{"id":"e4d6fa94-b7e0-4d3b-96ac-f2c9e5f24f73","arxiv_id":"2501.18372","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Including the many low-lying local minima of a molecular crystal shifts the predicted Form III to Form II transition temperature from 470 K to 270 K, showing configurational entropy cannot be ignored.","lead":"This paper counts many nearly identical crystal arrangements, not just the single lowest energy one, when computing which crystal form is stable at a given temperature. For one molecular crystal, this changes the predicted transition between two forms by about 200 K.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 200 K shift rests on a harmonic superposition of shallow basins separated by 10–100 meV barriers, comparable to kBT at the claimed 270 K crossing, which strains the paper's own rare-event validity condition.","rationale":"The reader's weakest assumption focuses on truncation of the sampled density of states; that is a genuine limitation, but its likely effect is to make the true crossing even lower, so it does not threaten the qualitative conclusion. A more decisive threat is the internal validity condition of the method itself: Eq. 5 assumes rare, harmonic basins, whereas the quoted barriers (10–100 meV) and hop times (~10 ps) at the relevant temperature (kBT ~ 23 meV) are borderline. If the harmonic-superposition sum overcounts shallow basins, the 470-to-270 K shift could be partly a methodological artifact rather than a physical configurational-entropy effect. The proposed metadynamics/umbrella-sampling comparison is a concrete check that would settle this. The paper does provide useful qualitative support for the multi-minima picture via DFT-level barriers, MD observation of Form III to Form II conversion, and convergence between 100 and 300 sampled structures; those support the general claim but do not pin down the magnitude. I therefore keep the reader's CONDITIONAL verdict unchanged, but would add the harmonic-superposition validation as an explicit condition before relying on the numerical 200 K estimate.","tokens_in":9156,"tokens_out":11236,"duration_ms":123928,"concrete_test":"Compute the Form II–III free-energy difference at 270 K with the same NequIP potential using a method that does not impose the basin decomposition, e.g., well-tempered metadynamics or umbrella sampling on the two collective variables of Eqs. 1–4, and compare directly with the delta-F implied by the Eq. 6 probability ratio at 270 K. If the two free-energy differences agree within about kBT/2 (~12 meV per molecule), the harmonic superposition treatment is validated; if they differ by more than that, the 200 K shift is not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The quantitative claim (Fig. 5: crossing moves from 470 K to 270 K when all sampled minima are included) rests on the quantum-superposition partition function, Eq. 5, with harmonic free energies F_j for each catchment basin. The paper states the method is valid only if transitions between neighboring basins are rare events, but its own MD and barrier data show Form I/II basins separated by 10–100 meV barriers and inter-basin hops on a ~10 ps timescale. At the claimed crossing temperature T = 270 K, kBT is about 23 meV, so the shallow basins are not clearly rare-event harmonic wells: the harmonic superposition sum can double-count or overcount entropy that a proper anharmonic free-energy calculation would already contain. The density-of-states truncation acknowledged in the Fig. 4 caption is a secondary uncertainty that would likely increase the effect; the harmonic/rare-event violation could act in the opposite direction and make the 200 K shift an artifact of the basin decomposition. This is the least secure link between the presented data and the headline error estimate.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper argues that the multi-minima character of molecular crystal potential energy surfaces contributes a significant configurational entropy that must be included in finite-temperature free-energy rankings. For N-(4'-methylbenzylidene)-4-methylalanine, the authors use minima hopping at the DFT level to sample 304 local minima, classify them into three experimentally known forms using collective variables, and compute harmonic free energies. They then apply the quantum superposition method (Eq. 5) to sum Boltzmann weights over all minima and obtain phase probabilities as a function of temperature. The central claim is that the Form II–III transition temperature moves from about 470 K when only the lowest minimum of each phase is used to about 270 K when all 304 minima are included, implying a roughly 200 K error if configurational entropy is neglected. The