Using complex analysis to evaluate $\int_0^\infty\frac{(\ln x)^3}{1+x^2}d x$

Here is my attempt:

Let $R>1>r$ and $C$ be the closed curve in $\mathbb{C}$ consists of the following pieces: $$C_1=\{Re^{it}: t\in(0,\pi)\},\quad C_2=[r,R],\quad C_3=\{re^{it}: t\in(0,\pi)\},\quad C_4=[-R,r]$$ all curves are oriented counterclockwise. It can be seen that $C$ is the boundary of the upper half of the annulus centred at $0$ with inner radius $r$ and outer radius $R$. Let $f(z)=\frac{(\text{Log } z)^3}{1+z^2},\quad Arg(z)\in\left(-\frac{\pi}{2},\frac{3\pi}{2}\right)$. Then $\int_Cf(z)d z=\sum_{i=1}^4\int_{C_i}f(z)d z$. Note that $f$ has a singularity $z=i$ inside $C$, thus by Cauchy's integral formula $$\int_Cf(z)dz=2\pi i\cdot\frac{(\text{Log } i)^3}{i+i}=\pi\left(\ln|i|+\frac{\pi}{2}i\right)^3=-\frac{\pi^4}{8}i$$ Now consider $\int_{C_i}f(z)d z$: \begin{aligned}&\left|\int_{C_1}f(z)d z\right|\leq\int_0^\pi\frac{(|\ln R|+|it|)^3}{|1-R^2e^{2it}|}|Re^{it}|dt\leq\int_0^\pi\frac{R(\ln R+t)^3}{R^2-1}|dt\leq \frac{\pi R(\ln R+\pi)^3}{R^2-1} \end{aligned} as $RHS\to 0$ as $R\to+\infty$, we have $\lim_{R\to\infty}\int_{C_1}f(z)d z=0$.

$$\int_{C_2}f(z)d z=\int_r^R\frac{(\ln x)^3}{1+x^2}d x\quad\Rightarrow\quad\lim_{r\to 0^+, R\to+\infty}\int_{C_2}f(z)d z=\int_0^\infty\frac{(\ln x)^3}{1+x^2}d x$$

\begin{aligned}&\left|\int_{C_3}f(z)d z\right|\leq\int_0^\pi\frac{(|\ln r|+|it|)^3}{|1+r^2e^{2it}|}|re^{it}|d t\leq\int_0^\pi\frac{r(\ln r+t)^3}{1-r^2}|d t\leq \frac{\pi r(\ln r+\pi)^3}{1-r^2} \end{aligned} as $RHS\to 0$ as $r\to0^+$, we have $\lim_{r\to 0^+}\int_{C_3}f(z)d z=0$

\begin{aligned}&\int_{C_4}f(z)d z=\int_{[-R,-r]}\frac{(\text{Log } z)^3}{1+z^2}d z=\int_{[-R,-r]}\frac{(\ln|z| +\pi i)^3}{1+z^2}d z=\int_r^R\frac{(\ln x +\pi i)^3}{1+x^2}d x\\ =&\int_r^R\frac{(\ln x)^3}{1+x^2}dx+3\pi i\int_r^R\frac{(\ln x)^2}{1+x^2}d x-3\pi^2\int_r^R\frac{\ln x}{1+x^2}d x-\pi^3 i\int_r^R\frac{1}{1+x^2}d x\end{aligned}

At this point I have to evaluate $\int_0^\infty\frac{(\ln x)^2}{1+x^2}d x$, which can be done by complex analysis again. However, this method is way too long and from my point of view, not the most efficient. Is there a shorter way to do this?

By the way, the answer is 0.

• @dustin Well, there it is, apply the same argument.I really should have thought about that.. – Frank Lu Feb 16 '15 at 4:58
• @dustin Exactly, this method works, but is not the most efficient, I believe. – Frank Lu Feb 16 '15 at 5:00
• @FrankLu: Are you interested in complex variables techniques only? – science Feb 16 '15 at 5:01
• @science It specifically said using complex analysis, so I have to. – Frank Lu Feb 16 '15 at 5:07
• @Lucian That's one way to see the answer, but the question said using methods of complex analysis – Frank Lu Feb 16 '15 at 16:25

