For non-negative integers $m$ and $n$, the double sum $$\sum_{j=0}^{n} \sum_{i=0}^{m} (-1)^{i+j} {m \choose i} {n \choose j} {i+j \choose i}$$ can be checked to be 0 or 1. How can one show this by hand, along with the conditions on $m$ and $n$?
-
What have you tried? Did you get anywhere? A little computer experiment suggests that the sum is $1$ precisely when $n=m$. (This true when $1\leq n,m\leq25$.) – saulspatz Jun 21 '19 at 03:57
-
2@saulspatz, Thanks, yes you are right, you may see my solution posted below. – Z Ahmed Jun 21 '19 at 05:20
-
Seems like you answer your own question...chill – Cloud JR K Jun 25 '19 at 13:23
3 Answers
The Laguerre Polynomials are defined as $$L_n(x)=\sum_{j=0}^{n} (-1)^{j} \frac{{n \choose j}}{j!} x^J,~~~(1)$$ these are well known to follow the orthogonality condition as $$\int_{0}^{\infty} e^{-x} L_m(x)~ L_n(x)~ dx=\delta_{m,n}~,~~~~(2)$$ where $\delta_{m,n}$ is the Kroneckor delta function, If we insert (1) in (2), we get $$\int_{0}^{\infty}\sum_{i=0}^{m} \sum_{0}^{n} (-1)^{i+j} \frac{{m \choose i}}{i!} \frac{{n \choose j}}{j!} x^{i+j} e^{-x}~ dx= \delta_{m,n}.$$ Finally, $\int_{0}^{\infty} x^k e^{-x}~dx= k!$, leads to $$\sum_{i=0}^{m} \sum_{j=0}^{n} (-1)^{i+j} {m \choose i} {n \choose j} {i+j \choose i} = \delta_{m,n}.$$ Hence the result.
- 43,235
-
1This is a much clearer solution. It lacks a proof of (1) but this can readily be found. – Paul Childs Jun 23 '19 at 06:13
This is a slight variation of @MarkoRiedels nice answer. We use the coefficient of operator $[z^k]$ to denote the coefficient of $z^k$ of a series. This way we can write for instance \begin{align*} [z^k](1+z)^n=\binom{n }{k}\tag{1} \end{align*}
We obtain for integral $m,n\geq 0$: \begin{align*} \color{blue}{\sum_{p=0}^n}&\color{blue}{\sum_{q=0}^m(-1)^{p+q}\binom{n}{p}\binom{m}{q}\binom{p+q}{q}}\\ &=\sum_{p=0}^n\binom{n}{p}(-1)^p\sum_{q=0}^m\binom{m}{q}(-1)^q[z^q](1+z)^{p+q}\tag{2}\\ &=[z^0]\sum_{p=0}^n\binom{n}{p}(-1)^p(1+z)^p\sum_{q=0}^m\binom{m}{q}\left(-\frac{1+z}{z}\right)^q\tag{3}\\ &=[z^0]\sum_{p=0}^n\binom{n}{p}(-1)^p(1+z)^p\left(1-\frac{1+z}{z}\right)^m\tag{4}\\ &=(-1)^m[z^m]\sum_{p=0}^n\binom{n}{p}(-1)^p(1+z)^p\tag{5}\\ &=(-1)^m[z^m]\left(1-(1+z)\right)^n\\ &=(-1)^{m+n}[z^m]z^n\\ &\,\,\color{blue}{=[[m=n]]}\tag{6} \end{align*}
and the claim follows.
Comment:
In (2) we use the coefficient of operator according to (1).
In (3) we do some rearrangements and apply the rule $[z^{p-q}]A(z)=[z^p]z^qA(z)$.
In (4) we apply the binomial theorem.
In (5) we do some simplifications and apply the rule from (3) again.
In (6) we use Iverson Brackets.
- 108,315
-
Thank you for your contribution. I hope it helps clarify the method that was used in the two combinatorial proofs so that we have a more readable page. To be precise the argument to the $[z^0]$ coefficient extractor on line three is not a formal power series, so for this to hold we must use the definition in terms of residues $${n\choose k} = [z^k] (1+z)^n = \mathrm{Res}_{z=0} \frac{1}{z^{k+1}} (1+z)^n.$$ The proof as I posted it uses formal power series only (no pole at zero). – Marko Riedel Jun 23 '19 at 20:24
-
@MarkoRiedel: You're welcome and thanks for the credit. I also hope, it helps to clarify the used technique. Thank you for the additional information which also makes the change of order of summation of the inner sum in your first line plausible. – Markus Scheuer Jun 23 '19 at 20:37
Starting from
$$\sum_{p=0}^n \sum_{q=0}^m (-1)^{p+q} {n\choose p} {m\choose q} {p+q\choose q}$$
we write
$$\sum_{p=0}^n (-1)^{p} {n\choose p} \sum_{q=0}^m (-1)^{m-q} {m\choose q} {p+m-q\choose m-q} \\ = \sum_{p=0}^n (-1)^{p} {n\choose p} (-1)^m [z^m] (1+z)^{p+m} \sum_{q=0}^m (-1)^{q} {m\choose q} z^q (1+z)^{-q} \\ = \sum_{p=0}^n (-1)^{p} {n\choose p} (-1)^m [z^m] (1+z)^{p+m} \left(1-\frac{z}{1+z}\right)^m \\ = (-1)^m [z^m] \sum_{p=0}^n (-1)^{p} {n\choose p} (1+z)^{p} \\ = (-1)^m [z^m] (1-(1+z))^n = (-1)^{m+n} [z^m] z^n = (-1)^{m+n} \delta_{n,m} = \delta_{n,m}.$$
- 61,317
-
-
-
-
1We are working with formal power series here. For another example and additional documentation consult this MSE link A and this MSE link B by Markus Scheuer. – Marko Riedel Jun 23 '19 at 14:02
-
@MarkoRiedel: Would you mind to look at this problem? Maybe you could help us to find a solution. Best regards, – Markus Scheuer Jul 06 '19 at 07:13