# Semi-Simple Lie Algebra Representations

1. Oct 29, 2007

### lion8172

I'm trying to prove that any representation of a semisimple Lie algebra can be uniquely decomposed into irreducible representations.

I have seen some sketches of proofs that show that any representation $$\phi$$ of a semisimple Lie algebra which acts on a finite-dimensional complex vector space $$V$$ is completely reducible (i.e. $$V = V_1 \oplus V_2 \oplus \cdots \oplus V_k$$, such that the restriction of $$\phi$$ to each $$V_i$$ is irreducible). But how do we know that this decomposition is unique?

Last edited: Oct 29, 2007
2. Oct 29, 2007

### matt grime

You show that in two decompositions

$$V\cong \oplus V_i \cong W_j$$

that one of the summands V_r is isomorphic to one W_s, wlog V_1 and W_1, then use induction on V/V_1.

3. Oct 29, 2007

### lion8172

question

So two representations $$\phi$$ and $$\phi '$$ for $$\mathfrak{g}$$ are said to be equivalent if there is an isomorphism $$E$$ between the underlying vector spaces such that $$E \phi (X) = \phi' (X) E$$, $$\forall X \in \mathfrak{g}$$. How do we know that the restriction of a representation to a given $$V_i$$ above is equivalent to a restriction of that representation to a given $$W_i$$?
Note that, by irreducible, I mean a vector space with no invariant subspaces under the given representation (except zero and itself).

Last edited: Oct 29, 2007
4. Oct 30, 2007

### matt grime

The identity map on V restricts to the identity map on V_1. This must then factor through one of the W_i, wlog W_1, hence V_1 is isomorphic to W_1.

5. Oct 31, 2007

### lion8172

Could you be a little more explicit?
Note that I'm trying to prove this for an arbitrary representation.

6. Oct 31, 2007

### matt grime

I know what you're trying to do; I was explicit; where do you see me do something not for an arbitrary representation? You want to show that V_1 is isomorphic to W_1 (after reordering the indices of the W_i). That is what I did.

1. You want to find an isomrphism from V_1 to W_1, that is maps f,g such that gf= Id on V_1 and fg=Id on W_1. That is the definition of isomorphism.

2. Consider Id on V. This maps \oplus V_i to \oplus W_j, and back again.

3. Look at V_1 mapping under Id to \oplus W_j.

4. The image must lie in at least one of the W_j. WLOG W_1.

5. So we have a map, Id restricted to V_1, call it f, that maps to W_1.

6. Call Id restricted to W_1 g.

7. What is gf? It is Id on V_1.

I really can't be any more explicit. In fact that is precisely what I wrote before.