023532 (2025)

27
Institució Catalana de Recerca i Estudis Avançats,
Passeig de Lluís Companys, 23, 08010 Barcelona, Spain
28
Institut de Física d’Altes Energies (IFAE), The Barcelona Institute of Science and Technology,
Campus UAB, 08193 Bellaterra Barcelona, Spain
29
Department of Physics and Astronomy, Siena College,
515 Loudon Road, Loudonville, New York 12211, USA
30
Departamento de Física, Universidad de Guanajuato—DCI, C.P. 37150, Leon, Guanajuato, México
31
Instituto Avanzado de Cosmología A. C., San Marcos 11 - Atenas 202. Magdalena Contreras,
10720. Ciudad de México, México
32
Department of Physics and Astronomy, University of Waterloo,
200 University Avenue W, Waterloo, Ontario N2L 3G1, Canada
33
Perimeter Institute for Theoretical Physics, 31 Caroline Street North, Waterloo,
Ontario N2L 2Y5, Canada
34
Waterloo Centre for Astrophysics, University of Waterloo,
200 University Avenue W, Waterloo, Ontario N2L 3G1, Canada
35
Instituto de Astrofísica de Andalucía (CSIC), Glorieta de la Astronomía, s/n, E-18008 Granada, Spain
36
Department of Physics and Astronomy, Sejong University, Seoul, 143-747, Korea
37
CIEMAT, Avenida Complutense 40, E-28040 Madrid, Spain
38
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, Maryland 21218, USA
39
Department of Physics, University of Michigan, Ann Arbor, Michigan 48109, USA
40
University of Michigan, Ann Arbor, Michigan 48109, USA
41
Department of Physics and Astronomy, Ohio University, Athens, Ohio 45701, USA
42
National Astronomical Observatories, Chinese Academy of Sciences,
A20 Datun Road, Chaoyang District, Beijing, 100012, People’s Republic of China

(Received 15 September 2024; accepted 15 November 2024; published 21 January 2025)

Baryon acoustic oscillation data from the first year of the Dark Energy Spectroscopic Instrument (DESI)
provide near percent-level precision of cosmic distances in seven bins over the redshift range z ¼ 0.1–4.2.
This paper is the follow-up to the original DESI BAO cosmology paper [A. G. Adame et al. (DESI
Collaboration), arXiv:2404.03002], which considered the conventional w0 wa cold dark matter (CDM)
model. We use the novel DESI data, together with other cosmic probes, to constrain the background
expansion history using some well-motivated physical classes of dark energy. In particular, we explore
three physics-focused behaviors of dark energy from the equation of state and energy density perspectives:
the thawing class (matching many simple quintessence potentials), emergent class (where dark energy
comes into being recently, as in phase transition models), and mirage class [where phenomenologically the
distance to cosmic microwave background (CMB) last scattering is close to that from a cosmological
constant Λ despite dark energy dynamics]. All three classes fit the data at least as well as ΛCDM, and
indeed can improve on it by Δχ 2 ≈ −5 to −17 for the combination of DESI BAO with CMB and supernova
data while having one more parameter. The mirage class does essentially as well as w0 wa CDM and exhibits
moderate to strong Bayesian evidence preference with respect to ΛCDM. These classes of dynamical
behaviors highlight worthwhile avenues for further exploration into the nature of dark energy.

DOI: 10.1103/PhysRevD.111.023532

I. INTRODUCTION manner. In the absence of a fundamental theory pointing to a

compelling model, and for robustness, many analyses take
Cosmic distances and the cosmic expansion rate contain
one of the latter two routes. The standard parametrization for
information on the matter and energy contents of the
the dark energy equation of state wðaÞ ¼ w0 þ wa ð1 − aÞ
Universe. Redshift surveys can measure these at many originated from exact solutions of the dark energy physical
epochs in cosmic history and so are especially valuable in dynamics and has been demonstrated to be accurate to
separating the contributions and studying their evolution. In ∼0.1% [1,2] in matching distances and expansion rates over
particular, a central question in cosmology is the nature of a wide array of models. Results for w0 wa CDM cosmology
cosmic acceleration: does it originate from a cosmological using DESI data are presented in [3]. Model-independent or
constant Λ or a dynamically evolving dark energy? at least model-agnostic approaches, e.g. Crossing Statistics,
Dark energy can be explored through the use of particular Gaussian Processes, values in redshift bins, etc., are inves-
models, general parametrizations, or in a model-independent tigated in [4] and forthcoming works for DESI data. Here, we

023532-2
DESI 2024: CONSTRAINTS ON PHYSICS-FOCUSED ASPECTS … PHYS. REV. D 111, 023532 (2025)

