{"id":"0123bf52-8ced-4112-9be3-81ad111d05ad","arxiv_id":"2412.14962","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"ThinCurr is an open-source boundary-element code that models thin-wall eddy currents with HODLR-compressed inductance matrices and automatic homology-based hole detection, verified against VALEN and Ansys.","lead":"ThinCurr is a new open-source computer code that models electric currents induced in thin metal walls by changing magnetic fields, as in fusion reactors. It uses a fast approximation that lets designers simulate million-element structures on a single workstation, and the authors show it agrees with established codes.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Thin-wall approximation is the load-bearing risk: the only volumetric-code benchmark (SPARC, Sec. 5.2) uses a 1.385 ms quench where skin depth (~3 cm) is comparable to the 2–4 cm walls, so 'excellent agreement' may not generalize to faster quenches or thicker walls.","rationale":"The reader's weakest-assumption analysis identified the thin-wall approximation as the load-bearing risk, and I agree. The numerical method itself is well verified: eigenvalue error converges at the expected O(Δx^2) rate, and time/frequency domain results match VALEN closely; the HODLR scaling studies are also internally consistent. But all of these tests except the SPARC Ansys case compare ThinCurr to another thin-wall model, and the single Ansys comparison sits close to the skin-depth/thickness crossover. A parameter sweep around that point would settle whether the headline 'excellent agreement' is robust or event-specific. This supports keeping the reader's CONDITIONAL verdict rather than accepting outright; it also points to a concrete, inexpensive benchmark the authors can add. The secondary caveat about linked holes in Sec. 2.1.3 is real but less central to the main claim.","tokens_in":15761,"tokens_out":12764,"duration_ms":115814,"concrete_test":"Re-run the Sec. 5.2 SPARC benchmark in both ThinCurr and Ansys for the same 2–4 cm wall model with quench decay times τ = 0.35, 1.385, and 5 ms (δ ≈ 1.4, 2.8, and 5.3 cm for typical stainless steel), and tabulate peak toroidal current and peak radial/vertical loads. If the ThinCurr/Ansys discrepancy stays below ~10% across all three τ, the concern is resolved; if it grows as τ decreases, the thin-wall approximation is regime-limited and the conclusions must be scoped accordingly.","verdict_should_be":"UNCHANGED","load_bearing_attack":"ThinCurr's central claim is that the thin-wall BFEM correctly predicts eddy-current loads in fusion devices. The VALEN comparisons (Sec. 5.1) verify the numerical implementation, but VALEN is also a thin-wall code, so they do not test the modeling assumption. The only test against a volumetric solver is the SPARC Ansys benchmark (Sec. 5.2), which is a single point in parameter space: 2–4 cm walls and an 8.7 MA quench with exponential decay time 1.385 ms. For typical stainless steel, the diffusion skin depth δ ≈ sqrt(τ/(μσ)) is ~3 cm, the same order as the wall thickness. Through-thickness current variation is therefore not negligible a priori, and agreement of total current/forces for one event does not establish validity for faster quenches, thicker sections, or the 3D current paths around ports/openings that motivate the code. If the approximation fails, ThinCurr's predictions are wrong regardless of HODLR accuracy or implementation quality.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents ThinCurr, an open-source boundary finite element code for thin-wall eddy current modeling in fusion devices, developed within the Open FUSION Toolkit. The authors describe the mathematical formulation using a surface current potential, the construction of inductance and resistance matrices, the treatment of multiply connected geometries via automatic homology-based hole and closure detection, and three solution modes (time domain, frequency domain, and eigenvalue). A central contribution is the use of hierarchical off-diagonal low-rank (HODLR) compression with ACA+ to replace the dense inductance matrix, with scaling tests showing O(N log N) memory and solution time. Verification is performed by cross-code comparisons against VALEN (eigenvalue, time-domain, and frequency-domain tests) and against Ansys for a SPARC tokamak disruption quench. The paper concludes that the numerical implementation is correct and that the HODLR compression enables scalability to whole-device models, with all comparisons showing excellent agreement.","tokens_in":15969,"tokens_out":5410,"duration_ms":44843,"significance":"If