Universal machine-learning interatomic potentials (uMLIPs) are emerging as foundation models for atomistic simulation, offering near-ab initio accuracy at far lower cost. Their safe, broad deployment is limited by the absence of reliable, general uncertainty estimates. We present a unified, scalable uncertainty metric, U, built from a heterogeneous ensemble that reuses existing pretrained MLIPs. Across diverse chemistries and structures, U strongly tracks true prediction errors and robustly ranks configuration-level risk. Using U, we perform uncertainty-aware distillation to train system-specific potentials with far fewer labels: for tungsten, we match full density-functional-theory (DFT) training using 4% of the DFT data; for MoNbTaW, a dataset distilled by U supports high-accuracy potential training. By filtering numerical label noise, the distilled models can in some cases exceed the accuracy of the MLIPs trained on DFT data. This framework provides a practical reliability monitor and guides data selection and fine-tuning, enabling cost-efficient, accurate, and safer deployment of foundation models.

Enhancing the hydrogen embrittlement (HE) resistance of alloys caters to the urgent needs of engineering safety and long-distance hydrogen transportation. Highly dense precipitates in the alloys act as H traps, however, some of them cannot strongly trap H thus failing to prevent its accumulation at the critical regions. Experimentally, it is challenging to expeditiously identify and generate phases causing strengthening and acting as strong H traps. Here, we demonstrate a computation-based design strategy to generate precipitates strongly trapping H. Based on the quantum machine learning Al-Sc-Cu potential, the optimal processing parameters of strong H trapping phases are determined, even though they are metastable in nature. Elemental mapping in electron microscope and atom probe tomography confirms the presence of Cu in Al₃Sc and its strong interaction with H. Hence, we envisage the proposed strategy will accelerate the design of HE-resistant microstructures of various technologically relevant materials via identification of desirable phases.

Carbide particles act as effective hydrogen traps that can mitigate hydrogen embrittlement in steels, with their trapping efficacy governed by the binding energy of hydrogen at atomic-scale sites within the carbide and at the carbide/matrix interface. Within the carbide, carbon vacancies are established trap sites, while at the carbide/matrix interface, multiple distinct trap sites may form depending upon the interface character. State-of-the-art experimental methods cannot attribute binding energies to individual sites unambiguously, particularly at the carbide/matrix interface, where distinct trap sites may exhibit similar binding energies. Density functional theory (DFT) calculations have partially addressed this limitation, but their computational cost prohibits accurate treatment of semicoherent and incoherent interfaces, for which hydrogen binding energies remain poorly characterised. In this work, we develop a DFT-accurate machine-learning Fe–TiC–H interatomic potential that enables simulations of hydrogen energetics and kinetics at various TiC/Fe interfaces in the Baker–Nutting orientation. Predictions of structural, dynamic and energetic properties of both bulk phases (i.e., bcc Fe and rocksalt TiC) and the coherent interface with and without hydrogen are in good agreement with DFT. We apply the potential to examine the atomic structure and hydrogen trapping at semicoherent and incoherent interfaces. We find that the binding energy of hydrogen at misfit dislocations is similar to that at coherent interfaces, with a difference of no more than 0.1 eV. However, the misfit strain field reduces carbon vacancy formation energies at the interface, where hydrogen can be trapped with similar binding energies as that in a bulk C vacancy (approximately –1 eV). Lower vacancy formation energies are also found at the incoherent interface, where hydrogen binding energies span a broader range, depending sensitively on the local atomic environment. Overall, the results enable interpretation of experimental observations of hydrogen segregation around semicoherent and incoherent TiC interfaces.

Iron oxidation is a complex process involving critical atomistic events, such as atomic adsorption, diffusion, and surface reconstruction, understanding of which is significant for both surface science and coating technology. Atomistic simulation serves as an useful tool to investigate the processes, where description of interatomic interactions is required. However, selecting appropriate force field or interatomic potential is not only difficult, but also essential for getting accurate result. In this work, we present a detailed benchmark of reactive force fields (ReaxFFs) and universal machine learning interatomic potentials (uMLIPs) against density functional theory (DFT) calculations of oxygen adsorption on various α-iron surfaces, which is the first yet crucial step towards oxidation. The comparisons show the coverage-dependent performance and improvable accuracy of both ReaxFFs and uMLIPs at reproducing DFT results, with ReaxFFs outperforming uMLIPs. Subsequently, iron oxidation is simulated using ReaxFF and uMLIP. The results reveal the strong capability of ReaxFF and poor stability of uMLIP for describing reactive process, i.e., the formation of iron oxide. This may be attributed to the suitable functional form of ReaxFF for the description of bond changes. The insights presented here not only provide an example of benchmarking force field or interatomic potential for system of interest, but also highlight the applicability of ReaxFF and scopes of improvement of uMLIP.

