Prepared for submission to JHEP

Pion dark matter in a θ vacuum: a thermal relic with

sharp velocity-dependent self-interactions
arXiv:2508.21121v1 [hep-ph] 28 Aug 2025

Camilo Garcı́a-Cely,a Giacomo Landini,b Luca Marsili,c Óscar Zapata d

a
Instituto de Fı́sica Corpuscular (IFIC), Consejo Superior de Investigaciones Cientı́ficas (CSIC)
and Universitat de València, C/ Catedratico Jose Beltran 2, E-46980 Paterna, Spain
d
Instituto de Fı́sica, Universidad de Antioquia,
Calle 70 # 52-21, Apartado Aéreo 1226, Medellı́n, Colombia

Abstract: As recently proposed, a non-vanishing topological angle may play a central

role in QCD-like theories of dark matter (DM). In this work, we introduce a dark pho-
ton portal to the Standard Model in order to establish thermal equilibrium in the early
Universe, and discuss the ensuing phenomenological constraints, including the stability of
DM. The resulting dynamics accounts for the observed DM relic abundance and yields
velocity-dependent DM self-interactions in astrophysical halos. Due to the sharp velocity
dependence arising from a Breit-Wigner resonance, dedicated studies are required to assess
the gravothermal evolution in detail, especially in the core-collapse regime. This is partic-
ularly timely in light of self-interacting DM interpretations of strong-lensing systems such
as SDSS J0946+1006, which can be naturally explained within our framework.

Keywords: Models for dark matter, Cosmology of Theories BSM, New Gauge Interactions
Contents

1 Introduction 1

2 Pion DM in a θ vacuum 2
2.1 Pion DM for unitary gauge groups 2
2.2 Our benchmark model 5

3 The relic density in the presence of the θ vacuum 8

3.1 The evolution with Tdark = T 8
3.2 Departure from kinetic equilibrium 13

4 Portal to the SM 14
4.1 Dark photon portal and DM decays 14
4.2 Thermalization 18
4.3 Viable parameter space 19

5 Impact of the θ vacuum in DM halos 22

5.1 Reconciling the relic density with velocity-dependent scattering 22
5.2 The SIDM parameter space 23

6 Conclusions 25

1 Introduction

Understanding dark matter (DM) remains one of the most pressing open problems in
modern science. DM in the form of dark pions [1] constitutes a compelling solution given
its similarity with the QCD sector of the Standard Model (SM). In a recent work [2],
it was pointed out that the topological theta angle of the dark sector can substantially
modify the dynamics of DM pions. Concretely, Ref. [2], hereafter referred to as Paper I,
demonstrated that a non-vanishing theta angle induces number-changing processes among
the DM pions. As a result, their freeze-out can thermally give rise to the observed DM relic
density, particularly in the presence of resonant effects. See also Ref. [3], which identified
the number-changing processes associated with the θ angle, but reproduces the relic density
with a mechanism independent of θ, as well as Refs. [4–29] for related work on MeV DM,
and Refs. [30–32] for the effect of the θ angle in QCD-like DM models in other contexts.
In this paper, we revisit the framework proposed in Ref. [2] and address an aspect
that remained unexplored, namely, the interactions between DM and ordinary particles of
the SM. A thorough investigation is crucial for the following reasons: on the one hand,
such interactions are essential for understanding the thermal processes leading to the relic
abundance, which inherently assume that the DM sector thermalizes with the SM sector.

–1–
On the other hand, with the portal interaction specified, the number-changing processes
in the DM sector can lead to indirect detection signatures of DM, potentially including its
decay. One must therefore ensure the DM meta-stability.
In contrast to the previous work [2], this analysis is necessarily model-dependent, as
multiple realizations are possible [3, 4, 19, 22] depending on the portal that communicates
DM and the SM fields. To carry out our study, and for the sake of demonstrating that there
exists at least one model that satisfies all current constraints, we focus on the benchmark
model BM1 introduced in the Paper I. We extend this model by connecting it to the SM
via a massive dark photon [33], endowing the dark sector with an additional U (1)d gauge
interaction.
We examine the Boltzmann equations relevant to determining the DM relic density
and justify several assumptions made in the Paper I. In fact, we will show that, although
correct, such assumptions were overly restrictive and, to a certain extent, unnecessary.
This refinement has implications for the allowed theta-angle parameter space and may
impact predictions for DM halo structures. For instance, we find a new family of solutions
reproducing the observed relic density.
Moreover, the theta angle can also lead to modifications in the structure of small-
scale DM halos today [34, 35], while still agreeing with the predictions of collisionless dark
matter (CDM) on larger scales [36–40]. Thus, the scenario naturally realizes the self-
interacting dark matter (SIDM) paradigm [34]. When both effects are taken into account,
one obtains a viable model of SIDM wherein the relic density is governed by the same
physics responsible for altering DM halo structures. This result is highly non-trivial, as it
is often believed that strongly coupled models of DM, while potentially predicting DM self-
interactions, cannot simultaneously account for velocity-dependent scattering and therefore
cannot capture the size-dependent effects in halo dynamics. Comprehensive reviews can
be found in Refs. [35, 41].
In particular, we now include a discussion of systems such as SDSS J0946+1006 [42],
whose properties suggest an SIDM halo undergoing gravothermal collapse, see e.g [43],
and which can be naturally explained within our framework through its sharp velocity
dependence with enhanced cross sections at intermediate scales and suppressed interactions
in clusters.
The structure of this paper is as follows. We begin by reviewing QCD-like theories
of DM in Section 2. In Section 3, we discuss the Boltzmann equations in full generality,
maintaining a model-independent perspective. The portal to the Standard Model is intro-
duced in Section 4, where we examine thermalization, DM stability, and indirect detection
constraints. In Section 5, we discuss the implications in DM halos. Finally, we present our
conclusions in Section 6.

2 Pion DM in a θ vacuum

2.1 Pion DM for unitary gauge groups

We consider QCD-like theories with gauge group SU (Nc ) and Nf flavors of light Dirac
quarks q in the fundamental representation. Our results can be easily generalized to other

–2–
gauge group choices such as SO(Nc ) and Sp(Nc ). Then, the Lagrangian of the theory is
given by
1 g2θ
L = − Fµν F µν + / − (q̄L M qR + h.c.) ,
Fµν Feµν + q̄iDq (2.1)
4 32π 2
where M = diag(m1 , . . . , mNf ) is the quark mass matrix with mq ̸= 0 and real for all
flavors. Strong interactions confine at some energy scale Λ and generate a fermion con-
densate, ⟨q̄q⟩ ∼ Λ3 [44], in analogy with standard QCD. If mq ≪ Λ, this Lagrangian
exhibits an approximate SU (Nf )L × SU (Nf )R symmetry, corresponding to independent
chiral transformations of the left- and right-handed fermion components in flavor space. In
particular, an axial transformation takes the form
qL → e−iθQ/2 qL , and qR → eiθQ/2 qR , (2.2)
where Q is an arbitrary Nf × Nf hermitian matrix. Being anomalous under the SU (Nc )
gauge interactions, such a transformation induces not only a change in the mass matrix
but also a shift in the θ angle. Concretely,
θ → θ(1 − Tr Q) , and M → Mθ = eiθQ/2 M eiθQ/2 . (2.3)
Hence, one can effectively move the θ parameter from the F F̃ term into the mass matrix
by imposing
Tr Q = 1 . (2.4)
In particular, one may take Q diagonal such that Q = diag(q1 , .., qNf ).
It is believed that the fermion condensate spontaneously breaks the chiral symmetry
to its diagonal subgroup, SU (Nf )L × SU (Nf )R → SU (Nf )V . This leads to the emergence
of Nf2 − 1 pseudo-Goldstone bosons, π a , whose dynamics at sufficiently low energies is
described in chiral perturbation theory [45, 46] by
fπ2 f2 iπ a λa
Leff = Tr[∂µ U † ∂ µ U ] + π B0 Tr[Mθ† U + U † Mθ ] , with U =e fπ . (2.5)
4 2
We normalize the generators of SU (Nf ) such that tr[λa λb ] = 2δ ab , and parametrize the
fermion condensate in terms of the dark meson constant, fπ , as ⟨q̄q⟩ = −B0 fπ2 . Note that
√
it is expected that the confinement scale is of the order Λ ∼ 4πfπ / Nc [47].
Let us emphasize a few points. The broken generators of the axial SU (Nf )A subgroup,
corresponding to the chiral transformations, qL,R → exp[∓iαa λa ]qL,R , are in a one-to-one
correspondence with the light pseudo-scalar mesons πa . Also, the SU (Nf )V symmetry is
an exact symmetry only in the limit of degenerate quarks; in general, it is explicitly broken
by the differing values of the quark masses and is only approximately conserved.

Cancellation of tadpoles. The vacuum is unstable unless the linear terms in π a are
absent from Leff . These terms cancel when Tr[Mθ λa − λa Mθ† ] = 0, or equivalently,
Tr[sin(θQ)M λa ] = 0 for every a. Since the matrices λa generate the space of traceless
Nf × Nf hermitian matrices, sin(θQ)M must be proportional to the identity. This hap-
pens if there exists a function α(θ) such that
α(θ)
sin(θQ) = α(θ) M −1 , or in components mi = . (2.6)
sin(qi θ)

–3–
In particular, one can cast Mθ in Eq. (2.3) as
Mθ = cos(θQ)M + iα(θ) 11 , (2.7)
Likewise, the M -dependent part of the effective Lagrangian is
f 2 B0  
Leff ⊃ π Tr[cos(θQ)M (U + U † )] − iα(θ)Tr[U − U † ] (2.8)
2
f 2 B0 α(θ)  
= π Tr[cot(θQ)(U + U † )] − iTr[U − U † ] .
2
The function α(θ) can be found imposing the condition in Eq. (2.4). One implicit solution1
is
 
X α(θ) 1
Q = arcsin α(θ)M −1 .


θ= arcsin , and (2.9)
mi θ
i
Explicit solutions can be found in two important cases:
• For a quark degenerate mass spectrum, mi = m, for which
1
α(θ) = m sin (θ/Nf ) , and Q= 11 . (2.10)
Nf
This regime was studied in Ref. [3]. Since all meson masses are identical in this case,
several effects are not captured; in particular, no resonant effects can arise.

• For θ ≪ 1, according to Eq. (2.6), α(θ) ≪ 1 and hence

θ M −1
α(θ) = , and Q = . (2.11)
TrM −1 TrM −1
This regime for θ was assumed in the Paper I ; however, it was merely a simplifying
assumption. As we shall see, allowing for arbitrary θ does not lead to significant
differences.
Note that the Nf +1 parameters θ and mi can be exchanged for θ, α and qi (i = 1, . . . , Nf −
1) taking them as independent.

Meson mass spectrum. According to Eq. (2.8), the meson masses follow from
2 1  
(Mmeson )ab = B0 α(θ)Tr cot(θQ) {λa , λb } , (2.12)
2
with a, b = 1, . . . , Nf . In the particular cases for which |qi θ| < π/2,
p
Eq. (2.6) allows to write cos(qi θ) = 1 − (α(θ)/mi )2 . This leads to α(θ) cot(qi θ) =
p
mi 1 − (α(θ)/mi )2 , and hence
2 1 p 
(Mmeson )ab = B0 Tr M 2 − α(θ)2 {λa , λb } . (2.13)
2
Since the function α(θ) is a number, in such cases, the sole effect of the θ angle on the
meson masses with respect to a theory with θ = 0 is to replace the quark masses by
q
mi → meff
i ≡ m2i − α(θ)2 . (2.14)

In particular, assuming m1 < m2 < . . . < mNf , we have meff eff eff
1 < m2 < . . . < mNf .
1
There are more solutions corresponding to the different branches of the arcsin.

