{"id":"545093ae-c15d-4058-a414-2867763a6e95","arxiv_id":"2412.12396","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A SUPG-stabilized continuous Galerkin mixed discretization for anisotropic heat conduction reduces spurious perpendicular heat loss in tokamak-style simulations from about 32-35% to 4%.","lead":"This paper introduces a new numerical method for simulating heat flow in magnetized fusion plasmas, where heat travels far faster along magnetic field lines than across them. The method cuts spurious heat leaking perpendicular to the field from roughly a third of stored energy down to a few percent in long tokamak simulations.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The consistency proof (Prop. 3.1) assumes continuous τ, but the implemented τ from (3.15) is cell-wise discontinuous on the unstructured mesh; the integration-by-parts step (3.9) then omits interface jump terms, so the proof does not cover the scheme used for the headline 4% tokamak result.","rationale":"The reader identified the heuristic τ choice and missing stability analysis as the main weakness; I agree, but the sharper problem is that the implemented τ is not the one analyzed. The theory in Definition 3.1 and Proposition 3.1 explicitly requires continuous τ, while Section 3 implements (3.15) with a cell-wise δx on an unstructured mesh. At every interior face the jump of τ generates terms missing from (3.9), so the consistency statement proven in the paper does not apply to the method that produces the 35%/32%-vs-4% numbers. This is an internal inconsistency between the mathematical claim and the numerical experiments, not a disagreement with a prevailing view. The numerical convergence test avoids the issue because its mesh is uniform, and the tokamak result is precisely where the mesh is unstructured and δx varies. I am not claiming the method fails; the empirical evidence is suggestive. But the paper should either use a continuous τ (projected or globally smoothed), extend the consistency analysis to discontinuous τ with explicit interface terms, or demonstrate numerically that the interface terms are negligible. Because the reader already returned CONDITIONAL, my concern does not change the overall verdict; it sharpens the condition that must be met before the method is adopted.","tokens_in":19712,"tokens_out":15787,"duration_ms":140551,"concrete_test":"Re-run the Section 4.3 tokamak equilibrium at the same resolution with two τ variants: (A) exactly as implemented, τ from (3.15) with cell-wise δx at quadrature points; (B) a continuous τ obtained by projecting the same expression into the P2 CG space (or using a globally continuous mesh-length function). Keep all other settings identical and compare the final relative total available temperature (4.7) for the P2×P2 SUPG method. If (A) and (B) differ by more than about 2 percentage points, or if (B) is unstable or qualitatively different, the 4% claim depends on the discontinuous-τ evaluation and Proposition 3.1 does not justify it. As a secondary check, repeat the §4.1 convergence study on a nonuniform mesh where δx varies; a significant drop in convergence order would confirm that the cell-wise τ evaluation changes the scheme's consistency.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Definition 3.1 defines M^f(χ,υ) := ⟨χ, υ + s·∇(τ υ)⟩ and the surrounding text states that τ must be continuous for M^f to be well-defined. The consistency proof of Proposition 3.1 integrates this term by parts to obtain (3.9); that step requires τ to be differentiable and produces no cross-element jumps. In the 'Stabilization parameter' subsection, τ is prescribed by (3.15) with a local cell length scale δx and, in practice, 'evaluated as an expression ... on quadrature points' on the unstructured tokamak mesh. Since δx varies from cell to cell, τ is generally discontinuous across cell faces. For a discontinuous τ, the distributional derivative of τ contains interface delta contributions; evaluating τ∇υ at quadrature points inside each cell drops these contributions. The equality (3.9) and the resulting consistency of the implemented bilinear form therefore do not follow from Proposition 3.1. The convergence test in §4.1 uses a uniform mesh, where δx (and hence τ) is constant, so it does not exercise the discontinuous-τ regime. The headline 4% spurious-loss result in §4.3 is computed with the cell-wise τ implementation, leaving the central accuracy claim outside the scope of the paper's consistency argument. The manuscript acknowledges that continuous τ is needed and says the quadrature evaluation 'found to work well', but no interface analysis or numerical evidence addresses the gap.