Non-standard neutrino cosmology dilutes the lensing anomaly

Ivan Esteban,1, 2, ∗ Olga Mena,3, † and Jordi Salvado4, ‡

1
Center for Cosmology and AstroParticle Physics (CCAPP), Ohio State University, Columbus, OH 43210
2
Department of Physics, Ohio State University, Columbus, OH 43210
3
IFIC, Universidad de Valencia-CSIC, 46071, Valencia, Spain
4
Departament de Física Quàntica i Astrofísica and Institut de Ciencies del Cosmos,
Universitat de Barcelona, Diagonal 647, E-08028 Barcelona, Spain
(Dated: February 11, 2022)
Despite the impressive success of the standard cosmological model, several anomalies defy its
triumph. Among them is the so-called lensing anomaly: the Planck satellite observes stronger
CMB gravitational lensing than expected. The role of neutrinos in this anomaly has been mostly
overlooked, despite their key role in CMB lensing, because in the standard scenario they tend to
arXiv:2202.04656v1 [astro-ph.CO] 9 Feb 2022

increase the tension. Here, we show that this strongly depends on the assumed neutrino equation of
state. We demonstrate that if neutrinos have yet undiscovered long-range interactions, the lensing
pattern is significantly affected, rendering the lensing anomaly as a pure statistical fluctuation. Our
results thus open up a window to link anomalous CMB lensing with present and future cosmological,
astrophysical, and laboratory measurements of neutrino properties.

I. INTRODUCTION at 3σ. CMB lensing also introduces a non-trivial four-

point correlation function, and therefore it can be inde-
The standard cosmological model, remarkably simple pendently measured. Adding this information somewhat
but with profound implications, remains strikingly suc- diminishes the tension, albeit the value of the lensing am-
cessful in many diverse environments [1–11]. The preci- plitude is still above the canonical one by about 2σ. The
sion is such, that self-consistency tests are feasible and ac- lensing anomaly is robust against changes in the fore-
curate. These tests check not only if individual datasets ground modeling in the baseline likelihood, and was al-
are consistent with the model, but also whether the same ready discussed in previous data releases, although it is
model simultaneously explains all observations or if, on currently more significant due to the lower reionization
the contrary, new ingredients are needed. optical depth preferred by the Planck 2018 data release.
It is in these consistency tests where the standard A recent result from the Atacama Cosmology Telescope
model may meet its match. Indeed, there exist sev- is compatible with Alens = 1 [11], but the results are
eral anomalies that can not be fully understood in the consistent with Planck within uncertainties.
minimal cosmological constant plus cold dark matter Barring systematic errors or a rare statistical fluc-
(ΛCDM) scenario. The most statistically significant one tuation, the lensing anomaly could have its origin in
is the discrepancy of the Hubble constant extracted from new physics scenarios. Attempts in the literature have
nearby universe probes and that derived from Cosmic Mi- changed either the geometrical or the gravitational sec-
crowave Background (CMB) measurements [12–14]. An- tors of the theory. A closed universe [18] has been shown
other one is related to the clustering parameter σ8 , whose to solve the internal tensions in Planck concerning the
values differ for CMB and weak lensing estimates [15]. cosmological parameter values at different angular scales,
The third one, the so-called lensing anomaly [7, 16], is alleviating the Alens anomaly. However, the positive cur-
the focus of this work. vature scenario is strongly rejected by other cosmolog-
CMB anisotropies get blurred due to gravitational lens- ical observations such as Baryon Acoustic Oscillations
ing by the large scale structure of the Universe: photons (BAO) [19]. Modified gravity theories or a modified
from different directions are mixed and the peaks at large primordial perturbation spectrum have also been con-
multipoles are smoothed. The amount of lensing is a pre- sidered as possible scenarios where to solve the Alens
cise prediction of the ΛCDM model: the consistency of anomaly [20–24], but it is also unclear if BAO measure-
the model can be checked by artificially increasing lensing ments are compatible with these solutions [23].
by a factor Alens [17] (a priori an unphysical parameter). Here, we show that non-standard neutrino properties
If ΛCDM consistently describes all CMB data, observa- can lead to unexpected lensing and dilute the preference
tions should prefer Alens = 1. for Alens 6= 1. Neutrinos are key to structure forma-
Intriguingly, Planck CMB data shows a preference for tion in our Universe, but due to a competition of differ-
additional lensing. Indeed, the reference analysis of tem- ent effects, in the standard paradigm they reduce CMB
perature and polarization anisotropies suggest Alens > 1 lensing [25, 26]. However, we demonstrate that this dra-
matically depends on the assumed equation of state for
the cosmological neutrino fluid. Under the presence of
∗ [email protected] long-range neutrino interactions [27], largely allowed by
† [email protected] other constraints [28–35], the generally assumed ideal gas
‡ [email protected] equation of state no longer holds. This pivotal difference
 2

makes neutrinos increase CMB lensing, notably allevi-

ating the Alens anomaly. Crucially, other cosmological Radiation
datasets such as BAO do not compromise this effect. 0.3
The structure of the manuscript is as follows. We start
in Sec. II by describing the neutrino interaction model 0.2
and its impact on the equation of state. In Section III, we

