{"id":"4ba9f728-0712-4049-b76c-93c47ff4b34e","arxiv_id":"2505.23028","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A two-site-correlation projection-operator method gives a beyond-mean-field treatment of the Dicke model with non-Markovian dephasing, curing the spurious initial-state dependence and predicting relaxation times proportional to N/g^2.","lead":"This paper builds a way to compute how quantum emitters inside an optical cavity behave when each emitter also feels its own noisy, memory-carrying environment, going beyond the usual mean-field approximation. The method removes a known artifact (results that wrongly keep depending on the emitters' starting orientation) and predicts finite-size relaxation times, which matters for molecular polaritons and cavity-modified chemistry.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The acknowledged production of Z2-broken steady states for large N contradicts the unqualified claim that the formalism resolves the initial-state dependence of mean-field theory, since non-unique long-time states can persist in the regime where the N/g^2 cure is asserted.","rationale":"The reader identified the two-site truncation and second-order memory-kernel expansion as the weakest assumption. This stress-test agrees but sharpens the concern: the paper's own Discussion contains a concrete manifestation of that truncation failing, namely Z2-broken steady states for large N. That failure directly impinges on the central claim that initial-state dependence is resolved, because it means the long-time state can still depend on initial conditions in the large-N superradiant regime. The reader's conditional verdict already reflects uncertainty about this truncation, so this concern does not require a change of verdict; it does require that the authors qualify the scope of the initial-state-independence claim and provide evidence, numerical or analytical, that the purported N/g^2 thermalization mechanism operates for the system sizes and parameter regimes where the pathology is claimed to be cured. Credit is due for the standard derivation of the projection-operator equations, the recovery of the mean-field limit in SM-IV at fixed time, and the honest statement of the large-N symmetry-breaking limitation. The concern is about the range of validity of the central claim, not about the derivation itself.","tokens_in":21558,"tokens_out":7644,"duration_ms":87441,"concrete_test":"Run the projection-operator equations for N=40, 160, and 640 at g=0.28 and g=0.16, starting from the four initial states of Fig. 2, and evolve to a time several times N/g^2. If the non-symmetric initial state converges to a symmetry-broken steady state different from the one reached from the symmetric initial condition, the claimed resolution of initial-state dependence fails in this regime. As a complementary check, compute the connected two-site correlators at Ωt = N/g^2 and verify that they remain O(1/N); if they grow, the two-site truncation is uncontrolled at the timescale of the claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that the projection operator in Eq. (3), with the memory kernel expanded to second order in the fluctuation Liouvillians (Eqs. (4)), erases the initial-state memory and restores thermalization in finite-N Dicke dynamics. The Discussion explicitly concedes, however, that for sufficiently large N the method produces symmetry-broken steady states when the initial state does not respect the Z2 symmetry, even though the exact finite-N steady state should be symmetric. In the mean-field pathology, different initial spin orientations freeze into different steady-state values of ⟨σz⟩; replacing that manifold by one of two symmetry-broken long-time states selected by the initial condition does not restore uniqueness in that regime. The convergence shown in Fig. 3(a,b) is only for N=10, and the N/g^2 relaxation in Fig. 3(d,e) is demonstrated only up to N=50 and for g<g_c. The large-N data in Fig. 3(c,f) concern the log-N rise time and the order parameter, not convergence to a common steady state. The self-consistency check in SM-IV is performed at Ωt=20 for g>g_c, whereas the N/g^2 timescale and steady-state crossover are long-time properties. The paper attributes the large-N failure to missing instanton processes of order e^{-N}, which are exactly the processes that would restore Z2 symmetry; therefore the central initial-state-independence claim is not supported in the large-N, long-time regime without an additional argument.