{"id":"13496e60-8bce-443e-96f7-219ef540f864","arxiv_id":"2411.08271","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A regularized IMEX relaxation Runge-Kutta scheme for the logarithmic Schrödinger equation conserves mass and reaches the same temporal order as its underlying Runge-Kutta method.","lead":"This paper builds high-order, mass-conserving time-stepping methods for the logarithmic Schrödinger equation by smoothing the logarithm near zero and using implicit-explicit relaxation Runge-Kutta time integrators with Fourier spectral space discretization. If the scheme works as claimed, it gives numerical analysts a more accurate tool for a singular nonlinear wave equation where most existing methods are only first or second order.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3.2's p+1 order claim assumes the ERLogSE solution is C^{p+1}, but the C^k energy regularization in (2.1)-(2.2) does not guarantee this for k<p; the paper never states a k-p relation or regularity condition on u0.","rationale":"The mass-conservation part of the construction is essentially the standard relaxation trick: choosing gamma by (3.4) forces the discrete norm to be invariant, and when the denominator vanishes the branch gamma=1 is consistent in the generic case where the stage increment is zero. The order result is the delicate point. Li et al.'s Lemma 3.1 is for p-th order IMEX RK methods under sufficient smoothness; the present paper imports it without stating the regularity that the C^k energy regularization can actually guarantee. Since f^k_epsilon is only C^k, the solution's time derivatives beyond order k are not controlled, and the Taylor argument in Theorem 3.2 cannot reach order p+1 when k<p. This is not an external-consensus disagreement; it is an internal gap between the assumptions stated (none on regularity) and the conclusion claimed. The numerical section does not contradict this: the Gaussian exact solution has time-independent density, so each grid point stays on one side of rho = epsilon^2 and the non-smooth switching surface is not crossed in time; the moving-Gausson experiment is only qualitative. I therefore regard the missing k-p relation as the single load-bearing concern. If the authors add the hypothesis k >= p (or prove the required regularity for the regularized solution) and verify convergence with a datum that actually crosses the threshold, the central claim should stand. The minor typo in (3.4) does not change this assessment but should be fixed.","tokens_in":7610,"tokens_out":14930,"duration_ms":163121,"concrete_test":"Run the temporal-convergence experiment with k=2 (the C^2 regularization), the RRK(8,5) method, and a moving or dynamic initial datum whose density |u|^2 crosses epsilon^2 during [0,T] (e.g., the two-Gaussian datum (4.3) or a single traveling Gaussian), using a highly resolved spatial grid and tau = 2^{-j} * 10^{-1}. If the observed convergence order is below 5, the missing regularity assumption in Theorem 3.2 is a real gap; if order 5 persists, the concern is not triggered and the theorem may only need an explicit smoothness hypothesis.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central order result, Theorem 3.2, depends on Lemma 3.1 (gamma_n = 1 + O(tau^{p-1})) and on expanding the exact ERLogSE solution to order p+1. That expansion is legitimate only if the solution has bounded time derivatives through order p+1. But the regularized nonlinearity f^k_epsilon in (2.2) is only C^k, and the equation u_t = iDelta u + i lambda u f^k_epsilon(u) forces d^{p+1}u/dt^{p+1} to involve derivatives of f^k_epsilon up to order p. For k<p, at points where the solution density crosses the switching surface rho = epsilon^2, the required derivatives do not exist in the classical sense, so the O(tau^{p+1}) local truncation error statement is not established and can fail. The paper neither states a condition k >= p nor gives regularity hypotheses on u0. The numerical convergence tests in Section 4.1 use the stationary Gaussian (4.1) whose density is time-independent at each point, so this regularity issue is never exercised; the two-Gaussian dynamics in Section 4.2 is not used for rate measurement. Separately, the printed relaxation coefficient (3.4) has a self-referential first branch (the 'tau gamma_n' inside the condition) that should be corrected, since Theorem 3.1's proof relies on this formula being solvable for gamma_n.