For typical organic systems, optimization+frequency using 6-31G* is sufficient. 6-311G++ will give you worse result, since the important polarization functions of heavy atoms are missing. Diffuse function will not noticably improve result of optimization+frequency, but significantly increase computational cost for large system.

Dear Renjith,

it would depend on how many processors are used for running the calculation and on the memory amount available, too. By the way, how many basis functions does this structure have? With this number of atoms and the basis set (6-311G**, right?) it should have pretty many of those, and frequency calculations could be quite long, too, especially with lots of basis functions. Also, does your structure have any symmetry? If not, then it can also make frequency calculations pretty long. The calculations are performed in the gas phase, right?

Best regards,

Aleksey.

An incredibly much faster alternative to compute vibrational spectra of large molecules is offered by the DFTB method, as implemented in GAMESS-US and described in the paper linked below and related publications.

http://dx.doi.org/10.1016/j.cplett.2016.11.014

Hi Renjith,

If you have already optimized the structure, you can do a single point calculation for the optimized coordinates using FREQ keyword. Since the system is not that big I guess Gaussian 9 can handle SP calculations relatively fast.  

Thanks!

For typical organic systems, optimization+frequency using 6-31G* is sufficient. 6-311G++ will give you worse result, since the important polarization functions of heavy atoms are missing. Diffuse function will not noticably improve result of optimization+frequency, but significantly increase computational cost for large system.

It's not really a problem, but a "feature" - frequency calculations in large basis sets can take up to a few times longer (2-3 on 4-core machines) than geometr optimizations. If you really want to do that in this specific basis set, there's nothing you can do, unless you get a faster computer (more cores won't really help, byt faster cpu clock will a lot).

please use geom=check, guess=read for the coordinates and int=grid=ultrafine, int=acc2e=11 keywords for calculating freq. I think you will be able to do it.

Thanks.

instead of obtaining IR with traditional way (optimization of molecular structure followed by diagonalization of mass weighted hessian) you can run ab inito MD, extract dipole moments from trajectory, then calculate dipole-dipole autocorrelation function and Fourrier transform it. Results will give you IR spectrum including anharmonic effects. It scales better than Hessian type calcs.

I will give typical reasons for such a problem

1- The version of Gaussian that you have, if you have 32bit version you will be able to use 4 core only and 1.8GB or RAM, so you need to use 64bit version to get more cores.

2- assuming that you have 64 bit Gaussian, than the problem is with your computer, if you have lets say Core i5 processor than you already limited to 4 core only, you need powerful processor such as Xeon with 6 core or higher, or you can use processor with hybrid technology so you get 4 core but 8 threat.

3- always optimized you structure at PM6 semi-empirical first, than use higher level of theory.

4-B3LYP 6-31G(d) is more than enough for optimization you really don't need the + unless you have cation or anion.

5- you can use SP calculation at higher level after you done you optimization.

I assumed all the above on the fact that you using PC not a cluster or server.

Best