{"id":"74c194e9-d678-4980-be1a-0b7232001578","arxiv_id":"2512.16026","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For a generalized Langevin particle in crossed E and B fields, the total entropy production is Gaussian with variance equal to twice its mean, giving P(Δs)/P(−Δs)=e^{Δs}.","lead":"The paper derives a detailed fluctuation theorem for the entropy production of a charged Brownian particle whose dynamics is non-Markovian (memory friction) and takes place in crossed electric and magnetic fields. It proves the theorem for two exactly solvable driving protocols by showing the entropy production is Gaussian with variance exactly twice its mean.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The variance–mean proof assumes two-time stationarity (Eq. 49) of non-Markovian fluctuations; this is not implied by the canonical one-time PSPD and likely fails for general memory kernels.","rationale":"The reader's conditional verdict correctly identifies the appendix-B canonical PSPD as a fragile step, but I believe the more load-bearing premise is the stationarity of two-time correlations assumed in Eq. (49) and Appendix C. Even if B14 is true, the variance–mean relation (58) requires the cross-correlation ⟨W R(τ)⟩ computed in Eq. (57), which is derived from the assumption that ⟨R(τ)R(t)⟩ depends only on τ−t. For a GLE with a memory integral truncated at t=0 and initial conditions independent of the future noise, the reduced process is generally not strictly stationary: this is the standard 'initial slip' problem of non-Markovian baths. The paper does not address this; it simply asserts time-translation invariance after Eq. (48). The entropy-definition issue raised by the reader remains a valid secondary concern about whether the physical quantity is the standard one, but it does not threaten the internal derivation as directly as the two-time stationarity assumption. A concrete exact or numerical check with a simple exponential memory kernel would settle whether Eq. (49) holds; if it fails, the central claim is not established for arbitrary memory kernels, and the conditional verdict should remain in place pending that test.","tokens_in":19642,"tokens_out":32877,"duration_ms":364072,"concrete_test":"Take a concrete non-Markovian kernel, e.g. γ(t)=γ0 e^{-t/τ}, and compute the exact two-time covariance ⟨X(t)X(t′)⟩ from the solution (26) with X0,V0 drawn from the canonical equilibrium and independent noise satisfying Eq. (2). Compare it with ⟨X0X(t−t′)⟩ from Eq. (49) for several t>t′. If they differ, Appendix C's identity (57) fails. Then simulate the GLE for protocol (i) with this kernel and check whether σ²_s equals 2⟨Δs_tot⟩ (Eq. 58) to numerical precision. If either check fails, the DFT of Eq. (59) is not established for general memory; if the equality holds, the stationarity assumption is at least consistent with this test.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The DFT follows from Gaussianity plus Eq. (58). Eq. (58) is obtained by inserting Eq. (57), whose derivation in Appendix C begins with the identity ⟨R(τ)R(t)⟩=⟨R0R(τ−t)⟩ (Eq. C1), i.e., strict stationarity of the two-time correlation of the autonomous fluctuation dynamics. The text asserts this after Eq. (48) \"for equilibrium initial conditions.\" But in the finite-horizon GLE (1), the memory integral starts at 0, and the noise in Eqs. (26)–(31) is treated as independent of the initial fluctuations (⟨X0f⟩=0). For a non-Markovian kernel γ(t), the reduced (r,v) process is not time-homogeneous; there is an initial slip in the bath degrees of freedom. The one-time canonical property B14 constrains only Ξ(τ)=⟨R(τ)R(τ)⟩ and does not determine ⟨R(τ)R(t)⟩. Therefore Eq. (49) and Eq. (C1) are an additional unproved assumption, not a consequence of Appendix B. If it fails, Eq. (57) is wrong, so Eq. (58) and the DFT do not follow. This is more load-bearing than the entropy-definition issue: even accepting the authors' definition of Δs_tot, the proof of the variance–mean relation is incomplete for arbitrary γ(t).","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a charged Brownian particle in a harmonic trap, driven by a time-dependent electric field in a constant magnetic field, with non-Markovian dynamics given by a generalized Langevin equation with memory and Gaussian noise. Using the exact linear solution, the authors obtain a Gaussian phase-space probability density and, from it, compute the trajectory-dependent total entropy production. For two driving protocols—direct forcing and dragged trap center—they claim to prove a detailed fluctuation theorem P(Δs_tot)/P(−Δs_tot)=e^{Δs_tot}, relying on Gaussianity of Δs_tot and a variance–mean identity σ²_s=2⟨Δs_tot⟩. The proof is built on two technical pillars: a canonical (equilibrium) phase-space density for the fluctuations at every time, and a two-time stationarity property of the reduced fluctuation dynamics.","tokens_in":19956,"tokens_out":15046,"duration_ms":164324,"significance":"If the central result were correct, it would be a useful exact