B3P86 Functional


Just Got Here
How would I request the B3P86 functional using NWChem?

Is it:

xc vwn_3 0.19 perdew86 0.81 HFexch 0.20 slater 0.80 becke88 nonlocal 0.72

?

Forum Vet
This is what you need to use to reproduce the B3P86 energies reported in two papers reported below

XC vwn_1_rpa perdew86 nonlocal 0.81 HFexch 0.20 slater 0.80 becke88 nonlocal 0.72

http://dx.doi.org/10.1016/0009-2614(96)00480-0
http://www.springerlink.com/content/7519420454541l6w

Just Got Here
The problem I have now is that:

xc vwn_3 0.19 perdew86 0.81 HFexch 0.20 slater 0.80 becke88 nonlocal 0.72

gives me the same result as:

xc vwn_3 0.19 perdew86 nonlocal 0.81 HFexch 0.20 slater 0.80 becke88 nonlocal 0.72

This should not be the case since I did not ask NWChem to scale the local correlation in perdew86 with the first XC line.

Forum Vet
LDA already defined as vwn_3
I am not 100% sure of what you are trying to do.
You have already defined the local part as vwn_3 0.19, therefore when you ask for perdew86, only the non-local bit of perdew86 is added.

Do you want to have of 0.81 contribution from the local part of perdew86 (This is NOT consistent with the B3P86, B3LYP, etc ... definition, by the way)? If this is the case, the following line will get you on this dangerous (and wrong) path by adding 0.81 of the LDA component of perdew86 (perdew81)
xc vwn_3 0.19 perdew86 nonlocal 0.81 perdew81 0.81 HFexch 0.20 slater 0.80 becke88 nonlocal 0.7

Just Got Here
According to the Gaussian manual:
http://www.gaussian.com/g_tech/g_ur/k_dft.htm

VWN provides the (1-C) amount of local correlation for B3P86 since P86 itself when scaled by C also includes C amount of local correlation.

I don't have the ability to run on Gau$$ian to verify this. Are you confirming that:

XC vwn_1_rpa perdew86 nonlocal 0.81 HFexch 0.20 slater 0.80 becke88 nonlocal 0.72

matches the B3P86 implementation in Gau$$ian?


Forum >> NWChem's corner >> Running NWChem