Re: [SIESTA-L] SIESTA-L Strange dependence of transiesta convergence on number of processors

[Guangping Zhang]

Dear Nuria,

I think it is not strange since you use version rc2, and I have find a similar 
problem before and post it to this forum which named "About the transiesta 
results for 3.0-b and rc2,not only a factor of 2".

I think there is some proble in the siesta part for rc2 version. At the same 
time, I think different initial guess for the NEGF may result in different 
convergence points.So the initial guess is also a very improtant factor.

You can have a try with siesta-3.1

Best

2012-02-03



Guangping Zhang



发件人：ngarcia
发送时间：2012-02-03 20:26
主题：[SIESTA-L] SIESTA-L Strange dependence of transiesta convergence on number 
of processors
收件人：siesta-l
抄送：

Dear Siesta users and developers, 

recently I found an strange behavior of the convergence of my system. The 
transiesta convergence part is very different depending on the number of 
processors I used in the calculation. 

Obviously the final transmission function is very different, too. But the 
part corresponding to the electrodes, see *.LDOS is the same in both 
calculations. 

I've attached the outputs of calculations with 30processors and 
56processors. (I've tried other number of processors, but they didn't 
converge!). 

Any suggestion will be kindly appreciated. 

Thank you in advance. 

Nuria 

Nuria Garcia Castelló 
PhD student 
University of Barcelona