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a time-dependent projection operator formalism for all-to-all spin-boson systems with non-Markovian local baths, retaining two-site spin correlations in the relevant subspace and expanding the memory kernel to second order in the light-matter coupling. The method is applied to the zero-temperature Dicke model with local Ohmic dephasing. The authors report that beyond-mean-field terms remove the conserved-σz obstruction of mean-field theory, producing a slow N/g^2 relaxation in the normal phase, a log-N photon-number rise time above threshold, and a sharpening of the steady-state crossover with N. The central claim is that this formalism resolves the spurious initial-state dependence found in mean-field dynamics.","tokens_in":21742,"tokens_out":3953,"duration_ms":39618,"significance":"If the truncation is controlled, the formalism is a potentially valuable new tool for molecular polaritonics and other one-to-all coupled open systems with structured environments. The explicit master equations for the second photon moments, the scaling predictions (τ ∝ N/g^2, t_r ∝ log N, and 1/N decay of connected two-site correlators), and the demonstration in SM-V that a naive projector fails are concrete and useful contributions. The central mechanism—two-site correlations converting a conserved mean-field quantity into a slow relaxation channel—is physically compelling. However, the significance is tempered by the acknowledged failure to capture Z2-symmetry-restoring instantons for large N, which directly limits the scope of the initial-state-independence claim.","major_comments":[{"comment":"The unqualified claim that the formalism 'resolves this pathology of initial state dependence' (Abstract and Introduction) is contradicted by the final paragraph of the 'Beyond mean-field' section, which concedes that for sufficiently large N the equations produce Z2-broken steady states when the initial state does not respect the Z2 symmetry, whereas the exact finite-N steady state is symmetric. In the normal phase, the mean-field pathology is a manifold of initial-condition-dependent ⟨σz⟩ values; replacing that manifold by two symmetry-broken states selected by the initial condition does not restore uniqueness in that regime. The claim of a cured initial-state dependence is therefore demonstrated only for the finite-N range shown in Fig. 3(a,b) (N=10), and the large-N, long-time regime requires an additional argument that is not provided.","section":"Beyond mean-field / Abstract"},{"comment":"The self-consistency check of the cluster expansion is performed at Ωt=20 and g=0.26>g_c, whereas the central long-time claims (N/g^2 relaxation and steady-state crossover) concern g<g_c and times t≫1/κ. The 1/N decay of connected two-site correlators at a single fixed time in the superradiant phase does not establish that three-site or higher irreducible correlations, or higher-order memory-kernel terms, remain negligible on the O(N/g^2) timescale in the normal phase. A self-consistency check in the actual relaxation regime, or an independent finite-N benchmark, is needed to support the central timescale claim.","section":"SM-IV"},{"comment":"The analytic argument that the memory terms are O(g^2/N) assumes that the irreducible part of ρ_ij is O(1/N) and that partial traces such as Tr_j Δσ^x_j ρ_ij inherit this smallness. This is the very property the derivation is meant to establish, so the estimate is partly self-referential. The numerical self-consistency check in SM-IV only partially mitigates this concern, because it is not performed in the regime where the N/g^2 relaxation is claimed to occur.","section":"SM-II, 'N-dependent relaxation time'"}],"minor_comments":[{"comment":"The sentence 'This can be can be understood as a pathology of mean-field theory' contains a duplicated word and should be corrected.","section":"Mean-field limit"},{"comment":"The word 'non-Marovian' should be 'non-Markovian'.","section":"Discussion"},{"comment":"The trace notation Tr0 and Tri,j is terse in the main text; a brief definition in the main text, rather than only in the supplement, would improve readability.","section":"Formalism, Eq. (4)"},{"comment":"The four initial spin states are described only in the Fig. 2 caption; restating them in the Fig. 3 caption would make the convergence claim easier to evaluate.","section":"Fig. 3(a,b)"},{"comment":"The notation Δ≡2g Re a_ss is used before it is introduced in Eq. (S3); consider defining it at first use.","section":"SM-I"}],"recommendation":"major_revision","confidential_remarks":"The paper is within scope for the journal and the central formalism is worth publishing if the claims are qualified to the regime where the method is actually controlled. The main revision needed is to align the abstract and introduction with the acknowledged large-N limitation, and to either provide a validity analysis for the long-time normal-phase regime or explicitly restrict the initial-state-independence claim to finite N up to the scale where the approximation holds. I do not see any concern about novelty or attribution; the authors clearly cite prior projection operator work and the related cumulant approaches."