Trying to run QM/MM simulation with calcium binding protein

Click For Summary
SUMMARY

This discussion focuses on running a QM/MM simulation using Orca for a calcium-binding protein that interacts with three Ca2+ ions. The user is experiencing difficulties with SCF convergence, trying methods such as DIIS and KDIIS without success. Key recommendations include plotting energy after each SCF cycle to assess convergence behavior and adjusting convergence criteria if oscillations occur. Additionally, starting with a lower level of theory, such as HF/STO-3G, can help establish a better initial guess for molecular orbitals in subsequent calculations.

PREREQUISITES
  • Understanding of QM/MM simulations
  • Familiarity with Orca software for quantum chemistry calculations
  • Knowledge of SCF convergence techniques, including DIIS and KDIIS
  • Basic principles of molecular orbital theory
NEXT STEPS
  • Learn how to adjust SCF convergence criteria in Orca
  • Explore methods for generating initial guesses for molecular orbitals
  • Investigate the use of lower-level theory calculations for initial setups
  • Study the implications of multi-configuration systems on SCF convergence
USEFUL FOR

Graduate students in biophysics, computational chemists, and researchers working on QM/MM simulations, particularly those dealing with complex systems involving metal ions.

nisler.1
Messages
2
Reaction score
0
Hello everyone, I'm a graduate student in a biophysics lab. I'm trying to run a QM/MM simulation with a protein that binds 3 Ca2+ ions, and as such I am having a devil of a time getting the SCF to converge. I've tried DIIS, KDIIS, damping, and a few other methods to try to get convergence with not much luck. The couple of times it has converged, it was after more than 100 iterations, which is just too expensive. Does anyone have any suggestions of things I could try? It doesn't matter much, but I'm using Orca for the QM portion. Thanks a bunch.
 
Chemistry news on Phys.org
Plot your energy after each SCF cycle. Are they oscillating or are they going down continually? If they're going down continually, then you probably just need more SCF cycles to reach convergence (I'm assuming the program terminates after a preset number of SCF cycles--which you should be able to adjust). If they're oscillating, you might need to loosen your convergence criteria, at least for preliminary runs. Sometimes if you start with a crappy initial guess for your molecular orbitals, it just takes a while and there's nothing you can do about it. But if you can get it to converge once, you can usually use these new MOs for a better initial guess. I'm sure there are ways to do this in Orca; it's a pretty robust program.

Of course, all this assumes that your system is reasonably well-behaved. If you have a system which is inherently multi-configuration and you're running an HF-type calculation, your SCF runs will fail to converge because your model isn't a reasonable approximation of the system.
 
TeethWhitener said:
But if you can get it to converge once, you can usually use these new MOs for a better initial guess.
I just realized this made no sense when I read it to myself. What I meant was: 1) get the system to converge with an obscenely low level of theory/small basis set like HF/STO-3G or 3-21G. Then 2) use the MOs from this calculation as an initial guess for a production run with a bigger basis set and/or a higher level of theory.