孫 程晗（スン・チェンハン）

thumbnail image
孫 程晗（スン・チェンハン）
博士研究員
関連ウェブサイト
連絡先

chsun atmark icredd.hokudai.ac.jp

瀧川 一学 グループ
主任研究者
教員
博士研究員

研究紹介

研究テーマ

Graph Network-based Simulators for Complex Reaction Networks, 3D Visualizations, Foundation Models for Chemistry

キーワード

Geometric Deep Learning, Graph Neural Networks, Visualizations, Foundation Models

研究概要

My research lies in the interdisciplinary fields of computational science and informatics, with a primary focus on developing cutting-edge geometric deep learning (GDL) models guided by physical principles. These models are applied to learn and simulate complex chemical reaction paths and networks. Additionally, I am particularly interested in developing representation tools for complex reaction networks through 3D visualization techniques (such as three.js) and augmenting GDL models with reinforcement learning techniques, such as world models. I have also been involved in several external collaborative projects focusing on building large foundation models for chemistry.

Prior to joining WPI-ICReDD, I gained rich experience in developing geometric-based descriptors and machine learning potentials (MLPs) by leveraging datasets from multiscale modeling for physical and chemical property predictions, in alignment with the AI4Science community.

代表的な研究成果

  • Developing Cheap but Useful Machine Learning-Based Models for Investigating High-Entropy Alloy Catalysts
    Chenghan Sun, R. Goel, Ambarish R. Kulkarni. Langmuir, 2024, 40, 7, 3691–3701.
    DOI: 10.1021/acs.langmuir.3c03401
  • Elucidating the Fluxionality and Dynamics of Zeolite-Confined Gold Nanoclusters Using Machine Learning Potentials
    Siddharth Sonti, Chenghan Sun (co-first author), Zekun Chen, Robert M. Kowalski, Joseph S. Kowalski, Davide Donadio, Surl-Hee Ahn, Ambarish R Kulkarni.
    Pending submission to Journal of the American Chemical Society
    ChemRxiv version: https://chemrxiv.org/engage/chemrxiv/article-details/6552db546e0ec7777fe3a056
  • Screening Cu-Zeolites for Methane Activation Using Curriculum-Based Training
    Jiawei Guo, Tyler Sours, Sam Holton, Chenghan Sun, Ambarish R. Kulkarni.
    ACS Catal., 2024, 14, 3, 1232–1242.
    DOI: 10.1021/acscatal.3c05275
  • Efficient Prediction of Partial Charges with a Size Extensive Multi-objective Deep Neural Network
    Wang-Yeuk Kong, Chenghan Sun (co-first author), Zekun Chen, Dean J. Tantillo, Davide Donadio.
    Manuscript in preparation