Author Topic: rij in rimp2 calculations  (Read 7714 times)

Addiw7

  • Jr. Member
  • **
  • Posts: 10
  • Karma: +0/-0
rij in rimp2 calculations
« on: February 13, 2013, 07:03:20 PM »
Hi,

When I try do to rimp2 geometry optimalization I get the error, that I should remove $rij from my control file. I do not have this problem while doing ricc2.

What is wrong than?



Hauke

  • Full Member
  • ***
  • Posts: 37
  • Karma: +0/-0
Re: rij in rimp2 calculations
« Reply #1 on: February 15, 2013, 11:31:19 PM »
For simple RI-MP2 you don't need the $rij keyword ($cbas is enough).

If you want to use "Coulomb- and Exchange-Fitting" (by using the rijk menu in define, setting the $jkbas and $rik keyword together with  $rij (One can also use  $rij without $rjk, see here ) you have to run jobex with the following parameters:
jobex -level cc2 -rijk
Even if you want to use mp2 and therefore have "geoopt model=mp2" is in the control file one has to use level -cc2, see jobex -help. I don't know why the combination of -level mp2 and -rijk is not considered in the jobex-script.

If this post of Christof Hättig is still up-to date it is also relevant to read.

Before you think about using rijk also on bigger jobs you should think about how to parallelise it:
  • In the SMP (GA) parallel version of TM 6.3.1 in combination with rijk the density is not beeing checked (I get "DENSITY CONVERGENCE CHECK UNAVAILABLE IN GA VERSION!" in the output ), and as always pointed out $denconv should be important for MP2 so I don't know how reliable the results are at the end
  • The mpi parallel version of TM 6.3.1 stops already at the beginning with the message "HF-exchange within ridft is not yet implemented in parallel mode".
Of course you can run it in serial but maybe it is faster to run it without rijk but several cpus.

If anybody knows a way to run parallel RI-SCF with check of density convergence, I would be really interested
« Last Edit: March 11, 2013, 03:04:45 PM by Hauke »