{"id":"bb3afec4-2764-4e1d-b26f-feb961cc1034","arxiv_id":"2605.29482","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":7.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"Fine-tuned MACE potentials screen 1500 equiatomic multicomponent carbides, showing group number controls stability/hardness, short-range order effects are small, and a new family of stacking-ordered phases appears in group 4/5-6 mixtures.","lead":"The paper fine-tunes a machine-learned interatomic potential on DFT calculations to screen thousands of multicomponent transition metal carbide compositions for stability and hardness. A smart generalist might read it to see how ML can accelerate discovery of materials for extreme environments and whether a new type of ordered phase emerges only when mixing many metals.","discovery_kind":"new_application","skeptic_critique":{"model":"grok-4.3","headline":"MACE ~10 meV/atom error may not reliably rank stability of new stacking-ordered phases absent targeted validation","rationale":"The reader's weakest_assumption directly identifies the same point; the full-text availability does not remove the need for explicit validation of the potential on the novel ordered structures that carry the headline result. Because DFT is stated to corroborate only the final candidates, the high-throughput screening step remains the load-bearing link.","tokens_in":1865,"tokens_out":377,"duration_ms":16814,"concrete_test":"Take the 8–10 equiatomic compositions where MACE reports the largest stabilization (>25 meV/atom) of a stacking-ordered phase relative to rocksalt; recompute their formation energies with the same DFT settings used in the original training set and check whether the ordering and magnitude of the difference are preserved within 10 meV/atom.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim identifies a new family of stacking-ordered phases (group 4/5 + group 6 metals) whose formation energies lie well below disordered rocksalt/hexagonal structures, enabling experimental accessibility. This ranking is produced by the fine-tuned MACE potential screened over 1500 compositions; the model reports average formation-energy error of ~10 meV/atom on thermodynamically relevant structures after training on only 20% of the 28k DFT set. No section details (i) error on stacking-ordered configurations specifically, (ii) whether the training set contained analogous long-period order, or (iii) how the 10 meV/atom uncertainty propagates into the decision that one prototype is “well below” another. If the true energy differences for the newly proposed phases fall inside or near this uncertainty, the ML-driven discovery of the family and the subsequent claim of synthesizability rest on an unverified extrapolation.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript claims that fine-tuning the MACE ML interatomic potential on ~28k DFT calculations for groups 4-6 transition metal carbides enables high-throughput screening of >1500 equiatomic compositions for thermodynamic stability and elastic hardness. It reports that metal group number governs both properties, that short-range order free-energy contributions are small (~few meV/atom), that synthesizability predictions at 1500°C match experiments, and that a new family of stacking-ordered phases in group 4/5 + group 6 mixtures has formation energies well below those of disordered rocksalt or hexagonal structures; DFT calculations on these phases are said to corroborate the ML predictions and indicate experimental accessibility.","tokens_in":2066,"tokens_out":583,"duration_ms":23690,"significance":"If the central claims hold, the work supplies a practical ML-accelerated framework for screening synthesizable multicomponent carbides with target mechanical properties across a nine-element space, demonstrates that group number is a dominant descriptor, and identifies a previously unreported class of stacking-ordered carbides whose stability appears accessible only in multicomponent compositions. The agreement between ML-based synthesizability predictions and existing experiments, together with the explicit DFT corroboration step for the new phases, strengthens the practical utility of the approach.","major_comments":[{"comment":"Abstract and methods: the reported average formation-energy error of ~10 meV/atom after training on only 20% of the DFT set is presented without (i) a breakdown of the train/validation split, (ii) error statistics specifically on stacking-ordered or long-period structures, or (iii) propagation of this uncertainty into the stability ranking that selects the new family; because the central discovery rests on ML-driven identification of phases whose energies are claimed to lie “well below” disordered prototypes, this omission is load-bearing.","section":"Abstract"},{"comment":"Results on new phases: while DFT is stated to corroborate the stacking-ordered predictions, the manuscript does not report how many candidate compositions were advanced from the 1500-composition ML screen to DFT, nor the magnitude of the formation-energy differences relative to the 10 meV/atom uncertainty; if those differences fall inside or near the model error, the claim that the new family is distinctly more stable (and therefore experimentally accessible) requires additional targeted validation.","section":"Results"}],"minor_comments":[{"comment":"The description of the elasticity-based hardness surrogate and its calibration against known