paper also reports exact saddle-point barriers along transformation pathways and MD-based estimates of inter-basin hopping times.","tokens_in":9341,"tokens_out":5438,"duration_ms":53314,"significance":"If correct, the result is important for crystal structure prediction: it would demonstrate that configurational entropy from multiple local minima can change polymorphic free-energy rankings by hundreds of kelvin, an effect currently ignored in standard CSP protocols. The study is valuable for its DFT-level minima-hopping sampling, exact saddle-point calculations, and the explicit connection to a crystal with a known disappearing-polymorphism problem. The methodology is standard but cleanly applied, and the observation that different forms have markedly different configurational densities of states is an interesting qualitative finding. However, the quantitative claim lacks uncertainty estimates and its central approximation is applied in a regime where the paper's own data indicate it may not be valid.","major_comments":[{"comment":"The paper states that the superposition method is valid only if transitions between neighboring catchment basins are rare events, but the data in Table I and Fig. 6 contradict this condition at the temperatures of interest. Table I shows inter-basin hops on a timescale of about 10 ps, while vibrational periods are reported to extend to about 1 ps; the 10–100 meV barriers in Fig. 6 are comparable to kBT ≈ 23 meV at the claimed 270 K crossing. Under these conditions, the harmonic superposition sum in Eq. (5) can double-count anharmonic contributions that a proper anharmonic free-energy calculation would already include, and the 200 K shift in Fig. 5 may be an artifact of the basin decomposition. I recommend that the authors demonstrate the validity of the rare-event harmonic description at the crossing temperature, for example by comparing with an anharmonic free-energy calculation for the lowest-energy minima or by showing that barriers are large compared to kBT over the relevant temperature range.","section":"Eq. (5) and text after it"},{"comment":"The configurational density of states shown in Fig. 4 is truncated to zero above the sampled energy range, as explicitly acknowledged in the caption. The convergence test between 100 and 300 structures (290 K vs 270 K) only varies the number of sampled minima and does not bound the contribution of unsampled higher-energy minima. Since the partition function in Eq. (5) is sensitive to the high-energy tail of the DOS, the reported 'about 200 K' error has no estimated uncertainty and may be an underestimate. The authors should either extrapolate the DOS, estimate the missing contribution, or provide a quantitative bound on the error introduced by the truncation.","section":"Fig. 4 and Eq. (5)"},{"comment":"The NequIP machine-learned potential is used to compute vibrational frequencies and to run longer MD, but the manuscript presents no validation of this potential against DFT (e.g., energy/force errors on a held-out test set). Because the harmonic free energies and the number of distinct minima visited depend directly on the accuracy of this potential, the quantitative claim currently lacks support. Please add a validation of the NequIP potential and an estimate of how the remaining DFT/ML discrepancy propagates into the computed transition temperatures.","section":"Machine-learned potential section"},{"comment":"The headline result of a 200 K error is presented as a single number without any uncertainty or sensitivity analysis. The 470 K and 270 K crossing temperatures are computed with specific choices of functional (PBE+D4), harmonic approximation, and a finite sample of 304 minima, but no error bar is given. At minimum, the authors should provide a crude estimate of the uncertainty based on, for example, variations of the DOS truncation or the ML potential accuracy, so that the central number can be meaningfully interpreted.","section":"Fig. 5"}],"minor_comments":[{"comment":"There is a notation inconsistency: the text describes a vector '®𝑤' spanned between the centers of mass of the two carbon rings, but Eq. (1) then uses '®𝑣' in the angle definition; please unify the notation.","section":"Eq. (1) and surrounding text"},{"comment":"The caption contains a typo: 'The ration' should be 'The ratio'. Please correct.","section":"Table I caption"},{"comment":"The molecule name appears inconsistently as 'N-(4-Methylbenzylidene)-4-methylalanine' in the abstract and 'N-(4’-Methylbenzylidene)-4-methylalanine' in the body; please use one name consistently.","section":"Title of molecule"},{"comment":"The manuscript does not specify how the harmonic free energies in Eq. (5) are computed, for