Cascade Decaying Dark Matter

• Cascade decaying dark matter is a framework where dark matter decays via a sequence of intermediate states, redistributing energy before final Standard Model interactions.
• The multi-step decay chains, featuring hierarchical and degenerate steps, modify photon spectra and neutrino signals while influencing relic abundance and cosmological evolution.
• Extended models include continuum decay spectra, instanton-mediated processes, and medium-induced disassembly, offering diverse observational signatures such as CMB distortions and IceCube cascades.

Searching arXiv for recent and relevant papers on cascade decaying dark matter and related frameworks. Cascade decaying dark matter denotes a class of dark-sector frameworks in which the observable consequences of dark matter are controlled by a sequence of decays, conversions, or induced breakups rather than a single direct transition to Standard Model states. In the literature, this includes multi-step dark-sector ladders ending in $f\bar f$ final states, relic-abundance mechanisms driven by decays of nearly degenerate partners, continuous decays in a gapped continuum, and medium-induced disassembly of composite dark matter while traversing the Earth (Elor et al., 2015, Dror et al., 2016, Csáki et al., 2021, Boukhtouchen et al., 17 Dec 2025).

1. Terminological scope and conceptual core

A common core runs through otherwise distinct constructions. Dark matter is not treated as an isolated stable particle with one dominant annihilation or decay channel into the Standard Model. Instead, the relevant dynamics are redistributed across intermediate states, so that either the relic abundance, the late-time signal, or the detector response is shaped by a cascade. In the simplest dark-sector ladder, a decay chain takes the form

$$\chi \to \phi_n\phi_n \to 2\times \phi_{n-1}\phi_{n-1}\to \cdots \to 2^{n-1}\times\phi_1\phi_1 \to 2^n\times f\bar f,$$

with $\phi_i$ denoting lighter dark states and $f\bar f$ the Standard Model endpoint (Elor et al., 2015).

The expression also appears in neutrino astronomy in a second, detector-level sense. PeV-scale decaying dark matter can yield neutrino fluxes that are dominated by IceCube cascade events, because $\nu_e$ charged-current interactions, most $\nu_\tau$ charged-current interactions, and all neutral-current interactions produce shower-like topologies rather than tracks (Esmaili et al., 2013). A third usage arises in medium-induced cascades: loosely bound composite dark matter can dissociate constituent by constituent while crossing the Earth, producing a scattering cascade whose signatures are non-collinear multiscatters, timing-separated events, and possible coincidences across underground laboratories (Boukhtouchen et al., 17 Dec 2025).

These usages are not identical, but they are structurally related. In all cases, the salient observable is set by energy flow through intermediate states or intermediate interactions, not by a one-step process.

2. Multi-step dark-sector cascades and spectral kinematics

The most developed kinematic framework is the multi-step dark-sector cascade studied for gamma-ray phenomenology. For decays, the first step is $\chi\to\phi_n\phi_n$, and the photon spectrum is obtained from the decay differential rate according to

$$\frac{1}{N_\gamma}\frac{dN_\gamma}{dE_\gamma} = \frac{1}{\Gamma}\frac{d\Gamma}{dE_\gamma}.$$

For the same internal cascade, the decay spectrum has the same shape as the annihilation spectrum, except that the initial dark matter particle is twice as heavy (Elor et al., 2015).

The cascade is parameterized by

$$\epsilon_i=\frac{2m_i}{m_{i+1}},\qquad \epsilon_f=\frac{2m_f}{m_1}.$$

A hierarchical step has $\epsilon_i\ll 1$, so the daughters are highly boosted; a degenerate step has $$\chi \to \phi_n\phi_n \to 2\times \phi_{n-1}\phi_{n-1}\to \cdots \to 2^{n-1}\times\phi_1\phi_1 \to 2^n\times f\bar f,$$0, so the daughters are nearly at rest in the parent frame. For an $$\chi \to \phi_n\phi_n \to 2\times \phi_{n-1}\phi_{n-1}\to \cdots \to 2^{n-1}\times\phi_1\phi_1 \to 2^n\times f\bar f,$$1-step chain ending in a fermion of mass $$\chi \to \phi_n\phi_n \to 2\times \phi_{n-1}\phi_{n-1}\to \cdots \to 2^{n-1}\times\phi_1\phi_1 \to 2^n\times f\bar f,$$2,