extension of detailed fluctuation theorems to non-Markovian charged-particle dynamics in crossed fields, going beyond the Markovian analyses of Refs. [31] and [34]. The paper is self-contained in its setup, provides explicit Laplace-domain response functions, and derives work-variance relations in closed form, with no free parameters. However, the two load-bearing technical steps—the stationary canonical PSPD of Appendix B and the two-time stationarity assumed in Eq. (49)/Appendix C—are not established for the stated model and are generally false under the manuscript's own treatment of initial conditions and noise. The result may be salvageable under a different, properly defined equilibrium-bath initialization, but as written the proof does not support the claim.","major_comments":[{"comment":"The variance–mean identity (58) depends on Eq. (57), which is derived from the two-time correlation identity ⟨R(τ)R(t)⟩=⟨R₀R(τ−t)⟩ asserted in Eq. (49) and used again in Eq. (C1). This is a strict time-translation-invariance property of the reduced fluctuation process. The text states it 'for equilibrium initial conditions' after Eq. (48), but Appendix B only establishes the one-time canonical marginal (B14). For the finite-horizon GLE with the memory integral starting at 0 and with noise treated as independent of the initial fluctuations (e.g., ⟨X₀f(t)⟩=0), the reduced (X,Y,Vx,Vy) process is not time-homogeneous: there is an initial-slip transient. The one-time canonical property constrains only equal-time covariances and does not determine two-time correlations. Without a proof—or an explicit change of initialization to a stationary process—Eq. (C1) is an additional unproved assumption","section":"§III.A.1, Eq. (49); Appendix C, Eq. (C1)"},{"comment":"The derivation of the canonical PSPD P(R,S) drops all initial-condition–noise cross-correlations (⟨X₀f(t)⟩=⟨Y₀f(t)⟩=⟨Vx₀f(t)⟩=⟨Vy₀f(t)⟩=0) and then asserts 'after a long algebra' that F=k_BT/ω², G=k_BT, H=I=0. This is not a consequence of the stated assumptions for a general memory kernel. For a colored-noise GLE with independent initial conditions, the equal-time variance has an initial-slip correction. For example, in the one-dimensional version with a smooth memory kernel, a small-t expansion of ⟨X²(t)⟩ gives an O(t⁴) contribution proportional to γ(0), while the deterministic initial-condition terms cancel at that order, so F(t)≠k_BT/ω² unless the initial conditions are correlated with the future noise. Thus the central premise (B14) is not merely undisplayed algebra; it is generally false under the paper's treatment of the noise. The 'equilibrium initial conditions' need to specify t","section":"Appendix B, Eq. (B14)"},{"comment":"The paper defines the system entropy as −ln P(r,t), using the marginal density of position only, and the first-law balance Q=W−ΔU with U taken as the potential energy only. In the underdamped stochastic thermodynamics of Seifert (as used in Refs. [27,30,31]), the system entropy is −ln P(x,v,t) and the internal energy includes kinetic energy. The manuscript does not justify why the position-marginal object is 'the' total entropy production. Even if the Gaussian variance–mean calculation were correct, the theorem would apply to an entropy-like functional constructed from the position marginal, not to the standard underdamped total entropy production. The terminology and scope should be clarified, or the definition should be matched to the standard one.","section":"§III, Eqs. (36)–(40)"}],"minor_comments":[{"comment":"The detailed fluctuation theorem is written as P(Δs_tot)/P(−Δs_tot)=e^{Δs_tot} with the same symbol on both sides; using, e.g., Σ on the left would avoid notational confusion.","section":"§III.A.2, Eq. (59)"},{"comment":"The appendix headings use Spanish ('Apéndice A', 'Apéndice B', 'Apéndice C'); these should be 'Appendix' in an English-language journal.","section":"Appendix headings"},{"comment":"Several references contain incomplete or malformed bibliographic data, e.g., Ref. [27] is cited as 'Phy. Rev. Lett.5, 040602' (should be PRL 95, 040602), and Refs. [7] and [9] have broken DOI strings. Please check all references.","section":"References [7], [9], [27]"},{"comment":"The notation ⟨W r̃(τ)⟩ appears before the covariance is defined; introduce the cross-covariance explicitly to avoid ambiguity.","section":"§III.A.2, Eq. (56)"},{"comment":"The factorization P(R,S)=P_z(Z,V_z)P(R̃,S̃) is stated but the z-component details are deferred; a sentence noting that the z-axis decouples because the magnetic force is perpendicular would help readability.","section":"§II, after Eq. (31)"}],"recommendation":"reject","confidential_remarks":"The core difficulty is not a stylistic or presentation issue: the proof assumes a stationary colored-noise GLE while the manuscript's own solution and independence assumptions describe a non-stationary process. The one-time canonical PSPD and the two-time stationarity of Eq. (49) are both load-bearing and, for the model as written, the latter is generally false. A proper repair would require reformulating the initial-condition/bath preparation and redoing Appendices B and C, which is beyond a routine revision. If the authors can supply such a corrected derivation under an explicitly stationary bath initialization, the result may still be of interest; as submitted, the central claim is not established."