betaFeedback

WorksPhysics and Astronomy

Dark matter evolution in stellar objects: A Boltzmann transport approach

Abstract

Dark matter (DM) captured by stars offers a way to search for physics beyond the Standard Model. We study the evolution of captured DM with the Boltzmann transport equation (BTE), which follows both the abundance and the phase-space distribution of the bound particles from capture toward a stationary state. Averaging over orbital motion, we write the BTE as a master equation in energy and angular momentum that includes capture, scattering between bound orbits, annihilation and evaporation. Applying it to the Sun, we find that captured DM does not thermalize within the solar age when its scattering on nuclei is rare. The bound DM then remains much more extended than the isothermal distribution assumed in the classical treatment, and its annihilation rate can be suppressed by many orders of magnitude. We also apply the BTE to the inelastic Higgsino interpretation of the recent LUX-ZEPLIN result, which involves two DM components coupled through inelastic scattering. The thermal motion of solar nuclei opens capture channels that are closed for nuclei at rest and strengthens the constraint on the mass splitting by tens of keV.

Introduction

Dark matter (DM) accounts for one-fourth of the total energy density in the Universe and plays a central role in the formation of cosmic structure [1]. Despite ample evidence from its gravitational effects, its particle nature and its interactions with ordinary matter remain unknown. Capture of DM in celestial bodies connects direct searches for DM scattering [2] to indirect searches for annihilation or decay products [3]. Annihilation of captured DM can yield signals from the Sun, Earth and other celestial bodies [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22]. DM annihilation may also power the first generation of stars and change the evolution of stars in DM-rich environments [23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37].

Neutron stars (NSs) extend these probes through kinetic and annihilation heating [38, 39, 40, 41, 42, 43, 44, 45, 46] that alters their cooling curves. Asymmetric DM can accumulate without annihilating and trigger gravitational collapse that engulfs the entire NS [47, 48, 49, 50, 51, 52]. White dwarfs (WDs) in environments that may be rich in DM, such as globular clusters and the Galactic center, provide another probe, since capture and annihilation can heat them and their observed luminosities constrain the scattering on nuclei [53, 54, 55, 56]. These studies have been extended to inelastic DM [57] and to the capture and thermalization of heavy particles [58], and annihilation into long-lived mediators that escape WDs can also produce gamma-ray signals [59]. All these signals depend on how much DM is bound and how it is distributed inside the host, so their interpretation requires following the captured population as it evolves.

The classical capture formalism of Refs. [60, 61] describes halo DM scattering into bound orbits, with extensions to multiple scattering [62, 63, 64, 65], more general interactions [66, 67, 68], and different objects and kinematic regimes [69, 70, 71]. The classical treatment evolves the particle number using evaporation and annihilation coefficients evaluated for an isothermal distribution [7, 72, 73, 74, 75], assuming that the captured DM thermalizes quickly with the stellar nuclei. Weak scattering [76, 77] or inelastic thresholds [78, 79, 80] can, however, delay thermalization beyond the stellar age. Since evaporation is sensitive to the weakly bound population and annihilation depends on the spatial overlap of particles, the abundance and the distribution must then be determined together.

Earlier studies have used the Boltzmann transport equation (BTE) and numerical phase-space evolution to investigate selected processes separately, including the redistribution of bound DM among orbits by scattering on nuclei, evaporation and heat transport [80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92]. In this work, we evolve capture, scattering between bound orbits, annihilation and evaporation together. When a bound particle completes many orbits between collisions, the captured population can be described by the number of particles on each orbit, and collisions move particles from one orbit to another. Following this population determines both the abundance and the distribution of captured DM without assuming thermalization, and integrating over orbits recovers the classical number evolution equation with coefficients set by the evolving distribution.

Throughout the paper, we take the Sun, described by the AGSS09 solar model [93, 94], as an example and consider spin-independent (SI) and spin-dependent (SD) DM-nucleon scattering, although the same approach applies to other stellar objects. Since each process enters as a separate term, DM with several components can be treated by adding the terms that couple them, such as scattering between the components or the conversion of one into another [95]. We demonstrate this multicomponent case with the inelastic Higgsino model, which has been proposed as one possible interpretation [96, 97] of the recent LUX-ZEPLIN (LZ) result [98].

The paper is organized as follows. In § 1, we introduce the BTE in position and velocity space, and in Sec. III we describe the orbital phase space, state the orbit-averaging approximation and reconstruct the spatial density. § 1.2 derives the master equation and its connection to the classical number equation, § 1.3 and VI apply the BTE to DM in the Sun and to the inelastic Higgsino model, respectively, and Sec. VII summarizes. Natural units c=ℏ=kB=1c = \hbar= k_B = 1 are used unless specified otherwise.

General formalism

The captured DM population is governed by the BTE,

∂f∂t+v⋅∇rf−∇rΦ⋅∇vf=C[f],(1)\frac{\partial f}{\partial t} + \mathbf{v} \cdot\nabla_{\mathbf{r}} f - \nabla_{\mathbf{r}}\Phi\cdot\nabla_{\mathbf{v}} f = C[f], \tag*{(1)}

where f=f(r,v,t)f = f(\mathbf{r}, \mathbf{v}, t) is the phase-space distribution of gravitationally bound DM, defined such that f d3rd3vf\,\mathrm{d}^{3}\mathbf{r}\mathrm{d}^{3}\mathbf{v} is the number of particles in the phase-space volume d3rd3v\mathrm{d}^{3}\mathbf{r}\mathrm{d}^{3}\mathbf{v}, and the two advection terms describe free motion and gravitational acceleration in the gravitational potential per unit mass, Φ\Phi. We assume a static, spherically symmetric Sun whose nuclei have isotropic thermal velocities. The interaction operator C=Ccap+Ccoll+Cevap+CannC = C_{\mathrm{cap}} + C_{\mathrm{coll}} + C_{\mathrm{evap}} + C_{\mathrm{ann}} contains three contributions from scattering on nuclei and one from annihilation. Scattering on nuclei can capture halo DM onto bound orbits, redistribute bound DM among orbits in phase space, and make bound DM evaporate by giving it enough energy to escape. The spatial number density of bound DM is

nχ(r,t)=∫d3v f(r,v,t),(2)n_{\chi}(\mathbf{r}, t) = \int\mathrm{d}^{3}\mathbf{v}\, f(\mathbf{r}, \mathbf{v}, t), \tag*{(2)}

which yields the total population

Nχ(t)=∫d3r d3v f(r,v,t).(3)N_{\chi}(t) = \int\mathrm{d}^{3}\mathbf{r}\,\mathrm{d}^{3}\mathbf{v}\, f(\mathbf{r}, \mathbf{v}, t). \tag*{(3)}

Bound orbits can extend beyond the solar surface, so nχ(r,t)n_{\chi}(\mathbf{r}, t) need not vanish outside the Sun. We suppress explicit time arguments when they are not needed.

DM-nucleus scattering operators

Scattering on a nucleus changes the velocity of a DM particle at fixed position. A particle with velocity v\mathbf{v} at radius rr scatters to velocity v′\mathbf{v}' with the rate per unit d3v′\mathrm{d}^{3}\mathbf{v}'

R(v→v′,r)=∑AnA(r)∫dΩu∫0∞du u2fA(r,u)σχA4π∫dΩg′ g′FA2(q)δ(3)(v′−v′),(4)R(\mathbf{v} \to\mathbf{v}', r) = \sum_A n_A(r) \int\mathrm{d}\Omega_u \int_0^\infty\mathrm{d}u\,u^2 f_A(r,u)\frac{\sigma_{\chi A}}{4\pi}\int\mathrm{d}\Omega_{g'}\,g'F_A^2(q)\delta^{(3)}(\mathbf{v}'-\mathbf{v}'), \tag*{(4)}

where nA(r)n_A(r) is the number density of nuclear species AA, fA(r,u)f_A(r,u) is the thermal velocity distribution of the nuclei, with u\mathbf{u} the velocity of a nucleus, g=∣v−u∣g = \lvert\mathbf{v} - \mathbf{u} \rvert is the relative speed and v′\mathbf{v}' is the outgoing DM velocity. We take the DM-nucleus cross section σχA\sigma_{\chi A} to be velocity independent, and FA(q)F_A(q) is the nuclear form factor at momentum transfer q=mχ∣v′−v∣q = m_\chi\lvert\mathbf{v}' - \mathbf{v} \rvert. Because the Sun is spherically symmetric, the rate depends on position only through rr. Elastic scattering preserves the relative speed, g′=gg' = g, and the incident velocities together with the outgoing relative direction Ωg′\Omega_{\mathbf{g}}' fix v′\mathbf{v}' through energy and momentum conservation, as given in App. A. The role of a collision depends on whether the initial and final speeds lie below the escape speed vesc(r)v_{\mathrm{esc}}(r) at rr. Scattering between bound states, v<vesc(r)v < v_{\mathrm{esc}}(r) and v′<vesc(r)v' < v_{\mathrm{esc}}(r), redistributes DM in phase space,

Ccoll[f]=∫v′<vesc(r)d3v′[R(v′→v,r)f(r,v′)−R(v→v′,r)f(r,v)],(5)C_{\mathrm{coll}}[f] = \int_{v'<v_{\mathrm{esc}}(r)} d^3\mathbf{v}' \left[R(\mathbf{v}' \to\mathbf{v},r)f(r,\mathbf{v}') - R(\mathbf{v} \to\mathbf{v}',r)f(r,\mathbf{v})\right], \tag*{(5)}

where the two terms describe particles scattered into and out of the state with velocity v\mathbf{v}. Scattering of a bound particle to v′>vesc(r)v' > v_{\mathrm{esc}}(r) gives evaporation,

Cevap[f]=−f(r,v)∫v′>vesc(r)d3v′R(v→v′,r),(6)C_{\mathrm{evap}}[f] = -f(r,\mathbf{v}) \int_{v'>v_{\mathrm{esc}}(r)} d^3\mathbf{v}' R(\mathbf{v} \to\mathbf{v}',r), \tag*{(6)}

where we assume that an evaporated particle leaves the Sun without another collision with a nucleus. Scattering of halo DM into a bound state gives capture. A halo particle with speed v∞v_\infty far from the Sun reaches radius rr with speed w(r)=v∞2+vesc2(r)w(r) = \sqrt{v_\infty^2 + v_{\mathrm{esc}}^2(r)} and is captured if its outgoing speed satisfies v<vesc(r)v < v_{\mathrm{esc}}(r). The capture operator is then

Ccap[f]=n∞∫0∞dv∞f⊙(v∞)v∞w(r)∫dΩw4πR(w→v,r),(7)C_{\mathrm{cap}}[f] = n_\infty\int_0^\infty d v_\infty\frac{f_\odot(v_\infty)}{v_\infty} w(r) \int\frac{d\Omega_{\mathbf{w}}}{4\pi} R(\mathbf{w} \to\mathbf{v},r), \tag*{(7)}

where n∞n_\infty is the halo DM number density far from the Sun, f⊙(v∞)f_\odot(v_\infty) is the halo speed distribution in the solar frame, and dΩw/4πd\Omega_{\mathbf{w}}/4\pi averages over the incident directions. Since we neglect DM self-capture [99] in this work, capture is an external source that does not depend on the bound population.

Annihilation

For ss-wave annihilation, σannvrel\sigma_{\mathrm{ann}}v_{\mathrm{rel}} does not depend on the relative speed vrelv_{\mathrm{rel}} of the annihilating pair, and we denote this constant by ⟨σv⟩\langle\sigma v \rangle. The annihilation operator then depends only on the local density of bound partners,

Cann[f]=−⟨σv⟩nχ(r)f(r,v).(8)C_{\mathrm{ann}}[f] = -\langle\sigma v \rangle n_\chi(r)f(r,\mathbf{v}). \tag*{(8)}

A velocity-dependent annihilation cross section would instead require an average over the velocities of both particles.

BTE in orbital phase space

Between collisions, a bound particle moves on a fixed orbit in the static potential, and its specific energy EE and angular-momentum magnitude LL are conserved. When the mean time between collisions is much longer than the radial period, τcoll≫τr\tau_{\mathrm{coll}} \gg\tau_r, a particle completes many orbits before it scatters again. We then assume that bound DM is distributed uniformly in radial phase space, the time since the last pericenter passage in units of the radial period, along each orbit and describe the captured population by the number of particles per orbit ξ=(E,L)\xi= (E,L). Below, we define the orbital coordinates and the orbital number density, and reconstruct the radial density with this approximation.

Orbital coordinates

With the gravitational potential set to zero at infinity, Φ(∞)=0\Phi(\infty) = 0, the specific energy and angular momentum of a phase-space point (r,v)(\mathbf{r},\mathbf{v}) are

E=v22+Φ(r),L=∣r×v∣,(9)E = \frac{v^2}{2} + \Phi(r), \qquad L = \lvert\mathbf{r} \times\mathbf{v} \rvert, \tag*{(9)}

and the bound orbits that cross the solar interior satisfy

Φ(0)<E<0,0<L<L⊙(E),(10)\Phi(0) < E < 0, \qquad0 < L < L_\odot(E), \tag*{(10)}

where

L⊙(E)=max⁡0≤r≤R⊙Φ(r)≤Er2[E−Φ(r)].(11)L_\odot(E) = \max_{\substack{0 \le r \le R_\odot\\ \Phi(r) \le E}} r\sqrt{2[E-\Phi(r)]}. \tag*{(11)}

This domain includes orbits confined to the Sun and those extending beyond its surface, and the boundary is discussed in App. B. We write ξ(r,v)≡(E,L)\xi(\mathbf{r},\mathbf{v}) \equiv(E,L) for the orbital coordinates of a phase-space point given by Eq. (9). At radius rr, the radial and tangential speeds, vrv_r and vtv_t respectively, are

vr=2[E−Φ(r)]−L2r2,vt=Lr,(12)v_r = \sqrt{2[E-\Phi(r)]-\frac{L^2}{r^2}}, \qquad v_t = \frac{L}{r}, \tag*{(12)}

with total speed v=vr2+vt2v = \sqrt{v_r^2+v_t^2}. Throughout this work, we take vr≥0v_r \ge0 and count the inward and outward passages together, since their radial speeds are equal in a spherically symmetric potential. A bound particle moves from the pericenter r−r_{-} to the apocenter r+r_{+} and back within the radial period

τr(ξ)=2∫r−r+drvr(r,ξ).(13)\tau_{r}(\xi) = 2 \int_{r_{-}}^{r_{+}} \frac{dr}{v_{r}(r,\xi)}. \tag*{(13)}

Orbital number density

The number density of bound particles per dEdLdE dL, derived together with the other quantities of this subsection in App. C § 1.9.2, is

nχ(ξ)=∫d3rd3v Δ(ξ−ξ(r,v))f(r,v),(14)n_{\chi}(\xi) = \int d^{3}\mathbf{r}d^{3}\mathbf{v}\,\Delta(\xi-\xi(\mathbf{r},\mathbf{v}))f(\mathbf{r},\mathbf{v}), \tag*{(14)}

where Δ(ξ−ξ(r,v))≡δ(E−E(r,v))δ(L−L(r,v))\Delta(\xi-\xi(\mathbf{r},\mathbf{v})) \equiv\delta(E-E(\mathbf{r},\mathbf{v}))\delta(L-L(\mathbf{r},\mathbf{v})) selects the phase-space points (r,v)(\mathbf{r},\mathbf{v}) whose orbital coordinates ξ(r,v)\xi(\mathbf{r},\mathbf{v}) equal ξ\xi. Since EE and LL are defined per unit mass, Δ\Delta has the dimension of 1/(rv3)1/(r v^{3}), and nχ(ξ)n_{\chi}(\xi) is the number of particles per dEdLdE dL. The phase-space volume of the orbits in dEdLdE dL is dVξ=8π2Lτr(ξ)dEdLdV_{\xi}=8\pi^{2}L\tau_{r}(\xi)dE dL, and Eq. (14) becomes

nχ(ξ)=8π2Lτr(ξ)f(ξ),(15)n_{\chi}(\xi) = 8\pi^{2}L\tau_{r}(\xi)f(\xi), \tag*{(15)}

where f(ξ)f(\xi) is the phase-space distribution averaged over the orientations and radial phases of the orbit. The total bound population is

Nχ=∫nχ(ξ)dEdL=∫f(ξ)dVξ.(16)N_{\chi} = \int n_{\chi}(\xi)dE dL = \int f(\xi)dV_{\xi}. \tag*{(16)}

Because the interaction rates vary along the orbit with the solar profile and the DM density, we evaluate them with the orbit-averaging approximation,

f(r,v,t)≃f(ξ(r,v),t),(17)f(\mathbf{r},\mathbf{v},t) \simeq f(\xi(\mathbf{r},\mathbf{v}),t), \tag*{(17)}

which requires τcoll≫τr\tau_{\mathrm{coll}} \gg\tau_{r} and a uniform spread of the particles in radial phase. The latter holds because particles reach an orbit at random times through capture or scattering.

