TDDFT excited state optimization fails

[basklau]

Hi,

whenever I try to optimize an excited state geometry with TDDFT and a hybid functional, I get after some egrad runs the following message:
vfile_a'   from previous  iteration not found
error in gradient step (1).

When ever I run the optimization with a non hybrid functional is converges.

Turbomole version 5.7

Is there any solution to this problem?

basklau

[doguapo]

Basklau,

From what I understand, the RI approximation is difficult to implement with hybrid functionals, but is possible.  I've been doing literature searches on Stefan Grimme - he has at least one paper where he mentions using rijk with b3-lyp, but i'm still not sure on how to do it.

also, from what i understand of Turbomole (which is still pretty limited), ridft is the correct command to use for TDDFT calcs, which is why i'm bringing up this whole RI stuff...and my guess is that you're getting this error message as a result of RI confusion.

[JakubV]

Hello,

ridft is the correct command to use for TDDFT calcs

so, when I want to execute RI-TDDFT geometry optimization, I should use

Code: [Select]

ridft > ridft.out
rdgrad > grad.out
jobex -ex -ri ...
NumForce -ex 1 -ri ... > force.out
, please ?

Best regards,
JakubV