{"id":"713838d3-64b4-43c8-bf70-f5712302aad8","arxiv_id":"2506.21314","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"An adaptive-rank, operator-splitting solver for the 1D1V Wigner-Poisson system achieves O(N) complexity with mass conservation and structure-preserving Fourier updates.","lead":"This paper builds a fast numerical solver for the Wigner-Poisson equation, a quantum kinetic model of plasmas, by keeping only the most important parts of the solution at every time step. The solver runs in near-linear time and preserves total mass, which could make high-dimensional quantum plasma simulations practical for fusion research.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The O(N) accuracy claim rests on greedy ACA pivot selection that the paper admits does not satisfy the maximal-volume assumptions of the only error bound offered; a missed row or column at late times would silently break the low-rank approximation, so the central claim is conditional.","rationale":"The reader's weakest assumption correctly identifies the decisive point: the O(N) accuracy claim and the long-time fidelity of the adaptive-rank solver both hinge on the greedy ACA pivot strategy, and the paper offers no error bound for that strategy in the Wigner-Poisson setting. The manuscript itself concedes in Section 3.2.2 that the maximal-volume hypotheses behind Theorem 3.2 are not satisfied and defers rigorous error analysis. That admission makes the central claim conditional rather than established, while the numerical evidence (matching damping rate gamma=0.1511 versus 0.1516, machine-precision mass conservation, bounded momentum error, and low-rank plots) provides real but finite support. I would not strengthen the verdict to reject: the experiments are consistent and the gap is addressable by a posteriori checks or a different pivoting strategy. I also note the structure-preserving theorem proof should explicitly track complex conjugation and clarify how the SVD truncation step preserves the paired-column symmetry; this is a secondary gap, but the ACA error-control issue is the more load-bearing one because it threatens the accuracy and O(N) claims directly. The verdict CONDITIONAL therefore remains appropriate.","tokens_in":23052,"tokens_out":16891,"duration_ms":200831,"concrete_test":"Fix the two-stream instability with H=1, Nx=Nv=512, epsilon_c=1e-4, epsilon_s=1e-3, and no maximum rank. Rerun Algorithm 3.2 with Phase I candidate sample sizes p=12, p=50, and p=200, using several random seeds, with all other settings identical. At T=45 and T=200, compute the L2 phase-space error against the full-rank reference, and on a few selected time steps assemble the full ACA residual matrix to check the largest entry after termination. If the error or the maximal residual varies by more than the SVD tolerance across seeds and p, the greedy pivot selection is not reliable; if all runs stay within tolerance, the concern is mitigated for this regime.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim depends on Algorithm 3.2's ACA-SVD compression being accurate enough over long times, but Section 3.2.2 explicitly states that the assumptions of Theorem 3.2 (Goreinov-Tyrtyshnikov maximal-volume bound) are not satisfied by the proposed greedy pivot search, and it defers error analysis to future work. This is an internal gap, not a disagreement with consensus: the only theoretical guarantee available for CUR-type approximations requires selecting a maximal-volume submatrix, and the paper's Phase I pivot selection (random candidate sample plus one row and one column refinement) does not enforce that condition. If at some later time the greedy search misses the row or column containing the dominant residual, the ACA approximation can lose the main physical feature while the reported rank, mass, and momentum diagnostics remain plausibly smooth. The numerical tests are encouraging, but they cover a finite set of initial data and parameters; they do not close the gap. The 'visually indistinguishable' and 'momentum accuracy up to the truncation error' claims are therefore contingent on an unproved sampling property, which is the most load-bearing assumption in the paper. A secondary concern is that Theorem 3.3 proves symmetry for a paired-column CUR decomposition, while Algorithm 3.2 applies an additional SVD truncation; this is likely repairable, but it is not fully spelled out in the proof.