{"id":"bd645745-11dc-4115-82f1-0c2fad0e4a41","arxiv_id":"2607.07692","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"Explicit commutator-scaling error bounds are derived for truncated BCH and Zassenhaus formulas in the skew-adjoint (unitary) setting, generalizing Lie–Trotter and Strang splitting bounds.","lead":"The paper derives explicit, rigorous error bounds for truncated Baker–Campbell–Hausdorff (BCH) and Zassenhaus formulas when the operators are skew-adjoint, as in quantum evolution. These bounds, expressed as linear combinations of nested commutator norms, are useful for analyzing quantum simulation gate complexity and Trotter error.","discovery_kind":"unclear","skeptic_critique":{"model":"glm-5.2","headline":"No significant objection identified. The proof structure is sound; the main limitation (finite-dimensional skew-adjoint restriction) is inherent to Lemma 2.1 and honestly stated.","rationale":"The reader correctly identified the two main limitations: (1) the restriction to finite-dimensional skew-adjoint operators, which is inherent to Lemma 2.1's use of unitarity, and (2) the unproven tightness of the triangle inequality step for general m. Neither of these constitutes a correctness issue — they are scope and sharpness limitations that the paper honestly acknowledges. The proof structure is sound: Lemma 2.1 provides the integral bound, Lemma 2.2 provides the Taylor remainder with integral form, and the combination yields explicit commutator bounds. The consistency checks (recovering Lie–Trotter for m=1, Strang for the symmetric case) and the numerical verification for m=2,3 provide empirical support. The Zassenhaus bounds appear genuinely novel. The work makes a solid contribution within the established program of product formula error analysis. No adjustment to the ACCEPT verdict is warranted.","tokens_in":24557,"tokens_out":727,"duration_ms":824447,"concrete_test":"Verify the m=3 bound in the Appendix (eq. 48) by independently recomputing C⁽³⁾₄ from the formula C^{[m]}_{m+1} = (1/(m+1))‖r_m‖ + (1/(m+1)!)(‖ad^m_A B‖ + (1/2)‖ad^{m-1}_{A+B} ad_A B‖) with m=3, and confirm it matches the first term of eq. (48) when expanded in the Hall basis. If the coefficients disagree, the general formula has an algebraic error.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim rests on Lemma 2.1, which requires unitarity of F₁(t) and F₂(t) to conclude ‖Ψ(t)‖ = 1 and ‖F₂(t)‖ = 1, yielding the integral bound ‖F₁(t) − F₂(t)‖ ≤ ∫₀ᵗ ‖M₁(s) − M₂(s)‖ ds. This is correct in the stated setting (finite-dimensional skew-adjoint operators). The subsequent application to BCH truncation (Theorem 2.3) and Zassenhaus truncation (eq. 47) follows by expanding M₁(s) − M₂(s) via Lemma 2.2 and bounding each remainder term using (14), which again uses skew-adjointness of the relevant operators to get ‖e^{s·ad_X}‖ = 1. The triangle inequality step from eq. (24) to eq. (25) introduces looseness but not incorrectness. The Corollary 2.4 claim that the bound is a polynomial in t of degree m²−1 follows from the algebraic structure of ad^k_{Z[m]} Ż[m] (eq. 27), which is a straightforward degree-counting argument. The m=1,2 cases recover known Lie–Trotter and Strang bounds, providing consistency checks. The numerical example in §3 confirms tightness within a factor of ~3 for m=2,3 on random matrices. No internal inconsistency or gap in the logic was identified. The restriction to finite dimensions is necessary for the spectral norm argument and is openly acknowledged.","agreement_with_reader":"agree"},"referee_report":{"model":"glm-5.2","summary":"The manuscript presents a general strategy for deriving rigorous, explicit error bounds on the truncation of the Baker–Campbell–Hausdorff (BCH) and Zassenhaus formulas when the operators involved are skew-adjoint (finite-dimensional unitary setting). The core technique (Lemma 2.1) relates the norm of the difference of two unitary evolution operators to the integral of the norm of the difference of their generator operators. By expanding these generators via integral-form Taylor remainders (Lemma 2.2) and exploiting the fact that