6
$\begingroup$

Showing

$$\pi\int_{0}^{\infty}[1+\cosh(x\pi)]^{-n}dx={(2n-2)!!\over (2n-1)!!}\cdot{2\over 2^n}\tag1$$

Recall

$$1+\cosh(x\pi)={(e^{x\pi}+1)^2\over 2e^{x\pi}}\tag2$$

$$I_n=2^n\pi\int_{0}^{\infty}{e^{xn\pi}\over (1+e^{x\pi})^{2n}}dx\tag3$$

$$I_n={2^n\pi\over n\pi}\int_{0}^{\infty}n\pi e^{nx\pi}(1+e^{x\pi})^{-2n}dx\tag4$$

$$I_n=\left.{2^n\over n}\cdot{1\over (1+e^{x\pi})^{2n-1}}\right|_{0}^{\infty}\tag5$$

$$I_n={1\over n(2n-1)2^{n-1}}\tag6$$

Help, where did I went wrong?


New Edit

From (3) we make a substitution

let $u=x\pi \rightarrow du=dx$

$$I_n=2^n\int_{0}^{\infty}e^{un}(1+e^u)^{-2n}du\tag{3a}$$

$\endgroup$
  • $\begingroup$ @AlexR. which is the same :) as the OP stated, it holds $1+\cosh(\pi x) = 1+\frac{e^{\pi x}+e^{-\pi x}}{2}={(e^{x\pi}+1)^2\over 2e^{x\pi}}$ $\endgroup$ – user190080 Jun 8 '16 at 22:11
  • 1
    $\begingroup$ You somehow introduced an $n$ into $(1+e^{n\pi x})^{2n}$. It should be $(1+e^{x\pi})^{2n}$ $\endgroup$ – Alex R. Jun 8 '16 at 22:16
  • 1
    $\begingroup$ The main issue is that $$ n x^{\color{red}{n}}\neq \frac{d}{dx} x^n.$$ $\endgroup$ – Jack D'Aurizio Jun 8 '16 at 22:25
  • $\begingroup$ @JackD'Aurizio which happens where exactly (I guess you refer to the substitution but this seems to be fine) $\endgroup$ – user190080 Jun 8 '16 at 22:32
  • $\begingroup$ @user190080: between $(4)$ and $(5)$. $\endgroup$ – Jack D'Aurizio Jun 8 '16 at 22:34
7
$\begingroup$

First: remove the useless constant by setting $x=\frac{z}{\pi}$. Then, through $z=\log u$ and $v=\frac{1}{u}$:

$$ \int_{0}^{+\infty}(1+\cosh(z))^{-n}\,dz = 2^n\int_{1}^{+\infty}\frac{\left(2+u+\frac{1}{u}\right)^{-n}}{u}\,du=2^n\int_{0}^{1}\frac{\left(2+v+\frac{1}{v}\right)^{-n}}{v}\,dv$$ so the LHS equals: $$ 2^{n-1}\int_{0}^{+\infty}\frac{u^{n-1} du}{(u+1)^{2n}}\,du = 2^n\int_{0}^{+\infty}\frac{t^{2n-1}\,dt}{(1+t^2)^{2n}}= 2^{n-1} B(n,n) = 2^{n-1}\frac{\Gamma(n)^2}{\Gamma(2n)}.$$

As an alternative, just apply IBP multiple times.
It leads to a recursion similar to the one for $\int_{0}^{\pi/2}\sin^{2n}(\theta)\,d\theta.$

$\endgroup$
  • $\begingroup$ +1, but I think you lack a $dt$ in the second integral of the second row $\endgroup$ – giobrach Jun 9 '16 at 11:42
  • $\begingroup$ @giobrach: thank you, now fixed. $\endgroup$ – Jack D'Aurizio Jun 11 '16 at 13:30
7
$\begingroup$

Take $u=\tanh\left(x\right) $, then $\cosh\left(x\right)=\frac{1+u^{2}}{1-u^{2}} $ and $dx=\frac{2du}{1-u^{2}} $ so $$\int_{0}^{\infty}\frac{1}{\left(\cosh\left(x\right)+1\right)^{n}}dx=\int_{0}^{1}\frac{2}{\left(1-u^{2}\right)\left(\frac{1+u^{2}}{1-u^{2}}+1\right)^{n}}du $$ $$=\frac{1}{2^{n-1}}\int_{0}^{1}\left(1-u^{2}\right)^{n-1}du=\frac{1}{2^{n}}\int_{0}^{1}\left(1-v\right)^{n-1}v^{-1/2}dv=\frac{B\left(1/2,n\right)}{2^{n}}$$ $$=\frac{1}{2^{n}}\frac{\sqrt{\pi}\left(n-1\right)!}{\Gamma\left(n+1/2\right)}.$$

$\endgroup$
  • 1
    $\begingroup$ (+1) There is a nice side-effect: our answers together prove the Legendre duplication formula :D $\endgroup$ – Jack D'Aurizio Jun 8 '16 at 22:36
1
$\begingroup$

Let $I(n)$ be the integral given by

$$\begin{align} I(n)&=\pi\int_0^\infty\frac{1}{\left(1+\cosh(\pi x)\right)^n}\,dx\\\\ &=\int_0^\infty\frac{1}{\left(1+\cosh(x)\right)^n}\,dx\tag 1 \end{align}$$

Then, using the identity $1+\cosh(x)=\frac12 e^{x}\left(1+e^{-x}\right)^2$ in $(1)$ reveals

$$\begin{align} I(n)&=2^n\int_0^\infty \frac{e^{-nx}}{\left(1+e^{-x}\right)^{2n}}\,dx\\\\ &=2^n\int_0^1 \frac{x^{n-1}}{(1+x)^{2n}}\,dx\tag 2 \end{align}$$

Now, making the substitution $x\to 1/x$ in $(2)$, we obtain

$$\begin{align} I(n)&=2^n\int_1^\infty \frac{x^{n-1}}{(1+x)^{2n}}\,dx \tag 3 \end{align}$$

Combining $(2)$ and $(3)$ yields

$$I(n)=2^{n-1}\int_0^\infty \frac{x^{n-1}}{(1+x)^{2n}}\,dx \tag 4$$

Then, enforcing the substitution $x\to \frac{x}{1-x}$ in $(4)$, we find that

$$\begin{align} I(n)&=2^{n-1}\int_0^1 x^{n-1}(1-x)^{n-1}\,dx\\\\ &=2^{n-1}B(n,n)\\\\ &=2^{n-1}\frac{\Gamma^2(n)}{\Gamma(2n)}\\\\ &=\frac{(n-1)!}{(2n-1)!!}\\\\ &=\frac{(2n-2)!!}{2^{n-1}(2n-1)!!} \end{align}$$

as was to be shown!

$\endgroup$
  • $\begingroup$ Are you sure, Mark? Numerics don't match. After "reveals" we should have $I(n)=2^n\left(\ldots\right)$, for instance. $\endgroup$ – Jack D'Aurizio Jun 8 '16 at 22:29
  • $\begingroup$ @JackD'Aurizio Working on it ... yes there is an error. Thanks!! $\endgroup$ – Mark Viola Jun 8 '16 at 22:32
  • $\begingroup$ @chinacat Please let me know how I can improve my answer. I really want to give you the best answer I can. -Mark $\endgroup$ – Mark Viola Jun 11 '16 at 3:48

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.