{"id":"717bcbef-b762-40d2-812e-b5fa64b6d93f","arxiv_id":"2506.15775","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"GIZMO now solves ideal general-relativistic magnetohydrodynamics in a mesh-free framework, validated by standard SRMHD and GRMHD test problems.","lead":"This paper adds magnetic fields to the general-relativistic version of the meshless hydrodynamics code GIZMO, producing the first mesh-free GRMHD solver. It passes a battery of special and general relativistic tests, which could let one code simulate from cosmological scales down to black hole horizons.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Divergence cleaning speed is a tuned fraction of the local magnetosonic speed, so cleaning vanishes where vms is small; no robustness study shows the chosen f/K values are safe for the advertised BH-accretion applications.","rationale":"The reader's weakest assumption identifies the most load-bearing risk in the paper. The central claim is not overturned by the lack of exact solutions in the multidimensional tests, nor by the acknowledged MFM magnetic-field growth in the TOV, because those are documented caveats. The load-bearing piece is the Dedner cleaning speed: it is the only active safeguard against div(B) corrupting the conservative-to-primitive inversion, it is not derived from first principles, and the paper itself shows a test-by-test choice of f and K. A parameter sweep on the magnetized TOV would settle whether the tuned values are safe or whether the advertised robustness for BH-accretion applications is unestablished. Since the concern is a robustness limitation rather than a demonstratable failure of the scheme, the conditional verdict remains appropriate.","tokens_in":26416,"tokens_out":21763,"duration_ms":224546,"concrete_test":"Rerun the Sec. 4.2.1 magnetized TOV in MFM mode with a grid in (f,K): (1,0.1), (2.5,0.1), (5,0.1), (2.5,0.75), at two resolutions, and record max |div(B)|, max(B), and central density rho_c at t=28 t_dyn. If max(B) or rho_c vary by more than about 5% across parameter choices, or if max |div(B)| grows instead of saturating, then the tuned cleaning speed is load-bearing and the conditional verdict should stand.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim (Sec. 1) is that GIZMO can evolve ideal GRMHD on a mesh-less discretization. That claim depends on keeping div(B) small enough for the conservative-to-primitive inversion (Sec. 3.2.3) to converge. The only active cleaning is the Dedner scheme of Sec. 3.3, whose advection and damping rates are set by c_h = vms(1+f)/(1+f vms^2). This is a user-tuned fraction of the local magnetosonic speed rather than the speed of light; the paper uses K=0.1 (0.75 for blast waves) and f=1 (f=2.5 for the TOV), and explicitly notes that ch=1 would break hierarchical time stepping. The risk is concrete: in regions where vms is small, c_h goes to zero, so the psi field neither advects nor damps divergence errors (the damping term alpha*sigma*c_h*D*psi in Eq. 30 vanishes). Appendix B shows the remaining Powell 8-wave source alone damps a strong monopole by only an order of magnitude in t=10. Thus in static or weakly magnetized regions, the scheme has no robust cleaning mechanism unless f is retuned. The paper does not include a parameter-robustness or convergence study that would show the chosen f/K are safe for the BH-accretion applications advertised in Sec. 1.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a general relativistic magnetohydrodynamics (GRMHD) extension to the mesh-less code GIZMO, building on the earlier GRHD implementation of Lupi (2023). The authors formulate ideal GRMHD in the 3+1 Valencia form, implement a single-state HLL Riemann solver with magnetosonic wave speeds, evolve particle volumes explicitly to handle the volume dependence of the magnetic conserved variables, and use a Noble et al. (2006) conservative-to-primitive inversion. To control div B, they implement Powell 8-wave source terms together with a hyperbolic Dedner cleaning field psi, with the cleaning speed set to an augmented local magnetosonic speed. The scheme is validated on a broad set of problems: monopole damping, Balsara shock tubes, loop advection, magnetic rotor, cylindrical and spherical blast waves in Minkowski spacetime, and magnetized TOV and Bondi accretion in Schwarzschild/Kerr-Schild backgrounds. The central claim is that this is the first GRMHD scheme in a mesh-free environment and that GIZMO can now evolve ideal GRMHD on a moving particle discretization.","tokens_in":26702,"tokens_out":5366,"duration_ms":58739,"significance":"If the implementation is robust, the significance is high: it would bring GRMHD into a widely used mesh-less astrophysical code, enabling quasi-Lagrangian simulations of black-hole