In this study, we propose an effective strategy for selecting alloying elements to suppress hydrogen diffusion in γ-uranium (γ-U) based on the first-principles investigation of the Niobium (Nb) influences on hydrogen diffusion behavior. The simulation results show that the substitution of Nb in the body-centered cubic (bcc) lattice of γ-U significantly reduces the hydrogen diffusion rate, driven by two key factors: the thermodynamic stabilization of the γ-U bcc lattice and Nb’s strong hydrogen trapping effect. Diffusion energy pathway and electronic structure analyses reveal the presence of energy wells around Nb atoms, causing hydrogen to form cage-like diffusion pathways centered on Nb atoms, which effectively restricts long-range hydrogen diffusion in γ-U. Although Nb’s hydrogen trapping ability decreases at higher hydrogen concentrations, it still plays a crucial role in preventing the nucleation of UH3. Based on these findings, we propose a strategy for predicting hydrogen diffusion kinetics in a series of U-X (X = Ti, Tc, Nb, Mo, Re, Zr, In, Tl) alloys using first-principles static calculations, and establish a near-linear correlation between diffusion energy barriers, X-H bond lengths, and alloy formation energies. Our study underscores the importance of first-principles calculations in selecting suitable alloying elements to regulate hydrogen diffusion in uranium alloys, offering valuable insights with significant implications for engineering applications.

The structural prediction of metal nanoclusters is hindered by the extremely complex potential energy surface and the prohibitive cost of first-principles calculations. Here, we develop an efficient structure-prediction framework that tightly integrates machine-learning interatomic potentials with global optimization. Neural network atomic potentials are iteratively trained to achieve density-functional-theory accuracy and coupled with a genetic algorithm to enable reliable exploration of complex energy landscapes. As a stringent benchmark, the framework is applied to neutral Aun clusters (n = 30-45), where it robustly identifies low-energy structures at an affordable computational cost and reveals a non-monotonic structural evolution from hollow cage-like motifs to multi-core-cage building blocks over a critical size range. Notably, this transition exhibits pronounced differences from that of the corresponding anionic clusters, highlighting the potential of the proposed active-learning workflow as an extensible strategy for investigating metal clusters with complex electronic structures.

Thermal management at Si/AlN heterointerfaces is essential for high-power electronics, yet the atomic-level mechanisms remain unclear. Here, we develop a physics-informed neural network atomic potential (NNAP) to uncover bonding-dependent thermal transport mechanisms of Si/AlN heterointerfaces. Non-equilibrium molecular dynamics simulations predict thermal conductivities of 148.4 W center dot m-1 center dot K-1 for Si and 208.8 W center dot m-1 center dot K-1 for AlN, consistent with reported density functional theory (DFT) and experimental values. We quantitatively map the Al-rich and N-rich chemical potential windows that stabilize Si-Al versus Si-N terminations at the (111)Si// (0001)AlN interface, offering a practical guideline for tuning the interfacial bonding type via growth conditions. The interfacial thermal conductance (ITC) of Si-N bonding interface is 591.2 MW center dot m- 2 center dot K-1, which is 2% higher than Si-Al interface. This enhancement is attributed to the stronger covalent bonding at the Si-N interface, which promotes phonon delocalization and improves phonon coupling, thereby facilitating more efficient heat transfer dominated by low-frequency phonons. In contrast, the Al defect layer at the Si-Al interface suppresses ITC by weakening vibrational matching across the interface and enhancing acoustic-optical phonon scattering. These findings offer atomic-level insights into tailoring interfacial bonding to optimize thermal transport and provide design principles for efficient thermal management in advanced electronic devices.

The hydrogenation-dehydrogenation (HDH) process is a cornerstone for the cost-effective production of high-quality titanium powder, yet its efficiency is traditionally constrained by the conventional assumption that effective embrittlement requires full conversion to stoichiometric delta-TiH2. This conventional route is inherently energy-intensive and time-consuming because it relies on high hydrogen uptake. Here, we propose a shift to stress-dominated embrittlement mechanism enabled by targeted multiphase microstructure design. Under low-pressure (<= 0.1 MPa) and short hydrogenation durations (<60 min), a unique three-phase architecture comprising alpha-Ti, a metastable body-centered tetragonal (BCT) hydride, and face-centered cubic (FCC) delta hydride was engineered. We demonstrate that the pronounced thermophysical mismatch among these coexisting phases generates substantial internal stress gradients, these stresses become the primary driving force for crack initiation and fast propagation, enabling efficient fragmentation at hydrogen concentrations far below those required for full delta-hydride saturation. Compared with traditional protocols, this strategy reduces total hydrogen consumption by similar to 50% and shortens processing time by >80%. This work provides an energy- and time-efficient route to titanium powder, and establishes a microstructure design framework for controlling hydride-induced failure in hexagonal close-packed (HCP) metals through phase-interaction-mediated stress engineering.

