{"id":"0cb47794-cd54-46fa-ad84-c60ec0a7e169","arxiv_id":"2508.20630","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A parameter-free 'total energy approach' using adiabatic first-principles MD is claimed to reproduce the specific heat of liquids and the glass-transition jump, and to resolve thermodynamic puzzles about equilibrium and hysteresis.","lead":"This review argues that the specific heat of liquids and glasses can be computed directly from first-principles molecular dynamics simulations of the total energy, without invoking phonon models, and that adiabatic (NVE) simulations are essential for correct temperature determination. The author's prior simulations of liquid sodium and glycerol are presented as evidence that the approach reproduces the temperature dependence of C_V and the specific-heat jump at the glass tran","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Parameter-free ΔCgl claim depends on unique-equilibration premise that finite-time NVE MD cannot certify; numerical fits are also underconstrained.","rationale":"The reader's verdict is already conditional; this stress-test identifies a specific, testable version of the same concern. The paper's headline numerical success (glycerol ΔCgl) is the only claimed quantitative prediction without empirical parameters. That success depends on two unsecured links: (i) the GB axiom applied to finite-time NVE MD, which is not verifiable from the published trajectories; and (ii) the choice of linear fits to scattered Est(T) data. Because Tm and Tg are both overestimated by roughly a factor of two, the temperature axis is not independently validated, so an accidental slope match is plausible. A rerun with controlled initial conditions and fit windows would directly settle whether C is history-independent and whether the fitted ΔCgl is robust. If the check passes, conditional acceptance is justified; if not, the universal claim should be weakened. This does not change the reader's conditional verdict, but it sharpens the condition that must be met.","tokens_in":46289,"tokens_out":6284,"duration_ms":71595,"concrete_test":"Take the glycerol system of Sec. 4.3 at a U value in the glassy regime just below Tg. Run at least three independent NVE MD trajectories with the same U and V but initial positions drawn from different cooling histories (e.g., different quench rates or different snapshots of the cooling run). If, after several times the apparent structural relaxation time, the time-averaged kinetic temperatures from Eq. (62) differ by more than the expected √N kinetic-energy fluctuation (Eq. 69), or if the linear-fit ΔCst changes by more than 0.1R when the fit windows above/below Tg are varied, the single-valuedness of Eq. (60) and the reliability of the reported ΔCgl are not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that NVE ('adiabatic') DFT-MD gives a parameter-free C(U) for liquids and glasses—requires the GB uniqueness axiom invoked in Sec. 3.3.3: for fixed U and constraints there is one equilibrium, so the mapping (U,{R_j(0)})→(T,{R_j^K}) in Eq. (60) is single-valued and T from Eq. (62) is a thermodynamic temperature. For a glass, structural relaxation times exceed accessible MD times by many orders of magnitude. The paper's equilibrium checks (Eq. 63: ⟨δR_j^2⟩ constant for solid-like, linear for liquid-like) only certify apparent stationarity on the simulation window; they cannot distinguish a metastable basin from the GB-unique equilibrium state. If two initial configurations with the same U relax to different T in finite time, U(T) is history-dependent and C is not a state function. Section 4.3 does not remove this possibility: Tm=635 K vs 291 K and Tg=310 K vs 185 K, and ΔCgl=0.73R is obtained from linear fits of scattered Est data above/below Tg plus an estimated ΔCte≤0.15R, with no error bars or fit-window sensitivity analysis. The agreement with the experimental 0.70R is therefore not yet evidence for the universal, history-independent claim.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a 'total energy approach' to computing the specific heat of liquids and glasses from first-principles molecular dynamics. Instead of decomposing the energy into phonon-like elemental excitations, the method uses direct adiabatic (NVE) DFT-MD simulations: for a fixed total energy U and initial atomic positions, the system is allowed to relax to an equilibrium state characterized by a temperature from the kinetic-energy average (Eq. 62). The GB statement of the second law is invoked to guarantee one stable equilibrium per (U, constraints), which makes the mapping (U,{R_j(0)})→(T,{R_j^K}) single-valued. The paper reviews experimental facts on C_V of liquids, emphasizes the role of atom relaxation and hysteresis, and presents two applications: the decreasing C_V(T) of liquid Na and the specific-heat jump ΔC_gl at the glass transition of glycerol. For glycerol, linear fits of the