{"id":"cd4a7969-b305-49a1-a375-a95e229a6f0e","arxiv_id":"2608.10809","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A QFT-based quantum Poisson solver with the Hockney method for free fields works in small ideal simulations, but the multiplication and state-preparation steps destroy any runtime advantage.","lead":"This paper gives a quantum circuit that solves the Poisson equation for periodic and free-field boundary conditions, using the quantum Fourier transform and the Hockney method for the free-field case. Two ways to implement the diagonal multiplication step are compared, and the full algorithm is found to give no runtime advantage over classical solvers in its current form.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Verification is qualitative only: the 2D analytic reference violates the Hockney zero-tail assumption and the LCU imaginary parts are unexplained, so the central claim is not yet established.","rationale":"The reader's weakest assumption concerned the Hockney zero-tail assumption and the regularization of the Green function. I agree those are real omissions, but I would sharpen the point: the same omission makes the comparison to analytic solutions invalid for the 2D tests, and the unexplained LCU imaginary parts suggest a possible implementation defect that no amount of qualitative agreement can rule out. This is not an internal inconsistency in the derivation; the mathematical structure of the algorithm is coherent, and the one-dimensional free-field tests with large L have negligible tails and show the expected behavior. Rather, the evidential support for the central claim is underdetermined. A direct comparison to the exact discrete Hockney solution is the natural and inexpensive check: it would separate algorithm correctness from test-design mismatch and would also reveal whether the imaginary parts are numerical artifacts or a real error in the LCU block. I therefore do not move the verdict to REJECT or UNVERDICTED; the conditional verdict remains appropriate. The paper deserves acceptance only after such a quantitative check, or after the abstract and verification claims are explicitly downgraded to qualitative demonstrations.","tokens_in":35648,"tokens_out":17907,"duration_ms":194979,"concrete_test":"Classically reproduce the 2D free-field Gaussian test (55): form S_pad by zero-padding to 16x16 per source quadrant, build D from the DFT of the 2D free Green function ln(r)/2pi over [-1,1]^2 with G(0)=0 and the q=0 component zeroed, using the wrap-around ordering described in Section IV, and compute phi_h = (1/sqrt(N)) QFT^dagger(D * QFT(S_pad)). Then run the corresponding Qiskit circuit for both multiplication variants and report the max and L2 error of the post-selected amplitudes relative to phi_h/||phi_h||, together with the norm of the imaginary part. Repeat for the 1D periodic LCU case (53), comparing the circuit output to the independently computed expected state (1/lambda) A QFT|S> from Eq. (36).","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing point is not the algebraic derivation but the verification protocol. The paper's central claim is that the post-selected state after the inverse QFT carries the discretized Hockney solution. The only evidence is visual agreement with infinite-domain analytic solutions (Section IV, Figs. 8-11), which the text itself calls 'qualitative evidence of the correct implementation.' For the 2D free-field tests with Lx=Ly=1 and sigma=0.3/sqrt(ln2), the Gaussian tail at the boundary is about 15% of its peak, so the zero-tail assumption behind the Hockney padding is violated; the analytic infinite-domain solution is therefore not the quantity the algorithm computes. This alone would be a test-design weakness, but it is compounded by the LCU runs producing imaginary parts of order 10^-5 in the 1D periodic case and 10^-3 in the 2D free-field case for a problem whose exact solution is real. The paper attributes these to classical simulation without giving a mechanism. If the circuit output had been compared against the exact discrete Hockney solution using the same padding, Green function regularization, and ordering, a mismatch would localize the error; without that check, the reported agreement cannot distinguish an incorrect LCU or Green-data construction from a correct algorithm compared against an inappropriate reference. Since the abstract's 'verified' claim rests entirely on these comparisons, this is the weakest load-bearing point.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a quantum algorithm for solving the Poisson equation with periodic boundary conditions and, via the Hockney method, free-field conditions. The algorithm amplitude-encodes the discretized source, applies a multi-dimensional QFT, multiplies the transformed state by a diagonal matrix containing the Green function or analytical Fourier factors, and applies the inverse QFT. Two implementations of the diagonal multiplication are studied: an LCU construction and a sequence of multi-controlled rotations. The authors report state-vector simulations for one- and two-dimensional test cases, compare the outputs with analytical solutions, and tabulate success probabilities and resource estimates. They conclude that the LCU variant has roughly twice the success probability but needs more ancilla qubits, while the rotation-gate variant is preferable in circuit depth; they also conclude that the current implementations of state preparation and diagonal multiplication prevent an end-to-end quantum runtime advantage.","tokens_in":35919,"tokens_out":4078,"duration_ms":44784,"significance":"If