driver: task gradient failed


Just Got Here
Hi

I would like to perform a geometry optimization of a quite large system. I'm getting this error after several hours of running:
[[File:


driver: task_gradient failed        0
------------------------------------------------------------------------
------------------------------------------------------------------------
current input line :
139: task dft optimize
------------------------------------------------------------------------
------------------------------------------------------------------------
]]
I guess this error comes from the fact that after 400 iterations the calculation did not converged. (although the integrated energy is not too far from the threshold I think) Which possibilities are available to avoid that problem?
(I have to say that is my first time using NWChem/6.0 )

The input looks like:

[[File:
start ZnO102.nw

title " SCF geometry optimization"
set geometry:actlist 1:58
geometry noautoz
C                 15.61693561    7.52673614   34.99168772
C 16.50777789 6.56286593 35.43699734
S 15.95533612 9.04085619 35.79846122
C 14.54927585 7.34057802 34.04427977
C 17.42705997 7.02802126 36.39012240
H 16.49218006 5.54166388 35.06345007
C 17.27377659 8.36220055 36.70965177
C 13.78037816 8.24763001 33.31969494
S 14.09806690 5.70256408 33.66132085
.
.
.
O 24.52696241 11.39622069 26.12607905
Zn                22.90196241   12.33422069   26.93307905
O 22.90196241 14.21022069 26.12607905
Zn 21.27696241 15.14822069 26.93307905
O 21.27796241 17.02422069 26.12607905
Zn 27.77496241 9.52122069 26.93307905
Zn 26.15096241 12.33422069 26.93307905
Zn 24.52596241 15.14822069 26.93307905
Zn 22.90196241 17.96222069 26.93307905

end
basis
  Zn library LANL2DZ_ECP
O library Stuttgart_RLC_ECP
C library 3-21g
S library Stuttgart_RLC_ECP
H library 3-21g
end
ecp
  Zn library LANL2DZ_ECP
O library Stuttgart_RLC_ECP
S library Stuttgart_RLC_ECP
end
dft
  iterations 400
end
task dft optimize

]]

the output does the 400 iterations and the last 20 os so have the same integrated energy (Shown are the last 5):

[[File:d= 0,ls=0.5,diis 396 -3380.2655940465 -3.79D-06 2.63D-06 6.49D-06 4634.1
WARNING: error on integrated density =  0.32D-05
greater than required accuracy of 0.10D-05
d= 0,ls=0.5,diis 397 -3380.2655973522 -3.31D-06 2.39D-06 9.85D-06 4648.2
WARNING: error on integrated density = 0.32D-05
greater than required accuracy of 0.10D-05
d= 0,ls=0.5,diis 398 -3380.2656006808 -3.33D-06 2.49D-06 1.19D-05 4658.2
WARNING: error on integrated density = 0.32D-05
greater than required accuracy of 0.10D-05
d= 0,ls=0.5,diis 399 -3380.2656043009 -3.62D-06 2.44D-06 1.29D-05 4674.4
WARNING: error on integrated density = 0.32D-05
greater than required accuracy of 0.10D-05
d= 0,ls=0.5,diis 400 -3380.2656078752 -3.57D-06 2.38D-06 1.38D-05 4682.0
WARNING: error on integrated density = 0.32D-05
greater than required accuracy of 0.10D-05

     Calculation failed to converge
------------------------------
Total DFT energy = -3380.265611647519
.
.
.
DFT Final Molecular Orbital Analysis
------------------------------------

Vector   27  Occ=2.000000D+00  E=-9.733423D+00
MO Center= -8.1D+00, -1.3D+00, 4.8D+00, r^2= 3.1D-02
Bfn. Coefficient Atom+Function Bfn. Coefficient Atom+Function
----- ------------ --------------- ----- ------------ ---------------
264 0.984507 42 C s
.
.
.]]

Gets Around
I guess playing with convergence keyword may help (i haven't tried that):
CONVERGENCE energy 5e-6
http://nwchemgit.github.io/index.php/Density_Functional_Theory_for_Molecules#CONVERGENCE_--_...
Note, that depending on the level of precision you aim at, you mey need to repeat such preliminary optimization (if succeeds) with tighter convergence parameters starting from the optimized "lousy" structure.


Forum >> NWChem's corner >> Running NWChem