34032_IntEqunsNeumannSeries
Transcript of 34032_IntEqunsNeumannSeries
MATH34032: Green’s Functions, Integral Equations and the Calculus of Variations 1
Section 4
Integral Equations: Neumann Series
As we have seen in Section 3, a Fredholm integral equation of the second kind may bewritten as
f (x) = λ
∫ b
a
K (x, y) f (y)dy + g (x) . (4.1)
We also saw, in section 3, that if the Kernel is separable, then we may take steps to solvethe integral equation analytically, by reducing it to a system of linear equations. However,what if the Kernel is not separable? In general these problems are much harder. Oneapproach exploits the fact that for sufficiently small values of λ it looks like we can neglectthe integral term to obtain the approximate solution
f (x) ≃ g (x) = f0 (x) . (4.2)
We can then improve on this approximation by substituting (4.2) into the right hand sideof (4.1), i.e.
f (x) ≃ f0 (x) + λ
∫ b
a
K (x, y) f0 (y) dy
= f0 (x) + λf1 (x) . (4.3)
Substituting (4.3) into the right hand side of (4.1) then yields an even better approximation
f (x) ≃ f0 (x) + λ
∫ b
a
K (x, y) [f0 (y) + λf1 (y)] dy
= f0 (x) + λ
∫ b
a
K (x, y) f0 (y)dy + λ2
∫ b
a
K (x, y) f1 (y) dy
= f0 (x) + λf1 (x) + λ2
∫ b
a
K (x, y) f1 (y) dy
= f0 (x) + λf1 (x) + λ2f2 (x) ,
say. This suggests letting
f (x) =∞
∑
n=0
λnfn (x) , (4.4)
which may be substituted into (4.1),
∞∑
n=0
λnfn (x) = λ
∫ b
a
K (x, y)
[
∞∑
n=0
λnfn (y)
]
dy + g (x)
=∞
∑
n=0
λn+1
∫ b
a
K (x, y) fn (y) dy + g (x) .
Equating coefficients of like powers in λ gives
λ0 : f0 (x) = g(x),
λn, n > 0 : fn (x) =∫ b
aK (x, y) fn−1 (y) dy.
(4.5)
MATH34032: Green’s Functions, Integral Equations and the Calculus of Variations 2
Definition 4.1 Formula (4.4), with (4.5), is called the Neumann series for the integralequation (4.1).
Example 1: Find the first 3 terms of the Neumann series for the integral equation
u(x) = 1 + ǫ
∫
1
0
(x − y)u(y) dy.
where ǫ ∈ R.
We note that λ = ǫ and K(x, y) = (x − y) so that
u(x) = u0(x) + ǫu1(x) + ǫ2u2(x) + ....
and it remains for us to determine each term. Clearly neglecting the integral term givesu0(x) = 1, and then
u1(x) =
∫
1/2
0
(x − y)u0(y)dy (4.6)
=
[
xy −y2
2
]1/2
0
(4.7)
=
(
x
2−
1
8
)
(4.8)
and then
u2(x) =
∫
1/2
0
(x − y)u1(y)dy (4.9)
=
∫
1/2
0
(x − y)
(
y
2−
1
8
)
dy (4.10)
=
[
xy2
4− x
y
8−
y3
6+
y2
16
]1/2
0
(4.11)
= −1
192(4.12)
so that the Neumann series takes the form
u(x) = 1 + ǫ
(
x
2−
1
8
)
−ǫ2
192+ .... (4.13)
But this approach is clearly only useful if the resulting series converges.....
MATH34032: Green’s Functions, Integral Equations and the Calculus of Variations 3
Iterated kernels and convergence
If the Neumann series is convergent it solves (4.1) uniquely. Now
fn (x) =
∫ b
a
K (x, y1) fn−1 (y1) dy1
=
∫ b
a
∫ b
a
K (x, y1) K (y1, y2) fn−2 (y2) dy2dy1,
replacing y by y1 in the first integral, and y2 in the second integral (for fn−1). We cancontinue replacing fm(x) by an integral containing fm−1(x) for each m down to m = 1,i.e.
fn(x) =
∫ b
a
∫ b
a
. . .