the adjoint action of a skew-adjoint operator is norm-preserving, the authors obtain bounds that are explicit polynomials in the parameter t with coefficients given by norms of nested commutators. The framework is applied to the two-operator BCH (Theorem 2.3, Corollary 2.4), the N-operator BCH (Theorem 4.1), the symmetric BCH (Theorem 5.1), and the Zassenhaus formula (Section 6). The m=1,2 cases recover known Lie–Trotter and Strang bounds. A numerical example on random 20×20 matrices confirms tightness within a factor of ~3 for m=2,3.","tokens_in":24773,"tokens_out":1569,"duration_ms":363087,"significance":"The problem addressed is well-motivated by quantum simulation and product-formula error analysis, where commutator-scaling bounds are of practical importance. The derivation is parameter-free and self-contained; the BCH and Zassenhaus coefficients are determined by standard recursive constructions, not fitted to data. The bounds are falsifiable and explicitly computable via computer algebra, with expressions provided for m=3 (Appendix) and m=4 (companion website). The recovery of known Lie–Trotter and Strang bounds at lowest order provides a consistency check. The Zassenhaus error bounds appear to be genuinely new. The restriction to finite-dimensional skew-adjoint operators is inherent to the method (Lemma 2.1 requires unitarity for the norm-preservation step) and is openly acknowledged.","major_comments":[{"comment":"§2.2, between Eqs. (18) and (19): The claim that M_1(t) − M_2(t) = O(t^m) is stated as following from the BCH construction, but the justification is only informal here. Since this order condition is load-bearing for Theorem 2.3 (it is what makes R^[m](s) = O(s^m) in the proof), a brief explicit verification or a forward reference to the proof of Theorem 2.3 (where it is re-asserted) would strengthen the logical flow. As written, the reader must take the claim on trust at the point it is first used.","section":null},{"comment":"§6, Zassenhaus bound for q=3: The bound includes the term (3/5) t^5 ‖[C_2, C_3]‖. The coefficient 3/5 arises from integrating s^4 and appears correct, but the claim immediately following that the q=2 bound (Eq. 10) constitutes an 'optimal bound' (i.e., 'it is not possible to get smaller coefficients') is stated without proof. This optimality claim is load-bearing for the framing of the contribution as sharp. Either a brief justification (e.g., by exhibiting a counterexample to any smaller coefficient) or a softening of the language to 'best known' or 'consistent with the structure of C_3' would be appropriate.","section":null},{"comment":"§4, Theorem 4.1: The remainder term V_{N,m}(t) is bounded via Eq. (37) using the triangle inequality on the sum over n. For general N and m, the looseness introduced by this step is not assessed. While the numerical example in §3 covers N=2, m=2,3, no numerical or analytical evidence is provided for N>2. A brief comment on whether the bounds remain tight for multiple operators (or at least an acknowledgment that tightness has only been verified for N=2) would set appropriate expectations.","section":null}],"minor_comments":[{"comment":"Abstract: The abstract states the operators are 'skew-adjoint,' but the introduction (§1, 'Main contribution') says 'self-adjoint.' Since e^{tA} is unitary for skew-adjoint A (and e^{itA} for self-adjoint A), this terminology inconsistency should be resolved. The body of the paper consistently uses 'skew-adjoint.'","section":null},{"comment":"§2.3, Eq. (24) to (25): The transition from the combined norm ‖[A,[A,B]] + [B,[A,B]]‖ to the sum of norms via the triangle inequality is noted, but the factor of 1/12 on the second term in (25) appears to combine the 1/12 from (24) with the split. This is correct but could confuse a quick reader; a one-line derivation would help.","section":null},{"comment":"§3, Figure 1 caption: 'truncated second-order BCH formula' should read 'truncated BCH formula after two terms' or 'second-order truncation,' as 'second-order' is ambiguous (the error is O(t^3)).","section":null},{"comment":"§5, Theorem 5.1: The statement