{"id":"e523d4b0-f6e3-4139-aaf6-8debf2990c2b","arxiv_id":"2509.09276","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A Fourier spectral scheme for the Landau-Coulomb equation is proven to converge with explicit error bounds: larger truncated domains and more Fourier modes push the error below any prescribed tolerance on any fixed time interval.","lead":"The paper proves a rigorous error estimate for a Fourier spectral scheme for the spatially homogeneous Landau equation with Coulomb interactions: for any fixed time interval and accuracy target, taking a large enough velocity box and enough Fourier modes makes the numerical solution arbitrarily close to the exact solution. The result is important for plasma simulation because it gives a provable convergence guarantee for a spectral method in the physically relevant Coulomb ca","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Nonnegativity proof of f^R uses a false identity; the central convergence theorem relies on it.","rationale":"The reader identified the imported uniform-in-time regularity M from preprint [32] as the weakest assumption. That is a legitimate validation risk, but the more immediate and concrete problem is internal: the nonnegativity proof of f^R in Section 3.3 uses a false algebraic identity. The claimed identity with a local 4π∫h³ term is not implied by the Landau operator's weak formulation; the operator is nonlocal and the negative-part energy satisfies a nonlocal double-integral expression. A direct counterexample (e.g., a negative constant profile) shows the identity cannot hold. Since Theorem 1.2's proof explicitly invokes f^R ≥ 0 to bound the term I111, and Theorem 1.3 is a direct corollary of Theorems 1.1 and 1.2, the central convergence guarantee loses its proof. This is not a disagreement with external consensus or a mere reliance on an unverified preprint; it is a gap in the manuscript's own derivation. The result might be true and repairable, but as submitted the main theorem is not rigorously established. Therefore the verdict should move from CONDITIONAL to REJECT.","tokens_in":40737,"tokens_out":30507,"duration_ms":295397,"concrete_test":"Check the identity in Section 3.3 with a concrete smooth negative function, e.g., h(v) = -e^{-|v|^2}. Compute the left side ∫Q(h,h)h dv using the weak-form double integral, and the right side -∫(a*h):∇h⊗∇h dv + 4π∫h³ dv. A numerical quadrature (or symbolic evaluation) will show they differ. For an even simpler analytical test, take h=-1 on B(0,R) with a smooth cutoff; as R grows, the left side scales at most like a surface term O(R^2), while 4π∫h³ scales like O(R^3), contradicting the identity.","verdict_should_be":"REJECT","load_bearing_attack":"In Section 3.3, the paper asserts an identity for the negative-part energy: ∫Q(h,h)h1_{h<0} dv = -∫(a*h):∇h⊗∇h 1_{h<0} dv + 4π∫h³ 1_{h<0} dv, where h=f^Rψ_R. This is algebraically false. Using the weak form of the Landau operator, ∫Q(h,h)φ(h)dv = -1/2∫∫ a(v-v*)(h'∇h - h∇h')·(∇φ(h) - ∇'φ(h')) dv dv*. For φ(h)=h1_{h<0}, this gives a nonlocal double integral, not the local expression with a 4π∫h³ term. For a counterexample, take h constant equal to -c on a large ball and zero outside; then Q(h,h) vanishes exactly where h is constant, and the left side is a surface term (or zero for a truly constant h), while 4π∫h³ is proportional to the ball's volume, so the identity fails for large balls. This matters because Theorem 1.2's proof uses the nonnegativity of f^R to make the leading term I111 nonpositive, and Theorem 1.3 relies on Theorem 1.2. Since nonnegativity of f^{R,N}_# is not claimed, the proof chain breaks at f^R. The paper gives no other argument for nonnegativity of f^R, so the error estimate (1.15) is not established as written.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes a Fourier-Galerkin spectral method for the spatially homogeneous Landau equation with Coulomb potential. The strategy is a two-step approximation: first, the equation is truncated in velocity with a smooth cutoff ψ_R, and an O(R^{-l}) error estimate between the truncated and exact solutions is derived (Theorem 1.1). The truncated solution is then periodized on a torus and discretized by N Fourier modes per direction, yielding an O((R/N)^{n-2} exp(κ R^{1/2} T)) spectral error (Theorem 1.2). Combining the two gives the main convergence statement (Theorem 1.3): for any fixed T and tolerance, sufficiently large L and N make the L2 error arbitrarily small. The proofs are based on weighted Sobolev energy estimates, commutator inequalities, a nonnegativity argument for the truncated solution, and exact Fourier symbol computations for the periodic Landau operator. Numerical experiments for Maxwellian and Coulombian interactions are included.","tokens_in":41000,"tokens_out":38001,"duration_ms":400046,"significance":"If the analysis is correct, this is a significant contribution: it gives the first