betaFeedback

WorksPhysics and Astronomy

Self-interacting dark matter from radiative bound-state-formation feedback

Abstract

Observations of galactic structure motivate sizable dark-matter self-interactions at low velocities, naturally realized through light force mediators. Such mediators can also make inelastic processes very efficient. Unitarity then requires their absorptive contributions to the two-particle self-energy kernel to be resummed, which in turn feeds back into both the elastic and inelastic scattering amplitudes. We study the feedback of radiative bound-state formation on elastic scattering and its implications for self-interacting dark matter. We focus on capture processes through emission of a charged scalar; such transitions are particularly efficient because the scattering and bound states are governed by different effective potentials, a feature also shared by radiative capture through emission of a non-Abelian gauge boson. We show that their feedback can generate sizable elastic scattering relevant for halo dynamics, even when the baseline elastic amplitude vanishes. Notably, the generated elastic cross section can grow toward low velocities, resembling long-range scattering. A companion paper examines the feedback from annihilation processes.

Contents

1Introduction3
2Bound-state formation5
2.1Model5
2.2Unregulated amplitudes and cross sections6
2.3Excited levels and unitarity violation8
2.4Validity of the Coulomb approximation with kinematic cutoff10
3Elastic scattering via bound-state formation11
3.1Unitarization11
3.2The effect of dispersive contributions13
3.3BSF-induced elastic scattering and its velocity dependence14
3.3.1Massless emitted boson, Wℓ=0W_\ell= 0 approximation14
3.3.2Massless emitted boson, including WℓW_\ell16
3.3.3Massive emitted boson19
3.3.4Comparison with nonresonant Yukawa scattering21
3.3.5Comparison near Yukawa antiresonances22
4Self-interacting dark matter phenomenology25
4.1Halo-averaged cross sections and SIDM benchmarks25
4.2Symmetric dark matter: BSF feedback versus Yukawa scattering25
4.3Asymmetric dark matter: BSF-generated self-interactions26
4.4Cosmological and astrophysical considerations28
4.4.1Dark-matter relic density28
4.4.2Other dark-sector relics31
4.4.3Asymmetric DM and dark charge neutrality32
4.4.4Inelastic processes inside halos32
5Conclusions33
ACapture via massive-boson emission: bound-level sum35
References40

Introduction

Galactic structure observations motivate dark matter (DM) self-interactions that are sizable at the low characteristic velocities of dwarf galaxies and subhalos, while remaining sufficiently weak at the larger velocities of galaxy clusters [1, 2, 3, 4, 5, 6, 7, 8, 9]. Such a velocity dependence arises naturally when DM interacts through light force mediators [10, 11, 12, 13]. If the mediator mass is comparable to or smaller than the characteristic momentum transfer, the interaction becomes effectively long-ranged and must generally be treated nonperturbatively. The resulting Coulomb- or Yukawa-like dynamics can produce large low-velocity self-scattering cross sections while remaining sufficiently weak at higher velocities, providing a particle-physics realization of self-interacting dark matter (SIDM).

The presence of light force mediators can also make inelastic processes important. Long-range interactions can strongly enhance annihilation and radiative processes such as bound-state formation (BSF), allowing their partial-wave cross sections to approach the limits imposed by unitarity [14, 15, 16, 17, 18]. In this regime, elastic and inelastic scattering cannot be treated independently. The optical theorem imposes coupled constraints on their amplitudes in each partial wave [18, 19, 20], such that a calculation that resums only the purely elastic interaction kernel is in general insufficient once inelasticity becomes sizable [18, 20, 21]. The absorptive contributions generated by open inelastic channels must also be incorporated into the two-particle self-energy kernel of the incoming state. Their resummation feeds back into both elastic and inelastic scattering and restores the coupled unitarity constraints. Importantly, this feedback need not merely suppress elasticity: when the baseline elastic amplitude is small or absent, inelastic transitions through intermediate states can instead generate a nonzero elastic cross section [18, 20].

Radiative BSF provides a particularly striking realization of these dynamics when the emitted boson carries charge, as can occur in Abelian and non-Abelian theories through the emission of a charged scalar [22, 23, 24] or a non-Abelian gauge boson [25, 26]. In this case, the incoming scattering state and outgoing bound state are governed by different interaction potentials, allowing very rapid capture transitions. Summing capture over the tower of excited bound levels leads to an especially strong low-velocity enhancement: the standard Coulombic calculation exceeds the global inelastic partial-wave unitarity limit at sufficiently low velocities, even for arbitrarily small nonzero couplings [27, 28]. This apparent unitarity violation signals the breakdown of treating BSF as an isolated process. A consistent resummation of the BSF-induced contribution to the incoming-state kernel is then essential [28] and, through the optical theorem, modifies the elastic scattering amplitude as well [18, 20]. A complementary effective-field-theory approach to the unitarization of radiative capture and the associated generation of elastic scattering has recently been developed [29].

In this work, we investigate the feedback of BSF via charged-scalar emission on elastic scattering. We consider scalar DM XX and a light scalar mediator ϱ\varrho, singly and doubly charged under a dark U(1)U(1) symmetry, respectively. We work in the limit in which this symmetry is global, such that no baseline long-range potential acts between same-charge particles, while particle-antiparticle pairs can bind into bound states B(XX†)\mathcal{B}(X X^\dagger) and same-charge particles can undergo the capture process XX→B(XX†)+ϱXX \to B(XX^\dagger) + \varrho [22]. While transitions proceeding via emission of a vector or neutral scalar boson are suppressed by dipole and quadrupole moments, respectively [16, 30, 31, 32, 33], the non-orthogonality of the scattering and bound-state wavefunctions in the charged-scalar process permits an unsuppressed monopole transition, making capture particularly efficient [22, 34]. This provides an additional enhancement on top of that arising from the different interaction potentials governing the incoming and outgoing states. On the other hand, the finite mass of the emitted scalar provides an important countervailing effect by imposing a kinematic cutoff on the accessible bound levels. As the velocity decreases, progressively fewer excited states remain kinematically accessible, eventually curtailing the growth that would otherwise occur for a massless emitted boson. Moreover, the unitarized result is sensitive to the nontrivial analytic structure of the underlying BSF amplitudes in the complex momentum plane [28].

We find that BSF feedback can generate sizable elastic self-scattering even when no baseline elastic potential acts in the incoming channel. The resulting elastic cross section can reach values relevant for halo dynamics in regions where the baseline BSF rate becomes very large. Importantly, it can grow toward low velocities, providing a novel realization of SIDM that can satisfy the elastic self-interaction benchmarks.

The phenomenological implications differ qualitatively between symmetric and asymmetric DM. For symmetric DM, the BSF-induced XXXX scattering supplements the conventional XX†XX^\dagger Yukawa scattering. In the parameter space we examine, however, the latter remains dominant once the phase-space suppression of the feedback contribution due to the finite mediator mass and the effects associated with the analytic structure of the BSF amplitudes are included. An exception occurs near Yukawa antiresonances, where the BSF-induced scattering can dominate; in the parameter space explored, however, these regions do not open additional viable SIDM parameter space. For asymmetric DM, by contrast, only the same-charge population survives, and the baseline long-range elastic interaction vanishes in the limit considered here. BSF feedback can nevertheless generate elastic self-scattering that grows toward low velocities, thereby opening a qualitatively different realization of SIDM. These results complement our companion study of annihilation feedback, where the irreducible annihilation interaction is short-distance and affects elastic scattering appreciably mainly near Sommerfeld resonances and zeros of the baseline elastic amplitude [35].

The remainder of this paper is organized as follows. In § 2, we introduce the dark-sector model and present the baseline BSF amplitudes and cross sections, including the sum over bound levels and the kinematic restrictions associated with the finite mass of the emitted scalar. In § 3, we unitarize the BSF amplitudes and investigate their feedback on elastic scattering, paying particular attention to the generation of elastic scattering in a channel with no baseline elastic potential and to the analytic structure of the capture amplitudes. In § 4, we assess the implications for the SIDM parameter space for symmetric and asymmetric DM. We conclude in § 5.

Bound-state formation

Model

We consider the scalar dark sector introduced in Ref. [22] and studied in our companion work [35], specializing from the outset to the global-U(1)DU(1)_{\mathrm{D}} limit. The gauged theory and possible realizations of this limit are discussed in Refs. [22, 35].

The DM field XX and the mediator ϱ\varrho are complex scalars with charges qX=1q_X = 1 and qϱ=2q_\varrho= 2 under the U(1)DU(1)_{\mathrm{D}} symmetry, respectively [22]. This charge assignment allows the trilinear interaction X2ϱ†X^2\varrho^\dagger. The terms relevant for the dynamics considered here are

L=∣∂μX∣2−mX2∣X∣2+∣∂μϱ∣2−mϱ2∣ϱ∣2−ymX2[X2ϱ†+(X†)2ϱ]+Lquartic.(1)\mathcal{L} = |\partial_\mu X|^2 - m_X^2|X|^2 + |\partial_\mu\varrho|^2 - m_\varrho^2|\varrho|^2 - \frac{y m_X}{2}\left[X^2\varrho^\dagger+ (X^\dagger)^2\varrho\right] + \mathcal{L}_{\mathrm{quartic}}. \tag*{(1)}

We take mϱ≪mXm_\varrho\ll m_X and define

αϱ≡y216π.(2)\alpha_\varrho\equiv\frac{y^2}{16\pi}. \tag*{(2)}

The quartic interactions are collected in

Lquartic=−λX4∣X∣4−λϱ4∣ϱ∣4−λXϱ∣X∣2∣ϱ∣2.(3)\mathcal{L}_{\mathrm{quartic}} = -\frac{\lambda_X}{4}|X|^4 - \frac{\lambda_\varrho}{4}|\varrho|^4 - \lambda_{X\varrho}|X|^2|\varrho|^2. \tag*{(3)}

We assume that the scalar-potential parameters are such that X=ϱ=0X = \varrho= 0 is a stable or cosmologically long-lived vacuum.

Long-range potentials and bound states

After partial-wave projection, the uu-channel ϱ\varrho exchange generates a Yukawa potential between XX and X†X^\dagger [22], while no long-range interaction exists between same-charge particles,

VXX†,ℓ(r)=−(−1)ℓαϱe−mϱrrandVXX(r)=VX†X†(r)=0.(4)V_{XX^\dagger,\ell}(r) = -(-1)^\ell\alpha_\varrho\frac{e^{-m_\varrho r}}{r} \qquad\text{and}\qquad V_{XX}(r) = V_{X^\dagger X^\dagger}(r) = 0. \tag*{(4)}

Quartic interactions and virtual ss-channel ϱ\varrho exchange also generate contact-like contributions to XXXX elastic scattering. Unlike BSF-induced scattering, these contributions are not enhanced at low velocities and can become subdominant in the light-mediator regime considered here, as follows from the results of § 3.3. We neglect them to isolate the effect of BSF feedback on elastic scattering.

The XX†XX^\dagger potential is attractive for even partial waves and repulsive for odd ones. Therefore, it supports XX†XX^\dagger bound states only in the even partial waves, provided that ϱ\varrho is sufficiently light. Despite the absence of a long-range interaction, same-charge pairs can form XX†XX^\dagger bound states through ϱ\varrho emission,

X+X→Bnℓ(XX†)+ϱ,(5)X + X \to B_{n\ell}(XX^\dagger) + \varrho, \tag*{(5)}

and the conjugate process, as shown in Figure 1. Here, n≥1+ℓn \ge1+\ell stands for the principal quantum number characterizing bound states in a central potential. We define

Capture into a bound state via charged-scalar emission

Figure 1. Capture into a bound state via charged-scalar emission, XX→B(XX†)+ϱXX \to\mathcal{B}(XX^\dagger) + \varrho. In the model considered here, the long-range XX†XX^\dagger kernel is generated by uu-channel ϱ\varrho exchange, whereas no baseline long-range kernel is present in the incoming XXXX channel. Nevertheless, BSF generates an absorptive contribution to KXX2PI\mathcal{K}^{2\mathrm{PI}}_{XX}, illustrated in Figure 3, which, upon resummation, modifies the BSF cross sections and induces elastic XXXX scattering.

mT=2mX,μ=mX/2,κ≡μαϱ,κn≡κ/n,k=μvrel,(6)m_T = 2m_X,\qquad\mu= m_X/2,\qquad\kappa\equiv\mu\alpha_\varrho,\qquad\kappa_n \equiv\kappa/n,\qquad\mathbf{k} = \mu\mathbf{v}_{\mathrm{rel}}, \tag*{(6)}

where mTm_T and μ\mu are the total and reduced masses of the interacting particles in the scattering or bound states, κ\kappa is the Bohr momentum, κn\kappa_n is the binding momentum of the nnth Coulombic bound level, and k\mathbf{k} is the momentum of one of the scattering-state particles in the center-of-momentum (CM) frame, with vrel\mathbf{v}_{\mathrm{rel}} being the relative velocity; we will also use the magnitudes k=∣k∣=μvrelk = |\mathbf{k}| = \mu v_{\mathrm{rel}} and vrel=∣vrel∣v_{\mathrm{rel}} = |\mathbf{v}_{\mathrm{rel}}|. The scattering-state kinetic energy in the CM frame and the Coulombic binding energies are

Ek=k22μandEn=−κn22μ,(7)\mathcal{E}_k = \frac{\mathbf{k}^2}{2\mu}\qquad\text{and}\qquad\mathcal{E}_n = -\frac{\kappa_n^2}{2\mu}, \tag*{(7)}

respectively. In the Coulomb approximation of the bound-state potential, the processes (5) are kinematically allowed if mϱ<ωn(k)m_\varrho< \omega_n(k), where ωn\omega_n is the energy radiated in the capture process [15, 16]

ωn(k)≡Ek−En=μ2(vrel2+αϱ2n2).(8)\omega_n(k) \equiv\mathcal{E}_k-\mathcal{E}_n = \frac{\mu}{2}\left(v_{\mathrm{rel}}^2+\frac{\alpha_\varrho^2}{n^2}\right). \tag*{(8)}

Capture processes are typically significant at low velocities, when Ek≲∣En∣\mathcal{E}_k \lesssim|\mathcal{E}_n|, so the kinematic condition becomes mϱ≲∣En∣m_\varrho\lesssim|\mathcal{E}_n|. For perturbative αϱ\alpha_\varrho, this implies

mϱ≲κ22n2μ≪κn2≲κn,(9)m_\varrho\lesssim\frac{\kappa^2}{2n^2\mu} \ll\frac{\kappa}{n^2} \lesssim\frac{\kappa}{n}, \tag*{(9)}

such that the interaction range, mϱ−1m_\varrho^{-1}, exceeds the characteristic bound-state length scales. We therefore use Coulombic bound-state wavefunctions in the BSF amplitudes, while retaining the finite mediator mass in the emission kinematics. We examine the validity of this approximation further in § 2.5.

