{"id":"43551d2e-b4e0-4333-99a8-e426291d9afb","arxiv_id":"1908.05812","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A donor-based scheme for dark matter annihilation feedback in cosmological simulations is implemented in Gizmo and validated against a receiver-based scheme, with a new Sedov-Taylor time-step limiter.","lead":"This paper implements and tests a new way to deposit energy from dark matter annihilation into gas in cosmological simulations: computing the energy at each dark matter particle and spreading it to nearby gas. The method matches an existing approach in realistic galaxy and cosmological tests, while offering more flexible injection choices and a new time-step limiter.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Untested DM timestep limiter (§3.2) leaves the donor-based validation dependent on an unspecified c_Δ choice in the common Δt_DM >> Δt_gas regime.","rationale":"The reader's weakest assumption (local instantaneous injection) is legitimate but is explicitly scoped by the paper: the default weights are local, non-local injection is described as future work, and the validation runs use local deposition. This limits the method's current applicability but does not threaten the numerical scheme's internal consistency for the tested local scenario. A more immediate gap is the untested c_Δ limiter. The paper itself admits in §3.2 that for Δt_DM ≫ Δt_gas (the common case, since gas has extra hydrodynamical timestep restrictions) gas particles can receive energy after leaving the donor's neighbourhood, and offers a limiter to prevent this. Yet no result in §5 indicates whether the limiter was enabled, and no sensitivity test is reported. This matters because the first-order scheme of Figure 2 holds the Eq. (4) energy rate fixed over the DM step, so without the limiter energy deposition can be non-local in a purely numerical sense. The stated 'good agreement' with the receiver-based method could thus be an artifact of an unstated timestep choice rather than a robust property of the method. The recommended rerun with c_Δ = 2, 4, 8 and off is a concrete, inexpensive check that would settle this. The reader's conditional verdict already calls for convergence and robustness tests, so I do not change the verdict; I would make the c_Δ test an explicit condition of acceptance.","tokens_in":23615,"tokens_out":14036,"duration_ms":140928,"concrete_test":"Rerun the cosmological simulation of §5.3 with the c_Δ limiter disabled and enabled for c_Δ = 2, 4, and 8, keeping all other settings fixed. Compare the halo mass function, halo velocity function, and the largest-cluster gas density and temperature profiles across these runs. If the HMF shifts by more than the Poisson error, or the central gas density changes by more than the 50% level already seen between donor and receiver methods, the reported validation is time-step dependent and a c_Δ choice must be specified. The paper should also state whether the originally reported runs used the limiter.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The donor-based method's handling of individual timesteps is not fully validated. In §3.2, the paper states that when Δt_DM ≫ Δt_gas, a gas particle may continue receiving energy from a DM particle's fixed neighbour list after moving far away, and an optional limiter Δt_DM ≤ c_Δ Δt_gas is introduced. However, the paper never states whether this limiter was enabled in the isolated-halo or cosmological runs (§5.2, §5.3), nor does it report sensitivity to c_Δ. Since gas timesteps are generally the smaller ones (the paper itself notes Δt_gas > Δt_DM is unlikely), the stale-neighbour regime is the typical case. The first-order scheme of Figure 2 holds the Eq. (4) energy rate constant for the entire DM step, so if the limiter is off, energy can be deposited at outdated locations. The reported 'good agreement' with the receiver-based method could therefore depend on an unspecified time-stepping choice, weakening the central claim that the method is self-consistent with adaptive individual timesteps.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a donor-based implementation of dark-matter annihilation feedback (DMAF) in the Gizmo code. Annihilation power is computed at each DM particle via Eq. (4) and distributed to neighbouring gas particles using either mass-weighted (Eq. 5) or solid-angle-weighted (Eq. 6) weights. The authors describe the interaction of this scheme with Gizmo's individual time stepping, introduce a Sedov--Taylor-based time step limiter, compare the donor approach with the receiver-based method of Iwanus et al. (2017), and report tests on a density jump, an isolated NFW halo, and a cosmological box. They claim that in realistic tests the donor and receiver methods agree well, while the donor method offers greater flexibility in energy-injection geometry and avoids suppression of energy in gas-poor regions.","tokens_in":23803,"tokens_out":7949,"duration_ms":76591,"significance":"If validated, the scheme is a useful addition to simulation toolkits: it enables channel-dependent energy deposition in a public code, is modular, and is built on straightforward energy-conserving arithmetic. The paper's comparisons to an independent receiver-based implementation and to an analytic homogeneous-universe solution (Appendix B) are appropriate checks and are genuine strengths, as is the explicit discussion of computational cost. The main limitations are the absence of any resolution/convergence study, the lack of statistical error