uses W^[p] in the exponent on the left-hand side but W^[p+1] in the integral on the right-hand side. While this is consistent with the definition (40) (truncation after p terms, indexed as p+1), the notation is potentially confusing. A clarifying remark would help.","section":null},{"comment":"§6, Algorithm 1, line 4: The formula C_n = (1/n) f_{⌊(n−1)/2⌋, n−1} for n≥5 should be checked for consistency with the indexing in line 3. The floor function ⌊(n−1)/2⌋ appears without prior motivation.","section":null},{"comment":"References: [11] (Casas & Murua, arXiv:2604.01026) is cited as a tech report from 2026. If this is a preprint, the journal reference should be updated upon publication.","section":null},{"comment":"The companion website (http://www.gicas.uji.es/Research/bch.html) is referenced multiple times for explicit bounds at higher m. The permanence of this URL cannot be verified at review time; depositing the code/expressions in a permanent repository (e.g., Zenodo) would improve reproducibility.","section":null}],"recommendation":"minor_revision","confidential_remarks":"The paper is a solid, methodical contribution. The main technical content is correct and the proofs are clean. The major comments are about tightening justification for two claims (the O(t^m) order condition and the 'optimal bound' language) and adding a caveat about N>2 tightness—none of these require new mathematics, just clearer presentation. I lean toward acceptance after these revisions. The authors should also be asked to deposit the companion-website material in a permanent archive."},"author_rebuttal":{"model":"glm-5.2","summary":"We thank the referee for a careful reading and for identifying three points where the manuscript can be improved. All three comments are well-taken. We will (1) add an explicit verification of the order condition M_1(t) - M_2(t) = O(t^m) at the point it is first used, (2) soften the optimality claim for the q=2 Zassenhaus bound to 'best known' with a brief justification, and (3) add a comment acknowledging that tightness has only been verified numerically for N=2. No standing objections remain.","responses":[{"response":"The referee is correct that the claim is stated informally at this point and only justified later in the proof of Theorem 2.3. We will add a brief explicit verification in §2.2. Specifically, the recursive construction (17) shows that at each order t^{n-1}, the coefficient of M_1(t) (namely (1/(n-1)!) ad_A^{n-1} B) is matched by the corresponding coefficient nΦ_n plus lower-order commutator terms F_{n-1}(Φ_1,...,Φ_{n-1}) in M_2(t). Since Φ_n is defined precisely to enforce this matching (equation (17)), all terms through order t^{m-1} cancel in M_1(t) - M_2(t), leaving the leading contribution at order t^m. We will include this one-paragraph verification between equations (18) and (19), and also add a forward reference to the proof of Theorem 2.3 where the same cancellation is re-examined in the context of R^[m](s).","revision_made":"yes","referee_comment":"§2.2, between Eqs. (18) and (19): The claim that M_1(t) − M_2(t) = O(t^m) is stated as following from the BCH construction, but the justification is only informal here. Since this order condition is load-bearing for Theorem 2.3 (it is what makes R^[m](s) = O(s^m) in the proof), a brief explicit verification or a forward reference to the proof of Theorem 2.3 (where it is re-asserted) would strengthen the logical flow. As written, the reader must take the claim on trust at the point it is first used."