EDIT: Instead of expressing the integral as the imaginary part of another integral, I instead expanded $\sin^{3}(x)$ in terms of complex exponentials and I don't run into problems anymore.

\begin{align} \int_{0}^{\infty} \frac{x^{3}-\sin^{3}(x)}{x^{5}} \ dx &= \frac{1}{2} \int_{-\infty}^{\infty} \frac{x^{3}-\sin^{3}(x)}{x^{5}} \ dx \\ &= \frac{1}{2} \ \int_{-\infty}^{\infty} \frac{x^{3}+\frac{1}{8i}(e^{3ix}-3e^{ix}+3e^{-ix}-e^{-3ix})}{x^{5}} \ dx \\ &= \frac{1}{2} \lim_{\epsilon \to 0^{+}} \ \int_{-\infty}^{\infty} \frac{x^{3}+\frac{1}{8i}(e^{3ix}-3e^{ix}+3e^{-ix}-e^{-3ix})}{(x-i \epsilon)^{5}} \ dx \\ &= \frac{1}{2} \lim_{\epsilon \to 0^{+}} \int_{-\infty}^{\infty} \frac{x^{3}+\frac{1}{8i} (e^{3ix}-3e^{ix})}{(x-i \epsilon)^{5}} + \frac{1}{16i} \lim_{\epsilon \to 0^{+}} \int_{-\infty}^{\infty} \frac{3e^{-ix}-e^{-3ix}}{(x-i \epsilon)^{5}} \ dx \end{align}

Then I integrated $ f(z) = \frac{z^{3}+ \frac{1}{8i}(e^{3iz}-3e^{iz})}{(z-i \epsilon)^{5}}$ around the upper half of $|z|=R$ and $ g(z) = \frac{3e^{-iz}-e^{-3iz}}{(z-i \epsilon)^{5}}$ around the lower half of $|z|=R$ and applied Jordan's lemma.

\begin{align} \int_{0}^{\infty} \frac{x^{3}-\sin^{3}x}{x^{5}} \ dx &= \frac{1}{2} \lim_{\epsilon \to 0^{+}}2 \pi i \ \text{Res}[f(z),i \epsilon] + \frac{1}{16i} \lim_{\epsilon \to 0^{+}} 2 \pi i (0) \\ &= \frac{1}{2} \lim_{\epsilon \to 0^{+}} \frac{2 \pi i}{4!} \lim_{z \to i \epsilon} \frac{d^{4}}{dz^{4}} \Big(z^{3}+\frac{1}{8i}e^{3iz}-\frac{3}{8i}e^{iz} \Big) \\ &= \frac{\pi i}{24} \lim_{\epsilon \to 0^{+}} \ \lim_{z \to i \epsilon}\Big( \frac{1}{8i}(3i)^{4}e^{3iz}- \frac{3}{8i} (i)^{4} e^{iz} \Big) \\ &= \frac{\pi i}{24} \lim_{\epsilon \to 0^{+}} \Big( \frac{81}{8i}e^{- 3\epsilon} - \frac{3}{8i}e^{- \epsilon} \Big) \\ &= \frac{\pi i}{24} \Big(\frac{81}{8i}-\frac{3}{8i} \Big) \\ &= \frac{13 \pi}{32} \end{align}

  • $\begingroup$ (I'm assuming you're using a semicircular contour). A somewhat related question: how would one justify that $\lim_{R \to \infty}\left|\int_0^\pi \frac{e^{i3Re^{it}}-3e^{iRe^{it}}}{(Re^{it}-i \epsilon)^5}\right|=0$? $\endgroup$ – Meow Jan 31 '14 at 10:28
  • $\begingroup$ So is there no question left? I don't see one. $\endgroup$ – robjohn Jul 24 '14 at 21:45
  • $\begingroup$ @robjohn Not long after I posted this question earlier this year, I realized why my original approach was wrong. So I modified my approach and explained what I did in the original post. Another user bumped this thread today, and when I looked at my post, I realized it could be made a lot clearer. So I edited it. If you have another way to evaluate it using contour integration, please post it. EDIT: I see you already have. :) $\endgroup$ – Random Variable Jul 24 '14 at 22:33
  • $\begingroup$ @RandomVariable: I don't think my approach is any different, now that I look more closely at yours. I should just delete mine. $\endgroup$ – robjohn Jul 24 '14 at 22:38
  • $\begingroup$ @robjohn Isn't the difference that you shifted the contour while I moved the pole? $\endgroup$ – Random Variable Jul 24 '14 at 22:41

