{"id":"c0255aed-72a1-428b-adf1-bad985d376f4","arxiv_id":"2501.18476","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Quenching a mixed-field Ising chain from paramagnetic to ferromagnetic parameters makes small subsystems display strong non-Markovian, memory-retaining dynamics, while the reverse quench is nearly Markovian.","lead":"By simulating a quantum spin chain, this paper shows that after a sudden change of parameters, small parts of the system can temporarily regain information from the rest, a signature of non-Markovian dynamics. The effect is much stronger when the system is driven from a disordered to an ordered magnetic phase than in the opposite direction.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central claim rests on unverified CPTP-ness of the subsystem dynamics; the paper explicitly measures only backflow, so trace-distance revivals may not establish non-Markovianity.","rationale":"The reader's weakest assumption correctly identifies the decisive issue: the paper does not verify that the subsystem evolution is a legitimate CPTP dynamical map, especially the intermediate maps, and explicitly states that only information backflow is measured. My analysis agrees and sharpens the point: because the initial state is a correlated ground state, even the initial-time map Λ_{t,0} may not be CPTP, so the whole divisibility/backflow framework is not automatically applicable. A direct Choi-matrix test on a small exactly-solvable instance would settle whether the revivals survive under a proper CPTP description. Since this is an addressable gap and the qualitative oscillations are still interesting, the conditional verdict remains appropriate; no change to the reader's recommendation is needed.","tokens_in":15038,"tokens_out":7672,"duration_ms":78812,"concrete_test":"For the same quench on a small chain (N=12, l=2) via exact diagonalization, construct the intermediate map Λ_{t+δ,t} by the Choi-state method: prepare the l-site subsystem maximally entangled with a reference ancilla, place the remaining N-l sites in the reduced state from the actual quench at time t, evolve the full N-site system for δ, partial-trace the environment, and compute the Choi matrix. Repeat for the times where α(t,δ)>0 in Fig. 1. If any Choi eigenvalue is negative, the intermediate map is not CPTP and the revival is not evidence of CP-indivisible dynamics under a legitimate CPTP map. Also construct Λ_{t,0} with the environment in the pre-quench reduced state; if its Choi matrix fails positive semidefiniteness, the subsystem dynamics is not a CPTP process at all and the non-Markovianity interpretation is unsupported. If all Choi matrices are PSD, the concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Sec. III assumes a family of CPTP maps Λ_{t,0} exists for the subsystem (Eq. 6 and surrounding text). For a closed system quenched from a fixed global pure state, such a family exists only if the initial global state is product. Here the initial state is the DMRG ground state at (J,h_x,h_z)=(0.2,1,0), which is not exactly product; so even Λ_{t,0} may fail to be CPTP. The author acknowledges in Sec. III that the numerics only measure information backflow and do not establish CPTP-ness of intermediate maps, relying on numerical invertibility to equate backflow with indivisibility. However, how the maps were constructed and inverted is not described, and for a correlated initial state the intermediate 'map' is not uniquely defined. If the reduced evolution is not CPTP, T_d(ρ^l_{t+δ},ρ^l_t) revivals can arise from initial correlations or assignment effects rather than genuine non-Markovianity. This gap between the observed oscillations and the central claim is load-bearing.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the (non-)Markovian nature of small-subsystem dynamics in an isolated spin-1/2 chain after quantum quenches. Using MPS/TEBD2 simulations with N=200 and Trotter step 0.01, it computes the trace distance T_d(ρ_l_{t+δ}, ρ_l_t) between temporally separated reduced density matrices of l=1..4 sites in the mixed-field Ising chain. It defines a degree of non-Markovianity by summing positive increments of this distance (Eq. 8), and claims that paramagnetic-to-ferromagnetic quenches produce strong revivals (hence strong subsystem non-Markovianity), while the reverse quench is effectively Markovian. The paper also studies the total variation distance between descendingly-ordered eigenvalue vectors of the same reduced states, reporting systematic revivals and two timescales, and offers heuristic arguments based on confinement and quasiparticle propagation.","tokens_in":15278,"tokens_out":15082,"duration_ms":146092,"significance":"If the interpretation were sound, the paper would provide a concrete, numerically accessible signature of whether small subsystems of a closed many-body system behave as Markovian or non-Markovian open systems, with potential implications for thermalization and for ultracold-atom experiments with site-resolved readout. The numerical machinery is standard, convergence checks with smaller Trotter steps are reported, and