CCL:G: "Low frequencies" in Gaussian Freq Jobs

 Sent to CCL by: "Brian Salter-Duke" [brian.james.duke::gmail.com]
 The frequencies for translational motion are generally small. Those
 for rotational motion can be large for two reasons:-
 1. For all methods they will not be close to zero if the geometry is
 not at a stationary point. If in doubt about their values repeat
 optimization with opt=tight in Gaussian.
 2. In DFT in Gaussian they can be non-zero, particularly for heavy
 atoms (1st transition row and heavier) because the way the grid is set
 up means the calculation is not rotationally invariant. You can try a
 bigger grid but this lowers the largest rotational frequency only
 slowly and the disk and memory use increases fast. You probably should
 use a grid larger than default for transition metals but the largest
 rotational frequency will still not be really satisfactory
 It is general considered that the "zero" frequencies should be in the
 range -10 to +10, but some people insist on even closer to zero such
 as in the -5 to +5 range. The sign is not important.
 Of course the big problem is that a large negative rotational
 frequency confuses seeing whether there is an imaginary frequency and
 hence not at local minimum.
 Hope this helps, Brian.
 On Dec 19, 2007 10:40 PM, Andrea Ciccioli
 andrea.ciccioli+*+uniroma1.it <owner-chemistry^ccl.net> wrote:
 >
 > Sent to CCL by: "Andrea  Ciccioli"
 [andrea.ciccioli_-_uniroma1.it]
 > Dear friends,
 >
 > it is commonly asserted in textbooks and software manuals that the obvious
 test to recognize minima in PES among stationary points is that the vibrational
 frequencies have to be real.
 > However, I wonder if besides this criterium one should also check carefully
 the values of the frequencies reported as "Low frequencies" in
 Gaussian outputs (just before the list of Harmonic frequencies). These are the
 "frequencies" actually corresponding to translations and rotations,
 and they should be ideally equal to zero, and indeed in many cases they are very
 low. However, it happens not seldom to me, e.g. for triatomic species containing
 heavy elements such as transition metals, to obtain outputs where, although the
 harmonic vibrational frequencies are all real (positive numbers in the Gaussian
 output), ie the structure should be a minimum, nevertheless one or two "Low
 Frequencies" are not that low. Furthermore, they are in general both
 positive and negative. For example, low frequencies as high as  +/- 10 to 40
 cm-1 are obtained. Moreover, these values are apparently larger for analytic
 second-derivative frequency calculations than for numerical calculations (I use
 DF!
 >  T methods).
 > As far as you know, these relatively high values of the "Low
 Frequencies" can indicate that the calculated structure is not a true
 minimum, in spite of having real harmonic frequencies ? What could be a
 reasonable criterium to consider the "Low frequencies" small enough to
 be sure that the stationary point is a true minimum ?
 > Has anyone some suggestions/indications to give me ?
 > Thanks to all, and season's greetings.
 > Andrea Ciccioli
 > University of Rome (ITALY)
 > Sapienza>
 >
 >
 --
 Brian Salter-Duke (aka Brian Duke)
 Brian.James.Duke^gmail.com