Reconstruction of radial DM number density

A particle on orbit ξ\xi spends a time dr/vrdr/v_{r} in a radial interval on each passage, so the probability of finding it between rr and r+drr+dr is

p(r,ξ)dr=2τr(ξ)drvr(r,ξ),(18)p(r,\xi)dr = \frac{2}{\tau_{r}(\xi)}\frac{dr}{v_{r}(r,\xi)}, \tag*{(18)}

with p=0p=0 outside r−<r<r+r_{-}<r<r_{+}. This probability is enhanced near the turning points r±r_{\pm}, where vr→0v_{r}\to0. To obtain a spherically symmetric density, we further assume that the orbits are isotropically oriented. Precession randomizes the pericenter direction within each orbital plane, as shown in the middle panel of Figure 1. The orientation of the orbital plane changes only through collisions, and repeated collisions erase any anisotropy inherited from the halo flux. Even when collisions are too rare to erase this anisotropy, the assumption of isotropically oriented orbits remains a good approximation, as we show in App. E. With these assumptions, the radial number density is

Radial probability along a bound orbit, precessing bound orbit, and mixed radial phase annulus

Figure 1. Left: Radial probability along a bound orbit. Middle: In the solar potential, a bound orbit generally precesses and does not close after one radial period τr\tau_r. Right: An ensemble with mixed radial phases fills an annulus in the orbital plane, with radial probability proportional to the time spent at each radius.

nχ(r)=14πr2∫Φ(0)0dE∫0L⊙(E)dL nχ(ξ)p(r,ξ),(19)n_{\chi}(r) = \frac{1}{4\pi r^{2}}\int_{\Phi(0)}^{0}dE\int_{0}^{L_{\odot}(E)}dL\,n_{\chi}(\xi)p(r,\xi), \tag*{(19)}

as illustrated in Figure 1. Integrating over all radii recovers the total bound population,

Nχ=∫0∞dr 4πr2nχ(r),(20)N_{\chi} = \int_{0}^{\infty}dr\,4\pi r^{2}n_{\chi}(r), \tag*{(20)}

which is equivalent to Eq. (3).

Master equation for DM evolution in orbital phase space

Multiplying Eq. (1) by Δ(ξ−ξ(r,v))\Delta(\xi-\xi(\mathbf{r},\mathbf{v})) and integrating over phase space, we obtain the master equation for the evolution of nχ(ξ)n_\chi(\xi),

∂nχ(ξ)∂t=∫d3rd3v Δ(ξ−ξ(r,v))C[f](r,v)≡K[nχ](ξ).(21)\frac{\partial n_\chi(\xi)}{\partial t}=\int\mathrm{d}^3r\mathrm{d}^3v\,\Delta(\xi-\xi(\mathbf{r},\mathbf{v}))C[f](\mathbf{r},\mathbf{v}) \equiv K[n_\chi](\xi). \tag*{(21)}

The advection terms drop out without approximation in this projection, because the motion between collisions conserves EE and LL and preserves phase-space volume (App. C1). The kernel KK gives the rate of population change per dEdL\mathrm{d}E\mathrm{d}L, and we derive the master equation and the collisional kernel discussed below in Apps. C3 and C4, respectively.

DM-nucleus scattering kernels

The rate per particle to scatter from orbit ξ\xi into the orbits within dE′dL′\mathrm{d}E'\mathrm{d}L' around ξ′=(E′,L′)\xi'=(E',L') is K(ξ→ξ′)dE′dL′K(\xi\to\xi')\mathrm{d}E'\mathrm{d}L'. Averaging the local rate RR, Eq. (4), along the incident orbit and selecting the final energy Ef′E'_f and angular momentum Lf′L'_f, given by Eq. (47), gives

K(ξ→ξ′)=2τr(ξ)∫r−r+drvr(r,ξ)∑AnA(r)σχA4π∫0∞du u2fA(r,u)∫dΩu∫dΩg gFA2(q)δ(E′−Ef′)δ(L′−Lf′).(22)K(\xi\to\xi')=\frac{2}{\tau_r(\xi)}\int_{r_-}^{r_+}\frac{\mathrm{d}r}{v_r(r,\xi)}\sum_A n_A(r)\frac{\sigma_{\chi A}}{4\pi}\int_0^\infty\mathrm{d}u\,u^2f_A(r,u)\int\mathrm{d}\Omega_u\int\mathrm{d}\Omega_g\,gF_A^2(q)\delta(E'-E'_f)\delta(L'-L'_f). \tag*{(22)}

Since nA(r>R⊙)=0n_A(r>R_\odot)=0, only the solar interior contributes to the radial integral. The average is nevertheless normalized by the full period τr(ξ)\tau_r(\xi) to account for the time a particle may spend outside the Sun between collisions.

As in § 1 A, the role of a collision depends on whether the initial and final orbits are bound. Scattering between bound orbits redistributes the population,

Kcoll[nχ](ξ)=∫d3rd3v Δ(ξ−ξ(r,v))Ccoll[f](r,v)=∫Φ(0)0dE′∫0L⊙(E′)dL′[K(ξ′→ξ)nχ(ξ′)−K(ξ→ξ′)nχ(ξ)].(23)K_{\mathrm{coll}}[n_\chi](\xi)=\int\mathrm{d}^3r\mathrm{d}^3v\,\Delta(\xi-\xi(\mathbf{r},\mathbf{v}))C_{\mathrm{coll}}[f](\mathbf{r},\mathbf{v}) =\int_{\Phi(0)}^0\mathrm{d}E'\int_0^{L_\odot(E')}\mathrm{d}L'\left[K(\xi'\to\xi)n_\chi(\xi')-K(\xi\to\xi')n_\chi(\xi)\right]. \tag*{(23)}

Otherwise, the particle becomes unbound, which gives the evaporation kernel,

Kevap[nχ](ξ)=−nχ(ξ)∫0∞dE′∫0L⊙(E′)dL′K(ξ→ξ′).(24)K_{\mathrm{evap}}[n_\chi](\xi)=-n_\chi(\xi)\int_0^\infty\mathrm{d}E'\int_0^{L_\odot(E')}\mathrm{d}L'K(\xi\to\xi'). \tag*{(24)}

where the integration over E′>0E'>0 selects the collisions that eject the particle.

Capture moves particles from unbound to bound orbits. An unbound particle with incident energy E′=v∞2/2>0E^\prime= v_\infty^2/2 > 0 at infinity crosses the solar interior once, with the passage time

τpass(ξ′)=2∫r−(ξ′)R⊙drvr(r,ξ′),(25)\tau_{\mathrm{pass}}(\xi^\prime) = 2 \int_{r_-(\xi^\prime)}^{R_\odot} \frac{\mathrm{d}r}{v_r(r,\xi^\prime)}, \tag*{(25)}

as illustrated in Figure 2. Averaging the scattering rate over this passage gives the source of bound DM,

Figure 2. Passage of an unbound DM particle through the solar interior (thick red line), with incident specific energy E′>0E' > 0.

Kcap[nχ](ξ)=2πn∞∫0∞dE′∫0L⊙(E′)dL′ L′τpass(ξ′)f⊙(2E′)2E′K(ξ′→ξ),(26)\mathcal{K}_{\mathrm{cap}}[n_\chi](\xi) = 2\pi n_\infty\int_0^\infty\mathrm{d}E^\prime\int_0^{L_\odot(E^\prime)} \mathrm{d}L^\prime\, L^\prime\tau_{\mathrm{pass}}(\xi^\prime) \frac{f_\odot(\sqrt{2E^\prime})}{2E^\prime} \mathcal{K}(\xi^\prime\to\xi), \tag*{(26)}

where K(ξ′→ξ)\mathcal{K}(\xi^\prime\to\xi) is given by Eq. (22) with τpass\tau_{\mathrm{pass}} in place of τr\tau_r and with the radial integral ending at R⊙R_\odot. The integral over E′E^\prime includes all incident unbound energies.

Annihilation kernel

Annihilation couples the populations of different bound orbits through their spatial overlap. Averaging the local loss rate ⟨σv⟩nχ(r)\langle\sigma v \rangle n_\chi(r) over the orbit with the radial probability of Eq. (18) and using Eq. (19), we obtain

Kann[nχ](ξ)=−nχ(ξ)∫Φ(0)0dE′∫0L⊙(E′)dL′nχ(ξ′)×⟨σv⟩∫0∞dr p(r,ξ)p(r,ξ′)4πr2,(27)\mathcal{K}_{\mathrm{ann}}[n_\chi](\xi) = -n_\chi(\xi) \int_{\Phi(0)}^0 \mathrm{d}E^\prime\int_0^{L_\odot(E^\prime)} \mathrm{d}L^\prime n_\chi(\xi^\prime) \times\langle\sigma v \rangle\int_0^\infty\mathrm{d}r \, \frac{p(r,\xi)p(r,\xi^\prime)}{4\pi r^2}, \tag*{(27)}

where the factor 1/(4πr2)1/(4\pi r^2) distributes the radial probability over a spherical shell.

Connection to the classical treatment of DM evolution

Integrating Eq. (21) over the bound orbital phase space gives the number evolution equation

dNχdt=Cc+CeNχ+CaNχ2,(28)\frac{\mathrm{d}N_\chi}{\mathrm{d}t} = C_c + C_e N_\chi+ C_a N_\chi^2, \tag*{(28)}

where the capture, annihilation and evaporation coefficients are given by the orbital kernels,

Cc=∫dEdL Kcap[nχ](ξ),(29)C_c = \int\mathrm{d}E\mathrm{d}L \, \mathcal{K}_{\mathrm{cap}}[n_\chi](\xi), \tag*{(29)}
Ca=1Nχ2∫dEdL Kann[nχ](ξ),(30)C_a = \frac{1}{N_\chi^2} \int\mathrm{d}E\mathrm{d}L \, \mathcal{K}_{\mathrm{ann}}[n_\chi](\xi), \tag*{(30)}
Ce=1Nχ∫dEdL Kevap[nχ](ξ).(31)C_e = \frac{1}{N_\chi} \int\mathrm{d}E\mathrm{d}L \, \mathcal{K}_{\mathrm{evap}}[n_\chi](\xi). \tag*{(31)}

The integrals cover Φ(0)≤E≤0\Phi(0) \le E \le0 and 0≤L≤L⊙(E)0 \le L \le L_\odot(E), and with this convention CaC_a and CeC_e are negative. Although capture does not depend on the bound population nχn_\chi, we write the capture kernels with an argument, such as Kcap[nχ]\mathcal{K}_{\mathrm{cap}}[n_\chi], to keep the notation uniform. Bound-to-bound scattering only redistributes particles in phase space and does not change NχN_\chi, so it vanishes in the number evolution equation. The classical treatment assumes that the distribution rapidly approaches a stationary, isothermal form, nχiso(r)∝exp⁡(−mχΦ(r)/Tχ)n_\chi^{\mathrm{iso}}(r) \propto\exp(-m_\chi\Phi(r)/T_\chi), where the DM temperature TχT_\chi is computed following [73], making these coefficients time independent. For this profile, about half of the DM lies within the thermal radius rth=3T⊙(0)/(2πGρcmχ)r_{\mathrm{th}} = \sqrt{3T_\odot(0)/(2\pi G\rho_c m_\chi)}, where GG is Newton’s constant and T⊙(0)T_\odot(0) and ρc\rho_c are the temperature and density at the solar center. Frequent collisions with nuclei can drive DM toward such a distribution in the solar core. At small scattering cross sections, however, relaxation can take longer than the solar age, and the evolving distribution must be obtained from the master equation.

Numerical results

To compare the BTE with the classical treatment of Ref. [73], in which frequent collisions with nuclei give an isothermal radial density nχiso(r)n_{\chi}^{\mathrm{iso}}(r) of captured DM, we evolve both from an empty population, Nχ=0N_{\chi}=0 at t=0t=0, under continuous capture, annihilation and evaporation up to one solar age, t⊙=4.6t_{\odot}=4.6 Gyr, with ⟨σv⟩=3×10−26 cm3s−1\langle\sigma v\rangle=3\times10^{-26}\ {\rm cm}^{3}{\rm s}^{-1}. In the BTE approach, we solve the master equation, Eq. (21), for nχ(ξ,t)n_{\chi}(\xi,t) on the 192×48192\times48 grid in (E,L)(E,L) described in App. F and evolve it with the implicit backward differentiation formula method,4 with the kernels of Eqs. (23), (24), (26) and (27), and obtain nχ(r,t)n_{\chi}(r,t) from Eq. (19). In the classical treatment, we solve Eq. (28) for Nχ(t)N_{\chi}(t), computing Cc,a,eC_{c,a,e} as in Ref. [73].

