WorksPhysics and Astronomy
Semi-Analytic Predictions for Strong-Lensing Perturbers in Cold and Self-Interacting Dark Matter
Abstract
Strong gravitational lensing is sensitive to the abundance and internal structure of dark matter halos, providing a probe of gravothermal evolution in self-interacting dark matter (SIDM). We combine semi-analytic cold dark matter (CDM) subhalo histories with a model for gravothermal evolution in SIDM to generate CDM and SIDM subhalo populations across SIDM parameter space. We predict their projected enclosed masses and deflection-based projected density slopes and compare them with three perturbers inferred through gravitational imaging in PJ011646, SDSS J0946+1006, and JVAS B1938+666. For SIDM model parameters cm/g and km/s, SIDM subhalos reproduce all three perturbers more often than CDM subhalos. For PJ011646, we find ~0.6 cored SIDM subhalos per host matching the perturber, compared with ~0.002 in CDM, while core-collapsed subhalos are too steep. For SDSS J0946+1006, we find ~3 core-collapsed and ~0.004 cored SIDM subhalos per host, compared with none in 900 CDM realizations, although baryons, neglected here, may matter for such massive subhalos. For "detection V" in JVAS B1938+666, we find ~220 core-collapsed and ~20 cored SIDM subhalos per host within the posterior region, compared with ~0.2 in CDM; the CDM candidates require exceptionally early accretion, at median redshift . Simultaneously reproducing all three perturbers requires km/s, so that the massive (~) progenitors of SDSS J0946+1006 candidates can collapse, and is achieved for cm/g in the parameter sets studied. Our results show how semi-analytic models can quantify how often subhalos reproduce individual perturbers in different dark matter models, and our modular framework can be readily updated as models of tidal stripping and late-stage gravothermal evolution in SIDM improve. (abridged)
Introduction
The particle nature of dark matter remains an open question, despite the success of the standard cold dark matter (CDM) paradigm in describing large-scale structure, the cosmic microwave background, and galaxy clustering [1, 2, 3]. In this paradigm, dark matter particles are cold and collisionless, and structure forms hierarchically: small dark matter halos collapse earlier and form larger halos through sequential mergers [4, 5, 6]. This leads to characteristic predictions for both the abundance and internal structure of dark matter halos, including a rapidly increasing abundance toward lower halo masses, a relationship between halo mass and concentration, and universal density profiles [7, 8, 9]. This process leaves behind a large population of bound subhalos inside galaxy- and group-scale host halos, with typical host masses of and , respectively [10, 11]. For both field halos and subhalos with masses below , inefficient gas cooling, reionization, and feedback strongly suppress gas retention and star formation, rendering these halos invisible to traditional luminous tracers [10, 12, 13, 14]. As a consequence, CDM has not been robustly tested on these mass scales, and alternative hypotheses for the particle nature of dark matter can modify both the abundance and internal structure of these invisible halos. For example, warm dark matter (WDM) can decouple from the primordial plasma while still relativistic, producing a free-streaming scale that suppresses small-scale density fluctuations, thereby reducing the abundance and concentrations of low-mass halos [15, 16, 17]. In fuzzy dark matter (FDM), the wave-like behavior of ultralight particles produces central solitonic cores in dark matter halos while suppressing their abundance at low masses [18]. In self-interacting dark matter (SIDM), the focus of this work, scattering between dark matter particles transports heat within halos. This initially produces lower-density cores, which can subsequently collapse and develop high central densities [19]. Measuring the abundance and internal structure of dark matter halos below would therefore provide a direct test of structure formation across different dark matter models.
Strong gravitational lensing can probe the abundance and density profiles of both field halos and subhalos through their gravitational effects on the light of background sources, without requiring these halos themselves to host stars or gas. Halos projected near lensed images can perturb the lensing potential and produce localized distortions in observations [20]. Flux-ratio anomalies of multiply imaged quasars can be used to statistically constrain the abundance and internal structure of low-mass halos [21, 22, 23]. In addition, gravitational imaging of extended lensed arcs and Einstein rings can reveal individual halos and constrain their masses and density slopes [24, 25]. With sufficiently high-resolution observations, these approaches can probe dark matter substructure down to mass scales of [26, 27, 28, 29]. Using gravitational imaging, several studies have reported candidate detections of individual dark matter perturbers in the strong-lens systems SDSS J0946+1006 [30], JVAS B1938+666 [29, 31], SDP.81 [32], SPT2147-50 [33], and PJ011646 [34], although the robustness and interpretation of these detections vary across analyses and the adopted lens and source models. Follow-up studies have argued that the inferred compactness of some of these perturbers may be difficult to reconcile with CDM [25, 28, 35, 36], a question we investigate in this work. With sufficiently robust theoretical predictions for substructure, strong lensing measurements can test CDM structure formation and distinguish between dark matter models that predict different internal density structures for low-mass halos [22, 37, 38, 39, 40].
The self-interacting dark matter paradigm introduces non-gravitational interactions between dark matter particles and was proposed to address small-scale tensions between CDM predictions and observations of dwarf and low-surface-brightness galaxies [41, 42]. Unlike collisionless dark matter, SIDM allows for heat transport within halos through particle self-interactions, which can significantly alter their inner density structure—a process known as gravothermal evolution [43, 44, 45, 46]. This tends to produce cored halos in its early stages, but can eventually lead to core collapse, characterized by a rapidly increasing central density, a shrinking core radius, and a steepening outer density profile [46, 47, 48, 49, 50, 51]. At late times, core collapse may even lead to black hole formation [52, 53, 54]. SIDM is especially relevant for strong-lensing analyses because gravothermal evolution can produce a greater diversity of subhalo density profiles than CDM, with core-collapsed halos generally producing stronger lensing perturbations than their CDM counterparts [19, 22, 55, 56, 57]. At the population level, a recent analysis of flux ratios from 29 quadruply imaged quasars constrained the abundance of core-collapsed halos, finding a preference for SIDM models with core collapse over CDM [23]. Due to the compactness of individual core-collapsed halos, SIDM has also been proposed as a possible explanation for inferred strong-lens perturbers whose projected masses and density slopes may be difficult to realize in CDM, including those in SDSS J0946+1006 and JVAS B1938+666 [35, 40, 58].
High-resolution -body simulations that include dark matter self-interactions and self-consistently follow halo merger histories, tidal evolution, and gravothermal evolution can make robust predictions for the (sub)halo populations inferred from strong-lensing data [19, 55]. For example, [40] identified strong-lens perturber candidates in the SIDM Concerto [19] simulations based on their projected masses and density slopes. However, resolving gravothermal collapse is itself numerically challenging because the central region of the halo becomes progressively denser and more compact, requiring increasingly stringent mass, spatial, and temporal resolution. Studies have shown that core-collapse rates can be sensitive to particle number, gravitational softening, and timestep choices, with the required particle resolution generally exceeding that achieved for low-mass subhalos in simulations [59, 60]. Moreover, cosmological SIDM -body simulations are computationally expensive, so they typically follow only a few host halos for a limited set of SIDM parameter choices. However, subhalos with projected masses and density slopes matching those of an observed perturber can be rare among simulated subhalo populations, and quantifying how often they arise, and how this frequency depends on the SIDM model, requires large ensembles of subhalo populations across many host realizations and SIDM parameter choices.
Semi-analytic models provide a much faster complementary approach because they sample halo merger histories from models calibrated against cosmological simulations and use analytic prescriptions for halo evolution calibrated against idealized -body simulations, thereby avoiding the computational cost of simulating individual dark matter particle trajectories [61, 62, 63]. This approach can efficiently generate large ensembles of CDM subhalo populations in group-scale hosts and track the accretion and tidal evolution histories of individual subhalos [64, 65]. When combined with models of gravothermal evolution in SIDM, these CDM subhalo histories can be mapped into SIDM counterparts using calibrated prescriptions for their structural evolution. For example, the model of [66], which is calibrated to SIDM -body simulations, describes both the core-expansion phase and the onset of core collapse during gravothermal evolution [45, 67]. In [68], this mapping was applied to semi-analytic CDM subhalo populations for Rutherford-like velocity-dependent cross sections, demonstrating that the fraction of core-collapsed subhalos peaks at a characteristic mass scale set by the SIDM parameters. This suggests that strong lensing may be sensitive to specific regions of SIDM parameter space, for which semi-analytic models provide a computationally efficient means of exploration.
Although previous studies have identified subhalos in CDM and SIDM -body simulations with projected structures consistent with inferred perturbers [28, 40, 69, 70, 71], semi-analytic predictions for the abundance of such SIDM subhalos across many realizations of strong-lens hosts remain largely unexplored. In this work, we use the open-source semi-analytic model SatGen [62, 63] to generate CDM subhalo realizations of host halos representative of observed strong-lens systems [20, 72]. We then apply the “parametric” model of [66] to construct corresponding SIDM populations across SIDM parameter space. We evaluate the projected enclosed masses and deflection-based projected density slopes of CDM and SIDM subhalos and compare these properties with those of three perturbers inferred through gravitational imaging in the strong-lens systems PJ011646, SDSS J0946+1006, and JVAS B1938+666. This approach allows us to quantify how frequently CDM and SIDM perturber candidates arise, examine which evolutionary histories produce them, and test whether the inferred structure of multiple perturbers can be simultaneously reproduced with a single set of SIDM model parameters.
This paper is organized as follows. § 2 defines our methodology for generating CDM and SIDM subhalo populations and making lensing-relevant predictions. In § 3, we explore the dependence of the projected structure of CDM and SIDM subhalos across SIDM parameter space and various host halo redshifts. In § 4, we compare CDM and SIDM subhalos with three candidate strong-lens perturbers inferred through gravitational imaging in PJ011646, SDSS J0946+1006, and JVAS B1938+666, focusing on the projected structure, abundance, and evolutionary histories of perturber candidates across SIDM parameter space and under alternative tidal stripping prescriptions in SIDM. § 5 discusses the main modeling uncertainties in our analysis, and § 6 summarizes our results and outlook. Appendix § 8 presents the annular number densities of CDM and SIDM subhalos for PJ011646 and JVAS B1938+666. Appendix § 9 tests the robustness of our results to the choice of collapsed density profile. Appendix § 10 compares the projected subhalo mass functions and distributions of averaged convergence for CDM subhalos generated using SatGen [62, 63] and pyHalo [21]. Appendix § 11 summarizes the current observations and analyses of the three inferred perturbers considered in this work.
Methodology
In this section, we describe how we generate CDM and SIDM subhalo populations, determine which SIDM subhalos undergo gravothermal collapse, and evaluate their lensing-relevant projected properties. We also describe how we convert the posteriors of three perturbers inferred through gravitational imaging in PJ011646, SDSS J0946+1006, and JVAS B1938+666 into the same projected properties used for comparison with CDM and SIDM subhalos in § 4. The detailed modeling of subhalo evolution in CDM and SIDM remains an active area of research. We make specific assumptions and choose specific prescriptions for subhalo evolution in both paradigms, and future work can modify our predictions. In § 2.1, we describe the merger and tidal evolution histories of CDM subhalos generated using the SatGen [62, 63] semi-analytic model. In § 2.2, we define the framework for mapping these CDM subhalo histories into SIDM counterparts using a model for gravothermal evolution in SIDM, which is calibrated to -body simulations [66]. In § 2.3, we describe the density profiles assigned to CDM subhalos, which evolve only through tidal evolution, and to SIDM subhalos, which evolve through both tidal and gravothermal evolution. In § 2.4, we examine the dependence of gravothermal evolution and collapse on the SIDM model parameters and subhalo infall properties. In § 2.5, we describe the formalism for evaluating the projected enclosed masses and deflection-based projected density slopes of CDM and SIDM subhalos. Lastly, § 2.6 defines the pivot radii at which these projected properties are evaluated for comparison between CDM and SIDM subhalos and the three strong-lens perturbers considered in this work.
CDM subhalo histories
We use the open-source code SatGen [62, 63] to generate CDM subhalo realizations of host halos representative of typical galaxy-galaxy strong lenses [20, 72]. including those of the lens systems PJ011646, SDSS J0946+1006, and JVAS B1938+666 for comparison between subhalos and inferred strong-lens perturbers in § 4. In this work, and denote the redshift-dependent virial mass and radius following the convention in [73]. SatGen generates halo merger trees using the extended Press–Schechter formalism [4, 74, 75] and the Monte Carlo algorithm of [76], with the calibration adopted by [77]. The host halo and pre-infall subhalos are modeled with Navarro–Frenk–White (NFW) [7] density profiles, with concentrations determined from their individual merger histories [78]. At infall, subhalo orbital initial conditions are sampled from distributions based on cosmological simulations [79]. After infall, each subhalo is evolved by integrating its orbit in the host potential from infall to the lens redshift, while modeling the effects of dynamical friction and mass loss due to tidal stripping [80]. Throughout this work, we use tidal stripping to refer to the removal of bound mass, and tidal evolution to refer to stripping together with the structural response of the bound remnant. If a subhalo’s bound mass falls below during its evolution, it is classified as tidally disrupted and its subsequent evolution is terminated. At the lens redshift, we only include subhalos that lie within the host virial radius and whose bound masses exceed the merger-tree resolution threshold. The dark-matter-only host and subhalos are modeled without baryonic physics such as a host disk potential or ram-pressure stripping of satellite gas. In Appendix § 10, we show that the projected subhalo mass functions and distributions of subhalo convergence are consistent between CDM subhalos generated using SatGen and pyHalo. We adopt the Planck 2018 cosmological parameters throughout our analysis [1]. In the following section, we use the resulting CDM mass-accretion, structural, and tidal-evolution histories as inputs to the SIDM mapping.
SIDM gravothermal evolution
We map SatGen CDM subhalo populations into corresponding SIDM populations by applying the “parametric” model of gravothermal evolution in [66]. This model assumes that the mass-accretion and tidal-stripping histories of SIDM halos approximately follow those of their CDM counterparts while self-interactions modify their internal structures. For certain velocity-dependent SIDM models, simulations show that the subhalo mass function is fairly consistent between CDM and SIDM [55, 67, 81]. However, this assumption can break down in regimes where scattering between host and subhalo particles induces additional mass loss [82, 83, 84], or where cored subhalos are more susceptible to tidal disruption, reducing their abundance [82, 84, 85]. Nonetheless, this model allows computationally efficient exploration of the velocity-dependent SIDM parameter space without requiring a separate cosmological SIDM simulation for each parameter choice, subject to the limitations of the model.
The model of [66] approximates the gravothermal evolution of SIDM halos by applying calibrated corrections to the structural evolution histories of their CDM counterparts, and , where is the maximum circular velocity, which can be defined for any density profile, and is the radius at which it occurs. For an NFW profile, these are related to the characteristic scale density and scale radius through and . The model assumes that gravothermal evolution is approximately self-similar when time is normalized by the halo collapse timescale [47, 49, 53].
where is calibrated against -body simulations [48, 50], and
where is the approximate one-dimensional halo velocity dispersion [45], while and are the effective NFW structural parameters. The dependence of gravothermal evolution on the SIDM particle model enters through the halo collapse timescale and the effective self-interaction cross section, which we adopt from Eq. (4.2) of [45],
where the viscosity cross section is
Here, is the relative velocity between scattered dark matter particles, is the cross-section normalization, and is the velocity scale controlling the turnover of the self-interaction cross section. In this work, we adopt the SIDM model parameters and from [55] as our “fiducial” choice, but we will present several examples of different parameter choices. These parameters were chosen in [55] to produce core-collapsed subhalos relevant for the perturber in SDSS J0946+1006 while simultaneously producing large cores in low-concentration isolated halos.
