Firefly and PC GAMESS-related discussion club



Learn how to ask questions correctly


Re: MP2 geo opt convergence with imaginary frequency problem (using ICUT=25-30, ITOL=30, INTTYP=HONDO, NCONV=10)

Alex Granovsky
gran@classic.chem.msu.su


Hi,

your starting geometry is almost symmetric. This approximate
symmetry is preserved for some time during optimization process,
so that optimizer may occasionally converges to "too symmetric"
critical point that is not the true minimum but rather a saddle
point.

To be more precise, molecular gradients are always transformed
according to the totally-symmetric representation of the molecular
point group. For symmetric starting geometry and optimization in
Cartesians, this means that the symmetry can only be destroyed
due to accumulations of round-off errors, and this typically
will need lots of geometry optimization steps so one has good
chance to converge to false minimum. The same is true for
internal coordinates that are symmetric. For DLCs, this can or
cannot be the case depending on the particular DLCs generated.

So the advice is - if you do not know anything specific on the
symmetry of the system of interest, use C1 group and start with
really not-too-symmetric structure. E.g., you can add some small
random displacements to all coordinates.  

Hope this helps,
Alex Granovsky


On Mon May 17 '10 8:10am, Gordon wrote
--------------------------------------
>Dear forum followers,

>I have been optimising geometries to calculate gas phase proton affinities with FF, using MP2 6-311+G(d,p) which will generally yield calc'd values within +/- 10 kJ/mol of experimental values. Now I am trying the same business using $BASIS NDFUNC=2 NFFUNC=1 NPFUNC=2 DIFFSP=.TRUE. DIFFS=.TRUE. $END. With this I can optimise the geometry, BUT, I also observe a single imaginary frequency. To try to fix this, I then used the same input and included $CONTRL RUNTYP=Optimize ICUT=30 ITOL=30 INTTYP=HONDO $END and $SCF DIRSCF=.TRUE. NCONV=10 FDIFF=.FALSE. $END.  (I have attached the two outputs for the acetate anion here).  I have also tried with ICUT=25, but I obtain the same result.

>I am not sure what I could do now to get further down the PE well!  Perhaps directed adjustment of these ICUT, ITOL, NCONV values are in order?  After 3 tries I was hoping to receive a suggestion from those who are more knowledgeable.  Perhaps there is another trick to try?

>I would be most grateful for any suggestions.

>Cheers,
>Gordon


[ Previous ] [ Next ] [ Index ]           Thu May 20 '10 5:37pm
[ Reply ] [ Edit ] [ Delete ]           This message read 973 times