5

Show that $$\int_{0}^{\infty} \frac{x - \sin(x)}{x^3} \, dx = \frac{\pi}{4}$$

My attempt is as follows: Let $$f(z) = \frac{z - i e^{iz}}{z^3}$$ and consider the contour on $[\epsilon, R]$ followed by a semicircular arc in the counter clockwise direction, then on $[-R, -\epsilon]$, then the semicircular clockwise contour avoiding the origin. We have, then, that

$$0 = \int_{\Gamma} f(z) dz = \int_{[\epsilon, R]} f(t) dt + \int_{C_R}f(Re^{it})Rie^{it}dt + \int_{[-R, -\epsilon]}f(t)dt + \int_{C_{\epsilon}}f(\epsilon e^{-it})\epsilon i e^{-it} dt$$

Then the first and third integrals ( $I_1$ and $I_3$) combine so that

$$I_1 + I_3 = 2\int_{\epsilon}^R \frac{t - \sin{t}}{t^3}\,dt$$

Further,

$$|I_{C_R}| \leq \int_0^\pi \left|\frac{Re^{it} - ie^{-R\sin{t}}e^{iRcos{t}}}{R^2 e^{2it}} \right|dt \rightarrow 0 \text{ as } R\rightarrow \infty$$

(I've omitted the details, it isn't too bad to bound)

However, I'm having trouble computing the limit

$$\lim_{\epsilon \rightarrow 0}\int_{C_{\epsilon}} f(\epsilon e^{-it})\epsilon i e^{-it} dt$$

No matter which way I look at it, it seems like this limit does not exist. Perhaps I'm seeing something wrong or have I chosen a bad $f(z)$?

Did
  • 279,727
DaveNine
  • 2,071
  • Your choice of function $f$ is odd. First, how is $$f(z) = \frac{z - i e^{iz}}{z^3}$$ even related to $$\frac{x-\sin x}{x^3}\ ?$$ The real parts when $z$ is real positive are not the same, right? Second, as you realize at the end, you need a bounded function near $z=0$, thus the expansion of your numerator as a power series at $z=0$ should have no $1$, $z$, $z^2$ term. Thus, it seems a natural choice of numerator would be $$z-\sin z=z-\frac{e^{iz}-e^{-iz}}{2i}$$ which is indeed $$\frac{z^3}6+o(z^3)$$ – Did Aug 16 '17 at 07:06
  • I chose that in order to make sure I get an integral that we want to compute along the real axes, but you're right, 0 is a problem. I'm thinking more and more to try a similar style of computation as in this post: https://math.stackexchange.com/questions/656757/evaluating-int-0-infty-fracx3-sin3xx5-dx-using-cont?rq=1 – DaveNine Aug 16 '17 at 07:13

6 Answers6

6

METHODOLOGY $1$: Straightforward Approach

We begin by letting $I$ be the integral of interest given by

$$\begin{align} I&=\int_0^\infty \frac{x-\sin(x)}{x^3}\,dx\\\\ &=\frac12 \text{Re}\left(\lim_{\varepsilon\to 0^+,R\to \infty}\left(\int_{-R}^{-\varepsilon} \frac{x+ie^{ix}}{x^3}\,dx+\int_\varepsilon^R \frac{x+ie^{ix}}{x^3}\,dx\right)\right) \end{align}$$


Next, we analyze the contour integral $J_{\varepsilon,R}$

$$\begin{align} J_{\varepsilon,R}&=\oint_{C_{\varepsilon,R}}\frac{z+ie^{iz}}{z^3}\,dz\\\\ &=\int_{-R}^{-\varepsilon} \frac{x+ie^{ix}}{x^3}\,dx+\int_\varepsilon^R \frac{x+ie^{ix}}{x^3}\,dx\\\\ &+\int_\pi^0 \frac{\varepsilon e^{i\phi}+ie^{i\varepsilon e^{i\phi}}}{(\varepsilon e^{i\phi})^3}\,i\varepsilon e^{i\phi}\,d\phi+\int_0^\pi \frac{Re^{i\phi}+ie^{iR e^{i\phi}}}{(R e^{i\phi})^3}\,iR e^{i\phi}\,d\phi \end{align}$$