example whether the vibrational frequencies are obtained at the Γ point only or with a k-point mesh, and whether corrections for imaginary modes are applied. This information is needed to reproduce the results.","section":"Methods details"},{"comment":"The caption has a typo: 'Methylkbenzylidene' should be 'Methylbenzylidene'.","section":"Fig. 1 caption"}],"recommendation":"major_revision","confidential_remarks":"This paper addresses an important and timely issue in crystal structure prediction, and the computational effort invested in DFT-level minima hopping and exact saddle-point calculations is substantial. However, the central quantitative claim rests on the validity of the harmonic superposition approximation at temperatures where the paper's own barrier data suggest the rare-event condition is not satisfied. This is fixable in revision by adding a validation test (e.g., anharmonic free energies for the lowest basins) or by tempering the headline claim. The absence of uncertainty estimates and the unvalidated ML potential are further issues that should be addressed. I recommend major revision rather than rejection, because the manuscript's core idea is sound and the data are valuable, but the 200 K number is not yet established."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe headline: this paper gives a concrete number for a real effect—configurational entropy from multiple minima shifts a polymorph transition temperature by about 200 K in a molecular crystal. The qualitative idea has been around (Dybeck, Butler, Francia), but the quantitative estimate for N-(4-methylbenzylidene)-4-methylalanine is new. If it holds up, it strengthens the case that finite-temperature CSP must include multi-minima contributions.\n\nWhat's good: the workflow is logical. Minima hopping on DFT, re-relaxation with PBE-D4, a NequIP potential for frequencies and MD, classification by collective variables, and a quantum-superposition partition function. The density of states per form (Fig. 4) makes the entropic argument visually clear. The convergence check with 100 vs 300 minima (290 vs 270 K) is a reasonable attempt. Exact saddle points are a nice addition.\n\nThe soft spots are real, though. The biggest one is the validity of the harmonic-superposition sum at the claimed crossing. The paper's own condition is that transitions between basins are rare events. But their barriers between Form I/II minima are 10–100 meV per molecule, and kBT at 270 K is about 23 meV. Their MD shows hops every few picoseconds across many distinct minima (Table I). That does not look like a rare-event regime. The harmonic partition function may be overcounting entropy that an anharmonic treatment would handle correctly, and the 200 K could be partly an artifact of the basin decomposition. The stress-test note is on target.\n\nSecond, the density of states is truncated at high energies—acknowledged in the Fig. 4 caption. The 100-vs-300 convergence is not a bound on unsampled minima. The direction is probably to increase the effect, but it's unquantified. Third, no error bars on the 200 K. Fourth, the NequIP potential is not validated against DFT for the specific observables (frequencies, MD barriers), so the whole free-energy calculation rests on an unbenchmarked model. Fifth, no code or data.\n\nNone of this kills the qualitative conclusion. The tolerant forms have more minima and lower frequencies, so they should be stabilized at high T. But the exact 200 K is not yet a reliable number. The paper deserves a serious referee; it's a methodologically important demonstration. The referee should ask for ML-potential validation, a check of the rare-event assumption (maybe via anharmonic free energies or longer MD), uncertainty quantification, and, ideally, release of the minima set and code.","headline":"A credible, important demonstration that configurational entropy can shift polymorph transition temperatures by ~200 K, but the central number needs stronger validation and a check of the harmonic-basin assumption.","tokens_in":9887,"tokens_out":3061,"would_cite":true,"duration_ms":30516,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Counting every crystal minimum shifts a phase transition by 200 K","keywords":["molecular crystals","polymorphism","configurational entropy","crystal structure prediction","free energy","quantum superposition method","potential energy landscape","solid-solid phase transition"],"falsifier":"Compute the density of states of each form more exhaustively, for example by enumerating all minima within 500 meV of the global minimum using longer enhanced-sampling dynamics, and recompute the Boltzmann crossing temperature; if the crossing stays at about 270 K, the 200 K