explicit, parameter-free a priori convergence rates for a spectral method applied to the full Landau-Coulomb equation, with the expected error structure R^{-l} + (L/N)^{n-2}. The proof is largely self-contained at the numerical-analysis level and carefully tracks all constants. The main external input is a global uniform Sobolev bound M imported from the analytic theory, in particular from the recent preprint [32] by He, Ji and Luo. This dependency is heavy and should be clearly stated, but it is not by itself a defect.","major_comments":[{"comment":"The proof states: 'Finally by applying Theorem A.2-(1) with r = 1/2, m = ... we obtain that (2.19) holds.' However Theorem A.2-(1) is stated only for r ∈ [-1/2, 0]. The desired membership (2.19) is exactly the r = -1/2 case of that theorem, giving C([0,∞); H^{-1/2}_m) ∩ L^2_loc([0,∞); H^{1/2}_{m-3/2}). Since (2.19) is used as the bootstrap hypothesis for the interval-induction yielding the uniform H^{n+2}_{k+l} bound, this is a load-bearing step. Please correct the value of r and verify the hypotheses (in particular f0 ∈ L^1_{2m+1}) under Assumption 1.1.","section":"§2.3 (proof of Proposition 1.1, final paragraph)"}],"minor_comments":[{"comment":"The identity ∫ Q(h,h) h 1_{h<0} dv = -∫ (a*h):∇h⊗∇h 1_{h<0} dv + 4π∫ h^3 1_{h<0} dv is correct for smooth h: the second term follows from ∇·(a*∇h) = -8πh and integration by parts with g = h^2 1_{h<0}/2. The stress-test counterexample with h a genuine step function is outside the admissible class, since products of surface distributions are not defined. Still, the paper's 'By further computation' is too terse; a short derivation should be added, especially because the negative-part function f^R 1_{f^R<0} is nonstandard (it is nonpositive rather than the usual positive negative part).","section":"§3.3 (nonnegativity identity)"},{"comment":"The condition R > 3 is not enough to guarantee ―f ψ_R―_{L1} > 1/2 from the energy bound ∫|v|^2 f = 3; Chebyshev gives mass outside |v|>R/2 ≤ 12/R^2, so one needs R > √24. Since the theorem takes R large anyway, this is a minor quantitative slip, but the threshold should be corrected or replaced by 'R sufficiently large'.","section":"Lemma 3.1"},{"comment":"Typographical issues: 'employee' should be 'employ' (§1.2), 'functionnal' → 'functional', 'moldecules' → 'molecules' (§1.3 and §5), 'W e' → 'We' (§4.2). Also, the remark after Theorem 1.2 that the estimate requires N > L is dimensionally odd (L has length units); the condition should be phrased as N sufficiently large relative to R, as in the theorem statement.","section":"Throughout"},{"comment":"The statement of Theorem A.2-(1) should be checked against [32] and its hypotheses listed with the range of r used later. In particular, the proof of Proposition 1.1 additionally invokes Theorem A.2-(2) and A.2-(5); making the exact dependencies explicit would help the reader, since all later constants depend on M.","section":"Appendix A / Theorem A.2"}],"recommendation":"minor_revision","confidential_remarks":"The main theorem is conditional on the correctness of a substantial analytic result imported from [32], an arXiv preprint by one of the authors. This is not itself a reason to reject, but the editor should be aware that if [32] is not yet firmly established, the convergence guarantee in this paper inherits that risk. A self-contained appendix stating the imported theorem and its proof status would materially strengthen the paper. The numerical section does not directly validate the error bound (it mostly shows qualitative behavior); that is acceptable, but a comparison of observed rates with Theorem 1.3 would be useful."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the paper targets a real gap—there is no previous error estimate with explicit rates for Fourier spectral methods applied to the Landau-Coulomb case. The two-level truncation (ψ_R inside and outside the collision operator) is a sensible device, and the final error structure (R^{-l} + (R/N)^{n-2}) matches what one expects. The exact formula for the periodic collision multiplier β(l,m) in Proposition 4.1 is a genuinely nice piece of algebra.\n\nBut there is a load-bearing flaw. Section 3.3 asserts the identity\n\n∫ Q(h,h) h 1_{h<0} dv = -∫ (a*h):∇h⊗∇h 1_{h<0} dv + 4π∫ h³ 1_{h<0} dv,\n\nwith h = f^R ψ_R. That is not the weak form. The correct expression contains the double integral ∫∫ a(v-v′) ∇h(v′) · ∇h(v) h(v) 1_{h(v)<0} plus surface terms living on the zero set of h; it does not collapse to a local cubic term. Try h = -c on a large ball and zero outside: Q(h,h) is identically zero on the ball, so the left side is zero, while the cubic term is a large negative number. The identity fails. The entire nonnegativity proof of f^R depends on this identity, and Theorem 1.2 later uses f^R ≥ 0 to make the leading term I111 nonpositive. I see no other argument for nonnegativity in the paper, so the convergence guarantee (1.15) is unsupported as written. This is not a typo; it is the central estimate.