I am studying Hematite (Fe2O3) Antiferromagnetic configuration using QE 7.0 (EDIT: compiled and tried 7.1) at the PBE+U level (U = 4.5).

I obtained an excellent agreement with experimental data on various properties, but there is one thing I can't wrap my head around.

The traces of the occupancy matrices after diagonalization for the minority spin on iron atoms are roughly close to 1 (0.88). I would expect them to be close to 0.

Iron atoms have a d5 configuration which I expect would lead to 5 electrons with either spin up or down. In my case, I get that the total number of electrons is 5 in the majority spin channel as expected, but roughly 1 in the minority channel. This is also the case in the example08 in the /PW/tutorials folder of QE for a FeO system.

I have been trying unsuccessfully to force the system into having the traces close to 0 for the minority channel.

I tried

  1. Applying starting magnetizations and charges on iron and oxygen atoms.
  2. Changing U values
  3. Reducing the Gaussian smearing to almost 0
  4. A mix of all the previous ones.
  5. Starting_ns_eigenvalues for the third eigenvalue in the minority channel initialised to 0
  6. Specified the Hubbard_occ to 5.00

I am now wondering if this is to be expected or maybe there is something I am missing about the theory or the implementation of DFT+U or maybe my understanding of the occupancy matrix is simply wrong.

EDIT: I am now trying the procedure following http://theossrv1.epfl.ch/Main/OxidationStates. I will try to investigate whether this works but also if there is a big difference in the results. I am wondering if, in the end, this occupancy problem is not a problem at all

  • 1
    $\begingroup$ In FeO yes, but In Hematite(Fe2O3) I would expect Iron to be 3+, losing the electrons in the 4s and then one in the d orbital to end up with a d5 configuration. $\endgroup$
    – Nekkrad
    Apr 28 at 8:09
  • $\begingroup$ Do you have an example of calculation with U correction leading to localized orbitals where the local population is modified ? I guess normally the local population of Iron should be almost the same compared to the isolated atom. $\endgroup$
    – M06-2x
    May 3 at 13:24


You must log in to answer this question.