9
$\begingroup$

I ran into this integral when computing the volume of a family of polytopes and I'm not sure how to evaluate it analytically (I know Wolframalpha says 2.27...). Any ideas? I tried using complex analysis (Cauchy Integral Formula, Residue Theorem, etc.) but nothing seemed applicable.

$$\frac{1}{\pi} \int_{0}^{\pi} e^{2\cos{\theta}} d\theta$$

$\endgroup$
  • $\begingroup$ Mathematica says it is equal to $I_0(2)$, where $I_0$ is the modified Bessel function of $0$th order. It is an easy consequence of the integral representation of this function (see formula (4) here). $\endgroup$ – Start wearing purple Jun 24 '13 at 19:06
  • $\begingroup$ According to Gradshteyn and Ryzhik, the integral evaluates to $I_0(2)$, where $I_0$ is the modified Bessel function of the first kind. $\endgroup$ – Cameron Williams Jun 24 '13 at 19:06
  • $\begingroup$ @Samuel Reid :$$\frac{1}{\pi} \int_{0}^{\pi} e^{2\cos{\theta}} d\theta=\frac{-1}{\pi}\int_{-1}^{0}\frac1ze^{z+\overline z}dz$$ $\endgroup$ – M.H Jun 24 '13 at 19:19
7
$\begingroup$

$$ I=\frac{1}{\pi}\int_0^\pi e^{2\cos\theta}\,d\theta=\frac{1}{\pi}\sum_{k=0}^\infty\frac{2^k}{k!}A_k, $$ with $$ A_k=\int_0^\pi\cos^k\theta\,d\theta \quad \forall k \ge 0. $$ We have $$ A_0=\int_0^\pi\,d\theta=\pi. $$ For every $k \ge 1$, if we set $\varphi=\pi-\theta$, then $$ \int_{\pi/2}^\pi\cos^k\theta\,d\theta=\int_0^{\pi/2}\cos^k(\pi-\varphi)\,d\varphi=(-1)^k\int_0^{\pi/2}\cos^k\varphi\,d\varphi. $$ Theorefore, for every $k \ge 1$ we have $$ A_k=\int_0^{\pi/2}\cos^k\theta\,d\theta+\int_{\pi/2}^\pi\cos^k\theta\,d\theta=[1+(-1)^k]\int_0^{\pi/2}\cos^k\theta\,d\theta. $$ We deduce that $A_{2k+1}=0$ for all $k \ge 0$. For every $k\ge 1$ we have \begin{eqnarray} B_k&:=&A_{2k}=2\int_0^{\pi/2}\cos^{2k}\theta\,d\theta=2\int_0^{\pi/2}(\sin\theta)'\cos^{2k-1}\theta\,d\theta\\ &=&2(2k-1)\int_0^{\pi/2}\sin^2\theta\cos^{2k-2}\theta\,d\theta=2(2k-1)\int_0^{\pi/2}(1-\cos^2\theta)\cos^{2k-2}\theta\,d\theta\\ &=&(2k-1)(B_{k-1}-B_k), \end{eqnarray} i.e. $$ B_k=\frac{2k-1}{2k}B_{k-1} \quad \forall k \ge 1. $$ Thus $$ B_k=\frac{(2k-1)\cdot(2k-3)\ldots3\cdot1}{(2k)\cdot(2k-2)\ldots4\cdot2}B_0=\frac{(2k)!}{[(2k)(2k-2)\ldots4\cdot2]^2}B_0=\frac{(2k)!}{2^{2k}(k!)^2}\pi. $$ The given integral is then $$ I=\frac{1}{\pi}\sum_{k=0}^\infty\frac{2^{2k}}{(2k)!}A_{2k}=\frac{1}{\pi}\sum_{k=0}^\infty\frac{2^{2k}}{(2k)!}\cdot\frac{(2k)!}{2^{2k}(k!)^2}\pi=\sum_{k=0}^\infty\frac{1}{(k!)^2}. $$

$\endgroup$
  • $\begingroup$ There is an easier way to obtain this. Using parity, extend the integration to interval $(-\pi,\pi)$, then expand the exponential in Taylor series (as you do), and then expand $\cos^n\theta$ using binomial theorem. All the exponentials $e^{ik\theta}$ with $k\neq0$ will give zero integrals (the new bounds are crucial for this), so each $\cos^n\theta$ will produce only one non-zero term which more or less coincides with your $A_{2n}$. But +1 anyway. $\endgroup$ – Start wearing purple Jun 24 '13 at 20:25
  • $\begingroup$ @Mercy: Thanks for this solution, I will combine O.L.'s idea with your solution here to figure it out! $\endgroup$ – Samuel Reid Jun 24 '13 at 23:53
7
$\begingroup$

Let me post my comment as an answer since there is not much more to say here.


Mathematica says that this integral is equal to $I_0(2)$, where $I_0$ denotes the modified Bessel function of $0$th order. It is an easy consequence of its standard integral representation (see formula (4) here).

Bessel functions do not reduce to simpler expressions at integer values of the argument (except $0$) nor for order $0$, which means that this expression cannot be simplified further.

$\endgroup$
  • $\begingroup$ I was already aware of the representation in terms of Bessel functions, I was looking for an explicit solution as given in the other answer. $\endgroup$ – Samuel Reid Jun 24 '13 at 23:53

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.