Skip to main content

Math 382-0: Kursobjekt

Section 1.23 Definite integrals

We assemble here a variety of complex integration techniques that can be used to compute definite integrals over real line intervals, both proper and improper.

Subsection Miscellaneous tools

Proof.

The inequality (1.83) is easily shown using calculus methods. (See Figure 1.23.2 for a fairly convincing visual argument.) We then have
\begin{align*} \int_0^{\pi}e^{-c\sin\theta}\, d\theta \amp =2\int_0^{\pi/2}e^{-c\sin\theta}\, d\theta \amp (\sin(\pi/2-t)=\sin(\pi/2+t))\\ \amp \leq 2\int_0^{\pi}e^{-2c\theta/\pi}\, d\theta \amp (\sin\theta\geq 2\theta/\pi)\\ \amp = -\frac{\pi}{c}(e^{-2c\theta/\pi})\Bigr\vert_0^{\pi/2}\\ \amp = \frac{\pi}{c}(1-e^{-c))\\ \amp \leq \frac{\pi}{c}\text{.} \end{align*}
Core idea of Jordan’s inequality
Figure 1.23.2. Core idea of Jordan’s inequality

Proof.

Subsection Trig integrals on \([0,2\pi]\)

The following theorem reduces the integral of any function \(F(\cos t, \sin t)\) over \([0,2\pi]\) to a complex line integral of a certain complex function \(G(z)\) over the unit circle (positively oriented). Of course, whether we can carry out that resulting line integral by hand depends on the complexity of \(F=F(x,y)\text{.}\) That computation is typically carried out using the Cauchy residue theorem.

Proof.

The proof is a simple matter of applying the definition of the line integral to right side of (1.86) and the fact that
\begin{align*} \frac{1}{2}(e^{it}+1/e^{it}) \amp = \cos t\\ \frac{1}{2i}(e^{it}-1/e^{it}) \amp = \sin t \text{.} \end{align*}

Example 1.23.5. Trig integral over \([0,2\pi]\).

Compute \(\displaystyle\int_{0}^{2\pi}\frac{1}{\cos t+\sin t+2}\, dt\text{.}\)
Solution.
We have Let \(I\) be the definite integral we wish to compute, let \(F(x,y)=1/(x+y+2)\text{,}\) and let
\begin{align*} f(z)=\frac{F(\tfrac{1}{2}(z+1/z),\tfrac{1}{2i}(z-1/z))}{iz}\amp = \frac{1}{iz}\cdot \frac{1}{\frac{1}{2}(z+1/z)+\frac{1}{2i}{z-1/z}+2} \\ \amp = \frac{2}{iz}\cdot \frac{1}{(z+1/z)+\frac{1}{i}{z-1/z}+4} \\ \amp = \frac{2}{iz^2+i+z^2-1+4iz}\\ \amp = \frac{2}{(i+1)z^2+4iz+(-1+i)} \end{align*}
Using the quadratic formula, the roots of denominator are
\begin{align*} \alpha \amp = \frac{-4i+2\sqrt{2}i}{2(1+i)} \amp \beta \amp = \frac{-4i-2\sqrt{2}i}{2(1+i)} \\ \amp =\frac{(-2+\sqrt{2})i}{1+i} \amp \amp =\frac{(-2-\sqrt{2})i}{1+i} \end{align*}
Looking at the moduli of \(\alpha\) and \(\beta\text{,}\) we see that \(\alpha\in B_1(0)\) and \(\beta\notin \overline{B}_1(0)\text{.}\) (In more detail, \(\abs{\alpha}=(2-\sqrt(2))/\sqrt{2} < 1\text{.}\)) Theorem 1.23.4 and the Cauchy residue formula then implies
\begin{align*} I =\int_\gamma f\, dz\amp =2\pi i \res_f \alpha\\ \amp =2\pi i \cdot \frac{2}{(1+i)(\alpha-\beta)} \amp (f(z)=\tfrac{2}{(1+i)(z-\alpha)(z-\beta)})\\ \amp = 2\pi i\cdot \frac{2}{2\sqrt{2}i}\\ \amp = \frac{2\pi}{\sqrt{2}}\\ \amp = \sqrt{2}\pi\text{.} \end{align*}

