scf convergence fails for Uranium based coordination compounds
Clicked A Few Times
3:19:24 AM PDT - Fri, May 20th 2016
Hi. I am trying to run a series of calculations on different uranium based coordination compounds, but I got all of them failing- the scf cycle does not converge

I am copying my input file

GEOMETRY
H    -7.232548    -2.275358     0.486014

C    -6.492637    -2.338257     1.270217

C    -5.497566    -3.323589     0.914568

C    -4.261059    -3.510227     1.476223

O    -3.832645    -2.851531     2.511246

U    -5.071889    -2.454629     4.342482

N    -3.643535    -3.269865     6.316560

C    -2.706916    -4.327936     6.039378

N    -6.500162    -1.639181     2.368390

C    -7.436677    -0.581028     2.645611

O    -5.978969    -3.919024     4.113230

O    -4.164847    -0.990206     4.571715

O    -6.311160    -2.057728     6.173702

C    -5.882773    -1.398936     7.208675

C    -4.646227    -1.585437     7.770299

C    -3.651084    -2.570705     7.414682

H    -3.553654    -4.202668     1.044363

H    -6.590210    -0.706504     7.640498

H    -2.911097    -2.633428     8.198828

H    -5.714970    -3.878461     0.004695

H    -4.428820    -1.030466     8.680109

H    -7.977207    -0.326872     1.735034

H    -8.142378    -0.910972     3.406560

H    -6.898684     0.294988     3.004888

H    -2.915897    -4.749701     5.057360

H    -1.693684    -3.929138     6.055021

H    -2.802846    -5.105133     6.796137
end

BASIS "ao basis" PRINT
U Library "Stuttgart RSC 1997 ECP"
H Library cc-pVDZ
C Library cc-pVDZ
O Library cc-pVDZ
N Library cc-pVDZ
END

CHARGE 0

DPLOT
SPIN spindens
END

scf
direct filesize 100000000

singlet
end

RELATIVISTIC
ZORA
END

DFT
xc pbe96
iterations 500
convergence energy 1e-6 density 1e-5
smear 1
grid fine
mulliken
end

TASK dft optimize
task dft freq


I have tried with and without smear, fine and xfine grid and changing the convergence criteria for energy and density. In all cases, the SCF behaves like this

  convergence    iter        energy       DeltaE   RMS-Dens  Diis-err    time

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

d= 0,ls=0.0,diis     1  -8641.0811366797 -1.14D+04  6.27D+00  1.13D+02   438.2

Grid integrated density:     198.001640727949

Requested integration accuracy:   0.10E-06

sigma= .10D+01 eFermi= -1.22871 T*S= -4.56E+01 nel(1)=  198.00 nel(

d= 0,ls=0.0,diis     2  -8339.8792074782  3.01D+02  6.48D+00  3.94D+02   560.1

Grid integrated density:     198.000395016022

Requested integration accuracy:   0.10E-06

sigma= .10D+01 eFermi= -1.11165 T*S= -5.67E+01 nel(1)=  198.00 nel(

d= 0,ls=0.0,diis     3  -8578.2631970768 -2.38D+02  1.32D+00  1.07D+02   687.6



        **************
Requested integration accuracy: 0.10E-06
sigma= .10D+01 eFermi= -0.96883 T*S= -6.04E+01 nel(1)=  198.00 nel(

d= 0,ls=0.0,diis   498  -8631.7121867927  6.15D-03  1.09D-03  1.79D+00 62222.7

Grid integrated density:     198.000545774563

Requested integration accuracy:   0.10E-06

sigma= .10D+01 eFermi= -0.96862 T*S= -6.04E+01 nel(1)=  198.00 nel(

d= 0,ls=0.0,diis   499  -8631.7236499945 -1.15D-02  9.17D-04  1.79D+00 62348.8

Grid integrated density:     198.000545975440

Any idea how I could get convergence?

Forum Regular
2:24:48 PM PDT - Fri, May 27th 2016
Hi,

A couple of questions:

1. Are you sure this is a charge 0 ?
2. Are you sure this is a singlet system ?

Best,
-Niri Govind

niri.govind@pnnl.gov

Clicked A Few Times
2:44:50 AM PDT - Mon, Jun 6th 2016
Hi Niri.

Thank you for your answer.

Yes, I am pretty sure it is a U(VI). What would you normally do when having convergence problems on the first SCF cycle?

Clicked A Few Times
2:45:02 AM PDT - Mon, Jun 6th 2016
Hi Niri.

Thank you for your answer.

Yes, I am pretty sure it is a U(VI). What would you normally do when having convergence problems on the first SCF cycle?

Forum Regular
8:25:52 AM PDT - Mon, Jun 6th 2016
I don't know if it is your only problem, but you don't specify an ECP block in the input file you have copied. This should be required for the basis set that you have specified for U.

Clicked A Few Times
5:06:03 AM PDT - Thu, Jun 9th 2016
Thank you for the answer, but that is not the problem.

I am sure the input file is correct, I am asking which are the usual steps to take when having convergence problems on the first SCF cycle. A combination of level shifts and turning on/off diis?

Forum Regular
8:21:50 AM PDT - Thu, Jun 9th 2016
I do believe that the lack of an ECP for uranium is a problem. By not specifying the ECP, you are trying to fit 92 electrons (core and valence) into a basis set that was designed for 32 valence electrons. Even if you could get the SCF to converge with such a setup, I would be highly skeptical of the usefulness of the results. With the ECP, the SCF converges in about 50 steps without the need for smearing or anything else. As far as general advice for difficulty SCF convergence problems, several steps of the quadratic convergence algorithm (keyword cgmin), followed by the standard diis algorithm is often helpful.

Clicked A Few Times
8:56:38 AM PDT - Thu, Jun 9th 2016
Thank you very much for your answer. I said that that was not the problem because in my input file I did have the ECP, I must have forgotten to paste it when I posed the question.

Thank you very much for the advice on cgmin, ill try this out.

Cheers

Forum Regular
4:45:56 AM PDT - Fri, Jun 10th 2016
I only kept on the point because the output you showed indicated that you had not specified the ECP since it said there were 198 electrons (nel(1) = 198, for this system there should be 138 since 60 of uranium's electrons are covered by the ECP).

Clicked A Few Times
6:08:34 AM PDT - Fri, Jun 10th 2016
Ah! I did not realize about that. My bad! Thank for your answers!


Forum >> NWChem's corner >> General Topics