Solu 1
Solu 1
homework assignment #5
Problem 1. Consider the non-Sturm-Liouville differential equation
d2 φ dφ
2
+ α(x) + (λβ(x) + γ(x))φ = 0.
dx dx
Multiply this equation by H(x). Determine H(x) such that the equation may be reduced to
the standard Sturm-Liouville form:
d dφ
p(x) + (λσ(x) + q(x))φ = 0.
dx dx
Given α(x), β(x), and γ(x), what are p(x), σ(x), and q(x)?
Solution: p(x) = eA(x) , σ(x) = eA(x) β(x), and q(x) = eA(x) γ(x), where A is an anti-derivative
of α.
It follows that p = H, p′ = Hα, σ = Hβ, and q = Hγ. We expect H to be positive. Then the first
two relations imply that
p′
Z Z
=α =⇒ log p = α(x) dx =⇒ p = exp α(x) dx .
p
Consequently, σ = pβ and q = pγ.
Problem 2. Consider
∂u ∂ ∂u
cρ = K0 + αu,
∂t ∂x ∂x
where c, ρ, K0 , α are functions of x, subject to
Solution: X∞
u(x, t) = cn e−λn φn (x),
n=1
where λ1 < λ2 < . . . are eigenvalues of the Sturm-Liouville eigenvalue problem
1
φ1 , φ2 , . . . are the corresponding eigenfunctions, and
Z L
f (x)φn (x)c(x)ρ(x) dx
cn = Z0 L .
2
|φn (x)| c(x)ρ(x) dx
0
Detailed solution: We search for the solution of the initial-boundary value problem as a
superposition of solutions u(x, t) = φ(x)G(t) with separated variables of the differential equation that
satisfy the boundary conditions.
Substituting u(x, t) = φ(x)G(t) into the differential equation, we obtain
Conversely, if functions G and φ are solutions of the above ODEs for the same value of λ, then
u(x, t) = φ(x)G(t) is a solution of the given PDE.
Substituting u(x, t) = φ(x)G(t) into the boundary conditions, we get
φ(0)G(t) = 0, φ(L)G(t) = 0.
It is no loss to assume that G is not identically zero. Then the boundary conditions are satisfied if
and only if φ(0) = φ(L) = 0.
To determine φ, we have a regular Sturm-Liouville eigenvalue problem
This problem has infinitely many eigenvalues λ1 < λ2 < . . . . Let φn denote an eigenfunction corre-
sponding to the eigenvalue λn . Then φn is determined up to multiplying by a scalar. λn and φn are
related through the Rayleigh quotient
L
Z L
−K0 φn φ′n + (K0 |φ′n |2 − α|φn |2 ) dx
0
λn = Z L0 .
2
cρ|φn | dx
0
Since φn (0) = φn (L) = 0, the non-integral term vanishes. If α < 0 then it follows that λn > 0.
The function G is to be determined from the equation G ′ = −λG. The general solution of this
equation is G(t) = c0 e−λt , where c0 is a constant.
Thus we obtain the following solutions of the heat equation satisfying the boundary conditions:
2
A superposition of these solutions is a series
X∞
u(x, t) = cn e−λn φn (x),
n=1
where c1 , c2 , . . . are constants. Substituting the series into the initial condition u(x, 0) = f (x), we get
X∞
f (x) = cn φn (x).
n=1
Hence the initial condition is satisfied if cn are coefficients of the generalized Fourier series of f , that
is, if Z L
f (x)φn (x)c(x)ρ(x) dx
0
cn = Z L .
2
|φn (x)| c(x)ρ(x) dx
0
If all eigenvalues are positive then the solution u(x, t) vanishes as t → +∞ because each term of
the corresponding series contains a decaying factor e−λn .
for any two functions u and v satisfying the boundary conditions. Show that the following yield
self-adjoint problems:
(i) φ′ (0) = 0 and φ(L) = 0;
(ii) φ′ (0) − hφ(0) = 0 and φ′ (L) = 0.
Solution: (i) Suppose that functions u and v satisfy the boundary conditions u′ (0) = v ′ (0) = 0
and u(L) = v(L) = 0. The first condition implies that u(0)v ′ (0) − v(0)u′ (0) = 0 while the second one
implies that u(L)v ′ (L) − v(L)u′ (L) = 0. It follows that
L
p (uv ′ − vu′ ) = 0.
0
(ii) Suppose that functions u and v satisfy the boundary conditions u′ (0)−hu(0) = v ′ (0)−hv(0) = 0
and u′ (L) = v ′ (L) = 0. The first condition implies that u(0)v ′ (0) − v(0)u′ (0) = u(0) · hv(0) − v(0) ·
hu(0) = 0 while the second one implies that u(L)v ′ (L) − v(L)u′ (L) = 0. It follows that
L
p (uv ′ − vu′ ) = 0.
0
3
√
Solution: λn ∼ (n − 1)π as n → ∞.
Detailed solution: (i) The Rayleigh quotient relates an eigenfunction φ to the corresponding
eigenvalue λ:
1
Z 1
−φφ ′
+ |φ′ (x)|2 dx
0 0
λ= Z 1 .
|φ(x)|2 dx
0
The boundary conditions imply that
1
−φφ′ = −φ(1)φ′ (1) + φ(0)φ′ (0) = |φ(1)|2 + |φ(0)|2 ≥ 0.
0
It follows that λ ≥ 0. Furthermore, if λ = 0 then φ(0) = φ(1) = 0 and φ′ is identically zero, which is
not possible since then φ is also identically zero. Thus λ > 0.
any λ > 0 the general solution of the equation φ′′ + λφ = 0 is φ(x) = c1 cos µx + c2 sin µx,
(ii) For √
where µ = λ and c1 , c2 are constants. Substituting this into the boundary conditions φ(0) = φ′ (0)
and φ(1) = −φ′ (1), we obtain
It follows that
c2 (µ cos µ + sin µ) = c2 (µ2 sin µ − µ cos µ),
c2 ((µ2 − 1) sin µ − 2µ cos µ) = 0.
A nonzero solution of the boundary value problem exists if and only if (µ2 − 1) sin µ = 2µ cos µ. Since
cos 1 6= 0, the latter equality implies that µ2 6= 1 and cos µ 6= 0. Hence it is equivalent to
2µ
tan µ = .
µ2 − 1
√
It remains to recall that µ = λ.
tan µ
2µ
µ2 −1
π 3π µ
1 2 2
4
√
Let 0√< λ1 < λ2 < . . . be the eigenvalues.
√ Graphically, we establish that 1 < λ1 < π/2,
π/2
√ < λ 2 < 3π/2, and (2n − 3)π/2 < λ n < (2n − 1)π/2 for n = 3, 4, . . . . Moreover,
λn ≈ (n − 1)π for a large n.
Solution: β = 2.
Detailed solution: In the case λ = 0, the general solution of the equation φ′′ + λφ = 0
is a linear function φ(x) = c1 + c2 x, where c1 , c2 are constants. Substituting it into the boundary
conditions φ(0) = φ′ (0) and φ(1) = βφ′ (1), we obtain equalities c1 = c2 , c1 + c2 = βc2 . They imply
that 2c1 = 2c2 = βc2 . If β 6= 2, it follows that c1 = c2 = 0, hence there are no eigenfunctions with
eigenvalue λ = 0. If β = 2 then φ(x) = 1 + x is indeed an eigenfunction.