M_INT versus MW_INT in constrained magnetic calculations
Posted: Tue Apr 23, 2024 2:04 pm
Hello,
I wanted to check something about constrained magnetic calculations using I_CONSTR_M=1 (constraining direction, not magnitude). I know that in the OSZICAR file, two values are printed related to magnetic moments, one being the M_INT, or the magnetization density integrated over RWIGS, and MW_INT which is the so-called "smoothed integrated moment" which is the value that actually enters the penalty energy. I know from the example on the VASP wiki that it is normal and expected for the magnitudes of these values to be slightly different, however, looking at my calculations recently, I noticed that my MW_INT is really quite a bit smaller than my non-smoothed, M_int (like 1.6 uB versus 2.6 muB). I never really worried about this because in my calculations I was able to get the moments to point exactly in the direction I wanted them, my penalty energies were really tiny, the total energies do what I expected as I change the direction of the moments, and the calculations are extremely reproducible across different VASP versions and architectures. But I thought I should check whether indeed, such a difference in the values of M_int and MW_int is not problematic provided that the value of MW_int is "large enough", and moreover, how I can tell what "large enough" constitutes.
I attached example input files and relevant outputs for reference.
Thanks very much in advance!
Best
Sophie
I wanted to check something about constrained magnetic calculations using I_CONSTR_M=1 (constraining direction, not magnitude). I know that in the OSZICAR file, two values are printed related to magnetic moments, one being the M_INT, or the magnetization density integrated over RWIGS, and MW_INT which is the so-called "smoothed integrated moment" which is the value that actually enters the penalty energy. I know from the example on the VASP wiki that it is normal and expected for the magnitudes of these values to be slightly different, however, looking at my calculations recently, I noticed that my MW_INT is really quite a bit smaller than my non-smoothed, M_int (like 1.6 uB versus 2.6 muB). I never really worried about this because in my calculations I was able to get the moments to point exactly in the direction I wanted them, my penalty energies were really tiny, the total energies do what I expected as I change the direction of the moments, and the calculations are extremely reproducible across different VASP versions and architectures. But I thought I should check whether indeed, such a difference in the values of M_int and MW_int is not problematic provided that the value of MW_int is "large enough", and moreover, how I can tell what "large enough" constitutes.
I attached example input files and relevant outputs for reference.
Thanks very much in advance!
Best
Sophie