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

[Stefaan Cottenier]

>     Before my new problems, I didn't agree with one of your point in 
> last email about the free Tb atom, you said:
>     free Tb atom f^9s^2 ,the free atom has (7-2)+(+3+2)=10 mu_B for 
> f-electrons only, and an extra 1 mu_B by the d-electron (ferromagnetic 
> intra-atomic 5d-4f coupling)
>     *Where is the "extra 1 mu_B by the d-electron" come from in free Tb 
> atom ? I think this ferromagnetic intra-atomic 5d-4f coupling only exist 
> in solid such as the compound of Tb, not in free Tb atom. If I am wrong, 
> please correct me, thank you.*

Right, I was mixing up two things. The d electron is there only in the
4f^8 configuration, not in the 4f^9.

>    *  <1>Since the total energy is smaller than last time, so can we say 
> placing the f_dn electron on m=3 is reasonable (or is the right ground 
> state) ?*

1) Whether or not the total energy is a suitable criterium for this, I
don't dare to say (see discussion in the papers I quoted before).

2) In any case, don't do any analysis on the result of an -orbc run.
This is only to prepare a suitable starting density for a normal -orb
run. Do all analysis on the -orb result only.

3) From your previous mails, it could be seen you use RLO's. That's fair
for lanthanides, but I vaguely remember (not sure) that this makes
things less stable. Perhaps try it once without RLO's as well. (For
sure, the RLO makes your result for the EFG and the orbital and dipolar
hyperfine field wrong, only the total energy is improved.)

> *    <2>I have no idea why :ORB was still the same as last time ?*
> 

It might happen that the electron "jumps back", even with -orbc: you
only freeze the potential, not the density matrix.

> *    *Next, I changed the mixing value in case.inm from 0.05 to 0.01, 
> and run runsp -ec 0.0001 -cc 0.0001 -so -orb -i 1000,
> 
> when I was about to keep my fingers crossed, but bang, scf stoped with 
> information like:
> 
>      LAPW0 END
> 
>      ORB   END
> 
>      >   stop error

Yes, that sounds familiar -- unfortunately. As far as I remember, there
was no clear clue. Playing around with different dmat occupations
sometimes results in something that works. Also, try to reduce your
mixing value (if still relevant with the new mixer) before doing the
-orbc run (that can prevent the electron from jumping back, and then
your frozen potential is less exotic for the subsequent -orb run then it
is now). This is far from systematic advice, I know...

Stefaan