{"id":"42bd0f89-5bad-4397-9bf2-433890065fed","arxiv_id":"2412.13820","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A dephasing spin chain hosts a pure scar-like stationary state alongside the infinite-temperature state, and the interface between them melts diffusively despite a finite Lindbladian gap.","lead":"This paper studies a line of quantum particles that lose information through dephasing, and shows it has two different long-time states: a featureless hot state and a special pure state that remembers the initial setup. It explains why the boundary between these two states melts slowly, even though the model's formal relaxation rate is fast, using a picture of a wandering membrane.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The membrane prediction is derived for a toy generator W-tilde with coefficients that are not close to W and at parameters violating the perturbative regime, so the analytic v=1/3, D=1 are never tested against the full Lindbladian.","rationale":"The central claim has two pillars: the exceptional stationary state and the slow membrane interface despite a finite Lindbladian gap. The first pillar is solid: the commutant-algebra proof in the SM rigorously gives C=Span{1,|⇑⟩⟨⇑|}, and the exponential bound η(t)≤e^{-4γt}η(0) is a clean algebraic inequality. The finite-gap claim is supported by exact diagonalization up to L=12 and by the Knabe bound for the effective model, so the gap's finiteness is credible. The second pillar is the soft spot. The analytic solution with v=1/3,D=1 is derived for W-tilde, which drops the three-site term of W and fixes parameters that violate the perturbative regime in which W is obtained from the Lindbladian. The full-model numerics at γ=1, g=J=0.5 are far from both the large-γ limit and the W-tilde solvable point, and the scale collapse is achieved by fitting v and D. This means the quantitative membrane prediction is not actually tested in the original model; the evidence is a flexible error-function fit, not a parameter-free check. The proposed simulation at the W-tilde point directly tests whether the toy generator is representative of the full Lindbladian at the parameters where the analytic result is claimed. This concern matches the reader's weakest assumption, so the reader's conditional verdict should stand: the qualitative phenomenology is plausible and worth reporting, but the quantitative membrane statement should be either validated or softened until the test is performed.","tokens_in":20222,"tokens_out":21112,"duration_ms":192217,"concrete_test":"Simulate the full Lindbladian interface protocol of Fig. 3 at the W-tilde solvable point, e.g., g=0.5, J=√2, γ=3 (so J^2/(4γ)=2g^2/γ=1/6), for L=40 up to t=30, and superimpose the unscaled numerical magnetization profiles on the analytic prediction Eq. (7) with v=1/3 and D=1, with no fitting. If the data collapse onto the analytic curve within deviations comparable to those in Fig. 4, then W-tilde is representative and the analytic membrane prediction is validated for the actual model. If the profiles instead require substantially different v and D (say, deviations beyond ±30%), the analytic derivation is disconnected from the full model, and the slow-interface claim in the original model rests only on the fitted error-function collapse.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's analytic membrane result (v=1/3, D=1) is obtained from W-tilde, which drops the three-site term in W and fixes J^2/(4γ)=2g^2/γ=1/6. Two uncontrolled steps separate this from the actual model. First, the dropped three-site term has the same coefficient (1/6) as the retained two-site terms, so W-tilde is not a controlled approximation to W; its single-domain-wall closure is the only justification. Second, the solvable point satisfies γ=1.5J^2, which violates the large-γ condition (J/γ≪1, g/γ≪1) under which the effective generator W was derived from the original Lindbladian. The full-model comparison in Figs. 3-4 instead uses J=g=0.5, γ=1, where both conditions fail and the three-site term is order 0.5; the observed collapse is obtained by fitting v and D rather than testing the predicted values. Thus the quantitative membrane claim and the skin-effect reconciliation (also derived for W-tilde) are not directly supported in the original model; the central assertion of slow interface melting rests on a two-parameter fit to an error function. If W-tilde is not representative, the paper's quantitative predictions lack controlled backing, leaving only the qualitative scaling observation.