Structural optimization with ATK-SE

[chp]

Dear QuantumWise Staff,

I am trying to use the ATK-SE package, but there are two questions which are not clear to me. Would you please give me some advice?

1.   I am going to perform structural optimization within the framework of ATK-SE: Slater-Koster, however, there is a choice of “No SCF iteration” in the settings of Slter-Koster Basis set. What’s the difference of optimization with and without “No SCF iteration”?

2.   In general, the ATK-SE method can run for a larger system than DFT method. Taking an example, if I run a two-dimensional system consisting of Boron Nitride and graphene in parallel (for example, with 8 CPUs), what’s the upper limit of the numbers of atoms in the system?

Thanks for the help.

[kstokbro]

1. Generally you should use SCF for the DFTB models. Our experience is that only the SCF models give good results for relaxations.

2. It is hard to say a general rule for how large systems you run, it depends on your computer and how long time you are willing to wait for the results. However, with the DFTB models you should be able to study systems in the 1000-10000 atoms range. se also
http://www.quantumwise.com/documents/tutorials/latest/MemoryUsage/index.html/

[chp]

Thanks a lot.

I also notice that there is only one choice in the list of built-in Slter-Koster Basis sets (i.e. “DFTB (CP2K, non-selfconsistent)”) to use when I perform DFTB calculations for BN system. According to your reply 1, can this set give good results for relaxations or properties if I don’t choose the “No SCF iteration” below?

[Anders Blom]

Probably doubtful. For this system I would recommend getting Slater-Koster files from http://dftb.org and use those instead (the "matsci" package is what you want). They are free for academic users. See http://quantumwise.com/documents/tutorials/latest/DFTB/ for instructions on how to install them in ATK.