{"id":"50edd0b3-5e40-4d11-82d7-717580ab9c3b","arxiv_id":"1908.05398","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"The GRILLIX code is extended to advanced divertor geometries by building grids and parallel operators from numerical equilibria and using conditional field-line tracing and penalisation, with Snowflake advection and diffusion tests shown.","lead":"This paper extends the GRILLIX turbulence code so it can simulate the edge of a fusion device with advanced snowflake divertor magnetic geometry. It tests the new grid and boundary-condition machinery on simple diffusion and advection cases, as a step toward full turbulence runs in such geometries.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Validation of the paper's boundary-treatment claim rests only on qualitative snapshots; the self-admitted fuzzy penalisation boundary is never checked for convergence in width W or relaxation epsilon.","rationale":"The reader's conditional verdict is appropriate and I see no reason to strengthen or weaken it. The strongest claim is modest, but the validation does not quantitatively establish the capability. The most load-bearing concern is the same one the reader identifies: the smoothed penalisation function is admitted in Section 3.4 to place the boundary effectively at different points for different quantities and to complicate boundary fluxes. My stress-test adds that the presented tests cannot even detect such a displacement because they are purely qualitative snapshots with no convergence check or analytic reference. The inverted parabolic/hyperbolic labels in the conclusions are a minor defect, not load-bearing. No code or data artifacts are provided, so independent replication is not possible from the paper alone. The concern does not rise to rejection because the approach is plausible, the mathematical construction is described in detail, and the paper honestly labels the work as initial testing. A quantitative convergence check along the lines proposed would either certify or falsify the boundary treatment. Therefore the conditional verdict remains unchanged.","tokens_in":6487,"tokens_out":4617,"duration_ms":54370,"concrete_test":"Run a 1D parallel-diffusion problem along a field line in the Snowflake equilibrium, using the same trace-angle construction and penalisation as in Section 3.4, for a case where the analytic solution with a Neumann boundary at the target is computable (e.g., method of images). Measure the L2 error in n and the error in the target boundary flux as functions of penalisation width W, relaxation epsilon, and grid spacing; require the errors to vanish at least linearly as W and epsilon go to zero. Repeat for the parallel-advection equation with the Bohm condition u_parallel = +/-1, comparing the time-integrated density loss to the analytic Bohm flux through the target. If the errors do not converge, the fuzzy boundary is not a benign approximation and the central claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central capability claim—that GRILLIX can treat advanced divertor geometries with the intended parallel boundary conditions—rests on the finite-width penalisation operator, yet the only supporting evidence is two qualitative snapshot tests. Section 3.4 explicitly concedes that the boundary condition is not applied exactly at the plate, that determination of boundary fluxes is complicated, and that the effective boundary may sit at different locations for different quantities. The parallel-diffusion and parallel-advection tests (Figures 3 and 4) are not compared with an analytic solution, a reference code, or a convergence study in penalisation width W, relaxation parameter epsilon, or grid spacing. As a result, a spurious shift of the effective boundary by O(W), or an incorrect Neumann/Bohm condition, would produce exactly the sort of sheared, decaying, advected structures shown. The tests certify that something diffuses and advects, not that the penalised boundary enforces the intended conditions at the intended location. This is the load-bearing weak point: the fuzzy-boundary approximation is both self-admitted and unquantified.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports modifications to the flux-coordinate-independent turbulence code GRILLIX that allow it to handle numerically defined magnetic equilibria of advanced divertor configurations. The modifications include preprocessing of EQDSK data, construction of poloidal-flux interpolators and wall/divertor polygons, field-line tracing with conditional stopping at the wall, and a generalization of the finite-width penalisation method to impose parallel boundary conditions at non-conformal divertor plates. The authors present the method as generally applicable to FCI codes and show initial qualitative results for a Snowflake equilibrium: a parallel-diffusion test with Neumann boundary conditions and a parallel-advection test with Bohm and