I have clusters of 38 FeCoNi atoms that were previously optimized using the Gupta semiempirical potential. I would like to perform a reoptimization, energy calculations, and frequency calculations using DFT with Gaussian 09. What potentials and basis sets do you recommend, as I need to obtain electronic and magnetic properties of these clusters? If there are any other parameters to consider, I would appreciate your guidance. Thank you.
Dear Emanuel, you may try long-range, dispersion corrected wB97XD functional along with a double zeta basis set for equilibrium geometries and thermochemistry:
#P wB97XD/genecp opt freq -other keywords you may need-
your project title
charge spin -remember to veirfy this by screening different configurations-
coordinates
Fe Co Ni 0
LANL2DZ
****
Fe Co Ni 0
LANL2DZ ECP
Thermochemistry is usually accurate with dobule zeta basis set. However, If you have enough computational resources, you may try first optimizing geometries with triple zeta basis set LANL2TZ or all-electron Def2TZVP (this latter one will be very expensive computationally). Also consider adding diffuse fucntions and PCM solvation (if needed). Then a freq job would follow with a double zeta basis set.
Tip: I usually start demonstrating that geometries optimized with double zeta are equivalent to those with triple zeta, then opt/freq jobs at double zeta are as valid as those with triple zeta.
For electronic energies and electronic/ molecular properties, I do suggest extending the basis set at minimal triple zeta including diffuse functions, like LANL2TZ+ (also consider PCM solvent if the case):
#P wB97XD/genecp sp -other keywords you may need-
your project title
charge spin -remember to veirfy this by screening different configurations-
coordinates
YOU NEED TO COPY/PASTE here the basis set and potentials
as given in https://www.basissetexchange.org/
for the three atoms Fe, Co and Ni. Gaussian does not have it.
So, overall the method I suggest is
wB97XD/LANL2TZ+ // wB97XD/LANL2DZ
You may compare this with all-electron basis set:
wB97XD/Def2TZVP // wB97XD/Def2SVP
Let me know if this was helpful.
Best, Pablo
Dear Pablo Mtz ,
Thank you for your comprehensive response. I appreciate your detailed recommendations. I will conduct tests as per your suggestions and keep you updated on the progress of my simulations. Your guidance is invaluable, and I'm grateful for your assistance.
Best regards, Emanuel
I have started with runs of 13 atoms; however, I am encountering this error:
FoFJK: IHMeth= 1 ICntrl= 0 DoSepK=T KAlg= 1 I1Cent= 0 FoldK=F
Rare condition: small coef for last iteration: 0.000D+00
Matrix for removal 19 Erem= -1873.68905131766 Crem= 0.000D+00
>>>>>>>>>> Convergence criterion not met.
SCF Done: E(RwB97XD) = -1873.68853228 A.U. after 129 cycles
Convg = 0.1089D-03 -V/T = 2.6614
Convergence failure -- run terminated.
Error termination via Lnk1e in /usr/local/g09/l502.exe at Mon Jan 29 14:35:55 2024.
My input file is this:
%CHK=./Fe5Co4Ni4.chk
# P wB97XD/LanL2DZ Opt=(maxcycles=512) Freq
Fe5Co4Ni4
0 1
...
Any idea to solve this problem?
OK, SCF did not converge. Try adding this keyword:
#P scf=(direct,maxcycle=200,XQC)
XQC means Gaussian will try to converge SCF within 200 cycles, if not automatically switches to QC quadratically convergent SCF procedure which usually does not fail.
Also, try another test with unrestricted UwB97XD, it might be that the system is open shell, you can check this when the last SCF is done, = 0.0 for closed singlet or = 1.0 for open singlet (sometimes this is lower in energy). Also do not forget triplet.
Best, Pablo
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs