I strongly suggest you consider to use the B97-3c proposed by Grimme in J. Chem. Phys., 148, 064104 (2018). Using B97-3c in ORCA is quite easy. The cost of B97-3c is fairly low, it is comparable with pure functional with def2-SVP and RI-J treatment, while the accuracy should be evidently better. According to the original paper, B97-3c works well for systems containing transition metals.

Well if you're dead set on B3LYP, just use the smallest basis set you can get away with (i.e. the double zeta one) and turn on all the density fitting approximations that work with hybrids. For example, ORCA supports RI-JK and RIJCOSX. I recommend the latter. These should speed up the calculations quite a bit.

I'd also reconsider if an expensive hybrid is really nescessary. Most people don't run these sorts of calculations on personal computers with consumer grade hardware.

Dear Melissa,

In addition to Jonathan David Colburn , transition metal complexes are usually better computed with pure functionals e.g. BP86 or BLYP. Take a look into these works:

Article Activity of phenoxy-imine titanium catalysts in ethylene pol...

Article Assessment of ten density functionals through the use of loc...

I guess RI formalism in ORCA is employed in hybrid DFT functionals by default (and even maybe in the pure ones) - check an ORCA manual:

https://orcaforum.kofo.mpg.de/app.php/dlext/?cat=1

As well, try to use recommended number of CPU cores on a cluster HPC for !opt and !freq tasks. This is also noted in the manual.

To my experience, Optimization taks took around 1-2 h for similar systems while the Hessian calculations were about 5-7 h at the BP86/SVP level on 4-8 cores CPU. So for larger bases (like triple-zeta TZVP or 6-311*G(d,p)) one could expect days of the calculation time. But consider that the BP86/SVP level might be enough for your tasks....

Regards,

Sergey

I've tried your advice on working with the pure functionals i.e. BP86 and BLYP.

1.) I previously used numfreq on RI BP86/SVP on a complex with a copper atom and it worked quite well. (this is the one that took 16 hours) However, on systems with just Sn and Fe replaced instead of copper, this kept on showing regardless of RI BP86 or BLYP:

ERROR

This wavefunction IS NOT CONVERGED!

You can't use it for a Numerical Calculation (NumFreq or NumGrad)!

Aborting...

Please restart calculation (with larger maxiter?)

What do I do about this?

2.) Also on some systems that I've tried computing for numfreq, this keeps on showing almost on every displaced atom:

An error has occured in the SCF module

CALLING COMMAND: orca_scf ligand5.proc0.gbw b>ligand5.proc0.lastscf

RETURN CODE : 125

Check for SCF-logfiles

ORCA_NUMFREQ: ORCA finished with an error in the energy calculation

and at the end:

WARNING: It seems that the Hessian Matrix ISN'T COMPLETE!

Please check your results VERY carefully!

For possible error information check files ligand5.last*.dpl*

If necessary - restart the job with local HessianFiles.

Modify your input options to ensure convergence and/or

avoid other sources of problems

What does this mean? I'm sorry and thank you in advance

3.) @Jonathan, I'm not actually deadset on B3LYP. It's just that its is described as one of the most popular out there. And me having little experience on ORCA, have thought that it might be better to get started on something. Do you recommend any method and basis set for the sole purpose of getting numerical frequencies (IR) that takes the least time for a medium sized compound that contains a transition metal (Cu, Ni, Cd, & Fe) and other metals (Sn& Zn)?

Dear Melissa Suelto ,

Did you manage to locate the optimized geometries? Or the error occurs just at the first optimization step while SCF iterations?

The one I mentioned in #1 is after 125 cycles of scf convergence (says ERROR SCF NOT CONVERGED AFTER 125 CYCLES). This puts the calculation to a halt. Meanwhile, the one mentioned in #2 didn't stop the calculation despite the errors. In fact, I have recently finished one at the moment!

It is because the max number of scf cycles is 125 by default.

So you have a classical scf convergence problem I consider.

Check the structure you are trying to calculate thoroughly. Then make sure that you set the appropriate charge and multiplicity values. If this doesn't help, then

Try to increase this default number of scf cycles.

You can analyze the scf energy vs step number plot as well. If you find plateaus, the try to restart the scf convergence from the cycle number that is in the middle of the last plateau (check for latest ORCA versions if this is possible. I know that in Gaussian you can do this).

Good luck!

I strongly suggest you consider to use the B97-3c proposed by Grimme in J. Chem. Phys., 148, 064104 (2018). Using B97-3c in ORCA is quite easy. The cost of B97-3c is fairly low, it is comparable with pure functional with def2-SVP and RI-J treatment, while the accuracy should be evidently better. According to the original paper, B97-3c works well for systems containing transition metals.

use

! opt RIjCOSX B3LYP def2-TZVPP def2/J

and more cpu

!PALX X=number of CPUs

Hello! Thank you for your help! I have successfully computed the aforementioned complexes. However, do you know why there are additional bonds in the metal atom of my ORCA .out file?

Here's my ORCA input to give you context of what I'm talking about:

! NumFreq RI BP86 slowconv

%scf MaxIter 500 end

! PrintBasis SVP

%output

print[p_mos] 1

end #output

* xyz 0 2

I have two complexes that were recently successfully computed. Both are almost identical except that the two differs in the central metal atom. Both have the central metal atom bonded to four N atoms (two of the N's are from two nitro group). Upon checking the geometry (I'm using gabedit to visualize my work), it still looks like the same molecule that I've input. However, upon checking the animations to compare with the generated IR spectra, many other bonds appeared out of nowhere. The metal ion has now bonds to the other O atoms in the nitro groups and nearby C atoms. Here's a picture to help you visualize what happened. https://imgbbb.com/image/Yc9Zy

How do you address problems like this? Did this affect the IR spectra I've generated? why there are additional bonds in the metal atom of my ORCA .out file?

Try using KDIIS in the first line and also give def2/J as auxiliary basis set.