QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: alokrs on February 4, 2021, 10:30

Title: scf convergence for finite bias device calculation
Post by: alokrs on February 4, 2021, 10:30

my zero bias calculation for both parallel and anti-parallel configuration, as well as finite bias 0.05v calculation for parallel configuration, converges well, but when I do 0.05V calculation for anti-parallel geometry, the calculation doesn't converge at all. I have used the zero bias calculation as the initial state for the finite bias 0.05V calculation and damping factor 0.05, history steps 12. I have attached my script file for your reference.

Title: Re: scf convergence for finite bias device calculation
Post by: Tue Gunst on February 15, 2021, 20:35

Hi,
Converging polarized calculations can be tricky.
What we usually observe is that it is system dependent but as a general guideline try one or more of the following:
1) Increase the k-point samplings.
2) Decrease the scf threshold of the electrode calculations.
3) Increase the electronic temperature smearing in the calculator.

In general, finite bias calculation will take longer and can suddenly converge while it spend a some time with little change in the scf loop.
If you didn't already do so then increase the number of scf steps allowed to a few thousand to be sure it will converge.
Best,
Tue