accretion and cosmological-AGN setups without the grid-alignment and conservation problems of fixed meshes. The paper's strengths are its validation against independent exact solutions (the Giacomazzo & Rezzolla RMHD Riemann solver, analytic TOV and Bondi solutions), the breadth of the test suite, and the unusually honest reporting of numerical artifacts and limitations. The central claim is not circular because the tests compare against externally computed exact solutions rather than fitted outputs. The main open questions concern robustness: the divergence-cleaning parameters and speed are tuned per test, the MFM TOV run shows a 1.6x magnetic-field growth at about 13 dynamical times, and the magnetized Bondi test requires the energy-entropy switch. These issues do not invalidate the proof-of-concept, but they need to be addressed before the scheme can be considered ready for the advertised BH-accretion applications.","major_comments":[{"comment":"The divergence-cleaning robustness for the advertised BH-accretion applications is not demonstrated. The Dedner speed is ch = vms(1+f)/(1+f vms^2) and the damping term in Eq. (30) is proportional to alpha*sigma*ch*D*psi; as vms goes to zero, both advection and damping of the psi field vanish. The paper chooses K=0.1 (0.75 for blast waves) and f=1 (2.5 for the TOV) and justifies them by 'extensive tests', but it provides no parameter-robustness or resolution study. Appendix B shows that with Powell terms alone a strong monopole is damped by only an order of magnitude at t=10, so in low-vms regions the only active error-control mechanism is weak. Because the conservative-to-primitive inversion in Sec. 3.2.3 is stated to be sensitive to nonzero div B, please add quantitative tests in low-magnetosonic-speed regions, with varied f and K, and a resolution study for the Bondi or another BH-accretion-like setup that reports div B and inversion failures.","section":"3.3, Eq. (30), Appendix B"},{"comment":"The MFM TOV evolution is not fully satisfactory as a validation of long-term equilibrium maintenance. The paper reports that after about 13 dynamical times the magnetic field undergoes a topological rearrangement and its maximum intensity grows by a factor of about 1.6, with a delayed central-density increase of about 2 percent. This is attributed to small perturbations introduced by the MFM frame-velocity prescription. Since MFM is the mass-conserving mode that the paper emphasizes, and since BH-accretion simulations are inherently long-time evolutions, this artifact should be quantified (e.g. with a convergence study and a time-to-failure measure) and, if possible, mitigated or bounded. Without such a study, the statement that the scheme can evolve GRMHD in MFM mode for astrophysical applications is not yet supported.","section":"4.2.1, Figs. 12-13"},{"comment":"The magnetized Bondi test requires the optional energy-entropy switch; without it, the specific internal energy is overestimated by a factor of 1.5 at the horizon and the radial velocity is underestimated. The switch is only valid for adiabatic, shock-free flows, yet realistic BH accretion flows contain shocks and dissipation. The paper should state the resulting scope limitation explicitly at the point where the Bondi test is used to support the BH-accretion motivation, and should either demonstrate that the switch is not needed for shock-containing flows or clarify that the present scheme is validated only for smooth, isentropic accretion regions. This is load-bearing for the claim that GIZMO-GRMHD can self-consistently simulate from cosmological scales down to the BH event horizon.","section":"4.2.2 and Appendix A"}],"minor_comments":[{"comment":"The name 'Schwarzchild' is misspelled; it should be 'Schwarzschild'.","section":"Abstract and throughout"},{"comment":"The sentence 'higher K values result in inefficient cleaning while too low K values lead to system instabilities' is confusing in view of Fig. 1, where K=1 visibly damps the monopole faster. Please rephrase to distinguish damping speed from the fidelity of the magnetic-field evolution.","section":"4.1.1"},{"comment":"The y-axis label 'log(Bmax)' is inconsistent with the linear tick labels shown in the figure; please correct the label or the axis.","section":"Fig. 13"},{"comment":"The 'first mesh-less GRMHD scheme' claim would be easier to evaluate if the term 'mesh-less' were explicitly defined in contrast to moving-mesh schemes, with a brief statement of how the cited moving-mesh GRMHD implementations (e.g. Fragile et al. 2019) differ from the present approach.","section":"1, Introduction"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is honest about its limitations and the test suite is strong, but the robustness issues around divergence cleaning and the MFM TOV artifact are central to the paper's astrophysical claims. I would support publication after the authors add a parameter-robustness/convergence study for the Dedner scheme and a more quantitative treatment of the MFM TOV magnetic-field growth. The novelty claim appears fair on its face, but the term 'mesh-less' should be defined carefully to avoid ambiguity with existing moving-mesh GRMHD codes."