 Re: Results of Mulliken populations from different of EigenvalueSolver, Band and Cluster. ( No.1 ) |
- Date: 2024/12/08 14:39
- Name: Naoya Yamaguchi
- Dear Kikuchi-san,
It is only natural that the results of the `Cluster` and `Band` calculations differ. For `Cluster`, `scf.Kgrid` is not used. Only the Γ point is used.
>I would like to know that what is the cause of different between cluster and band, also the reasons for differences from experimental values and calculation results of other programs.
It should be confirmed whether SCF convergence was obtained in the first place.
Regards,
Naoya Yamaguchi
|
| Re: Results of Mulliken populations from different of EigenvalueSolver, Band and Cluster. ( No.2 ) |
- Date: 2024/12/09 03:56
- Name: Soichiro Kikuchi
- Dear Dr.Yamaguchi,
Thank you very much for your reply.
>It is only natural that the results of the `Cluster` and `Band` calculations differ. For `Cluster`, `scf.Kgrid` is not used. Only the Γ point is used.
When are 'cluster' and 'band' used differently? As far as I can see from the example file, the “Mol_MnO_NC.dat”, which was calculating spin moments, confirms that clusters were used.
>It should be confirmed whether SCF convergence was obtained in the first place.
Is SCF convergence considered to have been achieved if the number of times the “SCF history” in the out file is less than the number set in “scf.maxIter”? If yes, both 'cluster' and 'band' appear to be converging SCF.
Regards,
Soichiro Kikuchi
|
| Re: Results of Mulliken populations from different of EigenvalueSolver, Band and Cluster. ( No.3 ) |
- Date: 2024/12/09 22:21
- Name: Naoya Yamaguchi
- Dear Kikuchi-san,
>When are 'cluster' and 'band' used differently? As far as I can see from the example file, the “Mol_MnO_NC.dat”, which was calculating spin moments, confirms that clusters were used.
This is because `Mol_MnO_NC.dat` is an input file for a MnO **molecule**.
Your computational model is the bulk system (solid), and you should use `Band` as appropriate.
>Is SCF convergence considered to have been achieved if the number of times the “SCF history” in the out file is less than the number set in “scf.maxIter”? If yes, both 'cluster' and 'band' appear to be converging SCF.
If so, the SCF calculations were converged.
On what points do you say the results do not match?
Regards,
Naoya Yamaguchi
|
| Re: Results of Mulliken populations from different of EigenvalueSolver, Band and Cluster. ( No.4 ) |
- Date: 2024/12/11 02:18
- Name: Soichiro Kikuchi
- Dear Dr.Yamaguchi,
Thank you very much for your reply.
>On what points do you say the results do not match?
First, the magnitude of the total spin moment is too small: according to VASP calculations, the total spin moment is about 3 mμB when the same lattice constant and atomic positions are used, and experimental results have been reported as 3.6 mμB.
Second, the direction of the total spin moment is different: in the VASP calculation results, the direction of the total spin moment is θ=90, Φ=0,or 180. That is, it was oriented toward the a-axis.
Final, the magnitude and orientation of the spin moment of the Mn site is different: VASP calculations show that the magnitude of the spin moment of the Mn site is about 2.9 μB, while experiments show it to be ~3 μB. The direction of the spin moment was equal to the initial Euler angle set in the input file.
I looked at the convergence to “scf.Energycutoff” and it seemed to converge from about 400, but the total spin moment was small and the direction was different.
・output file
scf.energycutoff 450
scf.kgrid 6 6 6
Total spin moment (muB) 0.334062054 Angles (Deg) 89.999701824 -41.453344008
Up Down Sum Diff theta phi
1 Mn 8.87736 5.17397 14.05133 3.70339 89.99997 -1.52704
2 Mn 8.87736 5.17399 14.05135 3.70337 90.00003 -1.52708
3 Mn 8.87895 5.17138 14.05033 3.70757 90.00003 242.13055
4 Mn 8.87895 5.17137 14.05032 3.70758 89.99998 242.13024
5 Mn 8.89357 5.16276 14.05633 3.73080 89.99999 119.42125
6 Mn 8.89357 5.16276 14.05633 3.73081 90.00000 119.42163
7 Sn 8.43073 8.41128 16.84201 0.01946 89.99396 124.58215
8 Sn 8.43072 8.41126 16.84199 0.01946 90.00194 124.58478
Also, I looked at convergence to “scf.kgird". But the results were different.