I like to calculate this integral as follows:

Let us note that

$$\frac{1}{x^5}=\frac{1}{4!}\int_0^\infty t^4e^{-xt}dt$$ So

$$I=\frac{1}{4!}\int_{0}^{\infty}(x^{3}-\sin^{3}x)\int_0^\infty t^4e^{-xt}\;dt\;dx$$


$$=\frac{1}{4!} \int_{0}^{\infty}t^4\left [\frac{6}{t^4}-\frac{6}{(t^2+1)(t^2+9)}\right ]dt$$


  • $\begingroup$ I considered doing it like that, but the problem came from a complex analysis textbook. But then again, you could evaluate that last integral by contour integration. Maybe that's what the authors had in mind. I don't know. Thanks. $\endgroup$ – Random Variable Feb 2 '14 at 15:55

$\newcommand{\+}{^{\dagger}} \newcommand{\angles}[1]{\left\langle\, #1 \,\right\rangle} \newcommand{\braces}[1]{\left\lbrace\, #1 \,\right\rbrace} \newcommand{\bracks}[1]{\left\lbrack\, #1 \,\right\rbrack} \newcommand{\ceil}[1]{\,\left\lceil\, #1 \,\right\rceil\,} \newcommand{\dd}{{\rm d}} \newcommand{\down}{\downarrow} \newcommand{\ds}[1]{\displaystyle{#1}} \newcommand{\expo}[1]{\,{\rm e}^{#1}\,} \newcommand{\fermi}{\,{\rm f}} \newcommand{\floor}[1]{\,\left\lfloor #1 \right\rfloor\,} \newcommand{\half}{{1 \over 2}} \newcommand{\ic}{{\rm i}} \newcommand{\iff}{\Longleftrightarrow} \newcommand{\imp}{\Longrightarrow} \newcommand{\isdiv}{\,\left.\right\vert\,} \newcommand{\ket}[1]{\left\vert #1\right\rangle} \newcommand{\ol}[1]{\overline{#1}} \newcommand{\pars}[1]{\left(\, #1 \,\right)} \newcommand{\partiald}[3][]{\frac{\partial^{#1} #2}{\partial #3^{#1}}} \newcommand{\pp}{{\cal P}} \newcommand{\root}[2][]{\,\sqrt[#1]{\vphantom{\large A}\,#2\,}\,} \newcommand{\sech}{\,{\rm sech}} \newcommand{\sgn}{\,{\rm sgn}} \newcommand{\totald}[3][]{\frac{{\rm d}^{#1} #2}{{\rm d} #3^{#1}}} \newcommand{\ul}[1]{\underline{#1}} \newcommand{\verts}[1]{\left\vert\, #1 \,\right\vert} \newcommand{\wt}[1]{\widetilde{#1}}$ $\ds{\int_{0}^{\infty}{x^{3}- \sin^{3}\pars{x} \over x^{5}}\,\dd x:\ {\large ?}}$

Lets $\ds{\color{#00f}{\fermi\pars{x}} \equiv x^{3} - \sin^{3}\pars{x} = \color{#00f}{x^{3} + {1 \over 4}\,\sin\pars{3x} - {3 \over 4}\,\sin\pars{x} }\tag{1}}$.

$$ \mbox{The integral in question becomes}\quad \int_{0}^{\infty}{\fermi\pars{x} \over x^{5}}\,\dd x $$