$$\chi \to \phi_n\phi_n \to 2\times \phi_{n-1}\phi_{n-1}\to \cdots \to 2^{n-1}\times\phi_1\phi_1 \to 2^n\times f\bar f,$$3

which implies the self-consistency bound

$$\chi \to \phi_n\phi_n \to 2\times \phi_{n-1}\phi_{n-1}\to \cdots \to 2^{n-1}\times\phi_1\phi_1 \to 2^n\times f\bar f,$$4

This fixes the maximum number of hierarchical steps allowed for a given $$\chi \to \phi_n\phi_n \to 2\times \phi_{n-1}\phi_{n-1}\to \cdots \to 2^{n-1}\times\phi_1\phi_1 \to 2^n\times f\bar f,$$5 and final state (Elor et al., 2015).

In the hierarchical limit, the central recursion is the boost convolution

$$\chi \to \phi_n\phi_n \to 2\times \phi_{n-1}\phi_{n-1}\to \cdots \to 2^{n-1}\times\phi_1\phi_1 \to 2^n\times f\bar f,$$6

which recursively maps the $$\chi \to \phi_n\phi_n \to 2\times \phi_{n-1}\phi_{n-1}\to \cdots \to 2^{n-1}\times\phi_1\phi_1 \to 2^n\times f\bar f,$$7-step spectrum to the $$\chi \to \phi_n\phi_n \to 2\times \phi_{n-1}\phi_{n-1}\to \cdots \to 2^{n-1}\times\phi_1\phi_1 \to 2^n\times f\bar f,$$8-step spectrum. Each hierarchical step doubles multiplicity, lowers the characteristic photon energy by roughly a factor of $$\chi \to \phi_n\phi_n \to 2\times \phi_{n-1}\phi_{n-1}\to \cdots \to 2^{n-1}\times\phi_1\phi_1 \to 2^n\times f\bar f,$$9, and broadens the spectrum. By contrast, a degenerate step does not broaden the spectrum; it only doubles multiplicity and halves the energy scale. This distinction is central to model building, because hierarchical steps control spectral broadening while degenerate steps raise the mass scale without changing the spectral shape (Elor et al., 2015).

This directly modifies indirect-detection expectations. Hard 0-step spectra such as $\phi_i$0 can become broad GeV-scale gamma-ray spectra after several steps, whereas soft 0-step spectra such as $\phi_i$1 require fewer steps. In the Galactic Center excess analysis, hierarchical cascades preferred $\phi_i$2 in the range $\phi_i$3–$\phi_i$4 for all final states, while degenerate steps admitted much higher masses; the paper also emphasized that the observed morphology disfavors decaying, rather than annihilating, explanations because decays trace $\phi_i$5 rather than $\phi_i$6 (Elor et al., 2015).

3. Decay-controlled relic abundance and cosmological histories

Several frameworks use decays not merely as a late-time signal, but as the mechanism setting the dark matter abundance.

Framework	Defining mechanism	Characteristic consequence
Co-decay	Stable $\phi_i$7, unstable $\phi_i$8, $\phi_i$9, out-of-equilibrium $f\bar f$0 decay	Relic abundance set by decay-driven depletion
Vev flip-flop	$f\bar f$1 decay allowed only while $f\bar f$2 between weak-scale transitions	Temporary instability reduces an initially excessive abundance
Top-down injection	Defect-sourced injection $f\bar f$3	Continuous non-thermal dark matter production

In co-decaying dark matter, the dark sector contains a stable state $f\bar f$4 and a nearly degenerate unstable state $f\bar f$5, with rapid $f\bar f$6 keeping them in chemical equilibrium while $f\bar f$7 decays out of equilibrium to the Standard Model. The density is depleted exponentially through the decay of $f\bar f$8, rather than by ordinary Boltzmann suppression, and the required annihilation cross section is therefore boosted relative to the standard WIMP case (Dror et al., 2016). In the cosmological extension of this framework, the non-relativistic dark sector naturally drives an early matter-dominated phase whose duration is

$f\bar f$9

and whose small-scale consequence is enhanced perturbation growth and microhalo formation. The characteristic microhalo mass is set by the reheating horizon,

$\nu_e$0

which can generate substantial boost factors for indirect detection and point-source-like gamma-ray signals (Dror et al., 2017).

