{"id":"6d497ff7-936d-4f0a-b282-2429aa1d9132","arxiv_id":"2502.04152","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"The paper derives the spectral form factor for generic kicked many-body systems in all three Dyson classes, matching RMT predictions to second order in time.","lead":"The paper derives the spectral form factor for generic kicked interacting quantum systems in all three Dyson universality classes, matching random matrix predictions up to second order in time. This gives a mechanism and a Thouless time scale for when chaotic many-body systems become universal, with concrete predictions for experiments.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The entire derivation hinges on the RPA phase-average in Eq. (14); if real H0 eigenphases are correlated with each other or with the V matrix elements, the permutation identities and the claimed universal SFF fail.","rationale":"The reader and I identify the same weakest point: the RPA in Eq. (14) is the single assumption on which the entire analytic machinery is built. The formal derivation from that assumption appears internally consistent: the inclusion-exclusion corrections for repeated states are handled to the required order in time, the reduced-diagram cancellations are argued in detail, and the final SFFs match the COE, CUE, and CSE RMT predictions. The numerical checks are real evidence, especially the RPA-versus-direct comparison in Fig. 28 and the direct SFF data for all three classes, even though the CSE comparison does not explicitly overlay the RPA formula. What is missing is proof, or a broad numerical demonstration, that RPA holds for generic H0, including clean systems and models with weaker disorder; this limits the scope of the genericity claim but does not invalidate the stated derivation. A targeted numerical experiment comparing exact SFF with the RPA expression for a CSE model and for a clean counterpart would settle whether the RPA caveat is merely formal or actually restricts the results. Until such a test is done, I would not change the accept verdict: the paper is explicit about its assumption and provides credible numerical support, so the appropriate action is to keep the reader's verdict while noting that the genericity claim is conditional on the RPA.","tokens_in":45514,"tokens_out":9337,"duration_ms":110831,"concrete_test":"Perform an independent RPA validation for the T^2=-1 spinful chain of Sec. X.C at L=8-10: (i) compute the exact disorder-averaged SFF from Eq. (9); (ii) compute the RPA prediction by substituting independent uniform phases (with Kramers degeneracy) into Eq. (14) and evaluating the same diagrammatic sum, including Type I, II, and III terms, for the same V; (iii) plot |Delta K(t)|/K(t) for 1<=t<=10^4 and extract the sign and magnitude of the t^2/N term. If the RPA curve does not match the exact SFF for t>t*, or if the CSE correction changes sign, the central claim is not validated for this class. Repeat the comparison for a clean (disorder-free) version of the H0 model to test whether the word 'generic' in the title is justified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing input is the random phase approximation in Sec. II: Eq. (14) replaces the actual eigenphases theta_n of H0 by independent uniform random variables, with Kramers degeneracy enforced for T^2=-1, and treats those phases as independent of the matrix elements V_{n,n'}=<n|e^{-iH1}|n'>. Every subsequent identity depends on this decoupling: the permutation-sum forms in Eqs. (17)-(22), the diagram rules in Secs. IV-VI, the reduced-diagram cancellations in Sec. IX, and the leading-order expressions for K(t) in terms of eigenvalues of M=V*V^*. If physical H0 eigenphases exhibit level repulsion, additional degeneracies beyond Kramers, or correlations with the H0 eigenbasis, then Eq. (14) is not exact and the claimed universal ramp and o(t^2/N) correction are not guaranteed. The paper's numerical support (Fig. 28, and the direct-SFF data in Figs. 17 and 20) is disorder-averaged and concentrated on models with random onsite potentials and long-range interactions; no direct RPA-versus-exact comparison is shown for the CSE/T^2=-1 case, and clean or weakly disordered systems are not tested. This is not an internal inconsistency, because the RPA is explicitly declared, but it is the precise point on which the genericity of the central claim rests.