w = P/ρ
show how different cosmological observables are affected,
describing in detail the effect on CMB lensing. Section 0.1
IV presents our quantitative results on the anomaly. We
draw our conclusions and highlight future directions in Dust
0.0 P
mν = 3 eV
Sec. V. P
mν = 1 eV
P
mν = 3 eV; gmν/Mφ = 10
−0.1 P
mν = 3 eV; gmν/Mφ = 102
P
mν = 1 eV; gmν/Mφ = 10
II. LONG-RANGE NEUTRINO INTERACTIONS 2 3
10 10 104
z
We consider neutrino self-interactions mediated by a
very light scalar field φ, so that neutrinos source a
classical field making long-range interaction effects rel- Figure 1. Modified neutrino equation of state w due to long-
evant [27, 36–41]. The corresponding interaction La- range interactions (see Ref. [27] for the numerical computa-
grangian is tion). As w controls the rate of neutrino energy density di-
lution, ρ̇/ρ = −3H(1 + w), the neutrino energy at different
redshifts is dramatically affected, which directly impacts many
Lint = −gφν̄ν , (1) cosmological observables.

and long-range effects are present if Mφ & 10−25 eV (oth-

erwise, the scalar field would act as quintessence). For
simplicity, we consider one scalar field universally cou- III. IMPACT ON CMB OBSERVABLES
pled to all three neutrino mass eigenstates, that we as-
sume to have a common mass mν (cosmological effects The modified equation of state directly impacts the
of neutrino mass splittings are very challenging to de- precisely measured CMB anisotropies. As we shall show
tect [42–44]). SN 1987A and other laboratory constraints below, it also increases the observed lensing.
impose g . 10−7 [28–35], which guarantees that the neu- Figure 2 shows how neutrino masses and interactions
trino Fermi-Dirac distributions are not distorted and that impact CMB temperature anisotropies. The bottom
scattering processes can be neglected [45]. In this sce- panel shows the anisotropy ratios to the canonical ΛCDM
nario, the interaction strength is governed by the ratio scenario with massless neutrinos, together with Planck
gmν /Mφ [27]. data. In this figure, we have fixed the cosmological pa-
The impact of long-range interactions on cosmic neu- rameters {θs , ωb , ωcdm , As , ns , τreio }.
trinos is simple (see Ref. [27] for more details): when The main effects of long-range interactions arise due to
the density is large and the typical energy is of the order the Integrated Sachs Wolfe effect (for ` . 500), directly
of the neutrino mass, neutrinos source a classical scalar sensitive to the neutrino equation of state; and a reduced
field. This, in turn, induces an effective neutrino mass Silk damping scale when the neutrino energy density red-
m̃ = mν + gφ 6= mν via the Lagrangian, see Eq. (1). The shifts as dust (visible as an enhancement for large `).
energy and pressure of the scalar field are also relevant, These are also the standard effects of neutrino masses on
and they modify the equation of state. Finally, when neu- CMB temperature anisotropies [27, 53], but since they
trinos become non-relativistic, the scalar field induces an can be traced back to how the neutrino equation of state
attractive force among them. This abruptly condenses gets modified when they become non-relativistic, they
all cosmic neutrinos into lumps of size . Mφ−1 [46–52]. are directly impacted by long-range interactions. As a
Figure 1 summarizes how long-range interactions af- consequence, strong long-range interactions completely
fect the evolution of cosmic neutrinos by changing their remove the CMB bound on neutrino masses [27].
equation of state (defined as the ratio between pressure Interestingly, we also observe additional wiggles at
and energy density) as a function of redshift z. Under the large `, that mimic the scenario with artificially modified
presence of a long-range interaction, neutrinos behave as lensing Alens 6= 1. The effect of non-interacting massive
radiation for longer times (because m̃ < mν due to scalar neutrinos is out of phase with Alens > 1, indicating that
interactions being attractive). Later, the negative pres- neutrino masses typically reduce CMB lensing as we de-
sure of the scalar field promptly changes the equation of scribe below. However, as we show below, introducing a
state to that of dust. This effect is further enhanced by long-range interaction increases the lensing. This effect
the abrupt formation of lumps. is key to our results.
 3

1.0 2

C`
ΛCDM
`(`+1)
2π
0.8

Cφφ × 107
1.5
1.04

0.6 1 ΛCDM
1.02
C` /C`ΛCDM

Alens = 1.07±0.04

[`(`+1)]2
P
mν = 3 eV

2π
0.4
1 P
mν = 3 eV; gmν /Mφ = 10
0.5
mν = 3 eV; gmν /Mφ = 102
P
Alens = 1.180±0.065
P P
0.98 mν = 3 eV; gmν/Mφ = 10 mν = 1 eV; gmν /Mφ = 10
0.2 P
mν = 3 eV; gmν/Mφ = 102
P 0
mν = 1 eV; gmν/Mφ = 10 1.1

Ratio
0.96 P
mν = 1 eV
0.0 1.0 CMB-S4
0.0 30 0.2500 0.4
1000 0.6
1500 0.8
2000 1.0
102
`
`

Figure 2. Modified CMB temperature anisotropies by mas-

Figure 3. Modified CMB lensing power spectrum by massive
sive and interacting neutrinos, together with the prediction
and interacting neutrinos, and the prediction for Alens 6= 1.
for Alens 6= 1. Neutrino long-range interactions mimic an
Unlike neutrino masses, neutrino long-range interactions en-
enhanced lensing contribution, as preferred by the data.
hance lensing, closely mimicking Alens > 1. Future observa-
tions could discriminate among these two effects (see text).