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear X,\n\nThe one thing you should know: this is the first GRMHD implementation in a meshless code, and the claim holds up. The authors implement the standard Valencia ideal MHD equations with an HLL solver, Powell source terms, and Dedner cleaning inside GIZMO's MFM and MFV frameworks, and the validation suite is broad enough that I trust the central claim.\n\nWhat is actually new is not the physics—those equations are standard—but the integration. The explicit volume update (Eq. 21) fixes the conservative-to-primitive inversion when B^2 makes the conserved variables nonlinear in V_i, which is a real technical step, and the energy-entropy switch for strongly magnetized, supersonic regions is useful. The test suite is the standard one and they run both methods: Balsara 1 and 4, loop advection, magnetic rotor, 2D and 3D blast waves, magnetized TOV, and Bondi accretion in Kerr-Schild coordinates. They report artifacts instead of hiding them: the MFM TOV shows magnetic field growth of about 1.6x around 13 dynamical times; the Bondi test needs the entropy switch; blast waves show oscillations. That honesty is credit.\n\nSoft spots, in order. The cleaning parameters K and f are tuned per test, and there is no parameter-robustness or convergence study showing the choices are safe for the advertised BH-accretion applications. The stress-test concern is real: setting ch to a fraction of the local magnetosonic speed means cleaning slows down in static or weakly magnetized regions, and the TOV test itself needed f=2.5 because signal speeds were slow. This is not fatal—they note f=1 on the TOV gives a stable run with slightly higher divB—but a referee should ask for scaling tests. Second, there is no public code yet. For a methods paper that is a genuine reproducibility gap. Third, the MFM TOV magnetic field rearrangement is described but not fully explained; calling it topological rearrangement is descriptive.\n\nWho this is for: anyone wanting to span cosmological scales down to the horizon in one Lagrangian simulation, particularly AGN disk, jet, and TDE modelers. It is a methods contribution, not a new physics result, but it opens a capability.\n\nI would send it to a serious referee, and I would expect acceptance after revision. The referee should push on the robustness study and code release, not on whether the method exists.\n\nBest","headline":"First meshless GRMHD implementation, and the first-mover claim holds up; the tuned cleaning speed and missing code release are the main soft spots.","tokens_in":27231,"tokens_out":3935,"would_cite":true,"duration_ms":38534,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper presents the first mesh-less solver for general-relativistic magnetohydrodynamics.","keywords":["GRMHD","mesh-less methods","divergence cleaning","HLL solver","GIZMO","magnetized accretion","black hole accretion disks","numerical astrophysics"],"falsifier":"Run the magnetic-rotor or Balsara4 test with the user factor lowered to $f=0.1$ while keeping $K=0.1$; if the maximum $|\\partial_i B^i|/|B|$ grows monotonically or primitive-variable recovery fails before the nominal end time, then the tuned cleaning speed is load-bearing in exactly the way the paper assumes. Alternatively, compare the magnetized TOV star at four times the resolution: the MFM mode shows a magnetic-field amplification by about a factor 1.6 after roughly 13 dynamical times, so if that amplification persists or grows with resolution rather than decaying, the scheme's long-term equilibrium behavior is not yet convergent.","tokens_in":2136,"feed_emoji":"🧲","tokens_out":8000,"duration_ms":135818,"temperature":0.7,"pith_summary":"The paper claims that the equations of general-relativistic magnetohydrodynamics (GRMHD) can be solved on the mesh-less, particle-like discretization used by the GIZMO code, and that this is the first mesh-free numerical scheme to do so. The motivation is physical: magnetized plasma around black holes, especially in active galactic nuclei, has so far required grid codes, while Lagrangian schemes preserve mass, angular momentum, and adaptive resolution more naturally. The authors extend GIZMO's existing general-relativistic hydrodynamics with an HLL Riemann solver that includes magnetic fields, an explicit volume evolution, and a