the central claim is established, the paper gives a useful modular construction for solving the Poisson equation in a free-field or periodic setting with amplitude encoding, and it provides an explicit comparison of two diagonal-multiplication subroutines. The derivations of the DFT/Hockney steps, the LCU post-selection formula (36), and the rotation-angle formulas (42)-(43) are standard and appear correct. The resource analysis in Section V, including Table III and the convergence data for success probabilities in Table II, is a concrete contribution that other researchers can build on. However, the algorithmic novelty is incremental: the circuit is a composition of known QFT, state preparation, and diagonal-multiplication techniques, and the paper itself honestly states that the verification is only qualitative and that no quantum runtime advantage is expected from the current implementation. The significance therefore rests on the correctness of the combination and on the reliability of the verification, which the current numerical evidence does not yet fully establish.","major_comments":[{"comment":"The central claim is that the post-selected inverse-QFT state encodes the discrete Hockney solution, but the verification is only qualitative. The text states that the plots 'were taken here just as a qualitative evidence of the correct implementation', yet the abstract claims functionality is 'verified' by comparison with analytical solutions. No error norms, maximum deviations, or resolution-convergence of the solution field are reported. The paper should quantify the agreement with a discrete reference solution and report error metrics per resolution and per boundary-condition type; otherwise the reader cannot distinguish a correct algorithm from an incorrect one that happens to agree visually.","section":"Section IV, Figs. 8-11; abstract"},{"comment":"For the 2D free-field test with source (55), the domain is [x_s-L/2, x_s+L/2] x [-L/2, L/2] with L=1 and sigma=0.3/sqrt(ln2). The Gaussian source has a value exp(-0.25/sigma^2) at the domain boundary, approximately 0.146 of its peak, so the source is not approximately zero outside the original interval. The Hockney padding assumption in Section IIB is therefore violated, and the infinite-domain analytical solution (56) is not the exact quantity the finite-domain Hockney algorithm computes. The comparison should be made against the exact discrete Hockney solution computed with the same padding, the same Green-function regularization, and the same ordering convention; only then can the agreement localize a possible error.","section":"Section IV, free-field 2D tests, Fig. 10 upper row and sources (55)-(56)"},{"comment":"The LCU simulations produce imaginary parts of the order 10^-5 in the 1D periodic case and 10^-3 in the 2D free-field cases, although the exact solutions are real. The text attributes these values to 'the respective numerical simulation processes on the classical computer' without giving a mechanism, and the alternative implementation produces imaginary parts three or more orders of magnitude smaller in the same tests. An unexplained, implementation-dependent imaginary component in a real-valued problem is a concrete correctness risk. The authors should either identify and fix the source (for example, an inconsistency in the SparsePauliOp decomposition or in the reordering of the QFT output) or demonstrate that the imaginary part converges to zero with increasing resolution and report quantitative error norms.","section":"Section IV, Figs. 9-11 and Table I"},{"comment":"The regularization choices - setting the singular Green-function value to zero and setting the k=0 Fourier component to zero - fix the additive constant of the solution, but the analytical references are not normalized in the same way. The text says a 'mismatching offset was aligned' by considering the real part at the first grid point. This offset procedure should be stated rigorously and verified: the authors should show that the only difference between the computed and analytical fields is an additive constant, or, if higher-order differences exist, quantify them. Otherwise the comparison step itself introduces an uncontrolled degree of freedom.","section":"Section IIB, Eq. (25); Section IV comparison procedure"}],"minor_comments":[{"comment":"The word 'additative' appears repeatedly (e.g., 'additative constant' in the captions of Figs. 8-11 and in the text around Eq. (50)); it should be 'additive'.","section":"Captions and text, Section IV"},{"comment":"The sentence 'a mismatching offset was aligned here via considering just the values of the real part for the first position' is imprecise. The authors presumably subtract a constant determined by that grid point; the procedure should be stated as an explicit formula.","section":"Section IV, comparison paragraph"},{"comment":"The text states that the multi-qubit SWAP gates are omitted and that the diagonal matrix entries are reordered to compensate. It would be helpful to specify exactly how the entries of the diagonal matrix and the second QFT are reordered, so that the description is implementable without ambiguity.","section":"Section IIIB and Section VC"},{"comment":"The resource estimates for the LCU SELECT block rely on the observation that only tensor products of identity and sigma_z appear in the Pauli decomposition. This is stated as an empirical observation in Section IV; the authors should state whether this sparsity is guaranteed for the diagonal matrices considered or only holds for