{"id":"d495b4c0-8c05-4183-bfc8-5bb9797d3ef2","arxiv_id":"1908.07010","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Typical product states, not sigma-z basis states, exhibit the same power-law entanglement growth exponent as the operator entanglement of the time evolution operator in disordered ergodic spin chains.","lead":"Using numerical simulations of disordered quantum spin chains, the authors find that entanglement growth after a quench from typical random product states matches the operator entanglement growth of the time evolution operator, with the same disorder-dependent power-law exponent.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed exponent equality rests on visual plateaus; no quantitative alpha(L) extraction or finite-size extrapolation is provided, so the L-vs-L/2 correspondence could be a finite-window artifact.","rationale":"The reader's weakest-assumption analysis correctly identifies the core soft spot: the paper's headline result is an equality of exponents, but the exponents are never extracted quantitatively. The derivative plateau plots are suggestive and the visual correspondence across two models and several disorder values is genuine supporting evidence, but the central claim is exactly the kind of statement that requires a numerical comparison with uncertainties. Without alpha values, error bars, and a finite-size check, the equality could be an artifact of the accessible time window or of the L-vs-L/2 size mapping. I find no independent flaw in the construction: the models are appropriate, the typical-product-state definition is clear, and the monogamy argument is a plausible mechanism. The missing quantitative analysis is therefore the single load-bearing concern. If the proposed check confirms agreement within error bars and shows convergence in L, the paper's claim would be substantially strengthened; if it does not, the conclusion would need to be weakened to 'consistent within the available window'. Since the reader already rendered a CONDITIONAL verdict, I recommend no change to that verdict.","tokens_in":15750,"tokens_out":9375,"duration_ms":107113,"concrete_test":"Reanalyze the raw data behind Figs. 2 and 3: for each disorder W and each system size, define an objective plateau criterion (e.g., maximal time window over which d ln<S>/d ln t varies by less than a tolerance), fit S(t) = a t^alpha + b over that window using bootstrap over disorder realizations, and produce alpha_psi(L) and alpha_U(L/2) as functions of 1/L. If the extrapolated L -> infinity exponents differ by more than the combined statistical error, the claimed equality is not established. A minimal version using only L=26 vs L=13 should still be reported, with the same plateau-selection rule applied to both curves.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim in Sec. V is that wave-function entanglement from typical product states and operator entanglement of U(t) share the same disorder-dependent power-law exponent across the ergodic phase. The only quantitative evidence is the visual overlap of plateaus in the logarithmic derivative d ln S / d ln t shown in Figs. 2(e-h) and 3(e-h). This is not sufficient to establish equality of exponents. First, the plateau windows are finite and system-size dependent; at W=1 the useful window before saturation is barely one decade, so the extracted exponent is sensitive to the chosen fitting range. Second, no exponent values or error bars are reported anywhere, so 'identical' cannot be distinguished from 'close but drifting with L'. Third, the wave-function data (L=20,24,26) and operator data (L=10,12,13,14) are never compared on a common scale, such as time normalized by the saturation time, and no finite-size extrapolation of alpha is attempted. The L vs L/2 mapping is justified only by equal maximal entanglement entropy, not by equal finite-time dynamics; if alpha_psi(L) and alpha_U(L/2) converge to different limits as L grows, the headline correspondence fails. The same quantitative gap affects the additional claim that sigma_z product states grow faster: the exponents in Fig. 4 are also read visually from derivative plateaus without fitting or uncertainty quantification.