# How to choose starting magnetization while doing spin polarised calculation in Quantum ESPRESSO?

I am trying to run spin polarised DFT calculation in Quantum ESPRESSO on system Co2MnSi. I kept starting magnetization for all these three species as,

starting_magnetization(Co) = 0.3,
starting_magnetization(Mn) = 0.5,
starting_magnetization(Si) = 0.2,


I have used the fact that Cobalt has 3 valance electrons in up, hence positive 0.3. Similarly for Mn and Si.Convergence is not achieved even after 100 iterations using these values as starting magnetization.

1)Is this the wrong assumption I am using for initialising starting_magnetization? If yes, What should be right fact for choosing initialization value?

2)Can I also use 'tetrahedra' smearing for magnetic systems?

Thanks!

Is this the wrong assumption I am using for initializing starting_magnetization? If yes, What should be the right fact for choosing the initialization value?

The initial magnetization is not so important. I will tell you how to generate the input file with a convenient method in which the initial magnetization will be considered automatically. I will assume that you are calculating the MnCo2Si in the ferromagnetic phase. The structure is shown below and can be download from the materials project database:

Once you obtain the structure, you can generate an input file for self-consistent calculation with this online website:

The input file is as follows:

In particular, the downloaded PWscf.zip file will contain the corresponding pseudopotential files.

Can I also use 'tetrahedra' smearing for magnetic systems?

The system is metal, you should use the smearing method properly.

• Do you mean that I should use 'cold' smearing and shouldn't use 'tetrahedra'? Of course this is a metal and 'cold' smearing is especially designed for metals. But I am trying to plot DOS also and if I'm using 'cold' smearing then I have to manually converge the degauss value which is done by 'tetrahedra' method automatically, I guess. Thanks for the answer. – Ujjawal M. Nov 21 at 18:00
• Isn't starting magnetization helps in converging the systems total energy? Why don't the you choose simply 1,2,3 as starting magnetization if it doesn't matter. Is the calculation independent of choosing these values? There should be some rule for choosing the right initialization, I guess. – Ujjawal M. Nov 21 at 18:06
• I believe you should do the smearing optimization even to the tetrahedra method. The MV cold smearing has a very flat response when compared to the gaussian one for example. Thus, the convergence process should be easy. – Anibal Bezerra Nov 21 at 19:20
• @UjjawalM. In VASP, we can use the smearing method to do the self-consistent calculation and then use the tetrahedra method to do the non-self-consistent calculation by reading the self-consistent charge density, like DOS. I think you can take the same strategy in QE. – Jack Nov 22 at 1:32