Antiferromagnetic Fe

[sponce24]

Hello,

Following yesterday's meeting, we tried to improve the results for Al and Fe for Abinit (these were off compared to other code).
The solution was to use PBE-PAW pseudopotential (like QE is using) rather than NC-PBE, see #113

With this we get:

As you can see all is good except Fe antiferromagnetic.

We tried a few things but cannot get it to agree with other codes. Also this case is the one with the largest spread for all codes.
I think this is due to the fact that we are doing a fake BCC antiferromagnetic Fe.
This is not the thermodynamically stable structure.

Therefore, I'm wondering if it would not be best to have a second Fe.cif with the FCC structure that we would use for the antiferromagnetic Fe ?

What do you think ?

[mbercx]

I would be in favour of at least checking this. 👍

I'm now running the EOS for this primitivized cell with Quantum ESPRESSO:

$ more Fe_afm.xyz
2
Lattice="2.57789919 0.0 0.0 0.0 2.57789919 0.0 0.0 0.0 3.6457" Properties=species:S:1:pos:R:3:tags:I:1 pbc="T T T"
Fe       0.00000000       0.00000000       0.00000000   1
Fe       1.28894960       1.28894960       1.82285000   2

This was obtained from here, and transformed into the cell above in order to have two site tags to assign the -2 and 2 magnetic moments to.

[sphuber]

@mbercx we had already actually discussed this and @bosonie has been running the EOS workchains for an AFM FCC Fe structure. Last time we spoke he didn't have all the results yet but maybe he has an update now?

[mbercx]

@sponce24 do you have a suitable reference for AFM FCC Fe? I suppose we would consider a planar AFM configuration?

[sponce24]

Hello,

Pure Fe is BCC ferromagnetic at low temperature.

The case of anti-ferromagnetism in iron is more delicate:

• In this paper, they suggest that a face-centred (gamma) iron would at low temperature result in anti-ferromagnetism.
• In this one they directly observe antiferromagnetism in FCC iron but it is a thin film.
• In this one by Pinski shows that the low temperature phase should be a type-II antiferromagnetic (alternating [111] ferromagnetic places) instead of type-I (alternating [100] plane).

So to answer @mbercx it guess it would not be planar AFM but in the 111 direction.

Again, if this is too complicated it is fine to use BCC and we can stress in the paper that the difference among codes is because we do an unstable materials that does not exists.

[bosonie]

To simulate the type-II antiferromagnetic structure we need a cell bigger than the conventional one with 4 atoms. The calculations are already quite demanding for being simple test case, therefore I would exclude this option.

Regarding the paper, we express in the final version that "the BCC structure is the most thermodynamically stable configuration only in the ferromagnetic arrangement", however Sebastian and I did not feel to create a logic linking between that and the fact that the results do not show such good agreement. Do we have any concrete proof that the comparison should be worse for unstable materials?

[yakutovicha]

Hey, I have a similar issue with CP2K, but in my case, the Fe-antiferro curve is completely off:

This is really strange because the ferro configuration with the same parameters comes out pretty well:

I tried to increase the precision by all the settings I am aware of, but it didn't help. Just for the record, this is what I tried:

1. Fix the total magnetization to 0.
2. Set a very large k-point grid: 20 20 20
3. Set MGRID CUTOFF to 1000
4. Use a very large basis: TZV2P-MOLOPT-SR-GTH-q16
5. Set EPS_DEFAULT to 1.0E-16
6. Use different starting magnetization: +4/-4, +2/-2.

The resulting EOS curve came out more-or-less the same.

I also contacted CP2K developers, and they couldn't obtain results that are different from the ones I have, even though the structure is perfectly antiferromagnetic:

Finally, I would like to share the absolute magnetization computed for the antiferro Fe with CP2K and QE:

    cell (ang^3)       QE (Bohr mag/cell)    CP2K (Bohr mag/cell)       Difference
     21.38                      2.34                   2.873               (+0.533)
     21.84                      2.58                   3.102               (+0.522)
     22.29                      2.81                   3.330               (+0.52)
     22.75                      3.03                   3.571               (+0.54)
     23.20                      3.25                   3.826              (+0.576)
     23.66                      3.48                   4.074              (+0.596)
     24.11                      3.72                   4.308               (+0.58)

I am currently at a loss. My next guess would be to try to compute FCC Fe. Did anyone try that already?

[bosonie]

@yakutovicha I tried the planar AFM arrangement. It is not the most stable configuration though. I will send you the data.

[yakutovicha]

@yakutovicha I tried the planar AFM arrangement. It is not the most stable configuration though. I will send you the data.

@bosonie, here are the results:

CP2K is doing much better here with respect to the BCC antiferro.

(the values on the x axis are wrong, but this is a minor detail)