{"id":"e7833665-8e4c-4263-afd5-5de3973fdc83","arxiv_id":"2506.22345","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A Carleman linearization maps 1D shallow water equations to a linear system for quantum solvers, but validation is limited to small-amplitude test cases and the speedup remains conditional.","lead":"The authors adapt a quantum-friendly way of rewriting fluid equations to the shallow water equations, turning the nonlinear problem into a large linear system. They test it on tiny classical simulations and study how the cost of a quantum solver scales, but the promised exponential speedup is not demonstrated end to end.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Missing Carleman closure-error analysis: truncation at order 3 drops couplings to orders 4 and 5, and the benchmark only tests the linear acoustic branch, so the claim that the linear system captures SWE dynamics is unverified.","rationale":"The reader's weakest_assumption identifies the same gap I consider most load-bearing: the unquantified Carleman closure error. Section IV.A.2 explicitly claims the truncation error reduces to the Taylor error of 1/h, but this is incorrect for any nonlinear polynomial system of degree at least 2: the k-th Carleman variable couples to orders k+1 and k+2, so truncating at k=3 introduces an additional error that is not estimated. The acoustic wave-speed benchmark only exercises the linear terms and cannot reveal this error. A direct convergence test over k would settle the matter. Because the paper is explicitly a proof-of-principle and the authors acknowledge the small-height restriction, the conditional verdict is appropriate; no verdict change is needed.","tokens_in":15375,"tokens_out":10637,"duration_ms":111019,"concrete_test":"Recompute the step-perturbation benchmark of Section VII.B with Carleman truncation orders k=3, 4, and 5, and also run a genuinely nonlinear SWE test (e.g., a small-amplitude dam break or solitary wave with |1-h| <= 0.05) against a standard finite-volume SWE solver. If the height and velocity profiles converge as k increases and agree with the reference solver, the closure error is negligible and the central claim holds. If the k=3 result differs from the k=5 result or the reference by more than a few percent, the mapping does not solve the SWE in the advertised regime.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's error analysis in Section IV.A.2 conflates two distinct sources of error. After replacing 1/h by the first-order Taylor polynomial 2 - h (Eq. 16), the DVBE collision operator becomes a polynomial of degree 3 in f. The authors state that 'truncation of the Carlemann matrices is reduced to truncation of this Taylor polynomial' and conclude an error O(|1-h|). For any polynomial system of degree d >= 2, the Carleman hierarchy does not close at order 3: the equation for the third-order Carleman variable V^(3) couples to V^(4) and V^(5) through the F^(2) and F^(3) terms in Eq. (14). Truncating at k=3 discards these couplings, introducing a closure error absent from the Taylor-error estimate. The exponential error bound attributed to Itani and Succi (Ref. 17) is not specialized to this equilibrium, and its hypotheses are never checked. Separately, the numerical validation in Section VII.B measures only the acoustic wave speed sqrt(g h0), which is governed by the linear terms; this test cannot detect errors in the nonlinear h u^2 coupling. The central claim that the linear system 'can reliably demonstrate key features of the SWE' therefore rests on an unverified assumption that the closure error is negligible in the tested regime.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper adapts a Carleman-linearization linearization scheme, originally developed for the Navier-Stokes equations by Li et al. (Phys. Rev. Research 7, 013036 (2025)), to the one-dimensional shallow water equations. The authors introduce a D1Q3 discrete velocity Boltzmann equilibrium, replace the 1/h nonlinearity by a first-order Taylor polynomial, truncate the resulting Carleman hierarchy at order 3, and convert the resulting ODE into a linear system via forward Euler. They then solve the linear system classically and benchmark it against the analytical small-amplitude wave speed v = sqrt(g h0). They also implement a QSVT-based quantum linear system solver and report scaling of the condition number with timesteps and grid points.","tokens_in":15681,"tokens_out":7796,"duration_ms":80021,"significance":"If the mapping is correct, this would be a useful extension of quantum CFD techniques from the Navier-Stokes equations to shallow water flows and would support the broader program of solving nonlinear fluid PDEs on quantum computers. The paper has several concrete strengths: it provides reproducible code on