{"id":"4dd72b72-8480-4612-9894-2c500bd577c4","arxiv_id":"2505.24343","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"For smooth accretion discs in SPH, the default quadratic artificial viscosity beta = 2 adds too much numerical viscosity, and linking beta to the switched alpha (beta = 2 alpha) gives the least numerical spreading.","lead":"This paper tests how the artificial viscosity settings in the SPH code phantom change the spread of matter in computer models of accretion discs, and finds the default quadratic viscosity is too strong for smooth discs. It recommends linking the quadratic term to the same time-dependent switch already used on the linear term, a fix that reduces unwanted numerical friction without sacrificing shock capture.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The β=2α switch recommendation is untested in shocks, which is the regime the switch exists for; the paper concedes this in footnote 9 and leaves a conflict with Price & Federrath (2010) unresolved.","rationale":"The reader's concern about the surface-density proxy is real but secondary: the relative ranking of β values is robust because non-viscous modeling differences (pressure gradients, boundary torque) should be nearly independent of the viscosity coefficients, so the comparison between β=2 and β=0.2 is still meaningful. The more load-bearing gap is that the paper's headline recommendation—a switch on β via β=2α—is validated only in quiescent discs, where it degenerates to a constant low β. The switch's purpose is to handle shocks, and the paper explicitly declines to test that regime (footnote 9), leaving a known contradiction with Price & Federrath (2010) unresolved. This supports the reader's CONDITIONAL verdict but for a different reason: the recommendation is not yet demonstrated for the main scenario it targets.","tokens_in":15288,"tokens_out":20497,"duration_ms":253847,"concrete_test":"Run a high-Mach shock tube (e.g., colliding streams with Mach ≥ 10) in phantom with the Cullen–Dehnen switch and β = 2α (α_min = 0), and compare post-shock density overshoot, particle penetration, and entropy error against the phantom default (switched α, constant β = 2). If the β = 2α run shows penetration or excessive overshoot, the recommendation fails in the regime it is meant for; if it matches the default, the recommendation is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 5's central recommendation—set β_SPH = 2 α_SPH with α_SPH from the Cullen–Dehnen switch—is supported only by smooth-disc simulations (Fig. 2), where the switch reduces β to ≈0.2 and the runs match the best constant-low-β cases. The switch exists to provide high β in strong shocks, but no shock test is presented. Footnote 9 explicitly concedes: 'It would therefore be worthwhile repeating the simulations of Price & Federrath (2010) with the Cullen-Dehnen switch to see if the requirement of a larger β_SPH value is necessary.' Price & Federrath (2010) found β=2 insufficient with the Morris–Monaghan switch, and the paper does not resolve this conflict. Without a shock test, the advice may fail in the high-Mach regime it is designed for; the smooth-disc evidence alone only shows that a constant small β is preferable, which is already established by Lodato & Rice (2004) and Meru & Bate (2012).","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper uses phantom SPH simulations of steady, planar, circular accretion discs with an explicit Navier-Stokes Shakura-Sunyaev viscosity (alpha_SS = 0.1) to study how the artificial-viscosity coefficients alpha_SPH and beta_SPH affect the effective numerical viscosity. By comparing the steady-state surface density profiles against a semi-analytic 1D solution at three resolutions (10^5, 10^6, 10^7 particles) and for a grid of constant coefficients and switched-alpha runs, the authors find that the default phantom value beta_SPH = 2 is unnecessarily dissipative in smooth discs, and that reducing beta (to ~0.2 or below) gives higher surface densities. They recommend applying a viscosity switch to the quadratic term by setting beta_SPH = 2 alpha_SPH with alpha_SPH from the Cullen-Dehnen switch, and they also recommend using an explicit Navier-Stokes viscosity rather than the 'artificial viscosity for a disc' method.","tokens_in":15528,"tokens_out":9985,"duration_ms":122588,"significance":"The paper addresses a practically important issue for the large community of phantom users: the default beta_SPH = 2 may dominate the numerical viscosity in well-resolved smooth discs. The study has strengths: it uses an external semi-analytic benchmark rather than fitting to simulation output, it covers three decades of particle number, and the qualitative ordering of models is consistent across resolutions. However, the main prescriptive result, beta_SPH = 2 alpha_SPH, is not new (it is already advocated in Morris & Monaghan 1997 and Cullen & Dehnen 2010), and the paper's smooth-disc experiments alone do not test the shock-capturing behaviour that motivates a time-dependent beta. The value of the paper is therefore primarily as a quantitative demonstration that the quadratic term is often the dominant numerical viscosity in disc simulations, rather than as a new numerical scheme.","major_comments":[{"comment":"The recommendation beta_SPH = 2 alpha_SPH is presented as a way to keep the quadratic viscosity low in smooth flows while allowing it to grow in strong shocks, but no shock test is provided. Footnote 9 explicitly concedes that Price & Federrath (2010) found beta = 2 insufficient with the Morris-Monaghan switch for strong shocks and that the question of whether the Cullen-Dehnen switch would behave differently is open. Because the smooth-disc results alone only show that constant low beta is preferable, the advantage of the proposed prescription over simply using beta = 0.02 rests on its shock behaviour, which is untested here. Please add a shock-tube or converging-flow test (e.g., the Price & Federrath 2010 setup) or clearly restrict the recommendation to smooth, planar discs.","section":"Section 5 (final paragraph), Section 6, footnote 9"},{"comment":"The inference that the surface-density deficit measures numerical viscosity assumes that other differences between the SPH simulations and the 1D semi-analytic model are subdominant. Pressure gradients, finite-width mass injection, and boundary torques are acknowledged in Section 4.1 but are not quantified, and no uncertainties are attached to the binned surface-density profiles. A control run with the physical viscosity disabled, or a residual plot normalised by the analytic profile at each resolution, would help separate the numerical-viscosity signal from these systematics; without it, the quantitative claim of having 'minimised' the numerical viscosity is not tightly constrained.","section":"Section 4.3, Eq. (4), Fig. 1"},{"comment":"The grid in beta_SPH (2, 0.2, 0.02) is coarse and no diagnostic of particle penetration or particle order is reported. The manuscript's abstract motivates an optimum by the competing effects of dissipation and particle disorder, but the simulations show a monotonic improvement down to the smallest tested beta (0.02), so the claimed optimum at alpha = 0.1, beta = 0.2 is not actually established by the data; it is consistent with earlier work but not demonstrated here. Please either extend the beta grid to lower values, add order/penetration diagnostics, or rephrase the conclusion as an upper bound rather than an optimum.","section":"Table 1, Section 4.3, Section 6"}],"minor_comments":[{"comment":"The statement that the 'optimal' case alpha = 0.1, beta = 0.2 overlaps essentially exactly with A9 (alpha = 0.01, beta = 0.02) is confusing; if both give the same profile, the optimum is a plateau, and the ratio beta/alpha rather than the absolute values may be the controlling parameter. Please clarify.","section":"Footnote 11"},{"comment":"The text says the beta = 2 alpha runs match the highest profiles, but footnote 10 reports that models B7 and B9 are slightly higher in R < 2; please adjust the wording to 'closely match' or quantify the difference.","section":"Fig. 2, Footnote 10"},{"comment":"The grouping of models into subgroups is useful, but the plots omit several models (A5, A6, A8, B5, B7, etc.) that are discussed; consider including them in an appendix or providing a table of the disc masses for all models so the grouping can be verified.","section":"Section 4.2 and Fig. 1"},{"comment":"The phrase that both Morris & Monaghan (1997) and Cullen & Dehnen (2010) switches apply to both the linear and quadratic terms is correct for the original formulations, but phantom's default implementation only switches alpha; readers may benefit from an explicit note that the BB runs here required modifying the standard phantom setup.","section":"Section 5, last paragraph"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid empirical study with a useful practical message, but the central recommendation is a known prescription and the shock-test gap is explicitly acknowledged in the manuscript. If the authors can add a shock test or substantially temper the claims to smooth discs, it would be publishable. The appendix's critique of the 'artificial viscosity for a disc' method is useful but somewhat orthogonal to the main thread."