Thank you Nick!
On Thursday, October 31, 2013 11:47 AM, Nick Papior Andersen <[email protected]> wrote: Yes, strictly speaking you don't need the same amount of points for different biases, you should just ensure that the result has converged with respect to the number of points, i.e. increasing the number of points on the contour should not change the results. For some systems one might experience that it is very hard to converge. And then you can either increase bias points or increase the imaginary value. There is no way for me to tell what is the "correct" settings, it is system dependent and one have to "fiddle" with the numbers. I will say that your bias point resolution is very fine (dE = 0.0053 eV), so I think that you are left with tweaking the imaginary part if you want it to converge. Kind regards Nick 2013/10/31 Alen H <[email protected]> That means it is unnecessary to use the same parameters for each bias? if so this could not affect the results? >I'm using 300 points for the non-equilibrium contour which I think overkill >but more accuracy. And if I increase the imaginary part value, then I'll get >less accuracy. > > >Thank you Nick! >Yours. > > > >On Thursday, October 31, 2013 12:26 AM, Nick Papior Andersen ><[email protected]> wrote: > >The system is harder to converge at positive bias. You can not expect the >converge to behave symmetric about 0V. > > > >You should check how the convergence is, try to increase the number of points >on the non-equilibrium contour to increase accuracy. >You could also start by adding a larger imaginary part and then later decrease >it while increasing the number of points on the non-equilibrium contour. > > >Kind regards Nick > > > >2013/10/30 Alen H <[email protected]> > >Hi all, >>I have a sequence of positive and negative bias applied on a system, for the >>negative bias( -1.6V) it took 25 steps for transiesta to converge but for >>the positive bias (1.6 V) it exceeds 100 steps without converging. >> >> >>Any one knows the reason? >> >> >>Regards, >>Alen > > >
