[CPMD-list] LSD

[Salomon Billeter]

Dear Natasa,

by default, CPMD does spin-unpolarized calculations, like most other DFT
programs do by default. This saves you almost half of the CPU time and
memory needed. There are two assumptions underlying this approach:

(1) The spatial distributions of the Kohn-Sham orbitals with spin up are
equal to the corresponding ones with spin down. This corresponds to the
assumption made in restricted Hartree-Fock and related methods. It means
you solve the KS equations for 2 electrons together.

(2) Each of these KS orbitals is equally populated by spin-up and spin-down
electrons, usually 1 electron each. This means you have to consider only
one density distribution instead of one for spin up and one for spin down.

Assumption (2) is certainly not valid if you are not calculating singlet
states (e.g., if you have upaired electrons, you'll end up with 0.5 up and
0.5 down electrons in the singly occupied KS orbital - clearly not a spin
state). Spin contamination can also be less straightforward to be detected
than in this obvious example.

By selecting LSD, CPMD does not make these two assumptions at the expense
of more CPU time and memory required for your calculation.

Note that you can also do restricted open-shell calculations in CPMD (but
only for the lowest triplet and the lowest excited singlet states). That
means assumption (1) without (2), so almost half of the work is saved and
you still have a defined spin state.

Best regards,
Salomon