[Wien] Problems of HCP Tb (The fifth time)

[Stefaan Cottenier]

> With the new mixer, reducing the mixing value is NOT relevant. If you
> feel that you need to use smaller mixing, use PRATT.

As you drop the name, my memory comes back -- yes, even with the old 
mixer it was sometimes needed to have many (=hundreds) of pratt 
iterations even before going to a small mixing value for broyden.

> N.B., this sounds to me like a case where the physical problem is
> badly posed with the flavor of DFT that you are using, so you should
> think about something like LDA/GGA+U or a hybrid -- if this is a
> better model of the physics (it might not be).

This *is* all about GGA+U.... (-orb).

Stefaan