We adopt a local halo density of ρ∞=n∞mχ=0.3 GeV cm−3\rho_{\infty}=n_{\infty}m_{\chi}=0.3\ {\rm GeV\,cm}^{-3}, and the DM halo speed v∞v_{\infty} in the solar frame follows a Maxwellian distribution [73],

f⊙(v∞)=32πv∞v⊙vd[e−3(v∞−v⊙)22vd2−e−3(v∞+v⊙)22vd2],(32)f_{\odot}(v_{\infty})=\sqrt{\frac{3}{2\pi}}\frac{v_{\infty}}{v_{\odot}v_{d}}\left[e^{-\frac{3(v_{\infty}-v_{\odot})^{2}}{2v_{d}^{2}}}-e^{-\frac{3(v_{\infty}+v_{\odot})^{2}}{2v_{d}^{2}}}\right], \tag*{(32)}

where v⊙=230 km s−1v_{\odot}=230\ {\rm km\,s}^{-1} is the speed of the Sun relative to the halo and vd=220 km s−1v_{d}=220\ {\rm km\,s}^{-1} is the root-mean-square speed of halo DM in the Galactic frame. The distribution is truncated at the Galactic escape speed of 544 km s−1544\ {\rm km\,s}^{-1} in the Galactic frame [101], which removes only a small high-speed tail from Eq. (32). We also neglect perturbations from the planets [76].5

The parameter space considered here is 0.1≤mχ/GeV≤1030.1\le m_{\chi}/{\rm GeV}\le10^{3} and 10−50≤σχp/cm2≤10−4010^{-50}\le\sigma_{\chi p}/{\rm cm}^{2}\le10^{-40}, where σχp\sigma_{\chi p} is the DM-proton cross section, for both SI and SD interactions. For SI scattering, we include all 29 elements of the AGSS09 solar model and assume isospin symmetry, giving σχA=σχpA2(μχA/μχp)2\sigma_{\chi A}=\sigma_{\chi p}A^{2}(\mu_{\chi A}/\mu_{\chi p})^{2}, where μχA,p\mu_{\chi A,p} are the DM-nucleus/nucleon reduced masses, and we use the Helm form factor

FA,SI(q)=3j1(qRA)qRAe−q2s2/2,(33)F_{\mathrm{A,SI}}(q)=\frac{3j_{1}(qR_{\mathrm{A}})}{qR_{\mathrm{A}}}e^{-q^{2}s^{2}/2}, \tag*{(33)}

where j1j_{1} is the spherical Bessel function of the first kind. The nuclear radius is RA=cA2+73π2a2−5s2R_{\mathrm{A}}=\sqrt{c_{\mathrm{A}}^{2}+\frac{7}{3}\pi^{2}a^{2}-5s^{2}}, with cA=(1.23A1/3−0.6) fmc_{\mathrm{A}}=(1.23A^{1/3}-0.6)\ {\rm fm}, a=0.52 fma=0.52\ {\rm fm}, s=0.9 fms=0.9\ {\rm fm} and FA(0)=1F_{\mathrm{A}}(0)=1.

For SD scattering, we take proton-only couplings, giving σχA=σχp⟨Sp,A⟩2(μχA/μχp)2(4/3)(JA+1)/JA\sigma_{\chi A}=\sigma_{\chi p}\langle S_{p,A}\rangle^{2}(\mu_{\chi A}/\mu_{\chi p})^{2}(4/3)(J_{\mathrm{A}}+1)/J_{\mathrm{A}}, where JAJ_{\mathrm{A}} is the nuclear spin and ⟨Sp,A⟩\langle S_{p,A}\rangle is the expectation value of the proton spin in the nucleus, taken from DarkSUSY 6.4.3 [69]. The momentum dependence is described by the form factor

FA,SD2(q)=exp⁡(−q2RA23),(34)F_{\mathrm{A,SD}}^{2}(q)=\exp\left(-\frac{q^{2}R_{\mathrm{A}}^{2}}{3}\right), \tag*{(34)}

where RA=κA(0.91A1/3+0.30) fmR_{\mathrm{A}}=\kappa_{\mathrm{A}}(0.91A^{1/3}+0.30)\ {\rm fm}. We include 1H{}^{1}\mathrm{H}, 3He{}^{3}\mathrm{He}, 13C{}^{13}\mathrm{C}, 14N{}^{14}\mathrm{N}, 15N{}^{15}\mathrm{N} and 17O{}^{17}\mathrm{O}, treating hydrogen as a free particle with FH,SD=1F_{\mathrm{H,SD}}=1 and using κA=(1,1,1.04,1.04,1.12)\kappa_{\mathrm{A}}=(1,1,1.04,1.04,1.12) for the other five nuclei in the order listed.6 With these spin factors, 3He{}^{3}\mathrm{He}, 13C{}^{13}\mathrm{C} and 17O{}^{17}\mathrm{O} carry their spin only in the unpaired neutron, so ⟨Sp,A⟩=0\langle S_{p,A}\rangle= 0 and they do not contribute for proton-only couplings. The thermal velocities of the nuclei follow the Maxwellian distribution

fA(r,u)=[mA2πT⊙(r)]3/2exp⁡[−mAu22T⊙(r)],(35)f_A(r,u) = \left[\frac{m_A}{2\pi T_\odot(r)}\right]^{3/2} \exp\left[-\frac{m_Au^2}{2T_\odot(r)}\right], \tag*{(35)}

where mAm_A is the nuclear mass and T⊙(r)T_\odot(r) is the solar temperature at rr. The two treatments give the same capture rate CcC_c, while the BTE also keeps the orbital distribution of newly captured DM through Kcap\mathcal{K}_{\rm cap}.

Validity of the orbit average

The orbit-averaging approximation requires τcoll≫τr\tau_{\rm coll} \gg\tau_r, so that the number of collisions per orbit, Ncoll=τr/τcollN_{\rm coll} = \tau_r/\tau_{\rm coll}, is small. Figure 3 shows the cross sections below which Ncoll<0.1N_{\rm coll} < 0.1 for 99% of the bound DM at all times during the evolution. Since the approximation fails only when NcollN_{\rm coll} approaches unity, this criterion leaves a margin of one order of magnitude, and even at the largest SI cross section we consider, σχp=10−40 cm2\sigma_{\chi p} = 10^{-40}\ {\rm cm}^2, NcollN_{\rm coll} remains far below one. At much larger cross sections, frequent collisions interrupt the orbital motion, and DM diffuses in the short-mean-free-path regime studied in Refs. [83, 102].

Figure 3. Below the SD (green) and SI (red) curves, 99% of the bound DM on our (E,L)(E, L) grid has Ncoll<0.1N_{\mathrm{coll}} < 0.1 per orbit at all times of the evolution.

Radial number density

Fig. 4 compares the radial densities for SD scattering at mχ=100 GeVm_\chi= 100\ {\rm GeV}, with σχp=10−40\sigma_{\chi p} = 10^{-40} and 10−50 cm210^{-50}\ {\rm cm}^2. The profiles of newly captured DM (orange curves), set by the capture kernel alone, have the same shape for the two cross sections, while their values reflect the different capture rates. The BTE profiles have tails that extend much farther than the classical isothermal profiles, because freshly captured particles are weakly bound (E→0E \to0) and occupy large orbits, often with apocenters beyond the solar surface.

For σχp=10−40 cm2\sigma_{\chi p} = 10^{-40}\ {\rm cm}^2, repeated collisions (∼109\sim10^9 per particle over t⊙t_\odot on average) transfer energy to solar nuclei and shrink the orbits, so that by one solar age the central population approaches the classical isothermal distribution. A dilute tail remains beyond the thermal radius, rth∼0.01R⊙r_{\rm th} \sim0.01R_\odot for this mass, because continuous capture keeps supplying particles that have not yet thermalized. For σχp=10−50 cm2\sigma_{\chi p} = 10^{-50}\ {\rm cm}^2, a particle undergoes much less than one collision on average over the solar age, so it does not thermalize, as anticipated in Refs. [76, 77]. The distribution then preserves the broad shape set by capture as NχN_\chi grows. At t⊙t_\odot, its central density is more than six orders of magnitude below the classical one, which suppresses the annihilation rate, as discussed next.

Annihilation rate

The annihilation rate follows from Kann\mathcal{K}_{\rm ann} given in Eq. (27),

ΓABTE=−12∫Φ(0)0dE∫0L⊙(E)dL Kann[nχ](ξ),(36)\Gamma_A^{\rm BTE} = -\frac{1}{2}\int_{\Phi(0)}^0 dE \int_0^{L_\odot(E)} dL\,\mathcal{K}_{\rm ann}[n_\chi](\xi), \tag*{(36)}

where the factor 1/21/2 accounts for the two particles removed in each annihilation. In evaluating ΓABTE\Gamma_A^{\rm BTE}, we restrict the radial integral in Eq. (27) to r<R⊙r < R_\odot, while the evolution itself includes annihilation at all radii.

Figure 5 shows the ratio of ΓABTE\Gamma_A^{\rm BTE} to the classical annihilation rate ΓAclass\Gamma_A^{\rm class}, obtained by assuming nχiso(r)n_\chi^{\rm iso}(r). Above a few GeV, where evaporation is weak, the BTE annihilation rate is suppressed most strongly at small σχp\sigma_{\chi p} and large mχm_{\chi}. In this regime, DM does not thermalize within the solar age and keeps the broad distribution set by capture, with most particles on orbits that extend well beyond the solar surface, as in the right panel of Figure 4. This dilute distribution reduces ΓA\Gamma_{\mathrm{A}} through its dependence on nχ2n_{\chi}^{2}. Annihilation outside the Sun, which is negligible for thermalized DM, would increase ΓABTE\Gamma_{\mathrm{A}}^{\mathrm{BTE}} in this regime by at most a factor of a few, but the distribution remains too dilute for this increase to overcome the suppression. The suppression is much stronger for SD than for SI scattering at the same σχp\sigma_{\chi p}, which reflects how efficiently collisions transfer energy. For SI scattering, the ratio stays roughly between 0.5 and 3 above about 1 GeV for σχp≥10−48 cm2\sigma_{\chi p}\geq10^{-48}\,\mathrm{cm}^{2}, because collisions with nuclei are both frequent and efficient in transferring energy. Energy exchange in an elastic collision is most efficient for comparable DM and nuclear masses, with a mean energy loss fraction of η=2mχmA/(mχ+mA)2\eta=2m_{\chi}m_{\mathrm{A}}/(m_{\chi}+m_{\mathrm{A}})^{2}, which favors helium, oxygen and iron. Helium and oxygen are among the most abundant elements in the Sun, and the coherent enhancement A2(μχA/μχp)2A^{2}(\mu_{\chi\mathrm{A}}/\mu_{\chi\mathrm{p}})^{2} of SI scattering compensates for the lower abundance of iron. At σχp=10−50 cm2\sigma_{\chi p}=10^{-50}\,\mathrm{cm}^{2}, collisions become too rare to thermalize DM within the solar age, although the suppression remains weakest where these nuclei are most effective, which gives the broad bump from a few to several tens of GeV. In contrast, SD scattering has no coherent A2A^{2} enhancement and is dominated by hydrogen for the interactions considered here. Energy exchange becomes inefficient when mχm_{\chi} is much larger than the proton mass, so the suppression grows rapidly with increasing mχm_{\chi}. Figure 6 shows ΓABTE/ΓAclass\Gamma_{\mathrm{A}}^{\mathrm{BTE}}/\Gamma_{\mathrm{A}}^{\mathrm{class}} over the mχm_{\chi}-σχp\sigma_{\chi p} plane for SI and SD interactions. In the region with ΓABTE/ΓAclass<0.5\Gamma_{\mathrm{A}}^{\mathrm{BTE}}/\Gamma_{\mathrm{A}}^{\mathrm{class}}<0.5, i.e., where the classical treatment overestimates the annihilation rate by more than a factor of two, the evolution of the distribution has to be followed with the BTE. Below a few GeV and at sufficiently large cross sections, evaporation becomes dominant and determines the captured population, as discussed next.

Figure 4. nχ(r,t)n_\chi(r,t) for the SD benchmark mχ=100 GeVm_\chi=100\ \mathrm{GeV}, with σχp=10−40 cm2\sigma_{\chi p}=10^{-40}\ \mathrm{cm}^2 (left) and 10−50 cm210^{-50}\ \mathrm{cm}^2 (right). Orange curves show the distribution of DM captured within one second, nχ(ξ)=Kcap(ξ)×1 sn_\chi(\xi)=K_{\mathrm{cap}}(\xi)\times1\ \mathrm{s}, and dark-blue curves the distribution at t=t⊙t=t_\odot. Solid curves are the BTE results and dotted curves the classical isothermal results under continuous capture and annihilation. Vertical dashed lines mark the solar surface.

Figure 5. BTE-to-classical ratio of the annihilation rate inside the Sun at t⊙t_\odot, for SI (left) and SD (right) interactions. Colors indicate the different cross sections.

Figure 6. Similar to Figure 5, with a scan of ΓABTE/ΓAclass\Gamma_{\mathrm{A}}^{\mathrm{BTE}}/\Gamma_{\mathrm{A}}^{\mathrm{class}} over the mχm_{\chi}-σχp\sigma_{\chi p} plane for SI (left) and SD (right) interactions. The contours indicate ratios of 0.9 (green), 0.5 (white), 0.1 (cyan), 0.01 (yellow), and 10−910^{-9} (pink).

Evaporation

To determine the DM mass at which evaporation becomes substantial, we use the evaporation mass mevapm_{\mathrm{evap}}, following the definition in Refs. [73, 75],

−Ce(t)Nχ(t)Cc∣mχ=mevap=1011,(37)-\left.\frac{C_{\mathrm{e}}(t)N_{\chi}(t)}{C_{\mathrm{c}}}\right|_{m_{\chi}=m_{\mathrm{evap}}}=\frac{10}{11}, \tag*{(37)}

where −Ce-C_{\mathrm{e}} is the evaporation rate per particle. The criterion therefore selects the mass at which the total evaporation rate is 10/1110/11 of CcC_{\mathrm{c}} at t⊙t_{\odot}, without assuming number equilibrium between capture and evaporation, N˙χ=0\dot{N}_{\chi}=0.7 Figure 7 shows the BTE (solid red) and classical (solid green) results for mevapm_{\mathrm{evap}}, while the green dashed curve assumes N˙χ=0\dot{N}_{\chi}=0. At σχp=10−44 cm2\sigma_{\chi p}=10^{-44}\,\mathrm{cm}^{2}, the largest cross section in Figure 7, the classical values agree with those of Ref. [73] for both SI and SD scattering. Since mevapm_{\mathrm{evap}} lies near the masses of hydrogen and helium, the most abundant nuclei in the Sun, η\eta is close to its maximum 1/21/2 at mχ∼mevapm_{\chi} \sim m_{\mathrm{evap}}, and DM thermalizes easily there once collisions are frequent enough. The BTE and classical values of mevapm_{\mathrm{evap}} thus agree for both SI and SD scattering down to σχp≈10−47 cm2\sigma_{\chi p} \approx10^{-47}\,\mathrm{cm}^{2}. At smaller cross sections, infrequent collisions keep nχ(r)n_{\chi}(r) away from an isothermal distribution, and less DM resides in the dense, hot core where most collisions occur. As a result, evaporation is less efficient, and the BTE value of mevapm_{\mathrm{evap}} falls significantly below the classical one.

Figure 7. Evaporation mass defined by Eq. (37) for SI (left) and SD (right) scattering. Red and green solid curves show the BTE and classical results at t⊙t_\odot, and green dashed curves show the classical stationary solution with N˙χ=0\dot{N}_\chi= 0.