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies entanglement growth after quenches from \"typical\" product states (random Haar states on each half of the bipartition, corresponding to maximal bond dimension in the MPS ansatz of Eq. (13)) in a disordered static XYZ chain and a Floquet counterpart, both chosen to lack conserved densities beyond energy (and energy in the Floquet case). The central claim is that the wave-function entanglement entropy S_psi(t) and the operator entanglement entropy S_U(t) of the time-evolution operator grow as power laws with the same disorder-dependent exponent alpha across the ergodic phase (W <= 4), whereas sigma_z basis states grow faster. The paper interprets this as resolving an apparent discrepancy between previous wave-function and operator entanglement studies and as evidence for slow information spreading on the ergodic side of the MBL transition.","tokens_in":15922,"tokens_out":3749,"duration_ms":43264,"significance":"If established, the claimed correspondence between state and operator entanglement growth would clarify an important conceptual point: the time-evolution operator's complexity is captured by typical product-state wave-function entanglement, not by special low-entanglement initial states. The paper has real strengths: it studies both a static and a driven model with no conserved densities, uses exact Krylov evolution up to L=26 for wave functions and exact operator evolution up to L=14, includes a Rényi-entropy appendix showing the robustness of the power-law behavior over Rényi indices, and proposes a physical mechanism (monogamy of entanglement) for the initial-state dependence. However, the central quantitative claim of identical exponents is supported only by visual inspection of plateaus in logarithmic derivatives, with no exponent values, error bars, fitting ranges, or finite-size extrapolation reported. The claim is therefore not yet established to the standard expected for a numerical demonstration.","major_comments":[{"comment":"The paper's headline claim that S_psi(t) and S_U(t) have identical power-law exponents is based entirely on visual overlap of plateaus in the discretized logarithmic derivative. No exponent values, uncertainties, fitting-window choices, or system-size dependence of the exponent are reported anywhere. At W=1 the useful window before saturation is barely one decade, so the extracted exponent is sensitive to the fitting range and to finite-size effects. Without a quantitative comparison, 'identical exponents' cannot be distinguished from 'close but drifting with L'. I request a table or plot of alpha_psi(L,W) and alpha_U(L,W) with fits and error bars, plus a finite-size extrapolation (or a scaling collapse) to support the equality claim.","section":"Sec. IV A, Figs. 2(e-h), 3(e-h)"},{"comment":"The comparison between S_psi for chains of length L and S_U for chains of length L/2 is justified in the text only by the equality of the maximal entanglement entropies (Hilbert-space dimensions). This is necessary but not sufficient: the power-law growth regime depends on the dynamics, not only on the saturation value. The wave-function and operator data are never shown on a common scale (for example, time normalized by the saturation time), and no finite-size extrapolation of alpha is made for either quantity. The apparent agreement of plateaus could therefore be an artifact of the different accessible system sizes and time windows. Please provide a quantitative finite-size comparison, e.g., alpha as a function of L for both quantities, or a collapse of d ln S/d ln t versus t/t_sat.","section":"Sec. III, comparison of L and L/2"},{"comment":"The additional claim that sigma_z product states grow faster and have a different exponent than typical product states is also read visually from the logarithmic-derivative plateaus in Fig. 4(c,d), without quoting exponents or uncertainties. Since this distinction is central to the paper's resolution of the earlier discrepancy, it requires the same quantitative fitting and uncertainty analysis as the main claim.","section":"Sec. IV C, Fig. 4"}],"minor_comments":[{"comment":"The MPS ansatz in Eq. (13) is not written cleanly: the bulk matrices M^{sigma_k} in C^{chi x chi} and the boundary vectors in C^chi are not distinguished in the notation, and the case chi=1 (including the pure sigma_z state) is not explicitly represented. Please clarify the definition of the state for all chi values.","section":"Eq. (13)"},{"comment":"The caption of Fig. 6 states that the Rényi operator entanglement entropies are for L=24, but the methods section limits exact operator evolution to L<=14. This appears to be a typo (likely L=12); please correct it or clarify the system size.","section":"Fig. 6 caption"},{"comment":"The text says results are averaged over '50-100 disorder realizations' in Sec. II B but over 'approximately 100 realizations' in Sec. IV A. Please make the number of realizations consistent and specify it separately for the static and Floquet cases.","section":"Sec. II B and Sec. IV A"},{"comment":"The statement that 'These results appear to be converged with system size at short enough times' is not supported by any quantitative measure of convergence (for example, comparing alpha at different L in a common time window). Please either add such a measure or soften the claim.","section":"Sec. IV A, Fig. 2"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the journal's scope and the conceptual claim is timely. The main issue is that the central numerical claim of identical exponents is not quantified. This is fixable by reanalysis of existing data, so I recommend major revision rather than rejection. I would also encourage the authors to state error bars on all disorder-averaged quantities, since they are currently absent."