Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.


Messages - gpenazzi

Pages: [1]
1
Hi,
there is no method to get the whole real space Hamiltonian. The closest you can do is to effectively zero out the fourier transform components  by repeating the configuration and extracting H, S in Gamma. Then you can access via Python the individual blocks of the output matrices and evaluate the onsite and hopping matrix elements.
And no, you can't really feed an external real space Hamiltonian, that's an internal quantity which can't be easily exposed or overriden.

2
Hi faxer92,
I am not the creator of the set, but I am familiar with the tight binding behavior in QuantumATK, hopefully I can shed some light.

As you can read in the documentation, in QuantumATK we always support self-consistent calculations. For binary compounds you will end up in some charge transfer between anion and cation during an scf calculation. The occupations are chosen by running reference bulk calculations and inspecting the mulliken population, to make sure that the starting point of an scf calculation is close to this solution. If you run an scf calculation for bulk GaAs and check the Mulliken population, you will get numbers which are very close. The end result is not very sensible to initial population, therefore as long as the total occupation is good, you can avoid to worry about it too much. For monoatomic systems where all atoms are chemically equivalent (like Silicon), no charge transfer is expected and the single occupations of the orbitals make no difference, as long as the total occupation is correct.

For the shifts (hartree and spin) the choices are to some extent arbitrary. While orbital resolved quantities can be taken, it is not uncommon to use the values corresponding to the highest occupied orbital (this is for example the typical approach in some SCF-TB schemes like doi:10.1103/PhysRevB.58.7260). Using a single quantity can be a very close approximation which gives better behavior in terms of convergence. If I'd be setting up a parametrization myself, I'd probably start with a single value for the highest occupied orbital and look at the result.

Pages: [1]