At fixed σχp\sigma_{\chi p}, reducing mχm_{\chi} far below the hydrogen mass gives η∼2mχ/mA≪1\eta\sim2m_{\chi}/m_{A} \ll1. With such inefficient energy exchange, the nonthermal distribution contains more weakly bound particles near the solar surface, which evaporate more easily. Since the ratio of the BTE evaporation coefficient, CeBTEC_{e}^{\mathrm{BTE}}, to the classical value, CeclassC_{e}^{\mathrm{class}}, is sensitive to the weakly bound tail, Figure 8 shows it only for the bound DM on orbits with r+<R⊙r_{+}<R_{\odot} (solid) or r+<3rthr_{+}<3r_{\mathrm{th}} (dotted), for different cross sections.8 This ratio refers only to the selected orbits, whereas mevapm_{\mathrm{evap}} depends on the total evaporation rate of all bound DM, much of which lies on orbits beyond the Sun that rarely scatter, so the BTE can give a lower mevapm_{\mathrm{evap}} even where the ratio exceeds unity. For mχ≲mevap∼1 GeVm_{\chi}\lesssim m_{\mathrm{evap}}\sim1\,\mathrm{GeV}, more of the weakly bound particles on the selected orbits evaporate than in the classical treatment, which lowers the density close to the solar surface, at r∼0.3r\sim0.3–1R⊙1R_{\odot}, and can also explain the suppression of the annihilation rate at mχ≲1 GeVm_{\chi}\lesssim1\,\mathrm{GeV} shown in Figure 5. When only orbits with r+<3rthr_{+}<3r_{\mathrm{th}} are included (dotted curves), the two treatments give similar nχn_{\chi}, and the ratio remains close to unity between about 0.5 GeV0.5\,\mathrm{GeV} and a few GeV\mathrm{GeV}.

Figure 8. BTE-to-classical ratio of the evaporation coefficient at t⊙t_{\odot}, for SI (left) and SD (right) scattering. Colors indicate the different cross sections. Solid and dotted curves include only orbits with r+<R⊙r_{+}<R_{\odot} and r+<3rthr_{+}<3r_{\mathrm{th}}, respectively, where rthr_{\mathrm{th}} is the thermal radius of the classical distribution. For each selection, the evaporation rate is divided by the number of selected particles in each treatment.

Once mχ≳m_{\chi}\gtrsim a few GeV, evaporation becomes negligible in the classical treatment, while the BTE keeps a weakly bound nonthermal tail that can still evaporate. For r+<R⊙r_{+}<R_{\odot}, this tail extends to the solar surface, and CeBTE/CeclassC_{e}^{\mathrm{BTE}}/C_{e}^{\mathrm{class}} rises rapidly and diverges as Ceclass→0C_{e}^{\mathrm{class}}\to0 (see App. § 1.3.4 for the nχn_{\chi} tail discussion). For SD scattering with σχp=10−50 cm2\sigma_{\chi p}=10^{-50}\,\mathrm{cm}^{2}, inefficient collisions leave such a tail even within 3rth3r_{\mathrm{th}}, so the dotted curve also reaches O(10)\mathcal{O}(10) around mχ∼5 GeVm_{\chi}\sim5\,\mathrm{GeV}. However, such a divergence does not mean that evaporation still dominates capture well above mevapm_{\mathrm{evap}} in the BTE approach. Figure 9 shows the evaporated fraction,

Figure 9. Accumulated evaporation fraction, Eq. (38), at t⊙t_{\odot} for SI (left) and SD (right) scattering. Solid and dotted curves show the BTE and classical results, respectively, with σχp\sigma_{\chi p} indicated by color.

fevap=NevapNevap+Nbound(38)f_{\mathrm{evap}}=\frac{N_{\mathrm{evap}}}{N_{\mathrm{evap}}+N_{\mathrm{bound}}} \tag*{(38)}

where NevapN_{\mathrm{evap}} is the number of evaporated particles and NboundN_{\mathrm{bound}} is the number of particles that remain bound. The fraction drops significantly at mχ≫mevapm_{\chi}\gg m_{\mathrm{evap}}, showing that evaporation has little impact on the captured DM there.

Generalization to two-component DM

As an example of DM with several components, we take the Higgsino model [96, 97], one of the inelastic DM interpretations [96, 97, 104, 105, 106] of the high-recoil candidate reported by LZ [98], for which solar-neutrino searches provide complementary tests [107, 108, 109, 110, 111]. The two neutral states, with masses mχ2=mχ1+δm_{\chi_{2}}=m_{\chi_{1}}+\delta, interact through inelastic scattering χ1A↔χ2A\chi_{1}A\leftrightarrow\chi_{2}A, elastic SI and SD scattering, radiative decay χ2→χ1+γ\chi_{2}\to\chi_{1}+\gamma and annihilation χ1χ1→W+W−/ZZ\chi_{1}\chi_{1}\to W^{+}W^{-}/ZZ, with the cross sections and

Coupled master equation

The halo contains a negligible abundance of χ2\chi_{2}, because the excited state decays promptly through χ2→χ1+γ\chi_{2} \to\chi_{1}+\gamma, with a lifetime τ2(δ)≃0.026 s (370 keV/δ)5\tau_{2}(\delta) \simeq0.026\ \mathrm{s}\,(370\ \mathrm{keV}/\delta)^{5} [112]. Halo χ1\chi_{1} is captured mainly through the inelastic scattering χ1A→χ2A\chi_{1}A \to\chi_{2}A on solar nuclei, because its elastic cross sections are much smaller than the inelastic ones (see Table I), and the χ2\chi_{2} produced in this way decays back to χ1\chi_{1} before it can scatter again. The populations of χ1\chi_{1} and χ2\chi_{2} on orbit ξ\xi then evolve according to the coupled master equations

Particle propertyValue
mχ1m_{\chi_1}1.1×103 GeV1.1 \times10^{3}\ \mathrm{GeV}
σpSI,inelastic\sigma_{p}^{\mathrm{SI,inelastic}}6.8×10−41 cm26.8 \times10^{-41}\ \mathrm{cm}^{2}
σnSI,inelastic\sigma_{n}^{\mathrm{SI,inelastic}}7.6×10−39 cm27.6 \times10^{-39}\ \mathrm{cm}^{2}
σpSI,elastic\sigma_{p}^{\mathrm{SI,elastic}}2.7×10−49 cm22.7 \times10^{-49}\ \mathrm{cm}^{2}
σnSI,elastic\sigma_{n}^{\mathrm{SI,elastic}}2.8×10−49 cm22.8 \times10^{-49}\ \mathrm{cm}^{2}
σpSD,elastic\sigma_{p}^{\mathrm{SD,elastic}}1.0×10−45 cm21.0 \times10^{-45}\ \mathrm{cm}^{2}
σnSD,elastic\sigma_{n}^{\mathrm{SD,elastic}}7.7×10−46 cm27.7 \times10^{-46}\ \mathrm{cm}^{2}
⟨σv⟩\langle\sigma v\rangle8.1×10−27 cm3s−18.1 \times10^{-27}\ \mathrm{cm}^{3}\mathrm{s}^{-1}
τ2(δ=370 keV)\tau_{2}(\delta= 370\ \mathrm{keV})0.026 s0.026\ \mathrm{s}

Table I. Higgsino benchmark parameters [96]. Proton (pp) and neutron (nn) couplings differ in both SI and SD scattering. Here ⟨σv⟩\langle\sigma v\rangle refers to χ1χ1\chi_{1}\chi_{1} annihilation, and τ2\tau_{2} is the excited-state lifetime. Other model parameters [96] listed in Tab. I.9

∂nχ1(ξ)∂t=Kcoll,11[nχ1](ξ)+Kcoll,21[nχ2](ξ)+Kevap,1[nχ1](ξ)+Kann,1[nχ1](ξ)+nχ2(ξ)τ2,(39)\frac{\partial n_{\chi_{1}}(\xi)}{\partial t} = \mathcal{K}_{\mathrm{coll},11}[n_{\chi_{1}}](\xi) + \mathcal{K}_{\mathrm{coll},21}[n_{\chi_{2}}](\xi) + \mathcal{K}_{\mathrm{evap},1}[n_{\chi_{1}}](\xi) + \mathcal{K}_{\mathrm{ann},1}[n_{\chi_{1}}](\xi) + \frac{n_{\chi_{2}}(\xi)}{\tau_{2}}, \tag*{(39)}
∂nχ2(ξ)∂t=Kcap,2[nχ2](ξ)+Kcoll,12[nχ1](ξ)+Kcoll,22[nχ2](ξ)+Kevap,2[nχ2](ξ)−nχ2(ξ)τ2,(40)\frac{\partial n_{\chi_{2}}(\xi)}{\partial t} = \mathcal{K}_{\mathrm{cap},2}[n_{\chi_{2}}](\xi) + \mathcal{K}_{\mathrm{coll},12}[n_{\chi_{1}}](\xi) + \mathcal{K}_{\mathrm{coll},22}[n_{\chi_{2}}](\xi) + \mathcal{K}_{\mathrm{evap},2}[n_{\chi_{2}}](\xi) - \frac{n_{\chi_{2}}(\xi)}{\tau_{2}}, \tag*{(40)}

where, for i≠ji \ne j, Kcoll,ij\mathcal{K}_{\mathrm{coll},ij} describes the production of bound χj\chi_{j} from χi\chi_{i}, while Kcoll,ii\mathcal{K}_{\mathrm{coll},ii} includes elastic redistribution and the loss through conversion into the other state. The last terms of the two equations describe the gain of χ1\chi_{1} and the loss of χ2\chi_{2} through the decay χ2→χ1+γ\chi_{2} \to\chi_{1}+\gamma. The capture of halo χ1\chi_{1} into bound χ2\chi_{2} is described by Kcap,2\mathcal{K}_{\mathrm{cap},2} in Eq. (37b), and we omit the small contributions from halo χ2\chi_{2} and from the elastic capture of χ1\chi_{1} from Eq. (37).

The photon emitted in the decay carries energy Eγ≃δE_{\gamma} \simeq\delta in the nonrelativistic limit, so χ1\chi_{1} keeps nearly the same orbital energy as χ2\chi_{2}, Eχ1≃Eχ2E_{\chi_{1}} \simeq E_{\chi_{2}}, and receives only a small recoil momentum ∣Δp∣∼δ|\Delta p| \sim\delta. The fractional change in angular momentum is then ∣ΔL∣/L∼∣r×Δp∣/(rmχ1vt)∼δ/(mχ1vt)|\Delta L|/L \sim|r \times\Delta p|/(r m_{\chi_{1}}v_{t}) \sim\delta/(m_{\chi_{1}}v_{t}), which is about 10−410^{-4} for vt∼103 km s−1v_{t} \sim10^{3}\ \mathrm{km}\ \mathrm{s}^{-1}. Our numerical calculation shows that this recoil has little impact, so Eq. (37) neglects it and places the daughter χ1\chi_{1} on the orbit of its parent χ2\chi_{2}.

Since χ2\chi_{2} decays promptly, nχ2n_{\chi_{2}} stays negligible, and the bound population consists of χ1\chi_{1}, whose annihilation rate is

ΓA=−12∫Φ(0)1dE∫0L∘(E)dL Kann,1[nχ1](ξ).(41)\Gamma_{A} = -\frac{1}{2}\int_{\Phi(0)}^{1} dE \int_{0}^{L^{\circ}(E)} dL\,\mathcal{K}_{\mathrm{ann},1}[n_{\chi_{1}}](\xi). \tag*{(41)}

Our calculation includes the excitation of bound χ1\chi_{1}, described by Kcoll,12\mathcal{K}_{\mathrm{coll},12}. It does not make the captured DM convert back and forth many times between χ1\chi_{1} and χ2\chi_{2}, because after three to four excitations χ1\chi_{1} has lost enough kinetic energy that the energy available in collisions with thermal nuclei no longer exceeds δ\delta.

Constraint

To constrain δ\delta we evolve Eq. (37) from nχ1=nχ2=0n_{\chi_{1}} = n_{\chi_{2}} = 0 at t=0t = 0 under continuous capture up to t⊙t_{\odot}, evaluate ΓA\Gamma_{A} at the present day with Eq. (41), and compare it with the upper limit on the total annihilation rate, ΓAIC=2.9×1019 s−1\Gamma_{A}^{\mathrm{IC}} = 2.9 \times10^{19}\ \mathrm{s}^{-1}, set by the 10-yr IceCube limit on the W+W−W^{+}W^{-} channel [12]. The W+W−W^{+}W^{-} channel, which we use alone for the neutrino signal, has a branching ratio of about 0.54 for this benchmark, while ZZZZ has 0.46. Here we use the same numerical setup as in § 1.3, generalized to the two components, with the 29 AGSS09 elements as targets. Figure 10 shows our lower limit on δ\delta (red) together with those of Refs. [96, 107, 108, 109], where the right-pointing arrows mark the allowed region. Our calculation excludes δ≲592 keV\delta\lesssim592\ \mathrm{keV} when the thermal motion of nuclei is included in the inelastic capture rate (filled circle) and δ≲552 keV\delta\lesssim552\ \mathrm{keV} when it is neglected (open circle). Adding Zn, Ge, Pb and U changes both circles by only about 0.1 keV.

Figure 10. Lower limits on δ\delta for the Higgsino model. Filled and open red circles show our results with thermal nuclei and nuclei at rest, respectively, during capture. Right-pointing arrows mark the allowed region. Limits from Refs. [96, 107, 108, 109] are shown for comparison.

Figure 11. Inelastic CcC_c vs. δ\delta. Our red solid and dashed curves use 33 nuclei, the 29 AGSS09 elements plus Zn, Ge, Pb and U, with and without thermal motion, respectively. The red dot-dashed curve includes thermal motion but omits Zn, Ge, Pb and U. Green and blue dashed curves show results for nuclei at rest from Refs. [107, 108] with heavy elements included.

The constraints of Refs. [96, 107, 108, 109] use the classical number equation with 2ΓA≃Cc2\Gamma_A \simeq C_c, which our evolution supports, but assume nuclei at rest in the capture rate. With nuclei at rest, our constraint weakens by about 40 keV (open circle) and becomes comparable to theirs. We show below that the thermal motion of the nuclei allows capture at larger δ\delta even without the additional heavy elements, such as Zn, Ge, Pb and U, that these studies include [96, 107, 108, 109].

Fig. 11 exhibits our inelastic CcC_c as a function of δ\delta for different treatments of the nuclei, with the results of Refs. [107, 108] for reference. With nuclei at rest, our CcC_c (red dashed curve) agrees with that of Ref. [108]. Near our constraint and with thermal motion included, the additional heavy elements barely change CcC_c (red solid and dot-dashed curves). Without thermal motion (red dashed curve), CcC_c is much lower for 500 keV ≲δ≲\lesssim\delta\lesssim 700 keV, because inelastic capture requires the center-of-mass (CoM) energy of the DM-nucleus pair,

Ek,cm=12μχAg2→u→012μχAw2,(42)E_{k,\mathrm{cm}} = \frac{1}{2}\mu_{\chi A}g^2 \xrightarrow{u\to0} \frac{1}{2}\mu_{\chi A}w^2, \tag*{(42)}

to exceed δ\delta, where g=∣w−u∣g = \lvert w-u\rvert is the relative speed, and the rightmost result is the limit of nuclei at rest. For a nucleus at rest, all the energy comes from the incident DM, so capture requires w>2δ/μχAw > \sqrt{2\delta/\mu_{\chi A}}, and heavier targets allow capture at larger δ\delta.