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Chris,\n\nHere is my take on Chen & Nixon. The paper is a clean, practical parameter study of SPH numerical viscosity in disc simulations. The headline result—that phantom's default constant beta_SPH=2 gives too much numerical viscosity in a smooth, planar disc, and that setting beta_SPH=2 alpha_SPH with the Cullen-Dehnen switch on alpha reproduces the best steady-state surface density profiles—is well supported by the three-resolution runs. That is worth knowing if you run phantom discs.\n\nWhat is genuinely new is not the beta=2alpha idea, which they correctly attribute to Morris & Monaghan (1997) and Cullen & Dehnen (2010). It is the quantitative demonstration in phantom, plus the warning that the \"artificial viscosity for a disc\" setup mis-shapes the surface density (Appendix A). The steady-state injection method from Drewes & Nixon is a good way to isolate numerical viscosity, and the authors are careful to separate cases where the linear or quadratic term dominates.\n\nThe soft spots are real but not fatal. The comparison rests on the surface density deficit relative to a 1D semi-analytic solution, with no error bars and a coarse beta grid (2, 0.2, 0.02), so the statement that beta~0.2 is optimal is a bit weak; the optimum could be anywhere between 0.2 and 2. The paper does not measure particle penetration or order at low beta, so we cannot tell if even lower beta would be better. And, as footnote 9 concedes, there is no shock test. The switch exists to respond to shocks, and the claim that beta=2alpha will behave in high-Mach flows is an extrapolation from Cullen & Dehnen's test plus an assumption. The authors do point out in footnote 3 that Price & Federrath's beta=4 requirement came from a different implementation where beta was multiplied by alpha<=1, so the conflict is softened, but the shock test remains undone.\n\nStill, the central argument is coherent and honestly scoped. I would send this to review. A referee should ask for a shock test (or a clearer statement that the recommendation is tentative for shocks), and for uncertainty quantification on the profiles. The paper is a useful practical contribution to the phantom community, and the appendix warning about the \"artificial viscosity for a disc\" method is a service.\n\n— [Your name]","headline":"Solid, useful demonstration that phantom's default beta=2 is over-viscous in smooth discs, but the shock regime is untested—the switched-beta recommendation is reasonable but extrapolated.","tokens_in":16025,"tokens_out":4679,"would_cite":true,"duration_ms":54023,"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":"For flat circular discs, the default quadratic artificial viscosity in SPH disc simulations is higher than needed, and the quadratic coefficient should be switched along with the linear one.","keywords":["smoothed particle hydrodynamics","numerical viscosity","artificial viscosity","accretion discs","viscosity switches","steady-state discs","Shakura-Sunyaev viscosity"],"falsifier":"Run the same steady-state disc setup at $\\beta_{\\rm SPH} = 0.2$ and $\\beta_{\\rm SPH} = 0.02$ while directly tracking particle interpenetration and measuring the shear stress: if lowering $\\beta$ again further does not raise the surface density, or raises it only through particle-reordering dissipation, the claimed minimum at $\\beta\\approx 0.2$ is not the true minimum of numerical viscosity.","tokens_in":15104,"feed_emoji":"🌀","tokens_out":11979,"duration_ms":117796,"temperature":0.7,"pith_summary":"This paper asks whether the artificial viscosity that SPH codes add to handle shocks and keep particles ordered is stronger than it needs to be for ordinary disc simulations. It answers by running steady-state accretion discs in the phantom code, feeding mass at a fixed rate and letting it flow out through inner and outer boundaries, then comparing the resulting surface density to the semi-analytic solution of the thin-disc diffusion