MATH 311
Topics in Applied Mathematics
Lecture 25:
Bessel functions (continued).
Bessel’s differential equation of order m ≥ 0:
d 2f
2 df
z 2
+ z + (z 2 − m2 )f = 0
dz dz
The equation is considered on the interval (0, ∞).
Solutions are called Bessel functions of order m.
Jm (z): Bessel function of the first kind,
Ym (z): Bessel function of the second kind.
The general Bessel function of order m is
f (z) = c1 Jm (z) + c2 Ym (z), where c1 , c2 are
constants.
Bessel functions of the 1st and 2nd kind
Asymptotics at the origin
Jm (z) is regular while Ym (z) has a singularity at 0.
As z → 0, we have for any integer m > 0
1 2m (m − 1)! −m
Jm (z) ∼ m z m , Ym (z) ∼ − z .
2 m! π
2
Also, J0 (z) ∼ 1, Y0 (z) ∼ log z.
π
To get the asymptotics for a noninteger m, we
replace m! by Γ(m + 1) and (m − 1)! by Γ(m).
Jm (z) is uniquely determined by this asymptotics
while Ym (z) is not.
Asymptotics at infinity
As z → ∞, we have
r
2 π mπ
Jm (z) = cos z − − + O(z −1 ),
πz 4 2
r
2 π mπ
Ym (z) = sin z − − + O(z −1 ).
πz 4 2
Both Jm (z) and Ym (z) are uniquely determined by
this asymptotics.
For m = 1/2, these are exact formulas.
Original definition by Bessel (only for integer m):
1 π
Z
Jm (z) = cos(z sin τ − mτ ) dτ
π 0
1 1
Z π Z π
= cos(z sin τ ) cos(mτ ) dτ + sin(z sin τ ) sin(mτ ) dτ .
π 0 π 0
The first integral is 0 for any odd m while the
second integral is 0 for any even m. It follows that
X∞
cos(z sin τ ) = J0 (z) + 2 J2n (z) cos(2nτ ),
n=1
X∞
sin(z sin τ ) = 2 J2n−1 (z) sin (2n − 1)τ .
n=1
Zeros of Bessel functions
Let 0 < jm,1 < jm,2 < . . . be zeros of Jm (z) and
0 < ym,1 < ym,2 < . . . be zeros of Ym (z).
′ ′
Let 0 ≤ jm,1 < jm,2 < . . . be zeros of Jm′ (z) and
′ ′
0 < ym,1 < ym,2 < . . . be zeros of Ym′ (z).
′ ′
(We let j0,1 = 0 while jm,1 > 0 if m > 0.)
Then the zeros are interlaced:
′ ′
m ≤ jm,1 < ym,1 < ym,1 < jm,1 <
′ ′
< jm,2 < ym,2 < ym,2 < jm,2 < . . .
Asymptotics of the nth zeros as n → ∞:
′
jm,n ≈ ym,n ∼ (n + 12 m − 34 )π,
′
ym,n ≈ jm,n ∼ (n + 21 m − 14 )π.
Dirichlet Laplacian in a circle
Eigenvalue problem:
∇2 φ + λφ = 0 in D = {(x, y ) : x 2 + y 2 ≤ R 2 },
φ|∂D = 0.
Separation of variables in polar coordinates:
φ(r , θ) = f (r )h(θ). Reduces the problem to two
one-dimensional eigenvalue problems:
r 2 f ′′ + rf ′ + (λr 2 − µ)f = 0, f (R) = 0, |f (0)| < ∞;
h′′ = −µh, h(−π) = h(π), h′ (−π) = h′ (π).
The latter problem has eigenvalues µm = m2 ,
m = 0, 1, 2, . . . , and eigenfunctions h0 = 1,
hm (θ) = cos mθ, h̃m (θ) = sin mθ, m ≥ 1.
The 1st intermediate eigenvalue problem:
r 2 f ′′ + rf ′ + (λr 2 − m2 )f = 0, f (R) = 0, |f (0)| < ∞.
√
New variable z = λ · r reduces the equation to
Bessel’s equation of order√m. Hence the√general
solution is f (r ) = c1 Jm ( λ r ) + c2 Ym ( λ r ),
where c1 , c2 are constants.
Singular condition |f (0)| < ∞√holds if c2 = 0.
Nonzero solution exists if Jm ( λ R) = 0.
Thus there are infinitely many eigenvalues λm,1 , λm,2 , . . . ,
p
where λm,n R = jm,n , i.e., λm,n = (jm,n /R)2 .
Associated eigenfunctions: fm,n (r ) = Jm (jm,n r /R).
The eigenfunctions fm,n (r ) = Jm (jm,n r /R) are
orthogonal relative to the inner product
Z R
hf , g ir = f (r ) g (r ) r dr .
0
Any piecewise continuous function g on [0, R] is
expanded into a Fourier-Bessel series
X∞ r hg , fm,n ir
g (r ) = cn Jm jm,n , cn = ,
n=1 R hfm,n , fm,n ir
that converges in the mean (with weight r ).
If g is piecewise smooth, then the series converges
at its points of continuity.
Eigenvalue problem:
∇2 φ + λφ = 0 in D = {(x, y ) : x 2 + y 2 ≤ R 2 },
φ|∂D = 0.
Eigenvalues: λm,n = (jm,n /R)2 , where
m = 0, 1, 2, . . . , n = 1, 2, . . . , and jm,n is the nth
positive zero of the Bessel function Jm .
Eigenfunctions: φ0,n (r , θ) = J0 (j0,n r /R).
For m ≥ 1, φm,n (r , θ) = Jm (jm,n r /R) cos mθ and
φ̃m,n (r , θ) = Jm (jm,n r /R) sin mθ.
Neumann Laplacian in a circle
Eigenvalue problem:
∇2 φ + λφ = 0 in D = {(x, y ) : x 2 + y 2 ≤ R 2 },
∂φ
∂n ∂D = 0.
Again, separation of variables in polar coordinates,
φ(r , θ) = f (r )h(θ), reduces the problem to two
one-dimensional eigenvalue problems:
r 2 f ′′ + rf ′ + (λr 2 − µ)f = 0, f ′ (R) = 0, |f (0)| < ∞;
h′′ = −µh, h(−π) = h(π), h′ (−π) = h′ (π).
The 2nd problem has eigenvalues µm = m2 ,
m = 0, 1, 2, . . . , and eigenfunctions h0 = 1,
hm (θ) = cos mθ, h̃m (θ) = sin mθ, m ≥ 1.
The 1st one-dimensional eigenvalue problem:
r 2 f ′′ + rf ′ + (λr 2 − m2 )f = 0, f ′ (R) = 0, |f (0)| < ∞.
For λ > 0, the √ general solution
√ of the equation is
f (r ) = c1 Jm ( λ r ) + c2 Ym ( λ r ), where c1 , c2 are
constants.
Singular condition |f (0)| < ∞√holds if c2 = 0.
Nonzero solution exists if Jm′ ( λ R) = 0.
Thus there are infinitely many eigenvalues λm,1 , λm,2 , . . . ,
′ ′
p
where λm,n R = jm,n , i.e., λm,n = (jm,n /R)2 .
′
Associated eigenfunctions: fm,n (r ) = Jm (jm,n r /R).
λ = 0 is an eigenvalue only for m = 0.
Eigenvalue problem:
∇2 φ + λφ = 0 in D = {(x, y ) : x 2 + y 2 ≤ R 2 },
∂φ
= 0.
∂n ∂D
′
Eigenvalues: λm,n = (jm,n /R)2 , where
′
m = 0, 1, 2, . . . , n = 1, 2, . . . , and jm,n is the nth
′ ′
positive zero of Jm (exception: j0,1 = 0).
′
Eigenfunctions: φ0,n (r , θ) = J0 (j0,n r /R).
In particular, φ0,1 = 1.
′
For m ≥ 1, φm,n (r , θ) = Jm (jm,n r /R) cos mθ and
′
φ̃m,n (r , θ) = Jm (jm,n r /R) sin mθ.
Laplacian in a circular sector
Eigenvalue problem:
∇2 φ + λφ = 0 in D = {(r , θ) : r < R, 0 < θ < L},
φ|∂D = 0.
Again, separation of variables in polar coordinates,
φ(r , θ) = f (r )h(θ), reduces the problem to two
one-dimensional eigenvalue problems:
r 2 f ′′ + rf ′ + (λr 2 − µ)f = 0, f (0) = f (R) = 0;
h′′ = −µh, h(0) = h(L) = 0.
The 2nd problem has eigenvalues µm = ( mπ 2
L ) ,
m = 1, 2, . . . , and eigenfunctions hm (θ) = sin mπθ
L .
The 1st one-dimensional eigenvalue problem:
r 2 f ′′ + rf ′ + (λr 2 − ν 2 )f = 0, f (0) = f (R) = 0.
Here ν 2 = µm . We may assume that λ > 0.
The general √ solution of the√equation is
f (r ) = c1 Jν ( λ r ) + c2 Yν ( λ r ), where c1 , c2 are
constants.
Boundary condition f (0) = 0√ holds if c2 = 0.
Nonzero solution exists if Jν ( λ R) = 0.
Thus there are infinitely many eigenvalues λm,1 , λm,2 , . . . ,
p
where λm,n R = jν,n , i.e., λm,n = (jν,n /R)2 .
Associated eigenfunctions: fm,n (r ) = Jν (jν,n r /R).
Note that ν = mπ/L.
Eigenvalue problem:
∇2 φ + λφ = 0 in D = {(r , θ) : r < R, 0 < θ < L},
φ|∂D = 0.
Eigenvalues: λm,n = (j mπ L ,n
/R)2 , where
m = 1, 2, . . . , n = 1, 2, . . . , and j mπ
L ,n
is the nth
positive zero of the Bessel function J mπ L
.
Eigenfunctions:
φm,n (r , θ) = J mπ
L
(j mπ
L ,n
· r /R) sin mπθ
L .