Theory of Odes 2008

Theory of Ordinary Differential Equations
Theory of Ordinary Differential Equations

C HRISTOPHER P. G RANT
Brigham Young University
Contents
Contents
1 Fundamental Theory
1.1 ODEs and Dynamical Systems
1.2 Existence of Solutions . . . .
1.3 Uniqueness of Solutions . . .
1.4 Picard-Lindelof Theorem . . .
1.5 Intervals of Existence . . . . .
1.6 Dependence on Parameters . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
2 Linear Systems
2.1 Constant Coefficient Linear Equations .
2.2 Understanding the Matrix Exponential .
2.3 Generalized Eigenspace Decomposition
2.4 Operators on Generalized Eigenspaces .
2.5 Real Canonical Form . . . . . . . . . .
2.6 Solving Linear Systems . . . . . . . . .
2.7 Qualitative Behavior of Linear Systems
2.8 Exponential Decay . . . . . . . . . . .
2.9 Nonautonomous Linear Systems . . . .
2.10 Nearly Autonomous Linear Systems . .
2.11 Periodic Linear Systems . . . . . . . .
3 Topological Dynamics
3.1 Invariant Sets and Limit Sets . .
3.2 Regular and Singular Points . .
3.3 Definitions of Stability . . . . .
3.4 Principle of Linearized Stability
3.5 Lyapunovs Direct Method . . .
i
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
1
6
9
13
15
18
.
.
.
.
.
.
.
.
.
.
.
25
25
27
31
34
37
39
46
50
52
56
59
.
.
.
.
.
65
65
69
72
77
82
C ONTENTS
3.6
4
ii
LaSalles Invariance Principle . . . . . . . . . . . . . . . . . .
Conjugacies
4.1 Hartman-Grobman Theorem:
4.6 Constructing Conjugacies . .
4.7 Smooth Conjugacies . . . .
85
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
91
. 91
. 92
. 95
. 98
. 101
. 104
. 107
Invariant Manifolds
5.1 Stable Manifold Theorem: Part 1 . . . .
5.7 Center Manifolds . . . . . . . . . . . .
5.8 Computing and Using Center Manifolds
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Part 1
Part 2
Part 3
Part 4
Part 5
. . . .
. . . .
.
.
.
.
.
.
.
113
113
116
119
122
125
129
132
134
Periodic Orbits
139
6.1 Poincare-Bendixson Theorem . . . . . . . . . . . . . . . . . . 139
6.2 Lienards Equation . . . . . . . . . . . . . . . . . . . . . . . . 143
6.3 Lienards Theorem . . . . . . . . . . . . . . . . . . . . . . . . 147
1
Fundamental Theory
1.1 ODEs and Dynamical Systems
Ordinary Differential Equations
An ordinary differential equation (or ODE) is an equation involving derivatives
of an unknown quantity with respect to a single variable. More precisely, suppose
j; n 2 N, E is a Euclidean space, and
n C 1 copies

F W dom.F / R E E ! Rj :
(1.1)
Then an nth order ordinary differential equation is an equation of the form
F .t; x.t/; x.t/;

P
x.t/;
R
x .3/ .t/; ; x .n/ .t// D 0:
(1.2)
If I R is an interval, then x W I ! E is said to be a solution of (1.2) on I if

x has derivatives up to order n at every t 2 I, and those derivatives satisfy (1.2).
Often, we will use notation that suppresses the dependence of x on t. Also, there
will often be side conditions given that narrow down the set of solutions. In these
notes, we will concentrate on initial conditions which prescribe x .`/ .t0 / for some
fixed t0 2 R (called the initial time) and some choices of ` 2 f0; 1; : : : ; ng. Some
ODE texts examine two-point boundary-value problems, in which the value of a
function and its derivatives at two different points are required to satisfy given
algebraic equations, but we wont focus on them in this one.
1
1. F UNDAMENTAL T HEORY
First-order Equations
Every ODE can be transformed into an equivalent first-order equation. In particular, given x W I ! E, suppose we define
y1 WD x
y2 WD xP
y3 WD xR
::
:
yn WD x .n
1/
and let y W I ! E n be defined by y D .y1 ; : : : ; yn /. For i D 1; 2; : : : ; n

define
Gi W R E n E n ! E
by
G1 .t; u; p/ WD p1
u2
G3 .t; u; p/ WD p3
::
:
u4
G2 .t; u; p/ WD p2
Gn
1 .t; u; p/
WD pn
u3
un ;
and, given F as in (1.1), define Gn W dom.Gn / R E n E n ! Rj by

Gn .t; u; p/ WD F .t; u1 ; : : : ; un ; pn /;
where

dom.Gn / D .t; u; p/ 2 R E n E n .t; u1 ; : : : ; un ; pn / 2 dom.F / :
Letting G W dom.Gn / R E n E n ! E n 1 Rj be defined by

0 1
G1
BG2 C
B C
B C
G WD BG3 C ;
B :: C
@ : A
Gn
we see that x satisfies (1.2) if and only if y satisfies G.t; y.t/; y.t//
P
D 0.
1,
ODEs and Dynamical Systems
Equations Resolved with Respect to the Derivative

Consider the first-order initial-value problem (or IVP)
8
P D0
<F .t; x; x/
x.t0 / D x0
:
x.t
P 0 / D p0 ;
(1.3)
where F W dom.F / R Rn Rn ! Rn , and x0 ; p0 are given elements of Rn

satisfying F .t0 ; x0 ; p0 / D 0. The Implicit Function Theorem says that typically
the solutions .t; x; p/ of the (algebraic) equation F .t; x; p/ D 0 near .t0 ; x0 ; p0 /
form an .n C 1/-dimensional surface that can be parametrized by .t; x/. In other
words, locally the equation F .t; x; p/ D 0 is equivalent to an equation of the
form p D f .t; x/ for some f W dom.f / R Rn ! Rn with .t0 ; x0 / in the
interior of dom.f /. Using this f , (1.3) is locally equivalent to the IVP
(
xP D f .t; x/
x.t0 / D x0 :
Autonomous Equations
Let f W dom.f / R Rn ! Rn . The ODE
xP D f .t; x/
(1.4)
is autonomous if f doesnt really depend on t, i.e., if dom.f / D R for some

Rn and there is a function g W ! Rn such that f .t; u/ D g.u/ for every
t 2 R and every u 2 .
Every nonautonomous ODE is actually equivalent to an autonomous ODE.
To see why this is so, given x W R ! Rn , define y W R ! RnC1 by y.t/ D
.t; x1 .t/; : : : ; xn .t//, and given f W dom.f / R Rn ! Rn , define a new
function fQ W dom.fQ/ RnC1 ! RnC1 by
0
1
1
Bf1 .p1 ; .p2 ; : : : ; pnC1 //C
B
C
fQ.p/ D B
C;
::
@
A
:
fn .p1 ; .p2 ; : : : ; pnC1 //
where f D .f1 ; : : : ; fn /T and

dom.fQ/ D p 2 RnC1 .p1 ; .p2 ; : : : ; pnC1 // 2 dom.f / :
Then x satisfies (1.4) if and only if y satisfies yP D fQ.y/.
Because of the discussion above, we will focus our study on first-order autonomous ODEs that are resolved with respect to the derivative. This decision
is not completely without loss of generality, because by converting other sorts
of ODEs into equivalent ones of this form, we may be neglecting some special
structure that might be useful for us to consider. This trade-off between abstractness and specificity is one that you will encounter (and have probably already
encountered) in other areas of mathematics. Sometimes, when transforming the
equation would involve too great a loss of information, well specifically study
higher-order and/or nonautonomous equations.
Dynamical Systems
As we shall see, by placing conditions on the function f W Rn ! Rn and
the point x0 2 we can guarantee that the autonomous IVP
(
xP D f .x/
(1.5)
x.0/ D x0
has a solution defined on some interval I containing 0 in its interior, and this solution will be unique (up to restriction or extension). Furthermore, it is possible
to splice together solutions of (1.5) in a natural way, and, in fact, get solutions to IVPs with different initial times. These considerations lead us to study a
structure known as a dynamical system.
Given Rn , a continuous dynamical system (or a flow) on is a function
' W R ! satisfying:
1. '.0; x/ D x for every x 2 ;
2. '.s; '.t; x// D '.s C t; x/ for every x 2 and every s; t 2 R;
3. ' is continuous.
If f and are sufficiently nice we will be able to define a function ' W
R ! by letting '.; x0 / be the unique solution of (1.5), and this definition will make ' a dynamical system. Conversely, any continuous dynamical
system '.t; x/ that is differentiable with respect to t is generated by an IVP.
Exercise 1 Suppose that:

Rn ;
4
ODEs and Dynamical Systems
' W R ! is a continuous dynamical system;

@'.t; x/
exists for every t 2 R and every x 2 ;
@t
x0 2 is given;
y W R ! is defined by y.t/ WD '.t; x0 /;
f W!
Rn
@'.s; p/
is defined by f .p/ WD
.
@s sD0
Show that y solves the IVP
Rn ,
yP D f .y/
y.0/ D x0 :
In these notes we will also discuss discrete dynamical systems. Given

a discrete dynamical system on is a function ' W Z ! satisfying:
1. '.0; x/ D x for every x 2 ;

2. '.`; '.m; x// D '.` C m; x/ for every x 2 and every `; m 2 Z;
3. ' is continuous.
There is a one-to-one correspondence between discrete dynamical systems ' and
homeomorphisms (continuous functions with continuous inverses) F W ! ,
this correspondence being given by '.1; / D F . If we relax the requirement of
invertibility and take a (possibly noninvertible) continuous function F W !
and define ' W f0; 1; : : :g ! by
n copies

'.n; x/ D F .F . .F .x// //;
then ' will almost meet the requirements to be a dynamical system, the only
exception being that property 2, known as the group property may fail because
'.n; x/ is not even defined for n < 0. We may still call this a dynamical system;
if were being careful we may call it a semidynamical system.
In a dynamical system, the set is called the phase space. Dynamical systems are used to describe the evolution of physical systems in which the state
of the system at some future time depends only on the initial state of the system and on the elapsed time. As an example, Newtonian mechanics permits us
5
to view the earth-moon-sun system as a dynamical system, but the phase space
is not physical space R3 , but is instead an 18-dimensional Euclidean space in
which the coordinates of each point reflect the position and momentum of each
of the three objects. (Why isnt a 9-dimensional space, corresponding to the three
spatial coordinates of the three objects, sufficient?)
1.2 Existence of Solutions

Approximate Solutions
Consider the IVP
xP D f .t; x/
x.t0 / D a;
(1.6)
where f W dom.f / R Rn ! Rn is continuous, and .t0 ; a/ 2 dom.f / is

constant. The Fundamental Theorem of Calculus implies that (1.6) is equivalent
to the integral equation
x.t/ D a C
f .s; x.s// ds:
(1.7)
t0
How does one go about proving that (1.7) has a solution if, unlike the case
with so many IVPs studied in introductory courses, a formula for a solution cannot be found? One idea is to construct a sequence of approximate solutions,
with the approximations becoming better and better, in some sense, as we move
along the sequence. If we can show that this sequence, or a subsequence, converges to something, that limit might be an exact solution.
One way of constructing approximate solutions is Picard iteration. Here, we
plug an initial guess in for x on the right-hand side of (1.7), take the resulting
value of the left-hand side and plug that in for x again, etc. More precisely, we
can set x1 .t/ WD a and recursively define xkC1 in terms of xk for k > 1 by
xkC1 .t/ WD a C
f .s; xk .s// ds:

t0
Note that if, for some k, xk D xkC1 then we have found a solution.
Another approach is to construct a Tonelli sequence. For each k 2 N, let
xk .t/ be defined by
8
if t0 t t0 C 1=k
<a;
Z t 1=k
xk .t/ D
(1.8)
f .s; xk .s// dx; if t t0 C 1=k

:a C
t0
Existence of Solutions
for t t0 , and define xk .t/ similarly for t t0 .
We will use the Tonelli sequence to show that (1.7) (and therefore (1.6)) has a
solution, and will use Picard iterates to show that, under an additional hypothesis
on f , the solution of (1.7) is unique.
Existence
For the first result, we will need the following definitions and theorems.
Definition. A sequence of functions gk W U R ! Rn is uniformly bounded if
there exists M > 0 such that jgk .t/j M for every t 2 U and every k 2 N.
Definition. A sequence of functions gk W U R ! Rn is uniformly equicontinuous if for every " > 0 there exists a number > 0 such that jgk .t1 / gk .t2 /j < "
for every k 2 N and every t1 ; t2 2 U satisfying jt1 t2 j < .
Definition. A sequence of functions gk W U R ! Rn converges uniformly to a
function g W U R ! Rn if for every " > 0 there exists a number N 2 N such
that if k N and t 2 U then jgk .t/ g.t/j < ".
Definition. If a 2 Rn and > 0, then the open ball of radius centered at a,
denoted B.a; /, is the set

x 2 Rn jx aj < :
Theorem. (Arzela-Ascoli) Every uniformly bounded, uniformly equicontinuous

sequence of functions gk W U R ! Rn has a subsequence that converges
uniformly on compact (closed and bounded) sets.
Theorem. (Uniform Convergence) If a sequence of continuous functions hk W

b; c ! Rn converges uniformly to h W b; c ! Rn , then
Z c
Z c
lim
hk .s/ ds D
h.s/ ds:
k"1 b
We are now in a position to state and prove the Cauchy-Peano Existence

Theorem.
Theorem. (Cauchy-Peano) Suppose f W t0 ; t0 C B.a; / ! Rn
is continuous and bounded by M > 0. Then (1.7) has a solution defined on
t0 b; t0 C b, where b D minf; =M g.
Proof. For simplicity, we will only consider t 2 t0 ; t0 C b. For each k 2 N, let
xk W t0 ; t0 C b ! Rn be defined by (1.8). We will show that .xk / converges to
a solution of (1.6).
Step 1: Each xk is well-defined.
Fix k 2 N. Note that the point .t0 ; a/ is in the interior of a set on which f is
well-defined. Because of the formula for xk .t/ and the fact that it is, in essence,
recursively defined on intervals of width 1=k moving steadily to the right, if xk
failed to be defined on t0 ; t0 C b then there would be t1 2 t0 C 1=k; t0 C b/ for
which jxk .t1 / aj D . Pick the first such t1 . Using (1.8) and the bound on f ,
we see that
Z
Z
t1 1=k
t1 1=k
jxk .t1 / aj D
f .s; xk .s// ds
jf .s; xk .s//j ds
t0
t0
Z t1 1=k

M ds D M.t1 t0 1=k/ < M.b 1=k/
t0
M=k < D jxk .t1 /
aj;
which is a contradiction.
Step 2: .xk / is uniformly bounded.
Calculating as above, we find that (1.8) implies that, for k 1=b,
jxk .t/j jaj C
bCt0 1=k
t0
jf .s; xk .s//j dx jaj C .b
Step 3: .xk / is uniformly equicontinuous.

If t1 ; t2 t0 C 1=k, then
Z t2
Z
jxk .t1 / xk .t2 /j D

f .s; xk .s// ds
t1
M jt2
t1 j:
t2
t1
1=k/M < jaj C :
jf .s; xk .s//j ds
Since xk is constant on t0 ; t0 C 1=k and continuous at t0 C 1=k, the estimate

jxk .t1 / xk .t2 /j M jt2 t1 j holds for every t1 ; t2 2 t0 ; t0 C b. Thus, given
" > 0, we can set D "=M and see that uniform equicontinuity holds.
Step 4: Some subsequence .xk` / of .xk / converges uniformly, say to x, on
the interval t0 ; t0 C b.
This follows directly from the previous steps and the Arzela-Ascoli Theorem.
Step 5: The function f .; x.// is the uniform limit of .f .; xk` ./// on the
interval t0 ; t0 C b.
Let " > 0 be given. Since f is continuous on a compact set, it is uniformly
continuous. Thus, we can pick > 0 such that jf .s; p/ f .s; q/j < " whenever
Uniqueness of Solutions
jp qj < . Since .xk` / converges uniformly to x, we can pick N 2 N such
that jxk` .s/ x.s/j < whenever s 2 t0 ; t0 C b and ` N . If ` N , then
jf .s; xk` .s// f .s; x.s//j < ".
Step 6: The function x is a solution of (1.6).
Fix t 2 t0 ; t0 C b. If t D t0 , then clearly (1.7) holds. If t > t0 , then for `
sufficiently large
xk` .t/ D a C
t
t0
f .s; xk` .s// ds
t
t 1=k`
f .s; xk` .s// ds:
(1.9)
Obviously, the left-hand side of (1.9) converges to x.t/ as ` " 1. By the Uniform Convergence Theorem and the uniform convergence of .f .; xk` .///, the
first integral on the right-hand side of (1.9) converges to
Z
f .s; x.s// ds;

t0
and by the boundedness of f the second integral converges to 0. Thus, taking

the limit of (1.9) as ` " 1 we see that x satisfies (1.7), and therefore (1.6), on
t0 ; t0 C b.
Note that this theorem guarantees existence, but not necessarily uniqueness,
of a solution of (1.6).
Exercise 2 How many solutions of the IVP

(
p
xP D 2 jxj
x.0/ D 0;
on the interval . 1; 1/ are there? Give formulas for all of them.
1.3 Uniqueness of Solutions

Uniqueness
If more than continuity of f is assumed, it may be possible to prove that
(
xP D f .t; x/
(1.10)
x.t0 / D a;
has a unique solution. In particular Lipschitz continuity of f .t; / is sufficient.
(Recall that g W dom.g/ Rn ! Rn is Lipschitz continuous if there exists a
constant L > 0 such that jg.x1 / g.x2 /j Ljx1 x2 j for every x1 ; x2 2
dom.g/; L is called a Lipschitz constant for g.)
One approach to uniqueness is developed in the following exercise, which
uses what are know as Gronwall inequalities.
Exercise 3 Assume that the conditions of the Cauchy-Peano Theorem

hold and that, in addition, f .t; / is Lipschitz continuous with Lipschitz
constant L for every t. Show that the solution of (1.10) is unique on t0 ; t0 C
b by completing the following steps. (The solution can similarly be shown
to be unique on t0 b; t0 , but we wont bother doing that here.)
(a) If x1 and x2 are each solutions of (1.10) on t0 ; t0 C b and U W t0 ; t0 C
b ! R is defined by U.t/ WD jx1 .t/ x2 .t/j, show that
Z t
U.t/ L
U.s/ ds
(1.11)
t0
for every t 2 t0 ; t0 C b.
(b) Pick " > 0 and let
V .t/ WD " C L
U.s/ ds:
t0
Show that
V 0 .t/ LV .t/
(1.12)
for every t 2 t0 ; t0 C b, and that V .t0 / D ".

(c) Dividing both sides of (1.12) by V .t/ and integrating from t D t0 to
t D T , show that V .T / " expL.T t0 /.
(d) By using (1.11) and letting " # 0, show that U.T / D 0 for all T 2
t0 ; t0 C b, so x1 D x2 .
We will prove an existence-uniqueness theorem that combines the results of
the Cauchy-Peano Theorem and Exercise 3. The reason for this apparently redundant effort is that the concepts and techniques introduced in this proof will be
useful throughout these notes.
10
Uniqueness of Solutions
First, we review some definitions and results pertaining to metric spaces.
Definition. A metric space is a set X together with a function d W X X ! R
satisfying:
1. d.x; y/ 0 for every x; y 2 X , with equality if and only if x D y;
2. d.x; y/ D d.y; x/ for every x; y 2 X ;
3. d.x; y/ C d.y; z/ d.x; z/ for every x; y; z 2 X .
Definition. A normed vector space is a vector space together with a function
k k W X ! R satisfying:
1. kxk 0 for every x 2 X , with equality if and only if x D 0;
2. kxk D jjkxk for every x 2 X and every scalar ;
3. kx C yk kxk C kyk for every x; y 2 X .
Every normed vector space is a metric space with metric d.x; y/ D kx
yk.
Definition. An inner product space is a vector space together with a function

h; i W X X ! R satisfying:
1. hx; xi 0 for every x 2 X , with equality if and only if x D 0;
2. hx; yi D hy; xi for every x; y 2 X ;
3. hx C y; zi D hx; zi C hy; zi for every x; y; z 2 X and all scalars
; .
p Every inner product space is a normed vector space with norm kxk equal to
hx; xi.
Definition. A sequence .xn / in a metric space X is said to be (a) Cauchy (sequence) if for every " > 0, there exists N 2 N such that d.xm ; xn / < " whenever
m; n N .
Definition. A sequence .xn / in a metric space X converges to x if for every

" > 0, there exists N 2 N such that d.xn ; x/ < " whenever n N .
11
Definition. A metric space is said to be complete if every Cauchy sequence in X
converges (in X ). A complete normed vector space is called a Banach space. A
complete inner product space is called a Hilbert space.
Definition. A function f W X ! Y from a metric space to a metric space is
said to be Lipschitz continuous if there exists L 2 R such that d.f .u/; f .v//
Ld.u; v/ for every u; v 2 X . We call L a Lipschitz constant, and write Lip.f /
for the smallest Lipschitz constant that works.
Definition. A contraction is a Lipschitz continuous function from a metric space
to itself that has Lipschitz constant less than 1.
Definition. A fixed point of a function T W X ! X is a point x 2 X such that
T .x/ D x.
Theorem. (Contraction Mapping Principle) If X is a nonempty, complete metric space and T W X ! X is a contraction, then T has a unique fixed point in
X.
Proof. Pick < 1 such that d.T .x/; T .y// d.x; y/ for every x; y 2 X .
Pick any point x0 2 X . Define a sequence .xk / by the recursive formula
xkC1 D T .xk /:
(1.13)
If k ` N , then
d.xk ; x` / d.xk ; xk
::
:
d.xk
k

1/
C d.xk
1 ; xk 2 /
C d.xk
d.x1 ; x0 / C k
N
d.x1 ; x0 /:
1
1 ; xk 2 /
C C d.x`C1 ; x` /
2 ; xk 3 /
C C d.x` ; x`
1/
d.x1 ; x0 / C C ` d.x1 ; x0 /
Hence, .xk / is a Cauchy sequence. Since X is complete, .xk / converges to some

x 2 X . By letting k " 1 in (1.13) and using the continuity of T , we see that
x D T .x/, so x is a fixed point of T .
If there were another fixed point y of T , then
d.x; y/ D d.T .x/; T .y// d.x; y/;
so d.x; y/ D 0, which means x D y. This shows uniqueness of the fixed
point.
12
Picard-Lindelof Theorem
1.4 Picard-Lindelof Theorem

Theorem. The space C.a; b/ of continuous functions from a; b to Rn equipped
with the norm

kf k1 WD sup jf .x/j x 2 a; b
is a Banach space.
Definition. Two different norms kk1 and kk2 on a vector space X are equivalent
if there exist constants m; M > 0 such that
mkxk1 kxk2 M kxk1
for every x 2 X .
Theorem. If .X ; k k1 / is a Banach space and k k2 is equivalent to k k1 on X ,
then .X ; k k2 / is a Banach space.
Theorem. A closed subspace of a complete metric space is a complete metric
space.
We are now in a position to state and prove the Picard-Lindelof ExistenceUniqueness Theorem. Recall that we are dealing with the IVP
(
xP D f .t; x/
(1.14)
x.t0 / D a:
Theorem. (Picard-Lindelof) Suppose f W t0 ; t0 C B.a; / ! Rn is
continuous and bounded by M . Suppose, furthermore, that f .t; / is Lipschitz
continuous with Lipschitz constant L for every t 2 t0 ; t0 C . Then (1.14)
has a unique solution defined on t0 b; t0 C b, where b D minf; =M g.
Proof. Let X be the set of continuous functions from t0
The norm
kgkw WD sup e
2Ljt t0 j
jg.t/j t 2 t0
b; t0 C b to B.a; /.
b; t0 C b
is equivalent to the standard supremum norm k k1 on C.t0 b; t0 C b/, so this

vector space is complete under this weighted norm. The set X endowed with this
norm/metric is a closed subset of this complete Banach space, so X equipped
with the metric d.x1 ; x2 / WD kx1 x2 kw is a complete metric space.
13
Given x 2 X , define T .x/ to be the function on t0
formula
Z
b; t0 C b given by the
T .x/.t/ D a C
f .s; x.s// ds:
t0
Step 1: If x 2 X then T .x/ makes sense.

This should be obvious.
Step 2: If x 2 X then T .x/ 2 X .
If x 2 X , then it is clear that T .x/ is continuous (and, in fact, differentiable).
Furthermore, for t 2 t0 b; t0 C b
Z t
Z t
jT .x/.t/ aj D
f .s; x.s// ds
jf .s; x.s//j ds M b ;
t0
t0
so T .x/.t/ 2 B.a; /. Hence, T .x/ 2 X .

