{"id":"cf59cf2f-360a-4c06-8d88-a7dfa1c45346","arxiv_id":"2607.08608","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":3,"one_line_summary":"Sparse-volume Grüneisen interpolation reconstructs QHA Gibbs free energies with sub-meV/atom errors for simple solids using only 3 phonon volume points, achieving 6-9x speedup.","lead":"This paper develops a method to compute Gibbs free energies under pressure using phonon calculations at only 3 volumes instead of 20+, cutting cost by 6-9x. It matters because finite-temperature phase stability verification is the bottleneck after AI-driven materials discovery generates candidate structures.","discovery_kind":"unclear","skeptic_critique":{"model":"glm-5.2","headline":"Headline accuracy depends on QHA-informed sparse-point selection (Eq. 14), creating partial circularity in the speedup claim: the 3 phonon volumes are chosen using the full QHA V_eq(P,T) path, which requires the expensive calculation the method claims to replace.","rationale":"The reader correctly identified the smooth-Grüneisen assumption as a weakness, and the paper itself tests and bounds it (Fig. 12). However, I believe the more load-bearing concern is the sparse-point selection circularity: the headline accuracy numbers are obtained under QHA-informed volume selection, which partially undermines the practical speedup claim. This is not a fatal flaw — the method is mathematically sound, the interpolation framework is reasonable, and the paper honestly reports limitations including Ta2O5 errors. But the practical deployment story is incomplete without demonstrating that comparable accuracy can be achieved with a selection protocol that does not require the dense QHA reference. The concern does not rise to the level of changing the verdict from ACCEPT because: (1) the paper is primarily a method validation against QHA benchmarks, and using QHA to define the validation window is a legitimate validation protocol; (2) the paper provides evidence (Appendix B, Fig. A4) that the method has some robustness to point selection within the bracket; (3) the paper explicitly states that broad compression ranges reduce to simple low/mid/high selection. The concern would become verdict-changing if a proxy-based selection test showed MAE degradation of an order of magnitude, but absent that evidence, the ACCEPT verdict is appropriate with the caveat that the practical speedup may be smaller than the headline 5.9–9.0× when QHA-guided selection is unavailable.","tokens_in":22926,"tokens_out":3042,"duration_ms":144185,"concrete_test":"For each of the six simple benchmark systems, select the 3 sparse phonon volumes using only the static equation of state U(V) (no phonon information whatsoever) — e.g., at V_min ≈ 0.85V0, V0, and V_max ≈ 1.02V0 — and recompute the full MAE_G on the same P-T grids as Table I. If the average MAE increases by more than 3× relative to the QHA-informed values (i.e., above ~0.45 meV/atom average), the practical speedup claim is materially weakened because a second round of phonon calculations or an iterative refinement would be needed to achieve the reported accuracy.","verdict_should_be":"ACCEPT","load_bearing_attack":"The central speedup claim (5.9–9.0×) is measured against a QHA benchmark, but the 3 sparse phonon volumes are themselves selected using QHA knowledge. Section II.A, Eq. 14 states: 'For a target pressure-temperature window, we first determine V_eq(P,T) = argmin_V [F_QHA(T,V) + PV] from the dense QHA reference and identify the range traversed by V_eq(P,T).' The three sparse points are then chosen to bracket this QHA-derived path. This means the headline MAE values in Table I (0.148 meV/atom average) are achieved under an optimal selection that presupposes the answer. The sensitivity tests mentioned in Section II.A ('we repeat the calculation for available three-point combinations satisfying this local bracket condition') only vary combinations within the QHA-defined bracket — they do not test selections made without phonon information. Appendix Fig. A4 shows error bars for different three-volume combinations, but again only among those satisfying the bracket condition derived from QHA. In a real deployment scenario, one would not have the dense QHA reference; one would need to select sparse volumes from cheaper proxies (static EOS, or 1–2 exploratory phonon points). If such proxy-based selection yields substantially larger MAEs, the practical speedup over QHA would be smaller than claimed because additional phonon calculations or iterations would be needed to identify good sparse points. The paper does note that 'in broad compression benchmarks… this procedure reduces to using representative low-, middle-, and high-volume points,' but the Table I accuracy numbers are not reported under this simpler selection protocol — they use the QHA-informed selection.","agreement_with_reader":"partial"},"referee_report":{"model":"glm-5.2","summary":"The manuscript presents a sparse-volume Grüneisen interpolation (GI) method for reconstructing Gibbs free energies under volume compression. The approach calculates phonon spectra at only 3 selected volumes and reconstructs the full free energy surface using a ZPE-level effective Grüneisen parameter for the static-ZPE branch and piecewise mode-resolved Grüneisen slopes for