Expanding $e^{i\varepsilon e^{i\phi}}$ as

$$e^{i\varepsilon e^{i\phi}}=1+i\varepsilon e^{i\phi}-\frac12 \varepsilon^2e^{i2\phi}+O\left(\varepsilon^3\right)$$

reveals that the integration over the semicircle of radius $\epsilon$ is

$$\begin{align} \int_\pi^0 \frac{\varepsilon e^{i\phi}+ie^{i\varepsilon e^{i\phi}}}{(\varepsilon e^{i\phi})^3}\,i\varepsilon e^{i\phi}\,d\phi&=\frac1{\varepsilon^2}\underbrace{\int_0^\pi e^{-i2\phi}\,d\phi}_{=0}-\frac12 \int_0^\pi (1)\,d\phi +O(\varepsilon)\\\\ &=-\frac\pi2 +O(\varepsilon) \end{align}$$


Furthermore, it is easy to show that the integration over the semi-circle of radisu $R$ is

$$\begin{align} \int_0^\pi \frac{Re^{i\phi}+ie^{iR e^{i\phi}}}{(R e^{i\phi})^3}\,iR e^{i\phi}\,d\phi=O\left(\frac1R\right) \end{align}$$


Since $\frac{z+ie^{iz}}{z^3}$ is analytic in and on $C_{\varepsilon,R}$, Cauchy's integral theorem guarantees that $J_{\varepsilon,R}=0$. Putting everything together, we see that

$$\begin{align} 0&=J_{\varepsilon,R}\\\\ &=\int_{-R}^{-\varepsilon} \frac{x+ie^{ix}}{x^3}\,dx+\int_\varepsilon^R \frac{x+ie^{ix}}{x^3}\,dx\\\\ &-\frac\pi2+O\left(\varepsilon\right)+\left(\frac1R\right) \end{align}$$

whereupon taking the limit as $\varepsilon\to 0^+$ and $R\to \infty$ yields

$$I=\frac\pi4$$

And we are done!


METHODOLOGY $2$: Simplifying Using Integration by Parts

We can make our life much easier if we apply successive integration by parts. We now proceed accordingly.

Let $I$ be the integral given by

$$\begin{align} I&=\int_0^\infty \frac{x-\sin(x)}{x^3}\,dx\tag1 \end{align}$$

Integrating by parts the integral on the right-hand side of $(1)$ with $u=x-\sin(x)$ and $v=-\frac{1}{2x^2}$, we find that

$$I=\frac12\int_0^\infty \frac{1-\cos(x)}{x^2}\,dx \tag2$$

Integrating by parts the integral on the right-hand side of $(2)$ with $u=1-\cos(x)$ and $v=-\frac1x$ reveals

$$\begin{align} I&=\frac12 \int_0^\infty \frac{\sin(x)}{x}\,dx\tag3 \end{align}$$

We will evaluate the integral in $(3)$ using contour integration.


We analyze the contour integral $J(\varepsilon,R)$, where $R>0$ and $\varepsilon>0$, as given by

$$\begin{align} J(\varepsilon,R)&=\oint_{C_{\varepsilon,R}}\frac{e^{iz}}{z}\,dz\\\\ &=\int_{-R}^{-\varepsilon} \frac{e^{ix}}{x}\,dx+\int_\pi^0 \frac{e^{i\varepsilon e^{i\phi}}}{\varepsilon e^{i\phi}}\,i\varepsilon e^{i\phi}\,d\phi+\int_{\varepsilon}^R \frac{e^{ix}}{x}\,dx+\int_\pi^0 \frac{e^{iR e^{i\phi}}}{R e^{i\phi}}\,iR e^{i\phi}\,d\phi\tag4 \end{align}$$

Since $\frac{e^{iz}}{z}$ is analytic in and on the contour defined by $C_{\varepsilon,R}$, Cauchy's Integral Theorem guarantees that $J(\varepsilon,R)=0$.

First, note from symmetry that

$$\int_{-R}^{-\varepsilon} \frac{e^{ix}}{x}\,dx+\int_{\varepsilon}^R \frac{e^{ix}}{x}\,dx=i2\int_{\varepsilon}^R \frac{\sin(x)}{x}\,dx$$

Furthermore, we have

$$\lim_{\varepsilon\to 0,R\to \infty}\int_{\varepsilon}^R \frac{\sin(x)}{x}\,dx=\int_0^\infty \frac{\sin(x)}{x}\,dx\tag5$$


Second, it is easy to see that

$$\lim_{\varepsilon\to 0}\int_\pi^0 \frac{e^{i\varepsilon e^{i\phi}}}{\varepsilon e^{i\phi}}\,i\varepsilon e^{i\phi}\,d\phi=-i\pi \tag6$$


