HF problems with diffuse basis set

[mpjohans]

Hello All!

I'm having problems with a dscf Hartree-Fock run on biphenyl with Jensen's uncontracted aug-pc-4 basis set ([19s12p7d4f3g2h] for C and [12s7p4d3f2g] for H). Usually, the first iteration is OK, but after that the SCF goes crazy. I've tried to play around with $scftol and $scfdamp, to no avail.

Are there any other general tuning parameters one could try to get dscf well-behaved? The Hückel starting guess is as good or bad as usually, so that shouldn't be the cause of the problems. I'm running in parallel if it makes a difference. Below a typical failure after the first iteration; with some combinations of $scftol/$scfdamp, a few more iterations are sane.

Code Select Expand


ITERATION  ENERGY          1e-ENERGY        2e-ENERGY     NORM[dD(SAO)]  TOL
   1  -458.77078360653    -1794.5079179     735.01061090    0.000D+00 0.864D-12
rdiag: n=746, sequential Div&Con
rdiag: n=746, sequential Div&Con
          irrep b2  : virtual orbitals shifted by   18.31413
rdiag: n=716, sequential Div&Con
          max. resid. norm for Fia-block=  7.039D+00 for orbital      3a         
          max. resid. fock norm         =  2.224D+05 for orbital    343a         
          irrep b3  : virtual orbitals shifted by   11.94400
rdiag: n=716, sequential Div&Con
          Delta Eig. = **************** eV

                                              current damping :  0.650
ITERATION  ENERGY          1e-ENERGY        2e-ENERGY     NORM[dD(SAO)]  TOL
   2  -5901851365.7466    -1733.6085875    -5901850232.9    0.167D+13 0.109D-12


Any help with this would be appreciated. Smaller basis sets, like pc-4 or aug-pc-3 run smoothly.

Have a nice day,
    Mikael J.

[christof.haettig]

It seems strange that the HF code should not work for this. Did you check if the basis set is provided in the corrected format? If any "tunig" parameter might helpt then the integral threshold $scftol (see manual for details).

Christof

PS: If this doesn't work, you should consider to submit the problem to the helpdesk (COSMOLogic).

[mpjohans]

Thanks, I'll send it to the helpdesk. The basis set format is OK in my eyes. I've tried $scftol down to 1.d-17, without luck. As seen from the original message, it is the 2-el energy that goes crazy. If it's a problem with the evaluation itself or the orbital changes between the cycles I don't know. If the latter, perhaps some DIIS options could help, too?

I'm attaching an excerpt from another slightly more succesful run below. Before it flips out, the orbital shifts get ridiculous, and so:

Code Select Expand


                                              current damping : 48.750
ITERATION  ENERGY          1e-ENERGY        2e-ENERGY     NORM[dD(SAO)]  TOL
  10  -459.35806020972    -1796.2372671     737.09237064    0.179D+05 0.126D-15
          Norm of current diis error:  17.546   
rdiag: n=746, sequential Div&Con
rdiag: n=746, sequential Div&Con
rdiag: n=716, sequential Div&Con
          max. resid. norm for Fia-block=  2.385D-01 for orbital      5b1       
          max. resid. fock norm         =  4.461D+01 for orbital     10b2       
          irrep b3  : virtual orbitals shifted by   89.26820
rdiag: n=716, sequential Div&Con
mo-orthogonalization: Cholesky decomposition
          Delta Eig. = **************** eV

                                              current damping : 48.800
ITERATION  ENERGY          1e-ENERGY        2e-ENERGY     NORM[dD(SAO)]  TOL
  11  -137554455.98691    -1773.9096748    -137553281.86    0.179D+05 0.100D-14
          Norm of current diis error:  280.03   
          irrep a   : virtual orbitals shifted by  151.39047
rdiag: n=746, sequential Div&Con
          irrep b1  : virtual orbitals shifted by   74.36259
rdiag: n=746, sequential Div&Con
          irrep b2  : virtual orbitals shifted by  235.67530
rdiag: n=716, sequential Div&Con
          max. resid. norm for Fia-block=  2.856D+03 for orbital      8b3       
          max. resid. fock norm         =  6.026D+03 for orbital     10b2       
          irrep b3  : virtual orbitals shifted by 8300.82825
rdiag: n=716, sequential Div&Con
          Delta Eig. = **************** eV

                                              current damping : 48.850
ITERATION  ENERGY          1e-ENERGY        2e-ENERGY     NORM[dD(SAO)]  TOL
  12   877004783.57657    -1613.1501465     877005796.94    0.639D+12 0.942D-15



    MJ

[prasanta13]

I am also having the same problem with HF/def2-QZVPPD. I have used this on small molecules which ran just okay, but for a large system I am getting this error.

Code Select Expand

current damping :  0.550
ITERATION  ENERGY          1e-ENERGY        2e-ENERGY     NORM[dD(SAO)]  TOL
  11  -6794821.8026532    -3601.5454102    -6793057.5872    0.324D+07 0.254D-11
          Norm of current diis error:  11.187   
          max. resid. norm for Fia-block=  1.023D+04 for orbital      1a         
          max. resid. fock norm         =  1.266D+04 for orbital      1a         
          irrep a   : virtual orbitals shifted by    4.02856
          Delta Eig. = **************** eV

                                              current damping :  0.600
ITERATION  ENERGY          1e-ENERGY        2e-ENERGY     NORM[dD(SAO)]  TOL
  12  -6793807.2562664    -3629.0976562    -6792015.4885    0.873D+06 0.253D-11
          Norm of current diis error:  2.6859   
          max. resid. norm for Fia-block=  1.414D+04 for orbital      1a         
          max. resid. fock norm         =  3.445D+04 for orbital      1a         
          Delta Eig. = **************** eV

I am using Turbomole 7.5