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Chen — quick take on 1908.07010. The useful result: in disordered ergodic spin chains, quenches from typical product states give wavefunction entanglement growth with the same disorder-dependent power-law exponent as the operator entanglement of U(t), while σz product states grow faster. That resolves a genuine discrepancy in the MBL literature. The static and Floquet models are well chosen, and the monogamy-based explanation for why σz states are unrepresentative is sensible.\n\nWhat's new is mostly the choice of initial ensemble. The paper makes a measure-based argument that maximal-bond-dimension product states dominate the space of separable states, and the numerics back that up. The correspondence with linear entanglement growth in clean systems (Jonay et al.) is extended to the sublinear power-law regime. Credit is due for testing both static and driven systems and for showing the same behavior across W = 1 to 4.\n\nThe soft spot is the one you'd expect from reading the figures: the central claim of identical exponents rests on visual plateaus in d ln S / d ln t. There are no quoted α values, no error bars, no finite-size extrapolation. The L vs L/2 mapping is motivated by equal Hilbert space dimension, but it is a choice; without comparing α_ψ(L) and α_U(L/2) as functions of system size, the equality is a plausible observation, not a demonstrated scaling result. At W=1 the time window before saturation is barely a decade, so fitting-range sensitivity is a real concern. The appendix on Rényi entropies is fine but doesn't address this.\n\nThat said, the claim is not load-bearing nonsense; the curves genuinely track each other. I'd send this to a referee rather than desk reject, but the referee should ask for quantitative exponent extraction with uncertainties, a finite-size analysis of α, and ideally a data-release note. Without those, the paper is a strong suggestion rather than a settled resolution.\n\nFor your reading group it's worth a slot; the initial-state dependence and monogamy discussion will generate good discussion. I'd cite it when discussing initial-state dependence in disordered dynamics.","headline":"A useful numerical resolution of the exponent discrepancy, provided you accept visual plateaus as evidence; the core idea is right but needs quantitative teeth.","tokens_in":16508,"tokens_out":1895,"would_cite":true,"duration_ms":20602,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Typical product states make the entanglement entropy of a quenched wave function and the entanglement entropy of the time-evolution operator grow with the same power-law exponent.","keywords":["entanglement entropy","operator entanglement entropy","typical product states","power-law growth","many-body localization","disordered spin chains","Floquet systems","monogamy of entanglement"],"falsifier":"For the static model at $W=2.0$, compare the logarithmic-derivative plateau for $S_\\psi$ at $L=26$ with that for $S_U$ at $L=13$ over a common time window extended by a factor of two; if the plateau values differ by more than the variation already seen across available system sizes, or if either plateau moves monotonically with $L$ at a fixed time, the claimed exponent equality is not established.","tokens_in":15482,"feed_emoji":"⚛️","tokens_out":10405,"duration_ms":99812,"temperature":0.7,"pith_summary":"Quantum many-body systems usually generate entanglement linearly in time, but near the many-body localization transition disordered spin chains are known to entangle more slowly, as a power law $t^\\alpha$ with an exponent $\\alpha$ that shrinks as disorder grows. This paper tries to establish that this power-law growth is a property of the time-evolution operator itself: the operator entanglement entropy $S_U(t)$ and the entanglement entropy $S_\\psi(t)$ of a wave function quenched from a typical product state grow with the same disorder-dependent exponent, in both a static and a periodically driven spin chain without conserved densities. A typical product state, in this paper's sense, is a product state across the entanglement cut that is random (Haar-distributed) inside each half, meaning maximally entangled within each subsystem. The authors show that special initial states such as $\\sigma_z$ basis states can entangle faster, and argue on the basis of monogamy of entanglement why they are unrepresentative. If this is right, it resolves the earlier mismatch between wave-function and operator entanglement exponents and identifies