Machine-learning interatomic potentials (MLIPs) with broad chemical flexibility are essential for atomistic simulations of compositionally complex alloys, but their deployment in large-scale molecular dynamics requires a balance among accuracy, efficiency, stability, transferability, and uncertainty quantification. Here, we benchmark two chemically scalable MLIP frameworks, neuroevolution potential (NEP) and graph atomic cluster expansion (GRACE), for 16 elemental metals and their multicomponent alloys. GRACE-FS shows higher training efficiency and generally better average accuracy, chemical transferability, and finite-temperature robustness, whereas UNEP-v1 provides substantially higher inference speed and remains competitive in selected stress and large-error metrics. We further show that chemical transferability is closely linked to high-temperature MD stability in highly multicomponent environments and that ensemble-based uncertainty provides a more reliable error indicator than D-optimality for the heterogeneous systems considered here. Finally, 3×106-atom shock simulations demonstrate that UNEP-v1, combined with ensemble uncertainty, enables uncertainty-aware simulations under extreme dynamic conditions, yielding robust global spall-strength predictions while revealing model sensitivity in local damage pathways. These results provide practical guidelines for selecting and deploying MLIPs in large-scale simulations of multicomponent alloys.

Nitrogen alloying improves the mechanical performance of martensitic stainless steel, while tempering is required to mitigate brittleness and enhance processability. However, tempering-induced microstructural changes markedly influence the semiconducting properties of the passive film, thereby affecting corrosion resistance. This work examines the passive film of high-nitrogen martensitic stainless steel (HNMSS) tempered at different temperatures. Tempering at 200 °C yields the highest pitting potential (461.28 mV), attributed to the transformation of CrN to Cr₂O₃, which enhances film protection. At higher temperatures, abundant Cr-rich nitride/precipitate (M₂N) depletes N and Cr in the matrix. The increase of precipitate/matrix interfaces and defect density in the passivation film impairs pitting corrosion resistance. First-principles calculations and quasi-in-situ scanning Kelvin probe force microscopy reveal that M₂N precipitates exhibit a higher Volta potential and the highest work function than those of the matrix, acting as cathodes to accelerate localized matrix dissolution, which reduces nitrogen incorporation into the passive film. These findings clarify the relationship of tempering, microstructure, and corrosion in the HNMSS.

Group contribution methods (GCMs) provide a practical and computationally efficient approach for predicting thermodynamic properties of hydrocarbons, especially when experimental data are scarce. This review evaluates the evolution of GCMs from classical first-order schemes (e.g. Lydersen method, Joback method) to more advanced second-order frameworks (e.g. CG94, Sharma method), hybrid extensions, and emerging machine learning integrations. While first-order models are simple and widely used, these models struggle with branched and long-chain molecules. Second-order approaches significantly improve structural sensitivity and predictive accuracy, achieving deviations below 2–3% for critical properties and within 1 kcal/mol for formation enthalpies of branched alkanes. Nevertheless, challenges remain in extrapolating to highly complex molecules, underrepresented functional groups, and extreme conditions. Promising directions include reinforcement of second-order GCMs with molecular theory, systematic expansion of experimental and quantum-based datasets, and hybrid GCM–machine learning models that retain interpretability while improving generalisability. We recommend prioritising models that balance accuracy, robustness, simplicity, and transferability to accelerate sustainable process and product designs, particularly in applications such as fuel upgrading including hydroisomerisation, separation processes, and green chemical development.