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a continuous Galerkin (CG) discretization for the anisotropic heat-conduction equation, using a mixed formulation in which an auxiliary variable approximates the directional derivative s·∇T and SUPG stabilization is applied to both the temperature and auxiliary-variable equations. The authors prove consistency of the bilinear form (Proposition 3.1), discuss the loss of symmetry and ellipticity of the resulting Petrov-Galerkin operator, and validate the method on a manufactured convergence test, a 2D flux-surface perturbation, a full-torus tokamak equilibrium sustainment case with κ⊥=0, and a tokamak flux-tube perturbation. The headline numerical claim is that the new method loses about 4% of available thermal energy in the long-run equilibrium test, versus about 35% and 32% for the primal and standard mixed CG formulations at matched resolution.","tokens_in":19993,"tokens_out":24997,"duration_ms":206272,"significance":"If the claims hold, this is a practically relevant contribution: it offers a CG-based alternative to the authors' earlier DG-upwind method, using the same auxiliary-variable idea but with lower implementation complexity for existing CG-based MHD codes. The test design is a strength: the κ⊥=0 equilibrium has an exact stationary solution, so the measured energy loss is directly attributable to spurious perpendicular transport; the comparisons are at matched degrees of freedom; and the consistency calculation in Proposition 3.1 is clean under its stated regularity assumptions. The authors also candidly acknowledge the loss of symmetry and ellipticity and the heuristic nature of τ. However, the consistency proof does not cover the quadrature-evaluated, cell-wise discontinuous τ actually used in the tokamak runs, and the definition of τ in Eq. (3.15) contains a dimensional inconsistency; these issues must be resolved before the central accuracy claim is fully supported.","major_comments":[{"comment":"The consistency proof assumes τ is continuous: Definition 3.1 states that M^f is well-defined 'provided that the stabilization parameter τ is continuous', and the integration by parts leading to Eq. (3.9) uses a global differentiable τ with no inter-element jumps. In the implementation, however, τ is prescribed by Eq. (3.15) with the local cell length δx and is 'evaluated as an expression ... on quadrature points'; on the unstructured tokamak mesh δx (and hence τ) is discontinuous across cell faces, so the distributional derivative s·∇(τ υ) contains interface contributions that the quadrature evaluation omits. Consequently Proposition 3.1 does not establish consistency of the scheme used for the headline 4% result in Section 4.3. The convergence study of Section 4.1 uses a uniform mesh, where δx (and τ) is constant, so it does not exercise this regime. Please either define τ through a continuous finite-element projection of Eq. (3.15), prove consistency for the discontinuous-τ form including jump terms, or add a numerical consistency test on an unstructured mesh.","section":"Definition 3.1; Stabilization parameter; Proposition 3.1"},{"comment":"The stabilization parameter in Eq. (3.15) is dimensionally inconsistent with the unit analysis given just above it. The authors state that τ must have units s^{1/2}, requiring the denominator to have units s^{-1/2}. The term k√κΔ/δx has units s^{-1/2}, but 2√δt has units s^{1/2} (if δt is a time), so the two terms cannot be added. If the intended formula is (2/√δt + k√κΔ/δx)^{-1}, as the dimensional argument and the remark that τ→0 under temporal refinement suggest, please correct it; as printed the formula is not reproducible and changes the time-step scaling of the stabilization.","section":"Eq. (3.15)"},{"comment":"The method is presented without a stability or coercivity analysis. The 'Ellipticity' paragraph acknowledges that the Petrov-Galerkin operator (3.13) is non-symmetric and possibly non-elliptic, and that 'potential instabilities' may arise if τ is not chosen carefully; the only guidance for τ is the heuristic formula (3.15), motivated by a dimensional argument and a CFD reference. Because the central accuracy claim (including the 4% loss in Section 4.3) depends on this parameter, the paper should provide either a coercivity/inf-sup analysis under explicit conditions on τ or a systematic numerical study (for example, varying τ around (3.15) and varying κ∥/κ⊥ and δt) to demonstrate robustness.","section":"Section 3, 'Ellipticity'"}],"minor_comments":[{"comment":"The text refers to 'our SUPG-modified spatial discretization (3.2)'; this should be (3.8) or Definition 3.2, since Eq. (3.2) is the transport-equation example.","section":"Remark 3.2"},{"comment":"The mesh description '10 × 2kx cells' should be typeset as 10 × 2^{k_x} cells for clarity.","section":"Section 4.1"},{"comment":"There are minor typos: 'Branginskii' should be 'Braginskii' in Appendix A, and 'discontinous' should be 'discontinuous' in Section 2.","section":"Appendix A and Section 2"},{"comment":"In Eq. (3.16b), 'were Trhs corresponds to the strong form' should read 'where Trhs corresponds to the