Third, noting that $\sin(\phi)\ge \frac{2\phi}{\pi}$ for $\phi\in [0,\pi/2]$, we see that

$$\begin{align} \left|\int_\pi^0 \frac{e^{iR e^{i\phi}}}{R e^{i\phi}}\,iR e^{i\phi}\,d\phi\right|&=\left|\int_0^\pi ie^{iR\sin(\phi)}e^{-R\cos(\phi)}\right|\\\\ &\le\int_0^\pi e^{-R\cos(\phi)}\,d\phi\\\\ &=2\int_0^{\pi/2}e^{-R\sin(\phi)}\,d\phi\\\\ &\le 2\int_0^{\pi/2}e^{-2R\phi/\pi}\,d\phi\\\\ &=\frac{\pi(1-e^{-R})}{R} \end{align}$$

Hence, we see that

$$\lim_{R\to \infty}\int_\pi^0 \frac{e^{iR e^{i\phi}}}{R e^{i\phi}}\,iR e^{i\phi}\,d\phi=0\tag 7$$


Finally, using $(5)-(7)$ in $(4)$ yields

$$\int_0^\infty \frac{\sin(x)}{x}\,dx=\frac{\pi}{2}$$

whence we find that

$$\int_0^\infty \frac{x-\sin(x)}{x^3}\,dx=\frac\pi4$$

Mark Viola
  • 179,405
  • +1, Mark , Do you know any theorems to evaluate vanishing contours around poles of any order ? I know there is one for simple pole. – Zaid Alyafeai Aug 17 '17 at 02:35
  • @zaidalyafeai Hi Zaid. Thank you for the up vote. I'm uncertain as to the meaning of your question. Help me understand. If we integrate a function $f$ on a closed contour, the only term in the Laurent series that impacts the integral's value is the first irder pole. – Mark Viola Aug 17 '17 at 02:41
  • Suppose we are integrating on the real axis and the function had a pole of order m that is on the real axis. Something like this theorem https://math.stackexchange.com/questions/83828/definite-integral-calculation-with-poles-at-0-and-pm-i-sqrt3/184874#184874 but for higher order poles. – Zaid Alyafeai Aug 17 '17 at 02:47
4

Here is a suggestion. Consider the function $f$ defined by $$f(z) = \frac{1+iz-e^{iz}}{z^3}.$$ On the real line, the imaginary part of $f$ will give you the integral you want and with this function the computation with residues will work.

C. Dubussy
  • 9,120
3

I guess that using the inverse Laplace transform counts as contour integration. In such a case:

$$ \int_{0}^{+\infty}\frac{x-\sin x}{x^3}\,dx = \int_{0}^{+\infty}\mathcal{L}(x-\sin x)(s)\,\mathcal{L}^{-1}\left(\frac{1}{x^3}\right)(s)\,ds$$ by an important property of the Laplace transform. The RHS equals: $$ \int_{0}^{+\infty}\left(\frac{1}{s^2}-\frac{1}{1+s^2}\right)\frac{s^2}{2}\,ds =\frac{1}{2}\int_{0}^{+\infty}\frac{ds}{1+s^2}=\color{blue}{\frac{\pi}{4}}.$$

Jack D'Aurizio
  • 353,855
0

If you denote by $\operatorname{Si}x$ the sinus integral of $x$, that is $$\operatorname{Si}x\colon = \int_0^x \frac{\sin t}{t} dt$$ then we have the equality for indefinite integrals $$\int \frac{x-\sin x}{x^3} = \frac{\operatorname {Si}x}{2} + \frac{\sin x + x \cos x - 2 x}{2 x^2}+ \mathcal{C} $$

Notice that the second term has limit $0$ at $0$ ( so removable singularity) and $\infty$. Moreover, $\operatorname{Si}(0)=0$ (clearly) and $\lim_{x \to \infty} \operatorname{Si} (x) = \int_{0}^{\infty} \frac{\sin t}{t} d t = \frac{\pi}{2}$ ( that is standard, proved using residues). We conclude:

