NaN energies in DFT calculation with M06 family of functionals


Clicked A Few Times
Hello,

I'm running into a strange problem when using NWChem 6.6 to calculate single-point energies with the DFT module for a fragment guess. When using the M06-HF, M06-L, or M06 xc functionals, I start to obtain NaN energies after several iterations.

A minimal input file that reproduces the issue:
geometry mn1
symmetry c1
Mn         1.51555        0.49368       -0.79695
Bq        -0.16540        1.26690       -0.21663 charge -1 # O phenoxide
Bq         2.80915       -1.21032       -0.74424 charge -1 # N backbone
Bq         3.44038        1.26002       -1.86467 charge -1 # N backbone
Bq         2.37297        1.59393        0.70502 charge -1 # N backbone
Bq         0.20258       -1.29026       -0.05376 charge -1 # O phenoxide
Bq         0.73640        0.02448       -2.38214 charge -1 # O peroxide
end

basis spherical
  Mn library "aug-cc-pvdz"
end

set geometry mn1
charge -3
dft
  xc m06-hf
  direct
  odft
  mult 5
  vectors input atomic output mn1.vec
  grid ultrafine
end
title "step 1 : spin-up Mn"
task dft energy

When I submit this input file (NWChem 6.6, LINUX64 platform, 16 MPI tasks), the following occurs:
   convergence    iter        energy       DeltaE   RMS-Dens  Diis-err    time
 ---------------- ----- ----------------- --------- --------- ---------  ------
 d= 0,ls=0.0,diis     1  -1152.7758220863 -1.11D+03  2.03D-01  3.20D+00    35.2
                                                     2.12D-01  1.53D+00
 d= 0,ls=0.0,diis     2  -1153.4568255155 -6.81D-01  2.19D-02  1.72D-01    45.9
                                                     3.18D-02  1.80D-01
 d= 0,ls=0.0,diis     3  -1153.4955096732 -3.87D-02  9.04D-04  9.58D-04    54.1
                                                     1.38D-03  1.52D-02
 d= 0,ls=0.0          4               NaN       NaN  7.40D-04              66.5
                                                     1.08D-03
 d= 0,ls=0.0          5               NaN       NaN  2.31D-04              75.0
                                                     4.22D-04

This continues until the maximum number of iterations is reached, at which point NWChem says the calculation failed to converge and exits.

I can only reproduce this problem with the M06 family of functionals - the calculation converges with the B3LYP and M11 functionals.

Any insights would be greatly appreciated!

Forum Vet
patch needed
Thanks for reporting this bug.
A patch is now available.
http://nwchemgit.github.io/download.php?f=Xccvs98.patch.gz
To apply

More details at
http://nwchemgit.github.io/index.php/Download#Patches_for_the_27746_revision_of_NWChem_6.6

Clicked A Few Times
Thanks
Thanks for the quick patch, much appreciated!


Forum >> NWChem's corner >> Running NWChem