Very confusing NSW = 0 and > 0

[yong_han]

Dear VASP developer,

Before I perform a SCF static self-consistent calculation for further calculating the band structure, I checked NSW = 0 and > 0. The only difference of input files in the attached two jobs is the NSW values in the INCAR files.

For NSW = 9999, I got:
FORCES: max atom, RMS 0.007869 0.004875
(This indicates the force convergence criterion is well reached: 0.007869 < |EDIFFG| = 0.008.)
and
energy without entropy= -600.15816739 energy(sigma->0) = -600.15816739

For NSW = 0, I got:
FORCES: max atom, RMS 0.152187 0.091868
(This indicates the force convergence criterion is NOT reached: 0.152187 > |EDIFFG| = 0.008.)
and
energy without entropy= -595.02541503 energy(sigma->0) = -595.02954398
(This energy value is significantly different from the above value from NSW = 9999.)

Questions:
1.
A SCF static self-consistent calculation requires NSW = 0, but it doesn't converge in this case at all. Doesn't it cause problems to use such a result (CHGCAR) to calculate energy bands?

2.
Can I directly use the output from NSW > 0, e.g., from NSW = 9999 (of course I should set LCHARG = T for getting CHGCAR), i.e., just skip the step of any SCF static self-consistent calculation with NSW = 0, to calculate the band structure?

3.
For anti-ferromagnetic systems containing many atoms, MAGMOM input on a single line in an INCAR file needs writing many 1 and -1. Do you have any simple ways (format) for writing this, e.g., like 20*(1, -1) instead of 1, -1, 1, -1, ......, 1, -1, or 20*(1, 1, -1, -1) instead of 1, 1, -1, -1, 1, 1, -1, -1, ......, 1, 1, -1, -1, etc.?

Thank You!
Yong

[pedro_melo]

Dear Yong,

Could you try both calculations with a different value for ALGO? Either Normal or, if the system isn't too big, Exact? Also, from your OUTCAR files I can see that even though for NSW=9999 you set IBRION=2, VASP did not perform any ionic relaxation.

Kind regards,
Pedro

[pedro_melo]

Dear Yong,

Regarding your other questions:

1. A SCF static self-consistent calculation requires NSW = 0, but it doesn't converge in this case at all. Doesn't it cause problems to use such a result (CHGCAR) to calculate energy bands?

If the SCF run did not reach convergence, it cannot be guaranteed that the bands you are going to plot have any meaning, since the ground-state density has not been optimised.

2. Can I directly use the output from NSW > 0, e.g., from NSW = 9999 (of course I should set LCHARG = T for getting CHGCAR), i.e., just skip the step of any SCF static self-consistent calculation with NSW = 0, to calculate the band structure?

Yes, provided that convergence was achieved, there should be no problems in doing so.

3. For anti-ferromagnetic systems containing many atoms, MAGMOM input on a single line in an INCAR file needs writing many 1 and -1. Do you have any simple ways (format) for writing this, e.g., like 20*(1, -1) instead of 1, -1, 1, -1, ......, 1, -1, or 20*(1, 1, -1, -1) instead of 1, 1, -1, -1, 1, 1, -1, -1, ......, 1, 1, -1, -1, etc.?

You can specify several values for the starting spin using the syntax described here wiki/index.php/MAGMOM. For the non-collinear case however, I am afraid that you must specify them line by line whenever they are different from zero.

Kind regards,
Pedro