2
$\begingroup$

I am seeking a closed-form (a form in terms of known special functions) to the sum $$ \sum_{n=0}^\infty\frac{(1/2)_n}{n!}(H_n-H_{n-1/2}). $$

Context:

I am searching for closed-forms to special cases of the double integral $$ I(\alpha_1,\alpha_2,\zeta,\nu)=\nu^2\int_0^1\int_0^1 {_2F_1}\left({1+\nu,1+\nu \atop \alpha_1};\frac{(1-x)(1-y)}{(1-(1-\zeta)x)(1-(1-\zeta)y)}\right) {_2F_1}\left({1-\nu,1-\nu \atop \alpha_2};(1-x)(1-y)\right)\left((1-(1-\zeta)x)(1-(1-\zeta)y\right)^{-\nu-1}\,\mathrm dx\mathrm dy, $$ which converges for $\alpha_1+\alpha_2>2$, $-\alpha_2<2\nu<\alpha_1$. Also, we require $\zeta\geq 0$, $\alpha_1,\alpha_2>0$, and $\nu\in\Bbb R$. This double integral represents the second moment of a continuous random variable I've been studying.

Substituting $\alpha_1=1$, $\alpha_2=3/2$, $\zeta=1$, and $\nu=-1/2$, I was able to reduce this double integral to $$ I(1,3/2,1,-1/2)=\frac{1}{2}\sum_{n=0}^\infty\frac{(1/2)_n}{n!}(H_n-H_{n-1/2}), $$ where $H_z$ is the Harmonic number. I can provide details to this calculation if desired. But how to proceed? The difference of the harmonic numbers can be reduce to the form $H_n-H_{n-1/2}=a_n+\log 4$, which may be helpful.

$\endgroup$

1 Answer 1

2
$\begingroup$

From the integral representation \begin{equation} H_p=\int_0^1\frac{1-t^p}{1-t}\,dt \end{equation} we have \begin{align} H_n-H_{n-1/2}&=\int_0^1\frac{t^{n-1/2}-t^n}{1-t}\,dt\\ &=\int_0^1\frac{t^{n-1/2}(1-\sqrt{t})}{1-t}\,dt\\ &=2\int_0^1\frac{u^{2n}}{1+u}\,du \end{align} The proposed series can be written as \begin{align} S&=\sum_{n=0}^\infty\frac{(1/2)_n}{n!}(H_n-H_{n-1/2})\\ &=2\sum_{n=0}^\infty\frac{(1/2)_n}{n!}\int_0^1\frac{u^{2n}}{1+u}\,du \end{align} by swapping integration and summation \begin{align} S&=2\int_0^1\frac{du}{1+u}\sum_{n=0}^\infty\frac{(1/2)_n}{n!}u^{2n}\\ &=2\int_0^1\frac{du}{(1+u)\sqrt{1-u^2}}\\ &=\int_0^1v^{-1/2}\,dv\\ &=2 \end{align} (the last integral was obtained by changing $u=(1-v)/(1+v)$).

$\endgroup$
1
  • 2
    $\begingroup$ I would have to say $I(1,3/2,1,-1/2)$ is the most complicated way to write the number one I've ever seen! $\endgroup$ Jan 2 at 19:13

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.