{"id":"860d98ce-f9a2-42db-a592-865d478cea64","arxiv_id":"2607.14797","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A small magnetic vector potential on one edge of a periodic Grover walk produces a first-order correction described by a Hermitian matrix H, and the discrete walk converges to a continuous-time quantum walk generated by τH.","lead":"This paper shows that adding a small magnetic phase to one edge of a graph whose Grover walk is periodic makes the walk behave, over long rescaled time, like a continuous-time quantum walk driven by an explicitly built Hermitian matrix. It also proposes a quantitative measure of how robust the walk's periodicity is to such a magnetic perturbation.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"B∂Xa is not a linear subspace, so κ is undefined; the unstated rank-restriction reading is load-bearing, and the C5 example computes an H whose kernel contradicts Theorem 3.2.","rationale":"The reader's weakest-assumption analysis correctly identifies that B∂Xa is not a linear subspace, making κ ill-defined as printed. This is genuinely load-bearing because the entire spectral classification and the kernel decomposition of H depend on κ. The C5 example sharpens the concern: the printed H matrix is full-rank, which would give Rp=0, contradicting the paper's own Table 1 and Theorem 3.2. A straightforward recomputation with the rank-restriction interpretation resolves the contradiction, suggesting the intended mathematics is recoverable but the manuscript must be corrected. These are addressable formal and computational errors rather than evidence that the central continuous-time limit is false, so the reader's CONDITIONAL verdict remains appropriate. I do not see grounds to accept as-is or to reject outright.","tokens_in":18435,"tokens_out":46431,"duration_ms":386768,"concrete_test":"Reformulate κ as rank(Π_{∂Xa}|_{ker(T−λT I)}), then independently recompute H for C5 directly from Definition 3.2 using T eigenvalues λT = cos(2πk/n), and diagonalize the resulting 10×10 matrix. Check whether dim kerH = 2 and whether the nonzero eigenvalues coincide with Theorem 3.2's µ± = ∓EλT(a,a)√(1−λT²). If both hold, the B∂Xa issue is a presentational defect; if the printed Proposition 5.1 matrix is reproduced instead, the spectral theorem is inconsistent with its own example.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Definitions 3.1 and 3.3 define B∂Xa = {f ∈ C^V | supp(f) ∩ ∂Xa ≠ ∅} and then use κ = dim(ker(T−λT I) ∩ B∂Xa). Since B∂Xa is not a linear subspace, this 'dimension' is mathematically undefined. For a two-dimensional eigenspace of T, one can choose two independent eigenvectors both with support meeting ∂Xa even when the restriction to ∂Xa has rank 1, so the literal reading does not even bound κ by 2. The intended meaning—rank of the projection onto the two endpoint coordinates, or equivalently the number of basis vectors with nonzero endpoint values—is never stated, yet it underpins the classification κ∈{0,1,2}, the definition of H, and the decomposition kerH=S_sim⊕T_per⊕L⊥ in Theorem 3.2. The problem is not cosmetic: for C5, the explicit H in Proposition 5.1 has the form (5I−2J)/25 in each orientation block and is full-rank, predicting dim kerH=0, while Theorem 3.2 and Table 1 require dim kerH = dimS_sim + dimL⊥ = 2. Recomputing H from Definition 3.2 under the rank-restriction reading gives (5I−J)/25, which has the required 2-dimensional kernel. Thus either the definition must be repaired or the example and theorem must be reconciled.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper considers the effect of a magnetic vector potential, supported on a single reference edge, on the discrete-time Grover walk on a finite graph whose unperturbed walk is τ-periodic. The central claim is that the stroboscopic full-period evolution U_β^τ admits the expansion U_β^τ = I + iβτH + O(β²), so that, after ⌈t/β⌉ full periods, the state converges as β→0 to the solution of the continuous-time Schrödinger equation generated by τH. A Hermitian matrix H is constructed from eigenspaces of the discriminant T whose restriction to the two endpoints of the reference edge is two-dimensional. The