{"id":"62a8310c-d235-4423-b867-1172be4cb6ae","arxiv_id":"2505.00376","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A hybrid NEP+ILP force field reproduces lattice constants, phonon spectra, and thermal conductivities of six transition metal dichalcogenides at far lower cost than density functional theory.","lead":"This paper tests a hybrid computer model that combines a machine-learned potential for atomic bonds inside layers with a registry-dependent potential for weak forces between layers, applied to six stacked transition metal dichalcogenides. The model reproduces many measured properties at low computational cost, which could accelerate design of 2D thermal devices.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Bulk modulus underpredictions of 30-59% in Section 3.2 contradict the abstract's unqualified 'accurately predicts bulk modulus' claim.","rationale":"The reader's weakest_assumption (PBE+MBD-NL vs HSE+MBD-NL for Te systems) is a real concern for two of the six compounds, but it cannot explain the bulk modulus discrepancies observed for MoS2, where the ILP was parameterized against HSE+MBD-NL in prior work. The internally inconsistent accuracy claim in the abstract is more load-bearing because it is directly falsified by the paper's own data across all five compounds with experimental references. The reader's rationale did mention the bulk modulus overstatement, but their formal weakest_assumption pointed elsewhere; hence partial agreement. The verdict should remain CONDITIONAL: the method is useful and many benchmarks are strong, but the central claim must be toned down and the bulk modulus limitation acknowledged. Verdict_should_be is therefore UNCHANGED relative to the reader's conditional recommendation.","tokens_in":23100,"tokens_out":5390,"duration_ms":50188,"concrete_test":"Recompute the pressure-volume curve for bulk MoS2 using the NEP+ILP model at 5, 10, and 14 GPa (as in Section 3.1 of the Supporting Information) and directly compare against the experimental equation-of-state data of Aksoy et al. (Ref. 83). If the NEP+ILP volume deviates by more than 5% at any of these pressures, the bulk modulus underestimation is confirmed as a systematic deficiency of the interlayer potential, and the abstract's 'accurately predicting ... bulk modulus' statement should be revised to a qualified claim about ambient-pressure behavior.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim of near-quantum-mechanical accuracy with accurate bulk modulus prediction is contradicted by the paper's own benchmark in Section 3.2. The Murnaghan bulk moduli are 35.1, 29.7, 24.5, 25.9, 29.8, and 23.8 GPa for MoS2, MoSe2, MoTe2, WS2, WSe2, and WTe2, underestimating experimental results by approximately 30.5%, 44.8%, 42.3%, 58.2%, and 58.6% (WTe2 lacks experimental reference). The authors explicitly state these values 'underestimate the experimental results.' This is not solely a PBE-vs-HSE artifact for the Te systems: MoS2 uses a previously benchmarked HSE+MBD-NL-based ILP, yet still shows a 30.5% error. Thus the abstract's unqualified claim that the framework 'accurately predicts' bulk modulus, and the conclusion's 'good agreement' phrasing, overstate the demonstrated accuracy. Because bulk modulus is one of the headline properties, the central claim is partially falsified by the paper's own data. The concern is load-bearing because it indicates the interlayer potential is systematically too soft under hydrostatic compression, which also affects pressure-dependent interlayer spacing, moire reconstruction energetics, and possibly cross-plane thermal transport in compressed heterostructures.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript extends a hybrid machine-learning/registry-dependent interlayer potential framework (NEP+ILP) to six TMD homojunctions (MoS2, MoSe2, MoTe2, WS2, WSe2, WTe2). The intralayer interactions are described by a neuroevolution potential (NEP) trained on monolayer DFT data, and the interlayer interactions by an anisotropic registry-dependent ILP fitted to binding-energy curves and sliding potential energy surfaces. The authors benchmark the combined potential against lattice constants, bulk modulus, moiré reconstruction, phonon spectra, and in-plane and cross-plane thermal conductivities, and also compare with an alternative NEP-D3 approach for