Subsection Integral of rational function over real line

Given a rational function of the form \(f(z)=p(z)/q(z)\) where \(p\) and \(q\) are polynomials and \(\deg q\geq \deg p+2\text{,}\) it is easy to see that
\begin{equation*} \lim\limits_{R\to \infty}\int_{\gamma_R}f(z)\, dz = 0\text{,} \end{equation*}
where \(\gamma_R(t)=Re^{it}\text{,}\) \(t\in [0,\pi]\) is the upper semicircle of radius \(R\) centered at the origin. (See Figure 1.23.6.) Since residues of rational functions are often easy to compute, the Cauchy residue theorem gives us a promising approach to these integrals.
Semicircular path for integrals over the real line
Figure 1.23.6. Semicircular path for integrals over the real line

Example 1.23.7. Rational function on real line.

Compute \(\displaystyle \int_{-\infty}^\infty \frac{1}{(x^2+4)^2(x^2+1)}\, dx\)
Solution.
Let \(f(z)=1/((z^2+4)^2(z^2+1))\text{.}\) The denominator factors as \((z-2i)^2(z+2i)^2(z-i)(z+i)\text{.}\) It follows that \(\pm 2i\) are poles of order 2 of \(f\text{,}\) and \(\pm i\) are simple poles. We compute, using Theorem 1.21.7 and Corollary 1.22.7,
\begin{align*} \res_f i \amp = \frac{1}{(i^2+4)^2(i+i)}\\ \amp = \frac{1}{18i}\\ \res_f 2i \amp = g'(2i), \ g(z)=1/((z+2i)(z^2+1))\\ \amp = -\frac{11}{288i}\text{.} \end{align*}
For each \(R> 0\text{,}\) let \(\gamma\) be the semicircular path in Figure 1.23.6. For \(R> 2\text{,}\) we have
\begin{equation*} \int_\gamma f\, dz=2\pi i(\res_f i+\res_f 2i)=2\pi(1/18-11/288)=\frac{5\pi}{144}\text{.} \end{equation*}
It follows that for all \(R> 2\text{,}\) we have
\begin{equation} \int_{-R}^Rf(z)\, dz=\frac{5\pi}{144}-\int_{\gamma_R}f(z)\, dz.\tag{1.87} \end{equation}
On \(\gamma_R\) we have, using an ML-inequality,
\begin{align*} \abs{\int_{\gamma_R}f\, dz} \amp \leq \max\{\abs{f(z)}\colon z\in \gamma_R\}\cdot \pi R \\ \amp \leq \frac{1}{(R^2-4)(R^2-1)}\cdot \pi R\\ \amp =\frac{\pi R}{(R^2-4)(R^2-1)}\text{,} \end{align*}
where we have used the fact that for \(z=Re^{it}\text{,}\) we have
\begin{align*} \abs{(z^2+4)(z^2+1)}=\abs{z^2+4}\abs{z^2+1} \amp\geq \abs{\abs{z^2}-4}{\abs{\abs{z}^2-1}} \\ \amp\geq (R^2-4)(R^2-1) \text{,} \end{align*}
for \(R> 2\text{.}\) Since \(\pi R/((R^2-4)(R^2-1))\to 0\) as \(R\to \infty\text{,}\) we see that
\begin{equation*} \lim\limits_{R\to \infty}\int_{\gamma_R} f\, dz=0\text{.} \end{equation*}
Taking limits of both sides of (1.87), we conclude that
\begin{equation*} \int_{-R}^Rf(z)\, dz=\frac{5\pi}{144}\text{.} \end{equation*}

Subsection Trig integral over real line

A semicircular path like Figure 1.23.6 is often useful also for integrating functions \(f\colon \R\rightarrow \R\) on the real line that involve trigonometric functions, and which have no singularities in \(\R\text{.}\) However, when choosing the complex function \(g\) we integrate over this path, we must take care that the integral over \(\gamma_R\) tends toward zero as \(R\to \infty\text{.}\) This is often accomplished by using \(e^{iz}\) instead of \(\cos z\) or \(\sin z\text{,}\) whose moduli are trickier to bound.

