I am reading the book "From Calculus to Cohomology: De Rham cohomology and characteristic classes"

Let $p, q, r$ be nonnegative integers.

It says, (for those who own the book, on pg 10) without justification that there are bijections $$S\left(p,q+r\right)\times S\left(\bar{p},q,r\right) \overset{\cong}{\longrightarrow} S\left(p,q,r\right),\ \left(\sigma,\tau\right)\mapsto \sigma \circ \tau$$ and $$S\left(p+q,r\right)\times S\left(p,q,\bar{r}\right) \overset{\cong}{\longrightarrow} S\left(p,q,r\right),\ \left(\sigma,\tau\right)\mapsto \sigma \circ \tau .$$

Here, the following notations are being used:

  • For any $n$, we let $S\left(n\right)$ denote the group of all permutations of $\left\{1,2,\ldots,n\right\}$.

  • For any $n$ and $m$, we let $S\left(n,m\right)$ denote the set of all permutations $\sigma\in S\left(n+m\right)$ with $$ \sigma\left(1\right) < \sigma\left(2\right) < \cdots < \sigma\left(n\right) \qquad \text{and} $$ $$\sigma\left(n+1\right) < \sigma\left(n+2\right) < \cdots < \sigma\left(n+m\right) . $$

  • We let $S\left(p,q,r\right)$ be the set of all permutations $\sigma \in S\left(p+q+r\right)$ with

$$\sigma\left(1\right)<\sigma\left(2\right)<\cdots<\sigma\left(p\right) ,$$


$$\sigma\left(p+q+1\right)<\sigma\left(p+q+2\right)<\cdots<\sigma\left(p+q+r\right) .$$

  • We let $S\left(\bar{p},q,r\right)$ be the set of all $\sigma \in S\left(p,q,r\right)$ such that $\sigma$ is the identity on $\left\{1,2,...,p\right\}$.

  • We let $S\left(p,q,\bar{r}\right)$ be the set of all $\sigma \in S\left(p,q,r\right)$ such that $\sigma$ is the identity on $\left\{p+q+1,p+q+2,...,p+q+r\right\}$.

I cant seem to get this, I am really stuck, I tried doing an example and breaking it down into transpositions but it shed no light on the situation.

Would rather a hint than an answer, although any help is greatly appreciated.


Define $n = p+q+r.$ First notice that $$|S(p,q,r)|=\binom{n}{p}\binom{n-p}{q}=\binom{n}{p,q,r},$$ and $|S(p,q+r)|=\binom{n}{p}$ and $|S(\overline{p},q,r)|=\binom{n-p}{q},$ so by multiplication rule $$|S(p,q+r)\times S(\overline{p},q,r)|=|S(p,q,r)|,$$ So if we prove just 1-1 or onto we are done.
Doing 1-1 is kind of painful, so let's see that the map is onto:r

Consider $\alpha \in S(p,q,r),$ we want to find $\sigma \in S(p,q+r),\tau \in S(\overline{p},q,r)$ such that $\sigma \tau = \alpha.$

From there is clear that $\sigma (i)=\alpha (i)$ for $i\leq p$ because if $\tau \in S(\overline{p},q,r),$ then $\tau (i)=i$ for $i\leq p$ but also, as $\sigma\in S(p,q+r)$, then we can uniquely determine $\sigma (j)$ for $j \geq p+1$ by a greedy approach, setting it to be the smallest element in the codomain that doesn't already appear among $\sigma(1), \sigma(2), \ldots, \sigma(j-1)$. In other words: $$\sigma (i) = \left\{ \begin{array}{ll} \alpha (i) & i\leq p \\ \min \{[n]\setminus \sigma ([i-1])\} & p+1\leq i \end{array} \right. $$ where $[n]=\{1,\cdots , n\}.$ From the definition of $\sigma$ is clear that $\sigma$ is in $S(p,q+r).$

Now that we have determined $\sigma,$ $\tau$ can be determined as $$\tau (i) = \left\{ \begin{array}{ll} i & i\leq p \\ \sigma ^{-1}(\alpha (i)) & p+1\leq i \end{array} \right.$$ So the only thing remaining is to show that $\tau \in S(\overline p,q,r),$ so you know that $\tau(i)=i$ for $i\leq p$ by definition. So take $p<i<j\leq p+q,$ so $\tau (i)=\sigma ^{-1}(\alpha (i))$ and $\tau (j)=\sigma ^{-1}(\alpha (j)),$ by definition of $\alpha$ we know that $\alpha (i)<\alpha (j)$ but by construction of $\sigma$ we are taking minimums so if $\sigma (k_1)=\alpha (i)$ and $\sigma (k_2)=\alpha (j)$ then $k_1<k_2$ which means $\sigma ^ {-1}(\alpha (i))<\sigma ^{-1}(\alpha (j))$ so $\tau (i)<\tau (j).$ Apply the same argument when $p+q<i<j\leq n$ and the result follows.

Hope it helps
Edit: An example when $p=q=r=3$,the $\color{red}{red}$ and the black lines to determine $\sigma$:


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.