Abstract:In this paper, we introduce a novel approach to solve the many-body Schrodinger equation by the tensor neural network. Based on the tensor product structure, we can do the direct numerical integration by using fixed quadrature points for the functions constructed by the tensor neural network within tolerable computational complexity. Especially, we design several types of efficient numerical methods to treat the variable-coupled Coulomb potentials with high accuracy. The corresponding machine learning method is built for solving many-body Schrodinger equation. Some numerical examples are provided to validate the accuracy and efficiency of the proposed algorithms.

What problem does this paper attempt to address?

The problem this paper attempts to solve is using Tensor Neural Networks (TNN) to solve the Many-Body Schrödinger Equation. Specifically, the paper proposes a new method that directly performs numerical integration through the tensor neural network structure to solve the Many-Body Schrödinger Equation within acceptable computational complexity. ### Background and Challenges 1. **High-Dimensional Optimization Problem**: - The Many-Body Schrödinger Equation is a high-dimensional optimization problem, where the dimension of the wave function \(\Psi(\mathbf{r})\) is \(3N\), with \(N\) being the number of electrons. - As \(N\) increases, the number of grid points and the computational complexity of direct integration schemes grow exponentially, making traditional numerical methods almost impossible to solve this problem, known as the "Curse of Dimensionality" (CoD). 2. **Pauli Exclusion Principle**: - The wave function must satisfy the Pauli Exclusion Principle, meaning it has antisymmetry. This implies that the sign of the wave function changes when any two electrons are exchanged. - This constraint limits the choice of trial function sets and adds extra computational workload. ### Solution 1. **Tensor Neural Networks (TNN)**: - The paper introduces a tensor neural network structure that leverages its low-rank properties to decompose high-dimensional integrals into a series of one-dimensional integrals, significantly reducing computational complexity. - Each sub-network of the TNN is a continuous mapping from a bounded closed set \(\Omega_i \subset \mathbb{R}\) to \(\mathbb{R}^p\), and the TNN structure is built by combining the output layers of these sub-networks. 2. **Efficient Numerical Integration Method**: - By using the tensor product Gaussian quadrature rule, the function constructed by the TNN can be integrated with high precision directly, without the need for Monte Carlo methods. - The paper details the design of the numerical integration method, ensuring the efficiency and accuracy of high-dimensional integration. 3. **Handling Antisymmetry**: - To ensure the antisymmetry of the wave function, the paper adds penalty terms to the loss function, which are the inner products of the wave function before and after exchanging electron positions. - This method allows the Pauli Exclusion Principle to be automatically satisfied during the optimization process. ### Application and Validation - The paper provides several numerical examples to verify the accuracy and efficiency of the proposed algorithm in solving the ground-state electronic structure of atomic systems. - This method is not only applicable to atomic systems but also has the potential to compute the ground-state structure of molecules without any prior information. ### Summary By introducing tensor neural networks and efficient numerical integration methods, the paper successfully addresses the high-dimensional optimization problem and the constraints of the Pauli Exclusion Principle in the Many-Body Schrödinger Equation, providing a new and effective approach to solving large-scale Many-Body Schrödinger Equations.

Solving Schrödinger Equation Using Tensor Neural Network

Solving High-dimensional Parametric Elliptic Equation Using Tensor Neural Network

Solving the Electronic Schrödinger Equation by Pairing Tensor-Network State with Neural Network Quantum State

Solving many-electron Schrödinger equation using deep neural networks

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

Solving Schrodinger equations using physically constrained neural network

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

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

Soliton, Breather and Rogue Wave Solutions for Solving the Nonlinear Schrödinger Equation Using a Deep Learning Method with Physical Constraints

Hermite Neural Network Simulation for Solving the 2D Schrodinger Equation

Solving Multi-Dimensional Schr\"{o}dinger Equations Based on EPINNs

Tensor neural networks for high-dimensional Fokker-Planck equations

SchrödingerNet: A Universal Neural Network Solver for The Schrödinger Equation

Solving High-Dimensional Partial Differential Equations Using Tensor Neural Network and A Posteriori Error Estimators

Deep learning neural network for approaching Schrödinger problems with arbitrary two-dimensional confinement

A clever neural network in solving inverse problems of Schrödinger equation

Solving High Dimensional Partial Differential Equations Using Tensor Neural Network and A Posteriori Error Estimators

Tensor Neural Network Interpolation and Its Applications

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

Physics-Informed Neural Networks as Solvers for the Time-Dependent Schrödinger Equation