Dear all,
Is there a way to use wB97X functionals with D3 or NL dispersion in Gaussian 16? wB97XD uses GD2 dispersion, and I would like to change the dispersion to D3, D3(BJ) or NL (with appropriate IOps I guess?). wB97M-V (or D3BJ) would also work, but as I understand, these functionals are not available in Gaussian 16.
Using other software is not really an option. Orca might be, but I cannot run it on the cluster, and it takes a lot of time to run a calculation on my personal computer. I was also considering double hybrid functionals (such as revDSD-PBE86-D3BJ and B2GP-PLYP with appropriate IOps), but the computational expenses are quite high, even for single point calculations.
Thanks in advance!
Hello Uros,
First of all, I suppose you have good reasons to use wB97XD(V) functional. The calculation options offered by G16 with respect to wB97X are in the papers shown in https://gaussian.com/dft/, although I suppose you already knew this. On the other hand, the selection of a DFT and the corresponding basis set is always a not intuitive task, and difficult to evaluate previously if you don't have a concrete and contrasted reference. It depends on the compound and what you really want to obtain. I can tell you that I have made a benchmaking between 14 DFTs for molecules with three carbons (covalent bonds) that we will publish soon, and the DFT that best describes the geometry (not the energy) with respect to the reference (CCSD(T)-F12/cc/cc-pVTZ-F12b) is not WB97XD(V), but TPSSh, with a cpu consumption time around 30% lower in optimization and calculation of harmonic frequencies procesess. Regarding the choice of the basis sets, you can find the one that best suits your compound at https://www.basissetexchange.org/. Finally Orca is definitely not an option unless you want to perform calculations at higher levels of theory as described above, but maybe Molpro (https://www.molpro.net/) could help you.
I hope this helps a bit.
Greetings, Juan
Even if you did manage to implement D3 or D3(BJ) parameters with the wB97X functional in Gaussian, there are no (to my knowledge) published parameters (i.e., s6, s8, etc) in combination with this functional (as implemented in Gaussian).
In 2013, Chai did obtain Grimme dispersion parameters for the wB97XD3 functional - however, the coefficients in the functional's form (e.g., such as the range separation parameter) were also reparameterized from scratch against benchmark data - whereas Gaussian's implementation of wB97X uses the original 2008 coefficients.
In short, wB97XD3 differs from wB97XD by more than just the type of dispersion correction, and it would be risky to create your own ad hoc version without extensive benchmarking.
Juan Carlos Ramal Olmedo Thank you, Juan, for your help and valuable insight. I am a bit reluctant about TPSSh functional, considering that my system features transition metals, so this functional might not be the best option. I guess the structures would be reasonable, but I think the barriers could suffer severely from SIE, or even BSIE and BSSE at TZ quality of basis set. The idea behind using wB97X-V and wB97M-V is a recent MOBH35 benchmark, where it was shown that these two are the best among hybrid-GGA or hybrid-meta-GGA functionals.
Robert Paton Thank you, Robert, for this insight. Based on your comment, I would say that it would be practically impossible to perform meaningful calculations with wB97X-D3 in Gaussian at this point. This is valuable information for me, thank you. Let's hope some of the future versions of Gaussian will have an extended selection of functionals
- Badges
- Science topic