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the Lindblad master equation (1) for a spin-1/2 chain with dephasing jump operators and a Hamiltonian consisting of hopping plus an East-West term. The main claims are: (i) for generic g,J≠0 the only stationary states are the infinite-temperature state and the pure fully-polarized state |⇑⟩⟨⇑|, with the latter a dark state protected by a non-extensive strong symmetry; (ii) coherences between the sector of |⇑⟩ and its orthogonal complement decay exponentially at a rate bounded by 4γ; (iii) the Lindbladian gap is finite in the thermodynamic limit but, despite this, an interface prepared between the two stationary states melts diffusively with drift on timescales growing with L; (iv) this apparent contradiction is reconciled through eigenvector localization, i.e., a non-Hermitian skin effect, of the effective Markov generator; and (v) |⇑⟩ should be regarded as an open-system quantum many-body scar. The analytic membrane solution with v=1/3 and D=1 is obtained for a simplified Markov generator W-tilde, while the full-model comparison in Figs. 3 and 4 uses fitted values of v and D.","tokens_in":20507,"tokens_out":16080,"duration_ms":157502,"significance":"If the claims hold, the paper provides a concrete dephasing model in which a finite Lindbladian gap coexists with system-size-dependent relaxation of a local observable, and it formulates a criterion for open-system many-body scars based on the absence of extensive strong symmetries. The strengths include the algebraic commutant calculation, the rigorous exponential bound on inter-sector coherence, the exact solvability of the auxiliary membrane generator, and the transparent tensor-network and exact-diagonalization numerics. These elements make the central existence claim credible and the proposed scar terminology worth discussing. However, the quantitative membrane prediction is not yet connected to the full model in a controlled way, so the broad significance of the paper depends on a reframing or on additional controlled numerics.","major_comments":[{"comment":"The analytic values v=1/3 and D=1 in Eq. (6) are derived for W-tilde, not for the model in Eq. (1). W-tilde is obtained from the effective generator W of Eq. (5) by dropping the three-site term 2(g^2/γ)∑_j π^z_{j-1}π^x_jπ^z_{j+1} and by fixing J^2/(4γ)=2g^2/γ=1/6. At that point the dropped term has the same coefficient as the retained two-site terms, so W-tilde is not a controlled approximation to W; its single-domain-wall closure is a property of W-tilde alone. Since the full-model comparison in Fig. 4 fits both v and D, the predicted constants v=1/3 and D=1 are never tested against the original Lindbladian. The paper should either present W-tilde as an auxiliary model that motivates the scaling form, or provide a controllability argument that connects it to Eq. (1).","section":"Membrane diffusion of the interface, Eqs. (5)-(7)"},{"comment":"The full-model simulations in Figs. 3 and 4 use J=g=0.5 and γ=1, i.e., g/γ=J/γ=0.5, outside the large-γ regime (g/γ, J/γ≪1) in which the effective generator W in Eq. (5) was derived; the authors acknowledge this in the text, but the consequence is that the error-function collapse with fitted v≃0.453 and D≃1.33 does not quantitatively test the membrane mechanism. A test in the controlled regime, or an analysis of the dependence of v and D on γ, would be needed to support the claim that the interface of the original model obeys the membrane dynamics rather than merely the diffusive scaling form.","section":"Membrane diffusion of the interface, Figs. 3 and 4"},{"comment":"The skin-effect reconciliation in the section 'Late-time asymptotics and spectral gap' and in the End Matter is carried out for the projected single-domain-wall matrix of W-tilde. The estimate in Eq. (14) shows that the matrix elements of W-tilde grow exponentially in L; no analogous statement is demonstrated for the low-lying eigenvectors of the actual Lindbladian L. Since the finite gap and the slow interface relaxation are both properties of Eq. (1), the reconciliation currently applies to the auxiliary model. The authors should state this limitation explicitly or provide numerical evidence for the corresponding eigenvector localization in the full Lindbladian.","section":"Late-time asymptotics and spectral gap, End Matter"},{"comment":"The claim that there are exactly two stationary states for all L is stronger than what is proven in the main text. The End Matter proves C=Span{1, |⇑⟩⟨⇑|} and notes the inclusion C⊆kerL, while the numerical double degeneracy of the zero eigenvalue is shown only for L=6. Unless a theorem ensures ker L=C for this class of dephasing Lindbladians (which the paper does not state), the uniqueness of the stationary states is not established in the thermodynamic limit. Please state the relevant theorem or extend the argument.","section":"Stationary states, End Matter"}],"minor_comments":[{"comment":"The sentence 'since C ⊆ ker L ... one safely concludes that C = Span{1, |⇑⟩⟨⇑|}' is logically circular and should be rewritten; the conclusion follows from the induction in the Supplemental Material, not from the observed degeneracy of the kernel.","section":"End Matter, Commutant