–4–
Number-changing interactions. While the first term in Eq. (2.8) describes meson
masses and interactions involving an even number of mesons, the second term leads to
interactions with an odd number of mesons. The leading two terms in the chiral expansion
of the latter give
 
B0 α(θ) cabcde
Leff, odd terms = − dabc πa πb πc − πa πb πc πd πe + O(fπ−5 ) . (2.15)
3fπ 10fπ2

With dabc = 14 Tr({λa , λb }λc ), and cabcde = (δab dcde + δcd dabe )/Nf + 21 dabm dcdn dmne , which
are non-zero only for Nf ≥ 3. As already mentioned, in Paper I it was assumed that
θ ≪ 1, which justifies using the expression for α(θ) given in Eq. (2.11). However, formally
speaking, the effective Lagrangian with the odd terms in π a remains the same for arbitrary
θ. Consequently, the results presented in that work concerning number-changing processes
remain largely valid even for large values of θ.
Finally, Eq. (2.15) must be compared against the Wess-Zumino-Witten (WZW) term [48,
49], an interaction of topological origin that reads
Z
Nc 
†
5
SWZW = −i Tr U dU (2.16)
240π 2
Z
Nc h i
= d 4
x ϵµνρσ
Tr λa b c d e
λ λ λ λ π a ∂µ π b ∂ν π c ∂ρ π d ∂σ π e + O(fπ−7 ) .
240π 2 fπ5

As Eq. (2.15), it gives rise to 3-to-2 meson annihilations.

2.2 Our benchmark model

The general properties outlined above were noted in Paper I and illustrated using two
benchmark models. In this paper, we focus on BM1. Let us briefly summarize this
model. It is based on a unitary gauge group with Nf = 3 and quark mass matrix
M = diag(m1 , m2 , m3 ). The resulting meson spectrum resembles that of the SM, fea-
turing mesons π 0 , π ± , K 0 , K̄ 0 , K ± , and η.2 The DM candidate is identified with the
lightest meson, π 0 , with the rest of the mesons eventually decaying or annihilating into
DM.

Mass spectrum for θ = 0. In this case,


2
mπ± = B0 (m1 + m2 ),


m2K ± = B0 (m1 + m3 ), (2.17)

m2

= B0 (m2 + m3 ),
K,K̄ 0

while π3 and π8 mix with their masses squared given by the eigenvalues of the matrix
B0
B0 (m1 + m2 ) √ (m1 − m2 )
 
3 . (2.18)
B0 B0
√
3
(m1 − m2 ) 3 (m1 + m2 + 4m3 )
2
Throughout this work, we will refer to the dark mesons with this notation; unless specified otherwise,
SM mesons will not be referred to in this way. In particular, the superscript does not refer to any charge.

–5–
 2.25 η 100

2.00 10−1 r12 = 0.33

0
K 10−2 2000km/s
1.75
mi/mπ0

K± r12 = 10−2

vR
1.50 10−3 200km/s

1.25 10−4
π± 20km/s
10−5 r12 = 10−4
1.00

0.0 0.2 0.4 0.6 0.8 1.0 10−3 10−2 10−1

r12 θ

Figure 1. Left: ratios of dark meson masses as a function of r12 for fixed values of the vacuum
angle: solid, dashed, and dotted lines respectively correspond to θ = 0, π/4, π/2. Right: mass
combination, vR , defined in Eq. (2.21) as a function of θ for fixed values of the indicated quark
mass ratio, r12 . For reference, the dashed lines correspond to typical velocities in DM halos when
expressed in natural units.

The physical states π 0 and η are given by

 0   
π cos θηπ sin θηπ π3
= , (2.19)
η − sin θηπ cos θηπ π8

with a mixing angle √

3(m2 − m1 )
tan(2θηπ ) = . (2.20)
2m3 − m1 − m2
Assuming 0 < m1 ≤ m2 ≤ m3 , the resulting mass spectrum exhibits a well-defined
hierarchy: the η meson is the heaviest state, while the π 0 is the lightest. The π ± mesons
are the second lightest states, with the mass splitting mπ± − mπ0 vanishing in the limit
m1 = m2 . This is illustrated in the left panel of Fig. 1 (solid line) for which we assume
mη = 2mπ0 when θ = 0. Note that then vR |θ=0 = 0 where
s
mη − 2mπ0
vR = 2 , (2.21)
mπ0
which, as we shall see, characterizes the resonant scatterings of DM [50]. More precisely,
the process π 0 π 0 → π 0 π 0 receives a resonant enhancement from the exchange of η meson
for non-relativistic velocities matching vR .
As in Paper I, we remain agnostic of the origin of values of vR of the order of non-
relativistic velocities in natural units, which arises from assuming vR |θ=0 = 0 above. For
examples of model-building in QCD-like models giving rise to this resonant behavior, see
e.g., Refs. [51, 52] (see also Ref. [53]).

The case of a non-vanishing θ angle. As explained above, for |qi θ| < π/2, the meson
spectrum can be readily computed from the formulae of θ = 0 using Eq. (2.14) . Moreover,
the four parameters m1 , m2 , m3 , and θ can be expressed as functions of the following four

–6–
 δ sin (θπη )
1.0 10−1 1.0 0.59

0.8 0.8
10−2 0.44
0.6 0.6

r12
10−4
r12

0.29
0.4 0.4
10−5 0.15
0.2 0.2

10−7 0.00
π/4 2π/3 π π/4 2π/3 π
θ θ

Figure 2. Left: Contours of pion mass splitting δ as a function of θ and r12 . The white regions
correspond to |qi θ| > π/2. See text for details. Right: Same as the left panel but for the π 0 -η
mixing angle θηπ .

quantities: θ, α, the velocity vR |θ=0 , and the mass ratio r12 = m1 /m2 . Notably, since α is
the only parameter out of the latter with mass dimension, any dimensionless combination
of the meson masses may depend only on θ, r12 , vR |θ=0 . Motivated by the resonant effect
associated with η, throughout we will focus on vR |θ=0 = 0. In Fig. 1 (left panel), we show
the corresponding mass spectrum of the mesons taking into account the contribution of
the θ angle.
Taking vR |θ=0 = 0, all dimensionless combinations of masses will be functions of θ and
r12 . Three dimensionless combinations of particular importance are the resonance velocity
vR , the mass splitting between π ± and π 0 parametrized by δ ≡ mπ± /mπ0 − 1, as well as the
mixing angle θπη . They are respectively shown in the Figs. 1 (right panel), 2 (left panel)
and 2 (right panel). In the white regions of Fig. 2 the assumption |qi θ| < π/2 does not
apply.

The role of η for number-changing processes and resonant scattering. The

trilinear interaction in Eq. (2.15) –only present for a non-vanishing θ angle– induces the
decay of η meson into two pions as well as resonant elastic scatterings between pions,
π 0 π 0 → η → π 0 π 0 . For these processes, the rate for η → π 0 π 0 enters in an important way,
and is given by s
α(θ)2 B 2 cos2 (3θ )
0 ηπ 4m2π0
Γ(η → π 0 π 0 ) = 1 − . (2.22)
24πfπ2 mη m2η
In addition to this decay and its inverse, the interactions in Eq. (2.15) induce 3 ↔ 2
self-annihilation processes. As such, the θ angle plays a central role in enabling efficient
number-changing dynamics within BM1.

–7–
3 The relic density in the presence of the θ vacuum

In our benchmark scenario, as will be explained in section 5, observations of galaxy clus-

ters impose an upper bound on the DM self-scattering cross section of approximately
0.5 cm2 /g [36–40]. In the region where chiral perturbation theory is valid, namely mπ0 /fπ <
√
4π/ Nc , and where this bound is satisfied, the maximum strength allowed for the 3 → 2
processes does not yield sufficient DM annihilation if vR ≫ 1, resulting in a relic abundance
significantly larger than the observed DM density. This issue becomes even more severe in
the limit θ → 0, where the WZW term is the sole source of number-changing interactions,
as the resulting 3 ↔ 2 processes are velocity suppressed and the scattering in halos is ve-
locity independent. [1, 47]. In contrast, in the regime vR ≲ 1, the exchange of the η meson
resonantly enhances the 3-to-2 processes, reconciling the predicted relic density with the
observed value. Concretely, the η effectively acts as a catalyzer [54], mediating number-
changing interactions such as π 0 π 0 → η followed by η π 0 → π 0 π 0 , effectively inducing a
resonant 3-to-2 process.
This section examines how via these number-changing processes a non-zero θ angle
impacts the relic abundances of the mesons. Before discussing the corresponding Boltz-
mann equations, let us define notation. As usual, H is the Hubble rate, which, in a
p
radiation-dominated Universe, is given by H ≃ 4π 3 g∗ /45 T 2 /MPl , with g∗ being the en-
ergy density degrees of freedom. Likewise, we find it convenient to normalize the number
density of mesons to the entropy density of the Universe
n
Y = , (3.1)
s(T )

where s(T ) = 2π 2 g∗s T 3 /45, and g∗s is the effective number of degrees of freedom in entropy.

3.1 The evolution with Tdark = T

We start by assuming that the dark sector and the SM are in kinetic equilibrium at the
same temperature. Later on, we will show that deviations from this assumption have a
minor effect on the relic density and do not significantly alter the results for benchmark
points of interest. Neglecting degeneracy effects, Tdark = T means that the dark mesons
are distributed in phase space according to
 
µ−E
f = exp . (3.2)
T

The corresponding densities are given by

d3 p d3 p
Z µ Z  
E
n= f = exp neq with neq ≡ exp − . (3.3)
(2π)3 T (2π)3 T

Hence, the number densities are determined by the chemical potentials, and vice versa, at
a given SM temperature. For simplicity, let us first consider a given meson, which could
be the DM or a heavier state. Quite generically, the Boltzmann equation determining its

–8–
number density, when one particle is destroyed in an individual reaction i → f , is given by
 ! 
3 3
Z Y  
d pa Y d pb
ṅ + 3Hn = − fa |Mi→f |2 (2π)4 δ 4 (pi − pf )
(2π)3 2Ea (2π)3 2Eb
a∈i b∈f
− (i ↔ f ) , (3.4)

where the second line accounts for the inverse reaction. In terms of the yield in Eq. (3.1)

dY mπ 0
ṅ + 3Hn = sHx , with x= . (3.5)
dx T
By substituting Eq. (3.2) into the Boltzmann equation, and using the principle of
detailed balance –which relates the squared amplitudes of a reaction and its inverse– we
obtain
 !  
dY 1 X 1 X
sHx = − exp µa − exp  µb  γ (i → f )
dx T T
a∈i b∈f
 
Y Ya Y Yb
= − −  γ (i → f ) . (3.6)
Ya,eq Ya,eq
a∈i b∈f

Here, the interaction rate density is

! 
d3 pa d3 p
Z Y Ea Y b
γ (i → f ) = e− T   |Mi→f |2 (2π)4 δ 4 (pi − pf ) . (3.7)
(2π)3 2Ea (2π)3 2Eb
a∈i b∈f

This quantity is related to the ordinary decay rates and thermally averaged cross sections
by

γ(1 → 23) = n1,eq K1 (m1 /T )

(
K2 (m1 /T ) Γ(1 → 23) ,
(3.8)
γ(12 → 34) = n1,eq n2,eq ⟨σ12→34 v⟩ .

