{"id":"02312fdf-0afc-4dbf-8802-32cbb1769007","arxiv_id":"2608.09675","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"An explicit quantum circuit implementation, built from level-set linearization and Schrödingerisation, estimates observables of nonlinear scalar conservation laws with complexity that beats classical finite differences in high dimensions.","lead":"This paper gives a step-by-step quantum circuit recipe for solving nonlinear conservation laws by first turning them into a linear problem and then simulating that linear problem on a quantum computer. It argues that for high-dimensional problems, the quantum version can estimate observables with fewer operations than classical solvers.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Algorithm's post-shock observable is a branch average of the level-set distribution, not the entropy solution; the quantum-advantage claim may only hold for smooth single-valued times.","rationale":"The reader's weakest assumption (state preparation, p-value oracles, and readout) is real, but it is explicitly listed in Remark 5 as an assumption and can be discharged in a standard oracle model; it affects the complexity comparison, not the identity of the quantity being computed. The shock/correctness issue is more load-bearing because it questions whether the algorithm solves the problem it claims to solve. In the level-set method for scalar conservation laws, after wave breaking the physically meaningful solution is the entropy solution; the branch-average moment produced by the quantum observable circuit is a different object. The paper gives no entropy-selection step, and its own description of the multi-valued observable in Section 3.3 acknowledges the average nature of the output. The numerical validation deliberately stays in the smooth regime, so it does not detect this gap. A single numerical experiment crossing the shock time would settle the concern. If the check shows a discrepancy, the appropriate fix is to scope the claim to smooth solutions, add an entropy-selection mechanism, or prove that the chosen moments coincide with the physically relevant observables. Since both the reader and this pass identify addressable issues rather than a mathematical contradiction, the CONDITIONAL verdict remains appropriate, hence UNCHANGED.","tokens_in":19594,"tokens_out":17816,"duration_ms":160741,"concrete_test":"Evolve 1D Burgers u_t + u u_x = 0 on [0,2π] with u_0(x) = sin(x) to T = 2, well past the shock formation time t ≈ 1. Run the full level-set/Schrödingerisation recovery of Section 3 (a classical simulation of the same circuit suffices) and compute ⟨p⟩_ψ(T,x) from Eq. (35). Compare pointwise with the entropy solution from a standard finite-volume Godunov solver. If the L1 error over the shock interval is not small, the algorithm does not estimate physical observables of the conservation law after shock, and the central claim must be restricted to the pre-shock/single-valued regime.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing issue is correctness in the multi-valued regime, not a complexity caveat. The algorithm defines its output, Eq. (35), as a moment of the level-set distribution, ⟨G⟩_j = (1/N_p)Σ_l G(p_l)ψ_{j,l}(T). For single-valued solutions ψ = δ(p−u) and this returns G(u). After shock formation the scalar conservation law has a unique entropy solution, while the level-set equation (4) develops multiple roots p_ν and ψ has the branch representation (6): ψ = Σ_ν |∂_pφ(p_ν)|^{-1}δ(p−p_ν). The paper explicitly states (Section 3.3) that the observable then represents a Jacobian-weighted ensemble average over all characteristic branches. That average is not the entropy solution of the conservation law, because entropy selection (Rankine–Hugoniot/Oleinik conditions) appears nowhere in the algorithm. Thus Theorem 2 bounds the cost of estimating an observable of the lifted multivalued problem, not an observable of the original nonlinear scalar conservation law in the regime where shocks exist. The numerical experiments in Section 5 use smooth initial data and times before shock formation, so they never test the case the level-set method is designed for. This is a correctness gap: even with free state preparation and readout, the method can compute the wrong physical quantity after wave breaking.