There was nothing "wrong" in wien2k_13 for default gga+u+so calculations.
As you can see when you look into the "updates" page in www.wien2k.at in
wien2k_16 we added the "cross-term" between up and dn spin states
(active only with SO) as the default behavior (while one had to activate
this by an extra switch in wien2k_13, and the total energy contribution
of this extra term had an error).
Obviously, this changed default can change the results in certain cases,
but usually I would expect a small effect.
I'd consider this similar as the difference of using an "effective U"
(J=0) or using U and J as two independent parameters. Calculations with
U_eff or with U - J (even when U_eff=U-J) will give slightly different
results.
Am 17.05.2017 um 03:21 schrieb Guo-ping Zhang:
Dear Peter,
Thank you very much for your help!
(Upon your request, I post this to the mailing list.)
Your suggestion (when running DOS for +U calculation I have to use x
lapw2 -qtl -p -up/dn -orb -so, followed by x tera -up/dn) solved my
problem.
I have two followup questions.
(1) Are GGA+U+SO results in Wien2k_13 trustworth? Or just the
calculation for DOS is incorrect, not the actual results such as spin
moment, optical properties, total energies?
I noticed that although Wien2k_13 (I am using) and Wien2k_16 give the
same peak position (energy) for 4F states, the shapes of DOS differ.
5D states are not affected by this since U is affed to 4F only, as
expected.
(2) I did notice a strange behavior with +U calculation.
Using Wien2k_13, after the second iteration, the output on the
terminal looks like this (suppose that I have three nodes)
ORB END
ORB END
LAPW1 END
LAPW1 END
LAPW1 END
LAPW1 END
LAPW1 END
LAPW1 END
... ...
But with Wien2k_16, I have
ORB END
ORB END
ORB END