Geometry Optimization with COSMO
Clicked A Few Times
10:59:42 AM PST - Fri, Jan 14th 2011
Hi,
I'm interested in running geometry optimization with COSMO. Is this possible? The code is doing something in my test jobs, but the "@" lines are not printed, the "sol phase energy" does not change, and the geometry coordinates do not change.

I assume the problem has something to do with the numerical gradients for COSMO. If someone has a test job that works, I'd be interested in trying it out.

Thanks,
--Niel

Forum Vet
11:36:41 AM PST - Tue, Jan 18th 2011
This should work. Could you post your input deck so we can take a look?

Bert

Clicked A Few Times
4:43:33 PM PST - Tue, Jan 18th 2011
Hi Bert,

Here is the input that I have most recently tested and have confirmed that it does not progress (the coordinates change only slightly). This was run with nwchem6.0 on a cray XT4 (athena at NICS). It's kinda big, but it contains all the functionality I would like to use. I have tested cosmo optimizations in the past with various methods and basis sets on various molecules (small and large). Nothing seems to work. Any help would be greatly appreciated.

Thanks,
--Niel

Clicked A Few Times
2:41:13 PM PST - Sun, Feb 13th 2011
I still haven't figured this out. Here is a much simpler input file for DFT/COSMO optimization of a small molecule.

http://www.amber.utah.edu/biodata/cosmo-opt.nw

(I would directly post it, but there seems to be a limit on the size of the post)

Using this input, the job will appear to be doing something, but the geometry does not change significantly and the "sol phase energy" doesn't either.

Any help would be great. Thanks,
--Niel

Clicked A Few Times
3:49:19 PM PST - Sun, Feb 13th 2011
Oops. I believe I have been mistaken here. The code finally produced a "@" line and appears to be running as expected. I guess I was confused by the COSMO output.

False alarm.

Sorry for the confusion,
--Niel

Clicked A Few Times
4:35:57 PM PDT - Fri, Jul 29th 2011
I'd be interested in knowing what is going on with the COSMO calculations here. Does it run it multiple times (20) in order to calculate the gradients for the next geometry step? Is there any way to reduce the number of iterations it runs through for each geometry step?

Thanks!

Just Got Here
9:35:37 AM PDT - Thu, May 23rd 2013
Geometry not updated
Hi I'm having a similar problem to the one outlined by Niel here. When I run Cosmo geometry
optimization jobs my geometry is not updated and my output also lacks the "@ lines" Niel mentions.
My input is as follows:

geometry full-system

C -1.0966 -2.5162 0.0025
C 1.1272 -1.2719 0.0017
C 0.4433 -2.4950 0.0026
C 1.1606 -3.6989 0.0037
C 0.3390 0.0511 0.0005
C -1.0623 0.0318 0.0004
C -1.8139 -1.3123 0.0014
C -1.7805 -3.7394 0.0034
C -0.9922 -5.0623 0.0046
C 0.4090 -5.0431 0.0047
C 2.5619 -3.6796 0.0038
C 3.2458 -2.4565 0.0029
C 2.5285 -1.2526 0.0018
H 3.0507 -0.3187 0.0011
H 4.3157 -2.4418 0.0030
H 3.1096 -4.5988 0.0046
H 0.9567 -5.9623 0.0055
H -1.5145 -5.9962 0.0053
H -2.8504 -3.7541 0.0033
H -2.8838 -1.3270 0.0013
H -1.6100 0.9510 -0.0004
H 0.8612 0.9850 -0.0002
Fe 0.4458 -2.5034 2.4377
Cl 2.5958 -2.5034 2.4377
Cl -0.6292 -0.6414 2.4377
Cl -0.6292 -4.3653 2.4377
end
geometry fe-only
Fe 0.4458 -2.5034 2.4377
end
geometry c-only
C -1.0966 -2.5162 0.0025
C 1.1272 -1.2719 0.0017
C 0.4433 -2.4950 0.0026
C 1.1606 -3.6989 0.0037
C 0.3390 0.0511 0.0005
C -1.0623 0.0318 0.0004
C -1.8139 -1.3123 0.0014
C -1.7805 -3.7394 0.0034
C -0.9922 -5.0623 0.0046
C 0.4090 -5.0431 0.0047
C 2.5619 -3.6796 0.0038
C 3.2458 -2.4565 0.0029
C 2.5285 -1.2526 0.0018
H 3.0507 -0.3187 0.0011
H 4.3157 -2.4418 0.0030
H 3.1096 -4.5988 0.0046
H 0.9567 -5.9623 0.0055
H -1.5145 -5.9962 0.0053
H -2.8504 -3.7541 0.0033
H -2.8838 -1.3270 0.0013
H -1.6100 0.9510 -0.0004
H 0.8612 0.9850 -0.0002
end
geometry cl-only
Cl 2.5958 -2.5034 2.4377
Cl -0.6292 -0.6414 2.4377
Cl -0.6292 -4.3653 2.4377
end

  1. Basis set #

basis
Cl library 6-31g
C library 6-31g
H library 3-21g
Fe library 6-31g
end
basis "cd basis"
  • library "Ahlrichs Coulomb Fitting"
end

  1. DFT on the fragments for initial guess for full system #

driver
maxiter 800
end

charge 3
set geometry fe-only
cosmo
dielec 78.0
rsolv 0.5
lineq 0
end
dft
direct
mult 6
XC b3lyp
vectors atomic output fe.mo
iterations 100
end
task dft

charge 0
set geometry c-only
cosmo
dielec 78.0
rsolv 0.5
lineq 0
end
dft
direct
mult -2
XC b3lyp
vectors atomic output c.mo
iterations 100
end
task dft

charge -3
set geometry cl-only
cosmo
dielec 78.0
rsolv 0.5
lineq 0
end
dft
direct
mult 1
XC b3lyp
vectors atomic output cl.mo
iterations 100
end
task dft

  2. DFT on the full system #

unset dft:*
driver
xyz temp
end
cosmo
dielec 78.0
rsolv 0.5
lineq 0
end
charge 0
set geometry full-system
dft
direct
mult 5
XC b3lyp
vectors fragment fe.mo c.mo cl.mo output fecl3.mo
convergence nolevelshifting
cdft 23 26 1 13 charge 0.0
iterations 800
end
task dft optimize

In this particular input I use the default values for the solute atom radius parameter, but I have
experimented with "set cosmo:map cosmo.par" as described in the documentation. Is there a highly
recommended fix for this problem?

Forum Vet
9:59:11 AM PDT - Thu, May 23rd 2013
NWChem 6.3 has analytic gradients for COSMO
Older versions of NWChem used numerical gradients for COSMO geometry optimizations, which obviously take a lot of time. The latest version has analytical gradients, which are much more efficient. Hence, the recommended fix is to download and compile NWChem 6.3.

Forum Vet
11:54:35 AM PDT - Fri, May 24th 2013
CDFT and COSMO
It looks like CDFT and COSMO do not work together properly. Without the CDFT the energy calculation will converge, with the CDFT the energy calculation will not converge in the COSMO step. There is as of now no quick fix for this problem.

Bert


Forum >> NWChem's corner >> NWChem functionality