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this paper is worth reading: it extends beyond-mean-field methods for the Dicke model to non-Markovian local baths, which prior 1/N methods could not handle. The core move—a self-consistent two-site projection operator plus a memory kernel expanded to O(g^2/N), solved with TTI-TEMPO—is new and, as far as I can tell, sound. The authors also deserve credit for disclosing the method's limitations, including the symmetry-broken steady states for large N.\n\nWhat's actually new: the projection goes beyond the naive Born approximation, the equations only need first and second photon moments, and they identify two finite-N timescales—a log-N rise time above threshold and an N/g^2 relaxation in the normal phase. The scaling arguments are plausible and the numerics support them, at least for the N ranges shown. The self-consistency check (connected two-site correlators decaying as 1/N) is the right check to do.\n\nThe soft spots are real but mostly in proportion. The claim that the method 'resolves' the initial-state dependence of mean-field theory is too strong as stated. For large N the method itself produces symmetry-broken steady states selected by the initial condition; the paper's own Discussion concedes this, but the abstract still says 'resolves' without qualification. The resolution is demonstrated only for N=10 in the normal phase. The N/g^2 derivation in SM-II is a verbal scaling argument, not a closed derivation. The convergence check in SM-IV is at one time (Ωt=20) and one coupling (g>g_c), which is not the regime where the N/g^2 relaxation claim lives. Numerical fits lack uncertainties. No code or data is posted.\n\nNone of that kills the paper. The method is a genuine advance for a problem people care about, and the limitations are openly acknowledged. It deserves a serious referee, but the referee should ask the authors to soften the abstract claim and either provide more extensive validation or at least error bars and a clearer statement of where the two-site truncation is trusted.","headline":"A useful and honest beyond-mean-field method for non-Markovian Dicke dynamics, but the headline claim overstates what is actually shown.","tokens_in":22505,"tokens_out":2834,"would_cite":true,"duration_ms":28626,"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 argues that a time-dependent projection operator retaining two-spin correlations, with the memory kernel expanded to second order in the light-matter coupling, gives a controlled beyond-mean-field description of the open Dicke…","keywords":["Dicke model","non-Markovian dephasing","beyond mean-field","projection operator formalism","finite-size dynamics","superradiance","tensor network influence functionals","open quantum systems"],"falsifier":"Run the same two-site projection-operator scheme but retain three-site irreducible correlations (or push the memory kernel to fourth order in $g$) for $N=10$ and $g=0.2\\,g_c$; if the convergence time of $\\langle\\hat{\\sigma}_z\\rangle$ no longer scales as $N/g^2$, or if initial-state dependence reappears, the truncation is the origin of the cure. Alternatively, measure $\\langle\\hat{\\sigma}_z\\rangle$ relaxation in a finite ensemble with tunable $N$ and coupling $g<g_c$ and test the predicted linear-in-$N$, inverse-quadratic-in-$g$ time.","tokens_in":21164,"feed_emoji":"⚛️","tokens_out":6633,"duration_ms":64319,"temperature":0.7,"pith_summary":"The paper aims to move beyond the mean-field approximation for a cavity coupled to many two-level emitters (the Dicke model) when each emitter also feels its own non-Markovian dephasing bath, a setup relevant to molecules in optical cavities. The authors argue that a time-dependent projection operator that keeps one- and two-site correlations, together with a memory kernel expanded to second order in the light-matter coupling, captures the leading $1/N$ corrections in a controlled way while still allowing structured baths to be treated by tensor-network influence functionals. Within this scheme, the spurious dependence of the mean-field steady state on the initial spin orientation disappears, and previously invisible finite-size time scales appear: spin relaxation in the normal phase takes a time of order $N/g^2$, while the photon-number rise