Indeed, the orange shaded region in Fig. 2 corresponds

to the massless neutrino case with the value of Alens in-
ferred from a fit to Planck temperature and polarization the last two effects are subdominant, particularly at large
measurements [7]. As we see, it lies very close to the `. Overall, neutrino masses tend to reduce CMB lensing.
interacting neutrino scenario. However, it is not fully de- However, long-range interactions flip this effect. The
generate as long-range interactions introduce additional relativistic to non-relativistic transition is more abrupt:
effects (particularly a reduced Silk damping). This allows neutrinos move close to the speed of light for a shorter
future observations to disentangle long-range interactions time (c.f. Fig. 1). The energy density in neutrinos is also
from other solutions to the lensing anomaly. smaller as scalar interactions are attractive [27]. There-
Figure 3 shows how neutrino masses and interactions fore, the physical effects that suppress CMB lensing are
affect the CMB lensing pattern. The bottom panel shows diminished; and those enhancing it are increased. The
the ratios to the canonical ΛCDM scenario with massless overall effect is now the opposite: neutrino masses and
neutrinos. The data points in the upper plot correspond interactions enhance the lensing power spectrum.
to Planck [54], while in the bottom panel they refer to Intriguingly, we note
P from Fig. 3 that the effect of mas-
the errors expected by the next-generation ground-based sive neutrinos with mν = 3 eV and gmν /Mφ = 102 is
CMB experiment CMB-S4 [55, 56]. The orange band almost degenerate with the massless neutrino case with
shows the enhanced lensing preferred by a fit to Planck Alens = 1.07 ± 0.04. Consequently, it is always possible
CMB temperature, polarization and lensing [7]. to tune the neutrino mass and the interaction strength
In the standard scenario, neutrinos reduce CMB lens- to produce a CMB lensing pattern very similar to that
ing because they are hot thermal relics with a very large obtained with a non-zero value of the Alens parameter.
velocity dispersion, and therefore they suppress cluster- Next-generation CMB observations, as those from CMB-
ing at small scales. Furthermore, when they become non- S4, might disentangle between these two effects, as the
relativistic their energy density dilutes slower, enhancing error bars in the bottom panel of Fig. 3 show [57].
the expansion of the Universe and further suppressing The presence of a long-range interaction in the neu-
structure formation. However, they also have lensing- trino sector thus generates very rich physics. Its effects
enhancing effects. On the one hand, they cluster as mat- turn on at a specific time (when neutrino masses start
ter at large scales, enhancing density perturbations and to be relevant), and make neutrinos behave as radiation
thus the lensing. On the other hand, their modified equa- even if they have large masses, at the same time en-
tion of state before recombination would affect the well- hancing CMB lensing as neutrinos increase perturbation
measured CMB sound horizon, which must be compen- growth in a different way than they would do in the non-
sated by slightly decreasing H0 , and a slower expansion interacting case. This makes long-range neutrino inter-
rate of the Universe also enhances structure formation. actions a very attractive and predictive scheme to solve
Without long-range interactions (dotted line in Fig. 3), the Alens anomaly.
 4

∆χ2eff = χ2eff (Alens = 1) − χ2eff (Alens 6= 1)

TTTEEE+lowl TTTEEE+lowl+lensing TTTEEE+lowl+lensing+BAO

9.66 3.43 4.26
Plik (reference): ΛCDM
p = 0.2% p = 6% p = 4%
4.87 0.76 2.71
Plik (reference): Self-interactions
p = 3% p = 38% p = 10%
4.82 2.01 1.96
CamSpec (alternative): ΛCDM
p = 3% p = 16% p = 16%
2.06 1.39 1.79
CamSpec (alternative): Self-interactions
p = 15% p = 24% p = 18%

Table I. Significance of the lensing anomaly without and with neutrino long-range self-interactions, for the two official Planck
likelihoods (see text for details). Introducing long-range interactions removes the significant preference for extra lensing, ren-
dering the CMB data self-consistent with the assumed cosmology.

