{"id":"3c948c7b-f23e-4736-9017-bc9207251895","arxiv_id":"2412.14727","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A new hierarchical equations of motion (LDUO-HEOM) combines an overdamped Lorentz-Drude bath with an undamped oscillator bath and shows system-mediated correlations between the two baths.","lead":"This paper derives new equations of motion for a quantum system coupled to two baths at once: an overdamped bath plus an undamped vibrational mode, and uses them to simulate 2D electronic spectra. It reports that the two baths exchange information indirectly through the system, and that the new model is about 180 times faster than the authors' previous two-bath model.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Non-additivity of bath-coordinate expectation values (Figs. 5f, 6f) is necessary but not sufficient evidence for inter-bath information transfer; an interventional or mutual-information test is required.","rationale":"The paper has two main claims: the LDUO-HEOM derivation removes superfluous damping at large computational savings, and the two baths communicate indirectly via the system. The first claim is supported by the derivation and the qualitative 2DES agreement in Fig. 2, and the reported 0.56% cost reduction is striking, though no code or data are shipped. The second claim, which is the advertised novelty, rests on the residual plots in Section V. The reader flagged the same weakest assumption: nonzero X_LDUO − X_UO − X_LD is interpreted as information transfer. I agree and sharpen it: the residual is computed across three different models, not as an interventional or correlational measure within one model. Any two baths coupled to the same system will alter each other's bath-coordinate expectation values through the system, even classically, so non-additivity alone cannot distinguish 'communication' from 'common cause'. This is load-bearing because if the concern lands, the paper's central conceptual claim is not established even though the HEOM construction may be correct. The concrete interventional test (vary only the UO initial state) would settle whether UO affects LD. If the test passes, the claim becomes much stronger. The reader's CONDITIONAL verdict is appropriate; my analysis does not change it.","tokens_in":13684,"tokens_out":6641,"duration_ms":62810,"concrete_test":"Keep the LDUO-HEOM Hamiltonian and all parameters fixed, but run two simulations that differ only in the initial state of the UO bath: thermal equilibrium at one temperature versus a displaced coherent state (or a second temperature), with the LD bath and system initial states identical. Compute the projected LD bath-coordinate expectation value X_LD^(1)(t) in both simulations. If X_LD^(1) is unchanged when only the UO initial state is varied, the residuals in Fig. 5f are not evidence of UO-to-LD information flow and the central claim is unsupported. If it does change, that would demonstrate a causal influence from the UO bath to the LD bath through the system; the residual diagnostic should then be supplemented by a direct information-theoretic witness such as I(UO:LD) to quantify the transfer.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim of Section V is that nonzero residuals in Figures 5f and 6f (X_LDUO − X_UO − X_LD for X^(1) and X^(2)) demonstrate information transfer between the undamped and overdamped baths. This inference is the weakest load-bearing step. X^(1) and X^(2), computed from Eqs. (51)–(56), are collective bath-coordinate expectation values; they are not measures of bipartite correlation such as mutual information, and they are evaluated on three different trajectories: the full two-bath model, the isolated UO model, and the isolated LD model. Because the system evolves under the combined influence of both baths in the full model but under only one bath in each isolated model, the sum of the isolated expectation values will generically differ from the full-model value whenever the reduced system dynamics are non-additive in the two couplings. This non-additivity is a common-cause effect of both baths acting on a shared system; it does not, by itself, show that information is transferred from one bath to the other. The plotted residuals could plausibly be reproduced in a purely classical model with two independent stochastic forces driving the same oscillator. Thus the presented evidence is necessary but not sufficient for the paper's headline claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper develops a new hierarchical equations of motion (HEOM) variant, LDUO-HEOM, for a two-level system coupled to a bath whose spectral density is a sum of a Lorentz-Drude overdamped component and an undamped oscillator delta-function component. The derivation follows standard Matsubara/path-integral methods and is presented in detail. The model is used to compute 2D electronic spectra, which are compared qualitatively to the authors' earlier bath vibration model (BVM), and a large computational speedup is reported. The paper then uses first- and second-order bath-coordinate expectation values, following Zhu et al., to argue that the two baths exchange information indirectly through the system, a claim framed as bath-system-bath communication.","tokens_in":13994,"tokens_out":9409,"duration_ms":84589,"significance":"If the derivation and implementation are correct, LDUO-HEOM is a useful and computationally efficient alternative to underdamped-bath HEOM for systems with a well-defined vibrational mode in a dissipative environment; the reported speedup of roughly two orders of magnitude is striking, and the derivation is self-contained and reproducible