follow the original DESI BAO cosmology paper and its f DE ðzÞ describes the dark energy density evolution. The
companion support papers, with general parametrizations dark energy equation of state, or pressure to energy density
informed by physics properties. Thus we do not use specific ratio, then comes from the continuity equation as
models, but are focused by the physics into classes of
1 d ln f DE ðzÞ
dynamical evolution. One perspective is to address the wðzÞ ¼ −1 þ : ð2Þ
evolution in terms of the dark energy equation of state, 3 d ln ð1 þ zÞ
specifically the thawing class, and another in terms of the For example,
dark energy density, specifically where dark energy arises
quickly, whether through a phase transition or a rapid growth wðaÞ ¼ w0 þ wa ð1 − aÞ ⇔ f DE ðaÞ ¼ a−3ð1þw0 þwa Þ e−3wa ð1−aÞ ;
(w ≪ −1). A third class is the mirage class [5], where an
apparent w ¼ −1 when forced to a constant value actually ð3Þ
hides dynamics. These are chosen because previous data
pointed to these physical properties as compatible with where the scale factor a ¼ 1=ð1 þ zÞ. The key then is seeing
how physics informs the dark energy density f DE ðzÞ or
observations, at least as well as the cosmological constant,
equation of state wðaÞ. We assume throughout that the dark
and they describe broad classes of characteristic behavior,
energy is fluid with positive energy density f DE ðzÞ > 0 and
rather than a single model.
sound speed c2s ¼ 1.
New, highly precise data have become available from
Data Release 1 of the DESI baryon acoustic oscillations
A. Dark energy equation of state: Thawing physics
(DESI BAO) measurements. The data and w0 wa CDM
results are discussed in detail in [3,6,7]. Here, in this While writing the dark energy equation of state as
companion work, we further investigate their implications wðaÞ ¼ w0 þ wa ð1 − aÞ [1,8] has been shown to be highly
for dark energy physics using the physics-focused classes accurate for a wide variety of models [1,2], the physics does
above. Section II describes the physics-focused classes for not actually predict that any arbitrary combination of w0
the dark energy equation of state and density. In Sec. III we and wa is equally valid [9,10]. Basic physics—evolution of
review the DESI distance data in seven redshift ranges, as the dark energy field through the long history of radiation
well as other data sets used in combination. Constraints on and matter domination in the presence of Hubble friction—
cosmology and the physics implications are discussed in calls out two regions of the phase space as preferred, known
Sec. IV, with conclusions in Sec. V. as the thawing and freezing regions [11]. Other regions of
the phase space arise only due to extraordinary circum-
II. PHYSICS-FOCUSED CLASSES stances, e.g., fine tuning, noncanonical kinetic structure, or
nongravitational interactions. The freezing region, with
The current data, including DESI BAO, favors dynamical wa > 0, tends not to be compatible with observations;
dark energy over a simple cosmological constant at various indeed DESI BAO plus other probes disfavors it at ∼3σ [3].
levels of significance when different data set combinations Thus we focus on exploring the thawing physics class.
are used, as shown in [3]. Additionally, the data shows a Thawing physics arises because, during the long cosmic
negligible preference for curvature, leading us to assume a history, the Hubble friction was high enough to overcome
spatially flat universe. This motivates consideration of a wide dark energy dynamics, causing it to act like a cosmological
variety of dark energy behaviors. constant. Only recently, as the Hubble expansion rate
However, there is no compelling physics-based theory declined sufficiently, was dark energy released to allow
for dark energy differing from Λ, so one tends to adopt a dynamics (“thawed”). This describes a broad variety of
more phenomenological approach. Here, we want to retain particle physics models for dark energy, including pseudo-
physics to a significant extent and use classes of dark Nambu-Goldstone bosons (PNGB [12]; e.g., axions), the
energy properties consistent with the data, and that are linear potential [13], and many monomial potentials (e.g.,
more general than specific models. the standard quadratic V ∼ m2 ϕ2 and quartic V ∼ λϕ4 ).
Dark energy properties enter the measurements through One of the great virtues of the w0 –wa parametrization is
their impact on cosmic distances (we do not here consider that it acts as a calibration relation for the physics. Not only
growth probes of large scale structures). This follows from are the thawing fields in the same class, but w0 –wa calibrates
the Friedmann equations, and we can write the dark energy their evolution wðaÞ, bringing their phase space tracks in
influence as (assuming a spatially flat universe) w–w0 into a universal relation [2]
H2 ðzÞ wa ≈ −1.58ð1 þ w0 Þ: ð4Þ
¼ Ωm;0 ð1 þ zÞ3 þ Ωr;0 ð1 þ zÞ4
H20
[The coefficient −1.58 comes from fits to the dynamics in [2],
þ ð1 − Ωm;0 − Ωr;0 Þf DE ðzÞ; ð1Þ
e.g., see Eq. (1) of [14].]
where Ωm;0 and Ωr;0 are the present fractions of the critical Another approach to thawing dynamics is to account for
energy density in matter and radiation respectively, and the Hubble friction freeze in the past plus an algebraic

023532-3
K. LODHA et al. PHYS. REV. D 111, 023532 (2025)



factor describing the thawing, roughly related to the ratio of Δ 1þz
the frozen dark energy density to the matter density, ∼a3 . wðzÞ ¼ −1 − 1 þ tanh Δlog10 : ð7Þ
3lnð10Þ 1 þ zt
Again, these are general characteristics of the thawing class
as a whole, and so not model dependent in the usual sense. Note that f DE ðzÞ goes from much less than one for z ≫ zt
Following [14,15] we have to one today to a finite value greater than one in the future

2=3 (de Sitter state), while wðzÞ goes from −1 − 2Δ=ð3 ln 10Þ at
3 z ≫ zt to −1 in the future. Please note that Eq. (6) can be
1 þ wðaÞ ¼ ð1 þ w0 Þa3 : ð5Þ
1 þ 2a3 reformulated in a different logarithmic base through an
appropriate rescaling of the Δ.
Note that both the calibration and algebraic forms have
simply one parameter more than ΛCDM. They have also
C. Dark energy: Mirage physics
both been demonstrated to have accuracy better than 0.1%
in matching distances dðzÞ and Hubble expansion rates Another, more phenomenological class is that of mirage
HðzÞ of the exact physics [14]. Describing the thawing dark energy [5]. This originated as a way to match the CMB
class by either the calibration or algebraic forms gives distance to last scattering from ΛCDM but for some
virtually identical results (see Fig. 7 in Appendix A), evolving dark energy equation of state wðaÞ. More gen-
adding support for the model-independent nature of the erally, it will appear to yield a constant w ¼ −1 for data
analysis. Thawing fields roll only a limited distance in field combinations with a pivot point, or greatest sensitivity to
space, covering only a small section of the potential, which dark energy equation of state, around a ≈ 0.7. The con-
makes effective reconstruction quite challenging. For a dition becomes
comprehensive analysis of the thawing class with an
effective quadratic potential, we direct readers to [16]. wa ¼ −3.66ð1 þ w0 Þ: ð8Þ