Next, we present the baseline amplitudes for the BSF processes (5). For convenience, we define the dimensionless parameters

ζ≡κk=αϱvrel,ζn≡κnk=αϱnvrel,ξ≡κmϱ=μαϱmϱ,(10)\zeta\equiv\frac{\kappa}{k} = \frac{\alpha_\varrho}{v_{\mathrm{rel}}},\qquad\zeta_n \equiv\frac{\kappa_n}{k} = \frac{\alpha_\varrho}{n v_{\mathrm{rel}}},\qquad\xi\equiv\frac{\kappa}{m_\varrho} = \frac{\mu\alpha_\varrho}{m_\varrho}, \tag*{(10)}
γ≡mϱ∣En=1∣=2mϱμαϱ2=2αϱξ.(11)\gamma\equiv\frac{m_\varrho}{|\mathcal{E}_{n=1}|} = \frac{2m_\varrho}{\mu\alpha_\varrho^2} = \frac{2}{\alpha_\varrho\xi}. \tag*{(11)}

Unregulated amplitudes and cross sections

The baseline (unregulated) amplitudes for the BSF processes (5), computed using the long-range potentials (4), while neglecting BSF feedback, are [22, 28]

Mnℓ,unregBSF(k)=(CℓXX)1/2αϱ1/2mT2μ2κn3/2rnℓM(ζ),(12)\mathcal{M}_{n\ell,\mathrm{unreg}}^{\mathrm{BSF}}(k) = \left(\mathcal{C}_{\ell}^{XX}\right)^{1/2}\alpha_\varrho^{1/2}\frac{m_T\sqrt{2\mu}}{2\kappa_n^{3/2}}r_{n\ell}^{\mathsf{M}}(\zeta), \tag*{(12)}

where CℓXX\mathcal{C}^{XX}_{\ell} is the symmetry factor for the identical particles in the incoming XXXX state,

CℓXX=2δℓ,even.(13)\mathcal{C}^{XX}_{\ell} = 2\delta_{\ell,\mathrm{even}}. \tag*{(13)}

The monopole transition obeys the selection rule Δℓ=0\Delta\ell= 0, so that the orbital angular momentum of the final bound state coincides with that of the incoming scattering state. The momentum dependence of the amplitudes is encapsulated in the functions

rnℓM(ζ)=iℓ(−1)n−ℓ[22ℓ+3ℓ!(2ℓ+1)!n(n+ℓ)!(n−ℓ−1)!]1/2ζnℓ+4(ζn∓i)2ℓ+4(ζn∓iζn±i)n+12F1(1+ℓ−n,1+ℓ;2ℓ+2;∓4iζn(ζn∓i)2).(14)r^{\mathcal{M}}_{n\ell}(\zeta) = \mathrm{i}^{\ell}(-1)^{n-\ell}\left[\frac{2^{2\ell+3}\ell!}{(2\ell+1)!}\frac{n(n+\ell)!}{(n-\ell-1)!}\right]^{1/2}\frac{\zeta_n^{\ell+4}}{(\zeta_n\mp\mathrm{i})^{2\ell+4}}\left(\frac{\zeta_n\mp\mathrm{i}}{\zeta_n\pm\mathrm{i}}\right)^{n+1} {}_2F_1\left(1+\ell-n,1+\ell;2\ell+2;\mp\frac{4\mathrm{i}\zeta_n}{(\zeta_n\mp\mathrm{i})^2}\right). \tag*{(14)}

The two signs in Eq. (14) arise from a Pfaff transformation of the hypergeometric function 2F1{}_2F_1. The corresponding cross sections are [22, 28]

σnℓ,unregBSF(k)=CℓXX4π(2ℓ+1)k2×Φn(k)×αϱ×rnℓ(ζ),(15)\sigma^{\mathrm{BSF}}_{n\ell,\mathrm{unreg}}(k) = \mathcal{C}^{XX}_{\ell}\frac{4\pi(2\ell+1)}{k^2}\times\Phi_n(k)\times\alpha_{\varrho}\times r_{n\ell}(\zeta), \tag*{(15)}

where Φn(k)\Phi_n(k) is the phase-space suppression factor,

Φn(k)≡1−mϱ2ωn2(k)=1−γ2ζ4(1+ζn2)2.(16)\Phi_n(k)\equiv\sqrt{1-\frac{m_{\varrho}^2}{\omega_n^2(k)}}=\sqrt{1-\frac{\gamma^2\zeta^4}{(1+\zeta_n^2)^2}}. \tag*{(16)}

and

rnℓ(ζ)≡1+ζn2ζn3×∣rnℓM(ζ)∣2=[22ℓ+3ℓ!(2ℓ+1)!]2n(n+ℓ)!(n−ℓ−1)!(1+ζn2)2ℓ+3∣2F1[−n+ℓ+12,n+ℓ+12;ℓ+32;4ζn2(1+ζn2)2]∣2.(17)\begin{aligned} r_{n\ell}(\zeta)\equiv\frac{1+\zeta_n^2}{\zeta_n^3}\times\left|r^{\mathcal{M}}_{n\ell}(\zeta)\right|^2 \\ &=\left[\frac{2^{2\ell+3}\ell!}{(2\ell+1)!}\right]^2\frac{n(n+\ell)!}{(n-\ell-1)!(1+\zeta_n^2)^{2\ell+3}}\left|{}_2F_1\left[\frac{-n+\ell+1}{2},\frac{n+\ell+1}{2};\ell+\frac{3}{2};\frac{4\zeta_n^2}{(1+\zeta_n^2)^2}\right]\right|^2. \tag*{(17)} \end{aligned}

where we used a quadratic transformation for 2F1{}_2F_1 of Eq. (14). The finite mediator mass can impose a kinematic cutoff on the levels that can form radiatively from free pairs at low velocities,

n<ncutBSF(k)≡{+∞,γζ2≤1,ζγζ2−1,γζ2>1.(18)n < n_{\mathrm{cut}}^{\mathrm{BSF}}(k) \equiv\begin{cases} +\infty, & \gamma\zeta^2 \le1, \\ \frac{\zeta}{\sqrt{\gamma\zeta^2 - 1}}, & \gamma\zeta^2 > 1. \end{cases} \tag*{(18)}

When μvrel2/2≥mφ\mu v_{\mathrm{rel}}^2/2 \ge m_\varphi, or equivalently γζ2≤1\gamma\zeta^2 \le1, the kinetic energy alone is sufficient to produce the emitted mediator. At lower velocities, the binding energy must contribute to reaching the emission threshold, restricting capture to sufficiently low-lying levels. In the ζ→∞\zeta\to\infty limit, ncutBSF→γ−1/2n_{\mathrm{cut}}^{\mathrm{BSF}} \to\gamma^{-1/2}. Since n≥1+ℓn \ge1+\ell, radiative capture into a state with angular momentum ℓ\ell remains kinematically possible only if γ<(1+ℓ)−2\gamma< (1+\ell)^{-2}.

Excited levels and unitarity violation

Partial-wave unitarity bounds

Unitarity sets coupled constraints on partial-wave elastic and inelastic cross sections [18, 19]. We define the partial-wave unitarity cross section [18, 20],

σℓU(k)≡Cℓ4π(2ℓ+1)k2,(19)\sigma_\ell^{\mathrm{U}}(k) \equiv\mathcal{C}_\ell\frac{4\pi(2\ell+1)}{k^2}, \tag*{(19)}

where Cℓ=1\mathcal{C}_\ell=1 for distinguishable particles and Cℓ=2\mathcal{C}_\ell=2 for surviving modes of identical particles; modes forbidden by exchange symmetry are omitted. We define the elastic and inclusive inelastic cross sections normalized to σℓU\sigma_\ell^{\mathrm{U}},

xℓelas≡σℓelasσℓU,yℓ≡σℓinelσℓU.(20)x_\ell^{\mathrm{elas}} \equiv\frac{\sigma_\ell^{\mathrm{elas}}}{\sigma_\ell^{\mathrm{U}}}, \qquad y_\ell\equiv\frac{\sigma_\ell^{\mathrm{inel}}}{\sigma_\ell^{\mathrm{U}}}. \tag*{(20)}

The unitarity constraints read [18, 20],

(xℓelas+yℓ)2≤xℓelas⇒xℓelas≤1,yℓ≤1/4.(21)\left(x_\ell^{\mathrm{elas}}+y_\ell\right)^2 \le x_\ell^{\mathrm{elas}} \quad\Rightarrow\quad x_\ell^{\mathrm{elas}} \le1,\quad y_\ell\le1/4. \tag*{(21)}

At the global inelastic maximum, yℓ=1/4y_\ell=1/4, the elastic cross section is fixed to xℓelas=1/4x_\ell^{\mathrm{elas}}=1/4.

In SIDM phenomenology, the transfer and viscosity cross sections are also commonly employed to characterize momentum and heat transport within halos. Unlike the elastic cross section, their partial-wave contributions involve interference between different angular modes. Unitarity implies corresponding term-by-term upper bounds [35],

xℓT≡σℓTσℓU≤1,σℓU,T(k)≡4πk2(ℓ+1),distinguishable particles only,(22)x_\ell^{\mathrm{T}} \equiv\frac{\sigma_\ell^{\mathrm{T}}}{\sigma_\ell^{\mathrm{U}}} \le1,\qquad\sigma_\ell^{\mathrm{U,T}}(k) \equiv\frac{4\pi}{k^2}(\ell+1),\qquad\text{distinguishable particles only,} \tag*{(22)}
xℓV≡σℓVσℓU≤1,σℓU,V(k)≡Cℓ4πk2(ℓ+1)(ℓ+2)2ℓ+3.(23)x_\ell^{\mathrm{V}} \equiv\frac{\sigma_\ell^{\mathrm{V}}}{\sigma_\ell^{\mathrm{U}}} \le1,\qquad\sigma_\ell^{\mathrm{U,V}}(k) \equiv\mathcal{C}_\ell\frac{4\pi}{k^2}\frac{(\ell+1)(\ell+2)}{2\ell+3}. \tag*{(23)}

For identical particles, the transfer cross section coincides with the elastic cross section, σT=σelas\sigma^{\mathrm{T}}=\sigma^{\mathrm{elas}}, so no separate transfer bound is required.

Bound-state tower and low-velocity growth

To obtain the inclusive inelastic cross section for a given partial wave, we must sum capture over the tower of bound levels accessible from the same partial wave,

σℓ,unregBSF(k)≡∑n=1+ℓncutBSF(k)σnℓ,unregBSF(k)=σℓU(k) αϱ rℓ(ζ,γ),(24)\sigma^{\mathrm{BSF}}_{\ell,\mathrm{unreg}}(k) \equiv\sum_{n=1+\ell}^{n_{\mathrm{cut}}^{\mathrm{BSF}}(k)} \sigma^{\mathrm{BSF}}_{n\ell,\mathrm{unreg}}(k) = \sigma^{\mathrm{U}}_{\ell}(k)\,\alpha_\varrho\,r_\ell(\zeta,\gamma), \tag*{(24)}

where1

rℓ(ζ,γ)≡∑n=1+ℓncutBSF(k)Φn(k) rnℓ(ζ).(25)r_\ell(\zeta,\gamma) \equiv\sum_{n=1+\ell}^{n_{\mathrm{cut}}^{\mathrm{BSF}}(k)} \Phi_n(k)\,r_{n\ell}(\zeta). \tag*{(25)}

Consequently,

yℓ,unregBSF=αϱ rℓ(ζ,γ).(26)y^{\mathrm{BSF}}_{\ell,\mathrm{unreg}} = \alpha_\varrho\,r_\ell(\zeta,\gamma). \tag*{(26)}

For emission of a massless mediator, Ref. [28] derived Kramers-like expressions for the sum over excited levels within each partial wave. At sufficiently low velocities,

rℓ(ζ,0)≈4ζ[ln⁡(ζbℓ)+12(1+ℓ)],ζ≫(1+ℓ)e,(27)r_\ell(\zeta,0) \approx4\zeta\left[\ln\left(\frac{\zeta}{b_\ell}\right) + \frac{1}{2(1+\ell)}\right], \qquad\zeta\gg(1+\ell)e, \tag*{(27)}

where bℓ≃(1+ℓ)/eb_\ell\simeq(1+\ell)/e. Thus rℓ(ζ,0)∼4ζln⁡ζr_\ell(\zeta,0) \sim4\zeta\ln\zeta asymptotically. This unbounded growth at large ζ\zeta implies that the unregulated partial-wave BSF cross sections exceed their global inelastic bounds at sufficiently low velocities for any nonzero αϱ\alpha_\varrho [28].

For finite mϱm_\varrho, however, the phase-space cutoff progressively renders the highly excited levels responsible for this growth inaccessible. The unregulated BSF cross section can consequently exceed the global bound only over a finite range of velocities before being curtailed again at lower velocities, as illustrated in Figure 2. In Appendix § 7, we derive analytic approximations for rℓ(ζ,γ)r_\ell(\zeta,\gamma) at finite mϱm_\varrho. Here we estimate when the cutoff curtails the low-velocity growth of the inclusive unregulated BSF cross section. In the limit mϱ→0m_\varrho\to0, Ref. [28] estimated that the maximum principal quantum number required to reproduce the low-velocity growth of yℓ,unregBSFy^{\mathrm{BSF}}_{\ell,\mathrm{unreg}}, for a given ℓ\ell, to fractional precision p is

Figure 2. The unregulated inclusive BSF cross section σℓ,unregBSF\sigma^{\mathrm{BSF}}_{\ell,\mathrm{unreg}} grows at low velocities due to the increasingly important contribution of excited levels. For finite mϱm_{\varrho}, the phase-space cutoff on the excited-state tower eventually curtails this growth. We have used representative values of ℓ\ell, αϱ\alpha_{\varrho}, and γ=2mϱ/(μαϱ2)\gamma=2m_{\varrho}/(\mu\alpha_{\varrho}^{2}), as denoted in the legends.

ngrowthBSF(p)≈ζ(ζ/bℓ)p−1≈ζpln⁡(ζ/bℓ),ζ>bℓ and p≪1.(28)n^{\mathrm{BSF}}_{\mathrm{growth}}(\mathrm{p}) \approx\frac{\zeta}{\sqrt{(\zeta/b_\ell)^{\mathrm{p}}-1}} \approx\frac{\zeta}{\sqrt{\mathrm{p}\ln(\zeta/b_\ell)}}, \qquad\zeta> b_\ell\ \text{and}\ \mathrm{p} \ll1. \tag*{(28)}

Retaining the Coulomb approximation for the capture amplitudes but taking into account the phase-space suppression, we estimate that a finite mϱm_\varrho affects this growth when ncutBSF<ngrowthBSF(p)n^{\mathrm{BSF}}_{\mathrm{cut}} < n^{\mathrm{BSF}}_{\mathrm{growth}}(\mathrm{p}). Combining Eqs. (18) and (28), this condition becomes

ζ≳(γbℓp)−1/(2−p)→p→0ζ≳γ−1/2orvrel≲2mϱμ,(29)\zeta\gtrsim\left(\gamma b_\ell^{\mathrm{p}}\right)^{-1/(2-\mathrm{p})} \xrightarrow{\mathrm{p}\to0} \zeta\gtrsim\gamma^{-1/2} \quad\text{or} \quad v_{\mathrm{rel}} \lesssim\sqrt{\frac{2m_\varrho}{\mu}}, \tag*{(29)}

independently of ℓ\ell.

Validity of the Coulomb approximation with kinematic cutoff

The validity of the Coulomb approximation for the capture amplitudes can be assessed from the characteristic scales of the Coulombic bound states,

⟨∣r∣⟩=12κ[3n2−ℓ(ℓ+1)],⟨r2⟩=n22κ2[5n2+1−3ℓ(ℓ+1)],⟨p2⟩=κ2n2.(30)\langle|r| \rangle= \frac{1}{2\kappa}\left[3n^{2}-\ell(\ell+1)\right], \qquad\langle r^{2} \rangle= \frac{n^{2}}{2\kappa^{2}}\left[5n^{2}+1-3\ell(\ell+1)\right], \qquad\langle p^{2} \rangle= \frac{\kappa^{2}}{n^{2}}. \tag*{(30)}

The Coulomb approximation requires the characteristic bound-state length scale to be within the screening radius, n2/κ≪mϱ−1n^{2}/\kappa\ll m_{\varrho}^{-1} or equivalently ξ≫n2\xi\gg n^{2}.

In our approximation, the levels that contribute significantly to the total capture rate are limited by the kinematic cutoff ncutBSFn_{\mathrm{cut}}^{\mathrm{BSF}} and by the scale ngrowthBSFn_{\mathrm{growth}}^{\mathrm{BSF}}. We therefore expect the Coulomb approximation to be accurate for our purposes provided that

ξ≫min⁡[ncutBSF,ngrowthBSF(p)].(31)\sqrt{\xi} \gg\min\left[n_{\mathrm{cut}}^{\mathrm{BSF}}, n_{\mathrm{growth}}^{\mathrm{BSF}}(p)\right]. \tag*{(31)}

where we take p∼0.1p\sim0.1. In the low-velocity regime, the kinematic cutoff controls the relevant levels, with ncutBSF→γ−1/2=(αϱξ/2)1/2≪ξn_{\mathrm{cut}}^{\mathrm{BSF}}\to\gamma^{-1/2}=(\alpha_{\varrho}\xi/2)^{1/2}\ll\sqrt{\xi}, i.e. the Coulomb approximation is well controlled for perturbative αϱ\alpha_{\varrho}. More broadly, we confirm that Eq. (31) is satisfied in the parameter range relevant for the BSF feedback considered below, although the approximation becomes less accurate when αϱ≳0.2\alpha_{\varrho}\gtrsim0.2 and 2mϱ/μ≲vrel≲10αϱmϱ/μ\sqrt{2m_{\varrho}/\mu}\lesssim v_{\mathrm{rel}}\lesssim\sqrt{10\alpha_{\varrho}m_{\varrho}/\mu}, where ngrowthBSFn_{\mathrm{growth}}^{\mathrm{BSF}} becomes large while remaining below ncutBSF→∞n_{\mathrm{cut}}^{\mathrm{BSF}}\to\infty.

Elastic scattering via bound-state formation

The BSF processes considered in § 2 generate an absorptive contribution to the XXXX self-energy kernel, as illustrated in Figure 3. Resumming it restores consistency with the coupled unitarity constraints of Eq. (21) and modifies both BSF and elastic scattering [18, 20]. We refer to quantities obtained after this resummation as regulated, in contrast to the baseline (unregulated) quantities of § 2. The effect of this resummation on the BSF cross sections was studied in Ref. [28]. Here, we focus on its feedback on elastic scattering. This effect is particularly transparent in the model considered here: the incoming XXXX state has no baseline long-range interaction, and we neglect the contact-like elastic contributions discussed in § 2.1, so its baseline elastic amplitude vanishes. Nevertheless, resumming the BSF-induced kernel generates a nonzero elastic XXXX amplitude. In § 4, we use the resulting elastic scattering cross sections to study the implications for SIDM.

Absorptive contribution to the XX self-energy kernel from BSF

Figure 3. Absorptive contribution to the XXXX self-energy kernel from BSF, XX→B(XX†)+ϱXX \to\mathcal{B}(XX^\dagger) + \varrho. The double line denotes an XX†XX^\dagger bound state, and the arrows on the field lines denote the U(1)DU(1)_D charge flow. The on-shell cut is denoted by the red dashed line.

