{"id":"8c5d8177-18cf-40b0-b0b2-2cc3a378dccc","arxiv_id":"1908.01682","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"SPH-IDIC, a particle-in-cell style implicit drag scheme, preserves the coupled dust-gas asymptotic solution on coarse smoothing lengths where the Monaghan-Kocharyan and ISPH schemes become dissipative.","lead":"This paper compares three ways of computing the drag force between gas and dust in two-fluid Smoothed Particle Hydrodynamics. A cell-based implicit scheme, SPH-IDIC, keeps solutions accurate at coarse resolution and large time steps, while two fully Lagrangian schemes dissipate.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"IDIC's asymptotic-preserving claim lacks support because the Euler-cell mean drag (Eqs. 19-21) has no error control and the Euler cell size is unreported; the two tests cannot separate AP behavior from grid smoothing.","rationale":"The reader identified the same load-bearing concern: replacing local gas velocity by the Euler-cell mean v* has no controlled error estimate, so the asymptotic-preserving behavior shown in two test problems may not persist in general. I agree with that assessment and add that the Euler cell size itself is not reported, making it impossible to determine whether h or the cell size is the resolution parameter responsible for the observed accuracy. The paper does provide legitimate evidence that IDIC beats MK and ISPH on Dustywave and Dustyshock, and the reported errors are consistent with the figures, so I am not rejecting the numerical comparison. However, the central claim is a property of a discretization family, not of two runs; without a formal AP/consistency argument or a cell-size convergence study, the claim is overstated. The parameter inconsistencies in the tables and figure captions further weaken the quantitative support. Since the reader's CONDITIONAL verdict already captures this situation, my stress-test does not move the verdict; the proposed convergence test would settle whether the concern actually lands.","tokens_in":6979,"tokens_out":15898,"duration_ms":156294,"concrete_test":"Compute the Dustywave error with fixed smoothing length h = 0.01 and the same physical parameters as Eq. (22), varying only the Euler cell size Delta x = h, h/4, h/16. For each Delta x, run t_stop = 0.002, 0.0002, and 0.00002 (e.g. by rescaling K) and report the L2 error of the dust velocity at t = 0.5. If the error at fixed Delta x grows as t_stop decreases, or if the error does not converge as Delta x -> 0 for the smallest t_stop, the claimed AP property is not established. Also report the Delta x used for the published h = 0.01 and h = 0.02 runs.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The Summary's central claim is that IDIC is asymptotic preserving and lets h and tau be chosen independently of drag intensity. The mechanism is the implicit cell-averaged drag in Eqs. (19)-(21): the gas velocity seen by dust and the dust velocity seen by gas are replaced by cell means v* and u*. In the stiff limit t_stop -> 0 these equations drive every gas and dust particle in a cell to a single common velocity, namely the cell-mass-weighted mean of the explicit updates, so the discretization becomes a cell-averaged mixture solver. The difference between v* and the local gas velocity inside a cell is O(Delta x^2) for smooth fields and O(Delta x) across discontinuities, and no bound is given in terms of t_stop, h, or the cell size. The Euler cell size is never specified in Section 2.3 or the test descriptions, so the numerical results cannot show that h, rather than the Euler cell size, is the resolution controlling the asymptotic error. In addition, the test parameters are not fully consistent (Table 1 and Figs. 1-2 report different h and tau values, and Fig. 2's t_stop = 0.00025 is not obviously compatible with K = 500, rho_d/rho_g = 1), so the empirical evidence for the strong-drag regime is not quantitatively reliable. The two test problems are good benchmarks and the reported IDIC errors are small, but the AP claim requires either a formal AP/consistency proof or a resolved cell-size convergence study; neither is provided.