I was hoping someone could help me with the following question.

Let $\rho$ be an irreducible presentation of a finite group $G.$ Prove
\begin{equation*} \sum_{g \in G} \rho(g) = 0 \end{equation*} unless $\rho$ is the trivial representation of degree $1$.

I think I have to use Schur's Lemma which states the following. Let $\rho: G \longrightarrow GL(n,\mathbb{C})$ be a representation of G. Then $\rho$ is irreducible if and only if every $n \times n$ matrix $A$ which satisfies \begin{equation*} \rho(g)A = A\rho(g) \ \ \ \forall \ g \in G
\end{equation*} has the form $A = \lambda I_n \, $ with $\lambda \in \mathbb{C}$.

But I am really not sure how the lemma can be applied to this question?


Let $t=\sum_{g\in G}\rho(g)$, which is an linear endomorphism of $V$. The subset $t(V)$ of $V$ is a $G$-submodule of $V$, as you can easily check. Moreover, $G$ acts trivially on all elements of $t(V)$.

If $V$ is irreducible, then either $t(V)=0$ or $t(V)=V$. In the first case, we have that in fact $t=0$. In the second one, we see that $G$ acts trivially on all of $V$, so $V$ must be of dimension $1$.


I think you can use this variant of Schur's lemma, yes! If $A$ is your sum, then $\rho(g)A = A\rho(g) = A$ for all $g \in G$: use the fact that $G$ is a group and that $\rho$ is a homomorphism. Thus $A = \lambda I$. If $\rho$ is not the trivial representation, then there exists a $g$ such that $\rho(g) \neq I$. Now you have $\lambda\rho(g) = \lambda I$. If $\lambda \neq 0$, then does this make any sense?

Added. That $\rho(g)\sum_{x \in G} \rho(x) = \sum_{x \in G} \rho(x)$ is a special case of the following fact: if I have a commutative monoid $A$, a finite set $I$, an indexing function $f\colon I \to A$, and a bijection $\mu\colon J \to I$ then $\sum_{i \in I} f(i) = \sum_{j \in J} f(\mu(j))$. This is just a pedantic way of changing variables. Here both $I$ and $J$ are $G$, and $\mu(x) = gx$.

  • $\begingroup$ Firstly how did you get $\rho(g)A = A\rho(g) = A$? As $\rho(g)A = \rho(g)(\rho(g_1) + \dots \rho(g_n)) = \rho(g)\rho(g_1) + \dots \rho(g)\rho(g_n) = \rho(gg_1) + \dots \rho(gg_n) \neq \rho(g_1g) + \dots \rho(g_ng) = (\rho(g_1) + \dots \rho(g_n))\rho(g) = A\rho(g).$ $\endgroup$
    – Anon
    Apr 1 '12 at 20:46
  • $\begingroup$ @AlexKite Is there any difference between the sets $\{h\}_{h \in G}$ and $\{gh\}_{h \in G}$? $\endgroup$ Apr 1 '12 at 20:51
  • $\begingroup$ No, I don't think there is? Also how do you get the @AlexKite sign by your comment? $\endgroup$
    – Anon
    Apr 1 '12 at 20:54
  • $\begingroup$ So this implies $\rho(g)A = A\rho(g) = A$? $\endgroup$
    – Anon
    Apr 1 '12 at 21:01
  • $\begingroup$ @AlexKite I just type @A in and your name pops up. People always get notified of comments on their answers though, so there's no need for you to do it. $\endgroup$ Apr 1 '12 at 21:34

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.