B. Dark energy density: Emergent physics [The coefficient −3.66 comes from Eqs. (1) and (3) of [5] and
In contrast to the previous subsection, we now consider varies by a couple percent over the range Ωm;0 ∈ ½0.25; 0.35.]
the dark energy density, rather than the equation of state, Interestingly, DESI BAO DR1 gives a confidence con-
and a rapid emergence or transition rather than a slow thaw. tour in the w0 –wa plane that follows this closely, and indeed
In this class of physical behavior, dark energy is negligible delivers a w ≈ −1 fit when assuming w ¼ const (e.g., see
(or vanishes) above moderate redshift (say z ≈ 2) but its Fig. 5 of [3]). We emphasize, however, as demonstrated in
energy density quickly grows at lower redshift (implying this article and [4], that this does not actually mean that
w < −1) before leveling off to a constant in the future (and w ¼ const. That may merely be a mirage, even for quite
so w → −1). Physical examples of this behavior include rapidly evolving wðaÞ. Note furthermore that since mirage
phase transitions such as vacuum metamorphosis [17,18] models match the CMB distance, they will also generally
and dark energy as a critical phenomena [19,20]. The closely match the growth of structure (within general
density of dark energy as a critical phenomenon can behave relativity), as the mirage holds for this as well (see [25]
similarly to the magnetization of the Ising model and and Fig. 6 of [5] for demonstration).
effectively emerges at a particular time (redshift) corre-
sponding to the critical temperature in the model [20]. TABLE I. Parameters and priors used in the analysis with our
Phenomenological emergent dark energy (PEDE) model modified version of the Boltzmann solver CLASS. All of the priors
[21,22] has been introduced as a zero freedom dark energy are uniform in the ranges specified below.
model where dark energy has no effective presence in the past
and effectively emerges in the late Universe. The model was Parameter Prior=Value
generalized as generalized emergent dark energy (GEDE) Background-only Ωm;0 U½0.01; 0.99
[23], to include both PEDE and Λ as two limits of the H0 rd ½kms−1  U½1000; 100000
parametric form and to include a larger class of emergent dark
CMB ωcdm ≡ Ωcdm h2 U½0.001; 0.99
energy behaviors.
ω b ≡ Ωb h 2 U½0.005; 0.1
The evolution of the energy density in GEDE is given
lnð1010 As Þ U½1.61; 3.91
by [23,24]
ns U½0.8; 1.2
  
1þz H0 ½km s−1 Mpc−1  U½20; 100
1 − tanh Δ × log10 1þz
τ U½0.01; 0.8
f DE ðzÞ ¼ ð6Þ
t
;
1 þ tanh ðΔ × log10 ð1 þ zt ÞÞ Thawing/Mirage w0 U½−3; 1
where Δ is a free parameter, determining the steepness of GEDE Δ U½−3; 10
the transition, and zt is a derived quantity determined by
w0 wa w0 U½−3; 1
solving ρm ðzt Þ ¼ ρDE ðzt Þ. The corresponding equation of wa U½−3; 2
state for GEDE is

023532-4
DESI 2024: CONSTRAINTS ON PHYSICS-FOCUSED ASPECTS … PHYS. REV. D 111, 023532 (2025)

FIG. 1. Marginalized constraints on the dark energy equation of state wðzÞ and energy density fDE ðzÞ for the thawing class,
parametrized by Eq. (4).

As to the physical mechanism behind the mirage, this is z ¼ 2.33 [6]. We refer the reader to [3,6,7,33–36] for
less clear. Such a crossing of wðaÞ ¼ −1 is a hallmark of further details.
perhaps a combination of multiple scalar fields, inter- (ii) Supernovae Ia (SNe Ia): We combine with super-
actions, or modified gravity generally involving nonca- nova data from three sets, one at a time: “Pantheon-
nonical kinetic terms and possibly braiding of the scalar Plus,” a compilation of 1550 supernovae spanning a
and tensor degrees of freedom [26–28]. Rapid emergence redshift range from 0.01 to 2.26 [37], “Union3,”
of the dark energy density [i.e., strongly negative wa and containing 2087 SNe Ia processed through the Unity
hence wðaÞ ≪ −1 at early redshifts] however is not a 1.5 pipeline based on Bayesian hierarchical model-
generic characteristic of such mechanisms, and if taken at ing [38], and “DES-SN5YR,” a compilation of 194
face value could point more to a phase transition mecha- low-redshift SNe Ia (0.025 < z < 0.1) and 1635
nism. Many of those, however, such as vacuum metamor- photometrically classified SNe Ia covering the range
phosis [17], tend toward a de Sitter (w ¼ −1) state not 0.1 < z < 1.3 [39].
w0 > −1. It is not clear what reasonable physics would (iii) Cosmic microwave background (CMB): We also
contain both a rapid emergence in density and a crossing of include temperature and polarization measurements
wðaÞ ¼ −1. [Speculatively, one could imagine the scalar of the CMB from the Planck satellite [40]. In particu-
field responsible for the phase transition having a negative lar, we use the high-lTTTEEE likelihood (planck_
potential or some interaction that would cause crossing 2018_highl_plik.TTTEEE), together with low-
of wðaÞ ¼ −1.] lTT (planck_2018_lowl.TT) and low-lEE
(planck_2018_lowl.EE) [41], as implemented
III. DATA AND METHODOLOGY in COBAYA [42]. Additionally, we include CMB
lensing measurements from the combination of
The cosmological probes and specific datasets used in NPIPE PR4 from Planck [43] and the Atacama
our analysis1 are Cosmology Telescope (ACT DR6) [44,45] using
(i) Baryon acoustic oscillations (BAO): We use the importance sampling. When using the combined
compilation of compressed distance quantities Planck þ ACT lensing likelihood denoted as
DM =rd , DH =rd , and DV =rd from the first year data variant:actplanck_baseline, which uti-
release of the Dark Energy Spectroscopic Instrument lizes multipoles in the range of (40 < L < 763), we
[29–33], where DM is the transverse comoving
set accurate_lensing:1 and delta_l_max:
distance, DH the Hubble distance, DV their isotropic
800 to match CAMB precision settings as recom-
average, and rd is the sound horizon at the baryon
mended by ACT.
drag epoch. This dataset, abbreviated as “DESI
In our analysis, we utilize Markov Chain Monte Carlo
BAO,” spans seven redshift bins from z ¼ 0.3 to
(MCMC) sampling to explore the parameter space using
the Metropolis-Hastings algorithm [46,47] as implemented
1 in COBAYA [42]. To facilitate efficient sampling of the CMB
All likelihoods utilized in this analysis are also available
within the publicly accessible version of COBAYA or can be Planck likelihoods, we employ the “fast-dragging” scheme
obtained from the GitHub repositories of the respective research [48]. We have adopted priors similar to [3], as presented in
teams. Table I, and have modified the Boltzmann solver CLASS

023532-5
K. LODHA et al. PHYS. REV. D 111, 023532 (2025)

IV. RESULTS AND DISCUSSIONS

We present the results for each class in turn, showing the
cosmological parameter joint posteriors and the recon-
structed probability mass function of dark energy equation
of state wðzÞ and energy density f DE ðzÞ for various
combinations of datasets. For the latter, we draw a large
number of samples from the posterior distribution of w0 .
For each sampled value of w0 , we compute wðzÞ and f DE
and then determine the quantiles for each redshift slice,
corresponding to the 1 and 2σ levels, respectively.

