\[\require{physics} \def\rn{\mathbb{R}^n} \def\half{\frac{1}{2}} \def\dd{\ \!\mathrm{d}} \def\dvvp#1#2{\frac{\partial #1}{\partial #2}} \def\dvvpn#1#2#3{\frac{\partial^#3 #1}{\partial #2^#3}} \def\of#1{\tilde{#1}} \def\ten#1{\mathsf{#1}} \def\diag#1{\mathrm{diag}{#1}} \def\abs#1{\left| #1 \right|} \def\pdv#1{\frac{\partial}{\partial #1}} %\def\dv#1{\frac{\dd}{\dd #1}} \def\ddv#1#2{\frac{\dd #1}{\dd #2}} \def\vdot{\mathbf{\cdot}}\]

Bessel Functions

Bessel functions are the solutions to Bessel’s differential equation,

(8)\[ x^2 \dv[2]{y}{x} + x \dv{y}{x} + (x^2 - \alpha^2)y = 0\]

for \(x \in \mathbb{C}\) and \(\alpha\) some arbitrary complex number.

When \(\alpha \in \mathbb{Z}\) the Bessel Functions are also Cylindrical Harmonics, representing the solutions to the Laplace equation in cylindrical coordinates. They can then be used to model the behaviour of wave equations in cylinders, such as heat propagation through a pipe.

When \(\alpha\) is of half-integer order the solutions represent the behaviour of functions in spherical coordinate systems.

More generally they represent a generalisation of the sine function, and can be thought of as representing the vibration of a string which has variable thickness or tension along it.

There are two families of Bessel function, which correspond to solutions to (8) which have a finite value at the origin (functions of the first kind) and those which have a singularity there (functions of the second kind).

../_images/bessel-1.png

Fig. 5 The Bessel Functions of the first kind over the real numbers .

../_images/bessel-2.png

Fig. 7 The Bessel Functions of the second kind over the real numbers .

../_images/bessel-3.png

Fig. 9 The Bessel Functions of the first kind with half-integer order over the complex number.

Bessel functions from the Generating Function

The Bessel functions can be described by a generating function,

(9)\[g(x,t) = \exp(\frac{x}{2t}(t^2-1)) = \sum_{\nu=-\infty}^{\infty} J_{\nu}(x) t^{\nu}\]

So, for Bessel functions of integer order we can expand this to form a series expansion,

(10)\[J_n(x) = \sum^{\infty}_{s=0} \frac{(-1)^s}{s! (n+s)!} \qty( \frac{x}{2} )^{n+2s} \approx \frac{x^n}{2^n n!}\]

for small \(x\).

Bessel functions with a negative index can be found from the relation

\[\label{eq:negativebessel} J_{-\nu}(x) = (-1)^{\nu} J_{\nu}(x)\]

Recurrence Relation for Bessel Functions

The Bessel functions can be descried by a pair of recurrence relations, found by differentiating with respect to \(t\),

\[\label{eq:recurrencebessel} J_{\nu-1}(x) + J_{\nu+1}(x) = \frac{2 \nu}{x} J_{\nu}(x)\]

and by differentiating with respect to \(x\),

\[\label{eq:recurrencebessel2} J_{\nu-1}(x) - J_{\nu+1}(x) = 2J_{\nu}^{\prime}(x)\]

A number of other integral relationships also exist.

\[\begin{split}\begin{aligned} \int x^n J_{n-1}(x) \dd{x} &= x^n J_n(x) \\ \int x^{-n} J_{n+1}(x) \dd{x} &= -x^{-n} J_n(x) \\ \int J_1(x) \dd{x} &= -J_0(x) \end{aligned}\end{split}\]

Orthogonality of the Bessel Functions

The orthogonality relations for Bessel functions are similar to those of the trigonometric functions, but they include an additional weighting factor, \(r\).

\[\label{eq:orthogbess} \int_0^a r J_p \qty( \frac{\alpha r}{a} ) J_p \qty(\frac{\beta r}{a}) \dd{r} = \delta_{\alpha \beta} \frac{a^2}{2} J_{p+1}^2(\alpha)\]

with

\[\begin{aligned} J_p(\alpha) = J_p(\beta) = 0\end{aligned}\]

Bessel Series

The orthogonality relations for Bessel functions allow the definition of Bessel series,

(11)\[ f(x) = \sum_0^{\infty} c_n J_p(k_n x)\]

with \(J_p(k_na)=0\).

Solving problems in cylindrical systems

Example 3 (Deriving the steady state inside an infinite cyclinder.)

With the curved sides kept at a temperature \(T_0\), and the base at \(T_1\).

We know \(\nabla^2 T =0\), and we can use seperation of variables to give a solution of the form \(T = R(r)\Theta(\theta)Z(z)\).