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper addresses two-fluid Smoothed Particle Hydrodynamics (TFSPH) simulations of monodisperse gas-solid mixtures with intense interphase interaction, i.e., drag relaxation times t_stop much smaller than the dynamical time. It compares three drag-discretization strategies: the explicit Monaghan–Kocharyan (MK) scheme, a semi-implicit interpolated SPH (ISPH) scheme, and the authors' previously introduced SPH-IDIC ('drag in cell') scheme, in which the drag force is evaluated using piecewise-constant Euler-cell-averaged velocities. The manuscript presents Dustywave (isothermal sound wave) and Dustyshock (Sod shock tube) test problems with reference solutions, reports L2 errors for the three methods at different smoothing lengths and time steps, and concludes that IDIC is asymptotic preserving and permits smoothing length and time step to be chosen independently of drag intensity, while MK and ISPH require h < c_s t_stop. The numerical evidence shows dramatically smaller IDIC errors in the reported cases, but the manuscript does not supply an asymptotic-preserving proof, a systematic parameter scan, or a cell-size convergence study, and the reported test parameters contain internal inconsistencies.","tokens_in":7289,"tokens_out":2273,"duration_ms":23645,"significance":"If the central claim is correct, the result is practically significant: it would remove the severe resolution constraint h < c_s t_stop for monodisperse gas-dust mixtures with linear drag, allowing affordable TFSPH simulations of stiff momentum exchange in astrophysical and engineering applications. The paper's positive features include the use of independent analytic reference solutions for both benchmark problems, explicit error tables, and a clear statement of the mechanism (cell-averaged implicit drag) underlying the proposed method. However, the significance is currently conditional on a missing asymptotic analysis or resolution study: the two tests, as presented, cannot separate genuine asymptotic-preserving behavior from favorable grid-smoothing effects, especially because the Euler-cell size is never specified.","major_comments":[{"comment":"The Euler-cell decomposition is introduced only abstractly ('disjoint volumes' with N gas and L dust particles); the actual cell construction, cell size, and any dependence of the results on cell resolution are never reported. Since the IDIC scheme replaces the gas velocity with the cell mean v* and the drag factor and densities with previous-time-layer cell quantities, the cell size is a discretization parameter that directly controls the error. Without a statement of the cell size used in Figures 1–2 and Tables 1–2, or a cell-size convergence study, the numerical results cannot establish that h, rather than the Euler-cell size, is the resolution that controls the asymptotic error. This omission is load-bearing for the paper's central claim.","section":"Section 2.3, Eqs. (19)–(21)"},{"comment":"The claim that IDIC is 'asymptotic preserving' is not supported by a proof or by an asymptotic-resolution study. The paper shows two test problems at fixed values of t_stop, h, and tau; there is no epsilon-asymptotic analysis as t_stop -> 0, no systematic variation of t_stop over several orders of magnitude, and no scan of the final time t_stop relative to t. The observed small IDIC errors at the reported parameters are consistent with the claim but do not demonstrate asymptotic preservation, which requires showing that the numerical solution converges to the equilibrium (zero-relative-velocity) solution at a controlled rate independent of the stiff parameter. The authors should either add a formal consistency/AP argument for Eqs. (19)–(21) or provide a resolved convergence study in t_stop, h, and cell size.","section":"Summary and Section 3"},{"comment":"The reported test parameters are internally inconsistent, which undermines the quantitative comparison. With the definitions in Eqs. (3) and (7), K = 500 and rho_d/rho_g = 1 imply t_stop = 0.002 (taking rho_g = 1), yet the Dustyshock description states t_stop = 0.00025. In addition, Table 1 lists h = 0.01 with tau = 0.00025 and h = 0.02 with tau = 0.001, while the Figure 1 caption states that for h = 0.025 and h = 0.01 the time step is tau = 0.001; Table 2 lists h = 0.01 with tau = 0.000025 and h = 0.02 with tau = 0.0001, while the Figure 2 caption states tau = 0.001 for h = 0.01 and tau = 0.0001 for h = 0.001. These inconsistencies must be resolved before the error values can be trusted as evidence for the AP claim.","section":"Tables 1–2 and Figures 1–2"}],"minor_comments":[{"comment":"The text contains numerous typographical errors and inconsistent capitalization, including 'bahaiviour', 'extreemly', 'assosiated', 'gravitaion', 'dymanics', and 'euler'; a careful proofreading pass is needed.","section":"Throughout"},{"comment":"The captions of Figures 1 and 2 describe smoothing lengths and time steps that do not match the corresponding rows of Tables 1 and 2; the captions and tables should be reconciled so that each reported error is unambiguously associated