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops an adaptive-rank solver for the 1D1V Wigner-Poisson system. It first improves a full-rank Strang-splitting solver by adding a structure-preserving Fourier update that keeps intermediate solutions real-valued. Numerical experiments show that Wigner-Poisson solutions exhibit low-rank structure for moderate to large dimensionless Planck constants (H ≥ 0.1). The authors then build an adaptive-rank solver that combines a semi-Lagrangian advection step (SLAR) with ACA-SVD compression and a structure-preserving Fourier update, together with a Lagrange-multiplier correction to conserve mass. The paper claims O(N) complexity in storage and time, mass conservation up to machine precision, momentum conservation up to truncation error, and rigorous structure preservation, supported by numerical tests for two-stream instability and strong Landau damping.","tokens_in":23348,"tokens_out":4109,"duration_ms":47745,"significance":"If the central claims hold, this is a substantial contribution to kinetic simulation of quantum plasmas: it offers the first adaptive-rank, arbitrary-CFL solver for the Wigner-Poisson system, with a plausible path toward higher-dimensional simulation for NIF-type stopping-power studies. The paper provides several strong forms of evidence: second-order temporal convergence tests (Figure 4.3a), mass and momentum diagnostics (Figures 4.2 and 4.6), a structure-preserving theorem for the CUR stage (Theorem 3.3), and a damping-rate check against an independent prediction (Figure 4.8b, γ=0.1511 vs. 0.1516). These are genuine falsifiable checks, and the absence of parameter fitting strengthens the empirical case. However, the theoretical support for the ACA pivot-selection step is admitted to be incomplete, and the structure-preserving proof does not cover the SVD-truncation stage of the actual algorithm; these gaps are load-bearing for the paper's headline accuracy and structure-preservation claims.","major_comments":[{"comment":"The central O(N)-accuracy claim rests on the greedy ACA pivot selection of Algorithm 3.2, yet the paper explicitly states that the assumptions of Theorem 3.2 (Goreinov-Tyrtyshnikov) do not hold and defers error analysis to future work. Because the only theoretical guarantee cited for CUR-type approximations requires a maximal-volume submatrix, and Phase I in Section 3.2.1 selects pivots by random sampling plus one row/column refinement, a dominant residual could be missed at later times, causing the low-rank approximation to silently lose a physical feature while diagnostics appear smooth. The numerical tests are encouraging but cover finitely many initial data and parameter values; they do not close this gap. The authors should either supply an a posteriori error estimator that monitors ACA approximation quality per step, or substantially qualify the accuracy claims to make their conditional nature explicit.","section":"§3.2.2 (Theorem 3.2 and following paragraph)"},{"comment":"Theorem 3.3 proves that the modified CUR stage preserves column-wise conjugate symmetry, but Algorithm 3.2 then applies an SVD truncation (Section 3.2.1) before the inverse FFT. The proof does not show that the truncated factors (U Σ V^T) retain the symmetry, so the statement that a 'rigorous proof of structure-preserving property is provided' for the implemented solver is not fully supported. This is likely repairable—the symmetric column pairing should survive the QR/SVD step under a suitable ordering—but as written, the proof covers only the intermediate CUR representation, not the final low-rank approximation used in the simulation.","section":"§3.2.3, Theorem 3.3"}],"minor_comments":[{"comment":"The keyword 'Stang splitting methods' is a typo and should read 'Strang splitting methods'.","section":"Abstract (keywords)"},{"comment":"The sentence containing 'the stincel shifts be one to the right' appears to be garbled; it should read 'the stencil shifts one index to the right'.","section":"§3.1.1"},{"comment":"The notation '1 ∈ J' is unclear because J is defined as a set of column indices; the condition should be phrased as 'if the index 1 is contained in J' to avoid ambiguity.","section":"§3.2.3, proof of Theorem 3.3"},{"comment":"The statement that low-rank results for H=1,8 are 'visually indistinguishable' from full-rank results is subjective; a quantitative comparison (e.g., L2 or L∞ difference between the two solutions) would strengthen the accuracy claim.","section":"§4.1"},{"comment":"The