{"id":"1b0045c0-0a2b-4139-8a27-e5ec5ba4cba4","arxiv_id":"2412.13750","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A new machine-learned potential for WTaCrV shows that experimentally observed CrV precipitates form semicoherent bcc-to-bcc interfaces with the matrix because coherent ones are destabilized by large lattice mismatch.","lead":"This paper builds a machine-learned atomistic model for a tungsten-based alloy and simulates how its atoms reorder and segregate around defects, comparing with atom probe experiments. It explains why the experimentally observed chromium-vanadium precipitates must form semicoherent interfaces (with a network of dislocations) rather than perfectly coherent ones.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Fig. 14 crossover rests on tabGAP-only semicoherent interface energies; a DFT cross-check of misfit dislocation energies could shift the crossover beyond observed precipitate sizes.","rationale":"The reader's weakest assumption identifies the same load-bearing concern: the semicoherent interfacial energies are computed only with tabGAP and not cross-checked against DFT, and the model systems are ideal equiatomic B2 binaries. My stress-test analysis confirms this is the central quantitative weakness of the strongest claim. The crossover in Fig. 14 is the linchpin: if the semicoherent branch is inaccurate, the conclusion that the experimentally observed CrV-rich precipitates must be semicoherent loses its thermodynamic basis. The sensitivity estimate (3*v*sigma/r) shows that modest errors in sigma can shift the crossover by the size scale of the observed precipitates, so the absence of any DFT validation for dislocation-bearing interfaces is a real gap. The paper is otherwise strong: the tabGAP is extensively validated for bulk, defects, surfaces, and grain boundaries, the DFT comparison for coherent interface properties is good, and the authors transparently acknowledge the tabGAP-only and equiatomic limitations. These honesty markers do not repair the gap but they do make the conditional nature of the conclusion explicit. A targeted DFT dislocation-dipole calculation would directly test the semicoherent sigma and could either support or shift the crossover. Since the reader already reached CONDITIONAL with the same reasoning, my read does not change the verdict; it reinforces it.","tokens_in":26952,"tokens_out":10481,"duration_ms":94351,"concrete_test":"Use DFT to compute the energy of a 1/2<111> edge dislocation dipole in bulk B2 CrV (and, for reference, B2 TaW), extract the dislocation core energy per unit length, and combine it with the already-computed coherent interfacial energy to obtain a DFT-based semicoherent interfacial energy for CrV/TaW. Compare this with the tabGAP values used in Fig. 14 (0.05-0.15 eV/A^2 for (100), 0.057 for (110), 0.096 for (111)). If the DFT-based sigma differs by more than about 0.03 eV/A^2, recompute the Fig. 14 crossover curves; if the crossover with coherent CrTa/VW moves above roughly 5 nm in radius, the claim that the observed 2-5 nm precipitates must be semicoherent is not quantitatively supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The conclusion that CrV-rich precipitates in WTaCrV must be semicoherent depends on the crossover in Fig. 14, where the semicoherent CrV/TaW branch is computed using interfacial energies from tabGAP only, as the paper states: \"Semicoherent interfacial energies are always from tabGAP.\" The tabGAP is validated against DFT for bulk lattices, point defects, surfaces, and grain boundaries, but not for the 1/2<111> misfit dislocation networks that define semicoherent interfaces. The semicoherent sigma values (0.05-0.15 eV/A^2 for (100), 0.057 for (110), 0.096 for (111)) enter the 3D precipitate formation energy through the S*sigma/N term (Eq. 9 with zeta = 0); for a sphere this is 3*v*sigma/r. At the observed 2-5 nm precipitate sizes, an uncertainty of only +/-0.03 eV/A^2 in sigma shifts the crossover between semicoherent CrV/TaW and coherent CrTa/VW by roughly a nanometer or more, potentially moving it outside the experimentally observed size range and undermining the \"must be semicoherent\" inference. The paper also acknowledges the equiatomic B2 idealization (\"the theoretical crossovers are for ideal equiatomic B2 binaries\") while the real precipitates are about 70% Cr and 30% V in a W38Ta36Cr15V11 matrix; this composition difference further clouds the quantitative crossover location, but the unvalidated semicoherent interfacial energy is the most directly load-bearing input.