bars on the halo statistics, and incomplete specification of the time-stepping settings used in the validation runs. These gaps, rather than the derivation of Eq. (4), are what prevent the central agreement claim from being fully established.","major_comments":[{"comment":"The manuscript does not state whether the optional Δt_DM ≤ c_Δ Δt_gas limiter of §3.2 was enabled in the isolated-halo or cosmological runs, nor which of the cosmological-expansion variants (I or II) of Appendix B was used. The paper itself notes that Δt_gas > Δt_DM is unlikely, so the stale-neighbour regime is the typical case rather than a corner case. Since Eq. (4) is held constant over the whole DM step in the first-order scheme of Figure 2, the reported agreement with the receiver-based method could depend on this unspecified time-stepping choice. The authors should state the c_Δ value used (or justify why the limiter was disabled) and the expansion variant, and ideally show sensitivity to c_Δ in at least one test.","section":"§3.2, Appendix B; runs in §5.2–5.3"},{"comment":"The central claim that donor and receiver methods 'agree well' is based on single-resolution runs without statistical uncertainties or a convergence study. The paper reports relative differences up to ~20 per cent in gas density within 10 kpc (Section 5.2) and ~50 per cent in the central gas density of the largest cluster (Figure 10), yet no Poisson or bootstrap errors are given on the HMF/HVF points in Figure 9 and no resolution sequence is presented. A reader cannot tell whether the residual method differences are physical or numerical artifacts; a resolution test (for example, rerunning the isolated halo with higher-resolution initial conditions) is needed to support the agreement claim.","section":"§5.2, §5.3, Figs. 7, 9, 10"},{"comment":"The paper motivates flexible weights by the desire to model channel-dependent energy deposition, but the validation exercises only test local injection with weights (5) and (6). The justification for local injection in §4.1 (relativistic products travel ~100 kpc within a time step) is not turned into a test for channels with different mean free paths or for f ≠ 1; B ≠ 1. At minimum, the authors should state explicitly that the numerical agreement is established only for the local-injection limit, and they should outline how the method would be tested against a known non-local deposition solution before being applied to photon or neutrino channels.","section":"§4.1 and §5"}],"minor_comments":[{"comment":"In the paragraph following Eq. (12), 'enumerator' should be 'numerator'.","section":"§4.1"},{"comment":"The phrase 'During the phase of linear evolution until z ∼ 100' is confusing because the simulation begins at z = 100; specify 'before the initial redshift' or 'for epochs prior to z = 100'.","section":"§5.3"},{"comment":"The solid-angle weight definition is not self-contained; the properties of A_ki and the convex-hull construction are only described in Hopkins et al. (2018a). A one-sentence summary of the relevant properties would improve readability.","section":"Eq. (6)"},{"comment":"The simulation is two-dimensional, so the black region marking the 3σ contour is a circle rather than a sphere; consider changing 'spherical' to 'circular'.","section":"Fig. 5 caption"},{"comment":"The layout of Table 1 is difficult to parse because the row labels and column entries are not clearly separated; reformatting would help the reader compare the neighbour-search requirements of the two methods.","section":"Table 1"}],"recommendation":"major_revision","confidential_remarks":"The paper is a suitable methods contribution for MNRAS. The referee's main concern is the under-specified validation, not the conceptual soundness of the donor-based construction. The authors should be encouraged to state exact run configurations (c_Δ, expansion variant, code version, and whether the code will be released) and to add at least a minimal resolution or convergence test before the agreement claim can be fully supported."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague—\n\nThe useful thing here is the donor-based DMAF implementation: compute annihilation power at each DM particle and deposit into gas with flexible weights (Eq. 4). That is a genuine step beyond the receiver-based method of Iwanus et al. (2017), and the paper is detailed enough that someone could reimplement it from the text. The main validation claim—that donor and receiver agree in an isolated halo and in a cosmological run—is believable. The toy density-jump test is the right kind of stress test: it shows where the methods differ, and it is honest about the receiver method's suppression of energy in steep-density regions.\n\nWhat is actually new: the donor-based formulation, the mass-weighted vs. solid-angle-weighted options, the time-binned energy storage for individual adaptive timesteps, and the Sedov-Taylor timestep limiter. The homogeneous-universe analytic comparison in Appendix B is a nice check. No code is released, but the algorithm is described cleanly.\n\nSoft spots, in rough order of importance. First, the timestep treatment is under-specified. The paper notes that Δtgas > ΔtDM is unlikely, so the usual case is ΔtDM >> Δtgas, exactly where a DM particle's neighbour list can go stale. An optional limiter ΔtDM ≤ cΔΔtgas is introduced in §3.2, and a Sedov–Taylor limiter in §3.3, but I could not find where the paper states whether these were enabled in the isolated-halo and cosmological runs, or what cΔ was used. The statement in §3.3 that the Sedov–Taylor limiter was not the dominant criterion in \"a halo with a gaseous fraction\" suggests it was on, but that is not made explicit for the production runs. This matters: the good agreement with the receiver-based method could be contingent on an unstated time-stepping choice. I would not call this fatal, but it should be fixed.