Conventional wisdom holds that hard grain-boundary (GB) precipitates embrittle structural alloys by acting as crack initiation sites. In this work, we overturn this paradigm through atomic-scale interfacial engineering, transforming brittle GB phases into ductility pathways in a machine-learning identified model complex concentrated alloy. By precisely tailoring thermomechanical processing, we fabricated compositionally and structurally graded interfaces (GIs) that enable sequential plasticity activation and coordinated deformation across GBs. This interfacial architecture converts an intrinsically brittle multi-phase alloy into a ductile material, achieving an exceptional yield strength of ∼1.2 GPa with a total elongation of ∼20%. The achieved strength-ductility synergy, realized via interfacial plasticity programming, establishes a generalizable materials design strategy to overcome the persistent challenge of GB embrittlement in precipitation-strengthened alloys.

Tungsten (W) plasma-facing components in ITER will undergo neutron irradiation, inducing transmutation to rhenium (Re) and osmium. These impurities critically alter microstructural evolution, but modeling this requires interatomic potentials with quantum-mechanical accuracy, a long-standing challenge for the W-Re system. We present a neural network atomic potential (NNAP). Trained on 43,603 ab-initio calculations, this potential achieves high accuracy on the test set, with mean absolute errors of 6.016 meV/atom for energy and 0.118 eV/& Aring; for forces, respectively. The potential uniquely reproduces not only fundamental properties (lattice parameters, defect energies, migration barriers) but also the correct thermodynamic hierarchy of competing body-centered cubic, hexagonal close packed, sigma-, and chi-phases, matching experimental phase boundaries. This fidelity enables predictive simulations beyond ab-initio scales. Large-scale Monte Carlo reveals a pronounced Re segregation to void surfaces, that is, a mechanism inaccessible to standard potentials but consistent with experiment, demonstrating how the NNAP captures coupled defect-impurity behaviors critical for irradiation response. The NNAP establishes a robust framework for modeling transmutation-driven evolution in W-based materials, providing a pathway to connect atomistic mechanisms with component-scale performance under fusion conditions.

Ultrastable glasses produced by physical vapor deposition exhibit remarkable thermodynamic and kinetic stability, yet their atomic-level formation mechanisms remain unclear. Using molecular dynamics simulations of the deposition of Zr-Cu-Al metallic glass films, we reveal that the growing surface exhibits diffusion coefficients one to three orders of magnitude higher than those of deposited film surface and shows stronger dynamic heterogeneity than the supercooled bulk when compared at identical diffusion coefficients. Notably, at the optimal deposition temperature, the timescale of dynamic heterogeneity reaches a maximum, while the spatial extent of correlated motion is minimized, indicating that more localized surface motions facilitate the formation of more stable glasses. In addition, the growing surface exhibits collective vibration modes, which efficiently dissipate the excess energy of newly deposited hot atoms into the film and further promote atomic mobility. These unique deposition-induced surface dynamics play a critical role in forming ultrastable glass films.

Elucidating the atomic-scale mechanisms governing shock-induced plasticity in metals under extreme conditions requires highly accurate descriptions of far-from-equilibrium states. Here, we develop a high-fidelity machine-learning potential for aluminum—the Neural Network Atomic Potential (NNAP)—trained on a first-principles dataset explicitly enriched with high-density configurations. The NNAP demonstrates near first-principles accuracy in reproducing phonon spectra and the overall features of pressure-temperature phase diagram from ambient conditions to extreme regimes. Employing the NNAP in large-scale non-equilibrium molecular dynamics simulations, we successfully reproduce the nanoscale merging trend of elastic and plastic waves observed in experiments. More importantly, these simulations reveal a pressure-dependent transition in deformation mechanisms: at lower pressures, plasticity is dominated by deformation twinning; whereas at higher pressures, a dynamic competition and reversible interconversion between twinning and a martensitic fcc-to-hcp transformation emerges. Generalized stacking fault energy calculations trace this mechanistic crossover to a pressure-induced shift in the relative energy barriers for twinning versus phase transformation. Our work not only provides a definitive atomistic explanation for shock-induced plasticity in aluminum but also establishes a reliable, data-driven framework for exploring complex material behaviors under extreme dynamic loading.