We focus on the feedback from the absorptive contributions to the optical potential generated by kinematically open BSF channels. The same inelastic vertices can in general also induce Hermitian dispersive contributions to the XXXX optical potential through off-shell intermediate states [20]. Unlike the absorptive contributions, these contributions need not vanish once a BSF channel becomes kinematically closed. They may provide additional contributions to elastic scattering and further modify the BSF amplitudes. We do not include these direct dispersive contributions in the present study, leaving their investigation to future work. These direct Hermitian optical-kernel contributions should be distinguished from the principal-value dispersive contributions generated upon resummation of the absorptive kernel, encoded below in the WℓW_\ell matrix, which we retain.

Unitarization

The general unitarization formalism was developed in Refs. [18, 20, 28] and reviewed in our companion study [35], where unitarization expressions for the transfer and viscosity cross sections employed in SIDM phenomenology were also derived. We specialize from the outset to the present case, for which the baseline XXXX phase shifts vanish, hence

xℓ,unregelas=xℓ,unregt=xℓ,unrege=0.x_{\ell,\mathrm{unreg}}^{\mathrm{elas}} = x_{\ell,\mathrm{unreg}}^{\mathrm{t}} = x_{\ell,\mathrm{unreg}}^{\mathrm{e}} = 0.

It is convenient to define the momentum-rescaled on-shell unregulated BSF amplitudes and collect them into a vector MℓBSF(k)\mathcal{M}^{\mathrm{BSF}}_\ell(k), with components

[MℓBSF(k)]n≡[4kωn(k)Φn(k)CℓXXmT2]1/2Mnℓ,unregBSF(k)=αϱ[Φn(k)1+ζnℓ2ζnℓ3]1/2rnℓM(ζ),(32)[\mathcal{M}^{\mathrm{BSF}}_\ell(k)]_n \equiv\left[\frac{4k\omega_n(k)\Phi_n(k)}{\mathcal{C}_\ell^{XX}m_T^2}\right]^{1/2}\mathcal{M}^{\mathrm{BSF}}_{n\ell,\mathrm{unreg}}(k)=\sqrt{\alpha_\varrho}\left[\Phi_n(k)\frac{1+\zeta_{n\ell}^2}{\zeta_{n\ell}^3}\right]^{1/2}r^\mathcal{M}_{n\ell}(\zeta), \tag*{(32)}

with nn running over the kinematically open bound levels, 1+ℓ≤n<ncutBSF(k)1+\ell\le n<n_{\mathrm{cut}}^{\mathrm{BSF}}(k). The matrices below act on this same set of levels. With this normalization,

Yℓ,unregBSF=(MℓBSF)†MℓBSF,(33)\mathcal{Y}^{\mathrm{BSF}}_{\ell,\mathrm{unreg}}=(\mathcal{M}^{\mathrm{BSF}}_\ell)^\dagger\mathcal{M}^{\mathrm{BSF}}_\ell, \tag*{(33)}

consistent with Eq. (26).

The momentum dependence of the BSF amplitudes renders their resummation ultraviolet convergent, so no renormalization is required [20, 28], in contrast to the contact-like annihilation processes considered in the companion paper [35]. Their nontrivial analytic structure in the complex momentum plane nevertheless generates finite dispersive contributions that affect the resummation. These contributions are encoded in the Hermitian matrix WℓW_\ell, obtained from the principal-value part of the Green’s function of the baseline problem projected onto the irreducible BSF amplitudes. Specializing the general result of Ref. [28] to the present model gives2

Wℓnn′(k)=2αϱπ(nn′)3/2[Φn(k)Φn′(k)(1+ζn2)(1+ζn′2)]1/2P.V.∫0+∞dz rnℓM(z)rn′ℓM∗(z)z2(z2−ζ2).(34)W_\ell^{nn'}(k)=\frac{2\alpha_\varrho}{\pi}(nn')^{3/2}\left[\Phi_n(k)\Phi_{n'}(k)(1+\zeta_n^2)(1+\zeta_{n'}^2)\right]^{1/2}\mathrm{P.V.}\int_0^{+\infty}\mathrm{d}z\,\frac{r^\mathcal{M}_{n\ell}(z)r^{\mathcal{M}*}_{n'\ell}(z)}{z^2(z^2-\zeta^2)}. \tag*{(34)}

Here, we have taken into account that the baseline XXXX Hamiltonian does not support bound states, so only the continuum part of the spectral decomposition of its Green’s function contributes. Although evaluated along the positive real axis, the principal-value integral in Eq. (34) encodes the analytic structure of the BSF amplitudes in the complex zz plane: after analytic continuation, poles and branch cuts of the off-shell amplitudes contribute to WℓW_\ell through their residues and discontinuities [20, 28]. In practice, we evaluate the WℓW_\ell elements numerically along the real axis.

For the purely absorptive separable kernel considered here, the two real quantities that enter the regulated cross sections are [20, 28]

wℓBSF=(MℓBSF)†(1+Wℓ2)−1MℓBSF,(35)w_\ell^{\mathrm{BSF}}=(\mathcal{M}^{\mathrm{BSF}}_\ell)^\dagger(1+W_\ell^2)^{-1}\mathcal{M}^{\mathrm{BSF}}_\ell, \tag*{(35)}
w~ℓBSF=(MℓBSF)†Wℓ(1+Wℓ2)−1MℓBSF.(36)\widetilde{w}_\ell^{\mathrm{BSF}}=(\mathcal{M}^{\mathrm{BSF}}_\ell)^\dagger W_\ell(1+W_\ell^2)^{-1}\mathcal{M}^{\mathrm{BSF}}_\ell. \tag*{(36)}

Since xℓ,unregelas=xℓ,unregT=xℓ,unregV=0x_{\ell,\mathrm{unreg}}^{\mathrm{elas}}=x_{\ell,\mathrm{unreg}}^{\mathrm{T}}=x_{\ell,\mathrm{unreg}}^{\mathrm{V}}=0, the regulated elastic, viscosity, and inclusive BSF cross sections reduce to [35]

xℓ,regelas=(wℓBSF)2+(w~ℓBSF)2(1+wℓBSF)2+(w~ℓBSF)2(37)x_{\ell,\mathrm{reg}}^{\mathrm{elas}}=\frac{\left(w_{\ell}^{\mathrm{BSF}}\right)^2+\left(\widetilde{w}_{\ell}^{\mathrm{BSF}}\right)^2}{\left(1+w_{\ell}^{\mathrm{BSF}}\right)^2+\left(\widetilde{w}_{\ell}^{\mathrm{BSF}}\right)^2} \tag*{(37)}
xℓ,regV=(wℓBSF−wℓ+2BSF)2+(w~ℓBSF−w~ℓ+2BSF)2[(1+wℓBSF)2+(w~ℓBSF)2][(1+wℓ+2BSF)2+(w~ℓ+2BSF)2](38)x_{\ell,\mathrm{reg}}^{\mathrm{V}}=\frac{\left(w_{\ell}^{\mathrm{BSF}}-w_{\ell+2}^{\mathrm{BSF}}\right)^2+\left(\widetilde{w}_{\ell}^{\mathrm{BSF}}-\widetilde{w}_{\ell+2}^{\mathrm{BSF}}\right)^2}{\left[\left(1+w_{\ell}^{\mathrm{BSF}}\right)^2+\left(\widetilde{w}_{\ell}^{\mathrm{BSF}}\right)^2\right]\left[\left(1+w_{\ell+2}^{\mathrm{BSF}}\right)^2+\left(\widetilde{w}_{\ell+2}^{\mathrm{BSF}}\right)^2\right]} \tag*{(38)}
yℓ,regBSF=wℓBSF(1+wℓBSF)2+(w~ℓBSF)2.(39)y_{\ell,\mathrm{reg}}^{\mathrm{BSF}}=\frac{w_{\ell}^{\mathrm{BSF}}}{\left(1+w_{\ell}^{\mathrm{BSF}}\right)^2+\left(\widetilde{w}_{\ell}^{\mathrm{BSF}}\right)^2}. \tag*{(39)}

As mentioned, for identical particles, σT=σelas\sigma^{\mathrm{T}}=\sigma^{\mathrm{elas}}, so no separate expression is required. Thus, BSF feedback generates nonzero elastic scattering despite the absence of a baseline elastic interaction.

If the WℓW_{\ell} corrections associated with the nontrivial analytic structure of the BSF amplitudes are neglected, formally taking Wℓ→0W_{\ell}\to0, then

wℓBSF→yℓ,unregBSF,w~ℓBSF→0.(40)w_{\ell}^{\mathrm{BSF}}\to y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}},\qquad\widetilde{w}_{\ell}^{\mathrm{BSF}}\to0. \tag*{(40)}

Note that w~ℓBSF→0\widetilde{w}_{\ell}^{\mathrm{BSF}}\to0 in this limit because we restrict the separable kernel to its absorptive part; a Hermitian contribution to the kernel would in general yield a non-vanishing w~ℓBSF\widetilde{w}_{\ell}^{\mathrm{BSF}}.

The effect of dispersive contributions

The quadratic-form representation in Eqs. (3.4) allows us to establish useful general bounds on the effect of the dispersive contributions encoded in WℓW_{\ell}. Defining

uℓ=(1+Wℓ2)−1/2MℓBSF,(41)u_{\ell}=\left(\mathbb{1}+W_{\ell}^{2}\right)^{-1/2}M_{\ell}^{\mathrm{BSF}}, \tag*{(41)}

and using Eq. (33), we can write

yℓ,unregBSF=uℓ†(1+Wℓ2)uℓ,wℓBSF=uℓ†uℓ,w~ℓBSF=uℓ†Wℓuℓ.(42)y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}}=u_{\ell}^{\dagger}\left(\mathbb{1}+W_{\ell}^{2}\right)u_{\ell},\qquad w_{\ell}^{\mathrm{BSF}}=u_{\ell}^{\dagger}u_{\ell},\qquad\widetilde{w}_{\ell}^{\mathrm{BSF}}=u_{\ell}^{\dagger}W_{\ell}u_{\ell}. \tag*{(42)}

Thus, yℓ,unregBSF=wℓBSF+∥Wℓuℓ∥2y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}}=w_{\ell}^{\mathrm{BSF}}+\lVert W_{\ell}u_{\ell}\rVert^2. The Cauchy–Schwarz inequality then implies

∣w~ℓBSF∣2=∣uℓ†Wℓuℓ∣2≤(uℓ†uℓ)(uℓ†Wℓ2uℓ)=wℓBSF(yℓ,unregBSF−wℓBSF).(43)\left|\widetilde{w}_{\ell}^{\mathrm{BSF}}\right|^2=\left|u_{\ell}^{\dagger}W_{\ell}u_{\ell}\right|^2\leq\left(u_{\ell}^{\dagger}u_{\ell}\right)\left(u_{\ell}^{\dagger}W_{\ell}^{2}u_{\ell}\right)=w_{\ell}^{\mathrm{BSF}}\left(y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}}-w_{\ell}^{\mathrm{BSF}}\right). \tag*{(43)}

It follows that

(wℓBSF)2+(w~ℓBSF)2≤wℓBSFyℓ,unregBSF,0≤wℓBSF≤yℓ,unregBSF.(44)\left(w_{\ell}^{\mathrm{BSF}}\right)^2+\left(\widetilde{w}_{\ell}^{\mathrm{BSF}}\right)^2\leq w_{\ell}^{\mathrm{BSF}}y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}},\qquad0\leq w_{\ell}^{\mathrm{BSF}}\leq y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}}. \tag*{(44)}

The absence of a baseline XXXX elastic amplitude then allows us to establish a stronger relation between the full result and the Wℓ=0W_{\ell}=0 approximation. Using Eq. (44) in Eq. (3.5a) gives

xℓ,regelas≤wℓBSFyℓ,unregBSF1+2wℓBSF+wℓBSFyℓ,unregBSF≤(yℓ,unregBSF)2(1+yℓ,unregBSF)2=[xℓ,regelas]Wℓ=0.(45)x_{\ell,\mathrm{reg}}^{\mathrm{elas}}\leq\frac{w_{\ell}^{\mathrm{BSF}}y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}}}{1+2w_{\ell}^{\mathrm{BSF}}+w_{\ell}^{\mathrm{BSF}}y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}}}\leq\frac{\left(y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}}\right)^2}{\left(1+y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}}\right)^2}=\left[x_{\ell,\mathrm{reg}}^{\mathrm{elas}}\right]_{W_{\ell}=0}. \tag*{(45)}

Thus, in the present model, the dispersive contributions encoded in WℓW_{\ell} can only reduce or leave unchanged the generated elastic cross section in each partial wave, relative to the Wℓ=0W_{\ell}=0 approximation. This inequality relies on the vanishing baseline elastic amplitude and the purely absorptive separable kernel considered here; it does not generally extend to a nonzero baseline amplitude or additional Hermitian kernel contributions. Even under the present assumptions, no analogous inequality holds for the viscosity cross section, which depends on interference between neighboring partial waves. These distinctions will be important in our discussion of BSF-induced scattering in § 3.3.

BSF-induced elastic scattering and its velocity dependence

Massless emitted boson, Wℓ=0W_{\ell}=0 approximation

To build intuition for the BSF feedback on elastic scattering, we first consider the Wℓ→0W_{\ell}\to0 limit, for which yℓ,regBSF→yℓ,unregBSFy_{\ell,\mathrm{reg}}^{\mathrm{BSF}}\to y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}} and w~ℓBSF→0\widetilde{w}_{\ell}^{\mathrm{BSF}}\to0 according to Eq. (40). As long as the unregulated BSF cross section is far below its unitarity limit, the elastic scattering generated by BSF feedback is quadratically suppressed,

xℓ,regelas∣Wℓ=0≃(yℓ,unregBSF)2,yℓ,regBSF∣Wℓ=0≃yℓ,unregBSF,for yℓ,unregBSF≪1.(46)\left.x_{\ell,\mathrm{reg}}^{\mathrm{elas}}\right|_{W_{\ell}=0}\simeq\left(y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}}\right)^2,\qquad\left.y_{\ell,\mathrm{reg}}^{\mathrm{BSF}}\right|_{W_{\ell}=0}\simeq y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}},\qquad\text{for } y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}}\ll1. \tag*{(46)}

The feedback becomes significant once yℓ,unregBSFy_{\ell,\mathrm{unreg}}^{\mathrm{BSF}} approaches unity. At yℓ,unregBSF=1y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}}=1, the regulated inelastic cross section reaches its global maximum, yℓ,regBSF=1/4y_{\ell,\mathrm{reg}}^{\mathrm{BSF}}=1/4, and the elastic cross section is correspondingly fixed to xℓ,regelas=1/4x_{\ell,\mathrm{reg}}^{\mathrm{elas}}=1/4. For even larger unregulated BSF cross sections,

xℓ,regelas∣Wℓ=0≃1,yℓ,regBSF∣Wℓ=0≃1yℓ,unregBSF,for yℓ,unregBSF≫1.(47)\left.x_{\ell,\mathrm{reg}}^{\mathrm{elas}}\right|_{W_{\ell}=0}\simeq1,\qquad\left.y_{\ell,\mathrm{reg}}^{\mathrm{BSF}}\right|_{W_{\ell}=0}\simeq\frac{1}{y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}}},\qquad\text{for } y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}}\gg1. \tag*{(47)}

The correlation between the unregulated BSF rate and the generated elastic scattering can be seen by comparing Figure 2 and Figure 4 for small mϱm_{\varrho}.

Figure 4. Generated elastic cross section for ℓ=0\ell= 0, as a function of ζ\zeta, for representative values of αϱ\alpha_{\varrho} and γ=2mϱ/(μαϱ2)\gamma= 2m_{\varrho}/(\mu\alpha_{\varrho}^{2}). The full result, including the principal-value contributions encoded in WℓW_{\ell}, is compared with the Wℓ=0W_{\ell}=0 approximation. As established in Eq. (45), the full result cannot exceed the Wℓ=0W_{\ell}=0 result for the vanishing baseline elastic amplitude considered here.

Although mϱ=0m_{\varrho}=0 removes the kinematic cutoff on the partial waves, the partial-wave sums remain convergent at fixed ζ\zeta. Their behavior differs substantially between the large- and low-velocity regimes, which we examine in turn.