strong form'.","section":"Remark 3.1"},{"comment":"The limiter parameters Tl and σl are introduced without discussion of how their values were chosen; a sentence explaining the choice would aid reproducibility.","section":"Eq. (4.6)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a serious candidate for publication after revision. The main numerical findings are interesting and the benchmark design is careful. The two issues that most concern me are the mismatch between the continuous-τ consistency proof and the discontinuous-τ implementation used for the headline tokamak result, and the dimensional error in Eq. (3.15). I would not reject on the basis of the missing stability analysis alone, but it should be addressed or explicitly scoped as future work."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. The headline result is believable: the SUPG-stabilized mixed CG scheme loses about 4% of the available temperature in the long-run tokamak equilibrium test, versus 32–35% for the standard primal and mixed CG formulations, and the test design earns that claim. Second, there is a real gap between the consistency proof and the implemented scheme, and the paper acknowledges but does not resolve it: tau must be continuous for the bilinear form M^f to be well-defined and for the integration-by-parts in Proposition 3.1 to go through, yet the implementation evaluates tau cell-wise from (3.15) with a local delta-x, so tau is discontinuous across faces on the unstructured tokamak mesh. The proof as written does not cover the scheme that produces the headline 4%.\n\nWhat is genuinely new is the specific combination: the mixed auxiliary-variable formulation with SUPG stabilization on both the temperature and auxiliary equations, with the auxiliary variable in a CG space. The paper positions it honestly as the CG analogue of the authors' earlier DG work, and the consistency proof is clean given standard regularity assumptions. The numerical test design is careful: kappa_perp = 0 means any heat loss is spurious, the DOF counts are matched across methods, and the improvement is consistent across two refinement levels. The convergence study shows the expected order, and the tokamak equilibrium test is a sensible benchmark for fusion-relevant regimes.\n\nSoft spots, in proportion. The tau consistency gap is the main one, but it is not fatal: the numerical evidence stands on its own, and the uniform-mesh convergence test, where tau is constant, is covered by the proof. A referee should ask the authors to close the gap—project tau to a continuous space, analyze the interface jump terms, or state plainly that the proof is for the idealized case. Also missing: stability or error analysis for the non-symmetric Petrov-Galerkin operator, which the paper freely admits, and the tau choice remains heuristic. And no code or data is provided, so the 4% figure is not independently reproducible.\n\nWho it is for: people building or using CG-based fusion MHD codes, and anyone working on anisotropic diffusion discretizations. It deserves a serious referee. My recommendation: send it to review, with the tau continuity question as the main thing to resolve.","headline":"The SUPG-mixed CG scheme is a genuine, well-tested improvement for anisotropic heat transport in fusion — the 4%-vs-32% result is credible — but the consistency proof does not cover the implemented scheme because tau is cell-wise discontinuous.","tokens_in":20598,"tokens_out":5947,"would_cite":true,"duration_ms":47129,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65M60","65N30","65N12","76X05"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper presents a stabilized continuous Galerkin mixed scheme for anisotropic heat conduction that cuts spurious tokamak heat loss to about 4%, versus 35% and 32% for standard CG formulations.","keywords":["anisotropic heat conduction","continuous Galerkin","SUPG stabilization","mixed formulation","magnetic confinement fusion","tokamak simulation","spurious cross-field diffusion","directional derivative auxiliary variable"],"falsifier":"Run the tokamak equilibrium test at the same resolution and time step but with a substantially larger parallel conductivity (say, 10 or 100 times) and check whether the relative available temperature at the final time stays near 0.96 or drops appreciably; a clear drop would indicate that the $\\tau$ formula, rather than the structural SUPG improvement, is carrying the result. A second check is to run the same test with $\\tau$ forced to a very small value, since the scheme then reduces to the standard mixed method (2.9), which loses about 32%.","tokens_in":19405,"feed_emoji":"🔥","tokens_out":10915,"duration_ms":83373,"temperature":0.7,"pith_summary":"The paper claims that a continuous Galerkin (CG) discretization of the anisotropic heat conduction equation can be made nearly as accurate as a discontinuous Galerkin one by adding