{"id":"0fd4941a-ab06-48e4-9f2b-37ed661cf5b3","arxiv_id":"2412.19703","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A numerical inverse scattering transform is implemented for the defocusing NLS equation with box-type initial data on a nonzero background, with accuracy demonstrated in both asymptotic regions for soliton-free data.","lead":"This paper builds a numerical method to compute solutions of the defocusing nonlinear Schrödinger equation when the initial profile is a box sitting on a constant nonzero background. The method solves the associated Riemann-Hilbert problem numerically and is tested in the two large-time asymptotic regions, where standard time-stepping methods become slow.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Unproven existence of RHP 5 (solvability condition (167)) leaves the method's arbitrary-(x,t) accuracy claim unsupported; a dense sweep of the Wronskian would test it.","rationale":"The reader's weakest_assumption identifies exactly the load-bearing concern: the deformed RHP 5 is central to the numerical method, and its existence for t>0 is openly left open in the paper (Remark 12, Remark 13, and the discussion after Eq. (167)). The reconstructed q(x,t) is trustworthy only if this RHP has a solution and the solvability condition (167) holds. My stress-test confirms that this is the weakest point in the argument, because the numerical solver's a posteriori check cannot distinguish a genuine solution from a numerical artifact when the true RHP is unsolvable. A concrete dense sweep of the Wronskian would empirically determine whether the condition fails for the tested parameter regimes. I agree with the reader's assessment that the paper is technically serious, honestly acknowledges this gap, and provides meaningful numerical evidence; therefore a CONDITIONAL verdict is appropriate. Since my concern is the same one already identified, no change to the reader's verdict is needed.","tokens_in":51149,"tokens_out":4451,"duration_ms":51509,"concrete_test":"Perform a systematic sweep over (x,t) with ξ=x/(2t) covering both regions, e.g., t=1,10,100,1000 and ξ in [-10,-1] ∪ [-0.9,0.9] ∪ [1,10], using box parameters (59) plus several other solitonless parameter sets from Sec. 4. At each point, compute the numerical solution of RHP 4/5 and evaluate the Wronskian W(ˆm_2(q_o), ˆm_1(-q_o)) from Eq. (167), as well as the maximum jump-matrix residual on the deformed contours. If any point yields |W| below the solver tolerance or a residual that does not converge with increasing n, the method's claimed domain is violated; if no failures occur across a dense grid, the existence obstruction is not observed in practice, though it remains unproven analytically.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim is that the numerical IST accurately solves the Cauchy problem at arbitrary (x,t) in the two asymptotic regions |x/(2t)|<q_o and |x/(2t)|>q_o. The numerical scheme solves the deformed RHP 5 for e-m, introduced to remove the z=0 singularity. As stated in Sec. 3.3.1 (Remark 12): 'The question of existence of solution of RHP 5 remains an open problem', and Remark 13 adds that a proof for t>0 'is currently not available'. Existence is tied to the solvability condition (167) (equivalently (94)): W(ˆm_2(q_o), ˆm_1(-q_o)) ≠ 0, which the code checks only a posteriori. Because the Chebyshev collocation method solves a finite-dimensional linear system that always has a solution, the a posteriori check can pass even if the true RHP 5 has no solution, yielding a spurious q(x,t). Thus the claimed validity for all (x,t) in the stated regions rests on an unproven analytic fact, and the numerical demonstrations, while suggestive, only cover finitely many points. This is the most load-bearing weakness: if the solvability condition fails anywhere in the claimed domain, the method silently produces incorrect values at those points.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript presents a numerical inverse scattering transform (IST) for the defocusing nonlinear Schrödinger equation with constant nonzero boundary conditions and a box-type initial condition, in the solitonless case. The scattering data are obtained in closed form through the uniformization variable z, and the inverse problem is reformulated as a sequence of Riemann--Hilbert problems deformed à la Deift--Zhou, with the singularity at z=0 removed by an auxiliary problem, RHP 5 for e_m. The final RHPs are solved by Chebyshev collocation using the authors' RHPackage and ISTPackage, and the method is tested by a finite-difference PDE residual, a damped Fourier spectral comparison, and