the directionality of the quench effect is an interesting observation. However, the central interpretation rests on two structural assumptions that are not established: the equivalence between revivals of T_d(ρ_{t+δ},ρ_t) and non-Markovianity, and the existence of a well-defined CPTP dynamical map for the subsystem. The TVD analysis additionally relies on an explicit hypothesis that is in fact false. These gaps prevent the paper, in its current form, from supporting its stated conclusions.","major_comments":[{"comment":"The paper asserts that for a divisible process the quantity T_d(ρ_{t+δ},ρ_t) must be non-increasing in t, and therefore any revival signals non-Markovianity. This does not follow from the data-processing inequality (Eq. 3), which bounds T_d(Λ(ρ),Λ(σ)) for two states ρ,σ, not T_d(Λ(ρ),ρ). In fact, for a divisible unitary process T_d(U_{t+δ}ρU_{t+δ}†, U_tρU_t†) is constant rather than necessarily decreasing, and no general monotonicity of T_d(Λ_{t+δ,t}(ρ),ρ) under divisible maps is known or proven here. Since Eq. (8) defines the degree of non-Markovianity entirely through revivals of this quantity, the central claim that the observed revivals establish non-Markovianity is not justified by the arguments given.","section":"Sec. III, Eq. (6) and surrounding text"},{"comment":"The reduced evolution ρ_l(t)=Tr_E[U_t ρ_global(0) U_t†] is written as Λ(t)[ρ_l(0)] with Λ(t) a CPTP map. For a correlated global initial state, such as the DMRG ground state used here, this linear map is not generally well defined: the mapping from ρ_l(0) to ρ_l(t) depends on the correlations with the environment, so a unique CPTP map independent of the subsystem state does not exist. The paper acknowledges that it only measures information backflow and does not verify CPTP divisibility, but it does not describe how the intermediate maps are constructed or how numerical invertibility is checked in a way that would resolve the ambiguity. Without a legitimate dynamical map, revivals of T_d(ρ_{t+δ},ρ_t) cannot be cleanly attributed to non-Markovianity rather than to initial correlations or the absence of a proper CPTP description.","section":"Sec. III, Eq. (6) and the paragraph starting 'Consider then a subsystem'"},{"comment":"The hypothesis that the total variation distance between descendingly-ordered eigenvalue vectors is non-increasing under CPTP maps is not only unproven but false. A concrete counterexample is provided by single-qubit dephasing: for ρ_1=|+><+| and ρ_2=|0><0|, both have eigenvalue vectors (1,0), so the eigenvalue TVD is 0. Under the dephasing channel D(ρ)=(1-p)ρ+p diag(ρ) with p>0, D(ρ_1) has eigenvalues (1-p/2, p/2) while D(ρ_2) remains |0><0|, giving eigenvalue TVD p/2>0. Thus the TVD can increase under a CPTP map. Consequently, the TVD revivals in Figs. 3-4 and the associated timescales do not constitute evidence of a 'classical characteristic of quantum non-Markovianity,' and the interpretational claim in that section should be removed or replaced with a justified quantity.","section":"Sec. IV, 'Let us hypothesize for the moment...' and Figs. 3-4"}],"minor_comments":[{"comment":"The reported factor-of-50 difference between quench directions is not accompanied by error bars or a sensitivity analysis; since the reverse quench yields degrees near zero, a statement of numerical uncertainty based on Trotter step and bond dimension would strengthen the claim.","section":"Sec. IV, Fig. 2"},{"comment":"The distinction between invertibility and reversibility is appropriate, but the text could clarify that numerical invertibility of a matrix representation does not imply physical positivity of the inverse map.","section":"Sec. III, 'invertible maps are not generally CPTP either'"},{"comment":"The notation for the degree of non-Markovianity varies between 𝒩, N, and script forms, and the sub/superscript for 𝒩_1 is introduced without a formal definition; a table of symbols or consistent notation would improve readability.","section":"Throughout"},{"comment":"The symbol Λ(t) is used both for the map from time 0 to t and in the divisibility discussion; an explicit two-argument notation Λ_{t,s} throughout would avoid ambiguity.","section":"Sec. III, Eq. (6)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is candid about its own limitations, which is commendable, but the combination of an unjustified monotonicity claim, an ill-defined dynamical map for a correlated initial state, and a false TVD hypothesis make the central claims overstated as written. The numerical observations of trace-distance revivals are themselves interesting and may be publishable if reframed as 'information backflow as measured by a specific quantity' without claiming equivalence to standard non-Markovianity, and if the TVD hypothesis is removed or corrected. The paper would also benefit from a standard BLP-style calculation using two distinct initial states (with the same environment preparation) to directly probe divisibility, although the authors note this is not the focus."