{"id":"dd6f8f6c-f521-4a9e-9036-3a0236aa3325","arxiv_id":"2412.05680","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"The DESC code can now compute free boundary 3D MHD equilibria with high order accuracy, with or without sheet currents.","lead":"This paper adds free boundary calculation capability to the DESC stellarator equilibrium code, using a high order method for the singular boundary integrals involved. The work matters because free boundary equilibria are needed for realistic stellarator design and for single stage coil and plasma optimization.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Quadrature convergence order is unverified in DESC's Fourier basis, so the high-order claim and VMEC superiority rest on an untested assumption.","rationale":"The reader's weakest_assumption is exactly the load-bearing concern I identify: the accuracy of the singular quadrature in DESC's Fourier setting is unverified. I agree that this is the most critical gap. The paper's vacuum benchmark against field-line tracing is strong evidence that the method works qualitatively, but it does not measure convergence order, and the finite-beta comparison with VMEC is self-referential because the same quadrature is used to evaluate both residuals. I considered whether the singular integral should be evaluated as an exterior limit rather than a principal value, since Eq. 4.4 uses Bout as the outside field; the paper mentions principal value but does not discuss the jump correction. However, the vacuum benchmarks suggest the implementation is behaving correctly, so I do not elevate this to a separate concern. The reader's conditional accept is appropriate: the method is plausible and the code is available, but the central accuracy claim needs a direct convergence test. No verdict change is needed; the concern strengthens the conditionality already expressed by the reader.","tokens_in":7421,"tokens_out":30219,"duration_ms":269840,"concrete_test":"Implement a standalone test of the singular Biot-Savart integral in Eq. 4.3 on a Fourier-parameterized toroidal surface with an analytic field (e.g., a known surface current distribution producing a uniform field or a dipole). Compute the error between the numerical integral and the analytic exterior field for increasing grid resolutions N (e.g., N=16,32,64,128) and fit the observed convergence order. If the order is 10th-12th, the high-order claim is supported; if it is lower (e.g., 2nd), the residual objective is corrupted and the VMEC comparison in Section 6 should be re-evaluated. Repeat with a realistic DESC boundary from the Landreman-Paul QA equilibrium to confirm the interpolation is not degrading performance.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim of high-order accuracy rests entirely on the partition-of-unity singular quadrature (Malhotra et al., 10th-12th order) as adapted to DESC's Fourier boundary representation. The adaptation replaces the original 12th-order Lagrange interpolation with Fourier interpolation on a shifted grid, and the authors state this makes the interpolation spectral. However, no convergence study of the resulting singular integral is presented. Figure 3 shows the free-boundary residual decreasing as spectral resolution increases, but that conflates the equilibrium solve, the boundary shape representation, and the sheet-current potential with the quadrature itself. The claimed 2-3x lower residual for DESC over VMEC in Section 6 is computed with this same untested quadrature, so if the quadrature is not actually high-order, that comparison does not establish superior accuracy. Furthermore, some benchmarks use MGRID external-field interpolation that the paper itself states is only first-order accurate, further obscuring the order of the overall method. The central load-bearing condition is thus that the adapted quadrature retains its high order on DESC's Fourier surfaces; this is neither proven nor independently tested.