{"id":"317d460d-8411-40f9-9fc9-288f80b5e212","arxiv_id":"2607.18594","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A 12,000-MOF phonon database plus experiments on Ce-UiO-66 variants show that heavy metal nodes and linker bromination tune negative thermal expansion, one variant reaching a record volumetric coefficient.","lead":"Using a machine-learned atomic model, the authors computed vibrations for more than 12,000 metal-organic frameworks and screened them for negative thermal expansion, the unusual property of shrinking when heated. Experiments on cerium-based UiO-66 confirm that heavier metal nodes and brominated linkers tune this contraction, with one variant reaching a record rate.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Record NTE claim for Ce-UiO-66-Br rests on a regime the authors flag as possible structural transition; single-phase Le Bail fits may not measure thermal expansion.","rationale":"The reader's weakest assumption was MACE-MP-MOF0 + QHA accuracy for NTE across diverse MOFs. That is a real concern for the computational screening, but the experimental low-temperature data on Ce-UiO-66-H/Br/2Br already validate the core recipe's monotonic trend (heavier node > Zr; bromination suppresses NTE), so the design recipe has independent support even if the MLIP has biases. The record claim, by contrast, is purely experimental and is undermined by the authors' own caveat that the high-T behavior may be a structural transition. Since the abstract advertises record-breaking coefficients, this is the single most load-bearing soft spot. I therefore recommend keeping the reader's CONDITIONAL verdict, but for a slightly different primary reason; the reader's rationale did mention this issue, so agreement is partial. A two-phase/finer-T PXRD analysis is the concrete check that would settle whether the record claim is real or an artifact.","tokens_in":14936,"tokens_out":3616,"duration_ms":46346,"concrete_test":"Re-analyze the 250-350 K synchrotron PXRD data for Ce-UiO-66-Br with a two-phase Le Bail/Rietveld model (e.g., parent fcu phase plus a possible high-T phase) and/or variable-temperature total scattering / pair distribution function analysis to detect an amorphous or second phase. Also collect finer temperature steps (e.g., every 10 K) across 250-350 K and check for a volume discontinuity or hysteresis. If the two-phase model yields stable cell parameters for one phase with the same contraction, the record claim survives; if the apparent contraction is a phase-mixture artifact, the abstract must be tempered.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The abstract's headline 'volumetric NTE coefficients surpassing current records' rests entirely on alpha_V = -593(25) (M K)-1 reported for Ce-UiO-66-Br above 250 K. Section 4 states this regime 'falls outside the QHA regime' and coincides with 'peak broadening and intensity loss... possibly due to a structural transition between 250 K and 350 K' (Fig. 3c,e). If a structural transition or phase mixture develops, the lattice parameter extracted from a single-phase Le Bail fit of a two-phase/broadened pattern is not a well-defined thermal expansion quantity; the finite-difference alpha_V across 250-350 K would then mix cell constants from different phases (or fit artifacts), producing an apparent 'colossal' contraction that is not the volumetric NTE of a single phase. The authors themselves flag this possibility yet retain the record claim in the abstract. This is load-bearing because the record is a prominent central claim and is not protected by the otherwise solid low-temperature validation of the design recipe.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a high-throughput workflow using a MOF-tuned machine-learned potential (MACE-MP-MOF0) to compute phonon properties for roughly 12,000 QMOF structures, forming the PhononMOFdb resource. Screening based on low-frequency phonon DOS together with quasi-harmonic approximation (QHA) yields 178 mechanically stable volumetric negative thermal expansion (NTE) candidates. Analysis of descriptors (crystal system, topology, metal node, linker length/functionalization) leads to a proposed design recipe: cubic topologies, heavier and/or lower-valent metal nodes, and linker functionalization as a tuning handle. The recipe is experimentally tested by synchrotron powder X-ray diffraction on Ce-UiO-66-H, -Br, and -2Br; the data confirm that bromination suppresses NTE below 250 K. The paper further reports a very large alpha_V = -593(25) (M K)^{-1} for Ce-UiO-66-Br above 250 K and claims this as a record volumetric NTE coefficient for MOFs.","tokens_in":15092,"tokens_out":5240,"duration_ms":60994,"significance":"If the design rules are valid, the work constitutes a substantial advance: it provides a large open phonon database for MOFs, a transferable MLIP-based screening strategy, and an experimental demonstration that linker functionalization can systematically tune NTE. Strengths include the release of the database and workflow, the INS benchmarking against