IV. METHODOLOGY AND RESULTS figure of merit also avoids the large Bayesian volume ef-
fects associated with exploring these interactions [27].
Below, we quantify how neutrino long-range interac- We compute cosmological observables with the ex-
tions alleviate the Alens anomaly. For this, we use the tension of CLASS [60, 61] from Ref. [27] (publicly
following datasets: available at https://github.com/jsalvado/class_
public_lrs). We compute the ∆χ2eff values by modi-
• The CMB anisotropies as observed by Planck. We fying the MontePython [62] code, and we carry out all
include temperature and polarization power spec- our minimizations with the BOBYQA minimizer [63] and
tra from the final 2018 data release, at both high its Python implementation [64]. This minimizer is ro-
and low multipoles (TTTEEE+lowl+lowE) [58]. bust against non-smooth likelihoods, making it ideal for
We employ both the reference Plik and alterna- the problem studied here [65, 66].
tive CamSpec likelihoods, which among other dif- Table I shows our main results. As described above,
ferences use different polarization masks, dust sub- we compute the values of ∆χ2eff for different datasets and
traction and calibration; and compare the results for the two Planck likelihoods, the reference Plik and
arising from both of them. Below, we refer to this the alternative CamSpec; for ΛCDM and with long-range
dataset as TTTEEE+lowl. neutrino interactions. In theP former case, we follow the
Planck analysis and assume mν =P0.06 eV [7]. In the
• The Planck CMB lensing reconstruction from the latter, both the total neutrino mass mν and the inter-
four-point correlation function [59]. Below, we refer action strength gmν /Mφ are free parameters to be deter-
to this dataset as Lensing. mined by data. We also show the frequentist p−values
for excluding Alens = 1. (Below, we further quantify the
• BAO observations from several galaxy clustering
tension with the Akaike Information Criterion.)
surveys. Specifically, we use the 6dFGS [1], the
We first note that when no long-range forces are
SDSS-DR7 MGS [2], and the BOSS DR12 [4] mea-
present in the neutrino sector, the fit to CMB tempera-
surements. These data points are isotropic BAO
ture and polarization anisotropy data greatly improves if
measurements of DV (z)/rd (with DV (z) and rd
we introduce the extra parameter Alens . From a frequen-
the spherically averaged volume distance and sound
tist point of view, there is a 3σ preference for Alens 6= 1
horizon at baryon drag, respectively) in the case
if we use the Plik likelihood, and somewhat smaller for
of 6dFGS and MGS, while BOSS DR12 is an
Camspec. Interestingly, this preference drops drastically
anisotropic BAO measurement of DH (z)/rd (with
when considering additional data sets, indicating some
DH (z) = c/H(z) the radial distance). Below, we
internal inconsistency.
refer to this dataset as BAO.
The preference for Alens 6= 1 significantly dilutes when
Following the work done by the Planck collaboration [7], we include neutrino long-range interactions. Its statisti-
we quantify the significance of the lensing anomaly by cal significance merely reaches the ∼ 1.5σ–2σ level (or
computing < 1σ if we consider Planck temperature, polarization
and lensing data), and therefore it can be interpreted
∆χ2eff = χ2eff (Alens = 1) − χ2eff (Alens 6= 1) , (2) as a simple statistical fluctuation. Furthermore, the val-
ues of ∆χ2eff become more uniform and consistent among
where χ2eff is the value of χ2 after minimizing over all the different datasets. This is unlike other proposed so-
cosmological and nuisance parameters. A larger ∆χ2eff lutions to the lensing anomaly such as curvature, where
corresponds to a more significant anomaly. Using this BAO data fully restores the preference towards Alens 6= 1.
 5

In order to further assess the impact of our results, we V. CONCLUSIONS AND FUTURE PROSPECTS
also use one of the information criteria widely exploited
in astrophysical and cosmological studies to discriminate The ΛCDM picture has provided an extremely good
between competing models (see e.g., Refs. [67, 68]): the explanation of present cosmological observations. Nev-
frequentist Akaike Information Criterion (AIC) ertheless, it relies on very simplistic assumptions on the
different ingredients, reflecting our ignorance on the mi-
AIC ≡ χ2eff + 2k , (3) crophysics of the different components of the Universe.
Namely, dark energy is simply described by a cosmologi-
where k is the number of free parameters in the model; cal constant representing the vacuum energy, dark matter
the second term penalizes models with more free pa- is merely assumed to be a non-interacting dust, and neu-
rameters. The best model is the one with the small- trinos are relativistic species that transition into dust as
est AIC. Following Jeffreys’ scale, we rate a difference an ideal gas.
∆AIC > 5 or > 10 as strong or decisive evidence, respec- Deviations from this simple and economical scenario
tively, against the disfavored model [67]. are thus natural as data gets more precise. These may
This statistical criterion also reflects that neutrino show up in self-consistency tests that start out as odd-
long-range interactions significantly dilute the lensing ball anomalies and then evolve into incontrovertible ev-
anomaly. Focusing on the TTTEEE+lowl dataset and idence pointing to either uncontrolled uncertainties or
the Plik likelihood, ∆AIC = 7.66, and therefore Alens = new physics. The fact that cosmology probes extreme
1 is penalized in a strong-to-decisive manner with re- densities also implies that phenomena beyond the Stan-
spect to the case in which the lensing amplitude is a free dard Model of Particle Physics could first appear there.
parameter. For the very same data set and likelihood, In this work, we have explored if the simplistic and
∆AIC < 5 when long-range interactions are switched on, minimal ΛCDM scenario could alleviate some of its cur-
and therefore no significant claim can be stated concern- rent anomalies if the neutrino sector is richer, leading to
ing a preference for Alens 6= 1. Similar results hold unexpected effects. This is somewhat expected, as neu-
for other datasets and the CamSpec likelihood. A non- trinos have provided so far the only laboratory evidence
minimal neutrino sector thus removes the preference for for physics beyond the Standard Model [70–78].
extra artificial lensing, making CMB data consistent with An especially rich cosmological observable is the CMB
the assumed cosmological history of the Universe. lensing. Neutrinos substantially affect it, to the ex-
tent that present and future CMB constraints on neu-
TTTEEE
TTTEEE trino masses strongly rely on measuring it [79]. And, as
TTTEEE
+lowl
+lowl we have shown, deviations from the minimal ideal gas
+lowl
+lensing
+ lensing scheme notably modify the canonical expectations.
+BAO A non-ideal neutrino equation of state thus naturally
provides a physical framework where to solve the so-
P
Plik mν [eV] 0.8 2.9 0.08
(reference) g mν /Mφ 13 317 2 · 103 called lensing (Alens ) anomaly. We have shown that the
presence of a long-range interaction in the neutrino sec-
P
CamSpec mν [eV] 0.7 0.07 0.06
(alternative) g mν /Mφ 16 7 15 tor makes CMB observations perfectly consistent with
the canonical value Alens = 1. In addition, the interac-
Table II. Neutrino mass and long-range interaction strength tion strength required to decrease the lensing anomaly
preferred by different datasets. These values alleviate the down to a pure statistical fluctuation is perfectly consis-
lensing anomaly. tent with other existing bounds.
Our scenario can be easily explored with other cos-
mological observations. Future CMB measurements will
Table II shows the neutrino parameters that allevi- have superb precision, and they could discriminate be-
ate the lensing anomaly. We P show the best-fit values tween neutrino interactions and uniformly enhanced lens-
for the total neutrino mass mν and the long-range in- ing [55, 57]. Furthermore, current surveys are rapidly im-
teraction strength gmν /Mφ assuming a canonical lensing proving our precision of the matter power spectrum [10,
amplitude Alens = 1, for both likelihoods and for the 80], to the extent that they could start being sensitive
three possible data sets here exploited. We note that, to effects induced by neutrino self-interactions [27]. The
in general, CMB data prefers moderate interaction rates light new scalar responsible for the long-range interac-
gmν /Mφ ∼ 10–103 . The preferred neutrino masses, still tion could also give rise to observable departures from
lying in the sub-eV region compatible with laboratory standard cosmology [36–41].
data [69], typically exceed the current cosmological lim- More interestingly, due to the direct link to particle
its for non-interacting neutrinos. More precise measure- physics, neutrino long-range interactions can leave im-
ments of the neutrino mass in the laboratory, which are prints on astrophysical and laboratory probes. Environ-
more model-independent [27], could thus shed comple- ments with a large neutrino density such as the Sun or
mentary light onto these models and their early Universe core-collapse supernovae might get affected [81], and the
phenomenology. precise data that future neutrino observatories will gather
 6

