Interestingly, almost all the work reported has been computational rather than
experimental. In experimental work, the acyclic N 5 cation has been made fairly
recently [ 23 – 26 ]), and the pentaaza analogue of the cyclopentadienyl anion has
been detected by mass spectrometry [ 27 , 28 ]; its generation in solution was claimed
[ 29 ], challenged [ 30 ], and eventually “proved unequivocally” by examination of its
labelled decomposition products (dinitrogen and azide ion) in redesigned experi-
ments [ 31 ]. The N 5 cation is stable in the sense that salts of it can be isolated at room
temperature, but it explodes capriciously. The N 5 anion was unstable at" 40 C[ 31 ]
and was not isolated or even seen spectroscopically by^15 N NMR. These two
species, and azide ion, known since 1890 [ 32 ], are the only polynitrogens to have
been prepared. We use “prepared” advisedly for N 5 ", and pass over Nxcations that
have been observed only in mass spectra [ 33 ].
Perhaps the first serious computational study of nitrogen oligomers was by
Engelke, who studied the N 6 analogues of the benzene isomers in Fig.9.4, first at
the uncorrelated [ 34 ] then at the MP2 [ 35 ] level. The uncorrelated calculations
suggested that 1 – 5 were “stable”, i.e. kinetically stable, although thermodynami-
cally much higher in energy than dinitrogen. However, on the MP2/6-31G* poten-
tial energy surface 1 is a hilltop (Section 2.2) and 5 is a transition state (Section 2.2).
This illustrates the not-so-rare fact that optimistic predictions at low levels of theory
may not be sustained at higher levels. Noncorrelated ab initio, and in particular,
semiempirical (Chapter 6) calculations, tend to be too permissive in granting reality
to exotic molecules. Hundreds of calculations on polynitrogens have been pub-
lished; a representative survey of these can be found in [ 33 ].
9.1.2 Mechanisms......................................................
We have seen, above, that computational chemistry can sometimes tell us with
good reliability whether a molecule can exist. Another important application is to
indicate how one molecule gets to be another: how chemical reactions occur.
Indeed, the prime architect of one of the most useful computational tools, the
AM1 method (Chapter 6), questioned “whether the mechanism ofanyorganic
reaction was really known.” [ 36 ] before the advent of computational chemistry!
1 2345
benzene Dewar benzene benzvalene prismane bicyclopropenyl
Fig. 9.4 Nitrogen analogs (CH!N) of these molecules have been investigated computationally
566 9 Selected Literature Highlights, Books, Websites, Software and Hardware