For two identical particles in the initial state, an additional 1/2 must be included. The
analogue factor for identical particles in the final state is included in the definition of Γ and
⟨σv⟩. Adding several reactions and particles is straightforward. For simplicity, throughout
we use the following notation
Ya
Ya = eµa /T = . (3.9)
Ya,eq

We are now in a position to write down the Boltzmann equations that determine the DM
density, including the effects of coannihilations. Concretely, the evolution of πi and η

–9–
populations is described by the following coupled Boltzmann equations
dY 0 X
sHx π = − γ(ηπ 0 → πi πi ) Yη Yπ0 − Yπ2i


dx
i=1,2
X
+ γ(ηπi → π 0 πi ) (Yη Yπi − Yπ0 Yπi )
i=0,1,2
X
γ(πi πi → π 0 π 0 ) Yπ2i − Yπ20 + 2γD (η → π 0 π 0 ) Yη − Yπ20 ,



+2 (3.10)
i=1,2
dYπi
= + 2γ(ηπ 0 → πi πi ) Yη Yπ0 − Yπ2i − 2γ(πi πi → π 0 π 0 ) Yπ2i − Yπ20



sHx
dx
+ 2γD (η → πi πi ) Yη − Yπ2i ,


for i = 1, 2; (3.11)
dYη X
γ(ηπ 0 → πi πi ) Yη Yπ0 − Yπ2i


sHx =−
dx
i=1,2
X
− γ(ηπi → π 0 πi ) (Yη Yπi − Yπ0 Yπi )
i=0,1,2
X
γD (η → πi πi ) Yη − Yπ2i .


− (3.12)
i=0,1,2

We neglect the effect of kaons because, although they are lighter than the η, there are
no resonant processes involving them that would impact the final DM abundance. The
remaining contributions are limited to conversion processes, which can be safely neglected
since mK ≳ 1.7 mπ0 . Furthermore, we also neglect the 3 ↔ 2 self-annihilations originated
from both the WZW term and the 5-point interactions induced by θ, as well as further
number-changing processes, such as 4-to-2 annihilations. This stems from the fact that
such processes are subdominant with respect to the resonant 3 → 2 interactions.
The thermally-averaged cross sections entering in the rates are the following. For the
semi-annihilations η πa → πb πc (a, b, c = 0, 1, 2)
√
5 δ m2π0
⟨σ(η πa → πb πc )v⟩ = βabc + O(θ2 ), (3.13)
64 π fπ4
where the nonzero values for βabc are
1 529 98
β000 = 1 , β011 = β022 = β0+− = , β101 = β202 = β±0± = . (3.14)
2 81 81
For the (co-)annihilations
√
9 δ m2 0
⟨σ(π1,2 π1,2 → π 0 π 0 )v⟩ = ⟨σ(π + π − → π 0 π 0 )v⟩ = √ π . (3.15)
32 2 π fπ4
In Fig. 3, we present the results of our numerical analysis for a representative pa-
rameter choice: (mπ0 , fπ ) = (20 MeV, 34 MeV). We numerically solve the Boltzmann
equations (3.10)–(3.12), varying θ and r12 within the ranges discussed above with the relic
density fixed to the observed value. Several comments are in order.
First, we can recognize two distinct families of solutions: those where the DM abun-
dance depends only on the value of r12 and is independent of θ (on the vertical line), and
those where the abundance is sensitive to both parameters (on the diagonal line).

– 10 –
 10−1

θ
10−2 Ω h2 = 0.12

10−3
mπ = 20 MeV, fπ = 34 MeV

10−3 10−2 10−1 100

r12

Figure 3. Parameter space leading to the observed DM relic abundance within 10% [55] for fixed
values of mπ and fπ , assuming kinetic equilibrium with the SM bath. See text for details.

• The former, studied in Paper I, are characterized by the fact that the processes
η ↔ π 0 π 0 are much faster than the Universe expansion when ηπ 0 → π 0 π 0 freezes
out, so that they keep chemical equilibrium among η and π 0 , which enforces µη = 2µπ
after freeze-out. See also Refs. [2, 56]. We illustrate this in the left panel of Fig. 4
for a representative choice of parameters. Concretely, we show the evolution of the
abundances of π 0 , π ± , η (upper left panel) and their chemical potentials (lower left
panel). Notice that µη = µπ0 = 0 when semi-annihilation and (inverse) decays are
simultaneously active, while µη = 2µπ0 after the decoupling of semi-annihilations.
These relations between chemical potentials allow us to write a simplified Boltzmann
equation for the combination Y = Yπ + 2Yη :

Yπ3 Yπ2
 
dY sYη,eq
= −⟨σηπ v⟩ 2
− , (3.16)
dx xH Yπ,eq Yπ,eq
where Yπ = Yπ0 +2Yπ± and ⟨σηπ v⟩ is the proper combination of cross sections. Notice
that, since the mass splitting between π 0 and π ± is less than 10%, the latter contribute
to the DM relic abundance, converting as π + π − → π 0 π 0 some time after freeze-
out. Most importantly, this simplified equation is independent of θ, in agreement
with our results in Fig. 3. As discussed in Paper I, the simplified equation can be
solved numerically, or alternatively, an approximate analytical solution of the form
Yπ,eq (xfo ) can be obtained solving for xfo from H(xfo ) ≃ nη,eq (xfo )⟨σηπ v⟩. Both of
these approaches are in excellent agreement with the numerical solution of the full
set of Boltzmann equations, thus corroborating the results of Paper I.

– 11 –
 r12 = 0.193, θ = 0.01, mπ = 20 MeV, fπ = 34 MeV r12 = 0.002, θ = 0.035, mπ = 20 MeV, fπ = 34 MeV

10−2 10−2

10−5 10−5

10−8 10−8
Y

Y
10−11 10−11

10−14 10−14

10−17 0 10−17 0
10 101 102 10 101 102

Yπ 0 Yπ + Yη YDM
2.0 2.0

1.5 1.5
µ/mπ

µ/mπ
1.0 1.0

0.5 0.5

0.0 0.0
100 101 102 100 101 102
x x
µπ 0 µη 2µπ0

Figure 4. Dark meson abundances and the respective chemical potentials for two representative
points. The point on the left lies on the vertical line of Fig. 3, while the one on the right belongs to
the diagonal line. The horizontal line in the upper plots represent the observed DM abundance [55].

• On the other hand, the points on the diagonal lines of Fig. 3 correspond to a new
family of solutions, not discussed in Paper I. In such points, the inverse decay pro-
cesses decouple before semi-annihilations. In the right panel of Fig. 4, we show the
abundances (upper right panel) and the chemical potentials (lower right panel) of the
different species for a representative choice of parameters. Once again, we observe
that all the chemical potentials vanish when both semi-annihilations and inverse de-
cays are in equilibrium. When the inverse decays become inefficient, the chemical
equilibrium between η and π is kept by the semi-annihilation, leading to µη = µπ0 .
This lasts until semi-annihilation freezes out. After that, both processes involved are
inactive and their rates are comparable in magnitude, such that the chemical equilib-
rium is not achieved. Eventually, the inverse decay becomes active again, restoring
the chemical equilibrium and fixing the chemical potentials to µη = 2µπ . A similar
behavior of the chemical potentials has been identified in other DM scenarios [57–59].
In this case, a simplified Boltzmann equation, such as Eq. (3.16), fails to provide
results in quantitative agreement with the full numerical solution. Nevertheless, we

– 12 –
 signal that a simplified equation involving only the (inverse) decay processes and
enforcing µπ = µη provides results in qualitative agreement with the full numerical
solution.

3.2 Departure from kinetic equilibrium

Before discussing the thermalization details—and thus the SM portal—let us emphasize
that the results so far assume T = Tdark . On the one hand, this requires strong interactions
for the dark mesons so that the dark sector develops a thermal distribution. On the other
hand, this also requires that the dark sector and the SM are kept in kinetic equilibrium,
occurring if the elastic scatterings among the dark mesons and SM particles are sufficiently
strong to allow an efficient energy transfer from one sector to the other.
If kinetic equilibrium between the two sectors is never realized, the dark sector devel-
ops its own temperature Tdark in view of elastic scattering among dark sector particles,
which evolves independently of the SM temperature T . We consider the more predictive
scenario in which the interactions between the SM and the dark sectors are strong enough
to keep kinetic equilibrium among them initially, thus enforcing Tdark = T until the Uni-
verse cools down to some temperature, Tdec . At that time, known as kinetic decoupling,
elastic scatterings of the dark mesons off SM particles become slower than the expansion
of the Universe and the dark sector evolves with its own temperature. If number-changing
processes within the dark sector are inefficient at the time of kinetic decoupling and after-
wards, the temperature evolves as Tdark ∼ 1/a if DM is relativistic or Tdark ∼ 1/a2 if DM
is non-relativistic [22, 60]. See Ref. [8] for a discussion in the context of SIDM. Notice that
in this scenario, the evolution of the dark temperature is predicted in terms of that of the
SM.
In the following, we denote the time of kinetic decoupling as xdec = mπ0 /Tdec and define
the time of DM freeze-out as xfo = mπ0 /Tfo . In this way, for x < xfo the dark mesons
have a vanishing chemical potential enforced by number-changing processes, while chemical
decoupling occurs for x > xfo with the dark mesons acquiring a chemical potential. For the
family of solutions corresponding to the vertical line of Fig. 3, in which semi-annihilations
decouple before inverse decays, xfo corresponds to the usual definition of the time at which
ηπ 0 → π 0 π 0 becomes inefficient.
In the next section, after introducing an explicit SM portal and the interactions be-
tween the SM and the dark mesons, we will be interested in the region of the parameter
space where kinetic decoupling occurs after DM freeze-out, xdec > xfo . Since xfo ∼ O(20)
and π 0 is the lightest dark particle, the dark sector is non-relativistic at the time of kinetic
decoupling, so that for x > xdec its temperature evolves as Tdark = T xdec /x ∼ 1/a2 . We
will take this effect into account by solving the Boltzmann equations for the dark meson
abundances in two steps:

• for x < xdec we solve the set of coupled Boltzmann equations in Eq. (3.10)-(3.12);

• for x > xdec we solve the same set of equations replacing the dark meson equilibrium
distributions as na,eq (T ) → na,eq (Tdark ), so that Ya,eq → na,eq (Tdark )/s(T ), with
Tdark = T xdec /x.