opens exciting prospects [82–88]. Our scenario also pre- ΛCDM paradigm will be profoundly scrutinized, and any
dicts that the entire cosmic neutrino background should inconsistency will shed light onto the most elusive part
be condensed in relatively dense lumps that could give of our Universe. Neutrinos, ordinarily ghostly particles,
additional signals [89–91]. Furthermore, in minimal re- may be the path to new cosmological models, with in-
alizations of our model laboratory neutrino mass con- triguing links to astrophysics and particle physics obser-
straints are not affected [27]. Thus, the relatively large vations.
neutrino masses required to alleviate the lensing anomaly
could generate unexpected signals in beta decay [92] or
neutrinoless double-beta decay [93] searches. Addition- ACKNOWLEDGMENTS
ally, variations of our model could also induce neutrino
decay [34, 45, 94–107]. We are grateful for helpful discussions with John
Finally, we have only considered the phenomenol- Beacom, Miguel Escudero, Chris Hirata and Gabriel
ogy induced by the minimal Lagrangian (1). Elec- Vasquez. OM is supported by the Spanish grants
troweak gauge invariance, or a mechanism that ensures FPA2017-85985-P, PROMETEO/2019/083 and by the
the lightness of the mediator, can only be achieved in European ITN project HIDDeN (H2020-MSCA-ITN-
a UV-complete model with extra ingredients (see, e.g., 2019//860881-HIDDeN). JS is supported from the
Refs. [29, 108–110]). This may bring up additional phe- European ITN project H2020-MSCAITN-2019/860881-
nomenology in other cosmological, astrophysical, or lab- HIDDeN, the Spanish grants PID2019-108122GBC32,
oratory observations. PID2019-105614GB-C21, and to the Unit of Excellence
We are living exciting times from cosmology. The María de Maeztu 2020-2023 (CEX2019-000918-M).

[1] F. Beutler, C. Blake, M. Colless, D. H. Jones, arXiv:1803.11526 [astro-ph.CO].