Large velocities, αϱ≪vrel≪1\alpha_{\varrho}\ll v_{\mathrm{rel}}\ll1

At ζ≪1\zeta\ll1, the unregulated BSF amplitudes are suppressed, and capture is dominated by the lowest partial waves. Expanding the exact Coulombic expression in Eq. (17), we find, for each fixed bound level,

rnℓ(ζ)∝ζ2ℓ+5,rn0(ζ)≃64ζ5n3.(48)r_{n\ell}(\zeta)\propto\zeta^{2\ell+5},\qquad r_{n0}(\zeta)\simeq\frac{64\zeta^5}{n^3}. \tag*{(48)}

Thus, summing over the bound levels, the ss-wave contribution is

y0,unregBSF≃64ζR(3)αϱζ5≪1,(49)y_{0,\mathrm{unreg}}^{\mathrm{BSF}}\simeq64\zeta_{\mathrm{R}}(3)\alpha_{\varrho}\zeta^5\ll1, \tag*{(49)}

where ζR\zeta_{\mathrm{R}} denotes the Riemann zeta function with ζR(3)≃1.2\zeta_{\mathrm{R}}(3) \simeq1.2. Higher partial waves are suppressed by additional powers of ζ≪1\zeta\ll1, so the ss wave dominates both the total elastic and viscosity cross sections. Using Eq. (46), we obtain

σregelas∣Wℓ=0≃215π[ζR(3)]2αϱ12μ2vrel12,σregv∣Wℓ=0≃2σregelas3∣Wℓ=0,αϱ≪vrel≪1.(50)\left.\sigma_{\mathrm{reg}}^{\mathrm{elas}}\right|_{W_{\ell}=0} \simeq\frac{2^{15}\pi[\zeta_{\mathrm{R}}(3)]^{2}\alpha_{\varrho}^{12}}{\mu^{2}v_{\mathrm{rel}}^{12}}, \qquad\left.\sigma_{\mathrm{reg}}^{\mathrm{v}}\right|_{W_{\ell}=0} \simeq\left.\frac{2\sigma_{\mathrm{reg}}^{\mathrm{elas}}}{3}\right|_{W_{\ell}=0}, \qquad\alpha_{\varrho} \ll v_{\mathrm{rel}} \ll1. \tag*{(50)}

Low velocities, vrel≪αϱv_{\mathrm{rel}} \ll\alpha_{\varrho}

To estimate the angular-momentum range contributing significantly to the partial-wave sums, we consider Eq. (A.6) for γ=0\gamma= 0. At ζ≫1\zeta\gg1, the turning-point behavior of the spherical Bessel functions implies that capture is significant for 1+ℓ≲2ζ1+\ell\lesssim2\zeta, while contributions from partial waves sufficiently above this range are exponentially suppressed.3 Considering these modes, we discern two regimes:

  • For 4αϱ2≪vrel≲αϱ4\alpha_{\varrho}^{2} \ll v_{\mathrm{rel}} \lesssim\alpha_{\varrho} (i.e. ζ≳1\zeta\gtrsim1 and 4αϱζ≪14\alpha_{\varrho}\zeta\ll1), Eqs. (A.18b) and (A.18c) give

yℓ,unregBSF∼4αϱζ≪1y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}} \sim4\alpha_{\varrho}\zeta\ll1, thus Eq. (46) applies, yielding

σregelas∣Wℓ=0∼4πk2(4αϱζ)2∑ℓ≲2ζCℓXX(2ℓ+1)≈4πk2(4αϱζ)2∫02ζdℓ (2ℓ+1)≈28παϱ6μ2vrel6,4αϱ2≪vrel≲αϱ.(51)\begin{aligned} \left.\sigma_{\mathrm{reg}}^{\mathrm{elas}}\right|_{W_{\ell}=0} \sim\frac{4\pi}{k^{2}}(4\alpha_{\varrho}\zeta)^{2}\sum_{\ell\lesssim2\zeta}\mathcal{C}_{\ell}^{XX}(2\ell+1) \approx\frac{4\pi}{k^{2}}(4\alpha_{\varrho}\zeta)^{2}\int_{0}^{2\zeta}\mathrm{d}\ell\,(2\ell+1) \\ &\approx\frac{2^{8}\pi\alpha_{\varrho}^{6}}{\mu^{2}v_{\mathrm{rel}}^{6}},\qquad4\alpha_{\varrho}^{2}\ll v_{\mathrm{rel}}\lesssim\alpha_{\varrho}. \tag*{(51)} \end{aligned}
  • For vrel≲4αϱ2v_{\mathrm{rel}}\lesssim4\alpha_{\varrho}^{2}, the unregulated BSF cross sections satisfy yℓ,unregBSF≫1y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}}\gg1 (cf. Eq. (A.18c)), except near the upper edge of the ℓ\ell range. Equation (47) thus applies, and summing over the saturated partial waves gives

σregelas∣Wℓ=0≈4πk2∑ℓ≲2ζCℓXX(2ℓ+1)≈4πk2∫02ζdℓ (2ℓ+1)≈16παϱ2μ2vrel4,vrel≲4αϱ2.(52)\begin{aligned} \left.\sigma_{\mathrm{reg}}^{\mathrm{elas}}\right|_{W_{\ell}=0}\approx\frac{4\pi}{k^{2}}\sum_{\ell\lesssim2\zeta}\mathcal{C}_{\ell}^{XX}(2\ell+1)\approx\frac{4\pi}{k^{2}}\int_{0}^{2\zeta}\mathrm{d}\ell\,(2\ell+1) \\ &\approx\frac{16\pi\alpha_{\varrho}^{2}}{\mu^{2}v_{\mathrm{rel}}^{4}},\qquad v_{\mathrm{rel}}\lesssim4\alpha_{\varrho}^{2}. \tag*{(52)} \end{aligned}

In both cases, the symmetry factor CℓXX=2δℓ,even\mathcal{C}_{\ell}^{XX}=2\delta_{\ell,\mathrm{even}} compensates, in the continuum approximation, for the spacing between the surviving even partial waves.

The viscosity cross section is instead controlled by the variation of the regulated elastic amplitudes between neighboring partial waves, as seen from Eqs. (3.5b) and (40) for Wℓ→0W_{\ell}\to0,

xℓ,regV∣Wℓ=0=(yℓ,unregBSF1+yℓ,unregBSF−yℓ+2,unregBSF1+yℓ+2,unregBSF)2.(53)\left.x_{\ell,\mathrm{reg}}^{\mathrm{V}}\right|_{W_{\ell}=0}=\left(\frac{y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}}}{1+y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}}}-\frac{y_{\ell+2,\mathrm{unreg}}^{\mathrm{BSF}}}{1+y_{\ell+2,\mathrm{unreg}}^{\mathrm{BSF}}}\right)^{2}. \tag*{(53)}

Although yℓ,unregy_{\ell,\mathrm{unreg}} varies with ℓ\ell throughout the range ℓ≲2ζ\ell\lesssim2\zeta, the differences above are strongly suppressed in the regime vrel≲4αϱ2v_{\mathrm{rel}}\lesssim4\alpha_{\varrho}^{2} where yℓ,unreg≫1y_{\ell,\mathrm{unreg}}\gg1. The viscosity cross section is therefore dominated by the transition region near ℓ≃2ζ\ell\simeq2\zeta, where saturation ceases. While an analytic estimate is possible, we do not pursue it here. Numerically, we find an approximate power-law behavior in the entire low-velocity range,

σregV∣Wℓ=0∝vrel−nwith8/3≲n≲5,vrel≲αϱ,(54)\left.\sigma_{\mathrm{reg}}^{\mathrm{V}}\right|_{W_{\ell}=0}\propto v_{\mathrm{rel}}^{-n}\quad\text{with}\quad8/3\lesssim n\lesssim5,\qquad v_{\mathrm{rel}}\lesssim\alpha_{\varrho}, \tag*{(54)}

with nn being lower for larger values of αϱ\alpha_{\varrho}. Thus, the viscosity cross section retains a strong velocity dependence, comparable to or milder than the vrel−4v_{\mathrm{rel}}^{-4} scaling of the total elastic cross section in the saturated low-velocity regime. There is also a significant numerical suppression of σregV\sigma_{\mathrm{reg}}^{\mathrm{V}} with respect to σregelas\sigma_{\mathrm{reg}}^{\mathrm{elas}}.

Massless emitted boson, including WℓW_{\ell}

We now retain the principal-value contributions encoded in WℓW_{\ell}. Their effect becomes important once the unregulated inclusive BSF cross section approaches the unitarity limit [28]. The full calculation then yields wℓBSF≤yℓ,unregBSFw_{\ell}^{\mathrm{BSF}}\leq y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}}, consistently with Eq. (44).

The off-diagonal elements of WℓW_\ell, which couple different intermediate bound levels, are essential for preventing wℓBSFw_\ell^{\mathrm{BSF}} from following the strong low-velocity growth of yℓ,unregBSFy_{\ell,\mathrm{unreg}}^{\mathrm{BSF}} found in the Wℓ=0W_\ell= 0 approximation [28]. A comparison between the two calculations, albeit for finite mϱm_\varrho, is shown in Figure 4.

Several limits of the regulated elastic cross section in Eq. (3.5a), illustrated in Figure 4 and Figure 5, are useful for interpreting the resulting elastic scattering. For yℓ,unregBSF≪1y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}} \ll1, wℓBSFw_\ell^{\mathrm{BSF}} and ∣w~ℓBSF∣|\widetilde{w}_\ell^{\mathrm{BSF}}| are small, and wℓBSF∼yℓ,unregBSFw_\ell^{\mathrm{BSF}} \sim y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}} while ∣w~ℓBSF∣/wℓBSF|\widetilde{w}_\ell^{\mathrm{BSF}}|/w_\ell^{\mathrm{BSF}} is suppressed by the effective radiative coupling, equal to 4αϱ4\alpha_\varrho in the present model. Then, for small couplings,

Figure 5. Full BSF unitarization, including the principal-value contributions encoded in WℓW_\ell, as a function of ζ\zeta, for representative values of ℓ\ell, αϱ\alpha_\varrho, and γ=2mϱ/(μαϱ2)\gamma=2m_\varrho/(\mu\alpha_\varrho^2). The panels show the generated elastic cross section xℓ,regelasx_{\ell,\mathrm{reg}}^{\mathrm{elas}} (top row) and the regularization parameters wℓBSFw_\ell^{\mathrm{BSF}} (middle row) and ∣w~ℓBSF∣|\widetilde{w}_\ell^{\mathrm{BSF}}| (bottom row). The phase-space cutoff induced by the finite ϱ\varrho mass truncates the tower of kinematically accessible bound levels entering both the absorptive and principal-value contributions and thereby modifies the low-velocity behavior.

xℓ,regelas≃(yℓ,unregBSF)2,yℓ,regBSF≃yℓ,unregBSF,for yℓ,unregBSF≪1,αϱ≪1.(55)x_{\ell,\mathrm{reg}}^{\mathrm{elas}} \simeq\left(y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}}\right)^2,\qquad y_{\ell,\mathrm{reg}}^{\mathrm{BSF}} \simeq y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}},\qquad\text{for } y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}} \ll1,\quad\alpha_\varrho\ll1. \tag*{(55)}

Once yℓ,unregBSF≳1y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}} \gtrsim1, and for sufficiently small αϱ\alpha_\varrho, we find numerically that wℓBSFw_\ell^{\mathrm{BSF}} tends to saturate near unity, and w~ℓBSF\widetilde{w}_\ell^{\mathrm{BSF}} remains coupling-suppressed. Consequently,

xℓ,regelas≃yℓ,regBSF≃1/4,for yℓ,unregBSF≫1,αϱ≪1.(56)x_{\ell,\mathrm{reg}}^{\mathrm{elas}} \simeq y_{\ell,\mathrm{reg}}^{\mathrm{BSF}} \simeq1/4,\qquad\text{for } y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}} \gg1,\quad\alpha_\varrho\ll1. \tag*{(56)}

Equation (56) differs qualitatively from the Wℓ=0W_\ell= 0 approximation (47), which instead gives xℓ,regelas→1x_{\ell,\mathrm{reg}}^{\mathrm{elas}} \to1 and yℓ,regBSF→0y_{\ell,\mathrm{reg}}^{\mathrm{BSF}} \to0 as yℓ,unregBSFy_{\ell,\mathrm{unreg}}^{\mathrm{BSF}} grows. At sufficiently large effective radiative couplings, however, the hierarchy between wℓBSFw_\ell^{\mathrm{BSF}} and w~ℓBSF\widetilde{w}_\ell^{\mathrm{BSF}} changes, and this limiting behavior no longer applies. For αϱ≪1\alpha_\varrho\ll1, the principal-value contributions encoded in WℓW_\ell thus suppress or leave unchanged the elastic scattering in each partial wave relative to the Wℓ=0W_\ell= 0 approximation, without altering its leading velocity scaling in the limits considered above.

At large velocities, Eqs. (46), (50) and (3.19a) then imply

σregelas∼215πζR(3)2αϱ12μ2vrel12,σregV∼2σregelas3,for αϱ≪vrel≪1.(57)\sigma_{\mathrm{reg}}^{\mathrm{elas}} \sim\frac{2^{15}\pi\zeta_R(3)^2\alpha_\varrho^{12}}{\mu^2v_{\mathrm{rel}}^{12}},\qquad\sigma_{\mathrm{reg}}^{V} \sim\frac{2\sigma_{\mathrm{reg}}^{\mathrm{elas}}}{3},\qquad\text{for } \alpha_\varrho\ll v_{\mathrm{rel}} \ll1. \tag*{(57)}

At low velocities, the estimates of Eqs. (3.17a) and (3.17b) and Eqs. (3.19) yield

σregelas∼{28παϱ6μ2vrel6,4αϱ2≪vrel≲αϱ,14σregelas∣Wℓ=0∼4παϱ2μ2vrel4,vrel≲4αϱ2.(58)\sigma_{\mathrm{reg}}^{\mathrm{elas}} \sim \begin{cases} \dfrac{2^8\pi\alpha_\varrho^6}{\mu^2v_{\mathrm{rel}}^6}, & 4\alpha_\varrho^2 \ll v_{\mathrm{rel}} \lesssim\alpha_\varrho, \\ \dfrac{1}{4}\left.\sigma_{\mathrm{reg}}^{\mathrm{elas}}\right|_{W_\ell=0} \sim\dfrac{4\pi\alpha_\varrho^2}{\mu^2v_{\mathrm{rel}}^4}, & v_{\mathrm{rel}} \lesssim4\alpha_\varrho^2. \end{cases} \tag*{(58)}

In this velocity range, the viscosity cross section remains sensitive to the variation of the regulated elastic amplitudes between neighboring partial waves, as seen from Eq. (3.5b). Numerically, we find that the WℓW_\ell contributions weaken these cancellations, allowing the viscosity cross section to exhibit a velocity dependence approaching that of its elastic counterpart,

σregV∝vrel−n,with 7/2≲n≲4,for vrel≲αϱ,(59)\sigma_{\mathrm{reg}}^{V} \propto v_{\mathrm{rel}}^{-n},\qquad\text{with } 7/2 \lesssim n \lesssim4,\qquad\text{for } v_{\mathrm{rel}} \lesssim\alpha_\varrho, \tag*{(59)}

with the variation of nn with αϱ\alpha_\varrho reversed with respect to the Wℓ=0W_\ell= 0 limit: here, nn is larger for larger αϱ\alpha_\varrho. Thus, while the WℓW_\ell contributions necessarily suppress or leave unchanged the total elastic cross section, as established in Eq. (45), their effect on the viscosity cross section can be opposite. Indeed, our numerical results show a significant enhancement of σregV\sigma_{\rm reg}^{\rm V} relative to the Wℓ=0W_\ell=0 approximation in the regime considered.

For sufficiently small but finite mϱm_\varrho, the low-velocity enhancement of the mϱ→0m_\varrho\to0 limit persists over a finite velocity range before the phase-space cutoff on the excited-state tower curtails it, as we discuss next. This is shown in Figure 6.

Figure 6. Total viscosity cross sections per unit DM mass, for αϱ=10−1\alpha_\varrho= 10^{-1} and mX=100m_X = 100 GeV. Left: As a function of vrelv_{\rm rel}, for γ≡2mϱ/(μαϱ2)=10−4,10−2\gamma\equiv2m_\varrho/(\mu\alpha_\varrho^2) = 10^{-4}, 10^{-2}, corresponding to mϱ=25m_\varrho= 25 keV and 2.5 MeV, respectively. Right: As a function of γ=2mϱ/(μαϱ2)\gamma= 2m_\varrho/(\mu\alpha_\varrho^2) for vrel=10,103v_{\rm rel} = 10,10^3 km/s. We compare BSF-induced XXXX scattering, including the WℓW_\ell contributions, with Yukawa XX†XX^\dagger scattering regulated by annihilation feedback [35]. For the BSF-induced contribution, we retain the first 100 allowed even partial waves and verify convergence of the viscosity sum. The BSF-induced contribution grows toward low velocities and smaller mϱm_\varrho, but generally remains subleading to Yukawa scattering, except near antiresonances. The narrowness of these antiresonances makes them difficult to resolve numerically, so only some of the expected regions of BSF dominance are visible.

Massive emitted boson

For finite mϱm_\varrho, only kinematically open BSF channels enter wℓBSFw_\ell^{\rm BSF}, w~ℓBSF\widetilde{w}_\ell^{\rm BSF}, and WℓW_\ell. The phase-space cutoff progressively removes the highly excited levels responsible for the low-velocity enhancement of the absorptive and principal-value contributions, eventually curtailing the BSF-induced elastic scattering, as seen in Figure 4 and Figure 5.

Low-velocity growth

The velocity scalings derived above for a massless emitted boson remain approximately applicable as long as the finite mediator mass does not significantly affect radiative capture. The onset of finite-mass effects can be estimated from the kinematic cutoff on the bound-level sum. For

vrel≳2mφμ,(60)v_{\mathrm{rel}} \gtrsim\sqrt{\frac{2m_{\varphi}}{\mu}}, \tag*{(60)}

we have ncutBSF=∞n_{\mathrm{cut}}^{\mathrm{BSF}} = \infty, so the tower of accessible bound levels is not truncated, although some phase-space suppression may still occur. At lower velocities, the kinematic cutoff progressively removes excited bound levels and restricts the partial waves that can contribute to BSF-induced elastic scattering. After an intermediate transition regime, the scattering becomes ss-wave dominated and approaches a low-momentum plateau, provided that the ground-state capture channel remains open at threshold (γ<1\gamma< 1), as discussed next.

