5
$\begingroup$

I am trying to show that,

\begin{align} I = \int_{-1}^1 x^nP_n(x)\,\mathrm{d}x = \frac{2^{n+1}n!n!}{(2n+1)!} \end{align}

So far I have done the following. Rodrigues formula is as follows:

\begin{align*} P_n(x) = \sum_{k=0}^N \frac{(-1)^k (2n-2k)!}{2^nk!(n-k)!(n-2k)!} x^{n-2k} \end{align*}

where,

\begin{align*} \begin{aligned} N&=n/2, && \text{if} \quad n=\text{even} \\ N&=(n-1)/2, && \text{if} \quad n=\text{odd} \end{aligned} \end{align*}

Substitute Rodrigues formula,

\begin{align*} I &= \int_{-1}^1 x^n \sum_{k=0}^N \frac{(-1)^k (2n-2k)!}{2^nk!(n-k)!(n-2k)!} x^{n-2k} \,\mathrm{d}x \\ &= \sum_{k=0}^N \frac{(-1)^k (2n-2k)!}{2^nk!(n-k)!(n-2k)!} \int_{-1}^1 x^{2n-2k} \,\mathrm{d}x \\ &= \sum_{k=0}^N \frac{(-1)^k (2n-2k)!}{2^nk!(n-k)!(n-2k)!} \left. \frac{x^{2n-2k+1}}{2n-2k+1}\right\rvert_{-1}^1 \\ &= \sum_{k=0}^N \frac{(-1)^k (2n-2k)!}{2^nk!(n-k)!(n-2k)!} \frac{1 - (-1)^{2n-2k+1}}{2n-2k+1} \\ &= \sum_{k=0}^N \frac{(-1)^k (2n-2k)!}{2^nk!(n-k)!(n-2k)!} \frac{1 + (-1)^{2n-2k}}{2n-2k+1} \end{align*}

Since $2n-2k$ is even,

\begin{align*} I &= \sum_{k=0}^N \frac{(-1)^k (2n-2k)!}{2^nk!(n-k)!(n-2k)!} \frac{2}{2n-2k+1} \\ &= \sum_{k=0}^N \frac{2^{1-n}(-1)^k (2n-2k)!}{k!(n-k)!(n-2k)!(2n-2k+1)} \end{align*}

Can someone give me a hint about how to proceed or do I need to prove with another way?

$\endgroup$
  • $\begingroup$ Try Bonnet’s recursion formula $(n+1)P_{n+1}(x)=(2n+1)xP_{n}(x)-nP_{n-1}(x)$. $\endgroup$ – marty cohen Jun 14 '16 at 20:06
  • $\begingroup$ I tried. After using recursion formula I inserted Rodrigues formula again for $P_{n+1}$ and $P_{n-1}$ but got stuck again. $\endgroup$ – Shibli Jun 14 '16 at 20:21
  • $\begingroup$ @Shibli: just keep the $\frac{d^n}{dx^n}$ in Rodrigues' formula and apply integration by parts $n$ times. $\endgroup$ – Jack D'Aurizio Jun 14 '16 at 20:46
7
$\begingroup$

It is faster to exploit the generating function that comes from Rodrigues' formula or Bonnet's recursion formula: $$\frac{1}{\sqrt{1-2xt+t^2}} = \sum_{n=0}^\infty P_n(x) t^n\tag{1} $$ By replacing $t$ with $xt$ we have: $$ \int_{-1}^{1}\frac{dx}{\sqrt{1-2x^2 t+x^2 t^2}}=\sum_{n\geq 0}\left(\int_{-1}^{-1}x^n P_n(x)\,dx\right)t^n\tag{2} $$ so our integrals can be computed through the Taylor series of the LHS, regarded as a function of $t$.
The LHS of $(2)$ is: $$ \frac{2\arcsin\!\sqrt{t(2-t)}}{\sqrt{t(2-t)}}=\frac{4\arcsin\sqrt{\frac{t}{2}}}{\sqrt{t(2-t)}} \tag{3} $$ and your initial claim turns out to be equivalent to a well-known identity.

Truth to be told, it is also very interesting sideways: some series related with the $\Gamma^2$ function can be easily computed by exploiting the mentioned property of Legendre polynomials.


Using Rodrigues' formula we have: $$\int_{-1}^{1}x^n P_n(x)\,dx = \frac{1}{2^n n!}\int_{-1}^{1}x^n\cdot\frac{d^n}{dx^n}(x^2-1)^n\,dx \stackrel{RIBP}{=}\frac{1}{2^n}\int_{-1}^{1}(1-x^2)^n\,dx\tag{4}$$ where $\text{RIBP}$ stands for repeated integration by parts.
But the RHS of $(4)$ is a value of the Euler's beta function: $$\frac{1}{2^n}\int_{-1}^{1}(1-x^2)^n\,dx=\frac{1}{2^n}\int_{0}^{1}z^{-1/2}(1-z)^n\,dz = \frac{\Gamma\left(\frac{1}{2}\right)\Gamma(n+1)}{2^n\,\Gamma\left(n+\frac{3}{2}\right)}\tag{5}$$ and we are equally done, since: $$\frac{\Gamma\left(\frac{1}{2}\right)\Gamma(n+1)}{2^n\,\Gamma\left(n+\frac{3}{2}\right)} = \frac{2\cdot n!}{(2n+1)!!} = \frac{2^{n+1} n!^2}{(2n+1)!}.\tag{6}$$

$\endgroup$
  • $\begingroup$ In the link you gave, there are two $\arcsin$ but here only one. I could not figure out the relation (sorry I am bad in math). $\endgroup$ – Shibli Jun 14 '16 at 20:41
  • $\begingroup$ @Shibli: It does not matter, please see my updated answer (the direct approach through Rodrigues' formula). $\endgroup$ – Jack D'Aurizio Jun 14 '16 at 20:44

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.