{"id":"cc92757c-3bc4-443b-9fbb-ffd05b3af774","arxiv_id":"2411.17952","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"An NMR experiment on two carbon-13 qubits shows that irreversible entropy production splits into a coherence part and a population mismatch part, and that both obey a known lower bound.","lead":"This paper uses an NMR quantum processor to measure how much irreversible entropy comes from quantum coherence during a fast drive of a two-qubit system. The experiment checks a theoretical lower bound on entropy production and separates coherence from population mismatch contributions.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The theoretical curves in Figs. 4–5 are computed from a single-qubit driving Hamiltonian (Eq. 13), but the experiment is a two-qubit system with J-coupling; until the full two-qubit drive is specified, the central comparison is under-determined.","rationale":"The reader correctly identifies the effectively unitary driving assumption as a weakness; if decoherence were significant, the experimental entropy production would include incoherent contributions and the clean comparison to the unitary model would be compromised. However, the more load-bearing issue is that the theoretical model itself is not fully specified for the actual two-qubit system. Equation (13) is a single-qubit Hamiltonian, while the experiment uses two coupled 13C spins with a J-coupling term in Eq. (10). Without a precise statement of how the driving Hamiltonian acts on both qubits and how the J-coupling is treated during the drive, the unitary U_{τ,0} of Eq. (14) is ambiguous, and the theoretical curves in Figs. 4 and 5 cannot be independently reproduced or checked. This concern is concrete and directly affects the central claim that experiment and theory agree and that the generalized Clausius inequality is verified. It does not necessarily invalidate the result, because the authors may have a well-defined two-qubit drive in mind, so the paper should be revised to specify it clearly. For that reason, I do not move the verdict away from CONDITIONAL; the appropriate action is to require the missing specification. This partially agrees with the reader's emphasis on the theoretical model, but the specific weakest point I identify is the Hamiltonian specification rather than the unitary-driving assumption.","tokens_in":7806,"tokens_out":8414,"duration_ms":79269,"concrete_test":"Ask the authors to provide the explicit two-qubit unitary U_{τ,0} (or the full two-qubit Hamiltonian including the J-coupling term and the drive term for each spin) that was used to compute the solid curves in Figs. 4 and 5. Then independently recompute ΔS_irr, C(ρ_τ), and the Bures-length bound from that specification for τ = 100, 400, and 800 μs and for both final gaps; if any recomputed value differs from the plotted theoretical curve beyond numerical precision, the claimed agreement is not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The theoretical model used to generate the solid curves in Figs. 4 and 5 is under-specified. Eq. (13) defines the driving Hamiltonian as a single-qubit Hamiltonian H(ν(t)) = −(1/2)hν(t)[cos(πt/2τ)σ_x + sin(πt/2τ)σ_y], but the experiment is performed on a two-qubit 13C-glycine system whose NMR Hamiltonian, Eq. (10), includes the scalar-coupling term ℏJ_{ij} I_z^i I_z^j. The text nowhere states how Eq. (13) acts on the two qubits: whether the same drive is applied independently to each spin, whether the J-coupling is included or refocused during the GRAPE pulse, or whether Eq. (13) is an effective single-qubit Hamiltonian describing only one of the two qubits. Consequently, the unitary U_{τ,0} in Eq. (14) is not uniquely defined, and the theoretical predictions for ΔS_irr, C(ρ_τ), and 8L^2/π^2 cannot be reproduced from the information given. Because the central claim is that the experimental data match these theoretical curves and thereby 'verify' the Bures-length bound, an incomplete specification of the drive Hamiltonian leaves the comparison—and hence the central claim—unsupported.