Let $m\in\mathbb{Z}_{>0}$ and $p\in\mathbb{Z}$ with $0\leqslant p\leqslant 2m.$ Then how could we simplify the following double sum? \begin{align*} \sum_{j=0}^{m}\sum_{\ell=0}^{p}(-1)^{j+\ell}\binom{j+k}{j}\binom{2(k+m)-p}{2(k+j)}\binom{\ell}{m-j}\binom{2m+k}{p-\ell} \end{align*}

- 1,366
-
1Can you use Python or something to find small values ? Say $k=0$, $m=0,1,2,3; p=0..2m$ Do you recognize the numbers you get out? Then try $k=1$ – Empy2 Jan 01 '23 at 14:31
-
1It is possible to compute the inner sum but it contains an hypergeometric function. – Claude Leibovici Jan 04 '23 at 10:20
1 Answers
Here is a starter. We can find a closed form for the inner sum.
We obtain \begin{align*} \color{blue}{\sum_{\ell=0}^p}&\color{blue}{\binom{l}{m-j}\binom{2m+k}{p-\ell}(-1)^{\ell}}\\ &=\sum_{\ell=m-j}^p\binom{\ell}{m-j}\binom{2m+k}{p-\ell}(-1)^{\ell}\tag{1}\\ &=\sum_{\ell=0}^{p-m+j}\binom{\ell+m-j}{m-j}\binom{2m+k}{p-\ell-m+j}(-1)^{\ell+m-j}\tag{2}\\ &=\sum_{\ell=0}^{p-m+j}\binom{\ell+m-j}{\ell}\binom{2m+k}{p-m+j-\ell}(-1)^{\ell+m-j}\tag{3}\\ &=\sum_{\ell=0}^{p-m+j}\binom{-m+j-1}{\ell}\binom{2m+k}{p-m+j-\ell}(-1)^{m-j}\tag{4}\\ &\,\,\color{blue}{=\binom{m+k+j-1}{p-m+j}(-1)^{m-j}}\tag{5} \end{align*}
Comment:
In (1) we set the lower index to $m-j$ since $\binom{p}{q}=0$ if $0\leq p<q$.
In (2) we shift the index to start with $\ell=0$.
in (3) we use $\binom{p}{q}=\binom{p}{p-q}$.
In (4) we use the binomial identity $\binom{-p}{q}=\binom{p+q-1}{q}(-1)^q$.
In (5) we use Vandermonde's identity.
We conclude \begin{align*} \color{blue}{\sum_{j=0}^{m}}&\color{blue}{\sum_{\ell=0}^{p}(-1)^{j+\ell}\binom{j+k}{j}\binom{2(k+m)-p}{2(k+j)}\binom{\ell}{m-j}\binom{2m+k}{p-\ell}}\\ &\,\,\color{blue}{=(-1)^m\sum_{j=0}^{m}\binom{j+k}{j}\binom{2(k+m)-p}{2(k+j)}\binom{m+k+j-1}{p-m+j}} \end{align*}

- 108,315
-
Great!, I have checked your equalities which are all valid. So, can we simplify this summation to a closed form from the last blue line ? This looks like a hypergeometric sum. – user1992 Jan 05 '23 at 08:47
-
1@user1992: Thanks! Unfortunately, I do not see any further simplifications at the moment. – Markus Scheuer Jan 05 '23 at 11:00
-
@user1992 Mathematica and Wolfram Alpha give a simplified form for $(-1)^m \sum\limits_{j=0}^m \binom{j+k}{j} \binom{2 (k+m)-p}{2 (k+j)} \binom{m+k+j-1}{p-m+j}$ (see https://www.wolframalpha.com/input?i=%28-1%29%5Em+Sum%5BBinomial%5Bj%2Bk%2Cj%5D+Binomial%5B2%28k%2Bm%29-p%2C2%28k%2Bj%29%5D+Binomial%5Bm%2Bk%2Bj-1%2Cp-m%2Bj%5D%2C%7Bj%2C0%2Cm%7D%5D). – Steven Clark Jan 06 '23 at 02:37
-
1@StevenClark: I would not consider this a simplification, but rather an alternative representation. The use of two generalized hypergeometric functions contains summation symbols, which is not a desired closed form. – Markus Scheuer Jan 06 '23 at 07:18
-
@epi163sqrt I understand your point, but in the first comment above user1992 expressed an interest in a hypergeometric function representation which user1992 seems to consider a closed form. Also, many functions typically included in closed form representations are evaluated with summation (approximations via truncated infinite series). Personally I consider any finite sum such as your derived representation a closed form. – Steven Clark Jan 06 '23 at 16:41
-
@StevenClark: I also understand your point of view. Nevertheless, I share the point of view of D. E. Knuth and R. P. Stanley, as indicated for example here. – Markus Scheuer Jan 06 '23 at 17:41
-
-