above threshold grows as $\\log N$. The paper thus claims a broadly applicable approach to finite-size quantum optical dynamics with non-Markovian local environments.","feed_headline":"Pair correlations cure the Dicke model's spurious memory","feed_subtitle":"The new scheme shows spins relaxing on an N/g^2 timescale that mean-field theory misses.","key_machinery":"The central object is a self-consistent, time-dependent projection operator $P(t)$ (built following the Willis-Picard construction) whose relevant state is the photon factor times a site cluster expansion cut off at two-site irreducible correlations, Eq. (3). Its companion is the exact Nakajima-Zwanzig equation for the relevant part, with the memory kernel expanded to second order in the fluctuation Liouvillians $\\Delta\\mathcal{L}'_i$, which yields corrections of order $g^2/N$. Because the photon sector enters only through first and second moments, the large cavity Hilbert space never needs to be represented, and the non-Markovian baths enter through the two-site equations via efficient bath representations (pseudo-mode or influence-functional/TEMPO methods). This machinery is what converts the otherwise-conserved $\\hat{\\sigma}^z$ of the mean-field normal phase into a decaying observable, producing the $N/g^2$ and $\\log N$ time scales.","core_discovery":"On its own terms, the paper establishes that the projected density matrix $\\hat{\\rho}_{\\mathrm{rel}}=\\hat{\\rho}_0\\otimes[\\bigotimes_i \\hat{\\rho}_i + \\sum_{j<i}(\\hat{\\rho}_{ij}-\\hat{\\rho}_i\\otimes\\hat{\\rho}_j)\\otimes\\bigotimes_{k\\neq i,j}\\hat{\\rho}_k]$ plus a Nakajima-Zwanzig memory kernel at second order in the fluctuation Liouvillians $\\Delta \\mathcal{L}'_i$ yields a closed, numerically treatable set of equations for the photon first and second moments and the two-site spin density matrices. Solving these equations for the Dicke model with zero-temperature Ohmic dephasing baths shows that different initial spin orientations no longer freeze to different steady states: the beyond-mean-field dynamics restores a unique steady state, with a relaxation time $\\tau\\propto N/g^2$ in the normal phase and a photon rise time above threshold that grows as $\\log N$. The same equations capture the sharpening of the steady-state photon crossover with increasing $N$ towards the mean-field superradiant transition. The paper also reports the method's known failure mode: for sufficiently large $N$ it produces $Z_2$-broken steady states from symmetry-breaking initial states, which a finite system should not do.","pith_inferences":["If the truncation is controlled, the $N/g^2$ relaxation time should be observable in finite-size cavity-QED or ensemble experiments: a direct measurement of $\\langle\\hat{\\sigma}_z\\rangle$ relaxation in the normal phase as a function of $N$ and $g$ would test the predicted scaling.","The mechanism suggests that for observables sensitive to three-site connected correlators (e.g., higher-order spin cumulants), the two-site truncation will eventually break down; extending the cluster expansion to three-site irreducible correlations would provide a sharper validity test.","The paper's admission of $Z_2$-broken steady states for large $N$ implies the method is blind to exponentially small ($e^{-N}$) instanton-like processes that restore symmetry; a future method capturing those would likely modify the late-time asymptotics, though probably not the early-time $N/g^2$ window.","The $\\log N$ photon rise time above threshold implies that finite-size corrections set a practical time beyond which mean-field predictions are unreliable, even when the initial state is a mean-field fixed point."],"forward_implications":["In the normal phase, the relaxation time of the spin imbalance grows linearly with $N$ and as $1/g^2$, a time scale absent from mean-field dynamics.","Above threshold, the scaled photon number first seeds at $O(1/N)$ then grows exponentially with an $N$-independent Lyapunov rate, so the rise time increases as $\\log N$.","Different initial spin orientations converge to a common steady state under the beyond-mean-field equations, curing the mean-field pathology of initial-state dependence.","The steady-state photon crossover sharpens with increasing $N$, consistent with a mean-field superradiant transition in the thermodynamic limit.","The same projection-operator formalism applies to other one-to-all coupled quantum optical systems with non-Markovian local baths, including finite temperature, and requires only photon second moments."],"supporting_citations":[{"why":"Supplies the time-evolving matrix product operator (TEMPO) mean-field treatment of non-Markovian