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript constructs an end-to-end quantum algorithm for scalar conservation laws by combining a level-set lifting to a linear Liouville equation, upwind finite-difference discretization, Schrödingerisation into a higher-dimensional Hamiltonian system, and explicit Trotterized circuits for the resulting evolution. It derives error bounds and gate-complexity estimates, with the main result in Theorem 2 stating that estimating an observable to precision ε costs eO(d^4 T^4 N_p N_w^2 ε^{-4}) gates, and Remark 5 uses this to claim a quantum advantage over classical finite-difference methods in sufficiently high spatial dimension. Numerical experiments for a one-dimensional traffic-flow flux and a two-dimensional Burgers equation, both with smooth data and before shock formation, are reported to validate accuracy and scaling.","tokens_in":19876,"tokens_out":10698,"duration_ms":111329,"significance":"If the algorithm computed the entropy solution of the scalar conservation law, the paper would be a valuable contribution: it gives an explicit circuit-level construction, imports and assembles Trotter error bounds in a coherent way, and improves the mollifier convergence by Richardson extrapolation. The use of the one-dimensional bounds from [21] as independent support for the p-multiplexed Hamiltonian is legitimate and not circular. However, the paper's central claim is compromised by a correctness gap in the shock regime: after wave breaking, the observable defined in Eq. (35) is a Jacobian-weighted average over all characteristic branches, and entropy selection conditions are absent. Since the numerical experiments are deliberately restricted to smooth solutions before shocks, they do not validate the claimed solution of the original nonlinear conservation law. The complexity result is also not yet an end-to-end bound because state preparation, oracle implementation, and measurement sampling are not costed. These issues are load-bearing for the claimed quantum advantage.","major_comments":[{"comment":"The algorithm does not compute the entropy solution of the conservation law after shock formation. By Eq. (6), for multivalued times the level-set distribution is ψ(t,x,p)=Σ_ν |∂_p φ(p_ν)|^{-1}δ(p−p_ν), and Eq. (35) defines the recovered observable as ⟨G⟩_j = (1/N_p)Σ_l G(p_l)ψ_{j,l}(T), which is the Jacobian-weighted ensemble average over all characteristic branches. The entropy selection conditions (Rankine–Hugoniot and Oleinik conditions) do not appear anywhere in the algorithm, so this branch average is not G(u_entropy(T,x_j)). Consequently Theorem 2 bounds the cost of estimating an observable of the lifted multivalued problem, not of the original nonlinear scalar conservation law in the regime where shocks exist. The numerical experiments in Section 5 use smooth initial data and times before shock formation (Section 5.1 explicitly states that the solution remains smooth over the tested interval), so they do not test the regime in which the level-set method is needed. This is a load-bearing correctness gap for the paper's central claim of quantum advantage for nonlinear scalar conservation laws.","section":"Section 3.3, Eqs. (6) and (35)"},{"comment":"The sign conventions in the Hermitian/skew-Hermitian decomposition and the Schrödingerised evolution are internally inconsistent. Equation (21) defines A_1 = (1/2)⊕_{α,l}(P^-_{α,l}−P^+_{α,l})(D_+−D_-) and A_2 = (1/(2i))⊕_{α,l}(P^-_{α,l}+P^+_{α,l})(D_++D_-). Equation (23) states ∂_t v̂ = i(ηA_1+A_2)v̂ = iH_C v̂, while Eq. (24) says d|v̂>/dt = −iH_C|v̂>. Equation (27) then sets U_C(τ)=exp(iH_Cτ). Moreover, the relative sign between H_1^{(l)} in Eq. (26) and β_{1,α,l} in Eq. (60) is not matched to the sign of P^-−P^+ in Eq. (21). As written, a reader cannot determine whether the circuit implements the intended Liouville evolution or its negative-time/adjoint version. Since the paper's contribution is an explicit gate-level implementation, these sign inconsistencies must be resolved before the complexity and correctness claims can be verified.","section":"Section 3.2.2–3.2.3, Eqs. (21), (23), (24), (27)"},{"comment":"The stated total complexity in Theorem 2 does not account for the full end-to-end procedure. Appendix E counts only the Hamiltonian evolution gates from Lemma 4, and Remark 5 assumes that preparing the initial encoded state |Ψ(0)⟩ from u_0 