with a single (h, tau) pair.","section":"Section 3, Figure captions"},{"comment":"It would help the reader to state explicitly whether the Euler cells are uniform, how their size is chosen relative to h, and whether the cell decomposition is recomputed each time step or held fixed.","section":"Section 2.3, Eqs. (19)–(21)"}],"recommendation":"major_revision","confidential_remarks":"This is a conference proceedings paper that rests heavily on the authors' own earlier work [9] for both the method and the benchmark parameters. The central contribution beyond [9] appears to be the head-to-head comparison of MK, ISPH, and IDIC, but the paper does not disclose what is new relative to [9] or provide the reproducibility details (cell size, parameter consistency) needed for independent verification. The editor may also wish to consider whether the paper's brevity is appropriate for the strength of the AP claim made in the Summary."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What you should know: this is a short conference paper that compares three ways to compute drag in two-fluid SPH on the Dustywave and Dustyshock benchmarks. The IDIC method itself is not new, the authors say it was introduced in their 2018 Astronomy and Computing paper. The new content is the head-to-head L2-error comparison with MK and ISPH. On the two tests, IDIC clearly wins: at h = 0.01–0.02, its L2 errors are 20–100 times smaller than MK or ISPH, and the figures agree with the tables. That is a real, useful result for people working on strongly coupled gas-dust SPH.\n\nThe soft spots are in the interpretation. The Summary says IDIC is 'asymptotic preserving' and lets time step and smoothing length be independent of drag intensity. That is a general claim, but the evidence is two test problems with no formal AP analysis, no systematic scan of t_stop, and no convergence study in the Euler cell size. In fact the Euler cell size is never reported in Section 2.3 or the test descriptions, so you cannot tell whether the good behavior comes from the smoothing length h or from the cell averaging. The cell-mean drag in Eqs. (19)–(21) replaces local velocities with cell averages, and there is no error bound in terms of t_stop, h, or cell size. So the 'asymptotic preserving' label is more than the paper justifies.\n\nThe parameter reporting is sloppy. Table 1 gives h = 0.01, τ = 0.00025, while the caption of Fig. 1 says h = 0.01 uses τ = 0.001; Table 2 and Fig. 2 similarly disagree about τ. The Dustyshock t_stop = 0.00025 in the figure caption is not obviously consistent with K = 500 and ρd/ρg = 1. These inconsistencies are minor in the sense that the qualitative conclusion probably survives, but they make the quantitative errors hard to trust. No code or data is shipped, so the numbers cannot be checked independently.\n\nCredit where earned: the comparison is against analytic solutions, so it is not circular. The IDIC momentum-conservation property was shown in the authors' earlier paper. For a conference proceedings, this is a legitimate extension, but the central claim needs to be toned down to 'IDIC behaves robustly on these two stiff test problems.' If this were submitted as a journal paper, I would send it to review with a request for a formal AP argument or a cell-size convergence study, and for corrected parameter reporting. As it stands, it is a useful, citable data point, but not a proof of asymptotic preservation.","headline":"A useful numerical comparison of three TFSPH drag schemes, but the paper's asymptotic-preserving claim for IDIC outruns the evidence: two test problems, no error control for the cell averaging, and parameter tables that do not match the figures.","tokens_in":7832,"tokens_out":2274,"would_cite":false,"duration_ms":24335,"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":"Cell-averaged drag lets two-fluid SPH capture strongly coupled gas-dust dynamics at coarse resolution.","keywords":["gas-solid mixtures","two-fluid smoothed particle hydrodynamics","drag force","stiff relaxation term","asymptotic preserving scheme","particle-in-cell","Epstein drag","Dustyshock test problem"],"falsifier":"Run IDIC on a Dustyshock-style setup over a range of shock strengths and dust-to-gas ratios while keeping the Euler cells fixed, and compare coarse-cell results against a converged high-resolution reference; if the error grows systematically with shock strength or with dust loading, the asymptotic-preserving property does not extend beyond the tested range. A second concrete check is a problem with nonlinear drag, for example $t_{\\rm stop}$ depending on relative speed, where the cell-averaged