SIMP Dark Matter: Number-Changing Dynamics

• SIMP dark matter is defined by dark-sector number-changing reactions (primarily 3→2 or 4→2) that set its relic density, unlike traditional 2→2 WIMP annihilations.
• The thermal history involves cannibalization where conversion of rest mass to kinetic energy requires efficient kinetic equilibrium via portals such as dark-photon, ALP, or Higgs couplings.
• Model variants include dark pion chiral theories with vector resonances, symmetry-driven 4→2 frameworks, and adjustments via non-standard cosmologies, which broaden the viable parameter space.

Strongly Interacting Massive Particle (SIMP) denotes a dark-matter paradigm in which the relic abundance is set primarily by dark-sector number-changing reactions rather than by conventional $2\to2$ annihilations into Standard Model (SM) states. In the canonical formulation, the dominant process is $3\to2$, while symmetry-restricted variants can instead realize $4\to2$. The resulting thermal history is characterized by strong self-interactions, cannibal heating, and a central requirement that the dark sector remain in kinetic equilibrium with a heat sink during freeze-out. In standard radiation domination this often points to MeV–GeV dark matter, but hidden sectors, resonant mediators, and non-standard pre-BBN cosmologies substantially broaden the viable parameter space (Chowdhury et al., 2024, Choi et al., 2018).

1. Thermal mechanism and cannibal dynamics

The defining kinetic equation of the SIMP framework is the number-changing Boltzmann equation. In the notation used for general $r\to2$ freeze-out, with $r=3,4$,

$$\frac{dn_\chi}{dt}+3H n_\chi=-(r-2)\,\langle \sigma_{r\to 2} v^{r-1}\rangle\left(n_\chi^r-n_\chi^2 n_{\chi,\rm eq}^{\,r-2}\right),$$

so the relic density is controlled by dark-sector self-annihilations rather than by annihilation into SM particles (Chowdhury et al., 2024). For $3\to2$, freeze-out occurs when $$n_{\rm eq}^2\langle \sigma_{3\to2}v^2\rangle \simeq H$$; for $4\to2$, when $$n_{\rm eq}^3\langle \sigma_{4\to2}v^3\rangle \simeq H$$ (Chowdhury et al., 2024).

A distinctive thermodynamic feature is cannibalization. Because a $3\to2$0 or $3\to2$1 reaction converts rest mass into kinetic energy of the remaining dark particles, the dark sector cools more slowly than an ordinary nonrelativistic gas unless it can dump entropy into another bath. In a hidden-sector SIMP with $3\to2$2, the nonrelativistic cannibal phase obeys

$3\to2$3

and the entropy-conserving yield is

$3\to2$4

with $3\to2$5 the SM-to-dark entropy ratio (Heikinheimo et al., 2018). This already shows that SIMP freeze-out is not only a question of cross sections; it is equally a question of thermal contact.

The usual contrast with WIMPs is therefore incomplete unless the thermal structure is stated explicitly. WIMP relic density scales inversely with a $3\to2$6 annihilation cross section into SM states, whereas SIMP relic density is controlled by dark-sector number-changing processes and by whether kinetic equilibrium is maintained long enough for the cannibal heat to be removed. A common symmetry-driven variant replaces $3\to2$7 by $3\to2$8: in $3\to2$9-stabilized models, vertices with an odd number of dark fields are forbidden, so $4\to2$0 is absent and $4\to2$1 dominates (Bernal et al., 2015).

2. Dark pions, chiral dynamics, and vector resonances

The best-developed ultraviolet picture of SIMP dark matter is a confining gauge theory in which the dark matter states are pseudo-Nambu–Goldstone bosons. In the dark-pion construction based on

$4\to2$2

the low-energy theory is a dark chiral perturbation theory with decay constant $4\to2$3, and the Wess–Zumino–Witten term generates the five-point interaction responsible for $4\to2$4 (Choi et al., 2018). In the pion-only effective theory, the relevant anomalous operator is

$4\to2$5

This is the canonical origin of SIMP number-changing dynamics in composite models (Choi et al., 2018).

