{"id":"bb974cf1-dfe2-44d0-866b-5c8adb87ce98","arxiv_id":"2507.19345","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A discrete clock plus Duhamel expansion and Gaussian quadrature yields a time-dependent Hamiltonian simulation algorithm with near-optimal query complexity.","lead":"This paper presents a quantum algorithm for simulating time-dependent Hamiltonians by adding a clock register and simulating a larger time-independent system. It claims near-optimal cost, improving on an earlier discrete-clock construction, though the proof has several gaps.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 9's error bound forces M = Omega(T^4 Hdot^2/eps^2), not Omega(T^4 Hdot/eps^2); the missing square propagates into the clock norm and quadrature order and is not accounted for in the proof of Theorem 1.","rationale":"The paper proposes a coherent discrete-clock algorithm for time-dependent simulation, and the central query complexity is plausibly correct: the number of segments T dHmax and the Duhamel truncation order K = O(log(TdHmax/eps)/loglog) are standard, and the query count follows from K block-encodings per segment. The main weakness is exactly what the reader flagged: Lemma 9's dimensional inconsistency. Recomputing the error from the commutator estimate shows the correct M is quadratically larger in Hdot, and this error propagates into ||Hclk|| and hence into the Gaussian quadrature order q. Because q appears logarithmically in the stated gate complexity, the asymptotic theorem may survive a repair, but the proof as written does not support it. This warrants a conditional verdict, consistent with the reader's assessment. The proposed numerical/analytical check would settle whether the corrected q breaks the theorem's stated bounds.","tokens_in":19291,"tokens_out":59734,"duration_ms":560700,"concrete_test":"Independently re-derive Lemma 9 for the test case H(t) = tA with T = 1, including the exact action of [Hclk,Hsys] on |0>|psi> and the first higher-order Zassenhaus terms, and confirm whether the total error is indeed O(T^2 Hdot / sqrt(M)) with threshold M = Theta(T^4 Hdot^2 / eps^2). Then substitute the corrected ||D|| = T^3 Hdot^2 / eps^2 into Eq. (70) and solve for the required quadrature order q; verify whether the resulting q satisfies the gate complexity bound in Eq. (7) and the ancilla bound in Eq. (8) of Theorem 1.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Lemma 9 (Eqs. 28-37) bounds the discretization error of replacing e^{-i(Hclk+Hsys)T}|0>|psi> by the M-step product formula. The commutator estimate in Eq. (35) gives per-step error epsilon_1 = O(T^2 Hdot / (2 M^{3/2})), so summing M steps yields total error epsilon_0 = O(T^2 Hdot / (2 sqrt(M))). Setting epsilon_0 <= eps requires M = Omega(T^4 Hdot^2 / eps^2). However, Eqs. (25), (37), (38), and (40) state M = O(T^4 Hdot / eps^2) and ||Hclk|| = O(T^3 Hdot / eps^2). Unit analysis confirms the printed bound is dimensionally inconsistent: M must be dimensionless, and T^4 Hdot has units of time^2 (with hbar = 1). This off-by-Hdot error understates the clock norm by a factor Hdot. Since ||D|| = ||Hclk|| enters the Gaussian quadrature error bound (Lemma 11, Eq. (70)) as ||D||^{2q}, the subsequent choices of q (Eqs. (71), (87)) and the gate count in Eq. (92) are derived from an underestimated ||D||. The theorem's gate and ancilla complexity, and potentially the LCU normalization if q grows, are therefore not established by the written proof.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a quantum algorithm for simulating a d-sparse time-dependent Hamiltonian H(t) on n qubits over t in [0,T] by embedding the evolution into a time-independent Hamiltonian in a larger space. The construction uses a discrete clock whose Hamiltonian is the logarithm of a cyclic increment operator, together with a system Hamiltonian diagonal in the clock register. The authors invoke Duhamel's principle to separate the clock and system parts, approximate the resulting multiple integrals by scaled Gauss-Legendre quadrature, and implement the resulting linear combination of unitaries via block-encoding and Grover-Rudolph state preparation. The stated main result, Theorem 1, gives O(dTHmax log(TdHmax/epsilon)/loglog(TdHmax/epsilon)) oracle queries plus the gate and ancilla counts shown in Eqs. (6)-(8). The paper is positioned as a proof of concept that the discrete-clock reduction can match the state-of-the-art complexity for time-dependent Hamiltonian simulation, improving on the earlier clock construction of Watkins et