– 13 –
This typically leads to a reduction of the DM relic abundance of a factor of O(10 − 30)%.3

4 Portal to the SM

4.1 Dark photon portal and DM decays

As mentioned above, we will assume a dark sector initially in thermal equilibrium with the
SM thermal bath. This is a crucial assumption as, in the absence of an explicit portal,
number changing processes within the dark sector would modify the evolution of the dark
sector temperature and the corresponding DM population.
In this section, we substantiate this assumption by constructing explicit realizations of
BM1 when it is coupled to the SM by means of a dark photon. This portal [33] has been
studied in the context of models of strongly interacting massive particle (SIMP) DM where
the relic abundance is obtained by 3-to-2 annihilations induced by the WZW term, see e.g.
Refs. [4] and [22]. Although other portals can be considered –notably those induced by
axion-like particles [3, 19], potentially related to the dark theta angle– we focus on dark
photons to demonstrate the existence of one possible realization of QCD-like DM produced
by the theta angle, as well as the internal consistency and phenomenological viability of
the framework.
For simplicity, in this section we restrict our analysis to small values of the θ angle,
namely θ ≪ 1. As discussed in the previous sections, we do not expect significant differences
for larger values, θ ∼ O(1). We extend the model by introducing a U (1)d gauge symmetry,
associated with a coupling constant ed . The corresponding dark photon, with mass mV
and kinetic mixing ε, is described by
1 1 ε
LDP = − Vµν V µν + m2V Vµ V µ + Vµν B µν , (4.1)
4 2 2
where Bµν is associated with the SM hypercharge. We remain agnostic regarding the origin
of the dark photon mass, which could arise either through the Stückelberg mechanism [62]
or via interactions with a dark Higgs field [63, 64].
On the one hand, as usual, the SM sector carries no dark charge and communicates
with the DM only through the parameter ε, that induces a mixing between the dark
photon and the SM photon. After diagonalization, the SM photon only couples to the SM
electromagnetic current JµSM as eAµ JµSM , while the physical dark photon couples to both
the dark current Jµdark and the SM current with coupling V µ (ed Jµdark + e ε cW JµSM ), where
cW denotes the cosine of the weak mixing angle. Since we are primarily interested in the
MeV scale, we ignore the small mixing between the dark photon and the Z gauge boson,
which is suppressed by ε (mV /mZ )2 [65].
On the other hand, the dark quarks are charged under the U (1)d gauge group, with
charges Qd = diag (q̃1 , q̃2 , q̃3 ). The chiral Lagrangian in Eq. (2.5) in presence of the dark
3
In the opposite regime, i.e xdec < xfo , number-changing processes are still active after kinetic decoupling
enforcing µi = 0 for all species and heating up the dark sector, whose temperature evolves logarithmically
with the scale factor, Tdark ∼ 1/ ln a [60, 61].

– 14 –
photon is extended to [22, 25, 46]

fπ2 h i B f2
0 π
   
Leff = Tr (Dµ U )† Dµ U + Tr Mθ† U + U † Mθ + cfπ4 e2d Tr U † Qd U Qd
4 2
ie2d Nc µναβ
 
2 † 2 † 1 † 1 †
− ϵ Vµν Vα Tr Qd ∂β U U − Qd ∂β U U − Qd U Qd ∂β U + Qd U Qd ∂β U
48π 2 2 2
ed Nc µνρσ 
† † † † † †

+ ϵ Vµ Tr Q ∂
d ν U U ∂ρ U U ∂σ U U + Q d U ∂ ν U U ∂ ρ U U ∂ σ U
48π 2
+ LWZW + higher order operators .
(4.2)
Here, the covariant derivative is defined as Dµ U = ∂µ U − ied Vµ [Qd , U ]. The last term
of the first line is responsible for the mass corrections of the dark mesons charged under
U (1)d , with c being a constant. Assuming a heavy dark photon, mV > Λ, it scales as
c ∼ (Λ/mV )2 ln(m2V /Λ2 ) [22].
In the presence of the portal, DM stability is no longer guaranteed, as the number-
changing processes might induce its decay. Note that this does not hold if a conserved
symmetry exists in the dark sector, but this is not the case for BM1, in which the flavor
symmetry is broken by quark mass differences, m1 ̸= m2 ̸= m3 . Thus, the DM particle
is meta-stable, and its interactions with SM particles determine whether it is sufficiently
long-lived.
DM decay is induced by the operator π 0 Vµν Ṽ µν and, in presence of a non-vanishing
π 0 − η mixing, by η Vµν Ṽ µν . For a generic choice of dark quark charges, both operators
are generated via the axial anomaly: the approximate (spontaneously broken) SU (3)A
global symmetry of the Lagrangian is anomalous under the dark U (1)d interactions.4 Note
that this closely resembles the anomalous decay of neutral mesons into two photons ob-
served in the SM. The presence of the anomaly is encoded in the second line of Eq. (4.2),
which provides interaction terms proportional to (e2d /fπ ) π3,8 Vµν Ṽ µν Tr[λ3,8 Q2d ]. These,
once expressed in terms of the mass eigenstates, induce the aforementioned operators with
couplings suppressed by only one power of fπ .
However, it is possible to choose the charge matrix Qd in such a way that the anomaly
of the axial currents vanishes, Tr[λ3,8 Q2d ] = 0. In such a case, the corresponding currents are
conserved in the massless limit, and the anomalous contribution to the operators discussed
in the previous paragraph vanishes. This requirement can only be fulfilled by assigning
the dark charges so that q̃12 = q̃22 = q̃32 , and hence Qd ∝ (±1, ±1, ±1). Without loss of
generality, there are only three charge assignments (up to a global normalization). Two of
(3)
them are listed in Table 1, while the third one, Qd ∝ diag(+1, −1, +1), is qualitatively
(2)
similar to Qd and provides analogous results.
Even for a vanishing anomaly, the interaction terms π 0 Vµν Ṽ µν and ηVµν Ṽ µν are gen-
erated by higher-order operators in the chiral Lagrangian expansion. More precisely, the
relevant operators arise at O(p6 ), that is, next-to-next-to-leading order. Following the
4
More precisely, even in the limit in which the dark quarks are massless, a subset of the currents
associated with the SU (3)A transformations are not conserved because of the quantum anomaly, namely
3,8 3,8
∂ µ Jµ,A ∝ e2d Vµν Ṽ µν Tr[λ3,8 Q2d ] with Jµ,A = q̄ γµ γ5 λ3,8 q.

– 15 –
 e−
γ

e+
V ×ε
π0

V ×ε e−

γ
e+
Figure 5. Feynman diagram for the main decay channel of π 0 in the mass range of interest in this
work. A similar diagram holds for the decay of η.

approach in Ref. [22], as a representative operator we consider

αd
iϵµνρσ V µν V ρσ Tr(Qd )Tr(Qd M U † ) + h.c. , (4.3)
(4π)2 fπ

where αd = e2d /4π. This operator induces the π 0 Vµν Ṽ µν and ηVµν Ṽ µν interactions with
couplings suppressed by an extra factor of (mπ0 /fπ )2 compared to the anomalous con-
tribution discussed before.5 Thus, in the case of vanishing anomaly, the DM decay rate
is significantly smaller close to the chiral limit, mπ0 ≪ fπ , providing a much larger DM
lifetime.
Finally, DM is absolutely stable in the limit in which the global SU (2)V symmetry
involving the first two flavors of dark quarks is conserved, i.e. dark isospin. This requires
both r12 = 0 and q̃1 = q̃2 . For the cases in which q̃1 = q̃2 but r12 ̸= 0, the decay of π 0
is induced by dark isospin breaking, or, in other words, by the π 0 − η mixing, as given in
Eq. (2.20). As shown in Fig. 2 (right panel), this is fixed by r12 and θ.
In the following, we assume mV > mη so that the decays π 0 , η → V V are kinematically
forbidden. As the dark mesons do not couple to the physical SM photon, the dark decays
π 0 , η → γγ are absent and DM decays to SM leptons. For mη ≲ 200 MeV, the relevant
DM decay channel is π 0 → e+ e− e+ e− , corresponding to a tree level diagram with two dark
photon propagators, see Fig. 5. This process involves the electron-dark photon coupling
twice, so that both DM and η decay rates are suppressed by ε4 e4d . An analogous decay
mode arises for the η meson. However, its branching ratio is extremely suppressed, as η
mostly decays to 2π 0 via the θ angle. We will briefly comment on this below.
In the following, we will describe the possible charge assignments presented in Table 1
for which the axial anomaly vanishes and the DM is long-lived.

• Charge assignment 1. The dark mesons π 0 , π ± and η are neutral under U (1)d ,
whereas K + and K 0 carry a charge of +2, while K − and K 0 carry a charge of −2.
5
Notice that the presence of a non-vanishing θ angle may induce additional CP-violating operators in the
chiral Lagrangian, potentially giving rise to θ η Vµν V µν and θ π 0 Vµν V µν interactions. Their contribution
to the DM decay rate is nevertheless suppressed for small values of θ. One could expect O(1) corrections
for large θ ∼ O(1).

– 16 –
 Charge Assignment P± DM Lifetime
(1)
Qd = diag(+1, +1, −1) K± τDM in Eq. (4.4)
(2)
Qd = diag(+1, −1, −1) π± τDM in Eq. (4.4) times sin2 θπη

Table 1. Charge assignments and associated properties for benchmark models. The two-step
thermalization between the DM and the SM bath is achieved by means of scatterings with the
meson P ± . See text for details.

Being charged, the four kaons are stable. Furthermore, the resulting Lagrangian is
invariant under the accidental global symmetry generated by Q′ = diag (1, −1, 0).
Being the lightest particles charged under this symmetry, the π ± are stable. In
contrast, η and π 0 are not protected by any symmetry, and therefore decay. DM
decays are induced via the operator in Eq. (4.3), with a lifetime estimated up to
O(1) factors as
2  −3 2  4
5 × 10−4

29 0.1 10
τDM ∼ 1.4 × 10 s
sin θπη αd ε
 m 
8 20 MeV  9   6
V 0.5
× . (4.4)
0.5 GeV mπ0 mπ0 /fπ
For reference, as shown in Fig. 2 (right panel), the range of values for the mixing
angle is 0.23 ≳ sin θπη ≳ 0.005 for θ ≪ 1. The decay is strongly suppressed close
to the chiral limit mπ0 ≪ fπ , as expected. Furthermore, the DM is stable in the
limit sin θπη = 0 (equivalent to r12 = 0) in view of the isospin-preserving charge
assignement, q̃1 = q̃2 = 1. The corresponding constraints from DM indirect searches
will be discussed in Section 4.3.
Moreover, the kaon masses receive a small correction due to the interaction with the
dark photon (similarly to the SM). As long as mV > Λ, this can be estimated as
m2K → m2K + ∆m2K with ∆m2K ∼ 8 c e2d fπ2 . Recall, nevertheless, that the DM relic
density is only marginally affected by this.

• Charge assignment 2. The charge of the dark mesons is analogous to the ordinary
electric charge in the visible sector: the mesons π 0 , η, K 0 , and K 0 are uncharged,
whereas π ± and K ± carry charge ±2. While the π ± states are stable as a result
of this, π 0 decays with a lifetime estimated –up to factors of O(1)– using Eq. (4.4)
without the (1/ sin2 θπη ) factor. In this case, as q̃1 ̸= q̃2 , the charge matrix explicitly
breaks isospin, so that the operator in Eq. (4.3) directly induces the decay of π 0
regardless of its mixing with η. For the same parameters used in Eq. (4.4), the
lifetime is shorter, namely, τDM ∼ 1027 s.
Along with this, the π ± interaction with dark photon shift their mass squared:
m2π± → m2π± + ∆m2π± . We can express this in terms of the pion splitting defined in
the previous sections as
mπ± − mπ0
δ= = δ(r12 ) + δV , (4.5)
mπ 0 (2)
Qd ed =0

– 17 –
 where δ(r12 ) e =0 is the correction in the absence of the portal, shown in Fig. 2 (left
d
panel). The other term can be estimated for mV > Λ as [22]

∆m2π± Λ2 m2V 4 e2d fπ2

 
δV ≃ ≃ ln , (4.6)
2m2π0 m2V Λ2 m2π0
where we used ∆m2π± ∼ 8ce2d fπ2 .
Notice that the additional source of mass splitting for the charged pions affects DM
relic density when δV ≳ 0.01. Indeed, the charged pions become sufficiently heavy
that their number density is significantly Boltzmann-suppressed at freeze-out, nπ± ≪
nπ0 , thereby reducing the total DM abundance, nDM = nπ0 + 2nπ± . Such a reduction
in the DM number density requires a larger value of fπ to match the cosmological relic
abundance for fixed DM mass and r12 compared to the case of δV = 0. Furthermore,
if δ ≳ 0.5, the total cross section for ηπ → ππ becomes smaller, as the ηπ 0 → π + π −
becomes kinematically closed.
We will take these effects into account in the following section while solving the Boltzmann
equations for the dark meson abundances for a representative benchmark point.