The same framework also makes clear why pure-pion SIMPs are often under pressure. To reproduce the observed relic density using only the WZW interaction, the chiral expansion parameter is driven toward the perturbative boundary. A practical criterion emphasized in the dark-pion literature is

$4\to2$6

and the WZW-only case violates this bound over the mass range relevant for Bullet Cluster and small-scale structure considerations (Choi et al., 2018). In a related hidden-sector analysis of composite SIMPs, the model-independent parametrization

$4\to2$7

makes the same tension manifest: large $4\to2$8 rates and acceptable $4\to2$9 self-interactions are difficult to reconcile if the dark sector temperature tracks the SM exactly (Heikinheimo et al., 2018).

A controlled resolution is to include dark vector resonances. In the hidden local symmetry construction, the vectors are massive gauge bosons of gauged $r\to2$0, with

$r\to2$1

The gauged WZW sector then generates vector-mediated five-point topologies, and near the “3-pion resonance”

$r\to2$2

the $r\to2$3 amplitude is resonantly enhanced while the underlying chiral couplings remain moderate (Choi et al., 2018). The narrow-width requirement

$r\to2$4

implies $r\to2$5 near the $r\to2$6 resonance (Choi et al., 2018).

This unitarization mechanism is quantitatively significant. With vectors, “in the mass range of light dark matter $r\to2$7,” one can maintain

$r\to2$8

and “parameter space is wider than the only WZW case” (Choi et al., 2018). The related analysis of dark vector resonances in hidden local symmetry reaches the same conclusion: resonances at $r\to2$9 or $r=3,4$0 enhance $r=3,4$1, reduce the required $r=3,4$2, and bring $r=3,4$3 into the $r=3,4$4–$r=3,4$5 range without leaving the domain of validity of dark ChPT (Choi et al., 2018).

A second way to regain perturbative control is to cool the dark sector relative to the SM. In hidden-sector composite SIMPs, a temperature ratio $r=3,4$6 lowers the required $r=3,4$7 strength. Explicitly, for $r=3,4$8, a viable mass range $r=3,4$9–$$\frac{dn_\chi}{dt}+3H n_\chi=-(r-2)\,\langle \sigma_{r\to 2} v^{r-1}\rangle\left(n_\chi^r-n_\chi^2 n_{\chi,\rm eq}^{\,r-2}\right),$$0 overlaps $$\frac{dn_\chi}{dt}+3H n_\chi=-(r-2)\,\langle \sigma_{r\to 2} v^{r-1}\rangle\left(n_\chi^r-n_\chi^2 n_{\chi,\rm eq}^{\,r-2}\right),$$1, while for $$\frac{dn_\chi}{dt}+3H n_\chi=-(r-2)\,\langle \sigma_{r\to 2} v^{r-1}\rangle\left(n_\chi^r-n_\chi^2 n_{\chi,\rm eq}^{\,r-2}\right),$$2 the viable range shifts to $$\frac{dn_\chi}{dt}+3H n_\chi=-(r-2)\,\langle \sigma_{r\to 2} v^{r-1}\rangle\left(n_\chi^r-n_\chi^2 n_{\chi,\rm eq}^{\,r-2}\right),$$3–$$\frac{dn_\chi}{dt}+3H n_\chi=-(r-2)\,\langle \sigma_{r\to 2} v^{r-1}\rangle\left(n_\chi^r-n_\chi^2 n_{\chi,\rm eq}^{\,r-2}\right),$$4 (Heikinheimo et al., 2018). This is one of the cleanest demonstrations that the standard “sub-GeV and strongly coupled” SIMP expectation is not universal, but thermal-history dependent.

