The heat equation is a fundamental model for transient diffusion processes, yet its numerical treatment becomes increasingly demanding when spatial dimensionality, source placement, and matrix conditioning interact. This study investigates the direct application of the Method of Fundamental Solutions (MFS) to the homogeneous heat equation with Dirichlet boundary conditions in two-dimensional circular and three-dimensional spherical domains. A fixed-source, meshless formulation was implemented in MATLAB using equal numbers of source and collocation points, two fictitious-boundary radii (R=1.5 and R=2), and the Moore-Penrose pseudo-inverse to stabilize ill-conditioned systems. Numerical accuracy was evaluated against analytical solutions through pointwise absolute and relative errors. In two dimensions, relative errors remained small and stable, ranging from 7.5×10−3 to 8.1×10−3 as M=N increased from 100 to 250. In three dimensions, the main improvement occurred between M=N=10 and M=N=15, after which the relative error plateaued near 3.62×10−1 to 3.63×10−1 despite further increases in point number. Changing the fictitious-boundary radius produced only modest differences in both dimensions. These findings show that increasing the number of approximation points does not necessarily yield monotonic accuracy gains and that MFS performance is shaped by the combined effects of point configuration, source-boundary placement, spatial dimensionality, and numerical conditioning. The study provides controlled comparative evidence on the practical behavior of direct fixed-source MFS for transient heat conduction and clarifies the conditions under which additional discretization offers diminishing numerical returns within the tested computational configurations.
Copyrights © 2026