A distinct implementation is the “vev flip-flop” scenario. Here a fermionic singlet $\nu_e$1 freezes out while still relativistic and is therefore initially overabundant. As the universe cools, a scalar mediator $\nu_e$2 develops a vacuum expectation value, breaks the stabilizing symmetry, mixes $\nu_e$3 with electroweak-triplet fermions, and opens decay channels. When the Higgs subsequently acquires its own vacuum expectation value, portal terms restore the dark symmetry and make $\nu_e$4 stable again. The relic density is therefore set by the integrated decay rate during a finite temperature window bounded by two weak-scale phase transitions (Baker et al., 2016).

A third possibility is non-thermal top-down production from decaying topological defects. In that case, the dark matter source term is time dependent, with injection power

$\nu_e$5

leading to a Boltzmann equation

$\nu_e$6

For all $\nu_e$7, the paper derives a closed-form asymptotic yield and shows that topological defects can be the principal source of dark matter even when standard freeze-out underproduces the relic density, potentially producing large annihilation boost factors (Hindmarsh et al., 2013).

Taken together, these constructions show that cascade-decay ideas are not restricted to late decays of a pre-existing relic; they can reorganize the entire thermal history.

4. Indirect signatures: gamma rays, neutrinos, morphology, and detector cascades

In gamma-ray phenomenology, the decay flux is obtained by replacing the annihilation kernel with the decay kernel,

$\nu_e$8

so the spectral machinery of dark-sector cascades carries over directly, but the morphology changes from $\nu_e$9 to $\nu_\tau$0. That distinction was explicitly noted as a reason the observed Galactic Center excess disfavors simple decaying explanations even though the cascade spectra themselves remain viable in other contexts (Elor et al., 2015).

For high-energy neutrinos, PeV-scale decaying dark matter was proposed as an explanation of the early IceCube events. A benchmark model with

$\nu_\tau$1

and branching fractions $\nu_\tau$2 to $\nu_\tau$3 and $\nu_\tau$4 to $\nu_\tau$5 produces a hard cutoff at $\nu_\tau$6, a bump near PeV energies, a dip between about $\nu_\tau$7 and $\nu_\tau$8, and a populated tail at $\nu_\tau$9–$\chi\to\phi_n\phi_n$0. After oscillations, the flavor ratio at Earth is approximately $\chi\to\phi_n\phi_n$1, so the signal is cascade dominated in IceCube topology as well as shaped by QCD and electroweak particle cascades in the decay products (Esmaili et al., 2013).

Subsequent dedicated IceCube searches used six years of track data and two years of cascade data and found no significant excess attributable to decaying dark matter. The resulting lower limits exclude lifetimes shorter than $\chi\to\phi_n\phi_n$2 at $\chi\to\phi_n\phi_n$3 CL for dark matter masses above $\chi\to\phi_n\phi_n$4, with hard channels most strongly constrained (Collaboration et al., 2018). KM3NeT was then projected to improve these tests in the PeV regime because of its Galactic Center visibility and improved cascade angular resolution; combining tracks and cascades, it was expected to produce world-leading limits on the decay lifetime and to test some of the dark matter interpretations of IceCube data (Ng et al., 2020).

A recurring source of confusion is that “cascade” in this literature can refer simultaneously to the dark-sector decay chain, the electroweak or hadronic shower generated by the final state, and the detector topology of the observed neutrino event. All three usages occur in the decaying-dark-matter neutrino literature.