Example 1.23.8. Trig integral over real line.

Compute \(\displaystyle\int_{-\infty}^{\infty}\frac{x\sin x}{x^2+1}\, dx\)
Solution.
Let \(\displaystyle f(z)=\frac{z e^{iz}}{z^2+1}\text{.}\) It is easy to see that \(f\) has simple poles at \(\pm i\) and we compute
\begin{equation*} \res_f i=\frac{ie^{i^2}}{2i}=\frac{e^{-1}}{2} \end{equation*}
with the help of Theorem 1.21.7.
For all \(R\) let \(\gamma\) be the usual semicircular path as in Figure 1.23.6. We have
\begin{align*} \int_{\gamma_R} f(z)\, dz \amp =\int_{-R}^0f(x)\, dx+\int_0^R f(x)\, dx+\int_{\gamma_R} f\, dz \\ \amp = -\int_0^{R}f(-x)\, dx+\int_0^Rf(x)\, dx+\int_{\gamma_R}f\, dz\\ \amp = \int_0^R f(x)-f(-x)\, dx+\int_{\gamma_R}f\, dz\\ \amp = 2i\int_0^R\frac{x\sin x}{x^2+1}\, dx+\int_{\gamma_R}f\, dz \text{.} \end{align*}
The Cauchy residue theorem tells us that \(\int_\gamma f\, dz=2\pi i\res_f i=\pi e^{-1}\text{,}\) and thus we have (after a little algebra)
\begin{equation} \int_0^{R}\frac{x\sin x}{x^2+1}\, dx=\frac{\pi}{e}-\frac{1}{2}\int_{\gamma_R} f\, dz\tag{1.88} \end{equation}
for all \(R\text{.}\) The ML-inequality is not delicate enough to show that \(\lim\limits_{R\to \infty}\int_{\gamma_R}f\, dz=0\text{,}\) and so we endeavor instead to use Jordan’s inequality:
\begin{align*} \abs{\int_{\gamma_R}f\, dz} \amp \leq \int_0^\pi \frac{\abs{Re^{it}}\abs{e^{iRe^{it}}}}{\abs{R^2e^{2it}+1}}\cdot \abs{iRe^{it}}\, dt\\ \amp \leq \int_0^\pi \frac{R^2e^{-R\sin t}}{R^2-1}\, dt\\ \amp \leq \frac{\pi R}{R^2-1}\amp (\text{Jordan's ineq.})\text{.} \end{align*}
Since \(\displaystyle \frac{\pi R}{R^2-1} \to 0\) as \(R\to \infty\text{,}\) it follows that \(\lim\limits_{R\to \infty}\int_{\gamma_R}f\, dz=0\text{,}\) using the squeeze theorem. Finally, taking limits of both sides of (1.88), we conclude that
\begin{equation*} \int_0^\infty \frac{x\sin x}{x^2+1}\, dx=\frac{\pi}{e}\text{.} \end{equation*}

Subsection Indented paths

To handle an integral of the form \(\int_0^\infty f(x)\, dx\) that is improper at \(0\text{,}\) we make use of an indented path like the one in Figure 1.23.9 and take the limit as \(R\to \infty\) and \(\epsilon\to 0\text{.}\) For this approach to be successful, we need to be able to say something meaningful about the limit of the integral around the bump as \(\delta\to 0\text{.}\) If the complex function we end up integrating over this path has an isolated singularity at \(z_0=0\text{,}\) then we might be able to make use of Theorem 1.23.3.
Indented path
Figure 1.23.9. Indented path

Example 1.23.10. Logarithmic improper integral.

