Author Topic: Triplet state optimization  (Read 9846 times)

Beverly

  • Newbie
  • *
  • Posts: 8
  • Karma: +0/-0
Triplet state optimization
« on: September 26, 2008, 11:24:03 PM »
Hi everyone
  I am new to turbomole and have been using the program to optimize some excited states. I have a relatively small molecule (31 atoms) and I am having a struggle to optimize the T3 state.
  I have been using SV(P) basis set, RI/DFT cist method and my structure has converged to a transition state with frequency -17 cm-1.
  I have used scfconv = 9, gcart4 and reduced dft grid size but nothing seems to work.
  Any advice would be greatly appreciated.
 Yours
  Beverly

uwe

  • Global Moderator
  • Sr. Member
  • *****
  • Posts: 487
  • Karma: +0/-0
Re: Triplet state optimization
« Reply #1 on: September 29, 2008, 01:45:36 PM »
Hi,

if you are new to Turbomole, I would like to ask some questions, just to be sure that you did what you intended to do - this is sometimes a little bit tricky...

  • you have been using DFT together with CIS, i.e. you have performed a Tamm-Dancoff Approximation calculation, right?
  • how did you calculate the frequencies? Did you use NumFroce?
  • what does 'reduced DFT grid size mean' - what is the gridsize in your $dft section?

Regards,

Uwe

Beverly

  • Newbie
  • *
  • Posts: 8
  • Karma: +0/-0
Re: Triplet state optimization
« Reply #2 on: September 29, 2008, 02:40:51 PM »
Hi there,
  Thanks for your reply.
  I have been using RI/DFT with cist
  I have been using NumForce to analyse frequencies
  My grid size for DFT is M3.
 
The method I have been using, described above, has proven successful for for the S1, S2, S3, T1 and T2 states so far.

Kind regards
Beverly

Arnim

  • Developers
  • Sr. Member
  • *
  • Posts: 238
  • Karma: +0/-0
Re: Triplet state optimization
« Reply #3 on: September 29, 2008, 05:04:10 PM »
Hi!

Are you sure?
escf/egrad actually quits, if "$scfinstab cist" and "$dft" aref found in the control file.

Cheers,
Arnim


Beverly

  • Newbie
  • *
  • Posts: 8
  • Karma: +0/-0
Re: Triplet state optimization
« Reply #4 on: September 30, 2008, 12:09:19 AM »
Hi Arnim
  I definately have $dft and $scfinstab in my control file, in define I have switched on dft and ri and each calculation has appeared to be successful, reporting the egrad has ended normally and reported reasonable results.
  Like I say I am very new to the program so am wondering if I should be using TDDFT instead if I am to call dft and ri.

Thanks for your reply
Beverly

Arnim

  • Developers
  • Sr. Member
  • *
  • Posts: 238
  • Karma: +0/-0
Re: Triplet state optimization
« Reply #5 on: October 01, 2008, 11:41:06 AM »
Hi,

sorry, I was wrong. I mixed up something. It runs with these settings.
Back to your original question... You could try to set the different convergence
thresholds higher. scfconv and gridsize in the control file, and use tighter
criteria for jobex, like jobex -gcart 5.

Best,
Arnim

Beverly

  • Newbie
  • *
  • Posts: 8
  • Karma: +0/-0
Re: Triplet state optimization
« Reply #6 on: October 01, 2008, 07:05:07 PM »
Hi Arnim
  i must admit I am having more success with TDDFT, however I was wondering if you could tell me if there are maximum values that can be selected for gcart and scfconv?

Kind wishes
Beverly

Arnim

  • Developers
  • Sr. Member
  • *
  • Posts: 238
  • Karma: +0/-0
Re: Triplet state optimization
« Reply #7 on: October 02, 2008, 01:56:06 PM »

The values are the decimal places, for scfconv in Hartree and gcart is the cartestian gradient norm.
I guess, there is no limit, but it is not useful to set them higher then the accuracy of your computer.
Just increasing the default alue by one or two should be enough in your case.