CCSD Divergence for Cation in presence of point charges


Clicked A Few Times
Hi, I am having an issue getting a CCSD calculation. The thing that is strange to me is with the same point charge setup the nuetral and anion versions of my molecule converge, but no the cation. Here is the input:

charge 1
memory stack 2500 mb heap 500 mb global 10000 mb
geometry nocenter noautosym noautoz
F 0 0 0
H 1.6 0 0
F 3.2 0 0
H 4.8 0 0
end
bq
-320.0 0.0 0.0 -0.768784
-318.4 0.0 0.0 0.659618
-316.8 0.0 0.0 -0.660149
-315.2 0.0 0.0 0.769315
-313.6 0.0 0.0 -0.768784
-312.0 0.0 0.0 0.659618
-310.4 0.0 0.0 -0.660149
-308.8 0.0 0.0 0.769315
... many more charges here.
basis
  • library 6-31g*
end
scf
rohf; doublet
end
tce
scf
diis 15
maxiter 10000
tilesize 30
2eorb
ccsd
end
task tce energy

Here is what occurs in the output:

CCSD iterations
-----------------------------------------------------------------
Iter Residuum Correlation Cpu Wall V2*C2
-----------------------------------------------------------------
1 0.2144145206232 -0.3429323564682 1.7 1.7 0.3
2 9.2785757312572 -0.3466589583034 1.7 1.7 0.3
3 751.9930037155181 -0.2830877173218 1.7 1.7 0.3
4****************** 46.8292921210168 1.7 1.7 0.3
5************************************ 1.7 1.7 0.3
6************************************ 1.7 1.7 0.3
7************************************ 1.7 1.7 0.3
8 NaN****************** 1.7 1.7 0.3
-----------------------------------------------------------------
Iterations converged
CCSD correlation energy / hartree = *************************
CCSD total energy / hartree = *************************

I have never seen this before, and am not sure why it is occuring specifcally with the cation.

But when I lower the diis steps to 3, it seems to converge to ~ -.354 but it doesn't actually ever converge.

MICROCYCLE DIIS UPDATE:                 2523                    3
2524 0.0001996156540 -0.3546342738788 1.7 1.7 0.3
2525 0.0129717480433 -0.3546419388683 1.7 1.7 0.3
2526 0.7737967986515 -0.3547375468700 1.7 1.7 0.3
MICROCYCLE DIIS UPDATE: 2526 3
2527 0.0001996156540 -0.3546342738788 1.7 1.7 0.3
2528 0.0129717480433 -0.3546419388683 1.7 1.7 0.3
2529 0.7737967986517 -0.3547375468700 1.7 1.7 0.3
MICROCYCLE DIIS UPDATE: 2529 3
2530 0.0001996156540 -0.3546342738788 1.7 1.7 0.3
2531 0.0129717480433 -0.3546419388683 1.7 1.7 0.3
2532 0.7737967986516 -0.3547375468700 1.7 1.7 0.3
MICROCYCLE DIIS UPDATE: 2532 3
2533 0.0001996156540 -0.3546342738788 1.7 1.7 0.3
2534 0.0129717480433 -0.3546419388683 1.7 1.7 0.3
2535 0.7737967986517 -0.3547375468700 1.7 1.7 0.3
MICROCYCLE DIIS UPDATE: 2535 3
2536 0.0001996156540 -0.3546342738788 1.7 1.7 0.3
2537 0.0129717480433 -0.3546419388683 1.7 1.7 0.3
2538 0.7737967986516 -0.3547375468700 1.7 1.7 0.3
MICROCYCLE DIIS UPDATE: 2538 3
2539 0.0001996156540 -0.3546342738788 1.7 1.7 0.3
2540 0.0129717480433 -0.3546419388683 1.7 1.7 0.3
2541 0.7737967986517 -0.3547375468700 1.7 1.7 0.3
MICROCYCLE DIIS UPDATE: 2541 3
2542 0.0001996156540 -0.3546342738788 1.7 1.7 0.3
2543 0.0129717480433 -0.3546419388683 1.7 1.7 0.3
2544 0.7737967986516 -0.3547375468700 1.7 1.7 0.3


And it just keeps running like this. Is there a simple fix to this?

Thanks.

Just Got Here
For those who face the same issue: I managed to cure that behavior by adding large levelshift.
Put "lshift 3" in the tce block. Now residuum decreases, but correlation energy also seem to decrease at each step.


Forum >> NWChem's corner >> Running NWChem