Step 3: T is a contraction on X .
Let x; y 2 X , and note that kT .x/ T .y/kw is
(
Z t
2Ljt t0 j
sup e
f
.s;
x.s//
f
.s;
y.s//
ds
t0
For a fixed t 2 t0
t 2 t0
b; t0 C b :
b; t0 C b,
Z t
2Ljt t0 j
e
f .s; x.s// f .s; y.s// ds
t0
Z t
2Ljt t0 j
jf .s; x.s// f .s; y.s//j ds
e
t
Z 0t
2Ljt t0 j
e
Ljx.s/ y.s/j ds
t
Z0 t
2Ljt t0 j
2Ljs t0 j
Le
kx ykw e
ds
t0
kx ykw
D
1
2
1
kx ykw :
2
2Ljt t0 j
Taking the supremum over all t 2 t0 b; t0 C b, we find that T is a contraction

(with D 1=2).
By the contraction mapping principle, we therefore know that T has a unique
fixed point in X . This means that (1.14) has a unique solution in X (which is the
only place a solution could be).
14
Intervals of Existence
1.5 Intervals of Existence

Maximal Interval of Existence
We begin our discussion with some definitions and an important theorem of real
analysis.
Definition. Given f W D RRn ! Rn , we say that f .t; x/ is locally Lipschitz
continuous with respect to x on D if for each .t0 ; a/ 2 D there is a number L and
a product set I U D containing .t0 ; a/ in its interior such that the restriction
of f .t; / to U is Lipschitz continuous with Lipschitz constant L for every t 2 I.
Definition. A subset K of a topological space is compact if whenever K is contained in the union of a collection of open sets, there is a finite subcollection of
that collection whose union also contains K. The original collection is called a
cover of K, and the finite subcollection is called a finite subcover of the original
cover.
Theorem. (Heine-Borel) A subset of Rn is compact if and only if it is closed and
bounded.
Now, suppose that D is an open subset of R Rn , .t0 ; a/ 2 D, and f W D !
is locally Lipschitz continuous with respect to x on D. Then the PicardLindelof Theorem indicates that the IVP
(
xP D f .t; x/
(1.15)
x.t0 / D a
Rn
has a solution existing on some time interval containing t0 in its interior and that
the solution is unique on that interval. Lets say that an interval of existence is an
interval containing t0 on which a solution of (1.15) exists. The following theorem
indicates how large an interval of existence may be.
Theorem. (Maximal Interval of Existence) The IVP (1.15) has a maximal interval of existence, and it is of the form .! ; !C /, with ! 2 1; 1/ and
!C 2 . 1; 1. There is a unique solution x.t/ of (1.15) on .! ; !C /, and
.t; x.t// leaves every compact subset K of D as t # ! and as t " !C .
Proof.
Step 1: If I1 and I2 are open intervals of existence with corresponding solutions x1 and x2 , then x1 and x2 agree on I1 \ I2 .
Let I D I1 \ I2 , and let I be the largest interval containing t0 and contained in
15
I on which x1 and x2 agree. By the Picard-Lindelof Theorem, I is nonempty. If
I I, then I has an endpoint t1 in I. By continuity, x1 .t1 / D x2 .t1 / DW a1 .
The Picard-Lindelof Theorem implies that the new IVP
(
xP D f .t; x/
(1.16)
x.t1 / D a1
has a local solution that is unique. But restrictions of x1 and x2 near t1 each
provide a solution to (1.16), so x1 and x2 must agree in a neighborhood of t1 .
This contradicts the definition of t1 and tells us that I D I.
Now, let .! ; !C / be the union of all open intervals of existence.
Step 2: .! ; !C / is an interval of existence.
Given t 2 .! ; !C /, pick an open interval of existence IQ that contains t, and
Q Because of step 1, this
let x.t/ D x.t/,
Q
where xQ is a solution to (1.15) on I.
determines a well-defined function x W .! ; !C / ! Rn ; clearly, it solves (1.15).
Step 3: .! ; !C / is the maximal interval of existence.
An extension argument similar to the one in Step 1 shows that every interval of
existence is contained in an open interval of existence. Every open interval of
existence is, in turn, a subset of .! ; !C /.
Step 4: x is the only solution of (1.15) on .! ; !C /.
This is a special case of Step 1.
Step 5: .t; x.t// leaves every compact subset K D as t # ! and as
t " !C .
We only treat what happens as t " !C ; the other case is similar. Furthermore,
the case when !C D 1 is immediate, so suppose !C is finite.
Let a compact subset K of D be given. Since D is open, for each point
.t; a/ 2 K we can pick numbers .t; a/ > 0 and .t; a/ > 0 such that
t
2.t; a/; t C 2.t; a/ B.a; 2.t; a// D:
Note that the collection of sets

.t .t; a/; t C .t; a// B.a; .t; a// .t; a/ 2 K
is a cover of K. Since K is compact, a finite subcollection, say
m
.ti .ti ; ai /; ti C .ti ; ai // B.ai ; .ti ; ai // iD1 ;
covers K. Let
K0 WD
16
m
[
iD1
ti

2.ti ; ai /; ti C 2.ti ; ai / B.ai ; 2.ti ; ai // ;
Intervals of Existence
let
m
Q WD min .ti ; ai / iD1 ;
and let
Q WD min .ti ; xi / iD1 :
Since K0 is a compact subset of D, there is a constant M > 0 such that f is

bounded by M on K0 . By the triangle inequality,
t0
Q K0 ;
;
Q t0 C
Q B.a; /
for every .t0 ; a/ 2 K, so f is bounded by M on each such product set. According

to the Picard-Lindelof Theorem, this means that for every .t0 ; a/ 2 K a solution
Q
Q =M
g units of time.
to xP D f .t; x/ starting at .t0 ; a/ exists for at least minf;
Q
Hence, x.t/ K for t > !C minf;
Q =M g.
Corollary. If D 0 is a bounded set and D D .c; d / D 0 (with c 2 1; 1/ and
d 2 . 1; 1), then either !C D d or x.t/ ! @D 0 as t " !C , and either
! D c or x.t/ ! @D 0 as t # ! .
Corollary. If D D .c; d / Rn (with c 2 1; 1/ and d 2 . 1; 1), then
either !C D d or jx.t/j " 1 as t " !C , and either ! D c or jx.t/j " 1 as
t #! .
If were dealing with an autonomous equation on a bounded set, then the first
corollary applies to tell us that the only way a solution could fail to exist for all
time is for it to approach the boundary of the spatial domain. (Note that this is
not the same as saying that x.t/ converges to a particular point on the boundary;
can you give a relevant example?) The second corollary says that autonomous
equations on all of Rn have solutions that exist until they become unbounded.
Global Existence
For the solution set of the autonomous ODE xP D f .x/ to be representable by
a dynamical system, it is necessary for solutions to exist for all time. As the
discussion above illustrates, this is not always the case. When solutions do die
out in finite time by hitting the boundary of the phase space or by going off to
infinity, it may be possible to change the vector field f to a vector field fQ that
points in the same direction as the original but has solutions that exist for all time.
For example, if D Rn , then we could consider the modified equation
f .x/
:
xP D p
1 C jf .x/j2
17
Clearly, jxj
P < 1, so it is impossible for jxj to approach infinity in finite time.
If, on the other hand, Rn , then consider the modified equation
d.x; Rn n /
f .x/
p
;
xP D p
1 C jf .x/j2
1 C d.x; Rn n /2
where d.x; Rn n / is the distance from x to the complement of . It is not

hard to show that it is impossible for a solution x of this equation to become
unbounded or to approach the complement of in finite time, so, again, we have
global existence.
It may or may not seem obvious that if two vector fields point in the same
direction at each point, then the solution curves of the corresponding ODEs in
phase space match up. In the following exercise, you are asked to prove that this
is true.
Exercise 4 Suppose that is a subset of Rn , that f W ! Rn and

g W ! Rn are (continuous) vector fields, and that there is a continuous
function h W ! .0; 1/ such that g.u/ D h.u/f .u/ for every u 2 . If x
is the only solution of
(
xP D f .x/
x.0/ D a
(defined on the maximal interval of existence) and y is the only solution of
(
yP D g.y/
y.0/ D a;
(defined on the maximal interval of existence), show that there is an increasing function j W dom.y/ ! dom.x/ such that y.t/ D x.j.t// for
every t 2 dom.y/.
1.6 Dependence on Parameters

Parameters vs. Initial Conditions
Consider the IVP
18
xP D f .t; x/
x.t0 / D a;
(1.17)
Dependence on Parameters
and the paramterized IVP
(
xP D f .t; x; /
x.t0 / D a;
(1.18)
where 2 Rk . We are interested in studying how the solution of (1.17) depends

on the initial condition a and how the solution of (1.18) depends on the parameter
. In a sense, these two questions are equivalent. For example, if x solves (1.17)
and we let xQ WD x a and fQ.t; x;
Q a/ WD f .t; xQ C a/, then xQ solves
(
xPQ D fQ.t; x;
Q a/
x.t
Q 0 / D 0;
so a appears as a parameter rather than an initial condition. If, on the other hand,
x solves (1.18), and we let xQ WD .x; / and fQ.t; x/
Q WD .f .t; x; /; 0/, then xQ
solves
(
xPQ D fQ.t; x/
Q
x.t
Q 0 / D .a; /;
so appears as part of the initial condition, rather than as a parameter in the
ODE.
We will concentrate on (1.18).
Continuous Dependence
The following result can be proved by an approach like that outlined in Exercise
3.
Theorem. (Grownwall Inequality) Suppose that X.t/ is a nonnegative, continuous, real-valued function on t0 ; T and that there are constants C; K > 0 such
that
Z
t
X.t/ C C K
X.s/ ds
t0
for every t 2 t0 ; T . Then

X.t/ C e K.t
t0 /
for every t 2 t0 ; T .
Using the Grownwall inequality, we can prove that the solution of (1.18)
depends continuously on .
19
Theorem. (Continuous Dependence) Suppose
f W t0
; t0 C 1 2 R Rn Rk ! Rn
is continuous. Suppose, furthermore, that f .t; ; / is Lipschitz continuous with

Lipschitz constant L1 > 0 for every .t; / 2 t0 ; t0 C 2 and f .t; x; /
is Lipschitz continuous with Lipschitz constant L2 > 0 for every .t; x/ 2 t0
; t0 C 1 . If xi W t0 ; t0 C ! Rn (i D 1; 2) satisfy
(
xPi D f .t; xi ; i /
xi .t0 / D a;
then
jx1 .t/
for t 2 t0
x2 .t/j
L2
j1
L1
2 j.e L1 jt
t0 j
1/
(1.19)
; t0 C .
This theorem shows continuous dependence on parameters if, in addition to

the hypotheses of the Picard-Lindelof Theorem, the right-hand side of the ODE
in (1.18) is assumed to be Lipschitz continuous with respect to the parameter (on
finite time intervals). The connection between (1.17) and (1.18) shows that the
hypotheses of the Picard-Lindelof Theorem are sufficient to guarantee continuous dependence on initial conditions. Note the exponential dependence of the
modulus of continuity on jt t0 j.
Proof. For simplicity, we only consider t t0 . Note that
Z t
jx1 .t/ x2 .t/j D f .s; x1 .s/; 1 / f .s; x2 .s/; 2 ds

C

Let X.t/ D L2 j1
t0
t
t0
Z t
t0
jf .s; x1 .s/; 1 /
f .s; x2 .s/; 2 /j ds
jf .s; x1 .s/; 1 /
f .s; x1 .s/; 2 /j ds
jf .s; x1 .s/; 2 /
f .s; x2 .s/; 2 /j ds
t0
t
t0
L2 j1
2 j C L1 jx1 .t/
X.t/ L2 j1
20
2 j C L1 jx1 .s/
x2 .t/j. Then
Z t
2 j C L1
X.s/ ds;
t0
x2 .s/j ds
so by the Gronwall Inequality X.t/ L2 j1
(1.19) holds.
2 je L1 .t
t0 / .
This means that
Exercise 5 Suppose that f W R R ! R and g W R R ! R are

continuous and are each Lipschitz continuous with respect to their second
variable. Suppose, also, that x is a global solution to
(
xP D f .t; x/
x.t0 / D a;
and y is a global solution to
(
yP D g.t; y/
y.t0 / D b:
(a) If f .t; p/ < g.t; p/ for every .t; p/ 2 R R and a < b, show that
x.t/ < y.t/ for every t t0 .
(b) If f .t; p/ g.t; p/ for every .t; p/ 2 R R and a b, show that
x.t/ y.t/ for every t t0 . (Hint: You may want to use the results
of part (a) along with a limiting argument.)
Differentiable Dependence
Theorem. (Differentiable Dependence) Suppose f W R R R ! R is a
continuous function and is continuously differentiable with respect to x and .
Then the solution x.; / of
(
xP D f .t; x; /
(1.20)
x.t0 / D a
is differentiable with respect to , and y D x WD @x=@ satisfies
(
yP D fx .t; x.t; /; /y C f .t; x.t; /; /
y.t0 / D 0:
(1.21)
That x , if it exists, should satisfy the IVP (1.21) is not terribly surprising;
(1.21) can be derived (formally) by differentiating (1.20) with respect to . The
real difficulty is showing that x exists. The key to the proof is to use the fact
21
that (1.21) has a solution y and then to use the Gronwall inequality to show that
difference quotients for x converge to y.
Proof. Given , it is not hard to see that the right-hand side of the ODE in (1.21)
is continuous in t and y and is locally Lipschitz continuous with respect to y,
so by the Picard-Lindelof Theorem we know that (1.21) has a unique solution
y.; /. Let
x.t; C / x.t; /
w.t; ; / WD
:

We want to show that w.t; ; / ! y.t; / as ! 0.
Let z.t; ; / WD w.t; ; / y.t; /. Then
dz
dw
.t; ; / D
.t; ; /
dt
dt
fx .t; x.t; /; /y.t; /
f .t; x.t; /; /;
and
f .t; x.t; C /; C / f .t; x.t; /; /
dw
.t; ; / D
dt

f .t; x.t; C /; C / f .t; x.t; /; C /
D

f .t; x.t; /; C / f .t; x.t; /; /
C

D fx .t; x.t; / C 1 x; C /w.t; ; / C f .t; x.t; /; C 2 /;
for some 1 ; 2 2 0; 1 (by the Mean Value Theorem), where
x WD x.t; C /
x.t; /:
Hence,
dz
.t; ; / jf .t; x.t; /; C 2 / f .t; x.t; /; /j

dt
C jfx .t; x.t; / C 1 x; C /j jz.t; ; /j

C jfx .t; x.t; / C 1 x; C /
C jfx .t; x.t; /; C /
fx .t; x.t; /; C /j jy.t; /j
fx .t; x.t; /; /j jy.t; /j
p.t; ; / C .jfx .t; x.t; /; /j C p.t; ; //jz.t; ; /j;

where p.t; ; / ! 0 as ! 0, uniformly on bounded sets.
Letting X.t/ D " C .K C "/jzj, we see that if
22
jfx .t; x.t; /; /j K
(1.22)
and
jp.t; ; /j < ";
then
so
(1.23)
Z t
Z t
dz
X.s/ ds
jz.t; ; /j
.s; ; / ds
t0 ds
t0
X.t/ " C .K C "/
which gives X.t/ "e .KC"/.t
t0 / ,
jzj
X.s/ ds;
t0
by Gronwalls inequality. This, in turn, gives
".e .KC"/.t t0 /
K C"
1/
Given t t0 , pick K so large that (1.22) holds. As ! 0, we can take "

arbitrarily small and still have (1.23) hold, to see that
lim z.t; ; / D 0:
!0
23
2
Linear Systems
2.1 Constant Coefficient Linear Equations
Linear Equations
Definition. Given
f W R Rn ! Rn ;
we say that the first-order ODE
xP D f .t; x/
(2.1)
is linear if every linear combination of solutions of (2.1) is a solution of (2.1).

Definition. Given two vector spaces X and Y, L.X ; Y/ is the space of linear
maps from X to Y.
Exercise 6 Show that if (2.1) is linear (and f is continuous) then there

is a function A W R ! L.Rn ; Rn / such that f .t; p/ D A.t/p, for every
.t; p/ 2 R Rn .
ODEs of the form xP D A.t/x C g.t/ are also often called linear, although
they dont satisfy the definition given above. These are called inhomogeneous;
ODEs satisfying the previous definition are called homogeneous.
Constant Coefficients and the Matrix Exponential

Here we will study the autonomous IVP
(
xP D Ax
x.0/ D x0 ;
25
(2.2)
2. L INEAR S YSTEMS
where A 2 L.Rn ; Rn /, or equivalently A is a (constant) n n matrix.
If n D 1, then were dealing with xP D ax. The solution is x.t/ D e t a x0 .
When n > 1, we will show that we can similarly define e tA in a natural way, and
the solution of (2.2) will be given by x.t/ D e tA x0 .
Given B 2 L.Rn ; Rn /, we define its matrix exponential
e
1
X
Bk
WD
:
k
kD0
We will show that this series converges, but first we specify a norm on L.Rn ; Rn /.
Definition. The operator norm kBk of an element B 2 L.Rn ; Rn / is defined by
jBxj
D sup jBxj:
x0 jxj
jxjD1
kBk D sup
L.Rn ; Rn / is a Banach space under the operator norm. Thus, to show that
the series for e B converges, it suffices to show that

m
X Bk

k
kD`
can be made arbitrarily small by taking m ` N for N sufficiently large.

Suppose B1 ; B2 2 L.Rn ; Rn / and B2 does not map everything to zero. Then
jB1 B2 xj
jB1 B2 xj jB2 xj
D
sup

jxj
jxj
x0
B2 x0;x0 jB2 xj
!
!
jB1 yj
jB2 xj
sup
sup
D kB1 k kB2 k:
y0 jyj
x0 jxj
kB1 B2 k D sup
If B2 does map everything to zero, then kB2 k D kB1 B2 k D 0, so kB1 B2 k

kB1 k kB2 k, obviously. Thus, the operator norm is submultiplicative. Using this
property, we have
m

m k
m
X B k X
k

B X kBk
:

k
k
k
kD`
kD`
kD`
Since the regular exponential series (for real arguments) has an infinite radius
of convergence, we know that the last quantity in this estimate goes to zero as
`; m " 1.
Thus, e B makes sense, and, in particular, e tA makes sense for each fixed
t 2 R and each A 2 L.Rn ; Rn /. But does x.t/ WD e tA x0 solve (2.2)? To check
that, well need the following important property.
26
Understanding the Matrix Exponential

Lemma. If B1 ; B2 2 L.Rn ; Rn / and B1 B2 D B2 B1 , then e B1 CB2 D e B1 e B2 .
Proof. Using commutativity, we have
e B1 e B2
1
!
1
1
1 X
1
1 X
j
j
j
X
X
X
X
B1
B2k
B1 B2k
B1 B2k
@
A
D
D
D
j
k
j k
j k
j D0
j D0 kD0
kD0
iD0 j CkDi
1 X
i
1 X
i j .i
j .i j /
X
X
B1 B2
i B1 B2
D
D
j
j .i j /
i
iD0 j D0
iD0 j D0
1
X
.B1 C B2 /i
iD0
j/
D e B1 CB2 :
Now, if x W R ! Rn is defined by x.t/ WD e tA x0 , then

d
x.t C h/
x.t/ D lim
dt
h
h!0
x.t/
e .t Ch/A
D lim
h
h!0
e tA
lim
h!0
1
X
hk
kD1
1 Ak
D lim
h!0
!
e .t Ch/A x0
h
e tA x0
e hA I
x0 D lim
h
h!0
e tA x0
e tA x0 D Ae tA x0 D Ax.t/;
so x.t/ D e tA x0 really does solve (2.2).
2.2 Understanding the Matrix Exponential

Transformations
Now that we have a representation of the solution of linear constant-coefficient
initial-value problems, we should ask ourselves: What good is it? Does the
power series formula for the matrix exponential provide an efficient means for
calculating exact solutions? Not usually. Is it an efficient way to compute accurate numerical approximations to the matrix exponential? Not according to
Matrix Computations by Golub and Van Loan. Does it provide insight into how
solutions behave? It is not clear that it does. There are, however, transformations
that may help us handle these problems.
27
2. L INEAR S YSTEMS
Suppose that B; P 2 L.Rn ; Rn / are related by a similarity transformation;
i.e., B D QPQ 1 for some invertible Q. Calculating, we find that
eB D
1
1
X
X
Bk
.QPQ
D
k
k
kD0
kD0
1 /k
1
X
QP k Q
k
kD0
1
X
Pk
DQ
k
kD0
D Qe P Q
It would be nice if, given B, we could choose Q so that P were a diagonal

matrix, since (as can easily be checked)
e diagfp1 ;p2 ;:::;pn g D diagfe p1 ; e p2 ; : : : ; e pn g:
Unfortunately, this cannot always be done. Over the next few sections, we will
show that what can be done, in general, is to pick Q so that P D S C N , where
S is a semisimple matrix with a fairly simple form, N is a nilpotent matrix with
a fairly simple form, and S and N commute. (Recall that a matrix is semisimple
if it is diagonalizable over the complex numbers and that a matrix is nilpotent if
some power of the matrix is 0.) The forms of S and N are simple enough that
we can calculate their exponentials fairly easily, and then we can multiply them
to get the exponential of S C N .
We will spend a significant amount of time carrying out the project described
in the previous paragraph, even though it is linear algebra that some of you have
probably seen before. Since understanding the behavior of constant coefficient
systems plays a vital role in helping us understand more complicated systems,
I feel that the time investment is worth it. The particular approach we will take
follows chapters 3, 4, 5, and 6, and appendix 3 of Hirsch and Smale fairly closely.
Eigensystems
28
Given B 2 L.Rn ; Rn /, recall that that 2 C is an eigenvalue of B if Bx D x

Q D x for some nonzero x 2 Cn , where BQ is
for some nonzero x 2 Rn or if Bx
the complexification of B; i.e., the element of L.Cn ; Cn / which agrees with B on
Rn . (Just as we often identify a linear operator with a matrix representation of it,
we will usually not make a distinction between an operator on a real vector space
and its complexification.) A nonzero vector x for which Bx D x for some
scalar is an eigenvector. An eigenvalue with corresponding eigenvector x
form an eigenpair .; x/.
If an operator A 2 L.Rn ; Rn / were chosen at random, it would almost surely
have n distinct eigenvalues f1 ; : : : ; n g and n corresponding linearly indepen-
Understanding the Matrix Exponential

dent eigenvectors fx1 ; : : : ; xn g. If this is the case, then A is similar to the (possibly complex) diagonal matrix
2
3
1 0 0
6
: 7
6 0 : : : : : : :: 7
6
7:
6 :: : : : :
7
4:
:
: 05
0 0 n
More specifically,
2
A D 4x1
2
3 1
6
60
xn 5 6
6 ::
4:
0
0
::
:
::
:

::
:
::
:
0
3
0
3
2
:: 7
: 7
7 4x1 xn 5
7
05
n
If the eigenvalues of A are real and distinct, then this means that
2
3
3 t1 0 0
2
3
2
6
:: 7
6 0 ::: :::
7
: 7 4
5
tA D 4x1 xn 5 6
6 ::
7 x1 xn
:: ::
4 :
:
: 0 5
0
tn
and the formula for the matrix exponential then yields

2 t
3
1
0
0
2
3 e
2
3
6
:: 7
:: ::
6
7
:
:
0
: 7 4
5
e tA D 4x1 xn 5 6
6 ::
7 x1 xn
:: ::
4 :
5
:
:
0
0
e t n
This formula should make clear how the projections of e tA x0 grow or decay as
t ! 1.
The same sort of analysis works when the eigenvalues are (nontrivially) complex, but the resulting formula is not as enlightening. In addition to the difficulty
of a complex change of basis, the behavior of e t k is less clear when k is not
real.
One way around this is the following. Sort the eigenvalues (and eigenvectors)
of A so that complex conjugate eigenvalues f1 ; 1 ; : : : ; m ; m g come first and
are grouped together and so that real eigenvalues fmC1 ; : : : ; r g come last. For
k m, set ak D Re k 2 R, bk D Im k 2 R, yk D Re xk 2 Rn , and
29
2. L INEAR S YSTEMS
zk D Im xk 2 Rn . Then
Ayk D A Re xk D Re Axk D Re k xk D .Re k /.Re xk /
D ak y k
.Im k /.Im xk /
bk zk ;
and
Azk D A Im xk D Im Axk D Im k xk D .Im k /.Re xk / C .Re k /.Im xk /
D bk yk C ak zk :
Using these facts, we have A D QPQ

2
1,
where
3
Q D 4z1 y1 zm ym xmC1 xr 5
and P is the .m C r/ .m C r/ block diagonal matrix, whose first m diagonal

blocks are the 2 2 matrices

bk
ak
Ak D
bk ak
for k D 1; : : : ; m, and whose last r m diagonal blocks are the 1 1 matrices
k for k D m C 1; : : : ; r.
In order to compute e tA from this formula, well need to know how to compute e tAk . This can be done using the power series formula. An alternative
approach is to realize that

x.t/
tAk c
WD e
y.t/
d
is supposed to solve the IVP
8
xP D ak x bk y
<yP D b x C a y
k
k
x.0/
D
c
:y.0/ D d:
(2.3)
Since we can check that the solution of (2.3) is

a t

x.t/
e k .c cos bk t d sin bk t/
D ak t
;
y.t/
e .d cos bk t C c sin bk t/
we can conclude that
e
30
tAk
a t
e k cos bk t
D ak t
e sin bk t
e ak t sin bk t
e ak t cos bk t
Generalized Eigenspace Decomposition