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper extends the DESC fixed-boundary equilibrium code to free-boundary problems. The exterior field is represented as the sum of the coil field, a virtual-casing field obtained from the fixed-boundary solution, and a surface-current field parameterized by a potential Φ; the free-boundary conditions (normal-field balance, pressure balance, and sheet-current relation) are then enforced by minimizing a residual objective in the boundary shape and Φ. Singular Biot–Savart integrals are evaluated with a partition-of-unity polar quadrature adapted from Malhotra et al., using Fourier interpolation on DESC's boundary grid. Results are shown for a vacuum quasi-axisymmetric equilibrium compared with direct field-line tracing, a finite-edge-pressure helical stellarator with sheet current and residual convergence, and a W7-X-like β=2% case compared with free-boundary VMEC. The paper concludes that DESC's high-order method is more accurate than VMEC.","tokens_in":7602,"tokens_out":4339,"duration_ms":43167,"significance":"If the accuracy claims are substantiated, this is a valuable capability: it avoids an explicit exterior Neumann solve and the associated large linear system inversion, handles finite edge pressure and sheet currents naturally, and is structured to fit into single-stage plasma–coil optimization. The open-source implementation and the effort to benchmark against field-line tracing and VMEC are strengths. However, the central high-order claim is not yet directly supported: the paper validates the coupled free-boundary solver, not the singular quadrature in its new Fourier-interpolated form, and the VMEC comparison uses the very method whose accuracy is at issue. The significance is therefore contingent on the additional verification requested below.","major_comments":[{"comment":"The central claim of high-order accuracy rests on the partition-of-unity singular quadrature retaining its 10th–12th order behavior when the original Lagrange interpolation is replaced by Fourier interpolation on DESC's boundary grid. The paper states that this interpolation is spectrally accurate and that derived quantities need not be recomputed, but it provides no convergence study of the singular integral itself. Figure 3 shows the full free-boundary residual decreasing with spectral resolution, which conflates the boundary representation, the equilibrium solve, and the quadrature behavior. I request a direct test, for example using a manufactured toroidal surface with a known Biot–Savart integral or a known harmonic field, reporting the error in Eq. (4.3) as a function of grid resolution and quadrature order, to establish the order of the adapted scheme.","section":"§5, Eqs. (5.1)–(5.6)"},{"comment":"The reported 2–3x lower residual for DESC over VMEC is computed with the same high-order boundary integral method that is the subject of the paper. If the quadrature has a systematic bias that favors DESC's Fourier boundary representation, this comparison would not establish superior physical accuracy. An independent error metric is needed, such as field-line tracing error, a known analytic free-boundary equilibrium, or residual evaluation with an independent method such as NESTOR. As written, the superiority claim is not independently corroborated.","section":"§6, Figure 4 and final paragraph"},{"comment":"The paper itself states that MGRID coil-field interpolation is only first-order accurate in the grid spacing. The helical stellarator example with sheet current (Figures 2 and 3) uses MGRID for the external field. This is inconsistent with the high-order claim unless the MGRID grid is fine enough that its interpolation error is negligible; no grid-resolution check for MGRID is shown. The vacuum QA benchmark uses direct filament Biot–Savart and is more convincing, but the sheet-current convergence demonstration should either use an accurate coil field or demonstrate that the MGRID error is subdominant at the reported resolutions.","section":"§5 and §6, MGRID-based examples"}],"minor_comments":[{"comment":"The displayed definition of ρ appears malformed: the factor s should presumably divide the normalized distance rather than multiply it. Please correct the equation or clarify the notation.","section":"Eq. (5.5)"},{"comment":"The entry 'Drevlak & Lobsien 2022' is cited as 'private communication'; this should be replaced with a citable document, a preprint, or removed from the formal reference list.","section":"References"},{"comment":"The vertical axis is labeled only as 'residual'; please specify the exact norm used (e.g., L2 norm of the boundary-condition vector), the normalization, and the discretization parameters held fixed while spectral resolution is increased.","section":"Figure 3"},{"comment":"The statement that 'results do not seem to depend on the choice of initialization' is anecdotal as written; a small scan over initial conditions or a more cautious phrasing would make the claim reproducible.","section":"§6, initialization statement"},{"comment":"The word 'efficient' in the abstract and conclusion is not supported by any