Plateau at very low velocities

At still lower velocities, the cutoff changes the behavior qualitatively, both in the Wℓ→0W_\ell\to0 approximation and when the WℓW_\ell contributions are retained. As shown in Appendix § 7, provided that γ<(1+ℓ)−2\gamma< (1+\ell)^{-2}, which allows for BSF channels to remain kinematically open in the limit k→0k \to0,

yℓ,unregBSF∝k2ℓ+1⇒Eq. xref‘(3.2)‘MℓBSF(k)∝kℓ+1/2.(61)y_{\ell,\mathrm{unreg}}^{\mathrm{BSF}} \propto k^{2\ell+1} \mathrel{\underset{\text{Eq. }{xref}`(3.2)`}{\Rightarrow}} M_{\ell}^{\mathrm{BSF}}(k) \propto k^{\ell+1/2}. \tag*{(61)}

At the same time, the matrix Wℓ\mathbf{W}_{\ell} approaches a finite kk-independent limit (cf. Eq. (34)). It then follows directly from the definitions in Eqs. (3.4) that

wℓBSF, w~ℓBSF∝k2ℓ+1.(62)w_{\ell}^{\mathrm{BSF}},\ \widetilde{w}_{\ell}^{\mathrm{BSF}} \propto k^{2\ell+1}. \tag*{(62)}

Since wℓBSFw_{\ell}^{\mathrm{BSF}} and w~ℓBSF\widetilde{w}_{\ell}^{\mathrm{BSF}} become small, the regulated cross-section formulas in Eqs. (3.5a) and (39) give

xℓ,regelas∝k4ℓ+2,σℓ,regelas∝k4ℓ,(63)x_{\ell,\mathrm{reg}}^{\mathrm{elas}} \propto k^{4\ell+2}, \qquad\sigma_{\ell,\mathrm{reg}}^{\mathrm{elas}} \propto k^{4\ell}, \tag*{(63)}
yℓ,regBSF∝k2ℓ+1,σℓ,regBSFvrel∝k2ℓ.(64)y_{\ell,\mathrm{reg}}^{\mathrm{BSF}} \propto k^{2\ell+1}, \qquad\sigma_{\ell,\mathrm{reg}}^{\mathrm{BSF}}v_{\mathrm{rel}} \propto k^{2\ell}. \tag*{(64)}

The BSF-induced elastic scattering therefore eventually exhibits the usual low-velocity behavior of a short-range interaction: the ss-wave cross section approaches a constant, while higher partial waves are velocity suppressed.

In the smooth approximation of Appendix § 7, the asymptotic regime that gives rise to Eqs. (3.25) corresponds parametrically to ζ≳(γbℓ)−1\zeta\gtrsim(\gamma b_{\ell})^{-1} (region A). For the ss wave, provided γ<1\gamma< 1 so that the n=1n=1 capture channel remains open as k→0k \to0, the onset of the asymptotic plateau is estimated to occur at

vrel≲mφμαφ.(65)v_{\mathrm{rel}} \lesssim\frac{m_{\varphi}}{\mu\alpha_{\varphi}}. \tag*{(65)}

This explains the low-velocity flattening observed in Figure 6. The height of the s-wave plateau can also be estimated. In the vrel→0v_{\mathrm{rel}} \to0 limit, Eq. (A.24) implies

y0,unregBSF≃A0(γ)vrel,A0(γ)=64∑n=11/γn31−γ2n4.(66)y^{\mathrm{BSF}}_{0,\mathrm{unreg}} \simeq A_0(\gamma) v_{\mathrm{rel}}, \qquad A_0(\gamma)=64\sum_{n=1}^{1/\sqrt{\gamma}} n^3\sqrt{1-\gamma^2 n^4}. \tag*{(66)}

For γ≪1\gamma\ll1, approximating the sum by an integral gives A0(γ)≃32/(3γ2)A_0(\gamma) \simeq32/(3\gamma^2).

For fixed γ\gamma, the low-energy W0W_0 matrix scales with αϱ\alpha_\varrho, so that its corrections become negligible at sufficiently small coupling. In this regime, x0elas∼(y0,unregBSF)2x^{\mathrm{elas}}_0 \sim(y^{\mathrm{BSF}}_{0,\mathrm{unreg}})^2, thus

σregelas≃σ0,regelas⟶W0→0σ0XX4πμ2A02(γ)≃213π32μ2γ4=29π32μ2αϱ8mϱ4,very small αϱ.(67)\sigma^{\mathrm{elas}}_{\mathrm{reg}} \simeq\sigma^{\mathrm{elas}}_{0,\mathrm{reg}} \underset{W_0\to0}{\longrightarrow} \sigma^{\mathrm{XX}}_0 \frac{4\pi}{\mu^2} A_0^2(\gamma) \simeq\frac{2^{13}\pi}{3^2\mu^2\gamma^4}=\frac{2^9\pi}{3^2}\frac{\mu^2\alpha_\varrho^8}{m_\varrho^4}, \qquad\text{very small }\alpha_\varrho. \tag*{(67)}

As αϱ\alpha_\varrho increases, the W0W_0 contributions become more important. The plateau remains velocity independent, but its height is reduced relative to the W0=0W_0=0 result, consistently with Eq. (45). Numerically, we find

σregelas≃σ0,regelas∝αϱ2mϱ2,larger αϱ.(68)\sigma^{\mathrm{elas}}_{\mathrm{reg}} \simeq\sigma^{\mathrm{elas}}_{0,\mathrm{reg}} \propto\frac{\alpha_\varrho^2}{m_\varrho^2}, \qquad\text{larger }\alpha_\varrho. \tag*{(68)}

Because higher partial waves are suppressed in this regime, σregv=(2/3)σregelas\sigma^{\mathrm{v}}_{\mathrm{reg}}=(2/3)\sigma^{\mathrm{elas}}_{\mathrm{reg}}.

Comparison with nonresonant Yukawa scattering

Massless mediator. The massless-mediator limit highlights an important distinction between BSF-induced and Coulomb scattering. For BSF-induced scattering, capture is dynamically suppressed for ℓ≳2ζ\ell\gtrsim2\zeta, so the partial-wave sums remain finite at fixed nonzero velocity, as seen in Eqs. (57) and (3.21). No analogous large-ℓ\ell cutoff exists in Coulomb scattering. Introducing ℓmax⁡\ell_{\max} as an infrared regulator, for ℓmax⁡≫ζ≫1\ell_{\max}\gg\zeta\gg1 one finds parametrically

σCoulelas∼ℓmax⁡2μ2vrel2andσCoulv∼αϱ2μ2vrel4ln⁡(ℓmax⁡ζ),(69)\sigma^{\mathrm{elas}}_{\mathrm{Coul}}\sim\frac{\ell_{\max}^2}{\mu^2v_{\mathrm{rel}}^2} \qquad\text{and}\qquad \sigma^{\mathrm{v}}_{\mathrm{Coul}}\sim\frac{\alpha_\varrho^2}{\mu^2v_{\mathrm{rel}}^4}\ln\left(\frac{\ell_{\max}}{\zeta}\right), \tag*{(69)}

up to an oscillatory order-one factor in the elastic cross section. While the BSF-induced partial-wave sums remain finite, the Coulomb cross sections are infrared divergent.

Massive mediator. For finite mϱm_\varrho, the BSF-induced scattering resembles Yukawa scattering more closely: in both cases the viscosity cross section can grow strongly toward low velocities before approaching a plateau. However, the parametric behavior differs.

In the non-perturbative regime, ξ≫1\xi\gg1 or μαϱ≫mϱ\mu\alpha_\varrho\gg m_\varrho, the partial waves contributing significantly to Yukawa scattering may be estimated classically as

ℓYuk∼ℓclas∼kmϱ=μvrelmϱ,μvrel≳mϱ.(70)\ell^{\mathrm{Yuk}}\sim\ell_{\mathrm{clas}}\sim\frac{k}{m_\varrho}=\frac{\mu v_{\mathrm{rel}}}{m_\varrho}, \qquad\mu v_{\mathrm{rel}}\gtrsim m_\varrho. \tag*{(70)}

At vrel≪αϱv_{\mathrm{rel}} \ll\alpha_\varrho, provided that ℓYuk≫ζ\ell^{\mathrm{Yuk}} \gg\zeta, the Coulomb estimates of Eq. (69) apply parametrically, with ℓmax⁡\ell_{\max} replaced by ℓYuk\ell^{\mathrm{Yuk}}. This gives

σYukelas∝1mϱ2andσYukv∝αϱ2μ2vrel4ln⁡(μvrel2mϱαϱ),αϱmϱμ≪vrel≪αϱ.(71)\sigma_{\mathrm{Yuk}}^{\mathrm{elas}} \propto\frac{1}{m_\varrho^2} \quad\text{and}\quad \sigma_{\mathrm{Yuk}}^{\mathrm{v}} \propto\frac{\alpha_\varrho^2}{\mu^2 v_{\mathrm{rel}}^4}\ln\left(\frac{\mu v_{\mathrm{rel}}^2}{m_\varrho\alpha_\varrho}\right), \qquad \sqrt{\frac{\alpha_\varrho m_\varrho}{\mu}} \ll v_{\mathrm{rel}} \ll\alpha_\varrho. \tag*{(71)}

This estimate ceases to be valid at ℓclas≲ζ\ell^{\mathrm{clas}} \lesssim\zeta, or vrel≲αϱmϱ/μv_{\mathrm{rel}} \lesssim\sqrt{\alpha_\varrho m_\varrho/\mu}, where the finite range of the Yukawa interaction increasingly affects the contributing partial waves. In the intermediate regime mϱ/μ≲vrel≲αϱmϱ/μm_\varrho/\mu\lesssim v_{\mathrm{rel}} \lesssim\sqrt{\alpha_\varrho m_\varrho/\mu}, the velocity dependence is generally nontrivial and can exhibit oscillatory features associated with the Yukawa phase shifts. As vrelv_{\mathrm{rel}} decreases below mϱ/μm_\varrho/\mu, higher partial waves become suppressed and, away from resonances, the cross sections approach an ss-wave plateau. Estimating its height from the Coulomb ss-wave cross section at the crossover scale k∼mϱk \sim m_\varrho gives, up to the oscillatory Coulomb phase-shift factor,

σ0elas≃32σ0v∼[σ0v(k=mϱ)]XX†=4πmϱ2,vrel≲mϱμ.(72)\sigma_0^{\mathrm{elas}} \simeq\frac{3}{2}\sigma_0^{\mathrm{v}} \sim\left[\sigma_0^{\mathrm{v}}(k=m_\varrho)\right]_{XX^\dagger} = \frac{4\pi}{m_\varrho^2}, \qquad v_{\mathrm{rel}} \lesssim\frac{m_\varrho}{\mu}. \tag*{(72)}

We summarize the comparison in Table 1. Both Yukawa and BSF-induced scattering can exhibit strong enhancements toward low velocities before approaching a plateau at sufficiently low velocities. Their velocity dependence, however, differs: BSF-induced scattering is more strongly suppressed at large velocities, while its long-range-like enhancement is restricted to a parametrically narrower velocity interval. Moreover, its low-velocity plateau is suppressed by higher powers of αϱ\alpha_\varrho at fixed masses. Consequently, BSF-induced scattering is typically subdominant to Yukawa scattering.

BSF-induced XXXX<br>mϱ≪μαϱ2/2m_{\varrho} \ll\mu\alpha_{\varrho}^{2}/2Yukawa XX†XX^\dagger<br>mϱ≪μαϱm_{\varrho} \ll\mu\alpha_{\varrho}, away from resonances and elastic zeros
Velocity rangeαϱ≪vrel≪1\alpha_{\varrho} \ll v_{\mathrm{rel}} \ll1αϱ≪vrel≪1\alpha_{\varrho} \ll v_{\mathrm{rel}} \ll1
σV\sigma^{\mathrm{V}} scalingαϱ12μ2vrel12\frac{\alpha_{\varrho}^{12}}{\mu^{2}v_{\mathrm{rel}}^{12}}αϱ2μ2vrel4\frac{\alpha_{\varrho}^{2}}{\mu^{2}v_{\mathrm{rel}}^{4}}
Velocity range2mϱμ≪vrel≲αϱ\sqrt{\frac{2m_{\varrho}}{\mu}} \ll v_{\mathrm{rel}} \lesssim\alpha_{\varrho}αϱmϱμ≪vrel≪αϱ\sqrt{\frac{\alpha_{\varrho}m_{\varrho}}{\mu}} \ll v_{\mathrm{rel}} \ll\alpha_{\varrho}
σV\sigma^{\mathrm{V}} scalingvrel−nv_{\mathrm{rel}}^{-n}, 3.5≲n≲43.5 \lesssim n \lesssim4αϱ2μ2vrel4\frac{\alpha_{\varrho}^{2}}{\mu^{2}v_{\mathrm{rel}}^{4}}
Plateau onset, vrelv_{\mathrm{rel}}mϱμαϱ\frac{m_{\varrho}}{\mu\alpha_{\varrho}}mϱμ\frac{m_{\varrho}}{\mu}
Plateau height, σV\sigma^{\mathrm{V}}μ2αϱ8mϱ4\frac{\mu^{2}\alpha_{\varrho}^{8}}{m_{\varrho}^{4}} (smaller αϱ\alpha_{\varrho}) to αϱ2mϱ2\frac{\alpha_{\varrho}^{2}}{m_{\varrho}^{2}} (larger αϱ\alpha_{\varrho})1mϱ2\frac{1}{m_{\varrho}^{2}}

Table 1. Parametric comparison of BSF-induced and nonresonant Yukawa viscosity cross sections. The intermediate scaling for the BSF-induced scattering is inferred numerically, including the WℓW_{\ell} contributions. Logarithmic factors for Yukawa scattering and crossover regions are not shown. The plateau scaling of BSF-induced scattering for larger αϱ\alpha_{\varrho} is also inferred numerically.

We reiterate that these estimates assume a vanishing baseline XXXX potential and a perturbative radiative coupling for BSF-induced scattering, and apply away from Sommerfeld resonances and cross-section zeros for Yukawa scattering. Relaxing these assumptions can alter the relative importance of the two contributions. Of particular interest are Yukawa antiresonances, to which we now turn.

Comparison near Yukawa antiresonances

For an attractive Yukawa potential, the baseline ss-wave elastic cross section exhibits velocity-dependent zeros that approach discrete values of ξ\xi in the zero-momentum limit. These antiresonances arise from the Ramsauer–Townsend effect [36], and their properties and velocity dependence are discussed in more detail in Refs. [35, 37]. Near an antiresonance, the baseline XX†XX^\dagger elastic amplitude is strongly suppressed, while higher partial waves remain negligible at sufficiently low velocities due to the centrifugal barrier. However, inelastic feedback can lift these zeros by regenerating a nonzero elastic amplitude. In the present model, XX†→ϱϱ†XX^\dagger\to\varrho\varrho^\dagger annihilation remains nonzero at the baseline elastic-scattering zeros and thereby lifts the Yukawa antiresonances through unitarization [35].

Nevertheless, sufficiently close to an antiresonance, the BSF-induced XXXX scattering can exceed the annihilation-regulated XX†XX^\dagger Yukawa contribution, as illustrated in Figure 7.

Figure 7. Viscosity cross sections per unit DM mass near an s-wave Yukawa antiresonance, for αϱ=10−1\alpha_{\varrho} = 10^{-1} and mX=100m_X = 100 GeV. Left: Cross sections as a function of mϱm_{\varrho} around the antiresonance, at vrel=10−1v_{\mathrm{rel}} = 10^{-1} km/s. Right: Cross sections as a function of vrelv_{\mathrm{rel}} at the antiresonant value of mϱm_{\varrho}. We compare the unregulated and annihilation-regulated XX†X X^{\dagger} Yukawa cross sections with the BSF-induced XXX X contribution. Near the antiresonance, the BSF-induced scattering can dominate over Yukawa scattering, despite the lifting of the latter’s elastic-scattering zero by annihilation feedback, XX†→ϱϱ†X X^{\dagger} \to\varrho\varrho^{\dagger}. This reverses the hierarchy between Yukawa and BSF-induced scattering found at generic parameter points.

The usual hierarchy between the two contributions is therefore reversed, in contrast to the behavior at generic points in parameter space, such as those shown in Figure 6.

Despite this reversal, antiresonant regions do not typically yield phenomenologically viable SIDM. As the velocity increases away from the antiresonance, the Yukawa contribution recovers and eventually overtakes the BSF-induced one. Consequently, parameter points where BSF-induced scattering provides sizable self-interactions at subhalo or dwarf-galaxy velocities tend to exhibit even larger Yukawa cross sections at cluster velocities, in conflict with cluster-scale constraints.

Self-interacting dark matter phenomenology

We now investigate the implications of the BSF-induced elastic scattering derived in § 3 for SIDM phenomenology. We consider two different possibilities for the DM composition: (i) symmetric DM, consisting of equal populations of XX and X†X^\dagger, where the BSF-induced XXXX scattering contributes alongside the baseline Yukawa scattering of XX†XX^\dagger, and (ii) asymmetric DM [38, 39], consisting only of XX, such that, under the assumptions of § 2, elastic self-scattering is generated entirely by BSF feedback. We focus here on the elastic momentum and energy transport induced by this feedback. BSF itself is an inelastic process and may additionally affect halo evolution through energy transfer and the production of bound states [40]; a treatment of these effects is beyond the scope of this work.

