{"id":"5dda78dd-9a83-4875-b90c-636e22820d14","arxiv_id":"2505.01602","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A quantum algorithm combining the Caffarelli-Silvestre extension with Schrödingerization solves fractional Poisson equations with mesh dependence independent of dimension.","lead":"This paper designs a quantum algorithm for solving high-dimensional fractional Poisson equations by converting the nonlocal problem into a local one and then into a quantum simulation. If the complexity analysis is correct, its cost per grid point does not grow exponentially with dimension, unlike classical solvers.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Headline h^{-2.5} query bound rests on an unconstructed variable-time amplitude amplification step that Remark 2.2 itself calls an open question; the paper's own un-amplified bound is h^{-4.5}.","rationale":"The reader's CONDITIONAL verdict is appropriate. The most load-bearing gap is that Theorem 4.1's proof invokes a VTAA procedure that Remark 2.2 explicitly declines to construct. Since Lemma 2.1 itself is imported from [36] without proof, even the un-amplified bound is not fully self-contained, but the specific factor that separates the advertised h^{-2.5} from the paper's own h^{-4.5} is precisely the missing VTAA. I do not see this as a reason to reject: the numerical experiments support the formulation, the block-encoding construction for the Kronecker-product stiffness matrix is concrete, and the un-amplified bound still gives an h-exponent independent of d. Supplying a VTAA construction, or honestly downgrading the headline to the un-amplified bound, would resolve the concern. Thus the reader's CONDITIONAL verdict should stand unchanged.","tokens_in":16999,"tokens_out":6877,"duration_ms":69646,"concrete_test":"Construct the VTAA procedure promised in Remark 2.2 for the Schrödingerization-based QLSP of Lemma 2.1: specify the stage-dependent success amplitudes and the variable-time amplitude amplification recursion, then recompute the total block-encoding and state-preparation query counts including all logarithmic factors and the O(g) trace-projection amplification of §4.2. If the resulting κ dependence is not near-linear, or if the Ω(1) success probability requires additional rounds, replace the h^{-2.5} claim with the paper's own un-amplified bound Õ(d^2 3^{5d/2} h^{-4.5}) and revise the abstract accordingly.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The advertised bound in Theorem 4.1, Õ(d 3^{3d/2} h^{-2.5}), is justified in the proof by a single sentence: 'combining Theorem 2.1, the VTAA procedure (see Remark 2.2), Theorem 3.1 and Eq. (4.5).' But Remark 2.2 explicitly says that the needed VTAA reduction 'requires substantial modifications to the LCU or QSP algorithms' and that a simple approach 'remains an open question'. No VTAA schedule is constructed, and no query count incorporating VTAA is given. The arithmetic is transparent: Theorem 3.1 gives κ(A)=O((d+1)3^{d+1}h^{-2}), and Eq. (4.5) contributes g=O(3^{d/2}h^{-1/2}) for trace extraction. Theorem 2.1's un-amplified complexity is quadratic in κ, yielding h^{-4}·h^{-1/2}=h^{-4.5}; the advertised h^{-2.5} requires replacing the quadratic κ dependence by a near-linear one via the missing VTAA. The qualitative exponential advantage in d may well survive at the un-amplified h^{-4.5} bound, but the concrete h^{-2.5} claim in the abstract and Theorem 4.1 is not established by this paper.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a quantum algorithm for the fractional Poisson equation on bounded domains. It first applies the Caffarelli–Silvestre extension to rewrite the nonlocal fractional Laplacian as a local degenerate/singular elliptic problem in one higher dimension, discretizes the resulting problem with a tensor-product finite element method, and then solves the resulting linear system Au=b by interpreting the solution as the steady state of the ODE du/dt=-Au+b and applying the Schrödingerization technique. The main claimed result is a query complexity of Õ(d3^{3d/2}h^{-2.5}) against block-encoding input models, to be compared with a classical conjugate-gradient cost of Õ(d^{1/2}3^{3d/2}h^{-d-2}), giving an exponential quantum advantage in the spatial dimension d for fixed mesh size h. Numerical experiments in one and two dimensions for both the extension formulation and a nonlocal variational formulation are reported.","tokens_in":17312,"tokens_out":6978,"duration_ms":67889,"significance":"If fully established, an exponential quantum advantage in d for a fractional PDE would be a significant result. The combination of the Caffarelli–Silvestre extension with Schrödingerization is natural and potentially impactful, and the manuscript contains genuinely useful components: a transparent block-encoding construction for the FEM stiffness matrix, an explicit