the finite-temperature vibrational branch. The method is validated against dense-grid QHA benchmarks for 7 systems (C, Al, Si, Ge, TiO2, PtO2, Ta2O5), achieving sub-0.53 meV/atom MAE for the six simpler systems and 5.9-9.0x speedup, while preserving phase-stability topology for the complex Ta2O5 polymorphs.","tokens_in":23696,"tokens_out":1259,"duration_ms":190059,"significance":"The paper addresses a genuine computational bottleneck in high-throughput materials discovery: the cost of dense-volume QHA phonon calculations for finite-temperature phase-diagram construction. The mathematical formulation (Eqs. 8-13) is internally consistent and builds on the standard Grüneisen relation. The validation is systematic, with explicit MAE metrics at both ZPE and Gibbs levels, pointwise error maps, phase-diagram comparisons, and thermal-expansion coefficient recovery. The honest reporting of breakdown for Ta2O5 (MAE up to 17.5 meV/atom) and the applicable compression range (Fig. 12) is commendable. The comparison with the VIP method (Appendix B) provides useful context. The speedup factors are practically significant for the materials discovery pipeline described in the introduction.","major_comments":[{"comment":"Section II.A, Eq. (14): The sparse-volume selection procedure uses the dense QHA reference to determine V_eq(P,T) and identify the volume range traversed by the equilibrium path. The three sparse phonon points are then chosen to bracket this QHA-derived path. This creates a practical circularity in the headline speedup claim: the method requires knowledge of the expensive QHA calculation it claims to replace in order to select its own input points. The sensitivity tests mentioned in Section II.A ('we repeat the calculation for available three-point combinations satisfying this local bracket condition') and Appendix Fig. A4 only vary combinations within the QHA-defined bracket, not selections made without QHA knowledge. In a real deployment scenario, one would not have the dense QHA reference and would need to select sparse volumes from cheaper proxies (static EOS, or 1-2 exploratory phon","section":null},{"comment":"The abstract states speedups of '5.911-9.023 times' relative to QHA workflows. However, as noted above, the sparse-point selection itself requires the QHA V_eq(P,T) path (Eq. 14). If the selection cost is included, the effective speedup is smaller. The paper should either (a) demonstrate that proxy-based selection (e.g., using static EOS or a single exploratory phonon point to estimate the relevant volume range) yields comparable MAEs, or (b) explicitly state that the reported speedup assumes the sparse points are selected without the full QHA reference, and provide a practical protocol for doing so. Without this, the speedup claim is conditional on information that would not be available in the intended use case.","section":null}],"minor_comments":[{"comment":"Table I: The R² values for Si (0.887) and Al (0.981) are notably lower than for other systems (0.999 for C, 0.998 for Ge). Given that Al and Si are among the systems with larger MAE_G values (0.52 and 0.19 meV/atom respectively), a brief discussion of whether the lower R² is correlated with the reconstruction error would strengthen the analysis.","section":null},{"comment":"Section II.A, Eq. (12): The notation uses subscripts a and b for sparse-volume endpoints, but the text also uses i for sparse volumes in general. Clarifying that a and b are specific instances of i would help readers.","section":null},{"comment":"Section III.C: The Ta2O5 errors range up to 17.5 meV/atom for the lambda phase. While the paper notes this is due to 'phase-specific soft modes, mode crossings, and shallow free energy separations,' it would be useful to specify which phases have known soft modes or mode crossings in the studied pressure range, to help users assess applicability.","section":null},{"comment":"Fig. 3: The pressure ranges are indicated outside each panel but are difficult to read. Consider using matching colors/symbols for pressure labels and curves, or adding a legend.","section":null},{"comment":"Section II.A, Eq. (15): The treatment of imaginary/zero-frequency modes (simple discarding) is stated without justification. For systems near dynamical instability, this could introduce non-negligible errors. A brief comment on the magnitude of this approximation, especially for the Ta2O5 phases where some may be near stability boundaries, would be valuable.","section":null},{"comment":"The abstract states speedups with excessive precision (5.911-9.023x). Rounding to 5.9-9.0x would be more appropriate for the abstract.","section":null},{"comment":"Reference [39] is cited as 'arXiv:2602.03649; Phys. Rev. B (in press).' If now published, the reference should be updated.","section":null}],"recommendation":"minor_revision","confidential_remarks":"The stress-test concern about QHA-informed sparse-point selection (Eq. 14) is valid and is the most important issue to address. The authors should be asked to either provide a proxy-based selection protocol or explicitly qualify the speedup claim. This is a presentation/qualification issue rather than a fundamental flaw, since the method itself is sound and the interpolation math is correct. The Ta2O5 results, while less accurate, are honestly reported and the phase-diagram topology is preserved, which is the practically relevant outcome. The paper is a solid contribution to the computational materials science toolkit and should be publishable after addressing the selection-circularity concern."