10
$\begingroup$

In this question, a proof using real analysis is given of the following identity $$ \sum_{n=1}^{\infty} \frac{(n-1)!}{n \prod_{i=1}^{n} (a+i)} = \sum_{k=1}^{\infty} \frac{1}{(a+k)^{2}}$$

Is there a combinatorial proof of this identity? If so, does the proof require that $a$ be a natural number?

$\endgroup$
3
  • 4
    $\begingroup$ Very good question! I'm also wondering if there are any known obstructions for an infinite-series identity to have a combinatorial proof. (Which means either a "finite partial sums" version or a "generating functions" version.) $\endgroup$ Commented Dec 30, 2012 at 1:15
  • 1
    $\begingroup$ You really mean "direct" instead of "combinatorial". It is a popular belief that elegant formulas must have a combinatorial proof, rather unsupported by the evidence. Negative examples include mathoverflow.net/questions/34145 and mathoverflow.net/questions/16129 $\endgroup$
    – Igor Pak
    Commented Dec 30, 2012 at 4:04
  • 3
    $\begingroup$ I believe that this identity is a limit of a finite sum that can be proved using the WZ method. Whether that makes it combinatorial is a matter of opinion. $\endgroup$
    – Ira Gessel
    Commented Jul 23, 2014 at 13:48

1 Answer 1

7
$\begingroup$

The proof below is a WZ method proof.

The pair of functions$G(n,k) = \dfrac{(1)_n}{(k+x+1)_n} \, \dfrac{1}{n^2}, \quad$ $F(n,k) = \dfrac{(1)_n}{(k+x+1)_n} \, \dfrac{1}{(k+x+1)n}$

is a WZ (Wilf and Zeilberger) pair. That is $\, G(n,k+1)-G(n,k) = F(n+1,k)-F(n,k)$.

WZ-pairs are gems because of their interesting properties. For example if $$ \lim_{n \to \infty} F(n,k)=0, \quad \text{and} \quad \lim_{k \to \infty} \sum_{n=1}^{\infty} G(n,k)=0.$$

Wilf and Zeilberger proved that

$$\sum_{n=1}^{\infty} G(n,0) = \sum_{k=1}^{\infty} F(1, k-1).$$

The pair of functions that we are considering satisfies these conditions. Hence

$$ \sum_{n=1}^{\infty} \frac{(1)_n}{n^2(x+1)_n} = \sum_{k=1}^{\infty} \frac{1}{(k+x)^2},$$ which is the identity we wanted to prove.

NOTE: With $(a)_n$ we mean the rising factorial.

$\endgroup$

You must log in to answer this question.

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