∫ b
a
K (x, y1)K (y1, y2) . . .K (yn−1, yn) f0 (yn) dyn . . . dy2dy1
=
∫ b
a
{∫ b
a
. . .
∫ b
a
K (x, y1) K (y1, y2) . . .K (yn−1, yn) dyn−1 . . . dy1
}
f0 (yn) dyn
=
∫ b
a
Kn (x, y) g (y) dy. (4.14)
Definition 4.2 The nth iterated kernel is
Kn (x, y) =
∫ b
a
. . .
∫ b
a
K (x, y1)K (y1, y2) . . .K (yn−1, y)dyn−1 . . . dy1.
Proposition 4.3 The nth iterated kernel satisfies the following inductive definition
K1 (x, y) = K (x, y) ,
Kn+1 (x, y) =∫ b
aK (x, z) Kn (z, y) dz.
Proof: Using Definition 4.2 with n replaced by n + 1 we have
Kn+1 (x, y) =
∫ b
a
. . .
∫ b
a
K (x, y1) K (y1, y2) . . .K (yn−1, yn) K (yn, y)dyn . . . dy1.
Now, setting y1 = z, say, and then relabelling each ym, m = 2, 3, . . . n, as ym−1 yields
Kn+1 (x, y) =
∫ b
a
K (x, z)
∫ b
a
. . .
∫ b
a
K (z, y1) . . .K (yn−2, yn−1) K (yn−1, y)dyn−1 . . . dy1dz,
or
Kn+1 (x, y) =
∫ b
a
K (x, z) Kn (z, y) dz
as required.From (4.4), (4.14) the Neumann series may be written
f (x) = f0 (x) +
∞∑
n=1
[
λn
∫ b
a
Kn (x, y) g (y)dy
]
MATH34032: Green’s Functions, Integral Equations and the Calculus of Variations 4
= f0 (x) + λ
∫ b
a
[
∞∑
n=1
λn−1Kn (x, y)
]
g (y) dy
= g (x) + λ
∫ b
a
R (λ, x, y) g (y)dy. (4.15)
Thus, comparing with our previous definition of the resolvent kernel (see Definition 3.11)we get the following proposition.
Proposition 4.4 The resolvent kernel R (λ, x, y) may be written
R (λ, x, y) =
∞∑
n=1
λn−1Kn (x, y) .
Theorem 4.5: If it exists, R (λ, x, z) is unique and solves (4.1) via (4.15).
No proof given.
Theorem 4.6: The Neumann series for the Fredholm equation
f (x) = λ
∫ b
a
K (x, y) f (y)dy + g (x) ,
where K (x, y) and g (x) are bounded and absolutely integrable, converges absolutely for allvalues of λ such that
|λ| <1
M (b − a),
whereM = sup
a≤x≤ba≤y≤b
|K (x, y)| .
No proof given.
Example 2: Determine for which values of ǫ the Neumann series converges in Example1.
Using Theorem 4.6, we see that b = 1, a = 0 and
M = supa≤x≤ba≤y≤b
|K (x, y)|
= supa≤x≤ba≤y≤b
|x − y|
= 1
so that the series converges for |ǫ| < 1.
MATH34032: Green’s Functions, Integral Equations and the Calculus of Variations 5
Theorem 4.7: The Neumann series for the Volterra equation
f (x) = λ
∫ x
a
K (x, y) f (y) dy + g (x) ,
where K (x, y) is bounded and g (x) is absolutely integrable, is absolutely convergent for allvalues of λ.
No proof given.
Theorem 4.8: The mth iterated kernel for a Volterra integral equation is given by
Km (x, y) =
{∫ x
yK (x, z) Km−1 (z, y) dz if y < x,
0 if y ≥ x.
Proof: We know from the definition of Volterra kernels that K1(x, y) = K(x, y) is zerofor y > x. So, start with m = 2:
K2 (x, y) =
∫ b
a
K (x, z) K (z, y) dz.
The first term in the integrand, K(x, z), is zero when z > x and the second, K(z, y), iszero when z < y. Hence
K2 (x, y) =
∫ x
y
K (x, z) K1 (z, y) dz.
The proof may now be completed by induction.
Example 3: Solve the Volterra integral equation
f (x) = λ
∫ x
0
f (y) dy + g (x)
and write down the resolvent kernel.
Solution: Here, K1 (x, y) = K (x, y) = 1.
K2 (x, y) =
∫ x
y
K (x, z) K1 (z, y) dz =
∫ x
y
dz = x − y,
K3 (x, y) =
∫ x
y
K (x, z) K2 (z, y) dz =
∫ x
y
1. (z − y)dz
=
[
1
2(z − y)2
]x
y
=1
2(x − y)2
,
K4 (x, y) =
∫ x
y
K (x, z) K3 (z, y) dz =
∫ x
y
1.1
2(z − y)2 dz
=
[
1
3!(z − y)3
]x
y
=1
3!(x − y)3 .
MATH34032: Green’s Functions, Integral Equations and the Calculus of Variations 6
Suppose that, in general,
Kn (x, y) =(x − y)n−1
(n − 1)!. (4.16)
Assume true for n = k, then for n = k + 1
Kk+1 =
∫ x
y
K (x, z) Kk (z, y) dz =
∫ x
y
(z − y)k−1
(k − 1)!dz
=
[
(z − y)k
k!
]x
y
=(x − y)k
k!.
Hence, (4.16) follows by induction. Therefore, from (4.15)
f (x) = g (x) +
∞∑
n=1
[
λn
∫ x
0
Kn (x, y) g (y) dy
]
= g (x) + λ
[
∫ x
0
∞∑
n=1
λn−1Kn (x, y) g (y)dy
]
= g (x) + λ
∫ x
0
[
∞∑
n=1
λn−1(x − y)n−1
(n − 1)!
]
g (y) dy
= g (x) + λ
∫ x
0
exp [λ (x − y)] g (y)dy.
Note, by comparison with (4.15), that the resolvent kernel is exp [λ (x − y)].The theory for degenerate kernels and for Neumann series may be used to prove thefollowing theorem, valid for an arbitrary kernel.
Theorem 4.9: The Fredholm Alternative for integral equations
Either the integral equation
f − λK̂f = g, f (x) − λ
∫ b
a
K (x, y) f (y)dy = g (x) , (4.17)
for fixed λ has precisely one solution f (x) for each arbitrary g (x); in particularf = 0 when g = 0.
The associated adjoint integral equation
f (x) − λ
∫ b
a
K (y, x) f (y) dy = g (x) ,
also has a unique solution. Note that the adjoint equation has kernel K(y, x) notK(x, y).
In this case λ is called a regular value.
MATH34032: Green’s Functions, Integral Equations and the Calculus of Variations 7
Or the associated homogeneous equation
f = λK̂f, f (x) = λ
∫ b
a
K (x, y) f (y)dy,
has m > 0, m ∈ N, linearly independent solutions f1, f2, . . . , fm.
The associated homogeneous adjoint integral equation
f (x) = λ
∫ b
a
K (y, x) f (y) dy,
also has m linearly independent solutions h1, h2, . . . , hm.
In this case equation (4.17) has a solution if and only if g (x) satisfies
〈g, hi〉 =
∫ b
a
g (y)hi (y)dy = 0
for i = 1, . . . , m, and the solution of (4.17) is then given by
f (x) = f0 (x) +
m∑
i=1
αifi (x)
where the αi are arbitrary constants.
In this case λ is called a characteristic value or eigenvalue.