Author Topic: Aoforce on linear CO2 - No degeneracy?  (Read 4272 times)


  • Full Member
  • ***
  • Posts: 37
  • Karma: +0/-0
Aoforce on linear CO2 - No degeneracy?
« on: July 04, 2013, 07:07:37 PM »

a colleague of mine did a calculation of CO2 and "for fun" started with a bend structure and just C1 symmetry. As expected the structure became linear after optimization but was of course not aligned on one of the coordinate axes. After this he did an aoforce run and got 4 modes with nonzero frequencies (3*3-5 =4, for me this is a clear sign that Turbomole also accepted the structure as linear). So far so good but the 4 modes had each totally different frequencies but one would expect the two bending modes to be degenerate. The expected result can be obtained by using NumForce (-central) or aligning the molecule along one axis

This behavior was also observed when using all the TM defaults (C1 symmetry, def-SV(P) basis set, HF, coord file
Code: [Select]
    0.00001227374264     -0.00000003075555      0.00000000000000  c
   -2.14857205963009      0.00000001537753      0.00000000000000  o
    2.14855978588746      0.00000001537802      0.00000000000000  o
$user-defined bonds
) where I got the following non-zero modes

aoforce 757.981054.691539.82 2630.16
grad, NumForce754.61763.471541.232631.03
grad, NumForce -central 760.82760.821540.30 2631.78
aoforce with molecule aligned on x axis758.04 758.04 1539.73 2629.94

Can you reproduce the results and maybe also give an explanation?
We are still using TM Version 6.3.1 (SMP parallel)
« Last Edit: July 04, 2013, 07:31:36 PM by Hauke »


  • Global Moderator
  • Sr. Member
  • *****
  • Posts: 494
  • Karma: +0/-0
Re: Aoforce on linear CO2 - No degeneracy?
« Reply #1 on: July 16, 2013, 11:54:18 AM »

the default in aoforce is to project out the translational and rotational modes, but it has to check whether the molecule is linear or not to take care of either two or three modes of rotation.

If your coordinates are almost linear, there is a range within which aoforce considers the input as non-linear and using all three moments of inertia for the projection. The numerical errors will be quite large because the third moment is almost zero.

So aoforce should be a bit less sloppy when checking for linear molecules in such cases.

The easiest solution is to use symmetry here (desy 0.1 in define) because of three reasons:
  • without symmetry usage of internal redundant coordinates for linear molecules is not possible - although for small molecules Cartesian optimization is not a problem...
  • symmetrization of the coordinates will avoid the problems with 'almost' linear structures
  • if you use thermodynamic data (with the module freeh), the correct symmetry number (sigma) for the linear molecule will be automatically set by freeh



P.S.: aoforce will be a bit more cautious in such cases in the next Turbomole release