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports an NMR experiment on a two-qubit 13C-glycine system in which the system is prepared at a thermal state and then driven out of equilibrium by a time-dependent Hamiltonian. The authors measure the irreversible entropy production ΔS_irr, the relative-entropy coherence C(ρτ), and the Bures-length lower bound 8L²/π² for two final energy gaps (3600 Hz and 5000 Hz) as functions of driving time. They observe that ΔS_irr decreases as driving time increases, that C(ρτ) follows the same trend, and that ΔS_irr ≥ 8L²/π² for the tested times, which they interpret as experimental verification of a generalized Clausius inequality.","tokens_in":8034,"tokens_out":6823,"duration_ms":62117,"significance":"If the central comparison is made fully specified, the experiment would provide a useful demonstration of how coherence contributes to irreversible entropy production and of the Bures-length bound in a controlled quantum platform. The paper benefits from explicit state tomography and from computing the coherence and entropy-production terms from the same reconstructed density operator. However, the current under-specification of the two-qubit drive Hamiltonian, the missing error-bar methodology, and the unquantified unitary assumption prevent the central claims from being fully supported.","major_comments":[{"comment":"The theoretical curves in Figs. 4 and 5 are generated from the driving Hamiltonian H(ν(t)) in Eq. (13), which is written as a single-qubit Hamiltonian acting on σ_x and σ_y. The experiment, however, is performed on two coupled 13C spins whose internal Hamiltonian in Eq. (10) includes the scalar-coupling term ℏJij I_z^i I_z^j. The text never specifies how Eq. (13) acts on the two qubits: whether the same drive is applied independently to each spin, whether the J-coupling is included in or refocused out of the GRAPE pulse, or whether Eq. (13) is an effective single-qubit model after tracing out the second spin. Consequently the unitary U_{τ,0} in Eq. (14) is not uniquely defined and the theoretical predictions for ΔS_irr, C(ρτ), and 8L²/π² in Figs. 4 and 5 cannot be reproduced from the information given. This under-specification affects the central claim that the data verify the generalized Clausius inequality, because the comparison is to curves whose generating Hamiltonian is ambiguous. Please specify the full two-qubit drive, including how the J-coupling is treated, and state explicitly how the computed observables are obtained from the two-qubit state.","section":"Section III, Eq. (13)"},{"comment":"No methodology is given for the experimental error bars shown in Figs. 4 and 5. The reader cannot tell whether the error bars are standard deviations over repeated experiments, uncertainties propagated from quantum state tomography, or instrumental estimates. Because the quantitative agreement between the measured points and the theoretical curves is part of the evidence for the paper's claims, the error-bar definition and propagation procedure must be stated.","section":"Section IV, Figs. 4 and 5"},{"comment":"The assumption that the driving is effectively unitary is stated in Section III ('This driving time is much less than the decoherence times...') but no numerical values for T1, T2, or T2* are given for the 13C spins under the experimental conditions, and the assumption is not tested, for example by checking the purity of the reconstructed states against the expected unitary evolution. Since decoherence and relaxation would add incoherent contributions to the measured irreversible entropy, the clean comparison to the unitary theoretical model requires either quantitative support for the assumption or a discussion of how such contributions were identified and excluded.","section":"Section III, unitary assumption"},{"comment":"The abstract and conclusions state that the experiment 'verified' the generalized Clausius inequality. Since Eq. (7) is a mathematical inequality derived from the formalism, the experiment cannot verify it in the sense of testing a conjectured physical law; the data can at most show that the reconstructed states satisfy the bound, which is an internal consistency check. The stronger and more informative experimental claims are the τ-dependence of ΔS_irr and C(ρτ) and their comparison to the unitary model. Please reframe the 'verification' language accordingly.","section":"Abstract and Conclusions"}],"minor_comments":[{"comment":"The sentence 'The irreversible entropy produced is more for a final energy gap of 5000 Hz (Figure 4) as compared to a final energy gap of 3600 