For example, at r=0.01R⊙r = 0.01R_\odot, w∼1410w \sim1410 km s−1^{-1} for a mean halo speed vˉ∞∼300\bar{v}_\infty\sim300 km s−1^{-1}, and the largest δ\delta accessible in scattering on iron at rest is about 550 keV. With thermal motion, a head-on collision with an iron nucleus moving at the mean thermal speed uˉ∼77\bar{u} \sim77 km s−1^{-1} could allow capture up to δ∼610\delta\sim610 keV. For isotropically distributed incident directions, Ek,cm>δE_{k,\mathrm{cm}} > \delta holds for about half of the directions at δ=550\delta= 550 keV and for about 10% at δ=600\delta= 600 keV. The thermal motion of the nuclei alone can therefore raise the δ\delta threshold for inelastic capture without the four heavy elements. Ignoring the thermal motion of the nuclei in CcC_c is also inconsistent with the classical isothermal treatment, in which the DM temperature TχT_\chi results from energy exchange with the same thermal nuclei and is thus closely tied to the solar temperature. With solar nuclei at rest, collisions would eventually drive Tχ→0T_\chi\to0, and nχn_\chi would collapse into the center of the Sun.

Summary

We have followed the evolution of DM captured by stars with the BTE approach, without assuming, as most studies do, that the captured population has already thermalized with the stellar nuclei. When a bound particle completes many orbits between collisions, the distribution can be averaged over the orbital motion, which leads to a master equation that evolves capture, scattering between bound orbits, annihilation and evaporation simultaneously. Integrating it over orbits recovers the classical number evolution equation, with the annihilation and evaporation coefficients set by the evolving distribution.

For DM captured by the Sun, the BTE reproduces the classical evaporation mass and, above a few GeV, the classical annihilation rate when scattering on nuclei is frequent, which supports the classical treatment in this regime. We find that the classical treatment breaks down when this scattering is rare or exchanges little energy per collision. The captured DM then does not thermalize within the solar age and keeps the broad distribution it had at capture, with many particles on orbits that extend far beyond the solar surface. At mχ=100m_\chi= 100 GeV, the classical treatment overestimates the annihilation rate inside the Sun by more than a factor of two for σχp\sigma_{\chi p} below about 10−46 cm210^{-46}\,\mathrm{cm}^2 for SD and 10−49 cm210^{-49}\,\mathrm{cm}^2 for SI scattering, and the suppression reaches many orders of magnitude at larger mχm_\chi and smaller σχp\sigma_{\chi p}. With less DM in the hot core, evaporation is also weaker, and the evaporation mass falls below the classical value.

As an example of DM with several components, we applied the BTE to the inelastic Higgsino model, a possible interpretation of the recent LZ result, and recomputed the constraint on the mass splitting δ\delta. The full evolution supports the relation 2ΓA≃Cc2\Gamma_A \simeq C_c assumed in recent studies [96, 107, 108, 109], and the IceCube limit excludes δ≲592\delta\lesssim592 keV. The constraint is stronger by tens of keV than in these studies, mainly because the thermal motion of solar nuclei in the capture rate allows inelastic capture at larger δ\delta without the inclusion of additional heavy elements.

The solar results show that, when scattering on nuclei is weak, the signals of captured DM should be computed from its evolving distribution rather than from an assumed thermal one. The BTE approach applies whenever bound DM completes many orbits before its orbit changes appreciably, a condition that is not specific to the Sun. One natural extension is DM self-interaction, which adds collisions among the bound particles that compete with scattering on nuclei in setting their distribution [22, 99]. Heavy DM changes its orbit only slightly in each collision, so the same approach can follow its capture and thermalization in WDs, NSs and massive stars [58, 113, 114] once the collision kernels are adapted to these environments. Coupling the BTE with stellar evolution would further replace the static star assumed here with a changing one. With these extensions, the BTE approach could provide the evolving distribution of captured DM, and hence its signals, in the other stellar objects used to search for DM.

Acknowledgments

Y.-H. L. and M.-R. W. acknowledge support of the National Science and Technology Council, Taiwan under Grants No. 115-2112-M-001-037-MY3, No. 111-2628-M-001-003-MY4, No. 115-2112-M-001-054-MY5, and Academia Sinica under Project No. AS-IV-114-M04. Y.-H. L. further acknowledges support from the Shui-Chin Lee Foundation through the Shui-Chin Lee Fellowship. M.-R. W. also acknowledges support of the Physics Division of the National Center for Theoretical Sciences, Taiwan. S. Y. is supported by IBS under the Project code No. IBS-R018-D1 and the Basic Science Research Program through the NRF funded by the Ministry of Education (No. RS-2026-25571203). We acknowledge the computing resources provided by the Academia Sinica Grid-computing Center (ASGC). We also acknowledge the use of Anthropic Claude and OpenAI Codex to optimize the code and manage parallel computing at ASGC.

Scattering kinematics

A collision changes the DM velocity at a fixed position, and the outgoing energy and angular momentum follow from the new velocity. At the collision point, we use an orthonormal basis (e^r,e^t,e^z)(\hat{\mathbf{e}}_r,\hat{\mathbf{e}}_t,\hat{\mathbf{e}}_z), where e^r\hat{\mathbf{e}}_r points radially outward, e^t\hat{\mathbf{e}}_t points along the tangential component of the incident DM velocity v\mathbf{v}, and e^z\hat{\mathbf{e}}_z is perpendicular to both. The velocities of the incident DM and the nucleus are then

v=(vrvt0),u=u(cos⁡θusin⁡θucos⁡ϕusin⁡θusin⁡ϕu)(43)\mathbf{v} = \begin{pmatrix} v_r \\ v_t \\ 0 \end{pmatrix}, \qquad \mathbf{u} = u \begin{pmatrix} \cos\theta_u \\ \sin\theta_u\cos\phi_u \\ \sin\theta_u\sin\phi_u \end{pmatrix} \tag*{(43)}

as shown in the left panel of Figure 12, and dΩu=dcos⁡θudϕud\Omega_u = d\cos\theta_u d\phi_u is the solid angle of u\mathbf{u}. In a spherically symmetric star, the results do not depend on the azimuth $\alpha of the tangential velocity about e^r\hat{\mathbf{e}}_r, and this choice of e^t\hat{\mathbf{e}}_t sets α=0\alpha= 0.

Figure 12. Collision kinematics. Left: the incident DM velocity has radial and tangential components vrv_r and vtv_t, and the thermal velocity u\mathbf{u} of the nucleus has polar angle θu\theta_u and azimuthal angle ϕu\phi_u. Right: elastic scattering rotates the relative velocity from g\mathbf{g} to g′\mathbf{g}', specified by θg′\theta_{g'} and ϕg′\phi_{g'}, while preserving its magnitude.

The relative and CoM velocities are

g=v−u,V=mχv+mAumχ+mA,(44)\mathbf{g} = \mathbf{v} - \mathbf{u}, \qquad\mathbf{V} = \frac{m_\chi\mathbf{v} + m_A \mathbf{u}}{m_\chi+ m_A}, \tag*{(44)}

with elastic scattering leaving V\mathbf{V} and the relative speed g=∣g∣g = \lvert\mathbf{g}\rvert unchanged while rotating g\mathbf{g} to

g′=g(cos⁡θg′sin⁡θg′cos⁡ϕg′sin⁡θg′sin⁡ϕg′),(45)\mathbf{g}' = g \begin{pmatrix} \cos\theta_{g'} \\ \sin\theta_{g'} \cos\phi_{g'} \\ \sin\theta_{g'} \sin\phi_{g'} \end{pmatrix}, \tag*{(45)}

where dΩg′=dcos⁡θg′dϕg′d\Omega_{g'} = d\cos\theta_{g'}d\phi_{g'} is the solid angle of the outgoing relative velocity. The outgoing DM velocity is

vf′=V+mAmχ+mAg′,(46)\mathbf{v}'_f = \mathbf{V} + \frac{m_A}{m_\chi+ m_A}\mathbf{g}', \tag*{(46)}

so the incident velocities and the outgoing direction determine Ef′E'_f and Lf′L'_f,

Ef′=vf22+Φ(r),Lf′=rvf,t′2+vf,z′2.(47)E'_f = \frac{v_f^2}{2} + \Phi(r), \qquad L'_f = r\sqrt{v_{f,t}^{\prime2} + v_{f,z}^{\prime2}}. \tag*{(47)}

Both components perpendicular to e^r\hat{\mathbf{e}}_r contribute to Lf′L'_f, so scattering can change the orbital plane as well as the orbital energy. The momentum transfer is q=mχ(vf′−v)\mathbf{q} = m_\chi(\mathbf{v}'_f - \mathbf{v}), whose squared magnitude is

q2=2μχA2g2(1−cos⁡θcm),(48)q^2 = 2\mu_{\chi A}^2 g^2(1-\cos\theta_{\mathrm{cm}}), \tag*{(48)}

where θcm\theta_{\mathrm{cm}} is the angle between g\mathbf{g} and g′\mathbf{g}', with cos⁡θcm=g⋅g′/g2\cos\theta_{\mathrm{cm}} = \mathbf{g}\cdot\mathbf{g}'/g^2 for g>0g > 0. This momentum transfer enters the nuclear form factor FA(q)F_A(q).

For inelastic scattering, χ1A→χ2A\chi_1 A \to\chi_2 A, we neglect the mass splitting in the inertial masses, and energy conservation in the CoM frame gives

g′2=g2−2δμχA,(49)g^{\prime2} = g^2 - \frac{2\delta}{\mu_{\chi A}}, \tag*{(49)}

which requires μχAg2/2>δ\mu_{\chi A}g^2/2 > \delta, while g′2=g2+2δ/μχAg^{\prime2} = g^2 + 2\delta/\mu_{\chi A} for the reverse process. Eq. (A4) still gives the outgoing DM velocity, now with ∣g′∣=g′\lvert\mathbf{g}'\rvert= g', and the squared momentum transfer becomes

q2=μχA2(g2+g′2−2gg′cos⁡θcm),(50)q^2 = \mu_{\chi A}^2\left(g^2 + g^{\prime2} - 2gg'\cos\theta_{\mathrm{cm}}\right), \tag*{(50)}

with cos⁡θcm=g⋅g′/(gg′)\cos\theta_{\mathrm{cm}} = \mathbf{g}\cdot\mathbf{g}'/(gg'). For g′=gg' = g, the above relation reduces to Eq. (A6). Since the final-state phase space scales with g′g', the factor gg in Eqs. (4) and (22) becomes g′g' for inelastic scattering, and the rate vanishes at threshold δ=μχAg2/2\delta= \mu_{\chi A}g^2/2.

Orbits that cross the solar interior

A bound particle can scatter on solar nuclei only if its orbit enters the Sun, r−<R⊙r_- < R_\odot. Figure 13 shows the three classes of bound orbits, namely orbits inside the Sun (left), orbits crossing its surface (middle) and orbits lying outside it (right). Particles that remain bound after capture or scattering inside the Sun move on orbits of the first two classes and make up the population considered in this paper.

Figure 13. Bound-orbit classes relative to the Sun (orange). From left to right: an interior orbit with r+<R⊙r_{+}<R_{\odot}, a solar-crossing orbit with r−<R⊙<r+r_{-}<R_{\odot}<r_{+}, and an exterior orbit with r−>R⊙r_{-}>R_{\odot}. Green and light-blue segments lie inside and outside the Sun, respectively. r−r_{-} and r+r_{+} mark the pericenter and apocenter. Only in the first two cases can DM scatter with solar nuclei.

At fixed energy, increasing LL raises r−r_- and lowers r+r_+ until they coincide on a circular orbit, r−=r+=rc(E)r_- = r_+ = r_c(E). The angular momentum of this circular orbit, Lc(E)L_c(E), is the largest allowed at that energy. If rc(E)<R⊙r_c(E) < R_\odot, every orbit at that energy has r−<R⊙r_- < R_\odot and crosses the solar interior, although its r+r_+ may lie outside the Sun. If rc(E)>R⊙r_c(E) > R_\odot, decreasing LL moves r−r_- inward until it reaches the solar surface at L=L⊙(E)L = L_\odot(E), as shown in Figure 14. At this grazing point the velocity is purely tangential, so L⊙(E)=R⊙vL_\odot(E) = R_\odot v with E=v2/2+Φ(R⊙)E = v^2/2 + \Phi(R_\odot). Including the case rc(E)≤R⊙r_c(E) \le R_\odot, we obtain

Figure 14. An orbit with a grazing point, rc>R⊙r_c > R_\odot and L=L⊙(E)<Lc(E)L = L_\odot(E) < L_c(E), will not cross the solar interior.

L⊙(E)={Lc(E),rc(E)≤R⊙,R⊙2[E−Φ(R⊙)],rc(E)>R⊙.(51)L_\odot(E) = \begin{cases} L_c(E), & r_c(E) \le R_\odot,\\ R_\odot\sqrt{2[E-\Phi(R_\odot)]}, & r_c(E) > R_\odot. \end{cases} \tag*{(51)}

In the second case, L⊙<LcL_\odot< L_c, and orbits with L⊙<L≤LcL_\odot< L \le L_c pass through the Sun, whereas those with L⊙<L≤LcL_\odot< L \le L_c remain outside it.

Derivations in orbital phase space

In this appendix, we collect the steps that take the BTE from position and velocity space to the orbital phase space of the main text, for the static, spherically symmetric potential assumed there. Each phase-space point (r,v)(\mathbf{r}, \mathbf{v}) lies on a single orbit ξ=(E,L)\boldsymbol{\xi} = (E,L), so integrals over (r,v)(\mathbf{r}, \mathbf{v}) can be organized orbit by orbit. For later use, primed variables denote a phase-space point (r′,v′)(\mathbf{r}',\mathbf{v}') with orbital coordinates ξ′=ξ(r′,v′)\boldsymbol{\xi}' = \boldsymbol{\xi}(\mathbf{r}',\mathbf{v}').

The vanishing advection terms

Multiplying the advection terms of Eq. (1) by Δ(ξ−ξ′)\Delta(\boldsymbol{\xi}-\boldsymbol{\xi}') and integrating over phase space, we obtain

∫d3r′d3v′ Δ(ξ−ξ′)(v′⋅∇r′f−∇r′Φ⋅∇v′f)=−∫d3r′d3v′ f(v′⋅∇r′−∇r′Φ⋅∇v′)Δ(ξ−ξ′)=0.(52)\int d^3r' d^3v'\,\Delta(\boldsymbol{\xi}-\boldsymbol{\xi}')(\mathbf{v}'\cdot\nabla_{\mathbf{r}'}f-\nabla_{\mathbf{r}'}\Phi\cdot\nabla_{\mathbf{v}'}f) = -\int d^3r' d^3v'\,f(\mathbf{v}'\cdot\nabla_{\mathbf{r}'}-\nabla_{\mathbf{r}'}\Phi\cdot\nabla_{\mathbf{v}'})\Delta(\boldsymbol{\xi}-\boldsymbol{\xi}')=0. \tag*{(52)}

The first equality follows from integration by parts with ∇r′⋅v′=0\nabla_{\mathbf{r}'}\cdot\mathbf{v}' = 0 and ∇v′⋅∇r′Φ=0\nabla_{\mathbf{v}'}\cdot\nabla_{\mathbf{r}'}\Phi= 0, and the boundary terms vanish because the collisionless motion conserves EE and does not leave the bound domain. The second equality holds because E′E' and L′L' are constant along the collisionless motion, so the operator in parentheses gives zero when acting on any function of (E′,L′)(E',L'). The projection therefore removes the advection terms without any assumption on f(r′,v′)f(\mathbf{r}',\mathbf{v}').

