I have the following error running doublet molecule using M062X\6-31G(d):
Annihilation of the first spin contaminant:
S**2 before annihilation 0.7554, after 0.7500
Convergence failure -- run terminated
What causes this error? and how could it be solved?
As Mark said spin contamination of that size shouldn't be a problem.
SCF=(maxcycles=128) is always a good idea, you can also try more than that but it rarely helps.
SCF=QC or SCF=XQC helps in many cases, as it used a completely different algorithm for SCF optimization, definitely worth a try.
SCF=NoVarAcc also helps in limited number of cases
If nothing above works, try adjusting slightly the geometry. Switching to a lower basis set is kind of tricky, I wouldn't try it since you may end up with a really bad geometry when using something like STO-3G or 3-21G*. But switching to different DFT functional, optimizing and then reverting back to M062X might also be a good option.
Ahmad,
The spin contamination becomes a concern when we are dealing with unrestricted calculations and the spin changes during the course of optimization. I do not really know what is the nature of your system but it is a common place error when transition metals are involved in the geometry. Based on my experience, the following options could solve your problem:
(1) Request quadratic convergence in Gaussian for your self consistent function using SCF=QC in your route line.
(2) Look at the structure at the step which gives you this error and try to see if there are any regions or very closely interacting atoms that belong to two different molecules. If yes, then you might need to tweak a bit. If no, then the first option would work.
(3) If in the case the first option does not work, I would recommend using a primitive basis set such as STO-3G and your current level of theory to obtain an initial estimate for your thorough basis set.
Hope this helps!
Regards,
Atiya
Hello Ahmed,
I'm not sure if this is all you're seeing in your output file, but if so my guess is that you are requesting standard or terse output from the calculation. If you use #P you'll get a much better idea of what Gaussian is doing at each stage of the calculation, though your log files will be much bigger (that said, most visualisation programs like extended output).
The spin contamination in itself is not a problem - the rule of thumb is that you can get up to 10% error from the expectation value (0.75 for your doublet) without the contaminant being considered too large to affect your structure.
In any case, it looks to me that as though you're just failing to converge your wavefunction in the standard number of SCF cycles - this is not related to the spin contamination; I think it's just unfortunate that the output you note above does seem to imply this.
In addition to Atiya's advice above, you can just try SCF=(maxcycles=128), where the default here is 64. This just increases the default number of SCF cycles before the run is considered to have failed.
My concern here is that, generally speaking, the default SCF maxcycles should really be enough (unless you have some peculiar, or perhaps very flexible structure or wavefunction - are you doing multireference calculations, for instance?) to tackle almost any problem. Either that, or you are a long way from any mimimum on your potential surface, which might also cause such a failure.
I think Atiya's advice to gain a well-converged structure at a lower level of theory, and ensuring it is in the correct symmetry, would be a good start.
Would you mind posting your input file? This might be helpful for further diagnosing your problem.
Hope this is of help!
Best Regards,
Mark.
As Mark said spin contamination of that size shouldn't be a problem.
SCF=(maxcycles=128) is always a good idea, you can also try more than that but it rarely helps.
SCF=QC or SCF=XQC helps in many cases, as it used a completely different algorithm for SCF optimization, definitely worth a try.
SCF=NoVarAcc also helps in limited number of cases
If nothing above works, try adjusting slightly the geometry. Switching to a lower basis set is kind of tricky, I wouldn't try it since you may end up with a really bad geometry when using something like STO-3G or 3-21G*. But switching to different DFT functional, optimizing and then reverting back to M062X might also be a good option.
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running