in structure. The absence of parameter fitting is a strength. However, the central conceptual claim of information transfer between baths is not established by the presented evidence: the difference plots in Section V demonstrate non-additivity of bath-coordinate moments, not communication in an information-theoretic sense. The manuscript is publishable after the central claim is either supported by a direct information-flow analysis or substantially reframed.","major_comments":[{"comment":"The residual plots in the bottom-right panels of Figs. 5 and 6 are presented as evidence that strong bath-system-bath coupling occurs and that information is transferred between the baths; this inference is not warranted. The quantities X^(1) and X^(2) are first and second moments of collective bath coordinates evaluated along three different trajectories: the full two-bath model and the two isolated single-bath models. Whenever the reduced system dynamics is not additive in the two couplings, the full-model expectation value will generically differ from the sum of the isolated values, because the system is a common cause that is itself affected by both baths. A nonzero residual is therefore necessary but not sufficient for inter-bath information transfer; a classical system driven by two independent stochastic forces would produce the same qualitative pattern. To support the headline claim, the authors should either perform an interventional test (e.g., decouple one bath at some time and show that the other bath's subsequent state changes) or compute a genuine measure of correlation and directionality between the two bath degrees of freedom, such as the mutual information of their reduced states or a bath-bath covariance. As written, the evidence does not distinguish 'communication through the system' from the trivial statement that two baths coupled to the same system both influence, and are influenced by, the system's state.","section":"Section V, Figures 5 and 6"},{"comment":"No convergence analysis is reported for the hierarchy truncation criterion Gamma_max = 10 max(I(gamma_k)). For the undamped component the Matsubara frequencies are purely imaginary, so the usual argument that high-tier ADOs decay and decouple is not available; the secular growth seen in the LDUO panels of Figs. 5 and 6 could in principle be a truncation artifact. Because these long-time bath-coordinate behaviours are the central evidence for the paper's main claim, the authors should show that X^(1), X^(2), and the residuals converge with respect to Gamma_max and hierarchy depth, and ideally compare with an independent numerically exact method or a substantially larger hierarchy for at least one test case.","section":"Section III, Eq. (42)"},{"comment":"The term 'information transfer' is never operationally defined. The abstract and Section V state that information exchange occurs between the baths, but no information-theoretic quantity is introduced, and the bath-coordinate moments used are not information measures. Consequently, the central claim is not falsifiable as stated. The authors should either define the measure of information transfer and compute it, or rewrite the abstract and conclusions to claim non-additive bath dynamics or system-mediated bath-coordinate correlations, which is what the present data actually support.","section":"Abstract and Section V"}],"minor_comments":[{"comment":"The claim of qualitative agreement between BVM and LDUO spectra is based on visual inspection; a quantitative measure, such as a normalized root-mean-square difference or spectral overlap, would make the comparison more convincing and reproducible.","section":"Section IV, Figure 2"},{"comment":"The reported speedup of 0.56% would be more convincing if the authors specified the number of cores, software/library versions, and how the equilibration and evolution times were measured, since these details affect reproducibility.","section":"Section IV, computational cost paragraph"},{"comment":"The captions of Figures 5 and 6 should state explicitly which difference is plotted in the bottom-right panels: for example, X_LDUO - X_UO - X_LD for both X^(1) and X^(2); the current wording is ambiguous.","section":"Section V, Figure captions"},{"comment":"The statement that the baths are combined via a 'direct sum' is confusing because the total Hamiltonian is simply a sum of independent bath oscillators with additive system-bath coupling; clarifying what 'direct sum' adds beyond this standard construction would help readers.","section":"Section II"},{"comment":"Please correct typographical errors, including 'incredibility powerful' (Introduction), 'superfluious' (Conclusion), 'bof' (Section I), and 'sate' (Eq. (28)).","section":"Throughout"},{"comment":"The recursion relations for L^(n+1)_i in Eq. (56) are stated without derivation; a brief explanation or a more explicit reference to the combinatorial factors would improve reproducibility.","section":"Section V, Eqs. (51)-(56)"}],"recommendation":"major_revision","confidential_remarks":"The LDUO-HEOM derivation and the computational speedup are valuable contributions, and I would support publication after revision. However, the abstract and Section V currently overstate the evidence: the difference plots show non-additivity of bath-coordinate expectation values, not information transfer in a well-defined sense. The authors should either add a direct information-theoretic or interventional analysis or soften the claim. The comparison to their own BVM is not circular, but the parameter matching between the BVM underdamped mode and the LDUO undamped mode should be clarified to make the 'qualitative agreement' meaningful."