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes an energy-regularized logarithmic Schrödinger equation (ERLogSE) and applies implicit-explicit relaxation Runge-Kutta (IMEX RRK) methods in time with Fourier pseudo-spectral discretization in space. The main theoretical claims are Theorem 3.1 (exact mass conservation of the semi-discrete IMEX RRK scheme for a specific relaxation coefficient) and Theorem 3.2 (temporal truncation error O(τ^{p+1}) for a p-th order base IMEX RK method). Numerical experiments report temporal and spatial convergence rates for four IMEX RK pairs and long-time dynamics for two colliding Gaussons.","tokens_in":7939,"tokens_out":6394,"duration_ms":69746,"significance":"If correct, the paper provides a practical linearly implicit, high-order time integrator family for the LogSE that conserves mass exactly, which is valuable because the logarithmic nonlinearity is singular and many existing schemes are low order or non-conservative. The algebraic mass-conservation mechanism is clean and the numerical experiments support the advertised orders in the tested cases. However, the order proof is semi-discrete and rests on unstated regularity assumptions tied to the polynomial degree k of the energy regularization; the fully discrete spatial error is not analyzed. The paper's main contribution is the method construction plus numerical validation, and that contribution is plausible but not yet fully rigorous.","major_comments":[{"comment":"Equation (3.4) as printed is not a well-defined definition: in the first branch the condition \"∑ (b^I_j g^I_j + τ γ_n ∑ b^E_j g^E_j) = 0\" contains γ_n on both sides, while the proof of Theorem 3.1 and the subsequent Lemma 3.1 require the simpler condition ∑ (b^I_j g^I_j + b^E_j g^E_j) = 0 with no γ_n inside. This display must be corrected before Theorem 3.1 is a precise statement.","section":"Section 3.1, Eq. (3.4)"},{"comment":"The proof expands the exact ERLogSE solution to order p+1, which requires the solution to have bounded time derivatives through order p+1. The regularized nonlinearity f^k_ε in (2.2) is only C^k in ρ, and when the density crosses ρ = ε^2 the (k+1)-st derivative jumps; for k < p the required Taylor expansion is not justified. The manuscript never states a condition such as k ≥ p, nor any regularity hypothesis on u0 that would imply C^{p+1} regularity of the ERLogSE solution. The rate tests in Section 4.1 use the Gaussian (4.1) with ς = 0, whose density is time-independent at each point, so they do not exercise the switching surface; the two-Gaussian experiment in Section 4.2 is not used for rate measurement. Please state the missing regularity assumption or prove the order result under weaker conditions.","section":"Section 3, Theorem 3.2"},{"comment":"The proof substitutes the exact solution into (3.3) and then invokes Lemma 3.1 for γ_n, but γ_n is defined through stage values of the numerical solution; it is not shown that the γ_n obtained from exact-solution substitutions has the same expansion. The displayed cancellation also drops the O((γ_n−1)^2 τ^2) term by using γ_n − 1 = O(τ^{p−1}), which is valid for p ≥ 2 but should be stated explicitly. As written, the argument is closer to a heuristic order estimate than to a self-contained local truncation error proof of the actual discrete map.","section":"Section 3, proof of Theorem 3.2"}],"minor_comments":[{"comment":"The introduction refers to a Section 6, but the paper has only five sections; the conclusion is Section 5.","section":"Introduction"},{"comment":"The polynomial degree k is denoted by n in the text and figure captions (for example, \"larger values of n\" and Fig. 4.3 captions with n = 2, n = 4); please use one symbol consistently.","section":"Section 4.1.1"},{"comment":"Theorems 3.1 and 3.2 are stated only for the semi-discrete system, while the abstract and Section 4 describe the fully discrete Fourier pseudo-spectral method; please state explicitly that the same algebraic mass-conservation identity holds for the spectral discretization, or add a remark on the fully discrete setting.","section":"Section 3 and 4"},{"comment":"The \"order\" column in Table 4.1 