divergence-cleaning scheme that combines Powell source terms with a general-relativistic version of Dedner hyperbolic cleaning. They demonstrate the scheme on special-relativistic shock tubes, loop advection, a magnetic rotor, and cylindrical and spherical blast waves, and on general-relativistic tests of a magnetized neutron star and Bondi accretion onto a black hole. If the claim holds, cosmological simulations that reach down to black-hole horizons can carry magnetic fields through the full dynamical range without switching to a grid.","feed_headline":"First mesh-less GRMHD scheme passes relativistic magnetized tests","feed_subtitle":"GIZMO now evolves magnetized plasma near black holes on moving particles without switching to a grid.","key_machinery":"The load-bearing mechanism is the coupling of GIZMO's mesh-less finite-volume discretization with a general-relativistic divergence-cleaning scheme. In the mesh-less method, each fluid element carries a volume defined by a kernel weight, and fluxes are computed with a one-dimensional HLL Riemann solver at element faces; the two modes, MFM and MFV, differ only in the assumed face velocity. The new GRMHD components are an explicit volume update over the timestep, needed because the magnetic terms in the conserved momentum and energy depend on volume-squared; a reconstruction of $W v^i$ rather than $v^i$ at faces to avoid superluminal velocities; and the Powell plus Dedner cleaning, in which the scalar field $\\hat{\\psi}$ obeys a mass-weighted advection equation and enters Rusanov fluxes for the field components. The cleaning parameters are the damping constant $K$ and the speed-increase factor $f$, with fiducial values $K=0.1$ and $f=1$, except $f=2.5$ for the neutron star and $K=0.75$ for the blast waves.","core_discovery":"The central discovery, stated on the paper's own terms, is that a mesh-less finite-volume scheme can evolve the ideal GRMHD equations in curved spacetime and pass the standard battery of relativistic MHD validation tests. Working in the 3+1 Valencia formulation, the authors write the equations in conservative form and solve them with a single-state HLL (Harten-Lax-van Leer) Riemann solver whose wave speeds come from the relativistic magnetosonic dispersion relation. To keep the magnetic-field divergence under control they combine Powell's eight-wave source terms with a hyperbolic divergence-cleaning scalar field, deriving both from a modification of Maxwell's equations consistent with the 3+1 decomposition; the scalar field is advected and damped at a local magnetosonic speed augmented by a user factor rather than at the speed of light, in order to preserve hierarchical time-stepping. The scheme reproduces exact or reference solutions for the special-relativistic shock tubes, loop advection, magnetic rotor, and blast waves, and it evolves both a magnetized Tolman-Oppenheimer-Volkoff star and magnetized Bondi accretion onto a non-spinning black hole in Kerr-Schild coordinates. A new energy-entropy switch cures internal-energy overestimation in strongly magnetized, supersonic, isentropic regions such as the Bondi flow inside the horizon.","pith_inferences":["If the first-implementation claim holds, the natural next comparison is against constrained-transport or vector-potential moving-mesh GRMHD codes at equal resolution; those alternative divergence controls may set the bar for how much the Powell plus Dedner cleaning costs in accuracy on long-run disk simulations.","The choice to cap the cleaning speed below the speed of light means residual divergence will never be exactly zero; one testable extension is to make the factor $f$ adaptive, high in strongly magnetized regions and low elsewhere, to see whether cleaning quality and step-size savings can both improve.","The MFM magnetic-topology drift seen in the TOV test suggests that mass-conserving schemes may need a small amount of numerical resistivity or a better face-velocity estimate before they can be trusted for decades-long accretion-disk evolution; the paper does not claim to solve that."],"forward_implications":["Both mesh-less modes, MFM and MFV, pass the SRMHD and GRMHD tests; MFV preserves magnetic topology better in the TOV equilibrium, while MFM conserves particle masses exactly and reaches higher density peaks in the rotor test.","Because the implementation accepts generic equations of state and user-supplied metrics, with flat and Kerr metrics already available in Boyer-Lindquist and Kerr-Schild coordinates, the same machinery can be applied to spinning black holes and, in future extensions, to dynamically evolving spacetimes.","The magnetized Bondi test in Kerr-Schild coordinates confirms that the scheme can carry a magnetized accretion flow smoothly across the event horizon, which