MoS2. The central claim, stated in the abstract and conclusions, is that the NEP+ILP framework achieves near-quantum-mechanical accuracy for key TMD properties while remaining scalable to large moiré superlattices.","tokens_in":23451,"tokens_out":2595,"duration_ms":27283,"significance":"If the hybrid framework delivers DFT-level accuracy for interlayer registry effects and thermal transport at classical MD cost, it would be a practically valuable tool for studying TMD heterostructures, which are difficult to treat by direct DFT at moiré length scales. The paper includes useful strengths: the potentials and code are made available, the cross-plane thermal conductivities agree well with available experiments, and the direct comparison with NEP-D3 provides a concrete assessment of an alternative approach. However, the paper's own benchmark data show that the bulk modulus is underestimated by 30–59% relative to experiments for five of the six systems, which directly contradicts the unqualified 'accurately predicts bulk modulus' claim in the abstract. This is a load-bearing discrepancy because bulk modulus is listed among the headline validated properties, and a too-soft interlayer response under compression may affect pressure-dependent properties and moiré reconstruction energetics.","major_comments":[{"comment":"The reported Murnaghan bulk moduli are 35.1, 29.7, 24.5, 25.9, 29.8, and 23.8 GPa for MoS2, MoSe2, MoTe2, WS2, WSe2, and WTe2, and the manuscript states that for MoS2, MoSe2, MoTe2, WS2, and WSe2 these underestimate experimental results by about 30.5%, 44.8%, 42.3%, 58.2%, and 58.6%, respectively. This is more than a minor deviation, and it contradicts the abstract's statement that the framework 'accurately predicts' the bulk modulus. Because the bulk modulus is one of the headline properties in the central claim, the overstatement should be corrected and, more importantly, the origin of the systematic softness of the interlayer potential under hydrostatic compression should be discussed or mitigated; if it cannot be mitigated, the claims must be limited to properties for which the demonstrated accuracy is supported.","section":"Section 3.2 and Fig. 2"},{"comment":"The ILP parameters for MoTe2 and WTe2 are fitted to PBE+MBD-NL reference data because HSE+MBD-NL is deemed computationally prohibitive, and the manuscript justifies this by stating that the disparity between PBE+MBD-NL and HSE+MBD-NL is 'small', with the quantitative comparison deferred to Section 2 of the Supporting Information. Since the Te-based systems comprise one third of the benchmark set, this substitution is load-bearing: any systematic difference between PBE and HSE interlayer binding or sliding energies would directly bias the lattice constants, bulk moduli, phonon spectra, and thermal conductivities reported for MoTe2 and WTe2. The quantitative evidence for this equivalence should be presented in the main text or the argument should be made convincingly in the Supporting Information; the current one-sentence assertion is not sufficient.","section":"Section 2.3"},{"comment":"The abstract claims 'near quantum mechanical accuracy' and 'accurately predicting key properties such as lattice constants, bulk modulus, moiré reconstruction, phonon spectra, and thermal conductivities', while the conclusions say the framework yields 'bulk modulus values and in-plane thermal conductivity in good agreement' with DFT and 'cross-plane thermal conductivity aligned with experimental results'. Given the measured bulk-modulus errors of 30–59% and the in-plane thermal conductivities that are systematically higher than most experiments (though consistent with BTE predictions), the phrase 'good agreement' is too strong. The claims should be recalibrated to state exactly which properties agree quantitatively and which agree only semi-quantitatively or with known systematic deviations.","section":"Abstract and Section 5"}],"minor_comments":[{"comment":"There are several typographical errors, including 'respecitvley' in Section 3.3, 'diomands' in Figure 8, and 'bashed' in the Table 1 caption; these should be corrected.","section":"Throughout"},{"comment":"The notation in Equation (5) is confusing: the neighbor sets Ω_k and Ω_l are introduced, but the text then refers to 'atom i in layer k and atom i in layer l' with inconsistent