3. Symmetry classes and representative model realizations

Symmetry determines which number-changing channel is even available. Bernal and Chu’s $$\frac{dn_\chi}{dt}+3H n_\chi=-(r-2)\,\langle \sigma_{r\to 2} v^{r-1}\rangle\left(n_\chi^r-n_\chi^2 n_{\chi,\rm eq}^{\,r-2}\right),$$5 singlet-scalar construction is the canonical $$\frac{dn_\chi}{dt}+3H n_\chi=-(r-2)\,\langle \sigma_{r\to 2} v^{r-1}\rangle\left(n_\chi^r-n_\chi^2 n_{\chi,\rm eq}^{\,r-2}\right),$$6 example: the potential

$$\frac{dn_\chi}{dt}+3H n_\chi=-(r-2)\,\langle \sigma_{r\to 2} v^{r-1}\rangle\left(n_\chi^r-n_\chi^2 n_{\chi,\rm eq}^{\,r-2}\right),$$7

forbids odd-$$\frac{dn_\chi}{dt}+3H n_\chi=-(r-2)\,\langle \sigma_{r\to 2} v^{r-1}\rangle\left(n_\chi^r-n_\chi^2 n_{\chi,\rm eq}^{\,r-2}\right),$$8 interactions, so $$\frac{dn_\chi}{dt}+3H n_\chi=-(r-2)\,\langle \sigma_{r\to 2} v^{r-1}\rangle\left(n_\chi^r-n_\chi^2 n_{\chi,\rm eq}^{\,r-2}\right),$$9 is absent and

$3\to2$0

sets the relic density, while

$3\to2$1

controls self-interactions (Bernal et al., 2015). In that scenario, a colder dark sector at freeze-out, with $3\to2$2–$3\to2$3, reduces the required $3\to2$4 and helps reconcile relic density with cluster bounds (Bernal et al., 2015).

Other symmetry choices support $3\to2$5 while embedding SIMPs in broader BSM structures. A radiative neutrino-mass model with a $3\to2$6 symmetry realizes resonantly enhanced $3\to2$7 through a complex scalar $3\to2$8 and mediator $3\to2$9, while the same Yukawa sector generates neutrino masses at two loops and keeps the dark sector in kinetic equilibrium through $$n_{\rm eq}^2\langle \sigma_{3\to2}v^2\rangle \simeq H$$0 scattering (Ho et al., 2017). Twin Higgs constructions identify the SIMP states with twin pions in a QCD-like sector, with the WZW term providing $$n_{\rm eq}^2\langle \sigma_{3\to2}v^2\rangle \simeq H$$1 and a kinetically mixed twin photon maintaining kinetic equilibrium; the preferred regime is $$n_{\rm eq}^2\langle \sigma_{3\to2}v^2\rangle \simeq H$$2 few $$n_{\rm eq}^2\langle \sigma_{3\to2}v^2\rangle \simeq H$$3 and $$n_{\rm eq}^2\langle \sigma_{3\to2}v^2\rangle \simeq H$$4 (Hochberg et al., 2018).

A different ultraviolet realization replaces non-Abelian confinement with a strongly interacting $$n_{\rm eq}^2\langle \sigma_{3\to2}v^2\rangle \simeq H$$5 sector and monopole condensation. In that construction, monopole condensation confines $$n_{\rm eq}^2\langle \sigma_{3\to2}v^2\rangle \simeq H$$6-charged matter, hidden pions emerge as the low-energy composites, and the radial monopole mode mixes with the Higgs so that no additional singlet portal is required (Kamada et al., 2016). The model is notable because baryons would remain $$n_{\rm eq}^2\langle \sigma_{3\to2}v^2\rangle \simeq H$$7-charged, so there are no low-energy baryon states competing with the pion SIMP.

Two-component SIMP sectors introduce additional dynamical structure. In the accidental $$n_{\rm eq}^2\langle \sigma_{3\to2}v^2\rangle \simeq H$$8 model with a complex scalar and a vector-like fermion, $$n_{\rm eq}^2\langle \sigma_{3\to2}v^2\rangle \simeq H$$9 interactions determine freeze-out, but an unavoidable two-loop induced $4\to2$0 process redistributes the component abundances after chemical freeze-out and “would significantly modify the predictions of the self-interacting cross section of DM compared with other SIMP models” (Ho et al., 2022). A related 2024 analysis considers a pFIMP in the presence of a SIMP and shows how SIMP–pFIMP conversion interpolates between a pure SIMP–FIMP limit and a SIMP–pFIMP limit in which DM–DM conversion reshapes the relic composition (Bhattacharya et al., 2024). These models make explicit that “SIMP” can denote a thermal mechanism rather than a single-particle species.

