Dear fiends, I am doing rob3lyp single point energy calculations in combination with solvation models. However, the tasks failed to converge though the following set was used, could anyone give me some advice about this question? My systems contain f-block metals, which increases the convergence difficulty.
rob3lyp/genecp scf=maxcycle=520,vshift,novaracc,conver=6,noincfock)
Make first a stability computation, save the orbitals.Use them as starting ones by setting the option, guess=read,mix in the input stream.
What is the error message that you get in the output file (eg. The .log file)? And what program are you using to run the calculation? I ask because sometimes the problem with challenging organometallic calculations is not an inherent flaw in the functional/basis set but rather in insufficient requested memory, wall time, or number of computer nodes. Depending on the configuration of the computer cluster you are using you may need to explicitly request your desired memory and wall time (eg -M 60GB -W 72:00) even though you have specified them in your .com or .PBS file.
I guess you are using Gaussian for your calculation.
I strongly advice you to use a tighter convergence threshold than 10^-6. The default (10^-8) is a much safer option if you want to trust your results...
As for the convergence problem, Fermi broadening might help (not sure whether it is compatible with vshift, though). If you can't get it to converge anyways, you can try with the additional keyword Use=l506. It will take a while, but is might solve your problem...
Good luck!
Try quadratic convergence or Fermi broadening, or use a slightly modified molecular structure. But never use iop(5/13)=1, using this iop indicates that the software can keep blind for the convergence failure.
- Badges
- Science topic
- Similar topics
- Biological Science
- Biochemistry
- Enzymology
- Enzymes