From (r,v) to orbital phase space

At a local phase-space point (r′,v′)(\mathbf{r}',\mathbf{v}'), the specific energy and angular momentum are

E′=v′22+Φ(r′),L′=∣r′×v′∣(53)E'=\frac{v'^{2}}{2}+\Phi(\mathbf{r}'),\qquad L'=|\mathbf{r}'\times\mathbf{v}'| \tag*{(53)}

with d3v′=vt′dvt′dvr′dαd^{3}v'=v'_{t}dv'_{t}dv'_{r}d\alpha. After integrating over the spatial orientation and α\alpha, we obtain

Δ(ξ−ξ′)d3r′d3v′=2×8π2r′2vt′Δ(ξ−ξ′)dr′dvr′dvt′.(54)\Delta(\boldsymbol{\xi}-\boldsymbol{\xi}')d^{3}r'd^{3}v' =2\times8\pi^{2}r'^{2}v'_{t}\Delta(\boldsymbol{\xi}-\boldsymbol{\xi}')dr'dv'_{r}dv'_{t}. \tag*{(54)}

The factor of two accounts for the inward and outward passages, which contribute equally, while the angular integrations give 4π4\pi from the direction of r′\mathbf{r}' and 2π2\pi from the azimuth α\alpha of v′\mathbf{v}' about it. The delta functions in Δ(ξ−ξ′)\Delta(\boldsymbol{\xi}-\boldsymbol{\xi}') fix the orbital coordinates to ξ\boldsymbol{\xi}, and the remaining Jacobian is

dvr′dvt′=∣∂(vr′,vt′)∂(E′,L′)∣dE′dL′=dE′dL′r′vr′(55)dv'_{r}dv'_{t}=\left|\frac{\partial(v'_{r},v'_{t})}{\partial(E',L')}\right|dE'dL'=\frac{dE'dL'}{r'v'_{r}} \tag*{(55)}

such that

16π2r′2vt′dr′dvr′dvt′=16π2L′dr′vr′(r′,ξ′)dE′dL′.16\pi^{2}r'^{2}v'_{t}dr'dv'_{r}dv'_{t}=16\pi^{2}L'\frac{dr'}{v'_{r}(r',\xi')}dE'dL'.

We then obtain

Δ(ξ−ξ′)d3r′d3v′=16π2L′dr′vr′(r′,ξ′)Δ(ξ−ξ′)dE′dL′(56)\Delta(\boldsymbol{\xi}-\boldsymbol{\xi}')d^{3}r'd^{3}v' =16\pi^{2}L'\frac{dr'}{v'_{r}(r',\xi')}\Delta(\boldsymbol{\xi}-\boldsymbol{\xi}')dE'dL' \tag*{(56)}
dVξ≡dEdL∫d3r′d3v′ Δ(ξ−ξ′)=dEdL∫16π2L′dr′vr′(r′,ξ′)Δ(ξ−ξ′)dE′dL′=8π2Lτr(ξ)dEdL(57)dV_{\xi}\equiv dEdL\int d^{3}r'd^{3}v'\,\Delta(\boldsymbol{\xi}-\boldsymbol{\xi}') =dEdL\int16\pi^{2}L'\frac{dr'}{v'_{r}(r',\xi')}\Delta(\boldsymbol{\xi}-\boldsymbol{\xi}')dE'dL' =8\pi^{2}L\tau_{r}(\xi)dEdL \tag*{(57)}

is the volume element of orbital phase space. Weighting Eq. (56) by the orbit-averaged distribution gives

dEdL∫d3r′d3v′ Δ(ξ−ξ′)f(r′,v′)=dEdL∫16π2L′dr′vr′(r′,ξ′)Δ(ξ−ξ′)f(ξ′)dE′dL′=8π2Lτr(ξ)f(ξ)dEdL=nχ(ξ)dEdL(58)dEdL\int d^{3}r'd^{3}v'\,\Delta(\boldsymbol{\xi}-\boldsymbol{\xi}')f(\mathbf{r}',\mathbf{v}') =dEdL\int16\pi^{2}L'\frac{dr'}{v'_{r}(r',\xi')}\Delta(\boldsymbol{\xi}-\boldsymbol{\xi}')f(\xi')dE'dL' =8\pi^{2}L\tau_{r}(\xi)f(\xi)dEdL=n_{\chi}(\xi)dEdL \tag*{(58)}

where f(ξ)f(\xi) is the distribution averaged over the orbit, that is, over its orientations and over the time spent at each radius. Since the integral covers the whole orbit, replacing ff by its orbit average in the second line does not require any approximation. The orbit-averaging approximation is needed only when ff multiplies a quantity that varies along the orbit, as in the collisional kernel of App. C 4. The number of particles in an orbital interval is then

dNχ=nχ(ξ)dEdL=f(ξ)dVξ.(59)dN_{\chi}=n_{\chi}(\xi)dEdL=f(\xi)dV_{\xi}. \tag*{(59)}

Equivalently,

f(ξ)=dNχdVξ=∫d3r′d3v′ Δ(ξ−ξ′)f(r′,v′)8π2Lτr(ξ)(60)f(\xi)=\frac{dN_{\chi}}{dV_{\xi}}=\frac{\int d^{3}r'd^{3}v'\,\Delta(\boldsymbol{\xi}-\boldsymbol{\xi}')f(\mathbf{r}',\mathbf{v}')}{8\pi^{2}L\tau_{r}(\xi)} \tag*{(60)}

relates the local distribution to its orbital average.

Derivation of the master equation

The phase-space volume of a fixed orbit is time independent. Differentiating Eq. (60) with respect to tt,

substituting Eq. (1) and using Eq. (52) gives

∫d3r′d3v′ Δ(ξ−ξ′)∂f(r′,v′)∂t=∫d3r′d3v′ Δ(ξ−ξ′)C[f](r′,v′).\int d^3r' d^3v'\,\Delta(\xi-\xi')\frac{\partial f(\mathbf{r}',\mathbf{v}')}{\partial t} = \int d^3r' d^3v'\,\Delta(\xi-\xi')\mathcal{C}[f](\mathbf{r}',\mathbf{v}').

Dividing by the orbital volume factor gives

∂f(ξ)∂t=∫d3r′d3v′ Δ(ξ−ξ′)C[f](r′,v′)8π2Lτr(ξ).(61)\frac{\partial f(\xi)}{\partial t} = \frac{\int d^3r' d^3v'\,\Delta(\xi-\xi')\mathcal{C}[f](\mathbf{r}',\mathbf{v}')}{8\pi^2L\tau_r(\xi)}. \tag*{(61)}

Multiplying by 8π2Lτr(ξ)8\pi^2L\tau_r(\xi) and using Eq. (15) then yields

∂nχ(ξ)∂t=∫d3r′d3v′ Δ(ξ−ξ′)C[f](r′,v′)(62)\frac{\partial n_\chi(\xi)}{\partial t} = \int d^3r' d^3v'\,\Delta(\xi-\xi')\mathcal{C}[f](\mathbf{r}',\mathbf{v}') \tag*{(62)}

which is the master equation, Eq. (21).

Derivation of the collisional kernel

Substituting Eq. (5) into the master equation, Eq. (21), gives the gain and loss terms below

∫d3r′d3v′ Δ(ξ−ξ′)Ccoll[f](r′,v′)=∫d3r′d3v′ Δ(ξ−ξ′)∫d3v′′[R(v′′→v′,r′)f(r′,v′′)−R(v′→v′′,r′)f(r′,v′)].(63)\int d^3r' d^3v'\,\Delta(\xi-\xi')\mathcal{C}_{\mathrm{coll}}[f](\mathbf{r}',\mathbf{v}') = \int d^3r' d^3v'\,\Delta(\xi-\xi')\int d^3v''\left[R(\mathbf{v}''\to\mathbf{v}',r')f(\mathbf{r}',\mathbf{v}'')-R(\mathbf{v}'\to\mathbf{v}'',r')f(\mathbf{r}',\mathbf{v}')\right]. \tag*{(63)}

For the loss term, Eq. (56) and the orbit-averaging approximation, Eq. (17), give

∫d3r′d3v′ Δ(ξ−ξ′)f(r′,v′)∫d3v′′R(v′→v′′,r′)=∫16π2L′dr′vr(r′,ξ′)Δ(ξ−ξ′) dE′dL′f(ξ′)∫d3v′′R(v′→v′′,r′).(64)\int d^3r' d^3v'\,\Delta(\xi-\xi')f(\mathbf{r}',\mathbf{v}')\int d^3v''R(\mathbf{v}'\to\mathbf{v}'',r') = \int16\pi^2L'\frac{dr'}{v_r(r',\xi')}\Delta(\xi-\xi')\,dE'dL'f(\xi')\int d^3v''R(\mathbf{v}'\to\mathbf{v}'',r'). \tag*{(64)}

After integrating over the outgoing velocity orientation, the delta function in R(v′→v′′)R(\mathbf{v}'\to\mathbf{v}'') gives

d3v′′ δ(3)(v′′−vf′′)=dE′′dL′′ δ(E′′−Ef′′)δ(L′′−Lf′′)(65)d^3v''\,\delta^{(3)}(\mathbf{v}''-\mathbf{v}''_f)=dE''dL''\,\delta(E''-E''_f)\delta(L''-L''_f) \tag*{(65)}

which is proved in App. C5. The final velocity vf′′\mathbf{v}''_f fixes (Ef′′,Lf′′)(E''_f,L''_f) through the kinematics given in App. A. Therefore

∫d3v′′R(v′→v′′,r′)=∫dE′′dL′′∑A∫dΩu∫0∞du u2fA(r′,u)nA(r′)×σχA4π∫dΩq′gFA2(q)δ(E′′−Ef′′)δ(L′′−Lf′′)(66)\int d^3v''R(\mathbf{v}'\to\mathbf{v}'',r') = \int dE''dL''\sum_A\int d\Omega_u\int_0^\infty du\,u^2f_A(r',u)n_A(r') \times\frac{\sigma_{\chi A}}{4\pi}\int d\Omega_{q'}gF_A^2(q)\delta(E''-E''_f)\delta(L''-L''_f) \tag*{(66)}

Substituting into Eq. (64) and integrating over dE′dL′dE'dL' yields

16π2L∫r−(ξ)r+(ξ)dr′vr(r′,ξ)f(ξ)∫dE′′dL′′∑AnA(r′)∫dΩu∫0∞du u2fA(r′,u)×σχA4π∫dΩq′gFA2(q)δ(E′′−Ef′′)δ(L′′−Lf′′).(67)16\pi^2L\int_{r_-(\xi)}^{r_+(\xi)}\frac{dr'}{v_r(r',\xi)}f(\xi)\int dE''dL''\sum_A n_A(r')\int d\Omega_u\int_0^\infty du\,u^2f_A(r',u) \times\frac{\sigma_{\chi A}}{4\pi}\int d\Omega_{q'}gF_A^2(q)\delta(E''-E''_f)\delta(L''-L''_f). \tag*{(67)}

Relabeling the integration variables (r′,E′′,L′′,Ef′′,Lf′′)→(r,E′,L′,Ef′,Lf′)(r', E'', L'', E_f'', L_f'') \to(r, E', L', E_f', L_f') gives

∫dE′dL′nχ(ξ)×[2τr(ξ)∫r−(ξ)r+(ξ)drvr(r,ξ)×∑AnA(r)∫dΩu∫0∞du u2fA(r,u)σχA4π∫dΩggFA2(q)δ(E′−Ef′)δ(L′−Lf′)]=∫Φ(0)0dE′∫0L⊙(E′)dL′K(ξ→ξ′)nχ(ξ′).(68)\begin{aligned} &\int dE' dL' n_\chi(\xi) \times\left[\frac{2}{\tau_r(\xi)}\int_{r_-(\xi)}^{r_+(\xi)} \frac{dr}{v_r(r,\xi)} \right.\\ &\qquad\qquad\left.\times\sum_A n_A(r)\int d\Omega_u\int_0^\infty du\,u^2 f_A(r,u)\frac{\sigma_{\chi A}}{4\pi}\int d\Omega_g gF_A^2(q)\delta(E'-E_f')\delta(L'-L_f')\right]\\ &= \int_{\Phi(0)}^0 dE'\int_0^{L_\odot(E')}dL' K(\xi\to\xi')n_\chi(\xi'). \tag*{(68)} \end{aligned}

The expression in brackets is K(ξ→ξ′)K(\xi\to\xi') in Eq. (22) and the integration limits select bound destinations. For the gain term,

∫d3r′d3v′ Δ(ξ−ξ′)∫d3v′′R(v′′→v′,r′)f(r′,v′′)=∫d3r′d3v′′f(r′,v′′)∫d3v′R(v′′→v′,r′)Δ(ξ−ξ′)=∫16π2L′′dr′vr(r′,ξ′′)f(ξ′′)dE′′dL′′∫d3v′R(v′′→v′,r′)Δ(ξ−ξ′)=∫dE′′dL′′nχ(ξ′′)×[2τr(ξ′′)∫r−(ξ′′)r+(ξ′′)dr′vr(r′,ξ′′)∫d3v′R(v′′→v′,r′)Δ(ξ−ξ′)].(69)\begin{aligned} \int d^3r' d^3v'\,\Delta(\xi-\xi')\int d^3v''R(v''\to v',r')f(r',v'') \\ &= \int d^3r'd^3v''f(r',v'')\int d^3v'R(v''\to v',r')\Delta(\xi-\xi') \\ &= \int16\pi^2L''\frac{dr'}{v_r(r',\xi'')}f(\xi'')dE''dL''\int d^3v'R(v''\to v',r')\Delta(\xi-\xi') \\ &= \int dE''dL''n_\chi(\xi'')\times\left[\frac{2}{\tau_r(\xi'')}\int_{r_-(\xi'')}^{r_+(\xi'')} \frac{dr'}{v_r(r',\xi'')}\int d^3v'R(v''\to v',r')\Delta(\xi-\xi')\right]. \tag*{(69)} \end{aligned}

The third line uses Eq. (56) for the incident state ξ′′\xi'' and the orbit-averaging approximation, f(r′,v′′)≃f(ξ′′)f(r',v'')\simeq f(\xi''). Applying the identity in Eq. (65), we obtain

∫d3v′R(v′′→v′,r′)Δ(ξ−ξ′)=∫dE′dL′∑AnA(r′)∫dΩu∫0∞du u2fA(r′,u)σχA4π×∫dΩggFA2(q)δ(E′−Ef′)δ(L′−Lf′)Δ(ξ−ξ′)=∑AnA(r′)∫dΩu∫0∞du u2fA(r′,u)σχA4π∫dΩggFA2(q)δ(E−Ef′)δ(L−Lf′).(70)\begin{aligned} \int d^3v'R(v''\to v',r')\Delta(\xi-\xi')&=\int dE'dL'\sum_A n_A(r')\int d\Omega_u\int_0^\infty du\,u^2f_A(r',u)\frac{\sigma_{\chi A}}{4\pi}\\ &\qquad\times\int d\Omega_g gF_A^2(q)\delta(E'-E_f')\delta(L'-L_f')\Delta(\xi-\xi')\\ &=\sum_A n_A(r')\int d\Omega_u\int_0^\infty du\,u^2f_A(r',u)\frac{\sigma_{\chi A}}{4\pi}\int d\Omega_g gF_A^2(q)\delta(E-E_f')\delta(L-L_f'). \tag*{(70)} \end{aligned}

Substituting into Eq. (69) and relabeling (r′,E′′,L′′,ξ′′,Ef′,Lf′)→(r,E′,L′,ξ′,Ef,Lf)(r', E'', L'', \xi'', E_f', L_f')\to(r, E', L', \xi', E_f, L_f) gives

∫dE′dL′nχ(ξ′)×[2τr(ξ′)∫r−(ξ′)r+(ξ′)drvr(r,ξ′)×∑AnA(r)∫dΩu∫0∞du u2fA(r,u)σχA4π∫dΩggFA2(q)δ(E−Ef)δ(L−Lf)]=∫Φ(0)0dE′∫0L⊙(E′)dL′K(ξ′→ξ)nχ(ξ′).(71)\begin{aligned} \int dE'dL'n_\chi(\xi')\times\left[\frac{2}{\tau_r(\xi')}\int_{r_-(\xi')}^{r_+(\xi')} \frac{dr}{v_r(r,\xi')}\right.\\ \left.\times\sum_A n_A(r)\int d\Omega_u\int_0^\infty du\,u^2f_A(r,u)\frac{\sigma_{\chi A}}{4\pi}\int d\Omega_g gF_A^2(q)\delta(E-E_f)\delta(L-L_f)\right]\\ =\int_{\Phi(0)}^0dE'\int_0^{L_\odot(E')}dL'K(\xi'\to\xi)n_\chi(\xi'). \tag*{(71)} \end{aligned}

Combining Eqs. (68) and (C20), we obtain the collisional kernel

Kcoll[nχ](ξ)=∫Φ(0)0dE′∫0L⊙(E′)dL′[K(ξ′→ξ)nχ(ξ′)−K(ξ→ξ′)nχ(ξ)](72)K_{\mathrm{coll}}[n_\chi](\xi)=\int_{\Phi(0)}^0dE'\int_0^{L_\odot(E')}dL'\left[K(\xi'\to\xi)n_\chi(\xi')-K(\xi\to\xi')n_\chi(\xi)\right] \tag*{(72)}

which reproduces Eq. (23), and the evaporation and annihilation kernels follow from the same projection.

