arXiv subjects

Yutack Park

Publications and source records attributed to Yutack Park.

9 recordsLinked to original sources

arXiv2026

SevenNet-Polar for MultiTask Prediction of Energy, Forces, Stress, and Born Effective Charges: Development and Application to ZrO$_2$, Li$_3$PO$_4$, and Perovskites

Accurate prediction of the Born effective charge (BEC) tensor is crucial for modeling materials under electric fields but remains computationally expensive. To bridge this gap, we present SevenNet-Polar, an equivariant graph neural network framework based on the SevenNet architecture for fast and accurate BEC predictions. Our BEC-only predictors can achieve an RMSE as low as 0.0043 e on ZrO$_2$, Li$_3$PO$_4$, and perovskites, despite the presence of high-temperature (up to 2,000 K) and defect-laden training data. Our all-in-one multitask models for predicting energy, forces, stress, and BEC in ZrO$_2$ and Li$_3$PO$_4$ achieve high accuracy with an RMSE of 1.0 meV/atom for energy, 12 meV/angstrom for forces, 0.05 GPa for stress, and 0.0029 e for BEC. BEC accuracy is not degraded by multitask training. Scaling analysis reveals distinct exponents for diagonal and off-diagonal BEC components, both of which exhibit less favorable scaling than energy, force and stress errors. SevenNet-Polar generalizes robustly when tested on scenarios containing structural environments absent from the training set, such as along nudged elastic band (NEB) trajectories or grain boundaries in ZrO$_2$. Accelerated by FlashTP, SevenNet-Polar enables simulations containing up to 1.5 million atoms on multi-GPU supercomputers and up to approximately 15,000 atoms on a single consumer-grade GPU. This makes charge-aware molecular dynamics simulations under electric fields more accessible.

cond-mat.mtrl-sci↗

arXiv2026

A Robust Agentic Framework for Expert-Level Automation of Atomistic Simulations

Traditionally, atomistic simulation has been constrained by the computational scaling limits of ab initio methods and the parameterization overhead of empirical force fields. The recent emergence of universal machine learning interatomic potentials has significantly mitigated these bottlenecks, offering near-quantum accuracy and generalizability across diverse chemical spaces at a fraction of the computational cost. However, this shift has relocated the bottleneck to the human dimension: time-consuming mechanical processes, such as input preparation and data analysis, now dominate the research lifecycle. We introduce Paimon, a Platform for Agentic Integration in Materials Optimization and Nanoscale-simulations. Through hundreds of trials on an expert-level liquid electrolyte simulation, we show that Paimon substantially improves the reliability of agentic workflows by suppressing silent errors: plausible yet physically incorrect results. We further demonstrate that Paimon can cooperate with an external scientific agent and autonomously reproduce simulation methodologies from the literature. As an agent harness for atomistic simulations, Paimon affords researchers a continuous, science-centric workflow throughout the entire simulation lifecycle.

cond-mat.mtrl-sci↗

arXiv2025

Optimizing Cross-Domain Transfer for Universal Machine Learning Interatomic Potentials

Accurate yet transferable machine-learning interatomic potentials (MLIPs) are essential for accelerating materials and chemical discovery. However, most universal MLIPs overfit to narrow datasets or computational protocols, limiting their reliability across chemical and functional domains. We introduce a transferable multi-domain training strategy that jointly optimizes universal and task-specific parameters through selective regularization, coupled with a domain-bridging set (DBS) that aligns potential-energy surfaces across datasets. Systematic ablation experiments show that small DBS fractions (0.1%) and targeted regularization synergistically enhance out-of-distribution generalization while preserving in-domain fidelity. Trained on fifteen open databases spanning molecules, crystals, and surfaces, our model, SevenNet-Omni, achieves state-of-the-art cross-domain accuracy, including adsorption-energy errors below 0.06 eV on metallic surfaces and 0.1 eV on metal-organic frameworks. Despite containing only 0.5% r$^2$SCAN data, SevenNet-Omni reproduces high-fidelity r$^2$SCAN energetics, demonstrating effective cross-functional transfer from large PBE datasets. This framework offers a scalable route toward universal, transferable MLIPs that bridge quantum-mechanical fidelities and chemical domains.

cond-mat.mtrl-sci↗

arXiv2025

Unveiling defect motifs in amorphous GeSe using machine learning interatomic potentials