appears to be computed from consecutive values of N; please state the formula used (for example, log(‖e_{N_j}‖/‖e_{N_{j+1}}‖)/log(N_{j+1}/N_j)) so the reader can interpret the very high orders at N = 20.","section":"Table 4.1"},{"comment":"For the exact Gaussian solution (4.1), the paper should state explicitly that the computational domain is chosen large enough so that the boundary truncation error is negligible, and should specify the boundary condition used in the finite-domain computation.","section":"Section 4.1"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of a numerical analysis journal. I did not find evidence of circularity or fabricated consistency; the mass conservation is an algebraic consequence of the method, and the main gap is the missing regularity condition in the order theorem plus a few presentation issues. The cited prior work [10] for Lemma 3.1 and [11] for a related first-order scheme are appropriate. The required fixes appear feasible within the manuscript's scope, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the genuinely new piece is applying Li et al.'s IMEX relaxation Runge-Kutta framework to Bao et al.'s energy-regularized LogSE, producing high-order, mass-conserving, linearly implicit time integrators. That is a real and useful extension — high-order options for LogSE are scarce, and mass conservation is exact by construction. I verified the mass argument: it's algebraic, hinges on the relaxation coefficient, and the proof in Theorem 3.1 is correct.\n\nThe numerics are honest and confirm the expected rates; the paper is readable and the test set is adequate. Credit is given where the ingredients come from, though the abstract's \"we propose an energy regularization technique\" overstates things — that's Bao et al.'s idea.\n\nWhere it gets sticky: Theorem 3.2's truncation order O(τ^{p+1}) relies on Lemma 3.1 from Li et al. and on a Taylor expansion of the exact ERLogSE solution. The expansion needs C^{p+1} time regularity, but the regularized nonlinearity f^k_ε is only C^k. Unless k ≥ p, derivatives above order k can blow up where |u|^2 crosses ε^2, so the claimed order is not actually established. The paper never states a relation between k and p, nor regularity hypotheses on u0. The numerical convergence tests dodge this by using a stationary Gaussian, whose density doesn't cross the switching surface; the two-Gaussian dynamics section never measures rates. This is a genuine gap, but likely fixable by requiring k ≥ p or by a more refined regularity analysis.\n\nThere's also a typo in (3.4): the first branch of γ_n contains τγ_n in its own condition. It's self-referential; presumably the condition should just be the stage combination being zero. A referee will catch it, and the authors should correct it.\n\nNo code or data are shipped, which is a minor reproducibility miss.\n\nBottom line: the paper deserves a serious referee. The construction is sound, the mass conservation is exact, and the numerical evidence is solid. The order proof needs a stated regularity condition and the typo needs fixing, but nothing here is load-bearing in the sense that the method fails. I'd send it out.","headline":"A solid, useful combination of existing regularization and IMEX relaxation RK ideas for LogSE, with a fixable gap in the order proof's regularity assumptions.","tokens_in":8439,"tokens_out":3469,"would_cite":true,"duration_ms":32476,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q41","65L06","65M12","65M70"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper develops energy-regularized implicit-explicit relaxation Runge-Kutta methods for the logarithmic Schrödinger equation that conserve mass exactly and keep the full temporal order of the underlying Runge-Kutta pair.","keywords":["logarithmic Schrödinger equation","energy regularization","implicit-explicit relaxation Runge-Kutta","mass conservation","Fourier pseudo-spectral method","high-order time integration","non-differentiable nonlinearity"],"falsifier":"Run the IMEX RRK(2,3) scheme on the ERLogSE with an initial datum that has a zero at some point, keep $\\varepsilon$ fixed, and measure the $L^2$ temporal error against a fine reference as $\\tau$ decreases. If the observed rate falls below 3, or if $\\max_n |\\gamma_n - 1|$ does