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. The numerical evidence for a strong quench-direction asymmetry in trace-distance revivals is real and cleanly shown. Standard MPS/TEBD2 with N=200, bond dimension 50, and explicit convergence checks; the data look believable. The finding directly challenges the \"almost Markovian\" claim of Figueroa-Romero et al. for generic closed dynamics, and the systematic TVD timescales are a genuinely new observation. The author also deserves credit for stating plainly in Sec. III that the numerics only measure information backflow and do not establish CPTP-ness of the intermediate maps. That caveat is honest.\n\nThe main weakness is more basic than the caveat suggests. The quantity T_d(rho_{t+delta}, rho_t) is not a standard non-Markovianity witness. The BLP-type measures use pairs of initial states evolved by the same map. Here, the paper compares states of a single trajectory separated by a time lag. For a Markovian but time-inhomogeneous divisible process, this time-lag distance need not decrease monotonically; simple dephasing with a time-dependent rate already gives a nonmonotone curve. So revivals in this measure do not by themselves imply non-Markovianity. The paper's Eq. (6) assumes a family of CPTP maps Λ(t) that is independent of the reference time, which is not true for reduced dynamics of a closed system after a quench, especially when the initial global state is correlated. The DMRG ground state is not a product state, so even the existence of a unique CPTP map from the initial reduced state to later times is not assured. The author acknowledges not checking intermediate maps, but the issue goes deeper: the measure itself may not be a faithful witness even when all maps are CPTP.\n\nThe missing error bars and lack of reproduction artifacts are minor by comparison. The TVD part rests on an explicitly unproven hypothesis about classical data-processing inequalities for eigenvalues; the author flags this, and it is presented as an open question rather than a result.\n\nIf the paper were reframed as reporting non-monotonic distinguishability of temporally separated states and its dependence on quench direction, the numerical content would be solid. As a claim about non-Markovianity in the established sense, it is not yet supported. Still, the question is important and the data are worth taking seriously. I would send this to a serious referee rather than desk reject, with the expectation of heavy revision on the interpretation.","headline":"The quench-direction asymmetry in trace-distance revivals is real and cleanly shown, but the paper's identification of those revivals with non-Markovianity rests on an unproven and possibly false assumption about the measure.","tokens_in":15762,"tokens_out":5226,"would_cite":false,"duration_ms":54579,"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":"In a quenched Ising chain, small subsystems can be strongly non-Markovian when the quench runs from paramagnetic to ferromagnetic order, while the reverse quench is effectively Markovian.","keywords":["non-Markovianity","information backflow","trace distance","quantum quench","mixed-field Ising chain","subsystem dynamics","open quantum systems","matrix product states"],"falsifier":"A direct numerical calculation would settle the claim: extract the intermediate maps $\\Lambda_{t,s}$ for the one- and two-spin subsystems in the paramagnetic-to-ferromagnetic quench and compute their Choi representations (whose positivity is equivalent to complete positivity) at the times when the trace distance revives; if any Choi matrix has negative eigenvalues, the revivals cannot be attributed to a violation of divisibility, so the claimed non-Markovianity would need re-qualification.","tokens_in":14839,"feed_emoji":"⚛️","tokens_out":13206,"duration_ms":109165,"temperature":0.7,"pith_summary":"An isolated quantum many-body system does not have to relax like a memory-less environment: this paper argues that after a quench, the reduced state of a small subsystem can measurably remember its own past. In the mixed-field Ising chain, the author follows the trace distance $T_d(\\rho^\\ell_{t+\\delta}, \\rho^\\ell_t)$ between reduced density matrices of a contiguous block of $\\ell=1,\\dots,4$ spins at two times separated by a lag $\\delta$. The distance revives strongly and persistently when the quench goes from the paramagnetic regime $(J,h_x,h_z)=(0.2,1,0)$ to the ferromagnetic regime $(1,0.1,0.5)$, and is roughly 50 times weaker for the reverse quench, making the reverse dynamics effectively Markovian. These revivals, the paper claims, are information backflow from the rest of the chain into the subsystem, i.e., non-Markovianity, giving an information-theoretic handle on how far-from-equilibrium many-body systems relax. A separate classical distance between ordered eigenvalue spectra shows systematic oscillations whose origin the paper leaves open.","feed_headline":"Quench direction flips whether spin subsystems forget","feed_subtitle":"Trace-distance revivals show strong memory backflow in paramagnetic-to-ferromagnetic quenches; the reverse is effectively Markovian.","key_machinery":"The central diagnostic is