Ovonic threshold switching (OTS) selectors play a critical role in non-volatile memory devices because of their nonlinear electrical behavior and polarity-dependent threshold voltages. However, the atomic-scale origins of the defect states responsible for these properties are not yet fully understood. In this study, we use molecular dynamics simulations accelerated by machine-learning interatomic potentials to investigate defects in amorphous GeSe. We begin by benchmarking several potential architectures-including descriptor-based models and graph neural network (GNN) models-and show that faithfully representing amorphous GeSe requires capturing higher-order interactions (at least four-body correlations) and medium-range structural order. We find that GNN architectures with multiple interaction layers successfully capture these correlations and structural motifs, preventing the spurious defects that less expressive models introduce. With our optimized GNN potential, we examine twenty independent 960-atom amorphous GeSe structures and identify two distinct defect motifs: aligned Ge chains, which give rise to defect states near the conduction band, and overcoordinated Ge chains, which produce defect states near the valence band. We further correlate these electronic defect levels with specific structural features-namely, the average alignment of bond angles in the aligned chains and the degree of local Peierls distortion around overcoordinated Ge atoms. These findings provide a theoretical framework for interpreting experimental observations and deepen our understanding of defect-driven OTS phenomena in amorphous GeSe.

cond-mat.mtrl-sci↗

arXiv2025

An efficient forgetting-aware fine-tuning framework for pretrained universal machine-learning interatomic potentials

Pretrained universal machine-learning interatomic potentials (MLIPs) have revolutionized computational materials science by enabling rapid atomistic simulations as efficient alternatives to ab initio methods. Fine-tuning pretrained MLIPs offers a practical approach to improving accuracy for materials and properties where predictive performance is insufficient. However, this approach often induces catastrophic forgetting, undermining the generalizability that is a key advantage of pretrained MLIPs. Herein, we propose reEWC, an advanced fine-tuning strategy that integrates Experience Replay and Elastic Weight Consolidation (EWC) to effectively balance forgetting prevention with fine-tuning efficiency. Using Li$_6$PS$_5$Cl (LPSC), a sulfide-based Li solid-state electrolyte, as a fine-tuning target, we show that reEWC significantly improves the accuracy of a pretrained MLIP, resolving well-known issues of potential energy surface softening and overestimated Li diffusivities. Moreover, reEWC preserves the generalizability of the pretrained MLIP and enables knowledge transfer to chemically distinct systems, including other sulfide, oxide, nitride, and halide electrolytes. Compared to Experience Replay and EWC used individually, reEWC delivers clear synergistic benefits, mitigating their respective limitations while maintaining computational efficiency. These results establish reEWC as a robust and effective solution for continual learning in MLIPs, enabling universal models that can advance materials research through large-scale, high-throughput simulations across diverse chemistries.

cond-mat.mtrl-sci↗

arXiv2025

Application of pretrained universal machine-learning interatomic potential for physicochemical simulation of liquid electrolytes in Li-ion battery

Achieving higher operational voltages, faster charging, and broader temperature ranges for Li-ion batteries necessitates advancements in electrolyte engineering. However, the complexity of optimizing combinations of solvents, salts, and additives has limited the effectiveness of both experimental and computational screening methods for liquid electrolytes. Recently, pretrained universal machine-learning interatomic potentials (MLIPs) have emerged as promising tools for computational exploration of complex chemical spaces with high accuracy and efficiency. In this study, we evaluated the performance of the state-of-the-art equivariant pretrained MLIP, SevenNet-0, in predicting key properties of liquid electrolytes, including solvation behavior, density, and ion transport. To assess its suitability for extensive material screening, we considered a dataset comprising 20 solvents. Although SevenNet-0 was predominantly trained on inorganic compounds, its predictions for the properties of liquid electrolytes showed good agreement with experimental and $\textit{ab initio}$ data. However, systematic errors were identified, particularly in the predicted density of liquid electrolytes. To address this limitation, we fine-tuned SevenNet-0, achieving improved accuracy at a significantly reduced computational cost compared to developing bespoke models. Analysis of the training set suggested that the model achieved its accuracy by generalizing across the chemical space rather than memorizing specific configurations. This work highlights the potential of SevenNet-0 as a powerful tool for future engineering of liquid electrolyte systems.

