Operator splitting for the bidomain model revisited (Q898977)

\textit{L. Tung} [A bi-domain model for describing ischemic myocardial D-C potential. Boston: MIT (Ph.D.\ Thesis) (1978), \url{http://hdl.handle.net/1721.1/16177}] introduced the bidomain model for the electrical activity of heart tissue, which is relevant within biology and medicine. This model represents an initial boundary value problem of a multi-scale system of parabolic partial differential equations (PDEs). A semi-implicit method, which performs explicit and implicit time integration for different parts of the system, see [\textit{J. A. Southern et al.}, ``Solving the coupled system improves computational efficiency of the bidomain equations'', IEEE Trans. Biomed. Eng. 56, No. 10, 2404--2412 (2009; \url{doi:10.1109/tbme.2009.2022548})], as well as a first-order Godunov operator-splitting method, where terms in the PDEs are separated, see [\textit{J. Sundnes} et al., Computing the electrical activity in the heart. Berlin: Springer (2006; Zbl 1182.92020)], are available for the numerical solution of the problem. In the paper under review, the authors investigate both the semi-implicit technique and the operator-splitting approach. In both methods, time derivatives are discretised first by forward or backward Euler schemes. Secondly, a finite element method yields the discretization of the spatial domain. The authors analyse the errors and show upper bounds of the error terms under simplifying assumptions. Therein, the local errors of the temporal discretisation and the total errors of the spatial discretisation are considered. It follows that the bounds in the semi-implicit method are larger than the bounds in the operator-splitting technique for the same step sizes in time and space. Finally, the authors present numerical simulations in the case of one, two and three space dimensions. The semi-implicit method as well as the operator-splitting approach are applied, where the results are compared to a reference solution. In all cases, the operator-splitting technique allows for larger temporal step sizes in comparison to the semi-implicit method, while the mean errors are nearly identical. Thus the Godunov operator-splitting scheme is more efficient for the considered class of problems. This performance agrees to the derived error bounds.