My Community

Dear VASP developers,

I did spin spiral calculations with q vector = (0,0,0) on LaNiO2 with PBE+U. It yields the magnetic moment  ± 0.648 μB and total energy: -53.49183176 eV. However when performing collinear calculations it yields: ± 0.975 μB and total energy: -53.53212588 eV. We expect the magnetic moment and total energy to be the same for both cases.
I also tested SCAN functional and got the same behavior.

Note: I used the interactive mode on NERSC-Perlmutter, so I do not attach any jobscript.

I was redoing your calculations and got the same results.
I noticed that in one case you are using spin-orbit coupling and in the other not. So I would expect that the results differ.

Dear VASP developers,
I tested 3 calculations of LaNiO2 (with no spin orbit coupling):
1. Collinear with PBE+U(4eV)
2. Non collinear only with PBE+U(4eV), with the tag LNONCOLLINEAR = .True.
3. Non collinear PBE+U(4eV), with the tag LNONCOLLINEAR = .True., LSPIRAL=.TRUE., and QSPIRAL = 0.0 0.0 0.0

I’d expect to get the same energy and magnetic moment for all the calculations, but this is not the case for the spiral.
For each case, respectively, I’m getting:
1. Collinear: -53.5321254 eV, 0.975 μB
2. Non collinear only: -53.53228102 eV, 0.976 μB
3. Non collinear with spiral: -52.87347456 eV, 0.654 μB

First of all, I am sorry for the very delayed answer.
I had a look at your calculations and I have a couple of recommendations:

1. I see you probably had some trouble with the SCF cycle so you tuned the mixing parameters. This is in principle ok but you need to be very carefull to ensure that the calculation does not stop too soon. If the mixing is too low then the changes in total energy will also be small and the SCF loop might stop earlier than it should. For these difficult to converge cases, I would suggest instead that you remove the custom mixing parameters and use ALGO=All instead.
2. I believe you should turn off symmetries (ISYM=-1) when performing spin spiral calculations.

Let me know if you recover the behaviour you would expect.