0
$\begingroup$

Can someone help me prove these two binomial identities using either walks in Pascal's triangle or a committee-selection model?

$(1)$ $\qquad$ $\displaystyle\sum_{k=0}^m {m\choose k}{n\choose r+k}={m+n\choose m+r}$

$(2)$ $\qquad$ $\displaystyle\sum_{k=n-s}^{m-r} {m-k\choose r}{n+k\choose s}={m+n+1\choose r+s+1}$

I couldn't find these identities in another question already asked on here. If I just wasn't looking in the right place, feel free to leave the link to the answer. Thanks!

$\endgroup$
2
  • $\begingroup$ The OP asks for a combinatorial proof so no algebraic proofs I believe. The second identity does not seem to hold (tested on some configurations). I might have entered it incorrectly though. $\endgroup$ Commented May 11, 2016 at 22:45
  • $\begingroup$ @Marko: The problem with $(2)$ is that the lower limit of summation should be $s-n$, not $n-s$. $\endgroup$ Commented May 11, 2016 at 23:09

1 Answer 1

4
$\begingroup$

You can rewrite the lefthand side of the first as

$$\sum_{k=0}^m\binom{m}k\binom{n}{n-r-k}\;;$$

the Vandermonde identity then says that this is equal to

$$\binom{m+n}{n-r}=\binom{m+n}{m+n-(n-r)}=\binom{m+n}{m+r}\;.$$

The linked Wikipedia article has a combinatorial proof of Vandermonde’s identity; try to adapt it to give the desired result. There are also combinatorial proofs of various forms of it on this site; searching for it by name will turn up most of them.

The second identity isn’t quite right: the lower limit of summation should be $s-n$, not $n-s$, making it

$$\sum_{k=s-n}^{m-r} {m-k\choose r}{n+k\choose s}={m+n+1\choose r+s+1}\;;\tag{1}$$

If we let $\ell=n+k$, so that $k=\ell-n$, $(1)$ can be rewritten

$$\sum_{\ell=s}^{m+n-r}\binom{m+n-\ell}r\binom{\ell}s=\binom{m+n+1}{r+s+1}\;.\tag{2}$$

Let $A=\{0,1,\ldots,m+n\}$, and note that $|A|=m+n+1$. Suppose that we’re to select $r+s+1$ elements of the set $A$, and we want to know how many different subsets are possible. (If you like, you can easily rephrase this in terms of selecting a committee of $r+s+1$ from a pool of $m+n+1$ candidates.) Of course the righthand side of $(2)$ is the correct number; our task is to find a way to count these subsets that naturally yields the lefthand side of $(2)$.

Suppose that $S$ is one of these subsets. There is a unique $\ell_S\in S$ such that exactly $s$ elements of $S$ are less than $\ell_S$, and exactly $r$ elements of $S$ are greater than $\ell_S$. Clearly $\ell_S$ must be at least $\ell$, the smallest member of $A$ such $A$ contains $\ell$ smaller numbers. On the other hand, $\ell_S$ cannot be any larger than $m+n-r$, the largest member of $A$ such that $A$ has $r$ larger elements.

For each integer $\ell$ such that $s\le\ell\le m+n-r$ let $\mathscr{S}_\ell$ be the set of all subsets $S$ of $A$ such that $\ell_S=\ell$; you can finish the argument by showing that

$$|\mathscr{S}_\ell|=\binom{m+n-\ell}r\binom{\ell}s\;.$$

$\endgroup$
2
  • $\begingroup$ I guess my textbook had a typo. I'll double check it with my professor tomorrow. Thanks $\endgroup$
    – BrianW
    Commented May 11, 2016 at 23:13
  • $\begingroup$ @BrianW: You’re welcome. It would be a very easy typo to make. $\endgroup$ Commented May 11, 2016 at 23:13

Not the answer you're looking for? Browse other questions tagged .