timing or complexity data. A short runtime or cost comparison with VMEC/NESTOR for one of the benchmark cases would substantiate this claim.","section":"Abstract and §7"}],"recommendation":"major_revision","confidential_remarks":"The manuscript fits the scope of J. Plasma Phys. and the core idea is promising. The main revision should focus on independent verification of the singular quadrature in its Fourier-interpolated form and on an unbiased accuracy comparison with VMEC. I did not find evidence of citation manipulation or overlapping publication concerns. The recommendation is major revision rather than rejection because the requested tests are within the scope of the manuscript and the central approach appears sound."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this if you work on stellarator equilibria. It delivers the first free-boundary capability in DESC, built on a partition-of-unity singular quadrature, and the benchmarks are honestly described and mostly convincing. The vacuum QA case is validated by direct field-line tracing; the finite-beta sheet-current case shows residuals converging as resolution increases. That is a real capability the DESC ecosystem did not have.\n\nWhat's new: casting the free-boundary conditions as a minimization over boundary shape and sheet-current potential, avoiding the exterior Neumann solve that VMEC/NESTOR does. Not a radically new idea—minimizing boundary-condition residuals is old—but this is the first implementation in DESC and it handles finite edge pressure with sheet currents, which NESTOR doesn't do explicitly. The code is open and the method section is clear.\n\nSoft spots, in order. (1) The high-order claim is under-supported. The paper swaps Malhotra's Lagrange interpolation for Fourier interpolation and calls it spectral, but there is no standalone convergence test of the adapted Biot-Savart integral on DESC's Fourier surfaces. Figure 3's residual curve blends the equilibrium solve, boundary shape, current potential, and quadrature, so it doesn't isolate the quadrature order. The VMEC comparison's 2–3x lower residual is computed with the same unverified quadrature, so it's not independent evidence of superior accuracy. This is a real gap, but it doesn't sink the paper: the field-line-tracing agreement in Figure 1 gives end-to-end confidence that the method produces correct equilibria. (2) The efficiency claim is asserted, not measured—no wall-clock comparison against NESTOR or VMEC. (3) The MGRID interpolation is first-order, which muddies the order of the overall method in the helical benchmark; that's acknowledged in the text, though.\n\nBottom line: a solid, useful capability paper for the DESC user base and for anyone doing single-stage optimization. The mathematics is standard, the citation pattern looks fair, and the missing pieces are numerical evidence, not correctness. A serious referee should ask for a focused convergence test of the singular integral and a runtime comparison; both are doable. I would send it to review.","headline":"First free-boundary DESC capability with a clean residual-minimizing formulation; the high-order quadrature claim is plausible but under-verified, and the paper deserves review with requests for standalone convergence and runtime tests.","tokens_in":8135,"tokens_out":2727,"would_cite":true,"duration_ms":25896,"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":"DESC computes free-boundary ideal MHD equilibria with a high-order singular boundary integral method, avoiding the exterior Neumann problem and achieving lower boundary error than VMEC.","keywords":["free boundary equilibrium","ideal MHD","stellarator equilibrium","singular boundary integrals","partition of unity quadrature","sheet current","virtual casing","DESC"],"falsifier":"Evaluate the boundary integral (4.3) on a surface with a known analytic field (for example a circular torus with a uniform sheet current) and measure the error as the grid is refined; if the observed order is far below the claimed 10th to 12th, the accuracy advantage over VMEC is not established. A complementary check is to reconstruct the same W7-X-like equilibrium with an independent high-accuracy method and compare flux surfaces and boundary residuals to the DESC and VMEC results.","tokens_in":7222,"feed_emoji":"🧲","tokens_out":9356,"duration_ms":77336,"temperature":0.7,"pith_summary":"This paper shows how to compute free-boundary ideal MHD equilibria in the DESC code without solving the exterior Neumann problem. The free-boundary conditions are