is what a full AGN-disk simulation inside the horizon will require.","The paper's own comparisons show that particle-motion artifacts, such as oscillations at strong shocks and slow magnetic noise in MFM mode, can be reduced by a close-packed lattice initialization, a smoother kernel, or more diffusive slope limiters, so the practical accuracy of the method is tunable."],"supporting_citations":[{"why":"Supplies the prior GRHD implementation in GIZMO that this scheme extends with magnetic fields.","marker":"Lupi 2023"},{"why":"Defines the mesh-less finite-volume and finite-mass discretization, kernel volumes, and Riemann-problem face estimation used throughout.","marker":"Hopkins 2015"},{"why":"Provides the Newtonian MHD version of GIZMO whose face-based divergence estimate and cleaning choices are adapted to general relativity here.","marker":"Hopkins & Raives 2016"},{"why":"Template for the Valencia-form GRMHD equations and for the general-relativistic divergence-cleaning formulation.","marker":"Mösta et al. 2014"},{"why":"Origin of the hyperbolic divergence-cleaning scalar field and its advection-damping equations.","marker":"Dedner et al. 2002"},{"why":"Supplies the eight-wave source-term prescription whose general-relativistic analogue stabilizes the induction equation.","marker":"Powell et al. 1999"},{"why":"Provides the two-dimensional Newton-Raphson conservative-to-primitive inversion used to recover primitive variables.","marker":"Noble et al. 2006"},{"why":"Exact relativistic MHD Riemann solver used as the reference solution for the shock-tube tests.","marker":"Giacomazzo & Rezzolla 2006"},{"why":"Defines the Balsara1 and Balsara4 shock-tube problems used as special-relativistic benchmarks.","marker":"Balsara 2001"}],"fun_headline_variants":["GIZMO mesh-free scheme passes first black hole MHD tests","Mesh-less GRMHD in GIZMO passes black hole accretion tests","First mesh-free GRMHD scheme for black hole jets passes tests","Black hole magnetized disks now simulated without a grid in GIZMO"],"cache_read_input_tokens":29312,"weakest_assumption_plain":"The scheme's safety rests on the assumption that a cleaning speed set to a local magnetosonic speed augmented by a user factor, rather than the speed of light, removes magnetic-divergence errors fast enough to protect the numerical step that recovers physical fluid quantities from the evolved conserved ones.","fun_headline_variants_meta":{"raw":{"variants":["GIZMO mesh-free scheme passes first black hole MHD tests","Mesh-less GRMHD in GIZMO passes black hole accretion tests","First mesh-free GRMHD scheme for black hole jets passes tests","Black hole magnetized disks now simulated without a grid in GIZMO"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.0006,"raw_usage":{"total_tokens":2828,"prompt_tokens":995,"completion_tokens":1833,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":611,"completion_tokens_details":{"reasoning_tokens":1756}},"tokens_in":611,"tokens_out":1833,"duration_ms":13290,"temperature":1.0,"reasoning_tokens":1756,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T19:30:43.632598+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the magnetic-rotor or Balsara4 test with the user factor lowered to $f=0.1$ while keeping $K=0.1$; if the maximum $|\\partial_i B^i|/|B|$ grows monotonically or primitive-variable recovery fails before the nominal end time, then the tuned cleaning speed is load-bearing in exactly the way the paper assumes. Alternatively, compare the magnetized TOV star at four times the resolution: the MFM mode shows a magnetic-field amplification by about a factor 1.6 after roughly 13 dynamical times, so if that amplification persists or grows with resolution rather than decaying, the scheme's long-term equilibrium behavior is not yet convergent.","supporting_citations":[{"cited_title":"C., Gammie , C","cited_arxiv_id":null,"evidence_quote":"Provides the two-dimensional Newton-Raphson conservative-to-primitive inversion used to recover primitive variables."},{"cited_title":"2023, Monthly Notices of the RAS, 519, 1115","cited_arxiv_id":null,"evidence_quote":"Supplies the prior GRHD implementation in GIZMO that this scheme extends with magnetic fields."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the mesh-less finite-volume and finite-mass discretization, kernel volumes, and Riemann-problem face estimation used throughout."},{"cited_title":"G., Roe , P","cited_arxiv_id":null,"evidence_quote":"Supplies the eight-wave source-term prescription whose general-relativistic analogue stabilizes the induction equation."},{"cited_title":"& Rezzolla , L","cited_arxiv_id":null,"evidence_quote":"Exact relativistic MHD Riemann solver used as the reference solution for the shock-tube tests."}],"review_version":2}