Are these definitions correct for the functionals I am trying to define?


Clicked A Few Times
Hi,

Can somebody let me know if the following definitions of the functionals are correct? Just want to make sure I am defining them correctly. I would really appreciate your help.

mPW1PW91: xc mpw91 0.75 HFexch 0.25 perdew91
mPW1PBE: xc mpw91 0.75 HFexch 0.25 cpbe96
B1LYP: xc becke88 0.75 HFexch 0.25 lyp
CAM-B3LYP: xc xcamb88 1.00 lyp 0.81 vwn_5 0.19 hfexch 1.00
                  cam 0.33 cam_alpha 0.19 cam_beta 0.46 

mPW1LYP: xc mpw91 0.75 HFexch 0.25 lyp
B3PW91: xc acm
X3LYP: xc vwn_1_rpa 0.129 lyp 0.871 hfexch 0.218 slater 0.782 \
          becke88 nonlocal 0.542  xperdew91 nonlocal 0.167

mPW3PBE: xc cpbe96 0.81 HFexch 0.2 slater 0.8 mpw91 nonlocal 0.72
B3P86: xc vwn_1_rpa 1.00 perdew86 0.81 HFexch 0.20 slater 0.80 becke88 nonlocal 0.72
TPSSh: xc xctpssh
VSXC: xc vs98
BP86: xc becke88 perdew86
BLYP: xc becke88 lyp

Thanks.

Clicked A Few Times
Well, I compared some single point energies of nwchem and G09 and they are the same. So I guess my definitions were right.

I have problem with two more functionals though which are BHandH and LC-omegaPBE. I defined them in NWCHEM as:
xc beckehandh

and


XC xwpbe 1.00 cpbe96 1.0 hfexch 1.00
cam 0.4 cam_alpha 0.00 cam_beta 1.00 

respectively. I got different single point energies from G09 and NWCHEM. I tried cam 0.3 cam_alpha 0.00 cam_beta 1.00 also for lc-omegaPBE, but still I get different single point energies. Can some one help? Below is a sample input file. My previous post on this was long time back (http://nwchemgit.github.io/Special_AWCforum/st/id472) and I was suggested to use cam 0.4 for lc-omegaPBE, but it did not work.




Here is my sample input file:

title "SCF geometry optimization"
memory 8000 mb
geometry units atomic
Ni 0.000000000000 0.000000000000 0.000000000000
C     2.100732792932     2.100732792932    -2.100732792932
O 3.315335275102 3.315335275102 -3.315335275102
C -2.100732792932 -2.100732792932 -2.100732792932
O -3.315335275102 -3.315335275102 -3.315335275102
C -2.100732792932 2.100732792932 2.100732792932
O -3.315335275102 3.315335275102 3.315335275102
C 2.100732792932 -2.100732792932 2.100732792932
O 3.315335275102 -3.315335275102 3.315335275102
end

DRIVER
tight
maxiter 50
END

dft
convergence energy 1e-8
direct
grid fine
iterations 300
XC xwpbe 1.00 cpbe96 1.0 hfexch 1.00
cam 0.4 cam_alpha 0.00 cam_beta 1.00
end

Forum Vet
beckehandh
Beulahsu,
After reading the G09 documentation, one can conclude that the bHandH functional in G09 and
the beckehandh functional in NWChem are different
in the correlation contribution.
To reproduce the G09 bHandH functional in NWChem you should write the following input lines.

dft
xc hfexch 0.5 slater 0.5 lyp
end

Just Got Here
LC-wPBE
Beulahsu,

Do you know what value of the cam parameter is used in G09 for LC-omegaPBE? That has to be matched in order to get the same results. I compared the results of the implementation in NWChem and Q-Chem for single point energy calculations and TDDFT excitations with LC-omegaPBE a while back and was able to get good agreement, so I'm not sure what's wrong without more information.

Daniel

Clicked A Few Times
The cam parameter in G09 is 0.4, same as the one I used in NWChem.

Just Got Here
LC-wPBE
Beulahsu,

I have been doing some tests using the G09 version we have at my school. Part of the problem is that the functional called "LC-wPBE" has gone through numerous revisions. In fact, LC-wPBE is based on the Heyd-Scurseria-Ernzerhof (HSE) functionals. What is called "LC-wPBE" in G09 may not be exactly the same functional we have implemented in NWChem, since the implementation in NWChem is based on the paper by Henderson et al. (J. Chem. Phys. 2008, 128, 194105). This is supposed to be the best version of the functional, and I believe it is what is implemented in Q-Chem.

You might need to try HSEh1PBE, HSE2PBE, or HSE1PBE as the functional. I don't use G09 enough to be more helpful with this problem, unfortunately.

Good luck,
Daniel

Forum Regular
Hi Beulahsu,

In line with Daniel's comments. Our implementation also uses constants from Weintraub, Henderson, Scuseria. J. Chem. Theory Comput., 2009, 5 , 754. The implementation in Gaussian may be based a different a different parameterisation.

Best regards,
-Niri

Gets Around
Quote:Beulahsu Jan 17th 6:46 pm
Hi,

Can somebody let me know if the following definitions of the functionals are correct? Just want to make sure I am defining them correctly. I would really appreciate your help.

B3PW91: xc acm


Note that b3pw91 in G09 is different from acm

To replicate energies obtained with b3pw91 using g09 in nwchem you need to use
xc HFexch 0.20 slater 0.80 becke88 nonlocal 0.72 perdew91 0.81 pw91lda 1.00

as per [[1]].

For the following functional input:
1. xc acm
2. xc HFexch 0.2 slater 0.8 becke88 nonlocal 0.72 vwn_5 1 Perdew91 0.81
3. xc HFexch 0.20 slater 0.80 becke88 nonlocal 0.72 perdew91 0.81 pw91lda 1.00
4. g09/b3pw91

using 6-31+G* on an unoptimised water molecule you get
1. -76.358375905073 Hartree
2. -76.358375905072
3. -76.355784373093
4. -76.3557851653

On bigger molecules the difference is even larger.

Clicked A Few Times
In the manual (http://nwchemgit.github.io/index.php/Release66:Density_Functional_Theory_for_Molecules#Range...) it says that cam should be set to 0.3 to obtain LC-wPBE.
In JCP 125, 074106 (2006) it appears that 0.4 is the optimal parameter.

If I set cam to 0.4, would I obtain the functional described in the aforementioned article?

xc xwpbe 1.00 cpbe96 1.0 hfexch 1.00
cam 0.4 cam_alpha 0.00 cam_beta 1.00


Thank you very much!
gncs

Forum Regular
Hi gncs,

Yes. That's correct. You can also optimize the parameter using the Koopman's condition.

Best,
-Niri

niri.govind@pnnl.gov


Forum >> NWChem's corner >> General Topics