There are two methods for obtaining a band structure
from CASTEP.
Method one
----------
In fort.14, set IPRINT=1 and ISBROT=1. Run for a single
iteration, having previously converged electronic structure
(NITER=1, ISTART=1). Band energies will be printed out at
k-points used by electronic struct calculation.
Method two
----------
In fort.14, set ISBROT=1 and IBANS=1. Castep will
read in fort.16 and keep this charge density fixed, solving
H|\psi> = E|\psi> for this fixed potential. The k-points at
which the band structure is to be calculated will be read from
fort.15, any old wavefunction files will not be used, and it
will need 10 to 20 iterations (NITER) to converge the wavefunctions
with the fixed potential. This method enables the band structure
at, e.g. the gamma point to be calculated without having to use
the gamma point in the kpoint sampling for the initial electronic
structure calculation. NBANDS=NBANOC can be increased at this point,
without setting IOCCUP to one, so that unoccupied bands can be calculated.
Other issues
------------
The band gap will be wrong. This is a feature of all LDA calculations.
The energy scale resulting is unusual. If one needs the vacuum level,
ensure there is some vacuum in the cell, and plot the hartree + local
pseudopotential, and take its value in the vacuum. Do not include the
XC pot - this is large for quite low charge densities (-0.1V at Rs=155A,
-0.2V at Rs=74A, -0.5V at Rs=28A). A "real" vacuum has Rs=inf and Vxc=0,
but no CASTEP cell has this region - minimum charge densities are
typically around Rs=25 to 30A.
(MJR 11/96)