Neumann conditions at the plates. The paper concludes that the approach is promising and that full turbulence simulations on advanced divertor cases are ongoing.","tokens_in":6666,"tokens_out":3754,"duration_ms":39668,"significance":"If the method works as claimed, this is a significant capability for edge/SOL turbulence modelling: FCI codes would be able to treat advanced divertor configurations without coordinate singularities and with flexible grid resolution, which is directly relevant to the design and interpretation of future divertor experiments. The reported generalization of penalisation to non-conformal boundaries and the automatic plate-direction identification via conditional field-line tracing are useful contributions, and the two test cases cover the two PDE classes affected by the parallel operators. However, the paper's support for the central claim is currently qualitative: the evidence consists of two snapshot figures with no error metrics, no convergence studies, and no comparison to reference solutions. The authors' own admission in Sec. 3.4 that the smoothed boundary is not applied exactly at the plate and may shift for different quantities makes quantitative validation particularly important.","major_comments":[{"comment":"The central claim that the finite-width penalisation enforces the intended parallel boundary conditions is supported only by qualitative snapshots. No comparison is made with an analytic solution or reference code, and there is no convergence study in the penalisation width W, the relaxation parameter epsilon, or the grid spacing. This is load-bearing because Sec. 3.4 explicitly concedes that the boundary condition is not applied exactly at the plate and that the effective boundary may lie at different locations for different quantities. A spurious O(W) shift of the boundary, or an incorrect Neumann/Bohm condition, would produce figures indistinguishable from those shown. I request a quantitative test, e.g., convergence of the diffusion profile to the analytic solution of a parallel-diffusion equation with a known target location, as W, epsilon, and grid spacing are varied, together with a report of the resulting effective boundary offset.","section":"Sec. 3.4 and Figs. 3-4"},{"comment":"The advection test is interpreted as 'expected behavior', but the conclusions drawn from it are not supported by any quantitative metric. Statements such as 'density loss near the X-point possibly indicating an ergodic region' and 'primary flow-channels aligned to the separatrix' are qualitative interpretations of a single snapshot; they do not establish that the boundary conditions or field-line operators are correct. Please add a quantitative comparison, for example by advecting a tracer along the traced field lines and comparing the numerical solution to the analytic projection, or by reporting a pointwise or integrated error norm against a reference solution.","section":"Fig. 4 and Sec. 4"},{"comment":"The definition of the smoothed penalisation function, chi = 1 + H3(zeta-, W) - H3(zeta+, W), and the companion function Xi = 1 - 2*H(zeta+ + zeta-) are not fully reproducible as written. The argument conventions for zeta+ and zeta-, the definition of the '3rd order smooth-step function' H3 and the Hermite step function H, and their behavior outside the transition region are not specified. Since the smooth width W is a central parameter of the method and the text emphasizes that its value should be minimized, I ask for explicit formulas and a clear sign convention so that the implementation can be checked and the width-controlled transition understood.","section":"Sec. 3.3-3.4"}],"minor_comments":[{"comment":"The labels 'parabolic' and 'hyperbolic' are swapped in the sentence 'applied to prototypical parabolic (parallel-advection) and hyperbolic (parallel-diffusion) test cases'; advection is hyperbolic and diffusion is parabolic.","section":"Sec. 5"},{"comment":"The phrase 'with n(tau = 0.5) set equal to 1' is unclear; please specify whether this is a normalization of the plotted profile or an imposed initial condition, and define the time normalization used in the right panel.","section":"Fig. 3 caption"},{"comment":"The caption states that chi equals 1 inside the divertor polygon, which appears inconsistent with the convention in Sec. 3.4 where chi = 1 in the ghost region and chi = 0 in the interior domain; please reconcile this notation.","section":"Fig. 2 caption"},{"comment":"Minor typographical issues: 'principle challenges' should be 'principal challenges' and 'diverter' in the abstract should be 'divertor'.","section":"Introduction"}],"recommendation":"major_revision","confidential_remarks":"This is a workshop proceedings paper, and the threshold for validation may reasonably be lower than for a full archival journal article. Nevertheless, the central claim regarding boundary treatment is currently underpinned only by qualitative