{"id":"eddffbf4-22d1-4c17-92a7-0501a73680c4","arxiv_id":"2608.07893","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Gevrey-smooth solutions of linear PDEs can be prepared by a quantum Fourier-basis algorithm with poly(d, log(1/eps)) resources.","lead":"This paper shows that quantum computers can solve high-dimensional linear PDEs with costs that grow polynomially in dimension, provided the solution belongs to a Gevrey smoothness class. It also gives explicit gate counts and applies the method to many-body linear response, though with several unresolved assumptions.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The Schrödinger application's sigma_min^+(L_N) converges to 0, not to E_1−E_0, so Corollary 5's polynomial complexity is unsupported and likely false.","rationale":"The Fourier-analytic core (Theorems 1–4, Corollaries 1–3) is carefully argued, and the Gevrey truncation analysis is a genuine contribution. However, the paper's flagship application, Corollary 5, hinges on an unresolved but crucial quantity, sigma_min^+(L_N). The reader flagged this as open; the stress-test shows it is actually degenerate: the projected Schrödinger operator has a near-null direction coming from the approximation of the unbandlimited ground state, so sigma_min^+(L_N) decays to zero and the conditioning explodes. This is load-bearing because (76)–(77) and the claimed polynomial-in-M,P,1/epsilon scaling depend on 1/sigma_min^+; with the paper's own N=Theta(poly(1/epsilon)), this factor becomes super-polynomial. It also threatens the correctness guarantee, since the near-null component of the source is amplified. The general theorem may stand if sigma_min^+ is treated as an input parameter, but the application's end-to-end claim needs major revision. A small numerical experiment can settle whether the concern lands.","tokens_in":37838,"tokens_out":15882,"duration_ms":185994,"concrete_test":"For a 1D model H=-d^2/dx^2+V on the torus with a non-trigonometric ground state (e.g., a Mathieu or soft-Coulomb potential with known eigenfunction), form L_N=P_N(H−E0)P_N in the Fourier basis for N=16,32,64,128 and compute its smallest singular value. If lambda_min(L_N) decays toward 0 roughly like the L2 Fourier tail of Psi0, rather than approaching E1−E0, then Remark 8's identification is false and Corollary 5's bound collapses. An analytic check: use the Rayleigh quotient phi_N=P_N Psi0/||P_N Psi0|| to show lambda_min(L_N) <= ||H−E0|| ||(I−P_N)Psi0||^2 -> 0.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Corollary 5 and Remark 8 claim that for the mollified Schrödinger operator H_gamma−E0, sigma_min^+(L_N) is Theta(E1−E0) as N grows, because Rayleigh–Ritz eigenvalues of L_N=P_N(H_gamma−E0)P_N converge from above to those of the continuous operator. This misidentifies which eigenvalue L_N approximates. The continuous operator has eigenvalue 0 in its kernel; the ground state Psi0 is not bandlimited, so the projected operator has a smallest positive eigenvalue lambda_0(N)=min_{phi in S_N, ||phi||=1} <phi,(H_gamma−E0)phi> that converges to 0 from above, not to E1−E0. Using phi_N=P_N Psi0/||P_N Psi0|| gives lambda_0(N) <= ||H_gamma−E0|| ||(I−P_N)Psi0||^2, which decays exponentially in N^{1/s'} for Gevrey/analytic Psi0. Hence sigma_min^+(L_N)->0, kappa(L_N)->infinity, and the factor 1/sigma_min^+ in (76)–(77) is not a harmless constant. Since Corollary 5 sets N=Theta(poly(1/epsilon)), the near-null eigenvalue is roughly exp(-c/epsilon^{2/s'}), so the claimed complexity is super-polynomial in 1/epsilon. Moreover, eta_N generically has a small component along the near-null vector phi_N, and L_N^{-1} eta_N is then dominated by that component, so the returned state need not approximate the true response. This is not an open convergence-rate question; the limit value asserted in Remark 8 is wrong for the smallest singular value.