cond-mat.mtrl-sci↗

arXiv2024

Data-efficient multi-fidelity training for high-fidelity machine learning interatomic potentials

Machine learning interatomic potentials (MLIPs) are used to estimate potential energy surfaces (PES) from ab initio calculations, providing near quantum-level accuracy with reduced computational costs. However, the high cost of assembling high-fidelity databases hampers the application of MLIPs to systems that require high chemical accuracy. Utilizing an equivariant graph neural network, we present an MLIP framework that trains on multi-fidelity databases simultaneously. This approach enables the accurate learning of high-fidelity PES with minimal high-fidelity data. We test this framework on the Li$_6$PS$_5$Cl and In$_x$Ga$_{1-x}$N systems. The computational results indicate that geometric and compositional spaces not covered by the high-fidelity meta-gradient generalized approximation (meta-GGA) database can be effectively inferred from low-fidelity GGA data, thus enhancing accuracy and molecular dynamics stability. We also develop a general-purpose MLIP that utilizes both GGA and meta-GGA data from the Materials Project, significantly enhancing MLIP performance for high-accuracy tasks such as predicting energies above hull for crystals in general. Furthermore, we demonstrate that the present multi-fidelity learning is more effective than transfer learning or $Δ$-learning an d that it can also be applied to learn higher-fidelity up to the coupled-cluster level. We believe this methodology holds promise for creating highly accurate bespoke or universal MLIPs by effectively expanding the high-fidelity dataset.

cond-mat.mtrl-sci↗

arXiv2024

Scalable Parallel Algorithm for Graph Neural Network Interatomic Potentials in Molecular Dynamics Simulations

Message-passing graph neural network interatomic potentials (GNN-IPs), particularly those with equivariant representations such as NequIP, are attracting significant attention due to their data efficiency and high accuracy. However, parallelizing GNN-IPs poses challenges because multiple message-passing layers complicate data communication within the spatial decomposition method, which is preferred by many molecular dynamics (MD) packages. In this article, we propose an efficient parallelization scheme compatible with GNN-IPs and develop a package, SevenNet (Scalable EquiVariance-Enabled Neural NETwork), based on the NequIP architecture. For MD simulations, SevenNet interfaces with the LAMMPS package. Through benchmark tests on a 32-GPU cluster with examples of SiO$_2$, SevenNet achieves over 80% parallel efficiency in weak-scaling scenarios and exhibits nearly ideal strong-scaling performance as long as GPUs are fully utilized. However, the strong-scaling performance significantly declines with suboptimal GPU utilization, particularly affecting parallel efficiency in cases involving lightweight models or simulations with small numbers of atoms. We also pre-train SevenNet with a vast dataset from the Materials Project (dubbed `SevenNet-0') and assess its performance on generating amorphous Si$_3$N$_4$ containing more than 100,000 atoms. By developing scalable GNN-IPs, this work aims to bridge the gap between advanced machine learning models and large-scale MD simulations, offering researchers a powerful tool to explore complex material systems with high accuracy and efficiency.

cond-mat.mtrl-sci↗

arXiv2022

$\textit{Ab initio}$ construction of full phase diagram of MgO-CaO eutectic system using neural network interatomic potentials

While several studies confirmed that machine-learned potentials (MLPs) can provide accurate free energies for determining phase stabilities, the abilities of MLPs for efficiently constructing a full phase diagram of multi-component systems are yet to be established. In this work, by employing neural network interatomic potentials (NNPs), we demonstrate construction of the MgO-CaO eutectic phase diagram with temperatures up to 3400 K, which includes liquid phases. The NNP is trained over trajectories of various solid and liquid phases at several compositions that are calculated within the density functional theory (DFT). For the exchange-correlation energy among electrons, we compare the PBE and SCAN functionals. The phase boundaries such as solidus, solvus, and liquidus are determined by free-energy calculations based on the thermodynamic integration or semigrand ensemble methods, and salient features in the phase diagram such as solubility limit and eutectic points are well reproduced. In particular, the phase diagram produced by the SCAN-NNP closely follows the experimental data, exhibiting both eutectic composition and temperature within the measurements. On a rough estimate, the whole procedure is more than 1,000 times faster than pure-DFT based approaches. We believe that this work paves the way to fully $\textit{ab initio}$ calculation of phase diagrams.

physics.comp-ph↗