Putting this all together and using the form of P , we have e tA D Qe tP Q 1 ,
where e tP is the .m C r/ .m C r/ block diagonal matrix whose first m diagonal
blocks are the 2 2 matrices

a t
e k cos bk t
e ak t sin bk t
e ak t sin bk t e ak t cos bk t
for k D 1; : : : ; m, and the last r m diagonal blocks are the 1 1 matrices e k t
for k D m C 1; : : : ; r.
This representation of e tA shows that not only may the projections of e tA x0
grow or decay exponentially, they may also exhibit sinusoidally oscillatory behavior.
2.3 Generalized Eigenspace Decomposition

Eigenvalues dont have to be distinct for the analysis of the matrix exponential
that was done last time to work. There just needs to be a basis of eigenvectors for
Rn (or Cn ). Unfortunately, we dont always have such a basis. For this reason,
we need to generalize the notion of an eigenvector.
First, some definitions:
Definition. The algebraic multiplicity of an eigenvalue of an operator A is the
multiplicity of as a zero of the characteristic polynomial det.A xI /.
Definition. The geometric multiplicity of an eigenvalue of an operator A is the
dimension of the corresponding eigenspace, i.e., the dimension of the space of
all the eigenvectors of A corresponding to .
It is not hard to show (e.g., through a change-of-basis argument) that the geometric multiplicity of an eigenvalue is always less than or equal to its algebraic
multiplicity.
Definition. A generalized eigenvector of A is a vector x such that .A I /k x D
0 for some scalar and some k 2 N.
Definition. If is an eigenvalue of A, then the generalized eigenspace of A belonging to is the space of all generalized eigenvectors of A corresponding to
.
Definition. We say a vector space V is the direct sum of subspaces V1 ; : : : ; Vm
31
2. L INEAR S YSTEMS
of V and write
V D V1 Vm
if for each v 2 V there is a unique .v1 ; : : : ; vm / 2 V1 Vm such that
v D v1 C C vm .
Theorem. (Primary Decomposition Theorem) Let B be an operator on E,
where E is a complex vector space, or else E is real and B has real eigenvalues.
Then E is the direct sum of the generalized eigenspaces of B. The dimension of
each generalized eigenspace is the algebraic multiplicity of the corresponding
eigenvalue.
Before proving this theorem, we introduce some notation and state and prove
two lemmas.
Given T W V ! V, let
N.T / D x 2 V T k x D 0 for some k > 0 ;
and let

R.T / D x 2 V T k u D x has a solution u for every k > 0 :
Note that N.T / is the union of the null spaces of the positive powers of T and
R.T / is the intersection of the ranges of the positive powers of T . This union
and intersection are each nested, and that implies that there is a number m 2 N
such that R.T / is the range of T m and N.T / is the nullspace of T m .
Lemma. If T W V ! V, then V D N.T / R.T /.
Proof. Pick m such that R.T / is the range of T m and N.T / is the nullspace
of T m . Note that T jR.T / W R.T / ! R.T / is invertible. Given x 2 V, let
m m
y D T jR.T /
T x and z D x y. Clearly, x D y C z, y 2 R.T /, and
T m z D T m x T m y D 0, so z 2 N.T /. If x D yQ C zQ for some other yQ 2 R.T /
and zQ 2 N.T / then T m yQ D T m x T m zQ D T m x, so yQ D y and zQ D z.
Lemma. If j ; k are distinct eigenvalues of T W V ! V, then
N.T
j I / R.T
Proof. Note first that .T k I /N.T

fact, .T k I /N.T j I / D N.T
.T
32
k I /jN.T
j I /
k I /:
j I / N.T j I /. We claim that, in

j I /; i.e., that
W N.T
j I / ! N.T
j I /
Generalized Eigenspace Decomposition

is invertible. Suppose it isnt; then we can pick a nonzero x 2 N.T j I / such
that .T k I /x D 0. But if x 2 N.T j I / then .T j I /mj x D 0 for some
mj 0. Calculating,
.T
j I /x D T x
2
.T
.T
j I / x D T .k
::
:
j x D k x
j /x
j I /mj x D D .k
j x D .k
j .k
j /x;
j /x D .k
j /2 x;
j /mj x 0;
contrary to assumption. Hence, the claim holds.

Note that this implies not only that
.T
k I /N.T
j I / D N.T
j I /
but also that

.T
k I /m N.T
j I / D N.T
j I /
for every m 2 N. This means that

N.T
j I / R.T
k I /:
Proof of the Principal Decomposition Theorem. The claim is obviously true if

the dimension of E is 0 or 1. We prove it for dim E > 1 by induction on dim E.
Suppose it holds on all spaces of smaller dimension than E. Let 1 ; 2 ; : : : ; q
be the eigenvalues of B with algebraic multiplicities n1 ; n2 ; : : : ; nq . By the first
lemma,
E D N.B q I / R.B q I /:
Note that dim R.B q I / < dim E, and R.B q I / is (positively) invariant
under B. Applying our assumption to BjR.B q I / W R.B q I / ! R.B
q I /, we get a decomposition of R.B q I / into the generalized eigenspaces
of BjR.B q I / . By the second lemma, these are just
N.B
1 I /; N.B
2 I /; : : : ; N.B
q
1 I /;
so
E D N.B
1 I / N.B
2 I / N.B
q
1I /
N.B
q I /:
33
2. L INEAR S YSTEMS
Now, by the second lemma, we know that BjN.B
eigenvalue, so dim N.B k I / nk . Since
q
X
kD1
nk D dim E D
we actually have dim N.B
q
X
dim N.B
k I /
has k as its only
k I /;
kD1
k I / D nk .
2.4 Operators on Generalized Eigenspaces

Weve seen that the space on which a linear operator acts can be decomposed
into the direct sum of generalized eigenspaces of that operator. The operator
maps each of these generalized eigenspaces into itself, and, consequently, solutions of the differential equation starting in a generalized eigenspace stay in that
generalized eigenspace for all time. Now we will see how the solutions within
such a subspace behave by seeing how the operator behaves on this subspace.
It may seem like nothing much can be said in general since, given a finitedimensional vector space V, we can define a nilpotent operator S on V by
1. picking a basis fv1 ; : : : ; vm g for V;
2. creating a graph by connecting the nodes fv1 ; : : : ; vm ; 0g with directed
edges in such a way that from each node there is a unique directed path
to 0;
3. defining S.vj / to be the unique node vk such that there is a directed edge
from vj to vk ;
4. extending S linearly to all of V.
By adding any multiple of I to S we have an operator for which V is a generalized
eigenspace. It turns out, however, that there are really only a small number of
different possible structures that may arise from this seemingly general process.
To make this more precise, we first need a definition, some new notation, and
a lemma.
Definition. A subspace Z of a vector space V is a cyclic subspace of S on V if
SZ Z and there is some x 2 Z such that Z is spanned by fx; Sx; S 2 x; : : :g.
34
Given S, note that every vector x 2 V generates a cyclic subspace. Call

it Z.x/ or Z.x; S/. If S is nilpotent, write nil.x/ or nil.x; S/ for the smallest
nonnegative integer k such that S k x D 0.
Operators on Generalized Eigenspaces

Lemma. The set fx; Sx; : : : ; S nil.x/
1 xg
is a basis for Z.x/.
Proof. Obviously these vectors span Z.x/; the question is whether they are linearly independent. If they were not, we could write down a linear combination
1 S p1 xC Ck S pk x, with j 0 and 0 p1 < p2 < < pk nil.x/ 1,
that added up to zero. Applying S nil.x/ p1 1 to this linear combination would
yield 1 S nil.x/ 1 x D 0, contradicting the definition of nil.x/.
Theorem. If S W V ! V is nilpotent then V can be written as the direct sum of
cyclic subspaces of S on V. The dimensions of these subspaces are determined
by the operator S.
Proof. The proof is inductive on the dimension of V. It is clearly true if dim V D
0 or 1. Assume it is true for all operators on spaces of dimension less than dim V.
Step 1: The dimension of SV is less than the dimension of V.
If this werent the case, then S would be invertible and could not possibly be
nilpotent.
Step 2: For some k 2 N and for some nonzero yj 2 SV, j D 1; : : : ; k,
SV D Z.y1 / Z.yk /:
(2.4)
This is a consequence of Step 1 and the induction hypothesis.

Pick xj 2 V such that Sxj D yj , for j D 1; : : : ; k. Suppose that zj 2 Z.xj /
for each j and
z1 C C zk D 0:
(2.5)
We will show that zj D 0 for each j . This will mean that the direct sum
Z.x1 / Z.xk / exists.
Step 3: Sz1 C C Szk D 0.
This follows from applying S to both sides of (2.5).
Step 4: For each j , Szj 2 Z.yj /.
The fact that zj 2 Z.xj / implies that
zj D 0 xj C 1 Sxj C C nil.xj /
1S
nil.xj / 1
xj
(2.6)
for some i . Applying S to both sides of (2.6) gives

Szj D 0 yj C 1 Syj C C nil.xj /
2S
nil.xj / 2
yj 2 Z.yj /:
35
2. L INEAR S YSTEMS
Step 5: For each j , Szj D 0.

This is a consequence of Step 3, Step 4, and (2.4).
If
Step 6: For each j , zj 2 Z.yj /.

zj D 0 xj C 1 Sxj C C nil.xj /
1S
nil.xj / 1
xj
then by Step 5
0 D Szj D 0 yj C 1 Syj C C nil.xj /
2S
nil.xj / 2
yj :
Since nil.xj / 2 D nil.yj / 1, the vectors in this linear combination are linearly
independent; thus, i D 0 for i D 0; : : : ; nil.xj / 2. In particular, 0 D 0, so
zj D 1 yj C C nil.xj /
1S
nil.xj / 2
yj 2 Z.yj /:
Step 7: For each j , zj D 0.

This is a consequence of Step 6, (2.4), and (2.5).
We now know that Z.x1 / Z.xk / DW VQ exists, but it is not necessarily
all of V. Choose a subspace W of Null.S/ such that Null.S/ D .VQ \ Null.S//
W. Choose a basis fw1 ; : : : ; w` g for W and note that W D Z.w1 / Z.w` /.
Step 8: The direct sum Z.x1 / Z.xk / Z.w1 / Z.w` / exists.
This is a consequence of the fact that the direct sums Z.x1 / Z.xk / and
Z.w1 / Z.w` / exist and that VQ \ W D f0g.
Step 9: V D Z.x1 / Z.xk / Z.w1 / Z.w` /.
Let x 2 V be given. Recall that Sx 2 SV D Z.y1 / Z.yk /. Write
Sx D s1 C C sk with sj 2 Z.yj /. If
nil.yj / 1
sj D 0 yj C 1 Syj C C nil.yj /
1S
uj D 0 xj C 1 Sxj C C nil.yj /
1S
yj ;
let
nil.yj / 1
xj ;
and note that Suj D sj and that uj 2 Z.xj /. Setting u D u1 C C uk , we

have
S.x
36
u/ D Sx
Su D .s1 C C sk /
.s1 C C sk / D 0;
Real Canonical Form

so x
u 2 Null.S/. By definition of W, that means that

x
u 2 Z.x1 / Z.xk / Z.w1 / Z.w` /:
Since u 2 Z.x1 / Z.xk /, we have

x 2 Z.x1 / Z.xk / Z.w1 / Z.w` /:
This completes the proof of the first sentence in the theorem. The second
sentence follows similarly by induction.
2.5 Real Canonical Form

Real Canonical Form
We now use the information contained in the previous theorems to find simple
matrices representing linear operators. Clearly, a nilpotent operator S on a cyclic
space Z.x/ can be represented by the matrix
3
2
0 0
6 ::
:: 7
61
:
:7
7
6
6 :: ::
:: 7 ;
60
:
:
:7
7
6
6 :: : : : : : : :: 7
4:
:
:
: :5
0 0
1 0
with the corresponding basis being fx; Sx; : : : ; S nil.x/ 1 xg. Thus, if is an
eigenvalue of an operator T , then the restriction of T to a cyclic subspace of
T I on the generalized eigenspace N.T I / can be represented by a matrix
of the form
2
3
0 0
6
:: 7
61 : : : : : :
:7
6
7
6
:7
(2.7)
6 0 : : : : : : : : : :: 7 :
7
6
6 :: : : : : : :
7
4:
:
:
: 05
0 0
1
If D a C bi 2 C n R is an eigenvalue of an operator T 2 L.Rn ; Rn /, and

Z.x; T I / is one of the cyclic subspaces whose direct sum is N.T I /, then
Z.x; T I / can be taken to be one of the cyclic subspaces whose direct sum
is N.T I /. If we set k D nil.x; T I / 1 and yj D Re..T I /j x/ and
zj D Im..T I /j x/ for j D 0; : : : ; k, then we get T yj D ayj bzj C yj C1
37
2. L INEAR S YSTEMS
and T zj D byj C azj C zj C1 for j D 0; : : : ; k 1, and T yk D ayk bzk
and T zk D byk C azk . The 2k C 2 real vectors fz0 ; y0 ; : : : ; zk ; yk g span
Z.x; T I / Z.x; T I / over C and also span a .2k C 2/-dimensional
space over R that is invariant under T . On this real vector space, the action of T
can be represented by the matrix
2
a
6 b
6
6
6 1
6
6
6 0
6
6
6 0
6
6
6 0
6
6 ::
6 :
6
6 ::
6 :
6
4 0
0
b
a
0
1
0
0
::
:
::
:
0
0
0
0
::
:
::
:
::
:
::
:
0
0
::
:
::
:
::
:
::
:

::
:
::
:
::
:
::
:

::
:
::
:
::
:
::
:

::
::
::
::
0
0
::
:
::
:
::
:
::
:
::
::
::
::
::
::
::
:
0
1
0
a
b
::
:

::
:

::
:
0
0
::
0
0
:
1
0
::
0
0
::
:
::
:
::
:
::
:
7
7
7
7
7
7
7
7
7
7
7:
7
7
7
7
0 7
7
7
0 7
7
b 5
a
(2.8)
The restriction of an operator to one of its generalized eigenspaces has a

matrix representation like
3
22

5
64 1
6
6
1
6

6

6
6
1
6
6

6
6

4
::
7
7
7
7
7
7
7
7
7
7
7
5
(2.9)
if the eigenvalue is real, with blocks of the form (2.7) running down the diagonal. If the eigenvalue is complex, then the matrix representation is similar to
(2.9) but with blocks of the form (2.8) instead of the form (2.7) on the diagonal.
Finally, the matrix representation of the entire operator is block diagonal,
with blocks of the form (2.9) (or its counterpart for complex eigenvalues). This
is called the real canonical form. If we specify the order in which blocks should
appear, then matrices are similar if and only if they have the same real canonical
form.
38
Solving Linear Systems
2.6 Solving Linear Systems

Exercise 7 Classify all the real canonical forms for operators on R4 . In
other words, find a collection of 4 4 matrices, possibly with (real) variable
entries and possibly with constraints on those variables, such that
1. Only matrices in real canonical form match one of the matrices in
your collection.
2. Each operator on R4 has a matrix representation matching one of the
matrices in your collection.
3. No matrix matching one of your matrices is similar to a matrix matching one of your other matrices.
For example, a suitable collection of matrices for operators on R2 would be:

0
0
a
b
I
I
; .b 0/:
1
0
b a
Computing e tA
Given an operator A 2 L.Rn ; Rn /, let M be its real canonical form. Write
M D S C N , where S has M s diagonal elements k and diagonal blocks

a
b
b a
and 0s elsewhere, and N has M s off-diagonal 1s and 2 2 identity matrices. If
you consider the restrictions of S and N to each of the cyclic subspaces of A I
into which the generalized eigenspace N.A I / of A is decomposed, youll
probably be able to see that these restrictions commute. As a consequence of this
fact (and the way Rn can be represented in terms of these cyclic subspaces), S
and N commute. Thus e tM D e tS e tN .
Now, e tS has e k t where S has k , and has

a t
e k cos bk t
e ak t sin bk t
e ak t sin bk t e ak t cos bk t
where S has
ak
bk

bk
:
ak
39
2. L INEAR S YSTEMS
The series definition can be used to compute e tN , since the fact that N is
nilpotent implies that the series is actually a finite sum. The entries of e tN will
be polynomials in t. For example,
3
2
3
2
0
1
7
6 ::
7
6
61
7
6 t :::
7
:
7
6
7 7! 6
7
6 :: ::
7
6 :: : : : :
4
4 :
:
: 5
:
: 5
tm t 1
1 0
and
2
0
6 0
6
6 1
6
6 0
6
6
6
6
4

0
0
::
0
:
1
::
:
::
:

1 0
0
0 1
0
2
7
7
7
7
7
7 7!
7
7
7
0 5
0

1 0
0 1
t 0
0 t
::
:
6
7
6
7
6
7
:
::
6
7
6
7
6
7:
6
7
::
::
6
7
:
:
6

7
4 t m =m
0
t 0
1 0 5

m
0
t =m
0 t
0 1
Identifying A with its matrix representation with respect to the standard basis, we have A D PMP 1 for some invertible matrix P . Consequently, e tA D
P e tM P 1 . Thus, the entries of e tA will be linear combinations of polynomials
times exponentials or polynomials times exponentials times trigonometric functions.
Exercise 8 Compute e tA (and justify your computations) if
2
3
0 0 0 0
61 0 0 17
7
1. A D 6
41 0 0 15
0
1 1 0
40

2
1
62
2. A D 6
43
4
1
2
3
4
1
2
3
4
3
1
27
7
35
4
Linear Planar Systems

A thorough understanding of constant coefficient linear systems xP D Ax in the
plane is very helpful in understanding systems that are nonlinear and/or higherdimensional.
There are 3 main categories of real canonical forms for an operator A in
L.R2 ; R2 /:

0
0
0
1
a
b

b
;
a
.b 0/
We will subdivide these 3 categories further into a total of 14 categories and

consider the corresponding phase portraits, i.e., sketches of some of the trajectories or parametric curves traced out by solutions in phase space.
u2
1

0
AD
0
. < 0 < /
u1
saddle
41
2. L INEAR S YSTEMS
u2
2
AD

0
0
. < < 0/
u1
stable node
u2

0
AD
0
. D < 0/
u1
stable node
u2

0
AD
0
.0 < < /
unstable node
42
u1

u2
5
AD

0
0
.0 < D /
u1
unstable node
6

0
AD
0
u2
b
. < D 0/
degenerate
7

0
AD
0
u1
u2
b
.0 D < /
degenerate
u1
43
2. L INEAR S YSTEMS
u2
8

0
0
.0 D D /
degenerate
AD
u1
u2

0
AD
1
. < 0/
u1
stable node
10
u2

0
AD
1
.0 < /
unstable node
44
u1

u2
11
AD
0
1
. D 0/
degenerate
u2
12

a
AD
b
u1
b
a
.a < 0 < b/
u1
stable spiral
u2
13

a
AD
b
b
a
.b < 0 < a/
u1
unstable spiral
45
2. L INEAR S YSTEMS
u2
14
AD
a
b
b
a

b
.a D 0; b > 0/
u1
center
If A is not in real canonical form, then the phase portrait should look (topologically) similar but may be rotated, flipped, skewed, and/or stretched.
2.7 Qualitative Behavior of Linear Systems

Parameter Plane
Some of the information from the preceding phase portraits can be summarized
in a parameter diagram. In particular, let D trace A and let D det A, so the
characteristic polynomial is 2 C . Then the behavior of the trivial solution
x.t/ 0 is given by locating the corresponding point in the .; /-plane:
stable node
degenerate
D
unstable node
degenerate
saddle
46
unstable spiral
center
stable spiral
Qualitative Behavior of Linear Systems
Growth and Decay Rates

Given A 2 L.Rn ; Rn /, let
(
)
M
Eu D
N.A I /
>0
8
< M
Re u u 2 N.A
:Re >0
Im 0
E D
N.A
9
>
>
=
I / ;
>
>
;
9 8
>
>
= < M
I /
Im u u 2 N.A
>
>
;
: Re <0
9
>
>
=
I / ;
>
>
;
Im 0
I /
<0
8
< M
Re u u 2 N.A
:Re <0
Im 0
and
E c D N.A/
8
< M
Re u u 2 N.A
:Re D0
Im 0
9 8
>
>
= < M
I /
Im u u 2 N.A
>
>
Re >0
;
:
Im 0
9
>
>
=
< M
I /
Im u u 2 N.A
>
>
;
:Re D0
Im 0
From our previous study of the real canonical form, we know that
I /
9
>
>
=
>
>
;
Rn D E u E s E c :
We call E u the unstable space of A, E s the stable space of A, and E c the center
space of A.
Each of these subspaces of Rn is invariant under the differential equation
xP D Ax:
(2.10)
In other words, if x W R ! Rn is a solution of (2.10) and x.0/ is in E u , E s ,

or E c , then x.t/ is in E u , E s , or E c , respectively, for all t 2 R. We shall see
that each of these spaces is characterized by the growth or decay rates of the
solutions it contains. Before doing so, we state and prove a basic fact about
finite-dimensional normed vector spaces.
Theorem. All norms on Rn are equivalent.
47
2. L INEAR S YSTEMS
Proof. Since equivalence of norms is transitive, it suffices to prove that every
norm N W Rn ! R is equivalent to the standard Euclidean norm j j.
Given an arbitrary norm N , and letting xi be the projection of x 2 Rn onto
the ith standard basis vector ei , note that
!
n
n
n
X
X
X
xi ei
jxi jN.ei /
jxjN.ei /
N.x/ D N

iD1
n
X
iD1
iD1
iD1
N.ei / jxj:
This shows half of equivalence; it also shows that N is continuous, since, by the
triangle inequality,
!
n
X
N.ei / jx yj:
jN.x/ N.y/j N.x y/
iD1

The set S WD x 2 Rn jxj D 1 is clearly closed and bounded and, therefore,
compact, so by the extreme value theorem, N must achieve a minimum on S.
Since N is a norm (and is, therefore, positive definite), this minimum must be
strictly positive; call it k. Then for any x 0,

x
x
D jxjN
kjxj;
N.x/ D N jxj
jxj
jxj
and the estimate N.x/ kjxj obviously holds if x D 0, as well.