4.2 Thermalization
In the rest of this section, we discuss the conditions for kinetic equilibrium between the SM
and the dark sector. Kinetic equilibrium is established via elastic scatterings of the dark
mesons with SM fermions mediated by the dark photon. Although this discussion shares
some similarities with previous studies of other SIMP candidates (see e.g. Refs. [4, 22]),
our DM candidate is neutral under U (1)d and therefore does not interact directly with SM
fermions. Thus, kinetic equilibrium is established in two steps:
I. The lightest dark mesons charged under U (1)d elastically scatter off SM leptons
(mainly electrons and positrons) via the t-channel exchange of a dark photon. If these
interactions are sufficiently strong, the charged mesons remain in kinetic equilibrium
with the SM bath, thereby maintaining a common temperature. The lightest charged
mesons are the kaons K ± within the charge assignment 1 and the pions π ± within
charge assignment 2, see Table 1. The relevant elastic cross section is given by
(
± ±
e2 e2d c2W Q2P ε2 p2f K, for charge assignment 1;
σv(P f → P f ) = 4 , with P = (4.7)
2πmV π, for charge assignment 2 ,

where pf is the momentum of the SM fermion f = e± and QK = Qπ = 2 is the dark

charge of the corresponding meson. They thermalize with the SM as long as [4, 22]
5ζ(5) T
Γscatt ≳ H , (4.8)
4 mP ±
where the scattering rate is given by
2 90ζ(5)e2 e2d c2W ε2 T 5
P
± QP
⟨neq
X
± ±
Γscatt = P f σv(P f → P f ) ⟩ = . (4.9)
2 − +
π 3 m4V
f =e ,e

– 18 –
 II. The DM particle π 0 thermalizes with the lightest meson charged under U (1)d , P ± ,
thus sharing their temperature, which is the same of the SM bath if the condition in
Eq. (4.8) is satisfied. The relevant thermal cross sections are given by

2 s
± 0 ± 0 m2π0 13 − 4m2K ± /m2π0 8 mπ 0
⟨σ(K π → K π )v⟩ = , (4.10)
9216π fπ4 (1 + mK ± /mπ0 )3/2 π x mK ±
2 2 /m2 2  4 s


m 0 13 − 4m ± 0 mπ ± 8 mπ0
⟨σ(π ± π 0 → π ± π 0 )v⟩ = π
4
π π
3/2
. (4.11)
1296π fπ (1 + mπ± /mπ0 ) mπ 0 π x mπ ±

We numerically find that these expressions are of the same order of magnitude. Once
established, the kinetic equilibrium is maintained as long as the elastic scatterings
are more rapid than the Hubble expansion, that is,
T
nP ± (T )⟨σv(P ± π 0 → P ± π 0 )⟩ ≳ H. (4.12)
mπ0

The main difference between the two charge assignments is that the charged pions
are nearly degenerate in mass with the neutral pion, i.e., mπ± ∼ mπ0 , resulting in a
significantly larger abundance of charged pions at freeze-out compared to that of kaons,
nπ± ≫ nK at T = Tfo . Consequently, the thermalization condition is more easily satisfied
for the charge assignment 2. Thus, for simplicity in the numerical analysis that follows, we
choose benchmark points within the charge assignment 2.
Finally, we compute the largest temperatures at which Eq. (4.8) or Eq. (4.12) no
longer holds, which corresponds to the kinetic decoupling Tdec as defined in Section 3.2.For
concreteness, we will focus on parameter space where kinetic decoupling occurs after DM
freeze-out, that is Tdec < Tfo .

4.3 Viable parameter space

Taking charge assignment 2, Fig. 6 shows the region of the parameter space (mV , ϵ) where
thermalization is achieved for a representative choice of the parameters mπ0 , fπ , r12 and
αd . In the purple region, denoted as early thermal decoupling, at least one of Eqs. (4.8)
and (4.12) is violated before DM freeze-out and kinetic equilibrium is lost. The yellow
and gray regions are excluded by indirect detection and collider searches to be discussed
below. Here, the annihilation channel π + π − → e+ e− is slower than the Universe expansion
rate at T = Tfo and is irrelevant for the evolution of the DM density6 . Furthermore, as
mV > 2mπ , the semi-annihilation channel ππ → πV is kinematically closed.
The Boltzmann equations for the dark meson abundances are solved for a representa-
tive benchmark point, indicated by a star in Fig. 6, with the corresponding results shown
in the left panel of Fig. 7. Although the behavior observed in the left panel of Fig. 7 is
6
The corresponding annihilation rate is given by
2
2 e2 e2d c2W mπ ε2 T
P
π Qπ
Γann = ⟨σv(π + π − → e+ e− ) nπ ⟩ = sYDM , (4.13)
Nπ 2 π(m2V − 4m2π )2

and satisfies Γann < H|Tfo for all the points displayed in Fig. 6.

– 19 –
 mπ = 20 MeV, fπ = 34 MeV, αd = 10−3 , r12 = 0.33

10−2
n
ec tio
t
t De )
c r
ire age
Ind(Voy
)
E RA
64 ( H BaBar
10−3 NA m
2 1c
ε

Belle II

10−4 MX Early Kinetic Decoupling

LD

300 500 1000 3000

mV [MeV]

Figure 6. Summary of constraints for the representative choice of the parameters indicated in the
legend. In the shaded region at the bottom (light purple), kinetic decoupling occurs before DM
freeze-out. The shaded regions in the upper part of the plot represent areas excluded by searches of
DM decay (yellow) as well as dark photon searches in collider and beam dump experiments (gray).
Future sensitivities are indicated by the dotted lines: LDMX and Belle II are shown in gray, while
21 cm prospects are indicated in orange.

similar to what has been studied in the literature for standard SIMP scenarios [4, 22, 25],
our value of mπ0 /fπ ∼ 0.6 lies well within the perturbative regime of chiral expansion.
This opens up parameter space in agreement with indirect detection constraints. In the
numerical solution of the Boltzmann equations, we take into account the kinetic decou-
pling discussed in the previous section as well as the correction to the charged pion masses,
which gives δV ∼ 0.05. Such corrections have a significant impact on the DM relic density.
In particular, the abundance of the charged pions (and therefore the total DM number
density) at xfo is reduced compared to the case without a portal. This effect, together
with the kinetic decoupling, explains the fact that this benchmark point –which belongs
to the vertical family of solutions discussed in Section 3– does not reproduce the DM relic
abundance for the value of r12 along the yellow vertical lines of Fig. 3, but for a slightly
larger one. Nevertheless, this effect on the relic density is at most at the tens-of-percent
level.

Constraints on DM decays. Current bounds on MeV DM decaying to electrons require

a lifetime τDM ≳ 1026 s [66]. More precisely, gas-rich dwarf galaxies with a low cooling rate
are extremely sensitive to heat injections from DM decays. The strongest constraint in the
mass range 1-15 MeV is set by observations of the Leo T dwarf galaxy [67]. Furthermore,

– 20 –
a DM candidate decaying to electrons would produce an imprint on charged cosmic rays.
While the solar magnetic field prevents their local observation, interstellar cosmic rays can
nonetheless be detected by Voyager [68], which sets the most stringent bound in the mass
range 15-50 MeV. Finally, larger masses in the sub-GeV range can be probed by X-ray
telescopes, which detect photons produced via Inverse Compton scattering of background
radiation off electrons originating from DM decay, with the strongest constraints provided
by XMM-Newton [69].
Thus, for the choice of DM mass of 20 MeV adopted in Fig. 6, the strongest constraint
is set by Voyager, τDM ≳ 3 × 1026 s. This is shown in Fig. 6, as the yellow exclusion region.
Notice that for the point indicated with the star, the DM lifetime is τDM ≃ 5 × 1026 s.
Concerning future prospects, HERA’s future measurements of the 21-cm power spectrum
will probe DM lifetimes, τDM ≲ 1028 s [70], potentially exploring most of the parameter
space in Fig. 6. We show this as a dotted orange line in the figure.
It is important to mention that all these bounds are obtained assuming that the DM
particle decays to an e+ e− pair. In our case, since the DM decays into e+ e− e+ e− , a
correction comparable in magnitude to the lifetime itself is expected [71]. Nevertheless,
given the large exponent of the DM mass in Eq. (4.4), this translates into a negligible
correction to the DM mass and it is therefore inconsequential for Fig. 6.
Finally, the η meson can be produced as a s-channel resonance from DM scatterings
in galactic halos, as we discuss in more detail in Section 5. If its branching ratio to SM
particles is large enough, severe constraints apply from DM indirect searches, see e.g.
Ref. [50]. However, the decay rate of the η meson is dominated by the θ-induced process
η → 2π 0 , while the branching ratio into electrons is extremely suppressed, yielding no
relevant experimental constraint. For the benchmark point shown in Fig. 7, we obtain
BR(η → leptons) ∼ O(10−30 ). Larger values of the θ angle further reduce this branching
ratio. Such minuscule sensitivities are far beyond the reach of any current experiment.

Bounds from dark photon searches. The most stringent bounds on the parameter
space under consideration arise from collider, fixed-target, and beam-dump experiments.
The constraints change depending on the branching ratio between visible and invisible
decays. For the choice of parameters of Fig. 6, we have mV > 2mη and αd ≫ αem ϵ,
hence the dark photon mainly decays to the dark mesons charged under U (1)d , V →
π + π − , K + K − for charge assignment 2.7 Thus, the relevant limits come from searches
for invisible decays. The BaBar collaboration [72] currently provides the most stringent
bounds on invisibly decaying dark photons in the mass range of 500 MeV to 8 GeV. Future
data from Belle II [73] are expected to significantly improve these bounds, reaching almost
ε ≃ 10−4 . On the other hand, among the beam dump experiments, currently the most
competitive constraint in the region relevant for thermalization below 500 MeV comes
from NA64 [74] In the future, both NA64 [75] and LDMX [76] will explore even further
this parameter space, see e.g. Ref. [77].
7
For charge assignment 1, the dark photon decays as V → K + K − , K 0 /K 0 , and the corresponding
constraints remain unchanged.

– 21 –
 mV = 500 MeV, mπ = 20 MeV, fπ = 34 MeV, ε = 0.0005, αd = 10−3, r12 = 0.33, θ = 0.0005
106

hσvi/mDM in cm2/g × km/s

10−4 105
2 /g
cm
π0 104 150
10−8
103
Y

10−12 102 2 g
m/
0. 5 c
101
10−16 η
π± 100 Ω h2 = 0.12, vR = 85 km/s
−20
xfo xdec
10
100 101 102 103 101 102 103 104
x hvi in km/s

Figure 7. Left panel: Abundance of DM and the various species for the benchmark point
indicated with a star in Figure 6. We use Tdark = T ∼ 1/a for x < xdec and Tdark = T xdec /x ∼ 1/a2
for x > xdec , as explained in Section 3.2. Right panel: Velocity dependence of the DM self-
scattering cross section for the same benchmark point. For reference, we show the upper bound
from cluster observations, σ/m ≲ 0.5 cm2 g−1 at ⟨v⟩ ∼ 2000 km s−1 , as well as the lower bound
σ/m ≳ 150 cm2 g−1 at ⟨v⟩ ≃ 60 km s−1 , seemingly explaining the lensing system SDSSJ0946+1006,
see text for details.

