QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: jchang on September 25, 2019, 23:59

Title: SCF converge issue in Optimization
Post by: jchang on September 25, 2019, 23:59

Hi

I am trying to do geometry optimization of a certain 3 layers of 2D materials.
I have issue in the scf convergence.
To resolve the scf convergence issue, I tried to increase k point and energy cut off. But both didn't work for my case.
I attached my input file of 9x9x1 k point and output log file.
In the first several scf loop converge, but it fails.
I have tested increased k point, but the result is same.

Please give me any advices to solve scf issue in optimization.

Thanks.
jiwon

Title: Re: SCF converge issue in Optimization
Post by: Petr Khomyakov on September 26, 2019, 10:03

- I would suggest first checking convergence of the total energy with respect to energy cutoff, k-point sampling in a systematic manner for a static configuration, i.e., without doing relaxation; you should aim at convergence of the total energy within 1-10 meV per atom.   

- You may also consider using a different pseudopotential, e.g., more modern one such as PseudoDojo.

- Another thing, you could try to enable Preconditioner in Iteration control in the Calculator. That might help improving SCF convergence for slab/surface/2D material PW calculations.