Theorem. Given A 2 L.Rn ; Rn / and the corresponding decomposition Rn D
E u E s E c , we have
[

Eu D
x 2 Rn lim je ct e tA xj D 0 ;
(2.11)
c>0
Es D
and
Ec D
c>0
t# 1

c>0
t "1
x 2 Rn lim je ct e tA xj D lim je
t# 1
t "1

e xj D 0 :
ct tA
(2.12)
(2.13)
Proof. By equivalence of norms, instead of using the standard Euclidean norm

on Rn we can use the norm
48
kxk WD supfjP1 xj; : : : ; jPn xjg;
Qualitative Behavior of Linear Systems

where Pi W Rn ! R represents projection onto the ith basis vector corresponding
to the real canonical form. Because of our knowledge of the structure of the real
canonical form, we know that Pi e tA x is either of the form
p.t/e t ;
(2.14)
where p.t/ is a polynomial in t and 2 R is an eigenvalue of A, or of the form

p.t/e at . cos bt C sin bt/;
(2.15)
where p.t/ is a polynomial in t, a C bi 2 C n R is an eigenvalue of A, and and

are real constants. Furthermore, we know that the constant or a is positive if
Pi corresponds to a vector in E u , is negative if Pi corresponds to a vector in E s ,
and is zero if Pi corresponds to a vector in E c .
Now, let x 2 Rn be given. Suppose first that x 2 E s . Then each Pi e tA x is
either identically zero or has as a factor a negative exponential whose constant is
the real part of an eigenvalue of A that is to the left of the imaginary axis in the
complex plane. Let .A/ be the set of eigenvalues of A, and set

max Re 2 .A/ and Re < 0
cD
:
2
Then e ct Pi e tA x is either identically zero or decays exponentially to zero as t "

1.
Conversely, suppose x E s . Then Pi x 0 for some Pi corresponding to a
real canonical basis vector in E u or in E c . In either case, Pi e tA x is not identically
zero and is of the form (2.14) where 0 or of the form (2.15) where a 0.
Thus, if c > 0 then
lim sup je ct Pi e tA xj D 1;
t "1
so
lim sup ke ct e tA xk D 1:
t "1
The preceding two paragraphs showed that (2.12) is correct. By applying

this fact to the time-reversed problem xP D Ax, we find that (2.11) is correct,
as well. We now consider (2.13).
If x 2 E c , then for each i, Pi e tA x is either a polynomial or the product of a
polynomial and a periodic function. If c > 0 and we multiply such a function of
t by e ct and let t # 1 or we multiply it by e ct and let t " 1, then the result
converges to zero.
If, on the other hand, x E c then for some i, Pi e tA x contains a nontrivial
exponential term. If c > 0 is sufficiently small then either e ct Pi e tA x diverges
as t # 1 or e ct Pi e tA x diverges as t " 1. This completes the verification of
(2.13).
49
2. L INEAR S YSTEMS
2.8 Exponential Decay

Definition. If E u D Rn , we say that the origin is a source and e tA is an expansion.
Definition. If E s D Rn , we say that the origin is a sink and e tA is a contraction.
Theorem.
(a) The origin is a source for the equation xP D Ax if and only if for a given
norm k k there are constants k; b > 0 such that
ke tA xk ke t b kxk
for every t 0 and x 2 Rn .
(b) The origin is a sink for the equation xP D Ax if and only if for a given norm
k k there are constants k; b > 0 such that
ke tA xk ke
tb
kxk

Proof. The if parts are a consequence of the previous theorem. The only if
parts follow from the proof of the previous theorem.
Note that a contraction does not always contract things immediately; i.e.,
jxj, in general. For example, consider

1=4
0
AD
:
1
1=4
je tA xj
If

x1 .t/
x.t/ D
x2 .t/
is a solution of xP D Ax, then

1=4
d
0
x1
2
jx.t/j D 2hx; xi
P D 2 x1 x2
1
1=4 x2
dt
1 2
1 2
1
D
x1 C 2x1 x2
x2 D x1 x2
.x1 x2 /2 ;
2
2
2
which is greater than zero if, for example, x1 D x2 > 0. However, we have the
following:
50
Exponential Decay
Theorem.
(a) If e tA is an expansion then there is some norm k k and some constant b > 0
such that
ke tA xk e t b kxk
(b) If e tA is a contraction then there is some norm k k and some constant b > 0
such that
ke tA xk e t b kxk
Proof. The idea of the proof is to pick a basis with respect to which A is represented by a matrix like the real canonical form but with some small constant
" > 0 in place of the off-diagonal 1s. (This can be done by rescaling.) If the
Euclidean norm with respect to this basis is used, the desired estimates hold. The
details of the proof may be found in Chapter 7, 1, of Hirsch and Smale.
Exercise 9
(a) Show that if e tA and e tB are both contractions on Rn , and BA D AB,
then e t .ACB/ is a contraction.
(b) Give a concrete example that shows that (a) can fail if the assumption
that AB D BA is dropped.
Exercise 10 Problem 5 on page 137 of Hirsch and Smale reads:

For any solution to xP D Ax, A 2 L.Rn ; Rn /, show that exactly one of
the following alternatives holds:
(a) lim x.t/ D 0 and lim jx.t/j D 1;
t "1
t# 1
(b) lim jx.t/j D 1 and lim x.t/ D 0;

t "1
t# 1
(c) there exist constants M; N > 0 such that M < jx.t/j < N for all
t 2 R.
51
2. L INEAR S YSTEMS
Is what they ask you to prove true? If so, prove it. If not, determine what
other possible alternatives exist, and prove that you have accounted for all
possibilities.
2.9 Nonautonomous Linear Systems

We now move from the constant coefficient equation xP D Ax to the nonautonomous equation
xP D A.t/x:
(2.16)
For simplicity we will assume that the domain of A is R.
Solution Formulas
In the scalar, or one-dimensional, version of (2.16)
xP D a.t/x
(2.17)
we can separate variables and arrive at the formula

x.t/ D x0 e
Rt
t0
a. / d
for the solution of (2.17) that satisfies the initial condition x.t0 / D x0 .
It seems like the analogous formula for the solution of (2.16) with initial
condition x.t0 / D x0 should be
x.t/ D e
Rt
t0
A. / d
x0 :
(2.18)
Certainly, the right-hand side of (2.18) makes sense (assuming that A is continuous). But does it give the correct answer?
Lets consider a specific example. Let

0 0
A.t/ D
1 t
and t0 D 0. Note that
Z
52

t2 0
0
0
A./ d D
D
t t 2 =2
2 2=t

0
;
1
Nonautonomous Linear Systems

and
2
0
0 5
4
t t 2 =2

1
D
0

1
D
0
2 2
t

2
t
2
0
0 0
0 0
C
C
C
1
2=t 1
2 2=t 1
2
"
#

0 0
1
0
2
0

2
C e t =2 1
D 2 t 2 =2
:
1
2=t 1
e
1
e t =2
t
On the other hand, we can solve the corresponding system

xP 1 D 0
xP 2 D x1 C tx2
directly. Clearly x1 .t/ D for some constant . Plugging this into the equation
for x2 , we have a first-order scalar equation which can be solved by finding an
integrating factor. This yields
Z t
2
t 2 =2
t 2 =2
x2 .t/ D e
C e
e s =2 ds
0
for some constant . Since x1 .0/ D and x2 .0/ D , the solution of (2.16) is

Since

1
x1 .t/
D t 2=2 R t
x2 .t/
e
0 e
e
t 2 =2
s 2 =2

x1 .0/
:
s 2 =2 ds e t 2 =2
x2 .0/
0
ds
2 t 2 =2
e
t

1
(2.18) doesnt work? What went wrong? The answer is that

d R t A. / d
e
e0
D lim
dt
h!0
e
lim
h!0
R tCh
0
Rt
0
A. / d
A. / d
h
h
Rt
R tCh
t
A. / d
A. / d
in general, because of possible noncommutativity.
Structure of Solution Set

We abandon attempts to find a general formula for solving (2.16), and instead
analyze the general structure of the solution set.
53
2. L INEAR S YSTEMS
Definition. If x .1/; x .2/ ; : : : ; x .n/ are linearly independent solutions of (2.16)
(i.e., no nontrivial linear combination gives the zero function) then the matrix

X.t/ WD x .1/ .t/ x .n/ .t/
is called a fundamental matrix for (2.16).
Theorem. The dimension of the vector space of solutions of (2.16) is n.

Proof. Pick n linearly independent vectors v .k/ 2 Rn , k D 1; : : : ; n, and let
x .k/ be the solution of (2.16) that satisfies the initial condition x .k/ .0/ D v .k/ .
Then these n solutions are linearly independent. Furthermore, we claim that
any solution x of (2.16) is a linear combination of these n solutions. To see
why this is so, note that x.0/ must be expressible as a linear combination of
fv .1/ ; : : : ; v .n/ g. The corresponding linear combination of fx .1/ ; : : : ; x .n/ g is, by
linearity, a solution of (2.16) that agrees with x at t D 0. Since A is continuous,
the Picard-Lindelof Theorem applies to (2.16) to tell us that solutions of IVPs are
unique, so this linear combination of fx .1/ ; : : : ; x .n/ g must be identical to x.
Definition. If X.t/ is a fundamental matrix and X.0/ D I , then it is called the
principal fundamental matrix. (Uniqueness of solutions implies that there is only
one such matrix.)
Definition. Given n functions (in some order) from R to Rn , their Wronskian
is the determinant of the matrix that has these functions as its columns (in the
corresponding order).
Theorem. The Wronskian of n solutions of (2.16) is identically zero if and only
if the solutions are linearly dependent.
Proof. Suppose x .1/ ; : : : ; x .n/ are linearly dependent solutions; i.e.,
n
X
kD1
54
k x .k/ D 0
P
P
for some constants 1 ; : : : ; n with nkD1 k2 0. Then nkD1 k x .k/ .t/ is 0
for every t, so the columns of the Wronskian W .t/ are linearly dependent for
every t. This means W 0.
Conversely, suppose that the Wronskian W of n solutions x .1/ ; : : : ; x .n/ is
identically zero. In particular, W .0/ D 0, so x .1/P
.0/; : : : ; x .n/ .0/ are linearly
dependent vectors. Pick constants 1 ; : : : ; n , with nkD1 k2 nonzero, such that
Nonautonomous Linear Systems

Pn
P
k x .k/ .0/ D 0. The function nkD1 k x .k/ is a solution of (2.16) that is
0 when t D
is the function that is identically zero. By uniqueness of
Pn0, but so.k/
solutions, kD1 k x D 0; i.e., x .1/ ; : : : ; x .n/ are linearly dependent.
kD1
Note that this proof also shows that if the Wronskian of n solutions of (2.16)
is zero for some t, then it is zero for all t.
What if were dealing with n arbitrary vector-valued functions (that are not
necessarily solutions of (2.16))? If they are linearly dependent then their Wronskian is identically zero, but the converse is not true. For example,

1
t
and
0
0
have a Wronskian that is identically zero, but they are not linearly dependent.
Also, n functions can have a Wronskian that is zero for some t and nonzero for
other t. Consider, for example,

1
0
and
:
0
t
Initial-Value Problems
Given a fundamental matrix X.t/ for (2.16), define G.t; t0 / to be the quantity
X.t/X.t0 / 1 . We claim that x.t/ WD G.t; t0 /v solves the IVP
(
xP D A.t/x
x.t0 / D v:
To verify this, note that
d
d
xD
.X.t/X.t0 /
dt
dt
v/ D A.t/X.t/X.t0 /
v D A.t/x;
and
x.t0 / D G.t0 ; t0 /v D X.t0 /X.t0 /
v D v:
Inhomogeneous Equations
Consider the IVP
xP D A.t/x C f .t/
x.t0 / D x0 :
(2.19)
In light of the results from the previous section when f was identically zero, its
reasonable to look for a solution x of (2.19) of the form x.t/ WD G.t; t0 /y.t/,
where G is as before, and y is some vector-valued function.
55
2. L INEAR S YSTEMS
Note that
x.t/
P
D A.t/X.t/X.t0 /
y.t/ C G.t; t0 /y.t/

P D A.t/x.t/ C G.t; t0 /y.t/I
P
therefore, we need G.t; t0 /y.t/

P D f .t/. Isolating, y.t/,
P
we need
1
y.t/
P D X.t0 /X.t/
f .t/ D G.t0 ; t/f .t/:
(2.20)
Integrating both sides of (2.20), we see that y should satisfy

Z t
y.t/ y.t0 / D
G.t0 ; s/f .s/ ds:
t0
If x.t0 / is to be x0 , then, since G.t0 ; t0 / D I , we need y.t0 / D x0 , so y.t/

should be
Z t
x0 C
G.t0 ; s/f .s/ ds;
t0
or, equivalently, x.t/ should be

G.t; t0 /x0 C
G.t; s/f .s/ ds;

t0
since G.t; t0 /G.t0 ; s/ D G.t; s/. This is called the Variation of Constants formula or the Variation of Parameters formula.
2.10 Nearly Autonomous Linear Systems

Suppose A.t/ is, in some sense, close to a constant matrix A. The question we
wish to address in this section is the extent to which solutions of the nonautonomous system
xP D A.t/x
(2.21)
behave like solutions of the autonomous system
xP D Ax:
(2.22)
Before getting to our main results, we present a pair of lemmas.

Lemma. The following are equivalent:
1. Each solution of (2.22) is bounded as t " 1.
2. The function t 7! ke tA k is bounded as t " 1 (where k k is the usual
operator norm).
56
Nearly Autonomous Linear Systems

3. Re 0 for every eigenvalue of A and the algebraic multiplicity of each
purely imaginary eigenvalue matches its geometric multiplicity.
Proof. That statement 2 implies statement 1 is a consequence of the definition of
the operator norm, since, for each solution x of (2.22),
jx.t/j D je tA x.0/j ke tA k jx.0/j:
That statement 1 implies statement 3, and statement 3 implies statement 2 are
consequences of what we have learned about the real canonical form of A, along
with the equivalence of norms on Rn .
Lemma. (Generalized Gronwall Inequality) Suppose X and are nonnegative, continuous, real-valued functions on t0 ; T for which there is a nonnegative
constant C such that
Z t
X.t/ C C
.s/X.s/ ds;
t0
for every t 2 t0 ; T . Then

X.t/ C e
Rt
t0
.s/ ds
Proof. The proof is very similar to the proof of the standard Gronwall inequality.
The details are left to the reader.
The first main result deals with the case when A.t/ converges to A sufficiently
quickly as t " 1.
Theorem. Suppose that each solution of (2.22) remains bounded as t " 1 and
that, for some t0 2 R,
Z 1
kA.t/ Ak dt < 1;
(2.23)
t0
where k k is the standard operator norm. Then each solution of (2.21) remains
bounded as t " 1.
Proof. Let t0 be such that (2.23) holds. Given a solution x of (2.21), let f .t/ D
.A.t/ A/x.t/, and note that x satisfies the constant-coefficient inhomogeneous
problem
xP D Ax C f .t/:
(2.24)
57
2. L INEAR S YSTEMS
Since the matrix exponential provides a fundamental matrix solution to constantcoefficient linear systems, applying the variation of constants formula to (2.24)
yields
Z
x.t/ D e .t
t0 /A
x.t0 / C
e .t
s/A
.A.s/
A/x.s/ ds:
(2.25)
t0
Now, by the first lemma, the boundedness of solutions of (2.22) in forward

time tells us that there is a constant M > 0 such that ke tA k M for every
t t0 . Taking norms and estimating, gives (for t t0 )
Z t
.t t0 /A
jx.t/j ke
k jx.t0 /j C
ke .t s/A k kA.s/ Ak jx.s/j ds
t0
M jx.t0 /j C
t0
M kA.s/
Ak jx.s/j ds:
Setting X.t/ D jx.t/j, .t/ D M kA.t/ Ak, and C D M jx.t0 /j, and
applying the generalized Gronwall inequality, we find that
jx.t/j M jx.t0 /je
Rt
t0
kA.s/ Ak ds
By (2.23), the right-hand side of this inequality is bounded on t0 ; 1/, so x.t/ is

bounded as t " 1.
The next result deals with the case when the origin is a sink for (2.22). Will
the solutions of (2.21) also all converge to the origin as t " 1? Yes, if kA.t/ Ak
is sufficiently small.
Theorem. Suppose all the eigenvalues of A have negative real part. Then there
is a constant " > 0 such that if kA.t/ Ak " for all t sufficiently large then
every solution of (2.21) converges to 0 as t " 1.
Proof. Since the origin is a sink, we know that we can choose constants k; b > 0
such that ke tA k ke bt for all t 0. Pick a constant " 2 .0; b=k/, and assume
that there is a time t0 2 R such that kA.t/ Ak " for every t t0 .
Now, given a solution x of (2.21) we can conclude, as in the proof of the
previous theorem, that
Z t
.t t0 /A
jx.t/j ke
k jx.t0 /j C
ke .t s/A k kA.s/ Ak jx.s/j ds
t0
for all t t0 . This implies that

jx.t/j ke
58
b.t t0 /
jx.t0 /j C
ke
t0
b.t s/
"jx.s/j ds
Periodic Linear Systems

for all t t0 . Multiplying through by e b.t
yield
t0 /
y.t/ kjx.t0 /j C k"
and setting y.t/ WD e b.t
t0 / jx.t/j
y.s/ ds
t0
for all t t0 . The standard Gronwall inequality applied to this estimate gives
y.t/ kjx.t0 /je k".t
t0 /
for all t t0 , or, equivalently,

jx.t/j kjx.t0 /je .k"
b/.t t0 /
for all t t0 . Since " < b=k, this inequality implies that x.t/ ! 0 as t "
1.
Thus, the origin remains a sink even when we perturb A by a small timedependent quantity. Can we perhaps just look at the (possibly, time-dependent)
eigenvalues of A.t/ itself and conclude, for example, that if all of those eigenvalues have negative real part for all t then all solutions of (2.21) converge to the
origin as t " 1? The following example of Markus and Yamabe shows that the
answer is No.
Exercise 11 Show that if

1 C 32 cos2 t
A.t/ D
1 32 sin t cos t
3
2
sin t cos t
1 C 32 sin2 t
then the eigenvalues of A.t/ both have negative real part for every t 2 R,
but

cos t t =2
x.t/ WD
e ;
sin t
which becomes unbounded as t ! 1, is a solution to (2.21).
2.11 Periodic Linear Systems

We now consider
xP D A.t/x
(2.26)
59
2. L INEAR S YSTEMS
when A is a continuous periodic n n matrix function of t; i.e., when there is a
constant T > 0 such that A.t C T / D A.t/ for every t 2 R. When that condition
is satisfied, we say, more precisely, that A is T -periodic. If T is the smallest
positive number for which this condition holds, we say that T is the minimal
period of A. (Every continuous, nonconstant periodic function has a minimal
period).
Let A be T -periodic, and let X.t/ be a fundamental matrix for (2.26). Define
Q
X W R ! L.Rn ; Rn / by XQ .t/ D X.t C T /. Clearly, the columns of XQ are
linearly independent functions of t. Also,
d Q
d
Q
X .t/ D
X.t C T / D X 0 .t C T / D A.t C T /X.t C T / D A.t/X.t/;
dt
dt
so XQ solves the matrix equivalent of (2.26). Hence, XQ is a fundamental matrix
for (2.26).
Because the dimension of the solution space of (2.26) is n, this means that
there is a nonsingular (constant) matrix C such that X.t CT / D X.t/C for every
t 2 R. C is called a monodromy matrix.
Lemma. There exists B 2 L.Cn ; Cn / such that C D e TB .
Proof. Without loss of generality, we assume that T D 1, since if it isnt we can
just rescale B by a scalar constant. We also assume, without loss of generality,
that C is in Jordan canonical form. (If it isnt, then use the fact that P 1 CP D
1
e B implies that C D e PBP .) Furthermore, because of the way the matrix
exponential acts on a block diagonal matrix, it suffices to show that for each
p p Jordan block
3
2
0 0
6
:: 7
61 : : : : : :
:7
6
7
6
7
:
:
:
CQ WD 6 0 : : : : : : ::: 7 ;
6
7
7
6 :: : : : : : :
4:
:
:
: 05
0 0
1
Q
CQ D e B for some BQ 2 L.Cp ; Cp /.
Now, an obvious candidate for BQ is the natural logarithm of CQ , defined in
some reasonable way. Since the matrix exponential was defined by a power series, it seems reasonable to use a similar definition for a matrix logarithm. Note
that CQ D I CN D I.I C 1 N /, where N is nilpotent. (Since C is invertible,
we know that all of the eigenvalues are nonzero.) We guess
60
BQ D .log /I C log.I C
N /;
(2.27)

where
log.I C M / WD
1
X
. M /k
;
k
kD1
in analogy to the Maclaurin series for log.1 C x/. Since N is nilpotent, this
series terminates in our application of it to (2.27). Direct substitution shows that
Q
e B D CQ , as desired.
The eigenvalues of C are called the Floquet multipliers (or characteristic
multipliers) of (2.26). The corresponding numbers satisfying D e T are
called the Floquet exponents (or characteristic exponents) of (2.26). Note that
the Floquet exponents are only determined up to a multiple of .2 i/=T . Given
B for which C D e TB , the exponents can be chosen to be the eigenvalues of B.
Theorem. There exists a T -periodic function P W R ! L.Rn ; Rn / such that
X.t/ D P .t/e tB :
Proof. Let P .t/ D X.t/e
P .t C T / D X.t C T /e
D X.t/e TB e
tB .
Then
.t CT /B
TB
tB
D X.t C T /e
D X.t/e
tB
TB
tB
D X.t/C e
TB
tB
D P .t/:
The decomposition of X.t/ given in this theorem shows that the behavior of
solutions can be broken down into the composition of a part that is periodic in
time and a part that is exponential in time. Recall, however, that B may have
entries that are not real numbers, so P .t/ may be complex, also. If we want
to decompose X.t/ into a real periodic matrix times a matrix of the form e tB
where B is real, we observe that X.t C 2T / D X.t/C 2 , where C is the same
monodromy matrix as before. It can be shown that the square of a real matrix
can be written as the exponential of a real matrix. Write C 2 D e TB with B real,
and let P .t/ D X.t/e tB as before. Then, X.t/ D P .t/e tB where P is now
2T -periodic, and everything is real.
The Floquet multipliers and exponents do not depend on the particular fundamental matrix chosen, even though the monodromy matrix does. They depend
only on A.t/. To see this, let X.t/ and Y.t/ be fundamental matrices with corresponding monodromy matrices C and D. Because X.t/ and Y.t/ are fundamental matrices, there is a nonsingular constant matrix S such that Y.t/ D X.t/S for
all t 2 R. In particular, Y.0/ D X.0/S and Y.T / D X.T /S. Thus, C D
X.0/
X.T / D SY.0/
Y.T /S
D SY.0/
Y.0/DS
D SDS
:
61
2. L INEAR S YSTEMS
This means that the monodromy matrices are similar and, therefore, have the
same eigenvalues.
Interpreting Floquet Multipliers and Exponents

Theorem. If is a Floquet multiplier of (2.26) and is a corresponding Floquet
exponent, then there is a nontrivial solution x of (2.26) such that x.t C T / D
x.t/ for every t 2 R and x.t/ D e t p.t/ for some T -periodic vector function
p.
Proof. Pick x0 to be an eigenvector of B corresponding to the eigenvalue ,
where X.t/ D P .t/e tB is the decomposition of a fundamental matrix X.t/. Let
x.t/ D X.t/x0 . Then, clearly, x solves (2.26). The power series formula for the
matrix exponential implies that x0 is an eigenvector of e tB with eigenvalue e t .
Hence,
x.t/ D X.t/x0 D P .t/e tB x0 D P .t/e t x0 D e t p.t/;
where p.t/ D P .t/x0 . Also,
x.t C T / D e T e t p.t C T / D e t p.t/ D x.t/:
Time-dependent Change of Variables

Let x solve (2.26), and let y.t/ D P .t/
ously. Then
1 x.t/,
where P is as defined previ-
d
d
P .t/y.t/ D
x.t/ D A.t/x.t/ D A.t/P .t/y.t/
dt
dt
D A.t/X.t/e tB y.t/:
But
d
P .t/y.t/ D P 0 .t/y.t/ C P .t/y 0 .t/
dt
D X 0 .t/e tB X.t/e tB By.t/ C X.t/e
D A.t/X.t/e
tB
y.t/
X.t/e
tB
tB 0
y .t/
By.t/ C X.t/e
tB 0
y .t/;
so
X.t/e
tB 0
y .t/ D X.t/e
tB
By.t/;
which implies that y 0 .t/ D By.t/; i.e., y solves a constant coefficient linear
equation. Since P is periodic and, therefore, bounded, the growth and decay of
62

x and y are closely related. Furthermore, the growth or decay of y is determined
by the eigenvalues of B, i.e., by the Floquet exponents of (2.26). For example,
we have the following results.
Theorem. If all the Floquet exponents of (2.26) have negative real parts then all
solutions of (2.26) converge to 0 as t " 1.
Theorem. If there is a nontrivial T -periodic solution of (2.26) then there must
be a Floquet multiplier of modulus 1.
Computing Floquet Multipliers and Exponents

Although Floquet multipliers and exponents are determined by A.t/, it is not
obvious how to calculate them. As a previous exercise illustrated, the eigenvalues
of A.t/ dont seem to be extremely relevant. The following result helps a little
bit.
Theorem. If (2.26) has Floquet multipliers 1 ; : : : ; n and corresponding Floquet exponents 1 ; : : : ; n , then
Z
1 n D exp
trace A.t/ dt
(2.28)
and
1
1 C C n
T
trace A.t/ dt mod
2 i
T
(2.29)
Proof. We focus on (2.28). The formula (2.29) will follow immediately from
(2.28).
Let W .t/ be the determinant of the principal fundamental matrix X.t/. Let
Sn be the set of permutations of f1; 2; : : : ; ng and let W Sn ! f 1; 1g be the
parity map. Then
W .t/ D
. /
2Sn
n
Y
Xi;.i/ ;
iD1
where Xi;j is the .i; j /-th entry of X.t/.

