7
$\begingroup$

Show that $$\begin{align*} \forall x \in [-1,1]: \int_0^{\pi} \frac{\sin^n \theta}{(1+x^2-2x \cdot \cos \theta)^{\frac{n}{2}}} \, d\theta &= c_n \tag{1} \\ \int_0^{\pi} \frac{\sin^{n+2} \theta}{(1+x^2-2x \cdot \cos \theta)^{\frac{n}{2}}} \, d\theta &= a_n \cdot x^2+b_n \tag{2} \end{align*}$$ where $a_n, b_n, c_n$ are constants (which do not depend on $x \in [-1,1]$), $n \geq 3$.

I tried several approaches (differentiation to show that the derivative of $(1)$ is equal to $0$, Weierstraß substitution, ...), but always got stuck. For example, one can show that the integrand in $(1)$ equals

$$\begin{align*} \frac{\sin^n \theta}{(1+x^2-2x \cdot \cos \theta)^{\frac{n}{2}}} &= \left( \sqrt{ \left( \frac{x-\cos \theta}{\sin \theta} \right)^2+1} \right)^{-n} \\ &= \left( \sqrt{4 \left( \frac{\sin \frac{\theta+\varrho}{2} \cdot \sin \frac{\theta-\varrho}{2}}{\sin \theta} \right)^2+1} \right)^{-n} \end{align*}$$ where $x=\cos \varrho$. I hoped to get some kind of symmetrization out of it, but (as far as I can see) it doesn't work.

Any ideas? (The aim is to find a rather quick or direct proof - a lengthy one is already known, using a recursive/inductive approach.)

Thanks!

$\endgroup$
1
  • $\begingroup$ Would the downvoter care to comment? Thanks. $\endgroup$
    – saz
    Mar 10, 2014 at 18:54

2 Answers 2

4
$\begingroup$

The Gegenbauer polynomials were built to solve problems like this. Suppose $m\in\mathbb{N}$. We have $$\begin{eqnarray*} \int_0^{\pi} d\theta \, \frac{\sin^{n+2m} \theta}{(1-2x \cos \theta+x^2)^{n/2}} &=& \int_0^\pi d\theta \, \sin^{n+2m} \theta \sum_{k=0}^\infty C_k^{(n/2)}(\cos\theta) x^k \\ &=& \int_{-1}^1 d u \, (1-u^2)^{n/2-1/2} (1-u^2)^m \sum_{k=0}^\infty C_k^{(n/2)}(u) x^k \hspace{5ex} (u=\cos\theta) \\ &=& \int_{-1}^1 d u \, (1-u^2)^{n/2-1/2} \sum_{l=0}^m \beta_{2l} C_{2l}^{(n/2)}(u) \sum_{k=0}^\infty C_k^{(n/2)}(u) x^k \\ &=& \sum_{l=0}^m \sum_{k=0}^\infty \beta_{2l} x^k \underbrace{\int_{-1}^1 d u \, (1-u^2)^{n/2-1/2} C_{2l}^{(n/2)}(u) C_k^{(n/2)}(u)}_{\gamma_{2l}^{(n/2)} \delta_{2l,k}} \\ &=& \sum_{l=0}^m \beta_{2l} \gamma_{2l}^{(n/2)} x^{2l}. \end{eqnarray*}$$ This proves the claim for any $m\in\mathbb{N}$. The restriction $n\ge 3$ is too loose. The result holds for any $n$ such that $n>-1$.

For a given $m$ it is a straightforward exercise to find the $\beta$s and $\gamma$s. For example, using the fact that $1=C_0^{(n/2)}(u)$ and the normalization relation for the Gegenbauer polynomials we find $$\int_0^{\pi} d\theta \, \frac{\sin^{n} \theta}{(1-2x \cos \theta+x^2)^{n/2}} = \frac{\sqrt{\pi}\Gamma\left(\frac{n+1}{2}\right)}{\Gamma\left(\frac{n}{2}+1\right)}.$$

Some details:

  • $1/(1-2x \cos \theta+x^2)^{n/2}$ is the generating function for the Gegenbauer polynomials.
  • The polynomials have the parity property.
  • The polynomials provide an orthogonal basis for functions on $[-1,1]$ with weight function $(1-u^2)^{n/2-1/2}$.
$\endgroup$
2
  • $\begingroup$ That's great, thanks! $\endgroup$
    – saz
    May 18, 2013 at 19:49
  • $\begingroup$ @saz: Glad to help. Thanks for the interesting question. $\endgroup$
    – user26872
    May 18, 2013 at 20:00
0
$\begingroup$

You are in the good way. The integral can be expressed as $$ I(\varphi)=\int_{0}^{\pi}f(\theta,\varphi)d\theta $$ where $$ f(\theta,\varphi)=\left[1+\left(\dfrac{2\sin\left(\dfrac{\theta+\varphi}{2}\right)\sin\left(\dfrac{\theta-\varphi}{2}\right)}{\sin\theta}\right)^{2}\right]^{-\frac{n}{2}} $$ where $x=\cos\varphi$.

Now, $f(\theta,\varphi)=f(-\theta,\varphi)$, so is no harm in writing the integral as $$ I(\varphi)=\dfrac{1}{2}\int_{-\pi}^{\pi}f(\theta,\varphi)d\theta $$ We want to prove that $$ \cfrac[l]{\partial I(\varphi)}{\partial\varphi}=\dfrac{1}{2}\int_{-\pi}^{\pi}\cfrac[l]{\partial f(\theta,\varphi)}{\partial\varphi}d\theta=0 $$ so it is enough to prove that $$ g(\theta,\varphi)=\cfrac[l]{\partial f(\theta,\varphi)}{\partial\varphi} $$ is antisymmetrical in $\theta$.

Deriving we obtaing $$ g(\theta,\varphi)=\dfrac{n\left(\dfrac{2\sin\left(\dfrac{\theta+\varphi}{2}\right)\sin\left(\dfrac{\theta-\varphi}{2}\right)}{\sin\theta}\right)}{\left[1+\left(\dfrac{2\sin\left(\dfrac{\theta+\varphi}{2}\right)\sin\left(\dfrac{\theta-\varphi}{2}\right)}{\sin\theta}\right)^{2}\right]^{\frac{n}{2}+1}} $$ and we see that $g(\theta,\varphi)=-g(-\theta,\varphi)$.

For expression (2), the proof is analogous.

$\endgroup$
2
  • $\begingroup$ How do you arrive at this expression for $g$? As far as I can see, we have $$g(\theta,\varphi) = \frac{n}{(\ldots)^{\frac{n}{2}+1}} \cdot \frac{2}{\sin \theta} \cdot \frac{\partial}{\partial \varphi} \left(\sin \frac{\theta+\varphi}{2} \cdot \sin \frac{\theta-\varphi}{2} \right)$$ ... along my calculations, I end up with $g(\theta,\varphi) = g(-\theta,\varphi)$. $\endgroup$
    – saz
    May 18, 2013 at 19:40
  • $\begingroup$ Excuse me, Saz. There was a mistake in my calculations. Forget my post. $\endgroup$
    – CarlesV
    May 18, 2013 at 20:00

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.