trace-extraction step, and numerical validation of the formulation. I also credit the authors for explicitly stating the un-amplified bound Õ(d^2 3^{5d/2}h^{-4.5}) at the end of Section 4.3; even at that bound, the h-dependence remains dimension-independent, so the qualitative exponential advantage may well survive. However, the headline h^{-2.5} bound depends on an essentially unproved lemma and on a VTAA procedure that the authors themselves describe as an open question, and several supporting estimates are incompletely justified. These gaps are load-bearing for the main complexity claim and must be closed before the result can be accepted as stated.","major_comments":[{"comment":"The claimed query complexity Õ(d3^{3d/2}h^{-2.5}) in Theorem 4.1 and the abstract is not established. The proof consists of the sentence 'combining Theorem 2.1, the VTAA procedure (see Remark 2.2), Theorem 3.1 and Eq. (4.5)', but Remark 2.2 states that a simple approach to reducing the κ-dependence from quadratic to near-linear remains an open question and that VTAA requires substantial modifications to LCU or QSP algorithms. No VTAA schedule is constructed and no VTAA query count is provided. The paper's own un-amplified bound is Õ(d^2 3^{5d/2}h^{-4.5}) (end of Section 4.3). Please either construct and analyze the VTAA step explicitly, or remove h^{-2.5} from the abstract and Theorem 4.1 and present the un-amplified bound as the main result.","section":"Theorem 4.1 and abstract"},{"comment":"Lemma 2.1 is the foundation of Theorem 2.1 but is not proved in this manuscript; it is introduced with 'Following the similar implementation in [36] with some modifications'. Since [36] is the authors' own concurrent work and the parameter dependence in Lemma 2.1, especially the claim η_max = O(log(1/ε)), is crucial for the final complexity, a self-contained proof or a precise statement of the relevant theorem in [36] is needed. Without this, the query-complexity chain lacks a verified base.","section":"Lemma 2.1 and Theorem 2.1"},{"comment":"The proof of the condition-number estimate treats only the case 0<s<1/2 (0<α<1) and asserts that s>1/2 'can be deduced in a similar manner'. This is not immediate because for α=1-2s<0 the weight y^α is singular at y=0, and the use of the second integral mean-value theorem to extract Y^α is not justified for a singular weight. In addition, the mesh in the extended direction is graded as in (3.9), not quasi-uniform; the quoted bound λ_min(G_ξ) ≳ (h/3)^{d+1} from [25] may not hold for the actual graded mesh. Since κ(A) enters the quantum complexity bound, a complete proof of Theorem 3.1 for all s∈(0,1) and for the graded mesh used by the algorithm is required.","section":"Theorem 3.1"},{"comment":"The proof of Lemma 4.1 is incomplete. For the 'second inequality' involving S(d+1), the text says only that one can apply the second integral mean-value theorem 'as done in Theorem 3.1', but S(d+1) is a weighted mass matrix and for α<0 the weight is singular, so the asserted bound h/3 ≲ λ(S(d+1)) ≲ h needs a derivation. Moreover, the conclusion g = O(3^{d/2}h^{-1/2}) requires a bound on ∥(uY)_h∥_{L2}/∥u_h∥_{L2}; this trace ratio is not obviously O(1) in the graded-mesh setting and is not discussed. An extra factor here would change the trace-extraction overhead in the final complexity.","section":"Lemma 4.1"},{"comment":"The block-encoding construction for A is described at the level of combining univariate block-encodings, but the query cost of constructing the block-encoding oracle for A is not quantified. The sentence 'the complexity of block-encoding A ... has only a quadratic dependence on d' is not accompanied by a count. Since Theorem 4.1 counts queries to the block-encoding oracle of A, the cost of implementing that oracle via the proposed LCU of tensor products must either be included in the total cost or stated as a separate input-model assumption. Without this, the claimed d-dependence of the quantum algorithm is not fully established.","section":"Section 4.1"}],"minor_comments":[{"comment":"In the paragraph following (3.6), 'the the Dirichlet boundary' should read 'the Dirichlet boundary'.","section":"Section 3.3"},{"comment":"The spelling of Schrödingerisation/Schrödingerization is inconsistent; please choose one convention and use it consistently.","section":"Throughout"},{"comment":"References [2] and [3] appear to be the same paper with different years; please merge or correct the entries.","section":"References"},{"comment":"The notation 'ui1,···,id,id+1=0' is ambiguous; it should read 'ui1,...,id,0' or clearly set i_{d+1}=0.","section":"Equation (4.4)"},{"comment":"The condition that the coefficient matrix is 'negative semi-definite over the interval [0,T]' is unclear as stated; if the authors mean that -A in (2.1) is negative definite for positive definite A, the wording should be made precise.","section":"Lemma 2.1"},{"comment":"The caption of Figure 6 refers to a fixed point (x1,x2)=(0,0) but the example is one-dimensional; please adjust the caption to match the plotted quantity.","section":"Figure 6"},{"comment":"The phrase 'which can show up to exponential advantage' is awkward; consider rewording to 'which yields an exponential advantage'.