subscripts. Please clarify the index convention.","section":"Equation (5) and surrounding text"},{"comment":"The monolayer thicknesses used for converting 2D to 3D thermal conductivity are defined (6.2, 6.6, 7.2 Å), but the choice of these values and their sensitivity to the reported thermal conductivities are not discussed; a brief justification or a reference for the effective thickness convention would be helpful.","section":"Section 3.5"},{"comment":"In Figure 2, the caption states that 'Reported experimental values and our MD simulation results are presented as blue and red circles', but the figure also includes DFT data; please clarify the symbol and color coding for all three data types.","section":"Figure captions"},{"comment":"The deviations of the lattice constants are reported as '0.01, 0.02, 0.02, 0.01, 0.02, and 0.02 Å from experiments and DFT data', but it is unclear whether these are maximum deviations, mean absolute deviations, or deviations from a specific reference; please specify.","section":"Section 3.1"}],"recommendation":"major_revision","confidential_remarks":"The paper is within scope for a computational physics journal and the hybrid method is a sensible extension of the authors' prior work. The main issue is the mismatch between the strongly worded abstract and the paper's own bulk-modulus benchmarks; this is fixable by either improving the potential or carefully qualifying the claims. I would also encourage the editor to ensure that the Supporting Information actually contains the quantitative PBE-vs-HSE comparison for MoTe2 and WTe2, since that is the main methodological risk for the Te-containing systems."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a useful methods paper with a real overstatement in the abstract. The genuinely new content is the Te-based ILP parameters, the systematic six-TMD benchmark, and the direct NEP-D3 comparison. Those are worth having.\n\nWhat the paper does well: the hybrid NEP+ILP partition is sensible. Training NEP on monolayers only and adding a registry-dependent interlayer potential cleanly avoids the long-range vdW problem in local MLPs. The benchmarks are broad: lattice constants, phonons, moiré reconstruction, in-plane and cross-plane thermal conductivity. The thermal conductivities look like genuine predictions—they are not fit to heat-transport data—and the cross-plane values agree well with experiment. The NEP-D3 comparison is fair and informative: it shows NEP-D3 overbinds the interlayer and gets the interlayer distance wrong, which is a real argument for ILP-style hybrids.\n\nThe soft spots are real but not fatal. The bulk modulus underestimates are large (30–59%) and the abstract's claim that the framework 'accurately predicts' bulk modulus is not supported by the paper's own numbers. The paper admits the underestimate but then calls the overall accuracy 'reasonable.' That characterization is defensible for lattice constants and thermal conductivity, but not for bulk modulus. The stress-test note is right that this is a load-bearing overstatement, not a cosmetic one. Also, the Te ILP parameters rest on PBE+MBD-NL rather than HSE+MBD-NL, with the justification deferred to the SI. The reader flags this as the weakest assumption. I agree it is a weaker point, but I would not call it fatal: PBE typically underestimates binding, and the lattice constants and phonons for MoTe2/WTe2 look consistent with experiment. Still, the authors should show the PBE-vs-HSE comparison for interlayer binding and sliding energies in the main text or make the SI comparison compelling.\n\nReproducibility is the other real soft spot. The ILP parameters are linked to a docs page and the code lives in a master branch, but there is no commit hash, no NEP parameter files, no training data release. For a methods paper, that is not acceptable; reviewers should ask for it.