5. Electromagnetic cascades before recombination and precision constraints

Cascade decays that inject energetic photons or $\chi\to\phi_n\phi_n$5 before recombination are constrained by CMB spectral distortions, but only after the full electromagnetic cascade is followed. For a decaying species $\chi\to\phi_n\phi_n$6 with fractional abundance $\chi\to\phi_n\phi_n$7, lifetime $\chi\to\phi_n\phi_n$8, and standard dark matter density $\chi\to\phi_n\phi_n$9, the injected power is

$$\frac{1}{N_\gamma}\frac{dN_\gamma}{dE_\gamma} = \frac{1}{\Gamma}\frac{d\Gamma}{dE_\gamma}.$$0

The channels explicitly studied were $$\frac{1}{N_\gamma}\frac{dN_\gamma}{dE_\gamma} = \frac{1}{\Gamma}\frac{d\Gamma}{dE_\gamma}.$$1 and $$\frac{1}{N_\gamma}\frac{dN_\gamma}{dE_\gamma} = \frac{1}{\Gamma}\frac{d\Gamma}{dE_\gamma}.$$2, with the subsequent electromagnetic cascade evolved exactly rather than collapsed into instantaneous heating (Acharya et al., 2019).

The distinction is important because relativistic electromagnetic cascades generate non-thermal relativistic spectral distortions, or $$\frac{1}{N_\gamma}\frac{dN_\gamma}{dE_\gamma} = \frac{1}{\Gamma}\frac{d\Gamma}{dE_\gamma}.$$3-type distortions, whose shape and amplitude differ from the standard $$\frac{1}{N_\gamma}\frac{dN_\gamma}{dE_\gamma} = \frac{1}{\Gamma}\frac{d\Gamma}{dE_\gamma}.$$4-, $$\frac{1}{N_\gamma}\frac{dN_\gamma}{dE_\gamma} = \frac{1}{\Gamma}\frac{d\Gamma}{dE_\gamma}.$$5-, and $$\frac{1}{N_\gamma}\frac{dN_\gamma}{dE_\gamma} = \frac{1}{\Gamma}\frac{d\Gamma}{dE_\gamma}.$$6-type templates. A significant fraction of the injected energy can be shifted into the high-frequency Wien tail, where COBE/FIRAS is less sensitive. As a result, the usual approximation that all injected energy becomes $$\frac{1}{N_\gamma}\frac{dN_\gamma}{dE_\gamma} = \frac{1}{\Gamma}\frac{d\Gamma}{dE_\gamma}.$$7-type heating can overstate the constraint substantially. For decays at $$\frac{1}{N_\gamma}\frac{dN_\gamma}{dE_\gamma} = \frac{1}{\Gamma}\frac{d\Gamma}{dE_\gamma}.$$8, the full treatment weakens the bound on $$\frac{1}{N_\gamma}\frac{dN_\gamma}{dE_\gamma} = \frac{1}{\Gamma}\frac{d\Gamma}{dE_\gamma}.$$9 by a factor of about $$\epsilon_i=\frac{2m_i}{m_{i+1}},\qquad \epsilon_f=\frac{2m_f}{m_1}.$$0 for $$\epsilon_i=\frac{2m_i}{m_{i+1}},\qquad \epsilon_f=\frac{2m_f}{m_1}.$$1 at $$\epsilon_i=\frac{2m_i}{m_{i+1}},\qquad \epsilon_f=\frac{2m_f}{m_1}.$$2, and by a factor of about $$\epsilon_i=\frac{2m_i}{m_{i+1}},\qquad \epsilon_f=\frac{2m_f}{m_1}.$$3 for $$\epsilon_i=\frac{2m_i}{m_{i+1}},\qquad \epsilon_f=\frac{2m_f}{m_1}.$$4 at $$\epsilon_i=\frac{2m_i}{m_{i+1}},\qquad \epsilon_f=\frac{2m_f}{m_1}.$$5; more generally, the paper states that the relaxation can be as large as a factor of $$\epsilon_i=\frac{2m_i}{m_{i+1}},\qquad \epsilon_f=\frac{2m_f}{m_1}.$$6 (Acharya et al., 2019).