a Cauchy error study. The central claims are that the method is accurate in both asymptotic regions |x/(2t)|<q_o and |x/(2t)|>q_o, and that evaluating q(x,t) costs effectively independent of x and t.","tokens_in":51460,"tokens_out":6183,"duration_ms":64691,"significance":"The paper attacks a genuinely new case for numerical IST: the defocusing NLS equation with a discontinuous, non-decaying initial condition. If the method is reliable, it goes substantially beyond prior work on Schwartz-class data and provides a practical tool for large-time asymptotics where direct time-stepping is prohibitive. The analytic development is detailed and mostly careful: explicit scattering coefficients, deformation contours with sign charts, analytic treatment of the logarithmic singularity in the delta function, and a clearly stated set of lemmas and propositions. The paper also honestly discloses its main limitation, the open existence question for RHP 5, rather than hiding it. These strengths make the manuscript a plausible important contribution, but the unresolved existence question and the narrow quantitative validation prevent acceptance in the present form.","major_comments":[{"comment":"Existence of a solution to the deformed RHP 5 is stated to be open, and the a posteriori solvability condition (167) is the only safeguard. The Chebyshev collocation discretization always yields a finite-dimensional linear system, so a numerically computed e_m can exist even when the true RHP 5 has no solution; checking (167) after the fact is not guaranteed to detect this. Because the reconstruction (93)--(100) uses this e_m, the advertised validity for all (x,t) in the two asymptotic regions rests on an unproved analytic fact. I ask the authors either to prove existence in the parameter regimes under consideration, or to state the main claim conditionally, or to provide a decisive numerical probe such as a dense sweep of the Wronskian (167) over the (x,t) plane in both regions, which would make silent failure implausible.","section":"Section 3.3.1, RHP 5, Remarks 12--13, Eq. (167)"},{"comment":"The only quantitative convergence and residual tests are performed in the solitonless regime: the Cauchy error is computed at x=-400, t=100 (so xi=-2), and the PDE residual is reported at x=-11, t=5 (so xi=-1.1). No equivalent error study is presented in the solitonic region |x/(2t)|<1, where the contour deformations and the removal of the singularity at z=0 are the most delicate. The abstract's phrase \"demonstrated to be accurate within the two asymptotic regions\" is therefore stronger than the numerical evidence supports; at least one quantitative convergence test for a point in the solitonic region is needed.","section":"Section 4, Fig. 18, finite-difference residual"}],"minor_comments":[{"comment":"Remark 17 appears self-contradictory: it states that the solvability condition (167) is satisfied at t=0 if and only if theta=+-pi/2, but then displays the unique e_m(0) for theta different from +-pi/2; the first clause should presumably read \"if and only if theta is different from +-pi/2.\"","section":"Remark 17, Appendix D"},{"comment":"Equation (100) uses e_m^(-1)(x,t) without defining it; specify that this is the coefficient of 1/z in the expansion e_m(z)=I + e_m^(-1)/z + O(z^(-2)).","section":"Eq. (100)"},{"comment":"The claim that evaluating the solution is \"a matter of seconds\" is not supported by timing data or a hardware description; a brief timing statement would make the computational advantage reproducible.","section":"Section 1, runtime claim"},{"comment":"The code is described as \"will be made available\" as electronic supplementary material; for reproducibility, include the archival link or DOI in the manuscript.","section":"Data availability"},{"comment":"The caption and surrounding text do not state explicitly which asymptotic regime the Cauchy-error point x=-400, t=100 belongs to; state xi=-2 so the reader can immediately see that this test is in the solitonless region.","section":"Figure 18"}],"recommendation":"major_revision","confidential_remarks":"I am sympathetic to the authors' position that the open existence problem can be isolated as a natural mathematical gap, and I see no evidence of circular parameter fitting. However, the central claim as stated is broader than what is proved, and the numerical demonstrations are too thinly distributed to establish both regions with equal confidence. A revised version that proves existence in the tested parameter regime, or substantially narrows and conditions the claims, and that adds quantitative validation in the solitonic region, could