Author Topic: how can we set the "initial_scaled_spin" reasonably for our calculation.  (Read 4403 times)

0 Members and 1 Guest are viewing this topic.

Offline alan

  • Regular QuantumATK user
  • **
  • Posts: 23
  • Reputation: 0
    • View Profile
Hi

I am confused with setting the value of the "initial_scaled_spin" for the spin-polarized calculation.
Does the different value for the initial_scaled_spin will effect the last  result ,such as the DOS or transmission?Is there any difference with the calculation between  the initial_scaled_spin =0 and 1(-1) for the reslut,besides the speed of  computing.
what is the reasonable understanding of the initial_scaled_spin in our calculation? which physical quantity should we caluate or analyse before?
If for the unhomogeneoustwo-probe system, can we set the uniform value for different atoms? is that correct? if we can not ,what is the Process for we get the initial_scaled_spin?
In a word, i want to know how can set the correct and reasonable value for initial_scaled_spin for any system.

Need your help

Offline zh

  • QuantumATK Support
  • Supreme QuantumATK Wizard
  • *****
  • Posts: 1141
  • Reputation: 24
    • View Profile
Before doing the spin-polarized calculation of a two-probe system, one had better separately do the spin-polarized calculations for the electrodes and the conductor object (in the center region and exclude the surface layers) to find out their most stable magnetic states( ferromagnetic, antiferromagnetic, non-magnetic, or other states). If the systems are magnetic, you would obtain the local magnetic moment for each atom in these systems. These obtained results will be good starting point for the setup of "initial_scaled_spin" for the spin-polarized calculation of the corresponding two-probe system.

The different values for the "initial_scaled_spin" of each atom could lead to the calculations being converged to different magnetic states.  They could also affect the convergence speed of self-consistent calculation.

For a two-probe system, the setup of "initial_scaled_spin" depends on the magnetic state of each part (electrodes and conductor) of two-probe system. It also depends on the problem under study.

For example, see the following cases:
i) the two electrodes are non-magnetic, and the conductor is magnetic:
  "initial_scaled_spin" for atoms in electrodes and the ones in the surface layers of center region:  0.0
ii) the two electrode are ferromagnetic, and the conductor is nonmagnetic:
  "initial_scaled_spin" for atoms in the conductor:  0.0
   There are two spin configurations for other parts:
    a) parallel spin alignment: "initial_scaled_spin" for atoms in left electrode and the ones in the left surface layers of center region are same as those of right electrode and the right surface layer, respectively.
   b) anti-parallel spin alignment: "initial_scaled_spin" for atoms in left electrode and the ones in the left surface layers of center region are opposite to those of right electrode and the right surface layer, respectively.



Offline alan

  • Regular QuantumATK user
  • **
  • Posts: 23
  • Reputation: 0
    • View Profile
Thanks for your reply.
another question: how to find their most stable magnetic states( ferromagnetic, antiferromagnetic, non-magnetic, or other states)?
to test and  find their most stable magnetic states.can we just by setting the different value (0.0,1.0,-1.0 or a random number among -1,1) as the initial_scaled_spin.and Comparing their total energy,chose the configuration with lowest energy?Is that OK? or we shold have other calculations?

Offline zh

  • QuantumATK Support
  • Supreme QuantumATK Wizard
  • *****
  • Posts: 1141
  • Reputation: 24
    • View Profile
Yes, do the thing as you said.

For the anti-ferromagnetic case, the unit cell may be larger than the one in the non-magnetic and the ferromagnetic cases.


Offline zhangguangping

  • QuantumATK Guru
  • ****
  • Posts: 187
  • Country: cn
  • Reputation: 2
    • View Profile
Dear all,

Is there a method to quick set the initial spin for a antiferromagnetic calculation if there are large number of atom in the unit cell? It is time consuming if one has to set the spin for each atom to have spin up and spin down alternately.

With best regards,

Guangping

Offline zh

  • QuantumATK Support
  • Supreme QuantumATK Wizard
  • *****
  • Posts: 1141
  • Reputation: 24
    • View Profile
You may consider the random setup of antiferromagnetic spin configuration. 

Offline zhangguangping

  • QuantumATK Guru
  • ****
  • Posts: 187
  • Country: cn
  • Reputation: 2
    • View Profile
You may consider the random setup of antiferromagnetic spin configuration.

Thanks for your reply.

With best.

Guangping

Offline zhangguangping

  • QuantumATK Guru
  • ****
  • Posts: 187
  • Country: cn
  • Reputation: 2
    • View Profile
You may consider the random setup of antiferromagnetic spin configuration.

Dear Zh,

I do not think a random setup is a good inital for antiferromagnetic or ferromagnetic. I do tests for a Ni(111) 5*5*3 super cell, and get a ferromagnetic configuration if I choose Nospin inital, while get a nonmagnatic configuration if I choose Random.

This indicates Nospin would possible give an energy lowest configuration, while one can not expect the results from a Random initial: maybe a ferromagnetic, antiferromagnetic, or nonmagnatic result. This will depend on what the "random" is. Also (I guess, because different random for differnt runs), differnet runs for the same input file using random initial would give different results. From this points, it is more useful for an antiferromagnetic calcualtion if there is an antiferromagnetic initial function. Of course, if the antiferromagnetic is not a stable configuration at all, then an antiferromagnetic, ferromagnetic or nonmagnatic would certainly do go to an antiferromagnetic results.

Please comment on my above understanding.

With best regards,

Guangping