For each CDM subhalo predicted by SatGen, we reconstruct the evolution histories of its structural parameters, and , from formation, through infall, and to the lens redshift of interest. After infall, these structural parameters evolve according to the tidal tracks described in [86], which depend on the subhalo’s bound mass fraction and infall concentration. Using Eqs. (1)–(6), we compute the collapse-timescale history of each subhalo and its corresponding gravothermal evolution history, characterized by the parameter
where we define the subhalo formation time as the snapshot at which its mass is closest to of its infall mass [78]. Since this integral increases monotonically, the model does not allow heating to reverse gravothermal evolution. [87] proposed a phenomenological extension of this model in which accretion-induced heating can delay core collapse, and showed that it better reproduces the evolution of some field halos in cosmological SIDM simulations, including a group-scale simulation with our fiducial SIDM parameters. Since subhalos stop accreting after infall, this effect mainly concerns their pre-infall evolution, and neglecting it may overestimate the abundance of collapsed subhalos, particularly among the massive progenitors of the SDSS J0946+1006 perturber candidates (§ 4.2).
Following [66], we map each subhalo’s CDM structural parameter histories into corresponding SIDM histories using
The functions and are the derivatives of the fitting functions in Eq. (2.4) of the arXiv version of [66], where . These fitting functions are calibrated against idealized dark-matter-only -body simulations with NFW initial conditions and dark matter self-interactions [45], and the resulting model is subsequently validated against a cosmological zoom-in SIDM simulation containing both isolated halos and subhalos [66, 67]. The second terms in Eqs. (2.8) and (2.9) capture the CDM evolution of the subhalo, which includes mass accretion and tidal stripping, while the third terms capture the changes to and induced by gravothermal evolution. In [66], the model fitting functions were calibrated through , beyond which the SIDM evolution terms were dropped. However, [88] validated the model through using an extended cosmological SIDM simulation. We therefore drop the SIDM evolution terms for , while retaining the CDM evolution terms. We additionally adopt the minimum values and to avoid numerical artifacts, following [66]. Using the predicted , , and histories of each subhalo, we invert Eq. (2.4) of [66] to obtain its fictitious initial NFW structural parameters and (and hence and ). These are then used to determine the structural parameters , , and of the cored SIDM density profile through Eq. (2.3) of the arXiv version of [66], as discussed in the next section.
Subhalo density profiles and tidal evolution
At the lens redshift , CDM subhalos are assigned density profiles based on their tidal evolution, while SIDM subhalos are assigned different density profiles based on their tidal and gravothermal evolution. After infall, the density profiles of SatGen CDM subhalos are modeled using the transfer function of [86], which is calibrated against the DASH library of idealized -body simulations [89]:
where the transfer function depends on each subhalo’s instantaneous bound mass fraction and infall concentration , capturing the response of the inner halo to tidal mass loss through revirialization as well as the truncation of the outer parts of the halo. SIDM subhalos with are assigned a cored profile:
The core radius , density normalization , and scale radius are the -dependent structural parameters determined from Eq. (2.3) in the arXiv version of [66], which capture both the core expansion and the onset of core collapse of the halo. Throughout this work, we use the term “cored” to denote all SIDM subhalos with . The model of [66] predicts that halos reach their “maximal core” at , after which the core contracts while its density rises, albeit much more slowly than in the later core-collapse phase. It is worth mentioning that the cored density profile in Eq. (2.11) slightly mismodels the density slope at (see, e.g., Figure 1 of [66]). Once core collapse begins, heat flows outward from the core, and the outer parts of the halo approach a self-similar density profile, i.e., for . Although the core continues to shrink and its density increases, with the exact profile in the innermost regions remaining uncertain, the self-similar outer region of the halo is relatively static [47]. Consequently, the time elapsed since the onset of core collapse has little effect on lensing observables probing physical scales .
We model collapsed halos with using a separate collapsed profile calculated via kinetic theory [90], which is also validated with idealized SIDM N-body simulations. This profile corresponds to a halo at . It also has a constant-density core below a radius , and is approximately described outside this core by a power law, . The enclosed mass of this collapsed profile agrees with that of the NFW profile to within approximately 10% at the virial radius, with a weak dependence on concentration. Throughout this work, we refer to SIDM subhalos assigned the profile from [90] as “collapsed” to distinguish them from the “cored” subhalos.
Figure 1 illustrates the density profiles assigned to SIDM subhalos under this binary prescription. For comparison, the purple curve shows a collapsed profile from [91], whose kinetic-theory simulations find departures from local thermodynamic equilibrium in the intermediate-mean-free-path regime that alter late-stage gravothermal evolution and the resulting density structure of collapsed halos relative to the conducting-fluid approximation.
This profile reaches much higher central densities than our adopted collapsed profile, but the two are similar at radii beyond a few percent of . In Appendix § 9, we show that our conclusions from § 4 are robust to the choice of collapsed profile.
The cored and collapsed SIDM density profiles described above do not explicitly include tidal truncation or mass loss due to scattering between subhalo and host particles. The detailed modeling of subhalo tidal evolution, mass loss, and survival in SIDM remains an active area of research (see also the discussion in § 5). We approximate tidal mass loss in SIDM by assuming that each SIDM subhalo retains the same fraction of its infall mass as its corresponding CDM counterpart. We implement this by applying a truncation factor of to the density profiles of cored and collapsed subhalos, where a time-dependent truncation radius is chosen to reproduce the bound mass fraction of the corresponding CDM subhalo at each time. The dotted red curve in Figure 1 shows an example of a truncated cored SIDM profile. Figure 7 in § 4 shows additional CDM and SIDM subhalo profiles that incorporate tidal evolution and, when relevant, gravothermal evolution.
Figure 1. Density profiles assigned to SIDM subhalos, compared with the initial NFW profile (blue), with radius and density in units of the initial NFW scale radius and scale density . Subhalos with are assigned the cored profile of Eq. (11) (red), and those with the collapsed profile of [90] (orange). The dashed red curve shows the cored profile at “maximal core”, , and the solid red curve shows it at the collapse threshold, , as predicted by the gravothermal evolution model of [66]. The collapsed profile has a constant-density core inside and is approximately outside it. The purple curve shows an alternative collapsed profile from Figure 1 of [91] at , where is the characteristic scattering timescale of the initial NFW halo; this corresponds to under the convention of Eq. (1). Appendix § 9 shows that our conclusions are robust to adopting this profile instead. All curves are shown without tidal truncation, except the dotted red curve, which shows the maximal-core profile truncated by the factor with .
Figure 7. Comparison of the projected structures of three inferred strong-lens perturbers with those of CDM and SIDM subhalo populations under the fiducial SIDM model parameters and . Left panels: projected enclosed mass (Eq. (13)) versus deflection-based projected density slope (Eq. (15)), obtained by fitting a power law to the deflection profile over . Blue, red, and orange points denote CDM, cored SIDM, and collapsed SIDM subhalos, respectively, across realizations. Horizontal bars show the central 68% interval of in bins of of width 0.3 dex. Purple contours show the 68% and 95% posterior regions of the perturbers in PJ011646 and SDSS J0946+1006; for “detection V” in JVAS B1938+666, the purple marker shows the region. Top row: PJ011646, , , . Middle row: SDSS J0946+1006, , , . Bottom row: JVAS B1938+666, , , . Right panels: posterior-median density profiles (dashed purple) with their 95% shaded regions ( for JVAS B1938+666), compared with the profiles of three individual subhalos marked by circles, squares, or triangles in the corresponding left panels. The dotted vertical line marks . The gray band shows the range of three-dimensional radii over which spherical shells contribute to the central 90% of for the posterior-median profile, i.e. the radii to which the projected-mass constraint is most sensitive; it extends beyond because shells at also contribute along the line of sight.
Incidence of gravothermal collapse
In this subsection, we examine how subhalo infall virial mass and infall redshift correlate with gravothermal evolution and core collapse. We consider SIDM subhalos with bound masses across 24 subhalo realizations of host halos with at . Figure 2, Figure 3, and Figure 4 show how the incidence of gravothermal collapse depends on the SIDM model parameters and subhalo infall properties. Using Eqs. (1)–(4), Eq. (7) can be written as
Figure 2. Fraction of gravothermally collapsed SIDM subhalos at the lens redshift , i.e., those with as calculated using Eq. (7), at each point in the parameter space (Eq. (6)). Panel (a): collapse fractions of subhalos with infall virial masses . Panel (b): corresponding collapse fractions for subhalos with . We consider SIDM subhalos with bound masses across 24 realizations of host halos with at . White contours mark constant collapse fractions of 0.1, 0.5, and 0.9.
Figure 3. Gravothermal evolution parameters (Eq. (7)) of SIDM subhalos at the lens redshift , under the fiducial SIDM model parameters and . We consider SIDM subhalos with bound masses across 24 realizations of host halos with at each lens redshift. Panel (a): versus subhalo infall virial mass at . Each point represents an individual subhalo, colored by its infall redshift . The dashed horizontal line marks , above which SIDM subhalos are classified as gravothermally collapsed. Panel (b): versus at . There are fewer collapsed subhalos at than at because subhalos have comparatively less time for both tidal and gravothermal evolution.
Figure 4. Fraction of gravothermally collapsed SIDM subhalos at the lens redshift , i.e., those with as calculated using Eq. (7), as a function of their infall virial masses. The cross-section normalization is fixed at , while the turnover velocity is varied over . Collapse fractions are evaluated in logarithmic infall mass bins of width 0.2 dex across 24 SIDM subhalo realizations of host halos with . The solid lines show the mean collapse fractions, while the shaded regions show the scatter across realizations. At fixed , the collapse fraction exhibits a characteristic peak as a function of infall mass that shifts toward larger masses as increases. Scatter in the CDM structural evolution histories broadens the infall mass range over which the collapse fraction peaks, because gravothermal evolution depends on the time-dependent effective cross section and halo structural parameters in Eq. (12).
which shows how gravothermal evolution depends on the full CDM merger and tidal evolution histories of each subhalo through the time-dependent halo structural parameters and .
Figure 2 shows how the SIDM model parameters (Eq. (6)) determine the fraction of gravothermally collapsed subhalos for two infall mass ranges, with the underlying CDM subhalo histories held fixed. At fixed turnover velocity , increasing the cross-section normalization increases the effective cross section (Eq. (5)), shortens the halo collapse timescale (Eq. (1)), and drives a transition from negligible to nearly complete collapse in both panels. At smaller , the self-interaction cross section is suppressed at higher velocities, which shifts contours of fixed collapse fraction toward larger , approximately following the high-velocity scaling , whereas at sufficiently large the interaction approaches the velocity-independent limit and the contours become nearly vertical. The lower-mass subhalos in panel (a) typically have smaller internal velocities and can therefore undergo gravothermal collapse at comparatively smaller than the higher-mass subhalos in panel (b).
We now examine the gravothermal state of SIDM subhalos at the lens redshift under the fiducial SIDM model parameters and . Figure 3 illustrates how subhalo infall mass and infall redshift correlate with . We consider SIDM subhalos with bound masses across 24 realizations of host halos with at in panel (a) and at in panel (b). At fixed infall mass, subhalos that fall into the host earlier generally experience greater cumulative tidal stripping and reach larger values of due to accelerated gravothermal evolution. However, there are fewer collapsed subhalos at than at because subhalos have comparatively less time for both tidal and gravothermal evolution. The broad distributions of at fixed or demonstrate that neither quantity uniquely determines the gravothermal state of SIDM subhalos; instead, gravothermal evolution depends on their full CDM structural evolution histories according to Eq. (12).
The relationship between the turnover velocity and the characteristic internal velocities of subhalos has important implications for gravothermal evolution. Figure 4 shows how the fraction of SIDM subhalos that are gravothermally collapsed at varies with infall mass and turnover velocity . We consider SIDM subhalos with bound masses across 24 realizations of host halos with at . At fixed cross-section normalization , the collapse fraction exhibits a characteristic peak as a function of infall mass that shifts toward larger masses as the turnover velocity increases, consistent with the behavior found by [68]. Increasing extends the range of internal halo velocities over which the effective cross-section tion in Eq. (5) remains large, allowing progressively more massive subhalos to undergo gravothermal collapse. The location of this peak does not simply correspond to subhalos with , for two reasons. First, the velocity-weighting in Eq. (5) peaks at a relative velocity , so the effective cross section becomes substantially suppressed when is of order . Second, also depends on the structural factor in Eq. (12), which generally decreases toward lower masses, so low-mass subhalos evolve more slowly even where their cross section is not velocity-suppressed. The peak lies where these two effects balance, and because and evolve through accretion and tidal stripping, scatter in these histories broadens the mass range over which collapse is most frequent.
Projected enclosed mass and deflection-based projected density slope
Strong gravitational lensing is sensitive to the mass projected onto the plane of the sky rather than to the three-dimensional subhalo density profile. To connect with lensing observations, we therefore characterize subhalos by their projected enclosed masses and deflection-based projected density slopes, which correspond closely to quantities inferred through gravitational imaging [28]. Rather than projecting each three-dimensional density profile along the line of sight and subsequently integrating the surface mass density , where is the line-of-sight coordinate, we calculate the projected enclosed mass directly from the three-dimensional profile using a spherical-shell projection. A spherical shell at three-dimensional radius contributes fully to a cylinder of projected radius when , and partially when , giving
where the projection weight is
For , is the fraction of the shell’s surface lying inside the cylinder, i.e. the two polar caps of half-angle , with combined area . This reduces the calculation to a single one-dimensional integral.
We characterize the projected density profiles of circularly symmetric subhalos via the radial dependence of their deflection angle , which, as for a point mass, depends only on the projected mass enclosed within , [92]. For an exact power-law surface density, (with so that the enclosed mass is finite), the enclosed mass and deflection angle scale as and , respectively. Therefore, we define the deflection-based projected density slope for an arbitrary projected subhalo profile as
More negative values of correspond to steeper projected profiles. For reference, a three-dimensional power law with projects to , so that ; an isothermal profile () thus has . For general profiles, this correspondence is only approximate, because receives contributions from all along the line of sight.
The deflection-based slope also has simple physical bounds. Using and , Eq. (15) can be written for any circularly symmetric profile as [92]
where . Thus, compares the local surface density at with the mean surface density within , rather than being the local logarithmic derivative of . For the non-negative, monotonically decreasing density profiles considered in this work, the surface density is also non-negative and decreasing, so and therefore
The upper bound is approached where the surface density is nearly constant within (a core). The lower bound is the point-mass limit: when lies well outside most of the mass (e.g., beyond the truncation radius of a compact, tidally stripped subhalo), , approaches a constant, and .
These limits provide a useful way to interpret the density profiles considered in this work. For all three subhalo families, tidal stripping removes mass along the line of sight at , steepening the deflection-based slope toward the point-mass limit as exceeds the truncation radius . CDM subhalos (Eq. (10)) retain an inner cusp, so is relatively shallow when lies within the scale radius . Before infall, an NFW profile has at and at . Cored SIDM subhalos (Eq. (11)) approach the upper bound, , for . Collapsed subhalos mostly span : the shallow end is set by the untruncated profile at , corresponding to an effective three-dimensional slope , with steeper slopes set by tidal truncation. In contrast to CDM and cored SIDM subhalos, this shallow end is nearly independent of subhalo mass, except for massive collapsed subhalos at small pivot radii, where approaches the core radius of the collapsed profile; see Figure 1 and Appendix § 9.