evidence, and the authors themselves identify the fuzzy-boundary approximation as a limitation. A modest quantitative validation, even for the simple diffusion/advection cases, would materially strengthen the paper and is within the scope of a revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Tom,\n\nHere's my read of Body et al. on GRILLIX for advanced divertor configurations. The paper does what it says: it extends a flux-coordinate-independent turbulence code to numerically defined equilibria like the Snowflake, and the two genuinely new pieces are the conditional field-line tracing that stops at divertor targets and the generalized smoothed penalisation for complex non-conformal boundaries. Both are sensible extensions of the group's earlier work, and the mathematical setup in Sections 3.2-3.4 is consistent with standard derivations.\n\nWhat I like: the paper is clearly scoped and honest. It does not oversell. The authors state plainly that the penalised boundary condition is not applied exactly at the plate, that boundary fluxes are complicated to determine, and that the effective boundary may sit at different locations for different quantities. The tests in Figures 3 and 4 show the expected qualitative behavior: diffusion decays symmetrically until the front reaches the boundary, advection produces blobs that are sheared along the field and a density pattern that maps onto the Snowflake flux geometry. For a code-development report, that is reasonable evidence of a working implementation.\n\nThe soft spot is the one you'd expect from the above: the validation is entirely qualitative. There is no convergence study in the penalisation width W, the relaxation parameter epsilon, or grid spacing, and no comparison with an analytic solution or a reference code. So I cannot fully certify that the smoothed boundary enforces the intended Neumann and Bohm conditions at the intended location. A shift of the effective boundary by O(W), or a slightly wrong boundary condition, would produce snapshots that look broadly like these. That is a genuine gap, but the paper itself flags it and explicitly positions this as initial testing, with saturated turbulence results still pending. So I would treat it as a reason for conditional acceptance, not rejection.\n\nTwo minor points: the conclusions swap the labels for parabolic and hyperbolic when referring to the advection and diffusion test cases — a clarity defect worth fixing. And no code or data beyond the figures is provided, which limits reproducibility. The citation pattern is fine; self-citation to Stegmeir et al. 2019 is appropriate because the penalisation method is extended from there.\n\nWho is this for: anyone working on edge/SOL turbulence codes, especially FCI approaches, and people assessing advanced divertor concepts who need to know what tools exist. It deserves a serious referee. My recommendation: send it to review, with a request that the authors add at least one quantitative check — a convergence scan in W or grid spacing, or a comparison with a known analytic case — before publication.\n\nBest,\n[Your name]","headline":"A modest, honest code-development paper that extends GRILLIX to advanced divertor geometries with two new algorithmic pieces; the main gap is qualitative-only validation of the penalised boundary.","tokens_in":7203,"tokens_out":2173,"would_cite":true,"duration_ms":21070,"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":"Plasma turbulence code now treats advanced divertor geometries","keywords":["flux-coordinate independent","GRILLIX","advanced divertor configurations","snowflake divertor","penalisation method","parallel advection","parallel diffusion","tokamak edge turbulence"],"falsifier":"A decisive test would be to run the parallel advection and diffusion cases on the same Snowflake equilibrium with progressively smaller penalisation widths and with an exactly conformal grid, and compare the density and velocity fields near the divertor plate. If the apparent boundary location shifts with the penalisation width by more than the grid spacing, or if the late-time density profiles differ noticeably from the conformal-grid result, the fuzzy-boundary approximation is not faithful enough for divertor turbulence studies.","tokens_in":6277,"feed_emoji":"🧲","tokens_out":6652,"duration_ms":61870,"temperature":0.7,"pith_summary":"This paper extends the flux-coordinate independent turbulence code GRILLIX so that it can treat advanced divertor configurations such as the Snowflake divertor, whose magnetic geometry is supplied as a numerical equilibrium rather than as an analytic field. The motivation is practical: advanced divertors are designed to spread heat and control detachment, and turbulent-transport simulations of these shapes have been difficult for field-aligned codes because of coordinate singularities at X-points and severe resolution requirements. The paper shows how the computational