Compute \(\displaystyle\int_0^\infty \frac{\ln x}{x^4+1}\, dx\text{.}\)
Solution.
With the idea of using a path as in Figure 1.23.9, we work with the branch \(\Log_{\alpha}\text{,}\) with \(\alpha=-\pi/2\text{.}\) This is defined on \(\C-R_{-\pi/2}\) and agrees with \(\ln x\) on the positive reals. Let \(f(z)=\frac{\Log_{\alpha}(z)}{z^4+1}\text{.}\) It is easy to see that \(f\) is analytic on \(U=\C-R_{\alpha}\) except at the roots \(z_k=e^{i\pi/4}e^{i\pi k/2}\text{,}\) \(0\leq k\leq 3\text{,}\) of \(z^4+1\text{.}\) With the help of Theorem 1.21.7, we compute
\begin{align*} \res_f z_0 \amp = \frac{\Log_\alpha(e^{i\pi/4})}{4e^{i3\pi/4}}\\ \amp = \frac{i\pi}{16}\cdot e^{-i3\pi/4} \\ \res_f z_1 \amp = \frac{\Log_\alpha(e^{i3\pi/4})}{4e^{i9\pi/4}}\\ \amp = \frac{i3\pi}{16}\cdot e^{-9i\pi/4} \\ \amp = \frac{i 3\pi}{16}\cdot e^{-i\pi/4}\text{.} \end{align*}
For all \(R > 1\) and \(0 < \delta < 1\text{,}\) let \(\gamma\) be the path as in Figure 1.23.9. On the one hand, since such a path only encloses the poles \(z_0\) and \(z_1\text{,}\) we have by the Cauchy residue theorem
\begin{align*} \int_{\gamma}f\, dz \amp = 2\pi i(\res_f z_0+\res_f z_1) \amp \\ \amp = -\frac{\pi^2}{8}(e^{-i3\pi/4}+3e^{-i\pi/4})\\ \amp = -\frac{\pi^2}{8}(\sqrt{2}-2\sqrt{2}i)\\ \amp = \frac{\pi^2}{8}(-\sqrt{2}+2\sqrt{2}i)\text{.} \end{align*}
On the other hand, breaking the path \(\gamma\) down into separate legs, we have
\begin{align*} \int_{\gamma}f\, dz \amp = \int_{-R}^{-\delta}\frac{\ln\abs{x}+\pi i}{x^4+1}\, dx+\int_{\delta}^R \frac{\ln x}{x^4+1}\, dx+\int_{\gamma_\delta}f\, dz+\int_{\gamma_R}f\, dz\\ \amp =2\int_{\delta}^R\frac{\ln x}{x^4+1}+i\int_{\delta}^R\frac{\pi}{x^4+1}\, dx +\int_{\gamma_\delta}f\, dz+\int_{\gamma_R}f\, dz\text{,} \end{align*}
where here \(\gamma_\delta(t)=\delta e^{i(\pi-t)}=i\delta e^{-it}\text{,}\) \(t\in [0,\pi]\text{,}\) and \(\gamma_R(t)=Re^{it}\text{,}\) \(t\in [0,\pi]\text{.}\) We will show that \(\lim\limits_{\delta\to 0^+}\int_{\gamma_\delta}f\, dz=0\text{,}\) and \(\lim\limits_{R\to \infty}\int_{\gamma_R}f\, dz=0\text{,}\) and hence, taking limits of the equality above,
\begin{equation*} \frac{\pi^2}{8}(-\sqrt{2}+2\sqrt{2}i)= 2\int_{\delta}^R\frac{\ln x}{x^4+1}+i\int_{\delta}^R\frac{\pi}{x^4+1}\, dx\text{.} \end{equation*}
Equating real and imaginary parts, we then conclude that
\begin{align*} \int_0^\infty\frac{\ln x}{x^4+1}\, dx \amp = \frac{\sqrt{2}\pi}{16} \amp \\ \int_0^\infty \frac{1}{x^4+1}\, dx \amp = \frac{\sqrt{2}\pi}{4} \amp (\text{bonus}) \text{.} \end{align*}
To see that \(\lim\limits_{\delta\to 0^+}\int_{\gamma_\delta}f\, dz=0\text{,}\) and \(\lim\limits_{R\to \infty}\int_{\gamma_R}f\, dz=0\text{,}\) it suffices to use ML-inequalities on both these paths. In general, for any circular path \(\phi(t)=se^{it}\text{,}\) \(t\in [0,\pi]\) we have
\begin{equation*} \abs{\Log_\alpha( se^{it})}=\abs{\ln \abs{s}+it}\leq \ln \abs{s}+\pi\text{.} \end{equation*}
Thus, we have
\begin{align*} \abs{\int_{\gamma_R}f\, dz} \amp\leq \frac{\ln R+\pi}{R^4-1}\cdot \pi R \\ \amp = \pi \cdot \frac{\ln R+\pi}{R^3-1/R}\\ \abs{\int_{\gamma_\delta}f\, dz} \amp\leq \frac{\ln \delta+\pi}{1-\delta^4}\cdot \pi \delta \\ \amp = \pi \frac{\delta \ln \delta+\delta\pi}{1-\delta^4}\text{.} \end{align*}
Some straightforward, if not altogether pleasant l’Hôpital’s rule arguments now show that
\begin{align*} \lim\limits_{R\to \infty} \pi \cdot \frac{\ln R+\pi}{R^3-\tfrac{1}{R}} \amp=0 \\ \lim\limits_{\delta\to 0^+} \pi \frac{\delta \ln \delta+\delta\pi}{1-\delta^4}\amp =0 \end{align*}
and hence that
\begin{align*} \lim\limits_{R\to \infty}\int_{\gamma_R}f\, dz \amp= 0 \\ \lim\limits_{\delta\to 0^+} \int_{\gamma_\delta}f\, dz \amp = 0 \text{,} \end{align*}
as desired.