Halo-averaged cross sections and SIDM benchmarks

To identify regions relevant for SIDM phenomenology, we use the representative benchmarks at galaxy-cluster, dwarf-galaxy and subhalo scales, as in the companion study [35],

σˉSIDMVmX≲1 barn GeV−1,at vc∼103 km/s,galaxy-cluster scale,(73)\frac{\bar{\sigma}_{\mathrm{SIDM}}^{\mathrm{V}}}{m_X} \lesssim1\ \mathrm{barn\ GeV}^{-1}, \qquad\text{at } v_c \sim10^{3}\ \mathrm{km/s}, \qquad\text{galaxy-cluster scale,} \tag*{(73)}
σˉSIDMVmX≳1 barn GeV−1,at vc∼10 km/s,dwarf-galaxy scale,(74)\frac{\bar{\sigma}_{\mathrm{SIDM}}^{\mathrm{V}}}{m_X} \gtrsim1\ \mathrm{barn\ GeV}^{-1}, \qquad\text{at } v_c \sim10\ \mathrm{km/s}, \qquad\text{dwarf-galaxy scale,} \tag*{(74)}
σˉSIDMVmX∼102 barn GeV−1,at vc∼1 km/s,subhalo scale,(75)\frac{\bar{\sigma}_{\mathrm{SIDM}}^{\mathrm{V}}}{m_X} \sim10^{2}\ \mathrm{barn\ GeV}^{-1}, \qquad\text{at } v_c \sim1\ \mathrm{km/s}, \qquad\text{subhalo scale,} \tag*{(75)}

where the halo-averaged viscosity cross section is defined via

σˉV≡⟨σVvrel3⟩4⟨vDM2⟩⟨vrel⟩,(76)\bar{\sigma}^{\mathrm{V}} \equiv\frac{\left\langle\sigma^{\mathrm{V}}v_{\mathrm{rel}}^{3}\right\rangle}{4\left\langle v_{\mathrm{DM}}^{2}\right\rangle\left\langle v_{\mathrm{rel}}\right\rangle}, \tag*{(76)}

as detailed in [35, 41]. We evaluate the averages assuming an isotropic Maxwellian velocity distribution truncated at vesc≃2vcv_{\mathrm{esc}} \simeq\sqrt{2}v_c, following the prescription derived in Ref. [35, Sec. 2.5 and App. B]. Here, vc/2v_c/\sqrt{2} is the one-dimensional velocity dispersion of the corresponding untruncated Maxwellian.

Symmetric dark matter: BSF feedback versus Yukawa scattering

For symmetric DM, the halo contains equal populations of XX and X†X^\dagger. Assuming that the two species follow the same velocity distribution and taking into account that σXXV=σX†X†V\sigma_{XX}^{\mathrm{V}}=\sigma_{X^\dagger X^\dagger}^{\mathrm{V}}, the halo-averaged viscosity cross section is

σˉSIDMV=12(σˉXX†V+σˉXXV).(77)\bar{\sigma}_{\mathrm{SIDM}}^{\mathrm{V}}=\frac{1}{2}\left(\bar{\sigma}_{XX^\dagger}^{\mathrm{V}}+\bar{\sigma}_{XX}^{\mathrm{V}}\right). \tag*{(77)}

The two contributions have different origins. The XX†XX^\dagger elastic scattering is generated by the Yukawa interaction, including the feedback from XX†→ϱϱ†XX^\dagger\to\varrho\varrho^\dagger annihilation studied in Ref. [35]. By contrast, the baseline XXXX elastic amplitude vanishes under the assumptions of § 2, and the corresponding scattering is generated entirely by BSF feedback, as studied in § 3.3.

The resulting SIDM parameter space is shown in Figure 8, using the benchmarks of Eqs. (4.1). The shading identifies regions exceeding the cluster-scale benchmark or exhibiting insufficient scattering at dwarf-galaxy velocities. The green contours correspond to the subhalo-scale benchmark for the full symmetric-DM cross section. As anticipated from the comparison in § 3.3, Yukawa scattering generally dominates the elastic transport of symmetric DM. Since the subhalo benchmark probes very low velocities, the green contours exhibit a pronounced resonance and antiresonance structure associated with the nonperturbative XX†XX^\dagger Yukawa interaction.

Figure 8. Halo-averaged viscosity cross section per unit DM mass, σˉSIDMV/mX\bar{\sigma}^{\mathrm{V}}_{\mathrm{SIDM}}/m_X, for symmetric DM. We show the DM mass–coupling plane for fixed mediator masses (upper row) and the DM mass–mediator mass plane for fixed couplings (lower row). The red-shaded regions exceed the cluster-scale bound of Eq. (4.1a), while the blue-shaded regions fall below the dwarf-galaxy benchmark of Eq. (4.1b). The white regions satisfy both criteria. The green contours correspond to the subhalo-scale benchmark of Eq. (4.1c). These results include both Yukawa XX†XX^\dagger scattering and BSF-induced XXXX scattering, with the relative weights appropriate for symmetric DM. The orange contours correspond to the subhalo benchmark (75), but retaining only the BSF-induced scattering. Where the orange and green contours coincide, BSF feedback dominates, curtailing the Yukawa antiresonance pattern. At other antiresonances, annihilation feedback is more important in lifting the suppression of Yukawa scattering [35]. The intricate structure of the green contours at large αϱ\alpha_\varrho and mXm_X is due to the increasing importance of higher partial waves. The gray regions marked “Numerical limit” correspond to parameters for which the large number of contributing bound levels renders the full calculation numerically challenging.

To assess the importance of BSF feedback, we also show in orange the subhalo-scale contours of Eq. (4.1c), obtained by retaining only the BSF-induced XXXX contribution in Eq. (77). Where the orange and green contours coincide, BSF-induced scattering dominates the total, effectively filling the suppression of Yukawa scattering near antiresonances. At other antiresonances, annihilation feedback plays a more important role in lifting the baseline Yukawa elastic-scattering zeros, while the BSF-induced contribution remains subdominant. The impact of annihilation feedback on the symmetric-DM parameter space is examined separately in Ref. [35], Fig. 15, which compares the Yukawa scattering predictions with and without this feedback.

Although BSF-induced scattering can locally dominate near Yukawa antiresonances, these regions do not generally open new parameter space satisfying the SIDM benchmarks. In particular, parameter choices yielding sufficiently large BSF-induced scattering at low velocities tend to produce excessive Yukawa scattering at cluster velocities, as discussed in § 3.6.2. Thus, in the parameter range explored here, BSF feedback modifies the antiresonance structure of symmetric-DM self-scattering without substantially altering its phenomenologically viable parameter space.

Asymmetric dark matter: BSF-generated self-interactions

We now consider the fully asymmetric limit, in which the DM population in the halo consists entirely of XX particles. The XX†XX^\dagger Yukawa-scattering channel is then absent, and the halo-averaged viscosity cross section relevant for elastic transport is

σˉSIDMV=σˉXXV.(78)\bar{\sigma}^{\mathrm{V}}_{\mathrm{SIDM}} = \bar{\sigma}^{\mathrm{V}}_{XX}. \tag*{(78)}

The resulting self-interactions can be strongly velocity dependent, as discussed in § 3.3. The cross section can grow toward lower velocities until the finite mediator mass curtails this enhancement, eventually approaching an ss-wave plateau if capture remains kinematically allowed. The velocity range over which this growth develops determines the relative strength of the interaction in clusters, dwarf galaxies, and smaller subhalos. BSF feedback can therefore generate sizable self-interactions on galactic and subgalactic scales while maintaining sufficiently weak scattering at cluster velocities.

We illustrate these possibilities in Figure 9 and Figure 10, using the cluster-, dwarf-, and subhalo-scale benchmarks of Eqs. (4.1). The panels exhibit three qualitatively distinct

Figure 9. Halo-averaged viscosity cross section per unit DM mass, σ‾SIDMv/mX\overline{\sigma}^{\mathrm{v}}_{\mathrm{SIDM}}/m_X, for asymmetric DM, in the DM mass–coupling plane, for fixed mediator masses as indicated. The elastic scattering is generated entirely by BSF feedback, including the principal-value contributions encoded in WℓW_\ell. Color coding as in Figure 8.

Figure 10. As in Figure 9, in the DM mass-mediator mass plane, for fixed values of the coupling αϱ\alpha_\varrho as indicated.

regimes. In the first, scattering can be sizable at dwarf-galaxy velocities while satisfying the cluster constraint, but falls short of the larger subhalo-scale cross section within this parameter space. In the second, all three benchmarks can be met: the scattering is weak enough at cluster scales, sizable at dwarf scales, and strongly enhanced at subhalo scales. In the third, the enhancement becomes significant only below dwarf-galaxy velocities, allowing strong subhalo-scale scattering to coexist with essentially collisionless DM in dwarf galaxies. Thus, BSF feedback can produce qualitatively distinct self-interaction patterns across halo scales. The contours include the full unitarized result, with principal-value contributions in WℓW_\ell. The “Numerical limit” marks computationally challenging regions due to numerous bound levels, not physical exclusions.

Cosmological and astrophysical considerations

We briefly comment on several cosmological and astrophysical considerations relevant to the scenarios discussed above. Their detailed implications depend on the thermal history of the dark sector and on interactions beyond those required for the scattering dynamics studied here, and are therefore beyond the scope of this work.

Dark-matter relic density

Assuming a standard thermal history in which the dark sector was populated and internally thermalized, the DM abundance is set by thermal freeze-out and the potential presence of an asymmetry. In the present model, depletion can receive important contributions from Sommerfeld-enhanced annihilation and from the exceptionally efficient BSF processes mediated by charged-scalar emission [22, 42]. Including these effects, Ref. [42] finds αφsym≃3×10−6\alpha_\varphi^{\rm sym} \simeq3 \times10^{-6} at mX≃1m_X \simeq1 GeV, assuming a common dark- and visible-sector temperature at DM decoupling. We adopt the numerical results of Ref. [42] for the masses studied there and extrapolate to lower masses using the local power law scaling, αφsym∝mX0.54\alpha_\varphi^{\rm sym} \propto m_X^{0.54}, inferred from these results around mX∼m_X \sim GeV.

As shown in Figure 11, along the symmetric thermal-relic estimate the coupling is too small for BSF feedback to generate appreciable elastic XXXX scattering, although sizable self-interactions may still arise from the baseline XX†XX^\dagger Yukawa interaction for sufficiently small mφm_\varphi. For the same thermal history, asymmetric DM requires more efficient depletion of the symmetric component and hence αφ≳αφsym(mX)\alpha_\varphi\gtrsim\alpha_\varphi^{\rm sym}(m_X), with the primordial XX–X†X^\dagger asymmetry adjusted to reproduce the observed abundance [17]. Increasing αφ\alpha_\varphi suppresses the residual X†X^\dagger population, and sufficiently large couplings can yield effectively fully asymmetric DM [17, 43]. This range overlaps with the parameter space where BSF feedback generates sizable elastic self-interactions, as illustrated in Figure 9 to Figure 11.

Figure 11. Left: Working estimate αϱsym(mX)\alpha_{\varrho}^{\mathrm{sym}}(m_X) for the coupling reproducing the observed symmetric thermal-relic abundance [42]; larger couplings may accommodate asymmetric thermal relics. Colored contours show σˉV/mX=1\bar{\sigma}^{V}/m_X = 1 barn/GeV for XX†X X^\dagger Yukawa scattering and for BSF-induced elastic XXX X scattering, at vc=10v_c = 10 km/s. Right: Same quantities at vc=103v_c = 10^3 km/s. In both panels, the cross section generally grows towards larger couplings and smaller masses, though resonances may introduce non-monotonicity.

The relation between αφ\alpha_\varphi and the relic abundance is cosmology dependent. A different dark-sector temperature at decoupling or subsequent entropy injection would modify the coupling required to reproduce the observed abundance. We therefore treat the self-interaction phenomenology independently of the relic-density requirement, both here and in the companion study [35].

Other dark-sector relics

The cosmological fate of the light dark-sector states depends on the realization of the U(1)DU(1)_D symmetry. We distinguish three possibilities.

Unbroken U(1)DU(1)_D. We require αV≪1\alpha_V \ll1 and, in particular, αV≪αϱ\alpha_V \ll\alpha_\varrho, for VV-mediated interactions to be negligible here. The precise condition depends on screening, e.g. Debye screening in a neutral plasma or stable bound-state formation (cf. § 4.4.3). The massless VV contributes to dark radiation if cosmologically populated, as does ϱ\varrho while relativistic. Planck CMB and BAO measurements constrain additional radiation to approximately ΔNeff≲0.5\Delta N_{\mathrm{eff}} \lesssim0.5 at the 3σ3\sigma level [44]. This can be accommodated if the dark sector decouples sufficiently early from the SM, so that subsequent entropy release in the visible sector, notably across the QCD transition, lowers the dark-sector temperature relative to the visible bath. The symmetric ϱ\varrho population may meanwhile be depleted through ϱϱ†→VV\varrho\varrho^\dagger\to V V, whose cross section scales as αV2/mϱ2\alpha_V^2/m_\varrho^2 and can remain efficient for sufficiently light ϱ\varrho even when αV≪αϱ\alpha_V \ll\alpha_\varrho. Charge neutrality for asymmetric DM is discussed below.

Broken U(1)DU(1)_D. We assume VV to be much heavier than ϱ\varrho and sufficiently heavy that vector-mediated interactions are short-ranged and negligible. Decays such as V→ϱϱ†V \to\varrho\varrho^\dagger remove the massive vector relic. Without lighter charged states, an accidental low-energy global symmetry stabilizes ϱ\varrho. For mϱ≲O(eV)m_\varrho\lesssim O(\mathrm{eV}), a thermal ϱ\varrho population can remain an acceptable subdominant hot relic if early decoupling sufficiently suppresses its temperature. For larger mϱm_\varrho, its abundance and conserved charge may instead be transferred to lighter states. For example, fermions χ\chi with qχ=1q_{\chi}=1 and mχ<mϱ/2m_{\chi}<m_{\varrho}/2 may couple through ϱ†χc‾χ+h.c.\varrho^\dagger\overline{\chi^c}\chi+\mathrm{h.c.}, analogously to Ref. [45], allowing the charge-preserving decay ϱ→χχ\varrho\to\chi\chi. This coupling can be weak enough to leave the self-interaction dynamics considered here unaffected. The resulting χ\chi population is constrained by NeffN_{\mathrm{eff}} while relativistic and by relic-abundance and structure-formation considerations if it later becomes nonrelativistic.

Global U(1)DU(1)_{D}. If U(1)DU(1)_{D} is global from the outset, there is no VV relic or associated dark radiation. As in the broken-gauge realization, ϱ\varrho may remain as a sufficiently light subdominant relic, or its abundance and charge may be transferred to lighter states while preserving the global symmetry.

Asymmetric DM and dark charge neutrality

If the gauged U(1)DU(1)_{D} is unbroken or mildly broken, an asymmetric XX population requires compensating dark charge to maintain gauge neutrality [46]. Minimally, this charge may be carried by ϱ†\varrho^\dagger. Since qX=1q_{X}=1 and qϱ=2q_{\varrho}=2, neutrality requires nϱ†−nϱ=(nX−nX†)/2n_{\varrho^\dagger}-n_{\varrho}=(n_{X}-n_{X^\dagger})/2. Once nonrelativistic, the mass density of this compensating population is suppressed by mϱ/(2mX)m_{\varrho}/(2m_{X}) relative to that of XX and can therefore be negligible.

Oppositely charged dark-sector species may form Coulombic bound states, potentially modifying the self-interactions [46]. Here, the very small αV\alpha_{V} and the potentially small mass of the compensating species imply a low binding scale and tend to suppress dark recombination, suggesting a small composite fraction, though the result depends on the thermal history. If the composite fraction is significant, the resulting multicomponent and composite-DM dynamics would need to be included explicitly in the analysis.

Inelastic processes inside halos

The inelastic processes whose feedback on elastic scattering we study here and in the companion paper [35] may also occur as real transitions inside DM halos.

Both XX†→ϱϱ†XX^\dagger\to\varrho\varrho^\dagger annihilation and radiative BSF can produce indirect-detection signals if a portal to the Standard Model allows the dark-sector radiation to yield visible particles. Bound states formed via XX→B(XX†)+ϱXX\to\mathcal{B}(XX^\dagger)+\varrho are unstable and can subsequently decay into ϱϱ†\varrho\varrho^\dagger. Long-range interactions have already been shown to enable indirect-detection signals in asymmetric DM, often involving BSF, despite suppressed conventional particle-antiparticle annihilation [17, 47, 48, 49, 50]. The efficient BSF processes considered here may provide an additional source of such signals.

If sufficiently efficient, these inelastic processes may deplete DM, potentially excluding parameter regions otherwise compatible with the elastic self-interaction benchmarks. A quantitative assessment lies beyond the scope of the present work and its companion, which focus on the feedback of inelasticity on elastic scattering. Establishing reliable constraints would require a more complete treatment of the inelastic dynamics, including the neglected dispersive optical-potential contributions from off-shell BSF, as well as the absorptive and dispersive contributions from ϱ\varrho bremsstrahlung, along with the evolution of the resulting particle populations. We leave this for future work.