the claims hold, ThinCurr is a useful, freely available tool for large-scale eddy current load calculations in fusion device design, and the paper provides a detailed and largely clear description of the numerical methods. The eigenvalue verification is a strength: the relative error between ThinCurr and VALEN decreases at the expected O(dx^2) rate across 100 modes and three geometries, indicating that the dense-matrix L and R assembly is correct. The scaling measurements for HODLR are also valuable and provide concrete performance data. The automatic homology-based hole detection is a practical usability contribution. However, two gaps currently limit the strength of the central claims: the HODLR/ACA+ path is not validated for accuracy, and the only volumetric-solver benchmark (Ansys) lacks quantitative error metrics and covers only a single operating point. These issues are fixable and do not appear to undermine the underlying method, but they need to be addressed before the paper can fully support its claims.","major_comments":[{"comment":"The HODLR/ACA+ compression is characterized only in terms of memory and time scaling; no test demonstrates that the compressed inductance matrix or the resulting eigenvalues, time traces, or forces match the dense solution to the specified tolerance. A user of the code needs to know that the compressed solution is accurate, not merely fast. Please add a convergence study on a moderately sized model (e.g., ~10^4 elements) comparing dense and HODLR results for a few eigenvalues and a representative time-domain response, with errors reported as a function of the SVD/ACA+ tolerance. Without this, the claim that HODLR 'enables scalability to whole device models' is a performance claim without a corresponding correctness verification.","section":"Sec. 4, Figs. 8-10"},{"comment":"The Ansys benchmark is the only test against a volumetric solver, yet the paper reports no quantitative error metric for the toroidal current and force traces, relying instead on the statement 'excellent agreement.' Moreover, the single 8.7 MA quench has a 1.385 ms exponential decay time, for which the diffusion skin depth in the 2-4 cm thick walls can be comparable to the wall thickness. The paper should report the skin depth-to-wall thickness ratio for this case and discuss the expected range of validity of the thin-wall approximation for faster quenches or thicker sections. Adding a second Ansys comparison with a faster decay time (or otherwise different skin-depth regime) would substantially strengthen the claim that the approximation is appropriate for design-limiting events.","section":"Sec. 5.2, Fig. 18"},{"comment":"The time-domain and frequency-domain comparisons against VALEN are presented only as overlaid plots with the qualitative statement 'excellent agreement.' The verification claim would be falsifiable and more convincing if the authors reported quantitative measures, such as the maximum or RMS relative error over the plotted sensor signals, or a normalized error norm. The eigenvalue comparison already has such metrics; the time- and frequency-domain tests should be held to the same standard.","section":"Secs. 5.1.2 and 5.1.3"}],"minor_comments":[{"comment":"The acronym 'HODLR' is introduced as 'HOLDR' in the first paragraph of Section 4 ('...utilizes an Hierarchical Off-Diagonal Low-Rank (HOLDR) approximation...'); the correct acronym is used elsewhere and should be made consistent.","section":"Sec. 4"},{"comment":"In the text describing the frequency-domain test, 'ThinCurr used the preconditioned GRMES approach' should be 'GMRES'.","section":"Sec. 5.1.3"},{"comment":"The text reads 'Adaptive Cross-Appoximation+ (ACA+)'; the word 'Approximation' is misspelled.","section":"Sec. 4.1"},{"comment":"The sentence 'Although not plotted, a comparison of the local current densities at the midplane of the inner and outer VVs also shows similar agreement' makes an unverifiable claim; either add the plot or remove the statement.","section":"Sec. 5.2"},{"comment":"The adaptive quadrature order formula p = log(err)/log(1 - dmin/dmax) would benefit from a brief definition of 'err' (target error) and a note on the expected range of p, since the denominator is negative and the intended behavior may not be immediately clear to readers.","section":"Sec. 2.1.1, Eq. (6)"}],"recommendation":"major_revision","confidential_remarks":"The paper fits the journal's scope well as a computational physics code description. The citation pattern is appropriate; the authors cite their own OFT framework where relevant