paper further defines a robustness measure Rp = dim kerH/|A|, claims an explicit spectral formula for H, and tests the theory on cycles, complete bipartite graphs, and paths.","tokens_in":18796,"tokens_out":34534,"duration_ms":292983,"significance":"If the main theorem is correct, the paper offers a genuinely novel mechanism: small magnetic flux on one edge converts the discrete-time periodic Grover walk into an effective continuous-time quantum walk, with a Hamiltonian expressed directly in terms of the spectral data of the underlying graph. The derivation is parameter-free and uses established spectral mapping theorems; the limiting statement is explicit and falsifiable through the examples. However, as printed, the formal definition of the key spectral classification is mathematically undefined, and the main spectral formula contains a factor error that propagates into the examples. These issues are repairable, but they are load-bearing rather than cosmetic.","major_comments":[{"comment":"Definition 3.1 defines B_{∂X_a} = {f∈C^V | supp f ∩ ∂X_a ≠ ∅}, which is not a linear subspace, yet the paper uses dim(ker(T−λ_T I)∩B_{∂X_a}) and writes this intersection as {f_1,f_2} in (3.1). The intersection of an eigenspace with this set is not a subspace; if the restriction to the two endpoints has rank 2 and the eigenspace has dimension >2, there are more than two linearly independent vectors in the set-theoretic intersection. The intended invariant is the rank of the restriction map from ker(T−λ_T I) to C^{∂X_a}, equivalently dim(ker(T−λ_T I)/(ker(T−λ_T I)∩B^⊥_{∂X_a})). Since the classification κ∈{0,1,2}, the definition of H, and the decomposition kerH=S_sim⊕T_per⊕L^⊥ all rest on this notion, the definition must be repaired.","section":"Definition 3.1, Eq. (3.1), Theorem 3.2"},{"comment":"The eigenvalue formula µ±_{λ_T} = ∓E_{λ_T}(a,a)√(1−λ_T²) is inconsistent with Definition 3.2 and with the final expression for H in Lemma 4.4. The correct formula is µ±_{λ_T} = ∓E_{λ_T}(a,a)/√(1−λ_T²). For C5, the printed formula gives nonzero eigenvalues of magnitude (1/5)sin²θ ≈ 0.181, whereas the H built from Definition 3.2 has eigenvalues ±1/5 = ±0.2. The proof contains the corresponding factor inversion: consistency of (4.3) with Lemma 2.3 requires M_σ^{(λ,λ)} = iE_{λ_T}(a,a)/√(1−λ_T²)[[0,1],[-1,0]], not iE√(1−λ_T²). This is load-bearing for the claimed spectrum, for Rp, and for Table 1.","section":"Theorem 3.2, Eq. (3.2); Lemma 4.4, Eqs. (4.3)–(4.6)"},{"comment":"The printed formula for the cycle graph C_n with odd n is wrong. For C5, the proof's own summation gives H' with diagonal 4/25 and off-diagonal −1/25, i.e., (5I−J)/25, which has a one-dimensional kernel in each orientation block, so dim kerH=2 and Rp=1/5 as in Table 1. The displayed formula (diag (n−2)/n², off-diag −2/n²) gives (5I−2J)/25, which is nonsingular and contradicts Theorem 3.2 and Table 1. For odd n the diagonal should be (n−1)/n² and the off-diagonal −1/n²; the even-n formula appears correct. This must be corrected for the main example to support the claimed classification.","section":"Proposition 5.1"},{"comment":"The K_{n,n} row of Table 1 is internally inconsistent. According to Lemma 2.2, dim L^⊥ = dim C_+ + dim C_- = (n−1)² + ((n−1)²+dim ker(T+I)) = 2(n−1)²+1 for K_{n,n}, not 2(n−1). Moreover, the listed Rp = 2(n²−2)/n² does not equal (dimS_sim+dimT_per+dimL^⊥)/|A| using the table's own dimensions. The asymptotic chain lim_{n→∞} Rp(K_{n,n})=1 requires an L^⊥ contribution of order n², so the Table and the surrounding robustness discussion need to be reconciled.","section":"Table 1 and Section 1.2 (K_{n,n} row)"}],"minor_comments":[{"comment":"The path graph is stated to be 2(n−1)-periodic in the introduction to the example, but Proposition 5.3 and its proof use period 2n. These must be reconciled, and the period should be checked against the cited reference.","section":"Section 5.3"},{"comment":"The heading says 'Harmitian H'; this should be 'Hermitian H'.","section":"Definition 3.2"},{"comment":"The derivation of (4.6) contains garbled factors: the line following (4.5) has the square-root factor in the wrong position, and (4.3) is typeset with an unclear 