The ABINIT Users Mailing List ( CLOSED )

[Valerio Olevano <valerio.olevano@polytechnique.fr>]

You didn't read my e-mail carefully.
yes it's true that the GW code now in that part has a specialization
for fcc (as well as in other parts not relevant like in the print of the
side a of the fcc cell somewhere that however is used nowhere).
But the error introduced it's small and as long as you increase your
mesh of k-points tends to 0.
That point (ndb the point for  rho_twiddle*rhotwiddle / q^2 for q->0 and 
G,G' = 0, n=n1 k=k1)
is treated like in Phillips Kleinman Phys Rev. 128, 2098 (1962),
 or if you want a more recent paper,
like in Hybertsen Louie, PRB 34, 5390 (1986).
Essentially you should calculate

1/eta \int_\eta d^3 q 1/q^2

where eta is the volume of the Brillouin zone divided by the number of 
k-points N_k.

If you suppose that your BZ (already a good approximation) is a sphere, 
then the result
you get is

7.79 \eta^(-2/3)

For the structure of the code, then at that point you should put
rhotwg(1) = the sqrt (of the previous result)    times   qpg(1)
(later re-stripped by a division by qpg(1)).

In the case of fcc the constant is 7.44
in the case of wurtzite I don't know or even if I know
I certainly don't remember.

Best regards.

Valerio Olevano