typical product states as the correct probe of entanglement production.","feed_headline":"Typical product states match operator entanglement growth","feed_subtitle":"Random typical states, not σz basis states, give the same power-law exponent as the evolution operator's entropy.","key_machinery":"The load-bearing object is the family of product initial states generated by the matrix-product-state ansatz in Eq. (13), with random Gaussian matrices $M^\\sigma$ of bond dimension $\\chi$ connecting the two halves; $\\chi=1$ gives $\\sigma_z$-type product states, while the typical product states used in the main comparison are the maximal-bond-dimension case $\\chi = 2^{\\ell_A/2}$. Two further ingredients carry the argument: the operator entanglement entropy of $U(t)$, computed by vectorizing the unitary into a rectangular matrix $u$ (Eq. 11) and taking its entanglement spectrum, and the monogamy of entanglement, which says the entanglement a subsystem shares with the rest is constrained by how much is already stored inside each half. The wave-function and operator entropies are compared at system sizes $L$ and $L/2$, respectively, so that both live in Hilbert spaces of the same dimension. Together these ingredients show why typical wave-function growth tracks $S_U(t)$: $U(t)$ must account for all initial states, and typical product states dominate the uniform measure over separable states.","core_discovery":"The central claim is a correspondence: for a disordered XYZ spin chain in the ergodic phase (disorder $W \\leq 4$), the disorder-averaged wave-function entanglement entropy $S_\\psi(t)$ after a quench from a typical product state—defined as a tensor product of random Haar states on the two subsystems, equivalently the maximal-bond-dimension limit of the paper's MPS ansatz in Eq. (13)—grows as $t^\\alpha$ with the same exponent $\\alpha$ as the operator entanglement entropy $S_U(t)$ of the time-evolution operator $U(t)$. The same power-law equality holds in the periodically driven Floquet version, where no energy conservation remains. The paper shows that $\\sigma_z$ product states (bond dimension $\\chi=1$) and other intermediate states can entangle faster, and explains the difference through monogamy of entanglement: a typical product state is already maximally entangled within each subsystem, so degrees of freedom must be freed before they can entangle across the central cut, which slows growth; the time-evolution operator, which must encode entanglement production for every initial state, reflects the typical, not the special, behavior.","pith_inferences":["The monogamy mechanism implies a quantitative ordering by initial bond dimension: the more entanglement stored inside each half at $t=0$, the slower the growth across the central cut; this could be tested by collapsing $S_\\psi(t)$ curves for several $\\chi$ values onto a master curve.","The same correspondence may hold for other entanglement measures: the appendix shows operator Rényi entropies with $0.4 \\leq \\alpha \\leq 2$ share the exponent, which suggests a family of universal exponents rather than a special property of the von Neumann entropy.","A cold-atom or trapped-ion experiment that prepares single-site Haar-random product states could test the prediction that the measured entanglement growth matches the operator entanglement entropy, estimated from randomized measurements of the evolution over equivalent times.","Because the models lack conserved densities, the exponent cannot be tied to subdiffusive transport; the paper's constraint picture points instead to the cost of untying intra-subsystem entanglement, which one would expect to persist in higher-dimensional slow-thermalizing systems."],"forward_implications":["The operator entanglement entropy of $U(t)$ can serve as a state-independent proxy for the entanglement growth seen from generic initial states in the ergodic phase of these disordered chains.","Quenches from $\\sigma_z$ basis states can entangle faster than typical states, so results based on such special initial states do not reflect the behavior encoded in the time-evolution operator.","Slow, disorder-dependent power-law entanglement growth occurs even with no conserved densities, in both static and Floquet settings, so it is a generic precursor of many-body localization rather than a transport signature.","The exponent $\\alpha$ decreases continuously as disorder grows toward the MBL transition and approaches the ballistic value $\\alpha=1$ for weak disorder, and this behavior shows up identically in wave-function and operator entanglement.","Because the time-evolution operator's entropy grows with the same exponent as typical wave functions, $S_U(t)$ can be used to estimate the entanglement growth expected from a generic quench without choosing an initial state."],"supporting_citations":[{"why":"supplies