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper derives the spectral form factor K(t) for periodically kicked interacting many-body systems in all three Dyson symmetry classes, working in the basis of the unperturbed Hamiltonian H0 and assuming a random phase approximation for its eigenphases. Under this assumption the authors obtain K(t) ≈ 2t − 2t^2/N for time-reversal-invariant systems with T^2 = 1, K(t) ≈ t for systems without time-reversal symmetry, and K(t) ≈ 2t + 2t^2/N for T^2 = −1, matching the COE, CUE, and CSE random-matrix predictions beyond a Thouless time t*. They also relate t* to the second-largest eigenvalue of the doubly stochastic matrix M = V • V* and derive system-size scalings t* ∝ L^2, L^0, or ln L for strongly interacting fermionic chains, with an extension to two dimensions. The technical core is a diagrammatic expansion organized into four theorems, with explicit second-order algebra in Appendices D and E.","tokens_in":45791,"tokens_out":7579,"duration_ms":85704,"significance":"If the results are correct, this is a valuable step beyond earlier work by Roy and Prosen and by Kos et al.: it extends the analytic RPA-based derivation of the RMT spectral form factor from the COE case to the CUE and, for the first time in this framework, CSE cases, and it gives a unified description of the Thouless-time scaling. The paper contains genuinely useful technical machinery: the diagrammatic rules in Secs. IV–VI, the reduced-diagram cancellations in Secs. VIII–IX, and the explicit second-order computations in Appendices D and E are detailed enough to be checked independently. The numerical checks in Figs. 17, 20, and 28 provide direct evidence for the leading-order ramp and for the t/L^2 collapse in the CUE and CSE cases. The main weakness is that the random phase approximation is assumed rather than derived, and the most novel quantitative claim, the CSE second-order correction, is not directly verified numerically. The paper does not provide code or data, which limits the reproducibility of the numerical figures.","major_comments":[{"comment":"The entire derivation is conditional on the random phase approximation, which is stated rather than derived. The paper's opening claim of 'generic periodically kicked interacting many-body systems' is therefore broader than what is shown: the numerical support (Fig. 28 for T^2 = 1; Figs. 17 and 20 for CUE and CSE ramps) covers disorder-averaged models with long-range interactions and random potentials, and no direct test of Eq. (14) against the actual eigenphase statistics is reported for the T^2 = −1 case. I request that the scope be stated as 'systems supporting the RPA', and that a direct diagnostic of Eq. (14) be added, at least for the CSE model, comparing the phase-averaged identity with exact averages over the H0 eigenphases used in the numerics. Without this, the genericity of the central claim is not established.","section":"Sec. II, Eq. (14)"},{"comment":"The central second-order result for the CSE class, K−1^(2)(t) = +2t^2/N, is not verified numerically. Figure 20 shows only the leading ramp and the t/L^2 collapse; it does not isolate the signed curvature of the SFF. Since the CSE calculation is a principal novelty of the paper, please add a plot of N(K(t) − 2t) versus t/N for the T^2 = −1 fermionic chain, together with the RPA prediction 2t^2/N and, ideally, a direct comparison to the exact Floquet SFF for the same parameters. This check is essential because the second-order sign distinguishes the CSE result from the COE result and is not visible in the leading-order collapse.","section":"Sec. XI C, Eqs. (107)–(110)"},{"comment":"The t* ∝ L^2 scalings rest on the assertion that M is SU(2) invariant and that its second-largest eigenvalue λ1 is doubly degenerate. This is supported by numerical observation and by Trotter-regime mappings (Eqs. (74), (81), (90)) valid for small |J|, |Δ|, but the direct SFF simulations use parameters outside that regime (e.g., J = 1, U0 = 22 in Fig. 28). Please either prove the SU(2) invariance from the structure of V, or provide a systematic