RI-MP2 PTED/COSMO SP computations

[acapobia]

I have optimized at RI-DFT/COSMO level a lot of structures [different conformers, 65 atoms system].

To assess and further validate my results, I considered to make Single Point RI-MP2/TZVP/COSMO [PTED approach] computations on a subset of the above structures. I used the mp2cosmo -ri script after adding the key $mp2energy in a new control file obtained from scratch (not the same of the previous DFT computation) and I noticed that mp2cosmo added the $cosmo_correlated key in my control file, which, if I understand the manual, relates to the PTE gradients, not needed to me at the moment.

Computation was very fast and appeared to go fine, I found the correlation energy in rimp2.log:


             *****************************************************
             *                                                   *
             *      SCF-energy   :   -1358.74399705800010        *
             *      MP2-energy   :      -4.84612067537233        *
             *      total        :   -1363.59011773337238        *
             *                                                   *
             *     (MP2-energy evaluated from T2 amplitudes)     *
             *                                                   *
             *****************************************************

         total  galoop  cpu-time : 22 minutes and 56 seconds
         total  galoop wall-time : 23 minutes and  4 seconds


    ------------------------------------------------------------------------
         total  cpu-time : 36 minutes and 57 seconds
         total wall-time : 37 minutes and 17 seconds
    ------------------------------------------------------------------------


   ****  rimp2 : all done  ****


    2010-05-02 07:46:35.882

Nevertheless the rimp2 step exited with an error, attached is mp2cosmo.out:


       Using: /home/programmi/TURBOMOLE6.1/bin/i786-pc-linux-gnu_mpi/dscf
       Using: /home/programmi/TURBOMOLE6.1/bin/i786-pc-linux-gnu_mpi/rimp2
       ccf/cosmo file: out(__mp2).ccf
       Max. cycles: 15
       Convergence crit:
          energy: 10**-6
          charges (RMS): 10**-6
          charges (max): 10**-5
   NOTE: $scfconv is automatically set to 7
cyc: 0
       Start: first full relaxed cosmo dscf
       Start: rimp2
\n     ERROR: CPHF not converged
         Please check rimp2.log \n

Now, since I'm not interested (at least at the moment) in MP2 gradients, is my computation reliable? Have I to modify the procedure and/or my control file in order to get RI-MP2/COSMO energies on RI-DFT/COSMO structures without errors?

Thank you all in advance.

Amedeo

[diedenho]

Hi Amedeo,

You should not set the $mp2cosmo keyword if you like to perform a PTED calculation.
In this iterative procedure the cosmo charges arising form the MP2 density have to be
calculated. In order to calculate the MP2 density, the CPHF equations have to be solved.
$mp2enegy quits after the energy calculation and skips the density calculation leading to the
error you got. Please don't abuse the mp2cosmo scripts, because it uses special keywords
for the frozen MP2 charges in the SCF (dscf). If you use it as proposed (mp2cosmo -h) it
arranges all keywords for the loop automatically and you can trust the results.
If you like to get PTE energies, just do a simple MP2 calculation on top of a dscf COSMO calculation.
Gradients are implemented for the PTE approach (as far as I know, there is no consistent gradient for the iterative schemes)
In this case provide the input for the gas phase rimp2 optimization, switch on cosmo (with $cosmo or cosmoprep) and add
$cosmo_correlated to the control file. Now you can start the normal rimpw jobex job.

hope this helps,
greetings,
Michael

[acapobia]

Thank you very much for your detailed reply... however I never added the key $mp2cosmo into my control file. Did you mean $mp2energy ?

Anyway I really went confused reading manual; possibly a few words more in the COSMO chapter and in the section of the keywords related to COSMO in the control chapter (for the next release) may be of help. However I understand now, reading manual again, after your clear explanation.

Amedeo