0% found this document useful (0 votes)
81 views5 pages

Picard-Lindelof Theorem: Martin Nemer April 22, 2015

This document presents the Picard-Lindelof theorem, which states that an ordinary differential equation has a unique solution if the operator defined by the equation is a contraction mapping on a complete metric space. The proof involves several lemmas: 1) The differential equation is equivalent to an integral equation if the function is continuous. 2) The operator maps continuous functions to continuous functions. 3) There exists a restriction of the domain where the operator maps within the range. 4) If the function satisfies a Lipschitz condition, successive approximations converge. 5) The operator is a contraction, so Banach's fixed point theorem guarantees a unique solution on the restricted domain.

Uploaded by

Kamal Gour
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
81 views5 pages

Picard-Lindelof Theorem: Martin Nemer April 22, 2015

This document presents the Picard-Lindelof theorem, which states that an ordinary differential equation has a unique solution if the operator defined by the equation is a contraction mapping on a complete metric space. The proof involves several lemmas: 1) The differential equation is equivalent to an integral equation if the function is continuous. 2) The operator maps continuous functions to continuous functions. 3) There exists a restriction of the domain where the operator maps within the range. 4) If the function satisfies a Lipschitz condition, successive approximations converge. 5) The operator is a contraction, so Banach's fixed point theorem guarantees a unique solution on the restricted domain.

Uploaded by

Kamal Gour
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

Picard-Lindelof Theorem

Martin Nemer
April 22, 2015

Definition 1 An operator F : X → Y is a mapping that maps a function in


the space of functions X to another function in the space of functions Y ,

F [g] = h. (1)

Definition 2 Let F be an operator which maps a function in the space X to


another function in X. Then F is a strict contraction if there exists a k, k ∈ R,
with 0 < k < 1 s.t.
dX (F [g], F [h]) ≤ k dX (g, h), (2)
where dX is a suitable metric on X.

Lemma 1 Let yi ∈ X, i ∈ N, with yi+1 := F [yi ], for some known y0 ∈ X. If


F is a strict contraction map then dX (yi+1 , yi ) ≤ k i dX (y1 , y0 ).

Proof. By induction. Consider a sequence of functions {yi }∞


i=0 , defined by

yi+1 = F [yi ], (3)

where F is a strict contracting. The base case is i = 0, 1,

y1 = F [y0 ], y2 = F [y1 ], (4)

which with equation (2) implies,

dX (y2 , y1 ) ≤ k dX (y1 , y0 ). (5)

Next make the inductive hypothesis

dX (yi+1 , yi ) ≤ k i dX (y1 , y0 ). (6)

In the case of i + 2, using the inductive hypothesis and the fact that F obeys
equation 2, we have

dX (yi+2 , yi+1 ) ≤ k dX (yi+1 , yi ) ≤ k i+1 dX (y1 , y0 ). ∴ (7)

Lemma 2 The sequence {yi }∞ i=0 with yi+1 = F [yi ], y0 - given, F - a strict
contraction, is Cauchy in (X, dX ).

1
Proof. Let m, n ∈ N with m > n. Then it will be shown by induction that
m
X
dX (ym , yn ) ≤ dX (yi , yi−1 ), (8)
i=n+1

The base case m = n + 1 follows immediately from equation (8). Assume that
equation (8) holds for m = n+a, a ≥ 1, a ∈ N. Next for the case of m = n+a+1
using the triangle inequality and then inserting the inductive hypothesis,

dX (yn+a+1 , yn ) ≤ dX (yn+a , yn ) + dX (yn+a+1 , yn+a )


n+a
X
≤ dX (yi , yi−1 ) + dX (yn+a+1 , yn+a )
i=n+1
n+a+1
X
= dX (yi , yi−1 ). ∴ (9)
i=n+1

Inserting equation (6) into equation (8) yields.


m
X m
X
dX (ym , yn ) ≤ dX (yi , yi−1 ) ≤ dX (y1 , y0 ) k i−1
i=n+1 i=n+1
m−n−1
X ∞
X
≤ dX (y1 , y0 ) k n k i < dX (y1 , y0 ) k n ki
i=0 i=0
kn
< dX (y1 , y0 ) . (10)
1−k
Then ∀  > 0,  ∈ R, we can find an N, N ∈ N such that ∀m, n ≥ N ,
dX (ym , yn ) < , namely that N such that

kN
dX (y1 , y0 ) < , (11)
1−k
which is possible ∀ > 0 since limN →∞ k N = 0. Thus k m < k N for 0 < k < 1
and m > N .
Theorem 1 If an operator F : X → X is a strict contraction with respect to
the metric dX and (X, dx ) is a complete metric space of functions, then F has
a unique fixed point y = F [y] with y ∈ X.
Proof. Given that (X, dX ) is complete and {yi }∞
i=0 is Cauchy, it follows that
there exists a unique limit in (X, dX ),

y = lim yi (12)
i→∞

or equivalently
y = lim F [yi ]. (13)
i→∞

2
Definition 3 An ordinary differential equation is an equation of the form

y 0 (t) = f (t, y(t)), y(a) = c, (14)