Posteriors of inferred strong-lensing perturbers
To compare strong-lens perturbers inferred in PJ011646, SDSS J0946+1006, and JVAS B1938+666 with simulated CDM and SIDM subhalo populations, we adopt the posterior samples from the corresponding gravitational-imaging analyses and convert them into posteriors of projected enclosed mass and deflection-based projected density slope using Eqs. (13) and (15). is the pivot radius at which the enclosed mass and density slope are most robustly constrained in each gravitational-imaging inference [25, 34], and at which their covariance vanishes [93]. It depends on the lensing geometry, perturber properties, signal-to-noise ratio, and angular resolution across each lens system. The pivot radius should not be confused with the perturbation radius of [36, 94], which is defined as the projected distance from the center of the perturber to the point on the critical curve that is most displaced by the perturber. The perturbation radius is set by the geometry of the perturbed critical curve, whereas the pivot radius is determined by where the inferred mass profile is best constrained, and the two need not coincide. For each CDM and SIDM density profile, we estimate by fitting the corresponding deflection profile over the radial window , i.e. a half-width of 25% of . This choice is motivated by previous lensing analyses of SDSS J0946+1006, which follow a similar convention [36, 58], since the enclosed mass and density slope are constrained in a narrow radial range as opposed to a single radius. We use this window consistently when evaluating the deflection-based density slopes of all subhalos and perturbers considered in this work.
We summarize here the relevant details of each candidate perturber, leaving additional details to Appendix Appendix § 11. For the inferred perturber in PJ011646, we adopt the NFW posterior samples from [34]. Assuming that the perturber lies at the lens redshift , the analysis reported and , with a concentration that they argued is consistent with a typically tidally stripped CDM subhalo (see § 4.1 for our interpretation), and a projected enclosed mass at an angular pivot radius of (corresponding to a physical radius of kpc at ) [34]. When we convert the NFW posterior samples into a joint posterior of and , we obtain a median projected enclosed mass and a median deflection-based projected density slope .
For the inferred perturber in SDSS J0946+1006 at lens redshift , we adopt the truncated NFW posterior samples from [36], whose lens model included both supersampling and multipoles; see their Table 4. At 95% confidence, the analysis constrained the projected enclosed mass within 1 kpc to – and the average logarithmic slope of the local surface density, , over 0.75–1.25 kpc to be steeper than . This differs from the deflection-based slope and radial window adopted in this paper. From the truncated NFW posterior samples, we determine a pivot radius of kpc, at which the fractional uncertainty in the projected enclosed mass is minimized and its covariance with the slope of the local surface density vanishes, following [93]. We convert these posterior samples into a joint posterior of and , obtaining a median projected enclosed mass of and a median deflection-based projected density slope of .
For “detection V” in JVAS B1938+666 at , we use the single-power-law (PL) inference reported in [35], which gives and a three-dimensional logarithmic density slope . We adopt this posterior because its pure power-law form allows the inferred three-dimensional slope to be directly related to the deflection-based projected density slope, . As shown in Supplementary Figure 12 of [35], the joint posterior of and is approximately Gaussian, with weak correlation between the two quantities. However, the highest-evidence model consists of a central point mass with a disk of uniform surface mass density. The data robustly constrain under both models and many alternative models at the lens redshift, despite their differing Bayesian evidences; see their Table 1 [35]. Although [29] identified 80 pc as the approximate radius at which the uncertainty in the enclosed mass is minimized for their pseudo-Jaffe model, we adopt pc following [35], who found that 10 of the 12 models with predict within 3% of their highest-evidence model. In our work, all conversions between angular and physical scales are performed using the gravitational-lensing software LENSTRONOMY [95, 96], which internally uses ASTROPY for cosmological distance calculations [97].
Projected structure of CDM and SIDM subhalos
Before comparing with specific inferred perturbers in § 4, we first examine the general predictions of our framework for the projected structure of CDM and SIDM subhalos. In § 3.1, we show how the distributions of projected enclosed masses and deflection-based projected density slopes change across SIDM parameter space, and in § 3.2, we examine their dependence on lens redshift and pivot radius.
Dependence across SIDM parameter space
In Figure 5, we show the distributions of projected enclosed mass and deflection-based projected density slope, at a fixed pivot radius kpc using Eqs. (13) and (15), for 24 CDM and SIDM subhalo realizations at and four sets of SIDM model parameters. The CDM and SIDM populations exhibit a positive correlation between and . At fixed pivot radius, more massive subhalos are generally probed farther within their characteristic structure and therefore have shallower projected slopes, whereas lower-mass or more strongly stripped subhalos are generally probed in their outer regions and tend to have steeper slopes. This trend is visible in the shift of the 68% intervals toward shallower density slopes with increasing enclosed mass, with the CDM and SIDM populations overlapping in parts of the – plane. Scatter in this correlation reflects differences in progenitor mass, concentration, and tidal evolution.
Figure 5. Dependence of the projected enclosed masses and deflection-based projected density slopes of CDM and SIDM subhalos on the SIDM model parameters (, ) at fixed lens redshift and pivot radius kpc. Panel (a): velocity-dependent viscosity cross section per unit mass (Eq. (6)) for four sets of SIDM model parameters. Panels (b)–(e): (Eq. (13)) versus (Eq. (15)), obtained by fitting a power law to the deflection profile over . We consider CDM and SIDM subhalos with bound masses across 24 realizations of host halos with . The SIDM model parameters are indicated in the upper-right corner of each panel, with colors matching the corresponding curves in panel (a). Blue, red, and orange points denote CDM, cored SIDM, and collapsed SIDM subhalos, respectively. Horizontal bars denote the central 68% interval of in bins of of width 0.3 dex. CDM and SIDM subhalos can be distinguished in observations provided that the pivot radius probes a region where their density profiles differ, rather than lying outside their truncation radii.
CDM and SIDM subhalos with comparable can be distinguished if their are steeper or shallower. Panel (a) considers four velocity-dependent viscosity cross sections (Eq. (6)), with colors corresponding to the parameter choices in the top right of panels (b)–(e). We consider subhalos with bound masses across 24 realizations of host halos with at . For and , panel (b) shows that the cored SIDM populations are largely degenerate with their CDM counterparts with comparable because inefficient self-scattering produces core radii that are typically much smaller than the pivot radius, yielding comparable . When the turnover velocity is increased to for fixed in panel (c), of SIDM subhalos are collapsed, predominantly at lower masses, while the most massive cored SIDM subhalos become shallower than their CDM counterparts because their core radii are closer to the pivot radius. Similar behavior is seen in panel (d), which uses a much larger cross-section normalization of with . In this case, of SIDM subhalos are collapsed while the most massive subhalos form cores comparable to the pivot radius because the cross section remains sufficiently large at higher internal halo velocities, as illustrated by the magenta curve in panel (a). When both the cross-section normalization and turnover velocity are increased to and in panel (e), the most massive cored subhalos exhibit both steeper and shallower slopes at the pivot radius compared to their CDM counterparts, while the lower-mass cored subhalos remain largely degenerate with their CDM counterparts because the pivot radius generally probes outside of their truncation radii. In addition, of SIDM subhalos are collapsed, including higher-mass subhalos, since the self-scattering cross section remains high at larger internal halo velocities. For sufficiently large cross-section normalizations and turnover velocities, gravothermal evolution produces SIDM subhalos with that can be substantially steeper or shallower than those of CDM subhalos with comparable . These subhalos can be distinguished in observations provided that the pivot radius probes a region where their density profiles differ, rather than lying outside their truncation radii.
Dependence on pivot radius and redshift
In Figure 6, we show the distributions of projected enclosed masses and deflection-based projected density slopes of CDM and SIDM subhalos under the SIDM model parameters and , evaluated at two pivot radii
Figure 6. Projected enclosed masses (Eq. (13)) and deflection-based projected density slopes (Eq. (15)) of CDM and SIDM subhalos for different lens redshifts and pivot radii , obtained by fitting a power law to the deflection profile over . Blue, red, and orange points denote CDM, cored SIDM, and collapsed SIDM subhalos, respectively, under the SIDM model parameters and . In each panel, we consider CDM and SIDM subhalos with bound masses across 5 realizations of host halos with at each lens redshift. Top panels: versus for lens redshifts from left to right. Bottom panels: versus for the same lens redshifts. Horizontal bars denote the central 68% interval of in bins of of width 0.4 dex. For each lens redshift, the bottom panel gives the total numbers of CDM, cored SIDM, and collapsed SIDM subhalos with across the 5 realizations. The number of CDM subhalos with bound masses increases with lens redshift, since subhalos at higher redshifts have had less time to undergo tidal disruption after accretion into their hosts than those at lower redshift, consistent with theoretical expectations [98]. In contrast, there are fewer collapsed subhalos at higher redshifts because subhalos have had less time to undergo tidal and gravothermal evolution.
(rows) and three lens redshifts (columns). We choose these redshifts because galaxy-scale strong lenses are observed over a broad redshift range, from in nearby spectroscopic samples [99] to in recent JWST discoveries [100, 101]. At each lens redshift and pivot radius, we consider CDM and SIDM subhalos with bound masses across 5 realizations of host halos with . At fixed pivot radius and SIDM model parameters, the CDM and SIDM subhalos span approximately the same regions in the plane over the range of lens redshifts considered. However, the distributions of and depend strongly on the pivot radius, which determines the physical scale over which the subhalo density profiles are probed. At , CDM and cored SIDM subhalos generally have shallower slopes than at 1 kpc because the smaller pivot radius probes regions well inside their scale radii and, for cored subhalos, inside their cores. Collapsed subhalos instead remain steep at both pivot radii, except for the most massive ones at , where the pivot radius approaches the small core of the collapsed profile and their slopes become shallower (§ 2.5). At each lens redshift, every CDM subhalo has either a cored or collapsed SIDM counterpart, and therefore the numbers of CDM and SIDM subhalos are equal. The number of CDM subhalos with bound masses increases with lens redshift, since subhalos at higher redshifts have had less time to undergo tidal disruption after accretion into their hosts than those at lower redshift, consistent with theoretical expectations [98]. In contrast, there are fewer collapsed subhalos at higher redshifts because subhalos have had less time to undergo tidal and gravothermal evolution.
Confronting strong-lensing perturbers with CDM and SIDM subhalos
In this section, we compare the projected enclosed masses (Eq. (13)) and deflection-based projected density slopes (Eq. (15)) of CDM and SIDM subhalo populations with those of three strong-lens perturbers inferred in PJ011646 [34], SDSS J0946+1006 [36], and JVAS B1938+666 [35]. For each system, we evaluate these properties at the pivot radius where they are most robustly constrained in each gravitational-imaging inference. The perturbers in PJ011646, SDSS J0946+1006, and JVAS B1938+666 were inferred from submillimeter (ALMA), optical (HST), and radio (VLBI) observations, respectively. Differences in the lensing geometry, perturber properties, signal-to-noise ratio, and angular resolution across the three systems lead to pivot radii ranging from 1.92 kpc to 90 pc (§ 2.6). In § 4.1–§ 4.3, we determine the numbers of CDM, cored SIDM, and collapsed SIDM subhalos across many realizations whose projected structures are approximately consistent with those of the inferred perturbers, i.e., those that approximately lie within the posterior regions in the – plane, under the fiducial SIDM model parameters and . We examine the evolutionary properties of these perturber candidates, focusing on their infall masses and bound mass fractions. In Section 4.4, we test whether reproducing the inferred structures of all three perturbers requires specific choices of the SIDM model parameters, while in Section 4.5, we test the robustness of our results to alternative tidal stripping prescriptions for SIDM subhalos.
The left panels of Figure 7 show versus for the three inferred strong-lens perturbers and the corresponding CDM and SIDM subhalo populations, under the fiducial SIDM model parameters and . The deflection-based projected density slope is estimated by fitting a power law to the deflection profile, , over . We consider subhalos with bound
masses across 24 realizations of host halos with for PJ011646 and SDSS J0946+1006, whereas for JVAS B1938+666 we consider subhalos with bound masses across 10 realizations. The right panels compare the posterior median density profile with those of a subhalo from each model family marked in the corresponding left panels. For PJ011646 and SDSS J0946+1006, we plot the density profiles of massive subhalos with enclosed masses consistent with the posteriors, while for JVAS B1938+666 we plot three subhalos with approximately the same enclosed mass and density slope. The shaded vertical gray bands highlight the range of three-dimensional radii over which spherical-shell density contributions account for the central 90% of for the median posterior density profiles. We note that the bands extend to radii beyond , since mass along the line of sight at three-dimensional radii contributes to the projected enclosed mass .
PJ011646
Under the fiducial SIDM model parameters, cored SIDM subhalos frequently reproduce the inferred structure of the perturber in PJ011646, whereas CDM subhalos only rarely do so and collapsed SIDM subhalos do not. As discussed in § 2.6, the posterior has a median projected enclosed mass and a median deflection-based projected density slope , as shown in panel (a) of Figure 7. To obtain more reliable counts of CDM and SIDM subhalos that lie within the approximate posterior region of the perturber, we generate 900 subhalo realizations under the fiducial model parameters and , independent of the 24 realizations shown in Figure 7. Within the rectangular region and , which approximately characterizes the 95% posterior region, we find 2 CDM, cored SIDM, and no collapsed SIDM subhalos. No collapsed subhalos lie within this region because their slopes at this pivot radius are steeper than (§ 2.5). To investigate the range of slopes exhibited by these massive subhalos, without attempting to represent the posterior region, we then widen the slope selection while keeping the mass selection fixed. For , we find CDM, cored SIDM, and still no collapsed SIDM subhalos. Further widening the selection to , which extends beyond the shallowest slopes of collapsed subhalos, yields CDM, cored SIDM, and collapsed SIDM subhalos.
The rarity of CDM candidates in our realizations is consistent with the high concentration inferred for this perturber. As the counts above show, massive CDM subhalos typically have slopes slightly shallower than the posterior. [34] found , about above the mass–concentration relation of CDM field halos [17], and argued that this is consistent with a tidally stripped CDM subhalo based on the typical increase in concentration by a factor of found for stripped subhalos in [102]. For a subhalo of this mass, this factor implies , still about a factor of two below the inferred value. Our CDM subhalos, whose tidal evolution is modeled with the transfer function of [86], similarly only rarely reach the steep slope required at the pivot radius. Baryonic contraction, which our dark-matter-only realizations neglect, could steepen the inner profiles of CDM subhalos of this mass and increase the number of CDM candidates (§ 5).
Panel (b) compares the median posterior NFW density profile with those of the three subhalos marked by circles in panel (a). These subhalos have consistent with the posterior but different . The pivot radius therefore probes a qualitatively different region of the density profile for each model family. For the CDM subhalo, it lies within the scale radius of the profile, whereas for the cored and collapsed SIDM subhalos, it lies beyond their cored regions, where the density profiles are steeper. The shaded gray band illustrates why agreement in the enclosed mass within the pivot radius does not require agreement in the density profiles at smaller radii. The radial range over which density contributes to the central 90% of for the posterior median lies largely outside the region where the selected profiles differ from the posterior.