Abstract Body-centered cubic (BCC) metals often exhibit limited ductility at finite temperatures, and elemental synergy in multi-principal element alloys (MPEAs) offers a promising route to overcome this limitation. Here, we employ atomistic simulations with a high fidelity machine-learning potential—which outperforms traditional empirical potentials at capturing complex chemical interactions—to investigate mode-I crack propagation in the engineering-relevant NbMoTaW MPEA versus pure Mo. While Mo fails by brittle cleavage, the NbMoTaW undergoes ductile fracture mediated by dislocation activity. We demonstrate that the average-based Rice criterion demands local corrections to sample diverse chemical environments along the advancing crack. Specifically, the MPEA's intrinsic chemo-mechanical heterogeneity produces atomic-scale stress delocalization at the crack tip, generating a chemically-driven shielding effect that mitigates the singular stress concentration responsible for cleavage. In particular, Ta-related local chemical heterogeneities act as precursors that nucleate localized shear and dissipate plastic energy. By embedding such heterogeneities into a brittle Mo matrix, we demonstrate that a sufficient level of interconnectivity is required to amplify stress delocalization and promote dislocation nucleation, enabling a brittle-to-ductile transition of the fracture mode. These findings reveal a stress delocalization mechanism by which local chemical heterogeneities control intrinsic ductility in BCC alloys and provide a theoretical guide for composition design.

This work develops a high-accuracy artificial neural network (ANN) potential for osmium (Os) to enable large-scale irradiation damage simulations in fusion materials. The potential employs spherical harmonic-Chebyshev polynomial descriptors within a Behler-Parrinello neural network architecture, trained on an extensive dataset generated via density functional theory and ab initio molecular dynamics. Comprehensive validations demonstrate excellent agreement with reference calculations and experimental data across multiple properties: lattice constants of diverse crystal structures, elastic constants for hexagonal close-packed Os, dimer interactions, and defect formation energies (vacancies, interstitials, surfaces). The ANN potential accurately reproduces key behaviors under extreme conditions, including melting characteristics, sputtering thresholds, and primary knock-on atom collision cascades. Simulations reveal defect evolution and clustering during radiation events. This transferable potential provides a critical computational tool for investigating Os precipitation effects on tritium retention and irradiation hardening in tungsten-based plasma-facing materials for fusion reactors.

Crystal nucleation critically governs the glass-forming ability of metallic melts, yet remains elusive due to the nanoscale size of nuclei and longtime timescales involved. While forward flux sampling (FFS) enables molecular dynamics (MD) studies of nucleation, complex crystallization pathways in metallic glass-forming liquids often compromise reliability. To address this, we introduce multi-type averaged bond-orientational order parameter (Multi-S6) optimized via cross-entropy, enabling robust differentiation between liquids and various crystalline phases in widely studied supercooled Cu-Zr model systems. We further develop an MPI-support parallel FFS program that combines temporally coarse-grained and jumpy FFS algorithms, dramatically improving stability and sampling efficiency. This approach achieves the first quantitative measurement of Laves phase nucleation rates. Meanwhile, this methodology is also applicable to Ni-Al system, enables accurate determination of Ni50Al50 nucleation rates near the melting point, which previously accessible only by extrapolation from classical nucleation theory using classical MD data at deep supercooling. These results demonstrate that Multi-S6 surpasses conventional order parameters for both simple and complex phases, while parallel FFS enhances computational efficiency. Our methodology offers unprecedented atomic-scale insights into nucleation mechanisms in glass-forming systems and is broadly applicable to complex nucleation processes.

Grain boundaries (GBs) in multi-principal element alloys (MPEAs) exhibit complex structural and chemical evolution under thermomechanical stimuli, yet the atomic-scale mechanisms coupling solute segregation, GB structural reconfiguration, and migration remain poorly understood. In this work, we employ hybrid Monte Carlo/molecular dynamics simulations combined with a first principles-accurate interatomic potential to investigate segregation-induced grain boundary dynamics, using a commercial Au–Ag–Cu alloy as a model system. Our results show that inhomogeneous Ag segregation consistently drives GBs toward lower-energy structural reconfigurations, including a reversible faceting–defaceting transition regulated by solute saturation. These transformations arise from the synergistic reduction of elastic strain energy and interfacial energy, which reshapes the local elastic fields at GBs and in turn controls the spatial partitioning of solutes. The GB dynamics exhibit two distinct modes depending on the distribution of Ag on either side of the boundary: one-side-dominant segregation creates a chemical potential imbalance, driving directional GB migration and eventual merging with phase boundaries; whereas segregation on both sides stabilizes GB within the Ag-rich precipitates. On this basis, we further propose a structure-chemistry interplay mechanism governing the evolution of asymmetric tilt GBs: the intrinsic structural asymmetry dictates the initial segregation preference, while local chemical potential imbalances provide the key driving force for the chemically actuated GB migration.

