Signature of a permutation couple

29 Views Asked by At

I was working on a problem when I found out I needed to know the signature of a permutation of the form : \begin{equation} (\sigma_A,\sigma_B) \end{equation} meaning that $(\sigma_A,\sigma_B)$ is defined from $A\times B$ where $A,B$ are finite ordered sets such that $(\sigma_A,\sigma_B)(a,b)=(\sigma_A(a),\sigma_B(b))$ where we consider the lexicographic order of $A\times B$. I intuitively find that the signature is : \begin{equation} \varepsilon(\sigma_A)^{|B|} \varepsilon(\sigma_B)^{|A|} \end{equation} Let me explain my reasoning. If I suppose $A=[|1,n|]$ and $B=[|1,m|]$, we get that the permutation couple is as follows : \begin{equation} \begin{pmatrix} (1,1) & \cdots &(1,m) & (2,1) &\cdots & (2,m) & & \cdots& & (n,1) & \cdots & (n,m) \\ (\sigma_A(1),\sigma_B(1)) & \cdots &(\sigma_A(1),\sigma_B(m)) & (\sigma_A(2),\sigma_B(1)) &\cdots & (\sigma_A(2),\sigma_B(m)) & & \cdots& & (\sigma_A(n),\sigma_B(1)) & \cdots & (\sigma_A(n),\sigma_B(m)) \end{pmatrix} \end{equation} We can easily see then that : \begin{equation} (\sigma_A,\sigma_B)=(\sigma_A,id_B)\circ (id_A,\sigma_B) \end{equation} The first permutation $(\sigma_A,id_B)$ is : \begin{equation} \begin{pmatrix} (1,1) & \cdots &(1,m) & (2,1) &\cdots & (2,m) & & \cdots& & (n,1) & \cdots & (n,m) \\ (\sigma_A(1),1) & \cdots &(\sigma_A(1),m) & (\sigma_A(2),1) &\cdots & (\sigma_A(2),m) & & \cdots& & (\sigma_A(n),1) & \cdots & (\sigma_A(n),m) \end{pmatrix} \end{equation} This problem seems like the mini-problem : \begin{equation} \begin{pmatrix} \text{Bloc 1} & \text{Bloc 2}\\ \text{Bloc 2} & \text{Bloc 1} \end{pmatrix} \end{equation} where the two blocs are of the same size $b$ obviously, the signature of this permutation would be : \begin{equation} (-1)^b \end{equation} because we'd have to switch each element of the bloc with the other to return to the identity permutation, which is a total of $b$ inversions. Hence, I can easily extend it to a random permutation (justified by the factorisation theorem of permutations): \begin{equation} \begin{pmatrix} \text{Bloc $1$} & \cdots & \text{Bloc $p$} \\ \text{Bloc $\sigma(1)$} & \cdots & \text{Bloc $\sigma(p)$} \end{pmatrix} \end{equation} whose signature would be : \begin{equation} \varepsilon(\sigma)^b \end{equation} then $\varepsilon(\sigma_A,id_B)$ is : $\varepsilon(\sigma_A)^{|B|}$.
For the second permutation $(id_A,\sigma_B)$, it's a little bit easier : \begin{equation} \begin{pmatrix} (1,1) & \cdots &(1,m) & (2,1) &\cdots & (2,m) & & \cdots& & (n,1) & \cdots & (n,m) \\ (1,\sigma_B(1)) & \cdots &(1,\sigma_B(m)) & (2,\sigma_B(1)) &\cdots & (2,\sigma_B(m)) & & \cdots& & (n,\sigma_B(1)) & \cdots & (n,\sigma_B(m)) \end{pmatrix} \end{equation} we can see that each block $(i,1) \cdots (i,m)$ is stable and has signature $\varepsilon(\sigma_B)$, then we'll have to multiply them all and get for the signature : \begin{equation} \varepsilon(\sigma_B)^{|A|} \end{equation} The first factorisation of $(\sigma_A,\sigma_B)$ shows the result.
The problem I find is that the formula isn't really nice for the usage I need. I'm working on figuring out the signature $\varepsilon_n(m)$ of the permutation : \begin{equation} \pi_n(m) : \begin{cases} \mathbb{Z}/n\mathbb{Z} \longrightarrow \mathbb{Z}/n\mathbb{Z} \\ x \longmapsto mx \end{cases} \end{equation} where $n$ and $m$ are coprime. I'm thinking on focusing on $n$ being a prime number, and for that I'll need to decompose $\pi_{ab}$ using $\pi_a$ and $\pi_b$ for $a$ and $b$ coprime which can be done using the chinese factorisation theorem, where if $\varphi : \mathbb{Z}/a\mathbb{Z} \times \mathbb{Z}/b\mathbb{Z} \to \mathbb{Z}/ab\mathbb{Z}$ is the bijection $(x,y)\to z$ and $z$ is the unique solution mod $ab$ of the equation $(z,z)=(x \text{ mod $a$} , y \text{ mod $b$})$, we'll have for all $m$ coprime with $ab$ : \begin{equation} (\pi_a(m),\pi_b(m))=\varphi \circ \pi_{ab}(m) \circ \varphi^{-1} \end{equation} where the usage of the established result becomes clear. If $n=p_1^{\alpha_1} \cdots p_r^{\alpha_r}$, then I feel the expression of $\varepsilon_n(m)$ would be ugly using the formula, so I wonder if what I've done has been right?