the trace distance between the reduced states of the same subsystem at two times separated by a lag, $T_d(\\rho^\\ell_{t+\\delta}, \\rho^\\ell_t)$. Because trace distance is contractive under every completely positive and trace-preserving (CPTP) map, any increase, or revival, of this distance during the evolution signals information backflow from the environment, here the rest of the chain, back into the subsystem. The paper quantifies the effect through the discrete slope $\\alpha(t,\\delta)$ of this distance and the cumulative degree of non-Markovianity $\\mathcal{N}(\\delta)=\\sum_t \\alpha(t,\\delta)$ over all positive slopes, with a same construction for the total-variation distance between ordered eigenvalue spectra. The Markovian-versus-non-Markovian reading is tied to CP-divisibility of the dynamical maps $\\Lambda_{t,s}$; because the numerics only check that the maps are invertible, the paper's stated working assumption is that information backflow and indivisibility coincide for invertible maps.","core_discovery":"The paper's central claim is that subsystem dynamics in an isolated, non-integrable quantum many-body system can be strongly non-Markovian, and that the effect is controlled by the direction of the quench in parameter space. For the mixed-field Ising chain with $N=200$ sites quenched from $(J,h_x,h_z)=(0.2,1,0)$ to $(1,0.1,0.5)$, the trace distance $T_d(\\rho^\\ell_{t+\\delta}, \\rho^\\ell_t)$ between temporally separated reduced states of contiguous blocks of $\\ell=1,\\dots,4$ spins is highly non-monotonic and revives repeatedly, which the author interprets as significant information backflow into the subsystem. The cumulative degree of non-Markovianity $\\mathcal{N}(\\delta)$ is about 50 times smaller for the reverse quench $(1,0.1,0.5)\\to(0.2,1,0)$, making that direction effectively Markovian; for the non-Markovian direction, smaller subsystems are considerably more non-Markovian than larger ones. The paper also reports that the total-variation distance between the ordered eigenvalue spectra of the time-separated reduced states is strongly oscillatory, with timescales of about 0.78 in evolution time and about 1.55--1.6 in lag $\\delta$, essentially independent of subsystem size; the author states that no interpretation of this classical-distance pattern is offered. A heuristic explanation is given for the asymmetry: the paramagnetic ground state is a product state while the ordered target state is highly entangled and GHZ-like, so the forward quench demands a difficult global restructuring, whereas the reverse quench lets quasiparticles spread almost ballistically through weak couplings.","pith_inferences":["Extension: the paper's heuristic implies a testable general principle, that quenches requiring assembly of long-range entangled ordered target states should generically produce subsystem non-Markovianity, while quenches into weakly coupled paramagnetic-like regimes should look Markovian; this could be checked in Heisenberg or long-range Ising chains, where the author says similar signatures alread","Extension: the unexplained TVD periodicities look like natural fingerprints of confined domain-wall, or meson, oscillations in the ferromagnetic regime; computing the revival spectrum as a function of the longitudinal field $h_z$ would test whether the period tracks the confinement energy scale.","Extension: because the paper's degree of non-Markovianity $\\mathcal{N}(\\delta)$ is computed for one fixed initial state rather than maximized over pairs of initial states, the reported asymmetry may be state-dependent; a maximization over initial pairs could confirm the factor-of-50 contrast or reveal hidden non-Markovianity in the reverse quench.","Extension: the protocol requires only reduced density matrices of a few sites, so the directional asymmetry is in principle measurable with quantum-gas microscopes on cold-atom or Rydberg arrays, where local site-resolved tomography is becoming routine."],"forward_implications":["For quenches from the paramagnetic side into the non-integrable ferromagnetic regime, even single-site reduced dynamics in an isolated chain is strongly non-Markovian, meaning information about a subsystem's past flows back from the rest of the system.","Smaller subsystems (one and two spins) show markedly larger degrees of non-Markovianity than three- and four-spin blocks, so the memory effect is strongest for the smallest experimentally accessible probes.","In the opposite quench direction, the subsystem dynamics is effectively Markovian, with a degree of non-Markovianity about 50 times smaller, so the direction of the quench in parameter space acts as a switch for memory effects.","Persistent revivals of the trace distance accompany slow relaxation and anomalous thermalization in the confined ferromagnetic regime; non-Markovianity is an obstruction to fast equilibration.","The total-variation distance between ordered eigenvalue spectra of the subsystem states shows systematic oscillations with timescales of about 0.78 in evolution time and about 1.55--1.6 in lag, largely independent of subsystem