structural energy Est above and below T_g give ΔC_st = 0.58R; adding an estimated ΔC_te ≤ 0.15R yields ΔC_gl = 0.73R, compared with the experimental 0.70R. The calculated T_m and T_g for glycerol are 635 K and 310 K, versus experimental values of 291 K and 185 K.","tokens_in":46598,"tokens_out":3299,"duration_ms":40713,"significance":"If the central claim held, the method would be a significant conceptual advance: a parameter-free, theory-agnostic route to the specific heat of non-periodic systems, avoiding the questionable phonon decomposition for liquids. The paper gives proper credit to prior phonon-based work and clearly identifies the finite-size rounding of first-order transitions and the multi-timescale nature of relaxation as key physical issues. The honest reporting of the overestimated T_m and T_g and of the negative-slope artifact in the Si melting curves is a strength, as is the explicit discussion of hysteresis as a state-space projection effect. However, the quantitative centerpiece—the parameter-free ΔC_gl—depends on linear fits to scattered data and an order-of-magnitude estimate, and the conceptual cornerstone (uniqueness of the equilibrium state reached by finite-time NVE MD) is asserted rather than verified. The paper is therefore a thought-provoking review with suggestive results, but the universal, parameter-free claim is not yet established to the standards implied by the abstract.","major_comments":[{"comment":"The central claim 'first time to calculate the jump ΔC_gl without any empirical parameter' is not supported with error analysis. The Est data are visibly scattered; ΔC_st = 0.58R comes from linear fits above and below T_g with no reported fit statistics, no error bars, and no sensitivity to the chosen T ranges. ΔC_te is 'estimated to be at most 0.15R', not calculated. The final ΔC_gl = 0.73R is thus a sum of an unquantified slope difference and an estimate. Without propagation of uncertainty, the agreement with the experimental 0.70R is suggestive but not demonstrated. The fact that T_g and T_m are overestimated by nearly a factor of two (310 K vs 185 K; 635 K vs 291 K) further leaves open the possibility that the agreement in ΔC_gl is coincidental.","section":"§4.3, Fig. 14"},{"comment":"The load-bearing premise is the GB uniqueness statement and the one-way mapping (U,{R_j(0)})→(T,{R_j^K}). The equilibrium checks in Eq. (63)—constant MSD for solids, linear MSD for liquids—only certify apparent stationarity on the simulation window. For a glass where structural relaxation times vastly exceed MD times, the check cannot distinguish a metastable basin from the unique stable equilibrium. The paper itself states that 'a large hysteresis indeed occurs' in Fig. 14 for cooling vs reheating, which demonstrates that different initial configurations at similar U reach different states. If the mapping is not single-valued on practical timescales, then U(T) is history-dependent and C = dU/dT is not a state function. A concrete test would be: for fixed U, start NVE runs from several independent configurations and show that the same T and the same {R_j^K} are reached within statistical","section":"§1.2 and §3.3.3, Eq. (60), Eq. (63)"},{"comment":"The liquid-sodium result, while qualitatively correct, is quantitatively off by about a factor of two: the computed dC_V/dT = −0.94×10^-3 R/K versus the experimental −1.6×10^-3 R/K, with the origin of the discrepancy left 'unclear'. Moreover, the region 370 < T < 500 K is handled by a linear interpolation of U(T), producing a constant C_V = 4.05R that the authors explicitly say 'does not have physical meaning and should not be compared with experiment'. Thus the universal, parameter-free prediction currently has predictive power only for the sign and rough magnitude of the temperature slope, not for the detailed shape of C_V(T) including the near-T_m region. This should be stated more prominently when the 'first time to demonstrate' claim is made for the decreasing behavior.","section":"§4.1.1, Fig. 10"},{"comment":"Convergence with respect to system size is not demonstrated for the glass calculation. For Si, the finite-size width W_m remains about 300 K even for N=512 atoms, and Eq. (69) shows slow convergence as N^-1/2. The glycerol calculation does not state the cell size, the MD duration, or how many independent runs were used, although the Est data in Fig. 14 are 'scattered'. Because the linear fits used for ΔC_st are performed over a temperature range that includes finite-size rounding effects, a convergence study with N (e.g., 64, 216, 512 atoms) and a reporting of simulation lengths are necessary to establish that the obtained slope difference is not a finite-size artifact.","section":"§3.3.3 and §4.2, Eq. (69), Fig. 13"}],"minor_comments":[{"comment":"Numerous typographical issues remain: 'a ffects', 'di fficult', 'Eherenfest' (should be Ehrenfest), 'di fference', 'visco se', 'overbinding