8
$\begingroup$

I was looking at this paper on section [17],

$$\sum_{n=1}^{\infty}\frac{H_n{2n \choose n}}{2^{2n}(2n-1)}=2\tag1$$

Let generalize $(1)$

$$\sum_{n=1}^{\infty}\frac{H_n{2n \choose n}}{2^{2n}(2n-1)(2n-3)(2n-5)\cdots [2n-(2k+1)]}\tag2$$

Where $k\ge 0$

I conjectured the closed form for $(2)$ to be

$$\sum_{n=1}^{\infty}\frac{H_n{2n \choose n}}{2^{2n}(2n-1)(2n-3)\cdots [2n-(2k+1)]}=\frac{2(-1)^k}{(2k+1)!!(2k+1)}\tag3$$

Here are a first few values of $k=1,2$ and $3$

$$\begin{align} \sum_{n=1}^{\infty}\frac{H_n{2n \choose n}}{2^{2n}(2n-1)(2n-3)}&=-\frac{2}{9}\tag4\\ \sum_{n=1}^{\infty}\frac{H_n{2n \choose n}}{2^{2n}(2n-1)(2n-3)(2n-5)}&=\frac{2}{75}\tag5\\ \sum_{n=1}^{\infty}\frac{H_n{2n \choose n}}{2^{2n}(2n-1)(2n-3)(2n-5)(2n-7)}&=-\frac{2}{735}\tag6 \end{align}$$

How do we go about to prove this conjecture $(3)?$

$\endgroup$
  • $\begingroup$ How did you determine (4), (5) and (6)? Have you tested them numerically? $\endgroup$ – Richard Jan 2 at 18:21
  • $\begingroup$ I tested them numerically. it seems correct $\endgroup$ – Endgame Jan 2 at 18:30
5
$\begingroup$

What about reindexing and induction? The terms $\frac{1}{(2n-1)\cdots(2n-2k-1)}$ have a nice telescopic structure: by the residue theorem $$ \frac{1}{(2n-1)(2n-3)\cdots(2n-2k-1)}=(-1)^k\sum_{h=0}^{k}\frac{(-1)^h}{(2n-2h-1)}\cdot\frac{1}{2^{k+1}(2h)!!(2k-2h)!!} $$ equals $$ \frac{(-1)^k}{2^{2k+1}k!} \sum_{h=0}^{k}\frac{(-1)^h}{(2n-2h-1)}\binom{k}{h}.$$ The natural temptation is now to compute $$ \sum_{n\geq 1}\frac{H_n}{4^{n}}\binom{2n}{n}\frac{1}{2n-2h-1}$$ through $\frac{-\log(1-z)}{1-z}=\sum_{n\geq 1}H_n z^n$ and $\frac{1}{4^n}\binom{2n}{n}=\frac{2}{\pi}\int_{0}^{\pi/2}\left(\cos\theta\right)^{2n}\,d\theta$, multiply both sides by $(-1)^k \binom{k}{h}$, sum over $h=0,1,\ldots,k$ and finish by invoking Fubini's theorem (allowing to switch the integrals with respect to $d\theta$ and $dz$) and the Fourier series $\sum_{m\geq 1}\frac{\cos(m\varphi)}{m}$ and $\sum_{m\geq 1}\frac{\sin(m\varphi)}{m}$.

The only obstruction is that $\frac{1}{2n-2h-1}=\int_{0}^{1}z^n\left[\frac{1}{2z^{h+3/2}}\right]\,dz$ does not hold unconditionally: we would have been happier in having rising Pochhammer symbols rather than falling ones. On the other hand, reindexing fixes this issue. Since $\binom{2n+2}{n+1} = \frac{2(2n+1)}{n+1}\binom{2n}{n}$, the original series can be written as

$$ \sum_{n\geq 1}\frac{H_n \binom{2n}{n}}{4^n(2n-1)} = \sum_{n\geq 0}\frac{2H_{n+1}\binom{2n}{n}}{4^{n+1}(n+1)}=-\frac{1}{\pi}\int_{0}^{1}\sum_{n\geq 0}\int_{0}^{\pi/2}z^n\left(\cos\theta\right)^{2n}\log(1-z)\,d\theta\,dz $$ or $$ -\frac{1}{\pi}\int_{0}^{1}\int_{0}^{\pi/2}\frac{\log(1-z)}{1-z\cos^2\theta}\,d\theta\,dz =-\frac{1}{2}\int_{0}^{1}\frac{\log(1-z)}{\sqrt{1-z}}\,dz,$$ clearly given by a derivative of the Beta function. This approach works also by replacing $(2n-1)$ with $(2n-1)\cdots(2n-2k-1)$, you just have to be careful in managing the involved constants depending on $k$.

On second thought, I have just applied the binomial transform in disguise.

$\endgroup$

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.