Recently a 30 band k.p hamiltonian is being used to calculate the full zone bandstructure, see for example Richard, et al. PHYSICAL REVIEW B70, 235204 (2004). I was trying to discretize this hamiltonian using finite difference for a quantum well and came up with spurious solution all over the band gap. Does anybody have any idea of how to eliminate these spurious solution? The hamiltonian has linear terms (kx, ky, & kz) on the off-diagonals and a hbar^2k^2/2m0 quadratic term on all the diagonals.