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":"The manuscript relies heavily on the authors' own earlier and concurrent work, notably [36] for Lemma 2.1 and [12] for VTAA. The editors may wish to verify that the proof of Lemma 2.1 has appeared before the present manuscript claims it. The headline complexity claim should not be emphasized in the abstract until the VTAA construction is either supplied or explicitly removed from the main result."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"One thing to know: this paper is a genuine algorithmic combination—Caffarelli-Silvestre localization plus Schrödingerization-based linear solving—but the headline query count is not actually proved. Theorem 4.1's h^{-2.5} bound is derived by a one-line reference to a variable-time amplitude amplification (VTAA) procedure that Remark 2.2 explicitly says requires substantial modifications and whose simpler form 'remains an open question.' The paper is honest about this: it immediately gives the un-amplified bound Õ(d^2 3^{5d/2} h^{-4.5}). That bound still implies exponential advantage in d against classical CG's h^{-d-2}, so the core claim survives in weakened form.\n\nWhat's new and good: the combination is new, as far as I can tell. The block encoding of the stiffness matrix from tensor-product FEM matrices is a clean way to avoid constructing the 3^{d+1}-sparse system explicitly, and the trace projection analysis (Lemma 4.1) gives a reasonable g = O(3^{d/2} h^{-1/2}) overhead for extracting the solution on the original domain. The numerical experiments are small but validate the formulation for both the extension and nonlocal variational forms. The paper also cites Nochetto et al. appropriately for the FEM side.\n\nSoft spots, in proportion. The main one is the missing proof chain for the advertised complexity. Lemma 2.1, the workhorse, is stated without proof and justified by 'following [36] with modifications.' That is acceptable if [36] is citable, but here [36] is a concurrent preprint; a referee should ask the authors to state exactly which theorem in [36] supplies what. The VTAA step is not just unconstructed—the authors themselves describe it as open. So the abstract and Theorem 4.1 overstate what the paper establishes. Second, Theorem 3.1's condition number bound is proved only for s<1/2 (α>0), with s>1/2 dismissed as 'similar manner.' That may be true, but for singular weights the second mean value theorem step does not carry over directly, so it needs a real argument. Minor: the numerical section reports no convergence rates or complexity measurements; it only plots solutions. That is fine for a formulation paper, but it does not test the theory.\n\nFor a reader: if you work on quantum PDE algorithms, this is worth knowing about. The un-amplified h^{-4.5} bound still gives exponential advantage in d, and the block-encoding construction is reusable. I would send it to a serious referee, but with a clear request to fix the complexity statement: either prove the VTAA reduction, or demote the headline to the un-amplified bound. As written, it is a conditional accept, not an accept.\n\nRecommendation: engage with it, but do not let the h^{-2.5} claim go to press unchallenged.","headline":"A legitimate new combination with a real gap: the advertised h^{-2.5} complexity rests on an unconstructed VTAA step that the authors themselves flag as open; the paper's own un-amplified bound still delivers the qualitative exponential advantage.","tokens_in":17804,"tokens_out":5055,"would_cite":true,"duration_ms":45296,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65N30","65N22","35R11","81P68"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims a quantum algorithm for the FEM-discretized fractional Poisson equation with query cost Õ(d 3^{3d/2} h^{-2.5}), independent of dimension in the mesh size, an exponential advantage over classical conjugate gradients in…","keywords":["fractional Laplacian","fractional Poisson equation","Caffarelli-Silvestre extension","Schrödingerization","quantum algorithm","finite element method","block-encoding","curse of dimensionality"],"falsifier":"Implement the proposed pipeline for a fixed dimension, say $d=2$ or $d=3$, on the FEM-discretized fractional Poisson equation over a sequence of mesh sizes $h$, and count block-encoding queries needed to prepare an $\\epsilon$-accurate solution with success probability $\\Omega(1)$. If the empirical exponent of $h^{-1}$ exceeds $2.5$, or if the success probability