not decay like $\\tau^2$, the smoothness premise behind Theorem 3.2 is contradicted; mass conservation would still hold algebraically.","tokens_in":7438,"feed_emoji":"⚛️","tokens_out":13243,"duration_ms":117885,"temperature":0.7,"pith_summary":"The logarithmic Schrödinger equation (LogSE) has a nonlinearity $\\ln|u|^2$ that is not differentiable at $u=0$, which blocks standard high-order numerical schemes. The paper removes the singularity by replacing the energy density with a $C^k$ polynomial near zero, then applies implicit-explicit relaxation Runge-Kutta (IMEX RRK) time stepping with Fourier pseudo-spectral discretization in space. Its central claim is that the relaxation coefficient can be chosen to make the semi-discrete scheme conserve mass exactly, while the local truncation error remains $O(\\tau^{p+1})$ for a base IMEX Runge-Kutta method of order $p$. If correct, this yields a practical class of linearly implicit, high-order, mass-conserving integrators for a singular nonlinear Schrödinger model, and the numerical experiments confirm orders 2 through 5.","feed_headline":"Mass-conserving high-order integrators for LogSE","feed_subtitle":"Energy regularization removes the log singularity; relaxation Runge-Kutta keeps mass exact and order p.","key_machinery":"The load-bearing object is the relaxation coefficient $\\gamma_n$ in (3.4), appended to an IMEX Runge-Kutta pair. It is computed from the discrete inner product and the stage derivatives after each step, and by construction it cancels the leading term in the mass variation, forcing exact conservation of the discrete $L^2$ norm. Because the base method is $p$-th order, Lemma 3.1 gives $\\gamma_n = 1 + O(\\tau^{p-1})$, so the correction is small for small steps, and the Taylor argument in Theorem 3.2 shows the truncation error remains $O(\\tau^{p+1})$. The second essential piece is the energy regularization (2.1), which replaces the singular energy density by a $C^k$ polynomial near zero, making the nonlinearity smooth enough for the expansion and for the Runge-Kutta analysis.","core_discovery":"The paper's central discovery is that any $p$-th order IMEX Runge-Kutta pair can be adapted to the energy-regularized logarithmic Schrödinger equation (ERLogSE) without losing order, by appending a relaxation factor $\\gamma_n$ that enforces exact mass conservation. The factor is chosen as in (3.4): it equals 1 when the stage update has zero mass derivative, and otherwise $\\gamma_n = -2\\,\\Re(\\sum_j (b^I_j g^I_j + b^E_j g^E_j), u^{\\varepsilon,n}_\\gamma) / (\\tau \\|\\sum_j (b^I_j g^I_j + b^E_j g^E_j)\\|^2)$. Theorem 3.1 shows this choice gives $\\|u^{\\varepsilon,n+1}_\\gamma\\|^2 = \\|u^{\\varepsilon,n}_\\gamma\\|^2$, and Theorem 3.2 shows the truncation error is $O(\\tau^{p+1})$, using the fact that $\\gamma_n = 1 + O(\\tau^{p-1})$ for a $p$-th order base method. The regularity that makes the analysis possible comes from the energy regularization (2.1), which replaces the non-differentiable $F(\\rho) = \\rho\\ln\\rho - \\rho$ by a $C^k$ polynomial on $0 \\le \\rho < \\varepsilon^2$.","pith_inferences":["The paper leaves implicit that the optimal balance between regularization and time step is $\\varepsilon \\sim \\tau^p$, since the ERLogSE-to-LogSE error is $O(\\varepsilon)$ while the temporal error is $O(\\tau^p)$; choosing $\\varepsilon$ much smaller wastes resolution, and choosing it larger makes the modeling error dominate.","The mass-conservation argument uses only the discrete inner product and the relaxation formula, so it should transfer to other spatial discretizations such as finite element or finite difference methods that preserve a compatible inner-product structure; the paper tests only Fourier pseudo-spectral, so this is an untested extension.","The same construction likely applies to the regularized sine-Gordon-Schrödinger equation (ERSSE) mentioned in the paper, whose results are stated to be similar but not shown; applying the scheme there would test the generality of the mechanism."],"forward_implications":["Any $p$-th order IMEX Runge-Kutta pair becomes a mass-conserving scheme for the ERLogSE with the same temporal order; the paper demonstrates this for schemes of orders 2 through 5.","The regularization parameter $\\varepsilon$ controls the modeling