5 Impact of the θ vacuum in DM halos

5.1 Reconciling the relic density with velocity-dependent scattering

Models in which DM production occurs via 3 → 2 annihilations, commonly referred to as
SIMP models, were initially proposed in the early 1990s [61]. These models experienced
a resurgence of interest roughly a decade ago, as they naturally predict the observed DM
relic abundance within regions of parameter space where the elastic self-interaction cross
section per unit mass is of the order O(cm2 /g). Scenarios with such interaction strengths
are known as SIDM models [34], which attracted considerable attention due to their ability
to reduce the central densities of DM halos. This effect appears to alleviate the tension
between predictions from collisionless CDM simulations [78–80] and certain astrophysical
observations [81–86].
However, it has become increasingly evident that SIDM models with constant elastic
cross sections face significant tension with observational data from galaxy clusters, which
impose upper bounds on the self-interaction cross section [36–40]:
σ
≲ 0.5 cm2 /g. (5.1)
m
As typical DM velocities in galaxy clusters are larger compared to those in smaller
halos, viable SIDM models now frequently incorporate velocity-dependent interactions [35,
41]. This poses a challenge for the original SIMP framework [1], which, in its simplest
form, does not accommodate velocity-dependent self-interactions. For example, in pion
DM scenarios with θ = 0, the elastic scattering cross section remains constant at non-
relativistic velocities. In contrast, the θ-induced DM production mechanism considered
in this work naturally yields velocity-dependent self-interaction cross sections that can

– 22 –
evade cluster constraints while remaining sizable in smaller halos. These interactions are
mediated by the η meson acting as a resonance when θ ̸= 0, and are described by the
non-relativistic Breit-Wigner formula [50, 87]:

128π Γ2
σ(v) = σ0 + 2 m2 (v 2 − v 2 )2 + 4Γ2 v 2 /v 2 , (5.2)
m2DM vR DM R R

where v is the relative velocity of DM particles, vR is the resonant velocity defined in

Eq. (2.21) and Γ is the decay rate of the mediator. For BM1

m2DM
σ0 = , (5.3)
128πfπ4

and Γ is given in Eq. (2.22). In BM1, all DM today consists of π 0 due to the efficient
conversion of π ± (and heavier stable mesons) into π 0 in the early universe. Furthermore,
we are assuming that the production of π ± from π 0 in astrophysical halos is kinematically
forbidden, which holds when δ ≫ v 2 ∼ 10−5 today, excluding the gray region shown in the
left panel of Fig. 2. As described in Section 4.1, in the presence of the dark photon portal,
π ± may become heavier due to radiative corrections. This makes its production in halos
even harder, enlarging the viable parameter space.
Following the analysis of Ref. [50], and for the benchmark point indicated by a star in
Fig. 6, Fig. 7 demonstrates that the velocity dependence encoded in σ(v) enables our QCD-
like DM model to realize SIDM while remaining consistent with both cluster constraints
and the observed relic density.

5.2 The SIDM parameter space

The ability of SIDM to explain cored density profiles arises from the predicted elastic self-
scattering between DM particles in halos, which thermalizes their velocities and enables
heat transport from the outer halo to the center, thereby flattening the central density
profile—a mechanism absent in standard collisionless cold DM. This thermal evolution
typically proceeds through two distinct phases. During the core formation phase, heat
transfer from the hotter outer halo to the cooler central region leads to the development
of an isothermal, constant-density core that replaces the steep central cusp predicted by
collisionless models. Over longer timescales, continued energy transport causes the core
to lose thermal support, transferring heat outward and initiating the core collapse phase,
also known as gravothermal collapse. In this runaway process, the core contracts while its
central density rises rapidly, analogous to the gravothermal catastrophe in globular clusters.
The transition between these two phases depends on not only σ(v)/mDM but also the halo
mass and assembly history, and can be influenced by environmental effects such as tidal
interactions. For a review with references to the original literature, we refer the reader to
Ref. [41].
Recent advances, largely based on non-resonant self-scattering cross sections, have
identified the preferred values for σ(v)/mDM . In particular, successful models require
strong velocity dependence, producing core formation in high-mass halos and gravothermal
core collapse in low-mass halos [88, 89], with outcomes influenced by assembly history and

– 23 –
halo concentration [90, 91] . Observational constraints span a broad range: as already
mentioned, galaxy clusters with v ∼ 1000 km s−1 imply σ/mDM ≲ 0.1–1 cm2 g−1 [36,
39, 83, 92, 93]; galaxies with v ∼ 100 km s−1 favor σ/mDM ≳ 3 cm2 g−1 [91, 94, 95];
and sub-galactic systems with v ∼ 10 km s−1 require σ/mDM ≳ 5–10 cm2 g−1 as well
as core collapse in some halos [96–99]. Furthermore, other analyses assuming the core
formation phase [100, 101] find that ultra-faint dwarf galaxies impose a stringent limit of
σ/mDM ≲ 0.1 cm2 g−1 at very low velocities, though uncertainties remain due to limited
stellar kinematic data.
In addition to this, in the last few years there is growing interest in certain diffuse satel-
lites and dense substructures suggesting σ/mDM ∼ 100 cm2 g−1 or greater. This includes
the lens system SDSS J0946+1006 [42], as it appears to host an unusually compact and
over-concentrated dark subhalo, one of the rare dark halos smaller than a galaxy that has
been directly detected by means of lensing [102]. Its steep density profile sharply deviates
from the predictions of collisionless DM, prompting investigations within the SIDM frame-
work [43], particularly in the core-collapse phase. For instance, Ref. [103] has analyzed
this system and found the observations could be explained by an SIDM halo undergo-
ing gravothermal collapse, provided the self-interaction cross section is much larger than
σ/mDM ∼ 100 cm2 g−1 . Another study has revisited the hypothesis of a core-collapsed
SIDM halo, modeling its lensing signature via an isothermal Jeans approach. This showed
that only a very massive core-collapsed halo (mass of order 1011 M⊙ ) can reproduce the
observations, but such a halo would be expected to host a luminous galaxy, which is not
observed, posing a challenge to the SIDM interpretation [104]. Overall, SDSS J0946+1006
remains a compelling but debated candidate for observing gravothermal collapse in SIDM.
Motivated by this, in Fig. 7 we indicate a lower bound [91, 104] of about σ/mDM ≃
150 cm2 g−1 at intermediate velocities of order 60 km s−1 , consistent with our benchmark
point. In fact, once vR ∼ 100 km s−1 is fixed, our QCD-like framework naturally predicts
cross sections of this magnitude at v ∼ vR as an unavoidable outcome of the model. Larger
values of the θ angle further enhance the likelihood of producing these effects as they lead
to larger cross sections. Interestingly, cross sections in this range have been proposed as a
mechanism for seeding black holes, particularly in the context of SDSS J0946+1006, which
may represent a striking phenomenological consequence of resonant SIDM in the early
Universe [105, 106].
Taken at face value, these results suggest σ/mDM ≲ 0.1 cm2 g−1 at both v ≳ 1000 km s−1
and v ≪ 30 km s−1 . This pattern implies a velocity-dependent cross section with a peak
near v ∼ 100 km s−1 and a steep drop toward lower velocities, consistent with models
in which DM self-scattering proceeds via a resonant intermediate state by means of a
Breit-Wigner cross section [50, 101]. While Fig. 7 is in qualitative agreement with the
aforementioned effects, in this work we do not attempt a comprehensive global fit, includ-
ing the relic density constraint, due to the ongoing investigations associated with resonant
self-interacting DM. For instance, at present there is no N -body simulation for models
exhibiting the velocity dependence of the Breit-Wigner form given in Eq. (5.2). Moreover,
while a study based on a gravothermal fluid approximation has been conducted [101], the
core-collapse phase has been examined only for models with velocity-independent cross

– 24 –
sections or for scenarios motivated by light-mediator interactions [107–110], remaining in-
sufficiently understood for Breit-Wigner cross sections. A detailed investigation will be
presented elsewhere.
Thus, QCD-like theories with a non-zero θ parameter provide a novel mechanism for
DM production through number-changing interactions, while naturally accommodating
velocity-dependent self-interactions. These features allow such models to evade stringent
cluster constraints and remain compatible with small-scale structure observations, poten-
tially revitalizing the SIDM framework in light of recent astrophysical data.

6 Conclusions

Understanding the nature of DM remains one of the most compelling and unresolved chal-
lenges in contemporary physics. Within this context, QCD-like theories that feature dark
pions as DM candidates offer a particularly attractive framework due to their structural
similarity with the visible sector. A recent advance in this direction [2] demonstrated that a
non-zero topological θ angle can qualitatively alter the dynamics of dark pions, notably by
inducing number-changing interactions that can drive thermal freeze-out consistent with
the observed relic abundance. Moreover, the same mechanism naturally leads to reso-
nant self-interactions that may influence the distribution of DM in galactic halos, thereby
realizing the SIDM paradigm in a theoretically coherent manner.
In this work, we revisited the framework introduced in Ref. [2] and addressed several
open issues, particularly those related to the connection between the dark sector and the
SM. By systematically analyzing the Boltzmann equations that govern the thermal history
of DM, we have shown that some of the simplifying assumptions made in the original
treatment, while valid, were unnecessarily restrictive.
To provide a concrete realization of the proposed scenario, we extended the benchmark
model BM1 introduced in Ref. [2] by introducing a dark photon mediator associated with an
additional U (1)d gauge symmetry. This portal interaction not only ensures thermal contact
between the dark and visible sectors but also has an impact on DM stability and indirect
detection signatures. We demonstrated that the extended model remains consistent with
current observational constraints, including those from indirect detection experiments and
astrophysical probes.
Importantly, our analysis confirms that the strongly coupled dark sector can simulta-
neously account for the observed relic abundance and the small-scale structure anomalies
that motivate SIDM models. The presence of resonant enhancements in the self-scattering
cross section enables the required velocity dependence: enhanced interactions on galactic
scales and suppressed interactions on cluster scales.
Looking forward, our results open several avenues for future research. From a theoret-
ical standpoint, it would be valuable to explore alternative portal interactions and gauge
group structures, particularly solving the Boltzmann equations for the DM distribution in
phase space. On the observational front, upcoming data from indirect detection experi-
ments, gravitational lensing studies, and precise measurements of dwarf galaxy dynamics
may offer critical tests of the velocity-dependent SIDM framework outlined in this work.

– 25 –
This is particularly timely in light of systems such as SDSS J0946+1006, which are claimed
to host an SIDM halo undergoing gravothermal collapse. A thorough assessment of this
requires a detailed study of the gravothermal evolution during the core-collapse phase of
DM halos, in which particles self-scatter via a Breit–Wigner resonance.
In summary, our findings demonstrate that QCD-like DM theories with a non-zero
θ angle furnish a viable and predictive framework that naturally unifies thermal freeze-
out with small-scale structure phenomenology. By establishing a connection between relic
abundance and halo dynamics through a common underlying mechanism, these models
provide a compelling DM sector.

