The combination between GGA, GGA+U and mBJ-GGA to get good gap for insulator ?

[Wien] confusion with GGA+U and mbj

shamik chakrabarti Sun, 08 Mar 2020 01:11:07 -0800

Dear Wien2k users,

                           If simulation of a material is to run with GGA+U
to get an approximate band gap, then *whether GGA+U or GGA* has to run as a
first step for running mbj?

As user guide says that we have to run a normal SCF prior to mbj, my
confusion is whether the normal SCF is to run with GGA or GGA+U if U is
mandatory for the material.

Looking forward to your response in this regard.

with regards,

-- 
Dr. Shamik Chakrabarti
Research Fellow
Department of Physics
Indian Institute of Technology Patna
Bihta-801103
Patna
Bihar, India

Re: [Wien] confusion with GGA+U and mbj

Tran, Fabien Sun, 08 Mar 2020 03:26:11 -0700

Hi,

It is recommended (but not mandatory) to do first a GGA calculation before 
using GGA+U or mBJ. By doing so, the starting electron density for the 
GGA+U/mBJ SCF iterations may be better than the one from init_lapw 
(lstart/dstart) in order to reduce the number of iterations and/or to reach the 
proper ground state.

It may also be a good idea (but again not mandatory) to start a mBJ calculation 
using the GGA+U electron density.



FT

Re: [Wien] confusion with GGA+U and mbj


Peter Blaha
Sun, 08 Mar 2020 03:41:07 -0700




It really depends on the specific example if you need to be carefully 
with the starting point, since both, LDA+U and mBJ can give multiple 
solutions in certain cases.

Eg: MnO or NiO do not require any special care. The dominant cubic 
crystal field is sufficient to drive the correct solution and both, 
GGA+U and mBJ will give only one solution, independent on the starting 
conditions.


Very different is eg. FeO: Both, GGA+U and in particular mBJ will NOT 
give a band gap when started with plain GGA (and probably also not when 
starting from scratch using atomic densities ??)

An analysis of the partial Fe-DOS is necessary, followed with a 
reoccupation of the density matrix by hand to occupy the proper single 
orbital. Using this starting dmat-files, GGA+U gives an insulator (and 
this solution should have lower energy - depending on U). Continuation 
with mBJ from such a solution, also gives a nice band gap.
Analyse your partial TM-d state DOS !