A. Thawing
Figure 1 illustrates the behaviors for the equation of state
wðzÞ and energy density evolution f DE ðzÞ for the thawing
class. Figure 2 shows the joint parameter constraints with
dashed black line corresponds to ΛCDM (w0 ¼ −1). In the
first few rows of Table II, we report the marginalized
constraints on some of the relevant cosmological parameters
and for various data combinations. The addition of CMB data
to DESI BAO significantly reduces the uncertainty in w0 ,
FIG. 2. Marginalized constraints within the thawing class of
dark energy described by Eq. (4), from different combinations of
shifting its value to < −1, which also results in wa > 0.
data sets. However, combining DESI BAO with PantheonPlus yields
w0 > −1. A combination of all three datasets provides even
tighter constraints, with posteriors peaking at w0 ≳ −1,
[49,50] incorporating a generalized equation of state for hence wa < 0. Using either of the other two supernova
dark energy for the theoretical prediction of observables. We datasets instead somewhat strengthens w0 > −1.
switched to the Recfast option for recombination as it does not We emphasize that the relations wa ðw0 Þ for both the
assume anything about the equation of state. We assumeP one thawing and mirage classes are designed to replicate
massive and two massless neutrino species a with mν ¼ observations, i.e., distances and Hubble parameters to
0.06 eV and N eff ¼ 3.044. For the supernovae likelihoods ∼0.1%, not the actual wðzÞ for a specific model within
(PantheonPlus, Union3, and DES-SN5YR), we analytically the class, and hence do not need to have wðz ≫ 1Þ → −1,
marginalize over the absolute magnitude M B . For clarity of say. Indeed wðzÞ can appear to cross −1 even if it actually
presentation, in the main text figures we use PantheonPlus does not, yet still fit the observations exquisitely—this is
but list constraints from each supernova set in the tables and well known: see Table 1 of [14] and Fig. 14 of [51] for a
show their contours in Appendix B. PNGB model, [52] for a hilltop model, etc. However, for

TABLE II. Constraints on the relevant cosmological parameters.

Model=Dataset H 0 ½km s−1 Mpc−1  Ωm;0 w0 Δ S8

Thawing
DESI BAO þ CMB 71.0þ2.4
−2.8
0.282  0.020 −1.17  0.15  0.812  0.011
DESI BAO þ CMB þ Union3 66.5  0.9 0.319  0.009 −0.91  0.06  0.822  0.009
DESI BAO þ CMB þ DES-SN5YR 66.5  0.6 0.320  0.006 −0.90  0.04  0.822  0.009
DESI BAO þ CMB þ PantheonPlus 67.3  0.7 0.312  0.007 −0.95  0.04  0.821  0.009
GEDE
DESI BAO þ CMB 71.0  1.5 0.282  0.012  0.81  0.40 0.816  0.009
DESI BAO þ CMB þ Union3 67.6þ1.0
−0.9
0.310  0.009  −0.08  0.25 0.820  0.009
DESI BAO þ CMB þ DES-SN5YR 66.8  0.7 0.317  0.007  −0.28  0.17 0.821  0.009
DESI BAO þ CMB þ PantheonPlus 67.7  0.7 0.309  0.007  −0.06  0.18 0.820  0.009
Mirage
DESI BAO þ CMB 66.2þ1.3
−0.8 0.327þ0.009
−0.015 −0.56þ0.15
−0.23
 0.842  0.013
DESI BAO þ CMB þ Union3 66.7  0.6 0.322  0.007 −0.65þ0.10
−0.11
 0.838  0.010
DESI BAO þ CMB þ DES-SN5YR 67.1  0.4 0.317  0.006 −0.74  0.07  0.833  0.009
DESI BAO þ CMB þ PantheonPlus 67.5  0.4 0.312  0.005 −0.84  0.06  0.827  0.009

023532-6
DESI 2024: CONSTRAINTS ON PHYSICS-FOCUSED ASPECTS … PHYS. REV. D 111, 023532 (2025)

FIG. 3. Marginalized constraints on the dark energy equation of state wðzÞ and energy density fDE ðzÞ for the GEDE model,
parametrized by Eq. (6).

jwa j ≳ 1 such unreal crossings of wðzÞ ¼ −1 tend to be in indicating that dark energy emerges at late times.
conflict with the CMB distance to last scattering, and such Adding PantheonPlus, GEDE prefers a negative value of
strong crossings tend in fact to be real. Δ, which corresponds to the injection of energy at earlier
redshifts. However, combining all three datasets results in a
B. Generalized emergent dark energy peak near Δ ¼ 0, indicating that dark energy density
remains roughly constant throughout evolution and that
The GEDE analysis proceeds similarly, with Fig. 3 GEDE is not preferred over ΛCDM. Note that the model-
exhibits the uncertainty bands for the reconstructed equa- agnostic reconstructions of dark energy in [4] do seem to
tion of state wðzÞ and energy density evolution f DE ðzÞ and indicate a sharp emergence of dark energy in the recent
Fig. 4 showing its constraints on parameters, using the past. The issue is that GEDE sharply emerges, but
same datasets. When considering DESI BAO alone, the asymptotes to w ¼ −1 rather than crossing it. Increasing
dataset is broadly consistent with Δ ¼ 0, corresponding to Δ fits the z ≈ 1 data better than ΛCDM but the z ≲ 0.5 data
ΛCDM. However, when combined with CMB, Δ peaks at a prefers w > −1 and so GEDE is worse than ΛCDM there,
positive value (Δ ≃ 1, corresponding to PEDE [21]), resulting in GEDE “mellowing” to approach ΛCDM
behavior (and so, as we will see, not having a particularly
advantageous goodness of fit). If there were an emergent
model that also crossed w ¼ −1 then this might provide a
superior fit to data, but physics motivation for such
behavior is not obvious.

C. Mirage
The mirage class has cosmological parameter constraints
illustrated in Fig. 5. Again, the preference is pulled off
ΛCDM (which is a member of this class, where the mirage
is real). A best fit of w0 ≈ −0.8, and hence wa ≈ −0.7, is
quite consistent with DESI BAO data, including in combi-
nation with other data sets such as CMB and supernovae.
One can make w0 even less negative (and hence wa more
negative), i.e. strengthen the mirage, if one compensates by
decreasing the late time dark energy density (increasing
Ωm;0 ), as seen in Fig. 5.
At earlier times, the strongly negative wa implies a
strongly negative wðaÞ, and hence very little dark energy
density, before rapidly increasing in energy density while
crossing wðaÞ ¼ −1. This effectively is acting like GEDE
FIG. 4. Marginalized constraints within the generalized emer- at higher redshift and the thawing class at lower redshift. In
gent dark energy (GEDE) class described by Eq. (6). Fig. 6, we show the reconstructed equation of state wðzÞ,

023532-7
K. LODHA et al. PHYS. REV. D 111, 023532 (2025)

TABLE III. Δχ 2 MAP ≡ χ 2 model − χ 2 ΛCDM values for the differ-

ent models and data combinations. The minimum χ 2 values were
obtained using IMINUIT [53,54] and Py-BOBYQA [55,56] mini-
mizer. Note that all data combinations include DESI BAO.