\n\nElsewhere, the quality is uneven but mostly real. The analytic machinery is sophisticated, and the imported regularity bound M from the authors' arXiv preprint [32] is a genuine validation risk—if that preprint does not survive review, the constants are not grounded—but the use of the Fisher information monotonicity from [27] is legitimate. The numerics are illustrative: no code or data are provided, so they do not independently verify the theorem. The introduction's promise that the scheme 'preserves non-negativity' is an overstatement; the discrete solution does not, and the authors concede this later in the text.\n\nMy bottom line: the paper's goal is important and several components are solid, but its headline result rests on a false computation. Send it to a serious referee by all means; the authors should be asked to replace the nonnegativity argument or to state the main theorem conditionally. I would not cite it as a proof until that is fixed.","headline":"The paper aims at the first explicit convergence bound for a spectral discretization of the Landau-Coulomb equation, but the nonnegativity proof in Section 3.3 relies on a false identity, so the main theorem is not established as written.","tokens_in":812,"tokens_out":723,"would_cite":false,"duration_ms":53754,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65M15","82C40","35B65","35R09","65M70","65N35"],"pacs":[],"model":"deepseek-v4-flash","headline":"A Fourier spectral method for the Landau–Coulomb equation is proven to converge with explicit error bounds: for any fixed time interval and any tolerance, sufficiently large domain and mode count bring the numerical solution within toleranc","keywords":["Landau equation","Coulomb potential","Fourier-Galerkin spectral method","error estimate","convergence analysis","kinetic theory","collision operator","spectral method"],"falsifier":"Run the scheme for a smooth admissible initial datum, and at a fixed time compare the $L^2$ error for $(L,N)$ pairs that double $L$ and $N$ according to Theorem 1.3 against a very high-resolution reference solution: the error should drop roughly like $\\max(L^{-l}, (L/N)^{n-2})$ with the proven rates; a slower rate would falsify the spectral projection or truncation estimates. A separate check: if one exhibits an admissible initial datum whose Landau-Coulomb solution develops a singularity or a growing $H^{n+2}_{k+l}$ norm in finite time, Proposition 1.1, and with it Theorem 1.3, is false.","tokens_in":40562,"feed_emoji":"⚛️","tokens_out":6081,"duration_ms":64497,"temperature":0.7,"texified_at":"2026-08-05T20:28:42.618043+00:00","pith_summary":"This paper proves convergence with explicit error bounds for a Fourier–Galerkin spectral method solving the spatially homogeneous Landau equation with Coulomb potential. The numerical solution is built in two steps: the infinite velocity space is truncated to a box of half-size $L$, and the periodized collision operator is discretized with $N$ Fourier modes per direction. The authors show that the $L^2$ distance between the numerical solution and the exact solution on a fixed time interval $[0,T]$ is bounded by a term decaying like $L^{-l}$ plus a term decaying like $(L/N)^{n-2}$, up to exponential-in-time factors. Hence for any tolerance, sufficiently large $L$ and $N$ bring the error below it. This gives the Coulomb case a rigorous error analysis that the literature had only provided for harder potentials.","texify_model":"deepseek-v4-flash","texify_usage":{"total_tokens":5099,"prompt_tokens":910,"completion_tokens":4189,"prompt_tokens_details":{"cached_tokens":0},"prompt_cache_hit_tokens":0,"prompt_cache_miss_tokens":910,"completion_tokens_details":{"reasoning_tokens":3286}},"feed_headline":"Spectral Landau-Coulomb solver gets rigorous error bound","feed_subtitle":"Proof: bigger velocity box and more Fourier modes push numerical error below any tolerance.","key_machinery":"The argument is carried by three pieces. (1) A coercivity/anisotropic-energy machinery for the Coulomb collision operator: with a weighted Sobolev norm and an angular-derivative norm $H^1_A$, the linearized collision operator gives a negative definite contribution, while commutator estimates control the remainder. (2) The periodic Landau operator $Q^\\#$, whose Fourier symbol is computed exactly in Proposition 4.1: $Q^\\#(e^{i\\pi \\ell \\cdot v/L}, e^{i\\pi m \\cdot v/L}) = \\beta(\\ell, m) e^{i\\pi (\\ell + m) \\cdot v/L}$ with $\\beta(\\ell, m)$ an explicit closed-form expression in $\\ell$ and $m$. This preserves the convolution structure and allows the spectral method to be analyzed mode-by-mode. (3) A two-level error decomposition: the truncation residual produces th","core_discovery":"The central result is Theorem 1.3. Under the paper's assumption that the initial datum lies in a weighted Sobolev space