\n\nBottom line: this deserves a serious referee report, but the authors need to either soften the bulk modulus claim or explain why the 30–59% errors are acceptable for the target applications. I would not cite the bulk modulus numbers, but I would cite the framework for thermal transport.","headline":"Solid six-TMD benchmark for a hybrid NEP+ILP approach, but the abstract oversells the bulk modulus accuracy and the reproducibility artifacts are incomplete.","tokens_in":23986,"tokens_out":1892,"would_cite":true,"duration_ms":19270,"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":"A hybrid machine-learning plus interlayer potential matches DFT for TMD heat transport.","keywords":["machine-learned potentials","neuroevolution potential","registry-dependent interlayer potential","transition metal dichalcogenides","moiré superlattices","thermal transport","van der Waals heterostructures"],"falsifier":"Recompute the ILP reference binding-energy curves and sliding potential energy surfaces for MoTe2 and WTe2 with HSE+MBD-NL and rerun the Section 3 benchmarks; if the equilibrium interlayer distances or sliding corrugations shift by more than the spread quoted in Supporting Information Section 2, the reported c0 values, bulk moduli, and cross-plane conductivities for these two compounds are biased. An experimental check is X-ray compression data: measured c0(P) of MoTe2 and WTe2 should match the predicted 14.29 Å and 14.34 Å spacings and the Murnaghan moduli of 24.5 and 23.8 GPa.","tokens_in":22912,"feed_emoji":"🌡️","tokens_out":4810,"duration_ms":46852,"temperature":0.7,"pith_summary":"This paper claims that the hybrid NEP+ILP framework, pairing a machine-learned neuroevolution potential for intralayer covalent bonds with a registry-dependent anisotropic interlayer potential for van der Waals interactions, can describe six layered TMDs (MoS2, MoSe2, MoTe2, WS2, WSe2, and WTe2) at near-quantum-mechanical accuracy while remaining fast enough for molecular dynamics. It reports agreement with experiment and DFT on intralayer and interlayer lattice constants, bulk modulus, moiré reconstruction in twisted bilayers, phonon spectra, and both in-plane and cross-plane thermal conductivities. The central point is to escape the tradeoff between DFT accuracy and classical-force-field speed, making large moiré superlattices and thermal device simulations practical.","feed_headline":"Hybrid ML plus interlayer potential matches DFT for TMD heat transport","feed_subtitle":"Six van der Waals crystals get accurate lattice constants, phonons, and in- and cross-plane conductivity at MD speed.","key_machinery":"The load-bearing object is the total energy $E_{\\text{tot}}^{\\text{NEP+ILP}}$, which splits pairwise interactions between layers: the neuroevolution potential handles intralayer chemistry through radial Chebyshev descriptors and angular spherical-harmonic descriptors up to degree $l=4$ with a 5 Å cutoff, while the registry-dependent interlayer potential handles van der Waals interactions with an anisotropic repulsive term that depends on lateral registry distances $\\rho_{ij}$ relative to the local layer normals and a damped $C_6/r^6$ attraction out to a 16 Å cutoff. That registry dependence is what lets the model distinguish AA, AB, and saddle stacking, so it carries the interlayer sliding energy surface and the moiré reconstruction; the NEP carries phonon and anharmonic properties.","core_discovery":"The central discovery is that decoupling intralayer and interlayer interactions—training the NEP on monolayer configurations only, and parameterizing the ILP on bilayer binding-energy curves and sliding potential energy surfaces—preserves accuracy while cutting training data. With this split, the NEP+ILP force field reproduces room-temperature in-plane conductivities (e.g., 150.4 W/m/K for monolayer MoS2 and 216.2 W/m/K for WS2), cross-plane bulk values that align with measurements (4.1 and 6.1 W/m/K, respectively), phonon dispersions close to PBE+MBD, and the triangular and hexagonal moiré reconstruction patterns seen in twisted TMD experiments. For the telluride systems the ILP parameters rest on PBE+MBD-NL reference data because HSE+MBD-NL is too costly; the paper argues the difference is small and benchmarks it in the supporting information. A head-to-head comparison for MoS2 against the NEP-D3 alternative shows the hybrid method captures interlayer binding and sliding energy surfaces with near-HSE+MBD-NL fidelity, while NEP-D3 overestimates them.","pith_inferences":["If the PBE-vs-HSE assumption holds, the same workflow should extend cleanly to other telluride or heavier chalcogenide pairs; if it fails, the MoTe2 and WTe2 predictions offer a clear way to detect the bias by measuring interlayer spacing under pressure.","The framework's registry sensitivity invites a