algebra"},{"comment":"There are several typos and minor notational slips: 'reconcyling' in the heading of the last section, 'isporphic' and 'ahereas' in the Supplemental Material, and '(σ^z_j(t))' in Eq. (11) should be 'σ^z_j'.","section":"Throughout"},{"comment":"The fit Δ(L)=Δ0+Δ1/L^2 is reported without fitted values or residuals; please provide the parameters and error bars, since the finite Δ0 is a central assertion.","section":"Fig. 1(b)"},{"comment":"The fitted values v≃0.453 and D≃1.33 differ from the analytic v=1/3 and D=1 by about 36% and 33%, respectively; the paper should comment on this discrepancy even under the interpretation that W-tilde is only an auxiliary model.","section":"Membrane diffusion of the interface, Fig. 4"}],"recommendation":"major_revision","confidential_remarks":"I recommend major revision rather than rejection: the exact algebraic claims and the numerical observation of slow interface relaxation are valuable, but the quantitative membrane claim needs to be either made controlled or explicitly demoted to a heuristic based on an auxiliary model. The paper appears to be within scope and I have no concerns about citation practices."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this one. It does something genuinely useful: exhibits a dephasing spin chain whose Lindbladian has a finite gap, yet local observables relax on timescales growing with L, via membrane-like interface diffusion. The pure dark stationary state |↑...↑> is protected by a non-extensive strong symmetry; the commutant algebra proof showing exactly two stationary states is clean, and the exponential coherence-decay bound is rigorous. The authors also honestly label 'scar' as a viewpoint, not a theorem. That is a solid theoretical-numerical package.\n\nWhere it gets soft: the analytic membrane result (v=1/3, D=1) is derived for W-tilde, a simplified generator with the three-site term dropped and rates hand-tuned. The full-model numerics are at J=g=0.5, gamma=1, outside the large-gamma regime where W was derived, and the dropped term is not small there. Then Figs. 3–4 verify the membrane picture by fitting v and D to the same data; the predicted values are never tested against the actual Lindbladian. So the quantitative part is a two-parameter fit to an error function, not an independent test. The qualitative claim—interface melts with diffusive broadening and drift on L-dependent timescales despite finite gap—is much better supported. The skin-effect reconciliation is also computed for W-tilde; the idea is plausible but not demonstrated for the full model.\n\nA smaller issue: the extrapolated gap ∆0 is fit over L=6..12 with a 1/L^2 ansatz. The finite-gap conclusion is probably right, but the fit alone is not decisive. That does not undermine the paper, just softens the headline.\n\nBottom line: the central physical claims are likely correct and worth saying. The paper is for researchers in many-body open quantum systems and weak ergodicity breaking. It deserves a serious referee; I would send it out. The referee should ask for either a full-model test of v,D at a parameter set where perturbation theory is controlled, or a clear statement that the analytic values apply only to W-tilde and the full model is a qualitative match.","headline":"A solid, honest paper with a genuinely interesting mechanism—finite gap yet size-dependent slow relaxation—but the quantitative membrane prediction is backed by fits rather than by testing the analytic values.","tokens_in":21066,"tokens_out":1755,"would_cite":true,"duration_ms":16671,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper shows that a one-dimensional dephasing Lindblad model has exactly two stationary states — the infinite-temperature state and the pure fully polarized state…","keywords":["open quantum systems","Lindblad master equation","quantum many-body scars","dephasing noise","dark states","strong symmetries","commutant algebras","non-Hermitian skin effect"],"falsifier":"Run the full Lindblad dynamics (or the full Markov generator $W$) from a sharp interface at small but non-special $g/\\gamma$ and $J/\\gamma$, and measure the weight that leaves the single-domain-wall subspace: if that weight grows with $L$ or the rescaled profile fails to collapse to the error function at large $t$, the membrane picture is not the correct mechanism. Alternatively, compute the Lindbladian gap for $L \\to \\infty$; if $\\Delta_0$ extrapolates to zero, the finite-gap premise fails.","tokens_in":19987,"feed_emoji":"🧲","tokens_out":8929,"duration_ms":74646,"temperature":0.7,"pith_summary":"This paper studies a one-dimensional spin chain with dephasing noise and shows that, besides the usual infinite-temperature state, it has a second stationary state: the pure fully polarized state with all spins up. That pure state is exceptional because it keeps memory of the initial condition, while