Semisimplicity and Schur’s Lemma
A representation
of a finite group is said to be irreducible if there is no nontrivial vector subspace
of
which is left stable under the action of
.
If there is an inner-product lurking in the background, and
is a unitary representation, then for every
there exists a unique
such that
. If
is such that:

Then it follows that
is such that:

But since the representation is unitary,
, thus every unitary representation of
decomposes into an orthogonal direct sum of irreducible ones. This is the definition of semisimplicity.
Now fix some representation
. If
is irreducible, then the only maps
with the property that for all
the following diagram commutes:
![<br />
\tikzstyle{labelled}=[circle,inner sep = 2pt]<br />
\begin{tikzpicture}[line width=0.5pt]<br />
\node (1a1) at ( 1,0) [labelled] {$V$};<br />
\node (2a1) at ( 4,0) [labelled] {$V$};<br />
\node (1b1) at ( 1,3) [labelled] {$V$};<br />
\node (2b1) at ( 4,3) [labelled] {$V$};<br />
\draw [->](1b1) -- (1a1);<br />
\draw [->](2b1) -- (2a1);<br />
\draw [->](1b1) -- (2b1);<br />
\draw [->](1a1) -- (2a1);<br />
\node (h1) at (2.5,3.25) {$\rho(g)$};<br />
\node (h2) at (2.5,-0.25) {$\rho(g)$};<br />
\node (v1) at (.75,1.5) {$\psi$};<br />
\node (v2) at (4.25,1.5) {$\psi$};<br />
\end{tikzpicture}<br />](http://analogical-engine.com/wordpress/wp-content/cache/tex_cc7945a0a6d9dfa11e75dab2136e872c.png)
are the scalar multiples of the identity.
To see this not that if
is an eigenvalue of
then the eigenspace of
corresponding to
is stable under all of
and must thus be the whole space. Over
every linear operator has at least one eigenvalue.
More generally, suppose that
decomposes as:

with each of
and
irreducible. If the representation of
restricted to
is not isomorphic to the representation of
on
, then the the only
which commutes with the action of
are those of the form:
where
denotes the identity on
and
denotes the identity of
.
If they are isomorphic, then we can write:

where
is a multiplicity space of dimension
. In this basis the representation looks like:
where
is the restriction of
to
and
is the identity matrix on
.
The
which commute with the action of
are now of the form:

where
denotes the identity matrix acting on
, and
is any matrix acting on
.
Finally if
is an arbitrary representation, and the underlying vector space decomposes as

with the
denoting pairwise non-isomorphic irreducible representations, and the
their multiplicity spaces, then the representation looks like:
and the
which commute with the action of
are those of the form:
This is more or less Schur’s lemma.