0

$$\int_{0}^{\infty}\dfrac{x^{n-1}}{1+x}dx=\dfrac{\pi}{\sin{n\pi}},\text{ where }0<n<1$$

I want to get the result on right. I substituted $x^{n-1}=t$ but when forwarded to transform in $t.$ The integral gets more complex. I do not know, how to proceed to get the desired result written on the right of the improper integral.

Arturo Magidin
  • 398,050

3 Answers3

2

Let $0<n<1$, $f(z)=\frac{z^{n-1}}{(z+1)}$ and $I= ∫_0^{∞}f(x)dx$. Consider the integral of $f(z)$ around the famous contour [given here] (https://en.wikipedia.org/wiki/Contour_integration#/media/File:Keyhole_contour.svg)

Then, $$∫_{C}f(z)dz=∫_{ε}^{M}f(x)dx+∫_{C_{R}}f(z)dz+∫_{M}^{ε}f(x)dx+∫_{C_{ε}}f(z)dz$$ which is equal to $2πi$ times residue of $f(z)$ at $z=-1$, that is, to $2πie^{2π(n-1)i}$. It can easily be shown that the integrals around circles tend to zero as $R→∞$ and $ε→0$. Hence, in the limit we have $(1+e^{2π(n-1)i})I=2πie^{2π(n-1)i}$, that is, $$I=\frac{-2πi}{e^{π(n-1)i}-e^{-π(n-1)i}}=\frac{-π}{\sin(πn-π)}.$$ In the end, $I=\frac{\pi}{\sin πn}.$

amWhy
  • 209,954
Bob Dobbs
  • 10,988
0

We consider two branches of the logarithm. One has a branch cut along the negative imaginary axis where arguments range from $-\frac{\pi}{2}$ to $\frac{3\pi}{2}$. The other has a branch cut along the positive imaginary axis where arguments range from $\frac{\pi}{2}$ to $\frac{5\pi}{2}$. Thus the two branches agree on the half plane $\operatorname{Re} x <0$. Now consider the following contours :

Two contours

Integrate $\frac{x^{n-1}}{1+x}$ on both contours, with the appropriate branch of $\log x$ then sum them up. After some simplifications (due to the fact that the two branches of $\log x$ agree on the left half plane) we get

$$ \int_{|x+1|=\delta} \frac{x^{n-1}}{1+x}\,\mathrm dx = \int_{|x|=R}\frac{x^{n-1}}{1+x}\,\mathrm dx - \int_{|x|=\delta} \frac{x^{n-1}}{1+x}\,\mathrm dx + \int_\delta^R \big(1- e^{2\pi i(n-1)}\big)\frac{x^{n-1}}{1+x}\,\mathrm dx$$ In the very last integral, we use the regular branch of the logarithm. Residue theorem says that $$ \int_{|x+1|=\delta} \frac{x^{n-1}}{1+x}\,\mathrm dx = 2\pi i e^{\pi i(n-1)} $$ The two remaining integrals can be estimated to be roughly $\delta^{n}$ and $R^{n-1}$ respectively (using the triangle inequality). Since $0<n<1$, these remaining integrals vanish as $\delta \to 0$ and $R\to\infty$, and at the end of the day, we have $$ 2\pi i e^{\pi i(n-1)} = \big(1-e^{2\pi i(n-1)}\big) \int_0^\infty \frac{x^{n-1}}{1+x}\,\mathrm dx $$ Equivalently... $$ \int_0^\infty \frac{x^{n-1}}{1+x}\,\mathrm dx = \frac{\pi}{\sin \pi n} $$

Célestin
  • 653
  • In other words, I integrate the analytic continuation of $\frac{x^{n-1}}{1+x}$ along a keyhole contour crossing the branch cut of the « regular » branch of $\log x$. That's why I introduce two different branches of $\log x$, which cover the whole plane (think of them as local charts of the Riemann surface of $\log x$). – Célestin May 16 '22 at 20:43
  • Note that this question is a duplicate. – Gary May 16 '22 at 22:11
-3

You should use a Gamma-Function. Especially that

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

bjcolby15
  • 3,599
zenitsu
  • 13
  • In my experience the reflection formula is proven by writing it's LHS as the desired integral, then evaluating the latter. – J.G. May 16 '22 at 19:00
  • @J.G. An interesting alternative is to apply the Euler-Weierstrass sine product to the Euler product for the gamma function. This makes the proof 2 lines long! – FShrike May 17 '22 at 05:56
  • @FShrike Yes, if you know how to prove that product, rather than just taking it on faith because it feels as morally true to us as it did to Euler. But proving the Weierstrass factorization theorem is overkill for the OP's problem. – J.G. May 17 '22 at 06:23
  • @J.G. indeed it is overkill. I do not even understand the factorisation theorem myself - if you’re interested, there is a nice real-analytic proof which is not too bad: https://www.researchgate.net/publication/323296523_Euler's_Sine_Product_Formula_An_Elementary_Proof – FShrike May 17 '22 at 06:24