Author Topic: How to do geometry optimization for OH molecule attached to Carbon Nanotube  (Read 442 times)

0 Members and 1 Guest are viewing this topic.

Offline agoldsto

  • New ATK user
  • *
  • Posts: 9
  • Country: us
  • Reputation: 0
    • View Profile
Hello, I'm trying to see the impact of attaching OH to carbon nanotubes of various chiralities. I want to perform geometry optimization for the OH molecule as the carbon nanotube has already been relaxed, however I'm not sure what the best approach is, currently I'm doing DFT-LDA with dirichlet boundary conditions in the A and B direction with periodic in the C direction, and constraining all directions in geometry optimization. Again, I performed the relaxation on the nanotube before adding the OH group. Any assistance would be appreciated!

Offline Jess Wellendorff

  • QuantumATK Staff
  • Supreme ATK Wizard
  • *****
  • Posts: 933
  • Country: dk
  • Reputation: 29
    • View Profile
Not clear what the problem is, please elaborate.

Offline agoldsto

  • New ATK user
  • *
  • Posts: 9
  • Country: us
  • Reputation: 0
    • View Profile
I'm not sure I'm doing the geometry optimization correctly, as is it's taking more than a day and the school network times out so I can't complete my research. If it simply can take over 24 hours to complete, then there maybe nothing I can do. Most geometry optimizations I've done take a maximum of 10 hours put typically less than 1. I can give the exact conditions of the optimization I'm running, but if it simply takes that long, then my hands are tied. I'm wondering if there is a faster way to do it

Offline Tue Gunst

  • Regular ATK user
  • **
  • Posts: 36
  • Country: dk
  • Reputation: 2
    • View Profile
Hi agoldsto,
First of all - optimizations can easily take more than 24 hours. It is only a matter of the size of the system and the degree of macroscopic change to the structure.
So to speed up the calculations are only two options:
1) Give a better guess for the initial structure.
If you look at the trajectory file from your first try you might see if you are doing a bad initial guess.
But generating better guesses can be tricky.
One way would be to use a force-field for pre-optimization but for complex structures the force-field might relax to different minima than DFT and the gain is low.
2) Reduce the system size/problem:
This can for instance be done by fixing/constraining part of the system during the relaxation.
For instance it could be a good first step to fix the CNT and let the molecule perform the macroscopic movement towards the CNT.

Alternatively notice that the OptimizeGeometry object has a restart_strategy option.
So if you resubmit the job it will continue the optimization from the existing trajectory file.

Best,
Tue