\n\nSecond, there is no convergence or resolution study, one cosmological realization, and no error bars on the HMF/HVF differences. The halo-number decrement of ~20% is presented as the headline effect, but with one box it is hard to know the variance. Third, the wall-time numbers contain an anomaly: the receiver-based run was faster than the fiducial no-DMAF run (6387 s vs 7965 s). That is odd and unexplained; it may be run-to-run noise, but it deserves a comment. Fourth, the locality assumption for energy injection is acknowledged but not tested for long-mean-free-path channels; that is a minor point since the weight framework is explicitly designed to be extended.\n\nWho this is for: simulation groups who want to put DMAF or DM decay into Gizmo, and anyone comparing implementations of feedback. It is a methods paper, not a discovery paper. The central argument holds up, with the timestep caveat above. I would send it to a serious referee and ask for the missing timestep details, a convergence test, and code release; it is not a desk reject.","headline":"Useful, clearly-written methods paper for donor-based DMAF in Gizmo; the validation is believable but the timestep-limiter settings and missing convergence tests leave some details unspecified.","tokens_in":24334,"tokens_out":3544,"would_cite":false,"duration_ms":34073,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A donor-based scheme for dark matter annihilation feedback: each dark matter particle deposits annihilation power into nearby gas via tunable weights, matching the receiver-based method in realistic tests while avoiding gas-poor…","keywords":["dark matter annihilation","cosmological simulations","Gizmo","feedback","N-body simulations","energy injection","Sedov-Taylor blast wave","meshless hydrodynamics"],"falsifier":"A side-by-side cosmological simulation using ray-based line-of-sight energy deposition for a long-mean-free-path channel (for example, photons) versus the local weights of equations (5) and (6): if the central gas density of the largest halo changes by more than the roughly 50 per cent method-to-method scatter already reported, the locality assumption is falsified for that channel.","tokens_in":23434,"feed_emoji":"☄️","tokens_out":11389,"duration_ms":90892,"temperature":0.7,"pith_summary":"The paper sets out to give cosmological N-body simulators a self-consistent way to include dark matter annihilation feedback (DMAF) by computing the annihilation power at each dark matter particle and depositing it into surrounding gas with weights that can be tailored to the annihilation channel. Implemented in the Gizmo code, this donor-based scheme is shown to agree well with the receiver-based method of Iwanus et al. (2017) for an isolated halo and a cosmological simulation, while avoiding that method's suppression of energy deposition in gas-poor or steep-density-gradient regions. The paper also introduces a time-step limiter derived from Sedov-Taylor blast-wave propagation to prevent large energy errors when strong feedback turns on. A sympathetic reader would care because the geometry of energy injection can change the predicted gas distribution in halo centres, and this scheme makes that geometry explicit and flexible.","feed_headline":"New scheme puts dark matter annihilation power at the donor particle","feed_subtitle":"A flexible Gizmo implementation matches receiver-based runs in tests while adding control over injection geometry.","key_machinery":"The central object is the donor-based energy rate of equation (4), a per-DM-particle annihilation power with normalised weights, together with the two weight choices of equation (5) (mass-weighted kernel) and equation (6) (solid-angle-weighted face areas of a convex hull around the donor). These weights set the geometry of energy injection and give the scheme its flexibility. Around this sit a time-binned energy-rate storage keyed to DM time steps, a bidirectional neighbour search, and the Sedov-Taylor time-step limiter of equation (9), which confines any strong shock to the gas smoothing length within one step.","core_discovery":"On its own terms, the central claim is that equation (4), $$dE_{i\\to j}/dt = \\frac{\\langle\\$\\sigma$ v\\rangle}{m_\\chi}\\,\\rho_{\\chi,i} M_i $c^{2}$\\, \\frac{w_j}{\\sum_k w_k},$$ gives a self-consistent donor-based DMAF rate: each dark matter particle $i$ deposits the annihilation power it generates into its gas neighbours $j$ with weights $w_k$, normalised so that the total injected energy equals the energy produced. With mass-weighted weights $w_k = M_k W(r_{ki}, h_i)$ the injection follows the local gas density; with solid-angle weights it is statistically isotropic. The paper demonstrates, in a contact-discontinuity toy model, that the donor-based total energy rate is independent of gas properties, while the receiver-based method deposits only 17.3 per cent of the donor-based total when steep DM density gradients are present. In the isolated-halo run the three DMAF variants agree to within 2 per cent for DM density, 5-20 per cent for gas density, and 13 per cent for temperature, and in the cosmological run they produce similar halo mass and velocity functions; the paper concludes from this that realistic simulation results are fairly robust to the choice of DMAF implementation, while the donor-based method adds flexibility in the injection geometry.","pith_inferences":["The locality assumption is the point I would press: for photon or neutrino channels the paper's own 100 kpc transport argument does not guarantee that nearest-neighbour deposition is accurate, and a ray-based weight scheme is the natural next test.","Because the donor-based total energy rate is independent of gas properties, this scheme is better suited than the receiver-based one for studying DMAF in gas-poor minihalos and at high redshift, where the receiver method would suppress the signal; the paper's toy example already points this way.","A resolution test comparing runs with gas spacings below and above the annihilation-product transport horizon would quantify when local injection fails; the paper does not report such a convergence test.","The Sedov-Taylor limiter could be repurposed for any feedback process whose power is known at the start of a step, not just DMAF; that generalisation is implicit in the derivation."],"forward_implications":["Simulators can change the energy-injection geometry for a given annihilation channel by swapping the weights in equation (4), without altering the rest of the code.","The donor-based total DMAF energy rate is independent of the gas distribution, so annihilation energy is not artificially suppressed in gas-poor haloes or regions with steep DM density gradients.","In the isolated-halo test, the three DMAF variants agree to within 2 per cent for DM density, 5-20 per cent for gas density, and 13 per cent for temperature; in the cosmological run they yield similar halo mass and velocity functions, with DMAF cutting the z=0 halo count by about 20 per cent for a 1 MeV c^-2 candidate.","The same machinery extends directly to velocity-dependent annihilation cross-sections, dark matter decay, and non-local (line-of-sight) energy injection.","The Sedov-Taylor time-step limiter prevents large early energy errors in scenarios with high annihilation rates, at little cost in the realistic tests where it is rarely the dominant criterion."],"supporting_citations":[{"why":"Supplies the receiver-based DMAF method and the analytical homogeneous-universe solution used for comparison and validation.","marker":"Iwanus et al. (2017)"},{"why":"The Gizmo code with adaptive individual time stepping and meshless hydrodynamics that hosts the implementation.","marker":"Hopkins (2015)"},{"why":"Source of the bidirectional neighbour search and solid-angle weight construction for statistically isotropic energy injection.","marker":"Hopkins et al. (2018a)"},{"why":"Provides the recommended c_Delta = 4 factor for limiting neighbour time steps, adapted here for DM donors.","marker":"Saitoh & Makino (2009)"},{"why":"Supplies the dimensionless beta coefficient in the Sedov-Taylor shock scaling used for the new time-step limiter.","marker":"Dokuchaev (2002)"},{"why":"Gives the electron-positron energy-loss mean free path used to justify instantaneous local injection.","marker":"Delahaye et al. (2010)"}],"fun_headline_variants":["Donor-based dark matter annihilation feedback matches standard runs","Annihilation power delivered at source: new feedback scheme","Flexible DM annihilation feedback for cosmological N-body runs","Self-consistent donor method for annihilation energy injection","Putting annihilation energy at the DM particle: new scheme"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The method assumes that the annihilation power released by a dark matter particle is deposited instantly and locally into the gas particles within its smoothing radius, with no transport along the line of sight or over distance.","fun_headline_variants_meta":{"raw":{"variants":["Donor-based dark matter annihilation feedback matches standard runs","Annihilation power delivered at source: new feedback scheme","Flexible DM annihilation feedback for cosmological N-body runs","Self-consistent donor method for annihilation energy injection","Putting annihilation energy at the DM particle: new scheme"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000213,"raw_usage":{"total_tokens":1433,"prompt_tokens":965,"completion_tokens":468,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":581,"completion_tokens_details":{"reasoning_tokens":392}},"tokens_in":581,"tokens_out":468,"duration_ms":4939,"temperature":1.0,"reasoning_tokens":392,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:04:10.470441+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A side-by-side cosmological simulation using ray-based line-of-sight energy deposition for a long-mean-free-path channel (for example, photons) versus the local weights of equations (5) and (6): if the central gas density of the largest halo changes by more than the roughly 50 per cent method-to-method scatter already reported, the locality assumption is falsified for that channel.","supporting_citations":[{"cited_title":"J., Lewis G","cited_arxiv_id":null,"evidence_quote":"Supplies the receiver-based DMAF method and the analytical homogeneous-universe solution used for comparison and validation."},{"cited_title":"I., 2002, @doi [A & A] 10.1051/0004-6361:20021305 , 395, 1023","cited_arxiv_id":null,"evidence_quote":"Supplies the dimensionless beta coefficient in the Sedov-Taylor shock scaling used for the new time-step limiter."}],"review_version":1}