GitHub, verifies the QSVT polynomial-degree scaling, and compares against an external analytical prediction rather than fitting parameters. However, the load-bearing error analysis is incomplete and the numerical validation exercises only the linear acoustic branch, so the central claim that the scheme 'can reliably demonstrate key features of the SWE' is not yet established.","major_comments":[{"comment":"The error estimate O(|1-h|^(k-2)) accounts only for the Taylor truncation of 1/h, not for the Carleman closure error. After replacing 1/h by 2-h, the collision operator becomes a polynomial of degree 3 in f; the equation for the third-order Carleman variable V^(3) therefore couples to V^(4) and V^(5) through the F^(2) and F^(3) terms in Eq. (14). Truncating at k=3 discards these couplings. The exponential error bound attributed to Itani and Succi is not specialized to this equilibrium, and its hypotheses are not checked. Consequently, the claim that the Carleman-truncated linear system solves the SWE in the benchmarked regime is unverified.","section":"§IV.A.2 and Eq. (14)"},{"comment":"The benchmark measures only the linear acoustic-wave speed v = sqrt(g h0) with a height perturbation δh0 = 0.01 h0. This test is governed by the linear terms of the equation and cannot detect errors in the nonlinear hu^2 coupling. The statement in §VII.B that the scheme 'can reliably demonstrate key features of the SWE' is therefore stronger than the evidence supports. A finite-amplitude test, such as a dam-break profile or a nonlinear wave steepening case, is needed to validate the nonlinear part of the model.","section":"§VII.B and Fig. 3"},{"comment":"The initial condition stated as f_eq in Eq. (33) does not equal the equilibrium distribution of Eq. (9) evaluated at the same h and u under the stated D1Q3 weights. Eq. (33) omits the gh/2 and u^2 terms present in Eq. (9), so, unless an unstated low-Froude approximation is intended, the initial state is not the equilibrium of the collision operator. In addition, the weights in Eq. (10) sum to 4/3 rather than 1, contradicting the conservation condition in Eq. (4); presumably the intended D1Q3 weights are (2/3, 1/6, 1/6). Since the collision matrices F^(j) are derived from this equilibrium, this discrepancy is load-bearing and must be resolved.","section":"§V.B, Eq. (33), and §III, Eq. (10)"},{"comment":"The recovery of the shallow water equations from the D1Q3 equilibrium with τ = 2ν/gh is asserted without a derivation or a precise reference, and the explicit forms of the matrices F^(1), F^(2), and F^(3) for this lattice are never given. The reader cannot verify that the Carleman collision matrices in Eq. (18) implement the intended equilibrium, nor reproduce the reported benchmarks from the text alone. The explicit F^(j) for D1Q3, or a supplementary derivation, should be provided.","section":"§III and §IV.A.3"}],"minor_comments":[{"comment":"The approximation in Eq. (16) requires h to be normalized by a reference depth, but the normalization is never defined; Section VII.A quotes heights in meters, which makes the O(|1-h|) error statement dimensionally ambiguous.","section":"§IV.A.2"},{"comment":"The description of the relative error as 'the sum of the deviations at each grid point, normalized by the total number of grid points' is unclear; the exact norm used should be specified (e.g., L1, L2, or maximum norm).","section":"§VII.A"},{"comment":"The phrase 'Normal models propagation speeds' in the section title should read 'Normal modes'.","section":"§VII.B"},{"comment":"The caption contains the typo 'T imesteps'; also, the symbol Nt is used for the number of timesteps in Section VI but the total number of timesteps is denoted n in Eq. (41), which is confusing.","section":"Fig. 1 and §IV.B"},{"comment":"The statement that κ is 'almost independent of the number of grid points' is based on only 3, 4, 5, and 6 grid points; this is too narrow a range to establish a scaling behavior, and the text appropriately notes the lack of explanation.","section":"§VIII.C and Fig. 5(c)"},{"comment":"The GitHub repository is cited without a version or commit identifier, which complicates reproducibility for the exact plots shown in the paper.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The manuscript reads partly as a workshop or course project, and the novelty over Li et al. is incremental. However, the central mapping is plausibly correct in the small-amplitude limit, so major revision is appropriate rather than rejection. The main issues to fix are the Carleman closure-error analysis, the equilibrium/initial-condition inconsistency, and a nonlinear validation test. I did not find evidence of circular fitting; the v = sqrt(g h0) benchmark is an external analytical prediction."