'q' symbol. The final expression (4.6) is correct, but the intermediate displayed equations should be cleaned up.","section":"Lemma 4.4, displayed algebra"},{"comment":"The informal description in Section 1.2 says B_{∂X_a} is the 'subspaces of vectors supported on the end points', while Definition 3.1 defines a set of vectors whose support intersects the endpoints. The terminology 'supported on' usually means support contained in the set; this discrepancy adds to the formal problem and should be fixed.","section":"Section 1.2 vs Definition 3.1"},{"comment":"The ceiling notation 'l n/2 −1 m' is poorly typeset and should be written as ⌈n/2−1⌉. Also, in the eigenvalue computation for T on the cycle, the text writes μ_k = 2cos(2πk/n); for the discriminant T (which is half the adjacency matrix of the 2-regular cycle) the eigenvalue should be cos(2πk/n). This typo should be corrected to avoid confusion.","section":"Proposition 5.1 proof"}],"recommendation":"major_revision","confidential_remarks":"The central construction and the limiting expansion appear to be a promising contribution, but the manuscript in its current form has a formally undefined key definition (B_{∂X_a}), a factor error in the main spectral formula of Theorem 3.2, and inconsistencies in the worked examples. These are not merely presentation issues; they affect the classification and table of robustness values. I recommend major revision. The paper may become acceptable after redefining the restriction-rank notion, correcting the eigenvalue formula to E/√(1−λ_T²), fixing Proposition 5.1 for odd cycles, and reconciling Table 1."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nRead this if you work on quantum walks and want a concrete perturbation result; just don't take the printed definitions and the odd-cycle example at face value. The genuinely new thing is the first-order expansion U_β^τ = I + iβτH + O(β^2) for a periodic Grover walk with a magnetic flux on one reference edge, with H given explicitly through spectral data of the discriminant T. From that they get a clean continuous-time limit and a robustness measure Rp = dim kerH/|A|. The derivation is real: no fitted parameters, Lemma 4.4 is carefully structured, and the examples P_n, C_n, K_{n,n} give concrete H matrices. The path example, where H = 0 and periodicity is perfectly robust, is a nice sanity check.\n\nTwo problems. First, Definition 3.1 defines B_{∂X_a} as the set of vectors whose support intersects {o(a), t(a)}. That is not a linear subspace, so κ = dim(ker(T−λ_T I)∩B_{∂X_a}) is not defined. The text's claim that at most two linearly independent eigenvectors lie in that set is false on the literal reading: in a 2D eigenspace you can pick three independent vectors, all with nonzero values at the endpoint. The intended reading — the restriction to the two endpoint coordinates has rank at most 2, or use vectors supported on ∂X_a — is obvious, but it is load-bearing: κ∈{0,1,2} drives the construction of H and the decomposition kerH = S_sim ⊕ T_per ⊕ L^⊥ in Theorem 3.2. This needs to be stated explicitly, not left implicit.\n\nSecond, Proposition 5.1 for odd cycles is numerically inconsistent with Theorem 3.2. For C5, the printed H is (5I−2J)/25, which is full rank, while Theorem 3.2 and Table 1 require dim kerH = 2. I recomputed from the proof's own formula: the off-diagonal should be −1/25 (not −2/25) and the diagonal 4/25 (not (n−2)/n^2 = 3/25), giving (5I−J)/25 with the required 1D kernel per orientation block. So this is typo-level, but it matters because the example is supposed to illustrate the classification.\n\nThe citation pattern is fine; the periodicity facts cited from the authors' earlier papers are published and parameter-free. No invented entities, no fitted constants.