equation. The lower the numerical viscosity, the closer the simulated disc stays to the semi-analytic density, so the coefficient values that maximise the steady-state surface density are the ones that minimise spurious dissipation. The finding is that for flat circular discs the default quadratic coefficient $\\beta_{\\rm SPH} = 2$ is too high, values near $\\alpha_{\\rm SPH} = 0.1$ and $\\beta_{\\rm SPH} = 0.2$ are closer to optimal, and the best balance comes from linking $\\beta_{\\rm SPH}$ to a switched $\\alpha_{\\rm SPH}$, e.g. $\\beta_{\\rm SPH} = 2\\alpha_{\\rm SPH}$. If this is right, many SPH disc simulations have been dissipating more angular momentum and heat than intended, and the remedy is cheap: apply the existing viscosity switch to the quadratic term as well.","feed_headline":"SPH disc defaults add extra viscosity — cut them","feed_subtitle":"Steady-state disc tests show beta=2 dissipates too much; switching beta with alpha keeps shocks while minimising drag.","key_machinery":"The central measuring device is the steady-state accretion disc: mass is injected at a fixed rate and removed at inner and outer boundaries, so the surface density profile that forms is a direct readout of the total (physical plus numerical) viscosity acting in the simulation. The benchmark is the semi-analytic steady-state solution of the 1D diffusion equation (Eq. 4): numerical viscosity lowers the simulated surface density below this solution, and lower coefficients keep the disc closer to it. The key identity is the Meru–Bate ratio (Eq. 6), $\\alpha_{\\rm quad}/\\alpha_{\\rm lin} = (135/62\\pi)(\\beta_{\\rm SPH}/\\alpha_{\\rm SPH})(\\langle h\\rangle/H)$, which the paper uses to show that when $\\alpha_{\\rm SPH}$ from a switch is $\\approx 0.1$ and $\\beta_{\\rm SPH}$ stays at 2, the quadratic term dominates the linear one unless the disc is resolved with $\\langle h\\rangle/H \\lesssim 0.1$ — a condition most simulations in the literature do not meet. This ratio is what motivates switching $\\beta_{\\rm SPH}$ as well.","core_discovery":"The paper's central claim is that, for planar and circular discs, the default values of the numerical viscosity parameters in the phantom code can be too high, particularly the quadratic term $\\beta_{\\rm SPH} = 2$, and that the quadratic coefficient should be made time-dependent through a switch just as the linear coefficient already is. The evidence comes from steady-state disc simulations at three resolutions: when the switch is active, varying the minimum $\\alpha_{\\rm SPH}$ between 0 and 0.1 makes almost no difference because the shell-averaged $\\alpha_{\\rm SPH}$ sits near 0.1, whereas the choice of $\\beta_{\\rm SPH} = 2$ versus $\\beta_{\\rm SPH} = 0.2$ splits the results into two clear groups, with the high-$\\beta$ runs falling well below the semi-analytic surface density. Runs with $\\beta_{\\rm SPH} = 2\\alpha_{\\rm SPH}$ reproduce the highest surface-density profiles of any configuration, showing that linking the quadratic term to the switched linear coefficient minimises numerical viscosity in smooth flows while preserving the ability to grow in shocks.","pith_inferences":["The same steady-state surface-density probe could be rerun for warped or eccentric discs, where the shear and compression structure differs; the planar circular optimum of $\\alpha\\approx 0.1$, $\\beta\\approx 0.2$ may not hold there.","A strong-shock test (Mach number $\\gtrsim 10$) with $\\beta_{\\rm SPH} = 2\\alpha_{\\rm SPH}$ is the natural check of whether the switch raises the quadratic term quickly enough to prevent particle interpenetration; the authors note the conflict between Price & Federrath (2010) and Cullen & Dehnen (2010) on this point remains open.","The Meru–Bate ratio can be read as a resolution requirement: with constant $\\beta_{\\rm SPH}=2$ and switched $\\alpha_{\\rm SPH}\\approx 0.1$, one needs $\\langle h\\rangle/H \\lesssim 0.1$ for the linear term to dominate, which is much stricter than the usual $\\langle h\\rangle/H < 1$ guidance.","Measuring the residual numerical viscosity as a function of resolution at fixed $(\\alpha_{\\rm SPH}, \\beta_{\\rm SPH})$ and comparing it with the $\\langle h\\rangle/H$ scaling of the Meru–Bate ratio would give a direct calibration of how much switching $\\beta$ actually saves."],"forward_implications":["SPH disc runs that keep $\\beta_{\\rm SPH}=2$ constant (the phantom default with a switch on $\\alpha$ only) dissipate more angular momentum and heat than necessary in smooth planar discs, pushing the steady-state surface density below the semi-analytic expectation.","Setting $\\beta_{\\rm SPH} = 2\\alpha_{\\rm SPH}$, with $\\alpha_{\\rm SPH}$ from a Cullen–Dehnen switch, reproduces the lowest numerical viscosity found in any configuration for smooth flows while keeping the viscosity available for strong shocks, and it removes one user-set parameter.","In well-resolved discs with switched $\\alpha_{\\rm SPH}\\approx 0.1$ and constant $\\beta_{\\rm SPH}=2$, the quadratic term contributes more numerical viscosity than the linear term unless $\\langle h\\rangle/H \\lesssim 0.1$, a condition few simulations satisfy, so the quadratic term should not be neglected when estimating numerical viscosity.","The 'artificial viscosity for a disc' method of Lodato & Price (2010) does not reproduce the steady-state surface density shape even at higher resolution; an explicit Navier-Stokes viscosity is the safer way to model a Shakura-Sunyaev $\\alpha$.","As resolution increases, numerical viscosity falls below the physical viscosity ($\\alpha_{\\rm SS}=0.1$ in these runs), so at high resolution the choice of $\\alpha_{\\rm SPH}$ and $\\beta_{\\rm SPH}$ becomes secondary to resolving the disc scale height."],"supporting_citations":[{"why":"Provides the analytic steady-state surface density solution (Eq. 4) that the simulations are measured against.","marker":"Nixon & Pringle (2021)"},{"why":"Describes the mass-injection method in phantom used to create and sustain the steady-state discs.","marker":"Drewes & Nixon (2021)"},{"why":"Supplies the ratio formula (Eq. 6) for quadratic-to-linear numerical viscosity and the prior choice of alpha=0.1, beta=0.2.","marker":"Meru & Bate (2012)"},{"why":"Introduces the viscosity switch that phantom adapts for alpha_SPH, and which the paper argues should also govern beta_SPH.","marker":"Cullen & Dehnen (2010)"},{"why":"First proposed a time-dependent viscosity switch acting on both linear and quadratic terms, cited as prior support.","marker":"Morris & Monaghan (1997)"},{"why":"Documents phantom's default numerical viscosity: constant beta_SPH = 2 with a modified Cullen-Dehnen switch on alpha_SPH only.","marker":"Price et al. (2018)"},{"why":"Source of both the criticised 'artificial viscosity for a disc' method (Sec. 3.2.3) and the recommended explicit Navier-Stokes viscosity (Sec. 3.2.4).","marker":"Lodato & Price (2010)"},{"why":"Claims beta=4 is needed to prevent particle penetration in strong shocks; the paper questions this and suggests re-testing with a switch on beta.","marker":"Price & Federrath (2010)"}],"fun_headline_variants":["Trim SPH viscosity: default beta too high for discs","Beta=2 wastes disc mass: switch quadratic SPH viscosity","Link SPH beta to alpha switch to cut disc viscosity","For smooth discs, default SPH beta adds excess drag"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the gap between the simulated steady-state surface density and the 1D semi-analytic solution is almost entirely due to numerical viscosity, with pressure gradients, boundary torques and finite-width mass injection contributing negligibly; the paper acknowledges these effects but does not quantify them.","fun_headline_variants_meta":{"raw":{"variants":["Trim SPH viscosity: default beta too high for discs","Beta=2 wastes disc mass: switch quadratic SPH viscosity","Link SPH beta to alpha switch to cut disc viscosity","For smooth discs, default SPH beta adds excess drag"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000298,"raw_usage":{"total_tokens":1785,"prompt_tokens":1063,"completion_tokens":722,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":679,"completion_tokens_details":{"reasoning_tokens":654}},"tokens_in":679,"tokens_out":722,"duration_ms":7644,"temperature":1.0,"reasoning_tokens":654,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T12:24:46.145252+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same steady-state disc setup at $\\beta_{\\rm SPH} = 0.2$ and $\\beta_{\\rm SPH} = 0.02$ while directly tracking particle interpenetration and measuring the shear stress: if lowering $\\beta$ again further does not raise the surface density, or raises it only through particle-reordering dissipation, the claimed minimum at $\\beta\\approx 0.2$ is not the true minimum of numerical viscosity.","supporting_citations":[],"review_version":1}