SDSS J0946+1006
Unlike for PJ011646, where a few CDM subhalos lie within the posterior region, we find no CDM subhalos within the posterior region of the perturber in SDSS J0946+1006, which simultaneously requires a large projected enclosed mass and a steep deflection-based projected density slope at a pivot radius of 0.63 kpc. However, collapsed SIDM subhalos frequently lie within the posterior region under the fiducial SIDM model parameters and . Since these parameters were chosen in [55] in part to produce core-collapsed subhalos relevant for this perturber (§ 2.2), we examine other parameter choices in § 4.4. As discussed in § 2.6, the posterior has a median and a median , as shown in panel (c) of Figure 7. To obtain more reliable counts of CDM and SIDM subhalos that lie within the approximate posterior region of the perturber, we generate 900 subhalo realizations, independent of the 24 realizations shown in Figure 7. Within the rectangular region and , which approximately characterizes the 95% posterior region, we find no CDM, 4 cored SIDM, and collapsed SIDM subhalos. To investigate the range of slopes exhibited by these massive subhalos, without attempting to represent the posterior region, we then widen the slope selection while keeping the mass selection fixed. For , we find no CDM, 42 cored SIDM, and the same collapsed SIDM subhalos. CDM subhalos appear only once the shallow edge of the slope selection exceeds ; for , we find CDM, cored SIDM, and again the same number of collapsed SIDM subhalos.
A CDM interpretation of the inferred perturber would require , whereas all CDM subhalos of comparable enclosed mass in our realizations have , because the pivot radius lies well within their scale radii (§ 2.5). Panel (d) of Figure 7 illustrates the steep density slope required by the inferred perturber by comparing the posterior-median truncated NFW profile with the density profiles of three subhalos marked by squares in the corresponding left panel. The CDM and cored SIDM examples have enclosed masses consistent with the posterior but slopes that are too shallow at the pivot radius, whereas the collapsed SIDM example has both an enclosed mass and a slope consistent with the posterior. Because each SIDM subhalo is assumed to share the accretion and tidal evolution history of its CDM counterpart, the difference in compactness arises primarily from self-interactions: gravothermal collapse makes these subhalos more compact, while their CDM counterparts remain far less concentrated.
We further investigate the evolutionary histories of the SIDM subhalos that lie within the approximate 95% posterior region of the inferred perturber across 900 realizations. Panel (a) of Figure 8 shows the infall times and infall virial masses of the 4 cored and collapsed subhalos. The 4 cored subhalos have core radii of and , just below the adopted collapse threshold . Consequently, the pivot radius of 0.63 kpc lies well outside their core radii, where their density profiles are steep. They also have bound mass fractions , corresponding to a loss of about 85% of their infall masses. The collapsed subhalos have a median infall virial mass , a median infall time (corresponding to ) and a median bound mass fraction , corresponding to a loss of about 70% of their infall masses. Relative to all SIDM subhalos (both cored and collapsed) within the enclosed mass selection, regardless of slope, these collapsed subhalos have median properties corresponding to the earliest of infall times and lowest of bound mass fractions. Our results suggest that the inferred structure of the perturber in SDSS J0946+1006 can frequently arise through the late-stage gravothermal evolution of massive subhalos, with infall masses of , without requiring unusual infall times or tidal histories. At such infall masses, however, these subhalos would likely host a stellar component, which is included in neither the adopted lens model nor our dark-matter-only realizations (see the discussion of baryonic effects in § 5).
Figure 8. Infall properties of selected CDM and SIDM subhalos whose projected enclosed masses (Eq. (13)) and deflection-based projected density slopes (Eq. (15)) lie approximately within the posterior regions of the inferred perturbers in SDSS J0946+1006 and JVAS B1938+666, as shown in panels (c) and (e) of Figure 7. Blue, red, and orange points denote CDM, cored SIDM, and collapsed SIDM subhalos, respectively, under the fiducial SIDM model parameters and . Panel (a): infall time versus infall virial mass of subhalos with and (SDSS J0946+1006; , , realizations). (Note that no CDM subhalos lie within the posterior of the perturber, and hence there are no blue points in this panel.) Panel (b): infall time versus infall virial mass of subhalos with and lying within the posterior region of “detection V” (JVAS B1938+666; , , realizations). Contours enclose 68% and 95% of the selected SIDM subhalo populations.
JVAS B1938+666
Under the fiducial SIDM model parameters and , we find that SIDM subhalos frequently reproduce the inferred structure of “detection V” in JVAS B1938+666, whereas CDM subhalos rarely do so and require unusual evolutionary histories. As discussed in § 2.6, the posterior of “detection ” has a mean projected enclosed mass of and a mean deflection-based projected density slope of , as shown in panel (e) of Figure 7. Across 10 realizations, only 2 CDM, cored SIDM, and collapsed SIDM subhalos lie within the posterior regions in both and . Within the posterior regions, we find 23 CDM, cored SIDM, and collapsed SIDM subhalos.
To examine which evolutionary histories produce these perturber candidates, we show in panel (b) of Figure 8 the infall times and infall virial masses of the subhalos within the posterior regions. The collapsed subhalos have a median infall virial mass of , a median infall time of (corresponding to ), and a median bound mass fraction of , corresponding to a loss of of their infall masses. The cored subhalos have a median infall virial mass of , a median infall time of (), and a median bound mass fraction of , corresponding to a loss of of their infall masses. They also have a median core radius of and a median , just below the adopted collapse threshold, so the 90 pc pivot radius lies well outside their cores, where is steep. The 23 CDM subhalos have a median infall time of (), a median infall virial mass of , and a median bound mass fraction of , corresponding to a loss of of their infall masses. At infall, before tidal stripping, their density profiles have a median and a median , about twice the posterior mean, so none of them would lie within the posterior region without tidal evolution.
To assess how unusual these evolutionary histories are, we compare each sample with its “parent” population: subhalos of the same family that lie within the posterior region of alone, without any selection on . Only of the CDM, of the cored, and of the collapsed parent subhalos are steep enough to also lie within the posterior region of . Relative to their parent populations, the steep CDM subhalos have median properties corresponding to the earliest of infall times and the lowest of bound mass fractions, compared with the earliest and lowest for cored subhalos, and the earliest and lowest for collapsed subhalos. A CDM interpretation of “detection ” therefore requires exceptionally early accretion, while the amount of tidal stripping is fairly typical; tidal stripping is necessary, since none of these subhalos would lie within the posterior region without it. A cored SIDM interpretation requires somewhat earlier infall and somewhat stronger tidal mass loss than typical, whereas a collapsed SIDM interpretation requires neither. Panel (f) of Figure 7 compares the posterior median density profile with one subhalo from each model family that approximately matches the posterior mean and , marked by the overlapping triangles in panel (e). The three profiles illustrate the compact and steep structure required for consistency with “detection ”.
However, an SIDM interpretation of “detection ” depends on the choice of SIDM model parameters . Panel (a) of Figure 9 shows the probability density functions (PDFs) of for the subhalos that lie within the posterior region under the fiducial SIDM model parameters, i.e., the same samples shown in panel (b) of Figure 8. In this case, collapsed subhalos dominate the SIDM sample near the posterior mean. Panel (b) shows the corresponding PDFs when the cross-section normalization is reduced to at fixed . Across the same 10 realizations, the number of SIDM subhalos within the posterior region then decreases by , from to . At this lower cross-section normalization, the remaining collapsed subhalos all have , steeper than the region of the posterior, so the SIDM subhalos closest to the posterior mean are almost exclusively cored. Even so, SIDM subhalos still outnumber the 23 CDM subhalos by more than an order of magnitude, so the constraints on the SIDM model parameters in § 4.4 are driven primarily by SDSS J0946+1006 rather than by “detection V”.
Figure 9. Probability density functions (PDFs) of the deflection-based projected density slope (Eq. (15)) for CDM and SIDM subhalos lying within the and posterior regions of “detection V” in JVAS B1938+666 (, , realizations). The PDFs for cored and collapsed subhalos together integrate to unity, while the PDF for CDM subhalos alone integrates to unity. The numbers of selected CDM (blue), cored SIDM (red), and collapsed SIDM (orange) subhalos are indicated in the lower-right corner of each panel. Panel (a): PDFs of for selected subhalos under the fiducial SIDM model parameters and , corresponding to the same subhalo samples in panel (b) of Figure 8. Panel (b): PDFs of for selected subhalos under and . The solid purple vertical line and shaded band denote the posterior mean and its region, respectively. At the lower cross-section normalization, the SIDM subhalos closest to the posterior mean of “detection V” are almost exclusively cored.
Varying the SIDM model parameters
We next investigate whether reproducing the structure of all three inferred perturbers with SIDM subhalos requires sufficiently large values of the SIDM model parameters. Figure 10 repeats the comparison of Figure 7 for three alternative sets of parameters. Each row corresponds to one of the perturbers in PJ011646, SDSS J0946+1006, and JVAS B1938+666, while each column corresponds to one parameter set: (left), (middle), and (right). The CDM populations do not depend on these parameters and are therefore identical across columns. In the first column, cored SIDM subhalos lie within the PJ011646 posterior, no subhalos lie
Figure 10. Comparison between the posteriors of inferred strong-lens perturbers and simulated CDM and SIDM subhalo populations, analogous to Figure 7, with each column corresponding to a different set of SIDM model parameters. Blue, red, and orange points denote CDM, cored SIDM, and collapsed SIDM subhalos, respectively. Top row: PJ011646, , , . Middle row: SDSS J0946+1006, , , . Bottom row: JVAS B1938+666, , , . Left column: and . Middle column: and . Right column: and . Horizontal bars denote the central 68% interval of in bins of of width 0.3 dex. Purple contours show the 68% and 95% posterior regions of the inferred perturbers in PJ011646 and SDSS J0946+1006; for “detection V” in JVAS B1938+666, the purple marker shows the 5 region. Among the parameter sets considered here and in Figure 7, simultaneous overlap with all three posteriors requires a turnover velocity , and is achieved for cross-section normalizations .
within the SDSS J0946+1006 posterior, and both cored and collapsed subhalos lie within the JVAS B1938+666 posterior. Reproducing the large enclosed mass of the SDSS J0946+1006 perturber requires massive progenitors (§ 4.2), whose internal velocities greatly exceed km/s. Their self-interactions are therefore strongly velocity-suppressed, and they do not undergo sufficient gravothermal evolution to collapse. Lower-mass subhalos with smaller internal velocities still scatter efficiently and collapse, which is why the JVAS B1938+666 posterior remains populated. In the second column, quadrupling the cross-section normalization to cm/g increases the abundance of collapsed subhalos, but massive subhalos still do not lie within the SDSS J0946+1006 posterior. For , Eq. (6) gives up to a logarithmic factor, so a fourfold increase in compensates for only a increase in velocity. In the third column, raising the turnover velocity to km/s at cm/g greatly reduces the velocity suppression for massive subhalos, which can then collapse and populate the SDSS J0946+1006 posterior, while cored subhalos continue to lie within the PJ011646 posterior and both cored and collapsed subhalos lie within the JVAS B1938+666 posterior. Since the PJ011646 and JVAS B1938+666 posteriors are populated by SIDM subhalos for all parameter sets considered here, including cm/g for JVAS B1938+666 (§ 4.3), the requirement for simultaneous overlap is set by SDSS J0946+1006. Among the parameter sets considered here, including the fiducial set, populating the SDSS J0946+1006 posterior requires a turnover velocity of approximately km/s, and is achieved for cross-section normalizations cm/g.
Impact of alternative tidal stripping prescriptions in SIDM
As discussed in § 4 and shown in Figure 7, under the fiducial SIDM model parameters cm/g and km/s, cored SIDM subhalos lie within the 95% posterior region of the PJ011646 perturber and both cored and collapsed SIDM subhalos lie within the 95% posterior region of the SDSS J0946+1006 perturber and the posterior region of “detection V” in JVAS B1938+666. These results assume the fiducial prescription for tidal mass loss in SIDM adopted thus far, in which every SIDM subhalo retains the same fraction of its infall mass as its CDM counterpart. In Figure 11, we consider two alternative tidal stripping prescriptions for SIDM subhalos that bracket this assumption, while the tidal evolution of CDM subhalos remains unchanged. These prescriptions modify only the bound mass fractions, and hence the truncation radii, of SIDM subhalos, not their gravothermal evolution histories. Since stronger tidal stripping would also accelerate gravothermal evolution, the severe-stripping scenario may underestimate the number of collapsed subhalos.
Figure 11. Comparison between the posteriors of inferred strong-lens perturbers and simulated CDM and SIDM subhalo populations, analogous to Figure 7, for two alternative prescriptions for SIDM tidal stripping, under the fiducial SIDM parameters and . Left column: severe tidal stripping, in which each SIDM subhalo retains only 10% of the bound mass fraction of its CDM counterpart. Right column: minimal tidal stripping, in which the bound mass fraction is held fixed once an SIDM subhalo exceeds the collapse threshold . Top row: PJ011646, , , . Middle row: SDSS J0946+1006, , , . Bottom row: JVAS B1938+666, , , . Horizontal bars denote the central 68% interval of in bins of of width 0.3 dex. Our conclusions from § 4 are robust to these alternative stripping prescriptions: SIDM subhalos reproduce the structures of the inferred perturbers more frequently than CDM subhalos under both prescriptions.
The left column shows a “severe” tidal stripping scenario in which each SIDM subhalo retains only 10% of the bound mass fraction of its CDM counterpart. This choice is motivated by the idealized SIDM -body simulations of [84], which adopt the same velocity-dependent SIDM model used in our analysis [45], but with cm/g and km/s. In their simulations, a subhalo evolved for 5 Gyr within a host on an orbit with an apocenter of and a pericenter of retained about an order of magnitude less bound mass than its CDM counterpart; see their Figure 4 [84]. As shown in the left column of Figure 11, severe stripping reduces the projected enclosed masses and steepens the deflection-based projected density slopes , because the truncation radii of many subhalos approach or fall below the pivot radius, driving their closer to the point-mass limit (§ 2.5). The right column shows a “minimal” tidal stripping scenario in which the bound mass fraction of an SIDM subhalo is held fixed once its gravothermal evolution parameter exceeds the collapse threshold . This choice is motivated
by cosmological SIDM zoom-in simulations showing that core-collapsed subhalos are more resilient to tidal disruption than cored ones [55], and by idealized -body simulations showing that tidal mass loss slows as cuspy subhalos approach stable bound remnants [85]. We therefore treat this scenario as a limiting case of strongly suppressed tidal stripping in SIDM, in which mass loss stops entirely after collapse. With larger truncation radii, collapsed subhalos exhibit shallower , as shown in the right column of Figure 11.
For each alternative prescription, we compare the number of subhalos lying approximately within the posterior regions of the three perturbers with that under the fiducial prescription. As in § 4.1–§ 4.3, we use the 900 realizations for PJ011646 and SDSS J0946+1006, which are independent of the 24 realizations shown in Figure 11, and the 10 realizations shown in the figure for JVAS B1938+666. Since minimal stripping applies only to subhalos with , the cored populations are also unchanged under this prescription.
Under severe stripping, the number of cored subhalos that lie within the PJ011646 posterior region decreases by , to , compared with 2 CDM subhalos. For SDSS J0946+1006, the number of collapsed subhalos decreases by , to , compared with no CDM subhalos, and the 4 cored subhalos discussed in § 4.2 no longer lie within the posterior region. For “detection V” in JVAS B1938+666, the numbers of cored and collapsed subhalos decrease by and , to and , respectively, compared with 23 CDM subhalos. In this case, cored subhalos lie within the posterior region more frequently than collapsed subhalos, and have a median , just below the adopted collapse threshold. Under minimal stripping, the number of collapsed subhalos that lie within the posterior region increases by for SDSS J0946+1006 and by for JVAS B1938+666, and remains zero for PJ011646. Although severe stripping substantially reduces the number of SIDM perturber candidates, our conclusions from § 4 are robust to both prescriptions: SIDM subhalos reproduce the inferred structures of all three perturbers more frequently than CDM subhalos.
