The pseudopotential method is one of the leading methods for
the calculation of the electronic structure properties of solids,
including the band structure and structural properties such as the
lattice constant.
In the pseudopotential method, a ``pseudoatom'' is created which has the
same valence properties as an all-electron atom, but which is much smoother
allows the use of Fourier series in the determination of the wave functions.
Over the past fifteen years, perhaps a dozen methods for generating
pseudopotentials have been proposed, with various claims being
made for each by their proponents.
It is timely to offer many of these methods in a unified and tested framework.
The pseudopotential approach is now migrating from academia into industry,
as electronic structure codes are turned into commercial software for materials
modelling applications. This increases the need for a well defined suite of
tests, so that an end-user who is not in a research environment can estimate
the reliability of a calculation. Working under a Cooperative Research and
Development Agreement with Biosym
Technologies, Inc., a leading marketer of scientific software, we are
developing a program to generate and evaluate pseudopotential properties.
Goal
Our goal is to develop a program in a mixture of C++
and legacy Fortran which can generate a variety of pseudopotentials,
using a variety of numerical methods.
We envision three levels of choices:
choice of approximation for the
all-electron atom (e.g., Hartree-Fock theory, Local Density Approximation,
Gradient-Corrected Local Density Approximation,
Local Spin Density Approximation)
choice of pseudizing method (e.g., Hamann-Schluter-Chiang,
Vanderbilt, Hamann, Rappe et al., Troullier and Martins)
choice of numerical methods
(e.g., low-order vs. high-order finite difference, basis function vs. grid-based, various public-domain packages)
We hope to have a public domain version of the program in 1996.
Persons interested in more information,
making suggestions,
or in collaborative opportunities,
particularly contributing legacy source code,
may contact
Zachary Levine
(zachary@molphys.nist.gov).