and implementing the coefficient oracles are efficient. No circuit or cost estimate is given for preparing the amplitude-encoded mollified initial data over N_p N_x^d grid points, for implementing the value oracle O_η in Eq. (50), or for converting the postselected states in Eqs. (39) and (47) into a classical estimate of ⟨G⟩ to precision ε. If any of these costs scales with N_x^d or with the required number of measurement repetitions, the claimed exponential-in-d advantage over classical finite differences disappears. The paper should either provide explicit circuits and costs for these components or state the quantum-advantage conclusion as conditional on them throughout the theorem and remark.","section":"Appendix E and Remark 5"}],"minor_comments":[{"comment":"The balancing statement at the end of Appendix D is inconsistent: from ω∼(dht)^{1/4}, the error estimate ω^2 + dht/ω^2 gives ε=O((dht)^{1/2}), not O((dht)^{1/4}) as written. The earlier balancing in Section 3.3 using ω=O(ε^{1/2}) and h=O(ε^2/(dT)) also differs from the Appendix D derivation and should be reconciled.","section":"Appendix D"},{"comment":"In the definition of V_1^{(l)}[s], the variable is written as a∈Z, but it should be s∈Z to match the subsequent usage V_1^{(l)}[−2^{n_w−1}] and V_1^{(l)}[2^m]; this is a typographical error, not a mathematical issue.","section":"Eq. (31)"},{"comment":"The line “≤Ch^{1/ω^2}” appears to be a typesetting error; the intended bound is O(h/ω^2), consistent with the local truncation error O(h∂_{xx}ψ).","section":"Appendix D, Part III"},{"comment":"The zeroth-moment expression in Eq. (38) omits the explicit 1/√(N_p) normalization from the Hadamard transform in Eq. (37) and does not state the post-selection success probability; these factors are needed for a complete accounting of the measurement cost.","section":"Section 3.3, Eq. (38)"}],"recommendation":"reject","confidential_remarks":"The decisive issue for my recommendation is the correctness gap in the shock regime: the algorithm estimates a branch-averaged level-set observable, not the entropy solution, and the numerical experiments avoid shocks by design. This cannot be repaired by local edits within the manuscript's current scope. I also note that the complexity comparison in Remark 5 is conditional on unproven state-preparation and readout costs; this would be significant even if the shock problem were resolved. Finally, the manuscript relies heavily on the authors' own work [21] and on the UnitaryLab library [23]; I do not see circularity in the technical dependence, but the overlap in funding and tools may merit an editorial disclosure check."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe headline: this is the first explicit gate-level construction for a nonlinear PDE class using level-set lifting plus Schrödingerisation. That is real and worth referee time. The authors assemble known pieces — the level-set linearization, Schrödingerisation, shift-operator circuits, and Trotter bounds — into an end-to-end algorithm with a complexity estimate and numerical support. The p-register block-diagonal structure and the binary-controlled implementation for affine flux derivatives are the genuinely new engineering.\n\nWhat the paper does well: the error analysis is concrete. Theorem 1 gives a Trotter bound with commutator scaling, and the p-block independence means no extra N_p factor in the evolution error. The gate-count lemmas are explicit enough to cost. The numerics are honest: they compare the Trotter–Schrödingerisation circuit to a matrix-exponential reference and confirm the N_w convergence, including the predicted first-order Fourier error for e^{-|w|} and the improvement from a smoother profile.\n\nThe soft spot is not a complexity caveat; it is what the algorithm computes after shocks. The observable in Eq. (35) is a moment of the level-set distribution. Before wave breaking that moment is G(u(x,T)) and everything is fine. After wave breaking the level-set distribution is a Jacobian-weighted sum over all characteristic branches, and the moment is the branch average. That is not the entropy solution of the scalar conservation law. Rankine–Hugoniot and Oleinik entropy selection appear nowhere in the circuit, the recovery step, or the tests. The numerical experiments use smooth initial data and stop before shock formation, so they never exercise the regime where the level-set method is needed. So Theorem 2 bounds the cost of estimating an observable of the lifted linear problem, not an observable of the nonlinear conservation law at times when shocks exist. The paper does state the branch-average interpretation in Section 3.3, but it does not draw the consequence for the quantum-advantage claim. This is the main thing a referee should push on.