This is similar to what Serre wrote in his book on linear representations of finite groups by Serre:

There is also the tensor product which has the properties of "multiplication". Let $V_1$ and $V_2$ be two vector spaces. The tensor product $\otimes: V_1 \times V_2 \rightarrow V_1 \otimes V_2$ constructs a new vector space $V_1 \otimes V_2$ with the map $(\xi_1,\xi_2) \rightarrow \xi_1 \cdot \xi_2$ which can be defined as the set of formal linear combinations $\xi_1 \otimes \xi_2$ subject to the conditions:

$(i)$ $\xi_1 \cdot \xi_2$ is linear in both variables $\xi_1$ and $\xi_2$.

$(ii)$ If $\lbrace e_{i_1} \rbrace$ and $\lbrace e_{i_2} \rbrace$ is a basis for $V_1$ and $V_2$ respectively then $e_{i_1} \otimes e_{i_2}$ is a basis for $V_1 \otimes V_2$.

We can show that this space exits and is unique. By condition $(ii)$ we have: $$ \dim(V_1 \otimes V_2) = \dim(V_1)\cdot \dim(V_2) $$

Now let $\rho^1\colon G \rightarrow GL(V_1)$ and $\rho^2\colon G \rightarrow GL(V_2)$ be two linear representations of a group $G$. For $s \in G$, define an element $\rho_s$ of $GL(V_1 \otimes V_2)$ by the condition: $$ \rho_s(\xi_1 \cdot \xi_2) = \rho_s^1(\xi_1)\cdot \rho_s^2(\xi_2) \quad \text{for } \xi_1 \in V,\; \xi_2 \in V_2. $$

We write: $$ \rho_s = \rho_s^1 \otimes \rho_s^2 $$

The $\rho_s$ above defines a linear representation of $G$ in $V_1 \otimes V_2$ which is called the tensor product of the given representations.

Defining the above using matrix notation:

let $\lbrace e_{i_1} \rbrace$ be a basis for $V_1$ and let $r_{i_1j_1}(s)$ be the matrix of $\rho_s^1$ with respect to this basis, define $\lbrace e_{i_2} \rbrace$ and $r_{i_2j_2}(s)$ in a similar manner. Now we have: $$ \rho_s^1(e_{i_1}) = \sum_{i_1} r_{i_1j_1}(s) \cdot e_{i_1}, \qquad \rho_s^2(e_{i_2}) = \sum_{i_2} r_{i_2j_2}(s) \cdot e_{i_2} $$

which implies: $$ \rho_s(e_{i_1}\cdot e_{i_2}) = \sum_{i_1,i_2} r_{i_1j_1}(s) \cdot r_{i_2j_2}(s)\cdot e_{i_1} \cdot e_{i_2} $$

Consequently the matrix of $\rho_s$ is $(r_{i_1j_1}(s) \cdot r_{i_2j_2}(s))$ which is the tensor product of the matrices $\rho_s^1$ and $\rho_s^2$.

We must add, the tensor product of two irreducible representations is not in general irreducible. It decomposes into a direct sum of irreducible representations which can be determined by means of character theory, which we shall discuss in the next chapter.

My question therefore is, can someone construct a concrete example to make me understand this

  • 1
    $\begingroup$ Have you seen tensor products of vector spaces (without a group acting) before? $\endgroup$ – Nate Apr 17 '19 at 13:41

Here's a rather simple but still informative example.

Let $\rho: \langle g: g^4=e \rangle \to GL(\mathbb{R}^2)$ be the representation of the cyclic group with four elements (i.e. $\mathbb{Z}/ 4\mathbb{Z}$) into $\mathbb{R}^2$ such that $$ \rho_g = \begin{bmatrix} 0 & -1 \\ 1 &0 \end{bmatrix} = M $$ and $\rho_{g^k}=M^k$. Note that there are no non-trivial $g$-invariant subspaces (one can verify this by showing $M$ does not have 1 as an eigenvalue) and so $\rho$ is irreducible.

However, let's look at what happens when we tensor $\rho$ with itself to get a representation of $\mathbb{Z}/ 4\mathbb{Z}$ into $\mathbb{R}^2 \otimes\mathbb{R}^2$. Letting $\rho' = \rho \otimes \rho$ we get that for $v,w\in \mathbb{R}^2$: $$ \rho'_g(v\otimes w) = \rho(v) \otimes \rho(w) $$ So your intuition should be that $\rho'$ mimics $\rho$ on each "copy" of $\mathbb{R}^2$. This is made clearer by actually computing the matrix of $\rho'_g$.

As you wrote, if $\{e_1,e_2 \}$ is the standard basis of $\mathbb{R}^2$ then the set $\{e_1\otimes e_1, e_1\otimes e_2, e_2\otimes e_1,e_2\otimes e_2\}$ forms a basis of $\mathbb{R}^2 \otimes \mathbb{R}^2$. Observe: $$ \rho'_g(e_1\otimes e_1) = Me_1 \otimes Me_1 = e_2 \otimes e_2 $$ $$ \rho'_g(e_1\otimes e_2) = Me_1 \otimes Me_2 = e_2 \otimes -e_1 = -e_2 \otimes e_1 $$ $$ \rho'_g(e_2\otimes e_1) = Me_2 \otimes Me_1 = -e_1 \otimes e_2 $$ $$ \rho'_g(e_2\otimes e_2) = Me_2 \otimes Me_2 = -e_1 \otimes -e_1 = e_1 \otimes e_1 $$ So the matrix of $\rho'_g$ is: $$ \rho'_g = \begin{bmatrix} 0 & 0 & 0 & 1 \\ 0 & 0 & -1 & 0 \\ 0 & -1 & 0 & 0 \\ 1 & 0 & 0 & 0 \end{bmatrix}. $$ This actually shows us something mildly interesting. While the original $\rho$ was an irreducible representation our new $\rho'$ is not since the subspace of $\mathbb{R}^2\otimes \mathbb{R}^2$ spanned by $\{e_1\otimes e_1, e_2\otimes e_2 \}$ is a proper invariant subspace. This confirms the statement at the end of your post.

Remark: I'm not sure if you know this but $\mathbb{R}^2\otimes \mathbb{R}^2 \cong \mathbb{R}^4$ as both are four dimensionals vector spaces over $\mathbb{R}$. So there really is nothing particularly fancy going on here, but I tried to keep everything in terms of tensors to help you see how things would work in a more general case.

| cite | improve this answer | |
  • $\begingroup$ A representation is just a group homomorphism from a group into $GL(V)$ for some vector space $V$. So I wouldn't say that I "came up" with a matrix $\rho_g$, but rather that this is how I defined the representation to begin with. I guess you could say that I came up with the idea for such a representation by trying to find a $2\times 2$ matrix that satisfied $M^4 =I$, since that would be enough for me to define my homomorphism. $\endgroup$ – Edgar Jaramillo Rodriguez Apr 24 '19 at 20:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.