Effective core potentials - convergence and freq issues

[Jakub]

Hi all,

I'm working on a neutral Gd(III) complex. The Gd itself is described using the Stuttgart relativistic ECP. I'm running into two problems:

1) The convergence in the first optimisation cycle looks incredibly erratic and takes a very long time (a few hundred iterations, sometimes doesn't converge at all). Any idea why that might be? Or

2) Even after successful optimization I get the following error when I try to perform a frequency calculation:

0: error ival=4
(rank:0 hostname:qnode6871 pid:2952):ARMCI DASSERT fail. ../../ga-5-4/armci/src/devices/openib/openib.c:armci_call_data_server():2209 cond:(pdscr->status==IBV_WC_SUCCESS)
24: error ival=10
(rank:24 hostname:qnode8061 pid:41107):ARMCI DASSERT fail. ../../ga-5-4/armci/src/devices/openib/openib.c:armci_call_data_server():2209 cond:(pdscr->status==IBV_WC_SUCCESS)

Thanks!

And here's the input file:

echo

start molecule

title "full"
charge 0

memory global 2400 mb stack 800 mb heap 100 mb noverify

geometry units angstroms print xyz autosym noautoz
  Gd        0.40951        0.07401        0.09204
   S        4.85393       -0.61518       -2.30596
----------------------details continue-------------------------

Symmetry C1
end

BASIS
  C  library 6-311G**
  O  library 6-311G**
  H  library 6-311G**
  N  library 6-311G**
  S  library 6-311G**
  Gd library stuttgart_rsc_1997_ecp
END

ECP
  Gd library stuttgart_rsc_1997_ecp
END

dft
  xc b3lyp
  mult 8
  maxiter 1000
end

driver
 maxiter 500
end

task dft optimize

task dft freq

[Edoardo Aprà]

You might want to post the full input and output files for a better analysis of your problem

[Jakub]

See attached.

[Edoardo Aprà]

Jakub

Here is a modified input file that should help for the convergence problems.

It uses a fragment guess, by computing first the molecular orbitals for Gd 3+ and the rest of the molecules (with a 3- charge).

I have also change the basis set to spherical (from the default cartesian).

I have tweaked accuracy of the evaluation of the Coulomb potential, too.

I was not able to reproduce your failure in the frequency part.

Since you seem to work with an older version of NWChem (6.6) my recommendation is to use the latest 7.0 release.

[Jakub]

Thank you!