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First: this is a careful, exact calculation that does what it advertises. For a non-Markovian GLE with memory, harmonic trap, constant magnetic field, and time-dependent electric driving, the authors prove a detailed fluctuation theorem for the total entropy production under two protocols. The mechanism is the usual one for linear systems: show Δs is Gaussian, compute the variance, and verify σ² = 2⟨Δs⟩. The genuinely new piece is the explicit phase-space density calculation in Appendix B, which shows the one-time distribution remains exactly the canonical Gibbs form despite the memory kernel and magnetic field. That is not trivial, and the paper gets credit for carrying it through. No free parameters, no fitted data; the work-variance relations are consistent with Jarzynski.\n\nWhere it gets softer. The biggest issue is the definition of total entropy production. The paper uses the position marginal P(r,t) in the system entropy and a first law with potential energy only. That is not the standard underdamped Seifert entropy, which uses the full phase-space distribution and includes kinetic energy. The authors have the phase-space distribution available—they just computed it—so this is a modeling choice, not an accident. But it is a substantial one: the object they prove the DFT for may not be the physically observed entropy production of the charged particle. This should be flagged and defended.\n\nSecond, the key algebraic cancellation in Appendix B (H=0, I=0, F=kBT/k, G=kBT) is reported as 'after a long algebra,' with intermediate steps suppressed. A referee cannot easily verify it. That is a presentation gap, and since the entire result rests on it, the authors should be asked to show more.\n\nThird, the two-time stationarity assertion leading to Eq. (49) and (C1)—that equilibrium initial conditions make correlation functions depend only on time differences—is stated without proof. I believe it is true for a GLE when the initial state is the full equilibrium, but the paper does not establish it, and it is load-bearing for Eq. (57). The text's casual wording invites exactly the objection the stress-test raises. It should be either proven or referenced.\n\nThe self-citations to [31] and [34] are appropriate; the method is a direct extension, and the paper says so. No fitting, no circularity. The novelty is real but incremental: this is a useful benchmark for non-Markovian charged-particle setups, not a conceptual shift. It belongs in a specialized journal; the referees should ask for the extra derivations and a clearer discussion of the entropy definition.","headline":"Careful exact DFT for the non-Markovian GLE; the result is right under the authors' position-marginal entropy definition, which itself needs a defense.","tokens_in":20468,"tokens_out":9684,"would_cite":true,"duration_ms":92572,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["05.10.Gg","05.40.Jc"],"model":"deepseek-v4-flash","headline":"For a charged Brownian particle in a harmonic trap driven by time-dependent electric fields and a constant magnetic field, the paper proves that the total entropy production obeys the detailed fluctuation theorem even when the surrounding b","keywords":["entropy production","detailed fluctuation theorem","generalized Langevin equation","non-Markovian dynamics","magnetic field","harmonic trap","Gaussian process","stochastic thermodynamics"],"falsifier":"Choose a specific memory kernel, such as an exponential γ(t) = (γ0/τ)e^{−t/τ}, compute the exact covariance matrix F, G, H, I from Eqs. (B2)–(B5) at a finite time, and check whether H = I = 0 and F = k_BT/ω², G = k_BT hold. Any deviation would break σ² = 2⟨Δs⟩ and, with it, the detailed fluctuation theorem. Numerically simulating the GLE for a non-harmonic trap and testing the ratio P(Δs)/P(−Δs) would also settle it.","tokens_in":19480,"feed_emoji":"🧲","tokens_out":4451,"duration_ms":42721,"temperature":0.7,"pith_summary":"This paper establishes that the detailed fluctuation theorem — P(Δs)/P(−Δs) = e^{Δs} — holds exactly for the total entropy production of a charged Brownian particle in a harmonic trap, driven by a time-dependent electric field, with a constant magnetic field and a non-Markovian (memory) bath. The memory kernel and the magnetic field complicate the dynamics, but the linearity of the generalized Langevin equation plus Gaussian noise keeps the phase-space distribution Gaussian. The authors prove the key relation σ² = 2⟨Δs⟩ by showing the fluctuation distribution stays canonical at all times, and this immediately gives the detailed fluctuation theorem. This matters because it extends exact entropy-production statistics