PHY 501, John Shumway Lec. 28: Sturm-Liouville Theory Nov.
1, 2002
Lecture 28: Sturm-Liouville Theory
1 Hermetian Operators
In matrix theory, Hermetian operators are matricies for which A = A
. These matricies have the
property that
ij
u
i
M
ij
v
j
=
ij
v
i
M
ij
u
j
. (1)
This equation follows immediately from the identity v
Au = [(v
Au)
and the properties of the
adjoint operator, .
Hermetian operators are nice for eigenvalue problems because they have the following properties,
All eigenvalues are real.
Eigenvectors with distinct (non-degenerate) eigenvalues are orthogonal.
Eigenvectors with degenerate eigenvalues can be made orthogonal.
The eigenvectors span the vector space (for the case of dierential operators, this means the
eigenfunctions form a complete set).
These properties can all be derived from the dention of a Hermetian operators, Eq. (??).
Note that with matrices, the terms Hermetian and self-adjoint both imply A = A
. When we deal
with dierential operators, we will have to distinguish the two properties. Specically, self-adjoint
operators satisfy
ij
u
i
M
ij
u
j
=
ij
u
i
M
ij
u
j
, (2)
while Hermetian operators satisfy
ij
u
i
M
ij
v
j
=
ij
v
i
M
ij
u
j
. (3)
Thus being Hermetian implies self-adjointness and a bit more. For dierential operators, we will
see that Hermetian operators need to be self-adjoint and have the right boundary conditions.
1
PHY 501, John Shumway Lec. 28: Sturm-Liouville Theory Nov. 1, 2002
2 Hermitian Diererential Operators (Sturm-Liouville Operators)
A dierential operator is Hermitian if it satises
_
b
a
u
(x)Lv(x)dx =
_
b
a
v(x)Lu
(x)dx. (4)
A necessary condition for being Hermetian is that the operator L be self-adjoint and functions u(x)
and v(x) have appropriate boundary conditions on the interval a < x < b.
2.1 Self-Adjoint Dierential Operators
We dene
u|Lui =
_
b
a
u
(x)
_
Lu(x)
_
(x)dx. (5)
The adjoint L
is then
L| =
_
b
a
_
L
(x)
_
u(x)
_
(x)dx. (6)
We want to see what self-adjointness means for a general second-order linear dierential operator,
L = p
0
(x)
d
2
dx
2
+ p
1
(x)
d
dx
+ p
2
(x). Using integration by parts, we nd,
L =
d
2
dx
2
p
0
(x)
d
dx
p
1
(x) + p
2
(x).
= p
0
(x)
d
2
dx
2
+ [2p
0
(x) p
1
(x)]
d
dx
+ [p
0
(x) p
1
(x) + p
2
(x)]
(7)
(there are boundary terms we omitted that disappear for self-adjoint operators). A necessary and
sucient condition for self-adjointness, Eq. (7), is for p
0
(x) = p
1
(x). In this case
L
y = Ly =
d
dx
_
p
0
(x)
dy(x)
dx
_
+ p
2
(x)y(x). (8)
All second-order dierential operators are either self-adjoint or can be made self-adjoint by scaling
by a function of x. If an operator L = p
0
(x)
d
2
dx
2
+ p
1
(x)
d
dx
+ p
2
(x) is not self-adjoint, it may be
made self-adjoint by multiplying by
1
p
0
(x)
exp
__
x
p
1
(t)
p
0
(t)
dt
_
. (9)
This is easily seen by
1
p
0
(x)
exp
__
x
p
1
(t)
p
0
(t)
dt
_
Ly =
1
p
0
e
p
1
p
0
dt
_
p
0
d
2
y
dx
2
+ p
1
dy
dx
+ p
2
y
_
= y
p
1
p
0
dt
+ y
p
1
p
0
e
p
1
p
0
dt
+
p
2
y
p
0
e
p
1
p
0
dt
=
d
dx
_
y
p
1
p
0
dt
_
+
p
2
y
p
0
e
p
1
p
0
dt
(10)
which is clearly self-adjoint by comparison to Eq. (8).
2
PHY 501, John Shumway Lec. 28: Sturm-Liouville Theory Nov. 1, 2002
2.2 Boundary conditions on Hermetian dierential operators
To get from self-adjoint operator to Hermitian operator, we need a restriction on the boundary
conditions, namely
_
y
i
py
x=b
x=a
= 0. (11)
This restriction is met by many common boundary conditions, such as y(a) = y(b) = 0 or y
(a) =
y
(b) = 0, or even p(a) = p(b) = 0.
2.3 Proof that Self-Adjointness and Proper Boundary Conditions Yeilds a Her-
mitian Operator
This proof relies on three steps: (1) write the operator in self-adjoint form, Eq. (8); (2) integrate
by parts twice; and (3) use the restriction on the proper boundary conditions to get rid boundary
terms from the integration-by-parts.
_
b
a
u Lvdx =
_
b
a
u
(pv
+ u
qvdx
=
_
b
a
u
(pv
dx +
_
b
a
vqu
dx
=
_
u
pv
b
a
_
b
a
vpu
dx +
_
b
a
vqu
dx
=
_
u
pv
b
a
_
u
pv
b
a
+
_
b
a
v(pu
dx +
_
b
a
vqu
dx
=
_
b
a
v(pu
dx +
_
b
a
vqu
dx
=
_
b
a
vLu
dx.
(12)
3 Sturm-Liouville Equations
A Sturm-Liouville equation is a second order ODE eigenvalue equation that is Hermetian. A
Sturm-Liouville equation takes the form
p(x)y
(x) + r(x)y
(x) + q(x)y(x) = (x)y(x), with r(x) = p
(x). (13)
From the last lecture we recall that the condition r(x) = p
(x) is just the requirement that L =
p
d
2
dx
2
+ r
d
dx
+ q be self-adjoint. We also need appropriate boundary [Link]
_
y
i
py
x=b
x=1
= 0. (14)
3
PHY 501, John Shumway Lec. 28: Sturm-Liouville Theory Nov. 1, 2002
This restriction is met by many common boundary conditions, such as y(a) = y(b) = 0 or y
(a) =
y
(b) = 0, or even p(a) = p(b) = 0.
Alternatively, we can write the Sturm-Liouville equation as
(py
+ qy = y. (15)
4 Examples of Common ODE in Sturm-Liouville Form
4.1 Simple Harmonic oscillator
The equation of motion for a simple harmonic oscillator with frequency =
_
k/m is
y
=
2
y, (16)
which is Eq. (15) with p(x) = 1, q(x) = 0, =
2
and = 1.
4.2 Legendre Equation
Legendres equation occurs when the Laplacian is evaluated in spherical coordinates (leading the
the spherical harmonics). The equation is
(1 x
2
)y
2xy
+ l(l + 1)y = 0 (17)
or
((1 x
2
)y
+ l(l + 1)y = 0 (18)
which is Eq. (15) with p(x) = (1 x
2
), q(x) = 0, = l(l + 1), and = 1.
4.3 Associated Legendre Equation
For spherical harmonics with m = 0 we have the associated Legendre Equation,
((1 x
2
)y
+
_
l(l + 1)
m
2
1 x
_
y = 0 (19)
which is Eq. (15) with p(x) = (1 x
2
), q(x) =
m
2
1x
, = l(l + 1), and = 1.
4
PHY 501, John Shumway Lec. 28: Sturm-Liouville Theory Nov. 1, 2002
4.4 Laguerre
Laguerres equation occurs in the radial part of the Schrodinger equation for the orbitals of the
hydrogen atom. The equation is
xy
+ (1 x)y
+ ny = 0. (20)
In this case we need an integrating factor, namely e
x
. Then
e
x
xy
e
x
(1 x)y
+ ne
x
y = (xe
x
y
+ ne
x
y (21)
which is Eq. (15) with p(x) = xe
x
, q(x) = 0, = n, and = e
x
2
.
4.5 Hermite
The Hermite equation occurs in the Schrodinger equation for the quantum mechanical solution of
a simple harmonic oscillator. The equation is
y
2xy
+ 2y = 0. (22)
In this case we need an integrating factor, namely e
x
2
. Then
e
x
2
y
e
x
2
2xy
+ 2e
x
2
y = (e
x
2
y
+ 2e
x
2
y (23)
which is Eq. (15) with p(x) = e
x
2
, q(x) = 0, = , and = e
x
2
.
4.6 Bessels equation
Bessels equation occurs when the Laplacian is evaluated in cylindrical coordinates, for the radial
dependence. The equation is
x
2
y
+ xy
+ (x
x
n
2
)y = 0. (24)
Performing a change of variable = x/a, we have
(y
n
2
y + a
2
y = 0, (25)
which is Eq. (15) with p() = , q() = , = n
2
, and = 1/x.
5
PHY 501, John Shumway Lec. 28: Sturm-Liouville Theory Nov. 1, 2002
5 Greens functions in terms of eigenvectors
Consider a matrix equation,
Ly = b. (26)
If we invert the matrix L, then the solution is trivial,
y = L
1
b. (27)
The matrix inversion is simple if we know the eigenvectors of L, in which case
L
ij
=
k
(v
i
)
k
k
(v
k
)
j
, (28)
where v
i
are eigenvectors and
i
are eigenvalues. The inverse is then
L
1
ij
=
k
(v
i
)
k
1
k
(v
k
)
j
. (29)
(If this isnt clear, recall that the orthogonal matrix O constructed from the eigenvectors rotates L
into a diagonal form, O
1
LO = , which has the eigenvalues
i
along the diagonal. Multiplying on
the left and right by O and O
1
gives L = OO
1
. Also, to take the inverse of a diagonal matrix,
simply take the reciprical of the diagonal elements.) Therefore, the solution is
y
i
=
k
(v
i
)
k
1
k
(v
k
)
j
b
j
. (30)
Now consider a second order linear ODE with a Hermetian operator (Sturm-Liouville equation).
In analogy with Eq. (30) the solution is
y(x) =
_
b
a
dx
n=0
y
n
(x)
1
n
y
n
(x
)f(x
)
=
_
b
a
dx
G(x, x
)f(x
),
(31)
where y
n
are properly normalized eigenfunctions and
G(x, x
) =
n=0
1
n
y
n
(x)y
n
(z), (32)
is the Green function for the operator . A Green function is to a linear dierential equation what
an inverse matrix is to set of linear equations the Green function inverts the eect of the linear
dierential operator.
6
