It is possible to provide a converged calculation as input to
calculateOptimizedAtomicGeometry(), the keyword is called
self_consistent_calculation. See
http://quantumwise.com/documents/manuals/ATK-2008.10/ref.calculateoptimizedatomicgeometry.html for details (section "Usage examples").
However, you cannot in this case change the geometry or the method, that is, you cannot change any of the parameters used for the calculation. The purpose of this functionality is thus really to restart an optimization that didn't converge. This is analogous to the way you can restart a non-converged SCF calculation via
executeSelfConsistentCalculation(self_consistent_calculation) (see
http://quantumwise.com/documents/manuals/ATK-2008.10/ref.executeselfconsistentcalculation.html, the third example in the beginning).
The simplest way to use a converged density, but also change some parameters, for instance k-point sampling or some other parameter that doesn't change the size of the density matrix (basically only the basis set size changes the matrix size), would be to run a single-point calculation with the new parameters, initialized from the old calculation. Then you have a
self_consistent_calculation object with the desired method, and with a (hopefully) reasonable density matrix, that you now can use to bootstrap the geometry optimization.
Meta-code:
scf_0 = restoreSelfConsistentCalculation("file.nc")
new_method = KohnShamMethod(...)
scf_1 = executeSelfConsistentCalculation(
new_geometry,
new_method,
initial_calculation=scf_0)
relaxed_geometry = calculateOptimizedAtomicGeometry(scf_1)
Do note that
new_geometry must have the same atoms, ordered in the same way, as originally used to compute "file.nc", or the whole thing breaks down.
I hope this helps; there are other solutions, which involve a bit more coding...