carbides should be expanded for reproducibility.","section":"Methods"},{"comment":"Figure captions and text should explicitly state the temperature and reference states used for the free-energy comparisons that underpin the synthesizability predictions.","section":"Figure captions"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for their thorough review and valuable feedback on our manuscript. The comments raise important points regarding the validation of our machine learning model and the robustness of our findings on the new stacking-ordered phases. We will revise the manuscript to incorporate the requested details and clarifications.","responses":[{"response":"We agree with the referee that these details are important for assessing the reliability of our predictions. In the revised manuscript, we will expand the Methods section to include a full breakdown of the train/validation split used in fine-tuning the MACE potential. We will also report error statistics broken down by structure type, including specifically for stacking-ordered and long-period structures. Furthermore, we will add an analysis in the Results section that propagates the model uncertainty into the stability rankings, demonstrating that the energy differences for the identified stacking-ordered phases remain significant relative to the ~10 meV/atom error. These revisions will directly address the load-bearing nature of this information for our central claims.","revision_made":"yes","referee_comment":"[Abstract] Abstract and methods: the reported average formation-energy error of ~10 meV/atom after training on only 20% of the DFT set is presented without (i) a breakdown of the train/validation split, (ii) error statistics specifically on stacking-ordered or long-period structures, or (iii) propagation of this uncertainty into the stability ranking that selects the new family; because the central discovery rests on ML-driven identification of phases whose energies are claimed to lie “well below” disordered prototypes, this omission is load-bearing."},{"response":"We acknowledge that the manuscript would benefit from more explicit reporting on the DFT validation step. In the revision, we will specify the number of candidate compositions (selected based on ML-predicted stability) that were advanced to full DFT calculations for corroboration. We will also provide the quantitative formation-energy differences between the stacking-ordered phases and the disordered rocksalt/hexagonal structures, along with a comparison to the model uncertainty. This will include showing that the differences are well outside the error margin, supporting the claim of distinct stability and experimental accessibility. We believe this additional information will strengthen the presentation without altering the conclusions.","revision_made":"yes","referee_comment":"[Results] Results on new phases: while DFT is stated to corroborate the stacking-ordered predictions, the manuscript does not report how many candidate compositions were advanced from the 1500-composition ML screen to DFT, nor the magnitude of the formation-energy differences relative to the 10 meV/atom uncertainty; if those differences fall inside or near the model error, the claim that the new family is distinctly more stable (and therefore experimentally accessible) requires additional targeted validation."}],"tokens_in":1592,"tokens_out":572,"duration_ms":26360,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The main takeaway is that mixing group 4/5 and group 6 metals produces a new class of stacking-ordered carbide phases whose formation energies sit well below those of disordered rocksalt or hexagonal structures.\n\nThe work fine-tunes MACE on roughly 28,000 DFT calculations across the nine-component space, reaches ~10 meV/atom error on relevant structures with only 20% of the data, then screens 1500 equiatomic compositions. It combines the potential with free-energy models and elasticity-based hardness estimates. Synthesizability predictions at 1500 °C line up with known experiments for both single-phase and multiphase cases. Short-range order contributions turn out small, so the disordered approximation is reasonable for screening. The group number of the metals correlates with both stability and hardness.\n\nThe new stacking-ordered family is the clearest addition; the abstract presents it as accessible only in multicomponent compositions and not seen in prior single-principal-element work. DFT checks on those phases provide independent grounding.\n\nThe soft spot is the transferability of the 10 meV/atom error to the new ordered structures and to the full nine-component space. The paper does not detail error specifically on stacking-ordered configurations or whether the training set contained similar long-period order, so the initial ranking step rests on an extrapolation whose size relative to the claimed energy gaps is not shown. That said, the DFT corroboration for the new phases keeps the central claim from collapsing.