},{"response":"The referee is right that the word 'optimal' is too strong as stated, since we do not provide a proof of optimality (e.g., by exhibiting a counterexample showing that smaller coefficients are impossible). We will soften the language. Specifically, we will replace 'this constitutes an optimal bound: it is not possible to get smaller coefficients in front of the respective commutators' with 'this constitutes the best bound achievable within our framework: the coefficients arise directly from the integral-form Taylor remainders without any further application of the triangle inequality, and are consistent with the structure of C_3.' We believe this accurately characterizes the situation—the bound (10) is obtained without any intermediate triangle-inequality step that would inflate coefficients—but we agree that a formal optimality proof would require a separate argument that we do not currently have.","revision_made":"yes","referee_comment":"§6, Zassenhaus bound for q=3: The bound includes the term (3/5) t^5 ‖[C_2, C_3]‖. The coefficient 3/5 arises from integrating s^4 and appears correct, but the claim immediately following that the q=2 bound (Eq. 10) constitutes an 'optimal bound' (i.e., 'it is not possible to get smaller coefficients') is stated without proof. This optimality claim is load-bearing for the framing of the contribution as sharp. Either a brief justification (e.g., by exhibiting a counterexample to any smaller coefficient) or a softening of the language to 'best known' or 'consistent with the structure of C_3' would be appropriate."},{"response":"This is a fair point. The numerical verification in §3 is limited to N=2, and we have not tested N>2 cases. The triangle inequality in equation (37) is applied to the sum over n inside the integral, and for larger N the recursive structure of C_{k,n} means that more terms accumulate, so some looseness is expected—though the N=2 results suggest it may be modest. We will add a sentence after Theorem 4.1 (or at the end of §4) acknowledging that tightness has only been verified numerically for N=2 and that the N>2 case may warrant further investigation. We will also note that for N=2 the bound reduces to Theorem 2.3 (as shown), so the N=2 numerical evidence provides at least a partial consistency check on the general formula.","revision_made":"yes","referee_comment":"§4, Theorem 4.1: The remainder term V_{N,m}(t) is bounded via Eq. (37) using the triangle inequality on the sum over n. For general N and m, the looseness introduced by this step is not assessed. While the numerical example in §3 covers N=2, m=2,3, no numerical or analytical evidence is provided for N>2. A brief comment on whether the bounds remain tight for multiple operators (or at least an acknowledgment that tightness has only been verified for N=2) would set appropriate expectations."}],"tokens_in":24445,"tokens_out":1190,"duration_ms":310827,"standing_objections":[]},"desk_editor":{"model":"glm-5.2","letter":"This paper gives explicit, computable error bounds for truncated BCH and Zassenhaus formulas when the operators are skew-adjoint (finite-dimensional). The main technique is straightforward: compare two unitary evolution operators satisfying different ODEs, use integral-form Taylor remainders, and exploit unitarity to get norm-1 factors for free. The result is a polynomial in t whose coefficients are linear combinations of nested commutator norms — exactly the commutator scaling that matters for quantum simulation and product-formula analysis. The Zassenhaus truncation bounds appear to be genuinely new; I'm not aware of prior work addressing this. The BCH bounds with explicit constants at arbitrary truncation order m are also more detailed than what exists in the literature, where the only general result I know is Theorem 5.31 in Bonfiglioli–Fulci, which bounds in terms of operator norms rather than commutators. Recovering the Lie–Trotter bound (eq. 21) and the Strang bound (§5) as special cases is a useful consistency check, not padding. The m=2 case giving the clean bound (9) is a nice concrete result. The extension to N operators (Theorem 4.1) and the symmetric BCH formula (Theorem 5.1) are natural and correct generalizations. The proofs of Lemmas 2.1, 2.2, and Theorems 2.3, 4.1, 5.1 check out. Standard techniques applied carefully. The degree-counting argument in Corollary 2.4 (bound is a polynomial of degree m²−1) is correct. The soft spots are real but minor. First, the restriction to finite-dimensional skew-adjoint operators is inherent to Lemma 2.1 — you need unitarity to get ‖Ψ(t)‖ = 1 and ‖F₂(t)‖ = 1. This is honestly acknowledged. Extension to unbounded operators or infinite