Discussion of modeling uncertainties
In this section, we discuss four modeling uncertainties that may affect our predictions and conclusions. Addressing these uncertainties in future work will allow for systematic improvements to our framework.
Late-stage gravothermal evolution: The detailed behavior of late-stage gravothermal evolution remains theoretically uncertain, and accurately modeling this regime is an active area of research. The gravothermal evolution model adopted in this work is calibrated only through the onset of core collapse and can underestimate the central density of halos by at [66]. This can result in underestimated projected enclosed masses and inaccurate deflection-based projected density slopes for subhalos with that are assigned the cored profile in Eq. (11). For , we assign SIDM subhalos the gravothermally collapsed profile from [90], which has a three-dimensional inner power-law slope outside its core of . This allows us to better approximate the lensing-relevant structure of collapsed subhalos without extrapolating the gravothermal evolution model beyond its calibrated range. Therefore, our binary assignment of density profiles to cored and collapsed subhalos in this work should be interpreted as an approximation of different stages of gravothermal evolution, rather than as a prediction of two distinct populations. In addition, the kinetic simulations of gravothermal evolution in [91] found an inner three-dimensional logarithmic density slope of at a late stage of gravothermal collapse, corresponding to under the collapse timescale definition in Eq. (1). In Appendix § 9, we demonstrate that our conclusions in § 4 are robust when assigning this alternative collapsed profile to subhalos with . However, as discussed in the appendix and shown in Figure 1, the two collapsed profiles differ at radii , affecting the distributions of projected enclosed masses and deflection-based projected density slopes of collapsed subhalos with larger masses and smaller pivot radii. Improved modeling of late-stage gravothermal evolution would yield more reliable predictions for the projected structures of SIDM subhalos and enable more robust comparisons with inferred strong-lens perturbers.
Tidal evolution, mass loss, and survival in SIDM: We assume that each SIDM subhalo retains the same fraction of its infall mass as its CDM counterpart by applying a time-dependent truncation radius described in § 2.3. This approach approximates the tidal evolution of SIDM subhalos while their internal density profiles evolve according to the gravothermal evolution model, but it does not capture how the evolving SIDM density profiles in turn affect their mass-loss histories. Two effects are therefore missing. First, the response of a subhalo to the host tidal field depends on its density profile: cored subhalos with shallower gravitational potentials are more vulnerable to tidal stripping and disruption [84, 85], whereas core-collapsed subhalos are more resilient [55]. Second, scattering between subhalo and host particles can cause additional mass loss and modify gravothermal evolution [82, 83, 84]. This effect is strongest when the cross section remains large at the relative velocities between subhalo and host particles, as for the large turnover velocities favored in § 4.4. The importance of both effects depends nontrivially on the subhalo orbit and density profile, the relative velocities between subhalo and host particles, and the velocity and angular dependence of the self-interaction cross section [103, 104, 105]. In addition, we assign an SIDM counterpart to every surviving CDM subhalo, whereas self-interactions can alter which subhalos survive [83]. The “severe” and “minimal” tidal stripping scenarios of § 4.5 provide approximate bounds on these effects, but a self-consistent treatment of mass loss and survival in SIDM is left to future work.
Baryonic effects in CDM and SIDM: All subhalos in this work are evolved in dark-matter-only hosts, without accounting for the baryonic potential of the host galaxy or hydrodynamic processes such as ram-pressure stripping, which can affect subhalo evolution and survival [106, 107, 108]. Baryons can also change the internal structure of subhalos. Hydrodynamic cosmological simulations show that baryonic processes can increase the central densities and concentrations of CDM field halos and subhalos and steepen their inner density slopes relative to dark-matter-only predictions [109, 110, 111, 112]. In particular, [111] found that most CDM subhalos with masses in the TNG50 simulation have three-dimensional logarithmic inner density slopes of . Their dark matter components exhibit similarly steep slopes due to their gravitational response to baryons. Such steepening would shift massive CDM subhalos toward steeper values than in our dark-matter-only realizations, which could make CDM interpretations of steep perturbers less rare. This is particularly relevant for the PJ011646 perturber, whose slope is only slightly steeper than those of typical CDM subhalos of comparable mass (§ 4.1). However, [69] found that the TNG100 subhalos steep enough to match the inferred perturber in SDSS J0946+1006 owe their slopes to adiabatic contraction by stellar masses an order of magnitude above the upper limit implied by the perturber’s inferred luminosity. In SIDM halos, baryons likewise modify the central dark matter profile, with a response that depends on the baryonic mass distribution and the strength of self-interactions [113, 114].
Baryons also affect the lensing inference itself, since stellar mass and light associated with with a perturber can alter its inferred projected enclosed mass and density slope [69, 115]. The collapsed SIDM subhalos that lie approximately within the SDSS J0946+1006 posterior have a median infall mass of (§ 4.2), for which a stellar population is plausible [25, 116, 117]. Such a population is included in neither the gravitational-imaging analysis adopted in this paper [36] nor our subhalo realizations, whereas [115] explicitly modeled a luminous perturber and showed that neglecting its light can substantially bias the inferred structure. Baryonic effects are therefore most important for the massive subhalos relevant to PJ011646 and SDSS J0946+1006. In contrast, the CDM and SIDM subhalos consistent with “detection V” in JVAS B1938+666 have median infall masses of (§ 4.3), well below the mass scale at which star formation becomes efficient [10, 13], so their own baryonic content is likely negligible, although the baryonic potential of the host can still affect their tidal evolution.
Host halo mass of strong-lens systems: The abundance of subhalos depends on the assumed host halo mass [11], which may differ among the strong-lens systems considered in this work. For all three systems, we assume a host virial mass of , representative of the group-scale halos associated with the massive early-type galaxies that dominate galaxy-scale strong-lens samples [20, 72]. Strong-lensing observations primarily constrain the mass distribution within the Einstein radius, while the virial mass of the host halo is typically inferred indirectly and remains uncertain [118]. For SDSS J0946+1006, [119] combined lensing, stellar kinematics, and a concentration–mass prior to infer , although this measurement is model-dependent and requires extrapolation beyond the radial range directly constrained by lensing. In [21], the host-mass and redshift dependence of the projected subhalo mass function normalization was parametrized through the scaling function
which is calibrated against the semi-analytic model Galacticus [61]. At fixed redshift, the projected subhalo abundance scales approximately linearly with host halo mass, . Therefore, the predicted number of subhalos that lie within the posteriors of the three perturbers is sensitive to the assumed host halo mass of each lens system. However, because each SIDM subhalo is mapped from a CDM counterpart, a different host halo mass rescales the abundances of CDM and SIDM subhalos together to first order, so the relative frequency of CDM and SIDM perturber candidates is less sensitive to this assumption than their absolute numbers.
Conclusions
We combined semi-analytic CDM subhalo histories with a model for gravothermal evolution in SIDM to efficiently generate CDM and SIDM subhalo populations across SIDM parameter space and evaluate their projected enclosed masses and deflection-based projected density slopes. We compared these populations with three perturbers inferred through gravitational imaging in PJ011646, SDSS J0946+1006, and JVAS B1938+666, at the pivot radii around which their inferred structures are most robustly constrained. We found that SIDM can produce structures consistent with the three inferred perturbers more frequently than CDM. Under the fiducial SIDM model parameters and , a strong-lens host contains on average CDM, cored SIDM, and no collapsed SIDM subhalos that reproduce the inferred structure of the PJ011646 perturber at a pivot radius of kpc (based on 900 realizations), since collapsed subhalos of comparable mass have steeper slopes than the perturber. In contrast, for SDSS J0946+1006, we find no CDM subhalos in 900 realizations, compared with cored and collapsed SIDM subhalos per host that reproduce the large enclosed mass and steep slope of the perturber at a pivot radius of kpc. For “detection V” in JVAS B1938+666, a host contains on average CDM, cored SIDM, and collapsed SIDM subhalos within the posterior region at a pivot radius of pc (based on 10 realizations). These per-host numbers refer to subhalos anywhere within the host virial radius; the number projected near the lensed images is much smaller (Appendix § 8). Within the wider region, the 23 CDM subhalos found across all realizations require unusually early accretion: relative to all CDM subhalos with projected enclosed masses within of “detection V”, their median infall time corresponds to the earliest , whereas their bound mass fractions are less unusual (lowest ). When varying the SIDM model parameters, we found that simultaneously reproducing the inferred structure of all three perturbers requires a turnover velocity of approximately km/s, so that the massive progenitors of SDSS J0946+1006 perturber candidates can gravothermally collapse, and is achieved for cross-section normalizations among the parameter sets considered. Our results are robust to the alternative tidal stripping prescriptions considered in § 4.5, although the abundance of SIDM subhalos that reproduce the inferred structure of the perturbers is substantially suppressed under severe tidal stripping.
These conclusions are subject to the modeling uncertainties discussed in § 5. In particular, the collapsed SIDM subhalos that reproduce the SDSS J0946+1006 perturber have a median infall mass of , for which a stellar component is plausible but is included in neither the adopted lens model nor our dark-matter-only realizations. Baryonic contraction, also neglected here, could steepen the inner profiles of massive CDM subhalos and increase the number of CDM candidates for the more massive perturbers. In addition, the abundance of SIDM perturber candidates depends on the treatment of late-stage gravothermal evolution and of tidal evolution in SIDM (§ 4.5). The framework developed here can be systematically improved by addressing these effects. For individual perturber detections, such improvements will sharpen comparisons between a detected object and the population of subhalos from which it is drawn, quantifying how typical its inferred structure is under a given dark matter model. For flux-ratio anomalies and other population-level inferences, they will enable more robust predictions of the lensing signal expected for a given SIDM model.
Both individual perturber detections and population-level inferences will become increasingly powerful as current and upcoming surveys substantially increase the known strong-lens population. The Vera C. Rubin Observatory is forecast to discover tens of thousands of galaxy–galaxy lenses, while the Euclid mission is expected to identify galaxy-scale strong lenses over its full survey [120, 121, 122]. Forecasts for the James Webb Space Telescope suggest hundreds of lenses per square degree in sufficiently deep observations [123]. The Nancy Grace Roman Space Telescope is predicted to discover strong lenses, of which may have sufficient signal-to-noise for substructure inference through gravitational imaging [124], with further deep-field observations in the extended lifetime of the telescope substantially increasing the expected sample. Together, these facilities will greatly increase the number of lens systems in which dark matter perturbers can potentially be discovered. Because our framework can efficiently predict the distributions of subhalo projected enclosed masses and deflection-based projected density slopes across SIDM parameter space, lens redshifts, and pivot radii, it is well suited for statistical comparisons between future perturber samples and the predictions of different dark matter models.
Acknowledgments
We thank Oren Slone for useful discussions and for providing the shape of the collapsed profile. RE acknowledges support from the DOE Grant DE-SC0025309 and the Simons Investigator in Physics award MPS-SIP-00010469. DA and JT were supported in part by DOE Grant DE-SC0025309 and the Simons Investigator in Physics award MPS-SIP-00010469. DA is also supported by NSF 2514888. JT and SB also acknowledge support from DOE Grant DE-SC0026113. DG gratefully thanks the Brinson Foundation for their continued support of his research, and acknowledges support from the Kavli Foundation under Grant No. PS-2025-GR-249-3072. ChatGPT (OpenAI) and Claude (Anthropic) were occasionally used in developing and testing code, making suggestions for clarity, grammar, and spelling, and checking the manuscript and citations for errors; the authors verified all output and take full responsibility for the content of this work.
Annular number density of subhalos with perturber-like enclosed masses
In this appendix, we evaluate the orientation-averaged annular number densities of mass-selected CDM and SIDM subhalos relevant to the inferred perturbers in PJ011646 and JVAS B1938+666, under the fiducial SIDM model parameters and . For a subhalo at three-dimensional host-centric radius , the probability that a random viewing orientation places it within a projected annulus is
where is the geometric projection kernel defined in Eq. (14). The number density within the annulus is then
In Figure 12, we show the annular number densities of mass-selected CDM and SIDM subhalos using a fixed annular width of 0.5 arcsec, with each plotted host-centric radius representing the midpoint of an annulus. Subhalos are selected by their projected enclosed mass within the pivot radius of each inferred perturber (§ 2.6), with no selection based on their deflection-based projected density slope (§ 2.5) because the enclosed mass of the inferred perturbers is more robustly constrained than their projected density slopes [34, 35]. Therefore, these annular number densities predict the projected abundances of CDM and SIDM subhalos capable of supplying the relevant enclosed masses. These predicted abundances should not be interpreted as detection rates, which depend on the subhalo positions relative to lensed images and the lensing selection functions determined by data quality and the adopted lens and source models [20, 125].
Figure 12. Orientation-averaged annular number densities (Eq. (A.2)) of mass-selected CDM and SIDM subhalos versus projected host-centric radius for PJ011646 and JVAS B1938+666, under the fiducial SIDM model parameters and . Subhalos are selected by their projected enclosed mass within the pivot radius alone, with no selection on , so these densities also include subhalos whose slopes lie outside the posterior regions of the inferred perturbers. Blue, red, and orange curves denote the annular number densities of CDM, cored SIDM, and collapsed SIDM subhalos, respectively. Panel (a): PJ011646 (, ), selecting subhalos with , averaged over realizations. Panel (b): JVAS B1938+666 (, ), selecting subhalos with , averaged over realizations. The annular width is fixed at 0.5 arcsec, with the host-centric radius corresponding to the midpoint of an annulus. Shaded bands denote the realization scatter. The upper axes show the corresponding host-centric radius in projected physical distances at the lens redshift of each system.
Panel (a) selects subhalos with projected enclosed masses at relevant for the inferred perturber in PJ011646. The fraction of SIDM subhalos under this mass selection that are collapsed at the lens redshift is , while the remainder are cored. Averaged over 900 realizations, the annular number densities of mass-selected CDM and collapsed SIDM subhalos are comparable, –, while those of cored SIDM subhalos are lower by a factor of , demonstrating the rarity of these massive subhalos within the annulus. The realization-scatter is comparable to or larger than the difference between the model families. These annular densities reflect the mass selection alone. When the slope is also required to lie within the posterior region (§ 4.1), the ranking of the three families changes: collapsed subhalos are too steep and CDM subhalos are mostly too shallow, whereas cored subhalos, despite being the least abundant in this mass-selected sample, most often have slopes consistent with the inferred perturber.
On the other hand, for “detection V” in JVAS B1938+666, the differences between the annular number densities of CDM, cored, and collapsed subhalos exceed the realization-scatter under the fiducial SIDM model parameters. Panel (b) selects subhalos with enclosed masses at . Under this mass selection, the fraction of SIDM subhalos that are collapsed at is , while the rest are cored. Across 10 realizations, the annular number densities at arcsec from the host center are arcsec for CDM subhalos, while their cored and collapsed counterparts have densities arcsec and arcsec, respectively. The number densities in JVAS B1938+666 are orders of magnitude larger than those in PJ011646, primarily because the mass selection probes much lower mass, more abundant subhalos. In addition, the fixed annular width of 0.5 arcsec corresponds to a larger projected physical distance at ( kpc) than at ( kpc). For the low-mass subhalos relevant for the inferred perturber in JVAS B1938+666 under the fiducial SIDM model parameters, the annular abundance of collapsed subhalos is significantly larger than that of CDM or cored subhalos.