Proof of Eq. (C14)

For a test function hh, the defining property of the delta function is

∫d3v δ(3)(v−v0)h(v)=h(v0)(73)\int d^3v\,\delta^{(3)}(\mathbf{v}-\mathbf{v}_0)h(\mathbf{v})=h(\mathbf{v}_0) \tag*{(73)}

Figure 15. nχ(r,t)n_\chi(r,t) at t⊙t_\odot for SD scattering with σχp=10−44 cm2\sigma_{\chi p}=10^{-44}\,\mathrm{cm}^2 and DM masses labeled 0.5 GeV (green) and 5 GeV (red). Solid and dotted curves show the BTE and classical results. Dot-dashed lines mark 3rth3r_{\mathrm{th}} for each mass and the gray dashed line marks the solar surface.

Using d3v=vt dvt dvr dαd^3v=v_t\,dv_t\,dv_r\,d\alpha, this becomes

∫dvt dvr dα vtδ(3)(v−v0)h(vt,vr,α)=h(vt,0,vr,0,α0)\int dv_t\,dv_r\,d\alpha\,v_t\delta^{(3)}(\mathbf{v}-\mathbf{v}_0)h(v_t,v_r,\alpha)=h(v_{t,0},v_{r,0},\alpha_0)

where the right-hand side can also be written as

h(vt,0,vr,0,α0)=∫dvt dvr dα δ(vt−vt,0)δ(vr−vr,0)δ(α−α0)h(vt,vr,α).h(v_{t,0},v_{r,0},\alpha_0)=\int dv_t\,dv_r\,d\alpha\,\delta(v_t-v_{t,0})\delta(v_r-v_{r,0})\delta(\alpha-\alpha_0)h(v_t,v_r,\alpha).

Comparing these expressions gives

δ(3)(v−v0)=1vtδ(vt−vt,0)δ(vr−vr,0)δ(α−α0).(74)\delta^{(3)}(\mathbf{v}-\mathbf{v}_0)=\frac{1}{v_t}\delta(v_t-v_{t,0})\delta(v_r-v_{r,0})\delta(\alpha-\alpha_0). \tag*{(74)}

From Eq. (C4), we have d3v=vt dE dL dα/(rvr)d^3v=v_t\,dE\,dL\,d\alpha/(r v_r), giving

d3vδ(3)(v−v0)=dE dL dαrvrδ(vt−vt,0)δ(vr−vr,0)δ(α−α0).(75)d^3v\delta^{(3)}(\mathbf{v}-\mathbf{v}_0)=\frac{dE\,dL\,d\alpha}{r v_r}\delta(v_t-v_{t,0})\delta(v_r-v_{r,0})\delta(\alpha-\alpha_0). \tag*{(75)}

We also have

δ(vt−vt,0)=δ(Lr−L0r)=rδ(L−L0)\delta(v_t-v_{t,0})=\delta\left(\frac{L}{r}-\frac{L_0}{r}\right)=r\delta(L-L_0)

and

E−E0=(vr2+vt2)−(vr,02+vt,02)2=vr2−vr,022,E-E_0=\frac{(v_r^2+v_t^2)-(v_{r,0}^2+v_{t,0}^2)}{2}=\frac{v_r^2-v_{r,0}^2}{2},

where we have used the constraint vt=vt,0v_t=v_{t,0} from δ(vt−vt,0)\delta(v_t-v_{t,0}). The delta function selects the radial direction of v0\mathbf{v}_0, so

δ(E−E0)=δ(vr2−vr,022)=δ(vr−vr,0)vr.\delta(E-E_0)=\delta\left(\frac{v_r^2-v_{r,0}^2}{2}\right)=\frac{\delta(v_r-v_{r,0})}{v_r}.

Finally, we obtain

d3vδ(3)(v−v0)=δ(E−E0)δ(L−L0)dE dL×∫δ(α−α0)dα=δ(E−E0)δ(L−L0)dE dL(76)\begin{aligned} d^3v\delta^{(3)}(\mathbf{v}-\mathbf{v}_0)&=\delta(E-E_0)\delta(L-L_0)dE\,dL\\ &\quad\times\int\delta(\alpha-\alpha_0)d\alpha\\ &=\delta(E-E_0)\delta(L-L_0)dE\,dL \tag*{(76)} \end{aligned}

which is Eq. (C14) with the direction of velocity integrated out.

n(r) tail at r>rth

Fig. 15 illustrates the nχn_\chi tail responsible for the evaporation behavior discussed in Sec. V D. For both masses, the BTE and classical radial densities agree closely within 3rth3r_{\mathrm{th}}. At mχ=0.5m_\chi=0.5 GeV they remain comparable out to the solar surface, consistent with the modest change in the evaporation coefficient ratio between the two orbital selections in Fig. 8. At mχ=5m_\chi=5 GeV, the profiles separate beyond roughly 0.3R⊙0.3R_\odot, where freshly captured, weakly bound particles sustain a tail that can still evaporate while the classical density is exponentially suppressed, which explains the large coefficient ratio in Fig. 8.

Anisotropy of rarely scattering DM

When collisions are rare, the orbital planes of captured DM keep a memory of the incident directions of halo DM. To estimate the largest possible effect, we assume that the initial anisotropy is never erased. We sample the positions and velocities of captured DM with a Monte Carlo simulation of the initial capture, based on the scattering kinematics of App. A. From this population, we compute the annihilation rate ΓAaniso\Gamma_{\mathrm{A}}^{\mathrm{aniso}} and compare it with ΓAiso\Gamma_{\mathrm{A}}^{\mathrm{iso}} obtained from the spherically symmetric density of Eq. (19). Figure 16 shows the density for SI scattering at mχ=10m_{\chi}=10 GeV, which is elongated along the solar motion. Figure 17 shows that ΓAaniso/ΓAiso\Gamma_{\mathrm{A}}^{\mathrm{aniso}}/\Gamma_{\mathrm{A}}^{\mathrm{iso}} stays within 4% of unity for all masses considered in this paper, and SD scattering gives the same result.

Figure 16. Density of initially captured DM for SI scattering at mχ=10m_\chi=10 GeV without subsequent collisions, normalized to its maximum. The black arrow indicates the motion of the Sun, the dashed circle marks the solar surface, and the contours mark 10−110^{-1}, 10−210^{-2} and 10−310^{-3}.

Figure 17. ΓAaniso/ΓAiso\Gamma_{\mathrm{A}}^{\mathrm{aniso}}/\Gamma_{\mathrm{A}}^{\mathrm{iso}} for SI and SD scattering. Both rates are integrated within 10R⊙10R_{\odot} (solid) and R⊙R_{\odot} (dotted).

Grid in orbital phase space

We evolve nχ(ξ)n_{\chi}(\xi) on 192 cells in EE and 48 cells in ℓ≡L/L⊙(E)\ell\equiv L/L_{\odot}(E), where L⊙(E)L_{\odot}(E) is given by Eq. (51), so that ℓ=0\ell=0 for radial orbits and ℓ=1\ell=1 at the largest angular momentum. The grid depends only on mχm_{\chi}, and all cross sections share it. Once the grid is set, we integrate Eq. (22) over the cells to obtain the rate of transfer between every pair of cells, which redistributes the bound population through Eq. (23), and the evaporation rate of each cell. Since these rates and the capture rate of each cell, obtained from Eq. (26), are proportional to σχp\sigma_{\chi p}, we compute them once for each mass and type of scattering, and the master equation becomes a set of coupled equations for the cell populations.

The energy range runs from the bottom of the potential, Φ(0)\Phi(0), to orbits whose radial turning point lies near 1100R⊙1100R_{\odot}, close to the orbit of Jupiter, and we split it into three regions. In terms of ϵ≡E−Φ(0)\epsilon\equiv E-\Phi(0), the first region covers 0≤ϵ≤10ϵ00 \leq\epsilon\leq10\epsilon_{0} with 120 cells, where ϵ0≡T⊙(0)/mχ\epsilon_{0} \equiv T_{\odot}(0)/m_{\chi} is the energy scale of DM thermalized with the solar core. Their edges are equally spaced in ln⁡(ϵ+ϵ0)\ln(\epsilon+\epsilon_{0}), so the cells are nearly uniform for ϵ≪ϵ0\epsilon\ll\epsilon_{0} and widen in proportion to ϵ\epsilon above it. Since ϵ0\epsilon_{0} scales as 1/mχ1/m_{\chi}, thermalized DM occupies a thinner layer for heavier DM, and we rebuild this region for each mass so that the same number of cells always resolves the thermal core. The second region contains 60 cells whose widths grow by a constant factor up to a turning radius near 12R⊙12R_{\odot}, which covers DM that cools after capture. The third region contains 12 cells with mass-independent edges between 12R⊙12R_{\odot} and 1100R⊙1100R_{\odot}, where newly captured DM resides. Bound DM on wider orbits beyond this range is kept in a reservoir, which counts as bound but does not scatter or annihilate. For mχ≲3m_{\chi} \lesssim3 GeV, 10ϵ010\epsilon_{0} is comparable to the depth of the potential, and the first region ends at a lower energy so that the cells of the second region have equal widths.

We space the 48 cells in ℓ\ell uniformly instead of logarithmically, because bound DM spreads over the whole range of ℓ\ell and dNχ/dℓdN_{\chi}/d\ell vanishes as ℓ→0\ell\to0, so logarithmic cells would crowd the sparsely populated radial orbits. We use this grid because a finer energy grid, at a higher computational cost, changes ΓA\Gamma_{\mathrm{A}} by at most a factor of O(1)\mathcal{O}(1), which is small compared with the suppression in Figure 5 and Figure 6.

