Problems with dscf when changing the basis set after geometry optimizations

[mapabst]

Hi,

I have the following problem with the dscf module, but only for larger molecules:
1. parallel (HPMPI) geometry optimization at ADC(2)/cc-pVDZ level works fine.
2. cpc to a new directory and changing the basis set via define to cc-pVTZ -> very crazy dscf, e.g. :

                                              current damping :  0.300
 ITERATION  ENERGY          1e-ENERGY        2e-ENERGY     RMS[dD(SAO)]   TOL
   1  -6186.3546927026    -10870.374305     238.12764096    0.000D+00 0.177D-11
          irrep ag  : virtual orbitals shifted by15270.55879
          irrep b1g : virtual orbitals shifted by 2674.11954
          irrep b2g : virtual orbitals shifted by 1734.98477
          irrep b3g : virtual orbitals shifted by15092.70862
          irrep au  : virtual orbitals shifted by 1463.05069
          irrep b1u : virtual orbitals shifted by21857.55067
          irrep b2u : virtual orbitals shifted by22737.20988
          max. resid. norm for Fia-block=  6.329D+02 for orbital      8b3g
          max. resid. fock norm         =  5.804D+04 for orbital    216b3g
          irrep b3u : virtual orbitals shifted by 1482.33860
          Delta Eig. = **************** eV
                                             
                                              current damping :  0.250
 ITERATION  ENERGY          1e-ENERGY        2e-ENERGY     RMS[dD(SAO)]   TOL
   2  -464624177.94668    -10822.260697    -464617801.58    0.913D+03 0.224D-12
          Norm of current diis error: 0.20723E+06   
          irrep ag  : virtual orbitals shifted by    0.20380
          max. resid. norm for Fia-block=  3.781D+02 for orbital      1b3g
          max. resid. fock norm         =  3.781D+02 for orbital      1b3g
          Delta Eig. = **************** eV
         
                                              current damping :  0.300
 ITERATION  ENERGY          1e-ENERGY        2e-ENERGY     RMS[dD(SAO)]   TOL
   3  -456864604.24760    -10804.411145    -456858245.73    0.142D+02 0.231D-12
          Norm of current diis error: 0.38609E+06
          irrep ag  : virtual orbitals shifted by  302.34931
          irrep au  : virtual orbitals shifted by    0.04962
          irrep b1u : virtual orbitals shifted by  280.85335
          max. resid. norm for Fia-block=  3.933D+02 for orbital      1b3g
          max. resid. fock norm         =  3.933D+02 for orbital      1b3g
          Delta Eig. = **************** eV
         
                                              current damping :  0.350
 ITERATION  ENERGY          1e-ENERGY        2e-ENERGY     RMS[dD(SAO)]   TOL
   4  -354188476.38160    -10802.568942    -354182119.70    0.507D+02 0.384D-12
          Norm of current diis error: 0.33262E+07
          irrep ag  : virtual orbitals shifted by   87.53253
          irrep b1g : virtual orbitals shifted by    0.06620
          irrep b2g : virtual orbitals shifted by    0.06975
          irrep b3g : virtual orbitals shifted by  627.95245
          irrep au  : virtual orbitals shifted by    0.07348
          irrep b1u : virtual orbitals shifted by  616.01609
          irrep b2u : virtual orbitals shifted by  625.75646
          max. resid. norm for Fia-block=  4.238D+02 for orbital      1b3g
          max. resid. fock norm         =  4.238D+02 for orbital      1b3g
          irrep b3u : virtual orbitals shifted by    0.07110
          Delta Eig. = 64414.5996942369 eV
          ....


This continues until maxiter is reached.

3. Some hints I tested so far:
a) A geometry optimization set up from scratch with the resulting geometry of the cc-pVDZ calculation works fine.
b) The two eht-mos for cc-pVTZ of 2. and 3.a) are different. Theire eigenvales differ in the third digit (maybe this is due to the accuracy of the guess and is not important?).
c) the optimizations are performed parallel.

Unfortunately this is not reproducable with smaller examples but happens quite often for larger molecules (10 carbon atoms+).

Has anyone an explenation for this issue, or even a solution?
Best regards, Mathias

[uwe]

Hi,

did you use the converged orbitals from the previous run with the 'use' option in define in the case where it did not work?

And did you remove all restart files after changing the basis set?

Regards,

Uwe

[mapabst]

Hi Uwe,

I did not use any restart files (new folder with only the input files).

The starting orbitals were generated by EHT. I also tried the 'use' option in order to curcumvent this problem but 'use' did not work - similar to one problem discussed in this forum (topic: "about single point calculation with larger basis set from smaller BS opt"). An error occurs like "mo 144 not found" and several others, but opposed to the other topic discussed earlier a second try did not work, too.

Regards, Mathias

[christof.haettig]

Dear Mathias,

if you kept the control file of the calculation that failed, please check if it contains any file names for scratch files with an absolute path...
A problem with the recycling or copy/paste generation of  the control file is (in particular for parallel calculations) that if one accidentally
starts two calculations with the same absolute file names for scratch files (e.g. twoint, or files for difference densities etc.) one gets
strange errors which are difficult to reproduce...

Christof

[mapabst]

Dear Christof,

I kept the control-file, but it does not contail any absolute paths, only the usual relative paths like "file=mos", "file=twoint" etc.
The error is, by the way, reproduceable in some way: for every calculation it occurs, it will happen every time I try to run the calculation...

If you have any idea what I should try to get another hint, please let me know.

Mathias

[uwe]

Hi,

please send the control file to the Turbomole Support - that is probably the fastest way to solve your problem.

Regards,

Uwe