Apr 24, 2024
2:15pm - 2:30pm
Room 320, Level 3, Summit
Wonseok Jeong1,Hyuna Kwon1,Yu-Ting Hsu1,Wenyu Sun1,Liwen Wan1,Michael Nielsen1,Tuan Anh Pham1
Lawrence Livermore National Laboratory1
Wonseok Jeong1,Hyuna Kwon1,Yu-Ting Hsu1,Wenyu Sun1,Liwen Wan1,Michael Nielsen1,Tuan Anh Pham1
Lawrence Livermore National Laboratory1
The ability to precisely determine the atomic structure of functional materials would have a transformative and broad impact on a broad range of emerging technologies, from energy storage and conversion to ion-selective membranes. In this talk I will summary our rencent activities in integrating atomistic simulations, data science, and spectroscopic measurements to elucidate structural and chemical heterogeneities in disordered systems. I will show how machine learning potential is used to efficiently explore the vast configuration space of amorphous carbon nitrides and to identify the local structural motifs of the systems. Density functional theory simulations were used to establish a correlation between the local structure motifs and X-ray absorption spectroscopic signatures, which then serves as the basis for interpreting and extracting chemical content from experimental data. Beyond predicting the chemical content, I will also discuss a strategy to predict the three-dimensional atomic structure of amorphous sytems from X-ray absorption spectroscopy via a generative diffusion model. Using amorphous carbon as a case study, it is found that the generative model exhibits a remarkable scale-agnostic property, enabling the generation of large-scale atomic structures, while being able to accurately predict the atomic structure from targeted spectroscopy. The methods developed here are general and can be broadly applied for inverse design of functional materials.