numerical check of the claimed degeneracy and of λ1 ≈ 1 − cβ/L^2 for the exact parameter values used in the SFF plots. The same comment applies to the d1 = O(L^ζ) rows in Table II, which drive the t* ∝ ln L scaling.","section":"Sec. X A–C, Table II"},{"comment":"The paper states that only double repetitions are retained in the overcounting corrections, and this truncation is used to derive the second-order terms. It would be helpful to show explicitly that triple and higher repetitions contribute only at higher order in the combined 1/N and t/N expansion when 1 ≪ t ≪ N, and to state the precise regime in which the 'up to second order in time' claim holds. As written, the truncation is an assumption, and a reader cannot tell whether omitted multiplicity corrections could shift the coefficient of the t^2/N term at t close to the Heisenberg time.","section":"Sec. II, Eqs. (20)–(22)"}],"minor_comments":[{"comment":"The symbol V^*_{−c,d} is not defined; if it denotes the matrix element with a time-reversed state, it should be written consistently as V^*_{Tc,d} or similar.","section":"Eq. (107)"},{"comment":"The column header 'Parameters (β)' is confusing: β appears in the text as part of cβ (a model-dependent constant) and is not defined as the Dyson index. Please rename the column or define β explicitly.","section":"Table II"},{"comment":"The statement that the RPA works better in higher dimensions is plausible but unquantified; consider adding a criterion or a short argument beyond the coordination-number remark.","section":"Sec. XII"},{"comment":"No code or data release accompanies the numerical claims; a data availability statement or a deposition of the exact SFF datasets would substantially improve reproducibility.","section":"Figs. 17, 20, 28"},{"comment":"The diagrammatic notation is dense; a combined glossary of the arc, arrow, and vertex conventions for all three symmetry classes would help readers verify the rules.","section":"Sec. IV–VI"},{"comment":"The proof is very brief; a sentence explaining why the two cases in Figs. 14 and 15 exhaust the possibilities would improve readability.","section":"Sec. IX, proof of Theorem 4"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a serious contribution that is likely publishable after revision, but the RPA conditionality and the absence of a direct numerical check of the CSE second-order correction should be addressed before acceptance. I saw no internal inconsistency in the main derivation; the concerns are about scope of the central claim and completeness of validation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a serious analytic derivation of the spectral form factor across all three Dyson classes, and the CSE result is genuinely new. It deserves a serious referee, not a desk reject.\n\nThe genuinely new pieces: the second-order SFF for T^2=1 in generic kicked interacting systems, the demonstration that the CUE second-order term is non-universal, the first CSE many-body SFF derivation, and the unified t* scaling story from the second-largest eigenvalue of M. The appendices are explicit and the reduced-diagram framework, while informal in places, is checkable. The numerics for the T^2=1 and CUE cases support the central claims, and Fig. 28 gives a direct RPA-versus-exact SFF comparison, which is the right kind of evidence. The derivation is not circular: it uses V and the RPA and produces RMT forms without fitted parameters.\n\nThe soft spots, in proportion. The RPA in Eq. (14) is the load-bearing assumption; the stress-test note is right that correlations among H0 eigenphases or with V would invalidate the permutation-sum identities. This is a stated assumption rather than a hidden one, prior work supports it for disorder-averaged long-range models, and the direct numerics back it for the other two classes, so I do not treat it as a fatal flaw. What is genuinely missing is a direct RPA-versus-exact test for the T^2=-1 case; Fig. 20 shows the SFF matches the expected scaling, but it does not isolate whether the phase factorization itself is accurate for CSE. That leaves the CSE genericity claim slightly less supported than the other two. The SU(2) invariance of M and the degeneracies d1 are numerical findings, buttressed by the Trotter-limit spin-chain mapping but not derived for general parameters. And there is no code or data release, which for a paper this long with this many figures is a minor but real reproducibility cost.