On regularization results for a two-dimensional nonlinear time-fractional inverse diffusion problem (Q6063657)

The authors consider the following ill-posed inverse problem for a two-dimensional nonlinear time-fractional inverse diffusion model involving the Caputo time-fractional derivative: for \(x,y>0\) and \(t>0\) \[ \partial_t^\beta u(x,y,t)=-a(x)\nabla\cdot u(x,y,t)+l(x,y,t,u(x,y,t)), \] where \(u(x,y,0)=0\) for \(x,y>0\). In addition, it is assumed that \(\lim_{x\to \infty}u(x,y,t)=0\) and \(u(1,y,t)=u_1(y,t)\) for all \(y,t>0\). The order \(\beta\) of the Caputo derivative \(\partial_t^\beta\) is fixed in \((0,1)\). The function \(a\) is a space-dependent diffusion coefficient, \(\ell\) is a nonlinear function satisfying a Lipschitz condition, and the data \(u_1\) is given approximately. The aim of this work is to determine \(u(x,y,t)\) for \(x\in [0,1)\). The authors weaken some existing a priori assumptions on the exact solution via two regularization strategies based on the integral equation method. Several explicit error estimates including an error estimate of the Hölder-Logarithmic type for all \(x\in [0,1)\), which can be seen as the improvement and the generalization of error estimates presented in the existing literature, are obtained. Some interesting numerical examples are presented to illustrate the theoretical results.