Let $N \lhd G$ (with $G$ finite) and let $\chi \in \mathrm{Irr}(G)$ be such that $\chi_N=\theta \in \mathrm{Irr}(N)$. Then the characters $\beta \cdot \chi$ for $\beta \in \mathrm{Irr}(G/N)$ are irreducible, distinct for distinct $\beta$ and are all of the irreducible constituents of $\theta^G$. This is Corollary 6.17 of Isaacs' character theory of finite groups.
My question is: Let $\chi \in \mathrm{Irr}(G)$ and set $\mathrm{Irr}(G/N)=\{\theta_1 \dots \theta_n\}$. Suppose that the $\{\theta_i \cdot \chi\}_{i=1}^n$ are distinct and irreducible. Can we conclude that $\chi_N$ is irreducible? If yes, why?

(Notation: with $\chi_N$ i indicate the restriction of $\chi$ to $N$ and with $\theta^G$ i indicate the extension of $\theta$ to $G$. With $\beta \cdot \chi$, $\theta_i \cdot \chi$ i indicate the product of those characters)

  • $\begingroup$ What's $\beta_\chi$ or $\theta_{i\chi}$ ? $\endgroup$ – Max Feb 18 at 20:00
  • $\begingroup$ @Max it is a product of characters: $\beta \cdot \chi$ and $\theta_i \cdot \chi$. I modified my question adding "\cdot". $\endgroup$ – ciccio Feb 18 at 20:04
  • $\begingroup$ Oh ok I thought it was an index, my bad $\endgroup$ – Max Feb 18 at 20:05
  • 1
    $\begingroup$ I don't know why, the $\chi$ stays a Little "lower". $\endgroup$ – ciccio Feb 18 at 20:08

Let us write and identify $Irr(G/N)=\{\beta \in Irr(G): N \subseteq ker(\beta)\}$. (Note that you are using $\theta_i$ in your question, but this is rather confusing, since you are using $\theta$ for an irreducible character of $N$).

Your assumption is that all the $\beta\chi$ are distinct and irreducible.

Let $\chi \in Irr(G)$ and $\theta \in Irr(N)$ be an irreducible constituent of $\chi_N$. Then $\chi_N=e\sum_{i=1}^{t}\theta_i$, where $t=|G:T|$, the index of the intertia subgroup of $\theta$ in $G$ and $e$ is a positive integer. So $\chi(1)=et\theta(1)$. We need that later.

Now let us have a look at the irreducible constituents of $\theta^G$. By Frobenius Reciprocity we have: $[\beta\chi,\theta^G]=[(\beta\chi)_N,\theta]=[\beta(1)\chi_N,\theta]=\beta(1)[\chi_N,\theta]=\beta(1)e.$ This means that all the different $\beta\chi$ appear with multiplicity $\beta(1)e$ as irreducible constituent of $\theta^G$.

Hence $\theta^G(1)=\theta(1)|G:N| \geq \sum_{\beta}\beta(1)e(\beta\chi)(1)=e\chi(1)\sum_\beta\beta(1)^2=e\chi(1)|G:N|=e^2t\theta(1)|G:N|.$ It follows that $e^2t \leq 1$, meaning $e=1=t$. But then $\chi_N=\theta$ and we are done.

  • $\begingroup$ All clear. Thank You! I did the same reasoning, but i cannot conclude since i couldn't see that inequality. $\endgroup$ – ciccio Feb 19 at 9:36
  • 1
    $\begingroup$ You are welcome. Good question and nice problem, never realized that Gallagher's result had a converse! $\endgroup$ – Nicky Hekster Feb 19 at 9:42

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.