\n\nWho this is for: anyone working on many-body quantum chaos, SFF, or RMT universality will want to read it, and it fills a genuine CSE gap. I would send it to a referee, ideally someone willing to check the CSE second-order algebra in Sec. XI C and the reduced-diagram cancellations.","headline":"Genuinely new CSE and beyond-leading-order SFF results with a mostly explicit derivation; the RPA assumption is the real soft spot, and the paper deserves refereeing.","tokens_in":46347,"tokens_out":2464,"would_cite":true,"duration_ms":25546,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81Q50","15B52"],"pacs":[],"model":"deepseek-v4-flash","headline":"For generic periodically kicked interacting many-body systems, the spectral form factor beyond a Thouless time exactly reproduces the random-matrix predictions of the circular orthogonal, unitary, and symplectic ensembles, up to the…","keywords":["spectral form factor","random matrix theory","Dyson symmetry classes","many-body quantum chaos","Floquet systems","Thouless time","random phase approximation","doubly stochastic matrix"],"falsifier":"Take a specific kicked model in which the eigenphases of $\\mathcal{H}_0$ are known to be correlated, for example a weakly disordered or integrable $\\mathcal{H}_0$, and compute the SFF directly: the RPA prediction $K^{(1)}(t)=2t\\,\\mathrm{tr}\\,\\mathcal{M}^t$ should already disagree in the linear-ramp regime. More sharply, numerically sample the joint distribution of the $\\mathcal{H}_0$ eigenphases and test the independence-and-uniformity assumption; if the distribution is not uniform on the torus, the derivation's starting point is violated.","tokens_in":45306,"feed_emoji":"🎲","tokens_out":6605,"duration_ms":60980,"temperature":0.7,"pith_summary":"Kumar, Prosen, and Roy set out to show that the spectral form factor (SFF) of generic periodically kicked interacting quantum many-body systems becomes universal in the chaotic phase, reproducing random matrix theory (RMT) predictions for all three Dyson symmetry classes. Assuming only a random phase approximation for the eigenphases of the undriven part, they derive $K(t)=2t-2t^2/\\mathcal{N}$ for time-reversal invariant systems with $\\mathcal{T}^2=1$ (COE), $K(t)=t$ with no universal second-order term in the absence of time reversal (CUE), and $K(t)=2t+2t^2/\\mathcal{N}$ for $\\mathcal{T}^2=-1$ (CSE), beyond a Thouless time $t^*$. They also show that $t^*$ is determined by the second-largest eigenvalue of the doubly stochastic matrix $\\mathcal{M}=\\mathcal{V}\\bullet\\mathcal{V}^*$, where $\\mathcal{V}$ is the kick operator in the eigenbasis of the unperturbed Hamiltonian. For strongly interacting fermionic chains, this gives $t^*\\propto L^2$ with $U(1)$ symmetry, and $t^*\\propto L^0$ or $t^*\\propto \\ln L$ without it, with the same scalings expected in higher dimensions.","feed_headline":"Form factor matches random-matrix curve in all three Dyson classes","feed_subtitle":"Beyond a Thouless time, kicked many-body spectra reproduce COE, CUE, and CSE form factors to second order.","key_machinery":"The central object is the doubly stochastic matrix $\\mathcal{M}=\\mathcal{V}\\bullet\\mathcal{V}^*$, the Hadamard (elementwise) product of the kick matrix $\\mathcal{V}$ with its complex conjugate, in the eigenbasis of $\\mathcal{H}_0$; its Hermitian counterpart $\\tilde{\\mathcal{M}}=\\mathcal{V}\\bullet\\mathcal{V}^\\dagger$ also appears when time reversal is absent or $\\mathcal{T}^2=-1$. The SFF is expanded as a sum over permutations of $t$ time steps, and the random phase approximation factorizes the phase average into permutation sums. A diagrammatic calculus assigns powers of $\\mathcal{M}$ and $\\tilde{\\mathcal{M}}$ to arcs, and passing to reduced diagrams (replacing an arc by $1/\\mathcal{N}$) isolates the universal Type I terms. Theorems 3 and 4 then show that most reduced diagrams cancel in identical pairs, leaving the linear ramp and two extra 'SR' diagrams that produce the second-order correction. The Thouless time is governed