Example 1.23.11. Sinc function.

The function \(f(x)=\frac{\sin x}{x}\) is a called the (unnormalized) sinc function. Compute \(\displaystyle\int_{0}^\infty f(x) \, dx\)
Solution.
The function \(f(z)=\frac{e^{iz}}{z}\) is analytic everywhere except at \(z=0\text{,}\) where it has a simple pole. An easy computation shows \(\res_f 0=1\text{.}\)
For all \(R> 1\) and \(0< \delta < 1\text{,}\) let \(\gamma\) be the indented path as in Figure 1.23.9. Since \(f\) is analytic on the interior of \(\gamma\) we have
\begin{equation*} \int_{\gamma}f\, dz=0 \end{equation*}
by Cauchy-Goursat. On the other hand, we have
\begin{align*} \int_\gamma f\, dz \amp = \int_{-R}^\delta f(x)\, dx+\int_\delta^Rf(x)\, dx +\int_{\gamma_\delta}f\, dz+\int_{\gamma_R}f\m dz\\ \amp = \int_{\delta}^Rf(x)-f(-x)\, dx+\int_{\gamma_\delta}f\, dz+\int_{\gamma_R}f\, dz \\ \amp = 2i\int_{\delta}^R \frac{\sin x}{x}\, dx+\int_{\gamma_\delta}f\, dz+\int_{\gamma_R}f\, dz \text{.} \end{align*}
Thus, using a little algebra, we see that
\begin{equation} \int_{\delta}^R \frac{\sin x}{x}\, dx= -\frac{1}{2i}\int_{\gamma_\delta}f\, dz-\frac{1}{2i}\int_{\gamma_R}f\, dz\text{.}\tag{1.89} \end{equation}
Using Theorem 1.23.3, we have
\begin{equation*} \lim\limits_{\delta\to 0} \int_{\gamma_\delta}f\, dz=-\pi i\res_f 0=-\pi i\text{.} \end{equation*}
(Recall \(\gamma_\delta\colon [0,\pi]\rightarrow \C\text{,}\) so the \(\beta-\alpha=\pi\) in this case.) The usual Jordan’s inequality argument shows that \(\lim\limits_{R\to \infty}f\, dz=0\text{.}\) Thus taking limits of both sides of (1.89), we see that
\begin{equation*} \int_{0}^\infty \frac{\sin x}{x}\, dx =-\frac{1}{2i}(-\pi i)=\frac{\pi}{2}\text{.} \end{equation*}

