The Bessel Equation
In this note we investigate series solutions to the Bessel equation
Solutions to the Bessel equation are transcendental functions called Bessel functions, and they have wide application. For instance they appear in the classical theory of modular forms on the upper half plane, and the solutions of the heat equation and Schrodinger equation in cylindrical coordinates. There are many different kinds of Bessel function. Each of these tends to be best suited to a specific type of application. In this note we treat the most common kinds of Bessel function.
For convenience, we introduce the notation
for the differential operator in \ref{eq:bessel}. The Bessel equation can then be rewritten as
Bessel Functions of the First Kind
Derivation of Solutions
Note that our equation has a regular singular point at \(z = 0\). In this section we follow the usual approach for finding series solutions to ODE with regular singularities. Namely, we look for solutions of the form
where \(a_0 \ne 0\). Substituting this into equation \ref{eq:bessel} yields
Comparing \(z^{n+\alpha}\)-coefficients in the equation above produces a series of identities involving the \(a_n\) terms. We consider the coefficients on \(z^\alpha\), \(z^{1+\alpha}\), and \(z^{n+\alpha}\) separately. To keep matters simple for the moment, it will be useful to assume that the terms \((n + \alpha)^2 - \nu^2\) appearing above do not vanish for any integer \(n \geq 1\); we treat the remaining cases later. Since
it will be enough to assume \(\alpha \pm \nu\) is not a negative integer.
Case \(z^{\alpha}\): Comparing these coefficients shows that
Since \(a _ 0 \ne 0\), this implies \(\alpha = \pm \nu\). The restriction \(\alpha \pm \nu \not\in \mathbb{Z} _ {<0}\) thus reduces to requiring \(-2\nu \not\in \mathbb{Z} _ {<0}\) when \(\alpha = -\nu\) (resp. \(2\nu \not\in \mathbb{Z}_{<0}\) when \(\alpha = \nu\)).
Case \(z^{\alpha + 1}\): From this coefficient we obtain
We have assumed \((\alpha \pm \nu) \ne - 1\), so the above implies \(a_1 = 0\).
Case \(z^{\alpha + n}\): This time we obtain a recurrence of the form
Rearranging, this is the same as
This allows us to quickly compute the odd-order terms \(a_n = a_{2k - 1}\). In particular, bootstrapping from the above case,
Repeating this for \(n = 5, 7, \dots\) demonstrates that \(a _ {2k-1} = 0\) for all \(k \in \mathbb{Z} _ {>0}\). Induction also allows us to compute the even-order terms \(a_n = a_{2k}\). Indeed, when \(n = 2\) we have
and in general
If we take \(\alpha = \nu\), this reduces to
Notice that there is no restriction on \(a_0\).
Combining all of our work so far, we arrive at our first candidate solution for the Bessel equation:
Since we have no restriction on \(a_0\), it is convenient to choose \(a_0 = \frac{1}{2^\nu \Gamma(\nu + 1)}\). The solution \(u\) in equation \ref{eq:candidate1} then simplifies to
If we had instead choosen \(\alpha = -\nu\) above, we could have set \(a_0 = \frac{1}{2^{-\nu} \Gamma(-\nu + 1)}\). This leads to a solution of the form
In all situations considered so far (ie. when \(2\nu\) is not an integer) the sums defining \(J_{\pm \nu}\) converge for all \(z \in \mathbb{C}\setminus \{0\}\).
Definition 1: The functions \(J_\nu\) and \(J_{-\nu}\) defined above are called Bessel functions of the first kind. ∎
It can be shown, eg. with the Weierstrass M-test, that the sum defining \(J_\nu\) converges absolutely and uniformly in a neighbourhood of any \(z \in \mathbb{C} \setminus \{ 0 \}\) and in any bounded domain of values of \(\nu\).
Fundamental System of \(1^{st}\)-kind solutions.
In this section we show that, provided \(\nu \not\in \mathbb{Z}\), the system \(\{ J_\nu, J_{-\nu}\}\) forms a linearly independent basis of solutions for the Bessel equation \ref{eq:bessel}. To do so, it suffices to show the Wronskian determinant \(\mathcal{W}(J_\nu, J_{-\nu}) \) does not vanish. In other words, we want to show
First, we attempt to extract a Wronskian-like quantity from the Bessel equation by computing
The whole expression equals zero since \(L_\nu[J_{-\nu}] = 0 = L_\nu[J_\nu]\). The bracketed term on the last line above somewhat resembles the product \(z \mathcal{W}(J_\nu, J_{-\nu})\). The main differences between the two are the second-order derivatives and factor of \(z\). Both of these differences can be accounted for by taking a derivative. We compute:
In other words, the function \(z \mapsto z\cdot \mathcal{W}(J_\nu(z), J_{-\nu}(z))\) is constant. Solving for the Wronskian, we have
In order to evaluate the constant \(C\), first let us note the easily obtained formulae, valid when \(\nu \not\in \mathbb{Z}\) and for small \(|z|\):
Similar formulae exist for \(J_{-\nu}\) and \(\frac{\partial J_{-nu}}{\partial z}\). This allows us to compute
In light of equation \ref{eq:frac} the \(O(z)\) term above must vanish. We conclude
Since \(\nu \not\in \mathbb{Z}\), \(\sin(\nu z\) does not vanish. We arrive at the following theorem:
Theorem 1: When \(\nu \not\in \mathbb{Z}\), the \(1^{st}\)-kind functions \(J_\nu\) and \(J_{-\nu}\) form a fundamental system of solutions to Bessel’s equation. ∎
On account of the relation \(J_n(z) = (-1)^n J_{-n}(z)\), these functions do not form a fundamental system when \(\nu = n \in \mathbb{Z}\).
Bessel Functions of the Second Kind
Derivation of Solutions
In the previous section we derived a fundamental system of solutions to the Bessel equation \ref{eq:bessel} subject to the condition \(\pm 2\nu \ne n\) for any integer \(n\in \mathbb{Z}\). That is, we required \(\nu\) not be an even integer, nor half of an odd integer. In this section we shall derive a solution that is valid even in the case \(\nu \in \mathbb{Z}\).
For the remainder of this section let us fix some \(n \in \mathbb{Z}\). Now, since \(J_\nu\) and \(J_{-\nu}\) solve the Bessel equation for non-integer \(\nu\), so does the linear combination
If we then set
it seems reasonable to believe that \(\mathbf{Y}\) is also a solution to Bessel’s equation. In order to check this, we recall \(J_\nu\) is analytic in \(\nu\) and compute
We will show this expression solves equation \ref{eq:bessel}. First we exploit the equality of mixed partials to write