Author Topic: TS optimization  (Read 9364 times)

iamxyj

  • Newbie
  • *
  • Posts: 6
  • Karma: +0/-0
TS optimization
« on: August 06, 2013, 03:33:05 pm »
I optimized TS by adding the following two lines in the control file.
$statpt
 itrvec 1

And submit the job using "jobexe -trans". the guess structure is from the surface. That is the maximum structure of the surface. But it turned out that the stationary point found is the minimum structure. How to fix this?

Thanks in advance. :)

Hauke

  • Full Member
  • ***
  • Posts: 37
  • Karma: +0/-0
Re: TS optimization
« Reply #1 on: August 06, 2013, 07:01:01 pm »
If I understand you correctly, you want to calculate the transition state of a (molecule on a) surface?
Did you calculate the frequencies after jobex-trans optimization or how do you know which is the "maximum/minimum" structure?

I'm not so much into surfaces but in principal your procedure looks alright for a TS search. To test it you could try the TS state search on acryloin following the example in the tutorial http://www.turbomole-gmbh.com/manuals/version_6_4/Turbomole_Tutorial_6-4.pdf (Section 7.6, there is no "on" needed to activate stp and the extra " dscf, grad and aoforce " step is also not necessary as jobex -trans automaticcally detectes that it needs a hessian to start and therefore calculates it itself. There you should be able to locate a TS.).

To find the transition state TM needs a good starting structure close too the TS.
« Last Edit: August 06, 2013, 07:10:02 pm by Hauke »

iamxyj

  • Newbie
  • *
  • Posts: 6
  • Karma: +0/-0
Re: TS optimization
« Reply #2 on: August 13, 2013, 10:23:54 am »
If I understand you correctly, you want to calculate the transition state of a (molecule on a) surface?
Did you calculate the frequencies after jobex-trans optimization or how do you know which is the "maximum/minimum" structure?

I'm not so much into surfaces but in principal your procedure looks alright for a TS search. To test it you could try the TS state search on acryloin following the example in the tutorial http://www.turbomole-gmbh.com/manuals/version_6_4/Turbomole_Tutorial_6-4.pdf (Section 7.6, there is no "on" needed to activate stp and the extra " dscf, grad and aoforce " step is also not necessary as jobex -trans automaticcally detectes that it needs a hessian to start and therefore calculates it itself. There you should be able to locate a TS.).

To find the transition state TM needs a good starting structure close too the TS.

Thank you for answering.
After jobex -trans optimization, I run aoforce and there was no image frequency. By good starting structure, do you mean that it should have image frequencies? And I optimized the TS on the excited state. Is there extra step should be taken?

Hauke

  • Full Member
  • ***
  • Posts: 37
  • Karma: +0/-0
Re: TS optimization
« Reply #3 on: August 14, 2013, 05:34:59 pm »

I meant by "good starting structure" that the initial structure should be closed to the expected TS (guess based on your chemical intuition, expectation, similar known TS...) because otherwise the optimizer will not find the expected TS.

I would expect that your initial structure has one/some imaginary modes (because how likely is it that you picked a minimum structure when you think it is not) but I'm not 100% sure if this is a mandatory condition.

I don't know if there are special considerations for a excited TS.

iamxyj

  • Newbie
  • *
  • Posts: 6
  • Karma: +0/-0
Re: TS optimization
« Reply #4 on: September 11, 2013, 01:19:29 pm »
My starting geometry has only one imaginary modes, which is reasonable chemically. But the optimization just did not converge. Can anyone give me some advice? 

uwe

  • Global Moderator
  • Sr. Member
  • *****
  • Posts: 418
  • Karma: +0/-0
Re: TS optimization
« Reply #5 on: September 12, 2013, 02:30:25 pm »
Hi,

did you look at the geometry during the optimization? Just read in the control file in the Turbomole GUI TmoleX or use any other viewer which allows visualization of multiple structures using xyz format (you get that if you run t2x) or molden (run tm2molden) format.

Often one can just see what happens and why.

Regards,

Uwe

yasmeen

  • Newbie
  • *
  • Posts: 7
  • Karma: +0/-0
Re: TS optimization
« Reply #6 on: September 09, 2014, 03:26:29 pm »
I have a similar problem while looking for TS in my system. I use statpt module with one parameter specified itrvec 1 (for my imaginary frequency). The initial geometry is close to the TS geometry so it should work.  But after optimization my system goes to the final products and no imaginary frequency is longer observed. It looks like the TS is missed during optimization.
I wonder if there are other parameters I can modify in order to catch the TS. I would appreciate any suggestion.

uwe

  • Global Moderator
  • Sr. Member
  • *****
  • Posts: 418
  • Karma: +0/-0
Re: TS optimization
« Reply #7 on: September 09, 2014, 03:44:45 pm »
Hi,

if you have the restart files from a former geometry optimization still in that directory, especially the hessapprox file, it is recommended to delete them before switching from geometry optimization to TS search. jobex -trans should remove the hessapprox file automatically, but it is safer to do it manually before starting a TS search.

Not sure, however, if that really was the problem in your case...

Regards,

Uwe

yasmeen

  • Newbie
  • *
  • Posts: 7
  • Karma: +0/-0
Re: TS optimization
« Reply #8 on: September 10, 2014, 11:14:18 pm »
Hi,

Thank you for the answer, but unfortunately the problem still exists.
Maybe there are other ideas - I hope.

Regards,
Agnieszka

Hauke

  • Full Member
  • ***
  • Posts: 37
  • Karma: +0/-0
Re: TS optimization
« Reply #9 on: September 10, 2014, 11:38:36 pm »

Just to be sure, you are using "jobex -trans", to optimise to the transition state?

To cite http://www.teokem.lu.se/~ulf/Methods/turbomole.html
Quote
For problematic ts cases, try:
$statpt
   itrvec       1
   hssfreq      1
   radmax      3.000000E-02
   tradius     3.000000E-02

You calculate the Hessian every step and have decreased the maximum trust radius by a factor of 10.

If you have a too big molecule for a full frequency calculations, you can use
$les all 1
$h0hessian
$nomw


The cited page is quite old so some information might be not longer up to date but at least the keywords still exist in the manual of Version 6.6 (there you can also find more information about their effects) therefore it might be worth giving it a try.
« Last Edit: September 10, 2014, 11:41:50 pm by Hauke »

bogac

  • Newbie
  • *
  • Posts: 5
  • Karma: +0/-0
Re: TS optimization
« Reply #10 on: March 09, 2015, 05:55:52 pm »
Hi everyone,

I am having a similar problem as well. My TS search, normally reaches to a very close step of TS structure but then, it drops to the other side of the PES unfortunately. Briefly it ends with an intermediate structure or something else irrelevant. In order to prevent that, I have changed the convergence criteria, I have set the gridsize to m5 and scfconv to 8.



I am using dft - b3lyp

$statpt
    itrvec 1

Right now I am also trying the hints in the post above. I will share the results if they help or not. And I guess my starting structures are alright.