4. Portals, kinetic equilibrium, and experimental probes

Kinetic equilibrium is structurally central because, as one dark-pion study puts it, “one of the pion masses goes to kinetic energy of dark matter in the $4\to2$1 channel” (Choi et al., 2018). Portal design is therefore not optional. Higgs- and singlet-scalar portals, $4\to2$2 portals, axion-like particles, and neutrino-coupled mediators have all been used to transfer entropy from the dark sector to the SM bath.

The dark-photon portal is the most extensively developed experimentally. In “SIMP Spectroscopy,” a kinetically mixed vector boson maintains thermal contact while allowing mono-photon missing-mass spectroscopy at $4\to2$3 colliders. The recoil photon energy directly measures the dark invariant mass,

$4\to2$4

so the dark resonance structure can, in principle, be reconstructed from the mono-photon spectrum (Hochberg et al., 2015). The 2023 “SIMPly add a dark photon” analysis pushes this idea further by asking whether one dark photon can simultaneously thermalize the dark sector, resonantly enhance $4\to2$5, and produce velocity-dependent self-interactions. The answer is model-dependent: for $4\to2$6 the minimal setup is “marginally excluded, as the required kinetic mixing is too small to maintain thermal equilibrium with the SM,” whereas adding an extra dark quark yields viable models with $4\to2$7–$4\to2$8 (Braat et al., 2023).

Axion and ALP portals were developed precisely to solve the heat-dump problem without sacrificing the SIMP mechanism. In the ALP-assisted model, semi-annihilation $4\to2$9 dominates the relic-setting dynamics, while the ALP maintains SM contact through

$$n_{\rm eq}^3\langle \sigma_{4\to2}v^3\rangle \simeq H$$0

and the region yielding $$n_{\rm eq}^3\langle \sigma_{4\to2}v^3\rangle \simeq H$$1–$$n_{\rm eq}^3\langle \sigma_{4\to2}v^3\rangle \simeq H$$2 was identified as reachable by SHiP (Kamada et al., 2017). A later axion-portal construction based on $$n_{\rm eq}^3\langle \sigma_{4\to2}v^3\rangle \simeq H$$3 dark pions likewise found viable parameter space when the ALP mass is close to the SIMP mass, with strong-scale masses of order a few hundred MeV and near-future coverage by beam-dump and collider experiments (Hochberg et al., 2018).

Higgs-portal thermalization remains viable in singlet-scalar and related models. In the reheating-era singlet-scalar analysis, direct detection, invisible Higgs decays, and future collider sensitivity are all mapped simultaneously: present LZ limits already exclude parts of the allowed SIMP region, while DARWIN/XLZD and precision Higgs measurements at HL-LHC and FCC-ee will probe additional regions (Bélanger et al., 5 Mar 2026). In the radiative neutrino-mass model, kinetic equilibrium is instead maintained by Yukawa scatterings off neutrinos until freeze-out (Ho et al., 2017). The portal question is thus inseparable from experimental strategy: electron recoils, mono-photons, missing momentum, invisible Higgs width, and displaced low-mass dileptons all arise as portal-specific probes.

5. Non-standard cosmologies and the enlargement of mass scales

Standard radiation domination leads to stringent mass limits from the combination of relic density and unitarity. For scalar SIMPs freezing out through $$n_{\rm eq}^3\langle \sigma_{4\to2}v^3\rangle \simeq H$$4 or $$n_{\rm eq}^3\langle \sigma_{4\to2}v^3\rangle \simeq H$$5, the radiation-dominated upper bounds quoted in the reheating analysis are

$$n_{\rm eq}^3\langle \sigma_{4\to2}v^3\rangle \simeq H$$6

assuming s-wave scalar dark matter and nonrelativistic freeze-out (Chowdhury et al., 2024). These values are often taken as indicative of the SIMP scale.