DFT and experiment, and the internal consistency of the low-temperature experimental trends with the computational predictions. The independent experimental check on three Ce-UiO-66 variants below 250 K is a genuine step beyond purely computational design rules. However, the record NTE claim and the generality of the screening-based recipe rest on assumptions that are not fully validated in the present manuscript; these issues are load-bearing for the central claims.","major_comments":[{"comment":"The headline claim of record volumetric NTE, alpha_V = -593(25) (M K)^-1 for Ce-UiO-66-Br above 250 K, is not supported by the experimental analysis as presented. The authors state that this regime 'falls outside the QHA regime' and coincides with 'peak broadening and intensity loss... possibly due to a structural transition between 250 K and 350 K' (Fig. 3c,e). Under a possible structural transition or phase mixture, the lattice parameter extracted from single-phase Le Bail refinement is not a well-defined single-phase lattice parameter, and a finite-difference alpha_V across 250-350 K mixes cell constants (or fit artifacts) from different phases. Since the abstract's 'surpassing current records' rests entirely on this number, either remove or appropriately qualify the record claim, or provide evidence for a single-phase assignment (e.g., two-phase or Pawley refinement, analysis of peak","section":"Section 4 / Abstract"},{"comment":"The screening filter and the ensuing design recipe depend on the accuracy of MACE-MP-MOF0 + QHA for NTE across diverse QMOF structures, but the validation in the main text is limited to DOS features for MOF-5, UiO-66, and MOF-74 and to QHA trends for 'several well-characterized MOFs' (Fig. S3). NTE is governed by the volume dependence of phonon frequencies (Gruneisen parameters), a more sensitive quantity than the 0-2 THz DOS; the INS benchmarks against MOF-5 and ZIF-8 validate spectral positions, not phonon volume derivatives. Since the 178-candidate pool and the metal/linker/length trends, including the choice of Ce-UiO-66, are obtained from this MLIP, a systematic comparison of computed alpha_V against DFT or experimental data for a more diverse set (especially non-cubic and heavy-metal frameworks) is needed to support the claim of a general, data-driven recipe.","section":"Section 3.1, Fig. S3"},{"comment":"The design recipe instructs readers to 'favor cubic crystal systems for larger NTE' (Section 3.2), but Section 3.1.1 and Fig. 2a report that the largest NTE magnitudes in the screened pool are predominantly triclinic and monoclinic, with only 0.5% cubic candidates, and conclude that crystal system alone is insufficient. If the recipe is instead intended to favor cubic systems for isotropic contraction or mechanical stability, that rationale is not stated. This internal contradiction should be resolved; otherwise the general recipe is ambiguous.","section":"Section 3.1.1 vs Section 3.2"}],"minor_comments":[{"comment":"The text cites 'Figure 2c' for the abrupt transition to colossal NTE; the correct reference is Figure 3c.","section":"Section 4"},{"comment":"The unit is written as '(M K)^{-1}'; if this means 10^-6 K^-1, please use conventional notation (e.g., MK^-1, ppm K^-1, or 10^-6 K^-1) consistently.","section":"Throughout"},{"comment":"Data Availability: 'The database and codes will are available' should be corrected to 'will be available' or 'are available'.","section":"Section 6"},{"comment":"The phrase 'heavier, lower-valent metal nodes' conflates mass and valency; Section 3.1.2 treats them as independent handles (lower charge and heavier mass both soften modes). Please rephrase to avoid an apparent contradiction with the tetravalent Ce/Hf examples.","section":"Abstract / Section 3.1.2"},{"comment":"The colossal-NTE threshold |alpha_V| >= 100 is not defined with units in the main text; please specify the units (presumably 10^-6 K^-1) and state the criterion explicitly.","section":"Section 3.1.1"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is ambitious and the low-temperature experimental data on Ce-UiO-66 are a genuine validation of the design trends. The main weaknesses are the overstatement of the record NTE claim based on a possibly multi-phase/highly anharmonic regime, the limited validation of the MLIP's Gruneisen parameters over diverse MOF space, and an internal inconsistency in the design recipe regarding crystal systems. These are fixable within the manuscript's scope by reframing claims and adding targeted validation, but they are load-bearing and currently prevent acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth reading for the database alone. PhononMOFdb, with ~12k MOFs and their phonon DOS, INS spectra, bulk moduli, and heat capacities, is a real resource. The validation of their MLIP against experimental INS for MOF-5 and ZIF-8 is good practice, and the screening protocol—low-frequency DOS filter, QHA, bulk-modulus cutoff—is sensible. The design recipe (heavy nodes, cubic topology, linker functionalization to tune NTE) is consistent with known mechanisms and gets genuine support from the low-temperature PXRD on Ce-UiO-66 variants. The cleanest result is Ce-UiO-66-H showing stronger NTE than Zr-UiO-66 at 250 K, exactly as the heavy-node argument predicts. That part is solid.