the operator entanglement entropy of the time-evolution operator whose power-law exponents are the reference values the wave-function data are compared against.","marker":"[42]"},{"why":"gives the earlier wave-function entanglement power-law exponents from σz product states, the apparent discrepancy this paper explains.","marker":"[65]"},{"why":"provides the coarse-grained entanglement phenomenology and the prior observation of matching linear growth for typical states.","marker":"[80]"},{"why":"establishes slow dynamics in a driven ergodic system without extensive conserved quantities, motivating the model choice and the claim that slow spreading is generic.","marker":"[66]"},{"why":"the monogamy of entanglement constraint used to explain why χ=1 product states entangle faster across the central cut.","marker":"[82]"},{"why":"the general monogamy inequality for qubits, supporting the same initial-state argument.","marker":"[83]"},{"why":"introduces the entanglement of quantum evolutions, the foundation for defining operator entanglement entropy used throughout.","marker":"[40]"}],"fun_headline_variants":["Typical states set the pace for operator entanglement growth","Same power-law for typical states and evolution operator","Disorder-driven power-law shared by states and operator","Random typical states match operator's entanglement growth","Typical product states mirror evolution operator's entropy"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The central assumption is that the plateau in the slope of log-entropy versus log-time, seen within the simulated time window and for the studied system sizes, already equals the true thermodynamic-limit power-law exponent; no scaling collapse or quantitative finite-size extrapolation is used to confirm that the wave-function and operator plateaus converge to the same number.","fun_headline_variants_meta":{"raw":{"variants":["Typical states set the pace for operator entanglement growth","Same power-law for typical states and evolution operator","Disorder-driven power-law shared by states and operator","Random typical states match operator's entanglement growth","Typical product states mirror evolution operator's entropy"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000345,"raw_usage":{"total_tokens":1894,"prompt_tokens":944,"completion_tokens":950,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":560,"completion_tokens_details":{"reasoning_tokens":878}},"tokens_in":560,"tokens_out":950,"duration_ms":9687,"temperature":1.0,"reasoning_tokens":878,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:29:19.327399+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For the static model at $W=2.0$, compare the logarithmic-derivative plateau for $S_\\psi$ at $L=26$ with that for $S_U$ at $L=13$ over a common time window extended by a factor of two; if the plateau values differ by more than the variation already seen across available system sizes, or if either plateau moves monotonically with $L$ at a fixed time, the claimed exponent equality is not established.","supporting_citations":[{"cited_title":"Operator space entan- glement entropy in a transverse ising chain,","cited_arxiv_id":null,"evidence_quote":"supplies the operator entanglement entropy of the time-evolution operator whose power-law exponents are the reference values the wave-function data are compared against."},{"cited_title":"Many-body localization and delocalization in large quantum chains,","cited_arxiv_id":null,"evidence_quote":"gives the earlier wave-function entanglement power-law exponents from σz product states, the apparent discrepancy this paper explains."},{"cited_title":"The finite group velocity of quantum spin systems,","cited_arxiv_id":null,"evidence_quote":"provides the coarse-grained entanglement phenomenology and the prior observation of matching linear growth for typical states."},{"cited_title":"Extended slow dynamical regime close to the many- body localization transition,","cited_arxiv_id":null,"evidence_quote":"establishes slow dynamics in a driven ergodic system without extensive conserved quantities, motivating the model choice and the claim that slow spreading is generic."},{"cited_title":"Entanglement spreading in a many-body localized sys- tem,","cited_arxiv_id":null,"evidence_quote":"the monogamy of entanglement constraint used to explain why χ=1 product states entangle faster across the central cut."},{"cited_title":"Distributed entanglement,","cited_arxiv_id":null,"evidence_quote":"the general monogamy inequality for qubits, supporting the same initial-state argument."},{"cited_title":"Phenomenology of fully many-body- localized systems,","cited_arxiv_id":null,"evidence_quote":"introduces the entanglement of quantum evolutions, the foundation for defining operator entanglement entropy used throughout."}],"review_version":1}