6

Apparently it is well known that the conjugacy class of maximum order is the one with one cycle of length $n-1$ and one element fixed. Which is somewhat intuitive when you already know the answer since we want $1^{c_1}c_1!\cdot2^{c_2}c_2!\cdot\ldots\cdot n^{c_n}c_n!$ to be minimal.

However I wasn't able to find a formal proof anywhere on the internet. This question was asked here once and wasn't answered. I was thinking that one probably could make an injective and non surjective mapping from all other conjugacy classes but I'm pretty sure that is not an elementary approach and I have no idea how I would even begin to do that. I'd rather have some kind of combinatorial proof if possible.

Edit: I know how to find order of the specific conjugacy class in $S_n$. What I want to know is how would someone deduce that the one of the highest order is the conjugacy class consisting of a cycle of length $n-1$ and one fixed element.

Derek Holt
  • 96,726
  • I know the formula. What I want to know is how would someone deduce that this specific conjugacy class above has the maximum order from the formula(or any other way) – Jurgen Klop Mar 14 '22 at 08:56
  • Closer is the question discussed here, but no proof is given. – kabenyuk Mar 14 '22 at 10:07
  • Yes, that is the one I was referring to in my question – Jurgen Klop Mar 14 '22 at 10:10
  • A similar question is also discussed here – kabenyuk Mar 14 '22 at 10:42
  • Why is this question closed?The link provided as a duplicate doesnt have an answer to it .It is not even the same question.How this site even work? One person decides the question has an answer already and reports it and it is automatically closed? – Jurgen Klop Mar 15 '22 at 07:54
  • I think it is wrong to close the issue and voted to reopen it. Those links I pointed out are only to point out to the questioner that this issue has been discussed in various aspects before. – kabenyuk Mar 15 '22 at 08:06
  • @kabenyuk I know you didnt report it, this comment wasnt meant for you, I didnt know where to write it so I wrote it in the comment section – Jurgen Klop Mar 15 '22 at 08:17

2 Answers2

6

This is not hard to prove, although it's a bit tedious. It might be easier to show that the $(n-1)$-cycle has the smallest centralizer, which has order $n-1$ when $n>2$, which we can do by induction on $n$. (When $n=2$, all centralizers have order $2$.)

Let $g \in S_n$. If $g$ is an $n$-cycle then its centralizer has order $n$ which is larger than $(n-1)$, so assume that $g$ is neither an $n$-cycle nor an $(n-1)$-cycle.

Then $g$ fixes two disjoint subsets $A$ and $B$ of $A \cup B =\{1,2,\ldots,n\}$ with $2 \le a \le b \le n-2$, where $a=|A|$ and $b=|B|$, and $$|C_{S_n}(g)| \ge |C_{{\rm Sym}(A)}(g|_A)||C_{{\rm Sym}(B)}(g|_B)|\ge (a-1)(b-1)$$ by inductive assumption, but when $a=2$ we can replace the right hand side by $2(b-1)$.

Now if $a=2$ and $b \ge 4$, then $2(b-1) > n-1 = a+b-1$, whereas if $a \ge 4$, or $a=3$ and $b \ge 5$, then $(a-1)(b-1) > a+b-1$, so this leaves only a few small cases like $a=b=3$ to check individually.

Derek Holt
  • 96,726
3

Consider the denominator $\prod_jj^{c_j}c_j!$ you mentioned, which needs to be minimized. Since this is larger for an $n$-cycle than for an $(n-1)$-cycle (for $n\ge3$), the maximal conjugacy class isn’t the class of $n$-cycles. Thus there is a largest cycle of length $k$, and at least one other cycle of length $l$. Combining these two cycles into a single cycle creates a cycle of a length $k+l$ that hadn’t occurred before. Thus, in the resulting permutation $c_{k+l}=1$, and $c_k$ and $c_l$ haven’t increased, so the product $\prod_jc_j!$ hasn’t increased. The product $\prod_jj^{c_j}$ has changed by a factor $\frac{k+l}{kl}=\frac1k+\frac1l$, which is $\le1$ unless one of $k$ and $l$ is $1$. Thus the denominator can always be decreased unless the only cycles except the largest one are $1$-cycles. Merging one of the $1$-cycles into the largest cycle (of length $k$) changes the denominator by a factor $\frac{k+1}{kc_1}$, which is $\gt1$ for $c_1=1$ but $\le1$ otherwise. Thus, the permutations in the maximal congugacy class have exactly one $1$-cycle in addition to the largest cycle of length $k=n-1$.

joriki
  • 242,601