to structured, viscoelastic-like baths, where standard Markovian proofs do not apply. The result holds for two protocols: a prescribed time-dependent force and an equivalently dragged trap center.","feed_headline":"Memory and magnets don't break the entropy fluctuation theorem","feed_subtitle":"Exact proof shows P(Δs)/P(−Δs) = e^{Δs} survives non-Markovian friction and crossed fields","key_machinery":"The proof hinges on the exact Laplace-transform solution of the linear generalized Langevin equation in crossed fields, expressed through response functions H0, H1, and H2. The covariance matrix of the phase-space fluctuations is computed and, after lengthy algebra, shown to reduce to the canonical form in Eq. (B14). This canonical-preservation identity is what forces the variance–mean relation σ²_s = 2⟨Δs_tot⟩, which, together with Gaussianity, yields the detailed fluctuation theorem. The same machinery yields a work-variance identity that provides the matching mean.","core_discovery":"The central discovery is that for equilibrium initial conditions, the phase-space probability density of the fluctuations (X, Y, Vx, Vy) is exactly the canonical Gibbs distribution at every time t, with position covariance k_BT/k and velocity covariance k_BT, independent of the memory kernel and the magnetic field. Consequently, the trajectory-dependent total entropy production is a Gaussian random variable whose variance equals twice its mean, σ²_s = 2⟨Δs_tot⟩. For a Gaussian variable this relation is equivalent to the detailed fluctuation theorem P(Δs)/P(−Δs) = e^{Δs}. The paper shows this explicitly for two protocols: direct time-dependent forcing and dragging the trap center.","pith_inferences":["If the canonical-preservation identity extends to any linear system whose response functions satisfy the same algebraic relations, the detailed fluctuation theorem would hold for a wider class of non-Markovian Gaussian processes; this is worth testing.","The definition of system entropy here uses only the position marginal; a phase-space definition that includes the velocity marginal would likely give the same variance–mean relation because the velocity marginal is also canonical, but this is not shown in the paper.","Anharmonic confining potentials would break the exact Gaussian-canonical structure; the detailed fluctuation theorem may then fail or require corrections, suggesting a testable boundary for the theorem's validity.","The magnetic field enters the proof only through the response-function algebra; the apparent irrelevance of B to the final distribution hints that the result may extend to non-uniform or time-dependent fields if time-reversal symmetry is respected."],"forward_implications":["The detailed fluctuation theorem and the integral fluctuation theorem ⟨e^{−Δs}⟩ = 1 hold for a non-Markovian harmonic charged particle in crossed electric and magnetic fields.","The result holds for arbitrary memory kernels consistent with the fluctuation–dissipation theorem, so it covers viscoelastic and other structured baths.","The two protocols—force driving and trap dragging—are proven thermodynamically equivalent at the level of entropy-production statistics.","The Gaussian character of entropy production is exact for this linear system, enabling testable predictions for the full distribution, not just the mean."],"fun_headline_variants":["Memory and magnets keep entropy FT exact","Entropy FT holds under memory and crossed fields","Non-Markovian, crossed fields: entropy FT proven","Exact FT for Langevin with memory and B-field"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that the fluctuation phase-space distribution remains exactly canonical at all times — that the cross-correlations H and I vanish and the variances stay fixed at their equilibrium values, which is asserted after lengthy algebra rather than displayed in full.","fun_headline_variants_meta":{"raw":{"variants":["Memory and magnets keep entropy FT exact","Entropy FT holds under memory and crossed fields","Non-Markovian, crossed fields: entropy FT proven","Exact FT for Langevin with memory and B-field"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000481,"raw_usage":{"total_tokens":2186,"prompt_tokens":685,"completion_tokens":1501,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":429,"completion_tokens_details":{"reasoning_tokens":1448}},"tokens_in":429,"tokens_out":1501,"duration_ms":12913,"temperature":1.0,"reasoning_tokens":1448,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T15:42:36.635071+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Choose a specific memory kernel, such as an exponential γ(t) = (γ0/τ)e^{−t/τ}, compute the exact covariance matrix F, G, H, I from Eqs. (B2)–(B5) at a finite time, and check whether H = I = 0 and F = k_BT/ω², G = k_BT hold. Any deviation would break σ² = 2⟨Δs⟩ and, with it, the detailed fluctuation theorem. Numerically simulating the GLE for a non-harmonic trap and testing the ratio P(Δs)/P(−Δs) would also settle it.","supporting_citations":[],"review_version":1}