{"id":"863970f5-650a-45f7-bbbd-a9f31a7891a6","arxiv_id":"2507.16637","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Thermal operations are uniquely the quantum channels that admit a dilation preserving the environment state at equilibrium, establishing an information-theoretic characterization.","lead":"This paper proves that thermal operations, the standard quantum channels used to model heat baths, are exactly those that leave the environment unchanged when acting on the equilibrium state. The result gives a purely information-theoretic characterization and resolves open questions about the hierarchy of doubly-stochastic quantum channels.","discovery_kind":"unification","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Core thermal-operations characterization is sound, but the advertised strict hierarchy (Eq. 12) rests on MIP*=RE, an unpeer-reviewed result; the paper's own footnote [40] confirms this dependency.","rationale":"The reader's verdict conditionally accepts; I agree. The single strongest claim—that thermal operations are precisely equilibrating dilations with full-rank fixed points—is internally consistent. The proof route is short: Lemma 3 (equilibrium implies product via vanishing mutual information) plus Proposition 4 gives [U, ωA⊗ωB] = 0; with ωA, ωB full-rank Gibbs states at a common β this is equivalent to [U, HA+HB] = 0, which is exactly the definition of a thermal operation. I found no hidden assumption in the finite-dimensional setting. The conditional element concerns the second advertised main result, Eq. (12). The strict inclusion EQ∩DS⊊F is not currently an unconditional theorem; it is a corollary of the claim that Connes' embedding problem has a negative answer, resting on [15]. The paper is exemplary in flagging this in footnote [40] and using 'If true' when deriving Eq. (15), but the figure and Eq. (12) present it as established. If [15] were wrong, the chain would still be nontrivial but with equality EQ∩DS=F, so the overclaim is localized yet real. The reader's weakest_assumption points to the same place, and I see no stronger concern. Hence the verdict is unchanged.","tokens_in":14873,"tokens_out":19970,"duration_ms":216130,"concrete_test":"Independently re-derive Eq. (15) (EQ∩DS⊊F) without invoking [15], by exhibiting an explicit finite-dimensional factorizable channel that is not strongly factorizable, using constructions from [13,14] or nonlocal games. If no such unconditional derivation exists, then the strict inclusion and Fig. 1 should be labeled conditional on the unpeer-reviewed MIP*=RE proof, and the abstract's hierarchy claim should be correspondingly qualified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that equilibrating dilations with full-rank fixed points characterize thermal operations (Proposition 4) is internally sound: Eq. (3) plus unitary invariance of entropy forces [U, ωA⊗ωB] = 0, and with ωA = e^{-βHA}/ZA and ωB = e^{-βHB}/ZB at a common β this is equivalent to [U, HA+HB] = 0, so the channel is a thermal operation. I found no gap in that argument. The load-bearing concern is a different advertised result: in the hierarchy Eq. (12), the strict inclusion EQ∩DS⊊F is derived from the MIP*=RE resolution of Connes' embedding problem [15], which footnote [40] itself states has no completely peer-reviewed proof. The main text says 'If true' before concluding Eq. (15), yet Fig. 1 and Eq. (12) print the strict inclusion as a result. If CEP were instead true, strongly factorizable and factorizable maps would coincide in finite dimensions and this inclusion would collapse, leaving MU⊊CAT⊊EQ∩DS=F⊊DS. The paper is transparent, but the advertised strict hierarchy is conditional on an external unverified proof.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a purely information-theoretic notion of equilibrium: two local states are in informational equilibrium relative to a unitary if each is preserved as the appropriate marginal of the joint time-evolved state (Eq. (1)). A channel admitting an equilibrating dilation (Definition 1) is shown, via Lemma 3 and Proposition 4, to commute with the fixed-point state in the dilation and to be covariant under the modular flow of that fixed point. Interpreting full-rank fixed points as Gibbs states, the authors conclude that thermal operations are precisely the channels admitting equilibrating dilations with full-rank fixed points. The paper then defines catalytic dilations (Definition 2), proves several structural results about catalytic and factorizable channels, and advertises a strict hierarchy MU⊊CAT⊊EQ∩DS⊊F⊊DS for doubly-stochastic channels (Eq. (12), Fig. 1). The strict inclusion EQ∩DS⊊F relies on the MIP*=RE