Error bounds of the finite difference time domain methods for the Dirac equation in the semiclassical regime (Q2291890)

Four finite difference time domain (FDTD) methods, the leap-frog, two semi-implicit, and the Crank-Nicolson method, are applied to solve the Dirac equation numerically in the semiclassical regime \(0 < \varepsilon \le 1\), where \(\varepsilon\) is representing the scaled Planck constant. The stability conditions and error estimates of the FDTD methods suggest that all these FDTD methods share the same \(\varepsilon\)-scalability at \(\tau = {\mathcal O}(\varepsilon^{3/2})\) and \(h= {\mathcal O}(\varepsilon^{3/2})\), where \(\tau\) is the time step size and \(h\) the mesh size in space. The computational cost for the leap-frog method is the lowest, and that for the Crank-Nicolson method is the highest. The total probability density and current density also show the same \(\varepsilon\)-scalability. The error estimates are validated by extensive numerical experiments.