Thanks for the help!


On 15-12-11 03:53 AM, wrote:
Just to add that the attached files are for AFM.

On Fri, 11 Dec 2015, wrote:

With strongly correlated solids, it is usually possible to
stabilize several electronic configurations. For FeO, if
you start a mBJ or LDA+U calculation from the PBE density,
then a metallic state is obtained. The state with a gap, which
is more stable, can be obtained by running first a constrained
LDA+U calculation with -orbc:

1) change manually the occupation (that you need to know) in
case.dmatup/dn of a LDA+U calculation,
2) execute "x orb -up/dn" to generate case.vorbup/dn
3) run LDA+U with -orbc instead of -orb
4) save the contrained calculation when it is finished
4) run LDA+U or mBJ as usual.

The struct and dmat files of a recent LDA+U calculation on FeO are

F. Tran

On Thu, 10 Dec 2015, John McLeod wrote:

Hello all,

I tried to calculate the electronic structure of FeO using mBJ.

1. I took the cubic FeO structure, made a 2x2x2 primitive supercell, relabeled the Fe sites "1" and "2" to get the appropriate AFM ordering, ran sgroup, and obtained a rhombohedral cell with 2 Fe sites and 1 O site. This structure looks correct when viewed with VESTA or xcrysden, so I think I am using the correct structure.

2. I run a spin-polarized PBE calculation, initializing the Fe to "up" and "down", and the O to "no spin", this converges quickly and obtains a zero band gap as expected.

3. I run an mBJ calculation using PRATT mixing, slowly increasing the mixing factor. The energy converges in 20 cycles or so, however the charge never converges - it keeps "sloshing" back and forth between the two Fe sites. The band gap for this system remains at zero.

In PRL 102 226401 (2009) and mBJ calculation on FeO reports a gap of 1.82 eV. May I ask Dr. Fabian Tran and/or Dr. Peter Blaha, to obtain these results did you:
  1. Perform an AFM calculation or just a spin-polarized calculation?
  2. Perform any sort of structural optimization?
  3. Use PRATT mixing only, or switch back to MSR1?

I also tried this approach with hematite (Fe2O3), obtaining a rhombohedral structure with 4 inequivalent Fe sites (as expected), and following the same steps I obtained good charge and energy convergence in mBJ as well as a reasonably accurate band gap - so I find it a bit curious that my approach fails for FeO.

I would greatly appreciate a quick tip on how the original calculations in PRL 102 226401 were performed.

John McLeod

Wien mailing list

Wien mailing list

Wien mailing list

Reply via email to