but not excessively. The thin-wall approximation is acknowledged as an assumption with stated limitations, so the main risk is not circularity but incomplete validation of the two load-bearing features: HODLR accuracy and the volumetric benchmark. Both are addressable with additional experiments and reporting, so I do not recommend rejection. The lack of quantitative error metrics in the Ansys and time/frequency comparisons is the most straightforward fix. I would suggest the editor ask for the additional validation before acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The key thing to know: this is a solid, honestly-reported software paper with genuinely new integration work, and the verification against VALEN and Ansys is the real deal. The automatic greedy homology-based hole/closure detection and the HODLR/ACA+ compression for arbitrary 3D thin-wall models are not just repackaged; they are implemented, open-sourced, and tested on problems up to ~10^5-10^6 elements with N log N scaling shown.\n\nThe eigenvalue verification is the strongest part: 100 modes, three geometries, and the error between ThinCurr and VALEN drops at the expected O(dx^2) rate. The Ansys SPARC benchmark adds a genuine volumetric-solver check on a full-device model, and the agreement on total current and forces is convincing. No free parameters, no fitting to make agreement happen. The self-citations to Open FUSION Toolkit are infrastructure references, not circular validation.\n\nSoft spots are real but not disqualifying. The time- and frequency-domain comparisons are shown as overlaid curves with 'excellent agreement' but no quantitative error metric; that should be an easy fix. The topology method has an explicitly stated gap: the text says 'we do not pursue proof or contradiction of such generality' for linked holes, which is honest but means the automatic hole detection isn't fully guaranteed for all geometries users might throw at it. The bigger physical risk is the thin-wall approximation itself, and the paper acknowledges it: thick-wall effects 'may be important in some configurations and events.' The Ansys benchmark uses a 1.385 ms quench with skin depth on the order of the 2-4 cm walls, so the single data point does not bound the error for faster quenches or thicker sections. I don't think this undercuts the paper — the approximation is standard for this class of tools — but the claims in the abstract and Sec. 6 about 'excellent agreement' should be scoped to the tested regime.\n\nMy take: it deserves a serious referee and should be published after the authors add a few numbers and a sentence of caution about thin-wall validity. I'd bring it to a fusion modeling reading group and cite it as the reference for ThinCurr.","headline":"Solid, honest software/methods paper for thin-wall eddy current modeling; the new integration and open-source HODLR implementation are verified well, though the thin-wall approximation is only tested at one point in parameter space.","tokens_in":16518,"tokens_out":1588,"would_cite":true,"duration_ms":13950,"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":"ThinCurr is an open-source boundary finite element code for 3D thin-wall eddy currents that compresses the dense inductance matrix to scale to whole-device tokamak models while matching established verification benchmarks.","keywords":["thin-wall approximation","eddy currents","boundary finite element method","hierarchical off-diagonal low-rank compression","adaptive cross approximation","tokamak disruption loads","current potential","open-source plasma modeling"],"falsifier":"Take a wall geometry like the paper's benchmark, with walls 2 to 4 cm thick and an 8.7 MA current quench whose exponential decay time is 1.385 ms, and compare ThinCurr's total currents and forces against a reference model that resolves the wall thickness across the same transient; if the disagreement exceeds the few-percent level the paper reports, the thin-wall approximation—not the solver implementation—is the failing link.","tokens_in":15573,"feed_emoji":"⚡","tokens_out":11366,"duration_ms":85501,"temperature":0.7,"pith_summary":"ThinCurr is an open-source solver for inductively coupled eddy currents in thin conducting structures, aimed at fusion devices where vacuum-vessel walls and other sheet-like conductors carry large transient currents during disruptions. The paper claims that a boundary finite element method with a surface current potential correctly captures these currents, and that hierarchical off-diagonal low-rank compression of the otherwise dense inductance matrix