・output file
scf.energycutoff 450
scf.kgrid 9 9 9
Total spin moment (muB) 0.269053458 Angles (Deg) 90.015060477 114.578895210
Up Down Sum Diff theta phi
1 Mn 8.87773 5.13444 14.01217 3.74329 89.91360 1.29113
2 Mn 8.87776 5.13442 14.01218 3.74334 90.08639 1.29209
3 Mn 8.87770 5.13448 14.01217 3.74322 90.01671 238.82282
4 Mn 8.87770 5.13448 14.01218 3.74321 89.98079 238.82347
5 Mn 8.87643 5.13694 14.01337 3.73949 89.91614 119.86283
6 Mn 8.87645 5.13692 14.01337 3.73954 90.08744 119.86085
7 Sn 8.48283 8.47946 16.96229 0.00337 110.73705 -67.24003
8 Sn 8.48255 8.47971 16.96226 0.00284 65.15074 -51.16351
scf.energycutoff 450
scf.kgrid 12 12 12
Total spin moment (muB) 0.141361498 Angles (Deg) 89.983072923 105.823877152
Up Down Sum Diff theta phi
1 Mn 8.87287 5.13739 14.01026 3.73548 89.99885 0.73203
2 Mn 8.87286 5.13744 14.01030 3.73542 90.00119 0.73271
3 Mn 8.87314 5.13695 14.01009 3.73620 89.99880 239.47068
4 Mn 8.87315 5.13694 14.01009 3.73621 90.00101 239.47040
5 Mn 8.87304 5.14007 14.01311 3.73297 89.99867 119.82519
6 Mn 8.87305 5.14004 14.01309 3.73301 90.00111 119.82488
7 Sn 8.48337 8.48317 16.96655 0.00020 95.99304 -10.69435
8 Sn 8.48330 8.48322 16.96652 0.00007 59.01283 -54.65826
scf.energycutoff 450
scf.kgrid 15 15 15
Total spin moment (muB) 0.126131219 Angles (Deg) 90.024145086 115.428302539
Up Down Sum Diff theta phi
1 Mn 8.87198 5.13844 14.01042 3.73353 90.00445 0.62639
2 Mn 8.87198 5.13843 14.01041 3.73355 89.99801 0.62605
3 Mn 8.87204 5.13834 14.01038 3.73370 89.99813 239.41579
4 Mn 8.87204 5.13836 14.01040 3.73368 90.00003 239.41628
5 Mn 8.87045 5.14013 14.01058 3.73032 90.00353 119.95053
6 Mn 8.87045 5.14013 14.01058 3.73032 89.99665 119.94923
7 Sn 8.48551 8.48312 16.96863 0.00239 90.91317 -46.98674
8 Sn 8.48551 8.48309 16.96860 0.00242 89.12571 -47.03991
scf.energycutoff 450
scf.kgrid 18 18 18
Total spin moment (muB) 0.236952624 Angles (Deg) 90.004090857 119.042462480
Up Down Sum Diff theta phi
1 Mn 8.87619 5.14184 14.01803 3.73435 89.99955 1.13081
2 Mn 8.87619 5.14185 14.01804 3.73434 90.00058 1.13047
3 Mn 8.87616 5.14196 14.01812 3.73421 89.99963 238.87284
4 Mn 8.87617 5.14195 14.01812 3.73421 90.00072 238.87443
5 Mn 8.87475 5.14425 14.01900 3.73051 89.99941 119.97682
6 Mn 8.87475 5.14424 14.01900 3.73051 90.00038 119.97599
7 Sn 8.47537 8.46948 16.94485 0.00589 90.04942 -58.09107
8 Sn 8.47536 8.46947 16.94483 0.00588 89.95098 -58.20305
 |
| Re: Results of Mulliken populations from different of EigenvalueSolver, Band and Cluster. ( No.5 ) |
- Date: 2024/12/11 22:59
- Name: Naoya Yamaguchi
- Dear Kikuchi-san,
>I looked at the convergence to “scf.Energycutoff” and it seemed to converge from about 400, but the total spin moment was small and the direction was different.
I don't think so. The initial condition in your input file was antiferromagnetic.
There is also a symmetry problem, so if you want to have a strict antiferromagnetic order, you should consider fixing the spin direction by putting a constraint condition:
https://www.openmx-square.org/openmx_man3.9/node113.html
`scf.ElectronicTemperature 0.0` may make some problems.
Regards,
Naoya Yamaguchi
|
| Re: Results of Mulliken populations from different of EigenvalueSolver, Band and Cluster. ( No.6 ) |
- Date: 2024/12/12 02:21
- Name: Soichiro Kikuchi
- Dear Dr.Yamaguchi,
Thank you very much for your reply.
>I don't think so. The initial condition in your input file was antiferromagnetic.
So should “scf.energycutoff” be set larger? Are there any problems if it is set too large?
