Calculation of optical matrix elements with nonlocal pseudopotentials

Electric dipole transition rates may be calculated using matrix elements of either the momentum or the position operator. If a nonlocal pseudopotential scheme is used, the two operator methods give different results, but Starace has shown that the correct procedure is to use the position operator. We present numerical calculations for atoms which illustrate the circumstances in which the differences between the two procedures are large. We also evaluate the residual errors due to the use of a pseudopotential which are found to be small in most cases.