If $G$ is a finite group and its conjugacy classes are known, can the conjugacy classes of the wreath product $G \wr S_n \cong G^n \rtimes S_n$ be determined?

  • 4
    $\begingroup$ Yes, this is well known. It is described in detail in James and Kerber's book on the representation theory of symmetric groups. $\endgroup$ – Christian Gaetz Oct 9 '18 at 15:59
  • 2
    $\begingroup$ Did you mean $G\wr S_n\cong G^n\rtimes S_n$? $\endgroup$ – Alex B. Oct 9 '18 at 16:14
  • $\begingroup$ Alex: I mean for $S_n$ to be acting on $n$ copies of $G$, in the natural way. Have I gotten the notation the wrong way round? $\endgroup$ – Chris Russell Oct 9 '18 at 16:19
  • 1
    $\begingroup$ In that case Chris, the notation is the wrong way round, and Alex's is what you meant ( and I think the group dealt with in James-Kerber as mentioned by Christian). $\endgroup$ – Geoff Robinson Oct 9 '18 at 16:25

This is indeed handled in Section 4.2 of James and Kerber's book on Representations of the Symmetric group.

We also considered this problem in Section 2 our paper describing a computer algorithm for computing conjugacy class representatives in permutation groups.

J. Cannon and D. Holt, Computing conjugacy class representatives in permutation groups, J. Algebra 300 (2006), 213--222.

I can cut and paste the main bit of theory you need. The problem easily reduces to finding those classes that map onto a specific element of $S_n$.

Let $A$ be any group, let $P$ be a permutation group acting on the set $\{ 1\ldots d\}$, and let $W := A \wr P$. Then $W$ is a semidirect product of $A^d$ by $P$, and elements of $W$ have the form $(g,x)$ with $g \in P$, $x \in A^d$, and $x = (x_1,\ldots,x_d)$ with $x_i \in A$. In general, for $x \in A^d$, we denote the $i$-th component of $x$ by $x_i$. The action of $P$ on $A^d$ in $W$ is given by: $$(g,1)^{-1}(1,(x_1,\ldots,x_d))(g,1) = (1,(x_{1^{g^{-1}}},\ldots, x_{d^{g^{-1}}})).$$

Hence, for $g \in P$, $x,z\in A^d$, we have \begin{equation}\label{conjeqn} (1,z)^{-1}(g,x)(1,z) = (g,y), \ \mathrm{where}\ y_i = z^{-1}_{i^{g^{-1}}}x_iz_i \ \mathrm{for}\ 1 \le i \le d. \end{equation}

$\textbf{Theorem}$ With the above notation, let $x^{(1)},\ldots,x^{(k)}$ be representatives of the conjugacy classes of $A$. Fix an element $g \in P$, and let $r_1,r_2,\ldots,r_s$ be representatives of the cycles of $g$ in its action on $\{ 1\ldots d\}$. Then $$ R := \{\,(g,x) \mid x_i = 1 \ \mathrm{for}\ i \not\in \{r_1,\ldots,r_s\},\, x_i \in \{x^{(1)},\ldots,x^{(k)}\}\ \mathrm{for}\ i \in \{r_1,\ldots,r_s\}\,\} $$ is a set of representatives of the $A^d$-classes of elements of $W$ of the form $(g,x)$ for $x \in A^d$.

The size of the $A^d$-class $R_{gx}$ containing $(g,x)$ is the product of the sizes of the classes of the $x_i$ in $A$ with $i \in \{r_1,\ldots,r_d\}$ with the lengths of the cycles of $g$ that contain $r_1,\ldots,r_d$.

The above describes the $A^d$-classes of $W$; i.e. the orbits under conjugation by the subgroup $A^d$ of $W$. To get the classes of $W$ itself, first of all we only consider those $R_{gx}$ as $g$ ranges over a set of class representatives of $P$. Then, for each such $g \in W$, the $A^d$-class representatives $R_{gx}$ are fused under the action of $C_P(g)$. This action is the same as the action of $C_P(g)$ on the cycles of $g$, so is easy to calculate.

Then the size of a $W$-class $R_{gx}$ is the product of the size of the class of $g$ in $P$ with the sum of the the sizes of the $A^d$-classes $R_{gx'}$ that lie in the orbit under the action of $C_P(g)$.

  • $\begingroup$ Thank you Derek, I see how an element $(g, x)$ is conjugate to something in the set $R$ you define. If I write $$R_g := \{\,(g,x) \mid x_i = 1 \ \mathrm{for}\ i \not\in \{r_1,\ldots,r_s\},\, x_i \in \{x^{(1)},\ldots,x^{(k)}\}\ \mathrm{for}\ i \in \{r_1,\ldots,r_s\}\,\}$$ is there a way to describe a subset of $$\bigcup_{g \in G}R_g$$ which contains only one representative from each conjugacy class? Also is there a way to determine how many elements are in a each conjugacy class? $\endgroup$ – Chris Russell Oct 10 '18 at 9:32
  • 1
    $\begingroup$ My answer was incomplete because it only described the $A^d$-classes. Calculating the $G$-classes from thsi information is not difficult - I have added some further explanation. $\endgroup$ – Derek Holt Oct 10 '18 at 10:48
  • $\begingroup$ I think I get the idea of how it works now but I'm not sure what $G$ is in the second last paragraph of your updated answer, which is confusing me. In one place I'm interpreting that $A^d$ is a subgroup of $G$ but in another $g \in G$ where I thought $g$ is an element of $P$. $\endgroup$ – Chris Russell Oct 10 '18 at 11:29
  • $\begingroup$ Sorry $G$ should be $W$. I don't have a $G$. I have $W = A^d \rtimes P$. Of course $P=S_n$ in your question. $\endgroup$ – Derek Holt Oct 10 '18 at 12:10

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.