velocity may not be the correct representative velocity; a systematic bias there would mark the linear-drag boundary of the method.","tokens_in":6770,"feed_emoji":"🪐","tokens_out":12409,"duration_ms":112138,"temperature":0.7,"pith_summary":"The paper compares three ways of computing the drag force between gas and dust in two-fluid Smoothed Particle Hydrodynamics, focusing on monodisperse mixtures where the velocity relaxation time $t_{\\rm stop}$ is much shorter than the dynamical time. In that regime, a classical explicit pairwise drag scheme and a semi-implicit interpolation-based scheme both develop strong dissipation unless the smoothing length obeys $h < c_s t_{\\rm stop}$, which forces very fine spatial resolution. The paper describes and tests a third scheme, 'drag in cell' (IDIC), in which the drag force is computed from cell-averaged gas and dust velocities; it argues that this scheme is asymptotic preserving, so the coupled gas-dust solution is captured even when the timestep $\\tau$ and smoothing length $h$ are much larger than the drag scale. On the Dustywave and Dustyshock tests with intense drag ($K=500$, $\\rho_d/\\rho_g=1$), IDIC reproduces the reference solutions at coarse resolution while the other two methods do not. If the claim holds, strongly coupled gas-dust simulations can run at ordinary SPH resolution, which matters for planet formation and industrial particle-flow applications.","feed_headline":"Coarse dust-gas SPH stays accurate when drag is computed per cell","feed_subtitle":"No fine resolution or tiny timesteps needed for strongly coupled dust-gas flows.","key_machinery":"The central object is the 'drag in cell' (IDIC) approximation. The domain is partitioned into disjoint Euler cells; within a cell containing $N$ gas particles of mass $m_g$ and $L$ dust particles of mass $m_d$, one defines the mass ratio $\\varepsilon^* = m_d L/(m_g N)$, the effective drag factor $K^* = \\rho_d^*/t_{\\rm stop}^*$, and the cell-averaged velocities $v_* = \\sum_a v_a/N$ and $u_* = \\sum_j u_j/L$. Drag on a gas particle uses $v_a - u_*$, drag on a dust particle uses $v_* - u_j$, and the drag factor and densities are evaluated on the previous time layer while the relative velocity is evaluated on the new one. This semi-implicit per-cell treatment of the relaxation term removes the $\\tau < t_{\\rm stop}$ stability restriction and, according to the comparison, the $h < c_s t_{\\rm stop}$ resolution restriction; the first-order-in-time version conserves momentum cell by cell and can be advanced by a direct update. In the stiff limit $t_{\\rm stop}\\to 0$, the scheme is designed to reproduce the fully coupled dusty-gas dynamics without resolving the relaxation scale.","core_discovery":"The central claim is that IDIC, whose drag term is evaluated with Eulerian cells, is asymptotic preserving for monodisperse gas-dust mixtures with linear drag. Inside each cell the gas velocity acting on dust is replaced by the cell mean $v_*$ and the dust velocity acting on gas by the cell mean $u_*$; the drag factor and densities are taken from the previous time layer and the relative velocity from the new one. The paper shows that this first-order semi-implicit treatment conserves momentum cell by cell and allows a direct update of $u^{n+1}$ and $v^{n+1}$. In the two test problems, IDIC keeps the $L_2$ error in dust velocity far below that of the two fully Lagrangian schemes when the smoothing length is raised from $h=0.001$ to $h=0.025$. The summary states that IDIC permits a timestep and smoothing length independent of drag intensity, whereas the classical pairwise and semi-implicit interpolation schemes require fine spatial resolution for intense interphase interaction.","pith_inferences":["A cell-averaged drag law acts as a spatial filter: if the relative velocity varies within a cell, the scheme responds to the mean rather than the local slip, so problems with sub-cell dust-gas velocity structure may need intra-cell velocity-gradient corrections.","The asymptotic-preserving property is demonstrated only for linear drag with $t_{\\rm stop}$ independent of relative velocity; the same cell-averaged implicit idea could plausibly extend to nonlinear drag laws, but the averaging error would then need separate analysis.","In the limit $t_{\\rm stop}\\to 0$ the mixture should behave as a single fluid, and IDIC should reduce naturally to the pure-gas SPH update; verifying this limit in the code would be a direct check of the asymptotic-preserving claim."],"forward_implications":["Strongly coupled gas-dust mixtures can be simulated with SPH resolution set by the gas flow features rather