JKJaesun Kim

Jaesun Kim

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Jaesun Kim has a total of 46 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • A Lightweight Universal Machine-Learning Interatomic Potential via Knowledge Distillation for Scalable Atomistic Simulations

    by Sangmin Oh, Jinmu You, Jaesun Kim, Jiho Lee, Hyungmin An, Seungwu Han, Youngho Kang

    We introduce a lightweight universal machine-learning interatomic potential (uMLIP), SevenNet-Nano, based on the graph neural network architecture SevenNet and enabled by a knowledge-distillation framework. The model inherits the broad generalization capability of a large multitask foundation model, SevenNet-Omni, trained on diverse materials data sets across chemical, configurational, and computational spaces. By learning chemical representations from high-quality inference data generated by the teacher model within a unified computational framework, SevenNet-Nano achieves high accuracy and strong transferability despite its compact architecture. The model also accurately captures a wide range of interatomic interactions, enabling reliable simulations under both equilibrium and extreme conditions, including plasma etching of SiO2. Comprehensive benchmarks on static and dynamical properties─such as Li-ion diffusion and liquid densities─demonstrate its broad applicability across diverse domains, with optional fine-tuning to further enhance performance if required. Importantly, SevenNet-Nano significantly reduces computational cost, achieving over an order-of-magnitude speedup and enabling large-scale atomistic simulations involving thousands of atoms.


  • Top Cited Papers
    Open Access

    Optimizing cross-domain transfer for universal machine learning interatomic potentials

    by Jaesun Kim, Jinmu You, Yutack Park, Yunsung Lim, Yujin Kang, Jisu Kim, Haekwan Jeon, Suyeon Ju, Deokgi Hong, Seung Yul Lee, et al.

    Accurate yet transferable machine-learning interatomic potentials are essential for accelerating materials and chemical discovery. However, many existing universal models are overfitted to narrow chemical spaces or computational protocols, limiting their reliability across diverse chemical and functional domains. Here, we introduce a transferable multi-domain training strategy that jointly optimizes parameters through selective regularization, coupled with a domain-bridging set that aligns potential-energy surfaces across datasets. Systematic ablation experiments show that suggested strategies synergistically enhance out-of-distribution generalization while preserving in-domain fidelity. Based on our observation, we train SevenNet-Omni on 15 open datasets spanning molecules, crystals, and surfaces. Our model achieves state-of-the-art accuracy in cross-domain benchmarks, reaching chemical accuracy in various scenarios including adsorption-energy in catalytic surfaces and metal–organic frameworks. SevenNet-Omni also accurately reproduces high-fidelity properties by effectively transferring knowledge learned from larger, lower-accuracy databases. This framework offers a scalable route toward universal, transferable models that bridge quantum-mechanical fidelities and chemical domains.


  • Unveiling Defect Motifs in Amorphous GeSe Using Machine Learning Interatomic Potentials

    by Minseok Moon, Seungwoo Hwang, Jaesun Kim, Yutack Park, Changho Hong, Seungwu Han

    Ovonic threshold switching (OTS) selectors are pivotal in nonvolatile memory devices due to their nonlinear electrical characteristics and polarity-dependent threshold voltages. However, the atomic-scale origins of the defect states responsible for these behaviors remain unclear. In this study, we systematically investigate defects in amorphous GeSe using molecular dynamics simulations accelerated by machine learning interatomic potentials (MLIPs). We first benchmark several MLIP architectures, including descriptor-based potentials and graph neural network (GNN)-based potentials. Our results demonstrate that capturing higher-order interactions, at least four-body correlations, and medium-range structural order is essential for accurately representing amorphous GeSe structures. Our analysis indicates that GNN architectures with multiple interaction layers effectively capture higher-order correlations and medium-range order, thereby preventing spurious defects easily introduced by less descriptive MLIPs. Utilizing the optimized GNN model, we identify two distinct defect motifs across 20 independent 960-atom amorphous GeSe structures: aligned Ge chains associated with defect states near the conduction band, and overcoordinated Ge chains linked to defect states near the valence band. Moreover, we establish correlations between electronic defect levels and specific structural features─namely, the average alignment of bond angles in aligned chains and the degree of local Peierls distortion around overcoordinated Ge atoms. These insights provide a theoretical framework for interpreting experimental observations and deepening the understanding of defect-driven OTS phenomena in amorphous GeSe.


  • Top Cited Papers
    Open Access

    Roadmap for the development of machine learning-based interatomic potentials

    by Yong-Wei Zhang, Viacheslav Sorkin, Zachary H Aitken, Antonio Politano, Jörg Behler, Aidan P Thompson, Tsz Wai Ko, Shyue Ping Ong, Olga Chalykh, Dmitry Korogod, et al.

    An interatomic potential, traditionally regarded as a mathematical function, serves to depict atomic interactions within molecules or solids by expressing potential energy concerning atom positions. These potentials are pivotal in materials science and engineering, facilitating atomic-scale simulations, predictive material behavior, accelerated discovery, and property optimization. Notably, the landscape is evolving with machine learning transcending conventional mathematical models. Various machine learning-based interatomic potentials, such as artificial neural networks, kernel-based methods, deep learning, and physics-informed models, have emerged, each wielding unique strengths and limitations. These methods decode the intricate connection between atomic configurations and potential energies, offering advantages like precision, adaptability, insights, and seamless integration. The transformative potential of machine learning-based interatomic potentials looms large in materials science and engineering. They promise tailor-made materials discovery and optimized properties for specific applications. Yet, formidable challenges persist, encompassing data quality, computational demands, transferability, interpretability, and robustness. Tackling these hurdles is imperative for nurturing accurate, efficient, and dependable machine learning-based interatomic potentials primed for widespread adoption in materials science and engineering. This roadmap offers an appraisal of the current machine learning-based interatomic potential landscape, delineates the associated challenges, and envisages how progress in this domain can empower atomic-scale modeling of the composition-processing-microstructure-property relationship, underscoring its significance in materials science and engineering.


  • Top Cited Papers

    Data-Efficient Multifidelity Training for High-Fidelity Machine Learning Interatomic Potentials

    by Jaesun Kim, Jisu Kim, Jaehoon Kim, Jiho Lee, Yutack Park, Youngho Kang, Seungwu Han

    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 multifidelity databases simultaneously. This approach enables the accurate learning of high-fidelity PES with minimal high-fidelity data. Employing the generalized gradient approximation (GGA) and meta-GGA as low- and high-fidelity approaches, respectively, we tested this framework on the Li6PS5Cl and InxGa1-xN systems. The results show that using a high-fidelity training set with a size approximately 10% of the low-fidelity set, the multifidelity training framework achieves excellent accuracy, with Li-ion conductivity predictions within 10% error and InxGa1-xN mixing energy showing an R2 of 0.98 compared to the reference high-fidelity MLIP results. It indicates that geometric and compositional spaces not covered by the high-fidelity meta-GGA database can be effectively inferred from low-fidelity GGA data, thus enhancing accuracy and molecular dynamics stability. We also developed 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 multifidelity learning is more effective than transfer learning or Δ-learning and 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 data set.


  • Top Cited Papers

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

    by Yutack Park, Jaesun Kim, Seungwoo Hwang, Seungwu Han

    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 SiO2, 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 pretrain SevenNet with a vast data set from the Materials Project (dubbed "SevenNet-0") and assess its performance on generating amorphous Si3N4 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.


  • Open Access

    Applications and training sets of machine learning potentials

    by Changho Hong, Jaehoon Kim, Jaesun Kim, Jisu Jung, Suyeon Ju, Jeong Min Choi, Seungwu Han