Firefly and PC GAMESS-related discussion club

Learn how to ask questions correctly  
We are NATO-free zone

Re^3: How do I get a better initial guess (SnI3)?

Alex Granovsky

Dear David,

the basis set you are using, namely aug-cc-pVTZ-pp was specifically
designed assuming exclusive use of pure spherical basis functions.
Use of d5=.t. option will indeed affect energies and other properties
but this is the only correct way how aug-cc-pVTZ-pp basis set should
be used.

There are several good reasons for this. First, this basis set becomes
nearly linearly dependent if Cartesian basis functions were used.
This causes various numerical instabilities and convergence problems.

Second, some QC programs would explicitly not allow one to use this
basis set in conjunction with Cartesian basis functions, so that the
only way to obtain results reproducible across multiple QC packages
is to use this basis set with d5 option in Firefly turned on.

Hope this helps.

Kind regards,
Alex Granovsky

On Thu Jan 25 '18 9:20pm, David Shobe wrote

>Follow-up question: Will $contrl d5=.t. $end affect the energy or other properties that are calculated?  In particular, will I have to recalculate SnI4 and I (atom) with $contrl d5=.t. $end to get an accurate bond dissociation enthalpy>

>On Thu Jan 25 '18 9:42am, Alex Granovsky wrote

>>First, you need to add $contrl d5=.t. $end option to your input file.
>>Second, you need to use either $contrl inttyp=hondo $end or direct scf
>>with fastints code ($scf dirscf=.t. $end, $contrl fstint=.t. $end ).

>>Hope this helps.

>>Kind regards,
>>Alex Granovsky
>>On Wed Jan 24 '18 11:37pm, David Shobe wrote
>>>How can I improve on GUESS=HUCKEL when it fails to perform?

>>>At least I'm assuming the initial guess is at fault and not the choice of parameters EXTRAP DAMP SHIFT DIIS and SOSCF. This is because from beginning to end of the DFT orbital calculation, I am getting wildly different energies from one iteration to the next.

>>>I tried SnI3+ (which is a closed shell), but got the same result. So my question is: now what?

>>>I could try to swap occupied and unoccupied orbitals with NORDER, but I'm flying blind.

>>>If you're wondering: Yes, I do have diffuse functions. SnI3 converges fine when you omit them.  But I think I need the diffuse functions for accuracy.

>>>Regards, and thanks for any help,
>>>--David Shobe

[ Previous ] [ Next ] [ Index ]           Fri Jan 26 '18 0:34am
[ Reply ] [ Edit ] [ Delete ]           This message read 382 times