makes full-device models practical by cutting memory and setup cost from quadratic to nearly linear scaling. The implementation is verified through cross-code comparisons with an established thin-wall research code and a commercial finite-element package, including a tokamak disruption benchmark, with close agreement in currents, forces, and time scales. A reader interested in tokamak structural design should care because this is a freely available route to eddy-current loads on realistic 3D geometry.","feed_headline":"Thin-wall eddy current code scales to full fusion device models","feed_subtitle":"Open-source boundary-element solver reproduces benchmark currents and forces while making matrix cost scale nearly linearly.","key_machinery":"The central object is the boundary finite element formulation on surfaces: the current is written as $\\mathbf{J}_s = \\nabla\\chi\\times\\hat{\\mathbf n}$, the inductance matrix $L$ is assembled from double surface integrals of the Biot-Savart kernel, and the resistance matrix $R$ uses surface resistivity $\\eta_s = \\eta/t_w$, where $t_w$ is wall thickness. Topological correctness is carried by \"hole\" elements, which represent current loops around distinct homological cycles, and \"closure\" elements, which fix the gauge on closed surfaces; these are located automatically by a greedy homology-basis algorithm. The scalability claim is carried by the HODLR/ACA+ compression of $L$, supported by block-Jacobi preconditioning so that iterative solves retain nearly linear scaling.","core_discovery":"The central claim is that representing each conducting structure as an oriented surface with a scalar current potential $\\chi$, so that the surface current is $\\mathbf{J}_s = \\nabla\\chi\\times\\hat{\\mathbf n}$, and discretizing that potential with linear finite elements on an unstructured triangular mesh yields a correct and practical model of eddy currents in large-scale systems. The dense inductance matrix $L$ that couples every surface element is compressed with a hierarchical off-diagonal low-rank approximation whose far blocks are built by adaptive cross approximation, restoring nearly linear scaling in memory and time. The paper reports that eigenvalue, time-domain, and frequency-domain results match a community thin-wall code, and that disruption-induced currents and forces computed for a full tokamak vacuum-vessel model match a commercial volume-conducting finite-element solver, with all comparisons showing close agreement.","pith_inferences":["I infer that users should check skin depth before applying the code: whenever the transient's skin depth is comparable to wall thickness, the thin-wall representation is the likely error source even though the numerical method is sound.","The compression tolerance gives a tunable accuracy-versus-speed knob, so a natural extension is to verify convergence of integrated quantities such as total force, not just eigenvalue spectra, as the tolerance is tightened.","The method is geometry-general, so the same open-source machinery could serve eddy-current studies outside fusion, such as induced loads in accelerator vacuum chambers, electromagnetic forming, shielding design, or nondestructive evaluation of thin sheet structures.","Because the code is open source, community additions such as higher-order elements, curved triangles, and source/sink support at T-junctions are plausible next steps that would expand the class of representable geometries."],"forward_implications":["Models with on the order of one hundred thousand elements fit in memory and solve in nearly linear time, bringing whole-device disruption load calculations onto workstation-class hardware.","Automatic detection of holes and closures lets users build solvable models directly from triangular meshes without manually specifying topological loops.","The same model supports time-domain, frequency-domain, and eigenvalue analyses, so design cycles can move from decay-time scoping to full transient loads without rebuilding the mesh.","Agreement with a commercial finite-element solver on a realistic tokamak disruption case supports using ThinCurr for design-cycle estimates of eddy-current forces on thin-walled structures.","Integration into workflows for resistive-wall mode stability, current reconstruction, and coil optimization is listed as ongoing work in the paper."],"supporting_citations":[{"why":"The community thin-wall code that provides the primary cross-code baseline for eigenvalue, time-domain, and frequency-domain verification.","marker":"[8]"},{"why":"The commercial finite-element