\n\nThis deserves a serious referee — conditional accept, not desk reject. The main theorem is likely correct after the B_{∂X_a} definition is repaired and the odd-cycle example is corrected. I would not cite it in its current form.","headline":"Genuinely new first-order perturbation result for periodic Grover walks under magnetic flux, but the printed definition of B∂Xa is not a subspace and the odd-cycle example contradicts Theorem 3.2 as written.","tokens_in":19272,"tokens_out":8207,"would_cite":false,"duration_ms":71111,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81Q99","05C50"],"pacs":[],"model":"deepseek-v4-flash","headline":"For a periodic Grover walk on a finite graph, a small magnetic vector potential on one edge makes the walk, over many periods, converge to a continuous-time quantum walk; this paper derives the effective Hamiltonian and the subspace where p","keywords":["Grover walk","quantum walk","periodicity","magnetic vector potential","continuous-time quantum walk","effective Hamiltonian","graph spectra","robustness"],"falsifier":"Take a small τ-periodic graph such as C_4 or K_{2,2}, choose a generic initial state, and compare the discrete-time state after ⌈t/β⌉ periods under U_β^τ with e^{-iτtH} for a decreasing sequence β→0; if the norm difference does not converge to zero, Theorem 3.1 fails. Independently, one can check Definition 3.1 algebraically: the set {f: supp f ∩ {o(a),t(a)} ≠ ∅} is not a linear subspace (two endpoint-supported vectors can sum to a vector supported elsewhere), so the dimension κ used in Theorem 3.2 must be read as the dimension of the eigenspace's image on the two endpoint coordinates; an eige","tokens_in":18322,"feed_emoji":"🧲","tokens_out":12934,"duration_ms":107944,"temperature":0.7,"pith_summary":"Grover walks on finite graphs can be exactly periodic, but real systems are subject to perturbations. This paper studies the simplest local perturbation available in the quantum-graph formalism: a magnetic vector potential of strength β placed on one reference edge. Its central claim is that for a τ-periodic Grover walk the perturbed evolution over one period is I + iβτH + O(β^2), so that the state after ⌈t/β⌉ periods converges, as β→0, to the continuous-time quantum walk generated by τH. H is built from the graph's degenerate eigenvalues and a geometric area term at the reference edge, and its kernel decomposes into three subspaces, one of which is the graph's cycle space. The paper interprets the fraction dim ker H / |A| as a robustness coefficient for periodicity, and computes it for paths, cycles, and complete bipartite graphs.","feed_headline":"Small magnetic flux turns periodic Grover walks into continuous time","feed_subtitle":"The paper derives the governing matrix and shows which states keep their periodic behavior under magnetic flux.","key_machinery":"The mechanism is a first-order expansion. The potential changes U_0 to U_β = U_0 + iβσU_0 + O(β^2), where σ is +1 on the reference edge and −1 on its reverse. Because the walk is τ-periodic, U_β^τ = I + iβ Σ_{j=0}^{τ−1} U_0^j σ U_0^{−j} + O(β^2). Periodicity makes all cross-eigenvalue terms cancel, so only diagonal blocks survive; those vanish for λ=±1 and for simple eigenvalues. For a degenerate λ≠±1 with two relevant eigenvectors f_1,f_2, the block is i E_{λ_T}(a,a)√(1−λ_T^2) [[0,1],[−1,0]]. The determinant E_{λ_T}(a,a) of f_1,f_2 at the two endpoints fixes the effective frequency and gives τH.","core_discovery":"On the paper's own terms, the main discovery is Theorem 3.1: if G induces a τ-periodic Grover walk, a magnetic vector potential of strength β on a reference edge a makes the state after ⌈t/β⌉ periods converge, as β→0, to the continuous-time solution of −i∂_t ψ_t = τH ψ_t. Theorem 3.2 adds that H has nonzero eigenvalues ∓E_{λ_T}(a,a)√(1−λ_T^2), coming only from degenerate eigenvalues λ_T of the discriminant matrix, and that ker H = S_sim ⊕ T_per ⊕ L^⊥. E_{λ_T}(a,a) is the determinant, at the two endpoints of a, of the two relevant eigenvectors — the area of the parallelogram they span. Eigenvalues ±1 and all simple eigenvalues contribute nothing to H, so only non-simple eigenvalues shape the","pith_inferences":["Extending the paper's logic, the same first-order expansion should apply to any small local unitary perturbation of a periodic quantum walk, not only a magnetic phase, so a continuous-time limit may be a general phenomenon in discrete-time quantum walks.","A testable consequence of the robustness formula is that among periodic graphs with the same number of edges, those with more degenerate eigenvalues