Hz (Figure 5)' has the figure numbers reversed; Fig. 4 corresponds to 3600 Hz and Fig. 5 to 5000 Hz.","section":"Section IV, text after Figs. 4 and 5"},{"comment":"The text attributes the inequality in Eq. (7) to 'Bures [21]' and calls it a 'Bures length inequality,' but Eq. (7) as a generalized Clausius inequality is due to Deffner and Lutz (Ref. [3]); Ref. [21] is Braunstein and Caves. Please correct the attribution.","section":"References, Eq. (7)"},{"comment":"The notation D(∆τ[ρτ || ρB]) in Eq. (3) is ambiguous and ρB is undefined. If the intended expression is D(∆τ[ρτ] || ρB), define ρB and the dephasing map Δτ explicitly, since this determines the decomposition.","section":"Equation (3)"},{"comment":"The symbol F is used both for Uhlmann fidelity in Eq. (9) and for the Hilbert-Schmidt inner-product fidelity in Eq. (15); use distinct notation or state explicitly that Eq. (15) is not the Uhlmann fidelity.","section":"Equations (9) and (15)"}],"recommendation":"major_revision","confidential_remarks":"This is a straightforward experimental paper whose main issues are under-specification and framing. The missing details of the two-qubit drive and error-bar methodology are fixable in revision, and the 'verification' language should be toned down. If the authors provide those details, the paper could become publishable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this one if you want a concrete NMR check of the Francica/Santos decomposition. The experiment is straightforward: pseudopure state on 13C-glycine, thermal initialization, GRAPE pulses that drive the energy gap from 2000 Hz to 3600/5000 Hz, tomography, and plots of ΔSirr, C(ρτ), and 8L²/π² versus driving time. The data track the theoretical curves, the bound is satisfied at every point, and the data points sitting slightly above the curves are consistent with the stated near-unitary assumption plus residual decoherence. That is genuine, useful experimental evidence for the quantum thermodynamics of small systems.\n\nWhat is new is limited but real: earlier NMR work measured irreversible entropy and work in similar driven qubits; this paper explicitly separates the coherence contribution and reports it as a function of driving time for two final gaps. The theory is not new (Deffner-Lutz, Francica et al., Santos et al.), and Eq. (3) is an identity for any state relative to a diagonal reference, so matching the two sides is mostly a consistency check. The nontrivial comparison is the one against the unitary model.\n\nSoft spots, in proportion:\n- The stress-test concern is legitimate. Eq. (13) is a single-qubit Hamiltonian, but the experiment uses two coupled 13C spins with J-coupling. The paper never states whether the drive is applied to both spins, whether the coupling is refocused, or whether Eq. (13) is an effective single-qubit description. Without that, the unitary in Eq. (14) and the solid curves in Figs. 4–5 are not uniquely defined. This is the main reason the verification is incomplete as written.\n- Error bars in Figs. 4–5 have no stated methodology. Not fatal, but a referee should ask.\n- The symbol τ is used both for the fixed 1/(2J) delay in the pulse sequence and for the variable driving time. Confusing, likely a typo.\n- “Verified a generalized Clausius inequality” is an overclaim. Better: observed data consistent with the bound for two protocols.\n\nThe citation pattern is fine: prior theory and experiments are credited, and the tomography refs are legitimate. The central claim is plausible and probably correct. It needs the Hamiltonian specification and error analysis before publication.\n\nThis is a paper for experimentalists in quantum thermodynamics and NMR quantum information. It deserves a serious referee; the verdict should be conditional on the drive Hamiltonian being fully specified and the error-bar method stated.","headline":"Real NMR data on the coherence/population split of irreversible entropy, but the two-qubit drive behind the theoretical curves is under-specified and the 'verification' claim is stronger than the evidence.","tokens_in":8589,"tokens_out":3892,"would_cite":false,"duration_ms":33041,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.65.-w","03.67.