The development of advanced catalysts with innovative nanoarchitectures is critical for addressing energy and environmental challenges such as the electrochemical CO₂ reduction reaction (CO₂ RR). Herein, the synthesis of an innovative copper–sulfur planar structure, Cu–S–BDC, within a metal–organic framework (MOF) catalyst is presented, which demonstrates 100% selectivity toward formate as the sole carbon product. Structural analysis and surface characterizations reveal that Cu–S–BDC exhibits quasi-2D inorganic building units, with Cu bonded to two S-CH (Formula presented.) groups and one BDC linker, while carboxylate groups adopt a bridging coordination mode. This unique arrangement not only imparts remarkable structural stability but also enhances the electronic properties of the MOF, as evidenced by a narrow bandgap of 1.203 eV that facilitates efficient charge transfer and increased electrochemical current density in CO (Formula presented.) RR. Notably, it offers a Faradaic efficiency of 92% for formate at an overpotential as low as −0.4 V versus the reversible hydrogen electrode (RHE) in an aqueous electrolyte of 1 m KOH, as well as a current density of −25.8 mA cm² at −0.9 V versus RHE, averaged over 24 h of electrolysis. This study highlights a fresh perspective in the field of MOF electrocatalysts by demonstrating that engineering the metal coordination environment can significantly enhance the electronic properties and consequently improve the electrocatalytic performance of these materials.

Multi-principal element alloys (MPEAs) are renowned for their enhanced mechanical strength relative to their constituent metals, as evidenced by various experimental techniques such as tension/compression tests and instrumental indentation. Nevertheless, atomistic simulations sometimes produce conflicting results, casting doubt on the consistently superior mechanical properties of MPEAs. In this study, machine-learning interatomic potentials (MLIPs) with first-principles accuracy were developed for body-centered cubic refractory MoNbTaW MPEAs, enabling systematic atomistic simulations under various deformation scenarios. The new MLIPs are supported by a comprehensive dataset encompassing extensive defects, and the established embedded-atom model (EAM) potential was benchmarked against both this dataset and the new MLIP. Simulations covering diverse compositions confirm that both MLIPs and EAM accurately capture the critical strengthening mechanisms in MoNbTaW MPEAs. It is revealed that MPEAs generally exhibit superior mechanical strength compared to their constituent metals in macroscale specimens, primarily due to solid solution strengthening during dislocation motion. However, at the nanoscale—where plasticity is predominantly governed by dislocation nucleation and grain boundary deformation—the constituent metals may outperform MPEAs. A critical length scale is identified above which MPEAs demonstrate enhanced mechanical strength relative to their constituent elements; below this scale, the advantage diminishes, underscoring a significant size-dependent effect that is crucial for optimizing MPEA applications, particularly at the nanoscale.

Understanding atomic hydrogen (H) diffusion in multi-principal element alloys (MPEAs) is crucial for enhancing hydrogen transport and storage technologies. However, the vast compositional space and complex chemical environments of MPEAs pose significant challenges. We develop highly accurate machine learning force field and neural network-driven kinetic Monte Carlo simulations to investigate H diffusion in body-centered cubic (BCC) MoNbTaW MPEAs. H diffusion exhibits super-Arrhenius behavior in MPEAs, dominated by the low percentile of the H solution energy spectrum. Robust analytical models are derived via machine learning symbolic regression to predict H diffusivity across general BCC MPEAs. Additionally, it is revealed that chemical short-range order (SRO) generally does not impact H diffusion in MoNbTaW MPEAs, except it enhances diffusion when H-favoring elements are present in low concentrations. These insights not only deepen our understanding of H diffusion dynamics in MPEAs but also guide the strategic development of advanced MPEAs for hydrogen-related applications by manipulating element type, composition, and SRO.