\n\nMinor issues: the sign conventions in Eqs. (21) and (27) don't line up (likely typos, since the Trotter bound is sign-insensitive), and the balancing paragraph at the end of Appendix D writes ε=O((dht)^{1/4}) where the algebra gives (dht)^{1/2}. State preparation and readout costs are assumed away in Remark 5, which is standard but makes the advantage conditional.\n\nBottom line: this deserves a serious referee. It is a solid paper about simulating the lifted level-set equation and estimating its observables; it is not yet a solution of nonlinear conservation laws in the shock regime. The authors should either restrict the claim to smooth times or add an entropy-selection step.\n\nRecommendation: send to peer review, require major revision to clarify the post-shock meaning of the observable.","headline":"Explicit circuit construction is serious and the numerics check out, but the algorithm computes the level-set branch average, not the entropy solution, so the quantum-advantage claim only holds before shock formation.","tokens_in":20405,"tokens_out":6921,"would_cite":true,"duration_ms":60693,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35L65","65M06","81P68"],"pacs":[],"model":"deepseek-v4-flash","headline":"For scalar conservation laws, the paper constructs explicit quantum circuits whose observable-estimation complexity is $\\widetilde{O}(d^4T^4N_pN_w^2\\varepsilon^{-4})$, beating classical finite differences when the spatial dimension $d$ is…","keywords":["scalar conservation laws","quantum algorithm","level-set method","Liouville equation","Schrödingerisation","Hamiltonian simulation","Trotter product formula","observable estimation"],"falsifier":"Build the complete circuit that includes state preparation and oracle loading for a specific flux such as Burgers and count its gates as a function of $N_x^d$; if the preparation or readout cost is $\\Omega(N_x^d)$ or $\\Omega(N_pN_x^d)$, the end-to-end complexity is no longer polynomial in $d$ and the Remark 5 advantage over classical cost does not hold.","tokens_in":19390,"feed_emoji":"⚛️","tokens_out":6868,"duration_ms":61224,"temperature":0.7,"pith_summary":"The paper tries to establish that nonlinear scalar conservation laws, a class that includes Burgers and traffic-flow equations, can be solved by a quantum computer with explicit gate-level circuits. The route is to lift the nonlinear equation into a linear Liouville equation in an extra variable, discretize it by upwind finite differences, and embed the resulting linear ODE into a unitary Schr\\\"odinger evolution. The authors derive end-to-end error bounds and a total gate count of $\\widetilde{O}(d^4T^4N_pN_w^2\\varepsilon^{-4})$ for estimating observables, and they argue this beats the classical first-order finite-difference cost when the spatial dimension $d$ is large. Numerical experiments in one and two dimensions confirm the recovery formulas and the predicted scaling in the auxiliary Fourier resolution. This matters because a polynomial-in-$d$ gate count would remove the exponential-in-dimension bottleneck for high-dimensional transport problems, at least for coherent observable estimation.","feed_headline":"High-dimensional conservation laws gain quantum advantage in observable estimates","feed_subtitle":"Explicit Schrödingerisation circuits lift nonlinear PDEs to linear quantum evolution, with d-polynomial gate counts.","key_machinery":"The mechanism that carries the argument is the level-set lifting: represent $u(t,x)$ as the zero level set of $\\phi(t,x,p)$ and use the delta-concentrated distribution $\\psi=\\delta(\\phi)$, whose dynamics are the linear Liouville equation $\\partial_t\\psi+(\\nabla_p F(p))\\cdot\\nabla_x\\psi=0$. The semi-discrete system is block diagonal in $p$, so each level-set slice evolves independently; Schr\\\"odingerisation introduces an auxiliary frequency register and turns the nonunitary ODE into the Hermitian Hamiltonian $H_C=\\eta