},"author_rebuttal":null,"desk_editor":{"model":"glm-5.2","letter":"This paper does something useful: it reduces the number of explicit phonon volume points needed for QHA Gibbs free energy calculations from ~20 down to 3, using a ZPE-derived effective Grüneisen parameter for the static branch and piecewise mode-resolved local Grüneisen slopes for the vibrational branch. The core idea is not deep — Grüneisen scaling is a 1912 textbook relation — but the specific assembly into a two-branch reconstruction workflow is new and the validation is thorough. Seven systems, explicit MAE metrics at both ZPE and Gibbs levels, pointwise error maps, phase-diagram comparisons, thermal-expansion recovery, and CPU-time benchmarks. The honest reporting of breakdown at strong compression (Fig. 12) and the large Ta2O5 errors (up to 17.5 meV/atom for the lambda phase) is commendable. For the six simple benchmark systems, sub-0.53 meV/atom accuracy with 6-9x speedup is a real practical advance for high-throughput screening pipelines. The VIP comparison in Appendix B is a nice touch — GI outperforms VIP in both compression-side and thermal-expansion regimes tested. The soft spot is the sparse-volume selection procedure. Eq. 14 uses the full dense QHA V_eq(P,T) path to choose the 3 interpolation volumes. The sensitivity tests vary combinations within this QHA-defined bracket but never test proxy-based selection without phonon information. In real deployment, you would not have the dense QHA reference — you would need cheaper proxies (static EOS, exploratory phonon points) to pick sparse volumes, and the paper does not report accuracy under that protocol. The authors do note that for broad compression benchmarks the procedure reduces to representative low/middle/high points, but the Table I numbers use the QHA-informed selection. This is a partial circularity in the speedup claim, not a fatal flaw — the method still works, but the practical speedup may be smaller than claimed if additional phonon calculations are needed to identify good sparse points. The Ta2O5 errors also show the method can fail silently when soft modes or mode crossings occur within the interpolation interval. This is a solid, incremental methods paper for researchers doing finite-temperature phase-diagram construction who need to cut computational cost. It deserves a serious referee who should push the authors to report accuracy under a QHA-free selection protocol.","headline":"Practical interpolation method that cuts QHA phonon calculations from ~20 to 3 volumes, validated well but with a selection-circularity caveat.","tokens_in":23758,"tokens_out":578,"would_cite":true,"duration_ms":93360,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["64.70.K-","71.15.-m"],"model":"glm-5.2","headline":"Three phonon calculations replace twenty for pressure-temperature phase diagrams","keywords":[],"falsifier":"Find a system where, within the moderate compression range (above 0.8*V0), a phonon mode undergoes a crossing or softening between two sparse volume endpoints such that the linear log-volume interpolation produces a qualitatively wrong frequency at an intermediate volume, leading to a Gibbs free energy error large enough to flip an assigned stable phase. The Ta2O5 lambda phase, with 17.5 meV/atom MAE, already hints that such failures occur for complex polymorphs.","tokens_in":23202,"feed_emoji":"📐","tokens_out":1077,"duration_ms":158851,"temperature":0.7,"pith_summary":"The paper addresses a computational bottleneck in materials science: constructing pressure-temperature phase diagrams requires Gibbs free energies, which in turn require phonon spectra calculated at many volume points (typically 20 or more). The authors show that phonon frequencies vary smoothly enough with the logarithm of volume that spectra at just three sparse volumes can reconstruct the entire free energy surface. They split the problem into two branches. For the zero-point energy contribution, a single effective Grüneisen parameter extracted from ZPE ratios at the sparse volumes scales the static energy branch. For the finite-temperature vibrational contribution, each phonon mode gets its own local Grüneisen slope between adjacent sparse volume pairs, and intermediate frequencies are reconstructed by exponential interpolation in log-volume space. The reconstructed free energies are then minimized over volume at each pressure and temperature point to yield the Gibbs free energy. For six benchmark systems (diamond, Al, Si, Ge, TiO2, PtO2), the mean absolute error relative to full dense-volume QHA stays below 0.53 meV/atom with an average of 0.148 meV/atom, while computational cost drops by a factor of 6 to 9. For structurally complex Ta2O5 polymorphs with up to 77 atoms per cell, the method still reproduces the correct phase-stability topology despite larger per-phase errors. The method breaks down at compressions beyond about 20 percent of the reference volume, where mode crossings and structural transformations invalidate the smooth-interpolation assumption.","feed_headline":"Three