by the second-largest eigenvalue $\\lambda_1$ of $\\mathcal{M}$ through $t^*\\simeq(\\ln d_1+1)/|\\ln\\lambda_1|$.","core_discovery":"The paper's central claim is that the late-time spectral statistics of a generic Floquet many-body system are identical to the RMT circular-ensemble SFFs, not merely similar. The linear ramp comes from the identity permutation and its cyclic and anticyclic variants, giving $K^{(1)}=2t\\,\\mathrm{tr}\\,\\mathcal{M}^t$ for COE, $K^{(1)}=t(\\mathrm{tr}\\,\\mathcal{M}^t+\\mathrm{tr}\\,\\tilde{\\mathcal{M}}^t)$ for CUE, and $K^{(1)}=2t(\\mathrm{tr}\\,\\mathcal{M}^t+\\mathrm{tr}\\,\\tilde{\\mathcal{M}}^t)$ for CSE. The second-order correction comes from two extra sub-sequence-reversal-with-repetition diagrams whose reduced diagrams do not cancel, yielding $-2t^2/\\mathcal{N}$ for COE, $+2t^2/\\mathcal{N}$ for CSE, and no universal term for CUE. Beyond $t^*$, all nonuniversal Type III terms decay exponentially, leaving exactly the COE, CUE, and CSE SFFs of RMT.","pith_inferences":["Because $\\lambda_1$ of $\\mathcal{M}$ controls both the ramp onset and the nonuniversal approach, one could invert the relation and use a measured SFF to extract $\\lambda_1$ and thereby infer the symmetry class and conserved-charge structure of an unknown many-body system; the authors do not state this inversion.","The reduced-diagram cancellation suggests the RPA-plus-permutation method may extend to other RMT symmetry classes, such as the Altland-Zirnbauer tenfold classification, since only the pairing structure of time reversal enters the diagram rules; the authors mention this extension only as a future direction.","A direct numerical test in two dimensions with short-range interactions and $U(1)$ symmetry would check the predicted $t^*\\propto L^2$ scaling and the universal second-order correction, thereby validating the higher-dimensional claim.","The RPA is defined for kicked systems, where the eigenbasis of $\\mathcal{H}_0$ supplies the random phases; the present results should not be applied to autonomous (non-kicked) systems without an additional averaging mechanism."],"forward_implications":["Beyond $t^*$, the full SFF curve up to second order in time is the RMT curve: $2t-2t^2/\\mathcal{N}$ for COE, $t$ for CUE, and $2t+2t^2/\\mathcal{N}$ for CSE; any sustained deviation flags nonuniversal physics.","The Thouless time is not an adjustable parameter; it is fixed by the second-largest eigenvalue $\\lambda_1$ of $\\mathcal{M}$, making $t^*$ computable for large systems from single-particle data when $\\mathcal{M}$ is $SU(2)$ invariant.","In strongly interacting fermionic chains, $U(1)$ conservation forces diffusive scaling $t^*\\propto L^2$; breaking $U(1)$ yields $t^*\\propto L^0$ or $t^*\\propto \\ln L$ depending on whether $\\lambda_1$ is nondegenerate or degenerate with a growing degeneracy.","The same diagrammatic machinery extends to higher spatial dimensions, giving the same $t^*$ scalings, so experimental probes in two- and three-dimensional systems are feasible.","The absence of a universal $1/\\mathcal{N}$ second-order term in the no-time-reversal case is itself a prediction that distinguishes CUE from COE and CSE in a physical system."],"supporting_citations":[{"why":"Introduced the random phase approximation route to the SFF for a kicked qubit chain with long-range interactions, the starting point this paper extends to second order and all Dyson classes.","marker":"[19]"},{"why":"Derived the leading-order SFF for kicked interacting fermionic chains and the $t^*\\propto L^2$ scalings that the present work generalizes and completes.","marker":"[26]"},{"why":"Provided a minimal bosonic model where the RPA and leading-order SFF were tested, supporting the same approximation used here.","marker":"[30]"},{"why":"Extended RPA-based SFF calculations to mixtures of multiple species, further validating the random phase approximation in interacting systems.","marker":"[37]"},{"why":"Defined the circular orthogonal, unitary, and symplectic ensembles whose SFFs are the universal targets of the derivation.","marker":"[9]"},{"why":"Supplied the eigenvalue-correlation functions and SFF formulas for the circular ensembles that the paper matches up to second