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript develops a tabulated Gaussian approximation potential (tabGAP) for the W–Ta–Cr–V system, validates it extensively against DFT for binary formation energies, energy-volume curves, point defects, surfaces, and grain boundaries, and then uses hybrid MC/MD simulations to study short-range order and segregation in WTaV, WTaCrV, and MoNbTaVW. Atom probe tomography of W38Ta36Cr15V11 thin films is used to characterize compositional striations in the as-deposited state and globular CrV-rich precipitates after irradiation. To explain these observations, the authors compute coherent and semicoherent interface energetics between B2-ordered CrV/TaW and CrTa/VW binaries, combining DFT coherency-strain energies with tabGAP interface energies. The resulting crossover curves in Fig. 14 indicate that coherent CrV precipitates in TaW are unstable at all sizes, and that semicoherent CrV/TaW becomes competitive with coherent CrTa/VW at thicknesses or radii around 2.5–3 nm, from which the authors conclude that the experimentally observed CrV-rich precipitates are semicoherent. The paper also draws general conclusions about segregation trends linked to pure-element surface, grain-boundary, and atomic-size properties.","tokens_in":27258,"tokens_out":8750,"duration_ms":82022,"significance":"If the central claim holds, the paper resolves a real discrepancy: coherent-lattice MC/MD simulations and cluster-expansion models disagree on whether CrV or CrTa ordering is preferred in WTaCrV, and the paper explains this by showing that the coherent lattice constraint selects CrTa/VW while the experimentally relevant state is semicoherent CrV/TaW. The work is strengthened by the breadth of DFT validation of the tabGAP, the iterative training on short-range-order-relevant structures, the cross-validation of two independent tabGAPs for WTaV, and the fact that the APT observations were not used as fitting input for any parameter. The conclusion that coherent CrV 3D precipitates are unstable is robust because the DFT coherency-strain energy (0.248 eV/atom) alone exceeds the combined B2 formation-energy gain of CrV and TaW (−0.08 eV/atom). The main uncertainty is quantitative: the semicoherent branch of Fig. 14, which is load-bearing for the semicoherency conclusion, rests on tabGAP-only interface energies for misfit-dislocation-bearing interfaces, and the crossover is computed for ideal equiatomic B2 binaries rather than the measured precipitate composition.","major_comments":[{"comment":"The semicoherent branch of Fig. 14 is the load-bearing input for the statement that CrV-rich precipitates in WTaCrV must be semicoherent, and it rests entirely on tabGAP interfacial energies: the text states that 'Semicoherent interfacial energies are always from tabGAP.' The tabGAP is validated against DFT for bulk phases, point defects, surfaces, and grain boundaries, but not for the 1/2<111> misfit dislocation networks that define these semicoherent interfaces. In Eq. (9) with ζ = 0, the interface term for a spherical precipitate is 3vσ/r; with v ≈ 12.2 Å^3 and σ in the 0.05–0.15 eV/Å^2 range reported, an uncertainty of only ±0.03 eV/Å^2 changes this term by roughly 0.04 eV/atom at r = 3 nm, which is larger than the 0.03 eV/atom asymptotic energy difference between the CrV/TaW and CrTa/VW combinations. Such an uncertainty can move the crossover outside the observed 2–5 nm precipitate-size range and thereby undermine the 'must be semicoherent' inference. I request either a DFT benchmark for at least one semicoherent interface (for example, the smallest commensurate 9:10 CrV/TaW supercell) or a systematic sensitivity analysis over σ demonstrating that the crossover location is stable to ±0.03 eV/Å^2.","section":"III D 2, Fig. 14, Table IV"},{"comment":"The crossover calculation is performed for ideal equiatomic B2 CrV/TaW and CrTa/VW binaries, as the paper itself acknowledges with 'the theoretical crossovers are for ideal equiatomic B2 binaries.' The APT-measured precipitates, however, are approximately 70% Cr and 30% V, and the matrix is W38Ta36Cr15V11 rather than equiatomic. Off-equiatomic compositions will change both the bulk formation-energy input to Eq. (9) (the −0.08 vs −0.05 eV/atom values) and the interfacial energies. A calculation at the measured precipitate composition, or at least an explicit demonstration that the crossover is insensitive to composition, is needed before the observed precipitates can be quantitatively identified with the calculated crossover.","section":"III D 2, Figs. 13–14"}],"minor_comments":[{"comment":"Equation (9) is derived for a periodic superlattice with two interfaces, but the 3D spherical-cluster application is described only as 'we use Eq. 9.' Please state explicitly how the 3D formula is obtained (one interface, S = 4πr^2, N = V/v) and whether the factor 2 in Eq. (9) is dropped for the spherical case.","section":"II F"},{"comment":"Several tabulated interfacial energies are negative (for example, σ(100) = −0.055 eV/Å^2 for CrV/TaW tabGAP). A sentence explaining that these negative values reflect favorable chemical mixing across the interface rather than an unphysical negative interface energy would prevent misinterpretation.","section":"Table IV"},{"comment":"The label 'DFT/tabGAP pred.' is used in the figure legend but is defined only in the main text; please define it in the caption as well, and state which quantities are from DFT and which are from tabGAP.","section":"Fig. 14 caption"},{"comment":"The phrase 'volemetric strains' should read 'volumetric strains.'","section":"III C 3"},{"comment":"The wording 'must be semicoherent' is stronger than the evidence supports, since APT cannot resolve the misfit dislocation network and the paper notes that no high-resolution imaging was possible. 