shift is robust, whereas if it moves back toward 470 K the truncation of unsampled minima was the deciding factor.","tokens_in":8957,"feed_emoji":"💎","tokens_out":5761,"duration_ms":51872,"temperature":0.7,"pith_summary":"The paper sets out to show that a polymorph of a molecular crystal is not a single structure but an ensemble of many local minima on the potential energy surface, and that the configurational entropy of this ensemble changes which phase is thermodynamically stable at finite temperature. Using the molecule N-(4-methylbenzylidene)-4-methylalanine, the authors find 304 distinct minima belonging to its three known forms, with the dense, low-energy Form III having only 84 minima while the looser Forms I and II have many more. Combining these minima through the quantum superposition method, the Boltzmann crossing between Form II and Form III moves from roughly 470 K to about 270 K when all sampled minima are included instead of just the lowest-energy structure of each phase. The authors conclude that neglecting configurational entropy introduces errors of about 200 K even in a relatively simple molecular crystal, and that finite-temperature crystal structure prediction must therefore include it.","feed_headline":"Counting every crystal minimum shifts a phase transition by 200 K","feed_subtitle":"A dense, low-energy polymorph loses to a looser one once configurational entropy is included.","key_machinery":"The load-bearing object is the quantum superposition method, which builds the total partition function as a sum over all sampled local minima: $Z = \\sum_j \\exp(-\\beta F_j(\\beta))$, where $F_j(\\beta)$ is the harmonic free energy of minimum $j$. This turns each phase into a weighted set of minima rather than a single structure. The machinery also includes: Minima Hopping at DFT level to find the 304 minima; a NequIP machine-learned potential fitted to DFT data, which makes molecular dynamics and vibrational frequencies affordable; two collective variables (angles between ring-link vectors and ring normal vectors) that classify the minima into the three experimental forms; and a configurational density of states, shown in Fig. 4, which quantifies how many minima each form has in each energy window. The method is valid when the system can visit all low-energy basins on experimental timescales but crossings between basins are rare events; the paper verifies this with MD timings.","core_discovery":"The central claim is that the multi-minima character of the phases is large enough to reverse free-energy rankings. For N-(4-methylbenzylidene)-4-methylalanine, Form III is the global minimum in potential energy with the highest density, but it is structurally rigid: rotating methyl groups or sliding layers is costly, so it has few nearby low-energy minima. Form II and Form I are structurally tolerant, generating 157 and 63 minima respectively at modest energy costs, and this difference in configurational density of states outweighs their higher potential energies at elevated temperature. Using the quantum superposition partition function, the probability of finding Form II overtakes Form III at about 270 K when all 304 minima are counted, whereas using only the lowest structures per phase puts the crossing at about 470 K. The paper argues that the entropy of the multi-minima ensemble is essential for accurate finite-temperature free energies and for explaining why 'lowest energy' ranking fails to predict the experimentally relevant phase.","pith_inferences":["This suggests 'over-prediction' in crystal structure prediction is partly a thermodynamic statement: many predicted minima are genuinely occupied at finite temperature, and what matters is the free energy of the ensemble, not the depth of one basin.","If correct, the rank-ordering of polymorphs by computed lattice energies should be accompanied by a measure of landscape 'tolerance' (e.g., density of low-energy minima) as a screening descriptor.","The same reasoning may apply to the phenomenon of disappearing polymorphism: a phase that is kinetically accessible but entropically disfavored could vanish from recrystallization even if it is the global energy minimum.","A testable extension is to run the same quantum-superposition analysis on other molecules with known polymorph pairs, checking whether the 200 K-scale shift generalizes."],"forward_implications":["Polymorph rankings based on the lowest-energy structure alone can be wrong at finite temperature; the correct ranking must sum over the minima of each phase.","Transition temperatures between polymorphs can be off by about 200 K when configurational entropy is neglected, so crystal structure prediction benchmarks that ignore it inherit this