0

Compute the integral $$\int_{-\infty}^\infty \frac{e^x}{1 + e^{4x}}dx$$

My attempt: I tried applying residue integration. There are four simple poles in the upper half plane: $z \in \{\pi/4, 3\pi/4, 5\pi/4, 7\pi/4\}$. However evaluating all of these singularities and summing them gives the value $0$ due to symmetry, so I get that the integral evaluates to $0$ which is incorrect. Not sure what I'm missing in my approach.

Any help is appreciated.

Pame
  • 1,033
  • 7
    You may first substitute $u=e^x$ to convert the integrand to a rational function and then try to compute that instead. The resulting rational function will have two simple poles above the real line and two below the real line. – Sangchul Lee Nov 18 '21 at 20:14
  • @SangchulLee I'll give that a try. I don't see what is wrong with my current approach though. – Pame Nov 18 '21 at 20:22
  • 4
    The current integrand has infinitely many simple poles at $(2k+1)\pi i/4$ for $k \in \mathbb{Z}$. – Sangchul Lee Nov 18 '21 at 20:23
  • 1
    Wolfy gives $\pi/(2\sqrt{2})$ for the definite integral and (after some manipulation), $\dfrac{ \log\left(1+\dfrac{2\sqrt{2} e^x}{-\sqrt{2} e^x + e^{2 x} + 1}\right)}{4 \sqrt{2}} +\dfrac{ \arctan\left(\dfrac{\sqrt{2} e^x}{1-e^{2x}}\right)}{2 \sqrt{2}}$ for the indefinite integral. – marty cohen Nov 18 '21 at 20:44
  • I concur with SangchulLee. You claim in your post there are four simple poles in the upper half plane, which is just not true. There are countably many. Your incorrect assumption regarding the number of poles is probably what resulted in the wrong final computation of the integral. – Angel Nov 18 '21 at 20:44
  • Also, following SangchulLee's suggestion, you get a much simpler integral that only requires you to consider one of the four poles in the entire plane. – Angel Nov 18 '21 at 20:46
  • I still wonder if we could substitute for $\tan\theta$? – Darshan P. Nov 18 '21 at 21:38
  • @SangchulLee Your substitution gives $\int\frac 1{1+u^4}$ we have seen this kind of integral, Don't we? I think we are good to go! – Darshan P. Nov 18 '21 at 21:40
  • @I_don't_know_maths_, Indeed, that integral can be tackled by many means, such as employing the tricky decomposition $$\frac{1}{1+u^4}=\frac{1}{2}\left(\frac{1-u^{-2}}{(u+u^{-1})^2-2}+\frac{1+u^{-2}}{(u-u^{-1})^2+2}\right)$$ or simply performing residue calculation. – Sangchul Lee Nov 18 '21 at 21:46
  • @SangchulLee When I looked for the question it scared me then $e^x = u$ later $u = \sqrt\tan\theta \implies e^x =\sqrt\tan\theta $ So, your comments can solve problems – Darshan P. Nov 18 '21 at 21:58
  • @SangchulLee I thought only the poles in the interval $[0,2\pi]$ were relevant? Due to periodicity. – Pame Nov 19 '21 at 08:57
  • Considering only those poles means that you are considering a contour that only encompasses those four poles. Whether that contour will help you compute the integral is quite questionable, though. If you don't want to make any substitution, consider using the rectangular contour $\mathcal{C}_R$ with corners $\pm R$ and $\pm R + i\pi$ as $R\to\infty$. – Sangchul Lee Nov 19 '21 at 09:04
  • @SangchulLee Would you like to suggest a book so that can learn the Contour method to solve integrals, I have heard we can solve complicated&tough problems. – Darshan P. Nov 19 '21 at 12:09
  • @I_don't_know_maths_ Im not the one u asked, but we use Kreyzig Advanced Engineering Mathematics in our complex analysis class. You can download a pdf version for free online. – Pame Nov 21 '21 at 11:48
  • @Pame I have heard of it, Thanks a lot Pame! – Darshan P. Nov 21 '21 at 12:04

5 Answers5

2

As @2 is even prime proceeds, that method is a bit complicated, imo. Here is a similar alternative.