Lagrange-multiplier correction rescales the entire solution by Lx/L^{n+1,*}; this conserves total mass but is a global scaling whose effect on local truncation error and on the structure-preserving property is not discussed.","section":"Algorithm 3.1, Step 5"}],"recommendation":"major_revision","confidential_remarks":"The manuscript fits the scope of math.NA, and the reuse of the SLAR method from [47] is legitimate since the Wigner-specific contributions (Fourier update, structure preservation, the ACA-SVD integration) are new. The main obstacle is the missing error analysis for the greedy ACA selection; this is an internal gap that should be repaired or explicitly bounded before publication. I would also ask the authors to extend the structure-preservation proof to the full ACA-SVD pipeline, or state the theorem for the exact algorithm actually implemented."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This is a solid numerical methods paper that deserves a serious referee, but the headline O(N) accuracy claim is conditional on an unproved sampling property of the ACA step, exactly as the paper concedes in Section 3.2.2. If that sampling property holds, the solver is a genuinely useful extension of the SLAR framework to Wigner-Poisson, and the numerical evidence suggests it does hold for the cases tested.\n\nWhat is new: adapting the semi-Lagrangian adaptive-rank scheme to the nonlocal Wigner operator is not trivial, and the conjugate-symmetric column pairing in the ACA step is a sensible, original construction. The structure-preserving Fourier update that keeps intermediate solutions real-valued is a clean improvement over the older Suh et al. approach. The rank-structure study across H = 0.1, 0.5, 1, 8 is informative and motivates the method. The numerical results are genuinely good: second-order convergence up to the truncation tolerance, mass conservation at machine precision, momentum within 1e-4, and the H=8 strong Landau damping rate 0.1511 matches the predicted 0.1516.\n\nThe soft spots are real but proportionate. The main one is that the only theoretical error bound for CUR approximation requires maximal-volume submatrices, and the greedy pivot search does not enforce that. The authors say so themselves and defer the analysis. That makes the O(N) accuracy claim an empirically supported conjecture rather than a proven property. The numerical tests are encouraging but finite; one can imagine a later-time snapshot where the greedy search misses the dominant row or column. This is not a fatal flaw in a numerical paper, but it should be stated more carefully as a conditional claim. Second, Theorem 3.3 proves symmetry for the CUR factorization, while Algorithm 3.2 applies an extra SVD truncation; the proof does not explicitly show that this truncation preserves the pairing. The numerics show it works, so this is likely repairable, but it needs to be spelled out. Minor: the abstract claims O(N) complexity without noting that this holds only in the low-rank regime H >= 0.1; the paper itself says H=0.1 approaches full rank.\n\nWho this is for: computational plasma physicists and numerical analysts working on low-rank kinetic equations. It is a useful stepping stone toward higher-dimensional Wigner-Poisson simulations, and it is honest about its limitations. I would send it to peer review rather than desk reject. The referee should ask for a more prominent statement of the conditional ACA guarantee, and for the SVD-truncation gap in the structure-preserving theorem to be closed or explicitly addressed.","headline":"Worth refereeing: the adaptive-rank Wigner-Poisson extension is real, but the O(N) claim rests on an admitted gap in ACA error control.","tokens_in":23869,"tokens_out":3524,"would_cite":true,"duration_ms":35785,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65M70","65F55","15A23"],"pacs":[],"model":"deepseek-v4-flash","headline":"Adaptive-rank solver cuts Wigner-Poisson cost to O(N) while conserving mass","keywords":["adaptive rank","Wigner-Poisson system","adaptive cross approximation","semi-Lagrangian method","operator splitting","low-rank approximation","structure-preserving","mass conservation"],"falsifier":"Run the two-stream instability test at H = 0.5 on a 2048x2048 mesh for T = 100 and compare the relative L2 error between the adaptive-rank solution (tolerance epsilon_s = 1e-3) and the full-rank solution: if the low-rank approximation accumulates