small Ni cluster convergence


Clicked A Few Times
I have a small Ni cluster as shown below and I cannot get even one complete SCF cycle converged. I am wondering what is wrong in input or else what I can add in order for the convergence to work?

charge 0

geometry units angstroms print xyz
Ni -0.04196 -0.13887 -0.06789
Ni -0.01871 -0.01555 2.40253
Ni 0.62215 -1.76237 1.00667
Ni -1.59185 -1.03913 0.90612
Ni -1.88742 1.21719 0.97918
Ni 0.67365 1.71725 1.09000
Ni 2.15725 -0.01555 1.05771
Ni -0.60253 -1.97149 -0.97491
Ni -1.84449 -0.03822 -0.95520
Ni -0.67889 1.90691 -1.09847
Ni 1.71989 1.21719 -1.25025
Ni 1.52236 -1.03913 -1.01856
Ni -0.02947 -0.03822 -2.07693
end

basis
   Ni library Def2-QZVPP
end

Driver
 Loose
end

dft
   MULT 1
iterations 2000
xc xpbe96 cpbe96
direct
smear
end


task dft optimize

Forum Vet
Please post the last lines of your output file

Clicked A Few Times
sigma= .10D-02 eFermi= -0.25236 T*S= -1.04E-03 nel(1)= 364.00 nel(
d= 0,ls=0.0,diis    40 -19511.3054019887  1.42D+01  7.35D-03  8.14D+02 28231.5


Forum Vet
Are you sure that there is error log? What about the computer system logs?

Forum Vet
The system has a strong linearly dependent problem

!! The overlap matrix has 71 vectors deemed linearly dependent with
   eigenvalues:
4.78D-09 1.01D-08 3.11D-08 4.39D-08 4.53D-08 6.21D-08 1.22D-07 1.25D-07
1.41D-07 1.42D-07 1.43D-07 1.43D-07 1.43D-07 1.43D-07 1.43D-07 1.43D-07
1.43D-07 1.43D-07 1.43D-07 1.43D-07 1.43D-07 1.44D-07 1.65D-07 1.94D-07
2.87D-07 3.09D-07 3.25D-07 4.58D-07 5.41D-07 5.52D-07 5.61D-07 5.65D-07
5.69D-07 5.73D-07 5.78D-07 5.87D-07 5.88D-07 5.91D-07 5.94D-07 5.95D-07
7.53D-07 9.25D-07 1.14D-06 1.15D-06 1.16D-06 1.43D-06 1.74D-06 2.25D-06
2.47D-06 2.77D-06 2.92D-06 2.93D-06 3.03D-06 3.68D-06 4.09D-06 4.46D-06
5.03D-06 5.25D-06 5.53D-06 5.70D-06 5.93D-06 6.04D-06 6.53D-06 6.56D-06
6.91D-06 7.30D-06 8.55D-06 8.82D-06 9.09D-06 9.54D-06 9.67D-06

Should we correct this or let it slide?

Clicked A Few Times
Yes I found this and have started with new geometry. However, I face the same problem everytime. Not sure if there is any problem in my geometry generating software.

Forum Vet
You might want to change the basis set to spherical (from the default cartesian) to cut down on the number of linear dependencies
basis spherical

   Ni library Def2-QZVPP

end

Forum Vet
Dear Dr. Edoapra
 When basis spherical is added to the input, the linear dependency  problem 
has been improved.

WARNING : Found 17 linear dependencies
S eigenvalue threshold:     1.00000E-05
Smallest S eigenvalue : 1.06803E-07
Largest S eigenvalue : 8.13807E-06


!! The overlap matrix has  17 vectors deemed linearly dependent with
eigenvalues:
1.07D-07 1.45D-07 1.80D-07 3.66D-07 3.71D-07 4.23D-07 4.55D-07 4.76D-07
5.06D-07 6.62D-07 1.45D-06 1.78D-06 1.84D-06 6.05D-06 6.17D-06 7.01D-06
8.14D-06

The warning will disappear if a smaller basis set is also utilized, but no tendency will be discovered toward the necessary convergency.

Forum Vet
Try the following input
start nininini

charge 0

geometry units angstroms print xyz
Ni        -0.04196       -0.13887       -0.06789
Ni        -0.01871       -0.01555        2.40253
Ni         0.62215       -1.76237        1.00667
Ni        -1.59185       -1.03913        0.90612
Ni        -1.88742        1.21719        0.97918
Ni         0.67365        1.71725        1.09000
Ni         2.15725       -0.01555        1.05771
Ni        -0.60253       -1.97149       -0.97491
Ni        -1.84449       -0.03822       -0.95520
Ni        -0.67889        1.90691       -1.09847
Ni         1.71989        1.21719       -1.25025
Ni         1.52236       -1.03913       -1.01856
Ni        -0.02947       -0.03822       -2.07693
end

basis spherical
    Ni library Def2-QZVPP
end

Driver
  Loose
end

dft
    MULT 1
    iterations 2000
    xc xpbe96 cpbe96
    direct
    convergence lshift 0.18
    tolerances tight
end

set quickguess t

task dft optimize

Forum Vet
Dear Dr. Edoapra
  Now the warning is the same  

WARNING : Found 17 linear dependencies
S eigenvalue threshold:     1.00000E-05
Smallest S eigenvalue : 1.06803E-07
Largest S eigenvalue : 8.13807E-06


!! The overlap matrix has  17 vectors deemed linearly dependent with
eigenvalues:
1.07D-07 1.45D-07 1.80D-07 3.66D-07 3.71D-07 4.23D-07 4.55D-07 4.76D-07
5.06D-07 6.62D-07 1.45D-06 1.78D-06 1.84D-06 6.05D-06 6.17D-06 7.01D-06
8.14D-06

Superposition of Atomic Density Guess
     -------------------------------------

Sum of atomic energies:      -19588.83253362
Time after variat. SCF: 18.0
Time prior to 1st pass: 18.1


Forum >> NWChem's corner >> Running NWChem