decays polynomially in $h$, the claimed dimension-independent advantage fails; a direct check of Lemma 2.1 against a stiff positive-definite matrix would settle the same question.","tokens_in":16815,"feed_emoji":"⚛️","tokens_out":9992,"duration_ms":88105,"temperature":0.7,"pith_summary":"This paper aims to establish that the high-dimensional fractional Poisson equation, whose nonlocal fractional Laplacian classically brings the curse of dimensionality, can be solved on a quantum computer with a cost that depends on the mesh size $h$ in a dimension-independent way. The route is to apply the Caffarelli–Silvestre extension, converting the nonlocal equation into a local weighted elliptic problem in one extra dimension, then to solve the finite-element linear system by interpreting the solution as the steady state of $du/dt=-Au+b$ and simulating that ODE with the Schrödingerization (warped-phase) method. The quantitative claim is $\\widetilde{\\mathcal{O}}(d 3^{3d/2} h^{-2.5})$ queries to block-encoding input models, against $\\widetilde{\\mathcal{O}}(d^{1/2}3^{3d/2} h^{-d-2})$ operations for classical conjugate gradients, an exponential quantum advantage in the dimension $d$ at fixed $h$. A sympathetic reader would care because such a result would make a class of nonlocal high-dimensional PDEs numerically accessible for the first time.","feed_headline":"Quantum algorithm claims exponential speedup for fractional Poisson","feed_subtitle":"The quantum cost's mesh-size dependence no longer grows with dimension, breaking the classical curse of dimensionality.","key_machinery":"The carrying mechanism is the pairing of the Caffarelli–Silvestre extension with the Schrödingerization transformation. The extension is the identity $d_s(-\\Delta)^s u = -\\lim_{y\\to 0^+} y^{\\alpha}u_y$ for $\\alpha=1-2s$, realized as the Dirichlet-to-Neumann map of the weighted elliptic equation $\\mathrm{div}(y^{\\alpha}\\nabla u)=0$ on $\\Omega\\times(0,\\infty)$; it replaces the nonlocal operator by a local problem in $d+1$ dimensions. Schrödingerization is the warped-phase change of variables $v(t,p)=e^{-p}u_f(t)$ that sends the dissipative ODE system $du_f/dt = A_f u_f$ into the Schrödinger-type system $\\partial_t v = -H_1\\partial_p v + iH_2 v$, whose Fourier discretization in the auxiliary variable $p$ is a Hamiltonian system suitable for quantum simulation. The final ingredient is the block-encoding of $A$ as a sum of Kronecker products $A^{(1)}\\otimes S^{(2)}\\otimes\\cdots\\otimes S^{(d+1)}$ plus cyclic permutations, which keeps the input-model cost quadratic in $d$ rather than exponential.","core_discovery":"The paper's central claim is that a nonlocal fractional problem can be solved through the local extension it inherits. After truncating the Caffarelli–Silvestre cylinder to $C_Y$ and using tensor-product piecewise-linear finite elements on graded meshes, the stiffness matrix $A$ has sparsity $3^{d+1}$ and condition number $\\kappa(A)=O((d+1)3^{d+1}h^{-2})$, and it can be block-encoded as a sum of Kronecker products of one-dimensional stiffness and mass matrices, so the exponentially large sparse matrix is never formed explicitly. Treating $Ax=b$ as the steady state of $du/dt=-Au+b$ and applying Schrödingerization produces a Hamiltonian system whose simulation gives an $\\epsilon$-approximation of $|x\\rangle$; combining the query bound with the trace projection onto the $y=0$ boundary and amplitude amplification yields the headline $\\widetilde{\\mathcal{O}}(d 3^{3d/2} h^{-2.5})$ query complexity. Without the variable-time amplitude amplification described in Remark 2.2, the paper's own estimate is $\\widetilde{\\mathcal{O}}(d^2 3^{5d/2} h^{-4.5})$, still independent of $d$ in the $h^{-1}$ factor.","pith_inferences":["Editorial inference: the same localization-plus-Schrödingerization template could be applied to other nonlocal operators admitting a local extension, such as fractional Laplacians with Neumann or Robin boundary conditions or regional fractional operators, but the paper does not treat those cases.","Editorial inference: the trace projection is a natural bottleneck, since the cost carries a $3^{d/2}$ factor from recovering the $d$-dimensional solution from the $d+1$-dimensional one; a block-encoding of the trace operator that avoids measurement-and-amplify could improve the dimension dependence.","Editorial inference: because the headline bound rests on an unproved lemma and an unconstructed subroutine, a direct proof or a numerical query-count test of Lemma 2.1 on stiff FEM matrices is the minimal next step that would make the exponential-advantage claim conclusive."],"forward_implications":["If the central claim is correct, the fractional Poisson equation loses the classical curse of dimensionality: for fixed mesh size the quantum cost is independent of $d$ in the $h^{-1}$ factor, while classical