\n\nThis is useful for people working on high-entropy ceramics or hard coatings for extreme conditions. Readers who need a reusable screening framework or concrete composition suggestions will find value. It deserves peer review because the scale of the screen plus the DFT follow-up gives it enough grounding to be worth referee time, even if the model-validation sections could be tightened.","headline":"The paper uses ML screening to flag a new family of stacking-ordered multicomponent carbides that DFT then backs as more stable than standard prototypes.","tokens_in":2585,"tokens_out":438,"would_cite":false,"duration_ms":19643,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"Multicomponent carbides mixing group 4/5 and group 6 metals form stacking-ordered phases with formation energies below those of disordered rocksalt or hexagonal structures.","keywords":["multicomponent carbides","transition metal carbides","machine-learned potentials","stacking-ordered phases","thermodynamic stability","elastic properties","synthesizability","high-throughput screening"],"falsifier":"Synthesis of one predicted stacking-ordered multicomponent carbide (for example, a Ti-Zr-Mo-C composition) at 1500 °C followed by X-ray diffraction that either confirms or rules out the ordered stacking sequence.","tokens_in":2746,"feed_emoji":"🔬","tokens_out":739,"duration_ms":21182,"temperature":0.7,"pith_summary":"This paper fine-tunes a machine-learned interatomic potential on density functional theory data to screen thermodynamic stability and elastic properties across thousands of equiatomic compositions in groups 4-6 transition metal carbides. The group number of the metals emerges as the main driver of both stability and hardness, while short-range order effects on free energy remain small enough that a disordered solid-solution model suffices for initial screening. The central finding is a new family of stacking-ordered phases that appear only in mixed group compositions and lie energetically below standard prototypes. Direct density functional theory checks support these predictions and indicate the phases should form under typical synthesis temperatures.","feed_headline":"Stacking order stabilizes new multicomponent carbide phases","feed_subtitle":"Mixed group 4/5 and group 6 metal carbides gain lower formation energies in ordered stacking structures than in disordered rocksalt or hexag","key_machinery":"The fine-tuned MACE machine-learned interatomic potential, which ranks formation energies and elastic properties for rapid screening of multicomponent compositions and identifies the stabilizing effect of stacking order in mixed-group carbides.","core_discovery":"Fine-tuning the MACE potential on roughly 28,000 density functional theory calculations enables accurate prediction of formation energies to about 10 meV per atom across the nine-component space. Screening more than 1500 compositions shows that mixing group 4/5 and group 6 metals stabilizes a previously unreported class of stacking-ordered carbide phases whose energies fall well below those of disordered rocksalt and hexagonal structures. Density functional theory calculations on the predicted structures confirm the energy ordering and suggest experimental accessibility at 1500 °C.","pith_inferences":["Targeted synthesis experiments could focus on equiatomic mixtures that combine early and late transition metals to test the predicted stacking order.","The same screening workflow could be applied to other multicomponent ceramics where stacking order might similarly lower energy.","Elastic-property predictions suggest these phases may offer hardness advantages that warrant direct nanoindentation measurements once synthesized."],"forward_implications":["Group number of the metals controls both thermodynamic stability and hardness across the full composition space.","Free-energy contributions from short-range order are only a few meV per atom, validating the disordered solid-solution approximation for high-throughput searches.","Synthesizability predictions at 1500 °C match known experimental single-phase and multiphase carbide behavior.","Stacking-ordered phases constitute a distinct, lower-energy family accessible only through multicomponent mixing."],"fun_headline_variants":["Stacking order lowers formation energy in mixed metal carbides","MACE potentials find stable ordered phases in multicomponent carbides","Screening uncovers stacking-ordered carbides in nine-component space","Group 4-6 mixes stabilize ordered stacking carbide structures"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The fine-tuned potential remains accurate enough to rank the thermodynamic stability of the new stacking-ordered structures correctly even when trained on only 20 percent of the density functional theory data.","fun_headline_variants_meta":{"raw":{"variants":["Stacking order lowers formation energy in mixed metal carbides","MACE potentials find stable ordered phases in multicomponent carbides","Screening uncovers stacking-ordered carbides in nine-component space","Group 4-6 mixes stabilize ordered stacking carbide structures"]},"model":"grok-4.3","cost_usd":0.006096,"raw_usage":{"total_tokens":2936,"prompt_tokens":780,"num_sources_used":0,"completion_tokens":66,"cost_in_usd_ticks":60962000,"prompt_tokens_details":{"text_tokens":780,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":2090,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":780,"tokens_out":66,"duration_ms":16877,"temperature":1.0,"reasoning_tokens":2090,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-29T06:49:58.883187+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"Synthesis of one predicted stacking-ordered multicomponent carbide (for example, a Ti-Zr-Mo-C composition) at 1500 °C followed by X-ray diffraction that either confirms or rules out the ordered stacking sequence.","supporting_citations":[],"review_version":1}