phonon calculations replace twenty for pressure-temperature phase diagrams","feed_subtitle":"Sparse-volume Grüneisen interpolation cuts Gibbs free energy cost 6-9x with sub-meV/atom error for moderate compression.","key_machinery":"The method operates by first selecting three stable volumes bracketing the equilibrium volume path traversed during Gibbs minimization. At these sparse volumes, full phonon calculations yield frequencies, ZPE, and static energies. The ZPE-level effective Grüneisen parameter is extracted as a through-origin slope of ln[ZPE(V)/ZPE(V0)] versus ln(V0/V). For each phonon mode j, a local Grüneisen slope is computed between each pair of adjacent sparse volumes. Target-volume frequencies are reconstructed by exponential interpolation: omega_j(V_m) = omega_j(V_a) * exp[gamma_j * (ln(V0/V_m) - ln(V0/V_a))]. Only positive-frequency modes enter the sums; imaginary or zero frequencies are discarded. The ","core_discovery":"The central finding is that the volume dependence of phonon frequencies in crystalline solids is smooth enough in the logarithmic volume coordinate that a piecewise mode-resolved Grüneisen interpolation using only three explicit phonon calculations can reconstruct Gibbs free energies to within sub-meV-per-atom accuracy for moderate compression ranges. The key object is the mode-resolved local Grüneisen slope, defined between two sparse volume endpoints for each phonon mode individually, which captures the mode-specific frequency shift without requiring a single global scaling parameter. Combined with a separate ZPE-level effective Grüneisen parameter for the static branch, this two-tiered重建y","pith_inferences":[],"forward_implications":["High-throughput materials discovery pipelines that currently stop at zero-temperature energy ranking could add finite-temperature and finite-pressure phase stability checks at roughly one-seventh the phonon calculation cost, making thermodynamic validation of AI-generated crystal structures tractable at scale.","The piecewise mode-resolved Grüneisen slopes themselves serve as a compact diagnostic: phases with large extracted gamma values flag strong volume sensitivity of specific modes, identifying candidates where denser sampling or explicit anharmonic treatment may be needed before running full calculations.","The three-point framework extends naturally to more sparse volumes when higher accuracy is required, providing a tunable accuracy-cost tradeoff rather than a fixed-cost method.","For systems where the method fails (strong compression, soft modes, mode crossings), the failure is detectable: the rapid error growth beyond 0.8*V0 shown in Figure 12 provides a quantitative criterion for when to switch to full dense-volume QHA."],"fun_headline_variants":["Three phonon volumes suffice for sub-meV Gibbs free energies under compression","Sparse-volume Grüneisen interpolation rebuilds Gibbs free energies from 3 phonon points","Mode-resolved Grüneisen slopes cut phonon calculations from 20 to 3","Gibbs free energy from 3 phonon volumes: sub-meV/atom error at 6-9x speedup","Piecewise Grüneisen interpolation reconstructs Gibbs free energies with 3 phonon points"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The method assumes that within a moderate compression range, phonon frequencies vary smoothly enough with the logarithm of volume that a locally linear Grüneisen slope between two sparse endpoints can faithfully reconstruct intermediate frequencies for every mode. This breaks down when mode crossings, soft modes, or incipient structural transformations within the interpolation interval change the topology of the phonon spectrum in ways that two endpoint frequencies cannot捕获.","fun_headline_variants_meta":{"raw":{"variants":["Three phonon volumes suffice for sub-meV Gibbs free energies under compression","Sparse-volume Grüneisen interpolation rebuilds Gibbs free energies from 3 phonon points","Mode-resolved Grüneisen slopes cut phonon calculations from 20 to 3","Gibbs free energy from 3 phonon volumes: sub-meV/atom error at 6-9x speedup","Piecewise Grüneisen interpolation reconstructs Gibbs free energies with 3 phonon points","Ab initio Gibbs free energy at 3 volumes matches 20-point quasi-harmonic benchmarks","Three phonon volumes rebuild pressure-temperature phase diagrams at 6-9x speedup","Sparse-volume phonon interpolation reaches sub-meV/atom Gibbs free energy error"]},"model":"glm-5.2","effort":"low","cost_usd":0.0,"raw_usage":{"total_tokens":1329,"prompt_tokens":627,"completion_tokens":702,"prompt_tokens_details":null},"tokens_in":627,"tokens_out":702,"duration_ms":19241,"temperature":1.0,"reasoning_tokens":512,"cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-10T04:31:14.899737+00:00","model_set":{"reader":"glm-5.2"},"falsifier":"Find a system where, within the moderate compression range (above 0.8*V0), a phonon mode undergoes a crossing or softening between two sparse volume endpoints such that the linear log-volume interpolation produces a qualitatively wrong frequency at an intermediate volume, leading to a Gibbs free energy error large enough to flip an assigned stable phase. The Ta2O5 lambda phase, with 17.5 meV/atom MAE, already hints that such failures occur for complex polymorphs.","supporting_citations":[],"review_version":1}