Subsection Keyhole (or Pac-Man) paths

To compute the integral in Example 1.23.13 we need to choose a branch of \(z^{1/3}\) that agrees with \(x^{1/3}\) on the positive real line: or what amounts to the same thing, a branch of \(\log z\) that agrees with the principle branch on the positive real line. We could theoretically proceed as in Example 1.23.10, choosing the branch \(\Log_{-\pi/2}\text{,}\) and integrating along the indented path Figure 1.23.9, but it turns out that the two integrals along the real line legs of this path do not combine nicely to give us the desired result. Instead, we choose a keyhole (or Pac-Man shaped path) \(\gamma_{R,\delta,\alpha}\) like that in Figure 1.23.12, where the parametrizations \(\gamma_R\) and \(\gamma_\alpha \) are restricted to inputs \(t\) lying in \([\alpha, 2\pi-\alpha]\text{,}\) and where the line segments \(\gamma_\alpha\) and \(\gamma_{-\alpha}\) are the line segments along the rays \(R_\alpha\) and \(R_{-\alpha}\) lying on and between these two circles. The idea is to use the Cauchy residue formula to compute the integral of the given function over a given \(\gamma_{R,\delta, \alpha}\text{,}\) and then investigate what happens as we let \(R\to \infty\text{,}\) \(\delta\to 0\text{,}\) and \(\alpha\to 0\text{.}\)
Keyhole (or Pac-Man) path
Figure 1.23.12. Keyhole (or Pac-Man) path

Example 1.23.13. Integral of power function.

