Complicated geomtry optimization with frozen internals and Cartesians

[evgeniy]

Hello,

I have a question concerning generating redundent internal coordinates,
needed for geometry optimization. In my case, I want to calculate
a minimum energy path along an internal coordinate. For this I need to do
several geometry optimization calculations, in which the value of the internal
coordinate will be varied, with each value being kept frozen during the optimization.
At the same time, I have two atoms that I want to freeze completely. I do
this by freezing their Cartesians. Now comes the question: I would like to
keep the values of these 6 frozen Cartesians the same at all points along the
minimum energy path. This seems to be not so trivial because, when
I change the value of the frozen internal coordiante (using define), define
also changes the vaules of the frozen Cartesians. I tried to replace the
new values of Cartesians by the old ones (i.e. those I want) and then run
the geometry optimization without repeating define. However, this doesn't work
in all cases. Sometimes when the initial geometry is to bizarre, due to the replacement,
the procedure which does convertion from internal to Cartesians does not converge.
Therefore I wonder if there is another, more appropriate way to keep the same
Cartesians while changing the value of an internal coordiante.

I would appreciate very much any idea/point to solve this difficulty.

Best wishes,
Evgeniy

[uwe]

Hi Evgeniy,

when you use internal coordinates in define and manipulate them, the program transforms in internal coordinates only. There is no way to take fixed cartesian coordinates into account. That would be possible in a post-processing step only, but this would also change the value of the internal coordinate - if they are somehow coupled (if they are not coupled there will be no problem anyway). Sounds like an iterative process to force both constraints, the internal and the cartesian ones.

I am not aware of any tool which is able to do that in a (even half-) automated way.

The question is how much does the distance between the fixed (in cartesian coordinates ) atoms change? And what kind of internal coordinate do you freeze? If you fix a bond length, this is going to be extremely hard to keep. If you fix an angle or a torsion, a simple idea would be to scale all coordinates such that the distance between the fixed cartesian atoms match your needs, because simple scaling (from the CMS as origin) should not change angles and torsions. After scaling, the fixed internal and cartesian coordinates can both be frozen and a usual optimization should work (although it might need a lot if cycles to converge).

Regards,

Uwe

[evgeniy]

Hi Uwe,

Many thanks for your reply.  Well, maybe it is overdoing to fix the Cratesians
precisely at the same values for all points, although I somehow managed to do it
before, but now do not remeber the whole procedure. In my case, the fixed internal
coordinates are two torsional angles and they are not coupled at all to the fixed Cartesians
(the Cartesians correspond to other, remote atoms). Of course the Cartesians change a little
when the values of the fixed internals vary, but the distance basically remains the same up to the
5th digit. So I think I will just use define without further manipulation of the fixed Cartesians.

Best regards,
Evgeniy