Firefly and PC GAMESS-related discussion club

Learn how to ask questions correctly

Alex Granovsky
gran@classic.chem.msu.su

I think all is OK with your hardware.

There are several points here to be mentioned.

1. Indeed, you changed the geometry and enforced Cs symmetry so it is natural to expect some differences.

2. Maximum and RMS gradients that are used by Firefly
to check for convergence are in internal coordinates
(if they are used for optimization naturally), rather
than in Cartesians. As DLCs initially constructed at
different geometries are different, the value of gradient
in internals can vary slightly even at the same geometry
point just because of this difference in the definition
of internals.

3. If the initial geometry is symmetric but no explicit
symmetry constrains are given (i.e., you are using C1 grooup),
in theory the symmetry will be preserved during geometry
optimization. This is because gradient in Cartesians always
belongs to the completely-symmetric representation of the
molecular point group. However, this is not always true in
practice! More precisely, this is usually true for optimization
in Cartesians (although even in Cartesians the accumulation of
various round-off errors can break symmetry), and is not usually
true for optimization in internals. For internals, this actually
depends on the particular molecule and particular choice of z-matrix
variables, with some of them being symmetry preserving while others
are not. In particular, any DLCs will not be symmetry-preserving!
However, if the initial structure is symmetric, the drift from
symmetric to non-symmetric structure will be slow at initial
stages of geometry optimization.

Hope this helps to some degree...

Regards,
Alex









On Tue Dec 15 '09 7:09pm, Valentin Monev wrote
----------------------------------------------
>Dear Alex,
>In the course of a geometry optimization (attached input and output files),
>the following point 9 was passed

>MAXIMUM GRADIENT =  0.0001013    RMS GRADIENT = 0.0000270
>          HESSIAN UPDATED USING THE BFGS FORMULA
>             ACTUAL ENERGY CHANGE WAS   0.0000008052
>          PREDICTED ENERGY CHANGE WAS  -0.0000006082 RATIO= -1.324
>            GDIIS STEP HAS LENGTH         =   0.002556
>          RADIUS OF STEP TAKEN=   0.00256  CURRENT TRUST RADIUS=   0.06250

>but without EQUILIBRIUM GEOMETRY LOCATED, and proceeded 15 more points, when I stopped it. In the meantime, I had taken the geometry of point 7 (energy min) as input, and (on another pc) immediately EQUILIBRIUM GEOMETRY LOCATED (MAXIMUM GRADIENT =  0.0000552    RMS GRADIENT = 0.0000142)!
>Is something wrong with my pc? What is the best course of action in such circumstances - stop the job and restart from the energy minimum?

>Thank you in advance.

>Valentin

>Edit: Actually, point 7 was taken as input, but all z coordinates were made pure zero, i.e. all 0.00xxx were cut to 0.00 to make the starting geometry completely planar. This forgotten detail may make my question meaningless...

[ This message was edited on Fri Dec 25 '09 at 5:32pm by the author ]