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a Fourier-based quantum algorithm for preparing amplitude encodings of periodic functions and for solving linear PDEs under Gevrey regularity assumptions. It derives explicit Fourier truncation thresholds and elementary gate counts (Theorem 5), supported by detailed Fourier-analytic lemmas in Appendix A and new block-encoding circuits for differential operators. The authors then apply the framework to the Poisson equation and to a mollified many-body Schrödinger linear-response problem, claiming polynomial complexity in the particle number and in 1/epsilon, and introduce a hierarchy of k-particle reduced-density-matrix pipelines.","tokens_in":38307,"tokens_out":12154,"duration_ms":140536,"significance":"The Fourier-analytic core is a useful contribution: it gives a clean regularity hierarchy from continuous functions through Gevrey to analytic and entire classes, with explicit tail bounds; the block encodings are concrete; and the Poisson corollary is genuinely parameter-free, with sigma_min^+(L_N) computed exactly. The paper is also unusually explicit about gate counts and about many of its residual open assumptions. If the general Theorem 5 is read as a conditional statement in terms of sigma_min^+(L_N) and known Gevrey parameters, it is a solid result. However, the advertised atomistic-application speedup rests on a false spectral claim about the smallest positive singular value of the projected Schrödinger operator, so the many-body application, which is a central selling point of the paper, is not currently supported.","major_comments":[{"comment":"The claim that sigma_min^+(L_N) tends to sigma_min^+(gamma) = Theta(E_1-E_0) via Rayleigh-Ritz monotonic convergence identifies the wrong eigenvalue. The continuous operator H_gamma-E_0 has eigenvalue 0 on Psi_0, so the projected operator L_N = P_N(H_gamma-E_0)P_N has a smallest positive eigenvalue lambda_0(N) = min_{phi in S_N, ||phi||=1} <phi,(H_gamma-E_0)phi>, which converges to 0 from above whenever Psi_0 is not exactly bandlimited; it does not converge to E_1-E_0. Consequently sigma_min^+(L_N) in Eqs. (76)-(77) is a near-null eigenvalue, and with N chosen as in Corollary 5 its reciprocal can grow super-polynomially in 1/epsilon, making the claimed polynomial complexity unsupported. This is not an open convergence-rate question: the limit asserted in Remark 8 is wrong for the smallest singular value.","section":"Section III.B.2, Remark 8 / Corollary 5"},{"comment":"The continuous-level lower bound sigma_min^+(gamma)=Theta(E_1-E_0) is not established. The trial-state min-max argument gives only upper bounds on lambda_0(H_gamma) and lambda_1(H_gamma); a lower bound on the gap requires the Feshbach-Schur calculation with a quantitative bound on the Schur complement, which is not supplied. Since W_gamma = V_gamma - V is pointwise unbounded, Weyl's inequality genuinely does not apply, and Eq. (84) is asserted rather than proved. Because the choice of gamma in Corollary 5 and the final epsilon-dependence rest on this gap, the application's claimed polynomial scaling is not supported.","section":"Section III.B.2, Remark 7 / Eq. (84)"},{"comment":"The invocation of Theorem 5 requires the mollified response deltaPsi_gamma to be s-Gevrey with radius Theta(gamma), but this is not proved. Analyticity of the mollified potentials V_gamma does not by itself imply the stated regularity of the eigenfunctions Psi_0, Psi_1, or of the response deltaPsi_gamma; Remark 8 concedes that the analyticity radius of the eigenfunctions remains open. Additionally, the discrete solvability condition eta_N perp ker(L_N) at the specific polynomial N is assumed rather than proved (see Section IV). These gaps are load-bearing for the atomistic-application claims.","section":"Section III.B.2, Corollary 5"}],"minor_comments":[{"comment":"The displayed formula for N is implicit because N appears on both sides through the (2N+1)^{d/2} factor; the proof later absorbs the resulting log N into eO(·), but the statement should say that N is implicitly defined and give the final explicit