of the LDA/GGA functional' is discussed without reference to the functional used in the glycerol calculations. A careful proofread is needed.","section":"Throughout"},{"comment":"The notation for the average in Eq. (62) is confusing: the brackets denote particle average and the bar denotes time average, but the expression ⟨M_j v_j(t)^2⟩_j mixes a particle index with a time argument. Clarify the order of averaging and the exact window used for the time average.","section":"§3.3.3, Eq. (62)"},{"comment":"The estimate ΔC_te ≤ 0.15R is stated without a derivation or a reference to a specific experimental value for glycerol's thermal expansion or compressibility near T_g. Since this term is added to obtain the headline number 0.73R, a one-source estimate is too thin; provide the underlying data or otherwise remove the term from the headline claim.","section":"§4.3"},{"comment":"The inequality G(gl) < G(li) for T < T_g is claimed to follow from d^2G/dT^2 = −C_P/T < 0. This is correct only if the liquid and glass curves are compared at the same T and if both are equilibrium branches; the derivation should explicitly state that the glass branch is treated as an equilibrium branch on the timescale of Eq. (41).","section":"§5.3.3, Eq. (77)"},{"comment":"Several key claims rely on the author's own arXiv preprints (Refs. [35], [56], [259]) that are not peer-reviewed or not yet published. When these are load-bearing (e.g., the definition of state variables for solids and the detailed hysteresis treatment), the reader should be told which results are established in the current paper and which are taken from those preprints.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"This is a review-style paper that re-presents results already published by the same author in a series of papers and preprints; the novelty lies mainly in the thermodynamic synthesis and the framing around the GB statement of the second law. The journal should consider whether such a synthesis is within scope. The strongest original element is the adiabatic NVE strategy for avoiding thermostat artifacts, but the load-bearing numerical evidence for glycerol is thin: no error bars, no system-size or run-length convergence, and a hand-waving estimate for one of the two terms. If the authors can supply a convergence study and a proper uncertainty estimate, the paper could become a valuable contribution; without those, the central quantitative claim exceeds the evidence."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The thing to know: this is a review article, not new primary work. The author's adiabatic NVE DFT-MD approach (compute U(T) directly, get CV from slopes) was already published, and this paper synthesizes that program, adds thermodynamic framing, and criticizes phonon-based models for liquids.\n\nWhat the paper does well: the critique of the phonon model for liquids is substantive. The point that the additive quasi-particle decomposition requires eigenstates, which liquids don't have, is a real and often glossed-over issue. The numerical example in Fig. 12, where phonon decomposition produces a negative structural contribution to CV, is a nice demonstration. The discussion of hysteresis—arguing that with a full set of state variables (including atom equilibrium positions) specific heat is a single-valued state function—is a legitimate conceptual contribution, even if it is developed more fully elsewhere.\n\nThe soft spots are where the reader's conditional score is earned. The central claim of a 'parameter-free' calculation of ΔCgl rests on linear fits to scattered Est data above and below Tg, an estimated upper bound on ΔCte, and a linear interpolation in the melting region of liquid Na. No error bars or sensitivity analysis are given. The large overestimates of Tm (635 vs 291 K) and Tg (310 vs 185 K) are acknowledged but left unresolved; if the functional overbinds, the 'universal' claim is premature, and the ΔCgl agreement (0.73R vs 0.70R) could be partly fortuitous.\n\nThe deeper issue, spelled out in the stress-test note, holds up: the framework depends on the Gyftopoulos-Beretta uniqueness axiom—for fixed U and constraints there is one equilibrium—but in a glass, structural relaxation times exceed MD times by orders of magnitude. The checks in Eq. (63) only certify stationarity on the simulation window; they cannot distinguish a metastable basin from the unique equilibrium. If different initial configurations with the same U relax to different temperatures in finite time, U(T) is history-dependent and CV is not a state function. The paper's own hysteresis loops show process dependence; that's consistent with the author's state-variable story, but it means the history-independence central to the 'parameter-free' claim is not demonstrated here.