1
$\begingroup$

I want to evaluate the following indefinite integral $$ \int_0^{\infty} x^{p - 1} \cos (ax) dx$$ where $0 < p < 1$ and $a > 0$. I was considering the function $f(z) = z^{p - 1} e^{iaz}$ and integrate it over the contour $\gamma = [-R, -\epsilon] - C_{\epsilon}^+ + [\epsilon, R] + C_R^+$. (Here $C_r^+$ denote the upper half circle centered at $0$ with radius $r$). However I only get $$ \int_0^{\infty} x^{p - 1} e^{iax} dx - e^{i \pi p} \int_0^{\infty} x^{p - 1} e^{-iax} dx = 0$$ instead, which only gives some relation between $\int_0^{\infty} x^{p - 1} \sin (ax) dx$ and $\int_0^{\infty} x^{p - 1} \cos (ax) dx$. I know the result is somehow related to gamma function that is $$ \int_0^{\infty} x^{p - 1} \cos(x) dx =\frac{ \pi }{ 2 \Gamma(1 - p) \sin( (1 - p) \pi / 2) }$$ So would the regular method using complex analysis can still evaluate this integral?

$\endgroup$
  • 2
    $\begingroup$ Terminology correction: this is not an indefinite integral, it is an improper integral. $\endgroup$ – Santiago Canez Apr 28 '14 at 23:52
  • $\begingroup$ Hint: $~\displaystyle\Gamma(p)=\int_0^\infty x^{p-1}e^{-x}dx\quad=>\quad\int_0^\infty x^{p-1}\cos\big(ax\big)dx=\Re\int_0^\infty x^{p-1}e^{-iax}dx.~$ Let $t=iax$. $\endgroup$ – Lucian Apr 29 '14 at 6:57
  • $\begingroup$ You might want to check the $\sin{(\pi (1-p)/2)}$ factor in your expression; I got $\sin{(\pi p/2)}$ instead. WA agrees. Also note that your expression remains finite as $p \to 0$. $\endgroup$ – Ron Gordon Apr 29 '14 at 17:55
3
$\begingroup$

Use a contour in the first quadrant, that is, consider

$$\oint_C dz \, z^{p-1} \, e^{i a z}$$

where $C$ is the contour consisting of $[\epsilon,R]$, the quarter-circle of radius $R$ in the first quadrant centered at the origin, $[i R,i \epsilon]$, and a quarter circle of radius $\epsilon$ about the origin.

The integral then becomes

$$\int_{\epsilon}^R dx \, x^{p-1} \, e^{i a x} + i R^p \int_0^{\pi/2} d\theta \, e^{i p \theta} \, e^{i a R e^{i \theta}} \\ + i^p \int_R^{\epsilon} dy \, y^{p-1} \, e^{-a y} + i \epsilon^p \int_{\pi/2}^0 d\phi \, e^{i p \phi} \, e^{i a \epsilon e^{i \phi}}$$

The fourth integral vanishes as $\epsilon \to 0$ as $0 \lt p \lt 1$. Also, the second integral vanishes as $R \to \infty$, as its magnitude is bounded by

$$R^p \int_0^{\pi/2} d\theta \, e^{-a R \sin{\theta}} \le R^p \int_0^{\pi/2} d\theta \, e^{-2 a R \theta/\pi} \le \frac{\pi}{2 a R^{1-p}}$$

and, again, $0\lt p \lt1$.

The contour integral is zero by Cauchy's theorem as there are no poles inside the contour. Therefore, in the above limits, we have

$$\int_0^{\infty} dx \, x^{p-1} \, e^{i a x} = i^p \int_0^{\infty} dy \, y^{p-1} \, e^{-a y} = e^{i p \pi/2} \frac{\Gamma(p)}{a^p}$$

Use the relationship

$$\Gamma(p) \Gamma(1-p) = \frac{\pi}{\sin{\pi p}}$$

and we have, taking real parts of the above,

$$\int_0^{\infty} dx \, x^{p-1} \, \cos{(a x)} = \frac{\pi \cos{(\pi p/2)}}{a^p \sin{(\pi p)} \Gamma(1-p)} = \frac{\pi}{2 a^p \sin{(\pi p/2)} \Gamma(1-p)}$$

$\endgroup$
  • $\begingroup$ Who downvoted this? Ridiculous. $\endgroup$ – Bennett Gardiner Apr 30 '14 at 1:15

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.