One of the most promising energy carriers for transport applications are hydrogen-based energy carriers. NaBH₄ is a hydrogen energy carrier and produces hydrogen bubbles when it is dissolved in water. The formation of hydrogen bubbles hinders experimental measurements of the thermodynamic and transport properties of aqueous NaBH₄ solutions at elevated temperatures. Accurate knowledge of these properties is essential for the NaBH₄ hydrolysis reactor modeling and design. Molecular dynamics (MD) simulations provide the option to study the thermodynamic and transport properties of NaBH₄ aqueous solutions without hindering hydrogen bubble formation. In this work, a new force field is developed for BH₄^-, namely, the Delft force field of BH₄^- (DFF/BH₄^-), which, combined with additional force fields, can accurately describe experimental densities and viscosities of 0 to 5 m (mol salt/kg water) NaBH₄, 0 to 3 m NaB(OH)₄, and 1 m NaOH aqueous solutions at 295 K within 1.8% and 10.8% maximum deviation, respectively. Empirical fitting correlations are created for densities, viscosities, and self-diffusivities obtained from the MD simulations of 0 to 5 m NaBH₄, 0 to 5 m NaB(OH)₄, and 0 to 1 m NaOH aqueous solutions at 295-363 K for NaBH₄ hydrolysis reactor modeling and design purposes.

Photocatalytic water splitting represents a promising approach for sustainable hydrogen production, with two-dimensional Janus materials offering unique advantages through intrinsic electric fields that enhance charge separation. We present a comprehensive first-principles investigation of Janus AlXY₂ (X = Ga, In; Y = S, Se, Te) monolayers using density functional theory and ab initio molecular dynamics simulations. All six systems exhibit excellent structural, thermal, and mechanical stability with HSE06 bandgaps of 2.029–2.969 eV suitable for UV-light absorption. The asymmetric structure generates strong intrinsic electric fields of 5.391–6.437 V perpendicular to the monolayer plane, significantly enhancing photogenerated charge carrier separation. While pristine monolayers show poor hydrogen evolution reaction (HER) activity with Gibbs free energies of 1.937–2.371 eV, strategic introduction of metal vacancies dramatically improves performance, reducing ΔG_H values to −0.371 to ＋0.607 eV and approaching optimal catalytic conditions. These findings demonstrate the potential of defect-engineered 2D Janus AlXY₂ materials for efficient photocatalytic hydrogen production.

Extended defects such as dislocation networks and general grain boundaries are ubiquitous in metals, and accurate modeling these extensive defects is crucial to elucidate their deformation mechanisms. However, existing machine learning interatomic potentials (MLIPs) often fall short in adequately describing these defects, as their large characteristic scales exceed the computational limits of first-principles calculations. To address this challenge, we present a computational framework combining a defect genome constructed via empirical interatomic potential-guided sampling, with an automated reconstruction technique that enables accurate first-principles modeling of general defects by converting atomic clusters into periodic configurations. The effectiveness of this approach was validated through simulations of nanoindentation, tensile deformation, and fracture in BCC tungsten. This framework enhances the modeling accuracy of extended defects in crystalline materials and provides a robust foundation for advancing MLIP development by leveraging defect genomes strategically.

In this study, we explore the mechanisms underlying the exceptional intrinsic strength of face-centered cubic (FCC) Multi-Principal Element Alloys (MPEAs) using a multifaceted approach. Our methods integrate atomistic simulations, informed by both embedded-atom model and neural network potentials, with first-principles calculations, stochastic Peierls-Nabarro (PN) modeling, and symbolic machine learning. We identify a consistent, robust linear correlation between the strength of MPEAs and the standard deviation of the maximum stacking-fault restoring force (τ_max,sd) across various potentials. This finding is substantiated by comparing the experimental strengths of Cantor alloys’ subsystems and Ni_62.5V_37.5 against τ_max,sd values from high-throughput first-principle calculations. Our theoretical insights are derived from integrating the stochastic Peierls-Nabarro model with a shearable precipitation hardening framework, demonstrating that lattice distortion alone does not directly enhance intrinsic strength. Instead, τ_max,sd emerges as a critical determinant, capable of boosting the strength of MPEAs by up to tenfold. Our analysis reveals the critical role of the exponential form of the PN model in achieving substantial strength improvement by transforming the Gaussian-like distribution of τ_max into an exponential-like distribution of local Peierls stress. Additionally, using an advanced symbolic machine learning technique, the sure independence screening and sparsifying operator (SISSO) method, we derive interpretable relationships between MPEA strength, elastic properties, and τ_max statistics, offering new insights into the design and optimization of advanced MPEAs. These findings highlight that the nonlinear physics and atomic fluctuations characterizing MPEAs not only underpin their unconventional intrinsic strength but also contribute to other complex properties such as sluggish diffusion and cocktail effect.