resolution of the Connes embedding problem, as the paper itself notes in footnote [40] and in the sentence immediately before Eq. (15). The paper also contains a multipartite equilibrium lemma (Lemma 7), a robust-catalysis result for thermal operations (Lemma 8), and a connection between catalytic unitaries and dual-unitary tensors (Appendix B).","tokens_in":15108,"tokens_out":10085,"duration_ms":118008,"significance":"The central characterization of thermal operations is conceptually attractive and, as far as the proofs in the End Matter show, correct: the argument from unitary invariance of entropy and mutual information to [U, ωA⊗ωB]=0 is elementary and does not presuppose thermodynamics. If the advertised hierarchy is understood with the appropriate caveats, the paper resolves previously open questions by separating mixed-unitary, catalytic, equilibrating, and factorizable channels, and it gives a useful new perspective on the gap between thermal operations and Gibbs-preserving maps. The proofs are largely self-contained, use no fitted parameters, and are honest about the external MIP*=RE dependency. The multipartite lemma and the robust-catalysis result are additional contributions that should be of interest to the quantum-thermodynamics community.","major_comments":[{"comment":"The strict inclusion EQ∩DS⊊F is derived from MIP*=RE, which the paper's footnote [40] explicitly says has no completely peer-reviewed proof. The sentence just before Eq. (15) says 'If true', but Eq. (12) and Fig. 1 state 'all inclusions are strict' as an unconditional result, and the Conclusions repeat that the classes are 'all distinct'. Because [13,14] imply that a positive resolution of the Connes embedding problem would make strongly factorizable and factorizable maps coincide in finite dimensions, this particular strict inclusion would collapse under the opposite external verdict. The paper should mark the inclusion as conditional in Eq. (12), in Fig. 1, and in the Conclusions, or supply a proof independent of [15].","section":"Hierarchy of doubly-stochastic channels, Eq. (12), Fig. 1, Eq. (15), footnote [40]"},{"comment":"In the final branch of the proof, where T_{|ψ⟩}=T for all |ψ⟩, the derivation reaches U^{⊤A}(ρ⊗1)(U^{⊤A})† = T(ρ)⊗1. The text then says this 'is possible only when ρ and T(ρ) have the same spectrum ... i.e. T(ρ)=WρW†.' As written, the unitary W is only shown to exist for each fixed ρ, and the proof does not establish that a unital completely positive channel preserving the spectrum of every density matrix must be a single unitary conjugation. Since Lemma 13 is load-bearing for the strict inclusion CAT⊊EQ∩DS in Eq. (12), this step needs either a complete argument or a precise citation to a theorem on spectrum-preserving completely positive maps.","section":"Appendix C, Lemma 13"}],"minor_comments":[{"comment":"In Eq. (10), the unitary U is defined on AB only, but it is applied to states on RAB; the expression should be written with an explicit extension such as (1_R⊗U)(ρ_RA⊗ω_B)(1_R⊗U†) or a convention should be stated.","section":"Lemma 5, Eq. (10)"},{"comment":"The derivation of Eq. (24) is compressed into 'using the same argument'; since Eq. (24) is essential for the application of Lemma 7, adding one or two intermediate lines would make the proof easier to verify.","section":"Lemma 8, proof around Eqs. (23)-(24)"},{"comment":"The tensor-network proof of Proposition 10 is difficult to parse in printed form; a sentence explaining the diagrammatic conventions, including the role of the suppressed W_j and the half-circles, would improve readability.","section":"Appendix A, Eq. (A10) and surrounding diagrams"},{"comment":"The caption states 'all inclusions are strict' without qualification; it should point to the conditional status of EQ∩DS⊊F that is discussed in the text and footnote [40].","section":"Fig. 1 caption"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is transparent about the MIP*=RE dependency, and the core thermal-operation characterization appears sound. My recommendation is driven by two load-bearing points: the advertised strict hierarchy should carry an explicit conditional marker at every place it is stated, and the proof of Lemma 13 needs a completed argument for its concluding step. Both are fixable within the scope of the manuscript, so I do not see grounds for rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The real result here is Proposition 4: thermal operations are exactly the channels that admit an equilibrating dilation with respect to a full-rank fixed point. That is new, self-contained, and I don't see a gap. The entropy argument for Lemma 3 is clean, and the step from vanishing mutual information to [U, ωA⊗ωB]=0 is correct. The paper also resolves two open strict inclusions from the authors' earlier work, MU⊊CAT⊊EQ, with proofs that are convincing even where terse.