Silicon nitride (Si3N4) is a strong, thermally stable covalent ceramic that is typically regarded as brittle with limited deformability. A recent experimental study combined with density functional theory (DFT) indicates that the α/β interface undergoes a β→α transformation via sliding followed by bond-switching, suggesting a pathway to achieve plasticity, but DFT’s spatiotemporal reach prevents a full mechanistic picture. Here, we develop a physics-informed high-accuracy neural network interatomic potential (NNAP) model with DFT-level accuracy for the phase transformations and use it to perform large-scale atomistic simulations. NNAP-guided simulations show that structural relaxation during the relative sliding between α- and β-phases at the interface triggers pronounced atomic-layer rearrangements and lowers the energy barrier by nearly 60

Enhancing the kinetic stability of glasses typically requires deepening their thermodynamic stability, which increases structural rigidity and degrades ductility; decoupling these properties remains a major challenge. Here, we demonstrate that spatial patterning in metallic glasses produces exceptional kinetic ultrastability that coexists with a thermodynamically metastable, high-energy state and excellent plasticity. Guided by atomistic simulations using replica exchange molecular dynamics and machine learning interatomic potentials, we reveal that oxygen, through reaction–diffusion-coupled pattern dynamics, self-organizes into oxygen-centered pinned structures (OPSs) that serve as localized kinetic constraints. These motifs drastically slow structural relaxation, delivering kinetic stability comparable to ultrastable glasses even as the system retains the high inherent energy of rapidly quenched states. The OPSs’ topology yields a spatially uniform activation of plastic events, promoting strain delocalization under mechanical load. By geometrically tailoring oxygen patterns, we increase the glass transition onset temperature ( T _onset ) by about 200 K with negligible loss of deformability. Our findings establish a practicable paradigm for decoupling kinetic and thermodynamic stability and point to a scalable, additive route for designing amorphous materials that combine hyperstability with plasticity.

As advanced amorphous materials with superior mechanical, physical, and chemical properties, metallic glasses (MGs) hold significant promise for a wide range of applications. However, their rational design and precise property control have long been impeded by notable challenges. These obstacles largely arise from the inherent compositional and structural complexity of MGs, which not only slows empirical trial-and-error experimentation but also limits the scalability of computationally intensive first-principles simulations. In recent years, machine learning has emerged as a transformative tool, offering unprecedented capabilities to decode these intricate relationships and overcome conventional research limitations. Here we provide a systematic overview of machine-learning-guided investigations of MGs and their associated data pipelines, centering on two key paradigms: the feature-driven ‘Keplerian’ data pipeline and a next-generation theory-experiment-aligned pipeline designed to close the gap between simulation and experiment. By breaking down both paradigms into modular workflow components, we underscore the essential roles of data standardization, interpretable feature engineering, and physics-informed validation in constructing a reliable research framework. We anticipate that a sufficiently robust, efficient, and generalizable data pipeline will not only unlock novel scientific insights and accelerate material discovery but also propel the field from a ‘trial-and-error’ approach toward an era of intelligent and principled design.

Cation-disordered AgBiS2 is a promising lead-free optoelectronic material, but both its ordered structure and the microscopic origin of its favorable electronic properties remain debated. Theory has proposed a mixed-coordination tendency with tetrahedral AgS4 and octahedral BiS6 units, whereas experiments mainly report octahedrally coordinated ordered and cation-disordered phases, together with local cation off-centering. Here, we combine a machine-learning interatomic potential with a deep-learning Hamiltonian to resolve the coupled structural and electronic evolution of AgBiS2 at large length scales. We identify the three-dimensional Bi-S network as the central structural motif governing both disorder stability and band-edge electronic states. At weak disorder, Ag/Bi exchange competes with the off-centering tendency of the Ag sublattice, producing strongly distorted local environments and convoluted diffraction signatures that hinder the identification of the ordered phase. With increasing disorder, BiS6-like units connect into a continuous Bi-S network, which stabilizes the rocksalt-like disordered phase. Despite strong cation disorder, AgBiS2 retains clear semiconductor-like band dispersion and develops a direct band gap. The connected Bi:p-S:p states supported by the Bi-S network preserve a dispersive conduction-band edge and a small electron effective mass. In contrast, mobile Ag disrupts the long-range periodicity of Ag-S bonding, leading to strongly localized valence states. These results clarify the structural controversy in ordered AgBiS2 and establish a unified physical picture of disorder stability and optoelectronic response in nonisovalent semiconductor alloys.