order after this absorption.","section":"Eq. (32) / Corollary 3"},{"comment":"The row for this work omits the prefactors ||g||/sigma_min^+ and the dependence on the Gevrey constants C and r; including these would make the comparison with classical methods more informative and more honest.","section":"Table I"},{"comment":"There are several typos, including 'aliaising' in Appendix A, 'Cauchy-Schwartz' in the proof of Corollary A.2, and 'operatr' in the Discussion; a careful proofread is needed.","section":"Appendix A / Discussion"}],"recommendation":"major_revision","confidential_remarks":"The Fourier-analytic core and the block-encoding constructions are solid and could support a good paper on Gevrey-regularity quantum PDE solvers. The atomistic linear-response section, however, contains a mathematically false spectral claim in Remark 8 about sigma_min^+(L_N), and Corollary 5's polynomial complexity is therefore unsupported. If the authors are unwilling to remove or fundamentally revise that section, I would not recommend acceptance. The manuscript also relies heavily on its own companion papers; the editor may wish to check novelty and attribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper does something genuinely useful. It extends quantum PDE solvers from analytic solutions to the Gevrey hierarchy, gives explicit Fourier truncation bounds for those classes, and provides concrete gate counts via new block encodings of higher-order differential operators. The Fourier analysis in Appendix A looks correct, and Theorem 5 is a solid conditional result: if the solution is s-Gevrey and the discrete operator's smallest nonzero singular value is under control, then the complexity is polynomial in d and log(1/eps) up to that constant. I would send this part to a serious referee.\n\nThe Poisson application is fine because sigma_min^+ is known exactly. The atomistic linear-response application is where the paper goes off the rails. Corollary 5 and Remark 8 claim that sigma_min^+(L_N) for the projected mollified Schrodinger operator is Theta(E1 - E0), citing Rayleigh-Ritz convergence of eigenvalues from above. That misses a basic point: the continuous operator H_gamma - E0 has a zero eigenvalue, the ground state. The projection of that ground state onto the degree-N space is a near-null vector phi_N = P_N Psi0 / ||P_N Psi0||. Its Rayleigh quotient is of order ||(I-P_N)Psi0||^2, which decays to zero as N grows. So the smallest positive singular value of L_N goes to 0, not to E1 - E0. The factor 1/sigma_min^+ in (76)-(77) is therefore exponentially large in the chosen N, and the claimed poly(1/eps) complexity is unsupported. This is not an open rate question; the limit value asserted in Remark 8 is wrong for the smallest singular value.\n\nI also note the paper assumes the Gevrey parameters (C,r) are known without deriving them from the PDE coefficients. That is a weaker concern, since Theorem 5 is conditional by design and says so. The paper is honest in Remark 8 about the analyticity radius of the mollified response being unproven, but that honesty does not fix the spectral misidentification.\n\nWho gets value from this? People working on quantum PDE solvers and Fourier-spectral methods will find the truncation bounds and block-encoding circuits worth citing. The many-body pipeline should not be cited as a proven speedup until the sigma_min^+ issue is resolved, likely by projecting out the approximate kernel before inversion. I would accept the paper for peer review but expect major revision, with the Schrodinger application either heavily revised or removed.","headline":"Gevrey Fourier framework is a real contribution; the many-body application's smallest-singular-value claim does not survive scrutiny, so the advertised end-to-end speedup is unsupported.","tokens_in":38764,"tokens_out":2560,"would_cite":true,"duration_ms":31043,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65T40","65N35","35B65","81P68"],"pacs":["03.67.Ac"],"model":"deepseek-v4-flash","headline":"Gevrey regularity—functions whose derivatives grow at most like $C(\\alpha!)