than by the dust stopping length.","The timestep can be chosen from the Courant condition for the gas instead of the much more restrictive $\\tau < t_{\\rm stop}$ condition.","Momentum is conserved per Euler cell, so the coarse-cell drag treatment does not introduce spurious momentum sources or sinks.","The comparison gives a concrete warning that the two fully Lagrangian drag formulations should be used for intense interphase interaction only when $h < c_s t_{\\rm stop}$ is satisfied.","For monodisperse solids with Epstein or Stokes drag, the method opens a practical simulation route for planet formation and particle-flow engineering."],"supporting_citations":[{"why":"Introduced the classical pairwise drag scheme for SPH multi-phase flow that is the baseline method shown to require fine resolution.","marker":"[10]"},{"why":"Defined the Dustywave test problem and identified the resolution condition $h < c_s t_{\\rm stop}$ for this class of schemes.","marker":"[7]"},{"why":"Proposed the semi-implicit interpolation-based drag approach that is compared here and also shown to need fine resolution.","marker":"[8]"},{"why":"Introduced the IDIC scheme and the detailed Dustywave and Dustyshock setups and parameters used for the comparison.","marker":"[9]"},{"why":"Provides the particle-in-cell idea of computing phase interaction through Eulerian cells that IDIC adapts.","marker":"[17]"},{"why":"Supplies the asymptotic-preserving framework for hyperbolic conservation laws with stiff relaxation terms that motivates the design goal.","marker":"[2]"}],"fun_headline_variants":["Cell-based drag makes dust-gas SPH stiff-proof","No tiny steps: per-cell drag tames stiff dust-gas","SPH dust-gas gets asymptotic preserving per-cell drag","Per-cell drag ends fine-tuning for coupled SPH flows"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that a single cell-averaged gas velocity, with drag coefficient and densities taken from the previous time level, faithfully represents the gas seen by dust inside that cell even when the cell contains steep velocity or density gradients; the paper's evidence for this rests on two test problems rather than a general error bound.","fun_headline_variants_meta":{"raw":{"variants":["Cell-based drag makes dust-gas SPH stiff-proof","No tiny steps: per-cell drag tames stiff dust-gas","SPH dust-gas gets asymptotic preserving per-cell drag","Per-cell drag ends fine-tuning for coupled SPH flows"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000192,"raw_usage":{"total_tokens":1349,"prompt_tokens":950,"completion_tokens":399,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":566,"completion_tokens_details":{"reasoning_tokens":333}},"tokens_in":566,"tokens_out":399,"duration_ms":3985,"temperature":1.0,"reasoning_tokens":333,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:06:51.485406+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run IDIC on a Dustyshock-style setup over a range of shock strengths and dust-to-gas ratios while keeping the Euler cells fixed, and compare coarse-cell results against a converged high-resolution reference; if the error grows systematically with shock strength or with dust loading, the asymptotic-preserving property does not extend beyond the tested range. A second concrete check is a problem with nonlinear drag, for example $t_{\\rm stop}$ depending on relative speed, where the cell-averaged velocity may not be the correct representative velocity; a systematic bias there would mark the linear-drag boundary of the method.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduced the classical pairwise drag scheme for SPH multi-phase flow that is the baseline method shown to require fine resolution."},{"cited_title":"Laibe, D","cited_arxiv_id":null,"evidence_quote":"Defined the Dustywave test problem and identified the resolution condition $h < c_s t_{\\rm stop}$ for this class of schemes."},{"cited_title":"Lor´ en-Aguilar, M","cited_arxiv_id":null,"evidence_quote":"Proposed the semi-implicit interpolation-based drag approach that is compared here and also shown to need fine resolution."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduced the IDIC scheme and the detailed Dustywave and Dustyshock setups and parameters used for the comparison."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the particle-in-cell idea of computing phase interaction through Eulerian cells that IDIC adapts."},{"cited_title":"Jin, C.D","cited_arxiv_id":null,"evidence_quote":"Supplies the asymptotic-preserving framework for hyperbolic conservation laws with stiff relaxation terms that motivates the design goal."}],"review_version":1}