conjugate gradients scale as $h^{-d-2}$.","The Kronecker-product block-encoding construction means the $3^{d+1}$-sparse FEM matrix never needs to be assembled or queried entrywise; block-encodings of one-dimensional stiffness and mass matrices suffice.","The same extension-plus-Schrödingerization pipeline is applied to a finite-difference discretization and to the nonlocal variational formulation, with Remark 5.1 observing that block-encoding of the dense nonlocal matrix remains possible on regular domains.","Even in the un-amplified version, the cost $\\widetilde{\\mathcal{O}}(d^2 3^{5d/2}h^{-4.5})$ is still dimension-independent in $h^{-1}$, so the quantum approach would avoid the exponential-in-$d$ mesh dependence even without variable-time amplitude amplification.","The explicit ingredients $\\kappa(A)=O((d+1)3^{d+1}h^{-2})$ and $g=O(3^{d/2}h^{-1/2})$ for the trace projection are what combine into the reported $3^{3d/2}h^{-2.5}$ exponent."],"supporting_citations":[{"why":"establishes the Caffarelli–Silvestre extension that converts the fractional Laplacian into a local Dirichlet-to-Neumann problem in one higher dimension.","marker":"[6]"},{"why":"supplies the FEM discretization on graded meshes, error estimates, and exponential decay of the extension, which justify truncation and the linear system.","marker":"[47]"},{"why":"provides the Schrödingerization-based linear-system solver whose query complexity Lemma 2.1 is asserted to follow from 'with some modifications'.","marker":"[36]"},{"why":"introduces the warped-phase Schrödingerization technique that turns the steady-state ODE into a Schrödinger-type Hamiltonian system.","marker":"[29,39,40]"},{"why":"is the source of the variable-time amplitude amplification invoked in Remark 2.2 to reach the improved h^{-2.5} scaling.","marker":"[12]"},{"why":"supplies the elliptic matrix spectral analysis used in Theorem 3.1 for the condition number and norm bounds.","marker":"[25]"}],"fun_headline_variants":["Quantum fractional Poisson: mesh cost independent of dimension","Schrödingerization speeds up fractional Poisson on quantum","Quantum algorithm for fractional Poisson dodges dimension curse","Fractional Poisson quantum solver: exponential mesh-size gain","Dimension no longer plagues quantum fractional Poisson cost"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is Lemma 2.1, the query-complexity estimate for the Schrödingerization-based linear-system solver, which the paper states as following 'with some modifications' from another work without proof, together with the variable-time amplitude amplification of Remark 2.2 that is invoked to reach the improved exponent but is explicitly not constructed; if either is unsound, the claimed $\\widetilde{\\mathcal{O}}(h^{-2.5})$ scaling gives way to the paper's own weaker $\\widetilde{\\mathcal{O}}(d^2 3^{5d/2}h^{-4.5})$ bound.","fun_headline_variants_meta":{"raw":{"variants":["Quantum fractional Poisson: mesh cost independent of dimension","Schrödingerization speeds up fractional Poisson on quantum","Quantum algorithm for fractional Poisson dodges dimension curse","Fractional Poisson quantum solver: exponential mesh-size gain","Dimension no longer plagues quantum fractional Poisson cost"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001012,"raw_usage":{"total_tokens":4345,"prompt_tokens":1085,"completion_tokens":3260,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":701,"completion_tokens_details":{"reasoning_tokens":3187}},"tokens_in":701,"tokens_out":3260,"duration_ms":20852,"temperature":1.0,"reasoning_tokens":3187,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T04:15:11.552725+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Implement the proposed pipeline for a fixed dimension, say $d=2$ or $d=3$, on the FEM-discretized fractional Poisson equation over a sequence of mesh sizes $h$, and count block-encoding queries needed to prepare an $\\epsilon$-accurate solution with success probability $\\Omega(1)$. If the empirical exponent of $h^{-1}$ exceeds $2.5$, or if the success probability decays polynomially in $h$, the claimed dimension-independent advantage fails; a direct check of Lemma 2.1 against a stiff positive-definite matrix would settle the same question.","supporting_citations":[{"cited_title":"Caffarelli and L","cited_arxiv_id":null,"evidence_quote":"establishes the Caffarelli–Silvestre extension that converts the fractional Laplacian into a local Dirichlet-to-Neumann problem in one higher dimension."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the FEM discretization on graded meshes, error estimates, and exponential decay of the extension, which justify truncation and the linear system."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the elliptic matrix spectral analysis used in Theorem 3.1 for the condition number and norm bounds."}],"review_version":1}