Frozen core approximation

[haakoek]

Thomas has suggested that frozen core approximation should be treated as early as possible at system level. However, this has some implications for how we should treat the system size. If a Hartree-Fock calculation is performed, we can treat frozen core by removing the columns of the coefficient matrix, C, corresponding to the frozen orbitals. However, we must change the occupied slice of system (system.n), while the number of virtual orbitals (system.m) remains unchanged.

I assume that system.l = number of basis functions (atomic orbitals). Let n_e be the number of electrons and suppose we have n_frozen orbitals. Furthermore, let C be the hartree-fock coefficient matrix used to change the system basis

- system.n = n_e - n_frozen
- system.m = system.l - n_e,
- C -> C[:,n_frozen:] and then change basis.

There might be other technical problems associated with this that I do not foresee at the current time.