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the paper does a real new thing—it takes Li et al.'s Carleman–LBM linearization, originally for the Navier–Stokes equations, and ports it to the 1D shallow water equations. The new bit that matters is the treatment of 1/h in hu^2: they Taylor-expand 1/h to first order, which caps the polynomial order in f and sets the Carleman truncation at 3. That observation is correct and is the actual contribution; it also explains honestly why the scheme is limited to small height perturbations. The paper is candid about the limits: 1D, linear regime, forward Euler, and it flags the readout and encoding costs that condition any real speedup.\n\nWhat is good: the derivation of the collision and streaming matrices is sketched rather than fully proven, but the structure is plausible and consistent with the appendix; the code is on GitHub; the QSVT condition-number study, though tiny, is a fair proof of principle and confirms known scalings. The measured v = sqrt(gh0) for the acoustic step is the right external check for a linear regime.\n\nThe soft spot is real and matches the stress-test note. Section IV.A.2's error analysis conflates two different truncations. Replacing 1/h by 2–h makes the DVBE collision term polynomial of degree 3 in f, but the Carleman equations do not close at order 3: V^(3) couples to V^(4) and V^(5), and truncating discards those couplings. The paper says the error is O(|1–h|) from the Taylor polynomial and cites the exponential Itani–Succi bound, but it never checks the hypotheses of that bound for this equilibrium. More importantly, the benchmark in VII.B only measures the linear acoustic wave speed. That test is blind to the nonlinear h u^2 coupling that the whole Carleman machinery is supposed to capture. The claim that the linear system 'can reliably demonstrate key features of the SWE' is therefore not yet demonstrated. A dam-break or nonlinear solitary-wave test on a few more grid points would have told us much more.\n\nMinor points: the recovery of the SWE from the D1Q3 equilibrium is asserted without the multiscale analysis; the stable-configuration test only checks drift, not dynamics; the condition-number study stops at 6 grid points. None of these are disqualifying.\n\nVerdict: a legitimate, clearly written contribution that deserves referee time, not a desk reject. The closure-error analysis and a nonlinear validation test are needed before I would believe the mapping captures anything beyond the linear regime. Send it out.","headline":"Legitimate Carleman–LBM adaptation to the shallow water equations; the linear-regime validation is too thin to support the 'reliably demonstrate' claim, but the issues are addressable.","tokens_in":16194,"tokens_out":2160,"would_cite":false,"duration_ms":23266,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["76B15","35Q35","81P68"],"pacs":[],"model":"deepseek-v4-flash","headline":"By writing the nonlinear shallow water equations as discrete-velocity Boltzmann equations and truncating a Carleman linearization at order 3, this paper produces a linear system that a quantum computer could in principle invert…","keywords":["shallow water equations","Carleman linearization","quantum linear system solvers","discrete velocity Boltzmann equations","quantum singular value transformation","computational fluid dynamics","D1Q3 lattice","tsunami modeling"],"falsifier":"Re-run the step-perturbation benchmark with $\\delta h_0 = 0.01 h_0$, four grid points, and four timesteps at Carleman truncation orders 4 and 5; if the measured wave speed deviates from $\\sqrt{g h_0}$ by more than the finite-difference discretization error, or if the order-3 results change materially when higher-order Carleman variables are retained, then the order-3 closure is not capturing the discrete-velocity Boltzmann dynamics.","tokens_in":15197,"feed_emoji":"🌊","tokens_out":8074,"duration_ms":78411,"temperature":0.7,"pith_summary":"The paper claims that the nonlinear shallow water equations can be rewritten as a large but structured linear system of equations, making them accessible to quantum linear-system solvers that are exponentially faster than classical solvers in principle. The route is a Carleman linearization of the discrete-velocity Boltzmann equations, extended from a prior Navier-Stokes scheme, truncated at order 3 under the assumption that water height deviates only slightly from a reference depth. Benchmarks on a one-dimensional D1Q3 lattice reproduce the analytical wave speed $\\sqrt{g h_0}$ for small perturbations, and a proof-of-principle quantum singular value transformation solver shows the condition number grows linearly with time steps but stays nearly flat with grid resolution. The practical value is that a class of geophysical flows, such as open-ocean tsunami propagation, enters the range of