every orthogonal state relaxes to the infinite-temperature state and loses all information. The authors argue that this pure dark state is an open-system analog of a quantum many-body scar, since it is not protected by any extensive conserved operator and no hydrodynamic slow modes exist. They then show that the interface between the two stationary states melts by diffusive broadening with a drift, on timescales that grow with system size even though the Lindbladian spectral gap is finite, calling the standard gap-to-relaxation-timescale identification into question.","feed_headline":"One pure state never forgets in a dephasing spin chain","feed_subtitle":"Its interface with the thermal state drifts and broadens like a membrane, on timescales the spectral gap alone cannot predict.","key_machinery":"The argument runs on the commutant algebra $\\mathcal{C} = \\mathrm{Span}\\{1, |\\!\\uparrow\\uparrow\\cdots\\uparrow\\rangle\\langle\\uparrow\\uparrow\\cdots\\uparrow\\!|\\}$, which fixes the stationary manifold, and on an effective Markov generator obtained by second-order perturbation theory in $g/\\gamma$ and $J/\\gamma$, acting on diagonal density matrices in the $z$ basis. In the Doi-Peliti representation this generator becomes a classical Markov process $W$. The paper introduces a simplified generator $\\widetilde W = W - (2g^2/\\gamma)\\sum_j \\pi_{j-1}^z\\pi_j^x\\pi_{j+1}^z$ with rates chosen so that $J^2/(4\\gamma) = 2g^2/\\gamma = 1/6$; in $\\widetilde W$ the single-domain-wall subspace is closed and the interface performs a biased random walk with drift $v = 1/3$ and diffusion constant $D = 1$, giving the error-function profile. The spectral decomposition of $\\widetilde W$, together with the non-Hermitian skin effect in its one-domain-wall sector, explains why the finite gap does not control the relaxation of local observables.","core_discovery":"The central claim is that the Lindblad dynamics of Eq. (1), with Hamiltonian $H = \\sum_j [J(\\sigma_j^+\\sigma_{j+1}^- + \\mathrm{H.c.}) + g(\\sigma_j^x\\pi_{j+1}^z + \\pi_j^z\\sigma_{j+1}^x)]$ and dephasing jump operators $L_j = \\sqrt{\\gamma}\\,\\sigma_j^z$, has exactly two stationary states: the infinite-temperature state and the pure fully polarized state $|\\!\\uparrow\\uparrow\\cdots\\uparrow\\rangle\\langle\\uparrow\\uparrow\\cdots\\uparrow\\!|$. The commutant algebra of the local terms is generated by these two operators, so no extensive conserved quantity protects the polarized state; the paper therefore calls it an open-system quantum many-body scar. Starting from an interface that joins the two stationary states, the magnetization profile evolves under an error-function scaling in $(x-vt)/\\sqrt{Dt}$ with nonzero drift and diffusion, even though no conserved densities exist and the Lindbladian gap is finite. The paper reconciles this apparent contradiction by showing that the overlaps of the initial state onto the Lindbladian eigenvectors grow exponentially with system size, a phenomenon tied to the non-Hermitian skin effect.","pith_inferences":["A direct test of the membrane mechanism would be to tune $g/\\gamma$ and $J/\\gamma$ to the perturbative regime and check whether fitted $v$ and $D$ approach $1/3$ and $1$; the paper's simulations sit at moderate couplings where the simplified generator is not guaranteed to dominate.","The non-Hermitian skin-effect explanation suggests boundary conditions are a switch: periodic boundary conditions force an even number of domain walls, so the slowest interface relaxation may look qualitatively different from the open-chain case at the same parameters.","If the scar classification by absence of extensive strong symmetries is adopted, other dephasing models with non-extensive symmetries should also display finite-gap, slow-interface relaxation, making the membrane scaling a generic experimental benchmark for weak ergodicity breaking in open systems."],"forward_implications":["Any initial state orthogonal to $|\\!\\uparrow\\uparrow\\cdots\\uparrow\\rangle$ eventually becomes locally indistinguishable from the infinite-temperature state, so all memory of the initial condition is erased except for the component along the polarized state.","The interface magnetization satisfies $\\langle\\sigma^z(x,t)\\rangle \\simeq \\mathrm{erf}((x-vt)/\\sqrt{Dt})$ with finite $v$ and $D$, so local probes see relaxation times that grow with system size rather than saturating at $1/\\Delta_0$.","Coherence between the scar subspace and its complement decays at least as $e^{-4\\gamma t}$, so the two sectors decouple exponentially fast.","A finite Lindbladian gap is not sufficient to conclude fast, size-independent thermalization; the size dependence of initial-state overlaps onto Lindbladian eigenvectors must be checked.","The same membrane-interface collapse appears in two further dephasing models (PXP-like and XPX-like), indicating the mechanism