\n\nWho is this for? Someone wanting a thorough statement of the total energy approach and its thermodynamic defense, or a compact account of why phonon models for liquids are suspect. It is not the place to find new data or new derivations. It deserves a serious referee—the claims are provocative and would benefit from careful checking—but the referee should push for tempering the universal claim and for real uncertainty quantification.\n\nMy advice: send it to review, with a recommendation for major revision.","headline":"A clear but overclaiming review of the author's adiabatic total-energy method; the phonon critique has teeth, the universal/parameter-free claim does not.","tokens_in":47057,"tokens_out":4389,"would_cite":true,"duration_ms":49124,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["82B30","82D30"],"pacs":["65.20.-w","64.70.Pc"],"model":"deepseek-v4-flash","headline":"The paper argues that the specific heat of liquids and glasses can be obtained directly from the total energy in adiabatic first-principles molecular dynamics, without phonon decompositions or empirical parameters, and that the resulting fr","keywords":["specific heat","liquids","glasses","first-principles molecular dynamics","density functional theory","glass transition","atom relaxation","hysteresis"],"falsifier":"Run two adiabatic NVE simulations of the same liquid with the same total energy and volume but very different initial atomic configurations, and wait until the structural relaxation completes. If the two runs settle at measurably different time-averaged temperatures, the uniqueness premise behind the method fails; if they converge to the same temperature, the central mapping survives.","tokens_in":46193,"feed_emoji":"🌡️","tokens_out":7737,"duration_ms":83265,"temperature":0.7,"pith_summary":"This review argues that the specific heat of liquids and glasses should be computed by the total energy approach: take the internal energy U directly from density-functional-theory molecular dynamics, then differentiate with respect to temperature. It claims that liquids have no true eigenstates, so the additive quasi-particle decomposition that underlies phonon theory is not valid, and the real obstacle is atom relaxation rather than the absence of periodicity. The paper uses the second law of thermodynamics to justify assigning temperature from the time-averaged kinetic energy in adiabatic NVE simulations, and shows that this reproduces the falling heat capacity of liquid sodium and, for the first time without empirical parameters, the specific-heat jump at the glass transition of glycerol. If the argument is right, it gives a universal, parameter-free route to the heat capacities of disordered materials and recasts the glass transition as a thermodynamic transition rather than a nonequilibrium state.","feed_headline":"No fitting parameters: specific heat for liquids and glasses","feed_subtitle":"Adiabatic first-principles runs reproduce the glass-transition jump and sodium's falling heat capacity with no fitting.","key_machinery":"The load-bearing object is the pair formed by the total energy approach and the adiabatic relaxation simulation: direct time-averaging of the density-functional total energy in constant-NVE molecular dynamics, with temperature assigned through the equipartition relation (3/2)kBT = (1/2)⟨M v^2⟩ from the time-averaged kinetic energy. The thermodynamic justification is a modern statement of the second law asserting that for fixed internal energy and fixed constraints there is exactly one stable equilibrium state; this licenses the mapping from (U, initial atomic positions) to (T, equilibrium atomic positions). For solids, the internal energy separates into a structural part Est and a phonon par","core_discovery":"The central claim is that the isochoric specific heat of any material phase, including liquids and glasses, is obtained by computing U(T) as the time average of the DFT total energy over adiabatic (NVE) molecular dynamics and taking the numerical derivative. The paper contrasts this total energy approach with the elemental excitation approach, in which total energy is the sum of independent quasi-particle energies; that additive property holds only for eigenstates, and liquids lack eigenstates because atom relaxation destroys them. The author's calculations show that the structural energy term Est, which carries configurational and relaxation contributions, is responsible for the specific-he","pith_inferences":["A natural next test is to apply the same adiabatic NVE protocol to molecular liquids with internal vibrational modes, especially water, whose large CP is unexplained by phonons; the sensitivity of the structural-energy term to long-range DFT functionals will likely decide whether the approach captures it.","Since the paper finds that the activation energy extracted from CV is much smaller than those from diffusion or viscosity, Arrhenius analysis of heat capacity could become a cheap experimental probe of the weakest relaxation channel in a liquid.","The