streamline upwind Petrov-Galerkin (SUPG) stabilization to a mixed formulation. The mixed formulation introduces an auxiliary field for the temperature's directional derivative along the magnetic field, and both the temperature and auxiliary equations receive SUPG stabilization. The authors prove that their discrete scheme is consistent with the strong form of the heat equation, and in a long-run tokamak equilibrium test the stabilized scheme loses only about 4% of the available internal energy to spurious cross-field heat transport, compared with 35% and 32% for the standard primal and mixed CG formulations at matched resolution. The payoff is a practical way to keep the favorable implementation properties of continuous finite elements while controlling the dominant numerical error in strongly anisotropic fusion simulations.","feed_headline":"SUPG-stabilized CG discretization cuts tokamak heat loss to 4%","feed_subtitle":"A continuous-Galerkin mixed scheme with streamline-upwind stabilization retains ~96% of stored energy in a long tokamak run.","key_machinery":"The workhorse is the SUPG-modified mixed CG scheme (3.8): the temperature $T_h$ and the auxiliary field $\\zeta_h$, a scaled directional derivative $\\sqrt{\\kappa_\\Delta}\\,b\\cdot\\nabla T$, both live in continuous Galerkin spaces, and test functions are modified to $\\gamma + \\tau s\\cdot\\nabla\\gamma$ with $s=\\sqrt{\\kappa_\\Delta}\\,b$. The SUPG-modified bilinear forms $M^a$, $M^f$, and $G^f_\\parallel$ carry the directional derivative and, after the auxiliary variable is eliminated, the anisotropic term has the form $(G^f_\\parallel)^T (M^f)^{-1}(G^f_\\parallel + G^f_{\\parallel,b})$, mirroring a discrete diffusion operator. The stabilization parameter $\\tau = (2/\\sqrt{\\delta t} + k\\sqrt{\\kappa_\\Delta}/\\delta x)^{-1}$ is chosen so the modification is nondimensional and vanishes in the space-time refined limit, in which the scheme reduces to the standard mixed method.","core_discovery":"The central claim is that applying SUPG-type stabilization to both equations of the mixed formulation produces a consistent spatial discretization whose discrete anisotropic operator keeps the same transpose-gradient / mass-inverse / gradient structure as the underlying parabolic operator, up to consistency-related modifications, and that preserving this structure is what keeps the directional derivative along the magnetic field accurately represented. With temperature in a continuous Galerkin space of degree k and the auxiliary variable also in a continuous space, the scheme is proven consistent with the strong solution (Proposition 3.1). In the full-torus tokamak equilibrium sustainment test at second order and matched resolution, the stabilized scheme keeps about 96% of the available temperature, while the primal and standard mixed CG formulations keep about 65% and 68%, respectively; at lower resolution the gap is even larger, with the stabilized scheme retaining roughly 84-86% versus roughly 22-29%.","pith_inferences":["Editorial inference: the same SUPG-modified mixed structure could be applied to other strongly anisotropic elliptic operators in plasma models, such as resistive diffusion or separate ion and electron energy equations, wherever a directional derivative must be represented accurately on meshes not aligned with the magnetic field.","Editorial inference: the stabilization parameter $\\tau$ is the main tuning knob, and a natural testable extension is the paper's own suggestion to replace the local cell length $\\delta x$ by the average length of the magnetic field line segment through each cell; one could check whether that makes the scheme less sensitive to mesh anisotropy.","Editorial inference: because the method loses symmetry and ellipticity, robust use in production codes will likely require solvers specialized for non-symmetric transport-dominated systems rather than standard symmetric multigrid, which the paper flags as future work.","Editorial inference: since the method reduces to the standard mixed scheme as $\\tau\\to 0$, an adaptive choice of $\\tau$ driven by a local error indicator could be tested to see whether the reported gains survive at even coarser grids or larger time steps."],"forward_implications":["Existing CG-based MHD codes can adopt the scheme with only a test-space modification and one extra continuous auxiliary field, without switching to discontinuous Galerkin degrees of freedom or interior penalty terms.","At the coarse resolutions typical of resistive MHD simulations run on dissipative time scales, the scheme keeps spurious thermal energy loss at a few percent rather than tens of percent.","The accuracy advantage persists under temperature-dependent parallel conductivity, where standard formulations are amplified by a self-reinforcing cross-diffusion cycle; with the conductivity limiter removed, the standard formulations deteriorate far more