Conclusions

Bound-state formation is a major source of inelasticity in theories with light force mediators. Beyond the direct phenomenological consequences of the inelastic processes themselves, inelasticity necessarily feeds back into elastic scattering, as mandated by the optical theorem. Self-interacting dark matter provides a particularly interesting setting in which this interplay can have important phenomenological consequences. In this work, we have investigated this connection for radiative bound-state formation and studied the elastic scattering generated by its feedback.

We focused on BSF through the emission of a charged boson, for which the interaction governing the incoming scattering state differs from that governing the produced bound state. Such transitions can be exceptionally efficient and arise in both Abelian and non-Abelian theories. In the scenario considered here, BSF proceeds through monopole transitions, which can be even more efficient than the familiar dipole transitions associated with vector emission. These features make this class of processes particularly well suited for generating sizable feedback effects.

We found that BSF necessarily generates an elastic contribution even in the absence of a baseline long-range interaction, as required by unitarity, and that this contribution can become sizable. Its origin differs qualitatively from ordinary long-range elastic scattering: the elastic interaction is generated by an absorptive kernel associated with the inelastic channel, rather than by a soft one-boson-exchange potential acting directly in the incoming state. The long-range dynamics instead enters through the final-state interaction responsible for forming the bound state. Together with the tower of bound levels, this can make capture increasingly efficient toward low velocities and generate a pronounced velocity dependence in the elastic feedback.

We derived analytic estimates for the velocity dependence of BSF-induced elastic scattering and investigated it numerically. For a massless emitted boson, the induced scattering is strongly suppressed at large velocities, but grows rapidly as the velocity decreases. At vrel≪αg2v_{\mathrm{rel}} \ll\alpha_g^2, the total elastic cross section scales as σregelas∝vrel−4\sigma_{\mathrm{reg}}^{\mathrm{elas}} \propto v_{\mathrm{rel}}^{-4}, while the viscosity cross section exhibits a similarly steep velocity dependence. This behavior arises from the increasing number of partial waves contributing to BSF feedback at low velocities, despite the absence of a baseline elastic potential in the incoming state. Unlike Coulomb scattering, however, the partial-wave sums remain convergent at any finite velocity. For a finite mediator mass, the kinematic closure of capture channels into excited bound levels progressively curtails this growth. If ground-state capture remains open at threshold, the induced scattering eventually approaches an ss-wave plateau; otherwise, it vanishes at zero velocity. The finite-mediator behavior is illustrated in Figure 6.

The BSF-induced elastic scattering is shaped by principal-value contributions associated with the analytic structure of the inelastic amplitudes, including the tower of bound-state poles. These dispersive effects, which arise from the resummation of the absorptive BSF kernel, can substantially modify the elastic response. Our calculation does not, however, include the direct Hermitian contributions to the optical potential generated by virtual bound-state–mediator intermediate states. Such contributions can persist even when the corresponding BSF channels are kinematically closed, and may further modify both the BSF amplitudes and the induced elastic scattering. Taken together, these features make BSF feedback qualitatively distinct from that of hard inelastic processes such as annihilation, studied in the companion work [35].

Applied to SIDM, these effects play different phenomenological roles in symmetric and asymmetric scenarios. For symmetric DM, the conventional particle-antiparticle Yukawa interaction typically provides the dominant elastic self-scattering, largely because the finite mediator mass cuts off the low-velocity growth of BSF feedback at higher velocities than that of the Yukawa scattering. The asymmetric scenario instead isolates the more distinctive possibility uncovered in this work: BSF feedback alone can generate elastic self-interactions of phenomenologically relevant magnitude even when the baseline elastic amplitude vanishes. Depending on the parameters, the resulting velocity dependence can satisfy the cluster constraint while producing sizable scattering at dwarf scales, and also reach the much larger cross sections considered at subhalo scales. This is shown in Figure 9 and Figure 10. Our analysis focuses on the elastic transport generated by the feedback of capture into particle-antiparticle bound states; the effects of real capture and the subsequent evolution of the bound states on the halo composition and energy budget lie beyond the scope of this work.

Several extensions are particularly well motivated. A natural next step is to incorporate the omitted Hermitian dispersive contributions into the BSF feedback calculation. Another direction is to consider BSF feedback in perturbative non-Abelian theories, where capture through the emission of a non-Abelian gauge boson may remain operative without the kinematic cutoff encountered here. More generally, realistic theories may contain multiple light mediators and several important sources of inelasticity acting on the same incoming state. It would therefore be interesting to study the feedback dynamics in this broader setting, allowing several mediators to participate and treating soft or ultrasoft processes such as BSF alongside hard processes such as annihilation, in order to determine how their combined contributions regulate the same elastic scattering amplitude.

Acknowledgments

This work was supported by the European Union’s Horizon 2020 research and innovation program under grant agreement No 101002846, ERC CoG CosmoChart. MMF is supported through a FRIPRO grant of the Norwegian Research Council (project ID 353561 ‘DarkTurns’).

Capture via massive-boson emission: bound-level sum

In this appendix, we derive a smooth analytic approximation for the bound-level sum entering the unregulated BSF cross section, and an exact expression for its leading asymptotic low-velocity coefficient within the Coulomb approximation. These results elucidate the behavior of rℓ(ζ,γ)r_{\ell}(\zeta,\gamma) and the effect of the finite-mediator phase-space cutoff. In the present model, only even ℓ\ell contribute, but we keep ℓ\ell general throughout.

General formulas

For a given incoming partial wave ℓ\ell, we recall from Eq. (25) that

rℓ(ζ,γ)=∑ℓ+1≤n<ncutBSF(k)Φn(k) rnℓ(ζ)≈∫ℓ+1ncutBSF(k)dn Φn(k) rnℓ(ζ),(79)r_{\ell}(\zeta,\gamma)=\sum_{\ell+1\leq n<n_{\mathrm{cut}}^{\mathrm{BSF}}(k)}\Phi_n(k)\,r_{n\ell}(\zeta)\approx\int_{\ell+1}^{n_{\mathrm{cut}}^{\mathrm{BSF}}(k)}\mathrm{d}n\,\Phi_n(k)\,r_{n\ell}(\zeta), \tag*{(79)}

where the phase-space suppression is defined in Eq. (16),

Φn(k)=1−[γζ21+ζ2/n2]2,(80)\Phi_n(k)=\sqrt{1-\left[\frac{\gamma\zeta^2}{1+\zeta^2/n^2}\right]^2}, \tag*{(80)}

and the strict kinematic cutoff in Eq. (18),

ncutBSF(k)={+∞,γζ2≤1,ζγζ2−1,γζ2>1.(81)n_{\mathrm{cut}}^{\mathrm{BSF}}(k)= \begin{cases} +\infty, & \gamma\zeta^2\leq1,\\ \frac{\zeta}{\sqrt{\gamma\zeta^2-1}}, & \gamma\zeta^2>1. \end{cases} \tag*{(81)}

In the second step of Eq. (A.1), the approximation of the discrete sum by the continuum integral assumes that many levels contribute; it becomes less accurate when the kinematic cutoff leaves only a few accessible levels. For the asymptotic low-velocity coefficient, we will return to the exact discrete sum.

The function rnℓ(ζ)r_{n\ell}(\zeta) has been defined in Eq. (17). Here, we will use the analytic approximation derived in Ref. [28] and employed in Ref. [42],

rnℓ(ζ)≃26ζ2ζn3(1+ζn2)3∣jℓ(2ζ1+ζn2)∣2,ζn≡ζn.(82)r_{n\ell}(\zeta)\simeq2^6\zeta^2\frac{\zeta_n^3}{(1+\zeta_n^2)^3}\left|j_{\ell}\left(\frac{2\zeta}{1+\zeta_n^2}\right)\right|^2,\qquad\zeta_n\equiv\frac{\zeta}{n}. \tag*{(82)}

This expression is formally obtained for n≫ℓn\gg\ell, although it provides a useful approximation to the envelope of the exact result more broadly [28, 42]. As in the main text, we approximate the attractive bound-state potential by its Coulomb limit, retaining the finite mφm_{\varphi} only in the kinematics of the emitted particle. With

z≡2ζ1+ζ2/n2,(83)z\equiv\frac{2\zeta}{1+\zeta^2/n^2}, \tag*{(83)}

the Jacobian and the factors in Eq. (A.4) combine to give

rℓ(ζ,γ)≃8ζ∫zmin⁡zmax⁡dz zjℓ2(z)1−(γζ2)2(84)r_{\ell}(\zeta,\gamma) \simeq8\zeta\int_{z_{\min}}^{z_{\max}} \mathrm{d}z\,zj_{\ell}^{2}(z)\sqrt{1-\left(\frac{\gamma\zeta}{2}\right)^{2}} \tag*{(84)}

where

zmin⁡=2ζ1+ζ2/(1+ℓ)2andzmax⁡=min⁡(2ζ,2γζ).(85)z_{\min}=\frac{2\zeta}{1+\zeta^{2}/(1+\ell)^{2}} \qquad\text{and}\qquad z_{\max}=\min\left(2\zeta,\frac{2}{\gamma\zeta}\right). \tag*{(85)}

The integral has non-empty support only if zmin⁡<zmax⁡z_{\min}<z_{\max}, equivalently if the lowest level accessible from the ℓ\ellth partial wave, n=1+ℓn=1+\ell, satisfies the phase-space condition. For γ>(1+ℓ)−2\gamma>(1+\ell)^{-2}, capture from the ℓ\ellth partial wave therefore shuts off completely at

ζ≥1γ−(1+ℓ)−2.(86)\zeta\geq\frac{1}{\sqrt{\gamma-(1+\ell)^{-2}}}. \tag*{(86)}

Smooth approximation

To evaluate Eq. (A.6) analytically, we use the small- and large-argument forms of the spherical Bessel function [28],

jℓ(z)≃{2ℓℓ!(2ℓ+1)!zℓ,0<z<2bℓ,sin⁡(z−ℓπ/2)z,z≥2bℓ,(87)j_{\ell}(z)\simeq \begin{cases} \dfrac{2^{\ell}\ell!}{(2\ell+1)!}z^{\ell}, & 0<z<2b_{\ell},\\[6pt] \dfrac{\sin(z-\ell\pi/2)}{z}, & z\geq2b_{\ell}, \end{cases} \tag*{(87)}

where bℓb_{\ell} is chosen such that the two approximations for jℓ2j_{\ell}^{2} match upon averaging the rapid oscillations in the second regime,

sin⁡2(z−ℓπ/2)⟶12.(88)\sin^{2}(z-\ell\pi/2)\longrightarrow\frac{1}{2}. \tag*{(88)}

This gives

bℓ≡12[(2ℓ+1)!2ℓ+1/2ℓ!]1/(ℓ+1)≃ℓ+1e.(89)b_{\ell}\equiv\frac{1}{2}\left[\frac{(2\ell+1)!}{2^{\ell+1/2}\ell!}\right]^{1/(\ell+1)} \simeq\frac{\ell+1}{e}. \tag*{(89)}

The averaging in Eq. (A.10) removes the rapid oscillations retained in Eq. (A.4) and yields a smooth approximation to the envelope of the level sum.4

Provided that zmin⁡<zmax⁡z_{\min}<z_{\max}, three regions arise,

region A:zmax⁡≤2bℓ⇒ζ≤bℓ or ζ≥(γbℓ)−1,region B:2bℓ≤zmin⁡⇒ζ−≤ζ≤ζ+,region C:zmin⁡<2bℓ<zmax⁡⇒bℓ<ζ<ζ− or ζ+<ζ<(γbℓ)−1.(90)\begin{aligned} \text{region A:}\quad&z_{\max}\leq2b_{\ell} &&\Rightarrow\quad\zeta\leq b_{\ell}\ \text{or}\ \zeta\geq(\gamma b_{\ell})^{-1},\\ \text{region B:}\quad&2b_{\ell}\leq z_{\min} &&\Rightarrow\quad\zeta_{-}\leq\zeta\leq\zeta_{+},\\ \text{region C:}\quad&z_{\min}<2b_{\ell}<z_{\max} &&\Rightarrow\quad b_{\ell}<\zeta<\zeta_{-}\ \text{or}\ \zeta_{+}<\zeta<(\gamma b_{\ell})^{-1}. \tag*{(90)} \end{aligned}

where

ζ±=(ℓ+1)22bℓ(1±1−4bℓ2(ℓ+1)2)≃(ℓ+1)e(1±1−4e−22),(91)\zeta_{\pm} = \frac{(\ell+1)^2}{2b_{\ell}}\left(1\pm\sqrt{1-\frac{4b_{\ell}^2}{(\ell+1)^2}}\right)\simeq(\ell+1)e\left(\frac{1\pm\sqrt{1-4e^{-2}}}{2}\right), \tag*{(91)}

or

ζ+≃2.28(1+ℓ)andζ−≃0.44(1+ℓ).(92)\zeta_{+}\simeq2.28(1+\ell)\qquad\text{and}\qquad\zeta_{-}\simeq0.44(1+\ell). \tag*{(92)}

It is convenient to introduce

umin⁡≡γζ2zmin⁡=γζ21+ζ2/(1+ℓ)2,(93)u_{\min}\equiv\frac{\gamma\zeta}{2}\qquad z_{\min}=\frac{\gamma\zeta^2}{1+\zeta^2/(1+\ell)^2}, \tag*{(93)}
umax⁡≡γζ2zmax⁡=min⁡(γζ2,1),(94)u_{\max}\equiv\frac{\gamma\zeta}{2}\qquad z_{\max}=\min\left(\gamma\zeta^2,1\right), \tag*{(94)}
ub=γζbℓ.(95)u_b=\gamma\zeta b_{\ell}. \tag*{(95)}

For γ>0\gamma>0 and within the kinematically allowed region, 0<umin⁡<umax⁡≤10<u_{\min}<u_{\max}\leq1. We further define

Fℓ(u)≡2F1(−12,1+ℓ;2+ℓ;u2),(96)F_{\ell}(u)\equiv{}_2F_1\left(-\frac{1}{2},1+\ell;2+\ell;u^2\right), \tag*{(96)}
G(u)≡arctanh⁡(1−u2)−1−u2.(97)G(u)\equiv\operatorname{arctanh}\left(\sqrt{1-u^2}\right)-\sqrt{1-u^2}. \tag*{(97)}

The resulting smooth approximations are

rℓA(ζ,γ)≃4ζ2ℓ+2[(zmax⁡2bℓ)2ℓ+2Fℓ(umax⁡)−(zmin⁡2bℓ)2ℓ+2Fℓ(umin⁡)],(98)r_{\ell}^{A}(\zeta,\gamma)\simeq\frac{4\zeta}{2\ell+2}\left[\left(\frac{z_{\max}}{2b_{\ell}}\right)^{2\ell+2}F_{\ell}(u_{\max})-\left(\frac{z_{\min}}{2b_{\ell}}\right)^{2\ell+2}F_{\ell}(u_{\min})\right], \tag*{(98)}
rℓB(ζ,γ)≃4ζ[G(umin⁡)−G(umax⁡)],(99)r_{\ell}^{B}(\zeta,\gamma)\simeq4\zeta\left[G(u_{\min})-G(u_{\max})\right], \tag*{(99)}
rℓC(ζ,γ)≃4ζ[G(ub)−G(umax⁡)+12ℓ+2[Fℓ(ub)−(zmin⁡2bℓ)2ℓ+2Fℓ(umin⁡)]].(100)r_{\ell}^{C}(\zeta,\gamma)\simeq4\zeta\left[G(u_b)-G(u_{\max})+\frac{1}{2\ell+2}\left[F_{\ell}(u_b)-\left(\frac{z_{\min}}{2b_{\ell}}\right)^{2\ell+2}F_{\ell}(u_{\min})\right]\right]. \tag*{(100)}

These expressions follow from

∫du u2ℓ+11−u2=u2ℓ+22ℓ+2Fℓ(u),(101)\int\mathrm{d}u\,u^{2\ell+1}\sqrt{1-u^2}=\frac{u^{2\ell+2}}{2\ell+2}F_{\ell}(u), \tag*{(101)}
∫du 1−u2u=−G(u).(102)\int\mathrm{d}u\,\frac{\sqrt{1-u^2}}{u}=-G(u). \tag*{(102)}

Equations (A.16) are Kramers-like relations for individual partial waves describing monopole capture into Coulombic bound states. They generalize the result of Ref. [28] to the case of a massive emitted particle.