In order to 'reduce' the $\ds{x^{-5}}$ power to a 'simple' $\ds{x^{-1}}$ power, we integrate by parts repeatedly: \begin{align} \color{#c00000}{\int_{0}^{\infty}{\fermi\pars{x} \over x^{5}}\,\dd x}&={1 \over 4}\int_{0}^{\infty}{\fermi'\pars{x} \over x^{4}}\,\dd x ={1 \over 12}\int_{0}^{\infty}{\fermi''\pars{x} \over x^{3}}\,\dd x={1 \over 24}\int_{0}^{\infty}{\fermi'''\pars{x} \over x^{2}}\,\dd x \\[3mm]&={1 \over 24}\int_{0}^{\infty}{\fermi^{\pars{\tt IV}}\pars{x} \over x} \,\dd x\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\quad\pars{2} \end{align}

From expression $\pars{1}$ we can evaluate $\ds{\fermi^{\pars{\tt IV}}\pars{x}}$: $$ \fermi^{\pars{\tt IV}}\pars{x} ={81 \over 4}\,\sin\pars{3x} - {3 \over 4}\,\sin\pars{x} $$

which is replaced in $\pars{2}$: \begin{align} \color{#c00000}{\int_{0}^{\infty}{\fermi\pars{x} \over x^{5}}\,\dd x}& ={27 \over 32}\int_{0}^{\infty}{\sin\pars{3x} \over x}\,\dd x -{1 \over 32}\int_{0}^{\infty}{\sin\pars{x} \over x}\,\dd x \\[3mm]&={27 \over 32}\int_{0}^{\infty}{\sin\pars{x} \over x}\,\dd x -{1 \over 32}\int_{0}^{\infty}{\sin\pars{x} \over x}\,\dd x \\[3mm]&=\underbrace{\pars{{27 \over 32} - {1 \over 32}}}_{\ds{13 \over 16}}\ \underbrace{\int_{0}^{\infty}{\sin\pars{x} \over x}\,\dd x}_{\ds{\pi \over 2}} \end{align}

$$\color{#00f}{\large% \int_{0}^{\infty}{x^{3}- \sin^{3}\pars{x} \over x^{5}}\,\dd x ={13 \over 32}\,\pi} \approx {\tt 1.2763} $$

  • $\begingroup$ The answer is accurate (+1), but it would really be nice to see more verbal explanation to your answers. Just a mention that you are integrating by parts 4 times would be helpful. Such a terse answer is sometimes hard to follow. $\endgroup$ – robjohn Jul 24 '14 at 22:44
  • $\begingroup$ @robjohn I just followed your advice. Thanks. $\endgroup$ – Felix Marin Jul 25 '14 at 2:16
  • $\begingroup$ How did the $\sin^3 x$ become $\frac{1}{4} \sin 3x - \frac{3}{4} \sin x$? $\endgroup$ – Jeel Shah Jul 25 '14 at 7:28