recast as a residual objective, minimized over the plasma boundary shape and a surface-current potential, and the needed singular Biot-Savart integral over the boundary is evaluated with a partition-of-unity quadrature that is claimed to be 10th to 12th order accurate. Benchmarks against direct field-line tracing and against VMEC free-boundary calculations support the method: a precise quasisymmetric equilibrium is reproduced, a helical stellarator converges in a few iterations, and a finite-$\\beta$ case with a nonzero edge pressure yields a sheet current of about 3.6 kA. For a W7-X-like equilibrium at $\\beta=2\\%$, the DESC boundary shape has roughly two to three times lower boundary-condition residual than the VMEC boundary. If these results hold, DESC gains a free-boundary capability that preserves the accuracy advantage it already has in fixed-boundary calculations.","feed_headline":"DESC out-accuracies VMEC on free-boundary stellarator equilibria","feed_subtitle":"A singular-integral method replaces the exterior Neumann solve and cuts boundary error two to three times.","key_machinery":"The load-bearing object is the partition-of-unity singular quadrature for the two-dimensional Biot-Savart integral (4.3). The integrand is split into a smooth part and a singular part supported near the evaluation point using $\\chi(\\rho)=e^{-36\\rho^8}$; the smooth part is integrated with the trapezoidal rule, which converges exponentially on the periodic domain, and the singular part is integrated in polar coordinates, whose Jacobian $\\rho$ cancels the $1/r$ singularity. DESC's adaptation replaces the original 12th-order Lagrange interpolation with Fourier interpolation, accounting for the shifted grid by a phase factor in the Fourier domain, so the interpolation is spectrally accurate and no metric quantities need be recomputed in polar coordinates. This quadrature supplies the external-field contribution at each optimization step, turning the free-boundary problem into minimization of the residual (4.4) over boundary shape and surface-current potential.","core_discovery":"The central claim is that the free-boundary equilibrium can be found by evaluating one singular boundary integral per iteration, the virtual-casing Biot-Savart integral $$B_{\\rm out}=B_{\\rm coil}+\\frac{\\mu_0}{4\\pi}\\int_{\\mathcal{D}}\\frac{[n\\times(B_{\\rm fixed}+\\nabla\\Phi)]\\times(r-r')}{|r-r'|^3}\\,$d^{2}$r',$$ instead of inverting the large linear operator of the exterior Neumann problem. The boundary conditions $B\\cdot n=0$, $[[p+B^2/2\\mu_0]]=0$, and $n\\times[[B]]=\\mu_0 K$ become the three rows of the residual objective (4.4), minimized over the boundary coefficients $(R_b,Z_b)$ and the current potential $\\Phi$. The paper reports that this reproduces a precise quasisymmetric equilibrium in agreement with field-line tracing, converges an initially circular boundary to a helical stellarator in about four steps, and for a W7-X-like case at $\\beta=2\\%$ gives a boundary whose residual is two to three times smaller than VMEC's, attributed to the higher order boundary integral method. The same formulation also handles the sheet current that must exist when edge pressure is finite, as demonstrated by the $\\sim 3.6$ kA surface current in a finite-$\\beta$ helical stellarator test.","pith_inferences":["If the quadrature really delivers 10th to 12th order accuracy on DESC surfaces, the same singular-integral machinery could also upgrade virtual-casing diagnostics and coil verification, not just equilibrium solves.","The residual-objective formulation invites a direct test: use the high-order integral to score boundary-condition residuals of equilibria produced by other solvers and see whether the error ordering persists across configurations and resolutions.","The explicit sheet-current potential may prove useful for modelling edge-localized current layers in finite-pressure devices, where the tangential-field jump is physically meaningful rather than a numerical artifact."],"forward_implications":["DESC free-boundary runs need only one singular surface integral per evaluation, so the exterior Neumann linear solve is removed from the equilibrium iteration.","Finite edge pressure is represented by a surface-current potential, so equilibria with sheet currents can be computed and the sheet current quantified, as in the $\\sim3.6$ kA helical-stellarator test.","Because the boundary conditions are a minimization objective, the same code path can be extended to single-stage optimization in which plasma boundary and