TURBOMOLE Users Forum

TURBOMOLE Modules => Mpgrad and Rimp2 => Topic started by: dlchen on October 19, 2009, 06:40:10 PM

Title: rimp2 optimization of O2
Post by: dlchen on October 19, 2009, 06:40:10 PM

Hi TURBOMOLE users,

I am running RIMP2 optimization calculation with cc2 model for O2 molecule. Open shell is adopted in the calculations.

After the dscf calculation, I performed the rimp2 calculation. However, the job ended with failure of convergence of CPHF.

Here is the error information:

---------------------------------------------------------------------------------------------------------------------------
          -------------------------------------------------
          -CPHF- contribution to MP2-Gradient via Z-vector
          -------------------------------------------------

   Requested convergence is     0.1000000000E-10  RMS
     and a residual norm of     0.1000000000E-03

      Iteration           RMS                   |R|

           1        0.3325913056E-04      0.1693203194E+00
           2        0.2140491776E-03      0.4295467275E+00
           3        0.1929694786E-02      0.1289727454E+01
           4        0.3796007945E-02      0.1808910957E+01
           5        0.1988337893E-02      0.1309178087E+01
           6        0.1946791192E-02      0.1295428117E+01
           7        0.2855523296E-02      0.1568904421E+01
           8        0.3639925464E-03      0.5601442448E+00
           9        0.3152515326E-04      0.1648474510E+00
          10        0.1742962203E-05      0.3876123604E-01
          11        0.5429684760E-05      0.6841336319E-01
          12        0.4536925095E-05      0.6253662472E-01
          13        0.8510575392E-07      0.8565112952E-02
          14        0.9766998823E-09      0.9175594251E-03
          15        0.2129574465E-08      0.1354877555E-02
          16        0.6579912303E-09      0.7531191410E-03
          17        0.2608865181E-08      0.1499613879E-02
          18        0.4477015640E-08      0.1964481479E-02
          19        0.6948120522E-09      0.7739043798E-03
          20        0.1847956619E-08      0.1262116716E-02
  the iterative expansion of z(ai) did not converge, abort !
  the convergence is :     0.1847956619E-08
error in gradient step (1)
-------------------------------------------------------------------------------------------------------------------------------

It does not converge in 20 iterations (in default).

1. Any suggestion on this kind of systems, i.e., open shell?
2. How to change the iteration numbers and the default convergence (0.1E-10 RMS) of CPHF? Will these changes help to the rimp2 optimization?

Thanks very much in advance! Any suggestions will be appreciated.

Best,

De-Li

Title: Re: rimp2 optimization of O2
Post by: dlchen on October 23, 2009, 09:07:18 PM

I tested another open shell system and the convergence is even worse, RMS increased gradually which should decrease.

--------------------------------------------------------------------------------------------------------------------------------
          -------------------------------------------------
          -CPHF- contribution to MP2-Gradient via Z-vector
          -------------------------------------------------

   Requested convergence is     0.1000000000E-10  RMS
     and a residual norm of     0.1000000000E-02

      Iteration           RMS                   |R|

           1        0.7357878135E-05      0.9869847607E+00
           2        0.1396933450E-04      0.1359947084E+01
           3        0.4868716966E-04      0.2538875566E+01
           4        0.6636556116E-03      0.9373581015E+01
           5        0.1569437979E-02      0.1441472066E+02
           6        0.2340346543E-02      0.1760249528E+02
           7        0.2077948490E-01      0.5245073045E+02
           8        0.6118689453E-01      0.9000432053E+02
           9        0.1518558537E+00      0.1417914098E+03
          10        0.1031661649E+01      0.3695751781E+03
          11        0.2080937145E+02      0.1659830089E+04
          12        0.5735030215E+02      0.2755510099E+04
          13        0.8516754566E+03      0.1061869674E+05
          14        0.5800860960E+04      0.2771279823E+05
          15        0.6775700081E+04      0.2995099392E+05
          16        0.2877669243E+05      0.6172407486E+05
          17        0.2389930437E+06      0.1778798612E+06
          18        0.4146561827E+06      0.2343032024E+06
          19        0.1503260686E+07      0.4461195975E+06
          20        0.4658843172E+07      0.7853679920E+06
  the iterative expansion of z(ai) did not converge, abort !
  the convergence is :     0.4658843172E+07
error in gradient step (1)
---------------------------------------------------------------------------------------------------------------------------------

Any idea and suggestions to rimp2 calculation for open shell system will be greatly appreciated.

Thanks,

De-Li

Title: Re: rimp2 optimization of O2
Post by: christof.haettig on November 24, 2009, 12:19:57 PM

Hi,

first of all, I recommend that you update you TURBOMOLE version. It seems that you use a pretty old version.

Concerning your convergence problems:
   * in particular for open-shell cases I would recommend to optimize the geometry first with DFT (e.g. B3-LYP or PBE0)
      and start the MP2 geometry optimization from the DFT geometry.
  *  more recent TURBOMOLE versions use a solver which is more stable and allows adapt the number of CPHF iterations.

Christof