2

can I evaluate this integration ?

$$\int_0^\infty\dfrac{x^2}{e^x-1}\mathrm{dx}$$

how? thank you

Nosrati
  • 29,995
Khaled
  • 21

2 Answers2

3

\begin{align} \int_0^\infty\dfrac{x^2}{e^x-1}\mathrm{dx} &= \int_0^\infty\dfrac{x^2e^{-x}}{1-e^{-x}}\mathrm{dx} \\ &= \int_0^\infty\mathrm{dx}~ x^2e^{-x}\sum_{n=0}^{\infty}e^{-nx} \\ &= \sum_{n=0}^{\infty}\int_0^\infty x^2e^{-(n+1)x}\mathrm{dx} \\ &= \sum_{n=0}^{\infty}\dfrac{1}{(n+1)^3}\Gamma(3) \\ &= \color{blue}{2\zeta(3)} \end{align}

Nosrati
  • 29,995
2

Here is an approach which avoids series and is based on the polylogarithm function.

We begin by rewriting the integral as $$I = \int_0^\infty \frac{x^2}{e^x - 1} \, dx = \int^\infty_0 \frac{x^2 e^{-x}}{1 - e^{-x}} \, dx.$$ Now as the polylogarithm function of order zero is given by $$\text{Li}_0 (x) = \frac{x}{1 - x},$$ if $x$ is replaced with $e^{-x}$ in the above expression, we have $$\text{Li}_0 (e^{-x}) = \frac{e^{-x}}{1 - e^{-x}},$$ and the integral can be rewritten in terms of the polylogarithm function as $$I = \int^\infty_0 x^2 \text{Li}_0 (e^{-x}) \, dx. \tag1$$

As the derivative for the polylogarithmic function of order $s$ is given by $$\frac{d}{dx} \text{Li}_s (x) = \frac{\text{Li}_{s - 1} (x)}{x},$$ one has $$\frac{d}{dx} \text{Li}_s (e^{-x}) = - \text{Li}_{s - 1} (e^{-x}),$$ and we immediately see that $$\int \text{Li}_s (e^{-x}) \, dx = - \text{Li}_{s + 1} (e^{-x}) + C. \tag2$$

We now make use of (2) and repeatedly integrate (1) by parts. Doing so yields \begin{align*} I &= 2 \int^\infty_0 x \text{Li}_1 (e^{-x}) \, dx \quad \text{(by parts)}\\ &= 2 \int^\infty_0 \text{Li}_2 (e^{-x}) \, dx \quad \text{(by parts again)}\\ &= 2 \, \text{Li}_3 (1). \end{align*}

Making use of the following result $$\text{Li}_s (1) = \zeta (s), \quad s > 1,$$ where $\zeta(x)$ is the Riemann zeta function we finally obtain $$\int^\infty_0 \frac{x^2}{e^x - 1} \, dx = 2 \zeta (3).$$

omegadot
  • 11,736
  • Why do you bother with $\text{Li}_s(z)$ instead of expanding $\frac{1}{e^x-1}$ directly ? In your other answer you use even more special functions. – reuns Dec 08 '17 at 19:34
  • I bother since the approach used here with polylogarithms is more general. In fact, the polylogarithmic approach allows one to find the indefinite integral, something not possible with the expansion method. Here $$\int \frac{x^2}{e^x - 1} , dx = -2 \text{Li}_3(e^{-x}) - 2x \text{Li}_2 (e^{-x}) - x^2 \text{Li}_1 (e^{-x})$$ Often where integrals of this type arise is in the study of blackbody radiation. Here integrals on a finite interval (say $[a,b]$ where $0 < a < b < \infty$) in addition to those on the interval $(0,\infty)$ are needed with the former being found in terms of polylogarithms. – omegadot Dec 09 '17 at 00:46