bonds broken while optimization

[Wanzhi Qiu]

Hi,

I am optimizing the geometry of a small protein molecule (with ~300 atoms) in vacuum. The self-consistent Slater Koster method is used. The optimization converses successfully achieving the requested max forces 0.05 eV/Ang. However, in the optimized geometry, one group of three atoms (O-C-O, they remain connected by themselves) disassociates with the rest of the molecule. This group was originally connected to the rest via a C-C bond. That is, this C-C bond was broken during the optimization process causing the group of three disassociating with the others.

Any suggestions and solutions on the causes of this problem. Many thanks.

 

[Anders Blom]

Which Slater-Koster method did you use? Normally these cannot be used for optimization since they do not contain a repulsive pair potential.

[Wanzhi Qiu]

Hi Anders, I am using the DFTB (CP2K, selfconsistent) basis set. Thanks..

[Anders Blom]

I have no direct experience with these basis sets but someone else noticed some similar issues with the CP2K sets so for now I can only recommend using other sets for the relaxation at least.

[Wanzhi Qiu]

Hi Anders, The other basis set available in ATK is CP2K non-selfconsistent; seems the DFT has to be used, which would take too long to converge for the ~300 atoms system..Thanks.. 

[Anders Blom]

We support the parameter sets from http://dftb.org, but we are not allowed to distribute them. So check if they have a set that contains all the elements of your system, then you can ask permission from them to obtain the set, and finally you can plug that into ATK (see http://quantumwise.com/documents/tutorials/latest/DFTB/).

[Wanzhi Qiu]

Thanks for advices. I'll have a go..