Page 1 of 1
Band gap of FeO using DFT+U calculations
Posted: Thu Aug 01, 2019 9:36 am
by ketong
Dear ABINIT friends,
I am trying to make some calculations with FeO. The DFT+U method is used in the literature to correct for the localized Fe d-electrons. Therefore, Same thing was done within ABINIT by adding a Hubbard potential U to the Fe d electrons. I am using the official recommended pseudopotentials table (the PAW JTH table: Fe.GGA_PBE-JTH.xml and O.GGA_PBE-JTH.xml).
FeO was computed in antiferromagnetic AFII structure, in which the magnetic moments are ferromagnetically coupled in the (111) planes and the (111) planes are antiferromagnetically coupled to each other.
I changed U_eff from 0.0 eV to 7.0 eV. I found the FeO is metallic whatever the U was used. The DOS figures are given in below when the U equals 0.0 eV and 4.0 eV. Although a gap can be found when the U equals 4.0 eV, this gap locates below the Fermi energy level. The input file can be found in the attachments.
In addition, I think this mistake maybe contribute to the spin-orbit coupling because I didn't consider the spin-orbit coupling in DFT+U calculations. Actually, when I try to consider spin-orbit coupling in DFT+U, I got such information from
Code: Select all
--- !WARNING
src_file: m_chkinp.F90
src_line: 3106
message: |
LDA+U+SpinOrbit is still on test
(not yet in production)
Has anyone else ever encountered such an issue? Any comments would be most appreciated.
Ke Tong
Re: Band gap of FeO using DFT+U calculations
Posted: Mon Aug 05, 2019 10:34 am
by ebousquet
Dear Ke Tong,
The error message with soc looks indeed a bug. Could you send also the output file to be sure the soc was not switched on?
Regarding the gap, I cannot see nor download the tif files you put but I can give the following comments:
- it is always better to initialise the spinat to a larger value than what it is supposed to be, put 4 or 5 could help in converging to the high-spin state.
- You have a smearing that could be too large and kill artificially the gap, test with a smaller value 1E-3 or 5E-4 Ha.
- Did you check that the final magnetic moment on the atoms is more or less the ones you expect?
Best wishes,
Eric
Re: Band gap of FeO using DFT+U calculations
Posted: Tue Aug 06, 2019 2:58 am
by ketong
ebousquet wrote:Dear Ke Tong,
The error message with soc looks indeed a bug. Could you send also the output file to be sure the soc was not switched on?
Regarding the gap, I cannot see nor download the tif files you put but I can give the following comments:
- it is always better to initialise the spinat to a larger value than what it is supposed to be, put 4 or 5 could help in converging to the high-spin state.
- You have a smearing that could be too large and kill artificially the gap, test with a smaller value 1E-3 or 5E-4 Ha.
- Did you check that the final magnetic moment on the atoms is more or less the ones you expect?
Best wishes,
Eric
Dear Eric,
Thank you very much for your suggestions.
1. I tuned the parameters of "spinat" (spinat was set to 5) and "tsmear" (tsmear was set to 5e-4 Ha) in the input file. There is a piece of bad news. The result shows the FeO is still metallic ... The DOS figure and DOS file (named "FeObulk_o_DOS_U_4_tsmear_5e-4.out") can be found in the attachments.
- DOS figure when the U equals 4.0 eV and tsmear equals 5e-4 Ha
2. For the DFT+U calculations, I check the final magnetic moment. The results are given in the "Magnetic moment results.out" file. The expected magnetic moment of FeO is 3.32 to 3.69 μB. Therefore, the calculation can obtain a reasonable magnetic moment even if U is 0.0 eV.
3. The output file of soc calculation is given in the "FeObulk_soc.out" file. In addition, the previous DOS figures (U = 4.0 eV) are attached once again using jpg format.
Best wishes
Ke Tong
Re: Band gap of FeO using DFT+U calculations [SOLVED]
Posted: Wed Aug 07, 2019 9:19 pm
by ebousquet
Dear Ke Tong,
I've not looked yet to the DOS problem within collinear calculations, but now seeing the plot it is clearly not a smearing problem, there is a big pick at the Fermi level. Did you test with the usual 4 atoms AFM cell before?
In the mean time, I can already tell you that for the non-collinear calculation+soc you have to remove all the symmetries (nsym=1) because the symmetries are not yet validated for non-collinear and fails in some specific cases. You are also going to face more converge problems and you'll have to reduce diemix (0.5 - 0.2) and diemixmag (0.4 - 0.1, the exact value depends on the system, if too small the convergence can works but could take much more SCF so the larger values that works is the best). To help more you can also put nline=10 or 12 and in the latest version (8.10.3) you can put nnsclo=-3 to -5, which improves a lot the stabilization of the magnetic moment in noncollinear.
Best wishes,
Eric
Re: Band gap of FeO using DFT+U calculations
Posted: Sun Aug 18, 2019 2:21 pm
by ketong
Dear Eric
ebousquet wrote:In the mean time, I can already tell you that for the non-collinear calculation+soc you have to remove all the symmetries (nsym=1) because the symmetries are not yet validated for non-collinear and fails in some specific cases. You are also going to face more converge problems and you'll have to reduce diemix (0.5 - 0.2) and diemixmag (0.4 - 0.1, the exact value depends on the system, if too small the convergence can works but could take much more SCF so the larger values that works is the best). To help more you can also put nline=10 or 12 and in the latest version (8.10.3) you can put nnsclo=-3 to -5, which improves a lot the stabilization of the magnetic moment in noncollinear.
Best wishes,
Eric
Thank you for your advice. I do the non-collinear+soc calculation according to your suggestion. The DOS figure can be found in the attachments. At this time, the right DOS of FeO can be obtained. The magnitude of the magnetic moment is about 3.59 μB. Therefore, this calculation implies that the spin-orbit coupling is essential to consider the FeO system. In addition, the input file is also attached.
ebousquet wrote:I've not looked yet to the DOS problem within collinear calculations, but now seeing the plot it is clearly not a smearing problem, there is a big pick at the Fermi level. Did you test with the usual 4 atoms AFM cell before?
Stable FeO was an antiferromagnetic AFII structure, in which the magnetic moments are ferromagnetically coupled in the (111) planes and the (111) planes are antiferromagnetically coupled to each other. In this case, there are at least 12 atoms in a FeO unit cell. I guess that "4 atoms AFM cell" is the unit cell in which the magnetic moments are ferromagnetically coupled in the (001) planes and the (001) planes are antiferromagnetically coupled to each other. Is it right?
Best regards
Ke Tong
