Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

For $ \nu \in \Bbb R$, I want to prove the well-known formula $$ J_\nu (x) \sim \sqrt{\frac{2}{\pi x}} \cos \left( x - \frac{2 \nu +1}{4} \pi \right) + O \left( \frac{1}{x^{3/2}} \right) \;\;\;\;(x \to \infty)$$ where $J_\nu$ denotes the Bessel function. How can I show this? Or would you tell me the Internet site which proves this formula? I could not find the proof of this.

share|cite|improve this question
See this. – J. M. Jul 31 '12 at 10:54
In the Olden Days we would find things like this in textbooks (such as G. N. Watson). Nowadays everyone wants things to be available on-line! – GEdgar Jul 31 '12 at 17:52
up vote 7 down vote accepted

First, some preliminary series expansions. Consider the substitution $$ \cos(t)=1-u^2/2\tag{1} $$ We get the power series for $u=2\sin(t/2)$: $$ u=t-t^3/24+t^5/1920-t^7/322560+t^9/92897280+O(t^{11})\tag{2} $$ and the inverse series for $t$; $$ t=u+u^3/24+3u^5/640+5u^7/7168+35u^9/294912+O(u^{11})\tag{3} $$

We need to concentrate on the stationary points at $t=\pi/2$ and $t=-\pi/2$, away from which the integral decays exponentially in $x$. The contribution at $-\pi/2$ is the conjugate of the contribution at $\pi/2$, so the whole contribution is twice the real part of the contribution at $\pi/2$. $$ \begin{align} J_\nu(x) &=\frac1{2\pi}\int_{-\pi}^\pi e^{-i(\nu t-x\sin(t))}\,\mathrm{d}t\\ &=2\mathrm{Re}\left(\frac1{2\pi}\int_0^\pi e^{-i(\nu t-x\sin(t))}\,\mathrm{d}t\right)\\ &=2\mathrm{Re}\left(e^{-i\nu\pi/2}\frac1{2\pi}\int_{-\pi/2}^{\pi/2} e^{-i\nu t}e^{ix\cos(t)}\,\mathrm{d}t\right)\\ &\sim2\mathrm{Re}\left(e^{-i\nu\pi/2}\frac1{2\pi}\int_{-\infty}^\infty\left(1-\nu^2u^2/2+O(u^4)\right)e^{ix(1-u^2/2)}\left(1+u^2/8+O(u^4)\right)\,\mathrm{d}u\right)\\ &=2\mathrm{Re}\left(e^{i(x-\nu\pi/2)}\frac1{2\pi}\int_{-\infty}^\infty\left(1-(4\nu^2-1)u^2/8+O(u^4)\right)e^{-ixu^2/2}\,\mathrm{d}u\right)\\ &=2\mathrm{Re}\left(e^{i(x-(2\nu+1)\pi/4)}\frac1{2\pi}\int_{-\infty}^\infty\left(1+i(4\nu^2-1)v^2/8+O(v^4)\right)e^{-xv^2/2}\,\mathrm{d}v\right)\\ &=2\mathrm{Re}\left(e^{i(x-(2\nu+1)\pi/4)}\frac1{2\pi}\left(\sqrt{\frac{2\pi}{x}}+i\frac{4\nu^2-1}{8}\frac1{2\pi}\sqrt{\frac{2\pi}{x}}^3+O\left(x^{-5/2}\right)\right)\right)\\ &=\cos\left(x-\frac{2\nu+1}{4}\pi\right)\sqrt{\frac{2}{\pi x}}-\sin\left(x-\frac{2\nu+1}{4}\pi\right)\frac{4\nu^2-1}{8}\sqrt{\frac{2}{\pi x^3}}+O\left(x^{-5/2}\right) \end{align} $$ In the $\sim$ step, we make the $u$ substitution whose series is given above, and integrate over $(-\infty,\infty)$ instead of $[-\sqrt{2},\sqrt{2}]$ since the part outside the compact interval decays exponentially.

We also use the substitution $u=e^{-i\pi/4}v$, so that $u^2=-iv^2$, and change the path of integration, which is allowed since there are no singularities.

In the end, we get the first two terms of the asymptotic expansion of $J_\nu(x)$.

share|cite|improve this answer
(+1) Love the saddle point method. – Jack D'Aurizio Aug 31 '14 at 2:58
It is a very powerful tool. I have used it in a lot of asymptotic expansions (my graduate advisor used it a lot in dealing with singular integrals). – robjohn Aug 31 '14 at 3:19

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.