Abstract:Green's function plays a significant role in both theoretical analysis and numerical computing of partial differential equations (PDEs). However, in most cases, Green's function is difficult to compute. The troubles arise in the following three folds. Firstly, compared with the original PDE, the dimension of Green's function is doubled, making it impossible to be handled by traditional mesh-based methods. Secondly, Green's function usually contains singularities which increase the difficulty to get a good approximation. Lastly, the computational domain may be very complex or even unbounded. To override these problems, we leverage the fundamental solution, boundary integral method and neural networks to develop a new method for computing Green's function with high accuracy in this paper. We focus on Green's function of Poisson and Helmholtz equations in bounded domains, unbounded domains. We also consider Poisson equation and Helmholtz domains with interfaces. Extensive numerical experiments illustrate the efficiency and the accuracy of our method for solving Green's function. In addition, we also use the Green's function calculated by our method to solve a class of PDE, and also obtain high-precision solutions, which shows the good generalization ability of our method on solving PDEs.

What problem does this paper attempt to address?

### Problems Addressed by the Paper This paper aims to address the computational challenges of Green's function in partial differential equations (PDEs). Specifically, the paper focuses on the computation of Green's functions for the Poisson equation and the Helmholtz equation in bounded domains, unbounded domains, and interface problems. ### Background and Challenges 1. **Increased Dimensions**: Compared to the original PDE, the dimension of the Green's function is doubled, making traditional grid-based methods difficult to handle. 2. **Singularity**: Green's functions typically contain singularities, which increases the difficulty of obtaining good approximations. 3. **Complex or Unbounded Computational Domains**: The computational domain can be very complex or even unbounded, further increasing the difficulty of computing Green's functions. ### Solutions To overcome the above problems, the authors propose two neural network-based methods: 1. **Derivative-Based Method (DB-GreenNet)**: - Uses neural networks to directly approximate the target function and constructs the loss function by computing the PDE residual and boundary condition loss through automatic differentiation. - This method performs well in bounded domains but is less effective in unbounded domains and interface problems. 2. **Boundary Integral-Based Method (BI-GreenNet)**: - Utilizes the boundary integral method to convert the PDE into a boundary integral equation, thereby automatically satisfying the PDE without requiring additional differentiation of the network input. - This method not only performs excellently in bounded domains but also handles unbounded domains and interface problems. ### Main Contributions 1. **Direct Computation of Green's Functions**: Proposes a neural network-based method that can directly compute Green's functions, which is a challenging problem for traditional methods. 2. **High-Precision PDE Solving**: Uses the computed Green's functions as solvers to solve a class of PDEs with high precision. 3. **No Need for Precise Datasets**: Compared to other methods that learn PDE solvers, this method does not require precise datasets, reducing the complexity of data preparation and improving computational generalization. 4. **Simple Network Structure**: Fully utilizes the properties of Green's functions, and simple hidden layer network structures such as MLP and ResNet are sufficient to fit the target function. 5. **Wide Applicability**: This method is not only applicable to bounded domains but also handles interface problems and unbounded domains. ### Experimental Results The paper validates the effectiveness and accuracy of the proposed methods through numerical experiments. For example, in the Poisson equation on a unit disk and the Poisson equation on an L-shaped domain, the relative L2 error of the BI-GreenNet method is significantly lower than that of the DB-GreenNet method, especially near the boundary where the BI-GreenNet method maintains a small error. ### Conclusion The BI-GreenNet method proposed in this paper has significant advantages in computing Green's functions, providing high-precision solutions in various scenarios, including bounded domains, unbounded domains, and interface problems.

BI-GreenNet: Learning Green's functions by boundary integral network

Numerical solutions of boundary problems in partial differential equations: A deep learning framework with Green's function

Learning Domain-Independent Green's Function For Elliptic Partial Differential Equations

Deep Generalized Green's Functions

Neural Green's Function Accelerated Iterative Methods for Solving Indefinite Boundary Value Problems

DeepGreen: deep learning of Green's functions for nonlinear boundary value problems

Learning Green's Function Efficiently Using Low-Rank Approximations

BINet: Learning to Solve Partial Differential Equations with Boundary Integral Networks

Multiscale Neural Networks for Approximating Green's Functions

Discovery of Green's function based on symbolic regression with physical hard constraints

A fully differentiable GNN-based PDE Solver: With Applications to Poisson and Navier-Stokes Equations

Machine Learning to Approximate Free-Surface Green's Function and Its Application in Wave-Body Interactions

Green Multigrid Network

Green's function for the poisson equation in the domains bounded by two intersecting spheres

Learning Green's functions associated with time-dependent partial differential equations

Green's Function and Pointwise Estimate for a Generalized Poisson‐Nernst‐Planck‐Navier‐Stokes Model in Dimension Three

Discrete null field equation methods for solving Laplace's equation: Boundary layer computations

PhyGNNet: Solving spatiotemporal PDEs with Physics-informed Graph Neural Network

Homotopy perturbation method based on Green function for solving non-linear singular boundary value problems

Deep surrogate model for learning Green's function associated with linear reaction-diffusion operator