Author Topic: jobex vs jobex -level mp2 vs jobex --ri level -mp2  (Read 13593 times)

kadlubanski

  • Newbie
  • *
  • Posts: 3
  • Karma: +0/-0
jobex vs jobex -level mp2 vs jobex --ri level -mp2
« on: May 08, 2013, 01:10:10 PM »
Hello,
I am new Turbomole user and I have very strange problem. I tried to optimize the structure of some In complex (open shells system), by using RIMP2 method. When I use jobex, which should optimize the structure on HF level of theory there is no problem, but when I use command jobex –ri –level mp2 , which should perform the optimization on rimp2 level of theory the program finish with error after first set of scf iteration. (the energy file is empty). I probably do something wrong. Could someone help me ?
Paweł Kadłubański     

Arnim

  • Developers
  • Sr. Member
  • *
  • Posts: 253
  • Karma: +0/-0
Re: jobex vs jobex -level mp2 vs jobex --ri level -mp2
« Reply #1 on: May 08, 2013, 02:37:42 PM »
Hello,

the options of jobex trigger which programs will be used. Presently, there are 3 MP2 programs in Turbomole available. mpgrad, rimp2, and ricc2.
'jobex -level mp2' will run mpgrad,  'jobex -ri -level mp2' rimp2, and 'jobex -level cc2' ricc2. Your control file has to be valid for the program, you want to use.

What error do you get? That could give some clue.

Arnim

kadlubanski

  • Newbie
  • *
  • Posts: 3
  • Karma: +0/-0
Re: jobex vs jobex -level mp2 vs jobex --ri level -mp2
« Reply #2 on: May 08, 2013, 08:53:00 PM »
During firs SCF point I received such error:

===============================================================================================================
convergence criteria cannot be satisfied within 30 iterations !


   *************************************************************************
   TmoleXProject                                                           
   *************************************************************************


                  ------------------------------------------
                 |  total energy      =  -1846.06595210128  |
                  ------------------------------------------
                 :  kinetic energy    =   1728.52459888813  :
                 :  potential energy  =  -3574.59055098941  :
                 :  virial theorem    =      1.93632873567  :
                 :  wavefunction norm =      1.00000000000  :
                  ..........................................


 ATTENTION : $energy will not be updated since
 this calculation did not converge (sob)

 A POSSIBLE REASON MAY BE THAT ...
 - THE GEOMETRY IS FAR FROM BEING REASONABLE
 - THIS IS AN OPEN SHELL CALCULATION (GOSH)
===============================================================================================================
This error is present only when I used jobex –ri –level mp2. When I use jobex everything goes correctly and I obtain the optimized geometry in 17 steps, but it is not the matter of rimp2 or mpgrad, I think, because the error is in SCF. 
Paweł Kadłubański 

Hauke

  • Full Member
  • ***
  • Posts: 37
  • Karma: +0/-0
Re: jobex vs jobex -level mp2 vs jobex --ri level -mp2
« Reply #3 on: May 10, 2013, 02:07:08 PM »

How does the energy in the first 30 steps change? Is it converging or oscillating? Maybe 30 iterations is just not enough to get everything fully converged and you could give more scf-iterations a try by setting $scfiterlimit to a higher value?

antti_karttunen

  • Sr. Member
  • ****
  • Posts: 227
  • Karma: +1/-0
Re: jobex vs jobex -level mp2 vs jobex --ri level -mp2
« Reply #4 on: May 10, 2013, 07:55:15 PM »
Hi Pawel,

Hauke's tips are spot on. You might also be interested in the Forum FAQ, which has several tips for solving SCF convergence issues:
http://www.turbo-forum.com/index.php?topic=195.0
http://www.turbo-forum.com/index.php?topic=401.0

The difference between HF and MP2 optimization might be that the $denconv convergence criterion is a bit tighter for MP2 and therefore SCF might not converge as quickly.

Antti