where f : R2 → R and y(a) are known, and the solution to (14), y : R → R,


is a continuous-differentiable function for t ∈ [a, b], with [a, b] ⊆ R, in order for
equation (14) to be well posed.
Lemma 3 Let y be the solution to the differential equation (14). Then he
differential equation (14) is equivalent to the integral equation,
Z t
y(t) = y(a) + f (s, y(s)) ds. (15)
a

if f : S → R is continuous with S ⊆ R2 , S := [a, b] × [d, g], and (t, y(t)) ∈ S for


t ∈ [a, b].
Proof. Given that f : S → R is continuous, and (t, y) : S → R is continuous, it
follows that the composition f ◦ (t, y) : [a, b] → R is continuous. Thus from the
FFTC, it follows that equation (15) is a solution to (14).

Lemma 4 If f : S → R is continuous, then the operator F defined by


Z t
F [y] = y(a) + f (s, y(s)) ds, (16)
a

maps any continuous-differentiable function y : [a, b] → [d, g] to another continuous-


differentiable function F[y], i.e. F : C 1 ([a, b], [d, g]) → C 1 ([a, b], R).
Proof. Let y : [a, b] → [d, g] be a continuous function. Let h : [a, b] → S be
h(t) := (t, y(t)), which is also continuous. Since f : S → R is continuous, the
composition f ◦ h is also continuous. Since f ◦ h is continuous it is Riemann
integrable and by the FFTC and the fact that sums of continuous-differentiable
functions are continuous and differentiable, it follows that F [y] : [a, b] → R is
continuous and differentiable.
Lemma 5 For any continuous y : [a, b] → [d, g], we can find a restriction y|[a,b0 ]
with b0 ≤ b s.t. the F [y|[a,b0 ] ] : [a, b0 ] → [d, g].

Proof. Given that f : S → R is continuous, there exists an M ∈ R with


0 ≤ M < ∞ s.t. |f (t, x)| ≤ M ∀ (t, x) ∈ S. Then
Z t

|F [y](t) − y(a)| = f (s, y(s)) ds (17)
a
Z t
≤ |f (s, y(s))| ds ≤ M |t − a|
a
⇒ y(a) − M |t − a| ≤ F [y](t) ≤ y(a) + M |t − a|,
∀t ∈ [a, b]. (18)

3
Let b0 be chosen s.t. the following inclusion is satisfied,

[y(a) − M |b0 − a|, y(a) + M |b0 − a|] ⊆ [d, g], (19)

assuming y(a) ∈ int([d, g]). ∴


Lemma 6 Let y0 (t) = y(a) be known (i.e. a boundary condition) and yi+1 =
F [yi ], with F [y] given by equation (16). If f : S → R satisfies the Lipschitz
condition
|f (t, x2 ) − f (t, x1 )| ≤ k |x2 − x1 |, (20)
where k ∈ R, 0 ≤ k < ∞, then
M k i (t − a)i+1
|yi+1 (t) − yi (t)| ≤ . (21)
(i + 1)!
Proof. By Induction. Consider the base case i = 0, then from Lemma 5 above
we have
|y1 (t) − y0 (t)| ≤ M (t − a), for a < t. (22)
Next make the inductive hypothesis that equation (21) holds and consider the
case of i + 1,
Z t

|yi+2 (t) − yi+1 (t)| = (f (s, yi+1 (s)) − f (s, yi (t)) ds
a
Z t
≤ |f (s, yi+1 (s)) − f (s, yi (t))| ds
a
Z t
≤ k |yi+1 (s) − yi (t)| ds
a
t
M k i (t − a)i+1
Z
≤ k ds
a (i + 1)!
M k i+1 (t − a)i+2
= , (23)
(i + 2)!
where we have made use of the inductive hypothesis. ∴
Theorem 2 The mapping yi+1 = F [yi ], y0 (t) = y(a), with F [y] given by
equation (16), is a strict contraction on the restriction t ∈ [a, b00 ] for some
a < b00 ≤ b0 ≤ b on a complete metric space of functions and therefore has a
unique solution y = F [y].
Proof. From equation (23) for t ∈ [a, b00 ] with b00 chosen s.t. b00 ≤ b0 where b0
satisfies equation (19), and b00 satisfies

(b00 − a)k < 1, (24)

the mapping F [y] is a strict contraction under the uniform metric

dX (y, h) = sup |y(t) − h(t)|. (25)


t∈[a,b00 ]

4
Given that y0 is continuous and differentiable it follows from Lemma 4 that
yi is continuous and differentiable. By Theorem 14.4.5 of Tao (Analysis II,
2009), the space of continuous functions with the uniform metric (25) form a
complete metric space. Therefore by Theorem 1, the initial value problem given
in equation (14) has a unique solution on t ∈ [a, b00 ].
Other Info. Further question might be to prove that one can repeat the Picard
procedure and obtain solutions arbitrarily close to t = b, starting from y(a), with
the initial condition for each additional Picard solution equal to y(b00 ).

You might also like