I came across an integral of this form:

$$\int_{0}^{a}\frac{dx}{(a^{n}-x^{n})^{1/n}}$$ How do I solve this integral? I tried using this substitution: $x=asin^{2}(x)$ in order to reduce this to a beta form integral, but I a man getting the value $\frac{B(0,1)}{2}$. This happens to tending towards minus infinity as $B(0,1)=\frac{\Gamma{(0)}\Gamma{(1)}}{\Gamma{(1)}}\rightarrow-\infty$

Any help is appreciated.


inserting ansatz, the new integral obtained is: $$\int_{0}^{\frac{\pi}{2}}\frac{2acos(\theta)sin(\theta)d\theta}{a^{n/n}(1-sin^{2n}(\theta))^{1/n}}$$ This further reduces to:

$$\int_{0}^{\frac{\pi}{2}}{2(cos(\theta))^{-1}(sin(\theta))^{1}d\theta}$$ Comparing this to the standard Beta integral we get p to be 0 and q to be 1 and thus the integral tends towards minus infinity.

  • $\begingroup$ Where is $n$ gone ? Double check your calculation. $\endgroup$ – Yves Daoust Feb 13 '17 at 13:22
  • $\begingroup$ I see that you ignored my comment. $\endgroup$ – Yves Daoust Feb 13 '17 at 13:37
  • $\begingroup$ Nope, I have edited in the calculations. $\endgroup$ – Spoilt Milk Feb 13 '17 at 13:39
  • $\begingroup$ Not at all, this is still wrong. $\endgroup$ – Yves Daoust Feb 13 '17 at 13:40
  • $\begingroup$ You mean the answer? $\endgroup$ – Spoilt Milk Feb 13 '17 at 13:41

WLOG, $a=1$.

Then with $x^n=t$,

$$I=\frac1n\int_0^1t^{1/n-1}(1-t)^{-1/n}dt=\frac1n\text{B}\left(\frac1n,1-\frac1n\right)=\frac1n\Gamma\left(\frac1n\right)\Gamma\left(1-\frac1n\right)=\frac{\dfrac\pi n}{\sin\dfrac\pi n}.$$

  • 2
    $\begingroup$ A natural question at this point is whether one can obtain this answer directly rather than through special function properties. I suspect the right approach is to use a substitution to map $[0,1]\to [0,\infty)$ and then use a keyhole contour. But this kind of argument can likely be found in a proof of Euler's reflection formula. $\endgroup$ – Semiclassical Feb 13 '17 at 13:52
  • 1
    $\begingroup$ @Semiclassical: the Beta/Gamma relation is usually established through a convolution-like identity involving the Laplace transform. Maybe some simplification can arise for these exponents (?) $\endgroup$ – Yves Daoust Feb 13 '17 at 13:57
  • 1
    $\begingroup$ @Semiclassical you an also use a dogbone contour directly on orignial integrand $\endgroup$ – tired Feb 13 '17 at 14:10

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.