Starting an unconstrained collinear magnetism calculation from a pre-converged constrained calculation

[mdh1992]

Dear All,

I am working on some calculations to explore surface magnetism in Ni. However, in some cases, I am struggling to obtain the correct ordering of magnetic moments (i.e. ferromagnetic ordering). I was hoping to perform some calculations applying magnetic constraints (namely, constraining the direction of magnetic moments with I_CONSTRAINED_M = 1), and then restart an unconstrained calculation from either the WAVECAR or CHGCAR, to ensure I get the correct magnetic ordering. I have tried a few times using MAGMOM with large initial moments in an attempt to predispose the calculation towards converging on the correct ordering, but this has so far failed, hence, the decision to try applying constraints.

As others have mentioned (forum/viewtopic.php?f=4&t=16929&sid=d78 ... 3e516144df), applying magnetic constraints requires one to invoke tags for non-collinear magnetism, even if one ideally would like to do constrained collinear calculations. I have tried this and the results are broadly acceptable - I set the constraints using M_CONSTR for each atom with a nominal value for M_x and M_y and a larger value for M_z in an attempt to initialise the magnetic moment along the z-axis i.e. the default SAXIS = 0 0 1. The calculation converged and the magnetisation data in the output file shows only small magnetic moments for each atom for the x and y direction (~0.015), with much larger moments along the z direction (~0.6), and all positively aligned, i.e. ferromagnetically ordered.

My issue is with taking the output from this calculation to hopefully converge on the correct magnetic ordering for an unconstrained collinear calculation. Naively trying to start the unconstrained collinear calculation from the WAVECAR obtained from the non-collinear constrained calculation fails and I get an error message indicating that the plane wave coefficients changed. Is there a way I can resolve this, or is it simply not possible for VASP to extract data from the WAVECAR to initialise a collinear calculation? I also tried starting from the CHGCAR, the calculation ran without encountering any errors, but the magnetic moments obtained were incorrect and did not resemble those obtained for the z direction in the constrained non-collinear calculation. My impression is that in this case, the magnetisation data read in from the CHGCAR file was that associated with either the x or y direction for the non-collinear calculation, and not the z direction, which is what I had hoped for. Is there a way to ensure that VASP reads the initial magnetic moments correctly?

I would be most appreciative of any advice or information. Any suggestions would be welcome regarding strategies for using pre-converged constrained calculations to start unconstrained magnetic calculations.

[henrique_miranda]

Is there a way I can resolve this, or is it simply not possible for VASP to extract data from the WAVECAR to initialise a collinear calculation?

In the current version of the code it is not possible to start a collinear calculation from a non-collinear one.
But you should be able to continue the non-collinear calculations without constrains from the calculation you performed with constraints.