software whose axisymmetric tokamak vacuum-vessel model provides the disruption current and force benchmark.","marker":"[9]"},{"why":"Supplies the adaptive cross approximation (ACA+) used to build low-rank far blocks of the HODLR inductance matrix, enabling nearly linear scaling.","marker":"[42]"},{"why":"The greedy homology-basis algorithm that the automatic hole-detection procedure adapts for finding supplemental elements on complicated surfaces.","marker":"[21]"},{"why":"Provides the compact tokamak device description used to construct the commercial finite-element benchmark model.","marker":"[44]"},{"why":"Supplies the 8.7 MA disruptive current quench with 1.385 ms decay time used in the tokamak time-dependent benchmark.","marker":"[47]"},{"why":"Gives the analytic form of the singular inner integral over a triangle used in assembling the inductance matrix.","marker":"[17]"},{"why":"Provides the triangular solid-angle formulas used in the analytic treatment of the 1/r singularity in the inductance integral.","marker":"[18]"}],"fun_headline_variants":["ThinCurr: open-source eddy current solver for large-scale fusion","Open-source thin-wall eddy current code for fusion plasma devices","New open-source code models eddy currents in large conducting structures","Scalable eddy current modeling for fusion devices goes open source","ThinCurr: fast eddy current modeling for large-scale conducting systems"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the thin-wall approximation: each structure is treated as a surface whose only thickness information is the surface resistivity, so currents flowing through the wall thickness are not resolved; if the electromagnetic skin depth is comparable to or smaller than the wall thickness during a transient, the model misses through-thickness current structure regardless of how accurately the matrices are built.","fun_headline_variants_meta":{"raw":{"variants":["ThinCurr: open-source eddy current solver for large-scale fusion","Open-source thin-wall eddy current code for fusion plasma devices","New open-source code models eddy currents in large conducting structures","Scalable eddy current modeling for fusion devices goes open source","ThinCurr: fast eddy current modeling for large-scale conducting systems"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000309,"raw_usage":{"total_tokens":1741,"prompt_tokens":902,"completion_tokens":839,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":518,"completion_tokens_details":{"reasoning_tokens":747}},"tokens_in":518,"tokens_out":839,"duration_ms":5421,"temperature":1.0,"reasoning_tokens":747,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T11:45:01.682191+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a wall geometry like the paper's benchmark, with walls 2 to 4 cm thick and an 8.7 MA current quench whose exponential decay time is 1.385 ms, and compare ThinCurr's total currents and forces against a reference model that resolves the wall thickness across the same transient; if the disagreement exceeds the few-percent level the paper reports, the thin-wall approximation—not the solver implementation—is the failing link.","supporting_citations":[{"cited_title":"URL https://www.ansys.com/","cited_arxiv_id":null,"evidence_quote":"The commercial finite-element software whose axisymmetric tokamak vacuum-vessel model provides the disruption current and force benchmark."},{"cited_title":"Bebendorf, S","cited_arxiv_id":null,"evidence_quote":"Supplies the adaptive cross approximation (ACA+) used to build low-rank far blocks of the HODLR inductance matrix, enabling nearly linear scaling."},{"cited_title":"Erickson, K","cited_arxiv_id":null,"evidence_quote":"The greedy homology-basis algorithm that the automatic hole-detection procedure adapts for finding supplemental elements on complicated surfaces."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the compact tokamak device description used to construct the commercial finite-element benchmark model."},{"cited_title":"Sweeney, A","cited_arxiv_id":null,"evidence_quote":"Supplies the 8.7 MA disruptive current quench with 1.385 ms decay time used in the tokamak time-dependent benchmark."},{"cited_title":"Ferguson, X","cited_arxiv_id":null,"evidence_quote":"Gives the analytic form of the singular inner integral over a triangle used in assembling the inductance matrix."},{"cited_title":"Van Oosterom, J","cited_arxiv_id":null,"evidence_quote":"Provides the triangular solid-angle formulas used in the analytic treatment of the 1/r singularity in the inductance integral."}],"review_version":1}