CCL:G: Summary: large imaginary frequency in coronene when diffuse functions are used

 Sent to CCL by: Fedor []
 Dear CCL-ers,
A summary of what I found out thusfar about the imaginary frequency in coronene
 for certain approaches.
Cory Pye and Marcel Swart helped to find the stable structures corresponding to
 the saddle point. The stable structure is of D3D-symmetry, which is found by
 manually displacing the structure along (bits of) the B2G-imaginary mode.
 Gaussian's IRC apparently has sufficient numerical noise to misjudge the
 symmetry after stepping along the imaginary mode (C2H). The
 symmetry-determining algorithms are of course very sensitive to slight
 numerical differences in cartesian coordinates.
 The resulting optimised stable structure is 'ruffled' with alternative upward
 and downward displacement out of the plane by ~0.003A for the central carbon
atoms and less for the other atoms. It is very comparable in energy to the flat
 structure (~0.3 kJ/mol).
 Thanks to Errol Lewars, Stefan Grimme, Steven Bachrach and Paul Schleyer for
 pointing out the paper
 J. Am. Chem. Soc., 2006, 128(9), 9342
where imaginary frequencies are observed for different methodologies/basis sets in benzene. These imaginary frequencies are also reported for large Pople basis
 sets by Steve Williams. This basis-set dependent effect is very likely related
 to that in coronene and, as Petar Todorov pointed out, in other PAHs as well.
In this paper, this basis set effect is attributed to a an intramolecular basis
 set incompleteness error.
 To illustrate how sensitive the basis set effect might be: with MPWB1K the
 ruffled structure is preferred if a 6-31+G* basis is used as well as when a
 split 6-31+G*(central)/6-31G*(peripheral) basis is used. However, as Grzegorz
Mazur pointed out, at the HF/6-31+G* level the flat D6H structure is preferred,
 while for the split basis it is unstable and the ruffled D3D structure is
 preferred instead.
In both my and previous (JACS-paper cited) calculations the aromatics are planar
 when Dunning's basis (aug-cc-pVnZ) are used, but not always when Pople basis
 sets with diffuse function are used.
 I have not fully pinned down yet what the underlying reason for the basis set
 dependent nonplanarity is, although I do feel that Isaac Bersuker's suggestion
 of a basis-set dependent pseudo Jahn-Teller effect is very plausible.
 A question that comes to mind is if this tendency for puckering may be
 especially for the larger PAHs in view of the structure of graphene recently
 determined to be corrugated (Nature, 2007, 446, 60)?
 Thanks for all your help,
 Kind regards,