quantum simulation, with the caveat that data loading and readout costs remain open.","feed_headline":"Shallow water equations recast as a quantum linear system","feed_subtitle":"Small-height tsunami flows become linear problems quantum solvers can solve fast.","key_machinery":"The central mechanism is Carleman linearization, a change of variables $V = (f, f \\otimes f, f \\otimes f \\otimes f, \\dots)$ that converts the polynomial nonlinearity of the Boltzmann collision operator into a linear ODE in a higher-dimensional vector space. A D1Q3 lattice, with three discrete velocities $c_i \\in \\{-1, 0, 1\\}$, supplies the discrete-velocity Boltzmann equations whose equilibrium distribution recovers the shallow water equations in the hydrodynamic limit. Truncation at order 3 keeps Carleman variables up to cubic products of the distribution functions, which corresponds to a first-order Taylor approximation $1/h \\approx 2-h$; this is the step that restricts validity to small height variations. Forward Euler time stepping turns the linear ODE into a sparse block-bidiagonal matrix $E$, and the resulting linear system $A x = b$ is what a quantum linear-system solver such as quantum singular value transformation would invert. The condition number $\\kappa$ of that matrix is the key cost driver, and the paper measures its scaling with timesteps and grid points.","core_discovery":"The core claim is that the one-dimensional shallow water equations, despite being nonlinear hyperbolic conservation laws, can be mapped, up to a controlled truncation, to a linear system of ordinary differential equations by first writing them as discrete-velocity Boltzmann equations and then applying Carleman linearization at order 3. Because the equilibrium distribution contains the term $h u^2$, the substitution of statistical moments requires a Taylor expansion of $1/h$; with truncation order 3 only the linear term $2-h$ is retained, which is accurate when $|1-h| \\ll 1$. The benchmarks, using a step perturbation in height, recover the propagation speed $v = \\sqrt{g h_0}$ predicted by linearized shallow-water analysis, and the associated quantum-singular-value-transformation condition-number studies show no bottleneck in the quantum part of the pipeline. The authors present this as extending the previously developed linearization approach for Navier-Stokes to the shallow water equations and as evidence that such linearization schemes can work beyond a single family of PDEs.","pith_inferences":["The exponential speedup claim concerns the quantum inversion step; the paper itself notes that state preparation and readout can offset the advantage, so an end-to-end speedup for realistic wave problems remains an open question.","A direct test of the truncation's validity would be to run the same step-perturbation benchmark at Carleman orders 4 and 5; if the recovered speed changes significantly, the order-3 closure is not the limiting error and the paper's error estimate is incomplete.","The near-constant condition number with grid points is unexplained; if it persists at larger resolutions, it would make the approach unusually attractive, but the current data, covering only 3 to 6 grid points, is too sparse to extrapolate.","The linearized system could be compared against a standard classical finite-volume shallow water solver at identical resolution to see whether the scheme is competitive classically before quantum hardware is needed."],"forward_implications":["If the mapping is correct, tsunami propagation with wave heights small compared to ocean depth can be phrased as a linear algebra problem, so quantum linear-system solvers become a candidate tool for that class of simulations.","The condition-number benchmarks imply that in the tested regime, the cost of quantum singular value transformation grows only linearly with the number of time steps and does not worsen with grid resolution, which is favorable for quantum speedup in the inversion step.","The same Carleman-plus-discrete-velocity-Boltzmann pipeline should apply to any conservation law whose discrete-velocity equilibrium is a low-order polynomial in the conserved moments, not only the shallow water equations.","Because only a small subset of Carleman variables corresponds to physical height and velocity, readout can be restricted to a few grid points or global observables rather than requiring full state tomography.","For larger height variations, the truncation order must be increased, raising the dimension of the Carleman vector exponentially; the paper positions this as a tunable accuracy-versus-cost tradeoff."],"supporting_citations":[{"why":"Supplies the original Carleman linearization of the lattice Boltzmann equations for Navier-Stokes that this paper extends to the shallow water equations.","marker":"[14]"},{"why":"Cited