error beyond the truncation tolerance or the rank grows linearly with time, the claim that ACA preserves accuracy at long times fails.","tokens_in":1581,"feed_emoji":"⚛️","tokens_out":4010,"duration_ms":48870,"temperature":0.7,"pith_summary":"This paper addresses the computational challenge of simulating the Wigner-Poisson system, a kinetic model for quantum plasmas that is nonlocal and therefore expensive to solve. The authors observe that in the quantum regime of interest, the solution is intrinsically low rank, so they develop an adaptive-rank solver that compresses the phase-space matrix at every step using a sampling-based cross approximation. The central claim is that this solver achieves linear complexity in storage and computation time while preserving the real-valuedness of the solution and conserving mass. If correct, this makes long-time, high-resolution Wigner-Poisson simulations affordable and opens a practical route toward the six-dimensional simulations needed for studying alpha-particle stopping power in warm dense plasmas.","feed_headline":"Adaptive-rank solver makes Wigner-Poisson O(N)","feed_subtitle":"A sampling-based compression scheme conserves mass and matches full-rank dynamics in the quantum regime.","key_machinery":"The load-bearing mechanism is the CUR decomposition built by adaptive cross approximation (ACA) with greedy pivot selection, coupled with SVD truncation. For each solution snapshot, the ACA procedure samples a small set of rows and columns of the full matrix, using the local numerical solver as a function handle, and constructs a rank-r approximation of the form A ≈ C U R. The SVD stage then compresses the core and stabilizes the representation. For the Fourier update, a modified column selection adds for each selected index its conjugate-symmetric counterpart, which guarantees that the CUR output preserves column-wise conjugate symmetry and therefore keeps the inverse-transformed solution real-valued. This structure-preserving property is proven in Theorem 3.3.","core_discovery":"The paper claims that a full-rank Strang-splitting solver for the 1D1V Wigner-Poisson system can be turned into a structure-preserving, mass-conserving adaptive-rank solver that runs in O(N) time and memory while matching full-rank dynamics, with rigorous preservation of the conjugate-symmetric structure that keeps the Wigner function real. The key numerical observation is that the singular value ranks of Wigner-Poisson solutions remain bounded and even decrease as the dimensionless Planck constant H grows, in contrast to Vlasov-Poisson solutions which develop fine filaments and growing rank. The proposed solver applies a semi-Lagrangian adaptive-rank (SLAR) update for advection and a modified adaptive cross approximation with symmetric column pairing for the Fourier-space nonlocal update, followed by SVD truncation and a Lagrange multiplier step to enforce mass conservation. Numerical tests on two-stream instability and strong Landau damping show visual agreement with full-rank results, mass conservation to machine precision, momentum conservation within $10^{-4}$, second-order temporal convergence, and approximately linear scaling of runtime with mesh size.","pith_inferences":["The paper does not prove that the greedy ACA pivot selection converges for Wigner-Poisson dynamics; it explicitly defers error analysis to future work, so the practical reliability of the method rests on empirical robustness rather than a theoretical guarantee.","A natural testable extension would be to compare the sampling-based ACA with a maximal-volume or rook-pivoting alternative on the same test problems, to see whether the O(N) cost savings come with a hidden accuracy cost at late times.","The connection to gamma = 0.1511 damping rate at H = 8 suggests the low-rank method captures not just solution morphology but also physically meaningful linear response, which could be validated against other quantum plasma diagnostics.","If the rank remains bounded as the spatial mesh is refined, the method may enable a practical tradeoff: smaller H values require higher rank, so the computational frontier will be at the transition to the near-classical regime."],"forward_implications":["If the claimed O(N) scaling holds, the bottleneck of Wigner-Poisson simulation moves from the cost of the nonlocal integral to the rank of the solution, making much larger meshes and longer time horizons feasible.","The observed low-rank