Then, in cylinderical coordinates,

\[\frac{1}{R}\frac{1}{r} \dv{r} \qty(r \dv{R}{r}) + \frac{1}{\Theta} \frac{1}{r^2} \dv[2]{\Theta}{\theta} + \frac{1}{Z} \dv[2]{Z}{z} = 0\]

We now have

\[\frac{1}{Z} \dv[2]{Z}{z} = k^2\]

implying

\[Z = \exp(\pm kz)\]

also,

\[\frac{1}{R}\frac{1}{r} \dv{r} \qty(r \dv{R}{r}) + \frac{1}{\Theta} \frac{1}{r^2} \dv[2]{\Theta}{\theta} + k^2 = 0\]

which we can multiply by \(r^2\),

\[\frac{r}{R} \dv{r} \qty(r \dv{R}{r}) + \frac{1}{\Theta} \dv[2]{\Theta}{\theta} + k^2 r^2 = 0\]

from which,

\[\frac{1}{\Theta} \dv[2]{\Theta}{\theta} = -n^2\]

implying that

\[\Theta = \{ \cos(n \theta), \sin(n \theta) \}\]

and the periodicity of \(\theta\) will force \(n\) to be a natural number. Then

\[\frac{r}{R} \dv{r} \qty(r \dv{R}{r}) + (k^2 r^2 - n^2) = 0\]

and letting \(kr = s\),

\[s \dv{s} \qty( s \dv{R}{s} ) + (s^2 - n^2)R = 0\]

which has the form of Bessel’s differential equation, equation (8), and thus the solutions are Bessel functions, \(J_n(s)\), the complete solution is thus

\[J_n (kr) \qty( A \sin(n\theta) + B \cos(n \theta) ) e^{-kz}\]

We can ignore the Bessel functions which are infinite at \(r=0\), as we need a finite solution there, so the functions of the first-kind are the appropriate solutions. We know that \(T_1 > T_0\), so \(T>T_0\) everywhere, and so \(T_0\) can be taken as a constant. The boundary condition of the curved surface at \(r=a\) is where \(J_n(ka) = 0\). We now need to know the zeros of the Bessel functions, and our solution becomes

\[T = T_0 + \sum_{m=0}^{\infty} c_m J_0 \qty(\alpha_{0m}\frac{r}{a}) \exp(-\qty(\frac{\alpha_{0m}z}{a}))\]

The boundary condition at \(z=0\) is that \(T=T_0\), so

\[T_1 - T_0 = \sum_m c_m J_0 \qty( \alpha_{0m} \frac{r}{a})\]

and using the orthogonality condition,

\[\int_0^a (T_1 - T_0) J_0 \qty( \alpha_{0m} \frac{r}{a} ) r \dd{r} = c_m \frac{a^2}{2} J_1^2(\alpha_{0m})\]

and then, from the indefinite integral relationship \(\int x J_0(x) \dd{x} = x J_1(x)\),

\[\begin{split}\begin{aligned} (T_1-T_0) \frac{a}{\alpha_{0m}} \qty[ r J_1 \qty( \alpha_{0m} \frac{r}{a})]_0^a &= (T_1 - T_0) \frac{a^2}{\alpha_{0m}} J_1 (\alpha_{0m})\\ &= c_m \frac{a^2}{2} J_1^2 (\alpha_{0m}) \end{aligned}\end{split}\]

with

\[c_m = \frac{2}{\alpha_{0m}} \frac{1}{J_1(\alpha_{0m})} (T_1-T_0)\]

and the overarching solution is thus

\[T = T_0 + \sum_m \frac{2 (T_1-T_0)}{\alpha_{0m}J_1(\alpha_{0m})} J_0 \qty( \alpha_{0m} \frac{r}{a}) \exp( - \qty(\frac{\alpha_{0m}z}{a}) )\]

Spherical Bessel Functions

The spherical Bessel functions are a class of Bessel function related to the half-integer order order Bessel functions by

(12)\[ j_n(x) = \sqrt{\frac{\pi}{2x}} J_{n+\frac{1}{2}}(x) = x^n \qty(- \frac{1}{x} \dv{x})^n \frac{\sin(x)}{x}\]

Example 4 (Finding energy levels of particles inside a spherical box using Schrodinger’s equation.)

Starting at

\[- \frac{\hbar^2}{2m} \nabla^2 \Psi = E \Psi\]

after seperating variables

\[\pdv{r} \qty(r^2 \pdv{R}{r}) + \qty( \frac{2mEr^2}{\hbar^2} - l(l+1) )R=0\]

letting

\[k^2 = \frac{2mE}{\hbar^2} \quad \text{and} \quad s=kr\]
\[s^2 \pdv[2]{R}{s} + 2s \pdv{R}{s} + \qty(s^2 - l(l+1))R = 0\]

and letting

\[R = \frac{Z}{s^{\frac{1}{2}}}\]
\[s^2Z^{\prime \prime} + s Z^{\prime} + (s^2 - \qty(l + \frac{1}{2})^2 ) Z = 0\]

Which is Bessel’s equation of order \(l+\half\), so

\[R = j_l \qty( \frac{\sqrt{2mE}}{\hbar} r)\]

which is a finite solution as \(r \to 0\). The lowest energy state will have \(l=0\) (so no angular variation), and to satisy the boundary condition of \(R=0\) when \(r=a\), we need

\[j_0 \qty( \frac{\sqrt{2mE}}{\hbar}a)=0\]

the zeros of \(j_0\) are the same as those of \(\sin(x)\), since

\[j_0(x) = \frac{\sin(x)}{x}\]

so

\[\frac{a \sqrt{2mE_{\rm min}}}{\hbar} = \pi\]

thus

\[E_{\rm min} = \frac{\pi^2 \hbar^2}{2ma^2}\]
© Copyright 2020-2022, Daniel Williams.
Created using Sphinx 4.5.0.