Comparison of collapsed SIDM profiles
In Figure 13, we test the sensitivity of the projected structure of SIDM subhalos to the choice of gravothermally collapsed density profile assigned to subhalos that have exceeded the adopted collapse threshold . In all three panels, we show in orange the projected enclosed masses and deflection-based projected density slopes of subhalos assigned a collapsed profile from [90] (§ 2.3). For comparison, we show in purple the corresponding projected structures of subhalos assigned the collapsed profile from Figure 1 of [91] at the final simulation snapshot , where is the characteristic scattering timescale of the initial NFW halo; under the collapse-timescale convention of Eq. (1), this corresponds to . In both cases, we apply the fiducial tidal truncation prescription described in § 2.3. The black posterior regions correspond to the inferred perturbers in PJ011646, SDSS J0946+1006, and JVAS B1938+666. The choice of collapsed density profile primarily affects the distributions of higher-mass subhalos at smaller pivot radii, since the two profiles differ significantly at radii , as shown in Figure 1. For example, at the pivot radius of 90 pc in panel (b), the higher-mass subhalos assigned the profile from [90] exhibit shallower projected density slopes than those assigned the profile from [91]. However, none of the inferred perturbers discussed in § 4 have a sufficiently large projected enclosed mass at a sufficiently small pivot radius for these differences to substantially affect our results. We therefore find that our conclusions from § 4 are robust to either choice of collapsed density profile.
Figure 13. Comparison of the projected enclosed masses (Eq. (13)) and deflection-based projected density slopes (Eq. (15)) of collapsed SIDM subhalos and inferred strong-lens perturbers, using two different collapsed density profiles and the same fiducial tidal truncation prescription described in § 2.3, under the fiducial SIDM model parameters and . Orange denotes subhalos assigned a collapsed density profile from [90], and purple denotes subhalos assigned a collapsed density profile from Figure 1 of [91]. Panel (a): versus for realizations alongside the posterior region of the inferred perturber in PJ011646 [34] (, ). Panel (b): versus for realizations alongside the posterior (PL region) of detection in JVAS B1938+666 [35] (, ). Panel (c): versus for realizations alongside the posterior region of the inferred perturber in SDSS J0946+1006 [36] (, ). Horizontal bars denote the central 68% interval of in bins of of width 0.3 dex. Our conclusions from § 4 are robust to the choice of collapsed density profile.
Comparison of SatGen and pyHalo CDM predictions
In this appendix, we generate CDM subhalo realizations of host halos with at using SatGen [62, 63] and pyHalo [21] to compare their lensing-relevant predictions. In Section C.1, we show that the tidally evolved, projected subhalo mass function is consistent between CDM realizations generated using both frameworks, and in Section C.2, we show that their distributions of subhalo lensing convergence are consistent within a circular subhalo-centric aperture with radius 10 milliarcseconds (mas). While these lensing-relevant quantities are not used in this paper for comparison with strong-lens perturbers, they provide a consistency check between the two distinct modeling frameworks.
Projected subhalo mass function
Strong-lensing analyses commonly describe the abundance of subhalos near the Einstein radius of a strong-lens system using a projected subhalo mass function (SHMF) [21, 126, 127].
We adopt the parameterization
where is the projected SHMF normalization, is the power-law slope, and we set . With the definition , Eq. (C.1) can be written as
We infer and from the number of CDM subhalos in the two adjacent mass bins and . For a circular aperture with area , the expected number of subhalos in mass bin is
where
Assuming that the counts in the two mass bins are independently Poisson distributed, maximizing their joint likelihood gives
and
where and are the subhalo counts in the two mass bins.
We estimate from 24 CDM subhalo realizations generated using SatGen and pyHalo. Figure 14 shows as a function of host-centric circular aperture radius. For each framework, we estimate using both the subhalo masses at infall and the evolved bound masses at the lens redshift. Using first-order error propagation, we verified that the uncertainty associated with the two-bin fit is negligible compared to the scatter across realizations for the aperture radii considered. The weak dependence of the inferred normalization on aperture radius indicates that the projected subhalo number density is approximately uniform over the radial range considered, consistent with theoretical expectations [64].
Figure 14. Projected subhalo mass function normalization (Eq. (C.6)) versus host-centric circular aperture radius for CDM subhalos generated using SatGen and pyHalo, averaged over 24 subhalo realizations of host halos with at . Solid lines use subhalo infall masses, while dashed lines use evolved bound masses at the lens redshift. The pyHalo curves correspond to abundance calibrations based on either the Symphony [128] cosmological simulations (orange) or the Galacticus [61] semi-analytic model (blue). Shaded bands show the realization scatter.
The SHMF normalizations inferred using infall masses from SatGen and pyHalo are not expected to coincide because the two frameworks generate subhalo populations differently. SatGen explicitly generates merger trees and tidally evolves individual subhalos from infall to the lens redshift, as described in § 2.1. In contrast, pyHalo generates subhalo populations at the lens redshift without following individual orbital histories. Infall masses are sampled in projection using Eq. (C.1), whose normalization is specified as an input parameter [126]. Bound mass fractions are sampled from the tidal-evolution model of [65], which is calibrated using Galacticus [61] and implemented in pyHalo. Tidal tracks then map the sampled bound mass fraction onto rescaled and truncated NFW profiles at the lens redshift. Following [126], we adopt the values of for their Galacticus [61] and Symphony [128] abundance priors, and , corresponding to the blue and orange solid lines in Figure 14, respectively. After applying the tidal-evolution prescription in pyHalo, we use Eq. (C.6) to infer the corresponding evolved normalizations, shown by the dashed blue and orange curves. Despite SatGen and pyHalo using different tidal evolution prescriptions, these frameworks predict consistent evolved SHMF normalizations, and hence consistent projected subhalo abundances over the aperture radii relevant for group-scale strong lenses.
Distribution of subhalo convergence
We compare the lensing strength of CDM subhalos between SatGen and pyHalo by considering their mean convergence within a 10-milliarcsecond subhalo-centric aperture, which is representative of the angular impact parameter over which low-mass halos produce several-percent perturbations to the magnifications of lensed-quasar images [126, 129]. For a circularly symmetric subhalo density profile, the convergence profile is defined as
where is the projected surface mass density profile and is the critical surface density for lensing [92]. The integrated convergence within a circular subhalo-centric aperture with radius is
The aperture-averaged convergence is then
and the corresponding magnitude of the deflection angle at radius satisfies
for an axisymmetric lens [92]. Thus, the aperture-averaged convergence determines the deflection angle produced by the subhalo mass profile, which is closely related to the localized image distortions generated by dark matter subhalos inferred in strong-lensing analyses [24, 38, 94].
Figure 15 shows distributions of aperture-averaged convergence for CDM subhalos generated using SatGen and pyHalo under both Symphony and Galacticus priors described in Sec. C.1. SatGen CDM subhalos are modeled using the density profile in Eq. (10), which is implemented in lenstronomy [95, 96]. We use this profile to construct their convergence profiles and calculate using Eq. (C.9) with . pyHalo CDM subhalos are modeled with rescaled truncated-NFW profiles described in [126], whose parameters are used to construct the corresponding truncated-NFW profile using lenstronomy. We then analogously calculate for each subhalo. We use 24 subhalo realizations of host halos with at lens redshift and source redshift , over which the counts in each convergence bin are averaged. We only use evolved subhalos with bound masses , whose projected positions lie within a circular 3.3 arcsec host-centric aperture. SatGen and pyHalo predict approximately consistent distributions of subhalo convergence despite using different tidal evolution prescriptions, and hence both frameworks predict comparable distributions of subhalo deflection angles at 10 mas.
Figure 15. Histograms of aperture-averaged convergence (Eq. (C.9)) within a circular 10 mas subhalo-centric aperture of CDM subhalos generated using SatGen and pyHalo under Symphony or Galacticus model priors. We use 24 subhalo realizations of host halos with at lens redshift and source redshift . is the realization-averaged number of subhalos in log-convergence bins of width 0.1 dex. At the lens redshift, only subhalos with evolved bound masses that lie within a circular projected host-centric aperture of 3.3 arcsec are considered.
Summary of PJ011646, SDSS J0946+1006, and JVAS B1938+666
We summarize the observational history and inferred properties of the three candidate strong-lens perturbers in PJ011646, SDSS J0946+1006, and JVAS B1938+666, emphasizing the gravitational imaging analyses from which we adopt their projected enclosed mass and density slope posteriors. PJ011646 is a strong-lens system with [130] and identified as part of the PASSAGES sample of Planck-selected sources [131]. Using high-resolution ALMA Band 7 dust-continuum observations, the detection of a potential dark matter perturber was reported in [34]. The interferometric data were modeled in the plane using a smooth elliptical power-law (EPL) mass distribution with external shear and third- and fourth-order multipoles. A grid-based search in the image plane for a spherical NFW halo identified a candidate, and the model including the perturber was preferred over the fiducial smooth macromodel by , corresponding to a detection significance of approximately . Assuming that the perturber lies at the lens redshift, the analysis reported and , with a concentration that they argued is consistent with a typically tidally stripped CDM subhalo (see § 4.1 for our interpretation). The projected enclosed mass was most robustly constrained at an angular pivot radius of , within which they inferred .
SDSS J0946+1006, also known as the “Jackpot” lens, is a strong-lens system with and three lensed background sources. The system was originally observed as a double Einstein ring in HST/ACS imaging with an inner source at redshift [132]. The second source redshift was subsequently measured to be using VLT/X-Shooter and MUSE observations [133], while a third multiply imaged source at was discovered using VLT/MUSE data [134]. The first gravitational-imaging detection of a dark perturber in this system was reported by [30] using HST/ACS F814W data. Their non-parametric reconstruction identified a localized perturbation near the inner Einstein ring, which they modeled with a tidally truncated pseudo-Jaffe profile. The analysis inferred a perturber mass of with no luminous counterpart detected in the image. The perturber was then analyzed in several subsequent studies using various lens and source models [25, 28, 58, 69, 117, 125]. However, its inferred structural properties remain sensitive to modeling assumptions such as macromodel flexibility and source regularization [25, 117], and whether the perturber is assumed to be completely dark or allowed to host faint stellar luminosity [115]. Subsequent analyses have investigated its consistency with CDM predictions and possible SIDM interpretations [25, 28, 58, 116]. The HST F814W data were reanalyzed by [36] using image-pixel supersampling and double-source-plane modeling. The analysis recovered a detection that is substantially less sensitive to the source-regularization prior, obtaining a log-Bayes factor , corresponding approximately to a detection. At 95% confidence, the projected enclosed mass within 1 kpc was constrained to , while the average logarithmic slope of the local surface density, , over 0.75–1.25 kpc was constrained to be steeper than .
JVAS B1938+666 is a radio and near-infrared strong-lens system with lens redshift and background source redshift [135, 136, 137, 138]. Using Keck-II/NIRC2 adaptive-optics observations, the detection of a perturber was reported in [31] with approximately confidence. This object, commonly referred to as detection A, was subsequently inferred to be a background halo at redshift [139]. More recently, however, the Keck adaptive-optics data were reanalyzed in [25] and detection A was found to be a foreground halo at redshift . A second substantially lower-mass perturber, commonly called detection V, was claimed to be discovered by [29] using 1.7 GHz VLBI observations with an angular resolution of approximately 5 mas. It was found along an extremely thin mas-long lensed arc using a non-parametric approach, while independent parametric modeling confirmed detection V with approximately confidence and measured a projected enclosed mass [29].
In [35], detection V was analyzed using a broad set of density profiles and perturber-redshift assumptions. Although its projected density slope is sensitive to the assumed profile, the projected enclosed mass is most robustly constrained at pc. They found that single-component NFW models require concentrations substantially larger than CDM expectations, and discussed whether the compact structure could instead be a core-collapsed SIDM halo, an ultra-compact dwarf galaxy containing a central black hole, or a nuclear star cluster.
References
- [1]Planck Collaboration, Aghanim, N., Akrami, Y., Ashdown, M., Aumont, J., Baccigalupi, C. et al., Planck 2018 results - VI. cosmological parameters, Astronomy & Astrophysics 641 (2020) A6.
- [2]T.M.C. Abbott, M. Aguena, A. Alarcon, S. Allam, Q. Alves, A. Amon et al., Dark energy survey year 3 results: Cosmological constraints from galaxy clustering and weak lensing, Phys. Rev. D 105 (2022) 023520.DOI
- [3]S. Alam, M. Aubert, S. Avila, C. Balland, J.E. Bautista, M.A. Bershady et al., Completed SDSS-IV extended baryon oscillation spectroscopic survey: Cosmological implications from two decades of spectroscopic surveys at the Apache Point Observatory, Phys. Rev. D 103 (2021) 085353.
- [4]C. Lacey and S. Cole, Merger rates in hierarchical models of galaxy formation, Monthly Notices of the Royal Astronomical Society 262 (1993) 627.
- [5]S.D.M. White and M.J. Rees, Core condensation in heavy halos: a two-stage theory for galaxy formation and clustering, Monthly Notices of the Royal Astronomical Society 183 (1978) 341.DOI
- [6]W.H. Press and P. Schechter, Formation of Galaxies and Clusters of Galaxies by Self-Similar Gravitational Condensation, The Astrophysical Journal 187 (1974) 425.DOI
- [7]J.F. Navarro, C.S. Frenk and S.D.M. White, A universal density profile from hierarchical clustering, The Astrophysical Journal 490 (1997) 493.
- [8]A.A. Dutton and A.V. Macciò, Cold dark matter haloes in the Planck era: evolution of structural parameters for Einasto and NFW profiles, Monthly Notices of the Royal Astronomical Society 441 (2014) 3359.DOI
- [9]J. Tinker, A.V. Kravtsov, A. Klypin, K. Abazajian, M. Warren, G. Yepes et al., Toward a halo mass function for precision cosmology: The limits of universality, The Astrophysical Journal 688 (2008) 709.DOI
- [10]L. Gao, S.D.M. White, A. Jenkins, F. Stoehr and V. Springel, The subhalo populations of ΛCDM dark haloes, Monthly Notices of the Royal Astronomical Society 355 (2004) 819.DOI
- [11]J. Han, S. Cole, C.S. Frenk and Y. Jing, A unified model for the spatial and mass distribution of subhaloes, Monthly Notices of the Royal Astronomical Society 457 (2016) 1208.DOI
- [12]J. Zavala and C.S. Frenk, Dark matter haloes and subhaloes, Galaxies 7 (2019) 81.arxiv.org/abs/1907.11775
- [13]T. Sawala, C.S. Frenk, A. Fattahi, J.F. Navarro, T. Theuns, R.G. Bower et al., The chosen few: the low-mass haloes that host faint galaxies, Monthly Notices of the Royal Astronomical Society 456 (2016) 85.DOI
- [14]T. Okamoto, L. Gao and T. Theuns, Mass loss of galaxies due to an ultraviolet background, Monthly Notices of the Royal Astronomical Society 390 (2008) 920.DOI
- [15]P. Bode, J.P. Ostriker and N. Turok, Halo formation in warm dark matter models, The Astrophysical Journal 556 (2001) 93.
