QuantumATK Forum
QuantumATK => General Questions and Answers => Topic started by: guyue03 on December 17, 2018, 10:00
-
Hi,
How to solve the convergence problem when doing IV calculation? SCF cannot be converged at 0.4 and 0.5V. (0.3V is okay) I have read the https://docs.quantumwise.com/technicalnotes/negf_convergence_guide/negf_convergence_guide.html#negf-convergence-guide-pulay
carefully. First, I increase the parameter Maximum steps to 250. (fail). Since I find dE and dH became oscillated. Then I just decrease the Damping factor to 0.01 and increase the number of History steps to 35. (still fail) I also try to increase the k-sampling factor a bit. (fail as well)
What am I supposed to do now?
Thank you so much!
-
Actually, you have to decrease (instead of increasing it as you did) History steps when decreasing Damping factor. In addition, you may want doing SCF calculations in smaller voltage steps for bias voltage > 0.3 V, as 0.1 V might be too large. Note that it depends on the system of study, so there is no universal bias step that would work for any system.
-
Okay, thanks! I will try this later. And I am confused about another problem. For example, at 0.3V, it is converged. But at 0.1V, it does not be converged. I wonder why the SCF can be converged at a larger bias, rather than a smaller one? I mean it may be like this trend: 0.1V(success), 02V(success), 0.3V(success), and 0.4V(fail). Now, I have this: 0.1V(fail), 02V(success), 0.3V(success), and 0.4V(fail)
-
It might not be not the most common behavior, but it can happen. For example, for 0.1 V you might have some resonance level in the bias energy window, and that creates some numerical instability that is not present for 0.3 V because this resonance is out of the bias energy range. I assume there might exist many other device electronic structure-related reasons giving rise to this convergence issue.