A_1+A_2$. The argument then flows through the tensor-product structure of $H_C$: each spatial dimension contributes shift operators implemented by the $V_1$ and $V_2$ circuits, the $p$-register and $w$-register supply controlled powers, and first-order Lie\\textendash{}Trotter steps with commutator bounds control the error. Observable recovery uses Hadamard summation and post-selection to read out moments without full state tomography.","core_discovery":"The central claim is that the full pipeline from initial data to observable estimate can be made explicit on a quantum computer. Starting from the level-set representation of the solution, the distribution function evolves under linear transport; after upwind discretization the ODE matrix is block diagonal in the level-set index, and Schr\\\"odingerisation with a warped-phase variable turns it into a Hermitian Hamiltonian evolution. The paper then writes that evolution as a first-order Lie\\textendash{}Trotter product, with each factor built from multi-controlled RZ gates and CNOTs, and recovers observables such as density, affine moments, and general nonlinear moments through Hadamard summation and post-selection. With a Richardson-corrected mollifier, the error budget yields Theorem 2: $\\widetilde{O}(d^4T^4N_pN_w^2\\varepsilon^{-4})$ gates per observable, which the authors compare with classical cost $O(d^{d+3}T^{d+2}\\varepsilon^{-(d+1)})$ to claim quantum advantage in high spatial dimensions.","pith_inferences":["The claimed advantage applies to observable estimation rather than full solution reconstruction; the paper itself notes that reading out all grid values would introduce a large measurement cost, and an end-to-end advantage would also require an efficient procedure for preparing the encoded initial state from $u_0$, which is assumed rather than constructed.","A natural next test is to count the state-preparation and oracle-implementation circuits explicitly for a concrete flux such as Burgers, since those costs are excluded from the complexity comparison; if they scale with $N_x^d$, the polynomial-in-$d$ conclusion would be limited.","The same level-set plus Schr\\\"odingerisation pipeline could plausibly extend to systems of conservation laws or Hamilton\\textendash{}Jacobi equations, but the block-diagonal structure and flux-derivative loading become more involved there.","The paper uses first-order Trotter formulas; higher-order product formulas would likely reduce the $T$ and $\\varepsilon$ exponents in the gate count at the cost of more complicated controlled gates."],"forward_implications":["Observables of the form $\\langle G\\rangle(t,x)=\\int G(p)\\psi(t,x,p)\\,dp$---including density, the solution value itself, and nonlinear functions such as entropy---can be estimated coherently with the constructed circuits.","With affine flux derivatives $F'_\\alpha(p_l)$ linear in $l$, the $p$-dependence is implemented by binary-controlled powers, reducing the $N_p$ multiplexing overhead to polylogarithmic cost as described in Remark 4.","Replacing the nonsmooth profile $e^{-|w|}$ by a smoother auxiliary profile $g(w)$ improves the $w$-discretization error from first order to higher order, giving the improved exponent stated in Remark 3.","For sufficiently high spatial dimension $d$, the quantum gate count $\\widetilde{O}(d^4T^4N_pN_w^2\\varepsilon^{-4})$ is smaller than the classical finite-difference cost $O(d^{d+3}T^{d+2}\\varepsilon^{-(d+1)})$, so the comparison favors quantum simulation for observable estimation.","The numerical experiments show that the Trotter\\textendash{}Schr\\\"odingerisation circuit reproduces the reference evolution for a one-dimensional traffic-flow flux and a two-dimensional Burgers equation as the auxiliary resolution $N_w$ increases."],"supporting_citations":[{"why":"It supplies the level-set observable-computation framework and the complexity comparison against classical finite differences.","marker":"[1]"},{"why":"It introduces the level-set method that reformulates a scalar conservation law as a linear Liouville equation.","marker":"[7]"},{"why":"It introduces Schr\\\"odingerisation, the warped-phase embedding used here to convert the nonunitary linearized system into Hermitian dynamics.","marker":"[13]"},{"why":"It provides the scalable circuit decomposition of finite-difference shift operators into