Comparison of several numerical solvers for a discretized nonlinear diffusion model with source terms (Q6126214)

The paper addresses the numerical solution of a system of two nonlinear integro-differential equations that arises from a real engineering problem related to diffusive processes in materials with memory. The properties of such materials depend on their prior states, leading to time-dependent nonlinear integro-differential equations or systems. The problem under consideration is a two-component magnetic field system based on a nonlinear diffusion model with source terms. The existence and uniqueness of solutions for this system have been previously established under certain conditions. To numerically solve the problem, the author employs finite differences or finite elements in space combined with backward Euler in time, resulting in a nonlinear system of equations at each time step. The main focus of the paper is on comparing the performance of several iterative methods for solving the nonlinear system of equations, namely Newton's method, secant method, Steffensen's method, and Traub's method. The author provides a detailed derivation of the fully discrete approximation for the case where the exponents in the model are \(p = 1\) and \(q = 2\), as well as for the more general case with \(0 < p < 1\) and \(q > 2\). The Jacobian matrices required for the implementation of Newton's method are also presented. Through numerical experiments, the author demonstrates that Newton's method is the fastest in all cases, while Steffensen's method may not converge unless the parameter p is sufficiently small. The secant method is slightly faster than Traub's method but requires more iterations overall. Initially, both secant and Traub's methods require more iterations to converge but eventually reach the same number of iterations as Newton's method. When the solution is not differentiable, Newton's method fails, and the author recommends using Traub's method in such cases. This research paper contributes to the field of numerical analysis and computational mathematics by providing valuable insights into the performance of various iterative methods for solving nonlinear systems arising from integro-differential models. The findings can guide researchers and practitioners in selecting the most appropriate numerical solver for similar problems, taking into account the differentiability of the solution and the values of the model parameters. Furthermore, the detailed derivations and Jacobian matrices presented in the paper can serve as a reference for future research on related problems.