[gpaw-users] [gpaw-developers] Functional specific convergence problem

[Lara Ferrighi]

No, we have discussed it here of course. TPSS and M06L are very different,  
both structurally and conceptually,
so I wouldn't even consider why one is converging and the other is not.
My idea was that it is a "eigensolver problem", oscillating between 2  
occupations and therefore 2 solutions. But it doesn't look so. Since
M06L works for clean slab and slab with vacancy+gold clusters on top I can  
just assume
that it's very related to that specific electronic structure of the slab  
with vacancy.
I thought that the eigensolver would kick toward one solution of the two  
or that could be some trick to make
it converge with some kind of "level shift" between orbitals. But I really  
don't know.

Best, Lara