'Are predicted to be semicoherent' would better match the computational nature of the evidence.","section":"III D 2"},{"comment":"Reference [43] is cited as 'W-Ta-Cr-V tabGAP: Potential files, training data, and input (2025)' without a repository URL or DOI; please provide the full citation so the potential and training data are actually accessible.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"This is a substantial computational paper with a valuable potential and a plausible resolution of a genuine discrepancy between experiments and coherent-lattice simulations. My concern is not circularity or fit, but that the central semicoherency claim is quantitatively tied to tabGAP-only semicoherent interface energies and to equiatomic B2 model systems. The requested sensitivity analysis or DFT spot check for at least one semicoherent interface is feasible and should be required before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper develops a new tabGAP potential for W-Ta-Cr-V, maps segregation to a broad set of defects in three refractory alloys, and offers a genuinely plausible resolution to the discrepancy between coherent-lattice simulations (which predict CrTa/WV clustering) and experiments (which see CrV-rich precipitates): coherent CrV/TaW is thermodynamically unstable because of large lattice mismatch, and semicoherent CrV/TaW becomes favorable above roughly 2.5-3 nm, matching the 2-5 nm precipitates seen in the APT data. That resolution is the real contribution, and it is well argued. The potential is validated extensively against DFT for binary formation energies, energy-volume curves, point defects, surfaces, and grain boundaries; the MC/MD energy and force RMSEs are excellent; and the thermodynamic decomposition of interface energetics is clean. The conclusion that coherent CrV 3D precipitates are unstable at all sizes is robust because it relies on DFT coherency strain and DFT interfacial energies. The work is also clearly new relative to the earlier Lyu et al. potential for the same quaternary system, which did not address segregation or precipitation.\n\nThe main soft spot is the quantitative crossover in Fig. 14. The semicoherent branch is computed using tabGAP only, with no DFT cross-check for the 1/2<111> misfit dislocation networks that define those interfaces. The paper states this openly. The quoted semicoherent interfacial energies are 0.05-0.15 eV/A^2 for the (100) termination, 0.057 for (110), and 0.096 for (111). An uncertainty of even +/-0.03 eV/A^2 shifts the crossover by roughly a nanometer, which is material when the observed precipitates are only 2-5 nm. The paper also acknowledges that the model systems are ideal equiatomic B2 binaries while the real precipitates are about 70% Cr and 30% V in a W38Ta36Cr15V11 matrix. These caveats make the phrase \"must be semicoherent\" a bit stronger than the evidence supports; \"likely semicoherent\" or \"consistent with semicoherent\" would be more accurate. I do not think the stress-test concern is fatal, because the qualitative mechanism is almost certainly right and the missing DFT check is a tractable follow-up, not a fundamental contradiction.\n\nOther limitations are minor: no direct experimental imaging of the interfaces, and the nucleation pathway for the precipitates is not established. The paper is honest about both. The segregation trends and the connection to pure-element surface and grain boundary energies are useful and likely transferable.\n\nWho should read this: computational materials scientists working on refractory high-entropy alloys, fusion first-wall alloy design, and anyone building or benchmarking ML interatomic potentials. The paper deserves a serious referee. I would send it to review, with the request that the authors add at least one DFT validation of a semicoherent interface and explicitly discuss the sensitivity of the crossover to the interfacial energy uncertainty.","headline":"A solid, well-validated tabGAP study that likely explains the CrV/TaW vs CrTa/WV discrepancy via semicoherent interfaces; the main weakness is that the crossover curve leans on tabGAP-only semicoherent interfacial energies.","tokens_in":860,"tokens_out":1151,"would_cite":true,"duration_ms":29589,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"CrV precipitates in WTaCrV are semicoherent, not coherent: simulations and atom probe data agree.","keywords":["refractory high-entropy alloys","machine-learned interatomic potential","tabGAP","short-range order","segregation","semicoherent interfaces","precipitation","atom probe tomography"],"falsifier":"A DFT calculation of a semicoherent CrV/TaW interface with a misfit dislocation network, or direct high-resolution transmission electron microscopy of the interfaces in the irradiated W38Ta36Cr15V11 