become acceptable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a careful read. This is the first numerical IST implementation for defocusing NLS with nonzero background and a discontinuous box IC, and the technical core is solid. The authors work out explicit scattering coefficients for box data, handle the z=0 singularity with a new transformation, and give a careful analytic treatment of the logarithmic singularities at ±q_o. The numerics are credible: Cauchy error to 1e-15, a PDE residual of 2.5e-5 at one point, and qualitative agreement with an independent Fourier method. The deformations in both the solitonic and solitonless regions are laid out with enough detail to reproduce.\n\nThe soft spot is the one the authors themselves flag: RHP 5, the deformed problem they actually solve, is not proven to have a solution for all (x,t) with t>0. Solvability condition (167) is checked only a posteriori, and the Chebyshev collocation solve of a finite linear system will always return something, so a silent failure is possible in principle. This is a real gap in the claim of accuracy at arbitrary (x,t) in the two asymptotic regions. It is not fatal: the paper openly states the issue, and analytic Fredholm theory suggests failure only on a measure-zero set, which matches the numerical experience. But the authors should either prove existence for the relevant parameter range or state a precise conjecture and back it with a dense numerical sweep of the Wronskian across both regions. Also worth noting: the box ICs are not in the functional class assumed in the scattering theory, and the code is promised but not yet shipped; both are minor but should be cleaned up.\n\nWho is this for? Anyone working on numerical IST or long-time asymptotics for integrable PDEs with non-decaying data. It deserves a serious referee, and with the existence gap addressed or scoped honestly, I would take it as is. Send it to peer review; ask the authors to tighten the claims around RHP 5 and ship the code.","headline":"Serious numerical IST paper for NLS on nonzero background with discontinuous box data; the main caveat is an openly admitted unproven existence statement for the deformed RHP, which should be fixed or clearly flagged as a conjecture.","tokens_in":51984,"tokens_out":1139,"would_cite":true,"duration_ms":14889,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q55","37K15","35Q15","45E05"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper presents a numerical inverse scattering transform for the defocusing NLS equation with box-type initial conditions on a nonzero background, and demonstrates accuracy in both asymptotic regions of the (x,t)-plane.","keywords":["numerical inverse scattering transform","defocusing nonlinear Schrödinger equation","nonzero background","box-type initial condition","Riemann-Hilbert problem","nonlinear steepest descent","Chebyshev collocation","solitonless regime"],"falsifier":"Evaluate the Wronskian $W(\\hat m_2(q_o),\\hat m_1(-q_o))$ from Eq. (167) on a fine grid of $(x,t)$ with $t>0$ for the box parameters (59); a single zero would mean RHP 5 has no solution there and the reconstructed $q(x,t)$ cannot be trusted. A sharper version is to test $\\theta=\\pi/2$, where the paper itself shows the solvability condition fails at $t=0$, and look for failure at small $t>0$.","tokens_in":50971,"feed_emoji":"🌊","tokens_out":8837,"duration_ms":422844,"temperature":0.7,"pith_summary":"The paper sets out to solve the Cauchy problem for the defocusing nonlinear Schrödinger equation when the initial data are piecewise constant (\"box-shaped\") and the solution approaches a constant nonzero background $q_o>0$ at spatial infinity. It does this by implementing the inverse scattering transform numerically, reducing the inverse step to a Riemann–Hilbert problem, a boundary-value problem for an analytic matrix function, which is solved with contour deformations in the spirit of the Deift–Zhou steepest descent method. The authors restrict to box parameters that produce no discrete spectrum (no solitons) and demonstrate that the computed $q(x,t)$ is accurate in both asymptotic regimes, $|x/(2t)|<q_o$ and $|x/(2t)|>q_o$, as $t\\to\\infty$. The payoff is that evaluating the solution at a point costs seconds rather than the hours needed by a Fourier time-stepping scheme with over a million modes.","feed_headline":"Numerical IST computes box-type NLS waves on a nonzero background","feed_subtitle":"Soliton-free box data solved by deformed Riemann-Hilbert contours, accurate in both asymptotic regions in seconds.","key_machinery":"The load-bearing object is the deformed