L. Staveley-Smith, L. Campbell, Q. Parker, W. Saun- [17] E. Calabrese, A. Slosar, A. Melchiorri, G. F. Smoot,
ders, and F. Watson, Mon. Not. Roy. Astron. Soc. 416, and O. Zahn, Phys. Rev. D 77, 123531 (2008),
3017 (2011), arXiv:1106.3366 [astro-ph.CO]. arXiv:0803.2309 [astro-ph].
[2] A. J. Ross, L. Samushia, C. Howlett, W. J. Percival, [18] E. Di Valentino, A. Melchiorri, and J. Silk, Nature As-
A. Burden, and M. Manera, Mon. Not. Roy. Astron. tron. 4, 196 (2019), arXiv:1911.02087 [astro-ph.CO].
Soc. 449, 835 (2015), arXiv:1409.3242 [astro-ph.CO]. [19] G. Efstathiou and S. Gratton, Mon. Not. Roy. Astron.
[3] A. G. Riess et al., Astrophys. J. 826, 56 (2016), Soc. 496, L91 (2020), arXiv:2002.06892 [astro-ph.CO].
arXiv:1604.01424 [astro-ph.CO]. [20] E. Di Valentino, A. Melchiorri, and J. Silk, Phys. Rev.
[4] S. Alam et al. (BOSS), Mon. Not. Roy. Astron. Soc. D 93, 023513 (2016), arXiv:1509.07501 [astro-ph.CO].
470, 2617 (2017), arXiv:1607.03155 [astro-ph.CO]. [21] H. Moshafi, S. Baghram, and N. Khosravi, Phys. Rev.
[5] D. M. Scolnic et al. (Pan-STARRS1), Astrophys. J. 859, D 104, 063506 (2021), arXiv:2012.14377 [astro-ph.CO].
101 (2018), arXiv:1710.00845 [astro-ph.CO]. [22] G. Domènech and M. Kamionkowski, JCAP 11, 040
[6] T. M. C. Abbott et al. (DES), Astrophys. J. Lett. 872, (2019), arXiv:1905.04323 [astro-ph.CO].
L30 (2019), arXiv:1811.02374 [astro-ph.CO]. [23] G. Domènech, X. Chen, M. Kamionkowski, and
[7] N. Aghanim et al. (Planck), Astron. Astrophys. 641, A6 A. Loeb, JCAP 10, 005 (2020), arXiv:2005.08998 [astro-
(2020), [Erratum: Astron.Astrophys. 652, C4 (2021)], ph.CO].
arXiv:1807.06209 [astro-ph.CO]. [24] D. K. Hazra, A. Antony, and A. Shafieloo, (2022),
[8] V. de Sainte Agathe et al., Astron. Astrophys. 629, A85 arXiv:2201.12000 [astro-ph.CO].
(2019), arXiv:1904.03400 [astro-ph.CO]. [25] M. Kaplinghat, L. Knox, and Y.-S. Song, Phys. Rev.
[9] M. Blomqvist et al., Astron. Astrophys. 629, A86 Lett. 91, 241301 (2003), arXiv:astro-ph/0303344.
(2019), arXiv:1904.03430 [astro-ph.CO]. [26] J. Lesgourgues, L. Perotto, S. Pastor, and
[10] T. M. C. Abbott et al. (DES), Phys. Rev. D 105, 023520 M. Piat, Phys. Rev. D 73, 045021 (2006), arXiv:astro-
(2022), arXiv:2105.13549 [astro-ph.CO]. ph/0511735.
[11] S. Aiola et al. (ACT), JCAP 12, 047 (2020), [27] I. Esteban and J. Salvado, JCAP 05, 036 (2021),
arXiv:2007.07288 [astro-ph.CO]. arXiv:2101.05804 [hep-ph].
[12] E. Di Valentino, O. Mena, S. Pan, L. Visinelli, [28] A. P. Lessa and O. L. G. Peres, Phys. Rev. D 75, 094001
W. Yang, A. Melchiorri, D. F. Mota, A. G. Riess, (2007), arXiv:hep-ph/0701068.
and J. Silk, Class. Quant. Grav. 38, 153001 (2021), [29] N. Blinov, K. J. Kelly, G. Z. Krnjaic, and S. D.
arXiv:2103.01183 [astro-ph.CO]. McDermott, Phys. Rev. Lett. 123, 191102 (2019),
[13] N. Schöneberg, G. Franco Abellán, A. Pérez Sánchez, arXiv:1905.02727 [astro-ph.CO].
S. J. Witte, V. Poulin, and J. Lesgourgues, (2021), [30] P. S. Pasquini and O. L. G. Peres, Phys. Rev. D
arXiv:2107.10291 [astro-ph.CO]. 93, 053007 (2016), [Erratum: Phys.Rev.D 93, 079902
[14] E. Di Valentino et al., Astropart. Phys. 131, 102605 (2016)], arXiv:1511.01811 [hep-ph].
(2021), arXiv:2008.11284 [astro-ph.CO]. [31] M. Agostini et al., Eur. Phys. J. C 75, 416 (2015),
[15] E. Di Valentino et al., Astropart. Phys. 131, 102604 arXiv:1501.02345 [nucl-ex].
(2021), arXiv:2008.11285 [astro-ph.CO]. [32] K. Blum, Y. Nir, and M. Shavit, Phys. Lett. B 785,
[16] P. Motloch and W. Hu, Phys. Rev. D 97, 103536 (2018), 354 (2018), arXiv:1802.08019 [hep-ph].
 7

