[Smeagol-discuss] Strange transmission

[XIAOHONG ZHENG]

Dear Ivan,

Thank you very much. Your result is very nice. I repeated the 
calculations using the input files you sent to me. The convergence is 
very good, like the following:


...
siesta:  216    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  217    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  218    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  219    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  220    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  221    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  222    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  223    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  224    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  225    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  226    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  227    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  228    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  229    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  230    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  231    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  232    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  233    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  234    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  235    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  236    -4031.8336    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  237    -4031.8337    -4031.8336    -4031.8336        
0.0001        0.0000
siesta:  238    -4031.8337    -4031.8337    -4031.8337        
0.0001        0.0000
siesta:  239    -4031.8337    -4031.8337    -4031.8337        
0.0001        0.0000
siesta:  240    -4031.8337    -4031.8337    -4031.8337        
0.0001        0.0000
siesta:  241    -4031.8337    -4031.8337    -4031.8337        
0.0001        0.0000
siesta:  242    -4031.8337    -4031.8337    -4031.8337        
0.0001        0.0000
siesta:  243    -4031.8337    -4031.8337    -4031.8337        
0.0001        0.0000

The output stops here. Then I got the following message:

[3]-  Segmentation fault      (core dumped) siesta < transport.fdf > 
transport.out

Maybe the convergence has been finished. When it starts to do next 
thing, it collapsed. But from this message, I can not tell in which part 
the problem is. Maybe there are some problem in my version or in my 
compilation. Are there anything that we should pay attention to after we 
download the source code and compile it, please? By the way, in  the 
input files you sent to me, there are some parameters that I cannot find 
in the manual like "MixHamiltonian   F" , "WriteDMT     T" 
,"Sigma.Method     1". What do they mean? Are there any other parameters 
that are not yet included in the manual, please?

Thanks a lot.

Sincerely,
Xiaohong

Ivan Rungger wrote:
> Dear Xiaonhong,
>
>   I checked your input file and there were indeed a few problems. I send
> you an updated file, for both leads and scattering region, together with
> the .TRC I obtained when running this test system. A note: this is an
> insulating 1-dimensional system, so that the screening is not very good.
> Your results will therefore depend on the length of the unit cell for
> the scattering region. This also causes some charge loss. Therefore also
> the TRC will depend on the length of the unit cell.
>
> I changed a few things, the main problem however was that you set the
> HartreeLeadsBottom in units of Ry, whereas this should be eV, because
> the Pot.exe file output for the potential is in eV.
>
> Regarding the convergence: in the current smeagol distribution version
> there is a problem when starting the pulay mixing scheme, this causes
> convergence problems. This has been fixed, and a new version will be
> distributed where this is fixed in some time. For now if you have
> convergence problems you might use linear mixing, i.e. set the number of
> Pulay steps to 1, at least for the first part of the calculation until
> the deltaDM is reasonably small, then you can switch on Pulay mixing. In
> your case the convergence problems were mainly caused by the fact that
> the HartreeLeadsBottom was wrong. In fact if there are some errors in
> the input file usually the convergence is not good.
>
> Cheers,
>
>  Ivan
>
>