# The sgn function and permutations

Let $P=\{(i, j)|1\leq i<j\leq n\}.$For $\sigma\in S_n$, define $\operatorname{sgn}\colon S_{n}\rightarrow \{\pm 1\}$ by $$\operatorname{sgn}(\sigma)=\prod_{(i, j)\in P}\operatorname{sgn}(\sigma(j)-\sigma(i)).$$ How to prove the following?

1. $\displaystyle{\operatorname{sgn}(\sigma)=\prod_{1\leq i<j\leq n}\left(\frac{\sigma(j)-\sigma(i)}{j-i}\right)}$.

2. Consider the polynomial $\displaystyle{p(x_{1},...,x_{n})=\prod_{1\leq i<j\leq n}(x_{j}-x_{i})}$.

For $\sigma\in S_n$ prove that $\displaystyle{\operatorname{sgn}(\sigma)=\frac{p(x_{\sigma(1)},\ldots,x_{\sigma(n)}) }{p(x_{1},\ldots,x_{n})}}$.

3. Let $N_{\sigma}$ be the number of orbits of $\sigma$ acting on $\{1,\ldots,n\}$. Show $\operatorname{sgn}(\sigma)=(-1)^{n-N_{\sigma}}$.

-
Surely you know how to solve some of these! Please show what you tried and where you are stuck. – Did May 24 '12 at 21:00
You have a conflicting notation: you use sgn for two different functions. – M Turgeon May 24 '12 at 21:05
Maybe you should first write them informally? – Phira May 24 '12 at 21:52
@Andres If $x\in\mathbb{R}$ is non zero, we have $sgn(x)=\frac{|x|}{x}$ (by definition). If this doesn't help, I don't know what will. – M Turgeon May 25 '12 at 1:45
Then write what you have for particular cases, but write something. Yourself. Mathematics is not a spectator sport. – Did May 25 '12 at 21:15

Pay attention to the fact that in the formula in $\,(1)\,$ , for any $\,1\leq i<j\leq n\,$, the denominator $\,j-i\,$ is always positive , so what "decides" the sign there is merely the numerator $\,\sigma(j)-\sigma(i)\,$ , thus: both "definitions" you wrote down of sgn are exactly the same.

Now, let us write down the given information: $$\frac{p\left(x_{\sigma(1)},...,x_{\sigma(n)}\right)}{p\left(x_1,...,x_n\right)}=\frac{\prod_{1\leq i<j\leq n}\left(x_{\sigma(j)}-x_{\sigma(i)}\right)}{\prod_{1\leq i<j\leq n}\left(x_j-x_i\right)}=\prod_{1\leq i<j\leq n}\frac{x_{\sigma(j)}-x_{\sigma(i)}}{x_j-x_i}\,\,(**)$$and as before: the factors in the denominator are the very same factors in the numerator up to sign, and there's a change of sign, say for some pairs $\,(r,s)\,,\,(i,j)\in P\,$: $$\frac{x_{\sigma(j)}-x_{\sigma(i)}}{x_r-x_s}=\frac{-(x_r-x_s)}{x_r-x_s}=-1$$iff exactly the same change exists in $$\frac{\sigma(j)-\sigma(i)}{r-s}=\frac{-(r-s)}{r-s}=-1$$ so from here clearly we have that $\,\,(**)=sgn(\sigma)\,$

Finally, we can try to attack $\,(3)\,$ with the following:

Lemma: For every permutation $\,\sigma\in S_n\,\,,\,n-N_\sigma=T_\sigma$ , with $\,T_\sigma:=\,$number of transpositions $\,(i \,j)\,,\,1\leq i\neq j\leq n\,$ , needed to express $\,\sigma\,$ as a product of transpositions.

Proof: It is enough to prove for cycles, as any permutation can be written as a product of (disjoint) cycles: let $\,\sigma = (i_1\,i_2\,,...,i_r)\,$ be an $\,r-\,$cycle , then $$\sigma=(i_1\,i_r)(i_1\,i_{r-1})\cdot...\cdot (i_1\,i_3)(i_1\,i_2)$$and we can see that the number of transpositions in the right hand side is $$\,r-1=n-N_\sigma\Longleftrightarrow N_\sigma=n-r+1$$ which is clearly the number of orbits of $\,\sigma\,$ , as it fixes $\,n-r\,$ points in $\,\{1,2,...,n\}\,$ Q.E.D.

Well, then we already are there since any transposition has sign $\,-1\,$, obviously, so that $$(-1)^{n-N_\sigma}=(-1)^{r-1}$$and let us not forget that the number of orbits of $\,\sigma\,$ is just another name for "the number of different (disjoint) cycles in the decomposition of $\,\sigma\,$"

-
@Andres Did you see my answer above? – DonAntonio Jun 10 '12 at 3:46