Reheating changes that conclusion because the Hubble rate and entropy evolution differ from radiation domination. During a reheating phase driven by a monomial inflaton potential $$n_{\rm eq}^3\langle \sigma_{4\to2}v^3\rangle \simeq H$$7, the paper derives different temperature and Hubble scalings for fermionic and bosonic reheating channels, and shows that entropy injection dilutes the final yield. Consequently, “a smaller cross-section” is needed than in the radiation-dominated case, and the unitarity bounds relax dramatically. For quadratic inflaton potential $$n_{\rm eq}^3\langle \sigma_{4\to2}v^3\rangle \simeq H$$8, the upper mass limits become

$$n_{\rm eq}^3\langle \sigma_{4\to2}v^3\rangle \simeq H$$9

whereas for quartic potential $3\to2$00 the allowed space is much smaller, with $3\to2$01 for bosonic reheating and $3\to2$02 for fermionic reheating in the $3\to2$03 case (Chowdhury et al., 2024).

A complementary result was obtained in the reheating-era singlet-scalar study of $3\to2$04 SIMP dark matter with $3\to2$05. In standard radiation domination, this realization “is not viable, as it requires sub-MeV masses and large quartic couplings in tension with bounds on dark matter self-interactions.” During early matter domination, however, the altered Hubble rate and entropy injection make the model viable over an enormous range: $3\to2$06 with

$3\to2$07

subject to BBN, perturbativity, and self-interaction bounds (Bélanger et al., 5 Mar 2026). This does not erase the standard sub-GeV SIMP regime; it shows instead that it is a statement about cosmological history, not a theorem about number-changing dark matter.

6. Conceptual tensions, terminology, and current status

Two persistent tensions define the modern SIMP literature. The first is the relic-density versus perturbativity problem: in pure WZW dark-pion models, the $3\to2$08 rate needed for $3\to2$09 frequently drives $3\to2$10 beyond the reliable regime of chiral perturbation theory (Choi et al., 2018). The second is the heat-dump problem: cannibalization requires kinetic equilibrium with some bath through freeze-out, but that same portal must not reintroduce dominant WIMP-like $3\to2$11 annihilation into the SM. Vector resonances, cooler hidden sectors, ALP portals, and non-standard cosmologies should therefore be regarded not as optional embellishments, but as the principal mechanisms by which the simplest SIMP tension is resolved.

This is also why “minimal” constructions can fail. The dark-photon-only setup studied in 2023 is explicitly “marginally excluded” for $3\to2$12 because the kinetic mixing that preserves the relic density is too small to thermalize the dark pions with the SM, while $3\to2$13 opens viable space and yields acceptable models for $3\to2$14–$3\to2$15 (Braat et al., 2023). The same study emphasizes that late-time annihilations are non-negligible, making the dark pion “a bit WIMPy” (Braat et al., 2023). That formulation captures a broader lesson: once resonant portals are added, the sharp WIMP–SIMP dichotomy becomes less rigid at the level of phenomenology even when the relic density is still primarily controlled by number-changing interactions.

A separate source of confusion is terminological. In the thermal-relic literature, SIMP almost always means a dark matter candidate whose abundance is set by $3\to2$16 or $3\to2$17 reactions inside the dark sector. In collider and direct-detection phenomenology, however, the same acronym is also used for dark matter with large nucleon cross sections. Surface $3\to2$18-cleus data exclude SIMP–nucleon cross sections up to approximately $3\to2$19 for masses above $3\to2$20 (Davis, 2017), and trackless-jet searches at the LHC were argued to tentatively exclude roughly $3\to2$21 for $3\to2$22 in simplified models (Daci et al., 2015). These studies concern a different interaction regime from the standard thermal SIMP mechanism, even though the acronym is shared.

The present status is therefore plural rather than singular. SIMP dark matter encompasses dark-pion WZW freeze-out, $3\to2$23 $3\to2$24 singlet scalars, hidden-sector cannibal models with $3\to2$25, resonantly unitarized pion theories with dark vectors, ALP- and dark-photon-mediated thermalization, Twin Higgs embeddings, and multicomponent sectors with post-freeze-out reshuffling (Bernal et al., 2015, Heikinheimo et al., 2018, Choi et al., 2018, Ho et al., 2022). What unifies these constructions is not a single mass range or mediator choice, but the replacement of WIMP-like chemical decoupling by number-changing dark-sector dynamics together with the requirement that the resulting cannibal heat be consistently managed.