References

References

  1. [1]N. Aghanim et al. (Planck), Astron. Astrophys. 641, A6 (2020), [Erratum: Astron.Astrophys. 652, C4 (2021)], arXiv:1807.06209 [astro-ph.CO].DOI
  2. [2]M. W. Goodman and E. Witten, Phys. Rev. D 31, 3059 (1985).
  3. [3]R. K. Leane and J. Tong, JCAP 12, 031 (2024), arXiv:2405.05312 [hep-ph].DOI
  4. [4]K. Freese, Phys. Lett. B 167, 295 (1986).
  5. [5]G.-L. Lin, Y.-H. Lin, and F.-F. Lee, Phys. Rev. D 91, 033002 (2015), arXiv:1409.3094 [hep-ph].DOI
  6. [6]J. Silk, K. A. Olive, and M. Srednicki, Phys. Rev. Lett. 55, 257 (1985).
  7. [7]K. Griest and D. Seckel, Nucl. Phys. B 283, 681 (1987), [Erratum: Nucl.Phys.B 296, 1034–1036 (1988)].
  8. [8]K. Choi et al. (Super-Kamiokande), Phys. Rev. Lett. 114, 141301 (2015), arXiv:1503.04858 [hep-ex].arxiv.org/abs/1503.04858
  9. [9]S. Adrian-Martinez et al. (ANTARES), Phys. Lett. B 759, 69 (2016), arXiv:1603.02228 [astro-ph.HE].DOI
  10. [10]M. G. Aartsen et al. (IceCube), JCAP 04, 022 (2016), arXiv:1601.00655 [hep-ph].
  11. [11]M. G. Aartsen et al. (IceCube), Eur. Phys. J. C 77, 146 (2017), [Erratum: Eur.Phys.J.C 79, 214 (2019)], arXiv:1612.05949 [astro-ph.HE].DOI
  12. [12]R. Abbasi et al. (IceCube), (2025), arXiv:2507.08457 [hep-ex].arxiv.org/abs/2507.08457
  13. [13]B. Batell, M. Pospelov, A. Ritz, and Y. Shang, Phys. Rev. D 81, 075004 (2010), arXiv:0910.1567 [hep-ph].DOI
  14. [14]P. Schuster, N. Toro, and I. Yavin, Phys. Rev. D 81, 016002 (2010), arXiv:0910.1602 [hep-ph].DOI
  15. [15]N. F. Bell and K. Petraki, JCAP 04, 003 (2011), arXiv:1102.2958 [hep-ph]
  16. [16]R. K. Leane, K. C. Y. Ng, and J. F. Beacom, Phys. Rev. D 95, 123016 (2017), arXiv:1703.04629 [astro-ph.HE].DOI
  17. [17]A. Albert et al. (HAWC), Phys. Rev. D 98, 123012 (2018), arXiv:1808.05624 [hep-ph].DOI
  18. [18]M. N. Mazziotta, F. Loparco, D. Serini, A. Cuoco, P. De La Torre Luque, F. Gargano, and M. Gustafsson, Phys. Rev. D 102, 022003 (2020), arXiv:2006.04114 [astro-ph.HE].DOI
  19. [19]D. Bose, T. N. Maity, and T. S. Ray, Phys. Rev. D 105, 123013 (2022), arXiv:2212.08826 [hep-ph].
  20. [20]R. K. Leane, T. Linden, P. Mukhopadhyay, and N. Toro, Phys. Rev. D 103, 075030 (2021), arXiv:2101.12213 [astro-ph.HE].
  21. [21]P. Bhattacharjee and F. Calore, Particles 7, 489 (2024), arXiv:2311.18455 [astro-ph.HE].
  22. [22]C.-S. Chen, G.-L. Lin, and Y.-H. Lin, Phys. Dark Univ. 14, 35 (2016), arXiv:1508.05263 [hep-ph].arxiv.org/abs/1508.05263
  23. [23]P. Salati and J. Silk, Astrophys. J. 338, 24 (1989).
  24. [24]M. Fairbairn, P. Scott, and J. Edsjo, Phys. Rev. D 77, 047301 (2008), arXiv:0710.3396 [astro-ph].DOI
  25. [25]P. Scott, M. Fairbairn, and J. Edsjo, Mon. Not. Roy. Astron. Soc. 394, 82 (2009), arXiv:0809.1871 [astro-ph].DOI
  26. [26]J. Casanellas and I. Lopes, Astrophys. J. 705, 135 (2009), arXiv:0909.1971 [astro-ph.CO].DOI
  27. [27]I. John, R. K. Leane, and T. Linden, Phys. Rev. D 109, 123041 (2024), arXiv:2311.16228 [astro-ph.HE].arxiv.org/abs/2311.16228
  28. [28]J. Lopes and I. Lopes, Astron. Astrophys. 651, A101 (2021), arXiv:2107.13885 [astro-ph.SR].DOI
  29. [29]H. Hong and A. C. Vincent, Phys. Rev. D 110, 103011 (2024), arXiv:2407.08773 [hep-ph].arxiv.org/abs/2407.08773
  30. [30]S. Ganguly, M. He, C. S. Shin, O. Straniero, and S. Yun, Phys. Rev. D 113, 075038 (2026), arXiv:2509.03388 [hep-ph].DOI
  31. [31]D. Spolyar, K. Freese, and P. Gondolo, Phys. Rev. Lett. 100, 051101 (2008), arXiv:0705.00521 [astro-ph].
  32. [32]F. Iocco, Astrophys. J. Lett. 677, L1 (2008), arXiv:0802.0941 [astro-ph].DOI
  33. [33]K. Freese, D. Spolyar, and A. Aguirre, JCAP 11, 014 (2008), arXiv:0802.1724 [astro-ph].DOI
  34. [34]F. Iocco, A. Bressan, E. Ripamonti, R. Schneider, A. Ferrara, and P. Marigo, Mon. Not. Roy. Astron. Soc. 390, 1655 (2008), arXiv:0805.4016 [astro-ph].DOI
  35. [35]M. Taoso, G. Bertone, G. Meynet, and S. Ekstrom, Phys. Rev. D 78, 123510 (2008), arXiv:0806.2681 [astro-ph].DOI
  36. [36]S. Sivertsson and P. Gondolo, Astrophys. J. 729, 51 (2011), arXiv:1006.0025 [astro-ph.CO].DOI
  37. [37]C. Ilie, C. Levy, P. Paliwa, and S. Zhang, Phys. Rev. D 104, 123031 (2021), arXiv:2009.11474 [astro-ph.CO].DOI
  38. [38]C. Kouvaris, Phys. Rev. D 77, 023006 (2008), arXiv:0708.2362 [astro-ph].DOI
  39. [39]A. de Lavallaz and M. Fairbairn, Phys. Rev. D 81, 123521 (2010), arXiv:1004.0629 [astro-ph.GA].DOI
  40. [40]C. Kouvaris and P. Tinyakov, Phys. Rev. D 82, 063531 (2010), arXiv:1004.0586 [astro-ph.GA].DOI
  41. [41]M. Baryakhtar, J. Bramante, S. W. Li, T. Linden, and N. Raj, Phys. Rev. Lett. 119, 131801 (2017), arXiv:1704.01577 [hep-ph].DOI
  42. [42]N. Raj, P. Tanedo, and H.-B. Yu, Phys. Rev. D 97, 043006 (2018), arXiv:1707.09442 [hep-ph].DOI
  43. [43]N. F. Bell, G. Busoni, and S. Robles, JCAP 09, 018 (2018), arXiv:1807.02840 [hep-ph].DOI
  44. [44]C.-S. Chen and Y.-H. Lin, JHEP 08, 069 (2018), arXiv:1804.03409 [hep-ph].DOI
  45. [45]J. F. Acevedo, J. Bramante, R. K. Leane, and N. Raj, JCAP 03, 038 (2020), arXiv:1911.06334 [hep-ph].DOI
  46. [46]G.-L. Lin and Y.-H. Lin, Phys. Rev. D 104, 063021 (2021), arXiv:2102.11151 [hep-ph].DOI
  47. [47]I. Goldman and S. Nussinov, Phys. Rev. D 40, 3221 (1989).
  48. [48]C. Kouvaris and P. Tinyakov, Phys. Rev. Lett. 107, 091301 (2011), arXiv:1104.0382 [astro-ph.CO].DOI
  49. [49]S. D. McDermott, H.-B. Yu, and K. M. Zurek, Phys. Rev. D 85, 023519 (2012), arXiv:1103.5472 [hep-ph].DOI
  50. [50]N. F. Bell, A. Melatos, and K. Petraki, Phys. Rev. D 87, 123507 (2013), arXiv:1301.6811 [hep-ph].DOI
  51. [51]R. Garani, Y. Gelonini, and T. Hambye, JCAP 05, 035 (2019), arXiv:1812.08773 [hep-ph].DOI
  52. [52]G.-L. Lin and Y.-H. Lin, JCAP 08, 022 (2020), arXiv:2004.05312 [hep-ph].DOI
  53. [53]I. V. Moskalenko and L. L. Wai, Astrophys. J. Lett. 659, L29 (2007), arXiv:astro-ph/0702654.DOI
  54. [54]G. Bertone and M. Fairbairn, Phys. Rev. D 77, 043515 (2008), arXiv:0709.1485 [astro-ph].DOI
  55. [55]T. J. Hurst, A. R. Zentner, A. Natarajan, and C. Badenes, Phys. Rev. D 91, 103514 (2015), arXiv:1410.3925 [astro-ph.CO].DOI
  56. [56]N. F. Bell, G. Busoni, M. E. Ramirez-Quezada, S. Robles, and M. Virgato, JCAP 10, 083 (2021), arXiv:2104.14367 [hep-ph].DOI
  57. [57]M. McCullough and M. Fairbairn, Phys. Rev. D 81, 083520 (2010), arXiv:1001.2737 [hep-ph].DOI
  58. [58]N. F. Bell, G. Busoni, S. Robles, and M. Virgato, JCAP 07, 051 (2024), arXiv:2404.16272 [hep-ph].DOI
  59. [59]J. F. Acevedo, R. K. Leane, and L. Santos-Olmsted, JCAP 03, 042 (2024), arXiv:2309.10843 [hep-ph].DOI
  60. [60]W. H. Press and D. N. Spergel, Astrophys. J. 296, 679 (1985).
  61. [61]A. Gould, Astrophys. J. 321, 571 (1987).
  62. [62]A. Gould, Astrophys. J. 387, 21 (1992).
  63. [63]J. Bramante, A. Delgado, and A. A. Martin, Phys. Rev. D 96, 063002 (2017), arXiv:1703.04043 [hep-ph].DOI
  64. [64]B. Dasgupta, A. Gupta, and A. Ray, JCAP 08, 018 (2019), arXiv:1906.04204 [hep-ph].DOI
  65. [65]C. Ilie and C. Levy, Phys. Rev. D 104, 083033 (2021), arXiv:2105.09765 [astro-ph.CO].DOI
  66. [66]Z.-L. Liang and Y.-L. Wu, Phys. Rev. D 89, 013010 (2014), arXiv:1308.5897 [hep-ph].DOI
  67. [67]R. Catena and A. B. Schwetz, JCAP 04, 042 (2015), arXiv:1501.03729 [hep-ph].DOI
  68. [68]B. Dasgupta, A. Gupta, and A. Ray, JCAP 10, 023 (2020), arXiv:2006.10773 [hep-ph].DOI
  69. [69]T. Bringmann, J. Edsjo, P. Gondolo, P. Ullio, and L. Bergström, JCAP 07, 033 (2018), arXiv:1802.03399 [hep-ph].DOI
  70. [70]N. A. Kozar, A. Caddell, L. Fraser-Leach, P. Scott, and A. C. Vincent, in Tools for High Energy Physics and Cosmology (2021) arXiv:2105.06810 [hep-ph].
  71. [71]R. K. Leane and J. Smirnov, JCAP 12, 040 (2023), arXiv:2309.00669 [hep-ph].arxiv.org/abs/2309.00669
  72. [72]G. Busoni, A. De Simone, and W.-C. Huang, JCAP 07, 010 (2013), arXiv:1305.1817 [hep-ph].DOI
  73. [73]R. Garani and S. Palomares-Ruiz, JCAP 05, 007 (2017) [Erratum: JCAP 08, E03 (2025)], arXiv:1702.02768 [hep-ph].DOI
  74. [74]G. Busoni, A. De Simone, P. Scott, and A. C. Vincent, JCAP 10, 037 (2017), arXiv:1703.07784 [hep-ph].DOI
  75. [75]R. Garani and S. Palomares-Ruiz, JCAP 05, 042 (2022), arXiv:2104.12757 [hep-ph].DOI
  76. [76]A. H. G. Peter, Phys. Rev. D 79, 103532 (2009), arXiv:0902.1347 [astro-ph.HE].DOI
  77. [77]A. Widmark, JCAP 05, 046 (2017), arXiv:1703.06878 [hep-ph].DOI
  78. [78]S. Nussinov, L.-T. Wang, and I. Yavin, JCAP 08, 037 (2009), arXiv:0905.1333 [hep-ph].DOI
  79. [79]A. Menon, R. Morris, A. Pierce, and N. Weiner, Phys. Rev. D 82, 015011 (2010), arXiv:0905.1847 [hep-ph].DOI
  80. [80]M. Blennow, S. Clementz, and J. Herrero-Garcia, Eur. Phys. J. C 78, 386 (2018), [Erratum: Eur.Phys.J.C 79, 407 (2019)], arXiv:1802.06880 [hep-ph].DOI
  81. [81]A. Gould, Astrophys. J. 321, 560 (1987).
  82. [82]D. N. Spergel and W. H. Press, Astrophys. J. 294, 663 (1985).
  83. [83]A. Gould and G. Raffelt, Astrophys. J. 352, 654 (1990).
  84. [84]A. Gould and G. Raffelt, Astrophys. J. 352, 669 (1990).
  85. [85]A. C. Vincent and P. Scott, JCAP 04, 019 (2014), arXiv:1311.2074 [astro-ph.CO].DOI
  86. [86]A. C. Vincent, A. Serenelli, and P. Scott, JCAP 08, 040 (2015), arXiv:1504.04378 [hep-ph].
  87. [87]A. C. Vincent, P. Scott, and A. Serenelli, JCAP 11, 007 (2016), arXiv:1605.06502 [hep-ph].
  88. [88]M. Nauenberg, Phys. Rev. D 36, 1080 (1987).
  89. [89]Z.-L. Liang, Y.-L. Wu, Z.-Q. Yang, and Y.-F. Zhou, JCAP 09, 018 (2016), arXiv:1606.02157 [hep-ph].DOI
  90. [90]Z.-L. Liang, Y.-L. Tang, and Z.-Q. Yang, JCAP 10, 035 (2018), arXiv:1802.01005 [hep-ph].DOI
  91. [91]H. Banks, S. Ansari, A. C. Vincent, and P. Scott, JCAP 04, 002 (2022), arXiv:2111.06895 [hep-ph].DOI
  92. [92]H. Banks, S. Beram, R. Reid, and A. C. Vincent, JCAP 06, 047 (2025), arXiv:2412.14342 [hep-ph].arxiv.org/abs/2412.14342
  93. [93]M. Asplund, N. Grevesse, A. J. Sauval, and P. Scott, Ann. Rev. Astron. Astrophys. 47, 481 (2009), arXiv:0909.0948 [astro-ph.SR].DOI
  94. [94]A. Serenelli, S. Basu, J. W. Ferguson, and M. Asplund, Astrophys. J. Lett. 705, L123 (2009), arXiv:0909.2668 [astro-ph.SR].DOI
  95. [95]C.-S. Chen and Y.-H. Lin, JHEP 04, 074 (2018), arXiv:1802.06956 [hep-ph].DOI
  96. [96]L. Wu, Y. Zhang, and B. Zhu, (2026), arXiv:2609.01590 [hep-ph].arxiv.org/abs/2609.01590
  97. [97]M. Di Mauro, (2026), arXiv:2609.02608 [hep-ph].arxiv.org/abs/2609.02608
  98. [98]D. S. Akerib et al., (2026), arXiv:2609.02823 [hep-ex].arxiv.org/abs/2609.02823
  99. [99]A. R. Zentner, Phys. Rev. D 80, 063501 (2009), arXiv:0907.3448 [astro-ph.HE].DOI
  100. [100]P. Virtanen et al., Nature Meth. 17, 261 (2020), arXiv:1907.10121 [cs.MS].DOI
  101. [101]M. C. Smith et al., Mon. Not. Roy. Astron. Soc. 379, 755 (2007), arXiv:astro-ph/0611671.
  102. [102]R. K. Leane and J. Smirnov, JCAP 10, 057 (2023), arXiv:2209.09834 [hep-ph].arxiv.org/abs/2209.09834
  103. [103]J. Binney and S. Tremaine, Galactic Dynamics, 2nd ed. (Princeton University Press, Princeton, 2008).
  104. [104]C. H. de Lima, (2026), arXiv:2609.05204 [hep-ph].arxiv.org/abs/2609.05204
  105. [105]Z.-T. Fan, H.-J. He, Y.-C. Wang, and Y. Zhao, (2026), arXiv:2609.10491 [hep-ph].arxiv.org/abs/2609.10491
  106. [106]H. Okada, Y. Shigekami, and J.-J. Wu, (2026), arXiv:2609.13038 [hep-ph].arxiv.org/abs/2609.13038
  107. [107]M. Pospelov and H. Ramani, (2026), arXiv:2609.02775 [hep-ph].arxiv.org/abs/2609.02775
  108. [108]M. Di Mauro and H. Shaikh, (2026), arXiv:2609.06760 [hep-ph].arxiv.org/abs/2609.06760
  109. [109]D. Bose et al., (2026), arXiv:2609.07807 [hep-ph].arxiv.org/abs/2609.07807
  110. [110]M. Di Mauro, (2026), arXiv:2609.19174 [hep-ph].arxiv.org/abs/2609.19174
  111. [111]D. Bose, T. T. Q. Nguyen, A. K. Saha, N. Raj, T. N. Maity, and R. Laha, (2026), arXiv:2609.21823 [hep-ph].arxiv.org/abs/2609.21823
  112. [112]R. Krall and M. Reece, Chin. Phys. C 42, 043105 (2018), arXiv:1705.04843 [hep-ph].DOI
  113. [113]N. F. Bell, G. Busoni, S. Robles, and M. Virgado, JCAP 04, 006 (2024), arXiv:2312.11892 [hep-ph].DOI
  114. [114]S. Robles, W. Tangarife, and G. Busoni, JCAP 03, 059 (2026), arXiv:2512.22727 [hep-ph].DOI

Paper details

vlnc:W_2tbrrxwx
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

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

  • Physics and Astronomy

Anthropic Claude, Codex

View statement
Looking for related papers…
Contents