baths that the paper compares against and extends.","marker":"[25]"},{"why":"Baseline result that Markovian local dephasing fully suppresses the superradiant transition, setting up the contrast with non-Markovian baths.","marker":"[29]"},{"why":"Provides the path-integral cumulant treatment and the mean-field transition analysis for Markovian baths that the present approach generalizes.","marker":"[33]"},{"why":"Introduces the self-consistent field expansion and time-dependent projection operator used to derive the beyond-mean-field equations.","marker":"[35]"},{"why":"Gives the projector construction whose properties (idempotence, commutation with the time derivative) the derivation relies on.","marker":"[36]"},{"why":"Documents the failure of a naive projector under the Born approximation to give the correct N-scaled steady state, motivating the two-site projector.","marker":"[38]"},{"why":"Supplies the time-translationally invariant tensor network influence functional method that enables long-time non-Markovian propagation.","marker":"[39]"},{"why":"States the Dicke model's critical coupling in the thermodynamic limit, which the paper's linear-response estimate generalizes to non-Markovian dephasing.","marker":"[44]"},{"why":"Shows that non-Markovian dephasing restores the superradiant transition in mean field, the starting point for the beyond-mean-field corrections.","marker":"[58]"}],"fun_headline_variants":["Non-Markovian dephasing: beyond mean-field fixes Dicke memory","Dicke model's spurious memory cured by pair correlations","New timescale for Dicke relaxation beyond mean-field","Unique steady state restored in Dicke model with baths","Memory kernel unlocks Dicke model's true dynamics"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole correction rests on the assumption that irreducible correlations involving three or more spins, and memory terms beyond second order in the light-matter coupling, are negligible at the couplings and timescales studied.","fun_headline_variants_meta":{"raw":{"variants":["Non-Markovian dephasing: beyond mean-field fixes Dicke memory","Dicke model's spurious memory cured by pair correlations","New timescale for Dicke relaxation beyond mean-field","Unique steady state restored in Dicke model with baths","Memory kernel unlocks Dicke model's true dynamics"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000724,"raw_usage":{"total_tokens":3249,"prompt_tokens":950,"completion_tokens":2299,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":566,"completion_tokens_details":{"reasoning_tokens":2217}},"tokens_in":566,"tokens_out":2299,"duration_ms":18464,"temperature":1.0,"reasoning_tokens":2217,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T12:57:07.927434+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same two-site projection-operator scheme but retain three-site irreducible correlations (or push the memory kernel to fourth order in $g$) for $N=10$ and $g=0.2\\,g_c$; if the convergence time of $\\langle\\hat{\\sigma}_z\\rangle$ no longer scales as $N/g^2$, or if initial-state dependence reappears, the truncation is the origin of the cure. Alternatively, measure $\\langle\\hat{\\sigma}_z\\rangle$ relaxation in a finite ensemble with tunable $N$ and coupling $g<g_c$ and test the predicted linear-in-$N$, inverse-quadratic-in-$g$ time.","supporting_citations":[{"cited_title":"Fowler-Wright, B","cited_arxiv_id":null,"evidence_quote":"Supplies the time-evolving matrix product operator (TEMPO) mean-field treatment of non-Markovian baths that the paper compares against and extends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the path-integral cumulant treatment and the mean-field transition analysis for Markovian baths that the present approach generalizes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the self-consistent field expansion and time-dependent projection operator used to derive the beyond-mean-field equations."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the projector construction whose properties (idempotence, commutation with the time derivative) the derivation relies on."},{"cited_title":"Degenfeld-Schonburg,Self-consistent projection opera- tor theory, Ph.D","cited_arxiv_id":null,"evidence_quote":"Documents the failure of a naive projector under the Born approximation to give the correct N-scaled steady state, motivating the two-site projector."},{"cited_title":"Simulating many-body non- markovian open quantum systems,","cited_arxiv_id":null,"evidence_quote":"Shows that non-Markovian dephasing restores the superradiant transition in mean field, the starting point for the beyond-mean-field corrections."}],"review_version":1}