coil degrees of freedom vary together.","The reported two-to-three-fold lower boundary residual than VMEC on the W7-X-like case suggests free-boundary calculations can inherit the accuracy DESC already shows for fixed-boundary equilibria."],"supporting_citations":[{"why":"Supplies the partition-of-unity singular quadrature that the paper adapts and the claimed 10th-12th order accuracy.","marker":"Malhotra et al. (2020)"},{"why":"Earlier high-order boundary integral solver for Taylor states in stellarators that motivates the method.","marker":"Malhotra et al. (2019)"},{"why":"Original three-dimensional free-boundary Green's function calculation that is the VMEC baseline.","marker":"Hirshman et al. (1986)"},{"why":"The NESTOR integral-equation technique for the exterior Neumann problem, the 2nd-order method being replaced.","marker":"Merkel (1986)"},{"why":"Proves the polar-coordinate quadrature yields the correct principal value for hypersingular kernels like Biot-Savart.","marker":"Ying et al. (2006)"},{"why":"Provides the precise QA equilibrium used as the vacuum benchmark.","marker":"Landreman & Paul (2022)"},{"why":"Shows DESC's fixed-boundary accuracy advantage over VMEC, motivating the extension.","marker":"Panici et al. (2023)"},{"why":"Supplies the constrained quasi-Newton optimization method used to minimize the residual.","marker":"Conlin et al. (2023)"}],"fun_headline_variants":["Boundary integrals beat Neumann solves for MHD equilibria","DESC's singular integral cuts free-boundary error below VMEC","One Biot-Savart integral per step yields stellarator equilibria","Sheet currents handled: DESC extends free-boundary MHD to finite beta"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument stands or falls on whether the partition-of-unity singular quadrature, as adapted with Fourier interpolation, actually retains high-order accuracy for the Biot-Savart integral on DESC's Fourier-parameterized boundary surfaces, and the paper does not report an independent convergence-order test in that setting.","fun_headline_variants_meta":{"raw":{"variants":["Boundary integrals beat Neumann solves for MHD equilibria","DESC's singular integral cuts free-boundary error below VMEC","One Biot-Savart integral per step yields stellarator equilibria","Sheet currents handled: DESC extends free-boundary MHD to finite beta"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000811,"raw_usage":{"total_tokens":3529,"prompt_tokens":891,"completion_tokens":2638,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":507,"completion_tokens_details":{"reasoning_tokens":2563}},"tokens_in":507,"tokens_out":2638,"duration_ms":19866,"temperature":1.0,"reasoning_tokens":2563,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T20:27:44.117811+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the boundary integral (4.3) on a surface with a known analytic field (for example a circular torus with a uniform sheet current) and measure the error as the grid is refined; if the observed order is far below the claimed 10th to 12th, the accuracy advantage over VMEC is not established. A complementary check is to reconstruct the same W7-X-like equilibrium with an independent high-accuracy method and compare flux surfaces and boundary residuals to the DESC and VMEC results.","supporting_citations":[{"cited_title":"Plasma Physics and Controlled Fusion 62 (2), 024004","cited_arxiv_id":null,"evidence_quote":"Supplies the partition-of-unity singular quadrature that the paper adapts and the claimed 10th-12th order accuracy."},{"cited_title":"Journal of Computational Physics 397 , 108791","cited_arxiv_id":null,"evidence_quote":"Earlier high-order boundary integral solver for Taylor states in stellarators that motivates the method."},{"cited_title":", van RIJ, W.I","cited_arxiv_id":null,"evidence_quote":"Original three-dimensional free-boundary Green's function calculation that is the VMEC baseline."},{"cited_title":"Journal of Computational Physics 66 (1), 83--98","cited_arxiv_id":null,"evidence_quote":"The NESTOR integral-equation technique for the exterior Neumann problem, the 2nd-order method being replaced."},{"cited_title":"Journal of Computational Physics 219 (1), 247--275","cited_arxiv_id":null,"evidence_quote":"Proves the polar-coordinate quadrature yields the correct principal value for hypersingular kernels like Biot-Savart."},{"cited_title":"Physical Review Letters 128 (3), 035001","cited_arxiv_id":null,"evidence_quote":"Provides the precise QA equilibrium used as the vacuum benchmark."}],"review_version":1}