Suppose
is odd. Then the fibres of
are
semisimple but not simple:
. We shall
restrict to the even subalgebras,
,
by demanding that
act as the identity in all cases.
Then we may adopt the convention that
We take
, but for
we now take
We classify the algebras
as follows. Taking
Here is a (rather pedestrian) sketch of how
is
constructed:
If
, take
to be a projector of rank
one, that is,
By local triviality, this can be done locally with varying
. If
is a ``good'' open cover
of
, we get local fields
with isomorphisms
of fields of
simple
-algebras. On nonempty intersections
, we get
-algebra isomorphisms
,
so there are fields of unitary maps
such that
.
On
, we see that
, and so
Suppose now that the Hilbert spaces
can be
chosen globally for
--not just locally for
-- that is, they are fibres of a vector bundle
(that may b gifted with a Hermitian metric) such that
, for
, via a single field of
isomorphisms
such that
for each
. Then
over
,
and so
over
, and
over
each
; hence
in
.
Conversely if
, so that
is trivial in
, i.e.,
is a 2-coboundary, then
there are maps
such that
on
. Setting
,
we get local fields of unitaries such that
on each
. These
are therefore
transition functions for a (Hermitian) vector bundle
such that
for each
. Let
denote the
-module of sections of
this bundle. Now the pointwise isomorphisms
, for
each
, imply that
as
-modules, and
indeed as
-algebras. We summarize all this in the following
Proposition.