al.","tokens_in":19636,"tokens_out":10473,"duration_ms":108901,"significance":"If the proof were completed, the result would be conceptually significant: it would show that a simple discrete-clock Hamiltonian, despite having a large norm, can support a simulation algorithm matching the best known time-dependent simulation bounds, rather than the quadratic-in-time scaling of the earlier discrete-clock work. The algorithmic structure -- Duhamel separation of clock and system, Gaussian quadrature in the interaction frame, and separate Grover-Rudolph preparation of quadrature amplitudes -- is a genuinely different route from the standard truncated Dyson series and interaction-picture algorithms. However, the significance is conditional because the central error analysis contains a dimensionally inconsistent scaling in Lemma 9, an unresolved inconsistency in the number of quadrature points between Eqs. (71) and (87), and a quadrature notation that is too garbled to verify Lemma 11. These issues are not cosmetic; they propagate into the clock norm, the quadrature order, and the final complexity statements.","major_comments":[{"comment":"The scaling of M in Lemma 9 is dimensionally inconsistent and understates the clock norm. The per-step error Eq. (35) is epsilon_1 = O(T^2 Hdotmax/(2 M^{3/2})). Summing over M steps gives epsilon_0 = O(T^2 Hdotmax/(2 sqrt(M))) in Eq. (36). Setting epsilon_0 <= epsilon therefore requires M = Omega(T^4 Hdotmax^2/epsilon^2), not M = O(T^4 Hdotmax/epsilon^2) as printed in Eq. (37) and Eq. (25). The same problem appears in Eq. (12), which states M = Theta(T^3 Hdotmax/epsilon^2). Consequently Eq. (38) and Eq. (40) bound ||Hclk|| = ||D|| by O(T^3 Hdotmax/epsilon^2) whereas the corrected M would give O(T^3 Hdotmax^2/epsilon^2). Since Lemma 11's quadrature error (Eq. (66)) scales as ||D||^{2q}, the subsequent choices of q in Eqs. (71) and (87) and the gate count in Eq. (92) are all derived from an underestimated ||D||. This is a load-bearing error in the proof of Theorem 1.","section":"Section 3, Lemma 9 (Eqs. 28-40)"},{"comment":"The manuscript gives two different lower bounds for the number q of Gauss-Legendre quadrature points. Eq. (71) states q >= Omega(T^6 alpha^2 Hdotmax/epsilon^2), while Eq. (87) states q >= Omega(T^4 Hdotmax/epsilon^2). These differ by a factor T^2 alpha^2 and are never reconciled. Moreover, as printed, both expressions have units of time^2 when Hdotmax has units of inverse time squared, so they are dimensionally inconsistent on their face; the corrected expressions should involve Hdotmax^2, as follows from the corrected ||D|| in Lemma 9. Because the state-preparation gate count in Eq. (92) and the ancilla count in Eq. (98) depend on log q, the proof of the final complexity statement is not established until this inconsistency is resolved.","section":"Section 4.2, Eqs. (71) and (87)"},{"comment":"The quadrature approximation that underlies Lemma 11 cannot be checked as written. Eq. (57) defines scaled nodes s_k x_j / t and scaled weights s_k w_j / t, but the subsequent definitions in Eqs. (58)-(60) introduce u_j(x) = x \\hat s_j / t and v_j(x) = x w_j / t, with \\hat x_{jk} := \\hat s_{jk}, without clarifying whether \\hat s_j denotes an unscaled Legendre root or a scaled root, and whether the composition in Eq. (62) respects the simplex 0 <= s_1 <= ... <= s_k <= t. This ambiguity directly impacts the correctness of Eq. (63) and the error bound Eq. (65). In addition, Lemma 12's proof is not fully rigorous: the statement claims preparation with O(log 1/epsilon) gates using q = Omega((1/epsilon^2) log(1/epsilon)), but the proof's reference to 'the full sum ... = t^{ell+1}/(ell+1)' introduces a parameter t that is not defined in the lemma and is not the integration interval of the unscaled Gauss-Legendre quadrature. Since Eq. (92) relies on Lemma 12 to claim O(log q) state preparation, this gap is load-bearing for the gate complexity claim.","section":"Section 4.2, Eqs. (57)-(62) and Lemma 12"}],"minor_comments":[{"comment":"The weight formula uses P'_n(x_j), but the quadrature has q points and should refer to P'_q(x_j); the notation should be made uniform with the rest of the section.","section":"Section 2.3, Eq. (14)"},{"comment":"The theorem says 'the evolution of H(T) for time T' should be simulated; the intended statement is likely the evolution generated by H(t) for t in [0,T]. This should be corrected.","section":"Theorem 1, statement"},{"comment":"The inequality 'M = O(T^2/epsilon max_s ||Hdot(s)||)' after Eq. (28) is stated as a sufficiency condition but should be