Bessel’s equation is given by
$$ x^2 \frac{\text d^2 y}{\text dx^2}+x\frac{\text dy}{\text dx}+(x^2-m^2)y=0 $$
subject to boundary conditions $y$ is finite at $x=0$
We are solving using it using the method of Frobenius which uses a series solution with a leading factor of $x^s$ to account for non-constant behaviour at small $x$
$$ y(x)=x^s\sum^\infin_{n=0}a_n x^n $$
🗒️ Note: $s$ could be $s<0$ and could be complex in general
Take the derivative
$$ \begin{aligned} x\frac{\text dy}{\text dx}&=\sum^\infin_{n=0}a_n(n+s)x^{n+s} \\ x^2 \frac{\text d^2 y}{\text dx^2}&=\sum^\infin_{n=0} a_n(n+s)(n+s-1)x^{n+s} \\ (x^2-m^2)y&=\sum^\infin_{n=0}a_n x^{n+s+2}-m^2 \sum^\infin_{n=0} a_n x^{n+s} \end{aligned} $$
To help equating coefficients we write this first term on the right hand side as
$$ \sum_{n=2}^\infin a_{n-2}x^{n+s} $$
Subbing in we get
$$ \begin{aligned} &\sum^\infin_{n=0} a_n(n+s)(n+s-1)x^{n+s}+\sum^\infin_{n=0} a_n(n+s)x^{n+s} \\&+\sum^\infin_{n=2}a_{n-2} x^{n+s}-m^2 \sum^\infin_{n=0} a_n x^{n+s}=0 \end{aligned} $$
We can now equate coefficients
when $n=0$
$$ \begin{aligned} a_0 s(s-1)+a_0s-m^2a_0&=0 \\ a_0(s^2-m^2)&=0 \\ \Rightarrow \qquad s&=\pm m
\end{aligned} $$
when $n=1$
$$ \begin{aligned} a_1s(s+1)+(1+s)a_1-m^2a_1&=0 \\ \left [ (s+1)^2-m^2\right ]a_1&=0 \\ a_1&=0 \end{aligned} $$
when $n\ge2$
$$ \begin{aligned} [(n+s)(n+s-1)+(n+s)-m^2]a_n+a_{n-2}=0 \\ \left [ (n+s)^2-m^2 \right ]a_n+a_{n-2}=0
\end{aligned} $$
For solutions that are regular at the origin we take $s=+m>0$ we have
$$ \begin{aligned} a_2&=-\frac{a_0}{(m+2)^2-m^2} \\ a_4&=-\frac{a_2}{(m+4)^2-m^2} \\ \vdots\;&\qquad \qquad \quad \vdots
\end{aligned} $$
We can expand these solutions as
$$ \begin{aligned} a_2&= -\frac{a_0}{2^2(1+m)}=-\frac{a_0}{2(2+2m)} \\ a_4&=-\frac{a_2}{2^3(2+m)}=\frac{a_0}{2.4(2+2m)(4+2m)}=\frac{a_0}{2^{2.2}2!(1+m)(2+m)} \end{aligned} $$
We can also evaluate $a_6$ by a similar procedure
$$ a_6=-\frac{a_0}{2^{2.3}3!(1+m)(2+m)(3+m)} $$
We also have $a_n=0$ is $n=$ odd. The series converges for all values $x$ since the ratio of the $(n+2)$th to the $n$th term in the series is
$$ \left | \frac{a_{n+2}x^{n+2}}{a_nx^n} \right |=\frac{x^2}{(n+2+m)^2-m^2} $$
which vanishes as $n\to\infin$ for all $x$
For positive integer values of $m$ we can write the expression in terms of factorials
$$ \begin{aligned} &=\frac{1}{(1+m)(2+m)\ldots(j+m)}\\ &=\frac{1.2.3\ldots (m-1)m}{1.2.3\ldots (m-1)(m)(1+m)\ldots (m+j)}\\&=\frac{m!}{(m+j)!} \end{aligned} $$
Thus in general for integer $m$ we have
$$ a_{2j}=\frac{(-1)^ja_0 m!}{2^{2j}j!(m+j)!} $$
$m=0,1,2,\ldots$
🗒️ Note: this can be generalised to non-integer $m$ using the gamma function
This gives a series of coefficients in the original power series and a solution to Bessel’s equation
$$ \begin{aligned} y(x)&=\sum^\infin_{n=0}a_nx^{m+n} \\ &=\sum^\infin_{j=0}a_{2j}x^{m+2j} \\ &=\sum^\infin_{j=0}\frac{(-1)^ja_0 m!}{2^{2j}j!(m+j)!}x^{m+2j} \\ &=a_0m! 2^m\sum^\infin_{j=0}\frac{(-1)^j}{2^{2j}2^mj!(m+j)!}x^{m+2j} \\ &=a_0m! 2^m\sum^\infin_{j=0}\frac{(-1)^j}{j!(m+j)!}\left ( \frac x2 \right )^{m+2j}
\end{aligned} $$
If $m$ is an integer it is customary to write
$$ a_0=\frac{1}{2^mm!} $$
Thus we have
$$ y(x)=a_02^mm! J_m(x) $$
where
$$ J_m(x)=\sum^\infin_{j=0}(-1)^j\frac{1}{j!(m+j)!}\left ( \frac x2 \right ) ^{m+2j} $$
This series solution corresponds to Bessel’s equation of the first kind
🗒️ Notes:
$y(x)\propto J_m(x)$
All others are zero at the origin $x=0$
$J_0(0)=1$
The orthogonality relation for these Bessel functions
$$ \int^\ell_0xJ_p\left ( \frac{\chi_n}{\ell}x\right ) J_p\left ( \frac{\chi_m}{\ell}x\right ) \,\text dx=\left \{ \begin{aligned} &\frac{\ell^2}{2}J^2_{p+1}(\chi_m) \; &n=m \\ &0 \quad &n\ne m \end{aligned}\right . $$
