Nanophotonics integrated circuits simulation: FDTD method

This paper considers the functionality of FDTD (Finite difference time-domain) method for parallel computing; the main difficulties that arise in its practical use are listed; a comparison with other numerical methods of electromagnetic phenomena simulation is made. The requirements to performance balance of supercomputer node, at which effective implementation of the parallel version of FDTD method is possible, are specified. The need for computing at exaflops scale for solving the tasks of integral nanophotonics is shown, an approach to solving such problems is proposed.