Acknowledgments

We are indebted to Ayuki Kamada for insightful and valuable discussions. GL is sup-
ported by the Generalitat Valenciana APOSTD/2023 Grant No. CIAPOS/2022/ 193.
C.G.C. is supported by a Ramón y Cajal contract with Ref. RYC2020-029248-I, the
Spanish National Grant PID2022-137268NA-C55 and Generalitat Valenciana through the
grant CIPROM/22/69. L.M. is supported by the Generalitat Valenciana through the
grant CIACIF/2023/91. O.Z. has been partially supported by Sostenibilidad-UdeA, the
UdeA/CODI Grant 2022-52380, and Ministerio de Ciencias Grant CD 82315 CT ICE-
TEX 2021-1080. O.Z. also acknowledges the support of the Fundación Carolina and the
hospitality of the Instituto de Fı́sica Corpuscular UV-CSIC in Valencia, Spain.

References
[1] Y. Hochberg, E. Kuflik, H. Murayama, T. Volansky and J.G. Wacker, Model for Thermal
Relic Dark Matter of Strongly Interacting Massive Particles, Phys. Rev. Lett. 115 (2015)
021301 [1411.3727].
[2] C. Garcı́a-Cely, G. Landini and O. Zapata, Dark matter in QCD-like theories with a theta
vacuum: Cosmological and astrophysical implications, Phys. Rev. D 111 (2025) 063044
[2405.10367].
[3] A. Kamada, H. Kim and T. Sekiguchi, Axionlike particle assisted strongly interacting
massive particle, Phys. Rev. D 96 (2017) 016007 [1704.04505].
[4] Y. Hochberg, E. Kuflik and H. Murayama, SIMP Spectroscopy, JHEP 05 (2016) 090
[1512.07917].
[5] E. Kuflik, M. Perelstein, N.R.-L. Lorier and Y.-D. Tsai, Elastically Decoupling Dark
Matter, Phys. Rev. Lett. 116 (2016) 221302 [1512.04545].
[6] N. Bernal, C. Garcia-Cely and R. Rosenfeld, WIMP and SIMP Dark Matter from the
Spontaneous Breaking of a Global Group, JCAP 04 (2015) 012 [1501.01973].
[7] N. Bernal and X. Chu, Z2 SIMP Dark Matter, JCAP 01 (2016) 006 [1510.08527].
[8] N. Bernal, X. Chu, C. Garcia-Cely, T. Hambye and B. Zaldivar, Production Regimes for
Self-Interacting Dark Matter, JCAP 03 (2016) 018 [1510.08063].
[9] S.-M. Choi and H.M. Lee, SIMP dark matter with gauged Z3 symmetry, JHEP 09 (2015)
063 [1505.00960].

– 26 –
[10] S.-M. Choi and H.M. Lee, Resonant SIMP dark matter, Phys. Lett. B 758 (2016) 47
[1601.03566].
[11] A. Soni and Y. Zhang, Hidden SU(N) Glueball Dark Matter, Phys. Rev. D 93 (2016)
115025 [1602.00714].
[12] A. Kamada, M. Yamada, T.T. Yanagida and K. Yonekura, SIMP from a strong U(1) gauge
theory with a monopole condensation, Phys. Rev. D 94 (2016) 055035 [1606.01628].
[13] N. Bernal, X. Chu and J. Pradler, Simply split strongly interacting massive particles, Phys.
Rev. D 95 (2017) 115023 [1702.04906].
[14] J.M. Cline, H. Liu, T. Slatyer and W. Xue, Enabling Forbidden Dark Matter, Phys. Rev. D
96 (2017) 083521 [1702.07716].
[15] S.-M. Choi, H.M. Lee and M.-S. Seo, Cosmic abundances of SIMP dark matter, JHEP 04
(2017) 154 [1702.07860].
[16] E. Kuflik, M. Perelstein, N.R.-L. Lorier and Y.-D. Tsai, Phenomenology of ELDER Dark
Matter, JHEP 08 (2017) 078 [1706.05381].
[17] M. Heikinheimo, K. Tuominen and K. Langæble, Hidden strongly interacting massive
particles, Phys. Rev. D 97 (2018) 095040 [1803.07518].
[18] S.-M. Choi, H.M. Lee, P. Ko and A. Natale, Resolving phenomenological problems with
strongly-interacting-massive-particle models with dark vector resonances, Phys. Rev. D 98
(2018) 015034 [1801.07726].
[19] Y. Hochberg, E. Kuflik, R. Mcgehee, H. Murayama and K. Schutz, Strongly interacting
massive particles through the axion portal, Phys. Rev. D 98 (2018) 115031 [1806.10139].
[20] N. Bernal, X. Chu, S. Kulkarni and J. Pradler, Self-interacting dark matter without
prejudice, Phys. Rev. D 101 (2020) 055044 [1912.06681].
[21] S.-M. Choi, H.M. Lee, Y. Mambrini and M. Pierre, Vector SIMP dark matter with
approximate custodial symmetry, JHEP 07 (2019) 049 [1904.04109].
[22] A. Katz, E. Salvioni and B. Shakya, Split SIMPs with Decays, JHEP 10 (2020) 049
[2006.15148].
[23] J. Smirnov and J.F. Beacom, New Freezeout Mechanism for Strongly Interacting Dark
Matter, Phys. Rev. Lett. 125 (2020) 131301 [2002.04038].
[24] C.-Y. Xing and S.-H. Zhu, Dark Matter Freeze-Out via Catalyzed Annihilation, Phys. Rev.
Lett. 127 (2021) 061101 [2102.02447].
[25] P. Braat and M. Postma, SIMPly add a dark photon, JHEP 03 (2023) 216 [2301.04513].
[26] E. Bernreuther, N. Hemme, F. Kahlhoefer and S. Kulkarni, Dark matter relic density in
strongly interacting dark sectors with light vector mesons, 2311.17157.
[27] R. Garani, M. Redi and A. Tesi, Dark QCD matters, JHEP 12 (2021) 139 [2105.03429].
[28] U.K. Dey, T.N. Maity and T.S. Ray, Light Dark Matter through Assisted Annihilation,
JCAP 03 (2017) 045 [1612.09074].
[29] R. Frumkin, Y. Hochberg, E. Kuflik and B. Vilk, Inelastically Decoupling Dark Matter,
2508.04772.
[30] M. Redi, A. Strumia, A. Tesi and E. Vigiani, Di-photon resonance and Dark Matter as
heavy pions, JHEP 05 (2016) 078 [1602.07297].

– 27 –
[31] P. Draper, J. Kozaczuk and J.-H. Yu, Theta in new QCD-like sectors, Phys. Rev. D 98
(2018) 015028 [1803.00015].
[32] T. Abe, R. Sato and T. Yamanaka, Composite dark matter with forbidden annihilation,
JHEP 09 (2024) 064 [2404.03963].
[33] B. Holdom, Two U(1)’s and Epsilon Charge Shifts, Phys. Lett. B 166 (1986) 196.
[34] D.N. Spergel and P.J. Steinhardt, Observational evidence for selfinteracting cold dark
matter, Phys. Rev. Lett. 84 (2000) 3760 [astro-ph/9909386].
[35] S. Tulin and H.-B. Yu, Dark Matter Self-interactions and Small Scale Structure, Phys.
Rept. 730 (2018) 1 [1705.02358].
[36] D. Harvey, R. Massey, T. Kitching, A. Taylor and E. Tittley, The non-gravitational
interactions of dark matter in colliding galaxy clusters, Science 347 (2015) 1462
[1503.07675].
[37] K. Bondarenko, A. Boyarsky, T. Bringmann and A. Sokolenko, Constraining self-interacting
dark matter with scaling laws of observed halo surface densities, JCAP 04 (2018) 049
[1712.06602].
[38] D. Harvey, A. Robertson, R. Massey and I.G. McCarthy, Observable tests of self-interacting
dark matter in galaxy clusters: BCG wobbles in a constant density core, Mon. Not. Roy.
Astron. Soc. 488 (2019) 1572 [1812.06981].
[39] L. Sagunski, S. Gad-Nasr, B. Colquhoun, A. Robertson and S. Tulin, Velocity-dependent
Self-interacting Dark Matter from Groups and Clusters of Galaxies, JCAP 01 (2021) 024
[2006.12515].
[40] DES collaboration, Examining the self-interaction of dark matter through central cluster
galaxy offsets, Mon. Not. Roy. Astron. Soc. 529 (2024) 52 [2304.10128].
[41] S. Adhikari et al., Astrophysical Tests of Dark Matter Self-Interactions, 2207.10638.
[42] S. Vegetti, L.V.E. Koopmans, A. Bolton, T. Treu and R. Gavazzi, Detection of a Dark
Substructure through Gravitational Imaging, Mon. Not. Roy. Astron. Soc. 408 (2010) 1969
[0910.0760].
[43] Q.E. Minor, S. Gad-Nasr, M. Kaplinghat and S. Vegetti, An unexpected high concentration
for the dark substructure in the gravitational lens SDSSJ0946+1006, Mon. Not. Roy.
Astron. Soc. 507 (2021) 1662 [2011.10627].
[44] Y. Bai and P. Schwaller, Scale of dark QCD, Phys. Rev. D 89 (2014) 063522 [1306.4676].
[45] A. Pich and E. de Rafael, Strong CP violation in an effective chiral Lagrangian approach,
Nucl. Phys. B 367 (1991) 313.
[46] S. Scherer, Introduction to chiral perturbation theory, Adv. Nucl. Phys. 27 (2003) 277
[hep-ph/0210398].
[47] A. Kamada, S. Kobayashi and T. Kuwahara, Perturbative unitarity of strongly interacting
massive particle models, JHEP 02 (2023) 217 [2210.01393].
[48] J. Wess and B. Zumino, Consequences of anomalous Ward identities, Phys. Lett. B 37
(1971) 95.
[49] E. Witten, Global Aspects of Current Algebra, Nucl. Phys. B 223 (1983) 422.