Using the contour I used here, we have that $$\int_0^{\infty}\frac{\log^3(x)}{1+x^2}dx = \int_{\Gamma}f(z)dz + \int_{-\infty}^{-\epsilon}f(z)dz + \int_{\gamma}f(z)dz + \int_{\epsilon}^{\infty}f(z)dz \tag{1}$$ where $f(z) = \frac{\log^3(z)}{z^2 + 1}$, $\Gamma$ is large semi circle, and $\gamma$ is small semi circle. As with the linked post, we will have the branch cut from $(-\infty, 0)$. By the estimation lemma, the first and third integrals of $(1)$ are zero. Now $$\log^3(z) = \log^3\lvert z\rvert - 3\arg^2(z)\log\lvert z\rvert + i(3\arg(z)\log^2\lvert z\rvert - \arg^3(z))$$ Additionally, we are taking the real Cauchy principal value of the integral. With that in mind and by the residue theorem, we have \begin{align} \int_{-\infty}^0\frac{\log^3\lvert z\rvert - 3\arg^2(z)\log\lvert z\rvert}{z^2 + 1} dz + \int_0^{\infty}\frac{\log^3\lvert z\rvert - 3\arg^2(z)\log\lvert z\rvert}{z^2 + 1}dz &= \text{Re}\left(2\pi i\sum\text{Res}\right)\\ 2\int_0^{\infty}\frac{\log^3(x)}{x^2 + 1}dz - 3\pi^2\int_0^{\infty}\frac{\log\lvert z\rvert}{z^2+1}dz & = 0\\ \int_0^{\infty}\frac{\log^3(x)}{x^2 + 1}dz&= 0\tag{2} \end{align} where $(2)$ occurs since $\int_0^{\infty}\frac{\log\lvert z\rvert}{z^2 + 1}dz = 0$ and the residue was purely imaginary so the real principal value is zero. Using the same contour for $\frac{\log(z)}{z^2+1}$, we obtain $$\int_0^{\infty}\frac{\log\lvert z\rvert}{z^2 + 1}dz = 2\int_0^{\infty}\frac{\log\lvert z\rvert}{z^2+1}dz + \int_0^{\infty}\frac{i\pi}{z^2+1}dz = \frac{i\pi^2}{2}\tag{3}$$ Again, we were seeking the real Cauchy principal value so $(3)$ becomes $$\int_0^{\infty}\frac{\log\lvert z\rvert}{z^2+1}dz = 0$$
• Sorry for the late comment. I am not very familiar with the Cauchy principal value since I've only heard about it few times, so I don't quite understand how you get $\int_0^\infty\frac{\log^3(x)}{x^2+1}dx=0$ if $\int_0^\infty\frac{\log|z|}{z^2+1}dz=0$ in (2) of your solution. Also, there is no upvotes or down votes to this post, so I don't know whether this is correct or not, so I left this for a long time. Sorry about that. – Frank Lu Mar 12 '15 at 5:16
• @FrankLu in equation $(3)$, you see $$\int_0^{\infty}\frac{i\pi}{z^2+1}dz$$ and $$\frac{i\pi^2}{2}\mbox{?}$$ Since both these terms are complex, we can disregard them since we want a real solutions so we conclude $$\int_0^{\infty}\frac{\log|z|}{z^2+1}dz =0$$ – dustin Mar 12 '15 at 5:21
• Then I think in (2), the RHS should be $0$ for all 3 equations, since you are taking real parts on both sides, that is, $\text{Re}\int_\gamma\frac{\text{Log}^3 z}{z^2+1}dz=\text{Re}(2\pi i\sum\text{Res})$, which means $2\int_0^{\infty}\frac{\log^3(x)}{x^2 + 1}dz - 3\pi^2\int_0^{\infty}\frac{\log\lvert z\rvert}{z^2+1}dz = 0$. Am I right? – Frank Lu Mar 12 '15 at 5:35
• @FrankLu yes. I just didn't show all steps. Also, the real log integral $\log^n(z)/(z^2+1)$ with odd powers is always zero since it produces purely complex solutions. You can get real solutions with even powers. – dustin Mar 12 '15 at 5:36
You don't have to use the contour integral to evaluate. In fact, you can split the integral into two parts $$\int_0^\infty \frac{\ln^3x}{x^2+1}dx=\int_0^1 \frac{\ln^3x}{x^2+1}dx+\int_1^\infty \frac{\ln^3x}{x^2+1}dx.$$ For the second part, use the substitute $x=\frac1u$ to obtain $$\int_1^\infty \frac{\ln^3x}{x^2+1}dx=-\int_0^1 \frac{\ln^3u}{u^2+1}du.$$ Hence $$\int_0^\infty \frac{\ln^3x}{x^2+1}dx=0.$$
• You should probably read the comments especially the one where it says @science It specifically said using complex analysis, so I have to. That is, I don't think this is what the OP wants. – dustin Feb 25 '15 at 16:18