the tested examples.","section":"Section VC, SELECT-gate resource estimates"},{"comment":"The decomposition H = R_x(pi) R_y(-pi/2) holds only up to the global phase i; this is correctly noted in the text, but the displayed equation itself would be clearer if it included the factor i or a note that equality is up to a global phase.","section":"Section VB, Eq. (62)"},{"comment":"The caption says the table documents 'the needed qubit numbers and the success probabilities', but the table also lists the resolution of the source domain; the caption should mention all columns.","section":"Table I caption"}],"recommendation":"major_revision","confidential_remarks":"The paper is honest about its own limitations, and the core derivation is standard, but the verification protocol does not yet support the abstract's 'verified' claim. The authors should be asked to add quantitative error tests against exact discrete Hockney solutions and to resolve the LCU imaginary-part anomaly. The topic is within the scope of the journal and may be of interest to the quantum-CFD community once the verification is tightened."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe core idea—using the Hockney domain-doubling trick to turn a free-field Poisson problem into a periodic one that a QFT-based quantum circuit can solve—is new in this combination and worth taking seriously. The derivations are clean: the DFT/Hockney steps, the LCU formula (36), the rotation-angle formulas (42)–(43), and the resource counts in Table III all check out as standard, careful work. I also appreciate the honest bottom line: no runtime advantage over classical FFT solvers with the current implementations, and the success probabilities are often tiny.\n\nWhere the paper falls short is the verification. The authors themselves call the plots 'qualitative evidence of the correct implementation,' yet the abstract says the algorithm is 'verified.' That overstates things. For the 2D free-field tests, the Gaussian source has tails of roughly 15% of its peak at the domain boundary, so the Hockney zero-tail assumption is violated. The analytic solution used for comparison is the infinite-domain one, not the discrete Hockney solution the algorithm actually computes. Matching the first point and eyeballing the rest is not enough, especially when the exact discrete reference is cheap to compute classically. The unexplained imaginary parts in the LCU variant (up to 1e-3 in 2D free field) are a symptom of the same problem: without a direct comparison to the exact discrete Hockney solution, you cannot tell whether the algorithm is correct and the reference is approximate, or vice versa. The multi-controlled-rotation variant looks cleaner in this respect, but the LCU issue needs a mechanism or a fix.\n\nNone of this is fatal. The algorithmic framework is sound and the resource analysis is genuinely useful. But the paper needs major revision before publication: add quantitative error norms, compare against the exact discrete solution for the same padding and Green-function regularization, explain or eliminate the LCU imaginary parts, and temper the abstract. With those changes, this will be a solid reference for quantum CFD.\n\nYes, I'd send it to peer review, but with a clear request for those revisions.","headline":"New and useful Hockney-QFT combination, but verification is qualitative and the abstract overclaims: needs major revision.","tokens_in":36464,"tokens_out":2336,"would_cite":true,"duration_ms":22651,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P68"],"pacs":["03.67.Lx"],"model":"deepseek-v4-flash","headline":"The paper claims the Poisson equation, periodic or free-field, can be solved by a quantum circuit of amplitude encoding, quantum Fourier transforms, and a diagonal-matrix multiplication, with Hockney zero-padding handling free-field cases.","keywords":["quantum algorithm","Poisson equation","Hockney method","quantum Fourier transform","amplitude encoding","linear combination of unitaries","free field boundary conditions","computational fluid dynamics"],"falsifier":"Run the algorithm on a source that is exactly zero outside a compact support strictly inside the domain (so no tails reach the boundary), and compare the post-selected amplitudes against a high-precision numerical evaluation of the infinite-domain convolution integral using the paper's offset convention; if the discrepancy does not shrink toward the discretization error as the grid refines, the claim that the circuit solves the free-field equation fails. A simpler version: repeat the 2D Gaussian test with the domain length $L$ increased to at least ten times the source width so that boundary truncation becomes negligible, and check that the computed field converges to the analytic solution rather than to some offset version of it.","tokens_in":35398,"feed_emoji":"⚛️","tokens_out":19529,"duration_ms":149894,"temperature":0.7,"pith_summary":"The paper claims that the Poisson equation — a ubiquitous step in fluid dynamics and field simulations — can be solved on a quantum computer by transcribing the classical Fourier-domain recipe into a quantum circuit: amplitude-encode the source, apply a multi-dimensional quantum Fourier transform, multiply each frequency component by the Fourier-transformed Green function, transform back, and read the solution from the post-selected ancilla subspace. The free-field, unbounded case is handled through the Hockney method, in which the source is zero-padded to double the computational domain so that the periodic discrete Fourier transform reproduces the