for the result that the Carleman truncation error improves exponentially with truncation order, underlying the error discussion in Section IV A 2.","marker":"[17]"},{"why":"Introduced the quantum linear-system solver whose exponential speedup motivates mapping the shallow water equations to a linear system.","marker":"[2]"},{"why":"Provides the high-order quantum algorithm for linear ordinary differential equations and the theoretical prediction that the condition number scales linearly with the number of time steps.","marker":"[4]"},{"why":"Constructs the shallow water lattice Boltzmann equilibrium used in the D1Q3 reduction, ensuring the discrete-velocity Boltzmann equations recover the shallow water equations.","marker":"[22]"},{"why":"Gives the analytical speed of sound $v = \\sqrt{g h_0}$ used as the benchmark for wave propagation.","marker":"[26]"},{"why":"Defines quantum singular value transformation, the quantum linear-system solver implemented as a proof of principle.","marker":"[30]"},{"why":"Cited for the $O(\\kappa \\log \\kappa)$ polynomial degree scaling and for sparse-matrix optimizations to block encoding.","marker":"[28]"},{"why":"Provides the numerical package used to approximate the inverse function and to study condition-number scaling.","marker":"[29]"}],"fun_headline_variants":["Tsunami equations linearized for quantum computing","Quantum-ready shallow water dynamics from a linearization","Navier-Stokes linearization trick extends to shallow water","Shallow water PDEs become quantum linear systems"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The mapping assumes that truncating the Carleman expansion at order 3, together with the linear approximation $1/h \\approx 2-h$, captures the full Boltzmann dynamics for the tested small height variations; the paper quantifies only the Taylor-expansion error, not the error from dropping order-4 and higher Carleman couplings.","fun_headline_variants_meta":{"raw":{"variants":["Tsunami equations linearized for quantum computing","Quantum-ready shallow water dynamics from a linearization","Navier-Stokes linearization trick extends to shallow water","Shallow water PDEs become quantum linear systems"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000964,"raw_usage":{"total_tokens":4094,"prompt_tokens":926,"completion_tokens":3168,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":542,"completion_tokens_details":{"reasoning_tokens":3108}},"tokens_in":542,"tokens_out":3168,"duration_ms":25846,"temperature":1.0,"reasoning_tokens":3108,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T22:06:29.798526+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the step-perturbation benchmark with $\\delta h_0 = 0.01 h_0$, four grid points, and four timesteps at Carleman truncation orders 4 and 5; if the measured wave speed deviates from $\\sqrt{g h_0}$ by more than the finite-difference discretization error, or if the order-3 results change materially when higher-order Carleman variables are retained, then the order-3 closure is not capturing the discrete-velocity Boltzmann dynamics.","supporting_citations":[{"cited_title":"Simulating fluid flows with quantum computing","cited_arxiv_id":"2409.09736","evidence_quote":"Cited for the result that the Carleman truncation error improves exponentially with truncation order, underlying the error discussion in Section IV A 2."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduced the quantum linear-system solver whose exponential speedup motivates mapping the shallow water equations to a linear system."},{"cited_title":"(3), the streaming matrix is built by first defining a matrix S such that: Sf = −c · ∇f (20) With x as the partial coordinate, in 1D this is equivalent to Sf =  ","cited_arxiv_id":null,"evidence_quote":"Provides the high-order quantum algorithm for linear ordinary differential equations and the theoretical prediction that the condition number scales linearly with the number of time steps."},{"cited_title":"Vadasz, Rendering the navier–stokes equations for a compressible fluid into the schr¨ odinger equation for quan- tum mechanics, Fluids 1 (2016)","cited_arxiv_id":null,"evidence_quote":"Constructs the shallow water lattice Boltzmann equilibrium used in the D1Q3 reduction, ensuring the discrete-velocity Boltzmann equations recover the shallow water equations."},{"cited_title":"Kinnmark, The Shallow Water Wave Equations: For- mulation, Analysis and Application(Springer, 1986)","cited_arxiv_id":null,"evidence_quote":"Defines quantum singular value transformation, the quantum linear-system solver implemented as a proof of principle."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the numerical package used to approximate the inverse function and to study condition-number scaling."}],"review_version":1}