Compute \(\displaystyle\int_{0}^\infty \frac{x^{1/3}}{x^2+4x+8}\, dx\text{.}\)
Solution.
Let \(f(z)=\frac{e^{\frac{1}{3}\Log_0 z}}{z^2+4z+8}\text{,}\) where as usual \(\Log_0 z=\ln\abs{z}+i\Arg_0 z\text{,}\) and \(\Arg_0\) is the branch of \(\arg\) defined on \(\C-R_0\) taking values in \((0,2\pi)\text{.}\) The function \(f\) is analytic everywhere on \(\C-R_0\) except for the roots \(z_1=-2+2i=2^{3/2}e^{3\pi i/4}\) and \(\overline{z_1}=2^{3/2}e^{5\pi i/4}\) of the denominator, which are simple poles of \(f\text{.}\) We compute
\begin{align*} \res_f(z_1) \amp = \frac{e^{\frac{1}{3}\Log_0 z_1}}{2z_1+4}\\ \amp = \frac{\sqrt{2}e^{\pi i/4}}{4i}\\ \res_f(\overline{z_1}) \amp = \frac{e^{\frac{1}{3}\Log_0 \overline{z}_1}}{2\overline{z}_1+4}\\ \amp = -\frac{\sqrt{2}e^{5\pi i/12}}{4i}\text{.} \end{align*}
For all \(\alpha> 0\text{,}\) and \(R\) and \(\delta\) satisfying \(0< \delta < 2\sqrt{2}< R\text{,}\) let \(\gamma\) be the path as in Figure 1.23.12. By the Cauchy residue theorem we have
\begin{align*} \int_\gamma f\, dz\amp =2\pi i(\res_f(z_1)+\res_f(\overline{z}_1)) \amp \\ \amp = \frac{\sqrt{2}}{2}\pi(e^{\pi i/4}-e^{5\pi i /12})\\ \amp = \frac{\sqrt{2}}{2}\pi e^{\pi i/3}(e^{-\pi i/12}-e^{\pi i/ 12})\\ \amp = -i\sqrt{2}\pi e^{\pi i/3}\sin(\pi/12)\text{.} \end{align*}
Yes, we resorted to some algebraic trickery there (factoring out the \(e^{\pi i/3}\)) in order to combine terms nicely. This was not strictly speaking necessary, but saves us some simplifying work at the end.
We now compute the integrals along the various subpaths that make up \(\gamma\text{.}\) Using straight line parametrizations for the integrals along \(\gamma_\alpha\) and \(\gamma_{-alpha}\) we see that
\begin{align*} \int_\gamma f\, dz\amp =\int_\delta^R \frac{x^{1/3}e^{i\alpha/3}}{x^2e^{2i\alpha}+4xe^{i\alpha}+8}\, dx-\int_\delta^R \frac{x^{1/3}e^{(2i(\pi-\alpha))/3}}{x^2e^{-2i\alpha}+4xe^{-i\alpha}+8}\, dx\amp \\ \amp +\int_{\gamma_{R,\alpha}}f\, dz+\int_{\gamma_{\delta,\alpha}}f\, dz\text{,} \end{align*}
where in our computation of \(f(\gamma_{-\alpha}(x))\text{,}\) we use the fact that
\begin{equation*} \Log_0(xe^{-i\alpha})=\ln x+i\Arg_0(e^{-i\alpha})=\ln x+i(2\pi -\alpha)\text{.} \end{equation*}
Taking the limit as \(\alpha\to 0\) we get
\begin{align*} \int_\gamma f\, dz\amp =(1-e^{2\pi i/3})\int_\delta^R \frac{x^{1/3}e^{i\alpha/3}}{x^2e^{2i\alpha}+4xe^{i\alpha}+8}\, dx++\int_{\gamma_R}f\, dz+\int_{\gamma_\delta}f\, dz \text{.} \end{align*}
Once again, the fact that the limit can be brought into the first two integrals relies on the convergence of the integrands to \(x^{1/3}/(x^2+4x+8)\) being uniform for all \(x\in [\delta, R]\text{.}\) The fact that
\begin{align*} \lim\limits_{\alpha\to 0}\int_{\gamma_{R,\alpha}}f\, dz \amp = \int_{\gamma_R}f\, dz\\ \lim\limits_{\alpha\to 0}\int_{\gamma_{\delta,\alpha}}f\, dz \amp = \int_{\gamma_{\delta}}f\, dz \end{align*}
can be shown using more elementary means.
Next, the usual ML-inequality arguments suffice to show that \(\int_{\gamma_R}f\, dz\to 0\) as \(R\to \infty\) and \(\int_{\gamma_\delta}f\, dz\to 0\) as \(\delta\to 0\text{.}\) Thus taking limits as \(R\to \infty\) and \(\delta\to 0\text{,}\) in our last equality, we have
\begin{align*} \int_{\gamma}f\, dz \amp =(1-e^{2\pi i/3})\int_0^\infty \frac{x^{1/3}}{x^2+4x+8}\, dx\text{.} \end{align*}
We conclude that
\begin{align*} \int_0^\infty \frac{x^{1/3}}{x^2+4x+8}\, dx\amp = \frac{1}{1-e^{2\pi i/3}}\int_{\gamma}f\, dz \\ \amp = \frac{-i\sqrt{2}\pi e^{\pi i/3}\sin(\pi/12)}{1-e^{2\pi i/3}}\\ \amp =i\frac{\sqrt(2)\pi\sin(\pi/12)}{e^{\pi i/3}-e^{-\pi i/3}} \\ \amp = \frac{\sqrt{2}\pi\sin(\pi/12)}{2\sin(\pi/3)} \text{.} \end{align*}
That is a fine answer. If, however, you insist on not having it expressed in terms of sine, we can use a trig identity on \(\sin(\pi/12)=\sin(\pi/3-\pi/4)\)
\begin{equation*} \sin(\pi/12)=\sin(\pi/3)\cos(-\pi/4)+\cos(\pi/3)\sin(-\pi/4)=\frac{1}{4}(\sqrt{6}-\sqrt{2})\text{,} \end{equation*}
and then conclude
\begin{equation*} \int_0^\infty \frac{x^{1/3}}{x^2+4x+8}\, dx=\frac{\pi(3-\sqrt{3})}{6} \text{.} \end{equation*}