63
2. L INEAR S YSTEMS
Differentiating yields
n
X
d W .t/
d Y
. /
Xi;.i/
D
dt
dt
iD1
2Sn

Y
n X
X
d
D
. /
Xj;.j /
Xi;.i/
dt
j D1 2Sn
ij
" n
#
n X
X
X
Y
D
. /
Aj;k .t/Xk;.j /
Xi;.i/
j D1 2Sn
n X
n
X
j D1 kD1
kD1
Aj;k .t/ 4
ij
. /Xk;.j /
2Sn
ij
Xi;.i/ 5 :
If j k, the inner sum is the determinant of the matrix obtained by replacing

the j th row of X.t/ by its kth row. This new matrix, having two identical rows,
must necessarily have determinant 0. Hence,
n
X
d W .t/
D
Aj;j .t/ det X.t/ D trace A.t/W .t/:
dt
j D1
Thus,
W .t/ D e
Rt
0
trace A.s/ ds
W .0/ D e
Rt
0
trace A.s/ ds
In particular,
e
RT
0
trace A.s/ ds
D W .T / D det X.T / D det.P .T /e TB / D det.P .0/e TB /

D det e TB D det C D 1 2 n :
Exercise 12 Consider (2.26) where

1
cos t
A.t/ D 2
a
3
2
b
C sin t
and a and b are constants. Show that there is a solution of (2.26) that becomes unbounded as t " 1.
64
3
Topological Dynamics
3.1 Invariant Sets and Limit Sets
We will now begin a study of the continuously differentiable autonomous system
xP D f .x/
or, equivalently, of the corresponding dynamical system '.t; x/. We will denote
the phase space and assume that it is an open (not necessarily proper) subset
of Rn .
Orbits
Definition. Given x 2 , the (complete) orbit through x is the set

.x/ WD '.t; x/ t 2 R ;
the positive semiorbit through x is the set
C .x/ WD '.t; x/ t 0g;
and the negative semiorbit through x is the set
.x/ WD '.t; x/ t 0g:
Invariant Sets
Definition. A set M is invariant under ' if it contains the complete orbit

of every point of M. In other words, for every x 2 M and every t 2 R,
'.t; x/ 2 M.
Definition. A set M is positively invariant under ' if it contains the positive
semiorbit of every point of M. In other words, for every x 2 M and every t 0,
'.t; x/ 2 M.
65
3. T OPOLOGICAL DYNAMICS
Definition. A set M is negatively invariant under ' if it contains the negative semiorbit of every point of M. In other words, for every x 2 M and every
t 0, '.t; x/ 2 M.
Limit Sets
Definition. Given x 2 , the !-limit set of x, denoted !.x/, is the set

y 2 lim inf j'.t; x/ yj D 0
t "1
D y 2 9t1 ; t2 ; : : : ! 1 s.t. '.tk ; x/ ! y as k " 1 :
Definition. Given x 2 , the -limit set of x, denoted .x/, is the set

y 2 lim inf j'.t; x/ yj D 0
t# 1
D y 2 9t1 ; t2 ; : : : ! 1 s.t. '.tk ; x/ ! y as k " 1 :
Lemma. If, for each A 2 , we let A represent the topological closure of A in

, then
\
!.x/ D
C .'.; x//
(3.1)
2R
and
.x/ D
.'.; x//:
(3.2)
2R
Proof. It suffices to prove (3.1); (3.2) can then be established by time reversal.
Let y 2 !.x/ be given. Pick a sequence t1 ; t2 ; : : : ! 1 such that '.tk ; x/ !
y as k " 1. Let 2 R be given. Pick K 2 N such that tk for all k K.
Note that '.tk ; x/ 2 C .'.; x// for all k K, so
y 2 C .'.; x//:
Since this holds for all 2 R, we know that
\
y2
C .'.; x//:
(3.3)
2R
Since (3.3) holds for each y 2 !.x/, we know that

!.x/
66
2R
C .'.; x//:
(3.4)
Invariant Sets and Limit Sets

Now, we prove the reverse inclusion. Let
\
y2
C .'.; x//
2R
be given. This implies, in particular, that

\
y2
C .'.; x//:
2N
For each k 2 N, we have
y 2 C .'.k; x//
so we can pick zk 2 C .'.k; x// such that jzk yj < 1=k. Since zk 2
C .'.k; x//, we can pick sk 0 such that zk D '.sk ; '.k; x//. If we set
tk D k C sk , we see that tk k, so the sequence t1 ; t2 ; : : : goes to infinity. Also,
since
j'.tk ; x/
yj D j'.sk C k; x/
< 1=k;
yj D j'.sk ; '.k; x//
yj D jzk
yj
we know that '.tk ; x/ ! y as k " 1. Hence, y 2 !.x/. Since this holds for
every
\
C .'.; x//;
y2
2R
we know that
2R
C .'.; x// !.x/:
Combining this with (3.4) gives (3.1).

We now describe some properties of limit sets.
Theorem. Given x 2 , !.x/ and .x/ are closed (relative to ) and invariant.
If C .x/ is contained in some compact subset of , then !.x/ is nonempty,
compact, and connected. If .x/ is contained in some compact subset of ,
then .x/ is nonempty, compact, and connected.
Proof. Again, time-reversal arguments tell us that it is only necessary to prove
the statements about !.x/.
Step 1: !.x/ is closed.
This is a consequence of the lemma and the fact that the intersection of closed
sets is closed.
67
Step 2: !.x/ is invariant.
Let y 2 !.x/ and t 2 R be given. Choose a sequence of times .tk / converging
to infinity such that '.tk ; x/ ! y as k " 1. For each k 2 N, let sk D tk C t,
and note that .sk / converges to infinity and
'.sk ; x/ D '.tk C t; x/ D '.t; '.tk ; x// ! '.t; y/
as k " 1 (by the continuity of '.t; /). Hence, '.t; y/ 2 !.x/. Since t 2 R and
y 2 !.x/ were arbitrary, we know that !.x/ is invariant.
Now, suppose that C .x/ is contained in a compact subset K of .
Step 3: !.x/ is nonempty.
The sequence '.1; x/; '.2; x/; : : : is contained in C .x/ K, so by the BolzanoWeierstrass Theorem, some subseqence '.t1 ; x/; '.t2 ; x/; : : : converges to some
y 2 K. By definition, y 2 !.x/.
Step 4: !.x/ is compact.
By the Heine-Borel Theorem, K is closed (relative to Rn ), so, by the choice of
K, !.x/ K. Since, by Step 1, !.x/ is closed relative to , it is also closed
relative to K. Since K is compact, this means !.x/ is closed (relative to Rn ).
Also, by the Heine-Borel Theorem, K is bounded so its subset !.x/ is bounded,
too. Thus, !.x/ is closed (relative to Rn ) and bounded and, therefore, compact.
Step 5: !.x/ is connected.
Suppose !.x/ were disconnected. Then there would be disjoint open subsets G
and H of such that G\!.x/ and H\!.x/ are nonempty, and !.x/ is contained
in G[H. Then there would have to be a sequence s1 ; s2 ; : : : ! 1 and a sequence
t1 ; t2 ; : : : ! 1 such that '.sk ; x/ 2 G, '.tk ; x/ 2 H, and sk < tk < skC1 for
each k 2 N. Because (for each fixed k 2 N)

'.t; x/ t 2 sk ; tk
is a (connected) curve going from a point in G to a point in H, there must be a

time k 2 .sk ; tk / such that '.k ; x/ 2 K n .G [ H/. Pick such a k for each
k 2 N and note that 1 ; 2 ; : : : ! 1 and, by the Bolzano-Weierstrass Theorem,
some subsequence of .'.k ; x// must converge to a point y in K n .G [ H/. Note
that y, being outside of G [ H, cannot be in !.x/, which is a contradiction.
68
Examples of empty !-limit sets are easy to find. Consider, for example, the
one-dimensional dynamical system '.t; x/ WD xCt (generated by the differential
equation xP D 1.
Regular and Singular Points

Examples of dynamical systems with nonempty, noncompact, disconnected
!-limit sets are a little harder to find. Consider the planar autonomous system
(
xP D y.1 x 2 /
yP D x C y.1 x 2 /:
After rescaling time, this differential equation generates a dynamical system on
R2 with

!.x/ D . 1; y/ y 2 R [ .1; y/ y 2 R
for every x in the punctured strip

.x; y/ 2 R2 jxj < 1 and x 2 C y 2 > 0 :
3.2 Regular and Singular Points
Consider the differential equation xP D f .x/ and its associated dynamical system
'.t; x/ on a phase space .
Definition. We say that a point x 2 is an equilibrium point or a singular point
or a critical point if f .x/ D 0. For such a point, '.t; x/ D x for all t 2 R.
Definition. A point x 2 that is not a singular point is called a regular point.
We shall show that all of the interesting local behavior of a continuous dynamical system takes place close to singular points. We shall do this by showing
that in the neighborhood of each regular point, the flow is very similar to unidirectional, constant-velocity flow.
One way of making the notion of similarity of flows precise is the following.
Definition. Two dynamical systems ' W R ! and
W R !
are topologically conjugate if there exists a homeomorphism (i.e., a continuous
bijection with continuous inverse) h W ! such that
h.'.t; x// D
.t; h.x//
for every t 2 R and every x 2 . In other words,

equivalently, the diagram

?
?
hy
(3.5)
.t; / D h '.t; / h
1,
or,
'.t;/
!
?
?
yh
.t;/
69
commutes for each t 2 R. The function h is called a topological conjugacy. If,
in addition, h and h 1 are r-times continuously differentiable, we say that ' and
are C r -conjugate.
A weaker type of similarity is the following.
Definition. Two dynamical systems ' W R ! and
W R !
are topologically equivalent if there exists a homeomorphism h W ! and
a time reparametrization function W R ! R such that, for each x 2 ,
.; x/ W R ! R is an increasing surjection and
h.'..t; x/; x// D
.t; h.x//
for every t 2 R and every x 2 . If, in addition, h is r-times continuously

differentiable, we say that ' and are C r -equivalent.
A topological equivalence maps orbits to orbits and preserves the orientation
of time but may reparametrize time differently on each individual orbit.
As an example of the difference between these two concepts, consider the
two planar dynamical systems

cos t
sin t
'.t; x/ D
x
sin t cos t
and
cos 2t
.t; y/ D
sin 2t

sin 2t
y;
cos 2t
generated, respectively, by the differential equations

0
1
xP D
x
1 0
and
yP D
0
2

2
y:
0
The functions h.x/ D x and .t; x/ D 2t show that these two flows are topologically equivalent. But these two flows are not topologically conjugate, since, by
setting t D we see that any function h W R2 ! R2 satisfying (3.5) would have
to satisfy h.x/ D h. x/ for all x, which would mean that h is not invertible.
Because of examples like this, topological equivalence seems to be the preferred concept when dealing with flows. The following theorem, however, shows
that in a neighborhood of a regular point, a smooth flow satisfies a local version
of C r -conjugacy with respect to a unidirectional, constant-velocity flow.
70
Regular and Singular Points

Theorem. (C r Rectification Theorem) Suppose f W ! Rn is r-times continuously differentiable (with r 1) and x0 is a regular point of the flow generated
by
xP D f .x/:
(3.6)
Then there is a neighborhood V of x0 , a neighborhood W of the origin in Rn ,

and a C r map g W V ! W witha C r inverse such that, for each solution x of
(3.6) in V, y.t/ WD g.x.t// satisfies the equation
2 3
1
607
6 7
yP D 6 : 7
(3.7)
4 :: 5
0
in W.
Proof. Without loss of generality, we shall assume that x0 D 0 and f .x0 / D

f .0/ D e1 for some > 0. Let W be a small ball centered at 0 in Rn ,
and define G.y/ WD G..y1 ; : : : ; yn /T / D '.y1 ; .0; y2 ; : : : ; yn /T /, where ' is
the flow generated by (3.6). (While ' might not be a genuine dynamical system
because it might not be defined for all time, we know that it is at least defined long
enough that G is well-defined if W is sufficiently small.) In words, G.y/ is the
solution obtained by projecting y onto the plane through the origin perpendicular
to f .0/ and locating the solution of (3.6) that starts at this projected point after
y1 units of time have elapsed.
Step 1: '.; p/ is C rC1 .
We know that
d
'.t; p/ D f .'.t; p//:
(3.8)
dt
If f is continuous then, since '.; p/ is continuous, (3.8) implies that '.; p/ is
C 1 . If f is C 1 , then the previous observation implies that '.; p/ is C 1 . Then
d
(3.8) implies that dt
'.t; p/ is the composition of C 1 functions and is, therefore,
C 1 ; this means that '.; p/ is C 2 . Continuing inductively, we see that, since f
is C r , '.; p/ is C rC1 .
Step 2: '.t; / is C r .
This is a consequence of applying differentiability with respect to parameters inductively.
Step 3: G is C r .
This is a consequence of Steps 1 and 2 and the formula for G in terms of '.
71
Step 4: DG.0/ is an invertible matrix.
Since
@G.y/
@
D
'.t;
0/
D f .0/ D e1
@y1 yD0
@t
t D0
and
@G.y/
@
@p
D
'.0; p/
ek D
ek D ek ;
@yk yD0
@p
@p pD0
pD0
for k 1, we have
DG.0/ D 4 e1 e2 en 5 ;
which is invertible since 0.
Step 5: If W is sufficiently small, then G is invertible.

This is a consequence of Step 4 and the Inverse Function Theorem.
Set g equal to the (locally defined) inverse of G. Since G is C r , so is g.
The only thing remaining to check is that if x satisfies (3.6) then g x satisfies
(3.7). Equivalently, we can check that if y satisfies (3.7) then G y satisfies (3.6).
Step 6: If y satisfies (3.7) then G y satisfies (3.6).
By the chain rule,
@
d
G.y.t// D
'.s; .0; y2 ; : : : ; yn //
yP1
dt
@s
sDy1
2 3
0
6
7
y
P
@
6 27
C
'.y1 ; p/
6 : 7
:
@p
pD.0;y2 ;:::;yn / 4 : 5
yPn
D f .'.y1 ; .0; y2 ; : : : ; yn /// D f .G.y//:
3.3 Definitions of Stability

In the previous section, we saw that all the interesting local behavior of flows
occurs near equilibrium points. One important aspect of the behavior of flows
72
Definitions of Stability
has to do with whether solutions that start near a given solution stay near it for all
time and/or move closer to it as time elapses. This question, which is the subject
of stability theory, is not just of interest when the given solution corresponds to
an equilibrium solution, so we study itinitially, at leastin a fairly broad context.
Definitions
First, we define some types of stability for solutions of the (possibly) nonautonomous equation
xP D f .t; x/:
(3.9)
Definition. A solution x.t/ of (3.9) is (Lyapunov) stable if for each " > 0 and
t0 2 R there exists D ."; t0 / > 0 such that if x.t/ is a solution of (3.9) and
jx.t0 / x.t0 /j < then jx.t/ x.t/j < " for all t t0 .
Definition. A solution x.t/ of (3.9) is asymptotically stable if it is (Lyapunov)
stable and if for every t0 2 R there exists D .t0 / > 0 such that if x.t/ is a
solution of (3.9) and jx.t0 / x.t0 /j < then jx.t/ x.t/j ! 0 as t " 1.
Definition. A solution x.t/ of (3.9) is uniformly stable if for each " > 0 there
exists D ."/ > 0 such that if x.t/ is a solution of (3.9) and jx.t0 / x.t0 /j <
for some t0 2 R then jx.t/ x.t/j < " for all t t0 .
Some authors use a weaker definition of uniform stability that turns out to
be equivalent to Lyapunov stability for autonomous equations. Since our main
interest is in autonomous equations and this alternative definition is somewhat
more complicated than the definition given above, we will not use it here.
Definition. A solution x.t/ of (3.9) is orbitally stable if for every " > 0 there
exists D ."/ > 0 such that if x.t/ is a solution of (3.9) and jx.t1 / x.t0 /j <
for some t0 ; t1 2 R then
[
[
x.t/
B.x.t/; "/:
t t1
t t0
Next, we present a couple of definitions of stability for subsets of the (open)

phase space Rn of a dynamical system '.t; x/. (In these definitions, a
neighborhood of a set A is an open subset of that contains A.)
Definition. The set A is stable if every neighborhood of A contains a positively
invariant neighborhood of A.
73
Note that the definition implies that stable sets are positively invariant.
Definition. The set A is asymptotically stable if it is stable and there is some
neighborhood V of A such that !.x/ A for every x 2 V. (If V can be chosen
to be the entire phase space, then A is globally asymptotically stable.)
Examples
We now consider a few examples that clarify some properties of these definitions.
y
1
(
xP D y=2
yP D 2x:
b
Orbits are ellipses with major axis along the y-axis. The equilibrium solution
at the origin is Lyapunov stable even though nearby orbits sometimes move away
from it.
y
2
(
rP D 0
P D r 2 ;
or, equivalently,
(
xP D .x 2 C y 2 /y
yP D .x 2 C y 2 /x:
The solution moving around the unit circle is not Lyapunov stable, since
nearby solutions move with different angular velocities. It is, however, orbitally
stable. Also, the set consisting of the unit circle is stable.
74
Definitions of Stability
y
3
(
rP D r.1 r/
P D sin2 .=2/:
b
The constant solution .x; y/ D .1; 0/ is not Lyapunov stable and the set
f.1; 0/g is not stable. However, every solution beginning near .1; 0/ converges to
.1; 0/ as t " 1. This shows that it is not redundant to require Lyapunov stability
(or stability) in the definition of asymptotic stability of a solution (or a set).
Stability in Autonomous Equations

When we are dealing with a smooth autonomous differential equation
xP D f .x/
(3.10)
on an open set Rn , all of the varieties of stability can be applied to essentially the same object. In particular, let x be a function that solves (3.10), and
let

A.x/ WD x.t/ t 2 R
be the corresponding orbit. Then it makes sense to talk about the Lyapunov,
asymptotic, orbital, or uniform stability of x, and it makes sense to talk about the
stability or asymptotic stability of A.x/.
In this context, certain relationships between the various types of stability
follow from the definitions without too much difficulty.
Theorem. Let x be a function that solves (3.10), and let A.x/ be the corresponding orbit. Then:
1. If x is asymptotically stable then x is Lyapunov stable;
2. If x is uniformly stable then x is Lyapunov stable;
3. If x is uniformly stable then x is orbitally stable;
4. If A.x/ is asymptotically stable then A.x/ is stable;
75
5. If A.x/ contains only a single point, then Lyapunov stability of x, orbital
stability of x, uniform stability of x, and stability of A.x/ are equivalent.
We will not prove this theorem, but we will note that parts 1 and 2 are immediate results of the definitions (even if we were dealing with a nonautonomous
equation) and part 4 is also an immediate result of the definitions (even if A were
an arbitrary set).
Exercise 13 In items 118, an autonomous differential equation, a phase

space (that is an open subset of Rn ), and a particular solution x of the
equation are specified. For each of these items, state which of the following
statements is/are true:
(a) x is Lyapunov stable;
(b) x is asymptotically stable;
(c) x is orbitally stable;
(d) x is uniformly stable;
(e) A.x/ is stable;
(f) A.x/ is asymptotically stable.
You do not need to justify your answers or show your work. It may
convenient to express your answers in a concise form (e.g., in a table of
some sort). Use of variables r and signifies that the equation (as well as
the particular solution) is to be interpreted as in polar form.
1. xP D x, D R, x.t/ WD 0
2. xP D x, D R, x.t/ WD e t
3. fxP 1 D 1 C x22 ; xP 2 D 0g, D R2 , x.t/ WD .t; 0/
4. frP D 0; P D r 2 g, D R2 , x.t/ WD .1; t/
5. xP D x, D .0; 1/, x.t/ WD e t
6. fxP 1 D 1; xP 2 D
x1 x2 g, D R2 , x.t/ WD .t; 0/
7. xP D tanh x, D R, x.t/ WD sinh

76
.e t /
Principle of Linearized Stability
8. fxP 1 D tanh x1 ; xP 2 D 0g, D .0; 1/ R, x.t/ WD .sinh

9. xP D tanh x, D .0; 1/, x.t/ WD sinh
.e t /; 0/
.e t /
10. fxP 1 D sech x1 ; xP 2 D x1 x2 sech x1 g, D R2 ,

x.t/ WD .sinh 1 .t/; 0/
p
11. xP D x 2 =.1 C x 2 /, D R, x.t/ WD 2=.t C t 2 C 4/
12. fxP 1 D sech x1 ; xP 2 D
x2 g, D R2 , x.t/ WD .sinh
13. xP D sech x, D R, x.t/ WD sinh
.t/; 0/
.t/
14. fxP 1 D 1; xP 2 D 0g, D R2 , x.t/ WD .t; 0/

15. xP D 0, D R, x.t/ WD 0
16. xP D 1, D R, x.t/ WD t
17. fxP 1 D
18. xP D
x1 ; xP 2 D
x2 g, D R2 , x.t/ WD .e t ; 0/
x, D R, x.t/ WD 0
3.4 Principle of Linearized Stability

Suppose f is a continuously differentiable function such that
xP D f .x/
(3.11)
generates a continuous dynamical system ' on Rn . Suppose, moreover,

that x0 2 is a singular point of '. If x solves (3.11) and we set u WD x x0
and A WD Df .x0/, we see that, by the definition of derivative,
uP D f .u C x0 / D f .x0 / C Df .x0/u C R.u/ D Au C R.u/;
(3.12)
where R.u/=juj ! 0 as juj # 0. Because R.u/ is small when u is small, it is

reasonable to believe that solutions of (3.12) behave similarly to solutions of
uP D Au
(3.13)
for u near 0. Equivalently, it is reasonable to believe that solutions of (3.11)

behave like solutions of
xP D A.x x0 /
(3.14)
77
for x near x0 . Equation (3.13) (or sometimes (3.14)) is called the linearization
of (3.11) at x0 .
Now, weve defined (several types of) stability for equilibrium solutions of
(3.11) (as well as for other types of solutions and sets), but we havent really
given any tools for determining stability. In this lecture we present one such tool,
using the linearized equation(s) discussed above.
Definition. An equilibrium point x0 of (3.11) is hyperbolic if none of the eigenvalues of Df .x0 / have zero real part.
If x0 is hyperbolic, then either all the eigenvalues of A WD Df .x0/ have
negative real part or at least one has positive real part. In the former case, we
know that 0 is an asymptotically stable equilibrium solution of (3.13); in the latter
case, we know that 0 is an unstable solution of (3.13). The following theorem
says that similar things can be said about the nonlinear equation (3.11).
Theorem. (Principle of Linearized Stability) If x0 is a hyperbolic equilibrium
solution of (3.11), then x0 is either unstable or asymptotically stable, and its stability type (with respect to (3.11)) matches the stability type of 0 as an equilibrium
solution of (3.13) (where A WD Df .x0 /).
This theorem is an immediate consequence of the following two propositions.
Proposition. (Asymptotic Stability) If x0 is an equilibrium point of (3.11) and
all the eigenvalues of A WD Df .x0/ have negative real part, then x0 is asymptotically stable.
Proposition. (Instability) If x0 is an equilibrium point of (3.11) and some eigenvalue of A WD Df .x0 / has positive real part, then x0 is unstable.
Before we prove these propositions, we state and prove a lemma to which we
have referred before in passing.
Lemma. Let V be a finite-dimensional real vector space and let L 2 L.V; V/.
If all the eigenvalues of L have real part larger than c, then there is an inner
product h; i and an induced norm k k on V such that
hv; Lvi ckvk2
78
for every v 2 V.

Proof. Let n D dim V, and pick " > 0 so small that all the eigenvalues of L have
real part greater than c C n". Choose a basis fv1 ; : : : ; vn g for V that puts L in
modified real canonical form with the off-diagonal 1s replaced by "s, and let
h; i be the inner product associated with this basis (i.e. hvi ; vj i D ij ) and let
k k be the induced
Pn norm on V.
Given v D iD1 i vi 2 V, note that (if L D .ìj /)
!
n
n X
n
n X
2
2
X
X
X
X
i
j
ì i i2 C
ìj i j
ì i i2
"
hv; Lvi D
2

iD1
iD1 j i
n
X
n
X
iD1
ì i i2
iD1
n"i2
iD1
n
X
.ì i
iD1 j i
n"/i2
iD1
n
X
iD1
ci2 D ckvk2 :
Note that applying this theorem to L also tells us that, for some inner product,
hv; Lvi ckvk2
(3.15)
if all the eigenvalues of L have real part less than c.

Proof of Proposition on Asymptotic Stability. Without loss of generality, assume
that x0 D 0. Pick c < 0 such that all the eigenvalues of A have real part strictly
less than c. Because of equivalence of norms and because of the lemma, we can
work with a norm k k and a corresponding inner product h; i for which (3.15)
holds, with L D A. Let r > 0 be small enough that kR.x/k c=2kxk for all
x satisfying kxk r, and let
Br WD x 2 kxk < rg:
If x.t/ is a solution of (3.11) that starts in Br at time t D 0, then as long as x.t/

remains in Br
d
kx.t/k2 D 2hx.t/; x.t/i
P
D 2hx.t/; f .x.t//i
dt
D 2hx.t/; Ax.t/i C 2hx.t/; R.x.t//i
2ckx.t/k2 C 2kx.t/k kR.x.t//k
2ckx.t/k2
ckx.t/k2 D ckx.t/k2 :
This means that x.t/ 2 Br for all t 0, and x.t/ converges to 0 (exponentially
quickly) as t " 1.
The proof of the second proposition will be geometric and will contain ideas
that will be used to prove stronger results later in this text.
79
Proof of Proposition on Instability. We assume again that x0 D 0. If E u ,E s ,
and E c are, respectively, the unstable, stable, and center spaces corresponding to
(3.13), set E WD E s E c and E C WD E u . Then Rn D E C E , all of the
eigenvalues of AC WD AjE C have positive real part, and all of the eigenvalues
of A WD AjE have nonpositive real part. Pick constants a > b > 0 such that
all of the eigenvalues of AC have real part larger than a, and note that all of the
eigenvalues of A have real part less than b. Define an inner product h; iC (and
induced norm k kC ) on E C such that
hv; AviC akvk2C
for all v 2 E C , and define an inner product h; i (and induced norm k k ) on
E such that
hw; Awi bkwk2
for all w 2 E . Define h; i on E C E to be the direct sum of h; iC and h; i ;
i.e., let
hv1 C w1 ; v2 C w2 i WD hv1 ; v2 iC C hw1 ; w2 i
for all .v1 ; w1 /; .v2 ; w2 / 2 E C E . Let k k be the induced norm, and note that
kv C wk2 D kvk2C C kwk2 D kvk2 C kwk2
for all .v; w/ 2 E C E .
Now, take (3.11) and project it onto E C and E
system for .v; w/ 2 E C E
(
to get the corresponding
vP D AC v C RC .v; w/
wP D A w C R .v; w/;
(3.16)
with kR .v; w/k=kvp

C wk converging to 0 as kv C wk # 0. Pick " > 0 small
enough that a b 2 2" > 0, and pick r > 0 small enough that kR .v; w/k
"kv C wk whenever

v C w 2 Br WD v C w 2 E C E kv C wk < r :
Consider the truncated cone
Kr WD v C w 2 E C E
80

kvk > kwk \ Br :
(See Figure 1.) Suppose x D v C w is a solution of (3.16) that starts in Kr at

w
Br
Kr
b
Figure 3.1: The truncated cone.
time t D 0. For as long as the solution remains in Kr ,

d
kvk2 D 2hv; vi
P D 2hv; AC vi C 2hv; RC .v; w/i
dt
2akvk2 2kvk kRC .v; w/k 2akvk2 2"kvk kv C wk
p
1=2
D 2akvk2 2"kvk kvk2 C kwk2
2akvk2 2 2"kvk2
p
D 2.a
2"/kvk2 ;
and
d
kwk2 D 2hw; wi
P D 2hw; A wi C 2hw; R .v; w/i
dt
2bkwk2 C 2kwk kR .v; w/k
2bkwk2 C 2"kwk kv C wk
D 2bkwk2 C 2"kwk kvk2 C kwk2

p
2bkvk2 C 2 2"kvk2
p
D 2.b C 2"/kvk2 :
1=2
The first estimate says that as long as the solution stays in Kr , kvk grows exponentially, which means that the solution must eventually leave Kr . Combining
the first and second estimates, we have
d
.kvk2
dt
kwk2 / 2.a
p
2 2"/kvk2 > 0;
81
so g.v C w/ WD kvk2 kwk2 increases as t increases. But g is 0 on the lateral surface of Kr and is strictly positive in Kr , so the solution cannot leave Kr
through its lateral surface. Thus, the solution leaves Kr by leaving Br . Since
this holds for all solutions starting in Kr , we know that x0 must be an unstable
equilibrium point for (3.11).
3.5 Lyapunovs Direct Method

Another tool for determining stability of solutions is Lyapunovs direct method.
While this method may actually seem rather indirect, it does work directly on the
equation in question instead of on its linearization.
We will consider this method for equilibrium solutions of (possibly) nonautonomous equations. Let Rn be open and contain the origin, and suppose
that f W R ! Rn is a continuously differentiable function. Suppose, furthermore, that f .t; 0/ D 0 for every t 2 R, so x.t/ WD 0 is a solution of the
equation
xP D f .t; x/:
(3.17)
(The results we obtain in this narrow context can be applied to determine the
stability of other constant solutions of (3.17) by translation.)
In this section, a subset of that contains the origin in its interior will be
called a neighborhood of 0.
Definition. Suppose that D is a neighborhood of 0 and that W W D ! R is
continuous and satisfies W .0/ D 0. Then:
If W .x/ 0 for every x 2 D, then W is positive semidefinite.
If W .x/ > 0 for every x 2 D n f0g, then W is positive definite.
If W .x/ 0 for every x 2 D, then W is negative semidefinite.
If W .x/ < 0 for every x 2 D n f0g, then W is negative definite.
Definition. Suppose that D is a neighborhood of 0 and that V W R D ! R is
continuous and satisfies V .t; 0/ D 0 for every t 2 R. Then:
If there is a positive semidefinite function W W D ! R such that V .t; x/
W .x/ for every .t; x/ 2 R D, then V is positive semidefinite.
82
If there is a positive definite function W W D ! R such that V .t; x/

W .x/ for every .t; x/ 2 R D, then V is positive definite.
Lyapunovs Direct Method

If there is a negative semidefinite function W W D ! R such that V .t; x/
W .x/ for every .t; x/ 2 R D, then V is negative semidefinite.
If there is a negative definite function W W D ! R such that V .t; x/
W .x/ for every .t; x/ 2 R D, then V is negative definite.
Definition. If V W R D is continuously differentiable then its orbital derivative
(with respect to (3.17)) is the function VP W R D ! R given by the formula
@V
@V
VP .t; x/ WD
.t; x/ C
.t; x/ f .t; x/:
@t
@x
(Here @V .t; x/=@x represents the gradient of the function V .t; /.)
Note that if x.t/ is a solution of (3.17) then, by the chain rule,
d
V .t; x.t// D VP .t; x.t//:
dt
A function whose orbital derivative is always nonpositive is sometimes called a
Lyapunov function.
Theorem. (Lyapunov Stability) If there is a neighborhood D of 0 and a continuously differentiable positive definite function V W R D ! R whose orbital
derivative VP is negative semidefinite, then 0 is a Lyapunov stable solution of
(3.17).
Proof. Let " > 0 and t0 2 R be given. Assume, without loss of generality, that
B.0; "/ is contained in D. Pick a positive definite function W W D ! R such that
V .t; x/ W .x/ for every .t; x/ 2 R D. Let

m WD min W .x/ jxj D " :
Since W is continuous and positive definite, m is well-defined and positive. Pick

> 0 small enough that < " and

max V .t0 ; x/ jxj < m:
(Since V is positive definite and continuous, this is possible.)

Now, if x.t/ solves (3.17) and jx.t0 /j < then V .t0 ; x.t0 // < m, and
d
V .t; x.t// D VP .t; x.t// 0;
dt
for all t, so V .t; x.t// < m for every t t0 . Thus, W .x.t// < m for every
t t0 , so, for every t t0 , jx.t/j ". Since jx.t0 /j < ", this tells us that
jx.t/j < " for every t t0 .
83
Theorem. (Asymptotic Stability) Suppose that there is a neighborhood D of
0 and a continuously differentiable positive definite function V W R D !
R whose orbital derivative VP is negative definite, and suppose that there is a
positive definite function W W D ! R such that V .t; x/ W .x/ for every
.t; x/ 2 R D. Then 0 is an asymptotically stable solution of (3.17).
Proof. By the previous theorem, 0 is a Lyapunov stable solution of (3.17). Let
t0 2 R be given. Assume, without loss of generality, that D is compact. By
Lyapunov stability, we know that we can choose a neighborhood U of 0 such that
if x.t/ is a solution of (3.17) and x.t0 / 2 U, then x.t/ 2 D for every t t0 . We
claim that, in fact, if x.t/ is a solution of (3.17) and x.t0 / 2 U, then x.t/ ! 0 as
t " 1. Verifying this claim will prove the theorem.
Suppose that V .t; x.t// does not converge to 0 as t " 1. The negative
definiteness of VP implies that V .; x.// is nonincreasing, so, since V 0, there
must be a number c > 0 such that V .t; x.t// c for every t t0 . Then
W .x.t// c > 0 for every t t0 . Since W .0/ D 0 and W is continuous,
(3.18)
inf jx.t/j t t0 "
for some constant " > 0. Pick a negative definite function Y W D ! R such that
VP .t; x/ Y.x/ for every .t; x/ 2 R D. The compactness of D n B.0; "/, along
with (3.18), implies that

Y.x.t// t t0
is bounded away from 0. This, in turn, implies that

VP .t; x.t// t t0
is bounded away from 0. In other words,
d
V .t; x.t// D VP .t; x.t//
(3.19)
dt
for some constant > 0. Clearly, (3.19) contradicts the nonnegativity of V for
large t.
That contradiction implies that V .t; x.t// ! 0 as t " 1. Pick a positive definite function W W D ! R such that V .t; x/ W .x/ for every .t; x/ 2 R D,
and note that W .x.t// ! 0 as t " 1.
Let r > 0 be given, and let

wr D min W .p/ p 2 D n B.0; r/ ;
84
which is defined and positive by the compactness of D and the continuity and
positive definiteness of W . Since W .x.t// ! 0 as t " 1, there exists T such
that W .x.t// < wr for every t > T . Thus, for t > T , it must be the case that
x.t/ 2 B.0; r/. Hence, 0 is asymptotically stable.
LaSalles Invariance Principle

It may seem strange that we ned to bound V by a time-independent, positive
definite function W from above. Indeed, some textbooks (see, e.g., Theorem 2.20
in Stability, Instability, and Chaos by Glendinning) contain asymptotic stability
theorems omitting this hypothesis. There is a counterexample by Massera that
demonstrates the necessity of the hypothesis.
Exercise 14 Show, by means of a counterexample, that the theorem on

asymptotic stability via Lyapunovs direct method fails if the hypothesis
about W is dropped.
(You may, but do not have to, proceed as follows. Let g W R ! R be a
function that is twice continuously differentiable and satisfies g.t/ e t
for every t 2 R, g.t/ 1 for every t 0, g.t/ D e t for every
[
t
.n 2 n ; n C 2 n /;
n2N
and g.n/ D 1 for every n 2 N. Let f W R R ! R be the function defined

by the formula
g 0 .t/
f .t; x/ WD
x;
g.t/
and let V W R R ! R be the function defined by the formula

Z t
x2
2
V .t; x/ WD
3
g./
d

:
g.t/2
0
Show that, for x near 0, V .t; x/ is positive definite and VP .t; x/ is negative
definite, but the solution 0 of (3.17) is not asymptotically stable.)
3.6 LaSalles Invariance Principle

Linearization versus Lyapunov Functions
In the previous two lectures, we have talked about two different tools that can be
used to prove that an equilibrium point x0 of an autonomous system
xP D f .x/
(3.20)
is asymptotically stable: linearization and Lyapunovs direct method. One might

ask which of these methods is better. Certainly, linearization seems easier to
apply because of its straightforward nature: Compute the eigenvalues of Df .x0 /.
85
The direct method requires you to find an appropriate Lyapunov function, which
doesnt seem so straightforward. But, in fact, anytime linearization works, a
simple Lyapunov function works, as well.
To be more precise, suppose x0 D 0 and all the eigenvalues of A WD Df .0/
have negative real part. Pick an inner product h; i and induced norm k k such
that, for some c > 0,
hx; Axi ckxk2
for all x 2 Rn . Pick r > 0 small enough that kf .x/ Axk .c=2/kxk whenever
kxk r, let
D D x 2 Rn kxk r ;
and define V W R D ! R by the formula V .t; x/ D kxk2 . Since k k is a norm,

V is positive definite. Also
VP .t; x/ D 2hx; f .x/i D 2.hx; Axi C hx; f .x/
2
2. ckxk C kxkkf .x/
Axk/
Axi/
ckxk2 ;
so VP is negative definite.
On the other hand, there are very simple examples to illustrate that the direct
method works in some cases where linearization doesnt. For example, consider
xP D x 3 on R. The equilibrium point at the origin is not hyperbolic, so linearization fails to determine stability, but it is easy to check that x 2 is positive
definite and has a negative definite orbital derivative, thus ensuring the asymptotic stability of 0.
A More Complicated Example

The previous example is so simple that it might make one question whether the
direct method is of any use on problems where stability cannot be determined by
linearization or by inspection. Thus, lets consider something more complicated.
Consider the planar system
(
xP D y x 3
yP D x 5 :
86
The origin is a nonhyperbolic equilibrium point, with 0 being the only eigenvalue,
so the principle of linearized stability is of no use. A sketch of the phase portrait
indicates that orbits circle the origin in the counterclockwise direction, but it is
not obvious whether they spiral in, spiral out, or move on closed curves.
The simplest potential Lyapunov function that often turns out to be useful is
the square of the standard Euclidean norm, which in this case is V WD x 2 C y 2 .

The orbital derivative is
VP D 2x xP C 2y yP D 2x 5 y
2xy
2x 4 :
(3.21)
For some points .x; y/ near the origin (e.g., .; /) VP < 0, while for other points
near the origin (e.g., .; /) VP > 0, so this function doesnt seem to be of much
use.
Sometimes when the square of the standard Euclidean norm doesnt work,
some other homogeneous quadratic function does. Suppose we try the function
V WD x 2 C xy C y 2 , with and to be determined. Then
VP D .2x C y/xP C .x C 2y/yP
D
.2x C y/.y C x 3 / C .x C 2y/x 5

2x 4 C x 6
2xy
x 3y C 2x 5 y
y 2 :
Setting .x; y/ D .; 2 / for positive and small, we see that VP is not going to
be negative semidefinite, no matter what we pick and to be.
If these quadratic functions dont work, maybe something customized for the
particular equation might. Note that the right-hand side of the first equation in
(3.21) sort of suggests that x 3 and y should be treated as quantities of the same
order of magnitude. Lets try V WD x 6 C y 2 , for some > 0 to be determined.
Clearly, V is positive definite, and
VP D 6x 5 xP C 2y yP D .2
6/x 5 y
6x 8 :
If 3, then VP is of opposite signs for .x; y/ D .; / and for .x; y/ D .; /

when is small. Hence, we should set D 3, yielding VP D 6x 8 0. Thus
V is positive definite and VP is negative semidefinite, implying that the origin is
Lyapunov stable.
Is the origin asymptotically stable? Perhaps we can make a minor modification to the preceding formula for V so as to make VP strictly negative in a deleted
neighborhood of the origin without destroying the positive definiteness of V . If
we added a small quantity whose orbital derivative was strictly negative when
x D 0 and jyj is small and positive, this might work. Experimentation suggests
that a positive multiple of xy 3 might work, since this quantity changes from positive to negative as we cross the y-axis in the counterclockwise direction. Also,
it is at least of higher order than 3y 2 near the origin, so it has the potential of
preserving the positive definiteness of V .
In fact, we claim that V WD x 6 C xy 3 C 3y 2 is positive definite with negative
definite orbital derivative near 0. A handy inequality, sometimes called Youngs
inequality, that can be used in verifying this claim (and in other circumstances,
as well) is given in the following lemma.
87
Lemma. (Youngs Inequality) If a; b 0, then
ab
ap
bq
C ;
p
q
(3.22)
for every pair of numbers p; q 2 .1; 1/ satisfying

1
1
C D 1:
p
q
(3.23)
Proof. Assume that (3.23) holds. Clearly (3.22) holds if b D 0, so assume that
b > 0, and fix it. Define g W 0; 1/ by the formula
g.x/ WD
xp
bq
C
p
q
xb:
Note that g is continuous, and g 0 .x/ D x p 1 b for every x 2 .0; 1/. Since
limx#0 g 0 .x/ D b < 0, limx"1 g 0 .x/ D 1, and g 0 is increasing on .0; 1/,
we know that g has a unique minimizer at x0 D b 1=.p 1/. Thus, for every
x 2 0; 1/ we see, using (3.23), that
g.x/ g.b
1=.p 1/
b p=.p
/D
p
1/
bq
C
q
p=.p 1/
1
1
C
p
q

1 b q D 0:
In particular, g.a/ 0, so (3.22) holds.

Now, let V D x 6 C xy 3 C 3y 2 . Applying Youngs inequality with a D jxj,
b D jyj3 , p D 6, and q D 6=5, we see that
jxy 3 j D jxjjyj3
if jyj 1, so
5jyj18=5
1
jxj6
5
C
x6 C y2
6
6
6
6
5
13
V x6 C y2
6
6
if jyj 1. Also,
VP D
88
6x 8 C y 3 xP C 3xy 2 yP D
6x 8
x 3 y 3 C 3x 6 y 2
6x 8
y 4:
y 3 .y C x 3 / C 3x 6 y 2
Applying Youngs inequality to the two mixed terms in this orbital derivative, we
have
3jxj8
5jyj24=5
3
5
j x 3 y 3 j D jxj3 jyj3
C
x8 C y4
8
8
8
8

if jyj 1, and
j3x 6 y 2 j D 3jxj6 jyj2 3
if jyj 1=2. Thus,

jyj8
9
3jxj8
3
9
3
C
D x8 C y8 x8 C y4
4
4
4
4
4
64
27 8
x
8
VP
21 4
y
64
if jyj 1=2, so, in a neighborhood of 0, V is positive definite and VP is negative

definite, which implies that 0 is asymptotically stable.

We would have saved ourselves a lot of work on the previous example if we could
have just stuck with the moderately simple function V D x 6 C 3y 2 , even though
its orbital derivative was only negative semidefinite. Notice that the set of points
where VP was 0 was small (the y-axis) and at most of those points the vector field
was not parallel to the set. LaSalles Invariance Principle, which we shall state
and prove for the autonomous system
xP D f .x/;
(3.24)
allows us to use such a V to prove asymptotic stability.

Theorem. (LaSalles Invariance Principle) Suppose there is a neighborhood D
of 0 and a continuously differentiable (time-independent) positive definite function V W D ! R whose orbital derivative VP (with respect to (3.24)) is negative
semidefinite. Let I be the union of all complete orbits contained in

x 2 D VP .x/ D 0 :
Then there is a neighborhood U of 0 such that for every x0 2 U, !.x0 / I.
Before proving this, we note that when applying it to V D x 6 C 3y 2 in the

previous example, the set I is a singleton containing the origin and, since D can
be assumed to be compact, each solution beginning in U actually converges to 0
as t " 1.
Proof of LaSalles Invariance Principle. Let ' be the flow generated by (3.24).
By a previous theorem, 0 must be Lyapunov stable, so we can pick a neighborhood U of 0 such that '.t; x/ 2 D for every x0 2 U and every t 0.
Let x0 2 U and y 2 !.x0 / be given. By the negative semidefiniteness of
VP , we know that V .'.t; x0 // is a nonincreasing function of t. By the positive
89
definiteness of V , we know that V .'.t; x0 // remains nonnegative, so it must
approach some constant c 0 as t " 1. By continuity of V , V .z/ D c for
every z 2 !.x0 /. Since !.x0 / is invariant, V .'.t; y// D c for every t 2 R. The
definition of orbital derivative then implies that VP .'.t; y// D 0 for every t 2 R.
Hence, y 2 I.
Exercise 15 Show that .x.t/; y.t// D .0; 0/ is an asymptotically stable
solution of
(
xP D x 3 C 2y 3
yP D 2xy 2 :
90
4
Conjugacies
4.1 Hartman-Grobman Theorem: Part 1
The Principle of Linearized Stability indicates one way in which the flow near
a singular point of an autonomous ODE resembles the flow of its linearization.
The Hartman-Grobman Theorem gives further insight into the extent of the resemblance; namely, there is a local topological conjugacy between the two. We
will spend the next 5 sections talking about the various forms of this theorem and
their proofs. This amount of attention is justified not only by the significance of
the theorem but the general applicability of the techniques used to prove it.
Let Rn be open and let f W ! Rn be continuously differentiable.
Suppose that x0 2 is a hyperbolic equilibrium point of the autonomous equation
xP D f .x/:
(4.1)
Let B D Df .x0 /, and let ' be the (local) flow generated by (4.1). The version of
the Hartman-Grobman Theorem were primarily interested in is the following.
Theorem. (Local Hartman-Grobman Theorem for Flows) Let , f , x0 , B,
and ' be as described above. Then there are neighborhoods U and V of x0 and
a homeomorphism h W U ! V such that
'.t; h.x// D h.x0 C e tB .x
whenever x 2 U and x0 C e tB .x
x0 //
x0 / 2 U.
It will be easier to derive this theorem as a consequence of a global theorem

for maps than to prove it directly. In order to state that version of the theorem,
we will need to introduce a couple of function spaces and a definition.
Let

Cb0 .Rn / D w 2 C.Rn ; Rn / sup jw.x/j < 1 :
x2Rn
91
4. C ONJUGACIES
When equipped with the norm
kwk0 WD sup kw.x/k;
x2Rn
where k k is some norm on Rn , Cb0 .Rn / is a Banach space. (We shall pick a
particular norm k k later.)
Let

Cb1 .Rn / D w 2 C 1 .Rn ; Rn / \ Cb0 .Rn / sup kDw.x/k < 1 ;
x2Rn
where k k is the operator norm corresponding to some norm on Rn . Note that

the functional
kw.x1 / w.x2 /k
Lip.w/ WD sup
kx1 x2 k
x1 ;x2 2Rn
x1 x2
is defined on Cb1 .Rn /. We will not define a norm on Cb1 .Rn /, but will often use
Lip, which is not a norm, to describe the size of elements of Cb1 .Rn /.
Definition. If A 2 L.Rn ; Rn / and none of the eigenvalues of A lie on the unit
circle, then A is hyperbolic.
Note that if x0 is a hyperbolic equilibrium point of (4.1) and A D e Df .x0/ ,
then A is hyperbolic.
Theorem. (Global Hartman-Grobman Theorem for Maps) Suppose that the
map A 2 L.Rn ; Rn / is hyperbolic and invertible. Then there exists a number
" > 0 such that for every g 2 Cb1 .Rn / satisfying Lip.g/ < " there exists a
unique function v 2 Cb0 .Rn / such that
F .h.x// D h.Ax/
for every x 2 Rn , where F D A C g and h D I C v. Furthermore, h W Rn ! Rn
is a homeomorphism.

Subspaces and Norms
We start off with a lemma that is analogous to the lemma in Lecture 21, except
this one will deal with the magnitude, rather than the real part, of eigenvalues.
92
Hartman-Grobman Theorem: Part 2

Lemma. Let V be a finite-dimensional real vector space and let L 2 L.V; V/. If
all the eigenvalues of L have magnitude less than c, then there is a norm k k on
V such that
kLvk ckvk
for every v 2 V.
Proof. As in the previous lemma, the norm will be the Euclidean norm corresponding to the modification of the real canonical basis that gives a matrix representation of L that has "s in place of the off-diagonal 1s. With respect to this
basis, it can be checked that
LT L D D C R."/;
where D is a diagonal matrix, each of whose diagonal entries is less than c 2 , and
R."/ is a matrix whose entries converge to 0 as " # 0. Hence, as in the proof of
the earlier lemma, we can conclude that if " is sufficiently small then
kLvk2 D hv; LT Lvi c 2 kvk2
for every v 2 V (where h; i is the inner product that induces k k).
Note that if L is a linear operator, all of whose eigenvalues have magnitude
greater than c, then we get the reverse inequality
kLvk ckvk
for some norm k k. This follows trivially in the case when c 0, and when
c > 0 it follows by applying the lemma to L 1 (which exists, since 0 is not an
eigenvalue of L).
Now, suppose that A 2 L.Rn ; Rn / is hyperbolic. Then, since A has only
finitely many eigenvalues, there is a number a 2 .0; 1/ such that none of the
eigenvalues of A are in the closed annulus
B.0; a
1/
n B.0; a/:
Using the notation developed when we were deriving the real canonical form, let
8
9
< M
=
E D
N.A I /
:
;
2. a;a/
8
9 8
9
>
>
>
>
< M
= < M
=
Re u u 2 N.A I /
Im u u 2 N.A I / ;
>
>
>
>
: jj<a
; : jj<a
;
Im 0
Im 0
93
4. C ONJUGACIES
and let
EC D
8
<
:
2. 1; a
9
=
N.A
I /
;
9 8
>
>
>
= < M
I /
Im u u 2 N.A
>
>
jj>a 1
>
;
:
1 /[.a 1 ;1/
< M
Re u u 2 N.A
:jj>a 1
Im 0
Im 0
9
>
>
>
=
I / :
>
>
>
;
Then Rn D E E C , and E and E C are both invariant under A. Define

P 2 L.Rn ; E / and P C 2 L.Rn ; E C / to be the linear operators that map
each x 2 Rn to the unique elements P x 2 E and P C x 2 E C such that
P x C P C x D x.
Let A 2 L.E ; E / and AC 2 L.E C ; E C / be the restrictions of A to E
and E C , respectively. By the lemma and the discussion immediately thereafter,
we can find a norm k k for E and a norm k kC for E C such that
kA xk akxk
for every x 2 E , and
kAC xkC a
kxkC
for every x 2 E C . Define a norm k k on Rn by the formula

kxk D maxfkP xk ; kP C xkC g:
(4.2)
This is the norm on Rn that we will use throughout our proof of the (global)
Hartman-Grobman Theorem (for maps). Note that kxk D kxk if x 2 E , and
kxk D kxkC if x 2 E C .
Recall that we equipped Cb0 .Rn / with the norm k k0 defined by the formula
kwk0 WD sup kw.x/k:
x2Rn
The norm on Rn on the right-hand side of this formula is that given in (4.2).
Recall also that we will use the functional Lip defined by the formula
Lip.w/ WD
sup
x1 ;x2 2Rn
x1 x2
kw.x1 /
kx1
The norm on Rn on the right-hand side of this formula is also that given in (4.2).
Let

Cb0 .E / D w 2 C.Rn ; E / sup kw.x/k < 1 ;
x2Rn
94
w.x2 /k
x2 k

and let

Cb0 .E C / D w 2 C.Rn ; E C / sup kw.x/kC < 1 :
x2Rn
Since Rn D E E C , it follows that
Cb0 .Rn / D Cb0 .E / Cb0 .E C /;

and the corresponding decomposition of an element w 2 Cb0 .Rn / is
wDP
w C P C w:
We equip Cb0 .E / and Cb0 .E C / with the same norm k k0 that we used on
thereby making each of these two spaces a Banach space. It is not hard
to see that
kwk0 D maxfkP wk0 ; kP C wk0 g:
Cb0 .Rn /,

Linear and Nonlinear Maps
Now, assume that A is invertible, so that
kAxk
> 0:
x0 kxk
inf
Choose, and fix, a positive constant

" < min 1

kAxk
a; inf
:
x0 kxk
Choose, and fix, a function g 2 C1b .Rn / for which Lip.g/ < ". The (global)
Hartman-Grobman Theorem (for maps) will be proved by constructing a map
from Cb0 .Rn / to Cb0 .Rn / whose fixed points would be precisely those objects v
which, when added to the identity I , would yield solutions h to the conjugacy
equation
.A C g/ h D h A;
(4.3)
and then showing that is a contraction (and that h is a homeomorphism).
Plugging h D I C v into (4.3) and manipulating the result, we can see that
that equation is equivalent to the equation
Lv D .v/;
(4.4)
95
4. C ONJUGACIES
where .v/ WD g .I C v/ A
Lv D v
and
AvA
DW .id A/v:
Since the composition of continuous functions is continuous, and the composition of functions is bounded if the outer function in the composition is bounded,
it is clear that is a (nonlinear) map from Cb0 .Rn / to Cb0 .Rn /. Similarly, A and,
therefore, L are linear maps from Cb0 .Rn / to Cb0 .Rn /. We will show that L can
be inverted and then apply L 1 to both sides of (4.4) to get
vDL
..v// DW .v/
(4.5)
as our fixed point equation.
Inverting L
Since A behaves significantly differently on E than it does on E C , A and, therefore, L behave significantly differently on Cb0 .E / than they do on Cb0 .E C /.
For this reason, we will analyze L by looking at its restrictions to Cb0 .E / and
to Cb0 .E C /. Note that Cb0 .E / and Cb0 .E C / are invariant under A and, therefore, under L. Therefore, it makes sense to let A 2 L.Cb0 .E /; Cb0 .E // and
AC 2 L.Cb0 .E C /; Cb0 .E C // be the restrictions of A to Cb0 .E / and Cb0 .E C /, respectively, and let L 2 L.Cb0 .E C /; Cb0 .E C // and LC 2 L.Cb0 .E C /; Cb0 .E C //
be the corresponding restrictions of L. Then L will be invertible if and only if L
and LC are each invertible. To invert L and LC we use the following general
result about the invertibility of operators on Banach spaces.
Lemma. Let X be a Banach space with norm k kX and corresponding operator
norm kkL.X ;X / . Let G be a linear map from X to X , and let c < 1 be a constant.
Then:
(a) If kGkL.X ;X / c, then id G is invertible and
k.id G/
(b) If G is invertible and kG
1k
L.X ;X /
k.id G/
96
kL.X ;X /
1
1
c, then id G is invertible and

kL.X ;X /
c
1
Proof. The space of bounded linear maps from X to X is a Banach space using
the operator norm. In case (a), the bound on kGkL.X ;X / , along with the Cauchy

convergence criterion, implies that the series
1
X
Gk
kD0
converges to a bounded linear map from X to X ; call it H . In fact, we see that

(by the formula for the sum of a geometric series)
kH kL.X ;X /
1
1
It is not hard to check that H .id G/ D .id G/H D id, so H D .id G/ 1 .

In case (b), we can apply the results of (a) with G 1 in place of G to deduce
that id G 1 is invertible and that
1
k.id G 1 / 1 kL.X ;X /
:
1 c
Since id G D G.id G 1 / D .id G 1 /G, it is not hard to check that
.id G 1 / 1 G 1 is the inverse of id G and that
c
:
k .id G 1 / 1 G 1 kL.X ;X /
1 c
The first half of this lemma is useful for inverting small perturbations of the
identity, while the second half is useful for inverting large perturbations of the
identity. It should, therefore, not be too surprising that we will apply the first half
with G D A and the second half with G D AC (since A compresses things in
the E direction and stretches things in the E C direction).
First, consider A . If w 2 Cb0 .E /, then
1
kA wk0 D kA w A
k0 D sup kAw.A
x2Rn
x/k D sup kAw.y/k

y2Rn
a sup kw.y/k D akwk0 ;

y2Rn
so the operator norm of A is bounded by a. Applying the first half of the lemma
with X D Cb0 .E /, G D A , and c D a, we find that L is invertible, and its
inverse has operator norm bounded by .1 a/ 1 .
Next, consider AC . It is not too hard to see that AC is invertible, and
C
.A / 1 w D A 1 w A. If w 2 Cb0 .E C /, then (since the eigenvalues of
the restriction of A 1 to E C all have magnitude less than a)
k.AC /
wk0 D kA
w Ak0 D sup kA
D sup kA
y2Rn
w.Ax/k
x2Rn
w.y/k a sup kw.y/k D akwk0 ;

y2Rn
97
4. C ONJUGACIES
so the operator norm of .AC / 1 is bounded by a. Applying the second half of the
lemma with X D Cb0 .E C /, G D AC , and c D a, we find that LC is invertible,
and its inverse has operator norm bounded by a.1 a/ 1 .
Putting these two facts together, we see that L is invertible, and, in fact,
D .L /
C .LC /
P C:
If w 2 Cb0 .Rn /, then
kL
wk0 D sup kL
w.x/k
x2Rn
D sup maxfkP L
x2Rn
w.x/k; kP CL
w.x/kg
D sup maxfk.L / 1 P w.x/k; k.LC / 1 P C w.x/kg

x2Rn

1
a
sup max
kw.x/k;
kw.x/k
1 a
1 a
x2Rn
1
1
D
sup kw.x/k D
kwk0 ;
1 a x2Rn
1 a
so the operator norm of L
is bounded by .1
a/
1.

The Contraction Map
Recall that we are looking for fixed points v of the map WD L 1 , where
.v/ WD g .I C v/ A 1 . We have established that L 1 is a well-defined
linear map from Cb0 .Rn / to Cb0 .Rn / and that its operator norm is bounded by
.1 a/ 1 . This means that is a well-defined (nonlinear) map from Cb0 .Rn / to
98

Cb0 .Rn /; furthermore, if v1 ; v2 2 Cb0 .Rn /, then
k.v1 /
.v2 /k0 D kL
D
1
1
..v1 /
kg .I C v1 / A
k.v1 /
g .I C v2 / A
sup kg.A 1 x C v1 .A 1 x//

a x2Rn
"

sup k.A 1 x C v1 .A 1 x//
1 a x2Rn
"
D
sup kv1 .A 1 x/
1 a x2Rn
"
sup kv1 .y/ v2 .y/k D
D
1 a y2Rn
.v2 //k0
g.A
.A
k0
"
a
x C v2 .A
x C v2 .A
v2 .A
1
.v2 /k0
x//k
x//k
x/k
kv1
v2 k0 :
This shows that is a contraction, since " was chosen to be less than 1 a.
By the contraction mapping theorem, we know that has a unique fixed point
v 2 Cb0 .Rn /; the function h WD I Cv satisfies F h D hA, where F WD A Cg.
It remains to show that h is a homeomorphism.
Injectivity
Before we show that h itself is injective, we show that F is injective. Suppose
it werent. Then we could choose x1 ; x2 2 Rn such that x1 x2 but F .x1 / D
F .x2 /. This would mean that Ax1 C g.x1 / D Ax2 C g.x2 /, so
kAx1
kA.x1 x2 /k
D
kx1 x2 k
kx1
kg.x1 /
Ax2 k
D
x2 k
kx1
kAxk
< " inf
;
x0 kxk
g.x2 /k
Lip.g/
x2 k
which would be a contradiction.

Now we show that h is injective. Let x1 ; x2 2 Rn satisfying h.x1 / D h.x2 /
be given. Then
h.Ax1 / D F .h.x1 // D F .h.x2 // D h.Ax2 /;
and, by induction, we have h.An x1 / D h.An x2 / for every n 2 N. Also,
F .h.A
x1 // D h.AA
D F .h.A
x1 / D h.x1 / D h.x2 / D h.AA
x2 /
x2 //;
99
4. C ONJUGACIES
so the injectivity of F implies that h.A 1 x1 / D h.A 1 x2 /; by induction, we
have h.A n x1 / D h.A n x2 / for every n 2 N. Set z D x1 x2 . Since I D h v,
we know that for any n 2 Z
kAn zk D kAn x1
An x2 k
D k.h.An x1 /
D kv.An x1 /
v.An x1 //
.h.An x2 /
v.An x2 //k
v.An x2 /k 2kvk0 :
Because of the way the norm was chosen, we then know that for n 0
kP C zk an kAn P C zk an kAn zk 2an kvk0 ! 0;
as n " 1, and we know that for n 0
kP zk a
as n #
kAn P zk a
kAn zk 2a
kvk0 ! 0;
1. Hence, z D P z C P Cz D 0, so x1 D x2 .
Surjectivity
It may seem intuitive that a map like h that is a bounded perturbation of the
identity is surjective. Unfortunately, there does not appear to be a way of proving
this that is simultaneously elementary, short, and complete. We will therefore
rely on the following topological theorem without proving it.
Theorem. (Invariance of Domain) Every continuous injective map from Rn to
Rn maps open sets to open sets.
In particular, this theorem implies that h.Rn / is open. If we can show that
h.Rn / is closed, then (since h.Rn / is clearly nonempty) this will mean that
h.Rn / D Rn , i.e., h is surjective.
So, suppose we have a sequence .h.xk // of points in h.Rn / that converges to
a point y 2 Rn . Without loss of generality, assume that
kh.xk /
yk 1
for every k. This implies that kh.xk /k kyk C 1, which in turn implies that
kxk k kyk C kvk0 C 1. Thus, the sequence .xk / is bounded and therefore
has a subsequence .xk` / converging to some point x0 2 Rn . By continuity of
h, .h.xk` // converges to h.x0 /, which means that h.x0 / D y. Hence, h.Rn / is
closed.
100
Continuity of the Inverse

The bijectivity of h implies that h 1 is defined. To complete the verification that
h is a homeomorphism, we need to confirm that h 1 is continuous. But this is an
immediate consequence of the the Invariance of Domain Theorem.

Modifying the Vector Field
Consider the continuously differentiable autonomous ODE
xP D f .x/
(4.6)
with an equilibrium point that, without loss of generality, is located at the origin.
For x near 0, f .x/ Bx, where B D Df .0/. Our goal is to come up with
a modification fQ of f such that fQ.x/ D f .x/ for x near 0 and fQ.x/ Bx
for all x. If we accomplish this goal, whatever information we obtain about the
relationship between the equations
xP D fQ.x/
(4.7)
xP D Bx
(4.8)
and
will also hold between (4.6) and (4.8) for x small.
Pick W 0; 1/ ! 0; 1 to be a C 1 function satisfying
(
1 if s 1
.s/ D
0 if s 2;
and let C D sups20;1/ j 0 .s/j. Given " > 0, pick r > 0 so small that
kDf .x/
Bk <
"
2C C 1
whenever kxk 2r. (We can do this since Df .0/ D B and Df is continuous.)
Define fQ by the formula

kxk
Q
f .x/ D Bx C
.f .x/
r
Bx/:
Note that fQ is continuously differentiable, agrees with f for kxk r, and agrees
with B for kxk 2r. We claim that fQ B has Lipschitz constant less than ".
101
4. C ONJUGACIES
Assuming, without loss of generality, that kxk and kyk are less than or equal to
2r, we have (using the Mean Value Theorem)
k.fQ.x/ Bx/ .fQ.y/ By/k

kxk
kyk

D
.f .x/ Bx/
.f .y/ By/

r
r

kxk
k.f .x/ Bx/ .f .y/ By/k

r

kxk
kyk
kf .y/ Byk
r
r
"
jkxk kykj
"

kx yk C C
kyk
2C C 1
r
2C C 1
"kx yk:
Now, consider the difference between e B and '.1; /, where ' is the flow
generated by fQ. Let g.x/ D '.1; x/ e B x. Then, since fQ.x/ D B.x/ for
all large x, g.x/ D 0 for all large x. Also, g is continuously differentiable, so
g 2 Cb1 .Rn /. If we apply the variation of constants formula to (4.7) rewritten as
xP D Bx C .fQ.x/
Bx/;
we find that
g.x/ D
so
e .1
s/B
fQ.'.s; x//
B'.s; x/ ds;
kg.x/ g.y/k
Z 1
ke .1 s/B kk.fQ.'.s; x// B'.s; x// .fQ.'.s; y// B'.s; y//k ds

0
Z 1
"
ke .1 s/B kk'.s; x/ '.s; y/k ds
0
Z 1
kx yk"
ke .1 s/B kke .kBkC"/s 1k ds;
0
by continuous dependence on initial conditions. Since

Z 1
"
ke .1 s/B kke .kBkC"/s 1k ds ! 0
0
as " # 0, we can make the Lipschitz constant of g as small as we want by making

" small (through shrinking the neighborhood of the origin on which fQ and f
agree).
102
Conjugacy for t D 1
If 0 is a hyperbolic equilibrium point of (4.6) (and therefore of (4.7)) then none
of the eigenvalues of B are imaginary. Setting A D e B , it is not hard to show
that the eigenvalues of A are the exponentials of the eigenvalues of B, so none of
the eigenvalues of A have modulus 1; i.e., A is hyperbolic. Also, A is invertible
(since A 1 D e B ), so we can apply the global Hartman-Grobman Theorem for
maps and conclude that there is a homeomorphism h W Rn ! Rn such that
'.1; h.x// D h.e B x/
(4.9)
for every x 2 Rn (where ' is the flow corresponding to (4.7)).
Conjugacy for t 1
For the Hartman-Grobman Theorem for flows, we need
'.t; h.x// D h.e tB x/
for every x 2 Rn and every t 2 R. Fix t 2 R, and consider the function hQ defined
by the formula
Q
h.x/
D '.t; h.e tB x//:
(4.10)
As the composition of homeomorphisms, hQ is a homeomorphism. Furthermore,
the fact that h satisfies (4.9) implies that
Q
'.1; h.x//
D '.1; '.t; h.e
B
D '.t; h.e e
tB
Q
so (4.9) holds if h is replaced by h.
Now,
hQ I D '.t; / h e tB I
D .'.t; /
tB
e tB / h e
x/// D '.t; '.1; h.e
x// D '.t; h.e
tB
C e tB .h
tB
x///
Q B x/;
e x/// D h.e
tB B
I/ e
tB
DW v1 C v2 :
The fact that '.t; x/ and e tB x agree for large x implies that '.t; / e tB is
bounded, so v1 is bounded, as well. The fact that h I is bounded implies that
v2 is bounded. Hence, hQ I is bounded.
The uniqueness part of the global Hartman-Grobman Theorem for maps now
implies that h and hQ must be the same function. Using this fact and substituting
y D e tB x in (4.10) yields
h.e tB y/ D '.t; h.y//
for every y 2 Rn and every t 2 Rn . This means that the flows generated by (4.8)
and (4.7) are globally topologically conjugate, and the flows generated by (4.8)
and (4.6) are locally topologically conjugate.
103
4. C ONJUGACIES
4.6 Constructing Conjugacies

The Hartman-Grobman Theorem gives us conditions under which a conjugacy
between certain maps or between certain flows may exist. Some limitations of
the theorem are:
The conditions it gives are sufficient, but certainly not necessary, for a
conjugacy to exist.
It doesnt give a simple way to construct a conjugacy (in closed form, at
least).
It doesnt indicate how smooth the conjugacy might be.
These shortcomings can be addressed in a number of different ways, but we wont
really go into those here. We will, however, consider some aspects of conjugacies.
Differentiable Conjugacies of Flows

Consider the autonomous differential equations
xP D f .x/
(4.11)
xP D g.x/;
(4.12)
and
generating, respectively, the flows ' and . Recall that the conjugacy equation
for ' and is
'.t; h.x// D h. .t; x//
(4.13)
for every x and t. Not only is (4.13) somewhat complicated, it appears to require
you to solve (4.11) and (4.12) before you can look for a conjugacy h. Suppose,
however, that h is a differentiable conjugacy. Then, we can differentiate both
sides of (4.13) with respect to t to get
f .'.t; h.x/// D Dh. .t; x//g. .t; x//:
Substituting (4.13) into the left-hand side of (4.14) and then replacing
x, we get the equivalent equation
f .h.x// D Dh.x/g.x/:
104
(4.14)
.t; x/ by
(4.15)
Note that (4.15) involves the functions appearing in the differential equations,
rather than the formulas for the solutions of those equations. Note, also, that
(4.15) is the same equation you would get if you took a solution x of (4.12) and
required the function h x to satisfy (4.11).
Constructing Conjugacies
An Example for Flows

As the simplest nontrivial example, let a; b 2 R be distinct constants and consider the equations
xP D ax
(4.16)
and
xP D bx
(4.17)
for x 2 R. Under what conditions on a and b does there exist a topological

conjugacy h taking solutions of (4.17) to solutions of (4.16)? Equation (4.15)
tells us that if h is differentiable then
ah.x/ D h0 .x/bx:
(4.18)
If b 0, then separating variables in (4.18) implies that on intervals avoiding

the origin h must be given by the formula
h.x/ D C jxja=b
(4.19)
for some constant C . Clearly, (4.19) does not define a topological conjugacy for
a single constant C , because it fails to be injective on R; however, the formula
(
xjxja=b 1 if x 0
(4.20)
h.x/ D
0
if x D 0;
which is obtained from (4.19) by taking C D 1 for positive x and C D 1
for negative x, defines a homeomorphism if ab > 0. Even though the function
defined in (4.20) may fail to be differentiable at 0, substitution of it into
e t a h.x/ D h.e t b x/;
(4.21)
which is (4.13) for this example, shows that it does, in fact, define a topological
conjugacy when ab > 0. (Note that in no case is this a C 1 -conjugacy, since
either h0 .0/ or .h 1 /0 .0/ does not exist.)
Now, suppose that ab 0. Does a topological (possibly nondifferentiable)
conjugacy exist? If ab D 0, then (4.21) implies that h is constant, which violates
injectivity, so suppose that ab < 0. In this case, substituting x D 0 and t D 1
into (4.21) implies that h.0/ D 0. Fixing x 0 and letting t sgn b # 1 in
(4.21), we see that the continuity of h implies that h.x/ D 0, also, which again
violates injectivity.
Summarizing, for a b there is a topological conjugacy of (4.16) and (4.17)
if and only if ab > 0, and these are not C 1 -conjugacies.
105
4. C ONJUGACIES
An Example for Maps

Lets try a similar analysis for maps. Let a; b 2 R be distinct constants, and
consider the maps F .x/ D ax and G.x/ D bx (for x 2 R). For what .a; b/combinations does there exist a homeomorphism h W R ! R such that
F .h.x// D h.G.x//
(4.22)
for every x 2 R? Can h and h 1 be chosen to be differentiable?

For these specific maps, the general equation (4.22) becomes
ah.x/ D h.bx/:
(4.23)
If a D 0 or b D 0 or a D 1 or b D 1, then injectivity is immediately violated.

Note that, by induction, (4.23) gives
an h.x/ D h.b n x/
for every n 2 Z. In particular, a2 h.x/ D h.b 2 x/, so the cases when a D
b D 1 cause the same problems as when a D 1 or b D 1.
So, from now on, assume that a; b f 1; 0; 1g. Observe that:
(4.24)
1 or
Setting x D 0 in (4.23) yields h.0/ D 0.

If jbj < 1, then fixing x 0 in (4.24) and letting n " 1, we have jaj < 1.
If jbj > 1, we can, similarly, let n #
1 to conclude that jaj > 1.
If b > 0 and a < 0, then (4.23) implies that h.1/ and h.b/ have opposite
signs even though 1 and b have the same sign; consequently, the Intermediate Value Theorem yields a contradiction to injectivity.
If b < 0 and a > 0, then (4.23) gives a similar contradiction.
Thus, the only cases where we could possibly have conjugacy is if a and b
are both in the same component of
. 1; 1/ [ . 1; 0/ [ .0; 1/ [ .1; 1/:
When this condition is met, experimentation (or experience) suggests trying h of
the form h.x/ D xjxjp 1 for some constant p > 0 (with h.0/ D 0). This is a
homeomorphism from R to R, and plugging it into (4.23) shows that it provides
a conjugacy if a D bjbjp 1 or, in other words, if
pD
106
log jaj
:
log jbj
Smooth Conjugacies
Since a b, either h or h 1 fails to be differentiable at 0. Is there some other
formula that provides a C 1 -conjugacy? No, because if there were we could differentiate both sides of (4.23) with respect to x and evaluate at x D 0 to get
h0 .0/ D 0, which would mean that .h 1 /0 .0/ is undefined.
Exercise 16 Define F W R2 ! R2 by the formula

x
x=2
F
D
;
y
2y C x 2
and let A D DF .0/.
(a) Show that the maps F and A are topologically conjugate.
(b) Show that the flows generated by the differential equations
zP D F .z/
and
zP D Az
are topologically conjugate.
(Hint: Try quadratic conjugacy functions.)
4.7 Smooth Conjugacies

The examples we looked at last time showing that topological conjugacies often cannot be chosen to be differentiable all involved two maps or vector fields
with different linearizations at the origin. What about when, as in the HartmanGrobman Theorem, we are looking for a conjugacy between a map (or flow)
and its linearization? An example of Hartman shows that the conjugacy cannot
always be chosen to be C 1 .
Hartmans Example
Consider the system
8
<xP D x
yP D . /y C "xz
:
zP D z;
107
4. C ONJUGACIES
where > > 0 and " 0. We will not cut off this vector field but will instead
confine our attention to x; y; z small. A calculation shows that the time-1 map
F D '.1; / of this system is given by
02 3 1 2
3
x
ax
F @4y 5A D 4ac.y C "xz/5 ;
z
cz
where a D e and c D e . Note that a > ac > 1 > c > 0. The time-1 map B
of the linearization of the differential equation is given by
2 3 2
3
x
ax
B 4y 5 D 4acy 5 :
y
cz
A local conjugacy H D .f; g; h/ of B with F must satisfy
af .x; y; z/ D f .ax; acy; cz/
acg.x; y; z/ C "f .x; y; z/h.x; y; z/ D g.ax; acy; cz/

ch.x; y; z/ D h.ax; acy; cz/
for every x; y; z near 0. Writing k.x; z/ for k.x; 0; z/, where k 2 ff; g; hg, we
have
af .x; z/ D f .ax; cz/
acg.x; z/ C "f .x; z/h.x; z/ D g.ax; cz/

ch.x; z/ D h.ax; cz/
(4.25)
(4.26)
(4.27)
for every x; z near 0.

Before proceeding further, we state and prove a lemma.
Lemma. Suppose that j is a continuous real-valued function of a real variable,
defined on an open interval U centered at the origin. Suppose that there are
constants ; 2 R such that
j.u/ D j.u/
(4.28)
whenever u; u 2 U. Then if jj < 1 < jj or jj < 1 < jj, j.u/ D 0 for

every u 2 U.
Proof. If jj < 1 < jj, fix u 2 U and apply (4.28) inductively to get
108
n j.u/ D j. n u/
(4.29)
Smooth Conjugacies
for every n 2 N. Letting n " 1 in (4.29), we see that j.u/ must be zero. If
jj < 1 < jj, substitute v D u into (4.28) to get
j.
for every v;
1v
v/ D j.v/
(4.30)
2 U. Fix v 2 U, and iterate (4.30) to get

n j.
v/ D j.v/
(4.31)
for every n 2 N. Letting n " N in (4.31), we get j.v/ D 0.

Setting x D 0 in (4.25) and applying the Lemma gives
f .0; z/ D 0
(4.32)
for every z near zero. Setting z D 0 in (4.27) and applying the Lemma gives
h.x; 0/ D 0
(4.33)
for every x near zero. Setting x D 0 in (4.26), using (4.32), and applying the
Lemma gives
g.0; z/ D 0
(4.34)
for every z near zero. If we set z D 0 in (4.26), use (4.33), and then differentiate
both sides with respect to x, we get cgx .x; 0/ D gx .ax; 0/; applying the Lemma
yields
(4.35)
gx .x; 0/ D 0
for every x near zero. Setting z D 0 in (4.34) and using (4.35), we get
g.x; 0/ D 0
(4.36)
for every x near zero.

Now, using (4.26) and mathematical induction, it can be verified that
an c n g.x; z/ C n"f .x; z/h.x; z/ D g.an x; c n z/
(4.37)
for every n 2 N. Similarly, mathematical induction applied to (4.25) gives

f .x; z/ D a
f .an x; c n z/
(4.38)
for every n 2 N. If we substitute (4.38) into (4.37), divide through by c n , and

replace x by a n x we get
an g.a
x; z/ C n"f .x; c nz/h.a
x; z/ D c
g.x; c n z/
(4.39)
109
4. C ONJUGACIES
for every n 2 N.
The existence of gx .0; z/ and gz .x; 0/ along with equations (4.34) and (4.36)
imply that an g.a n x; z/ and c n g.x; c n z/ stay bounded as n " 1. Using this
fact, and letting n " 1 in (4.39), we get
f .x; 0/h.0; z/ D 0;
so f .x; 0/ D 0 or h.0; z/ D 0. If f .x; 0/ D 0, then, in combination with (4.33)
and (4.36), this tells us that H is not injective in a neighborhood of the origin.
Similarly, if h.0; z/ D 0 then, in combination with (4.32) and (4.34), this implies
a violation of injectivity, as well.
Poincares Linearization Theorem

Suppose that f W Rn ! Rn is analytic and satisfies f .0/ D 0. It is possible to
establish conditions under which there is an analytic change of variables that will
turn the nonlinear equation
xP D f .x/
(4.40)
into its linearization
uP D Df .0/u:
(4.41)
Definition. Let 1 ; 2 ; : : : ; n be the eigenvalues of Df .0/, listed according to

multiplicity. We say that Df .0/ is resonant if there are nonnegative integers
m1 ; m2 ; : : : ; mn and a number s 2 f1; 2; : : : ; ng such that
n
X
kD1
mk 2
and
s D
n
X
mk k :
kD1
If Df .0/ is not resonant, we say that it is nonresonant.