\n\nThe soft spot is the headline. The -593(25) MK^-1 for Ce-UiO-66-Br above 250 K is not a reliable expansion coefficient. The paper itself says that regime “falls outside the QHA regime” and coincides with peak broadening and intensity loss “possibly due to a structural transition.” If a structural transition or phase mixture is present, a single-phase Le Bail fit does not give a thermal expansion coefficient—you could be mixing cell constants from different phases or chasing fit artifacts. The abstract presents this as a record without the caveat, which is misleading. Either they need two-phase refinement or variable-temperature total scattering to back the claim, or the abstract should say the low-temperature trend is the validated result and the high-temperature behavior is unresolved.\n\nSecond, the computational alpha_V values driving the screening are not systematically benchmarked. The INS/DOS comparisons for MOF-5, UiO-66, and MOF-74 are encouraging, but NTE depends on Grüneisen parameters—volume derivatives of phonon frequencies—which are far more sensitive than a DOS match. Validating on three MOFs is thin. The reliance on the authors’ own MLIP is not itself a problem; the circularity charge doesn’t stick because the design rules emerge from phonon screening, not from fitting NTE. But an independent DFT or experimental benchmark on a diverse set would substantially strengthen the quantitative claims.\n\nMinor point: the screening pool inherits QMOF’s skew toward triclinic/monoclinic frameworks, so the “colossal NTE in non-cubic space” observation may partly reflect database composition. The authors acknowledge this.\n\nI couldn’t check the SI, so some of this rests on the authors’ description. Overall, this is a serious paper with a valuable resource and a plausible recipe, but the record claim needs either data or a retraction. Deserves peer review; a careful referee should push on the high-temperature analysis and ask for broader alpha_V validation.","headline":"Valuable 12k-MOF phonon database and a plausible design recipe, but the -593 MK^-1 record claim rests on a phase transition the authors themselves flag.","tokens_in":15697,"tokens_out":2526,"would_cite":true,"duration_ms":32705,"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":"The paper argues that negative thermal expansion in metal-organic frameworks can be designed rather than stumbled upon, and demonstrates the recipe with a record-setting measurement.","keywords":["negative thermal expansion","metal-organic frameworks","phonons","machine learning interatomic potential","high-throughput screening","linker functionalization","quasi-harmonic approximation","UiO-66"],"falsifier":"Compute the low-frequency Grüneisen parameters from density-functional-theory phonons for a sample of the screened candidates (especially triclinic and heavy-metal structures) and compare with the database: mismatches would invalidate the screen. Alternatively, an independent high-resolution diffraction or single-crystal study of the mono-brominated cerium-based UiO-66 variant between 250 and 350 K would test whether the −593(25) × 10⁻⁶ K⁻¹ value is a genuine bulk phase transition or an artifact of peak broadening and Le Bail fitting.","tokens_in":14762,"feed_emoji":"🔥","tokens_out":6071,"duration_ms":68215,"temperature":0.7,"pith_summary":"Metal–organic frameworks (MOFs) are porous crystals built from metal nodes and organic linkers, and many of them shrink when heated — negative thermal expansion. The paper tries to turn this phenomenon into a design rule. It computes phonon spectra for more than 12,000 MOFs using a machine-learned interatomic potential, screens for intense low-frequency vibrations, and extracts three guidelines: cubic, highly porous frameworks; heavier, lower-valent metal nodes; and linker functionalization as a continuous dial for the expansion coefficient. The rules are tested on a cerium-based UiO-66 framework and two brominated variants: bromine suppresses contraction below 250 K, and one brominated form then undergoes an abrupt transition to a massive contraction of −593(25) × 10⁻⁶ K⁻¹ above 250 K, the largest MOF volumetric negative thermal expansion the authors report on record. If the computational screen is trustworthy, the same recipe can be applied across a broad MOF chemical space to design materials with a target thermal expansion.","feed_headline":"Heated MOF sets a record for shrinking","feed_subtitle":"Data-driven screening plus lab tests yield design rules for tunable negative thermal expansion.","key_machinery":"The argument is carried by a phonon-centric screening workflow. A machine-learned interatomic potential fitted to MOF energetics is used within the quasi-harmonic approximation to compute harmonic phonons, phonon densities