Shear band propagation and interaction are critical to the mechanical performance of metallic glasses and are strongly governed by thermal history, yet their microscopic mechanisms remain unclear. Here, using molecular dynamics simulations combined with a state-of-the-art annealing protocol, we systematically investigate these behaviors in a model metallic glass across effective quenching rates spanning six orders of magnitude. Through a double-notch model, we show that the normalized interaction distance relative to the single shear band width is significantly larger in slowly quenched samples than in rapidly quenched ones. Atomic-scale analysis reveals that rapidly quenched samples exhibit a high density of pre-existing soft regions, which trigger correlated shear transformation zones through local vortex fields, resulting in propagation path locking and weak inter-band coupling. In contrast, slowly quenched samples exhibit enhanced structural heterogeneity and a right-shifted activation energy spectrum, promoting a single large-scale vortex field ahead of the shear band front. This field facilitates long-range stress transmission and induces shear band deflection, convergence, and coalescence, a transition resembling a "shielding effect" in fracture mechanics, where vortex-mediated disturbances destabilize the advancing shear band front. Our findings establish a direct microscopic connection between glass stability and shear-band-mediated plasticity and suggest that regulating shear band interactions offers a promising route to enhance the room-temperature ductility of metallic glasses.

Establishing a quantitative structure-property relationship is essential for the development and design of new materials. However, this approach faces significant challenges in amorphous materials, where even a quantitative description of atomic structure is nearly impossible. In this study, we examined the packing characteristics of atoms based on their contributions to excess low-frequency vibrational modes in a model metallic glass. Our investigation spans more than eight orders of magnitude in effective cooling rates, ensuring the exploration of a broader range of thermal history states and their associated properties. We found that atoms with smaller contributions tend to cluster spatially, while those with larger contributions form branched, quasi-twodimensional structures with fractal characteristics. As a result, the critical fraction of atoms required to form a percolated network is significantly lower for high-contribution atoms than for low-contribution ones. In both types of networks, the correlation between connectivity and contribution follows an exponential relationship, with higher sensitivity in networks composed of large-contribution atoms. As the system's energy decreases, the intensity of the low-frequency excess peak diminishes, yet the critical fraction of atoms remains constant, irrespective of whether the networks are composed of high- or low-contribution atoms. This reveals a hidden topological invariance in the atomic packing features of metallic glasses.

Ductilizing amorphous metals without sacrificing strength is challenging due to unclear plasticity carriers. This study attempts to mimic the microalloying strategy of physical metallurgy in computer simulations by selectively pinning a small fraction of typical atoms in metallic glass, which is targeted to efficiently optimize the mechanical properties. We found that pinning atoms with high participation in the low-frequency vibrational modes are more effective in strengthening, attributing to a mechanism of scale-dependent pinning effect. By pinning only 2 % atoms in the unstable glassy samples, one can achieve shear modulus comparable to samples prepared with cooling rates that are eight orders of magnitude slower, highlighting the validity of microalloying over thermal treatment. Moreover, this microalloying approach not only control elastic properties, but also mitigates the failure mode of metallic glass. It demonstrates that restricting the motion of atoms in regions external to the shear band plays a critical role in inhibiting the propagation of the shear band.

The functional properties of glasses are governed by their formation history and the complex relaxation processes they undergo. However, under extreme conditions, glass behaviors are still elusive. In this study, we employ simulations with varied protocols to evaluate the effectiveness of different descriptors in predicting mechanical properties across both low- and high-pressure regimes. Our findings demonstrate that conventional structural and configurational descriptors fail to correlate with the mechanical response following pressure release, whereas the activation energy descriptor exhibits robust linearity with shear modulus after correcting for pressure effects. Notably, the soft mode parameter emerges as an ideal and computationally efficient alternative for capturing this mechanical behavior. These findings provide critical insights into the influence of pressure on glassy properties, integrating the distinct features of compressed glasses into a unified theoretical framework.