error: the ERLogSE solution and its density converge linearly in $\\varepsilon$ to the LogSE solution, while the regularized energy converges quadratically.","Mass conservation holds exactly at the semi-discrete level for the chosen relaxation coefficient, independently of the time-step size.","Combined with Fourier pseudo-spectral spatial discretization, the method gives spectral accuracy in space, so the overall solver can be high order in both time and space."],"supporting_citations":[{"why":"Supplies the IMEX relaxation Runge-Kutta framework and Lemma 3.1, the key tool behind the $O(\\tau^{p+1})$ truncation-error argument.","marker":"[10]"},{"why":"Introduces relaxation Runge-Kutta step-rescaling for conservation of inner-product norms, which the paper adapts to the implicit-explicit setting.","marker":"[9]"},{"why":"Provides the local energy regularization technique ($C^k$ polynomial smoothing) that the paper uses to remove the log singularity.","marker":"[3]"},{"why":"Baseline regularized numerical method for LogSE; the paper's scheme is positioned as a higher-order, mass-conserving alternative to this semi-implicit finite difference approach.","marker":"[2]"},{"why":"Supplies the exact Gaussian solution used as the reference for the convergence-rate experiments in Section 4.","marker":"[5]"},{"why":"Provides the particular high-order additive Runge-Kutta pairs ARK4(3)6L[2]SA and ARK5(4)8L[2]SA tested in the numerical experiments.","marker":"[8]"}],"fun_headline_variants":["Exact mass in LogSE: high-order relaxation RK","Relaxation RK tames log singularity, keeps mass","Mass-exact high-order IMEX for LogSE","Energy-regularized LogSE: mass-conserving RK"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the exact ERLogSE solution is smooth enough in time for a Taylor expansion of order $p+1$, and that the base IMEX Runge-Kutta method is at least second-order accurate; the paper takes this smoothness as given rather than deriving it from the $C^k$ energy regularization.","fun_headline_variants_meta":{"raw":{"variants":["Exact mass in LogSE: high-order relaxation RK","Relaxation RK tames log singularity, keeps mass","Mass-exact high-order IMEX for LogSE","Energy-regularized LogSE: mass-conserving RK"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00063,"raw_usage":{"total_tokens":2917,"prompt_tokens":961,"completion_tokens":1956,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":577,"completion_tokens_details":{"reasoning_tokens":1889}},"tokens_in":577,"tokens_out":1956,"duration_ms":14763,"temperature":1.0,"reasoning_tokens":1889,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T21:47:03.494079+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the IMEX RRK(2,3) scheme on the ERLogSE with an initial datum that has a zero at some point, keep $\\varepsilon$ fixed, and measure the $L^2$ temporal error against a fine reference as $\\tau$ decreases. If the observed rate falls below 3, or if $\\max_n |\\gamma_n - 1|$ does not decay like $\\tau^2$, the smoothness premise behind Theorem 3.2 is contradicted; mass conservation would still hold algebraically.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the IMEX relaxation Runge-Kutta framework and Lemma 3.1, the key tool behind the $O(\\tau^{p+1})$ truncation-error argument."},{"cited_title":"Ketcheson","cited_arxiv_id":null,"evidence_quote":"Introduces relaxation Runge-Kutta step-rescaling for conservation of inner-product norms, which the paper adapts to the implicit-explicit setting."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the local energy regularization technique ($C^k$ polynomial smoothing) that the paper uses to remove the log singularity."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Baseline regularized numerical method for LogSE; the paper's scheme is positioned as a higher-order, mass-conserving alternative to this semi-implicit finite difference approach."},{"cited_title":"Carles and I","cited_arxiv_id":null,"evidence_quote":"Supplies the exact Gaussian solution used as the reference for the convergence-rate experiments in Section 4."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the particular high-order additive Runge-Kutta pairs ARK4(3)6L[2]SA and ARK5(4)8L[2]SA tested in the numerical experiments."}],"review_version":1}