MP2 density in gaussian cube format
Just Got Here
10:38:23 AM PDT - Mon, Jul 7th 2014
I am trying to output the MP2 density in gaussian cube file format. More specifically, I am trying to reproduce the results of the following Gaussian input file:

  1. MP2/6-31G** guess=core nosym density=MP2 pop=esp cube=cards cube=frozencore

water001 user i fragm 1

0 1

O    0.000000    0.000000    0.000000

H    0.756950    0.585880    0.000000

H   -0.756950    0.585880    0.000000

c:\clp\doc\dens\water001a.den
   0   -3.240000   -3.240000   -3.240000

  80    0.080000    0.000000    0.000000

  80    0.000000    0.080000    0.000000

  80    0.000000    0.000000    0.080000

I do not own a gaussian license, therefore I am trying to reproduce this particular calculation using free software. I have tried the following syntax, but the density is slightly different to the original gaussian cube's (while the grid's dimensions are identical). Is there something wrong with the syntax or is it just a matter of different numerical precision/approximations? I have been able to reproduce the exact same result in pcgamess/firefly, but I have no experience with nwchem. Can someone please help me with this?

start molecule
title "water nwchem mp2 density"
charge 0
geometry units angstroms print xyz nocenter noautosym
  O        0.00000        0.00000        0.00000

  H        0.75695        0.58588        0.00000

  H       -0.75695        0.58588        0.00000
end
basis Cartesian
 * library 6-31G**
end
scf
 vectors output molecule.movecs
end
mp2
 vectors molecule.movecs

 tight

 freeze core

 vectors output molecule.mp2nos
end
dplot
 title "water cube"

 vectors molecule.mp2nos

 limitxyz

 -3.24 3.08 79

 -3.24 3.08 79

 -3.24 3.08 79

 gaussian

 where grid

 output water.cube
end
task scf
task mp2 gradient
task dplot

Gets Around
12:00:38 PM PDT - Mon, Jul 7th 2014
Can you provide your Firefly input? I think it should work in GAMESS-US, and would help to diagnose differences.

Just Got Here
12:07:15 PM PDT - Mon, Jul 7th 2014
Quote:Mernst Jul 7th 12:00 pm
Can you provide your Firefly input? I think it should work in GAMESS-US, and would help to diagnose differences.


Here it is:

$CONTRL COORD=UNIQUE RUNTYP=ENERGY MPLEVL=2 EXETYP=RUN $END

$BASIS GBASIS=N31 NGAUSS=6 NDFUNC=1 NPFUNC=1 NFFUNC=0 DIFFSP=0 DIFFS=0 $END

$GUESS GUESS=HUCKEL $END

$LOCAL FCORE=1 $END

$ELDENS IEDEN=1 WHERE=CUBE $END

$ELPOT IEPOT=1 WHERE=CUBE $END

$PDC PTSEL=CONNOLLY $END

$CUBE CUBE=1

NXCUBE=80 NYCUBE=80 NZCUBE=80   

X0CUBE=-6.122713 Y0CUBE=-6.122713 Z0CUBE=-6.122713   

XINCUB=0.151178 YINCUB=0.151178 ZINCUB=0.151178 

$END

$SYSTEM TIMLIM=10000 MWORDS=80 FASTF=.T. WSCTL=.T.

$END

$DATA
Water
C1
O    8.0 0.000000    0.000000    0.000000

H    1.0 0.756950    0.585880    0.000000

H    1.0 -0.756950    0.585880    0.000000

$END

Gets Around
2:11:58 PM PDT - Mon, Jul 7th 2014
Well, that didn't work. GAMESS-US and Firefly have diverged more than I thought.

After some reading and looking I noticed your "freeze core" directive. This doesn't actually do anything -- you need to put a greater-than-zero number after "core" for it to have an effect, e.g. "freeze core 1".

I read the Gaussian manual about frozen cores and there sure are a lot of options! I don't know if NWChem's "freeze atomic" is equivalent to the default Gaussian freezing scheme for post-HF calculations, but it's worth a quick try.

If the results still don't match it is time to carefully read your Firefly and NWChem output files side by side. Are the numbers of orbitals and basis functions the same for both calculations? Do the energies for SCF and MP2 agree to a reasonable number of decimal places? You may also want to manually compare the basis set definitions that are echoed in the output files. You would think that the definition of "6-31G**" was set in stone long ago, but I have been surprised in the past to find how software has interpreted named Pople basis sets.

The earliest significant numerical divergence between the two jobs needs to be corrected or explained before downstream results. If everything agrees closely except the final cube, that would be surprising.

Just Got Here
12:55:45 AM PDT - Tue, Jul 8th 2014
The freeze core / freeze atomic keywords don't seem to affect the result. The cube files are similar but they should be identical to several decimal digits, not just the first, which makes me wonder whether the dplot command prints the scf density instead of the MP2 density in the final cube file. Perhaps I should have mentioned earlier that I am using a precompiled Windows executable of nwchem 6.0.

The gaussian input is an example file from A. Gavezzotti's clp package. Clp automatically generates input for gaussian, in order to obtain electron densities. Firefly can be used as a substitute for gaussian for that particular calculation, but I was hoping that it would be just as easy to perform the same calculations with nwchem, because nwchem has more extended capabilities than firefly.

I will proceed as you suggest when I have a little more time. Thanks for your help.


Forum >> NWChem's corner >> General Topics