films, would settle whether the precipitates are semicoherent. If DFT placed semicoherent CrV/TaW above coherent CrTa/VW at all sizes up to 5 nm, or if imaging showed no misfit dislocation network while precipitates were still present, the central claim would fail.","tokens_in":26744,"feed_emoji":"⚛️","tokens_out":4804,"duration_ms":38918,"temperature":0.7,"pith_summary":"This paper argues that the CrV-rich precipitates observed in irradiated WTaCrV refractory alloys are not coherently embedded in the matrix; instead they form semicoherent bcc-to-bcc interfaces carrying networks of misfit dislocations. The authors build a machine-learned interatomic potential for the W-Ta-Cr-V system, validate it against density functional theory, and use hybrid Monte Carlo/molecular dynamics simulations plus interface-energy calculations to show that coherent CrV precipitates in a TaW matrix are thermodynamically unstable at all sizes. Above roughly 2.5 to 3 nm in slab thickness or radius, a semicoherent CrV/TaW configuration becomes the lowest-energy structure, matching the 2 to 5 nm precipitates seen in atom probe tomography. This resolves a discrepancy in which coherent-lattice simulations predicted CrTa/WV short-range order while experiments and an earlier cluster-expansion model see CrV precipitates.","feed_headline":"CrV precipitates in WTaCrV are semicoherent, simulations show","feed_subtitle":"Coherent CrV clusters are unstable at any size; misfit-dislocation interfaces win above 3 nm, matching experiments.","key_machinery":"The load-bearing machinery is the tabGAP machine-learned interatomic potential for the full W-Ta-Cr-V composition space, trained on about 13,483 density functional theory structures. To explain precipitation, the authors combine the interfacial superlattice formation energy E_f/N = 2Sσ/N + ζ + $ΔE_f^{{AB}}$/N with coherency strain energies computed from Eq. 10, obtaining coherent interface energies with both DFT and tabGAP and semicoherent interface energies with tabGAP for the mismatched CrV/TaW system. The crossover curves in Fig. 14 are generated from these ingredients and are what place semicoherent CrV/TaW below coherent CrTa/VW above 2.5 to 3 nm.","core_discovery":"The central claim is that the experimentally observed CrV-rich precipitates in WTaCrV form semicoherent bcc-to-bcc interfaces with the surrounding TaW-rich matrix, because coherent precipitates are unstable due to excessive lattice mismatch. The evidence is a crossover analysis: coherent CrV three-dimensional precipitates in a TaW matrix have positive formation energy of about 0.17 eV/atom at large radius, while semicoherent CrV/TaW becomes the lowest-energy configuration above 2.5 to 3 nm in slab thickness or 3 nm in radius, in the size range of the 2 to 5 nm precipitates observed by atom probe tomography. A Monte Carlo/molecular dynamics simulation starting from a random W38Ta36Cr15V11 solution with a compressed embedded precipitate produces a semicoherent precipitate with roughly 70% Cr and 30% V composition and a network of misfit dislocations, quantitatively matching the earlier experiments. The paper therefore concludes that earlier coherent-lattice simulations predicted the wrong short-range order because they excluded semicoherent interfaces.","pith_inferences":["If the crossover depends on the equiatomic B2 idealization, then for the real roughly 70% Cr and 30% V precipitates the crossover size and interfacial energy may shift; extending the calculation to non-equiatomic compositions would be a direct test.","The semicoherent interfacial energies are currently computed only with the machine-learned potential; a DFT calculation of a misfit-dislocation-bearing CrV/TaW supercell would either confirm or revise the 2.5 to 3 nm crossover.","The paper leaves open how semicoherent precipitates form; the proposed pathway through segregation to interstitial dislocation loops could be tested by kinetic simulations of CrV nucleation at interstitial clusters.","If semicoherent interfaces are indeed the stable form, mechanical properties of irradiated WTaCrV such as hardening may be controlled more by misfit dislocation networks at precipitate interfaces than by short-range order alone, which would affect alloy design considerations."],"forward_implications":["If the central claim is correct, the CrV-rich precipitates seen in irradiated WTaCrV are surrounded by misfit dislocation networks, and any simulation that fixes atoms on a single coherent lattice will miss the observed short-range order.","The crossover at 2.5 to 3 nm implies that small coherent CrV clusters are not thermodynamically stable, explaining why the experimental precipitates are 2 to 5 nm and why no smaller precipitates are observed.","The uniform segregation rule that small atoms (Cr, V) prefer compressed regions and large atoms (Ta, Nb) prefer tensile regions extends across WTaV, WTaCrV, and MoNbTaVW, offering a simple size-based guideline for other refractory alloys.","Grain boundary segregation in WTaV approaches TaV2 composition, suggesting grain