I run geometry optimization and Hessian calculations for free radicals using Gaussian v. 16.
However, these calculations are way too exhaustive: they took about 30-40 hours (Bip_cat-rad.log in the attachment) while the same calculations for the normal ground state molecules took about 2 hours (Bip_S0.log in the attachment).
So my questions are following:
1) What does Gaussian do during these additional calcluations for the free radical?
2) How to turn off these unnecessary calculations?
Regarding Bip_S0:
1. Are the charge and multiplicity correctly mentioned? The job is properly done, I guess. Please give some RAM by using %mem=16GB or whatever suitable to you.
Regarding Bip_cat-rad:
2. Same points here, use %mem=something GB
rather than using ROHF, you can use UHF. It has faster convergence behavior.
while using minnessota functionals, always use int=grid=ultrafine/superfine.
these functionals are highly sensitive on the DFT integration grids.
Cheers
Open shell methods require more calculations as a matter of principle. There isn't much you can do about it.
Whereas restricted Hartree-Fock methods get by with only calculating doubly occupied orbitals (i.e. 1 orbital for every pair of electrons), unrestricted HF calculates the orbital separately for every electron. This already massively increases the calculation time.
In addition, UHF has to take spin-spin interactions into account, which normally cancel out in closed shell systems. These require a bunch of additional time-consuming calculations.
ROHF is simply speaking a mixture of RHF and UHF, using RHF's doubly occupied orbitals for paired electrons and distinct orbitals for the unpaired ones, but it has some drawbacks in comparison to UHF, which usually make the latter more preferable.
Simon Janich To be precise, UHF or UDFT also have drawbacks. E.g. one should carefully examine spin contamination, which is present for the unrestricted methods. I would also rather avoid describiing ROHF as a mixture of RHF and UHF
Dear Andrey А. Buglak ,
I could recommend to use the keyword freq=noraman . This option could save you even 20% of the time calculation in the freq subroutine.
On the other hand, the calculation hasn't been done in a cluster, so larger calculation times are normal. Although, to improve the time, you should specify the memory (i.e %mem=1GB). If you use a PC, take into account the memory available and if you will use it while job is running. For a 16Gb RAM PC, I recommend to add at least 8Gb. (4Gb for windows (2Gb for Linux, if I remember correctly), 8Gb for the G16 and 2Gb free for PC uses).
Hope it helps you,
Best regards,
Joaquim
Tatiana Korona Yes, absolutely. Precise calculations of radicals are always tricky, especially with such methods.
My description of of ROHF as "mixture of RHF and UHF" was deliberately simplified. There's certainly more to it.
- Badges
- Science topic
- Similar topics
- Clinical Pharmacology
- Dosing