34
MATH 3700 Ordinary Dierential Equation Notes 1 Ex is tence a nd Uniq ueness We are interested in the following dierential equation y = f (x, y), y (x 0 ) = y 0 . You have already met a theorem about existence and uniqueness but we will go into this in more detail here. 1.1 Exampl es We will begin with some examples. These illustrate various aspects of the existence and uniquen ess theorem. 1.  y = y 2 2.  y = y 2/3 1

Lectures Weeks 1-4

Embed Size (px)

Citation preview

Page 1: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 1/34

Page 2: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 2/34

Page 3: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 3/34

Page 4: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 4/34

1.4 Picard’s Theorem

Let  f (x, y) and  ∂f/∂y  be continuous functions of  x and  y  on a closed rectangle  R  withsides parallel to the axes. If (x0, y0) is any interior point of  R  then there exists a numberh > 0 with the property that the initial value problem

y′ = f (x, y), y(x0) = y0

has one and only one solution  y  =  y(x) on the interval |x − x0| ≤ h.Proof . We write  yn  as

yn(x) = y0(x) +n

m=1

[ym − ym−1]

We therefore define

y(x) = y0(x) +∞

m=1

[ym − ym−1]

and we need to show that

1.   y(x) exists

2.   y(x) is a solution of the DE

3.   y(x) is the only solution of the DE

We begin by proving (1) that  y(x) exists, i.e. we need to show that the series aboveconverges.

We know that|f (x, y)| ≤ M 

|  ∂ 

∂yf (x, y)| ≤ K 

on the interval  R. The mean value theorem tell us that

|f (x, y1) − f (x, y2)| ≤ ∂ 

∂y f (x, y⋆) |y1 − y2|

for some number  y⋆ between y1  and  y2. Therefore

|f (x, y1)− f (x, y2)| ≤ K |y1 − y2|

We define h < 1/K  and define R′ = {(x, y), |x−x0| ≤ h, |y−y0| ≤ Mh}. We requirethat  h  is sufficiently small that  R ′ lies inside  R.

We can show that all of the functions  yn  are in the interval  R′.Define  a = max |y1(x) − y0(x)|. Then

|f (t, y1)− f (t, y0)| ≤ K |y1(x) − y0(x)| ≤ Ka

4

Page 5: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 5/34

Page 6: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 6/34

1.5 Lipschitz condition

If we look at the proof for the existence and uniqueness solution we see that we actuallyonly required that

|f (x, y1)− f (x, y2)| ≤ K |y1 − y2|This is a weaker condition than requiring that the function   f (x, y) has a continuouspartial derivative with respect to  y  and the existence and uniqueness theorem can im-mediately be generalized.

Definition A function is Lipschitz continuous in an interval  I   if 

|f (y1)− f (y0)| ≤ C |y1 − y0|

for all  y1, y0

 ∈I   for some  C > 0.

What does it mean to be Lipschitz continuous? It is a stronger condition thancontinuity but it does not imply that the derivative is bounded. Consider the examples

1.   f (y) = |y|

2.   f (y) = y2/3

6

Page 7: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 7/34

We can extend the theorem to the case where   f   is only Lipshitz continuous withrespect to y. We will prove a slightly different version of the theorem

We have the following theorem: If   f (y) is differentiable in   I   then   f   is Lipschitzcontinuous with constant  C  if and only if 

supy∈I 

∂f 

∂y(y)

≤ C.

Proof:

7

Page 8: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 8/34

1.6 Existence and Uniqueness for the Case of Lipshitz Conti-

nuity

The existence and uniqueness proof can be generalized in many ways. Let us considerthe following theorem.

Theorem Let  f (x, y) be a continuous function that satisfies a Lipshitz condition

|f (x, y1)− f (x, y2)| ≤ K |y1 − y2|

on a strip defined by a ≤ x ≤ b and −∞ < y < ∞ (note that this is not true for  y ′ = y2).Then, if (x0, y0) is any point in the strip, then the initial value problem

y′ = f (x, y), y(x0) = y0.

has a unique solution on the interval  a ≤ x ≤ b.Proof  We define the constants

M 0 = |y0|, M 1 = max |y1(x)|, M  = M 0 + M 1

It follows that |y0(x)| ≤ M   and |y1(x) − y0(x)| ≤ M . If  x0 ≤ x ≤ b  it follows that

|y2(x)− y1(x)|   =

   xx0

f (t, y1)− f (t, y0) dt

≤    x

x0 |f (t, y1)− f (t, y0)|  dt

≤   K 

   xx0

|y1(x) − y0(x)| dt

≤   KM (x− x0)

Likewise

8

Page 9: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 9/34

Page 10: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 10/34

Therefore we obtain the following result. The difference between the true solution  yand the Picard Iterate  yn   is

|y(x)− yn(x)| ≤ M (K (x− x0))n

n!  eK (x−x0)

Note this is an upper bound and the actual iterates may be much more accurate. Thereare also bounds on the Picard iterates which can be found for the first case we considered.

Example Consider the initial value problem

y′ = x2y;   y(0) = 1.

one the interval 0 ≤  x ≤  1. Find the Picard iterates y1,  y2  etc. Find an upper boundon the error for these iterates.

10

Page 11: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 11/34

2 Autonomous Systems

A system of DE’s is autonomous if the right hand side does not depend on the variableof differentiation. In two dimensions they can be written as

dx

dt  = f (x, y)

dy

dt  = g(x, y)

Example

dx

dt

  = 0.5x

−0.4xy

dy

dt  =   −y + 0.2xy

2.1 Direction fields

At each point in the (x, y) plane we can determine the slope of the solution. We canthen plot these slopes to give a direction field.

So to sketch solution curves to DE,

1. plot direction field, then

2. starting at some initial point, sketch a smooth curve that follows vectors in direc-tion field.

11

Page 12: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 12/34

Example:  Sketch some representative solutions for the system

dx

dt   =   y

dy

dt  = sin(x)

The direction field is given below.

12

Page 13: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 13/34

2.2 Equilibrium solutions

The point (x0, y0) is an  equilibrium point   for the system

dx

dt  = f (x, y)

dy

dt  = g(x, y)

if  f (x0, y0) = g(x0, y0) = 0.Example 1

dx

dt  = 2x + y

dy

dt  = 2y + x

Example 2

dx

dt  =   x + y

dy

dt  =   y (2 − x)

Behaviour of solutions near equilibriums can be observed with   pplane . Note that

1. direction of vectors in direction field changes dramatically near an equilibriumpoint, and

2. solutions passing near an equilibrium go very slowly (because all components of 

vector field → 0 near an equilibrium).

13

Page 14: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 14/34

Page 15: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 15/34

Example Determine the behaviour of solutions to the system

dx

dt   = −4   −2−1   −3

x.

 dx/dt = − 4 x − 2 y dy/dt = − x − 3 y

−2 −1.5 −1 −0.5 0 0.5 1 1.5 2

−2

−1.5

−1

−0.5

0

0.5

1

1.5

2

x

      y

The equilibrium point at the origin in this type of system is called a   sink. If theflow is reversed then the equilibrium is a  source.

15

Page 16: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 16/34

Page 17: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 17/34

Linear systems with repeated eigenvalues

Example Consider the system

dx

dt  =

  2 00 2

x

 dx/dt = 2 x dy/dt = 2 y

 

−2 −1.5 −1 −0.5 0 0.5 1 1.5 2

−2

−1.5

−1

−0.5

0

0.5

1

1.5

2

x

      y

17

Page 18: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 18/34

Example Consider the system

dx

dt   = −5 08   −5

x.

 dx/dt = − 5 xdy/dt = 8 x − 5 y

 

−2 −1.5 −1 −0.5 0 0.5 1 1.5 2

−2

−1.5

−1

−0.5

0

0.5

1

1.5

2

x

      y

Note that system has only one straight line solution.

18

Page 19: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 19/34

Example  : Consider the one-parameter family of linear systems

dx

dt   =

  1 2a   0

x

where  a  is a parameter. Determine the type of equilibrium at the origin for all values of a. Sketch the phase portrait for representative values of  a.

Eigenvalues of matrix

A =

  1 2a   0

are

12 ± 1

2√ 1 + 8a

so the type of equilibrium at the origin depends on a. Also, det(A) = −2a and trace(A) =1.

19

Page 20: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 20/34

Page 21: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 21/34

Example 1 Analyze the equilibrium points for

dx

dt   = 2− x− y

dy

dt  =   x2 − y

 dx/dt = 2 − x − y

 dy/dt = x2 − y

−4 −3 −2 −1 0 1 2 3 4

−2

−1

0

1

2

3

4

5

6

x

      y

21

Page 22: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 22/34

Unfortunately, linearization does not always work. In particular, if the Jacobianmatrix has a zero eigenvalue or a purely imaginary eigenvalue, then we cannot predict

the behaviour in the nonlinear system based on linearization alone.Example

dx

dt  =   −x3

dy

dt  =   −y + y2

 dx/dt = − x3 

dy/dt = − y + y2

 

−2 −1.5 −1 −0.5 0 0.5 1 1.5 2

−1

−0.5

0

0.5

1

1.5

2

x

      y

Notice that in this phase portrait, (0, 0) looks like a sink and (0, 1) looks like a saddle.

These results were not predicted by the corresponding linearized systems. Linearizationdoes not work in these cases because of the zero eigenvalues of the Jacobian.

22

Page 23: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 23/34

Page 24: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 24/34

3 Hamiltonian and Lyapunov functions

3.1 PendulumThe equation for a pendulum which is free to rotate is given by

d2θ

dt2  = −g

l sin θ

We write this (assuming  g/l  = 1) as

dx

dt  =   y

dy

dt   =   − sin(x)

24

Page 25: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 25/34

3.2 Hamiltonian Systems

A system is Hamiltonian if it can be written in the form

dx

dt  =

 ∂H 

∂y

dy

dt  = −∂H 

∂x

If we consider the equations for the pendulum we can see that H  = − cos(x) + y2/2 givesthe desired equations.

One important consequence of this is that the quantity H  must be conserved for oursystem. This has important consequences.

25

Page 26: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 26/34

3.3 KdV equation

The KdV (Korteweg de Vries) equations is one of the most important nonlinear partialdifferential equations. When looking for travelling wave solutions the problem is reducedto the following (assuming again all constants are one)

dx

dt  =   y

dy

dt  =   x − x2

What is the Hamiltonian for this system.

26

Page 27: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 27/34

Page 28: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 28/34

3.5 Damped pendulum

We can modify the pendulum problem by including damping. This leads to

dx

dt  =   y

dy

dt  =   − sin(x) − y/10

28

Page 29: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 29/34

3.6 Lyapunov function

We assume that we have a dynamical system

dx

dt  =   f (x, y)

dy

dt  =   g(x, y)

We assume we have an equilibrium point at (0 , 0) (we can always transform any otherequilibrium point to the origin). A Lyapunov function is a function with the propertiesthat

1.   V  (x, y)

≥0

2.   V  (0, 0) = 0

3.   dV dt ≤ 0 along solutions of the DE, in some neighbourhood of the equilibrium point.

If such a function can be found then the equilibrium point must be stable.If the function  V   is strictly decreasing then all solutions in the neighbourhood tend

to the equilibrium point.Note that in general it is not easy to find a Lyapunov function.

29

Page 30: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 30/34

Consider the equation for a spring with linear damping.

md2y

dt2   + cdy

dt   + ky  = 0

30

Page 31: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 31/34

Consider the system

dx

dt   =   −2xy

dy

dt  =   x2 − y3

Try a function of the formV  (x, y) = ax2m + by2n

31

Page 32: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 32/34

Page 33: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 33/34

4.2 Pitchfork Bifurcation

In this bifurcation a source (or sink) is transformed into three equilibriums, two sources(sinks) and one saddle.

dx

dt  =   αx− x3

dy

dt  =   −y

33

Page 34: Lectures Weeks 1-4

8/12/2019 Lectures Weeks 1-4

http://slidepdf.com/reader/full/lectures-weeks-1-4 34/34

4.3 Hopf Bifurcation

In this bifurcation the real part of complex eigenvalues for the linearized system changefrom negative to positive. In this case, under fairly general conditions a closed periodicorbit is created.

dx

dt  = 1 + x2y − (α + 1)x

dy

dt  =   αx − x2y