>There is also a symmetry problem, so if you want to have a strict antiferromagnetic order, you should consider fixing the spin direction by putting a constraint condition:
Recalculate with constraints. But does it mean that it is difficult to converge to strict antiferromagnetic order without constraints?
>`scf.ElectronicTemperature 0.0` may make some problems.
The calculation results showed that the spin moment was in the wrong direction even at 300. But, recalculate with `scf.ElectronicTemperature 300.0`
Regards,
Soichiro Kikuchi
|
| Re: Results of Mulliken populations from different of EigenvalueSolver, Band and Cluster. ( No.7 ) |
- Date: 2024/12/12 15:55
- Name: Naoya Yamaguchi
- Dear Kikuchi-san,
>The calculation results showed that the spin moment was in the wrong direction even at 300.
I think that the spin direction of each atom is reasonable. On what points do you say the spin moment was in the wrong direction?
My comment “I don't think so.” was given for “the total spin moment was small and the direction was different.”
>So should “scf.energycutoff” be set larger? Are there any problems if it is set too large?
>Recalculate with constraints. But does it mean that it is difficult to converge to strict antiferromagnetic order without constraints?
It could resolve the symmetry problem to some extent, but the constraint is a good way because OpenMX doesn’t consider such symmetries, and the difference of the spin direction may give too small change to find the global minimum solution.
Regards,
Naoya Yamaguchi
|
| Re: Results of Mulliken populations from different of EigenvalueSolver, Band and Cluster. ( No.8 ) |
- Date: 2024/12/12 17:16
- Name: Soichiro Kikuchi
- Dear Dr.Yamaguchi,
Thank you very much for your reply.
>I think that the spin direction of each atom is reasonable. On what points do you say the spin moment was in the wrong direction?
In the VASP calculation results, for the atomic configuration and lattice constants used, the spin moment orientation of each Mn site was the orientation of the initial configuration I set, and the total spin moment orientation was parallel to the x axis, or a axis, Φ = 180.
This result was confirmed to some extent by experiments.
However, the OPENMX calculation results showed that the orientation of each Mn site was also slightly rotated and the direction of the total spin moment was Φ≠ 180, which did not agree with the other calculations.
If it is possible to reproduce the results of the VASP calculations and results close to the experiments, I would like to reproduce them.
Regards,
Soichiro Kikuchi
|
| Re: Results of Mulliken populations from different of EigenvalueSolver, Band and Cluster. ( No.9 ) |
- Date: 2024/12/12 19:13
- Name: Naoya Yamaguchi
- Dear Kikuchi-san,
>In the VASP calculation results, for the atomic configuration and lattice constants used, the spin moment orientation of each Mn site was the orientation of the initial configuration I set, and the total spin moment orientation was parallel to the x axis, or a axis, Φ = 180.
The initial spin of Mn is given by the antiferromagnetic order, and the total spin moment is expected to be zero in magnitude, so the orientation of the total spin moment should not be determined, Obviously, the total sum is zero, since a triangle can be formed by the vectors. Therefore, I do not know what you mean in your last sentence in the above.
Regards,
Naoya Yamaguchi
|
| Re: Results of Mulliken populations from different of EigenvalueSolver, Band and Cluster. ( No.10 ) |
- Date: 2024/12/12 22:40
- Name: Soichiro Kikuchi
- Dear Dr.Yamaguchi,
Thank you very much for your reply.
I did understand that the vectors would cancel each other out and become 0, given the direction of the vectors.
However, the experimental results and the VASP calculations I have seen confirm a moment of about 3 mμB in magnitude, which I hope to reproduce.
Should I try to change the spin initial configuration of each Mn site to produce a total spin moment?
Regards,
Soichiro Kikuchi
|
| Re: Results of Mulliken populations from different of EigenvalueSolver, Band and Cluster. ( No.11 ) |
- Date: 2024/12/12 23:43
- Name: Naoya Yamaguchi
- Dear Kikuchi-san,
>I did understand that the vectors would cancel each other out and become 0, given the direction of the vectors.
>However, the experimental results and the VASP calculations I have seen confirm a moment of about 3 mμB in magnitude, which I hope to reproduce.
If you understand that, your previous comment below seems to be a misunderstanding.
>First, the magnitude of the total spin moment is too small: according to VASP calculations, the total spin moment is about 3 mμB when the same lattice constant and atomic positions are used, and experimental results have been reported as 3.6 mμB.
Since your OpenMX results show 3.7 μB, the correct interpretation is that it is rather LARGE than small.
Although the results are closer to the experimental values (3.6 μB) than VASP (3 μB), it is sometimes difficult to reproduce the magnetic moment values due to differences in detailed parameters and conditions between first-principles calculation software.
First you said the experimental value was 3.6 μB, and now you say it is about 3 μB, so finally which is correct?
