Convergence problems with a large system

[viscio]

Hi, i'm Michele from Italy and I'm using ATK for my thesis work.

First of all I want to thank everyone in this forum, because it is really helpful: I've digged in almost every post and I found really interesting and helpful topics that have helped me a lot in my work! 

I'm having some problems with a rather complex system: I'm trying to do a geometric optimization, but it seems that it won't take less than a week to complete the calculations (if they converge at all), and I think that there's something wrong, maybe I'm not tuning the correct parameters in order to get a fast convergence. I've tried almost everything (I've read your tutorial about convergence tricks, which is fantastic and saved me many times!), but it seems that there aren't any changes in the convergence speed. I'm currently using ATK version 2008.10. I'm also using an 8-processor PC, with 2GB of RAM.

I've attached the script that I've prepared using VNL, maybe if someone has some time, he can help me... 

Thank you in advance
Michele

[Anders Blom]

Hi Michele,

thanks for your nice words about the Forum - it's nice to know that it serves its intended purpose

About your calculation, I didn't understand whether it's the SCF calculation itself that fails to converge, or the relaxation?

[viscio]

Hi Anders!

The problem is that the relaxation doesn't converge. The scf converges, in about 10-15 steps, but the entire process is really slow... I think  it takes about 2-3 hour for every step of the scf.

I don't know if ti is a problem at all... in fact, we are talking about a rather complex system, but I thought to ask you anyway, because i'm not so used to the software and maybe I don't have much experience... Maybe it's just a problem of some parameters that I'm not tuning in the proper way...

Thank you again! 
Michele

[Anders Blom]

One of the problems might be that the surface atoms are not constrained. Thus, all your atoms in the central region are free to move. I suggest you try keeping all the gold atoms fixed and just optimize the molecule between the electrodes. Also, you may want to pre-optimize the molecule in vacuum.

You could try the steepest descent approach instead, I honestly prefer it myself.

[viscio]

Thank you very much for your fast replies! 

Unfortunately, I cannot set the atoms in the central region fixed, because my study is also related on the changes of the surface caused by the bonding with the tiophene, but I'll try the Steepest Descent approach! If I'll encounter other problems, I'll write here again!

Thank you very much, again!

Michele

[Anders Blom]

Yes, the outermost layers of the surface can be left free, but those closest to the electrode are  best left fixed, for many reasons.

[viscio]

ok, I'll try to fix these atoms and use the steepest descent approach.