- [16]A. Schneider, R.E. Smith, A.V. Macciò and B. Moore, Non-linear evolution of cosmological structures in warm dark matter models, Monthly Notices of the Royal Astronomical Society 424 (2012) 684.DOI
- [17]A.D. Ludlow, S. Bose, R.E. Angulo, L. Wang, W.A. Hellwing, J.F. Navarro et al., The mass–concentration–redshift relation of cold and warm dark matter haloes, Monthly Notices of the Royal Astronomical Society 460 (2016) 1214.DOI
- [18]S. Elgamal, M. Nori, A.V. Macciò, M. Baldi and S. Waterval, No catch-22 for fuzzy dark matter: testing substructure counts and core sizes via high-resolution cosmological simulations, Monthly Notices of the Royal Astronomical Society 532 (2024) 4050.arxiv.org/abs/2311.03591
- [19]E.O. Nadler, D. Kong, D. Yang and H.-B. Yu, SIDM Concerto: Compilation and data release of self-interacting dark matter zoom-in simulations, The Astrophysical Journal 991 (2025) 69.DOI
- [20]S. Vegetti, S. Birrer, G. Despali, C.D. Fassnacht, D. Gilman, Y. Hezaveh et al., Strong gravitational lensing as a probe of dark matter, Space Science Reviews 230 (2024) 58.
- [21]D. Gilman, S. Birrer, A. Nierenberg, T. Treu, X. Du and A. Benson, Warm dark matter chills out: constraints on the halo mass function and the free-streaming length of dark matter with eight quadruple-image strong gravitational lenses, Monthly Notices of the Royal Astronomical Society 491 (2020) 6077–6101.DOI
- [22]D. Gilman, J. Bovy, T. Treu, A. Nierenberg, S. Birrer, A. Benson et al., Strong lensing signatures of self-interacting dark matter in low-mass haloes, Monthly Notices of the Royal Astronomical Society 507 (2021) 2432.DOI
- [23]D. Gilman, A.M. Nierenberg, J. Gurian, H. Paugnat, C. Gannon, M.N. Martinez et al., JWST lensed quasar dark matter survey V: Hints of self-interacting dark matter from 29 quadruply imaged quasars, arXiv e-prints (2026) arXiv:2609.35974.arxiv.org/abs/2609.35974
- [24]S. Vegetti and L.V.E. Koopmans, Bayesian strong gravitational-lens modelling on adaptive grids: objective detection of mass substructure in galaxies, Monthly Notices of the Royal Astronomical Society 392 (2009) 945.DOI
- [25]M. Tajalli, S. Vegetti, C.M. O’Riordan, S.D.M. White, C.D. Fassnacht, D.M. Powell et al., SHARP–IX: the dense, low-mass perturbers in B1938+666 and J0946+1006: implications for cold and self-interacting dark matter, Monthly Notices of the Royal Astronomical Society 543 (2025) 540–564.DOI
- [26]D. Gilman, A. Benson, J. Bovy, S. Birrer, T. Treu and A. Nierenberg, The primordial matter power spectrum on sub-galactic scales, Monthly Notices of the Royal Astronomical Society 512 (2022) 3163.DOI
- [27]D. Gilman, X. Du, A. Benson, S. Birrer, A. Nierenberg and T. Treu, Constraints on the mass–concentration relation of cold dark matter haloes with 11 strong gravitational lenses, Monthly Notices of the Royal Astronomical Society: Letters 492 (2020) L12.DOI
- [28]G. Despali, F.M. Heinze, C.D. Fassnacht, S. Vegetti, C. Spingola, R. Klessen et al., Detecting low-mass haloes with strong gravitational lensing: II. constraints on the density profiles of two detected subhaloes, Astronomy & Astrophysics 699 (2025) A222.DOI
- [29]D.M. Powell, J.P. McKean, S. Vegetti, C. Spingola, S.D.M. White and C.D. Fassnacht, A million-solar-mass object detected at a cosmological distance using gravitational imaging, Nature Astronomy 9 (2025) 1714–1722.DOI
- [30]S. Vegetti, L.V.E. Koopmans, A. Bolton, T. Treu and R. Gavazzi, Detection of a dark substructure through gravitational imaging, Monthly Notices of the Royal Astronomical Society 408 (2010) 1969–1981.DOI
- [31]S. Vegetti, D.J. Lagattuta, J.P. McKean, M.W. Auger, C.D. Fassnacht and L.V.E. Koopmans, Gravitational detection of a low-mass dark satellite galaxy at cosmological distance, Nature 481 (2012) 341–343.
- [32]Y.D. Hezaveh, N. Dalal, D.P. Marrone, Y.-Y. Mao, W. Morningstar, D. Wen et al., Detection of lensing substructure using ALMA observations of the dusty galaxy SDP.81, The Astrophysical Journal 823 (2016) 37.DOI
- [33]S.C. Lange, A. Amvrosiadis, J.W. Nightingale, Q. He, C.S. Frenk, A. Robertson et al., Galaxy mass modelling from multiwavelength JWST strong lens analysis: dark matter substructure, angular mass complexity, or both?, Monthly Notices of the Royal Astronomical Society 539 (2025) 704.arxiv.org/abs/2410.12987
- [34]A. Amvrosiadis, J.W. Nightingale, Q. He, A. Robertson, S. Cole, C.S. Frenk et al., Detection of a dark matter subhalo in the strongly lensed system J9011646, arXiv e-prints (2026) arXiv:2605.21212.
- [35]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 Astronomy 10 (2026) 440–447.DOI
- [36]Q.E. Minor, High significance detection of the dark substructure in gravitational lens SDSS J0946+1006 by image pixel supersampling, The Astrophysical Journal 981 (2025) 2.DOI
- [37]A. Diaz Rivero, F.-Y. Cyr-Racine and C. Dvorkin, Power spectrum of dark matter substructure in strong gravitational lenses, Physical Review D 97 (2018) 023001.DOI
- [38]A.Ç. Şengül and C. Dvorkin, Probing dark matter with strong gravitational lensing through an effective density slope, Monthly Notices of the Royal Astronomical Society 516 (2022) 336.DOI
- [39]K.E. Kollmann, J.W. Nightingale, M. Lisanti, A. Robertson and O. Slone, Using strong lensing to detect subhaloes with steep inner density profiles, Monthly Notices of the Royal Astronomical Society 546 (2026) stag066.DOI
- [40]D. Kong, E.O. Nadler and H.-B. Yu, Strong-lensing perturber signatures in self-interacting dark matter simulations, Phys. Rev. D 114 (2026) 043011.DOI
- [41]D.N. Spergel and P.J. Steinhardt, Observational evidence for self-interacting cold dark matter, Physical Review Letters 84 (2000) 3760–3763.pubmed.ncbi.nlm.nih.gov/11019199
- [42]S. Tulin and H.-B. Yu, Dark matter self-interactions and small scale structure, Physics Reports 730 (2018) 1–57.DOI
- [43]D. Lynden-Bell and R. Wood, The gravo-thermal catastrophe in isothermal spheres and the onset of red-giant structure for stellar systems, Monthly Notices of the Royal Astronomical Society 138 (1968) 495.
- [44]D. Lynden-Bell and P.P. Eggleton, On the consequences of the gravothermal catastrophe, Monthly Notices of the Royal Astronomical Society 191 (1980) 483.
- [45]D. Yang and H.-B. Yu, Gravothermal evolution of dark matter halos with differential elastic scattering, Journal of Cosmology and Astroparticle Physics 09 (2022) 077.DOI
- [46]N.J. Outmezguine, K.K. Boddy, S. Gad-Nasr, M. Kaplinghat and L. Sagunski, Universal gravothermal evolution of isolated self-interacting dark matter halos for velocity-dependent cross-sections, Monthly Notices of the Royal Astronomical Society 523 (2023) 4786.DOI
- [47]S. Balberg, S.L. Shapiro and S. Inagaki, Self-interacting dark matter halos and the gravothermal catastrophe, The Astrophysical Journal 568 (2002) 475.DOI
- [48]J. Koda and P.R. Shapiro, Gravothermal collapse of isolated self-interacting dark matter haloes: N-body simulation versus the fluid model, Monthly Notices of the Royal Astronomical Society 415 (2011) 1125.
- [49]R. Essig, S.D. McDermott, H.-B. Yu and Y.-M. Zhong, Constraining dissipative dark matter self-interactions, Physical Review Letters 123 (2019) 121102.DOI
- [50]H. Nishikawa, K.K. Boddy and M. Kaplinghat, Accelerated core collapse in tidally stripped self-interacting dark matter halos, Phys. Rev. D 101 (2020) 063009.DOI
- [51]S. Adhikari, A. Banerjee, K.K. Boddy, F.-Y. Cyr-Racine, H. Desmond, C. Dvorkin et al., Astrophysical tests of dark matter self-interactions, Reviews of Modern Physics 97 (2025) 045004.DOI
- [52]S. Balberg and S.L. Shapiro, Gravothermal collapse of self-interacting dark matter halos and the origin of massive black holes, Phys. Rev. Lett. 88 (2002) 101301.pubmed.ncbi.nlm.nih.gov/11909338
- [53]J. Pollack, D.N. Spergel and P.J. Steinhardt, Supermassive black holes from ultra-strongly self-interacting dark matter, The Astrophysical Journal 804 (2015) 131.DOI
- [54]F. Jiang, Z. Jia, H. Zheng, L.C. Ho, K. Inayoshi, X. Shen et al., Formation of the little red dots from the core collapse of self-interacting dark matter halos, The Astrophysical Journal Letters 996 (2025) L19.arxiv.org/abs/2503.23710
- [55]E.O. Nadler, D. Yang and H.-B. Yu, A self-interacting dark matter solution to the extreme diversity of low-mass halo properties, The Astrophysical Journal Letters 958 (2023) L39.DOI
- [56]C. Mace, B. Dhanasingham, Z.C. Zeng, F.-Y. Cyr-Racine, X. Du, A.H.G. Peter et al., The sensitivity of substructure lensing to SIDM core-collapse model variation, arXiv e-prints (2026) arXiv:2605.24174.arxiv.org/abs/2605.24174
- [57]D. Yang and H.-B. Yu, Self-interacting dark matter and small-scale gravitational lenses in galaxy clusters, Phys. Rev. D 104 (2021) 103031.DOI
- [58]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?, Monthly Notices of the Royal Astronomical Society 540 (2025) 247.arxiv.org/abs/2411.08565
- [59]C. Mace, Z.C. Zeng, A.H.G. Peter, X. Du, S. Yang, A. Benson et al., Convergence tests of self-interacting dark matter simulations, Phys. Rev. D 110 (2024) 123024.arxiv.org/abs/2402.01604
- [60]I. Palubski, O. Slone, M. Kaplinghat, M. Lisanti and F. Jiang, Numerical challenges in modeling gravothermal collapse in self-interacting dark matter halos, Journal of Cosmology and Astroparticle Physics 09 (2024) 074.DOI
- [61]A.J. Benson, Galacticus: A semi-analytic model of galaxy formation, New Astronomy 17 (2012) 175–197.DOI
- [62]F. Jiang, A. Dekel, J. Freundlich, F.C. van den Bosch, S.B. Green, P.F. Hopkins et al., SatGen: a semi-analytical satellite galaxy generator – I. the model and its application to local-group satellite statistics, Monthly Notices of the Royal Astronomical Society 502 (2021) 621–641.DOI
- [63]S.B. Green, F.C. van den Bosch and F. Jiang, SatGen – II. assessing the impact of a disc potential on subhalo populations, Monthly Notices of the Royal Astronomical Society 509 (2022) 2624–2636.DOI
- [64]C. Gannon, A. Nierenberg, A. Benson, R. Keeley, X. Du and D. Gilman, Dark matter substructure: A lensing perspective, Phys. Rev. D 112 (2025) 023532.arxiv.org/abs/2501.17362
- [65]X. Du, D. Gilman, T. Treu, A. Benson and C. Gannon, Empirical model for the tidal evolution of dark matter substructure around strong gravitational lenses, Physical Review D 112 (2025) 023009.DOI
- [66]D. Yang, E.O. Nadler, H.-B. Yu and Y.-M. Zhong, A parametric model for self-interacting dark matter halos, Journal of Cosmology and Astroparticle Physics 02 (2024) 032.DOI
- [67]D. Yang, E.O. Nadler and H.-B. Yu, Strong dark matter self-interactions diversify halo populations within and surrounding the Milky Way, The Astrophysical Journal 949 (2023) 67.DOI
- [68]S. Ando, S. Horigome, E.O. Nadler, D. Yang and H.-B. Yu, SASHIMI-SIDM: semi-analytical subhalo modelling for self-interacting dark matter at sub-galactic scales, Journal of Cosmology and Astroparticle Physics 02 (2025) 053.DOI
- [69]Q. Minor, S. Gad-Nasr, M. Kaplinghat and S. Vegetti, An unexpected high concentration for the dark substructure in the gravitational lens SDSSJ0946+1006, Monthly Notices of the Royal Astronomical Society 507 (2021) 1662.DOI
- [70]X. Zhang and H.-B. Yu, SIDM and CDM interpretations of the million-solar-mass lensing perturber JVAS B1938+666-6V, Phys. Rev. D 114 (2026) 083015.
- [71]H.-B. Yu, Core-collapsed SIDM halos as the common origin of dense perturbers in lenses, streams, and satellites, Phys. Rev. Lett. 136 (2026) 141001.
- [72]A. Sonnenfeld, A. Leauthaud, M.W. Auger, R. Gavazzi, T. Treu, S. More et al., Evidence for radial variations in the stellar mass-to-light ratio of massive galaxies from weak and strong lensing, Monthly Notices of the Royal Astronomical Society 481 (2018) 164.DOI
- [73]G.L. Bryan and M.L. Norman, Statistical properties of x-ray clusters: Analytic and numerical comparisons, The Astrophysical Journal 495 (1998) 80–99.
- [74]J.R. Bond, S. Cole, G. Efstathiou and N. Kaiser, Excursion Set Mass Functions for Hierarchical Gaussian Fluctuations, The Astrophysical Journal 379 (1991) 440.DOI
- [75]R.G. Bower, The evolution of groups of galaxies in the press–schechter formalism, Monthly Notices of the Royal Astronomical Society 248 (1991) 332.DOI
- [76]H. Parkinson, S. Cole and J. Helly, Generating dark matter halo merger trees, Monthly Notices of the Royal Astronomical Society 383 (2008) 557.