Massless-emitted-boson limit

The function G(u)G(u) diverges logarithmically as u→0u \to0, although the differences appearing in Eqs. (A.16) have finite limits. It is therefore convenient to state the massless case, γ=0\gamma= 0, separately. Then zmax⁡=2ζz_{\max} = 2\zeta, and

rℓA(ζ,0)≃4ζ2ℓ+2[(zmax⁡2bℓ)2ℓ+2−(zmin⁡2bℓ)2ℓ+2](103)r_{\ell}^{A}(\zeta,0) \simeq\frac{4\zeta}{2\ell+2}\left[\left(\frac{z_{\max}}{2b_{\ell}}\right)^{2\ell+2}-\left(\frac{z_{\min}}{2b_{\ell}}\right)^{2\ell+2}\right] \tag*{(103)}
rℓB(ζ,0)≃4ζln⁡(zmax⁡zmin⁡)(104)r_{\ell}^{B}(\zeta,0) \simeq4\zeta\ln\left(\frac{z_{\max}}{z_{\min}}\right) \tag*{(104)}
rℓC(ζ,0)≃4ζ[ln⁡(zmax⁡2bℓ)+12ℓ+2[1−(zmin⁡2bℓ)2ℓ+2]](105)r_{\ell}^{C}(\zeta,0) \simeq4\zeta\left[\ln\left(\frac{z_{\max}}{2b_{\ell}}\right)+\frac{1}{2\ell+2}\left[1-\left(\frac{z_{\min}}{2b_{\ell}}\right)^{2\ell+2}\right]\right] \tag*{(105)}

These expressions reproduce the result of Ref. [28], Eqs. (49).

Low-velocity limit

For any fixed nonzero γ\gamma, the asymptotic low-velocity behavior differs qualitatively from the massless-emission limit. We first consider 0<γ<(1+ℓ)−20 < \gamma< (1+\ell)^{-2}, for which at least one bound level remains kinematically accessible as ζ→∞\zeta\to\infty. In this limit,

zmin⁡≃2(1+ℓ)2ζ,zmax⁡≃2γζ,(106)z_{\min} \simeq\frac{2(1+\ell)^2}{\zeta}, \qquad z_{\max} \simeq\frac{2}{\gamma\zeta}, \tag*{(106)}

so the smooth approximation eventually lies entirely in region A. To determine not only the velocity dependence but also its coefficient, we return to the discrete sum (the first equality) in Eq. (A.1), rather than approximating it by an integral. For each fixed bound level nn, the exact Coulombic expression in Eq. (17) gives, at ζ→∞\zeta\to\infty,

rnℓ(ζ)≃[22ℓ+3ℓ!(2ℓ+1)!]2n2ℓ+2(n+ℓ)!(n−ℓ−1)!1ζ2ℓ+1,(107)r_{n\ell}(\zeta) \simeq\left[\frac{2^{2\ell+3}\ell!}{(2\ell+1)!}\right]^2\frac{n^{2\ell+2}(n+\ell)!}{(n-\ell-1)!}\frac{1}{\zeta^{2\ell+1}}, \tag*{(107)}

while, for γn2<1\gamma n^2 < 1, the phase-space factor satisfies

Φn(k)⟶1−γ2n4.(108)\Phi_n(k) \longrightarrow\sqrt{1-\gamma^2 n^4}. \tag*{(108)}

Since the number of accessible levels approaches a finite value at fixed γ>0\gamma> 0, the leading asymptotic behavior can be summed term by term:

rℓ(ζ,γ)≃Aℓ(γ)ζ2ℓ+1,ζ→∞,0<γ<(1+ℓ)−2,(109)r_{\ell}(\zeta,\gamma) \simeq\frac{A_{\ell}(\gamma)}{\zeta^{2\ell+1}}, \qquad\zeta\to\infty,\quad0 < \gamma< (1+\ell)^{-2}, \tag*{(109)}
Aℓ(γ)≡[22ℓ+3ℓ!(2ℓ+1)!]2∑ℓ+1≤n<γ−1/2n2ℓ+2(n+ℓ)!(n−ℓ−1)!1−γ2n4.(110)A_{\ell}(\gamma) \equiv\left[\frac{2^{2\ell+3}\ell!}{(2\ell+1)!}\right]^2 \sum_{\ell+1\le n<\gamma^{-1/2}}\frac{n^{2\ell+2}(n+\ell)!}{(n-\ell-1)!}\sqrt{1-\gamma^2n^4}. \tag*{(110)}

The coefficient Aℓ(γ)A_{\ell}(\gamma) is exact at leading order in 1/ζ1/\zeta within the Coulomb and phase-space approximations employed here, even when only a few bound levels remain accessible. The γ→0\gamma\to0 and ζ→∞\zeta\to\infty limits do not commute; in particular, the above expansion should not be applied at γ=0\gamma= 0.

For fixed ℓ\ell and γ→0\gamma\to0, many levels contribute, with the dominant terms occurring at n∼γ−1/2≫ℓn \sim\gamma^{-1/2} \gg\ell. Using (n+ℓ)!/(n−ℓ−1)!≃n2ℓ+1(n+\ell)!/(n-\ell-1)! \simeq n^{2\ell+1} in Eq. (A.22b) and replacing the sum by an integral, we obtain

Aℓ(γ)≃[22ℓ+3ℓ!(2ℓ+1)!]2B(ℓ+1,32)4γ2ℓ+2,γ(1+ℓ)2≪1,(111)A_{\ell}(\gamma) \simeq\left[\frac{2^{2\ell+3}\ell!}{(2\ell+1)!}\right]^2 \frac{B\left(\ell+1,\frac{3}{2}\right)}{4\gamma^{2\ell+2}}, \qquad\gamma(1+\ell)^2 \ll1, \tag*{(111)}

where BB is the Euler beta function. This approximation is not uniform at large ℓ\ell. In particular,

A0(γ)=64∑1≤n<γ−1/2n31−γ2n4≃323γ2,γ≪1,(112)A_0(\gamma)=64\sum_{1\leq n<\gamma^{-1/2}} n^3\sqrt{1-\gamma^2 n^4}\simeq\frac{32}{3\gamma^2}, \qquad\gamma\ll1, \tag*{(112)}

which is the coefficient used for the s-wave plateau in § 3.4.2.

Thus, at fixed γ>0\gamma>0 with open capture channels, the finite mediator mass replaces the massless-emission enhancement by the threshold scaling σℓ,unregBSFvrel∝vrel2ℓ\sigma_{\ell,\mathrm{unreg}}^{\mathrm{BSF}}v_{\mathrm{rel}}\propto v_{\mathrm{rel}}^{2\ell} and suppresses higher partial waves at very low velocities. If γ>(1+ℓ)−2\gamma>(1+\ell)^{-2}, capture from the ℓ\ellth partial wave instead closes completely at the finite value of ζ\zeta given in Eq. (A.8). At the boundary γ=(1+ℓ)−2\gamma=(1+\ell)^{-2}, the n=1+ℓn=1+\ell level remains accessible for every finite ζ\zeta, while higher levels eventually close; its phase-space factor approaches Φℓ+1(k)≃2(ℓ+1)/ζ\Phi_{\ell+1}(k)\simeq\sqrt{2}(\ell+1)/\zeta, yielding the stronger suppression

rℓ(ζ,(1+ℓ)−2)≃24ℓ+62(ℓ!)2(1+ℓ)2ℓ+3(2ℓ+1)!ζ2ℓ+2.(113)r_{\ell}\left(\zeta,(1+\ell)^{-2}\right)\simeq\frac{2^{4\ell+6}\sqrt{2}(\ell!)^2(1+\ell)^{2\ell+3}}{(2\ell+1)!\zeta^{2\ell+2}}. \tag*{(113)}

References

References

  1. [1]D.N. Spergel and P.J. Steinhardt, Observational evidence for selfinteracting cold dark matter, Phys. Rev. Lett. 84 (2000) 3760 [astro-ph/9909386].pubmed.ncbi.nlm.nih.gov/11019199
  2. [2]S. Tulin and H.-B. Yu, Dark Matter Self-interactions and Small Scale Structure, Phys. Rept. 730 (2018) 1 [1705.02358].DOI
  3. [3]J.S. Bullock and M. Boylan-Kolchin, Small-Scale Challenges to the ΛCDM Paradigm, Ann. Rev. Astron. Astrophys. 55 (2017) 343 [1707.04256].DOI
  4. [4]M. Markevitch, A.H. Gonzalez, D. Clowe, A. Vikhlinin, L. David, W. Forman et al., Direct constraints on the dark matter self-interaction cross-section from the merging galaxy cluster 1E0657-56, Astrophys. J. 606 (2004) 819 [astro-ph/0309303].DOI
  5. [5]S.W. Randall, M. Markevitch, D. Clowe, A.H. Gonzalez and M. Bradac, Constraints on the Self-Interaction Cross-Section of Dark Matter from Numerical Simulations of the Merging Galaxy Cluster 1E 0657-56, Astrophys. J. 679 (2008) 1173 [0704.0261].DOI
  6. [6]I. Dutra, P. Natarajan and D. Gilman, Self-interacting Dark Matter, Core Collapse, and the Galaxy–Galaxy Strong-lensing Discrepancy, Astrophys. J. 978 (2025) 38 [2406.17024].DOI
  7. [7]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?, Mon. Not. Roy. Astron. Soc. 540 (2025) 247 [2411.08565].arxiv.org/abs/2411.08565
  8. [8]S. Vegetti, S.D.M. White, J.P. McKean, D.M. Powell, C. Spingola, D. Massari et al., A possible challenge for cold and warm dark matter, Nature Astron. 10 (2026) 440 [2601.02466].DOI
  9. [9]D. Gilman et al., JWST lensed quasar dark matter survey V: Hints of self-interacting dark matter from 29 quadruply imaged quasars, 2609.35974.
  10. [10]J.L. Feng, M. Kaplinghat and H.-B. Yu, Halo Shape and Relic Density Exclusions of Sommerfeld-Enhanced Dark Matter Explanations of Cosmic Ray Excesses, Phys. Rev. Lett. 104 (2010) 151301 [0911.0422].pubmed.ncbi.nlm.nih.gov/20481979
  11. [11]A. Loeb and N. Weiner, Cores in Dwarf Galaxies from Dark Matter with a Yukawa Potential, Phys. Rev. Lett. 106 (2011) 171302 [1011.6374].pubmed.ncbi.nlm.nih.gov/21635025
  12. [12]S. Tulin, H.-B. Yu and K.M. Zurek, Resonant Dark Forces and Small Scale Structure, Phys. Rev. Lett. 110 (2013) 111301 [1210.0900].pubmed.ncbi.nlm.nih.gov/25166522
  13. [13]S. Tulin, H.-B. Yu and K.M. Zurek, Beyond Collisionless Dark Matter: Particle Physics Dynamics for Dark Matter Halo Structure, Phys. Rev. D 87 (2013) 115007 [1302.3898].DOI
  14. [14]B. von Harling and K. Petraki, Bound-state formation for thermal relic dark matter and unitarity, JCAP 12 (2014) 033 [1407.7874].DOI
  15. [15]K. Petraki, M. Postma and M. Wiechers, Dark-matter bound states from Feynman diagrams, JHEP 06 (2015) 128 [1505.00109].DOI
  16. [16]K. Petraki, M. Postma and J. de Vries, Radiative bound-state-formation cross-sections for dark matter interacting via a Yukawa potential, JHEP 04 (2017) 077 [1611.01394].DOI
  17. [17]I. Baldes and K. Petraki, Asymmetric thermal-relic dark matter: Sommerfeld-enhanced freeze-out, annihilation signals and unitarity bounds, JCAP 09 (2017) 028 [1703.00478].DOI
  18. [18]M.M. Flores and K. Petraki, Unitarity in the non-relativistic regime and implications for dark matter, Phys. Lett. B 858 (2024) 139022 [2405.02222].DOI
  19. [19]L. Hui, Unitarity bounds and the cuspy halo problem, Phys.Rev.Lett. 86 (2001) 3467 [astro-ph/0102349].DOI
  20. [20]M.M. Flores and K. Petraki, Unitarizing non-relativistic scattering, JHEP 06 (2026) 148 [2512.02097].DOI
  21. [21]K. Blum, R. Sato and T.R. Slatyer, Self-consistent Calculation of the Sommerfeld Enhancement, JCAP 06 (2016) 021 [1603.01383].DOI
  22. [22]R. Oncala and K. Petraki, Dark matter bound state formation via emission of a charged scalar, JHEP 02 (2020) 036 [1911.02605].DOI
  23. [23]R. Oncala and K. Petraki, Bound states of WIMP dark matter in Higgs-portal models. Part I. Cross-sections and transition rates, JHEP 06 (2021) 124 [2101.08666].DOI
  24. [24]R. Oncala and K. Petraki, Bound states of WIMP dark matter in Higgs-portal models. Part II. Thermal decoupling, JHEP 08 (2021) 069 [2101.08667].DOI
  25. [25]J. Harz and K. Petraki, Radiative bound-state formation in unbroken perturbative non-Abelian theories and implications for dark matter, JHEP 07 (2018) 096 [1805.01200].DOI
  26. [26]T. Binder, M. Garny, J. Heisig, S. Lederer and K. Urban, Excited bound states and their role in dark matter production, Phys. Rev. D 108 (2023) 095030 [2308.01336].
  27. [27]M. Beneke, T. Binder, L. de Ros, M. Garny and S. Lederer, Perturbative unitarity violation in radiative capture transitions to dark matter bound states, JHEP 02 (2025) 189 [2411.08737].arxiv.org/abs/2411.08737
  28. [28]M.M. Flores and K. Petraki, Unitarity violation and restoration in radiative bound-state formation, JHEP 06 (2026) 231 [2602.20243].DOI
  29. [29]L. De Ros and J. Reig Navarro, A Unitary Fixed Point Away from Threshold: Momentum-Surface EFT for Strong Mixing, 2609.25219.
  30. [30]M. Pospelov and A. Ritz, Astrophysical Signatures of Secluded Dark Matter, Phys. Lett. B 671 (2009) 391 [0810.1502].DOI
  31. [31]J.D. March-Russell and S.M. West, WIMPonium and Boost Factors for Indirect Dark Matter Detection, Phys. Lett. B 676 (2009) 133 [0812.0559].
  32. [32]M.B. Wise and Y. Zhang, Stable Bound States of Asymmetric Dark Matter, Phys. Rev. D 90 (2014) 055030 [1407.4121].DOI
  33. [33]R. Oncala and K. Petraki, Dark matter bound states via emission of scalar mediators, JHEP 01 (2019) 070 [1808.04854].DOI
  34. [34]P. Ko, T. Matsui and Y.-L. Tang, Dark matter bound state formation in fermionic Z₂ DM model with light dark photon and dark Higgs boson, JHEP 10 (2020) 082 [1910.04311].DOI
  35. [35]K. Petraki and T. Pignard, Annihilation feedback on elastic scattering in self-interacting dark matter, to appear (2026) .arxiv.org/abs/2610.27056
  36. [36]J.J. Sakurai and J. Napolitano, Modern quantum mechanics, Cambridge University Press (2020).
  37. [37]M.M. Flores, K. Petraki and T. Pignard, Variable-phase methods and data for long-range scattering, to appear (2026) .arxiv.org/abs/2610.30261
  38. [38]K. Petraki and R.R. Volkas, Review of asymmetric dark matter, Int. J. Mod. Phys. A 28 (2013) 1330028 [1305.4939].DOI
  39. [39]K.M. Zurek, Asymmetric Dark Matter: Theories, Signatures, and Constraints, Phys. Rept. 537 (2014) 91 [1308.0338].DOI
  40. [40]M. Rosenlyst, V. Tran, O. Zier, X. Shen, M. Vogelsberger, T. Brinckmann et al., Self-Interacting Dark Matter: From Particle Physics Theories to Structure Formation Simulations, 2610.16968.arxiv.org/abs/2610.16968
  41. [41]B. Colquhoun, S. Heeba, F. Kahlhoefer, L. Sagunski and S. Tulin, Semiclassical regime for dark matter self-interactions, Phys. Rev. D 103 (2021) 035006 [2011.04679].DOI
  42. [42]K. Petraki, A. Socha and C. Vasilaki, The role of unitarisation on dark-matter freeze-out via metastable bound states, JCAP 09 (2025) 026 [2505.20443].DOI
  43. [43]M.L. Graesser, I.M. Shoemaker and L. Vecchi, Asymmetric WIMP dark matter, JHEP 10 (2011) 110 [1103.2771].DOI
  44. [44]Planck collaboration, Planck 2018 results. VI. Cosmological parameters, Astron. Astrophys. 641 (2020) A6 [1807.06209].
  45. [45]M.M. Flores, K. Petraki and A. Socha, Critical and super-critical scatterings in baryogenesis and leptogenesis, 2507.11532.arxiv.org/abs/2507.11532
  46. [46]K. Petraki, L. Pearce and A. Kusenko, Self-interacting asymmetric dark matter coupled to a light massive dark photon, JCAP 07 (2014) 039 [1403.1077].DOI
  47. [47]L. Pearce and A. Kusenko, Indirect Detection of Self-Interacting Asymmetric Dark Matter, Phys. Rev. D 87 (2013) 123531 [1303.7294].DOI
  48. [48]L. Pearce, K. Petraki and A. Kusenko, Signals from dark atom formation in halos, Phys. Rev. D 91 (2015) 083532 [1502.01755].DOI
  49. [49]I. Baldes, M. Cirelli, P. Panci, K. Petraki, F. Sala and M. Taoso, Asymmetric dark matter: residual annihilations and self-interactions, SciPost Phys. 4 (2018) 041 [1712.07489].
  50. [50]I. Baldes, F. Calore, K. Petraki, V. Poireau and N.L. Rodd, Indirect searches for dark matter bound state formation and level transitions, SciPost Phys. 9 (2020) 068 [2007.13787].DOI

Paper details

vlnc:W_5nxgnakb
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 1

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents