Professional Documents
Culture Documents
Bessel Functions
63
8.1 Bessel functions of integer order
The functions Jm (kρ) are the analogues of the associated Legendre functions Plm (cos θ) ,
except that k is a continuous index (unlike l), if the domain of definition is 0 ≤ ρ ≤ ∞.
As in the case of Plm and Pl−m , negative values of m give nothing new and we can choose
J−m (kρ) = (−1)m Jm (kρ). The functions Jm have series expansions and recursion relations
(given later), but there is no Rodrigues formula for them, since they are not reducible to
polynomials. Instead, there is the very useful integral representation
1 Z 2π dϕ
Jm (x) = m exp (ix cos ϕ − imϕ) , (8.6)
i 0 2π
that can be written in many other ways, including
Z π dϕ
Jm (x) = cos (x cos ϕ − mϕ) . (8.7)
0 π
Closely related to these formulas (through Fourier’s theorem) is the generating function
expansion
∞
X
exp(ikρ cos ϕ) = im eimϕ Jm (kρ). (8.8)
n=−∞
The useful lore about Bessel functions includes a large number of special integrals that
we will discuss as needed. We note here only the orthogonality and normalization integral
over the domain (0, ∞)
Z ∞ 1
Jm (kρ)Jm (k 0 ρ) ρ dρ = δ(k − k 0 ). (8.9)
0 k
Successive Jm functions have interlacing zeros and maxima, like the Legendre func-
tions. It is important to remember that Jm (x) behaves like xm for small x, so it is regular
at the origin.
For a given m2 there is also the “second solution” of the Bessel differential equation,
which for small x behaves like x−m for m > 0 and like ln x for m = 0. Its standard form is
known as the Neumann function, or the Bessel function of the second kind, and is denoted
with Nm (x) by some authors, Ym (x) by others. It is discussed further here below. It is used
in problems where ρ = 0 is not part of the domain.
We close this section by writing down the general “interior” solutions of the Laplace
equations in cylindrical polar coordinates. They are, for k > 0,
e±kz e±imφ Jm (kρ), (8.10)
and represent standing waves in the radial direction, evanescent waves in the z direction. If
instead of the Laplace equation we had solved the Helmholtz equation, ∇2 ψ = −(ω/c)2 ψ,
the solutions would be
e±ikz z e±imφ Jm (kρ ρ) (8.11)
with kz2 + kρ2 = (ω/c)2 .
64
8.2 General Bessel and Hankel functions
Many properties of the Bessel functions hold for any value of the index ν and are sometimes
easier to discuss for non-integer ν.
• Recursion relations that hold for Jν as well as for Nν , Hν(1) , Hν(2) are
2ν
Jν−1 + Jν+1 = Jν , (8.17)
x
dJν
Jν−1 − Jν+1 = 2 . (8.18)
dx
They are useful to compute integrals, but are numerically unstable as step-up relations.
• The behavior for small x can be read from the series (see J3.89-90). Note however the
special case
2 x
N0 (x) → ln + γ , (8.19)
π 2
where γ ' 0.5772 is the Euler-Mascheroni constant.
65
• Of special interest for fields confined inside a conducting cylinder are the zeroes of
Jm (x). They are tabulated by Jackson. The orthogonality and normalization relations
over a finite domain are also given. These relations are of interest to people who
play (round) drums. They will also be useful to discuss the propagation of waves in
cylindrical fibers and waveguides (when we get to it).
all we have to require in the zero-frequency limit is that kz2 + kρ2 = 0. Clearly, in addition to
the option kz = ik, kρ = k, we have the option kz = k, kρ = ik, which gives the electrostatic
solutions
e±ikz e±imφ Im (kρ), (8.22)
where, by definition,
Im (x) = i−m Jm (ix) (8.23)
is a modified Bessel function. I0 (x) resembles cosh x, I1 (x) resembles sinh x. Every Im is well-
behaved at the origin, but not at infinity: thus these functions are to be used for “interior”
problems. The standard “second solution” is the modified Hankel function
π m+1 (1)
Km (x) = i Hm (ix). (8.24)
2
This blows up at the origin and vanishes at infinity: thus it is to be used for “exterior”
problems. The above definitions are valid for any ν, not just for integer values m. The
prefactors are chosen to make Iν (x) and Kν (x) real for real ν and x.
66