than the SUPG one.","A fourth-order primal CG run at the same degree-of-freedom budget still loses about 10% of the available temperature, five times more than the lowest-order SUPG scheme in the same tokamak test.","In the flux-tube perturbation test, the stabilized scheme keeps the temperature perturbation between the expected bounding flux surfaces, while the standard mixed formulation leaks heat across them."],"supporting_citations":[{"why":"The DG-upwind predecessor whose upwind-stabilized mixed structure this paper adapts to continuous Galerkin spaces; it supplies the design target and the symmetric-positive-definite elimination structure.","marker":"[38]"},{"why":"Introduces the mixed formulation with an auxiliary directional-derivative variable that the present scheme stabilizes; it is also one of the two baseline methods compared in the tests.","marker":"[19]"},{"why":"The original SUPG method for convection-dominated flows, which supplies the Petrov-Galerkin test-space modification used in the new discretization.","marker":"[8]"},{"why":"The collisional transport closure giving $\\kappa_\\parallel \\propto T^{5/2}$, used to set the temperature-dependent parallel conductivity in the tokamak equilibrium test.","marker":"[6]"},{"why":"The equilibrium solver that generates the initial MHD equilibrium and magnetic field data for the tokamak test cases.","marker":"[28]"},{"why":"The flux-limiting argument used to justify the capped parallel conductivity model in (4.6).","marker":"[3]"}],"fun_headline_variants":["SUPG-stabilized CG reduces tokamak heat loss from 35% to 4%","Mixed CG with SUPG stabilization cuts spurious loss to 4%","New SUPG-CG scheme retains 96% of tokamak stored energy","SUPG-stabilized CG keeps 96% energy in long tokamak run","Anisotropic heat flux tamed by SUPG-stabilized continuous Galerkin"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The scheme's accuracy rests on the heuristic formula for the stabilization parameter $\\tau$ in (3.15); the paper proves consistency but not stability or ellipticity, and warns that an uncareful choice of $\\tau$ can introduce instabilities, so if (3.15) is not robust across the conductivity and time-step regimes used in fusion simulations, the reported 4% heat loss may not generalize.","fun_headline_variants_meta":{"raw":{"variants":["SUPG-stabilized CG reduces tokamak heat loss from 35% to 4%","Mixed CG with SUPG stabilization cuts spurious loss to 4%","New SUPG-CG scheme retains 96% of tokamak stored energy","SUPG-stabilized CG keeps 96% energy in long tokamak run","Anisotropic heat flux tamed by SUPG-stabilized continuous Galerkin"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000854,"raw_usage":{"total_tokens":3721,"prompt_tokens":967,"completion_tokens":2754,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":583,"completion_tokens_details":{"reasoning_tokens":2658}},"tokens_in":583,"tokens_out":2754,"duration_ms":17855,"temperature":1.0,"reasoning_tokens":2658,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T14:07:36.037358+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the tokamak equilibrium test at the same resolution and time step but with a substantially larger parallel conductivity (say, 10 or 100 times) and check whether the relative available temperature at the final time stays near 0.96 or drops appreciably; a clear drop would indicate that the $\\tau$ formula, rather than the structural SUPG improvement, is carrying the result. A second check is to run the same test with $\\tau$ forced to a very small value, since the scheme then reduces to the standard mixed method (2.9), which loses about 32%.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The DG-upwind predecessor whose upwind-stabilized mixed structure this paper adapts to continuous Galerkin spaces; it supplies the design target and the symmetric-positive-definite elimination structure."},{"cited_title":"G¨ unter, Q","cited_arxiv_id":null,"evidence_quote":"Introduces the mixed formulation with an auxiliary directional-derivative variable that the present scheme stabilizes; it is also one of the two baseline methods compared in the tests."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The original SUPG method for convection-dominated flows, which supplies the Petrov-Galerkin test-space modification used in the new discretization."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The collisional transport closure giving $\\kappa_\\parallel \\propto T^{5/2}$, used to set the temperature-dependent parallel conductivity in the tokamak equilibrium test."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The equilibrium solver that generates the initial MHD equilibrium and magnetic field data for the tokamak test cases."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The flux-limiting argument used to justify the capped parallel conductivity model in (4.6)."}],"review_version":1}