2

$$S=\sum_{r=1}^{\infty}\sum_{s=1}^{\infty} \frac{1}{rs(r+s)}=\int_{0}^1 \sum_{r=1}^{\infty}\sum_{s=1}^{\infty}\frac{x^{r+s-1}}{rs} dx$$ $$=\int_{0}^{1}\frac{\log^2(1-x)}{x} dx= \int_{0}^{1} \frac{\log^2 t}{1-t}dt= \int_{0}^{\infty} \frac{u^2}{e^u-1} du=2\zeta(3)=2\sum_{k=1}^{\infty}\frac{1}{k^3}.$$

The question is: what are other ways of proving the above summation without using integration?

Z Ahmed
  • 46,319

1 Answers1

2

We group terms in the double sum by setting $n=r+s$. For each $n\geq2$, the pairs of positive integers $(r,s)$ satisfying $r+s=n$ are exactly those with $r=1$ to $n−1$, and $s=n−r$. With this we rewrite the double sum as a single sum over $n$ with an inner sum over $r$.

$$ \sum_{r=1}^\infty \sum_{s=1}^\infty \frac{1}{rs(r+s)} = \sum_{n=2}^\infty \sum_{r=1}^{n-1} \frac{1}{r(n - r)n} $$

We can then manipulate the expression

$$ \frac{1}{r(n - r)} = \frac{1}{n} \left( \frac{1}{r} + \frac{1}{n - r} \right) $$

So we get

$$ \sum_{r=1}^{n-1} \frac{1}{r(n - r)} = \frac{1}{n} \sum_{r=1}^{n-1} \left( \frac{1}{r} + \frac{1}{n - r} \right) = \frac{2}{n} \sum_{r=1}^{n-1} \frac{1}{r} = \frac{2H_{n-1}}{n} $$

Therefore we can write that

$$ \sum_{r=1}^\infty \sum_{s=1}^\infty \frac{1}{rs(r+s)} = \sum_{n=2}^\infty \frac{1}{n} \cdot \frac{2H_{n-1}}{n} = 2 \sum_{n=2}^\infty \frac{H_{n-1}}{n^2} $$

We now shift the index by letting $m = n - 1$, so:

$$ 2 \sum_{n=2}^\infty \frac{H_{n-1}}{n^2} = 2 \sum_{m=1}^\infty \frac{H_m}{(m+1)^2} $$

It is known(1.12, page 3) that

$$ \sum_{m=1}^\infty \frac{H_m}{(m+1)^2} = \zeta(3) $$

Hence, we can conclude that

$$ \sum_{r=1}^{\infty} \sum_{s=1}^{\infty} \frac{1}{rs(r+s)} = 2\zeta(3) $$

  • But can the series $\sum_{m=1}^\infty \frac{H_m}{(m+1)^2}$ be calculated at the precalculus level? Looking at just the paper you linked, I'm not so sure... – David H May 31 '25 at 18:58
  • @DavidH I believe that at the pre-calc level, the best that can be done is to introduce students to the ideas of convergence and harmonic numbers. In regards to the series, it would just be them "knowing" the result $$\sum_{m=1}^\infty \frac{H_m}{(m+1)^2} = \zeta(3)$$ – Bumblebee Jun 01 '25 at 00:21