\n\nThe soft spot is the one the stress-test flags, and the paper itself confirms it in footnote [40]: the strict inclusion EQ∩DS⊊F in the advertised hierarchy relies on MIP*=RE, which has no completely peer-reviewed proof. Figure 1 and Eq. (12) print the full hierarchy as a result, while the main text only says \"If true\" before that particular step. That is a presentational inconsistency worth fixing: either keep the conditional flag on the figure, or state explicitly in the caption that one arrow depends on an external unverified result. If Connes' embedding problem turns out to be true, that one arrow collapses, leaving MU⊊CAT⊊EQ∩DS=F⊊DS. The unconditional part of the paper is untouched.\n\nTwo minor things. Lemma 13's final step, deducing that T is a mixture of unitaries from the product form, is stated too quickly; the authors should expand it. And the full-rank assumption is stated but worth being explicit that it is generic, not universal. Neither is a problem for the main claim.\n\nI agree with the reader's conditional verdict. This is a strong paper for quantum thermodynamics and quantum information. The central characterization is original and sound, and the conditional hierarchy is clearly labeled as depending on an external result. The authors are not hiding anything. I would send it to a serious referee, with a note to check the MIP*=RE dependency and to ask for the hierarchy to be presented with the same conditional caveat throughout. The paper deserves to be in the literature once that is cleaned up.","headline":"Clean new characterization of thermal operations via equilibrating dilations; the advertised strict hierarchy is partly conditional on MIP*=RE, but the paper is honest about that and the core result is solid.","tokens_in":15631,"tokens_out":1455,"would_cite":true,"duration_ms":19275,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.65.Ta"],"model":"deepseek-v4-flash","headline":"Thermal operations are exactly the quantum channels that admit an equilibrating dilation: a unitary interaction with an environment left invariant whenever the system starts in equilibrium, provided the fixed point has full rank.","keywords":["thermal operations","equilibrating dilation","catalytic channels","Gibbs-preserving maps","doubly-stochastic channels","factorizable channels","Connes embedding problem","dual-unitary circuits"],"falsifier":"Take a known enhanced thermal operation that is not a thermal operation and numerically search over all dilations $(U,\\omega_B)$ of the channel for one where $\\operatorname{Tr}_A(U\\omega_A\\otimes\\omega_B U^\\dagger)=\\omega_B$ with a full-rank Gibbs fixed point; the paper's Eq. (9) claims no such dilation exists, so finding one would refute the central characterization.","tokens_in":14684,"feed_emoji":"⚖️","tokens_out":11101,"duration_ms":100935,"temperature":0.7,"pith_summary":"This paper argues that the standard allowed operations of quantum thermodynamics—thermal operations—are nothing more than the channels that admit an 'equilibrating dilation': a unitary coupling to an environment whose reduced state is left unchanged whenever the system is prepared in its equilibrium state. The authors prove that if a channel has such a dilation with respect to a full-rank fixed point, then the unitary must commute with the joint state and the channel must be time-translation covariant; interpreting full-rank states as Gibbs states, this is exactly the defining property of thermal operations. Consequently, any candidate thermalization channel that is Gibbs-preserving but not a thermal operation inevitably disturbs the environment in every dilation, which means the environment must contain non-equilibrium resources. The paper extends the same idea to 'catalytic channels', where the environment is invariant for every input state, and proves a strict hierarchy of doubly-stochastic channels in the fully degenerate case. It also shows that thermal operations already capture all robust catalytic advantages, so no catalyst is needed.","feed_headline":"Thermal operations are exactly the equilibrium-preserving channels","feed_subtitle":"If true, any non-thermal thermalization channel must disturb its environment, revealing non-equilibrium resources.","key_machinery":"The load-bearing object is the 'equilibrating dilation' (Definition 1): a dilation $(U,\\omega_B)$ of a channel $T$ on system $A$ such that $T(\\rho)=\\operatorname{Tr}_B(U\\rho\\otimes\\omega_B U^\\dagger)$ and the environment