grid and parallel operators are built from numerically defined equilibria, and how parallel boundary conditions are imposed at non-conformal divertor boundaries by a generalised finite-width penalisation method. Prototypical advection and diffusion tests in the Snowflake scenario behave as expected, and work on the full turbulence model is reported as ongoing.","feed_headline":"Plasma turbulence code now treats advanced divertor geometries","feed_subtitle":"Snowflake-config tests pass advection and diffusion checks, opening advanced divertors to turbulence simulation.","key_machinery":"The load-bearing object is the flux-coordinate independent (FCI) discretisation together with the penalisation boundary treatment. In FCI, the turbulence equations are solved on a Cartesian-like poloidal grid that is independent of the magnetic field, and the parallel structure of the field is encoded through operators built by tracing field lines from each grid point. The new machinery for advanced divertors consists of: (1) a bicubic spline interpolator for $\\Psi(R,Z)$, from which the field components follow by $B_Z = (1/2\\pi R)\\partial_R \\Psi$ and $B_R = -(1/2\\pi R)\\partial_Z \\Psi$; (2) polygon objects using the winding algorithm to mark interior and exterior regions; (3) conditional field-line tracing with the dop853 integrator, producing interior/exterior trace angles $\\zeta^+$ and $\\zeta^-$ to the targets; and (4) a smoothed penalisation characteristic function $\\chi = 1 + \\sigma_3(\\zeta^-,\\chi_W) - \\sigma_3(\\zeta^+,\\chi_W)$ with a Hermite smooth-step of width $\\chi_W$, plus a sign function $\\Xi = 1 - 2\\sigma(\\zeta^+ + \\zeta^-)$ for the direction of the nearest plate. The penalisation layer couples the staggered and full grids at the boundary and enforces the parallel boundary conditions as the relaxation parameter tends to zero.","core_discovery":"The central claim is that the flux-coordinate independent approach, in which the grid is chosen independently of the magnetic field and the field is encoded in parallel operators, can be applied to numerically defined advanced divertor equilibria. The paper establishes the construction pipeline: a bicubic spline interpolation of the poloidal flux $\\Psi$ from an eqdsk file, polygon-based identification of the first wall, divertor, flux-limit regions and end-of-domain, field-line tracing with an adaptive 8th-order integrator, and conditional traces that determine the toroidal angle to the divertor target in each direction. Boundary conditions are then imposed through a smoothed characteristic function $\\chi$ built from smooth-step functions of the trace angles, with a second function $\\Xi$ giving the direction to the nearest plate. The authors report that parallel diffusion of a blob decays symmetrically until the front reaches the boundary, and that parallel advection with Bohm and Neumann conditions produces the expected outflow in the SOL, density accumulation near the X-point and poloidal flux plateau, and flow channels aligned with the separatrix.","pith_inferences":["If the tests carry over to the full model, the same preprocessing and tracing code could be applied directly to the other advanced divertor designs (X-divertor, Super-X, double-null), since they are also supplied as numerical equilibria; the only geometry-specific pieces are the polygons and flux limits.","The paper's own concern about the smoothed boundary suggests a quantitative convergence study in penalisation width; a code that used strong-sink or other exact boundary conditions would provide a clean benchmark for whether the fuzzy boundary changes turbulent fluxes.","A geometry scan enabled by this method could isolate the effect of flux expansion and connection length on SOL turbulence, comparing Snowflake against conventional single-null on grids of identical resolution."],"forward_implications":["The same pipeline of $\\Psi$-interpolation, polygon preprocessing, and conditional field-line tracing can in principle be reused for any advanced divertor equilibrium supplied as eqdsk data.","FCI avoids coordinate singularities at X-points, so the Snowflake's extra X-point needs no special numerical treatment.","Parallel boundary conditions at complex non-conformal divertor targets are imposed through the penalisation layer without changing the grid structure.","In the diffusion test the density decays symmetrically until the front reaches the penalisation boundary, then the Neumann condition alters the evolution, matching expected behaviour.","In the advection test with Bohm velocity and Neumann density boundary conditions the density is expelled from the SOL and accumulates near the X-point and the poloidal flux plateau, with flow channels aligned to the separatrix."],"supporting_citations":[{"why":"Supplies the GRILLIX FCI scheme, the support-operator parallel discretisation, and the original penalisation method that this paper