convergence threshold affected by property calculation


  • Guest -
1:36:15 PM PST - Wed, Jan 11th 2012
Hi,

I am doing a DFT property calculation to evaluate the Mulliken population (Property ; Mulliken ; end [...] Task DFT property). I noticed that the property module changes the convergence criteria from the defaults to make them more stringent as you can see from the following output lines :

                                NWChem Property Module

                                ----------------------



 itol2e modified to match energy

 convergence criterion.



                                NWChem DFT Module

                                -----------------



         Convergence on energy requested: 1.00D-07

         Convergence on density requested: 1.00D-05

         Convergence on gradient requested: 1.00D-06

Screening Tolerance Information
     -------------------------------

         Density screening/tol_rho: 1.00D-11

Is there a way to override this in order to keep the default thresholds and speed up the calculation?

Thanks,

Chloe

  • Guest -
12:02:46 PM PST - Fri, Jan 13th 2012
Any idea? Somebody?

Chloé

Forum Vet
5:55:23 PM PST - Fri, Jan 13th 2012
Insert "task dft energy"
Chloe,
If you inser the line "task dft energy", the energy calculation will be run with the default convergence threshold for energy (or even looser, if you re-define them in the dft input field), e.g.

start hh

geometry
h 0. 0. 0.
h 0. 0. 1.
end
basis
* library sto-3g
end

Property ; Mulliken ; end
dft
convergence energy 1d-6
end

task dft energy

task dft property

  • Guest -
8:13:17 AM PST - Mon, Jan 16th 2012
Thanks, but this still reconverges the DFT with an energy threshold of 1e-7 when the property is evaluated. That is what I am trying to avoid. Is there a way of only evaluating the property using the vectors converged during the energy calculation?

Chloé

Forum Vet
3:43:28 PM PST - Tue, Jan 24th 2012
You may want to try setting "mulliken" in the DFT block itself. It should perform a basic Mulliken analysis there too.

Alternatively, there is a little known feature in the code that would prevent the recalculation of the vectors. When you specify a specific set of vectors in the property block it should just use those without recomputing.


start hh

geometry
  h 0. 0. 0.

  h 0. 0. 1.
end

basis
  * library sto-3g
end

property 
  Mulliken 

  vectors hh.movecs
end

dft
  convergence energy 1d-6
end

task dft energy

task dft property


A second option is to use the following input block, where the prop:nocheck parameter that's get set avoids DFT setting a tighter convergence for the property calculation.


start hh

geometry
  h 0. 0. 0.

  h 0. 0. 1.
end

basis
  * library sto-3g
end

property 
  Mulliken
end

dft
  convergence energy 1d-6
end

task dft energy

set prop:nocheck .true.

task dft property



Bert




Quote: Jan 16th 3:13 pm
Thanks, but this still reconverges the DFT with an energy threshold of 1e-7 when the property is evaluated. That is what I am trying to avoid. Is there a way of only evaluating the property using the vectors converged during the energy calculation?

Chloé

  • Guest -
12:49:24 PM PST - Wed, Jan 25th 2012
Thanks a lot Bert!

Chloé


Forum >> NWChem's corner >> NWChem functionality