is invariant when the system sits in the fixed point $\\omega_A$, i.e. $\\operatorname{Tr}_A(U\\omega_A\\otimes\\omega_B U^\\dagger)=\\omega_B$. Lemma 3—the 'informational zeroth law'—shows that equilibrium already forces global stationarity, $U\\omega_A\\otimes\\omega_B U^\\dagger = \\omega_A\\otimes\\omega_B$, from which Proposition 4 derives the commutation and covariance properties that make the channel a thermal operation. The second main object is the 'catalytic dilation' (Definition 2), where the environment is invariant for every input state; Proposition 10 characterizes these dilations by the partial transpose $U^{\\top_A}$ being unitary, which connects catalytic channels to dual-unitary tensors. The strict inclusions in the doubly-stochastic hierarchy are then proven by extremality arguments (Lemma 13) and Schur multipliers (Proposition 15).","core_discovery":"The central claim is an exact characterization: a quantum channel is a thermal operation if and only if it admits an equilibrating dilation with respect to a full-rank fixed point $\\omega_A = e^{-\\beta H_A}/Z_A$, meaning there exist a unitary $U$ and environment state $\\omega_B$ such that $T(\\rho)=\\operatorname{Tr}_B(U\\rho\\otimes\\omega_B U^\\dagger)$ for all $\\rho$ and $operatorname{Tr}_A(U\\omega_A\\otimes\\omega_B U^\\dagger)=\\omega_B$. Proposition 4 derives from this equilibrium condition that $[U,\\omega_A\\otimes\\omega_B]=0$ and that $T$ is covariant under the modular flow $\\rho\\mapsto \\omega_A^{it}\\rho\\omega_A^{-it}$; when $\\omega_A$ is full rank it can be read as a Gibbs state, so these are precisely time-translation covariance and total energy conservation. Conversely, every thermal operation obviously satisfies the equilibrium condition. A corollary is that the gap between thermal operations and looser classes such as enhanced thermal operations or Gibbs-preserving maps is exactly the failure of equilibrium preservation: for any dilation of such a channel, the environment state must change, even when the system is at equilibrium. For the fully degenerate case the paper determines the structure of doubly-stochastic channels, proving the strict chain $\\mathrm{MU}\\subsetneq\\mathrm{CAT}\\subsetneq\\mathrm{EQ}\\cap\\mathrm{DS}=\\text{strongly factorizable}\\subsetneq F\\subsetneq \\mathrm{DS}$, and links catalytic channels to dual-unitary circuits.","pith_inferences":["If the characterization holds, 'equilibrium' is a more primitive notion than temperature: any full-rank state can serve as the equilibrium state, and the Hamiltonian and inverse temperature emerge from $-\\log\\omega_A$ rather than being put in by hand; this could extend thermal operations to systems without a pre-specified Hamiltonian.","The F-layer of the hierarchy depends on the negative resolution of the Connes embedding problem via MIP*=RE, which the paper itself flags as not yet fully peer-reviewed; if that resolution ever fails, the hierarchy could collapse to MU $\\subsetneq$ CAT $\\subsetneq$ EQ$\\cap$DS = F $\\subsetneq$ DS.","The duality between catalytic and dual-unitary tensors suggests a two-way transfer: known results about dual-unitary quantum circuits (e.g., integrability and spectral properties) might yield new constraints on catalytic channels, and vice versa.","A natural next test is whether the hierarchy survives for non-degenerate Hamiltonians through Gibbs-embedding, a route the paper mentions but leaves open; if it does, the 'environment must change' signature could be used experimentally to certify whether a purported heat bath is truly thermal."],"forward_implications":["Thermal operations are the only channels whose environment can stay in equilibrium with a full-rank fixed point; every Gibbs-preserving non-thermal channel has dilations that must disturb the environment, so non-equilibrium resources are unavoidable in any implementation.","Enhanced thermal operations and general Gibbs-preserving maps are strictly larger than thermal operations, and the difference is exactly this environmental disturbance.","Among doubly-stochastic channels, mixed-unitary, catalytic, equilibrating, and factorizable channels form a strict hierarchy; in particular catalytic channels are not all mixed unitary, and not every doubly-stochastic channel is catalytic.","Thermal operations gain nothing from robust catalysis: every robust catalytic thermal operation can be implemented without a catalyst.","Catalytic channels correspond to dual-unitary circuits and catalytic unitaries, linking thermodynamic channel structure to exactly solvable circuit models."],"supporting_citations":[{"why":"This reference defines thermal operations, the class of channels under