– 28 –
[50] X. Chu, C. Garcia-Cely and H. Murayama, Velocity Dependence from Resonant
Self-Interacting Dark Matter, Phys. Rev. Lett. 122 (2019) 071103 [1810.04709].
[51] Y.-D. Tsai, R. McGehee and H. Murayama, Resonant Self-Interacting Dark Matter from
Dark QCD, Phys. Rev. Lett. 128 (2022) 172001 [2008.08608].
[52] C. Csáki, A. Gomes, Y. Hochberg, E. Kuflik, K. Langhoff and H. Murayama,
Super-resonant dark matter, JHEP 11 (2022) 162 [2208.07882].
[53] T. Lee and Y.-D. Tsai, Naturally Resonant Dark Matter from Extra Dimensions,
2504.00076.
[54] X. Chu, M. Nikolic and J. Pradler, Even SIMP miracles are possible, 2401.12283.
[55] Planck collaboration, Planck 2018 results. VI. Cosmological parameters, Astron.
Astrophys. 641 (2020) A6 [1807.06209].
[56] R. Frumkin, Y. Hochberg, E. Kuflik and H. Murayama, Thermal Dark Matter from
Freeze-Out of Inverse Decays, Phys. Rev. Lett. 130 (2023) 121001 [2111.14857].
[57] M. Garny, J. Heisig, B. Lülf and S. Vogl, Coannihilation without chemical equilibrium,
Phys. Rev. D 96 (2017) 103521 [1705.09292].
[58] R.T. D’Agnolo, D. Pappadopulo and J.T. Ruderman, Fourth Exception in the Calculation
of Relic Abundances, Phys. Rev. Lett. 119 (2017) 061102 [1705.08450].
[59] L. Puetter, J.T. Ruderman, E. Salvioni and B. Shakya, Bouncing dark matter, Phys. Rev.
D 109 (2024) 023032 [2208.08453].
[60] D. Pappadopulo, J.T. Ruderman and G. Trevisan, Dark matter freeze-out in a
nonrelativistic sector, Phys. Rev. D 94 (2016) 035005 [1602.04219].
[61] E.D. Carlson, M.E. Machacek and L.J. Hall, Self-interacting dark matter, Astrophys. J.
398 (1992) 43.
[62] E.C.G. Stueckelberg, Interaction energy in electrodynamics and in the field theory of
nuclear forces, Helv. Phys. Acta 11 (1938) 225.
[63] F. Englert and R. Brout, Broken Symmetry and the Mass of Gauge Vector Mesons, Phys.
Rev. Lett. 13 (1964) 321.
[64] P.W. Higgs, Broken Symmetries and the Masses of Gauge Bosons, Phys. Rev. Lett. 13
(1964) 508.
[65] K.S. Babu, C.F. Kolda and J. March-Russell, Implications of generalized Z - Z-prime
mixing, Phys. Rev. D 57 (1998) 6788 [hep-ph/9710441].
[66] M. Cirelli, A. Strumia and J. Zupan, Dark Matter, 2406.01705.
[67] D. Wadekar and Z. Wang, Strong constraints on decay and annihilation of dark matter
from heating of gas-rich dwarf galaxies, Phys. Rev. D 106 (2022) 075007 [2111.08025].
[68] M. Boudaud, J. Lavalle and P. Salati, Novel cosmic-ray electron and positron constraints on
MeV dark matter particles, Phys. Rev. Lett. 119 (2017) 021103 [1612.07698].
[69] J. Koechler, X-rays constraints on sub-GeV Dark Matter, PoS TAUP2023 (2024) 044
[2309.10043].
[70] Y. Sun, J.W. Foster, H. Liu, J.B. Muñoz and T.R. Slatyer, Inhomogeneous energy injection
in the 21-cm power spectrum: Sensitivity to dark matter decay, Phys. Rev. D 111 (2025)
043015 [2312.11608].

– 29 –
[71] T.R. Slatyer and C.-L. Wu, General Constraints on Dark Matter Decay from the Cosmic
Microwave Background, Phys. Rev. D 95 (2017) 023010 [1610.06933].
[72] BaBar collaboration, Search for Invisible Decays of a Dark Photon Produced in e+ e−
Collisions at BaBar, Phys. Rev. Lett. 119 (2017) 131804 [1702.03327].
[73] Belle-II collaboration, The Belle II Physics Book, PTEP 2019 (2019) 123C01
[1808.10567].
[74] NA64 collaboration, Search for Light Dark Matter with NA64 at CERN, Phys. Rev. Lett.
131 (2023) 161801 [2307.02404].
[75] NA64 collaboration, Shedding light on dark sectors with high-energy muons at the NA64
experiment at the CERN SPS, Phys. Rev. D 110 (2024) 112015 [2409.10128].
[76] LDMX collaboration, Light Dark Matter eXperiment (LDMX), 1808.05219.
[77] C. Antel et al., Feebly-interacting particles: FIPs 2022 Workshop Report, Eur. Phys. J. C
83 (2023) 1122 [2305.01715].
[78] J. Dubinski and R.G. Carlberg, The Structure of cold dark matter halos, Astrophys. J. 378
(1991) 496.
[79] J.F. Navarro, C.S. Frenk and S.D.M. White, The Structure of cold dark matter halos,
Astrophys. J. 462 (1996) 563 [astro-ph/9508025].
[80] J.F. Navarro, C.S. Frenk and S.D.M. White, A Universal density profile from hierarchical
clustering, Astrophys. J. 490 (1997) 493 [astro-ph/9611107].
[81] R. Dave, D.N. Spergel, P.J. Steinhardt and B.D. Wandelt, Halo properties in cosmological
simulations of selfinteracting cold dark matter, Astrophys. J. 547 (2001) 574
[astro-ph/0006218].
[82] M. Vogelsberger, J. Zavala and A. Loeb, Subhaloes in Self-Interacting Galactic Dark Matter
Haloes, Mon. Not. Roy. Astron. Soc. 423 (2012) 3740 [1201.5892].
[83] M. Rocha, A.H.G. Peter, J.S. Bullock, M. Kaplinghat, S. Garrison-Kimmel, J. Onorbe
et al., Cosmological Simulations with Self-Interacting Dark Matter I: Constant Density
Cores and Substructure, Mon. Not. Roy. Astron. Soc. 430 (2013) 81 [1208.3025].
[84] A.H.G. Peter, M. Rocha, J.S. Bullock and M. Kaplinghat, Cosmological Simulations with
Self-Interacting Dark Matter II: Halo Shapes vs. Observations, Mon. Not. Roy. Astron. Soc.
430 (2013) 105 [1208.3026].
[85] O.D. Elbert, J.S. Bullock, S. Garrison-Kimmel, M. Rocha, J. Oñorbe and A.H.G. Peter,
Core formation in dwarf haloes with self-interacting dark matter: no fine-tuning necessary,
Mon. Not. Roy. Astron. Soc. 453 (2015) 29 [1412.1477].
[86] A.B. Fry, F. Governato, A. Pontzen, T. Quinn, M. Tremmel, L. Anderson et al., All about
baryons: revisiting SIDM predictions at small halo masses, Mon. Not. Roy. Astron. Soc.
452 (2015) 1468 [1501.00497].
[87] X. Chu, C. Garcia-Cely and H. Murayama, A Practical and Consistent Parametrization of
Dark Matter Self-Interactions, JCAP 06 (2020) 043 [1908.06067].
[88] N. Shah and S. Adhikari, The abundance of core-collapsed subhaloes in SIDM: insights from
structure formation in ΛCDM, Mon. Not. Roy. Astron. Soc. 529 (2024) 4611 [2308.16342].

– 30 –
 [89] S. Ando, S. Horigome, E.O. Nadler, D. Yang and H.-B. Yu, SASHIMI-SIDM:
semi-analytical subhalo modelling for self-interacting dark matter at sub-galactic scales,
JCAP 02 (2025) 053 [2403.16633].
[90] D. Yang, E.O. Nadler and H.-B. Yu, Strong Dark Matter Self-interactions Diversify Halo
Populations within and surrounding the Milky Way, Astrophys. J. 949 (2023) 67
[2211.13768].
[91] E.O. Nadler, D. Yang and H.-B. Yu, A Self-interacting Dark Matter Solution to the
Extreme Diversity of Low-mass Halo Properties, Astrophys. J. Lett. 958 (2023) L39
[2306.01830].
[92] M. Kaplinghat, S. Tulin and H.-B. Yu, Dark Matter Halos as Particle Colliders: Unified
Solution to Small-Scale Structure Puzzles from Dwarfs to Clusters, Phys. Rev. Lett. 116
(2016) 041302 [1508.03339].
[93] K.E. Andrade, J. Fuson, S. Gad-Nasr, D. Kong, Q. Minor, M.G. Roberts et al., A stringent
upper limit on dark matter self-interaction cross-section from cluster strong lensing, Mon.
Not. Roy. Astron. Soc. 510 (2021) 54 [2012.06611].
[94] T. Ren, A. Kwa, M. Kaplinghat and H.-B. Yu, Reconciling the Diversity and Uniformity of
Galactic Rotation Curves with Self-Interacting Dark Matter, Phys. Rev. X 9 (2019) 031020
[1808.05695].
[95] M.G. Roberts, M. Kaplinghat, M. Valli and H.-B. Yu, Gravothermal collapse and the
diversity of galactic rotation curves, Phys. Rev. D 111 (2025) 103041 [2407.15005].
[96] M. Valli and H.-B. Yu, Dark matter self-interactions from the internal dynamics of dwarf
spheroidals, Nature Astron. 2 (2018) 907 [1711.03502].
[97] J.I. Read, M.G. Walker and P. Steger, The case for a cold dark matter cusp in Draco, Mon.
Not. Roy. Astron. Soc. 481 (2018) 860 [1805.06934].
[98] O. Sameie, H.-B. Yu, L.V. Sales, M. Vogelsberger and J. Zavala, Self-Interacting Dark
Matter Subhalos in the Milky Way’s Tides, Phys. Rev. Lett. 124 (2020) 141102
[1904.07872].
[99] M. Silverman, J.S. Bullock, M. Kaplinghat, V.H. Robles and M. Valli, Motivations for a
large self-interacting dark matter cross-section from Milky Way satellites, Mon. Not. Roy.
Astron. Soc. 518 (2022) 2418 [2203.10104].
[100] K. Hayashi, M. Ibe, S. Kobayashi, Y. Nakayama and S. Shirai, Probing dark matter
self-interaction with ultrafaint dwarf galaxies, Phys. Rev. D 103 (2021) 023017
[2008.02529].
[101] A. Kamada and H.J. Kim, Evolution of resonant self-interacting dark matter halos, Phys.
Rev. D 109 (2024) 063535 [2304.12621].
[102] W.J.R. Enzi, C.M. Krawczyk, D.J. Ballard and T.E. Collett, The overconcentrated dark
halo in the strong lens SDSS J0946+1006 is a subhalo: evidence for self interacting dark
matter?, 2411.08565.
[103] M. Tajalli, S. Vegetti, C.M. O’Riordan, S.D.M. White, C.D. Fassnacht, D.M. Powell et al.,
SHARP – IX. The dense, low-mass perturbers in B1938+666 and J0946+1006: implications
for cold and self-interacting dark matter, 2505.07944.
[104] S. Li et al., The ”Little Dark Dot”: Evidence for Self-Interacting Dark Matter in the Strong
Lens SDSSJ0946+1006?, 2504.11800.

– 31 –
[105] F. Jiang, Z. Jia, H. Zheng, L.C. Ho, K. Inayoshi, X. Shen et al., Formation of the Little Red
Dots from the Core-collapse of Self-interacting Dark Matter Halos, 2503.23710.
[106] M.G. Roberts, L. Braff, A. Garg, S. Profumo, T. Jeltema and J. O’Donnell, Early
formation of supermassive black holes from the collapse of strongly self-interacting dark
matter, JCAP 01 (2025) 060 [2410.17480].
[107] V. Tran, D. Gilman, M. Vogelsberger, X. Shen, S. O’Neil and X. Zhang, Gravothermal
Catastrophe in Resonant Self-interacting Dark Matter Models, 2405.02388.
[108] D. Yang and H.-B. Yu, Gravothermal evolution of dark matter halos with differential elastic
scattering, JCAP 09 (2022) 077 [2205.03392].
[109] S. Yang, X. Du, Z.C. Zeng, A. Benson, F. Jiang, E.O. Nadler et al., Gravothermal Solutions
of SIDM Halos: Mapping from Constant to Velocity-dependent Cross Section, Astrophys. J.
946 (2023) 47 [2205.02957].
[110] N.J. Outmezguine, K.K. Boddy, S. Gad-Nasr, M. Kaplinghat and L. Sagunski, Universal
gravothermal evolution of isolated self-interacting dark matter halos for velocity-dependent
cross-sections, Mon. Not. Roy. Astron. Soc. 523 (2023) 4786 [2204.06568].

– 32 –