Riemann–Hilbert problem RHP 5 for the transformed unknown $\\tilde m(z)$ in the solitonic region and its analogue in the solitonless region. RHP 5 is obtained after three moves: opening lenses along rays where the oscillatory exponentials $\\exp(\\pm 2ti\\Theta(\\xi,z))$ decay; solving a scalar Riemann–Hilbert problem for $\\delta(z)$ that absorbs the jump on part of the real axis; and multiplying by $E(z)=I_2 + q_o \\sigma_2 e^{-i\\theta\\sigma_3} z^{-1} \\tilde m(0)^{-1}$ to remove the $1/z$ singularity at $z=0$. The payoff is a regular Riemann–Hilbert problem with determinant one, whose jump matrices are exponentially close to the identity for large $t$, so Chebyshev collocation converges rapidly and the computational cost is essentially independent of $x$ and $t$.","core_discovery":"The central discovery is that the inverse scattering transform for the defocusing NLS equation on a nonzero background can be made fully numerical even when the initial data are discontinuous and do not decay at infinity, provided the scattering data are known explicitly and no solitons are present. Working in the uniformization variable $z=k+\\lambda$, $\\lambda=\\sqrt{k^2-q_o^2}$, removes the square-root branching of the spectral problem. The paper then applies a sequence of lens-opening deformations, a scalar Riemann–Hilbert problem for a function $\\delta$ that removes the jump on part of the real axis, and a final transformation $E(z)$ that removes the remaining singularity at $z=0$; the resulting regular Riemann–Hilbert problem for the unknown $\\tilde m$ is solved by Chebyshev collocation. Reconstructing $q(x,t)$ from the large-$z$ behavior of $\\tilde m$ yields the solution at any $(x,t)$. The method is verified by residual checks against the NLS equation, by Cauchy-error convergence in the collocation degree, and by comparison with a heavily damped Fourier method; the presented runs reach a residual of $2.5\\times 10^{-5}$ at $(x,t)=(-11,5)$ and show the claimed seconds-scale cost.","pith_inferences":["If RHP 5 is later proved to have a solution for all $t>0$ (currently open), the method would become a fully rigorous numerical IST for this class of data; until then, a zero of the solvability Wronskian would mark a genuine boundary of the method's validity.","The same contour-deformation and singularity-removal machinery should extend to initial data with a few solitons, by converting residue conditions into jumps, and to smooth non-decaying data, at the cost of adding a $\\bar\\partial$-problem for the reflection coefficient.","Because evaluation cost is independent of $x$ and $t$, the method is well suited to generating tables of \"nonlinear special functions\" for the defocusing NLS, analogous to the role numerical Riemann–Hilbert solvers play for Painlevé transcendents.","A direct test of the open existence question would be to run the solver for phases near $\\theta=\\pi/2$, where the paper shows the solvability condition already fails at $t=0$, and see whether failures persist at small $t>0$."],"forward_implications":["For any box initial condition free of discrete spectrum, the solution $q(x,t)$ can be evaluated at a single point in seconds, with accuracy that improves as $t$ grows because the deformed jumps decay exponentially.","The approach removes the need to resolve the nonlinear Gibbs phenomenon: discontinuous initial data are handled directly, whereas a Fourier method required a period of length 800 and over a million modes to reach $10^{-3}$ accuracy at $t=1$.","The solvability condition $W(\\hat m_2(q_o),\\hat m_1(-q_o))\\neq 0$ can be checked after each computation, giving a built-in test of whether the reconstructed $q(x,t)$ is trustworthy at that $(x,t)$.","The same deformed-RHP construction works in both asymptotic regions $|x/(2t)|<q_o$ and $|x/(2t)|>q_o$, and the numerical solutions from the two regions agree across the collisionless-shock boundary for bounded times."],"supporting_citations":[{"why":"Supplies the long-time asymptotic analysis in the solitonic region and the weighted Sobolev setting used to state the IST.","marker":"[6]"},{"why":"Supplies the long-time asymptotics in the solitonless region, the template for the two-stationary-phase-point deformations.","marker":"[38]"},{"why":"Provides the explicit scattering coefficients for a box initial condition and the conditions on parameters that exclude discrete eigenvalues.","marker":"[5]"},{"why":"Introduces the uniformization variable that removes spectral branching, on which the whole formulation rests.","marker":"[17]"},{"why":"Contributes the contour-deformation technique for logarithmic singularities