Magnetic moments of Ce2In


#1

I am validating some entries in Materials Project. I found the volume of Ce2In (icsd: 621365) is very smaller than one in OQMD.

mp_volume_per_atom: 29.044630
oqmd_volume_per_atom: 48.785477

I got the structure of Ce2In as follows:

mag_s = m.get_structure_by_material_id('mp-19733')
print(mag_s)

Full Formula (Ce4 In2)
Reduced Formula: Ce2In
abc   :   5.482131   5.482131   6.695572
angles:  90.000000  90.000000 120.000000
Sites (6)
  #  SP           a         b     c    magmom
---  ----  --------  --------  ----  --------
  0  In    0.666667  0.333333  0.75     0.795
  1  In    0.333333  0.666667  0.25     0.795
  2  Ce    0.333333  0.666667  0.75     0.981
  3  Ce    0.666667  0.333333  0.25     0.981
  4  Ce    0         0         0.5     -0.045
  5  Ce    0         0         0       -0.045

In sites have large local magnetic moment. Is this magnetic structure correct? The corresponding entry in OQMD has the following structure, and the moments at In sites are smaller than ones at Ce sites. I think the magnetic structure of OQMD is usual.

Full Formula (Ce4 In2)
Reduced Formula: Ce2In
abc   :   6.391284   6.393282   8.270970
angles:  90.170184  90.000000 119.989650
Sites (6)
  #  SP           a         b         c    magmom
---  ----  --------  --------  --------  --------
  0  Ce    0.999958  0.999916  0.999965     0.243
  1  Ce    0.666172  0.332341  0.249714     0.165
  2  Ce    0.999924  0.99985   0.500045     0.241
  3  Ce    0.33379   0.667581  0.750109     0.166
  4  In    0.333531  0.667062  0.250201    -0.016
  5  In    0.666626  0.33325   0.749965    -0.016

#2

Hi @Tony-Y,

This is a good question! Without an experimental value (which I don’t have to hand), I wouldn’t like to claim which is correct.

There are a lot of A_2B compounds in P6_3/mmc that have interesting magnetic characteristics, and I wouldn’t be surprised if the computation is not getting the correct magnetic ground state. This is for three reasons:

  1. Magnetism is only handled in the collinear case without spin-orbit coupling for scale reasons in the Materials Project (and also in OQMD, to my knowledge). This could be significant in this system.

  2. Choice of initial magnetic moments matters, since getting trapped in a local minima is very easy. There is no correct answer for initial magnetic moments to choose; we choose values that typically give the correct results, but there are edge cases.

  3. Ce in particular is not a nice element to simulate using DFT.

So, in summary, I’d say it’s entirely possible that both MP and OQMD are incorrect in this case, though whichever database more accurately matches the experimental lattice parameters I would favor. However, there are other calculation factors that can affect volume change, and this could be something entirely unrelated to the magnetic order. I would also say that if you’re interested in a specific material in depth, it is worth doing your own calculations explicitly including SOC and trying a variety of magnetic orderings (I would not be surprised if the true ground state here has a lot more ferrimagnetic character).

Please do let us know if you find anything more about this system though! I would be interested to hear about it.

Best,

Matt


#3

Thank you for a nice explanation. However, I can’t figure out a source of the local magnetic moments at In sites in the element spin densities of states.

https://materialsproject.org/electronic_structure/dos/plot/mp-19733

Is there something wrong with the local spin density of states?

Edit

Complete DOS for Full Formula (Ce4 In2)
Reduced Formula: Ce2In
abc   :   5.482131   5.482131   6.695572
angles:  90.000000  90.000000 120.000012
Sites (6)
  #  SP           a         b     c
---  ----  --------  --------  ----
  0  Ce    0.333333  0.666667  0.75
  1  Ce    0.666667  0.333333  0.25
  2  Ce    0         0         0.5
  3  Ce    0         0         0
  4  In    0.666667  0.333333  0.75
  5  In    0.333333  0.666667  0.25

The order of atoms is different from one in the structure given by get_structure_by_material_id. Which order of atoms does the order of magmom match?

Edit2

I have made site-projected DOSs, and plotted them as a stacked area graph:

site_dos_Ce2In

This graph clearly shows the order of magmom matches the order of atoms in Complete DOS. So, the following magnetic structure would be correct.

  #  SP           a         b     c    magmom
---  ----  --------  --------  ----  --------
  0  Ce    0.333333  0.666667  0.75     0.795
  1  Ce    0.666667  0.333333  0.25     0.795
  2  Ce    0         0         0.5      0.981
  3  Ce    0         0         0        0.981
  4  In    0.666667  0.333333  0.75    -0.045
  5  In    0.333333  0.666667  0.25    -0.045

#4

Ah, so the magmom tags are ‘correct’ but have been permuted so they’re attached to the wrong atomic sites. This is a bug we saw previously but I thought had been fixed (note that this bug should be relatively rare, in the majority of cases the magnetic moments are attached to the correct site).

Thanks very much for investigating further and reporting back! We’ll address this in the next database release.


#5

I’ll update this thread when this has been resolved.