an Omega-lower bound on M to make the Riemann-sum error at most epsilon; the direction of the asymptotic notation is misleading.","section":"Section 3, Eq. (29)"},{"comment":"The displayed gate complexity O(K + K log q) is then simplified to a form containing log(T Hdotmax/epsilon); after the corrected scaling in Lemma 9 and the corrected q, the simplification should be re-derived explicitly, since the current text jumps from q = O(T^4 Hdotmax log^3(1/epsilon)/epsilon^2) to a final expression without showing the intervening steps.","section":"Section 4.3, Eq. (92)"},{"comment":"The proof swaps between P'_n and P'_q in the weight formula and in the asymptotic expansion; using a single symbol for the degree would improve readability.","section":"Appendix A, Lemma 7 proof"}],"recommendation":"major_revision","confidential_remarks":"The paper presents an interesting algorithmic idea and a plausible route to matching state-of-the-art time-dependent simulation complexity via a discrete clock, but the proof of the main theorem is currently not correct as written. The missing square in Lemma 9 is a concrete, load-bearing arithmetic error, and the quadrature section needs substantial reorganization before the central claim can be verified. I do not see evidence of a fatal conceptual obstruction, so a careful major revision that corrects the scaling analysis and re-derives the propagated constants would be appropriate."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nQuick take: the paper is worth taking seriously. The central idea is to use a discrete clock Hamiltonian with a log-increment clock, then apply Duhamel's principle and scaled Gaussian quadrature to keep the clock norm from killing the gate count. If the proof can be repaired, this looks like the first clock-based reduction that matches the near-optimal complexity of time-dependent Hamiltonian simulation from other approaches. The authors are honest that their method is formally equivalent to interaction-picture simulation, which is the right thing to say.\n\nWhat is genuinely new: the combination of the simple log-clock with Duhamel plus quadrature is not in WWRL24, and the improvement over that prior work is real (polynomial in time, exponential in precision). This is a proof of concept, not a new lower bound or a complexity breakthrough, and the authors say as much.\n\nThe soft spots are mostly in the derivations. The dimensional concern about Lemma 9 checks out: Eq. (35) gives a per-step error of O(T^2 Hdot / M^{3/2}), so summing M steps gives O(T^2 Hdot / sqrt(M)), requiring M = Omega(T^4 Hdot^2 / eps^2). The printed M = O(T^4 Hdot / eps^2) is dimensionally inconsistent (T^4 Hdot has units of time^2, while M should be dimensionless). That missing square propagates into the bound on ||Hclk|| and the quadrature order q. The theorem's query complexity may still survive because it is stated in terms of Hmax and log factors, but the gate and ancilla counts as written are not established.\n\nAlso, the required number of quadrature points q differs between Eq. (71) and Eq. (87) without explanation, and the state-preparation lemma (Lemma 12) is more of a sketch than a full proof. Lemma 8's use in Lemma 12 needs care: the partial-sum error is O(l(b-a)/q^2), but the normalization constant in Eq. (140) involves a square root that could amplify errors. These are fixable, but they are load-bearing.\n\nBottom line: the central idea is coherent and the errors look repairable without changing the asymptotic claim. The authors should be asked to redo the error analysis carefully. I would send it to peer review; a serious referee can sort out the constants and the missing square. I would also bring it up in the reading group—the clock construction discussion is worth having.","headline":"A genuinely promising clock-based time-dependent Hamiltonian simulation algorithm whose proof has a real dimensional error in Lemma 9 and a few other gaps, but the central idea is coherent and likely fixable.","tokens_in":20167,"tokens_out":2978,"would_cite":true,"duration_ms":29153,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P68"],"pacs":[],"model":"deepseek-v4-flash","headline":"Time-dependent Hamiltonian simulation reduces to time-independent evolution on a larger clock space at the same asymptotic cost as the best known algorithms.","keywords":["time-dependent Hamiltonian simulation","discrete clock Hamiltonian","Duhamel's principle","Gaussian quadrature","linear combination of block-encodings","sparse-access model","quantum simulation complexity","ancilla