infinite-space convolution. Two implementations of the non-unitary multiplication step are constructed and compared: a linear-combination-of-unitaries version and a single-ancilla sequence of multi-controlled rotations. The authors report verification against analytical solutions in one and two dimensions via state-vector simulations, noting in Section IV that the comparison plots were taken as qualitative evidence of correct implementation and that ideal simulations give no information about real-hardware performance. They also find that the LCU variant's success probability is about twice that of the multi-controlled-rotation variant at the cost of many more ancilla qubits, and conclude that the current state-preparation and multiplication circuits prevent a quantum speed advantage over classical FFT-based solvers.","feed_headline":"Solve free-field Poisson equations with a quantum Fourier circuit","feed_subtitle":"Hockney zero-padding becomes a one-qubit-per-dimension step, verified in 1D and 2D simulations.","key_machinery":"The load-bearing object is the Hockney method, a classical construction that converts a free-field Poisson problem into a periodic discrete Fourier transform: the source is assumed to vanish outside $[x_0, x_0+L]$, it is zero-padded to an interval of length $2L$, and the Green function's Fourier transform is evaluated over $[-L, L]$ rather than $[0, L]$. The paper's quantum version of this method is the circuit of Fig. 2: amplitude encoding of the padded source, a multi-dimensional quantum Fourier transform (built from one-dimensional QFT circuits applied to subregisters in parallel), a multiplication by the diagonal matrix whose entries are the Green function's DFT values, the inverse multi-dimensional QFT, and an ancilla measurement that post-selects the valid output subspace. Two concrete implementations of the non-unitary diagonal multiplication are given: the LCU method with preparation and selection gates (needing up to $n$ ancilla qubits and decomposing the diagonal into tensor products of the identity and Pauli-Z), and a single-ancilla sequence of multi-controlled $R_y$ and $R_z$ gates that factors each diagonal entry into an absolute value $|A_{i,i}|$ and a complex phase $e^{i\\varphi_i}$. The regularization choices — setting the singular value of the 2D Green function at zero distance to zero, and zeroing the $k=0$ Fourier component so that the computed field has zero spatial mean — are what fix the otherwise free additive constant.","core_discovery":"The paper's central claim is that a Poisson solver can be built from the same three operations as the classical Fourier method — quantum Fourier transform, pointwise multiplication by the Green function's Fourier transform, and inverse quantum Fourier transform — and that this circuit solves the discretized equation for periodic boundary conditions and, through the Hockney method, for free-field conditions. Concretely, the source is amplitude-encoded, the domain is doubled by adding one qubit per dimension whose zero state implements the zero padding, and the multiplication step is embedded in a unitary by way of ancilla qubits whose measured configuration selects the valid output subspace. The abstract states that functionality is 'verified' by comparing state-vector simulations with analytical solutions; in Section IV the authors qualify this, writing that the plots were taken 'just as a qualitative evidence of the correct implementation' and that ideal-simulation results give no insight into the quantitative performance of a real quantum computer. On the two multiplication variants, the paper reports that the LCU version yields a success probability roughly a factor of two higher than the multi-controlled-rotation version in all tested cases, at the price of up to as many ancilla qubits as there are source qubits, while the one-ancilla version allows higher-resolution simulations and has smaller native-gate counts.","pith_inferences":["A decisive test the paper does not run is a source that is exactly zero at and beyond the domain boundary, so the Hockney assumption holds to machine precision; as run, the 2D Gaussian sources have boundary tails of roughly 15% of the peak, making the comparison with the infinite-domain analytical solution approximate rather than exact.","Since the post-selection probabilities converge but remain small (roughly $10^{-2}$ down to $10^{-5}$ in the tested cases), amplitude amplification could in principle turn the conditioned operation into a near-deterministic one; the paper does not explore this route.","The circuit skeleton of amplitude encoding, Fourier transform, diagonal multiplication, and inverse transform applies to any translation-invariant linear equation whose Green function is known in Fourier space; the paper names the Helmholtz equation only as a future step, but the construction itself is not restricted to Poisson problems."],"forward_implications":["A free-field Poisson problem can be embedded in a quantum circuit with only one extra qubit per spatial dimension, since the Hockney zero-padding is realized simply by the state of the added most-significant qubit.","The post-selection success probability converges as the grid is refined for a fixed source distribution, so increasing the resolution does not systematically increase the measurement overhead.","The LCU variant gives roughly twice the success probability of the multi-controlled-rotation variant in all tested cases, but needs up to as many ancilla qubits as the source register, whereas the rotation variant needs a single ancilla