of states, heat capacities, and bulk moduli for roughly 12,000 MOFs. The key proxy is the low-frequency (0–2 THz) density of states: modes in this window include the transverse linker vibrations and rigid-unit modes whose Grüneisen parameters (the volume sensitivity of each mode's frequency) are negative; a high density of such modes is taken as a fingerprint for negative thermal expansion. This proxy turns an otherwise prohibitive first-principles phonon calculation into","core_discovery":"The central claim is that the magnitude and sign of volumetric thermal expansion in MOFs can be predictably engineered by choosing topology, metal node, and linker chemistry. Analyzing a new database of phonon-derived properties for over 12,000 MOFs, the paper finds that frameworks with cubic topologies, high porosity, and heavier, lower-valent metal nodes concentrate their low-frequency vibrations — transverse linker modes and rigid-unit rotations — into modes with strongly negative Grüneisen parameters, producing large negative thermal expansion. Linker functionalization, such as adding bromine to the BDC linker, progressively suppresses these modes and can flip negative expansion to posit","pith_inferences":["The 0–2 THz density-of-states proxy, if it holds up, could be exported to any phonon-producing machine-learned potential, letting other groups screen their own MOF libraries without re-running the full quasi-harmonic workflow.","The high-temperature colossal contraction in the brominated variant sits outside the quasi-harmonic approximation; anharmonicity or a structural phase transition is probably at work, and that regime may be a separate, designable phenomenon.","The mass–charge argument suggests mixed-metal-node or isotope studies could directly test whether heavier nodes always enhance negative thermal expansion.","The screening's reliance on one machine-learned potential is the evident risk: a density-functional-theory re-check of low-frequency Grüneisen parameters for even ten of the candidates would show whether the trends are robust."],"forward_implications":["If the recipe holds, engineers can choose an expansion coefficient for a MOF — negative, near-zero, or positive — by picking a topology and adding substituents, without redesigning the framework class.","Heavy-node substitution (cerium or hafnium instead of zirconium) should enhance negative thermal expansion in other fcu frameworks, not just UiO-66.","Linker elongation boosts negative thermal expansion but softens the framework; the 1 GPa bulk-modulus cutoff marks a practical stability boundary for screening.","More than ten colossal-negative-thermal-expansion candidates (|αV| ≥ 100 × 10⁻⁶ K⁻¹) are flagged for experimental follow-up.","The database extends to anisotropic, non-cubic MOFs, where volumetric negative thermal expansion is less studied than in cubic isoreticular families."],"fun_headline_variants":["Data-driven MOF design achieves record negative thermal expansion","ML-screened MOFs yield design rules for tunable thermal shrinkage","12,000 MOFs screened to crack negative thermal expansion code","Brominated MOF sets record for thermal contraction on heating","Linker tweaks flip MOF thermal expansion from negative to positive"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that the machine-learned potential's low-frequency phonons — the volume-dependent mode softening that actually produces negative thermal expansion — are accurate across the diverse chemical space of the database; if those modes are biased for triclinic or heavy-metal frameworks, the candidate pool and the record material could be artifacts.","fun_headline_variants_meta":{"raw":{"variants":["Data-driven MOF design achieves record negative thermal expansion","ML-screened MOFs yield design rules for tunable thermal shrinkage","12,000 MOFs screened to crack negative thermal expansion code","Brominated MOF sets record for thermal contraction on heating","Linker tweaks flip MOF thermal expansion from negative to positive"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00021,"raw_usage":{"total_tokens":1263,"prompt_tokens":775,"completion_tokens":488,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":519,"completion_tokens_details":{"reasoning_tokens":411}},"tokens_in":519,"tokens_out":488,"duration_ms":6198,"temperature":1.0,"reasoning_tokens":411,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T14:56:43.019333+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the low-frequency Grüneisen parameters from density-functional-theory phonons for a sample of the screened candidates (especially triclinic and heavy-metal structures) and compare with the database: mismatches would invalidate the screen. Alternatively, an independent high-resolution diffraction or single-crystal study of the mono-brominated cerium-based UiO-66 variant between 250 and 350 K would test whether the −593(25) × 10⁻⁶ K⁻¹ value is a genuine bulk phase transition or an artifact of peak broadening and Le Bail fitting.","supporting_citations":[],"review_version":1}