study.","marker":"[1]"},{"why":"This reference gives the original energy-preserving dilation form of thermal operations.","marker":"[9]"},{"why":"This reference introduces Gibbs-preserving maps, the broader class thermal operations are contrasted with.","marker":"[5]"},{"why":"This reference develops factorizable channels and their link to Connes' embedding problem.","marker":"[13]"},{"why":"This reference shows factorizable and strongly factorizable maps coincide iff Connes' embedding problem holds.","marker":"[14]"},{"why":"This reference is the MIP*=RE result giving a negative resolution of Connes' embedding problem, and the paper notes no fully peer-reviewed proof exists.","marker":"[15]"},{"why":"This reference shows catalytic channels are doubly-stochastic and posed the open separation problem.","marker":"[18]"},{"why":"This reference characterizes catalytic dilations by unitarity of the partial transpose, a central tool.","marker":"[19]"},{"why":"This reference supplies extremal exactly factorizable channels that are not catalytic, which proves CAT $\\subsetneq$ EQ$\\cap$DS.","marker":"[41]"},{"why":"This reference raised the robust catalysis question that Lemma 8 answers.","marker":"[34]"}],"fun_headline_variants":["Thermal ops = equilibrium-preserving channels","Equilibrium preservation uniquely defines thermal operations","Thermal ops are the only channels with equilibrating dilations","Bath invariance at equilibrium pins down thermal operations","Thermal ops uniquely preserve system-bath equilibrium"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof that thermal operations equal equilibrating channels assumes the fixed point is a full-rank state so that it can be written as a Gibbs state, and the strict inclusion of factorizable channels in the hierarchy rests on the Connes embedding problem having a negative resolution, which the paper notes is not yet fully peer-reviewed.","fun_headline_variants_meta":{"raw":{"variants":["Thermal ops = equilibrium-preserving channels","Equilibrium preservation uniquely defines thermal operations","Thermal ops are the only channels with equilibrating dilations","Bath invariance at equilibrium pins down thermal operations","Thermal ops uniquely preserve system-bath equilibrium"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001275,"raw_usage":{"total_tokens":5229,"prompt_tokens":974,"completion_tokens":4255,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":590,"completion_tokens_details":{"reasoning_tokens":4183}},"tokens_in":590,"tokens_out":4255,"duration_ms":34435,"temperature":1.0,"reasoning_tokens":4183,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T15:06:13.239876+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a known enhanced thermal operation that is not a thermal operation and numerically search over all dilations $(U,\\omega_B)$ of the channel for one where $\\operatorname{Tr}_A(U\\omega_A\\otimes\\omega_B U^\\dagger)=\\omega_B$ with a full-rank Gibbs fixed point; the paper's Eq. (9) claims no such dilation exists, so finding one would refute the central characterization.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"This reference defines thermal operations, the class of channels under study."},{"cited_title":"Quantum Metrology Theory Based on Temporal Correlation Opera- tors","cited_arxiv_id":null,"evidence_quote":"This reference gives the original energy-preserving dilation form of thermal operations."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"This reference introduces Gibbs-preserving maps, the broader class thermal operations are contrasted with."},{"cited_title":"Faist, F","cited_arxiv_id":null,"evidence_quote":"This reference develops factorizable channels and their link to Connes' embedding problem."},{"cited_title":"Faist, J","cited_arxiv_id":null,"evidence_quote":"This reference shows factorizable and strongly factorizable maps coincide iff Connes' embedding problem holds."},{"cited_title":"Faist and R","cited_arxiv_id":null,"evidence_quote":"This reference is the MIP*=RE result giving a negative resolution of Connes' embedding problem, and the paper notes no fully peer-reviewed proof exists."},{"cited_title":"Janzing, P","cited_arxiv_id":null,"evidence_quote":"This reference shows catalytic channels are doubly-stochastic and posed the open separation problem."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"This reference supplies extremal exactly factorizable channels that are not catalytic, which proves CAT $\\subsetneq$ EQ$\\cap$DS."},{"cited_title":"Borsi and B","cited_arxiv_id":null,"evidence_quote":"This reference raised the robust catalysis question that Lemma 8 answers."}],"review_version":1}