where we have
$$ J_p(\chi_n)=0 \qquad n\in \N^* $$
which corresponds to locating the zeros of the Bessel function $J_p$
Start with the wave equation
$$ \nabla^2\phi=\frac{1}{c^2}\frac{\partial^2\phi}{\partial t^2} $$
We use separation of variables $\phi=F(r,\theta)T(t)$
$$ \frac{1}{F}\nabla^2F=\frac{1}{C^2 T}\frac{\partial^2 T}{\partial t^2}=-k^2 $$
where $k^2>0$ is a constant to be determined
$$ \begin{aligned} \nabla^2F+k^2F&=0 \\ \frac{\partial^2 T}{\partial t^2}+k^2c^2T&=0 \end{aligned} $$
write the equation for $F$ in polar coordinates
$$ \frac{1}{r}\frac{\partial}{\partial r}\left ( r\frac{\partial F}{\partial r} \right ) + \frac{1}{r^2}\frac{\partial^2F}{\partial \theta^2}+k^2 F=0 $$
make another separation of variables $F(r,\theta)=R(r)\Theta(\theta)$
$$ \underbrace{\frac{r}{R}\frac{\text d}{\text dr} \left ( r\frac{\text dR}{\text dr} \right ) +k^2 r^2}{=n^2} +\underbrace{\frac{1}{\Theta}\frac{\text d^2 \Theta}{\text d \theta^2}}{-n^2}=0 $$
We can now solve
$$ \begin{aligned} \frac{1}{\Theta}\frac{\text d^2 \Theta}{\text d \theta^2}+n^2&=0 \quad \Theta=\left \{ \begin{matrix} \sin n\theta \\ \cos n\theta \end{matrix} \right . \\ \frac{r}{R}\frac{\text d R}{\text dr}\left ( r \frac{\text d R}{\text dr}\right )-n^2+k^2r^2&=0 \end{aligned} $$
🗒️ Note: we have the boundary condition $\Theta(\theta+2\pi n)=\Theta(\theta)$ so $n\in\Z$
We can rewrite the equation for $R$ as
$$ r\frac{\text d }{\text dr}\left ( r \frac{\text d R}{\text dr}\right )+(k^2r^2-n^2)=0 $$
Thus its solutions are
$$ R(r)=J_n(kr) $$
which are of the form
$$ \phi=J_n(kr)\left \{ \begin{matrix} \sin n\theta \sin kct \\ \sin n\theta \cos kct \\ \cos n\theta \sin kct \\ \cos n\theta \cos kct \end{matrix} \right . $$
If we apply the boundary condition that the circumference at $r=a$ is fixed then we have
$$ J_n(ka)=0 $$
and hence $k_{n,m}a=\chi_{n,m}$ is the $m$th root of the Bessel function $J_n(x)$. The angular frequency of the resonant modes is given by
$$ \omega_{n,m}=k_{n,m}c=\frac{\chi_{n,m}c}{a} $$
🗒️ Notes:
The lowest angular frequency of vibration, the fundamental mode ($\omega_0,1=\frac{\chi_{0,1}c}a\approx \frac{2.405c}{a}$) corresponds to the membrane vibrating as a whole with the peak in the center
The $(0,2)$ mode oscillates at an angular frequency $\omega_{0,2}=\frac{\chi_{0,2}c}{a}\approx \frac{5.520c}{a}$ and one half is up whilst the other half of the membrane is down. There is a circle in-between them which remains at rest and this is known as a node. We can see where this node is located according to $J_0(\chi_{0,2}r/a)=0$ if we set $r=a\chi_{0,1}/\chi_{0,2}$ then we have $J_0(\chi_{0,2}\chi_{0,1}/\chi_{0,2})=J_0(\chi_{0,1})=0$ and hence the nodal circle must be located at $r=a\chi_{0,1}/\chi_{0,2}\approx 0.436 a$ (where $\chi_{0,2}>\chi_{0,1}$)
Finally we can note that the solution is given by the doubly infinite sum
$$ \begin{aligned} \phi&=\sum^\infin_{n=0}\sum^\infin_{m=0}J_n(k_{n,m}r)(A_{n,m}\cos n\theta+B_{n,m} \sin n\theta)\cos\omega_{n,m} t \\ \tilde \phi&=\sum^\infin_{n=0}\sum^\infin_{m=0}J_n(k_{n,m}r)(\tilde A_{n,m}\cos n\theta+\tilde B_{n,m} \sin n\theta)\sin \omega_{n,m} t
\end{aligned} $$
where
$$ \omega_{n,m}=ck_{n,m}=c\chi_{n,m}/a $$
and the initial conditions will determine the constants
💼 Case: Consider vibrations of a membrane fixed to a rectangular frame.
$$ \frac{\partial^2\phi}{\partial x^2}+\frac{\partial^2 \phi}{\partial y^2}=\frac 1{c^2}\frac{\partial ^2 \phi}{\partial t^2} $$
Subject to the boundary conditions
$$ \phi(0,y,t)=\phi(a,y,t)=\phi(x,0,t)=\phi(x,b,t)=0 $$