SO-DFT stability
Clicked A Few Times
10:32:40 AM PDT - Thu, Sep 18th 2014
Dear all,

I'm trying to do some spin-orbit DFT calculations on Br atom using NWCHEM. I found that I got different results by the same input:

echo
Memory 2048 mw
charge 0
geometry noautoz noautosym units angstrom
Br                 0.00000000    0.00000000    0.00000000
end
basis "ao basis" spherical print
Br library crenbl_ecp
END
ECP
Br library crenbl_ecp
END
dft
mult 2

xc m06

grid xfine

convergence energy 1.000E-08

tolerances acccoul 13

iterations 200

vectors input atomic  output irat.mos
end
task dft
SO
.......
END
dft
vectors input irat.mos output irso.mos
end
task sodft

I followed the way suggested by "Edoapra" from old post. The energies are different by 3.8e-3 hartree between two identical inputs. The only difference between my two jobs is that one uses single core while another uses 16 cores. Is there a method to check the stability of the sodft wavefunction? Or the reliability of the sodft single point energy?
Thanks a lot!

Forum Vet
1:21:32 PM PDT - Thu, Sep 18th 2014
What version of NWChem have you been you using for this input?
Could you provide details about your installation, both hardware and software?
Thanks, Edo

Clicked A Few Times
1:59:34 PM PDT - Thu, Sep 18th 2014
I'm using NWCHEM 6.1.1. I'm using external computing resource from a supercomputing center that has intel/gnu C, C++, Fortran compiler installed. I believed that I could run NWCHEM using mpi on a Unix-based paralell machine by:

mpirun -np (# of core) nwchem input_file > output_file


Thank you a lot!

Forum Vet
3:53:56 PM PDT - Thu, Sep 18th 2014
Could you try the input below to see if you still get such a pretty large variation in results between parallel and serial jobs?
Thanks

start brso
echo
memory 2048 mw 
charge 0
geometry 
 Br                 0.00000000    0.00000000    0.00000000
symmetry oh
end
basis "ao basis" spherical  print
Br library crenbl_ecp
END
ECP
Br library crenbl_ecp
END

dft
 mult 2
 xc m06
 fon alpha filled 9 electrons 0.  partial 1
 fon beta filled 7 electrons 1. partial 2
 grid xfine
 direct
 convergence energy 1.000E-09  
 iterations 200
 vectors input atomic  output irat.mos
end
task dft
SO
  Br ul
    2 2.54780006 -0.039374
    2 9.90429974 0.262533
    2 32.94390106 -0.562529
    2 98.63279724 2.001467
    1 287.78079224 0.751481
  Br P
    2 2.34380007 -2.114400
    2 2.78999996 8.375434
    2 3.77500010 -12.007087
    2 5.22399998 5.418710
    1 15.22420025 -0.570016
    0 14.20930004 0.023705
  O ul
    2 10.028600 0.007783
    2 34.198000 -0.006240
    1 100.003900 0.036949
END
dft
 vectors input irat.mos output irso.mos
 fon  filled 16 electrons 1.  partial 2
end
task sodft

Forum Regular
8:31:51 AM PDT - Fri, Sep 19th 2014
Hi,

We have reproduced the variation here as well. We'll investigate and get back to you.

Best,
-Niri

Clicked A Few Times
3:27:19 PM PDT - Fri, Sep 19th 2014
Thank you for your reply! But my two jobs still have different energies by 8e-4 Hartree, both run in parallel.


Forum >> NWChem's corner >> General Topics