error.","Structurally tolerant, less dense phases are stabilized by configurational entropy at high temperature, acting together with their larger vibrational entropy.","The quantum superposition method, combined with a machine-learned potential, is a practical route to finite-temperature free energies for molecular crystals.","For more complex molecular crystals, where the multi-minima character is presumably more pronounced, the effect on phase stability should be even larger."],"supporting_citations":[{"why":"Supplies the quantum superposition partition function (Eq. 5) that sums harmonic free energies of all minima.","marker":"[30]"},{"why":"MD evidence that molecular crystal trajectories visit catchment basins of many local minima.","marker":"[17]"},{"why":"Threshold algorithm results showing barriers between minima are small enough to be crossed at room temperature.","marker":"[19]"},{"why":"Minima hopping method that generated the 304 local minima for the molecule.","marker":"[34]"},{"why":"ASE implementation of minima hopping used for variable-cell structure prediction.","marker":"[36]"},{"why":"NequIP machine-learned potential fitted to DFT data, enabling MD and vibrational frequency calculations.","marker":"[46]"},{"why":"Compass saddle-point search used to compute exact barriers along transformation pathways.","marker":"[51]"},{"why":"Overlap matrix fingerprint used to distinguish and classify the minima into the three forms.","marker":"[52]"}],"fun_headline_variants":["Crystal's many minima shift phase transition by 200 K","Ignoring multi-minima entropy causes 200 K phase error","Multi-minima free energy flips crystal phase at 270 K","Hidden minima in crystal phases change stability by 200 K","All 304 minima count: phase ranking shifts 200 K"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The 304 minima found by the search, and the assumption that no unsampled higher-energy minima contribute, faithfully represent the phase space of the three forms at the temperatures of interest.","fun_headline_variants_meta":{"raw":{"variants":["Crystal's many minima shift phase transition by 200 K","Ignoring multi-minima entropy causes 200 K phase error","Multi-minima free energy flips crystal phase at 270 K","Hidden minima in crystal phases change stability by 200 K","All 304 minima count: phase ranking shifts 200 K"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000416,"raw_usage":{"total_tokens":2151,"prompt_tokens":955,"completion_tokens":1196,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":571,"completion_tokens_details":{"reasoning_tokens":1110}},"tokens_in":571,"tokens_out":1196,"duration_ms":10320,"temperature":1.0,"reasoning_tokens":1110,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T23:43:19.924937+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the density of states of each form more exhaustively, for example by enumerating all minima within 500 meV of the global minimum using longer enhanced-sampling dynamics, and recompute the Boltzmann crossing temperature; if the crossing stays at about 270 K, the 200 K shift is robust, whereas if it moves back toward 470 K the truncation of unsampled minima was the deciding factor.","supporting_citations":[{"cited_title":"Wales, Energy Landscapes: Applications to Clusters, Biomolecules and Glasses, 1st ed","cited_arxiv_id":null,"evidence_quote":"Supplies the quantum superposition partition function (Eq. 5) that sums harmonic free energies of all minima."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"MD evidence that molecular crystal trajectories visit catchment basins of many local minima."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Threshold algorithm results showing barriers between minima are small enough to be crossed at room temperature."},{"cited_title":"Goedecker, Minima hopping: An efficient search method for the global minimum of the potential energy surface of complex molecular systems, J","cited_arxiv_id":null,"evidence_quote":"Minima hopping method that generated the 304 local minima for the molecule."},{"cited_title":"Krummenacher, M","cited_arxiv_id":null,"evidence_quote":"ASE implementation of minima hopping used for variable-cell structure prediction."},{"cited_title":"Batzner, A","cited_arxiv_id":null,"evidence_quote":"NequIP machine-learned potential fitted to DFT data, enabling MD and vibrational frequency calculations."},{"cited_title":"Sommer-Jörgensen and S","cited_arxiv_id":null,"evidence_quote":"Compass saddle-point search used to compute exact barriers along transformation pathways."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Overlap matrix fingerprint used to distinguish and classify the minima into the three forms."}],"review_version":1}