^s/r^{|\\alpha|}$—suffices to solve linear PDEs on a quantum computer with complexity polynomial in dimension and polylogarithmic in precision, breaking the curse…","keywords":["quantum PDE solvers","Gevrey regularity","Fourier spectral methods","curse of dimensionality","block encoding","linear response","first quantization","quantum linear systems"],"falsifier":"Take a specific periodic $s$-Gevrey solution of a linear PDE with known parameters $(C,r)$—for example the Poisson equation on the torus with a known Gevrey right-hand side—and check numerically whether the state returned by Algorithm 1 is $\\varepsilon$-close at the cutoff $N$ given by Eq. (32); if the actual error exceeds $\\varepsilon$ because $\\sigma_{\\min}^+(L_N)$ decays with $N$ much faster than the continuous-level gap, the theorem's bound fails. A more targeted calculation is to compute $\\sigma_{\\min}^+(L_N)$ for an elliptic operator with non-constant coefficients and compare its $N$-dependence with the Rayleigh-Ritz monotone convergence; a rate slower than $e^{-cN^{1/s}}$ would force $N$ to grow beyond the polylogarithmic threshold and break the claimed precision scaling.","tokens_in":37678,"feed_emoji":"⚛️","tokens_out":6333,"duration_ms":64611,"temperature":0.7,"pith_summary":"This paper claims that the computational bottleneck for quantum linear PDE solvers is determined by the smoothness class of the solution, and that the Gevrey hierarchy provides the correct classification. It establishes that for an $s$-Gevrey solution, $s\\ge 1$, a quantum Fourier-spectral algorithm can prepare an $\\varepsilon$-accurate amplitude encoding of the solution with complexity polynomial in the dimension $d$, polylogarithmic in $1/\\varepsilon$, and independent of the curse of dimensionality, using $O(d\\log N)$ qubits. The paper further converts query-complexity bounds into explicit elementary gate counts and demonstrates the framework on Poisson-type equations and on linear-response Schr\\\"odinger equations in first-quantized atomistic simulation. The central methodological step is a block encoding of the Fourier-truncated differential operator $\\sum_\\alpha \\mathrm{diag}\\{g_{\\alpha,N}\\}\\widetilde D^\\alpha_N$ that costs $\\widetilde O(d(|J|+\\lceil J\\rceil))$ gates once the truncation threshold $N$ is chosen through the Gevrey decay estimate.","feed_headline":"Gevrey regularity alone breaks quantum PDE curse of dimensionality","feed_subtitle":"Weaker than analyticity, s-Gevrey smoothness gives polylog(1/eps) quantum PDE solve time in any dimension.","key_machinery":"The machinery is the Gevrey regularity class defined by $\\|D^\\alpha f\\|_\\infty \\le C(\\alpha!)^s/r^{|\\alpha|}$, together with the aliasing relation $\\tilde f_\\omega = \\sum_{m\\in\\mathbb Z^d} \\hat f_{\\omega+(2N+1)m}$ and the diagonal form of differentiation in the Fourier basis. The key identity is that for an $s$-Gevrey function the discrete Fourier coefficients converge to the continuous coefficients with error at most $C e^{-(r/2)N^{1/s}}$ once $N\\in \\widetilde\\Omega((ds/r)^s)$, and the same bound holds for the approximate derivative operator $\\widetilde D^\\alpha_N$ acting on the discretized function. This converts Fourier-coefficient decay into a rigorous truncation threshold for the PDE solver. On the circuit side, the paper constructs a block encoding of the full differential sum $\\sum_{\\alpha\\in J}|\\bar\\alpha\\rangle\\langle\\bar\\alpha|\\otimes \\widetilde D^\\alpha_N$ using a conditional swap network and a single block encoding $\\Delta_N$ of the approximate first derivative $\\widetilde\\partial_N = F_N^{-1}\\mathrm{diag}\\{i\\omega\\}F_N$, costing $O(\\lceil J\\rceil(d\\log\\lceil J\\rceil\\log N+\\log N\\log\\log N+R_N))$ gates with $R_N=O(\\log N)$.","core_discovery":"The central discovery is that the Gevrey scale interpolates cleanly between