generalises.","marker":"[1]"},{"why":"Earlier description of the GRILLIX code and flux-coordinate independent approach that the present modifications build on.","marker":"[2]"},{"why":"The penalisation technique for imposing boundary conditions in fluid codes, the basis for the finite-width generalisation.","marker":"[3]"},{"why":"Defines the Snowflake divertor configuration used as the test geometry.","marker":"[7]"},{"why":"Provides the bicubic spline interpolation of the poloidal flux used to build field components and grid quantities.","marker":"[13]"},{"why":"The winding algorithm used by the polygon object to decide whether points lie inside the first wall, divertor, flux-limits, or end-of-domain regions.","marker":"[15]"},{"why":"Supplies the adaptive 8th-order Runge-Kutta integrator used for field-line tracing and conditional traces to the divertor targets.","marker":"[16]"},{"why":"The strong-sink boundary conditions discussed as a possible future alternative to the smoothed penalisation layer.","marker":"[18]"}],"fun_headline_variants":["GRILLIX turbulence code now handles snowflake divertors","Advanced divertor geometries simulated in GRILLIX","Flux-coordinate independent method tests snowflake divertor","GRILLIX code extends to advanced divertor configurations","Snowflake divertor test passed by GRILLIX turbulence code"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The method stands or falls on whether the smoothed, finite-width penalisation layer enforces the intended parallel boundary conditions faithfully; the paper itself notes the boundary condition is not applied exactly at the plate, and the effective boundary may sit at slightly different locations for different quantities.","fun_headline_variants_meta":{"raw":{"variants":["GRILLIX turbulence code now handles snowflake divertors","Advanced divertor geometries simulated in GRILLIX","Flux-coordinate independent method tests snowflake divertor","GRILLIX code extends to advanced divertor configurations","Snowflake divertor test passed by GRILLIX turbulence code"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000153,"raw_usage":{"total_tokens":1182,"prompt_tokens":896,"completion_tokens":286,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":512,"completion_tokens_details":{"reasoning_tokens":203}},"tokens_in":512,"tokens_out":286,"duration_ms":3402,"temperature":1.0,"reasoning_tokens":203,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:31:10.114368+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A decisive test would be to run the parallel advection and diffusion cases on the same Snowflake equilibrium with progressively smaller penalisation widths and with an exactly conformal grid, and compare the density and velocity fields near the divertor plate. If the apparent boundary location shifts with the penalisation width by more than the grid spacing, or if the late-time density profiles differ noticeably from the conformal-grid result, the fuzzy-boundary approximation is not faithful enough for divertor turbulence studies.","supporting_citations":[{"cited_title":"Stegmeir, et al.,Physics of Plasmas2019,26 (5), 052517","cited_arxiv_id":null,"evidence_quote":"Supplies the GRILLIX FCI scheme, the support-operator parallel discretisation, and the original penalisation method that this paper generalises."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Earlier description of the GRILLIX code and flux-coordinate independent approach that the present modifications build on."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The penalisation technique for imposing boundary conditions in fluid codes, the basis for the finite-width generalisation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the Snowflake divertor configuration used as the test geometry."},{"cited_title":"Williams,Multidimensional B-Spline Interpolation of Data on a Regular Grid, github.com/jacobwilliams/bspline-fortran,2015–2019","cited_arxiv_id":null,"evidence_quote":"Provides the bicubic spline interpolation of the poloidal flux used to build field components and grid quantities."},{"cited_title":"Naresh Kumar, Mallikarjun Bangi,IFAC-PapersOnLine2018, 51(1), 548 – 553","cited_arxiv_id":null,"evidence_quote":"The winding algorithm used by the polygon object to decide whether points lie inside the first wall, divertor, flux-limits, or end-of-domain regions."},{"cited_title":"Williams,Modern Fortran Edition of Hairer’s DOP853 ODE Solver, github.com/jacobwilliams/dop853,2015–2017","cited_arxiv_id":null,"evidence_quote":"Supplies the adaptive 8th-order Runge-Kutta integrator used for field-line tracing and conditional traces to the divertor targets."},{"cited_title":"Paredes, et al.,Journal of Computational Physics2014,274, 283–298","cited_arxiv_id":null,"evidence_quote":"The strong-sink boundary conditions discussed as a possible future alternative to the smoothed penalisation layer."}],"review_version":1}