is not an accident of the East-West Hamiltonian."],"supporting_citations":[{"why":"Supplies the second-order perturbative construction of the effective Markov generator $W$ from the dephasing Lindbladian.","marker":"[18]"},{"why":"Provides the commutant-algebra viewpoint used to identify stationary states of open quantum systems.","marker":"[21]"},{"why":"Defines quantum many-body scars via non-extensive symmetries and commutant algebras, the criterion the paper adapts to open systems.","marker":"[26]"},{"why":"Establishes the Liouvillian skin effect that explains how slow relaxation coexists with a finite spectral gap.","marker":"[29]"},{"why":"Introduces the East-West Hamiltonian whose constrained spin-flip structure makes the polarized state a stationary dark state.","marker":"[31]"},{"why":"Supplies the commutant-algebra machinery used in the proof that the algebra is spanned by the identity and the polarized projector.","marker":"[36]"},{"why":"Provides the Knabe-type finite-size bound used to prove that the simplified Markov generator has a finite spectral gap.","marker":"[63]"}],"fun_headline_variants":["Exceptional pure state defies dephasing in open quantum system","Membrane-like interface between two stationary states in dephasing chain","Open-system scar: one pure state remembers while thermal state forgets","Dephasing many-body system hosts a second, pure stationary state"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The quantitative membrane prediction is derived from a simplified generator $\\widetilde W$ that drops the three-site coupling and pins rates to a special value; the paper assumes this simplified generator faithfully describes the interface dynamics of the full model at the parameters actually simulated, where the dropped terms are not small.","fun_headline_variants_meta":{"raw":{"variants":["Exceptional pure state defies dephasing in open quantum system","Membrane-like interface between two stationary states in dephasing chain","Open-system scar: one pure state remembers while thermal state forgets","Dephasing many-body system hosts a second, pure stationary state"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000463,"raw_usage":{"total_tokens":2316,"prompt_tokens":949,"completion_tokens":1367,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":565,"completion_tokens_details":{"reasoning_tokens":1292}},"tokens_in":565,"tokens_out":1367,"duration_ms":9080,"temperature":1.0,"reasoning_tokens":1292,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T12:45:55.752912+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the full Lindblad dynamics (or the full Markov generator $W$) from a sharp interface at small but non-special $g/\\gamma$ and $J/\\gamma$, and measure the weight that leaves the single-domain-wall subspace: if that weight grows with $L$ or the rescaled profile fails to collapse to the error function at large $t$, the membrane picture is not the correct mechanism. Alternatively, compute the Lindbladian gap for $L \\to \\infty$; if $\\Delta_0$ extrapolates to zero, the finite-gap premise fails.","supporting_citations":[{"cited_title":"Algebraic versus exponen- tial decoherence in dissipative many-particle systems,","cited_arxiv_id":null,"evidence_quote":"Supplies the second-order perturbative construction of the effective Markov generator $W$ from the dephasing Lindbladian."},{"cited_title":"Hilbert space fragmentation in open quantum systems,","cited_arxiv_id":null,"evidence_quote":"Provides the commutant-algebra viewpoint used to identify stationary states of open quantum systems."},{"cited_title":"Exhaust- ive characterization of quantum many-body scars using commutant algebras,","cited_arxiv_id":null,"evidence_quote":"Defines quantum many-body scars via non-extensive symmetries and commutant algebras, the criterion the paper adapts to open systems."},{"cited_title":"Liouvillian Skin Effect: Slowing Down of Relaxation Processes without Gap Closing,","cited_arxiv_id":null,"evidence_quote":"Establishes the Liouvillian skin effect that explains how slow relaxation coexists with a finite spectral gap."},{"cited_title":"Quantum east model: Localization, nonthermal eigenstates, and slow dynamics,","cited_arxiv_id":null,"evidence_quote":"Introduces the East-West Hamiltonian whose constrained spin-flip structure makes the polarized state a stationary dark state."},{"cited_title":"Hilbert space fragmentation and commutant algebras,","cited_arxiv_id":null,"evidence_quote":"Supplies the commutant-algebra machinery used in the proof that the algebra is spanned by the identity and the polarized projector."},{"cited_title":"Energy gaps and elementary excitations for certain VBS-quantum antiferromagnets,","cited_arxiv_id":null,"evidence_quote":"Provides the Knabe-type finite-size bound used to prove that the simplified Markov generator has a finite spectral gap."}],"review_version":1}