$$\int_{0}^{\infty} \frac{x - \sin x}{x^3} d x =\frac{ \pi}{4}$$

orangeskid
  • 53,909
0

$\newcommand{\bbx}[1]{\,\bbox[15px,border:1px groove navy]{\displaystyle{#1}}\,} \newcommand{\braces}[1]{\left\lbrace\,{#1}\,\right\rbrace} \newcommand{\bracks}[1]{\left\lbrack\,{#1}\,\right\rbrack} \newcommand{\dd}{\mathrm{d}} \newcommand{\ds}[1]{\displaystyle{#1}} \newcommand{\expo}[1]{\,\mathrm{e}^{#1}\,} \newcommand{\ic}{\mathrm{i}} \newcommand{\mc}[1]{\mathcal{#1}} \newcommand{\mrm}[1]{\mathrm{#1}} \newcommand{\on}[1]{\operatorname{#1}} \newcommand{\pars}[1]{\left(\,{#1}\,\right)} \newcommand{\partiald}[3][]{\frac{\partial^{#1} #2}{\partial #3^{#1}}} \newcommand{\root}[2][]{\,\sqrt[#1]{\,{#2}\,}\,} \newcommand{\totald}[3][]{\frac{\mathrm{d}^{#1} #2}{\mathrm{d} #3^{#1}}} \newcommand{\verts}[1]{\left\vert\,{#1}\,\right\vert}$ With $\ds{0 < \epsilon < R}$, I'll perform an integration along a quarter circle -of radius $\ds{R}$- in the complex plane first quadrant which include an indent -an arc of radius $\ds{\epsilon}$- about the origin of coordinates. Namely, \begin{align} &\int_{\epsilon}^{R}{\expo{\ic x} - 1 - \ic x \over x^{3}}\,\dd x \,\,\,\stackrel{\mrm{as}\ \epsilon\ \to\ 0^{+}}{\sim}\,\,\, \left.-\int_{0}^{\pi/2}{\expo{\ic z} - 1 - \ic z \over z^{3}}\,\dd z \,\right\vert_{z\ =\ R\exp\pars{\ic\theta}} \\[2mm] - &\ \int_{R}^{\epsilon}{\expo{-y} - 1 + y \over -\ic y^{3}}\,\,\ic\,\dd y - \int_{\pi/2}^{0}{-\epsilon^{2}\expo{2\ic\theta}/2 \over \epsilon^{3}\expo{3\ic\theta}}\,\epsilon\expo{\ic\theta}\ic\,\dd\theta \end{align} The RHS first integral vanishes out as $\ds{R \to \infty}$. Then, \begin{align} &\int_{\epsilon}^{\infty}{\expo{\ic x} - 1 - \ic x \over x^{3}}\,\dd x \,\,\,\stackrel{\mrm{as}\ \epsilon\ \to\ 0^{+}}{\sim}\,\,\, -\int_{\epsilon}^{\infty}{\expo{-y} - 1 + y \over y^{3}}\,\dd y - {1 \over 2}\ic\int_{0}^{\pi/2}\dd\theta \\[5mm] &\int_{0}^{\infty}{\sin\pars{x} - x \over x^{3}}\,\dd x = - {\pi \over 4}\quad\mrm{as}\ \epsilon \to 0^{+} \\[5mm] & \bbx{\int_{0}^{\infty}{x - \sin\pars{x} \over x^{3}}\,\dd x = {\pi \over 4}} \\ & \end{align}

Felix Marin
  • 89,464
0

We can evaluate the integral using Feynman's trick:

$$ I = \int_0^\infty \frac{x - \sin(x)}{x^3}dx = I(1) $$

$$ I(a)= \int_0^\infty \frac{ax - \sin(ax)}{x^3}dx $$

$$ \frac{dI(a)}{da}= \int_0^\infty \frac{\partial}{\partial a} \Bigg(\frac{ax - \sin(ax)}{x^3}\Bigg)dx = \int_0^\infty \frac{1 - \cos(ax)}{x^2}dx $$

$$ \frac{d^2I(a)}{da^2}= \int_0^\infty \frac{\partial}{\partial a} \Bigg(\frac{1 - \cos(ax)}{x^2}\Bigg)dx = \int_0^\infty \frac{\sin(ax)}{x}dx = \frac{\pi}{2}$$

$$ \frac{dI(a)}{da} = \frac{\pi}{2}a+C_1 $$ $$ \frac{dI(0)}{da} = 0 = 0+C_1 \to C_1 = 0$$ $$ I(a) = \frac{\pi}{4}a^2+C_2 $$ $$ I(0) = 0 = 0+C_2 \to C_2 = 0 $$ $$ I = I(1) = \frac{\pi}{4} $$