Here is another contour integration approach. Note that the integrand is even, so using the contours $\gamma=[-R,R]\cup Re^{i[0,\pi]}$ as $R\to\infty$ and $\beta=[-R,R]\cup Re^{-i[0,\pi]}$, we have $$ \begin{align} \int_0^\infty\frac{x^3-\sin^3(x)}{x^5}\mathrm{d}x &=\frac12\int_{-\infty}^\infty\frac{x^3-\sin^3(x)}{x^5}\mathrm{d}x\tag{1}\\ &=\lim_{\epsilon\to0^+}\frac12\int_{-\infty-i\epsilon}^{\infty-i\epsilon}\frac{z^3-\sin^3(z)}{z^5}\mathrm{d}z\tag{2}\\ &=\lim_{\epsilon\to0^+}\frac12\int_{\gamma-i\epsilon}\frac{\frac1{8i}(e^{3iz}-3e^{iz})}{z^5}\mathrm{d}z\tag{3}\\ &+\lim_{\epsilon\to0^+}\frac12\int_{\beta-i\epsilon}\frac{z^3+\frac1{8i}(3e^{-iz}-e^{-3iz})}{z^5}\mathrm{d}z\tag{4}\\ &=\lim_{\epsilon\to0^+}\frac1{16i}\int_{\gamma-i\epsilon}\frac{e^{3iz}-3e^{iz}}{z^5}\mathrm{d}z\tag{5}\\[3pt] &=\frac\pi8\frac1{4!}\left((3i)^4-3i^4\right)\tag{6}\\[6pt] &=\frac{13\pi}{32}\tag{7} \end{align} $$ Explanation:
$(1)$: integrand is even, duplicate the domain and divide by $2$
$(2)$: offset the path of integration since there are no singularities and the integrand decays at $\infty$
$(3)$: take some of the terms along the upper contour
$(4)$: take the rest along the lower contour
$(5)$: there are no singularities inside $\beta-i\epsilon$ and the respective integrands decay appropriately on the circular arcs
$(6)$: the residues depend on the $z^4$ terms in the expansions of the exponentials

  • $\begingroup$ Shoot... now that I look at this, I think this may be the same as RandomVariable's approach except that I moved $\frac{z^3}{z^5}$ to the lower contour. $\endgroup$ – robjohn Jul 24 '14 at 22:34
  • $\begingroup$ I've undeleted since Random Variable points out that I did use a different contour (however slightly different). $\endgroup$ – robjohn Jul 24 '14 at 22:47

Another approach :

Sorry Random Variable, this is not using contour integration technique since I don't know how to approach the integral using that way. $\ddot\smile$

Consider $$ \mathcal{I}(\alpha)=\int_0^\infty\frac{(\alpha x)^3-\sin^3\alpha x}{x^5}dx.\tag1 $$ Differentiating $(1)$ four times yields \begin{align} \frac{d^4\mathcal{I}}{d\alpha^4}&=\int_0^\infty\frac{\partial^4}{\partial\alpha^4}\left(\frac{(\alpha x)^3-\sin^3\alpha x}{x^5}\right)dx\\ &=\color{green}{\int_0^\infty\left(\frac{81\sin3\alpha x-3\sin\alpha x}{4x}\right)dx}\\ &=\frac{81}{4}\cdot\frac\pi2-\frac{3}{4}\cdot\frac\pi2\\ &=\frac{39\pi}{4},\tag2 \end{align} where $$ \int_0^\infty\frac{\sin\alpha x}{x}dx=\frac\pi2\qquad\text{for }\alpha\neq0. $$ Then from $(2)$ we obtain $$ \large\color{blue}{\mathcal{I}(\alpha)=\frac{13\pi}{32}a^4},\tag3 $$ where $\mathcal{I}(0)=\mathcal{I'}(0)=\mathcal{I''}(0)=\mathcal{I'''}(0)=0$. Thus

$$ \mathcal{I}(1)=\int_0^\infty\frac{x^3-\sin^3 x}{x^5}dx=\large\color{blue}{\frac{13\pi}{32}}. $$


I think you can easily apply the contour integration technique in line $2$ (green-colored) equation $(2)$.

  • $\begingroup$ What integration technique is this? $\endgroup$ – Jeel Shah Jul 25 '14 at 10:28
  • $\begingroup$ @gekkostate Differentiation under the integral sign a.k.a. Feynman's method. $\endgroup$ – Tunk-Fey Jul 25 '14 at 11:02
  • $\begingroup$ Are you aware of any tutorials which explain this method in detail? For example, how it works and the motivation behind it? Thanks! $\endgroup$ – Jeel Shah Jul 25 '14 at 11:44
  • 1
    $\begingroup$ @gekkostate You may refer here: $(1)$ and $(2)$. $\endgroup$ – Tunk-Fey Jul 25 '14 at 11:50

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.