Data Δχ 2Thaw Δχ 2GEDE Δχ 2Mirage Δχ 2w0 wa

DESI BAO −0.2 −0.04 −5.0 −3.8
þCMB −0.6 −5.7 −7.6 −8.9
þPantheonPlus −3.2 −3.0 −3.5 −3.5
þUnion3 −6.3 −5.2 −8.7 −8.9
þDES-SN5YR −8.8 −7.7 −10.7 −11.1
þCMB þ PantheonPlus −0.6 −1.7 −9.0 −9.6
þCMB þ Union3 −3.0 −3.2 −15.2 −15.6
þCMB þ DES-SN5YR −5.0 −4.8 −17.7 −18.3

is mostly consistent with zero (hence Λ). The mirage class

is where the strongest deviation from Λ is seen, and as we
discuss next is where the goodness of fit is best as well.
Consistency between a class and ΛCDM should be viewed
not through the 1D confidence intervals, however, but
FIG. 5. Marginalized constraints within the mirage class of dark through the joint parameter constraints and the Δχ 2
energy described by Eq. (8). quantification.
Table III presents how the three classes compare to each
other, relative to ΛCDM, and w0 wa CDM in goodness of fit
crossing the w ¼ −1 threshold near z ≈ 0.4, along with the
(Δχ 2 ), for various combinations of datasets. Here, we report
corresponding evolution of energy density f DE ðzÞ; these
Δχ 2MAP , the difference between the maximum a posteriori
results agree with the DESI results in [3,4].
of the model and the maximum of the posterior fixing w0 ¼
−1 or Δ ¼ 0 (i.e., to the cosmological constant case). Note
D. Comparison and discussion that the three classes have one more parameter than ΛCDM
Table II summarizes the constraints on relevant cosmo- and one less than w0 wa CDM. The first general result of
logical parameters. One general aspect to note is that the interest is that all three classes have better χ 2 than ΛCDM.
clustering amplitude parameter S8 has a rather consistent They do have one more parameter but in the combination of
and reasonable value for all three classes (note that no direct all three cosmological probes, the improvement is notably
galaxy clustering growth data is used, only BAO). When more than one. Note that w0 wa CDM has a significantly
combining all three cosmological probes, the values of H0 better χ 2 than ΛCDM, even taking into account its two
and Ωm are also quite consistent both across classes and more parameters, as discussed in [3]. However, the χ 2 for
regardless of which supernova dataset is included. For the the physics-focused classes are often close to the
thawing class, w0 is pulled somewhat off Λ. For GEDE, Δ w0 wa CDM values, while having one less parameter. This

FIG. 6. Marginalized constraints on the equation of state wðzÞ for the mirage class, parametrized by Eq. (8).

023532-8
DESI 2024: CONSTRAINTS ON PHYSICS-FOCUSED ASPECTS … PHYS. REV. D 111, 023532 (2025)

is especially true for the mirage class, while the thawing preference (depending on the SNe Ia dataset considered) for
class and GEDE appear to be less favored. As thawing and the Mirage class over ΛCDM. However, the significance of
mirage are subsets of w0 wa CDM, their χ 2 cannot go below the Bayesian evidence has to be interpreted cautiously
that of w0 wa ; for the case of fitting DESI BAO data alone, [62,63]. We leave a detailed model-selection analysis for
this appears not to hold for the mirage class, but this is due future works.
to the limited prior range of −3 < wa < 2 (also used for Other cosmology parameters such as H0 rd , Ωm , and S8
w0 wa in [3])—see the extended degeneracies in Fig. 5. remain near ΛCDM values when using any of the three
When combining DESI BAO with other data this influence classes with the full dataset combination. Please note the
of the prior no longer matters. preference for dynamical dark energy is primarily driven by
The promising Δχ 2MAP for the mirage class led us to background expansion, and our findings will largely hold
conduct additional nested-sampling runs using the PolyChord even if the assumption of
sampler [57] to calculate the Bayesian evidence using the
ANESTHETIC package [58]. We report the Bayes factors of c2s ¼ 1
j ln B21 j ¼ 2.8ð0.65Þ, 4.2 (2.4), and 6.4 (2.8) in favor of the
mirage class (compared to w0 wa CDM) over ΛCDM for the is relaxed. Together with model independent analyses, such
DESI þ CMB with PantheonPlus, Union3, and DES- physics-focused classes provide important clues to the
SN5YR data combinations, respectively. These findings physical properties we should seek in dark energy models,
suggest a moderate preference for the mirage class over beyond the “blank slate” characterization of w0 –wa .
ΛCDM by the PantheonPlus combination and a strong The dark energy properties indicated by the data—
preference by the Union3 and DES-SN5YR on a Jeffreys’ consistent with [3,4]—are rapid evolution from wða ≪
scale [59,60]. In addition to this, we also check the 1Þ ≪ −1 across w ¼ −1 to more recent wða ≈ 1Þ > −1,
posterior average of the log-likelihood (hln LiP ) [61], and hence emergent dark energy density at modest redshift
which removes the prior-dependent Occam penalty while at low redshift an energy density bump together with
(DKL ) contribution from the log-evidence to provide a a slowing down of recent cosmic acceleration [see, e.g., the
quantitative assessment of how well the model fits the data. qðzÞ reconstruction in [4] ]. These characteristics do not
We observe that hln LiP trends remains consistent with the generally exist simultaneously in the usual dark energy
frequentist Δχ 2 presented in Table III. models. Phase transition-type behavior does not generally
give w evolving away from Λ today, and the thawing class
V. CONCLUSION evolving away from Λ today does not generally give
w < −1, let alone w ≪ −1, in the past.
Physics-focused classes can give insight into the nature If the data and its analysis hold, then we are facing a
of dynamical dark energy. Using DESI BAO data combined more complicated dark energy sector than generally treated,
with different state-of-the-art supernovae compilations possibly involving multiple components or involving spe-
(PantheonPlus, Union3, DES-SN5YR) and CMB (Planck cial nonlinearities in the action (modified gravity or
and ACT) observations, our main result indicates a pref- couplings). Fortunately, further data is imminent, with
erence for evolving dark energy rather than a cosmological more DESI BAO data, as well as DESI measurements of
constant. This behavior can be very well captured by the redshift space distortions and peculiar velocities that can
mirage class, evolving from w < −1 and low energy test cosmic growth and gravity.
density at z ≳ 1 to w > −1 more recently. Note that this
also gives a hump in the dark energy density at z ≈ 0.3–0.5,
ACKNOWLEDGMENTS
in agreement with our previous model-agnostic findings
[4]. The mirage class of dark energy models has a The authors thank Luis Urena-Lopez for his help with
comparable Δχ 2 with that of the w0 wa CDM model, while the PolyChord runs. We acknowledge the use of the
having one less degree of freedom. Boltzmann solver CLASS [49,50] for the computation of
The mirage class combines the emergence of dark energy theoretical observables, COBAYA [42] for the sampling and
density, perhaps indicative of a phase transition, with the GetDist [64] for the post-processing of our results. We also
recent evolution of wðaÞ to less negative values than the acknowledge the use of the standard PYTHON libraries for
cosmological constant of the thawing class. With DESI þ scientific computing, such as NUMPY [65], SCIPY [66] and
SNe Ia, consistently across the three supernova sets, all three MATPLOTLIB [67]. This work was supported by the high-
classes have better fits than ΛCDM and come close to performance computing cluster Seondeok at the Korea
w0 wa CDM (which has one more parameter). Neither thaw- Astronomy and Space Science Institute. A. S. would like to
ing nor GEDE have a strong advantage for DESI þ CMB, acknowledge the support by National Research Foundation
however, and the combination of all three cosmology probes of Korea 2021M3F7A1082056, and the support of the
gives a clear advantage to the mirage class over the other two Korea Institute for Advanced Study (KIAS) grant funded
(which are still better fits than ΛCDM). This preference is by the government of Korea. This material is based upon
reflected in the Bayes factor, showing a moderate to strong work supported by the U.S. Department of Energy (DOE),

