Abstract:Obtaining many excited-state solutions of the multidimensional time-independent Schrodinger equation (TISE) has turned out to be crucial for the development of quantum chemistry, computational materials science, and atomic structure theory. In this work we report a numerical method based on neural networks to solve many excited states of multidimensional TISE. The effectiveness of our approach in eigenvalue problems of TISE has been demonstrated via numerical experiments on typical analytically solvable TISE, i.e., harmonic oscillators and hydrogen-like atoms. The results indicate that introducing variance of energy as the primary loss function is capable of solving the convergence imbalance problem when multiple loss terms coexist, thereby improving the accuracy of wave functions. We propose a method to explain the physical significance of neural network wave functions, laying a foundation for neural networks in atomic structure calculations. Furthermore, we find the following features: the convergence accuracy decreases as the absolute value of the system's energy eigenvalue becomes smaller, from which we develop a scheme to prevent the decreases of the convergence accuracy. Our work offers a viable insight for further study of multielectron Schrodinger equations for atom and molecules. These findings may enrich the understanding of intrinsic mechanisms for specific physical, chemical, and biological processes.

Spin-symmetry-enforced Solution of the Many-Body Schrödinger Equation with a Deep Neural Network

Symmetry enforced solution of the many-body Schrödinger equation with deep neural network

Symmetry enforced solution of the many-body Schr\" odinger equation with deep neural network

Teaching Spin Symmetry While Learning Neural Network Wave Functions

Symmetries and many-body excited states with neural-network quantum states

Learning eigenstates of quantum many-body Hamiltonians within the symmetric subspaces using neural network quantum states

Solving many-electron Schrödinger equation using deep neural networks

Ju l 2 01 8 Solving Many-Electron Schrödinger Equation Using Deep Neural Networks

Neural-network Variational Quantum Algorithm for Simulating Many-Body Dynamics

Unsupervised Deep Neural Network Approach To Solve Fermionic Systems

Ab initio solution of the many-electron Schrödinger equation with deep neural networks

Approach to Solving Spin-Boson Dynamics Via Non-Markovian Quantum Trajectories

Neural-network-based Multistate Solver for a Static Schrödinger Equation

Calculating many excited states of the multidimensional time-independent Schr?dinger equation using a neural network

Symmetrical Graph Neural Network for Quantum Chemistry, with Dual R/K Space

Neural-network quantum states for many-body physics

Deep-learning electronic-structure calculation of magnetic superstructures

Solving the Quantum Many-Body Problem with Artificial Neural Networks

A neural network approach for two-body systems with spin and isospin degrees of freedom