overhead"],"falsifier":"Evaluate the bound in Lemma 9 on a concrete Hamiltonian with known $\\dot{H}_{\\max}$, e.g. $H(t)=\\sigma_z\\cos(\\omega t)$, by computing the norm difference $\\|e^{-i(H_{\\mathrm{clk}}+H_{\\mathrm{sys}})T}|0\\rangle|\\psi\\rangle-|T\\rangle U(T,0)|\\psi\\rangle\\|$ as $M$ grows. If the error follows $T^2\\dot{H}_{\\max}/(2\\sqrt{M})$, then the paper's $M=O(T^4\\dot{H}_{\\max}/\\epsilon^2)$ is dimensionally off and the clock register must scale as $T^4\\dot{H}_{\\max}^2/\\epsilon^2$; a more favorable scaling would support the stated lemma.","tokens_in":19065,"feed_emoji":"⏱️","tokens_out":12965,"duration_ms":128412,"temperature":0.7,"pith_summary":"Time-dependent Hamiltonians are more realistic than static ones but harder to simulate because the evolution is time-ordered. This paper shows that the problem can be reduced to time-independent evolution in a larger space: a coherent clock register advances stepwise and controls which $H(t)$ acts, so the whole joint system evolves under a single time-independent Hamiltonian. The authors prove that this discrete-clock strategy, with Duhamel's principle and Gaussian quadrature, reaches the same asymptotic query complexity as the best known time-dependent simulation methods, $O(dTH_{\\max}\\log(TdH_{\\max}/\\epsilon)/\\log\\log(TdH_{\\max}/\\epsilon))$ queries for $d$-sparse $n$-qubit Hamiltonians. This matters because it says the intuitive “external clock” picture is not just a proof of concept with extra cost; it is a computationally competitive way to model and simulate time-dependent systems.","feed_headline":"Discrete clock reaches top speed for time-dependent simulation","feed_subtitle":"Coupling the system to a stepwise clock proves as efficient as the best existing simulation methods.","key_machinery":"The load-bearing object is the discrete clock: $M$ clock states $|n\\rangle$ label slices of width $\\delta=T/M$, with $H_{\\mathrm{sys}}$ applying $H(n\\delta)$ on slice $n$ and $H_{\\mathrm{clk}}=(1/\\delta)\\log U_+$ shifting the clock by one step per slice. Its norm is large, about $M/T$, so the paper never simulates $H_{\\mathrm{clk}}$ directly in the time domain; instead a Fourier transform converts it to a diagonal operator $D$ that can be fast-forwarded. Duhamel's principle then separates $D$ from the system part $B$, writing $e^{-i(D+B)t}$ as $e^{-iDt}$ times a series of nested integrals of rotated block-encodings $e^{iDs}Be^{-iDs}$. Scaled Gauss–Legendre quadrature converts each nested integral into a weighted sum of unitary products, and the weights are prepared by an efficient state-preparation routine based on Legendre-weight asymptotics. This combination is what removes the earlier quadratic-in-time overhead and keeps the precision dependence polylogarithmic.","core_discovery":"The central claim, Theorem 1, is that for a $d$-sparse time-dependent Hamiltonian $H(t)$ on $n$ qubits, $t\\in[0,T]$, the time-ordered evolution $U(T)=\\mathcal{T}\\exp(-i\\int_0^T H(t)\\,dt)$ can be simulated to precision $\\epsilon$ with $O(dTH_{\\max}\\log(TdH_{\\max}/\\epsilon)/\\log\\log(TdH_{\\max}/\\epsilon))$ queries to the sparse-access oracles, where $H_{\\max}$ bounds the max-norm of $H(t)$, plus the gate and ancilla costs stated in the theorem. The strategy is to prepare $|T\\rangle\\otimes U(T)|\\psi\\rangle$ by evolving the joint state under $H_{\\mathrm{clk}}+H_{\\mathrm{sys}}$, with $H_{\\mathrm{sys}}=\\sum_n |n\\rangle\\langle n|\\otimes H(n\\delta)$ and $H_{\\mathrm{clk}}=(1/\\delta)\\log U_+$ on $M$ clock states; after a Fourier transform the clock part becomes a large diagonal operator $D$, and Duhamel's principle expands $e^{-i(D+B)T}$ into $e^{-iDT}$ times a series of multiple integrals. Scaled Gauss–Legendre quadrature approximates those integrals as a weighted sum of products of block-encoded unitaries, implemented as a linear combination of block-encodings. The claimed upshot is that the discrete-clock framework matches the performance of interaction-picture and truncated-Dyson methods, and that it improves on the earlier clock construction's quadratic-in-$T$ and polynomial-in-$1/\\epsilon$ overhead.","pith_inferences":["The proof's Lemma 9 appears dimensionally inconsistent: the displayed error $T^2\\dot{H}_{\\max}/(2\\sqrt{M})$ forces $M=\\Omega(T^4\\dot{H}_{\\max}^2/\\epsilon^2)$, not $M=O(T^4\\dot{H}_{\\max}/\\epsilon^2)$; if so, the clock-register size changes but the theorem's query complexity likely survives.","The same Duhamel-plus-quadrature structure should apply to any time-ordered product in which