boundaries can act as nucleation sites for Laves phases.","In WTaCrV, Cr segregation to defects is accompanied by V to form stable CrV mixtures, making it the most segregation-resistant of the three alloys studied."],"supporting_citations":[{"why":"Supplies the experimental WTaCrV thin-film samples, the observation of CrV-rich precipitates after irradiation, and the atom probe tomography data reanalyzed in this paper.","marker":"[5]"},{"why":"Supplies the ordered binary phases and cluster-expansion short-range-order predictions that conflict with coherent-lattice ML simulations and motivate the interface analysis.","marker":"[13]"},{"why":"Supplies the other machine-learned potential for WTaCrV that also predicts CrTa/WV short-range order, establishing the discrepancy the paper resolves.","marker":"[20]"},{"why":"Supplies the tabGAP framework and prior ordering and segregation results for MoNbTaVW that the present simulations extend.","marker":"[11]"},{"why":"Supplies the improved MoNbTaVW tabGAP and the hyperparameter and training methodology used for the new WTaCrV potential.","marker":"[19]"},{"why":"Supplies DFT vacancy and interstitial formation and migration energies used to validate the new WTaCrV potential.","marker":"[30]"},{"why":"Supplies the interfacial superlattice formation-energy expression used to extract interfacial energies.","marker":"[56]"},{"why":"Supplies the coherency strain energy formula used for coherent interfaces and precipitates.","marker":"[57]"}],"fun_headline_variants":["Semicoherent CrV precipitates win in WTaCrV above 3 nm","Coherent CrV clusters unstable; semicoherent interfaces match experiments","Misfit dislocations stabilize CrV precipitates in WTaCrV","Why CrV precipitates in WTaCrV are semicoherent","Semicoherent CrV/TaW interfaces beat coherent above 3 nm"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The conclusion that the observed precipitates must be semicoherent rests on semicoherent interfacial energies computed only with the machine-learned potential, without a DFT cross-check, and on model systems that are ideal equiatomic B2 binaries rather than the real roughly 70% Cr and 30% V precipitates.","fun_headline_variants_meta":{"raw":{"variants":["Semicoherent CrV precipitates win in WTaCrV above 3 nm","Coherent CrV clusters unstable; semicoherent interfaces match experiments","Misfit dislocations stabilize CrV precipitates in WTaCrV","Why CrV precipitates in WTaCrV are semicoherent","Semicoherent CrV/TaW interfaces beat coherent above 3 nm"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000274,"raw_usage":{"total_tokens":1643,"prompt_tokens":952,"completion_tokens":691,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":568,"completion_tokens_details":{"reasoning_tokens":592}},"tokens_in":568,"tokens_out":691,"duration_ms":5563,"temperature":1.0,"reasoning_tokens":592,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T12:49:53.194102+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A DFT calculation of a semicoherent CrV/TaW interface with a misfit dislocation network, or direct high-resolution transmission electron microscopy of the interfaces in the irradiated W38Ta36Cr15V11 films, would settle whether the precipitates are semicoherent. If DFT placed semicoherent CrV/TaW above coherent CrTa/VW at all sizes up to 5 nm, or if imaging showed no misfit dislocation network while precipitates were still present, the central claim would fail.","supporting_citations":[{"cited_title":"Based on Tab","cited_arxiv_id":null,"evidence_quote":"Supplies the experimental WTaCrV thin-film samples, the observation of CrV-rich precipitates after irradiation, and the atom probe tomography data reanalyzed in this paper."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the ordered binary phases and cluster-expansion short-range-order predictions that conflict with coherent-lattice ML simulations and motivate the interface analysis."},{"cited_title":"Modeling refractory high-entropy alloys with efficient machine-learned interatomic potentials: defects and segregation","cited_arxiv_id":"2106.03369","evidence_quote":"Supplies the other machine-learned potential for WTaCrV that also predicts CrTa/WV short-range order, establishing the discrepancy the paper resolves."},{"cited_title":"Cheng, J","cited_arxiv_id":null,"evidence_quote":"Supplies the tabGAP framework and prior ordering and segregation results for MoNbTaVW that the present simulations extend."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the improved MoNbTaVW tabGAP and the hyperparameter and training methodology used for the new WTaCrV potential."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the interfacial superlattice formation-energy expression used to extract interfacial energies."},{"cited_title":"Stukowski, Visualization and analysis of atomistic simulation data with OVITO–the Open Visualization Tool, Modelling Simul","cited_arxiv_id":null,"evidence_quote":"Supplies the coherency strain energy formula used for coherent interfaces and precipitates."}],"review_version":1}