023532-9
K. LODHA et al. PHYS. REV. D 111, 023532 (2025)

Office of Science, Office of High-Energy Physics, under

Contract No. DE–AC02–05CH11231, and by the National
Energy Research Scientific Computing Center, a DOE Office
of Science User Facility under the same contract. Additional
support for DESI was provided by the U.S. National Science
Foundation (NSF), Division of Astronomical Sciences under
Contract No. AST-0950945 to the NSF’s National Optical-
Infrared Astronomy Research Laboratory; the Science and
Technology Facilities Council of the United Kingdom;
the Gordon and Betty Moore Foundation; the Heising-
Simons Foundation; the French Alternative Energies and
Atomic Energy Commission (CEA); the National Council
of Humanities, Science and Technology of Mexico
(CONAHCYT); the Ministry of Science and Innovation of
Spain (MICINN), and by the DESI Member Institutions [68].
The DESI collaboration is honored to be permitted to conduct
scientific research on Iolkam Du’ag (Kitt Peak), a mountain
with particular significance to the Tohono O’odham Nation.

Any opinions, findings, and conclusions or recommen-

dations expressed in this material are those of the author(s)
and do not necessarily reflect the views of the U.S. National FIG. 7. Comparison of the calibration Eq. (4) vs algebraic
Science Foundation, the U.S. Department of Energy, or any Eq. (5) parametrizations of the thawing class of dark energy,
using DESI BAO þ PantheonPlus.
of the listed funding agencies.
symmetric linear potential, and the classic ϕ2 and ϕ4
DATA AVAILABILITY potentials. The calibration Eq. (4) and algebraic Eq. (5)
forms were shown to accurately describe the exact numeri-
The data that support the findings of this article are
cal solutions for observables to ∼0.1% (e.g., Table 1 of
openly available [69].
[2,14]). The former emphasizes the calibration, i.e. model
independent, aspects of the physics, and the latter the
APPENDIX A: THAWING CLASS: CALIBRATION dynamical evolution flow, but each captures the essential
VS ALGEBRAIC FORMS physics and Fig. 7 confirms that the two forms give nearly
The thawing class encompasses rich physics, including identical results. The specific numbers quoted in this work
pseudo-Nambu Goldstone bosons (PNGB axions), the shift use the calibration form.

TABLE IV. Constraints on the relevant cosmological parameters for DESI BAO and DESI BAO þ SNe Ia
datasets.

Model/Dataset Ωm;0 H0 rd [km s−1 ] w0 Δ

Thawing
DESI BAO 0.300þ0.020
−0.023
10080  410 −0.94  0.20 
DESI BAO þ PantheonPlus 0.303  0.013 9994  110 −0.90  0.06 
DESI BAO þ Union3 0.309  0.013 9835  140 −0.81þ0.08
−0.07

DESI BAO þ DES-SN5YR 0.306  0.012 9890  100 −0.84  0.05 
GEDE
DESI BAO 0.296þ0.014
−0.016
10170  330  0.12þ0.81
−1.20
DESI BAO þ PantheonPlus 0.299  0.013 9993  110  −0.49þ0.27
−0.32
DESI BAO þ Union3 0.303  0.014 9845  140  −0.86þ0.30
−0.40
DESI BAO þ DES-SN5YR 0.302  0.014 9871  99  −0.80þ0.24
−0.27
Mirage
DESI BAO 0.430þ0.066
−0.044 8813þ300
−630
> 0.08 
DESI BAO þ PantheonPlus 0.311  0.012 10010  100 −0.88  0.06 
DESI BAO þ Union3 0.331  0.015 9768  150 −0.67  0.11 
DESI BAO þ DES-SN5YR 0.322  0.011 9875  99 −0.77  0.07 

023532-10
DESI 2024: CONSTRAINTS ON PHYSICS-FOCUSED ASPECTS … PHYS. REV. D 111, 023532 (2025)

FIG. 8. Marginalized posterior distributions using DESI BAO and each supernova dataset. The vertical dashed line indicates ΛCDM.

