1

While reviewing contour integration, I found the following question: prove that if $0<s<4$ then $$ \int_0^\infty \frac{x^{s-1}}{x^4+1}dx=\frac{\pi}{8\sin(\frac{\pi s}{4})} $$ This seemed like an easy enough integral to evaluate using residues. I considered the following contour $\gamma$

The contour <span class=$\gamma$ in pink with the segments $\gamma_r$, $\gamma_R$, and $\gamma_v$ referred to below." />

Now, I can estimate that for $r<1$ and $R>1$ we have (by the ML inequality) that $$ \left|\int_{\gamma_r}f(z) dz\right| \leq \frac{r^s}{1-r^4}, \quad \left|\int_{\gamma_R}f(z) dz\right| \leq \frac{R^s}{R^4-1} $$ hence $$ \lim_{r\rightarrow 0}\int_{\gamma_r}f(z) dz=\lim_{R\rightarrow \infty}\int_{\gamma_R}f(z) dz= 0 $$ Moreover, by parameterizing the integral we can see that $$ \int_{\gamma_v}f(z)dz = ie^{\frac{\pi is}{2}} \int_0^\infty f(x)dx $$ Then the residue theorem implies that $$ \int_\gamma f(z) dz =2\pi i \operatorname{Res}(f,e^{\frac{\pi i}{4}}) = -\frac{\pi i}{2}e^{\frac{\pi is}{4}} $$ and putting this together with the above reasoning we see that $$ \int_0^\infty \frac{x^{s-1}}{x^4+1}dx=-\frac{\pi i}{2(1+ie^{\frac{\pi is}{2}})}e^{\frac{\pi is}{4}} $$ It seems that all is left is some algebraic manipulation, but I cannot seem to work it out myself. Any hint would be much appreciated.

  • 3
    I think that $\int_0^\infty \frac{x^{s-1}}{x^4+1}dx=\frac{\pi}{4\sin(\frac{\pi s}{4})}$ – Quý Nhân Aug 13 '24 at 04:18
  • 1
    It may be easier to note that $$ \newcommand{\M}[2]{\mathcal{M} \left{ #1 \right} \left( #2 \right)} \M{f(x^c)}{s} = \frac{1}{|c|} \M{f}{\frac s c} $$ provided that $\M{f}{s}$ converges for $\mathfrak{Re}(s) \in (\alpha,\beta)$ and $c \ne 0$, in which case the fundamental strip for the above is given by $ \frac{\mathfrak{Re}(s)}{c} \in (\alpha,\beta) $.

    And then noting that $$ \M{\frac{1}{1+x^2}}{s} = \frac{\pi}{2 \sin \frac{\pi s}{2}} \text{ if } \mathfrak{Re}(s) \in (0,2) $$ the initial property with $c=2$ allows an easy conclusion, in line with the previous comment.

    – PrincessEev Aug 13 '24 at 04:47
  • 1
    If you substitute $x=e^{t/2}$ and use symmetry to get an integral over $t\in (0,\infty)$, this integral is equivalent to the one found here: https://math.stackexchange.com/q/241922/137524. – Semiclassical Aug 13 '24 at 05:26

0 Answers0