[33] T. Brune and H. Päs, Phys. Rev. D 99, 096005 (2019), constrained optimization without derivatives, Tech. Rep.
arXiv:1808.08158 [hep-ph]. DAMTP 2009/NA06 (University of Cambridge, 2009).
[34] M. Escudero and M. Fairbairn, Phys. Rev. D 100, [64] C. Cartis, J. Fiala, B. Marteau, and L. Roberts, ACM
103531 (2019), arXiv:1907.05425 [hep-ph]. Trans. Math. Softw. 45 (2019), 10.1145/3338517.
[35] V. Brdar, M. Lindner, S. Vogl, and X.-J. Xu, Phys. [65] P. A. R. Ade et al. (Planck), Astron. Astrophys. 571,
Rev. D 101, 115001 (2020), arXiv:2003.05339 [hep-ph]. A16 (2014), arXiv:1303.5076 [astro-ph.CO].
[36] R. Fardon, A. E. Nelson, and N. Weiner, JCAP 10, 005 [66] J. Prasad, (2014), arXiv:1412.3298 [astro-ph.CO].
(2004), arXiv:astro-ph/0309800. [67] A. R. Liddle, Mon. Not. Roy. Astron. Soc. 377, L74
[37] D. B. Kaplan, A. E. Nelson, and N. Weiner, Phys. Rev. (2007), arXiv:astro-ph/0701113.
Lett. 93, 091801 (2004), arXiv:hep-ph/0401099. [68] R. Trotta, Contemp. Phys. 49, 71 (2008),
[38] P. Gu, X. Wang, and X. Zhang, Phys. Rev. D 68, arXiv:0803.4089 [astro-ph].
087301 (2003), arXiv:hep-ph/0307148. [69] M. Aker et al. (KATRIN), Phys. Rev. Lett. 123, 221802
[39] R. Peccei, Phys. Rev. D 71, 023527 (2005), arXiv:hep- (2019), arXiv:1909.06048 [hep-ex].
ph/0411137. [70] B. Pontecorvo, Zh. Eksp. Teor. Fiz. 53, 1717 (1967).
[40] R. Bean and J. Magueijo, Phys. Lett. B 517, 177 (2001), [71] V. N. Gribov and B. Pontecorvo, Phys. Lett. B 28, 493
arXiv:astro-ph/0007199. (1969).
[41] R. Bean, Phys. Rev. D 64, 123516 (2001), arXiv:astro- [72] Q. R. Ahmad et al. (SNO), Phys. Rev. Lett. 89, 011301
ph/0104464. (2002), arXiv:nucl-ex/0204008 [nucl-ex].
[42] J. Lesgourgues and S. Pastor, Phys. Rept. 429, 307 [73] Q. R. Ahmad et al. (SNO), Phys. Rev. Lett. 89, 011302
(2006), arXiv:astro-ph/0603494. (2002), arXiv:nucl-ex/0204009 [nucl-ex].
[43] J. R. Pritchard and E. Pierpaoli, Phys. Rev. D 78, [74] S. N. Ahmed et al. (SNO), Phys. Rev. Lett. 92, 181301
065009 (2008), arXiv:0805.1920 [astro-ph]. (2004), arXiv:nucl-ex/0309004 [nucl-ex].
[44] R. Jimenez, T. Kitching, C. Pena-Garay, and L. Verde, [75] B. Aharmim et al. (SNO), Phys. Rev. C72, 055502
JCAP 05, 035 (2010), arXiv:1003.5918 [astro-ph.CO]. (2005), arXiv:nucl-ex/0502021 [nucl-ex].
[45] S. Hannestad and G. Raffelt, Phys. Rev. D 72, 103514 [76] R. Becker-Szendy et al., Phys. Rev. D46, 3720 (1992).
(2005), arXiv:hep-ph/0509278. [77] Y. Fukuda et al. (Kamiokande), Phys. Lett. B335, 237
[46] N. Afshordi, M. Zaldarriaga, and K. Kohri, Phys. Rev. (1994).
D 72, 065024 (2005), arXiv:astro-ph/0506663. [78] Y. Fukuda et al. (Super-Kamiokande), Phys. Rev. Lett.
[47] L. Beca and P. Avelino, Mon. Not. Roy. Astron. Soc. 81, 1562 (1998), arXiv:hep-ex/9807003 [hep-ex].
376, 1169 (2007), arXiv:astro-ph/0507075. [79] K. N. Abazajian et al. (Topical Conveners: K.N. Abaza-
[48] M. Kaplinghat and A. Rajaraman, Phys. Rev. D 75, jian, J.E. Carlstrom, A.T. Lee), Astropart. Phys. 63, 66
103504 (2007), arXiv:astro-ph/0601517. (2015), arXiv:1309.5383 [astro-ph.CO].
[49] O. E. Bjaelde, A. W. Brookfield, C. van de [80] M. E. Levi et al. (DESI), (2019), arXiv:1907.10688
Bruck, S. Hannestad, D. F. Mota, L. Schrempp, [astro-ph.IM].
and D. Tocchini-Valentini, JCAP 01, 026 (2008), [81] M. Cirelli, M. C. Gonzalez-Garcia, and C. Pena-Garay,
arXiv:0705.2018 [astro-ph]. Nucl. Phys. B 719, 219 (2005), arXiv:hep-ph/0503028.
[50] R. Bean, E. E. Flanagan, and M. Trodden, New J. [82] K. Abe et al., (2011), arXiv:1109.3262 [hep-ex].
Phys. 10, 033006 (2008), arXiv:0709.1124 [astro-ph]. [83] K. Abe et al. (Hyper-Kamiokande), (2018),
[51] C. Gao and A. Stebbins, (2021), arXiv:2110.07012 [hep- arXiv:1805.04163 [physics.ins-det].
ph]. [84] S. Andringa et al. (SNO+), Adv. High Energy Phys.
[52] A. Y. Smirnov and X.-J. Xu, (2022), arXiv:2201.00939 2016, 6194250 (2016), arXiv:1508.05759 [physics.ins-
[hep-ph]. det].
[53] J. Lesgourgues, G. Mangano, G. Miele, and S. Pas- [85] F. Capozzi, S. W. Li, G. Zhu, and J. F. Beacom, Phys.
tor, Neutrino Cosmology (Cambridge University Press, Rev. Lett. 123, 131803 (2019), arXiv:1808.08232 [hep-
2013). ph].
[54] N. Aghanim et al. (Planck), Astron. Astrophys. 641, A8 [86] K. Abe et al. (Hyper-Kamiokande), Astrophys. J. 916,
(2020), arXiv:1807.06210 [astro-ph.CO]. 15 (2021), arXiv:2101.05269 [astro-ph.IM].
[55] K. N. Abazajian et al. (CMB-S4), (2016), [87] B. Abi et al. (DUNE), (2020), arXiv:2002.03005 [hep-
arXiv:1610.02743 [astro-ph.CO]. ex].
[56] T. Brinckmann, D. C. Hooper, M. Archidiacono, J. Les- [88] F. An et al. (JUNO), J. Phys. G 43, 030401 (2016),
gourgues, and T. Sprenger, JCAP 01, 059 (2019), arXiv:1507.05613 [physics.ins-det].
arXiv:1808.05955 [astro-ph.CO]. [89] Y. Ayaita, M. Baldi, F. Führer, E. Puchwein, and
[57] F. Renzi, E. Di Valentino, and A. Melchiorri, Phys. Rev. C. Wetterich, Phys. Rev. D 93, 063511 (2016),
D 97, 123534 (2018), arXiv:1712.08758 [astro-ph.CO]. arXiv:1407.8414 [astro-ph.CO].
[58] N. Aghanim et al. (Planck), Astron. Astrophys. 641, A5 [90] S. Casas, V. Pettorino, and C. Wetterich, Phys. Rev.
(2020), arXiv:1907.12875 [astro-ph.CO]. D 94, 103518 (2016), arXiv:1608.02358 [astro-ph.CO].
[59] N. Aghanim et al. (Planck), Astron. Astrophys. 641, A6 [91] J. Alvey, M. Escudero, N. Sabti, and T. Schwetz,
(2020), arXiv:1807.06209 [astro-ph.CO]. (2021), arXiv:2111.14870 [hep-ph].
[60] D. Blas, J. Lesgourgues, and T. Tram, JCAP 07, 034 [92] A. Osipowicz et al. (KATRIN), (2001), arXiv:hep-
(2011), arXiv:1104.2933 [astro-ph.CO]. ex/0109033.
[61] J. Lesgourgues, (2011), arXiv:1104.2932 [astro-ph.IM]. [93] A. Giuliani, J. J. Gomez Cadenas, S. Pascoli, E. Pre-
[62] T. Brinckmann and J. Lesgourgues, Phys. Dark Univ. vitali, R. Saakyan, K. Schäffner, and S. Schönert
24, 100260 (2019), arXiv:1804.07261 [astro-ph.CO]. (APPEC Committee), (2019), arXiv:1910.04688 [hep-
[63] M. J. D. Powell, The BOBYQA algorithm for bound ex].
 8