This result has direct relevance for cascade-decaying dark matter more broadly. If a multi-step dark-sector chain ends in electromagnetic injection before recombination, the detailed internal cascade matters only through the final redshift distribution and energy spectrum of SM photons and $$\epsilon_i=\frac{2m_i}{m_{i+1}},\qquad \epsilon_f=\frac{2m_f}{m_1}.$$7. Once those are specified, the observable CMB distortion must be computed from the full electromagnetic cascade, not inferred from a pure-heating ansatz.

6. Extended variants: continuum spectra, instanton-mediated decay, and medium-induced disassembly

The idea of cascade-decaying dark matter extends beyond finite ladders of particles. In gapped continuum dark matter, the dark sector consists of a continuum of states above a mass gap $$\epsilon_i=\frac{2m_i}{m_{i+1}},\qquad \epsilon_f=\frac{2m_f}{m_1}.$$8, with spectral density

$$\epsilon_i=\frac{2m_i}{m_{i+1}},\qquad \epsilon_f=\frac{2m_f}{m_1}.$$9

In the weakly interacting continuum model, dark matter interacts through a $\epsilon_i\ll 1$0-portal, reproduces the observed relic density, avoids direct detection because continuum kinematics render low-energy scattering intrinsically inelastic, and exhibits continuous decays throughout cosmological history as well as cascade decays of continuum states produced at colliders (Csáki et al., 2021). Here the cascade is not a finite chain but a continuous flow in mass space from heavier continuum modes toward the gap.

A different route to metastable decays uses non-perturbative dark dynamics. In the dark-instanton model, a global symmetry protects a TeV-scale dark matter candidate $\epsilon_i\ll 1$1 at the perturbative level, but an $\epsilon_i\ll 1$2 instanton induces an operator violating that symmetry and generating leptophilic decays. The dominant channel is $\epsilon_i\ll 1$3, and the lifetime can naturally be tuned to $\epsilon_i\ll 1$4 by choosing the dark gauge coupling $\epsilon_i\ll 1$5 near $\epsilon_i\ll 1$6 for $\epsilon_i\ll 1$7 and $\epsilon_i\ll 1$8 (Carone et al., 2010). The same construction can be generalized into a genuine dark-sector cascade if some of the heavier $\epsilon_i\ll 1$9 fermions that appear in the instanton vertex are taken lighter than $$\chi \to \phi_n\phi_n \to 2\times \phi_{n-1}\phi_{n-1}\to \cdots \to 2^{n-1}\times\phi_1\phi_1 \to 2^n\times f\bar f,$$00, so that the instanton first produces on-shell dark daughters which then decay to the Standard Model.

The most radical extension is medium-induced. For loosely bound composite dark matter, each constituent–nucleus scatter in the Earth can impart enough energy to dissociate a constituent, so the Earth acts as a cascade medium. The mean free path is

$$\chi \to \phi_n\phi_n \to 2\times \phi_{n-1}\phi_{n-1}\to \cdots \to 2^{n-1}\times\phi_1\phi_1 \to 2^n\times f\bar f,$$01

the distance to the next scatter is sampled as

$$\chi \to \phi_n\phi_n \to 2\times \phi_{n-1}\phi_{n-1}\to \cdots \to 2^{n-1}\times\phi_1\phi_1 \to 2^n\times f\bar f,$$02

and the resulting random walk produces a spread radius

$$\chi \to \phi_n\phi_n \to 2\times \phi_{n-1}\phi_{n-1}\to \cdots \to 2^{n-1}\times\phi_1\phi_1 \to 2^n\times f\bar f,$$03

Underground signatures then include non-collinear multiple scatters in a single detector, parameter-dependent timing separations of multiscatter events, and coincident signals in different laboratories (Boukhtouchen et al., 17 Dec 2025). In this setting, “decay” is not an intrinsic instability but an interaction-triggered disassembly of the dark state.

The literature therefore uses cascade decaying dark matter as a genuinely broad organizing concept. It encompasses finite dark-sector ladders, decay-driven relic-abundance mechanisms, continuum spectra with perpetual intra-sector decays, and composite states whose cascades are activated by ordinary matter. What unifies these otherwise disparate constructions is the replacement of one-step dark-matter phenomenology by hierarchical energy transport through dark-sector intermediates.