Note that while the algebraic form Eq. (5) does not APPENDIX B: SUPERNOVA DATA
cross w ¼ −1 while the calibration form Eq. (4) does, they COMPARISON
both describe the observations nearly identically, with The figures in the main section of the paper use
excellent accuracy. Thus not every wðaÞ that crosses −1 PantheonPlus as the supernova dataset for clarity of presen-
indicates a physical crossing of −1; the forms are designed tation by limiting the number of contours. Tables II and IV list
to describe the observations, not wðaÞ itself, as empha- the parameter constraints for each supernova dataset in turn.
sized in Sec. IVA. However, for w0 too far from −1, i.e., Here, Fig. 8 presents the joint confidence contours using
jwa j large enough, such apparent crossings will not fit DESI BAO in combination with each supernova set in turn.
certain observations like the distance to CMB last scatter- The results are quite consistent between each supernova
ing, and so jwa j ≳ 1 often does point to a real crossing of dataset and a similar trend can also be seen in Figure 6 of [3]
−1 by wðaÞ. for w0 wa CDM.

[1] E. V. Linder, Exploring the expansion history of the uni- [11] R. R. Caldwell and E. V. Linder, The limits of quintessence,
verse, Phys. Rev. Lett. 90, 091301 (2003). Phys. Rev. Lett. 95, 141301 (2005).
[2] R. de Putter and E. V. Linder, Calibrating dark energy, J. [12] J. A. Frieman, C. T. Hill, A. Stebbins, and I. Waga,
Cosmol. Astropart. Phys. 10 (2008) 042. Cosmology with ultralight pseudo Nambu-Goldstone bo-
[3] A. G. Adame et al. (DESI Collaboration), DESI 2024 VI: sons, Phys. Rev. Lett. 75, 2077 (1995).
Cosmological constraints from the measurements of baryon [13] A. Linde, Inflation and quantum cosmology., in Three
acoustic oscillations, arXiv:2404.03002. Hundred Years of Gravitation (Academic Press, New York,
[4] R. Calderon et al., DESI 2024: Reconstructing dark energy 1987), pp. 604–630.
using crossing statistics with DESI DR1 BAO data, J. [14] E. V. Linder, Quintessence’s last stand?, Phys. Rev. D 91,
Cosmol. Astropart. Phys. 10 (2024) 048. 063006 (2015).
[5] E. V. Linder, The mirage of w ¼ −1, arXiv:0708.0024. [15] E. V. Linder, The dynamics of quintessence, the quintes-
[6] A. G. Adame et al. (DESI Collaboration), DESI 2024 III: sence of dynamics, Gen. Relativ. Gravit. 40, 329 (2008).
Baryon acoustic oscillations from galaxies and quasars, [16] W. J. Wolf, C. García-García, D. J. Bartlett, and P. G.
arXiv:2404.03000. Ferreira, Scant evidence for thawing quintessence, Phys.
[7] A. G. Adame et al. (DESI Collaboration), DESI 2024 IV: Rev. D 110, 083528 (2024).
Baryon acoustic oscillations from the Lyman alpha forest, [17] L. Parker and A. Raval, New quantum aspects of a vacuum
arXiv:2404.03001. dominated universe, Phys. Rev. D 62, 083503 (2000); 67,
[8] M. Chevallier and D. Polarski, Accelerating universes with 029903(E) (2003).
scaling dark matter, Int. J. Mod. Phys. D 10, 213 (2001). [18] R. R. Caldwell, W. Komp, L. Parker, and D. A. T. Vanzella,
[9] S. Peirone, M. Martinelli, M. Raveri, and A. Silvestri, Impact A sudden gravitational transition, Phys. Rev. D 73, 023513
of theoretical priors in cosmological analyses: The case of (2006).
single field quintessence, Phys. Rev. D 96, 063524 [19] A. Banihashemi, N. Khosravi, and A. H. Shirazi, Ginzburg-
(2017). Landau theory of dark energy: A framework to study both
[10] E. V. Linder, Interpreting dark energy data away from Λ, temporal and spatial cosmological tensions simultaneously,
arXiv:2410.10981. Phys. Rev. D 99, 083509 (2019).

023532-11
K. LODHA et al. PHYS. REV. D 111, 023532 (2025)