connected prediction not made in the paper: the cross-plane thermal conductance of a twisted TMD bilayer should be tunable by sliding one layer at fixed twist angle, since the sliding potential corrugation is captured accurately.","The NEP-D3 comparison suggests a general design principle for dispersion-corrected machine-learned potentials: including sliding configurations at the true equilibrium interlayer distance in the training set is necessary for transferability."],"forward_implications":["Researchers can simulate micron-scale twisted TMD stacks with DFT-level forces, allowing direct prediction of how moiré reconstruction alters interfacial thermal conductance with twist angle.","The decoupled training strategy means adding a new TMD pair requires only monolayer DFT frames plus a few bilayer energy curves, drastically lowering the cost of building accurate potentials across the MX2 family.","Cross-plane conductivity values near experimental data imply the interlayer potential is reliable for through-plane phonon transport, so the method can be used to design van der Waals thermal metamaterials by stacking and twisting layers.","The MoS2 comparison indicates NEP+ILP outperforms NEP-D3 in sliding-energy fidelity, making it an appropriate base for friction and superlubricity simulations of TMD interfaces."],"supporting_citations":[{"why":"Introduces the hybrid NEP+ILP framework that this work extends to TMD homostructures.","marker":"[38]"},{"why":"Establishes the anisotropic interlayer potential for MoS2 with HSE+MBD-NL reference data used here for comparison.","marker":"[24]"},{"why":"Provides the earlier anisotropic ILP parameterization for group-VI TMDs that the present ILP builds on.","marker":"[53]"},{"why":"Defines the neuroevolution machine learning potential used for intralayer interactions.","marker":"[62]"},{"why":"Presents the NEP-D3 approach compared against NEP+ILP for MoS2 in this paper.","marker":"[33]"},{"why":"Supplies the homogeneous nonequilibrium molecular dynamics method used to compute thermal conductivities.","marker":"[64]"},{"why":"Provides the nonlocal many-body dispersion correction used in the PBE+MBD-NL reference calculations.","marker":"[75]"},{"why":"Supplies the DFT reference data for lattice constants and bulk moduli used in the benchmark comparisons.","marker":"[82]"}],"fun_headline_variants":["Hybrid ML-ILP force field matches DFT for TMD heat transport","Split ML and ILP yields DFT-grade TMD thermal conductivity","Machine learning plus registry potential: TMD heat at DFT accuracy","Interlayer potential + ML achieves DFT-level TMD heat flow"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that PBE+MBD-NL density functional theory can stand in for the more expensive HSE+MBD-NL when fitting the interlayer potential of MoTe2 and WTe2, so if the two functionals diverge more than the supporting comparison indicates, every computed property for those two materials inherits that bias.","fun_headline_variants_meta":{"raw":{"variants":["Hybrid ML-ILP force field matches DFT for TMD heat transport","Split ML and ILP yields DFT-grade TMD thermal conductivity","Machine learning plus registry potential: TMD heat at DFT accuracy","Interlayer potential + ML achieves DFT-level TMD heat flow"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000278,"raw_usage":{"total_tokens":1680,"prompt_tokens":1000,"completion_tokens":680,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":616,"completion_tokens_details":{"reasoning_tokens":606}},"tokens_in":616,"tokens_out":680,"duration_ms":6346,"temperature":1.0,"reasoning_tokens":606,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T04:44:12.443034+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the ILP reference binding-energy curves and sliding potential energy surfaces for MoTe2 and WTe2 with HSE+MBD-NL and rerun the Section 3 benchmarks; if the equilibrium interlayer distances or sliding corrugations shift by more than the spread quoted in Supporting Information Section 2, the reported c0 values, bulk moduli, and cross-plane conductivities for these two compounds are biased. An experimental check is X-ray compression data: measured c0(P) of MoTe2 and WTe2 should match the predicted 14.29 Å and 14.34 Å spacings and the Murnaghan moduli of 24.5 and 23.8 GPa.","supporting_citations":[],"review_version":1}