structure for H ≥ 0.1 means the method is most effective exactly in the quantum regime relevant to warm dense plasma studies, where classical kinetic models fail.","The mass-conservation Lagrange multiplier combined with SVD truncation bounds momentum error at the truncation tolerance, giving predictable accuracy limits over the 50-plasma-period simulations shown.","The success of the structure-preserving Fourier update suggests that the same symmetric-pairing ACA strategy can be applied to other equations with conjugate-symmetric Fourier structures.","The method's linear complexity in N provides evidence that the extension to a tensor-based 6D Wigner-Poisson solver is a viable research direction."],"supporting_citations":[{"why":"Provides the original splitting framework for the Wigner-Poisson system and the benchmark damping rate gamma = 0.1516 used for comparison.","marker":"[43]"},{"why":"Gives the stability analysis for the second-order Strang splitting method on which the full-rank solver is built.","marker":"[3]"},{"why":"Introduces the SLAR adaptive-rank approach and the ACA-SVD compression strategy that this paper extends to the Wigner-Poisson system.","marker":"[47]"},{"why":"Supplies the Goreinov-Tyrtyshnikov CUR convergence theorem that motivates the approach, though its assumptions are not satisfied by the greedy ACA.","marker":"[20]"},{"why":"Defines the conservative semi-Lagrangian WENO scheme used for the advection subproblem in both full-rank and adaptive-rank solvers.","marker":"[35]"},{"why":"Provides the Cheng-Knorr splitting idea that underlies the operator-splitting treatment of the kinetic equation.","marker":"[10]"}],"fun_headline_variants":["Adaptive-rank Wigner-Poisson solver hits linear time","Sampling-based rank adaptation speeds Wigner-Poisson","Wigner-Poisson solved in O(N) with adaptive rank","Mass-conserving adaptive-rank Wigner-Poisson in O(N)"],"cache_read_input_tokens":25984,"weakest_assumption_plain":"The greedy ACA pivot selection, without rook pivoting or the maximal-volume condition, produces a low-rank approximation accurate enough for long-time Wigner-Poisson dynamics, even though the theoretical convergence theorem does not apply.","fun_headline_variants_meta":{"raw":{"variants":["Adaptive-rank Wigner-Poisson solver hits linear time","Sampling-based rank adaptation speeds Wigner-Poisson","Wigner-Poisson solved in O(N) with adaptive rank","Mass-conserving adaptive-rank Wigner-Poisson in O(N)"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000246,"raw_usage":{"total_tokens":1593,"prompt_tokens":1053,"completion_tokens":540,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":669,"completion_tokens_details":{"reasoning_tokens":467}},"tokens_in":669,"tokens_out":540,"duration_ms":5118,"temperature":1.0,"reasoning_tokens":467,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T22:26:55.700185+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the two-stream instability test at H = 0.5 on a 2048x2048 mesh for T = 100 and compare the relative L2 error between the adaptive-rank solution (tolerance epsilon_s = 1e-3) and the full-rank solution: if the low-rank approximation accumulates error beyond the truncation tolerance or the rank grows linearly with time, the claim that ACA preserves accuracy at long times fails.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the original splitting framework for the Wigner-Poisson system and the benchmark damping rate gamma = 0.1516 used for comparison."},{"cited_title":"1622–1643","cited_arxiv_id":null,"evidence_quote":"Gives the stability analysis for the second-order Strang splitting method on which the full-rank solver is built."},{"cited_title":"Zheng, D","cited_arxiv_id":null,"evidence_quote":"Introduces the SLAR adaptive-rank approach and the ACA-SVD compression strategy that this paper extends to the Wigner-Poisson system."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Goreinov-Tyrtyshnikov CUR convergence theorem that motivates the approach, though its assumptions are not satisfied by the greedy ACA."},{"cited_title":"Qiu and A","cited_arxiv_id":null,"evidence_quote":"Defines the conservative semi-Lagrangian WENO scheme used for the advection subproblem in both full-rank and adaptive-rank solvers."},{"cited_title":"Cheng and G","cited_arxiv_id":null,"evidence_quote":"Provides the Cheng-Knorr splitting idea that underlies the operator-splitting treatment of the kinetic equation."}],"review_version":1}