[Wien] a story about in1new

[Stefaan Cottenier]

Dear all,

I had a funny experience with in1new recently, which I report below. 
Both as a kind of warning and to have it available in the archives if 
somebody later would meet something similar.

A case didn't want to converge, it kept oscillating with a large :DIS. 
All the usual tests didn't help. After a while, it turned out that it 
converged smoothly with the 'old' linearization scheme, however (no -
in1new switch used). As soon as in1new was turned on -- even from the 
converged situation -- large oscillations were there again. The reason 
for this turned out to be a small gap in the broad band region of the 
DOS: the bands crossing E_F were separated from another set of 
dispersed bands by (exactly) 0.05 Ry, such that the criterium set by 
ESEPER0 (top lines of case.in2) was just met. As a result, these lower 
bands were incorrectly considered as semi-core states for the 
determination of the linearization energies. Increasing ESEPER0 a bit 
fully solved the problem.

The UG tells that this has to be checked each time. But although I use 
in1new often, I never met any problem with it. And that makes one to 
forget the good advice of the UG after a while...

Stefaan