- [77]A.J. Benson, The mass function of unprocessed dark matter haloes and merger tree branching rates, Monthly Notices of the Royal Astronomical Society 467 (2017) 3454.DOI
- [78]D.H. Zhao, Y.P. Jing, H.J. Mo and G. Börner, Accurate universal models for the mass accretion histories and concentrations of dark matter halos, The Astrophysical Journal 707 (2009) 354–369.DOI
- [79]Z.-Z. Li, D.-H. Zhao, Y.P. Jing, J. Han and F.-Y. Dong, Orbital distribution of infalling satellite halos across cosmic time, The Astrophysical Journal 905 (2020) 177.DOI
- [80]S.B. Green, F.C. van den Bosch and F. Jiang, The tidal evolution of dark matter substructure – II. the impact of artificial disruption on subhalo mass functions and radial profiles, Monthly Notices of the Royal Astronomical Society 503 (2021) 4075.DOI
- [81]H.C. Turner, M.R. Lovell, J. Zavala and M. Vogelsberger, The onset of gravothermal core collapse in velocity-dependent self-interacting dark matter subhaloes, Monthly Notices of the Royal Astronomical Society 505 (2021) 5327.DOI
- [82]Z.C. Zeng, A.H.G. Peter, X. Du, A. Benson, S. Kim, F. Jiang et al., Core-collapse, evaporation, and tidal effects: the life story of a self-interacting dark matter subhalo, Monthly Notices of the Royal Astronomical Society 513 (2022) 4845.DOI
- [83]E.O. Nadler, A. Banerjee, S. Adhikari, Y.-Y. Mao and R.H. Wechsler, Signatures of velocity-dependent dark matter self-interactions in Milky Way-mass halos, The Astrophysical Journal 896 (2020) 112.DOI
- [84]D. Klemmer, M.S. Fischer, K.K. Boddy, M. Kaplinghat and L. Sagunski, Survival of the most compact: the life and death of satellite halos in self-interacting dark matter, arXiv e-prints (2026) arXiv:2603.19362.arxiv.org/abs/2603.19362
- [85]R. Errani, J.F. Navarro, J. Peñarrubia, B. Famaey and R. Ibata, Dark matter halo cores and the tidal survival of Milky Way satellites, Monthly Notices of the Royal Astronomical Society 519 (2023) 384–396.DOI
- [86]S.B. Green and F.C. van den Bosch, The tidal evolution of dark matter substructure – I. subhalo density profiles, Monthly Notices of the Royal Astronomical Society 490 (2019) 2091–2101.DOI
- [87]S. Raut, E. Nadler and A. Benson, An extended parametric model for self-interacting dark matter halos, arXiv e-prints (2026) arXiv:2604.22013.arxiv.org/abs/2604.22013
- [88]D. Yang, E.O. Nadler and H.-B. Yu, Testing the parametric model for self-interacting dark matter using matched halos in cosmological simulations, Physics of the Dark Universe 47 (2025) 101807.DOI
- [89]G. Ogiya, F.C. van den Bosch, O. Hahn, S.B. Green, T.B. Miller and A. Burkert, DASH: a library of dynamical subhalo evolution, Monthly Notices of the Royal Astronomical Society 485 (2019) 189–202.DOI
- [90]O. Slone, F. Jiang, I. Palubski, M. Lisanti and M. Kaplinghat, “A toolbox for predicting signatures of self-interacting dark matter halos.” 2026 (in preparation).
- [91]J. Gurian and S. May, Core collapse beyond the fluid approximation: The late evolution of self-interacting dark matter halos, Phys. Rev. Lett. 135 (2025) 222001.DOI
- [92]P. Saha, D. Sluse, J. Wagner and L.L.R. Williams, Essentials of strong gravitational lensing, Space Science Reviews 220 (2024) 12.arxiv.org/abs/2401.04165
- [93]M. Tajalli and S. Vegetti, On what scales does strong lensing robustly constrain the mass profile of low-mass perturbers?, arXiv e-prints (2026) arXiv:2608.20454.DOI
- [94]Q.E. Minor, M. Kaplinghat and N. Li, A robust mass estimator for dark matter subhalo perturbations in strong gravitational lenses, The Astrophysical Journal 845 (2017) 118.DOI
- [95]S. Birrer and A. Amara, lenstronomy: Multi-purpose gravitational lens modelling software package, Physics of the Dark Universe 22 (2018) 189–201.
- [96]S. Birrer, A. Shajib, D. Gilman, A. Galan, J. Aalbers, M. Millon et al., lenstronomy II: A gravitational lensing software ecosystem, Journal of Open Source Software 6 (2021) 3283.
- [97]Astropy Collaboration, A.M. Price-Whelan, P.L. Lim, N. Earl, N. Starkman, L. Bradley et al., The Astropy Project: Sustaining and Growing a Community-oriented Open-source Project and the Latest Major Release (v5.0) of the Core Package, Astrophys. J. 935 (2022) 167 [2206.14220].
- [98]F. Jiang and F.C. van den Bosch, Statistics of dark matter substructure – I. model and universal fitting functions, Monthly Notices of the Royal Astronomical Society 458 (2016) 2848.arxiv.org/abs/1403.6827
- [99]R.J. Smith, J.R. Lucey and C. Conroy, The SINFONI nearby elliptical lens locator survey: discovery of two new low-redshift strong lenses and implications for the initial mass function in giant early-type galaxies, Monthly Notices of the Royal Astronomical Society 449 (2015) 3441.DOI
- [100]Mercier, W., Shuntov, M., Gavazzi, R., Nightingale, J. W., Arango, R., Ilbert, O. et al., The COSMOS-Web ring: In-depth characterization of an Einstein ring lensing system at z ∼ 2, Astronomy & Astrophysics 687 (2024) A61.DOI
- [101]J.W. Nightingale, G. Mahler, J. McCleary, Q. He, N.B. Hogg, A. Amvrosiadis et al., The cosmos-web lens survey COWLS I: discovery of > 100 high redshift strong lenses in contiguous JWST imaging, Monthly Notices of the Royal Astronomical Society 543 (2025) 203.
- [102]V. Springel, J. Wang, M. Vogelsberger, A. Ludlow, A. Jenkins, A. Helmi et al., The Aquarius Project: the subhalos of galactic halos, Mon. Not. Roy. Astron. Soc. 391 (2008) 1685 [0809.0898].DOI
- [103]J. Kummer, F. Kahlhoefer and K. Schmidt-Hoberg, Effective description of dark matter self-interactions in small dark matter haloes, Monthly Notices of the Royal Astronomical Society 474 (2018) 388.
- [104]O. Slone, F. Jiang, M. Lisanti and M. Kaplinghat, Orbital evolution of satellite galaxies in self-interacting dark matter models, Phys. Rev. D 107 (2023) 043014.arxiv.org/abs/2108.03243
- [105]M. Shirasaki, T. Okamoto and S. Ando, Modelling self-interacting dark matter substructures – I. calibration with N-body simulations of a Milky-Way-sized halo and its satellite, Monthly Notices of the Royal Astronomical Society 516 (2022) 4594.DOI
- [106]J. Richings, C. Frenk, A. Jenkins, A. Robertson and M. Schaller, A high-resolution cosmological simulation of a strong gravitational lens, Monthly Notices of the Royal Astronomical Society 501 (2021) 4657.DOI
- [107]S. Garrison-Kimmel, A. Wetzel, J.S. Bullock, P.F. Hopkins, M. Boylan-Kolchin, C.-A. Faucher-Giguère et al., Not so lumpy after all: modelling the depletion of dark matter subhaloes by Milky Way-like galaxies, Monthly Notices of the Royal Astronomical Society 471 (2017) 1709.
- [108]G. Despali and S. Vegetti, The impact of baryonic physics on the subhalo mass function and implications for gravitational lensing, Monthly Notices of the Royal Astronomical Society 469 (2017) 1997.DOI
- [109]A.R. Duffy, J. Schaye, S.T. Kay, C. Dalla Vecchia, R.A. Battye and C.M. Booth, Impact of baryon physics on dark matter structures: a detailed simulation study of halo density profiles, Monthly Notices of the Royal Astronomical Society 405 (2010) 2161.
- [110]Q. Zhu, F. Marinacci, M. Maji, Y. Li, V. Springel and L. Hernquist, Baryonic impact on the dark matter distribution in Milky Way-sized galaxies and their satellites, Monthly Notices of the Royal Astronomical Society 458 (2016) 1559.DOI
- [111]F.M. Heinze, G. Despali and R.S. Klessen, Not all subhaloes are created equal: modelling the diversity of subhalo density profiles in TNG50, Monthly Notices of the Royal Astronomical Society 527 (2024) 11996.DOI
- [112]O.Y. Gnedin, A.V. Kravtsov, A.A. Klypin and D. Nagai, Response of dark matter halos to condensation of baryons: Cosmological simulations and improved adiabatic contraction model, The Astrophysical Journal 616 (2004) 16.DOI
- [113]G. Despali, M. Sparre, S. Vegetti, M. Vogelsberger, J. Zavala and F. Marinacci, The interplay of self-interacting dark matter and baryons in shaping the halo evolution, Monthly Notices of the Royal Astronomical Society 484 (2019) 4563.DOI
- [114]O. Sameie, M. Boylan-Kolchin, R. Sanderson, D. Vargya, P.F. Hopkins, A. Wetzel et al., The central densities of Milky Way-mass galaxies in cold and self-interacting dark matter models, Monthly Notices of the Royal Astronomical Society 507 (2021) 720.DOI
- [115]Q. He, A. Robertson, J.W. Nightingale, A. Amvrosiadis, S. Cole, C.S. Frenk et al., Not so dark, not so dense: An alternative explanation for the lensing subhalo in SDSS J0946+1006, The Astrophysical Journal Letters 991 (2025) L53.arxiv.org/abs/2506.07978
- [116]S. Li, R. Li, K. Wang, Z. Jia, X. Cao, C.S. Frenk et al., The “little dark dot”: Evidence for self-interacting dark matter in the strong lens SDSS J0946+1006?, The Astrophysical Journal 994 (2025) 201.DOI
- [117]D.J. Ballard, W.J.R. Enzi, T.E. Collett, H.C. Turner and R.J. Smith, Gravitational imaging through a triple source plane lens: revisiting the ΛCDM-defying dark subhalo in SDSSJ0946+1006, Monthly Notices of the Royal Astronomical Society 528 (2024) 7564.DOI
- [118]W. Enzi, R. Murgia, O. Newton, S. Vegetti, C. Frenk, M. Viel et al., Joint constraints on thermal relic dark matter from strong gravitational lensing, the Lyα forest, and Milky Way satellites, Monthly Notices of the Royal Astronomical Society 506 (2021) 5848.DOI
- [119]T. Li, T.E. Collett, C.M. Krawczyk, G. Granata, W.J.R. Enzi, D.J. Ballard et al., A Salpeter IMF and an NFW halo: disentangling the dark and stellar mass of an elliptical galaxy through precise lens modelling of a double-source-plane system, Monthly Notices of the Royal Astronomical Society 548 (2026) stag683.DOI
- [120]T.E. Collett, The population of galaxy–galaxy strong lenses in forthcoming optical imaging surveys, The Astrophysical Journal 811 (2015) 20.DOI
- [121]Euclid Collaboration, Holloway, P., Verma, A., Walmsley, M., Marshall, P. J., More, A. et al., Euclid quick data release (q1) - XXX. the strong lensing discovery engine e – ensemble classification of strong gravitational lenses: lessons for data release 1, Astronomy & Astrophysics 711 (2026) A30.DOI
- [122]A.J. Shajib, G.P. Smith, S. Birrer, A. Verma, N. Arendse, T. Collett et al., Strong gravitational lenses from the Vera C. Rubin Observatory, Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 383 (2025) 20240117.DOI
- [123]E. Silver, R. Wang, X. Huang, A.S. Bolton, C.J. Storfer and S. Banka, ML-driven strong lens discoveries: Down to θ_E ∼ 0.03″ and M_halo < 10¹¹M☉, The Astrophysical Journal 994 (2025) 117 [2507.01943].
- [124]B. Wedig, T. Daylan, S. Birrer, F.-Y. Cyr-Racine, C. Dvorkin, D.P. Finkbeiner et al., The Roman view of strong gravitational lenses, The Astrophysical Journal 986 (2025) 42.arxiv.org/abs/2506.03390
- [125]J.W. Nightingale, Q. He, X. Cao, A. Amvrosiadis, A. Etherington, C.S. Frenk et al., Scanning for dark matter subhaloes in Hubble Space Telescope imaging of 54 strong lenses, Monthly Notices of the Royal Astronomical Society 527 (2024) 10480.DOI
- [126]D. Gilman, A.M. Nierenberg, T. Treu, C. Gannon, X. Du, H. Paugnat et al., JWST lensed quasar dark matter survey. IV. stringent warm dark matter constraints from the joint reconstruction of extended lensed arcs and quasar flux ratios, Phys. Rev. D 114 (2026) 043044.arxiv.org/abs/2511.07513
- [127]E.O. Nadler, S. Birrer, D. Gilman, R.H. Wechsler, X. Du, A. Benson et al., Dark matter constraints from a unified analysis of strong gravitational lenses and Milky Way satellite galaxies, The Astrophysical Journal 917 (2021) 7.DOI
- [128]E.O. Nadler, P. Mansfield, Y. Wang, X. Du, S. Adhikari, A. Banerjee et al., Symphony: Cosmological zoom-in simulation suites over four decades of host halo mass, The Astrophysical Journal 945 (2023) 159.
- [129]A.M. Nierenberg, R.E. Keeley, D. Sluse, D. Gilman, S. Birrer, T. Treu et al., JWST lensed quasar dark matter survey – I. description and first results, Monthly Notices of the Royal Astronomical Society 530 (2024) 2960.arxiv.org/abs/2309.10101
- [130]D. Liu, N.M. Förster Schreiber, K.C. Harrington et al., Detailed study of a rare hyperluminous rotating disk in an Einstein ring 10 billion years ago, Nature Astronomy 8 (2024) 1181.DOI
- [131]D.A. Berman, M.S. Yun, K.C. Harrington, P. Kamieneski, J. Lowenthal, B.L. Frye et al., PASSAGES: the large millimeter telescope and ALMA observations of extremely luminous high-redshift galaxies identified by the Planck, Monthly Notices of the Royal Astronomical Society 515 (2022) 3911.DOI
- [132]R. Gavazzi, T. Treu, L.V.E. Koopmans, A.S. Bolton, L.A. Moustakas, S. Burles et al., The Sloan Lens ACS Survey. VI. discovery and analysis of a double Einstein ring*, The Astrophysical Journal 677 (2008) 1046.DOI
- [133]R.J. Smith and T.E. Collett, A fully-spectroscopic triple-source-plane lens: the Jackpot completed, Mon. Not. Roy. Astron. Soc. 505 (2021) 2136 [2104.12790].DOI
- [134]T.E. Collett and R.J. Smith, A triple rollover: a third multiply imaged source at z ≃ 6 behind the Jackpot gravitational lens, Mon. Not. Roy. Astron. Soc. 497 (2020) 1654 [2004.00649].DOI
- [135]D.A. Riechers, Molecular gas in lensed z > 2 quasar host galaxies and the star formation law for galaxies with luminous active galactic nuclei, The Astrophysical Journal 730 (2011) 108.DOI
- [136]J.L. Tonry and C.S. Kochanek, Redshifts of the gravitational lenses MG 1131+0456 and B1938+666†, The Astronomical Journal 119 (2000) 1078.DOI
- [137]L.J. King, I.W.A. Browne, T.W.B. Muxlow, D. Narasimha, A.R. Patnaik, R.W. Porcas et al., Multifrequency radio observations of the gravitational lens system 1938 + 666, Monthly Notices of the Royal Astronomical Society 289 (1997) 450.DOI
- [138]L. King, N. Jackson, R. Blandford, M. Bremer, I. Browne, A. De Bruyn et al., A complete infrared Einstein ring in the gravitational lens system b1938 + 666, Monthly Notices of the Royal Astronomical Society 295 (1998) L41.
- [139]A.Ç. Şengül, C. Dvorkin, B. Ostdiek and A. Tsang, Substructure detection reanalysed: dark perturber shown to be a line-of-sight halo, Monthly Notices of the Royal Astronomical Society 515 (2022) 4391.DOI














