Getting started with NWChem - Convergence problem


Clicked A Few Times
Hi, I'm new to NWChem and, as a self-taught ORCA user, trying to learn as much as I can while I'm preparing to my masters degree in chemistry.
I'm very sorry if this is a stupid question but, after some research, I wasnt able to find a solution.

I've decided to start from the very basics with NWChem, so, my first experiment was to try to calculate the transition state for H2SO4. Here is my input:

# ================================================================
# NWChem input file made in Gabedit
# ================================================================

charge 0
memory 3000 mb

geometry autosym
 S  -1.139420 -0.181760 -0.065320
 O  -0.603930 0.894620 -0.859210
 O  -0.327300 -1.144590 0.633970
 O  -2.146750 0.489400 0.987400
 O  -2.158340 -1.022960 -0.975080
 H  -2.473210 1.320810 0.584900
 H  -2.293600 -1.887270 -0.534180
end

basis  "ao basis" spherical
 * library 6-31G 
end
scf
 Singlet
 maxiter 999
thresh 1e-2
end

task SCF  saddle


And here is the error:

 ------------------------------------------------------------------------[color=red]
 scf saddle failed                                                                                  0[/color]
 ------------------------------------------------------------------------
 ------------------------------------------------------------------------
  current input line : 
    27: task SCF  saddle
 ------------------------------------------------------------------------
 ------------------------------------------------------------------------
 This type of error is most commonly associated with calculations not reaching convergence criteria
 ------------------------------------------------------------------------
 For more information see the NWChem manual at http://nwchemgit.github.io/index.php/NWChem_Documentation


 For further details see manual section:                                                                                                                                                                                                                                                                
0:0:scf saddle failed:: -1
(rank:0 hostname:localhost.localdomain pid:10924):ARMCI DASSERT fail. src/common/armci.c:ARMCI_Error():208 cond:0
--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 0 in communicator MPI COMMUNICATOR 4 DUP FROM 0 
with errorcode -1.

NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
You may or may not see output from other processes, depending on
exactly when Open MPI kills them.
--------------------------------------------------------------------------


It is a convergence issue, right? I was thinking that the problem was going to be solved using more iterations for scf: (lets say, maxiter 999), but, still the convergence problem is not fixed.
What is the correct way to force the scf saddle task to have more iteration in order to reach a convergence?
Henrique C. S. Junior
Inorganic Chemist - UFRRJ - Rio de Janeiro

Gets Around
You have increased the allowable iterations for SCF electronic convergence but you need to increase the allowable iterations for geometry optimization. You should also check your transition state solution by running a frequency calculation and looking for a single imaginary frequency. To get a good TS solution you need to make the geometry convergence tighter than default. Your SCF threshold was too loose for this job. Here's an input file that should work better:

charge 0
memory 3000 mb

geometry autosym
 S  -1.139420 -0.181760 -0.065320
 O  -0.603930 0.894620 -0.859210
 O  -0.327300 -1.144590 0.633970
 O  -2.146750 0.489400 0.987400
 O  -2.158340 -1.022960 -0.975080
 H  -2.473210 1.320810 0.584900
 H  -2.293600 -1.887270 -0.534180
end

basis  "ao basis" spherical
 * library 6-31G 
end

scf
 singlet
 thresh 1e-4
end

driver
  tight
  maxiter 999
end

task scf saddle
task scf frequencies

Clicked A Few Times
Thank you for your reply and sorry for my delay (I went on vacation).
So, this block
driver
  tight
  maxiter 999
end

Is where I can set more iterations for geometry optimization?

Gets Around
Yes, you can set more iterations for geometry optimization in the driver block. Or if you are using the non-default stepper geometry optimizer, you can do the same thing in a stepper block.

Also the default scf thresh value is 1e-4, as I made explicit above; I'm not sure why it was loosened in the original input file. Maybe Gabedit put that in for you.


Forum >> NWChem's corner >> General Topics