Core Concepts
The RBF-FD method exhibits superconvergence, where the solution error converges at a higher order than expected, for certain monomial augmentation degrees when solving the Poisson equation on an irregularly discretized unit disc.
Abstract
The paper investigates the error convergence properties of the Radial Basis Function-generated Finite Difference (RBF-FD) method when solving the Poisson equation on an irregularly discretized unit disc.
The key findings are:
The operator approximation error, i.e., the error in the discretization of the Laplacian operator, scales as expected, with the order determined by the degree of monomial augmentation.
However, the solution error exhibits superconvergence for even monomial augmentation degrees, where the convergence order is approximately one higher than the operator approximation error.
The authors analyze this phenomenon by studying Bayona's explicit formula for the RBF interpolation error. They show that the extra order of convergence in the solution error arises from the way the terms in the error formula behave after the global sparse system is inverted to obtain the numerical solution.
The authors suggest that further investigation into the properties of the global matrix A and the process of its inversion could provide more insights into the observed superconvergence behavior.
Stats
The paper does not contain any explicit numerical data or statistics. The key results are presented through convergence plots showing the scaling of the operator approximation error and the solution error with respect to the internodal distance h and the scaling factor R.
Quotes
The paper does not contain any direct quotes that are particularly striking or support the key arguments.