with sufficiently high regularity and moment decay, the exact solution is globally regular with a uniform bound $M$. For any time $T>0$, once the velocity box half-size $L$ and the number of Fourier modes $N$ satisfy the explicit conditions $L \\ge 2\\tilde R(T)$ and $N/L \\ge \\tilde N(T,L/2)$, the $L^2(\\mathbb{R}^3)$ error between the numerical solution $f^{R,N}$ and the exact solution $f$ obeys $\\| f^{R,N}(t) - f(t) \\|_{L^2} \\leq C \\left[ e^{\\kappa_0 t}/L^\\ell + (L/N)^{n-2} e^{\\kappa L^{1/2} t} \\right]$ for all $t \\in [0,T]$, with constants depending only on $M,n,k,l$. Because both terms can be made arbitrarily small by increasing $L$ and $N$, the sche","pith_inferences":["The exponential factor exp(κ L^{1/2} t) in the spectral error suggests that for very long time horizons the number of modes must grow rapidly with the domain size; in practice one might need adaptive or time-dependent truncation to keep the provable bound useful.","The error estimate is in L2; because the scheme also propagates high Sobolev regularity, an L∞ or pointwise error bound should be derivable by Sobolev embedding, though the paper does not state one.","The explicit β(l,m) formula, independent of the rest of the analysis, could be reused to build conservative or entropy-stable spectral discretizations of related kinetic equations, or to analyze time-discretization errors in the Fourier basis.","The proof relies on the global-in-time regularity bound M imported from analytic theory; if future work weakens that hypothesis, the same error structure would carry over, making the numerical analysis conditional only on the regularity of the exact solution."],"forward_implications":["For any fixed time interval and any error tolerance, the spectral method with sufficiently large L and N is guaranteed to approximate the Landau-Coulomb solution within that tolerance in L2.","The truncated equation preserves nonnegativity of the distribution function for R large enough, so the physical meaning of the numerical solution is retained at the truncation level.","The error rate in the number of modes is spectral in the ratio L/N: for solutions with H^n regularity, the discretization error scales like (L/N)^{n-2}, so smoother solutions give faster convergence.","The explicit Fourier symbol β(l,m) supplies a rigorous justification of the fast spectral algorithms used in practice, which evaluate the collision operator through convolutions and FFT.","The numerical experiments for the Coulomb case show that the scheme reproduces the expected decay of entropy, relative entropy, and Fisher information, and that moderate N (e.g., 48 per direction) already matches a reference solution on long time intervals."],"fun_headline_variants":["Landau equation solver: explicit error bound guarantees accuracy","New proof: spectral method for Landau-Coulomb converges","Rigorous error control for Landau equation simulations","Spectral Landau solver: error below any tolerance with enough modes","Guaranteed accuracy for Landau-Coulomb with Fourier modes"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is the uniform-in-time weighted Sobolev bound $\\| f \\|_{L^\\infty([0,\\infty); H^{n+2}_{k+l})} < M$ for the exact solution (Proposition 1.1), imported from prior analytic results: if the Landau-Coulomb solution can lose this regularity or the bound is not finite, the constants in Theorems 1.1–1.3 do not exist and the convergence guarantee collapses.","fun_headline_variants_meta":{"raw":{"variants":["Landau equation solver: explicit error bound guarantees accuracy","New proof: spectral method for Landau-Coulomb converges","Rigorous error control for Landau equation simulations","Spectral Landau solver: error below any tolerance with enough modes","Guaranteed accuracy for Landau-Coulomb with Fourier modes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000584,"raw_usage":{"total_tokens":2566,"prompt_tokens":710,"completion_tokens":1856,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":454,"completion_tokens_details":{"reasoning_tokens":1774}},"tokens_in":454,"tokens_out":1856,"duration_ms":14332,"temperature":1.0,"reasoning_tokens":1774,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T19:22:48.706289+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the scheme for a smooth admissible initial datum, and at a fixed time compare the $L^2$ error for $(L,N)$ pairs that double $L$ and $N$ according to Theorem 1.3 against a very high-resolution reference solution: the error should drop roughly like $\\max(L^{-l}, (L/N)^{n-2})$ with the proven rates; a slower rate would falsify the spectral projection or truncation estimates. A separate check: if one exhibits an admissible initial datum whose Landau-Coulomb solution develops a singularity or a growing $H^{n+2}_{k+l}$ norm in finite time, Proposition 1.1, and with it Theorem 1.3, is false.","supporting_citations":[],"review_version":1}