-a","05.70.Ln"],"model":"deepseek-v4-flash","headline":"In a two-qubit NMR processor, irreversible entropy from unitary driving is shown to split into coherence and population-mismatch parts, with the total always exceeding a Bures-length lower bound.","keywords":["irreversible entropy production","quantum coherence","generalized Clausius inequality","Bures length","NMR quantum processor","nonequilibrium thermodynamics","unitary driving","quantum state tomography"],"falsifier":"Repeat the protocol with drive durations comparable to the qubits' decoherence time, or add a variable delay after the drive before tomography: the unitary prediction of monotonically decreasing entropy with $\\tau$ will break down, and the bound would be falsified if the measured $\\Delta S_{\\mathrm{irr}}$ falls below $(8/\\pi^2)L^2(\\rho_\\tau, \\rho_f)$ within experimental uncertainty.","tokens_in":7573,"feed_emoji":"⚛️","tokens_out":9799,"duration_ms":79413,"temperature":0.7,"pith_summary":"This paper reports an experiment on a two-qubit NMR processor in which a thermal equilibrium state is driven out of equilibrium by a time-dependent unitary Hamiltonian, with driving times from 100 to 800 microseconds. The authors measure the irreversible entropy produced in the process and show that it splits into two positive parts: one from quantum coherence created during the drive and one from a population mismatch between the actual final state and the target equilibrium state. They also verify a generalized Clausius inequality: in every run, the irreversible entropy remains above a nonzero lower bound set by the Bures length between the actual and equilibrium states. The result matters because it tests a thermodynamic bound that is much tighter than the classical statement that entropy production is merely nonnegative, on actual quantum hardware rather than in theory alone.","feed_headline":"NMR spins confirm irreversible entropy clears a geometric bound","feed_subtitle":"Driven carbon-13 nuclei split entropy into two parts and stay above a tighter-than-Clausius lower bound.","key_machinery":"The argument rests on two identities. First, the irreversible entropy $\\Delta S_{\\mathrm{irr}} = D(\\rho_\\tau \\| \\rho_f)$ is decomposed as $C(\\rho_\\tau) + D(\\Delta_\\tau[\\rho_\\tau] \\| \\rho_B)$, where $C(\\rho_\\tau) = S(\\Delta_\\tau[\\rho_\\tau]) - S(\\rho_\\tau)$ is the relative entropy of coherence of the actual state and the second term is the population mismatch with the target equilibrium state. Second, the generalized Clausius inequality bounds this entropy from below by $(8/\\pi^2) L^2(\\rho_\\tau, \\rho_f)$, where $L(\\rho_\\tau, \\rho_f) = \\arccos \\sqrt{F(\\rho_\\tau, \\rho_f)}$ is the Bures length built from the fidelity $F$. Experimentally, the machinery is an NMR processor that prepares a Gibbs state at fixed pseudospin temperature, applies a GRAPE-optimized unitary drive with time-dependent energy gap, and reconstructs the final density matrix by quantum state tomography so that all three quantities can be computed and compared.","core_discovery":"On the paper's own terms, the central discovery is that the unitary-driving protocol on 13C-labeled glycine qubits produces entropy production $\\Delta S_{\\mathrm{irr}} = D(\\rho_\\tau \\| \\rho_f)$ whose two contributions, coherence generation $C(\\rho_\\tau)$ and population mismatch $D(\\Delta_\\tau[\\rho_\\tau] \\| \\rho_B)$, follow the predicted decomposition (Eq. 3), and that the measured values satisfy the generalized Clausius inequality $\\Delta S_{\\mathrm{irr}} \\geq (8/\\pi^2) L^2(\\rho_\\tau, \\rho_f)$ (Eq. 7) for both final energy gaps studied. The data show that entropy production decreases as driving time increases toward the quasi-static limit, that the coherence contribution tracks the total and becomes almost negligible at long times, and that driving farther from equilibrium produces more total entropy but a smaller share from coherence. Experimental points lie slightly above the theoretical curves, which the authors attribute to pulse calibration, state preparation, field inhomogeneities, and GRAPE optimization errors, all of which add extra entropy.","pith_inferences":["A testable extension the authors leave implicit: deliberately adding pulse noise or using non-optimal drives should increase the gap between $\\Delta S_{\\mathrm{irr}}$ and its Bures-length bound while leaving the inequality intact, which would probe how tight the bound is.","If