Regards,
Naoya Yamaguchi
|
| Re: Results of Mulliken populations from different of EigenvalueSolver, Band and Cluster. ( No.12 ) |
- Date: 2024/12/13 01:24
- Name: Soichiro Kikuchi
- Dear Dr.Yamaguchi,
Thank you very much for your reply.
>Since your OpenMX results show 3.7 μB, the correct interpretation is that it is rather LARGE than small.Although the results are closer to the experimental values (3.6 μB) than VASP (3 μB),
I believe the 3.7μB is the magnitude of 'the spin moment of the Mn site'. The 3.6 mμB and 3 mμB I mentioned in my previous comment are the magnitude of the total spin moment. The units are 'm'μB.
The problem with the magnitude of the spin moment of the Mn site is that:
--Final, the magnitude and orientation of the spin moment of the Mn site is different: VASP calculations show that the magnitude of the spin moment of the Mn site is about 2.9 μB, while experiments show it to be ~3 μB. The direction of the spin moment was equal to the initial Euler angle set in the input file.--
>First you said the experimental value was 3.6 μB, and now you say it is about 3 μB, so finally which is correct?
The explanation was wrong. The experimental value is 3.6 mμB and the VASP calculation is 3 mμB. This is the magnitude of the total spin moment for both.
To summarize again, despite the antiferromagnetic order, the total spin moment of 3-3.6'm'μB is confirmed by experimental values and VASP calculations, but my calculations do not reproduce it. Additionally, the direction of the total spin moment and the magnitude of the spin moment at the Mn site also do not reproduce the experimental and calculated values.
Sorry for the confusing explanation and poor understanding. I am extremely grateful for your response.
Regards,
Soichiro kikuchi
|
| Re: Results of Mulliken populations from different of EigenvalueSolver, Band and Cluster. ( No.13 ) |
- Date: 2024/12/15 16:51
- Name: Naoya Yamaguchi
- Dear Kikuchi-san,
I see. I missed "milli-". Although I've not been familiar with Mn3Sn, you meant the weak ferromagnetism. I obtained the following result of the total spin moment, and it is reasonable and almost reproduced.
```
Total spin moment (muB) 0.023347652 Angles (Deg) 90.001312157 118.034143300
```
The following are the part including changes made in the input files.
```
Species.Number 2
<Definition.of.Atomic.Species
Mn Mn6.0-s3p2d1 Mn_PBE19
Sn Sn7.0-s3p2d2 Sn_PBE19
Definition.of.Atomic.Species>
Atoms.Number 8
Atoms.SpeciesAndCoordinates.Unit FRAC
<Atoms.SpeciesAndCoordinates
1 Mn 0.8418997760617231 0.6837995580507723 0.2499994261181939 9.0 6.0 90.0 0.0 90.0 0.0 0 off
2 Mn 0.1580997540495446 0.3161995080990891 0.7500005346670946 9.0 6.0 90.0 0.0 90.0 0.0 0 off
3 Mn 0.3162006822288461 0.1581007910616492 0.2499994261181939 9.0 6.0 90.0 -120.0 90.0 -120.0 0 off
4 Mn 0.6837998848945261 0.8419003491124215 0.7500005346670946 9.0 6.0 90.0 -120.0 90.0 -120.0 0 off
5 Mn 0.8419001029054840 0.1581007910616492 0.2499994261181939 9.0 6.0 90.0 120.0 90.0 120.0 0 off
6 Mn 0.1581004642178954 0.8419003491124215 0.7500005346670946 9.0 6.0 90.0 120.0 90.0 120.0 0 off
7 Sn 0.3333330220499562 0.6666660440999053 0.2499994261181939 7.0 7.0 0.0 0.0 0.0 0.0 0 off
8 Sn 0.6666665080613186 0.3333330220499562 0.7500005346670946 7.0 7.0 0.0 0.0 0.0 0.0 0 off
Atoms.SpeciesAndCoordinates>
scf.ElectronicTemperature 100.0
scf.EigenvalueSolver band
scf.Kgrid 9 9 11
scf.Mixing.Type rmm-diisv
scf.Init.Mixing.Weight 0.01
scf.Min.Mixing.Weight 0.001
scf.Max.Mixing.Weight 0.01
scf.Mixing.History 40
scf.Mixing.StartPulay 20
```
The cutoff radii of PAOs were set to that for bulk systems. You can refer to https://www.openmx-square.org/openmx_man3.9/node27.html for the guideline for choice of PAOs. Since it is a fairly small value, I think it is necessary to investigate the dependence on cutoff energy, k-point sampling, and PAOs.
Regards,
Naoya Yamaguchi
|
|
Results of Mulliken populations from different of EigenvalueSolver, Band and Cluster.