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Hi [Colleague],\n\nThe useful thing here is the LDUO-HEOM derivation. Combining an exactly undamped oscillator bath with an overdamped Lorentz-Drude bath in one hierarchy is new, and the derivation follows the standard path-integral/Matsubara route cleanly. The projected bath-coordinate expectation values (Eqs. 51-56) are a nice addition. On the practical side, the 2DES spectra are qualitatively compatible with the authors' earlier BVM and the reported 0.56% compute time is a real selling point for people who want an undamped vibration plus an overdamped environment without the cost of ASD-HEOM. The citation pattern is honest: they cite the existing undamped HEOM variants and position the novelty as the specific combination.\n\nThe soft spot is Section V. The paper claims 'information transfer between the baths' based on the residual X_LDUO − X_UO − X_LD being nonzero. That residual is non-additivity of expectation values, but the three trajectories are not interchangeable: in the full model the system is driven by both baths, while the isolated runs have only one bath each. So the residual can be entirely a common-cause effect of two independent baths acting on the same system; it need not mean one bath sends information to the other. This is the classic difference between correlation and causation. I don't think the paper has to be wrong in its intuition, but the evidence doesn't support the strong statement 'it is therefore clear that strong bath-system-bath coupling occurs.' A projection within the same two-bath trajectory, or a mutual-information-like measure, would do the job.\n\nAlso, validation is qualitative against their own BVM; there is no independent baseline, no code or data shipped. That is a real limitation for reproducing the speedup and spectra. The derivation itself is self-contained and the hierarchy truncation condition (Eq. 42) is stated, so a competent referee can check it.\n\nOverall: worth sending to peer review, because the model is new, the derivation appears sound, and the computational advantage is potentially useful. But the communication claim needs to be either reworded or supported with a proper interventional/correlation measure. I'd suggest the referee focus on Section V and on requesting code/data availability.","headline":"A useful HEOM extension for undamped mode plus overdamped bath, but the information-transfer claim rests on non-additivity that doesn't establish it.","tokens_in":14482,"tokens_out":3123,"would_cite":true,"duration_ms":25376,"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":"Two baths exchange information without touching","keywords":["non-Markovian","HEOM","undamped oscillator bath","Lorentz-Drude bath","bath-system-bath coupling","2D electronic spectroscopy","information transfer","open quantum systems"],"falsifier":"Run the same LDUO-HEOM simulation but compute the mutual information between the collective coordinates of the LD and UO baths directly from the full density matrix; if that mutual information stays at zero while the difference-plots in Figures 5 and 6 remain nonzero, the claim of bath-system-bath information transfer would be falsified. Alternatively, replace the quantum system with a classically driven linear oscillator coupled to the two baths; if the same residual pattern appears, the 'communication' is a classical driving artefact, not quantum information transfer.","tokens_in":13461,"feed_emoji":"🔄","tokens_out":6276,"duration_ms":48766,"temperature":0.7,"pith_summary":"This paper develops a new hierarchical equations of motion model, LDUO-HEOM, in which a quantum system couples to two baths: an overdamped Lorentz-Drude bath and an undamped oscillator bath, combined by a direct sum rather than a tensor product. The authors aim to establish that information is transferred between these two baths even though they do not interact directly, the quantum system serving as the mediator through a bath-system-bath coupling. They also argue that this model removes the spurious damping of their earlier two-bath bath vibration model while producing qualitatively similar 2D electronic spectra at about 0.56% of the computational cost. If correct, the result implies that multi-bath open quantum systems can exhibit hidden channels of information exchange that are invisible in single-bath descriptions.","feed_headline":"Two baths exchange information without touching","feed_subtitle":"New model also reproduces 2D spectra at 0.56% of previous compute.","key_machinery":"The central object is the LDUO-HEOM hierarchy, a set of equations of motion for auxiliary density operators obtained from a path-integral influence functional whose spectral density is split into an overdamped Lorentz-Drude component and an undamped oscillator component $J(\\omega)=J_{\\mathrm{LD}}(\\omega)+J_{\\mathrm{UO}}(\\omega)$, with $J_{\\mathrm{UO}}$ proportional to $\\omega(\\delta(\\omega-\\omega_{\\mathrm{UO}})+\\delta(\\omega+\\omega_{\\mathrm{UO}}))$. Each component is Matsubara-decomposed separately, giving one hierarchy axis for the undamped pair of imaginary frequencies $\\pm i\\omega_{\\mathrm{UO}}$ and additional axes for the overdamped and Matsubara terms; the undamped axis is never Markovianized, which is what removes the spurious damping. Information flow between the baths is then analysed by projecting the collective bath coordinate expectation values $X^{(1)}$ and $X^{(2)}$ onto the separate hierarchy faces (UO and LD planes) and comparing the full model with isolated single-bath runs.","core_discovery":"The paper claims that in the LDUO-HEOM