GW bug: reordering G-vectors
Posted: Tue Jan 10, 2012 7:09 pm
Dear developers,
I've noticed some bug from the GW part, that would prevent you from using ecutsigx > ecutwfn in some cases (typically, anisotropic systems).
In setup_sigma, whenever you generate a gamma-centered G-sphere, the vectors are reordered so as to match those found in the kss. This creates the problem that G-vectors are no longer ordered by increasing modulus, and results in a "G-shell not closed" error.
As far as I can tell, removing this step makes no difference at all, as opposed to removing the assumption of an ordered set of G-vectors. I believe there have been some development to handle this gamma-centerd / k-centered duality, and that the reordering is no longer relevant.
Does anyone see a problem with this modification?
Find here-attached an example of an input file that would trigger the bug.
I've noticed some bug from the GW part, that would prevent you from using ecutsigx > ecutwfn in some cases (typically, anisotropic systems).
In setup_sigma, whenever you generate a gamma-centered G-sphere, the vectors are reordered so as to match those found in the kss. This creates the problem that G-vectors are no longer ordered by increasing modulus, and results in a "G-shell not closed" error.
As far as I can tell, removing this step makes no difference at all, as opposed to removing the assumption of an ordered set of G-vectors. I believe there have been some development to handle this gamma-centerd / k-centered duality, and that the reordering is no longer relevant.
Does anyone see a problem with this modification?
Find here-attached an example of an input file that would trigger the bug.