order.","marker":"[10]"},{"why":"Gave the diagonal approximation leading-order SFF, the semiclassical analogue of the identity-permutation contribution.","marker":"[11]"},{"why":"Introduced Sieber-Richter pairs, the semiclassical counterpart of the off-diagonal diagrams that yield the second-order correction.","marker":"[12]"},{"why":"Computed the leading off-diagonal approximation for hyperbolic systems, analogous to the two extra SR diagrams here.","marker":"[13]"},{"why":"Showed $t^*\\propto \\ln L$ behavior in a spatially extended circuit model, the comparison point for the logarithmic scaling found here.","marker":"[21]"}],"fun_headline_variants":["All three Dyson classes match RMT spectral form factor","Beyond Thouless, SFF equals COE, CUE, CSE","Exact RMT form factor in each Dyson class","Kicked many-body SFF becomes universal across symmetries"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole derivation rests on the random phase approximation: the eigenphases of $\\mathcal{H}_0$ are treated as independent and uniformly distributed on $[0,2\\pi)$, with only Kramers degeneracy when $\\mathcal{T}^2=-1$. If physical eigenphases carry correlations or extra degeneracies, the factorization behind Eq. (14) and every permutation-sum identity built on it fails.","fun_headline_variants_meta":{"raw":{"variants":["All three Dyson classes match RMT spectral form factor","Beyond Thouless, SFF equals COE, CUE, CSE","Exact RMT form factor in each Dyson class","Kicked many-body SFF becomes universal across symmetries"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000281,"raw_usage":{"total_tokens":1807,"prompt_tokens":1232,"completion_tokens":575,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":848,"completion_tokens_details":{"reasoning_tokens":504}},"tokens_in":848,"tokens_out":575,"duration_ms":6082,"temperature":1.0,"reasoning_tokens":504,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T23:19:57.436705+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a specific kicked model in which the eigenphases of $\\mathcal{H}_0$ are known to be correlated, for example a weakly disordered or integrable $\\mathcal{H}_0$, and compute the SFF directly: the RPA prediction $K^{(1)}(t)=2t\\,\\mathrm{tr}\\,\\mathcal{M}^t$ should already disagree in the linear-ramp regime. More sharply, numerically sample the joint distribution of the $\\mathcal{H}_0$ eigenphases and test the independence-and-uniformity assumption; if the distribution is not uniform on the torus, the derivation's starting point is violated.","supporting_citations":[{"cited_title":"Sieber, Leading off-diagonal approximation for the spectral form factor for uniformly hyperbolic systems, J","cited_arxiv_id":null,"evidence_quote":"Introduced the random phase approximation route to the SFF for a kicked qubit chain with long-range interactions, the starting point this paper extends to second order and all Dyson classes."},{"cited_title":"Richter, J","cited_arxiv_id":null,"evidence_quote":"Extended RPA-based SFF calculations to mixtures of multiple species, further validating the random phase approximation in interacting systems."},{"cited_title":"Casati, F","cited_arxiv_id":null,"evidence_quote":"Defined the circular orthogonal, unitary, and symplectic ensembles whose SFFs are the universal targets of the derivation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplied the eigenvalue-correlation functions and SFF formulas for the circular ensembles that the paper matches up to second order."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gave the diagonal approximation leading-order SFF, the semiclassical analogue of the identity-permutation contribution."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduced Sieber-Richter pairs, the semiclassical counterpart of the off-diagonal diagrams that yield the second-order correction."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Computed the leading off-diagonal approximation for hyperbolic systems, analogous to the two extra SR diagrams here."},{"cited_title":"M¨ uller, S","cited_arxiv_id":null,"evidence_quote":"Showed $t^*\\propto \\ln L$ behavior in a spatially extended circuit model, the comparison point for the logarithmic scaling found here."}],"review_version":1}