model, the undamped oscillator bath and the overdamped Lorentz-Drude bath exchange information indirectly, despite having no direct coupling term, because each is coupled to the same quantum system. This is detected through first- and second-order bath coordinate expectation values: subtracting the isolated-bath responses from the full two-bath response leaves substantial nonzero residuals, showing that the full bath coordinate is not a sum of independent parts. The paper further claims that the LDUO-HEOM reproduces qualitatively the 2D electronic spectra of the earlier BVM while eliminating the extra damping that the finite linewidth of the underdamped mode had introduced, at a computational saving of at least 99.4%.","pith_inferences":["This result suggests that any open system coupled to two independent reservoirs may generically mediate information exchange between them, so the usual assumption of statistically independent baths should be checked whenever the system is non-Markovian.","A natural test would be to compute a direct information-theoretic measure, such as the mutual information between the two bath coordinates or a quantum channel capacity, and see whether it matches the residual signal; the paper uses expectation-value residuals as a proxy rather than a direct measure.","The difference plots in Figures 5 and 6 might be reconciled with a classical analogue: if a single driven oscillator coupled to two damped degrees of freedom reproduces the same residuals, the 'communication' could be a semiclassical interference effect rather than quantum information exchange."],"forward_implications":["If the claim holds, two-bath open-system simulations must treat bath-system-bath coupling as a genuine physical channel, not as an artifact of the system-bath split.","The model gives a computationally cheap route (about 0.56% of BVM cost) for 2D electronic spectra of systems with an undamped intramolecular vibration in an overdamped environment, making realistic spectra calculations far more accessible.","The existence of indirect bath-bath information transfer means that non-Markovianity in multi-bath systems can be amplified or redirected through the system, with implications for quantum heat engines and refrigerators where multiple reservoirs are present.","The removal of superfluous damping sharpens the earlier conclusion that the Hamiltonian vibration model and the bath vibration model are not equivalent in practice, clarifying when a vibration can be safely moved from the Hamiltonian into the bath."],"supporting_citations":[{"why":"Supplies the earlier two-bath BVM model that LDUO-HEOM is benchmarked against for spectra and computational cost.","marker":"22"},{"why":"Provide the bath-coordinate expectation-value formulas used to detect information transfer between the baths.","marker":"43,44"},{"why":"Gives the undamped-oscillator spectral density form that the LDUO decomposition is built from.","marker":"17"},{"why":"Establishes the HEOM construction that the combined influence-functional and auxiliary-density-operator derivation follows.","marker":"40"},{"why":"Provides the two-level system model and bath parameters used in the simulations.","marker":"31"},{"why":"Supply the Tanimura hierarchy derivation methods that the new LDUO-HEOM equations extend.","marker":"23-25"}],"fun_headline_variants":["Baths swap info with no direct contact","Indirect bath-bath coupling alters dynamics","Two baths share secrets via shared system","Mediated bath transfer without any touching"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The analysis assumes that a nonzero difference between the full two-bath response and the sum of the isolated-bath responses demonstrates information transfer between the baths; if the residuals only reflect that the system evolves differently when a second bath is present, the communication claim is not established.","fun_headline_variants_meta":{"raw":{"variants":["Baths swap info with no direct contact","Indirect bath-bath coupling alters dynamics","Two baths share secrets via shared system","Mediated bath transfer without any touching"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000215,"raw_usage":{"total_tokens":1385,"prompt_tokens":855,"completion_tokens":530,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":471,"completion_tokens_details":{"reasoning_tokens":477}},"tokens_in":471,"tokens_out":530,"duration_ms":4471,"temperature":1.0,"reasoning_tokens":477,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T11:57:58.260688+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same LDUO-HEOM simulation but compute the mutual information between the collective coordinates of the LD and UO baths directly from the full density matrix; if that mutual information stays at zero while the difference-plots in Figures 5 and 6 remain nonzero, the claim of bath-system-bath information transfer would be falsified. Alternatively, replace the quantum system with a classically driven linear oscillator coupled to the two baths; if the same residual pattern appears, the 'communication' is a classical driving artefact, not quantum information transfer.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the earlier two-bath BVM model that LDUO-HEOM is benchmarked against for spectra and computational cost."},{"cited_title":"Seibt \\ and\\ author O","cited_arxiv_id":null,"evidence_quote":"Gives the undamped-oscillator spectral density form that the LDUO decomposition is built from."},{"cited_title":"Green , author B","cited_arxiv_id":null,"evidence_quote":"Provides the two-level system model and bath parameters used in the simulations."}],"review_version":1}