size, a pattern the paper reports but does not explain."],"supporting_citations":[{"why":"Defines the trace-distance-based information backflow measure that the paper adapts to fixed initial subsystem states.","marker":"[30]"},{"why":"Supplies the review-level definitions and characterizations of Markovianity and CP-divisibility used to frame the diagnostics.","marker":"[15]"},{"why":"Provides the connection between information backflow and non-Markovianity that justifies reading trace-distance revivals as memory effects.","marker":"[16]"},{"why":"Establishes the equivalence between information-backflow and indivisibility notions for invertible dynamical maps, which the paper relies on.","marker":"[18]"},{"why":"Argues that almost-Markovian behavior is generic in closed quantum dynamics, the baseline against which the paper's strong directional exception stands out.","marker":"[20]"},{"why":"Gives the criterion for dividing quantum channels into CPTP steps, the formal backbone of the divisibility notion used in the paper.","marker":"[25]"},{"why":"Locates the Ising critical point separating the paramagnetic and ferromagnetic regimes, fixing where the quenches start and end.","marker":"[36]"},{"why":"Supplies the TEBD algorithm used for the real-time matrix-product-state simulations of the quench dynamics.","marker":"[52]"},{"why":"Supplies the DMRG and matrix-product-state framework used to prepare ground states and construct reduced density matrices.","marker":"[53]"}],"fun_headline_variants":["Quench direction flips whether spin subsystems forget","Memory backflow in spin subsystems flips with quench direction","Reverse quench wipes non-Markovianity from spin subsystems","One quench direction makes spin subsystems non-Markovian","Quench asymmetry controls subsystem memory in spin chains"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The non-Markovianity reading rests on the assumption that the reduced dynamics of a subsystem is a legitimate completely positive and trace-preserving (CPTP) process with invertible intermediate maps, so that information backflow can be equated with CP-indivisibility, and the paper explicitly states that it does not verify complete positivity of the intermediate maps $\\Lambda_{t,s}$.","fun_headline_variants_meta":{"raw":{"variants":["Quench direction flips whether spin subsystems forget","Memory backflow in spin subsystems flips with quench direction","Reverse quench wipes non-Markovianity from spin subsystems","One quench direction makes spin subsystems non-Markovian","Quench asymmetry controls subsystem memory in spin chains"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00077,"raw_usage":{"total_tokens":3513,"prompt_tokens":1147,"completion_tokens":2366,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":763,"completion_tokens_details":{"reasoning_tokens":2284}},"tokens_in":763,"tokens_out":2366,"duration_ms":18159,"temperature":1.0,"reasoning_tokens":2284,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T23:19:47.801500+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct numerical calculation would settle the claim: extract the intermediate maps $\\Lambda_{t,s}$ for the one- and two-spin subsystems in the paramagnetic-to-ferromagnetic quench and compute their Choi representations (whose positivity is equivalent to complete positivity) at the times when the trace distance revives; if any Choi matrix has negative eigenvalues, the revivals cannot be attributed to a violation of divisibility, so the claimed non-Markovianity would need re-qualification.","supporting_citations":[{"cited_title":"Breuer, E.-M","cited_arxiv_id":null,"evidence_quote":"Defines the trace-distance-based information backflow measure that the paper adapts to fixed initial subsystem states."},{"cited_title":"Rivas, S","cited_arxiv_id":null,"evidence_quote":"Supplies the review-level definitions and characterizations of Markovianity and CP-divisibility used to frame the diagnostics."},{"cited_title":"Chruściński, Dynamical maps beyond markovian regime, Physics Reports 992, 1 (2022)","cited_arxiv_id":null,"evidence_quote":"Establishes the equivalence between information-backflow and indivisibility notions for invertible dynamical maps, which the paper relies on."},{"cited_title":"Figueroa-Romero, K","cited_arxiv_id":null,"evidence_quote":"Argues that almost-Markovian behavior is generic in closed quantum dynamics, the baseline against which the paper's strong directional exception stands out."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the criterion for dividing quantum channels into CPTP steps, the formal backbone of the divisibility notion used in the paper."},{"cited_title":"Pfeuty, The one-dimensional ising model with a transverse field, Annals of Physics 57, 79 (1970)","cited_arxiv_id":null,"evidence_quote":"Locates the Ising critical point separating the paramagnetic and ferromagnetic regimes, fixing where the quenches start and end."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the TEBD algorithm used for the real-time matrix-product-state simulations of the quench dynamics."}],"review_version":1}