Without separating the integral, we substitute $e^x=t$.

$$I = \int\limits_0^\infty \frac{\mathrm dt}{1+t^4}$$

Now we substitute $t\mapsto \frac1t $.

$$I = \int\limits_0^\infty \frac{t^2}{1+t^4}\,\mathrm dt$$

Adding the two equations,

$$\begin{align} I &= \frac12 \int\limits_0^\infty\frac{1+t^2}{1+t^4}\,\mathrm dt \\ &= \frac12 \int\limits_0^\infty \frac{1+\frac1{t^2}}{(t-\frac1t)^2+2}\,\mathrm dt \\ &\overset{t-1/t =y}{=} \frac12 \int\limits_{-\infty}^\infty \frac{\mathrm dy}{y^2+2} \\ &= \left. \frac1{2\sqrt 2}\arctan \Big(\frac y{\sqrt 2}\Big) \right|_{-\infty}^\infty \\ I &= \frac\pi{2\sqrt2} \end{align}$$

0

$$\int_{-\infty}^{\infty}\frac {e^x}{1 + e^{4x}}dx$$ Substitute: $ e^x = \sqrt{\tan\theta} \implies e^xdx = \frac {\sec^2{\theta}d\theta}{2\sqrt{\tan\theta}}$

$$\begin{align*} \frac12\int_0^{\frac\pi2} \sin^{-\frac12}\theta\cos^{\frac12}\theta d\theta & = \frac12\frac12\beta(\frac14, \frac34)\\ & = \frac14\frac\pi{\sin(\frac\pi4)} \end{align*}$$ Which is the Beta function

Darshan P.
  • 1,181
0

$$\int_{-\infty}^\infty \frac{e^x}{1 + e^{4x}}dx$$ Split the integral at $0$ and bring them between same bounds i.e $[0,\infty]$. $$\int_{0}^{\infty}\frac{e^{3x}+e^{x}}{e^{4x}+1} dx$$ Substitute $u=e^{x}\implies dx=\frac{du}{u}$. $$\int_{1}^{\infty}\frac{u^2+1}{u^{4}+1} du$$ Substitute $k=u^2\implies du=\frac{dk}{2\sqrt{k}}$. $$\frac{1}{2}\int_{1}^{\infty}\frac{\sqrt{k}+\frac{1}{\sqrt{k}}}{k^2+1} dk$$ Put, $ k=\frac{1}{t} \implies dk=\frac{-dt}{t^2} $ $$\int_{0}^{1}\frac{\sqrt{t}+\frac{1}{\sqrt{t}}}{t^{2}+1} dt$$ This step allows us to use formula for infinite geometric sums but the problem is at the boundary condition. To solve that take a limit to $1$ from negative as follows, $$\lim_{a\to 1^{-}} \int_{0}^{a}\frac{\sqrt{t}+\frac{1}{\sqrt{t}}}{t^{2}+1} dt$$ Write, $\frac{1}{t^2+1}=\sum_{n\ge 0}t^{2n}$. $$\lim_{a\to 1^{-}} \sum_{n\ge 0}\int_{0}^{a} t^{2n-\frac{1}{2}}+t^{2n+\frac{1}{2}} dt$$ I think you can proceed from here.

0

We start with the substitution $e^x=t$. This yields

$$I = \int\limits_0^\infty\frac{\mathrm dt}{1+t^4} $$

Now using $t^4=y$,

$$\begin{align}I &= \frac14\int\limits_0^\infty \frac{ y^{1/4-1}}{1+y}\,\mathrm dy \\ &= \frac14\mathrm B(1/4,3/4) \\ &= \frac{ \Gamma(1/4)\Gamma(3/4)}{\Gamma(1)}\\ &= \frac{\pi}{4\sin\frac{\pi}4} \\ I &= \frac{\pi}{2\sqrt2}\end{align}$$

0

$$ \begin{aligned} I &=\int_{-\infty}^{\infty} \frac{d y}{1+y^{4}}, \quad \text{where } y=e^{x} \\ &=\int_{-\infty}^{\infty} \frac{\frac{1}{y^{2}}}{y^{2}+\frac{1}{y^{2}}} d y \\ &=\frac{1}{2} \int_{-\infty}^{\infty} \frac{\left(1+\frac{1}{y^{2}}\right)-\left(1+\frac{1}{y^{2}}\right)}{y^{2}+\frac{1}{y^{2}}} d y \\ &=\frac{1}{2} \int_{-\infty}^{\infty} \frac{d\left(y-\frac{1}{y}\right)}{\left(y-\frac{1}{y}\right)^{2}+2}-\frac{1}{2} \int_{-\infty}^{\infty} \frac{d\left(y+\frac{1}{y}\right)}{\left(y+\frac{1}{y}\right)^{2}-2} \\ &=\frac{1}{2 \sqrt{2}}\left[\tan ^{-1}\left(\frac{y-\frac{1}{y}}{\sqrt{2}}\right)\right]_{-\infty}^{\infty}-\frac{1}{2 \sqrt{2}}\left[\ln \left| \frac{y+\frac{1}{y}-\sqrt{2}}{y+\frac{1}{y}+\sqrt{2}}\right|\right]_{-\infty}^{\infty}\\ &=\frac{\pi}{2 \sqrt{2}} \end{aligned} $$

Lai
  • 20,421