UPPSALA UNIVERSITY
DEPARTMENT OF MATHEMATICS Applied Mathematics
Jörgen Östensson Fall 2012
Integral equations problems
1. Prove the Leibniz formula
Z b(x) Z b(x)
d
F (x, y) dy = Fx (x, y) dy + F (x, b(x))b0 (x) − F (x, a(x))a0 (x).
dx a(x) a(x)
2. Solve the Volterra equation Z x
u(x) = 1 + λ u(y) dy
0
(a) by computing the Neumann series for u.
(b) using the Laplace transform.
(c) by first converting it into an equivalent initial value problem.
3. Reformulate the initial value problem
( 00
u (x) − λu(x) = f (x), x > 0,
u(0) = 1, u0 (0) = 0,
as an equivalent Volterra integral equation. (Use the Leibniz formula to verify that the solution
of the Volterra equation indeed satisfies the initial value problem.)
4. Solve the Fredholm equation Z 1
u(x) − λ u(y) dy = 1
0
(a) using a Neumann series.
(b) by a direct approach.
5. Solve the Fredholm integral equation
Z 1
u(x) = λ (1 − 3xy)u(y) dy
0
for all values of λ.
6. Discuss the solvability of the Fredholm integral equation
Z 1
u(x) − λ (1 − 3xy)u(y) dy = f (x).
0
1
Answers or hints:
Rβ d
1. Let G(x, α, β) = α
F (x, y) dy. Compute dx G(x, a(x), b(x)) using the chain rule.
λn xn
P∞
2. u(x) = eλx = n=0 n! .
Rx Rx
3. u(x) − λ 0
(x − y)u(y) dy = 1 + 0
(x − y)f (y) dy.
1
P∞
4. The integral equation is solvable iff λ 6= 1; then u(x) = 1−λ . u(x) = n=0 λn if |λ| < 1.
5. We have the following cases:
• If λ 6= ±2, then u(x) ≡ 0.
• If λ = 2, then u(x) = c(1 − x) where c is an arbitrary constant.
• If λ = −2, then u(x) = c(1 − 3x) where c is an arbitrary constant.
(See 6. below for further explanation.)
6. We have the following cases:
• If λ 6= ±2, then the integral equation is uniquely solvable for any f .
R1 R1
• If λ = 2, then the integral is solvable if and only if 0 f (x) dx = 0 xf (x) dx. If this condition
is satisfied, then the integral equation has infinitely many solutions.
R1 R1
• If λ = −2, then the integral is solvable if and only if 0 f (x) dx = 3 0 xf (x) dx. If this
condition is satisfied, then the integral equation has infinitely many solutions.
Below follows a derivation of the above statements.
P2
Note that the kernel is degenerate. Indeed, k(x, y) = 1 − 3xy = j=1 αj (x)βj (y), where say
α1 (x) = 1, β1 (y) = 1, α2 (x) = x and β2 (y) = −3y.
The algebraic system associated with the Fredholm equation is
(I − λA)c = f , (1)
where
R1 R1 ! !
0
β1 (x)α1 (x) dx 0
β1 (x)α2 (x) dx 1 1/2
A= R1 R1 = ,
β2 (x)α1 (x) dx β2 (x)α2 (x) dx −3/2 −1
0 0
R1 ! R1 !
0
f (x)β1 (x) dx 0
f (x) dx
f= R1 = R1 .
0
f (x)β2 (x) dx −3 0
xf (x) dx
If this system is solvable, then solutions of the given integral equation are given by
u(x) = f (x) + λ(c1 α1 (x) + c2 α2 (x)) = f (x) + λ(c1 + c2 x). (2)
Note that
1−λ −λ/2 λ2
det(I − λA) = =1− .
3λ/2 1+λ 4
So the system (1), hence the integral equation, is uniquely solvable if λ 6= ±2.
2
If λ = 2 the system (1) becomes
! R1 !
−1 −1 0
f (x) dx
c= R1 .
3 3 −3 xf (x) dx
0
R1 R1
This is solvable iff 0 f (x) dx = 0 xf (x) dx, and the integral equation then has infinitely many
solutions.
If λ = −2 the system (1) becomes
! R1 !
3 1 0
f (x) dx
c= R1 .
−3 −1 −3 xf (x) dx
0
R1 R1
This is solvable iff 0 f (x) dx = 3 0 xf (x) dx, and the integral equation then has infinitely many
solutions.
Of course, in case solutions do exist, they can be explicitly computed by solving the algebraic
system (1) and using (2). For example, consider the homogenous case f = 0 of Problem 5. Then
f = 0. If λ 6= ±2 the homogeneous system (1) has only the trivial solution, and (2) gives u(x) ≡ 0.
If λ = 2 putting c1 = a gives c2 = −a. If we write c = 2a, then equation (2) gives u(x) = c(1 − x).
If λ = −2 putting c1 = a gives c2 = −3a. Writing c = −2a, equation (2) gives u(x) = c(1 − 3x).