smooth and analytic functions and gives the right measure of quantum Fourier-encoding hardness. An $s$-Gevrey $d$-variate periodic function has Fourier coefficients decaying as $e^{-r\\|\\omega\\|_\\infty^{1/s}}$, which makes the aliasing error between discrete and continuous Fourier coefficients exponentially small once the cutoff $N$ is chosen as in Eq. (32); the same decay controls the error of the Fourier-truncated differential operator. The paper proves in Theorem 5 that for a unique $s$-Gevrey solution of the linear PDE in (8), with the discrete source orthogonal to the kernel, Algorithm 1 returns an $\\varepsilon$-close normalized state using $\\widetilde O\\big(\\lceil g\\rceil(|J|+\\lceil J\\rceil)d^{\\lceil J\\rceil s+1}(s\\log(1/(\\|\\hat u\\|\\varepsilon))){}^{\\lceil J\\rceil s}\\log(1/\\varepsilon)/\\sigma_{\\min}^+\\big)$ elementary gates and $O(d\\log N)$ qubits. Thus Gevrey regularity, which is weaker than analyticity, is sufficient to break the exponential dependence on dimension in quantum PDE solvers.","pith_inferences":["If the Gevrey parameters $(C,r)$ of the solution cannot be certified from the PDE coefficients, then Eq. (32) is not directly executable; a practical deployment would need a certified upper bound on $(C,r)$ or an adaptive scheme that estimates them from the Fourier tail, which the paper leaves open.","The framework suggests that any function class between smooth and analytic that lacks exponential Fourier decay will force $N$ to grow polynomially in $1/\\varepsilon$, re-introducing the dimension in the exponent; proving a lower bound for non-Gevrey classes would complete the classification that Table II begins.","The mollification radius $\\gamma$ shrinks with $\\varepsilon$, so the Gevrey radius $r(\\gamma)=\\Theta(\\gamma)$ and the smallest nonzero singular value $\\sigma_{\\min}^+(\\gamma)$ trade off; optimizing this trade-off against the condition number could improve the stated $\\varepsilon^{-6}$ dependence in the atomistic application.","The same Fourier-Gevrey machinery could likely be adapted to Chebyshev pseudo-spectral methods on bounded domains, as the paper speculates; the testable step is to derive an analogue of Theorem 3 for Chebyshev coefficients and check whether the exponential-in-$N^{1/s}$ convergence survives the loss of periodicity."],"forward_implications":["For the anisotropic Poisson equation with an $s$-Gevrey solution, Algorithm 1 solves the equation to precision $\\varepsilon$ with $\\widetilde O(\\|\\Sigma\\|_{1,1}s^{2s}d^{2s+1}\\log^{2s}(1/(\\|\\hat u\\|\\varepsilon))\\log(1/\\varepsilon)/\\sigma_{\\min}^+)$ elementary gates and $O(d\\log N)$ qubits.","For the inhomogeneous Schr\\\"odinger equation arising in linear response, the mollified Coulomb problem is solved to accuracy $\\varepsilon$ with a gate count polynomial in the particle number $M$, the number of nuclei $P$, and $1/\\varepsilon$, so the curse of dimensionality is avoided even though the dependence on precision is $\\mathrm{poly}(1/\\varepsilon)$ rather than polylogarithmic.","Each level $k$ of the $k$-particle reduced-density-matrix hierarchy yields a further polynomial-degree quantum speedup in the PDE dimension, so the framework provides a gradual improvement in simulation efficiency as quantum computers scale.","The paper supplies explicit elementary gate and qubit counts, not just query complexity, for high-precision quantum PDE solvers, and it identifies the Gevrey threshold (32) as the quantity that fixes the Fourier cutoff $N$ in terms of the target precision.","For sign-definite (bosonic) targets the method achieves polylogarithmic state-preparation infidelity in $1/\\varepsilon$, compared with the $O(1/\\varepsilon^2)$ sample cost of Monte Carlo methods; for fermionic targets the exponential sign problem is bypassed, giving an exponential advantage over both grid-based and Monte Carlo classical methods."],"supporting_citations":[{"why":"Supplies the classical exponential-accuracy result