QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: M.Albert on April 10, 2009, 10:47

Title: Can we change some parameter setting to make convergence faster
Post by: M.Albert on April 10, 2009, 10:47

Hi
How to change some settings which  do not  impact the final result much and make the convergence faster?
I want to calculate transmission spectrum at 0v,not a high bias. The scattering region is a nanotube ,and electrode is Li.
enandter
Thank you!

Title: Re: Can we change some parameter setting to make convergence faster
Post by: Nordland on April 10, 2009, 13:21

First of all, it is a good idea to calculate the transmission spectrum and the selfconsistent part as two seperate calculation, as you will avoid to perform the selfconsistent calculation again, if you want to adjust the transmission spectrum parameters.

When that is being set, based on your input, i would go for a smaller basis set, and perhaps increase the diagonal mixing parameter to a slightly higher value - like 0.2.

However the missing information is - how much time is spent in the calculation of the transmission spectrum versus the selfconsistent part?

Title: Re: Can we change some parameter setting to make convergence faster
Post by: M.Albert on April 12, 2009, 12:50

Hi!I will try to change diagonal mixing parameter as you said.
Time spent in calculation of the transmission spectrum is very short compared with time spent in selfconsistent.
What I want to know is whether can we make convergence faster by change some  parameter  settings when  we perform selfconsistent.

Title: Re: Can we change some parameter setting to make convergence faster
Post by: Nordland on April 12, 2009, 14:19

Okay, it is part of a broader study of an ensemble of systems that are similar in nature or it is a single system?