the decomposition of Eq. (3) is generic, control strategies that target coherence suppression separately from population targeting could be used to design low-dissipation unitary strokes for quantum engines.","Extending the same measurement to open-system settings, where a bath is present during the drive, would require modifying the second term to include bath-induced transitions; whether the Bures bound survives that generalization is not addressed by this experiment."],"forward_implications":["Longer driving times monotonically reduce irreversible entropy production, so the quasi-static limit is approached from above as the control parameter is changed more slowly.","The coherence contribution to entropy production follows the same trend as the total, so reducing coherence generation is a practical route to lowering irreversibility in driven quantum processes.","Driving the system farther from equilibrium raises total entropy production but lowers the fraction of it that comes from coherence; population mismatch then dominates.","The Bures-length lower bound holds for both final gaps examined and for every driving time, making it a valid tighter alternative to the plain Clausius inequality in these unitary-driven processes."],"supporting_citations":[{"why":"Derives the generalized Clausius inequality $\\Delta S_{\\mathrm{irr}} \\geq (8/\\pi^2)L^2$ that the experiment verifies.","marker":"[3]"},{"why":"Supplies the decomposition of irreversible entropy into coherence and population-mismatch contributions.","marker":"[5]"},{"why":"Defines the relative entropy of coherence used to quantify the coherence contribution $C(\\rho_\\tau)$.","marker":"[18]"},{"why":"Underlies the Bures-length bound by relating fidelity to a statistical distance between quantum states.","marker":"[21]"},{"why":"Supplies the work fluctuation relation used to identify irreversible work with relative entropy and justify $\\langle w \\rangle \\geq \\Delta F$.","marker":"[23]"},{"why":"Defines the quantum relative entropy $D(\\rho_\\tau \\| \\rho_f)$ that measures the entropy production.","marker":"[24]"},{"why":"Provides the GRAPE optimal-control method used to implement the high-fidelity driving unitary.","marker":"[37]"},{"why":"Supplies the quantum state tomography procedure used to reconstruct the experimental density matrices.","marker":"[38]"}],"fun_headline_variants":["NMR experiment dissects entropy into coherence and mismatch","Quantum test: coherence drives entropy production bound","NMR spins split entropy, beat Clausius bound","Coherence contribution to entropy measured in NMR","Driven nuclei: entropy from coherence obeys new bound"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The comparison to the unitary model assumes the drive is effectively unitary because 100 to 800 microseconds is far shorter than the qubits' decoherence times, so any dissipative error is neglected.","fun_headline_variants_meta":{"raw":{"variants":["NMR experiment dissects entropy into coherence and mismatch","Quantum test: coherence drives entropy production bound","NMR spins split entropy, beat Clausius bound","Coherence contribution to entropy measured in NMR","Driven nuclei: entropy from coherence obeys new bound"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000209,"raw_usage":{"total_tokens":1356,"prompt_tokens":843,"completion_tokens":513,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":459,"completion_tokens_details":{"reasoning_tokens":440}},"tokens_in":459,"tokens_out":513,"duration_ms":5141,"temperature":1.0,"reasoning_tokens":440,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T11:42:57.841631+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the protocol with drive durations comparable to the qubits' decoherence time, or add a variable delay after the drive before tomography: the unitary prediction of monotonically decreasing entropy with $\\tau$ will break down, and the bound would be falsified if the measured $\\Delta S_{\\mathrm{irr}}$ falls below $(8/\\pi^2)L^2(\\rho_\\tau, \\rho_f)$ within experimental uncertainty.","supporting_citations":[{"cited_title":"Umegaki, Kodai Mathematical Seminar Reports 14, 59 (1962)","cited_arxiv_id":null,"evidence_quote":"Defines the quantum relative entropy $D(\\rho_\\tau \\| \\rho_f)$ that measures the entropy production."}],"review_version":1}