This is a standard result in representation theory of finite groups. A poof of the column orthogonality of the character table can be found here. By orthogonality of columns I mean that $$\sum_{V}\overline{\chi}_V(g)\chi_V(h)=\begin{cases} \vert G \vert/\vert c_g\vert& \text{if }g\sim h\\ 0& \text{otherwise. }\end{cases} $$ Here $G$ is a finite group, and $c_g$ is the conjugacy class of $g\in G$. Also, $V$ is supposed to be running in the irreducible representations of $G$.The prove mentioned before, although clever, it seems a bit artificial. In the sense that treating a character table as a matrix is not obvious at all. I think there must be a more pedestrian way to deduce this result from the row orthogonality $$\sum_g\overline\chi_V(g)\chi_W(g)=\begin{cases}\vert G\vert&\text{if }V\simeq W\\ 0& \text{otherwise.} \end{cases}$$ There must be an algebraic manipulation of equation that lead to this result. Is there one?

  • 2
    $\begingroup$ What is wrong with treating it as a matrix? Anyway, what is a rigorous definition for the character table itself? Answer: you define it as a matrix. If $\chi_1,...,\chi_r$ are the irreducible characters and $g_1,...,g_r$ are representatives for the conjugacy classes of $G$ then you simply define a matrix by $(A)_{ij}=\chi_i(g_j)$. This is absolutely well defined. So the character table is a matrix by definition. (if you want to define it formally) $\endgroup$
    – Mark
    Sep 1 '20 at 21:45
  • $\begingroup$ @Mark Yes. It is well defined, but I mean that it is not obvious, right away, that we should construct a table, let alone a matrix. If all we knew was that characters of different irreducible representation are orthonormal; how could we come up with the other result? $\endgroup$ Sep 1 '20 at 22:02
  • 1
    $\begingroup$ I don't know other proofs. I mean, multiplying matrices is exactly the "algebraic manipulation" that you ask about. When you multiply matrices, the corresponding equations give you the result. I think it is a pretty natural idea to construct a matrix when we work with finite orthogonal sequences. $\endgroup$
    – Mark
    Sep 1 '20 at 22:10

Here is an alternative proof:

Lemma: The characters of irreducible representations form a basis for the space $\mathbb C_{class}(G)$ of class functions $f:G\to \mathbb C$.

Proof: We prove that whenever a class function $f$ is orthogonal to every character of irreducible representations, then it is the zero map. Let $\rho:G\to\text{GL}(V)$ be an irreducible representation of $G$. If $f\in \mathbb C_{class}(G)$, then the map $\phi=\sum_{g\in G}f(g)\rho(g^{-1})$ is a representation homomorphism. By Schur's Lemma $\phi=\lambda I$ for some $\lambda\in \mathbb{C}$. The trace of $\phi$ vanishes by assumption. Thus $\phi$ is the zero map in any irreducible representation and therefore in any representation of $G$. Setting the representation $\rho $ as the regular representation we have $\sum_{g\in G}f(g)\rho(g^{-1})=0$, and all group actions in the regular representation are linearly independent. Then $f=0$. $\blacksquare$

I sketch what is needed to prove your result, you can fill in the details.

  1. Fix $g\in G$, let $f(h)=1$ if $h\sim g$ and $f(h)=0$ otherwise.
  2. Write $f=\sum_{i}a_i\chi_i$, we consider the irreducible representations $V_i$ with characters $\chi_i$.
  3. Compute $a_i$ by means of the Hermitian product (use orthonormality).

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.