should have smaller R_p; the paper's P_n/C_n/K_{n,n} table is consistent with this but does not prove it.","The robustness measure depends on the choice of reference edge, since H changes when the flux is placed elsewhere; comparing effective Hamiltonians for different reference edges on the same graph would quantify how the continuous-time limit and R_p vary with the gauge choice.","The geometric reading of E_{λ_T}(a,a) as an area suggests that the frequencies of the emerging continuous-time walk are set by how degenerate eigenvectors project onto the two flux-carrying vertices, a feature that could be probed numerically on graphs with movable degenerate subspaces."],"forward_implications":["Every τ-periodic Grover walk has a universal small-flux limit: the discrete-time walk over many periods becomes the continuous-time walk generated by τH, connecting the two quantum-walk paradigms through a perturbation expansion.","The robustness coefficient R_p = dim kerH/|A| can be computed from spectral degeneracies of the discriminant matrix; paths have R_p=1, even cycles have R_p=1/m, and odd cycles R_p=1/(2m+1), so simple spectra are the most stable.","The kernel decomposition kerH = S_sim ⊕ T_per ⊕ L^⊥ exposes three sources of robustness: simple eigenvalues, eigenvectors supported away from the reference edge, and the graph's cycle space; since dim L^⊥ = |E|−|V|+1, edge-rich graphs can be robust for a different reason than spectrally simple graphs.","The effective eigenfrequencies are ±E_{λ_T}(a,a)√(1−λ_T^2), so the rate at which periodicity degrades is controlled by a local geometric area at the reference edge rather than by global graph size alone.","For trees such as the path P_n, H=0 to first order, so periodicity is robust to leading order in the magnetic flux and the graph remains essentially periodic even under the perturbation."],"fun_headline_variants":["Magnetic flux turns periodic Grover walks into continuous-time quantum walks","Small magnetic perturbation drives periodic Grover walks to continuous time","Non-simple eigenvalues control Grover walk periodicity under magnetic flux","Magnetic flux on Grover walks yields continuous-time limit via non-simple eigenvalues"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that each eigenvalue of the discriminant matrix contributes at most a two-dimensional 'endpoint-relevant' space; as printed in Definition 3.1, that space is the set of vectors whose support merely intersects {o(a),t(a)}, which is not closed under addition or scalar multiplication, so the dimension κ and the two-vector basis entering H are only meaningful under a corrected definition (the rank of the eigenspace restricted to the two endpoint coordin","fun_headline_variants_meta":{"raw":{"variants":["Magnetic flux turns periodic Grover walks into continuous-time quantum walks","Small magnetic perturbation drives periodic Grover walks to continuous time","Non-simple eigenvalues control Grover walk periodicity under magnetic flux","Magnetic flux on Grover walks yields continuous-time limit via non-simple eigenvalues"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001002,"raw_usage":{"total_tokens":4046,"prompt_tokens":686,"completion_tokens":3360,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":430,"completion_tokens_details":{"reasoning_tokens":3284}},"tokens_in":430,"tokens_out":3360,"duration_ms":22459,"temperature":1.0,"reasoning_tokens":3284,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T01:10:06.827326+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a small τ-periodic graph such as C_4 or K_{2,2}, choose a generic initial state, and compare the discrete-time state after ⌈t/β⌉ periods under U_β^τ with e^{-iτtH} for a decreasing sequence β→0; if the norm difference does not converge to zero, Theorem 3.1 fails. Independently, one can check Definition 3.1 algebraically: the set {f: supp f ∩ {o(a),t(a)} ≠ ∅} is not a linear subspace (two endpoint-supported vectors can sum to a vector supported elsewhere), so the dimension κ used in Theorem 3.2 must be read as the dimension of the eigenspace's image on the two endpoint coordinates; an eige","supporting_citations":[],"review_version":1}