Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Integrating Graph Neural Networks and Many-Body Expansion Theory for Potential Energy Surfaces

Siqi Chen, Zhiqiang Wang, Xianqi Deng, Yili Shen, Cheng-Wei Ju, Jun Yi, Lin Xiong, Guo Ling, Dieaa Alhmoud, Hui Guan, Zhou Lin

TL;DR

This study developed a new computational tool integrating fragment-based graph neural networks (FBGNN) into the fragment-based many-body expansion (MBE) theory, referred to as FBGNN-MBE, and demonstrated its capacity to reproduce full-dimensional potential energy surfaces (FD-PES) for hierarchic chemical systems with manageable accuracy, complexity, and interpretability.

Abstract

Rational design of next-generation functional materials relied on quantitative predictions of their electronic structures beyond single building blocks. First-principles quantum mechanical (QM) modeling became infeasible as the size of a material grew beyond hundreds of atoms. In this study, we developed a new computational tool integrating fragment-based graph neural networks (FBGNN) into the fragment-based many-body expansion (MBE) theory, referred to as FBGNN-MBE, and demonstrated its capacity to reproduce full-dimensional potential energy surfaces (FD-PES) for hierarchic chemical systems with manageable accuracy, complexity, and interpretability. In particular, we divided the entire system into basic building blocks (fragments), evaluated their single-fragment energies using a first-principles QM model and attacked many-fragment interactions using the structure-property relationships trained by FBGNNs. Our development of FBGNN-MBE demonstrated the potential of a new framework integrating deep learning models into fragment-based QM methods, and marked a significant step towards computationally aided design of large functional materials.

Integrating Graph Neural Networks and Many-Body Expansion Theory for Potential Energy Surfaces

TL;DR

This study developed a new computational tool integrating fragment-based graph neural networks (FBGNN) into the fragment-based many-body expansion (MBE) theory, referred to as FBGNN-MBE, and demonstrated its capacity to reproduce full-dimensional potential energy surfaces (FD-PES) for hierarchic chemical systems with manageable accuracy, complexity, and interpretability.

Abstract

Rational design of next-generation functional materials relied on quantitative predictions of their electronic structures beyond single building blocks. First-principles quantum mechanical (QM) modeling became infeasible as the size of a material grew beyond hundreds of atoms. In this study, we developed a new computational tool integrating fragment-based graph neural networks (FBGNN) into the fragment-based many-body expansion (MBE) theory, referred to as FBGNN-MBE, and demonstrated its capacity to reproduce full-dimensional potential energy surfaces (FD-PES) for hierarchic chemical systems with manageable accuracy, complexity, and interpretability. In particular, we divided the entire system into basic building blocks (fragments), evaluated their single-fragment energies using a first-principles QM model and attacked many-fragment interactions using the structure-property relationships trained by FBGNNs. Our development of FBGNN-MBE demonstrated the potential of a new framework integrating deep learning models into fragment-based QM methods, and marked a significant step towards computationally aided design of large functional materials.

Paper Structure

This paper contains 39 sections, 6 equations, 5 figures, 7 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Methods
  3. Many-Body Expansion Theory
  4. Fragment-Based Graph Neural Networks
  5. General Architecture of GNNs
  6. Backbone MXMNet and PAMNet Models
  7. Experiments
  8. Benchmark Systems
  9. Molecular Dynamics and Quantum Mechanical Calculations.
  10. Inputs and Outputs
  11. Hyperparameter Tuning
  12. Results and Discussions
  13. Overall Performance Assessment
  14. Comparison between MXMNet-MBE and PAMNet-MBE
  15. Comparison across Data Sets
  16. ...and 24 more sections

Figures (5)

  • Figure 1: Schematic strategy of our FBGNN-MBE approach integrating FBGNNs into the MBE theory, using a water cluster as an illustrative example.
  • Figure 2: Schematic design of MXMNet (burgundy boxes) and PAMNet (orange boxes) for a multi-fragment complex system zhang2020molecularzhang2023universal.
  • Figure 3: Comparison between MXMNet-predicted and MP2/DFT-evaluated 2B and 3B energies for all three benchmark systems.
  • Figure 4: Comparison between PAMNet-predicted and MP2/DFT-evaluated 2B and 3B energies for all three benchmark systems.
  • Figure 5: Effects of $N_\text{layer}$ values on MXMNet-trained and PAMNet-trained values of $R^2$ and $MAE$ for all benchmark systems.