Convergence issue - geometry optimisation of excited state


Just Got Here
I've been trying to model the Stokes shift of the first excited singlet of a small organic probe molecule in solvents.

The issue is that when modelling the molecule excited state in water, the geometry optimisation of the excited state is not well behaved. A minimum is shown after 2 steps (which is significantly higher than previous studies with same XC, solvent model and basis), and then the energy starts to oscillate and increase. How can I improve convergence of this calculation?

In vacuum, and for apolar COSMO-SMD models, the geometry optimisation has been well behaved. Convergence is achieved in around 8 optimisation steps.

I've tried multiple approaches. I've attempted with both COSMO and COSMO SMD. I've made the trust radius larger and smaller. I've used alternate starting geometries and restarts (such as the relaxed geometry of vacuum excited state). I've also checked for implemented patches in NWchem.

My initial guess is that the forces act to optimise the molecule and that when the new strongly interacting COSMO cavity is constructed it is energetically dis-favourable.

Here's an example input:

geometry units angstroms
---
end

basis
---
end

dft
 XC B3LYP
 iterations 50
 decomp
end

TDDFT
 RPA
 NROOTS 10
 PRINT HIGH
 CIVECS
 NOTRIPLET
 TARGET 1
 TARGETSYM a
 GRAD
  ROOT 1
 END
END

DRIVER
MAXITER 50
XYZ MOL-B3LYP-6-311+gdp
END

COSMO
 do_cosmo_smd true
 solvent water
END

task TDDFT optimize



The corresponding example output :

@ Step       Energy      Delta E   Gmax     Grms     Xrms     Xmax   Walltime
@ ---- ---------------- -------- -------- -------- -------- -------- --------
@    0    -711.93801005  0.0D+00  0.03302  0.00519  0.00000  0.00000   3582.9
@    1    -711.94026873 -2.3D-03  0.01163  0.00227  0.05829  0.18416   8252.7
@    2    -711.94019608  7.3D-05  0.00748  0.00145  0.01870  0.04729  12887.2
@    3    -711.94013249  6.4D-05  0.00606  0.00118  0.00770  0.02076  17597.9
@    4    -711.94008912  4.3D-05  0.00554  0.00107  0.00365  0.01090  22224.8
@    5    -711.94006818  2.1D-05  0.00524  0.00102  0.00245  0.00777  26823.9
@    6    -711.94003572  3.2D-05  0.00499  0.00097  0.00285  0.00931  31527.7
@    7    -711.94001458  2.1D-05  0.00473  0.00092  0.00281  0.00927  36081.6
@    8    -711.93997759  3.7D-05  0.00450  0.00088  0.00336  0.01119  40650.6
@    9    -711.93995431  2.3D-05  0.00431  0.00084  0.00308  0.01022  45185.9
@   10    -711.93991563  3.9D-05  0.00408  0.00079  0.00365  0.01222  49788.4
@   11    -711.93989370  2.2D-05  0.00391  0.00076  0.00296  0.00989  54345.7
@   12    -711.93985394  4.0D-05  0.00370  0.00072  0.00411  0.01369  58954.7
@   13    -711.93983413  2.0D-05  0.00357  0.00070  0.00274  0.00915  63516.5
@   14    -711.93978598  4.8D-05  0.00336  0.00065  0.00470  0.01559  68200.9
@   15    -711.93976709  1.9D-05  0.00325  0.00063  0.00258  0.00863  72740.6
@   16    -711.93971708  5.0D-05  0.00307  0.00060  0.00471  0.01568  77396.3
@   17    -711.93970510  1.2D-05  0.00297  0.00058  0.00242  0.00804  82011.4


Thank you for any help with this issue,

Duncan


Forum >> NWChem's corner >> Running NWChem