and permits simulations at higher resolution.","As a standalone solver the algorithm offers no quantum runtime advantage: the QFT step alone scales better than a parallelized FFT, but the state-preparation and diagonal-multiplication routines dominate the resource count, and the paper concludes that the multiplication step must be substantially improved.","Embedded as a building block in a time-marching CFD solver, the observed low success probabilities would be highly problematic, an obstacle the paper explicitly identifies."],"supporting_citations":[{"why":"supplies the QFT-based periodic-domain Poisson solver in a vortex-in-cell context that this work extends to a full quantum algorithm.","marker":"[1]"},{"why":"supplies the Hockney method, the classical free-field solver that the quantum algorithm implements.","marker":"[3]"},{"why":"provides the free-field Green functions of the Poisson equation used for the analytical reference solutions.","marker":"[5]"},{"why":"provides the multi-dimensional quantum Fourier transform circuit used for the Fourier steps of the algorithm.","marker":"[9]"},{"why":"the earlier quantum Poisson solver with Dirichlet boundary conditions, the main prior approach against which this QFT-based algorithm positions itself.","marker":"[10]"},{"why":"the quantum simulation framework in which the state-vector verification tests were implemented and run.","marker":"[18]"},{"why":"supplies the periodic Green functions used to build the analytical solutions for the periodic-boundary test case.","marker":"[19]"},{"why":"provides the amplitude-encoding state preparation and the native-gate decomposition rules on which the circuit construction and resource estimates rely.","marker":"[20]"},{"why":"defines the linear-combination-of-unitaries method used as the first multiplication variant.","marker":"[21]"}],"fun_headline_variants":["Quantum Poisson solver uses Hockney trick for free fields","Free-field Poisson solved via QFT and Hockney padding","LCU variant doubles success in quantum Poisson solver","Quantum Poisson solver: LCU wins success, costs ancillas"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The free-field claim rests on the Hockney assumption that the source is exactly zero outside $[x_0, x_0+L]$, so zero-padding to double length and using the periodic DFT of the Green function over $[-L, L]$ reproduces the infinite-space convolution; for the 2D tests the Gaussian sources still have visible tails at the boundary, and this discrepancy is never quantified in the paper.","fun_headline_variants_meta":{"raw":{"variants":["Quantum Poisson solver uses Hockney trick for free fields","Free-field Poisson solved via QFT and Hockney padding","LCU variant doubles success in quantum Poisson solver","Quantum Poisson solver: LCU wins success, costs ancillas"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000735,"raw_usage":{"total_tokens":3373,"prompt_tokens":1120,"completion_tokens":2253,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":736,"completion_tokens_details":{"reasoning_tokens":2186}},"tokens_in":736,"tokens_out":2253,"duration_ms":24364,"temperature":1.0,"reasoning_tokens":2186,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T16:53:21.664776+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the algorithm on a source that is exactly zero outside a compact support strictly inside the domain (so no tails reach the boundary), and compare the post-selected amplitudes against a high-precision numerical evaluation of the infinite-domain convolution integral using the paper's offset convention; if the discrepancy does not shrink toward the discretization error as the grid refines, the claim that the circuit solves the free-field equation fails. A simpler version: repeat the 2D Gaussian test with the domain length $L$ increased to at least ten times the source width so that boundary truncation becomes negligible, and check that the computed field converges to the analytic solution rather than to some offset version of it.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the QFT-based periodic-domain Poisson solver in a vortex-in-cell context that this work extends to a full quantum algorithm."},{"cited_title":"The solutions were obtained based on the relation (14)","cited_arxiv_id":null,"evidence_quote":"supplies the Hockney method, the classical free-field solver that the quantum algorithm implements."},{"cited_title":"Steijl and G","cited_arxiv_id":null,"evidence_quote":"provides the free-field Green functions of the Poisson equation used for the analytical reference solutions."},{"cited_title":"Delfs,Basics of Aeroacoustics(lecture notes, 2023)","cited_arxiv_id":null,"evidence_quote":"provides the multi-dimensional quantum Fourier transform circuit used for the Fourier steps of the algorithm."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"the earlier quantum Poisson solver with Dirichlet boundary conditions, the main prior approach against which this QFT-based algorithm positions itself."},{"cited_title":"Bengoechea, P","cited_arxiv_id":null,"evidence_quote":"supplies the periodic Green functions used to build the analytical solutions for the periodic-boundary test case."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"provides the amplitude-encoding state preparation and the native-gate decomposition rules on which the circuit construction and resource estimates rely."},{"cited_title":"Dewitte, J","cited_arxiv_id":null,"evidence_quote":"defines the linear-combination-of-unitaries method used as the first multiplication variant."}],"review_version":1}