dimensions would require different tools. Second, the triangle inequality step from eq. (24) to (25) introduces looseness, and tightness for general m is not proven — only checked numerically for m=2,3 on random 20×20 matrices, where the bound overestimates by a factor of about 3. That's acceptable for this type of result; the bound is meant to be rigorous, not tight. Third, the explicit expressions grow exponentially in complexity with m, but the authors provide computer-algebra implementations, which is the right approach. One minor terminology issue: the main-contribution paragraph says 'self-adjoint' but the paper works with skew-adjoint operators throughout. This is presumably the standard iA convention but should be clarified. This paper is for researchers in quantum simulation, geometric numerical integration, and Lie-group methods who need explicit error constants rather than big-O statements. It deserves a serious referee. I'd recommend acceptance pending a check of the explicit m=3 appendix formulas and the online code.","headline":"Solid contribution: explicit commutator-scaling error bounds for truncated BCH and Zassenhaus formulas in the unitary setting. Zassenhaus bounds appear genuinely new. Proofs are clean, limitations honestly stated. Deserves a serious referee.","tokens_in":25268,"tokens_out":1131,"would_cite":true,"duration_ms":94641,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["17B66","34L30","81Q05"],"pacs":[],"model":"glm-5.2","headline":"Explicit error bounds for truncated BCH and Zassenhaus formulas","keywords":["Baker-Campbell-Hausdorff formula","Zassenhaus formula","error bounds","commutator scaling","unitary evolution","truncation error","skew-adjoint operators","quantum simulation"],"falsifier":"If one constructs skew-adjoint matrices A, B where the commutator norms ||[A,[A,B]]|| and ||[B,[A,B]]|| are small but the actual error ||e^A e^B - e^{A+B+[A,B]/2}|| is large, the m=2 bound would be violated.","tokens_in":24624,"feed_emoji":"","tokens_out":1274,"duration_ms":197715,"temperature":0.7,"pith_summary":"The Baker--Campbell--Hausdorff (BCH) formula says that the product of two operator exponentials, e^A times e^B, can be written as a single exponential whose exponent is A + B plus an infinite series of nested commutators. In practice one truncates this series, and the authors ask: how large is the resulting error? They prove that for skew-adjoint operators A and B (the setting of quantum evolution, where exponentials are unitary), the error of truncating the BCH series after m terms is bounded by an explicit polynomial in t whose coefficients are norms of nested commutators of A and B. The degree of this polynomial is m^2 - 1, and its leading term is of order t^{m+1}. For example, keeping terms through the first commutator [A,B] (m=2) yields the bound (1/4)||[A,[A,B]]|| + (1/12)||[B,[A,B]]||. The authors establish analogous bounds for the Zassenhaus formula (the dual decomposition of e^{A+B} into a product of exponentials of commutators), for the symmetric BCH formula arising in Strang splitting, and for products of N operators. In every case the bound depends on commutator norms rather than on the raw norms of A and B, which means the bounds tighten automatically when the operators nearly commute.","feed_headline":"Explicit commutator-based error bounds for truncated BCH and Zassenhaus formulas","feed_subtitle":"For skew-adjoint operators in quantum evolution, truncating the BCH or Zassenhaus series produces errors bounded by explicit polynomials in嵌","key_machinery":"Lemma 2.1: for unitary F_1, F_2 solving dF_i/dt = M_i(t) F_i with skew-adjoint M_i, the bound ||F_1(t) - F_2(t)|| <= integral ||M_1(s) - M_2(s)|| ds. Lemma 2.2: integral-form Taylor remainder for e^{s ad_X} Y, bounded by (1/n!)