[20] A. Banihashemi, N. Khosravi, and A. Shafieloo, Dark [41] N. Aghanim et al. (Planck Collaboration), Planck 2018
energy as a critical phenomenon: A hint from Hubble results. V. CMB power spectra and likelihoods, Astron.
tension, J. Cosmol. Astropart. Phys. 06 (2021) 003. Astrophys. 641, A5 (2020).
[21] X. Li and A. Shafieloo, A simple phenomenological [42] J. Torrado and A. Lewis, COBAYA: Code for Bayesian
emergent dark energy model can resolve the Hubble tension, analysis of hierarchical physical models, J. Cosmol.
Astrophys. J. Lett. 883, L3 (2019). Astropart. Phys. 05 (2021) 057.
[22] S. Pan, W. Yang, E. Di Valentino, A. Shafieloo, and S. [43] J. Carron, M. Mirmelstein, and A. Lewis, CMB lensing
Chakraborty, Reconciling H 0 tension in a six parameter from Planck PR4 maps, J. Cosmol. Astropart. Phys. 09
space?, J. Cosmol. Astropart. Phys. 06 (2020) 062. (2022) 039.
[23] X. Li and A. Shafieloo, Evidence for emergent dark energy, [44] M. S. Madhavacheril, F. J. Qu, B. D. Sherwin, N.
Astrophys. J. 902, 58 (2020). MacCrann, Y. Li, I. Abril-Cabezas, P. A. R. Ade, S.
[24] W. Yang, E. Di Valentino, S. Pan, A. Shafieloo, and Aiola, T. Alford, M. Amiri et al., The Atacama cosmology
X. Li, Generalized emergent dark energy model and telescope: DR6 gravitational lensing map and cosmological
the Hubble constant tension, Phys. Rev. D 104, 063521 parameters, Astrophys. J. 962, 113 (2024).
(2021). [45] F. J. Qu et al. (ACT Collaboration), The Atacama cosmol-
[25] M. J. Francis, G. F. Lewis, and E. V. Linder, Power spectra ogy telescope: A measurement of the DR6 CMB lensing
to 1% accuracy between dynamical dark energy cosmolo- power spectrum and its implications for structure growth,
gies, Mon. Not. R. Astron. Soc. 380, 1079 (2007). Astrophys. J. 962, 112 (2024).
[26] A. Vikman, Can dark energy evolve to the phantom?, Phys. [46] A. Lewis and S. Bridle, Cosmological parameters from
Rev. D 71, 023515 (2005). CMB and other data: A Monte Carlo approach, Phys. Rev. D
[27] C. Deffayet, O. Pujolas, I. Sawicki, and A. Vikman, 66, 103511 (2002).
Imperfect dark energy from kinetic gravity braiding, J. [47] A. Lewis, Efficient sampling of fast and slow cosmological
Cosmol. Astropart. Phys. 10 (2010) 026. parameters, Phys. Rev. D 87, 103529 (2013).
[28] W. Hu, Crossing the phantom divide: Dark energy internal [48] R. M. Neal, Taking bigger metropolis steps by dragging fast
degrees of freedom, Phys. Rev. D 71, 047301 (2005). variables, arXiv:math/0502099.
[29] DESI Collaboration, A. Aghamousa et al., The DESI [49] J. Lesgourgues, The cosmic linear anisotropy solving
experiment part I: Science, targeting, and survey design, system (CLASS) I: Overview, arXiv:1104.2932.
arXiv:1611.00036. [50] D. Blas, J. Lesgourgues, and T. Tram, The cosmic linear
[30] B. Abareshi et al. (DESI Collaboration), Overview of the anisotropy solving system (CLASS). Part II: Approximation
instrumentation for the dark energy spectroscopic instru- schemes, J. Cosmol. Astropart. Phys. 07 (2011) 034.
ment, Astron. J. 164, 207 (2022). [51] E. V. Linder, Mapping the cosmological expansion, Rep.
[31] A. G. Adame et al. (DESI Collaboration), Validation of the Prog. Phys. 71, 056901 (2008).
scientific program for the dark energy spectroscopic instru- [52] D. Shlivko and P. Steinhardt, Assessing observational
ment, Astron. J. 167, 62 (2024). constraints on dark energy, Phys. Lett. B 855, 138826
[32] A. G. Adame et al. (DESI Collaboration), The early data (2024).
release of the dark energy spectroscopic instrument, Astron. [53] H. Dembinski and P. Ongmongkolkul et al., scikit-hep/
J. 168, 58 (2024). iminuit (Zenodo, 2020), 10.5281/zenodo.3949207.
[33] DESI Collaboration, DESI 2024 I: Data release 1 of the dark [54] F. James and M. Roos, Minuit: A System for
energy spectroscopic instrument. Function Minimization and Analysis of the Parameter
[34] DESI Collaboration, DESI 2024 II: Sample definitions, Errors and Correlations, Comput. Phys. Commun. 10, 343
characteristics and two-point clustering statistics, (1975).
arXiv:2411.12020. [55] C. Cartis, J. Fiala, B. Marteau, and L. Roberts, Improving
[35] DESI Collaboration, DESI 2024 V: Analysis of the full the flexibility and robustness of model-based derivative-free
shape of two-point clustering statistics from galaxies and optimization solvers, arXiv:1804.00154.
quasars, arXiv:2411.12021. [56] C. Cartis, L. Roberts, and O. Sheridan-Methven, Escaping
[36] DESI Collaboration, DESI 2024 VII: Cosmological con- local minima with local derivative-free methods: A numeri-
straints from full-shape analyses of the two-point clustering cal investigation, Optimization 71, 2343 (2021).
statistics measurements, arXiv:2411.12022. [57] W. J. Handley, M. P. Hobson, and A. N. Lasenby, PolyChord:
[37] D. Brout et al., The Pantheon þ analysis: Cosmological Nested sampling for cosmology, Mon. Not. R. Astron. Soc.
constraints, Astrophys. J. 938, 110 (2022). 450, L61 (2015).
[38] D. Rubin et al., Union through UNITY: Cosmology with [58] W. Handley, ANESTHETIC: Nested sampling visualisation,
2,000 SNe using a unified Bayesian framework, arXiv: J. Open Source Software 4, 1414 (2019).
2311.12098. [59] H. Jeffreys, The Theory of Probability, Oxford Classic Texts
[39] T. M. C. Abbott et al. (DES Collaboration), The dark energy in the Physical Sciences (Oxford University Press Inc, New
survey: Cosmology results with ˜1500 new high-redshift York, 1939).
Type Ia supernovae using the full 5-year dataset, Astrophys. [60] R. Trotta, Applications of Bayesian model selection to
J. Lett. 973, L14 (2024). cosmological parameters, Mon. Not. R. Astron. Soc. 378,
[40] N. Aghanim et al. (Planck Collaboration), Planck 2018 72 (2007).
results. VI. Cosmological parameters, Astron. Astrophys. [61] L. T. Hergt, W. J. Handley, M. P. Hobson, and A. N.
641, A6 (2020); 652, C4(E) (2021). Lasenby, Bayesian evidence for the tensor-to-scalar ratio

023532-12
DESI 2024: CONSTRAINTS ON PHYSICS-FOCUSED ASPECTS … PHYS. REV. D 111, 023532 (2025)

r and neutrino masses mν : Effects of uniform vs logarithmic N. J. Smith et al., Array programming with NUMPY, Nature
priors, Phys. Rev. D 103, 123511 (2021). (London) 585, 357 (2020).
[62] H. Koo, R. E. Keeley, A. Shafieloo, and B. L’Huillier, [66] P. Virtanen, R. Gommers, T. E. Oliphant, M. Haberland,
Bayesian vs frequentist: Comparing Bayesian model selec- T. Reddy, D. Cournapeau, E. Burovski, P. Peterson,
tion with a frequentist approach using the iterative smooth- W. Weckesser, J. Bright (SciPy 1.0 Contributors), SCIPY
ing method, J. Cosmol. Astropart. Phys. 03 (2022) 047. 1.0: Fundamental algorithms for scientific computing in
[63] R. E. Keeley and A. Shafieloo, On the distribution of Bayesian PYTHON, Nat. Methods 17, 261 (2020).
evidence, Mon. Not. R. Astron. Soc. 515, 293 (2022). [67] J. D. Hunter, MATPLOTLIB: A 2d graphics environment,
[64] A. Lewis, GetDist: A PYTHON package for analysing Comput. Sci. Eng. 9, 90 (2007).
Monte Carlo samples, arXiv:1910.13970. [68] https://www.desi.lbl.gov/collaborating-institutions
[65] C. R. Harris, K. J. Millman, S. J. van der Walt, R. Gommers, [69] DESI Collaboration, DESI Data Release 1, https://data.desi.
P. Virtanen, D. Cournapeau, E. Wieser, J. Taylor, S. Berg, lbl.gov/doc/releases/ (2025).

023532-13