one factor is diagonal and fast-forwardable, for instance Lindblad or Floquet propagators, giving a testable extension beyond unitary Hamiltonian simulation.","A direct numerical check of the commutator error on a simple two-level time-dependent Hamiltonian would settle whether the $M^{-1/2}$ improvement over naive step splitting in Eq. (36) is real or an artifact of the bound."],"forward_implications":["The discrete-clock reduction now carries the same worst-case query complexity as interaction-picture and truncated-Dyson algorithms, so viewing time dependence as a larger time-independent system costs nothing asymptotically in queries.","The ancillary cost is only polylogarithmic in $1/\\epsilon$ plus a logarithm of the clock dimension, so high-precision simulations remain practical in this framework.","Because $H_{\\mathrm{clk}}$ is never simulated directly—only its diagonal Fourier image is fast-forwarded—the method is a concrete template for compiling time-ordered dynamics into a time-independent circuit.","With Hamiltonian rescaling, the method adapts to $L^1$-norm dependence, making it competitive for Hamiltonians whose norm fluctuates strongly over time."],"supporting_citations":[{"why":"Introduces the discrete-clock construction this paper builds on; its quadratic-in-time complexity is the baseline to improve.","marker":"[WWRL24]"},{"why":"Supplies the interaction-picture simulation algorithm and the time-dependent block-encoding HAMT whose query complexity Theorem 1 matches.","marker":"[LW19]"},{"why":"Establishes the $L^1$-norm-scaling benchmark and is the adaptation target named in Remark 2.","marker":"[BCS+20]"},{"why":"Provides the truncated-series LCU and oblivious amplitude amplification framework used to realize the quadrature sum as a quantum circuit.","marker":"[BCC+15]"},{"why":"Gives the efficient state-preparation method used to build the quadrature-weight superposition L.","marker":"[GR02]"},{"why":"Supplies the block-encoding composition lemma used to encode each product $F_k$ of rotated block-encodings.","marker":"[GSLW19]"},{"why":"Lemma 5 on linear combinations of block-encodings is the central tool for assembling the LCU.","marker":"[LW23b]"},{"why":"Asymptotics of Legendre polynomials and roots support the quadrature-weight estimates behind Lemmas 6–8 and Lemma 12.","marker":"[Sze75]"}],"fun_headline_variants":["Discrete clock matches best time-dependent simulation costs","Time-dependent simulation via static clock in larger space","Clock construction makes time-dependent simulation efficient","Discrete clock reaches state-of-the-art simulation efficiency","Static clock boosts time-dependent Hamiltonian simulation"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is Lemma 9's claim that a clock dimension $M=O(T^4\\dot{H}_{\\max}/\\epsilon^2)$ suffices—where $\\dot{H}_{\\max}$ is the maximum rate of change of $H(t)$—to make the discrete-clock evolution approximate the true time-ordered evolution within $\\epsilon$; the paper's own displayed error $T^2\\dot{H}_{\\max}/(2\\sqrt{M})$ actually forces $M=\\Omega(T^4\\dot{H}_{\\max}^2/\\epsilon^2)$, so this premise, as stated, needs repair.","fun_headline_variants_meta":{"raw":{"variants":["Discrete clock matches best time-dependent simulation costs","Time-dependent simulation via static clock in larger space","Clock construction makes time-dependent simulation efficient","Discrete clock reaches state-of-the-art simulation efficiency","Static clock boosts time-dependent Hamiltonian simulation"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000197,"raw_usage":{"total_tokens":1404,"prompt_tokens":1027,"completion_tokens":377,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":643,"completion_tokens_details":{"reasoning_tokens":309}},"tokens_in":643,"tokens_out":377,"duration_ms":4034,"temperature":1.0,"reasoning_tokens":309,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T17:58:20.910511+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the bound in Lemma 9 on a concrete Hamiltonian with known $\\dot{H}_{\\max}$, e.g. $H(t)=\\sigma_z\\cos(\\omega t)$, by computing the norm difference $\\|e^{-i(H_{\\mathrm{clk}}+H_{\\mathrm{sys}})T}|0\\rangle|\\psi\\rangle-|T\\rangle U(T,0)|\\psi\\rangle\\|$ as $M$ grows. If the error follows $T^2\\dot{H}_{\\max}/(2\\sqrt{M})$, then the paper's $M=O(T^4\\dot{H}_{\\max}/\\epsilon^2)$ is dimensionally off and the clock register must scale as $T^4\\dot{H}_{\\max}^2/\\epsilon^2$; a more favorable scaling would support the stated lemma.","supporting_citations":[],"review_version":1}