multi-controlled rotations and CNOT gates.","marker":"[20]"},{"why":"It gives the Schr\\\"odingerisation circuit construction and per-operator Trotter error bounds that the present complexity analysis builds on.","marker":"[21]"},{"why":"It supplies the first-order Trotter formula with commutator scaling used to bound the product-formula error.","marker":"[15]"},{"why":"It suggests replacing $e^{-|w|}$ by a smoother profile to improve the Fourier discretization error, which underlies Remark 3.","marker":"[22]"}],"fun_headline_variants":["Explicit quantum circuits show advantage for nonlinear conservation laws","Schrödingerisation gives explicit quantum algorithm with edge for PDEs","Quantum advantage made explicit for scalar conservation laws","Gate-level quantum solution for nonlinear PDEs with proven advantage"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The quantum-advantage comparison assumes that preparing the encoded initial state from $u_0$ and implementing the $p$-dependent coefficient oracles cost negligibly compared with the evolution, and that observables can be read out without paying for the full $N_x^d$ spatial grid; if those costs scale with the grid size, the claimed polynomial-in-$d$ advantage disappears.","fun_headline_variants_meta":{"raw":{"variants":["Explicit quantum circuits show advantage for nonlinear conservation laws","Schrödingerisation gives explicit quantum algorithm with edge for PDEs","Quantum advantage made explicit for scalar conservation laws","Gate-level quantum solution for nonlinear PDEs with proven advantage"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000613,"raw_usage":{"total_tokens":2820,"prompt_tokens":888,"completion_tokens":1932,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":504,"completion_tokens_details":{"reasoning_tokens":1867}},"tokens_in":504,"tokens_out":1932,"duration_ms":12947,"temperature":1.0,"reasoning_tokens":1867,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T12:56:31.840955+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Build the complete circuit that includes state preparation and oracle loading for a specific flux such as Burgers and count its gates as a function of $N_x^d$; if the preparation or readout cost is $\\Omega(N_x^d)$ or $\\Omega(N_pN_x^d)$, the end-to-end complexity is no longer polynomial in $d$ and the Remark 5 advantage over classical cost does not hold.","supporting_citations":[{"cited_title":"Quantum algorithms for nonlinear partial differential equations,","cited_arxiv_id":null,"evidence_quote":"It supplies the level-set observable-computation framework and the complexity comparison against classical finite differences."},{"cited_title":"A level set method for the computation of multi-valued solutions to quasi-linear hyperbolic PDEs and Hamilton–Jacobi equations,","cited_arxiv_id":null,"evidence_quote":"It introduces the level-set method that reformulates a scalar conservation law as a linear Liouville equation."},{"cited_title":"Quantum simulation of partial differential equations via Schr¨ odingerization,","cited_arxiv_id":null,"evidence_quote":"It introduces Schr\\\"odingerisation, the warped-phase embedding used here to convert the nonunitary linearized system into Hermitian dynamics."},{"cited_title":"Hamiltonian simulation for hy- perbolic partial differential equations by scalable quantum circuits,","cited_arxiv_id":null,"evidence_quote":"It provides the scalable circuit decomposition of finite-difference shift operators into multi-controlled rotations and CNOT gates."},{"cited_title":"Quantum circuits for partial differential equations via Schr¨ odingerisation,","cited_arxiv_id":null,"evidence_quote":"It gives the Schr\\\"odingerisation circuit construction and per-operator Trotter error bounds that the present complexity analysis builds on."},{"cited_title":"Theory of Trotter error with commutator scaling,","cited_arxiv_id":null,"evidence_quote":"It supplies the first-order Trotter formula with commutator scaling used to bound the product-formula error."},{"cited_title":"Schr¨ odingerization based computationally stable algorithms for ill-posed problems in partial differential equations,","cited_arxiv_id":null,"evidence_quote":"It suggests replacing $e^{-|w|}$ by a smoother profile to improve the Fourier discretization error, which underlies Remark 3."}],"review_version":1}