that is generalized here to the points ±q_o.","marker":"[4]"},{"why":"Establishes the Chebyshev collocation framework for solving singular integral equations equivalent to Riemann-Hilbert problems.","marker":"[27]"},{"why":"Provides the numerical Riemann-Hilbert theory, including convergence criteria, used to justify the collocation solver.","marker":"[33]"},{"why":"Demonstrates the numerical nonlinear steepest descent approach for decaying data that this paper adapts to non-decaying backgrounds.","marker":"[29]"},{"why":"Supplies the artificial-damping Fourier method used as the baseline comparison for accuracy and runtime.","marker":"[23]"}],"fun_headline_variants":["Numerical IST for box-type NLS on nonzero background","Soliton-free box waves solved by numerical IST","Deformed RH contours solve box NLS on background","Accurate numerical IST for defocusing NLS with box IC","Numerical inverse scattering for box-type NLS on nonzero background"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the deformed Riemann–Hilbert problem for the final unknown $\\tilde m$ has a solution for every $t>0$ in the tested regions; the paper proves this only at $t=0$ and otherwise checks the solvability condition numerically after the fact.","fun_headline_variants_meta":{"raw":{"variants":["Numerical IST for box-type NLS on nonzero background","Soliton-free box waves solved by numerical IST","Deformed RH contours solve box NLS on background","Accurate numerical IST for defocusing NLS with box IC","Numerical inverse scattering for box-type NLS on nonzero background"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000415,"raw_usage":{"total_tokens":2166,"prompt_tokens":994,"completion_tokens":1172,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":610,"completion_tokens_details":{"reasoning_tokens":1090}},"tokens_in":610,"tokens_out":1172,"duration_ms":9470,"temperature":1.0,"reasoning_tokens":1090,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T23:57:29.695885+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the Wronskian $W(\\hat m_2(q_o),\\hat m_1(-q_o))$ from Eq. (167) on a fine grid of $(x,t)$ with $t>0$ for the box parameters (59); a single zero would mean RHP 5 has no solution there and the reconstructed $q(x,t)$ cannot be trusted. A sharper version is to test $\\theta=\\pi/2$, where the paper itself shows the solvability condition fails at $t=0$, and look for failure at small $t>0$.","supporting_citations":[{"cited_title":"On the asymptotic stability of𝑁-soliton solutions of the defocusing nonlinear Schr¨dinger equation","cited_arxiv_id":null,"evidence_quote":"Supplies the long-time asymptotic analysis in the solitonic region and the weighted Sobolev setting used to state the IST."},{"cited_title":"DefocusingNLSequationwithnonzerobackground:Large-timeasymptoticsinasolitonlessregion","cited_arxiv_id":null,"evidence_quote":"Supplies the long-time asymptotics in the solitonless region, the template for the two-stationary-phase-point deformations."},{"cited_title":"OnthespectrumoftheDiracoperatorandtheexistenceofdiscreteeigenvaluesforthedefocusing nonlinear Schr¨dinger equation","cited_arxiv_id":null,"evidence_quote":"Provides the explicit scattering coefficients for a box initial condition and the conditions on parameters that exclude discrete eigenvalues."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the uniformization variable that removes spectral branching, on which the whole formulation rests."},{"cited_title":"Numerical Inverse Scattering for the Toda lattice","cited_arxiv_id":null,"evidence_quote":"Contributes the contour-deformation technique for logarithmic singularities that is generalized here to the points ±q_o."},{"cited_title":"A general framework for solving Riemann–Hilbert problems numerically","cited_arxiv_id":null,"evidence_quote":"Establishes the Chebyshev collocation framework for solving singular integral equations equivalent to Riemann-Hilbert problems."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the numerical Riemann-Hilbert theory, including convergence criteria, used to justify the collocation solver."},{"cited_title":"Nonlinear steepest descent and the numerical solution of Riemann–Hilbert problems","cited_arxiv_id":null,"evidence_quote":"Demonstrates the numerical nonlinear steepest descent approach for decaying data that this paper adapts to non-decaying backgrounds."},{"cited_title":"Anartificially-dampedFouriermethodfordispersiveevolutionequations","cited_arxiv_id":null,"evidence_quote":"Supplies the artificial-damping Fourier method used as the baseline comparison for accuracy and runtime."}],"review_version":1}