[94] Z. G. Berezhiani, G. Fiorentini, M. Moretti, and [102] M. Lindner, T. Ohlsson, and W. Winter, Nucl. Phys.
A. Rossi, Z. Phys. C 54, 581 (1992). B 622, 429 (2002), arXiv:astro-ph/0105309.
[95] J. F. Beacom and N. F. Bell, Phys. Rev. D 65, 113009 [103] M. Kachelriess, R. Tomas, and J. W. F. Valle, Phys.
(2002), arXiv:hep-ph/0204111. Rev. D 62, 023004 (2000), arXiv:hep-ph/0001039.
[96] V. D. Barger, J. G. Learned, S. Pakvasa, and T. J. [104] S. Ando, Phys. Lett. B 570, 11 (2003), arXiv:hep-
Weiler, Phys. Rev. Lett. 82, 2640 (1999), arXiv:astro- ph/0307169.
ph/9810121. [105] G. L. Fogli, E. Lisi, A. Mirizzi, and D. Montanino,
[97] G. L. Fogli, E. Lisi, A. Marrone, and G. Scioscia, Phys. JCAP 06, 012 (2006), arXiv:hep-ph/0603033.
Rev. D 59, 117303 (1999), arXiv:hep-ph/9902267. [106] M. Maltoni and W. Winter, JHEP 07, 064 (2008),
[98] V. D. Barger, J. G. Learned, P. Lipari, M. Lusignoli, arXiv:0803.2050 [hep-ph].
S. Pakvasa, and T. J. Weiler, Phys. Lett. B 462, 109 [107] M. Escudero, J. Lopez-Pavon, N. Rius, and S. Sandner,
(1999), arXiv:hep-ph/9907421. JHEP 12, 119 (2020), arXiv:2007.04994 [hep-ph].
[99] S. Choubey, N. P. Harries, and G. G. Ross, Phys. Rev. [108] K. Blum, A. Hook, and K. Murase, (2014),
D 74, 053010 (2006), arXiv:hep-ph/0605255. arXiv:1408.3799 [hep-ph].
[100] S. Palomares-Ruiz, S. Pascoli, and T. Schwetz, JHEP [109] J. M. Berryman, A. De Gouvêa, K. J. Kelly,
09, 048 (2005), arXiv:hep-ph/0505216. and Y. Zhang, Phys. Rev. D 97, 075030 (2018),
[101] M. Dentler, I. Esteban, J. Kopp, and P. Machado, Phys. arXiv:1802.00009 [hep-ph].
Rev. D 101, 115013 (2020), arXiv:1911.01427 [hep-ph]. [110] K. J. Kelly, M. Sen, W. Tangarife, and Y. Zhang, Phys.
Rev. D 101, 115031 (2020), arXiv:2005.03681 [hep-ph].