- Professor Yong-Hoon Kim's team from the School of Electrical Engineering succeeded in accelerating calculations for electronic structure in quantum mechanics for the first time in the world using a convolutional neural network (CNN) model
- Presenting the learning principle of quantum mechanical 3D chemical bonding information through AI, expected to be applied to next-generation material and device computer design fields
The close relationship between AI and highly complicated scientific computing can be seen in the fact that both the 2024 Nobel Prizes in Physics and Chemistry were awarded to scientists for devising the AI for their respective fields of study. KAIST researchers succeeded in dramatically shortening the calculation time of highly sophisticated quantum mechanical computer simulations by predicting atomic-level chemical bonding information distributed in 3D space using a novel approach to teach AI.
KAIST (President Kwang-Hyung Lee) announced on the 30th of October that Professor Yong-Hoon Kim's team from the School of Electrical Engineering developed a 3D computer vision artificial neural network-based calculation methodology for the first time in the world that bypasses the complex algorithms required for atomic-level quantum mechanical calculations performed using supercomputers to derive the properties of materials.
< Figure 1. Various methodologies are utilized in the simulation of materials and materials, such as quantum mechanical calculations at the nanometer (nm) level, classical mechanical force fields at the scale of tens to hundreds of nanometers, continuum dynamics calculations at the macroscopic scale, and calculations that mix simulations at different scales. These simulations are already playing a key role in a wide range of basic research and application development fields in combination with informatics techniques. Recently, there have been active efforts to introduce machine learning techniques to radically accelerate simulations, but research on introducing machine learning techniques to quantum mechanical electronic structure calculations, which form the basis of high-scale simulations, is still insufficient. >
The density functional theory (DFT) calculations in quantum mechanics using supercomputers have become an essential and standard tool in a wide range of research and development fields, including advanced materials and drug design, as they allow for fast and accurate prediction of quantum properties.
*Density functional theory (DFT): A representative theory of ab initio (first principles) calculations that calculate quantum mechanical properties from the atomic level.
However, in actual density functional theory (DFT) calculations, a complex self-consistent field (SCF) process of generating three-dimensional electron densities and solving quantum mechanical equations must be repeated tens to hundreds of times, which limits its application to hundreds to thousands of atoms.
*Self-consistent field (SCF): A scientific computing method widely used to solve complex many-body problems that must be described by a number of interconnected simultaneous differential equations.
Professor Yong-Hoon Kim's research team asked whether it is possible to avoid the self-consistent field process using the artificial intelligence technique that has recently been rapidly developing. As a result, the DeepSCF model was developed to accelerate calculations by learning chemical bond information distributed in three-dimensional space through a neural network algorithm in the field of computer vision.
< Figure 2. The deepSCF methodology developed in this study provides a way to rapidly accelerate DFT calculations by avoiding the self-consistent field process (orange box) that had to be performed repeatedly in traditional quantum mechanical electronic structure calculations through artificial neural network techniques (green box). The self-consistent field process is a process of predicting the 3D electron density, constructing the corresponding potential, and then solving the quantum mechanical Cohn-Sham equations, repeating tens to hundreds of times. The core idea of the deepSCF methodology is that the residual electron density (δρ), which is the difference between the electron density (ρ) and the sum of the electron densities of the constituent atoms (ρ0), corresponds to chemical bonding information, so the self-consistent field process is replaced with a 3D convolutional neural network model. >
The research team focused on the fact that according to density functional theory, electron density contains all quantum mechanical information of electrons, and in addition, the residual electron density, which is the difference between the total electron density and the sum of the electron densities of the constituent atoms, contains chemical bond information, and selected it as a target for machine learning.
Afterwards, a data set of organic molecules containing various chemical bond characteristics was adopted, and the atomic structures of the molecules included in it were subjected to arbitrary rotations and deformations to further improve the accuracy and generalization performance of the model. Finally, the research team demonstrated the validity and efficiency of the DeepSCF methodology for complex and large systems.
< Figure 3. An example of applying the deepSCF methodology to a carbon nanotube-based DNA sequence analysis device model (top left). In addition to classical mechanical interatomic forces (bottom right), the residual electron density (top right) and quantum mechanical electronic structure properties such as the electronic density of states (DOS) (bottom left) containing information on chemical bonding are rapidly predicted with an accuracy corresponding to the standard DFT calculation results that perform the SCF process. >
Professor Yong-Hoon Kim, who led this research, said, "We have found a way to correspond quantum mechanical chemical bonding information distributed in three-dimensional space to an artificial neural network," and added, "Since quantum mechanical electronic structure calculations are the basis for all-scale material property simulations, we have established the overall basic principles for accelerating material calculations through artificial intelligence."
The results of this research, conducted by Ryong-Gyu Lee, a Ph.D. candidate of the School of Electrical Engineering, as the first author, was published online on October 24th in the authoritative journal in the field of materials calculation, Npj Computational Materials. (Paper title: Convolutional network learning of self-consistent electron density via grid-projected atomic fingerprints)
This research was conducted with support from the KAIST Venture Research Program for Graduate and PhD Students and the National Research Foundation of Korea's Mid-career Researcher Support Program.