for Fourier and Chebyshev differencing of analytic functions, which the paper extends to the full Gevrey hierarchy and to arbitrary dimension.","marker":"[Tad86]"},{"why":"Provides the Lemma 48 construction that turns bit-oracles for coefficient functions into block encodings of diagonal matrices with $O(\\log^{2.5}(1/\\delta))$ gates, used for the coefficient terms of the PDE.","marker":"[GSLW19]"},{"why":"Supplies the adiabatic quantum linear-system solver with query complexity $O(\\kappa\\log(1/\\varepsilon))$, which is the inversion subroutine in Algorithm 1 and in the proof of Theorem 5.","marker":"[CAS+22]"},{"why":"Supplies the lattice-sum bound (Lemma E.5) used to control the aliasing error between discrete and continuous Fourier coefficients in Proposition A.3 and its corollaries.","marker":"[MR24]"},{"why":"Supplies the norm-distance inequality (Lemma 13) used to pass from convergence of unnormalized vectors to convergence of normalized amplitude-encoding states in Corollary 2 and Corollary 3.","marker":"[BCOW17]"},{"why":"Provides the prior high-precision quantum PDE algorithms whose regularity assumptions the paper identifies as implicitly requiring uniform bounds on arbitrarily high derivatives, motivating the explicit Gevrey analysis.","marker":"[CLO21]"},{"why":"Supplies the quantum linear-systems algorithm with exponentially improved precision dependence that underlies several of the compared prior quantum PDE solvers and the precision scaling targeted by this work.","marker":"[CKS17]"}],"fun_headline_variants":["Smoothness, not analyticity, kills quantum PDE dimension curse","Gevrey functions end exponential quantum PDE slowdown","Polylog precision quantum PDE solves for smooth periodic data","Weaker than analytic: Gevrey regularity defeats curse of dimensionality","Quantum PDEs: Gevrey smoothness is enough for polynomial speedup"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the solution $u$ is known to be $s$-Gevrey with known parameters $(C,r)$, so that the Fourier cutoff $N$ can be set by Eq. (32), yet the paper does not derive these parameters from the PDE coefficients; if they are unavailable or the discrete smallest nonzero singular value $\\sigma_{\\min}^+(L_N)$ decays faster than assumed, the stated complexity bound cannot be put into practice.","fun_headline_variants_meta":{"raw":{"variants":["Smoothness, not analyticity, kills quantum PDE dimension curse","Gevrey functions end exponential quantum PDE slowdown","Polylog precision quantum PDE solves for smooth periodic data","Weaker than analytic: Gevrey regularity defeats curse of dimensionality","Quantum PDEs: Gevrey smoothness is enough for polynomial speedup"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000655,"raw_usage":{"total_tokens":3007,"prompt_tokens":961,"completion_tokens":2046,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":577,"completion_tokens_details":{"reasoning_tokens":1959}},"tokens_in":577,"tokens_out":2046,"duration_ms":14770,"temperature":1.0,"reasoning_tokens":1959,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T00:42:11.293705+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a specific periodic $s$-Gevrey solution of a linear PDE with known parameters $(C,r)$—for example the Poisson equation on the torus with a known Gevrey right-hand side—and check numerically whether the state returned by Algorithm 1 is $\\varepsilon$-close at the cutoff $N$ given by Eq. (32); if the actual error exceeds $\\varepsilon$ because $\\sigma_{\\min}^+(L_N)$ decays with $N$ much faster than the continuous-level gap, the theorem's bound fails. A more targeted calculation is to compute $\\sigma_{\\min}^+(L_N)$ for an elliptic operator with non-constant coefficients and compare its $N$-dependence with the Rayleigh-Ritz monotone convergence; a rate slower than $e^{-cN^{1/s}}$ would force $N$ to grow beyond the polylogarithmic threshold and break the claimed precision scaling.","supporting_citations":[],"review_version":1}