finite-size argument suggests that broadened specific-heat peaks near transitions with disorder may contain a latent-heat component; comparing the integrated peak area with directly measured latent heat in a material such as quartz would test this interpretation.","If the glass transition is truly thermodynamic, the Prigogine-Defay ratio inequality should follow from the structural-energy change alone; the paper asserts this connection, and a numerical survey across many glass formers would strengthen it."],"forward_implications":["For any material whose DFT forces are affordable, CV and, by integration, CP and free energies become computable with no empirical input, including for supercooled liquids and glasses.","The glass transition is classified as a structural, thermodynamically driven transition: below Tg the glass, not the liquid, is the stable equilibrium state.","Observed hysteresis in C-T curves is compatible with equilibrium thermodynamics once equilibrium atomic positions are counted as state variables; history dependence no longer forces a nonequilibrium label.","Phonon-based estimates of liquid heat capacities may capture qualitative trends but cannot be the foundation of the theory, because the Bose-Einstein occupation of liquid phonons is not justified.","Finite-size rounding of melting converts part of latent heat into apparent specific heat, so sharp lambda-like peaks in disordered or small systems should be interpreted cautiously."],"supporting_citations":[{"why":"Supplies the statement of the second law and the definition of stable equilibrium on which the temperature assignment in adiabatic MD rests.","marker":"[30]"},{"why":"Supplies the compiled experimental CV data for liquid metals near melting that the theory must reproduce.","marker":"[13]"},{"why":"The author's adiabatic-relaxation calculation of liquid sodium's CV, the main quantitative demonstration of the decreasing temperature dependence.","marker":"[52]"},{"why":"The author's parameter-free glycerol calculation of the glass-transition specific-heat jump, the central new result.","marker":"[50]"},{"why":"Extends the glass-transition result to the Prigogine-Defay ratio and the structural-energy interpretation.","marker":"[51]"},{"why":"Provides the independent-particle and Bose-Einstein machinery whose validity for liquids the paper contests, plus the master-equation background.","marker":"[12]"},{"why":"The phonon-based model of liquid CV that serves as the main alternative baseline compared against the total energy approach.","marker":"[17]"},{"why":"Establishes the plural order parameters and the thermodynamic treatment of the glass transition and hysteresis that the paper reinterprets.","marker":"[92]"},{"why":"Experimental thermodynamic data for sodium used as the comparison for the calculated CV.","marker":"[70]"},{"why":"The author's silicon melting study demonstrating finite-size transition width and the role of relaxation in melting-temperature calculations.","marker":"[180]"}],"fun_headline_variants":["Heat capacity from DFT time averages, no fits","Beyond eigenstates: heat capacity for liquids and glasses","Specific heat for glass and liquid from adiabatic MD","Liquids and glasses: specific heat from time-averaged energy","Total energy approach to heat capacity in disordered solids"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The load-bearing premise is that an isolated system with fixed total energy and fixed constraints has exactly one stable equilibrium state, and that a finite-size adiabatic molecular-dynamics run actually reaches it, so the time-averaged kinetic energy of the simulated atoms can be read as the thermodynamic temperature.","fun_headline_variants_meta":{"raw":{"variants":["Heat capacity from DFT time averages, no fits","Beyond eigenstates: heat capacity for liquids and glasses","Specific heat for glass and liquid from adiabatic MD","Liquids and glasses: specific heat from time-averaged energy","Total energy approach to heat capacity in disordered solids"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000645,"raw_usage":{"total_tokens":2821,"prompt_tokens":786,"completion_tokens":2035,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":530,"completion_tokens_details":{"reasoning_tokens":1957}},"tokens_in":530,"tokens_out":2035,"duration_ms":16154,"temperature":1.0,"reasoning_tokens":1957,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T14:56:25.678469+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run two adiabatic NVE simulations of the same liquid with the same total energy and volume but very different initial atomic configurations, and wait until the structural relaxation completes. If the two runs settle at measurably different time-averaged temperatures, the uniqueness premise behind the method fails; if they converge to the same temperature, the central mapping survives.","supporting_citations":[],"review_version":1}