<p>The rearrangements
of ethynamine <b>3</b> (H-CºC-NH<sub>2</sub>) to
ketenimine <b>4</b> (CH<sub>2</sub>=C=NH)
and acetonitrile <b>5</b> (CH<sub>3</sub>CN)
were investigated computationally up to the MP4(SDTQ)/6-31G*//MP2(FU)/6-31G*
level. The calculated barrier for a concerted reaction <b>3</b> -> <b>4</b> is very high, 74
kcal/mol, the structure of the transition state very unusual, and this path is
discredited. A lower barrier of about 60 kcal/mol via aminovinylidene <b>2</b> and imidoylcarbene <b>15</b> has been found. The calculated barrier
for a concerted second step <b>4 </b>-><b> 5</b> is 61 kcal/mol, and the transition state structure is again very
unusual with a virtually linear CCN backbone, but this does not appear to correspond
to physical reality. Instead, CASPT2 calculations predict reaction via vinylnitrene
<b>9</b> and/or homolysis of <b>4 </b>to the radical pair ·CH<sub>2</sub>CN + H· (<b>11</b>)
with a barrier of 67-70 kcal/mol in agreement with experimental shock-tube data.
Recombination (maybe via roaming) affords acetonitrile <b>5</b>. There is strong experimental evidence for homolytic paths in
pas-phase pyrolyses of ketenimines.</p>