||ad^n_X Y|| when X is skew-adjoint. Theorem 2.3: general m-term BCH bound. Corollary 2.4: polynomial structure with coefficients C_j^{[m]}. Theorem 4.1: N-operator generalization. Theorem 5.1: symmetric BCH. Equation 47: Zassenhaus bound.","core_discovery":"The central mechanism is a comparison of two unitary evolution operators via their generator equations. The authors write e^{tA}e^{tB} and e^{Z[m](tA,tB)} as solutions of linear ODEs with skew-adjoint generator matrices M_1(t) and M_2(t). A lemma shows that the norm of the difference of two unitary evolution operators is bounded by the time-integral of the norm of the difference of their generators. By expanding M_1 and M_2 in powers of t with integral remainders (using the Taylor expansion of the adjoint action e^{s ad_X} Y), the difference M_1 - M_2 becomes an explicit function of nested commutators. Integrating term by term produces a polynomial in t whose coefficients are linear combicor","pith_inferences":["The restriction to finite-dimensional skew-adjoint operators is essential because the proof uses ||F_2(t)|| = 1 and ||Psi(t)|| = 1, which holds for unitary operators. Extending to unbounded or infinite-dimensional settings would require substitute estimates for the evolution operator norms.","The triangle-inequality step that separates combined commutator norms (e.g., ||[A,[A,B]] + [B,[A,B]]|| into individual terms) introduces looseness. The numerical examples for m=2,3 show the overestimation factor is roughly 2-3, but whether this remains controlled for large m is not established.","The computational complexity of extracting explicit C_j^{[m]} coefficients grows exponentially with m, which may limit practical use to moderate truncation orders unless algorithmic improvements are made."],"forward_implications":["Quantum simulation algorithms using Trotter or product-formula decompositions can use these bounds to determine circuit depth requirements with explicit constants, rather than asymptotic order estimates.","Systems where operators nearly commute benefit directly: the bounds scale with commutator norms, so near-commutativity yields tighter error control without additional analysis.","The Zassenhaus bounds are the first published error estimates for truncated Zassenhaus expansions, filling a gap for applications in periodically driven quantum systems and quantum nonlinear optics.","The polynomial structure (degree m^2-1, leading term t^{m+1}) gives a practical criterion for choosing the truncation order m: one selects m so that the leading commutator norm is small enough for the target accuracy."],"fun_headline_variants":["Bounding truncation errors in BCH and Zassenhaus formulas for unitary evolution","Polynomial error bounds for truncated BCH and Zassenhaus expansions","Explicit truncation error bounds for BCH and Zassenhaus in quantum evolution","Generator comparison yields error bounds for truncated BCH and Zassenhaus","Quantifying truncation error in BCH and Zassenhaus for skew-adjoint operators"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The entire framework depends on the operators being skew-adjoint in finite dimensions, which ensures that the evolution operators in the comparison are unitary and have norm exactly one. This is what makes the integral bound on the generator difference directly control the operator difference without extra factors.","fun_headline_variants_meta":{"raw":{"variants":["Bounding truncation errors in BCH and Zassenhaus formulas for unitary evolution","Polynomial error bounds for truncated BCH and Zassenhaus expansions","Explicit truncation error bounds for BCH and Zassenhaus in quantum evolution","Generator comparison yields error bounds for truncated BCH and Zassenhaus","Quantifying truncation error in BCH and Zassenhaus for skew-adjoint operators"]},"model":"glm-5.2","effort":"high","cost_usd":0.0,"raw_usage":{"total_tokens":1020,"prompt_tokens":522,"completion_tokens":498,"prompt_tokens_details":null},"tokens_in":522,"tokens_out":498,"duration_ms":41188,"temperature":1.0,"reasoning_tokens":440,"cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-09T02:01:04.296338+00:00","model_set":{"reader":"glm-5.2"},"falsifier":"If one constructs skew-adjoint matrices A, B where the commutator norms ||[A,[A,B]]|| and ||[B,[A,B]]|| are small but the actual error ||e^A e^B - e^{A+B+[A,B]/2}|| is large, the m=2 bound would be violated.","supporting_citations":[],"review_version":1}