Note that in Hartmans example there is resonance. A study of normal forms
reveals that nonresonance permits us to make changes of variable that remove
nonlinear terms up to any specified order in the right-hand side of the differential equation. In order to be able to guarantee that all nonlinear terms may be
removed, some extra condition beyond nonresonance is required.
110
Smooth Conjugacies
Definition. We say that .1 ; 2 ; : : : ; n / 2 Cn satisfy a Siegel condition if there
are constants C > 0 and > 1 such that
X
C
mk k Pn
s
. kD1 mk /
kD1
for all nonnegative integers m1 ; m2 ; : : : ; mn satisfying

n
X
kD1
mk 2:
Theorem. (Poincares Linearization Theorem) Suppose that f is analytic, and

that all the eigenvalues of Df .0/ are nonresonant and either all lie in the open
left half-plane, all lie in the open right half-plane, or satisfy a Siegel condition.
Then there is a change of variables u D g.x/ that is analytic near 0 and that
turns (4.40) into (4.41) near 0.
111
5
Invariant Manifolds
5.1 Stable Manifold Theorem: Part 1
The Hartman-Grobman Theorem states that the flow generated by a smooth vector field in a neighborhood of a hyperbolic equilibrium point is topologically
conjugate with the flow generated by its linearization. Hartmans counterexample shows that, in general, the conjugacy cannot be taken to be C 1 . However, the
Stable Manifold Theorem will tell us that there are important structures for the
two flows that can be matched up by smooth changes of variable. In this section,
we will discuss the Stable Manifold Theorem on an informal level and discuss
two different approaches to proving it.
Let f W Rn ! Rn be C 1 , and let ' W R ! be the flow generated
by the differential equation
xP D f .x/:
(5.1)
Suppose that x0 is a hyperbolic equilibrium point of (5.1).
Definition. The (global) stable manifold of x0 is the set
n
o
W s .x0 / WD x 2 lim '.t; x/ D x0 :

t "1
Definition. The (global) unstable manifold of x0 is the set
n
o
W u .x0 / WD x 2 lim '.t; x/ D x0 :

t# 1
Definition. Given a neighborhood U of x0 , the local stable manifold of x0 (relative to U) is the set
n
o
s
Wloc
.x0 / WD x 2 U C .x/ U and lim '.t; x/ D x0 :
t "1
113
5. I NVARIANT M ANIFOLDS
Definition. Given a neighborhood U of x0 , the local unstable manifold of x0
(relative to U) is the set
n
o
u
.x0 / WD x 2 U .x/ U and lim '.t; x/ D x0 :
Wloc
t# 1
Note that:
s .x / W s .x /, and W u .x / W u .x /.
Wloc
0
0
0
loc 0
s
u
Wloc
.x0 / and Wloc
.x0 / are both nonempty, since they each contain x0 .
W s .x0 / and W u .x0 / are invariant sets.

s .x / is positively invariant, and W u .x / is negatively invariant.
Wloc
0
loc 0
s .x / is not necessarily W s .x / \ U, and W u .x / is not necessarily
Wloc
0
0
loc 0
W u .x0 / \ U.
s .x / is not necessarily invariant, since it might not be negatively invariWloc
0
u .x / is not necessarily invariant, since it might not be positively
ant, and Wloc
0
invariant. They do, however, possess what is known as relative invariance.
Definition. A subset A of a set B is positively invariant relative to B if for every

x 2 A and every t 0, '.t; x/ 2 A whenever '.0; t; x/ B.
Definition. A subset A of a set B is negatively invariant relative to B if for every
x 2 A and every t 0, '.t; x/ 2 A whenever '.t; 0; x/ B.
Definition. A subset A of a set B is invariant relative to B if it is negatively
invariant relative to B and positively invariant relative to B.
s .x / is negatively invariant relative to U and is therefore invariant relative
Wloc
0
u .x / is positively invariant relative to U and is therefore invariant
to U. Wloc
0
relative to U.
Recall that a (k-)manifold is a set that is locally homeomorphic to an open
s
subset of Rk . Although the word manifold appeared in the names of Wloc
.x0 /,
u
s
u
Wloc .x0 /, W .x0 /, and W .x0 /, it is not obvious from the defintions of these sets
that they are, indeed, manifolds. One of the consequences of the Stable Manifold
s .x / and W u .x / are manifolds.
Theorem is that, if U is sufficiently small, Wloc
0
loc 0
(W s .x0 / and W u .x0 / are what are known as immersed manifolds.)
For simplicity, lets now assume that x0 D 0. Let E s be the stable subspace
of Df .0/, and let E u be the unstable subspace of Df .0/. If f is linear, then
114
Stable Manifold Theorem: Part 1

W s .0/ D E s and W u .0/ D E u . The Stable Manifold Theorem says that in the
nonlinear case not only are the Stable and Unstable Manifolds indeed manifolds,
but they are tangent to E s and E u , respectively, at the origin. This is information
that the Hartman-Grobman Theorem does not provide.
More precisely there are neighborhoods U s of the origin in E s and U u of
the origin in E u and smooth maps hs W U s ! U u and hu W U u ! U s such
that hs .0/ D hu .0/ D 0 and Dhs .0/ D Dhu .0/ D 0 and the local stable and
unstable manifolds of 0 relative to U s U u satisfy

s
Wloc
.0/ D x C hs .x/ x 2 U s
and

u
.0/ D x C hu .x/ x 2 U u :
Wloc
Furthermore, not only do solutions of (5.1) in the stable manifold converge to 0

as t " 1, they do so exponentially quickly. (A similar statement can be made
about the unstable manifold.)
Liapunov-Perron Approach
This approach to proving the Stable Manifold Theorem rewrites (5.1) as
xP D Ax C g.x/;
(5.2)
where A D Df .0/. The Variation of Parameters formula gives

Z t2
.t2 t1 /A
x.t2 / D e
x.t1 / C
e .t2 s/A g.x.s// ds;
(5.3)
t1
for every t1 ; t2 2 R. Setting t1 D 0 and t2 D t, and projecting (5.3) onto E s

yields
Z t
tAs
xs .t/ D e xs .0/ C
e .t s/As gs .x.s// ds;
0
where the subscript s attached to a quantity denotes the projection of that quantity
onto E s . If we assume that the solution x.t/ lies on W s .0/, set t2 D t, let t1 " 1,
and project (5.3) onto E u , we get
Z 1
xu .t/ D
e .t s/Au gu .x.s// ds:
t
Hence, solutions of (5.2) in W s .0/ satisfy the integral equation

Z t
Z 1
x.t/ D e tAs xs .0/ C
e .t s/As gs .x.s// ds
e .t s/Au gu .x.s// ds:
0
115
Now, fix as 2 E s , and define a functional T by
Z t
Z
tAs
.t s/As
.T x/.t/ D e as C
e
gs .x.s// ds
0
e .t
s/Au
gu .x.s// ds:
A fixed point x of this functional will solve (5.2), will have a range contained in
the stable manifold, and will satisfy xs .0/ D as . If we set hs .as / D xu .0/ and
define hs similarly for other inputs, the graph of hs will be the stable manifold.
Hadamard Approach
The Hadamard approach uses what is known as a graph transform. Here we
define a functional not by an integral but by letting the graph of the input function
move with the flow ' and selecting the output function to be the function whose
graph is the image of the original graph after, say, 1 unit of time has elapsed.
More precisely, suppose h is a function from E s to E u . Define its graph
transform F h to be the function whose graph is the set

'.1; C h.// 2 E s :
(5.4)
(That (5.4) is the graph of a function from E s to E u if we identify E s E u

with E s E u is, of course, something that needs to be shown.) Another way of
putting this is that for each 2 E s ,
F h..'.1; C h.///s / D .'.1; C h.///u I
in other words,
F h s '.1; / .id Ch/ D u '.1; / .id Ch/;
where s and u are projections onto E s and E u , respectively. A fixed point

of the graph transform functional F will be an invariant manifold, and it can be
show that it is, in fact, the stable manifold.

Statements
Given a normed vector space X and a positive number r, we let X .r/ stand for
the closed ball of radius r centered at 0 in X .
The first theorem refers to the differential equation
116
xP D f .x/:
(5.5)

Theorem. (Stable Manifold Theorem) Suppose that is an open neighborhood of the origin in Rn , and f W ! Rn is a C k function (k 1) such that
0 is a hyperbolic equilibrium point of (5.5). Let E s E u be the decomposition
of Rn corresponding to the matrix Df .0/. Then there is a norm k k on Rn ,
a number r > 0, and a C k function h W E s .r/ ! E u .r/ such that h.0/ D 0
s .0/ of 0 relative to
and Dh.0/ D 0 and such that the local stable manifold Wloc
s
u
B.r/ WD E .r/ E .r/ is the set

vs C h.vs / vs 2 E s .r/ :
Moreover, there is a constant c > 0 such that
n
o
s
Wloc
.0/ D v 2 B.r/ C .v/ B.r/ and lim e ct '.t; v/ D 0 :
t "1
Two immediate and obvious corollaries, which we will not state explicitly,
describe the stable manifolds of other equilibrium points (via translation) and
describe unstable manifolds (by time reversal).
We will actually prove this theorem by first proving an analogous theorem
for maps (much as we did with the Hartman-Grobman Theorem). Given a neighborhood U of a fixed point p of a map F , we can define the local stable manifold
of p (relative to U) as
n
o
s
Wloc
.p/ WD x 2 U F j .x/ 2 U for every j 2 N and lim F j .x/ D p :
j "1
Theorem. (Stable Manifold Theorem for Maps) Suppose that is an open

neighborhood of the origin in Rn , and F W ! is an invertible C k function
(k 1) for which F .0/ D 0 and the matrix DF .0/ is hyperbolic and invertible.
Let E s E u .D E E C / be the decomposition of Rn corresponding to the matrix
DF .0/. Then there is a norm k k on Rn , a number r > 0, a number Q 2 .0; 1/,
and a C k function h W E s .r/ ! E u .r/ such that h.0/ D 0 and Dh.0/ D 0 and
s .0/ of 0 relative to B.r/ WD E s .r/E u .r/
such that the local stable manifold Wloc
satisfies

s
Wloc
.0/ D vs C h.vs / vs 2 E s .r/
n
o
D v 2 B.r/ F j .v/ 2 B.r/ for every j 2 N
n
o
D v 2 B.r/ F j .v/ 2 B.r/ and kF j .v/k Q j kvk for all j 2 N :
Preliminaries
The proof of the Stable Manifold Theorem for Maps will be broken up into a
series of lemmas. Before stating and proving those lemmas, we need to lay a
117
foundation by introducing some terminology and notation and by choosing some
constants.
We know that F .0/ D 0 and DF .0/ is hyperbolic. Then Rn D E s E u , s
and u are the corresponding projection operators, E s and E u are invariant under
DF .0/, and there are constants < 1 and > 1 such that all of the eigenvalues
of DF .0/jE s have magnitude less than and all of the eigenvalues of DF .0/jE u
have magnitude greater than .
When we deal with a matrix representation of DF .q/, it will be with respect
to a basis that consists of a basis for E s followed by a basis for E u . Thus,
3
2
Ass .q/ Asu .q/
5;
DF .q/ D 4
Aus .q/ Auu .q/
where, for example, Asu .q/ is a matrix representation of s DF .q/jE u in terms
of the basis for E u and the basis for E s . Note that, by invariance, Asu .0/ D
Aus .0/ D 0. Furthermore, we can pick our basis vectors so that, with k k being
the corresponding Euclidean norm of a vector in E s or in E u ,
kAss .0/vs k
<
kvs k
vs 0
kAss .0/k WD sup

and
kAuu .0/vu k
> :
kvu k
vu 0
m.Auu .0// WD inf
(The functional m./ defined implicitly in the last formula is sometimes called
the minimum norm even though it is not a norm.) For a vector in v 2 Rn , let
kvk D maxfks vk; ku vkg. This will be the norm on Rn that will be used
throughout the proof. Note that B.r/ WD E s .r/ E u .r/ is the closed ball of
radius r in Rn by this norm.
Next, we choose r. Fix > 0. Pick " > 0 small enough that
C " C " < 1 <
"=
Pick r > 0 small enough that if q 2 B.r/ then

kAss .q/k < ;
m.Auu .q// > ;

kAsu .q/k < ";
118
kDF .q/
kAus .q/k < ";

DF .0/k < ";
2":

and DF .q/ is invertible. (We can do this since F is C 1 , so DF ./ is continuous.)
Now, define
1
\
s
Wr WD
F j .B.r//;
j D0
Wrs
and note that

is the set of all points in B.r/ that produce forward semiorbits
(under the discrete dynamical system generated by F ) that stay in B.r/ for all
s .0/ W s ; we will show that these two sets
forward iterates. By definition, Wloc
r
are, in fact, equal.
Two other types of geometric sets play vital roles in the proof: cones and
disks. The cones are of two types: stable and unstable. The stable cone (of
slope ) is

C s ./ WD v 2 Rn ku vk ks vk ;
and the unstable cone (of slope ) is

C u ./ WD v 2 Rn ku vk ks vk :
An unstable disk is a set of the form
vu C .vu / vu 2 E u .r/
for some Lipschitz continuous function

stant (less than or equal to) 1 .
W E u .r/ ! E s .r/ with Lipschitz con-

The Action of DF .p/ on the Unstable Cone
The first lemma shows that if the derivative of the map is applied to a point in the
unstable cone, the image is also in the unstable cone.
Lemma. (Linear Invariance of the Unstable Cone) If p 2 B.r/, then
DF .p/C u ./ C u ./:
Proof. Let p 2 B.r/ and v 2 C u ./. Then, if we let vs D s v and vu D u v,
we have kvu k kvs k, so
ku DF .p/vk D kAus .p/vs C Auu .p/vu k
kAuu .p/vu k
kAus .p/vs k
m.Auu .p//kvu k
.
"=/kvu k;
kAus .p/kkvs k kvu k
"kvs k
119
and
ks DF .p/vk D kAss .p/vs C Asu .p/vu k kAss .p/vs k C kAsu .p/vu k
kAss .p/kkvs k C kAsu .p/kkvu k kvs k C "kvu k
.= C "/kvu k:
Since
"= .= C "/,

ku DF .p/vk ks DF .p/vk;
so DF .p/v 2 C u ./.
The Action of F on Moving Unstable Cones

The main part of the second lemma is that moving unstable cones are positively
invariant. More precisely, if two points are in B.r/ and one of the two points is
in a translate of the unstable cone that is centered at the second point, then their
images under F satisfy the same relationship. The lemma also provides estimates
on the rates at which the stable and unstable parts of the difference between the
two points contract or expand, respectively.
In this lemma (and later) we use the convention that if X and Y are subsets
of a vector space, then

X C Y WD x C y x 2 X and y 2 Y :
Lemma. (Moving Unstable Cones) If p; q 2 B.r/ and q 2 fpg C C u ./, then:

(a) ks .F .q/
F .p//k .= C "/ku .q
(b) ku .F .q/
F .p//k .
"=
p/k;
"/ku .q
p/k;
(c) F .q/ 2 fF .p/g C C u ./.

Proof. We will write differences as integrals (using the Fundamental Theorem
of Calculus) and use our estimates on DF .v/, for v 2 B.r/, to estimate these
integrals.
120

Since B.r/ is convex,
Z

F .p//k D

d
s F .tq C .1
0 dt

t/p/.q p/ dt

ks .F .q/
Z 1

s DF .tq C .1
D

0
Z 1

D
Ass .tq C .1 t/p/s .q

0

t/p/ dt

p/ dt
Asu .tq C .1
t/p/u .q
kAss .tq C .1 t/p/kks .q

Z 1
C
kAsu .tq C .1
p/k dt

p/ dt

t/p/kku .q
p/k dt
ks .q
0
p/k C "ku .q
p/k dt .= C "/ku .q
p/k:
This gives (a).

Similarly,
ku .F .q/ F .p//k
Z 1

D
Aus .tq C .1 t/p/s .q p/ dt

0

Z 1

C
Auu .tq C .1 t/p/u .q p/ dt

0
Z 1
Z 1

Auu .0/u .q p/ dt

Aus .tq C .1 t/p/s .q p/ dt
0
0
Z 1

.Auu .tq C .1 t/p/ Auu .0//u .q p/ dt

0
Z 1
m.Auu .0//ku .q p/k
kAus .tq C .1 t/p/kks .q p/k dt
0
Z 1
kAuu .tq C .1 t/p/ Auu .0/kku .q p/k dt
0
ku .q
.
"=
p/k
"ks .q
"/ku .q
p/k
"ku .q
p/k
p/k:
This gives (b).

121
From (a), (b), and the choice of ", we have
ku .F .q/
F .p//k .
"=
"/ku .q
. C "/ku .q
ks .F .q/
so F .q/
p/k
p/k
F .p//k;
F .p/ 2 C u ./, which means that (c) holds.

Stretching of C 1 Unstable Disks
The next lemma shows that if F is applied to a C 1 unstable disk (i.e., an unstable
disk that is the graph of a C 1 function), then part of the image gets stretched out
of B.r/, but the part that remains in is again a C 1 unstable disk.
Lemma. (Unstable Disks) Let D0 be a C 1 unstable disk, and recursively define
Dj D F .Dj
1/
\ B.r/
for each j 2 N. Then each Dj is a C 1 unstable disk, and

0
1
j
\
F i .Di /A 2. "= "/
diam @u
(5.6)
iD0
for each j 2 N.
Proof. Because of induction, we only need to handle the case j D 1. The estimate on the diameter of the u projection of the preimage of D1 under F is
a consequence of part (b) of the lemma on moving invariant cones. That D1 is
the graph of an 1 -Lipschitz function 1 from a subset of E u .r/ to E s .r/ is a
consequence of part (c) of that same lemma. Thus, all we need to show is that
dom. 1 / D E u .r/ and that 1 is C 1 .
Let 0 W E u .r/ ! E s .r/ be the C 1 function (with Lipschitz constant less
than or equal to 1 ) such that

D0 D vu C 0 .vu / vu 2 E u .r/ :
Define g W E u .r/ ! E u by the formula g.vu / D u F .vu C 0 .vu //. If we can
show that for each y 2 E u .r/ there exists x 2 E u .r/ such that
122
g.x/ D y;
(5.7)

then we will know that dom. 1 / D E u .r/.
Let y 2 E u .r/ be given. Let L D Auu .0/. Since m.L/ > , we know
that L 1 2 L.E u ; E u / exists and that kL 1 k 1=. Define G W E u .r/ ! E u
by the formula G.x/ D x L 1 .g.x/ y/, and note that fixed points of G
are solutions of (5.7), and vice versa. We shall show that G is a contraction and
takes the compact set E u .r/ into itself and that, therefore, (5.7) has a solution
x 2 E u .r/.
Note that
Dg.x/ D u DF .x C
so
D Auu .x C
0 .x//.I
0 .x// C
CD
0 .x//
Aus .x C
0 .x//D
0 .x/;
kDG.x/k D kI L 1 Dg.x/k kL 1 kkL Dg.x/k

1
.kAuu .x C 0 .x// Auu .0/k

C kAus .x C 0 .x//kkD 0 .x/k/
" C "=

< 1:

The Mean Value Theorem then implies that G is a contraction.
Now, suppose that x 2 E u .r/. Then
kG.x/k kG.0/k C kG.x/
kL
G.0/k
k.kg.0/k C kyk/ C
" C "=
kxk

1
.kg.0/k C r C ." C "=/r/:

Let W E s .r/ ! E u .r/ be defined by the formula .vs / D u F .vs /. Since
.0/ D 0 and, for any vs 2 E s .r/, kD.vs /k D kAus .vs /k ", the Mean Value
Theorem tells us that

kg.0/k D ku F .
0 .0//k
D k.
0 .0//k
"k
0 .0/k
"r:
(5.8)
Plugging (5.8) into the previous estimate, we see that

kG.x/k
1
1 C "= C 2"
."r C r C ." C "=/r/ D
r < r;

so G.x/ 2 E u .r/.
That completes the verification that (5.7) has a solution for each y 2 E u .r/
and, therefore, that dom. 1 / D E u .r/. To finish the proof, we need to show that
1
Q be the restriction of g to g 1 .D1 /, and observe that
1 is C . Let g
1
gQ D s F .I C
0 /:
(5.9)
123
We have shown that gQ is a bijection of g 1 .D1 / with D1 and, by the Inverse
Function Theorem, gQ 1 is C 1 . Thus, if we rewrite (5.9) as
1
we can see that
1,
D s F .I C
0/
gQ
as the composition of C 1 functions, is indeed C 1 .
Wrs is a Lipschitz Manifold

Recall that Wrs was defined to be all points in the box B.r/ that produced forward
orbits that remain confined within B.r/. The next lemma shows that this set is a
manifold.
Lemma. (Nature of Wrs ) Wrs is the graph of a function h W E s .r/ ! E u .r/ that
satisfies h.0/ D 0 and that has a Lipschitz constant less than or equal to .
Proof. For each vs 2 E u .r/, consider the set
D WD fvs g C E u .r/:
D is a C 1 unstable disk, so by the lemma on unstable disks, the subset Sj of D
that stays in B.r/ for at least j iterations of F has a diameter less than or equal to
2. "= "/ j r. By the continuity of F , Sj is closed. Hence, the subset S1 of
D that stays in B.r/ for an unlimited number of iterations of F is the intersection
of a nested collection of closed sets whose diameters approach 0. This means
that S1 is a singleton. Call the single point in S1 h.vs /.
It should be clear that Wrs is the graph of h. That h.0/ D 0 follows from the
fact that 0 2 Wrs , since F .0/ D 0. If h werent -Lipschitz, then there would
be two points p; q 2 Wrs such that p 2 fqg C C u ./. Repeated application of
parts (b) and (c) of the lemma on moving unstable cones would imply that either
F j .p/ or F j .q/ is outside of B.r/ for some j 2 N, contrary to definition.
s
Wloc
.0/ is a Lipschitz Manifold
s
Our next lemma shows that Wloc
.0/ D Wrs and that, in fact, orbits in this set
converge to 0 exponentially. (The constant Q in the statement of the theorem can
be chosen to be C " if 1.)
Lemma. (Exponential Decay) If 1, then for each p 2 Wrs ,

kF j .p/k . C "/j kpk:
124
s .0/.
In particular, Wrs D Wloc
(5.10)

Proof. Suppose that 1 and p 2 Wrs . By mathematical induction (and the
positive invariance of Wrs ), it suffices to verify (5.10) for j D 1. Since 1
the last lemma implies that Wrs is the graph of a 1-Lipschitz function. Since
F .p/ 2 Wrs , we therefore know that
kF .p/k D max fks F .p/k; ku F .p/kg D ks F .p/k:
Using this and estimating, we find that
Z 1
Z 1

d

kF .p/k D
s F .tp/ dt D
s DF .tp/p dt

0 dt
0
Z 1

D
Ass .tp/s p C Asu .tp/u p dt

0
1
kAss .tp/kks pk C kAsu .tp/kku pk dt
ks pk C "ku pk . C "/kpk:

s
Wloc
.0/ is C 1
Lemma. (Differentiability) The function h W E s .r/ ! E u .r/ for which

s
Wloc
.0/ D vs C h.vs / vs 2 E s .r/
is C 1 , and Dh.0/ D 0.
Proof. Let q 2 Wrs be given. We will first come up with a candidate for a plane
that is tangent to Wrs at q, and then we will show that it really works.
For each j 2 N and each p 2 Wrs , define
C s;j .p/ WD D.F j /.p/
C s ./;
and let
C s;0 .p/ WD C s ./:
By definition (and by the invertibility of DF .v/ for all v 2 B.r/), C s;j .p/ is the
image of the stable cone under an invertible linear transformation. Note that
C s;1 .p/ D DF .p/
C s ./ C s ./ D C s;0 .p/
125
by the (proof of the) lemma on linear invariance of the unstable cone. Similarly,
C s;2 .p/ D D.F 2 /.p/
D DF .p/
C s ./ D DF .F .p//DF .p/
1
DF .F .p//
DF .p/
C s;3 .p/ D D.F 3 /.p/
C s ./ D DF .p/
C s ./
C s;1 .F .p//
C s ./ D C s;1 .p/
and
C s ./ D DF .F 2 .p//DF .F .p//DF .p/
D DF .p/
DF .F .p//
DF .F 2 .p//
DF .F .p//
C s;1 .F 2 .p//
DF .p/
DF .F .p//
D DF .p/
C s ./
C s ./
C s ./ D C s;2 .p/:
Recursively, we find that, in particular,

C s;0 .q/ C s;1 .q/ C s;2 .q/ C s;3 .q/ :
The plane that we will show is the tangent plane to Wrs at q is the intersection
C
s;1
.q/ WD
1
\
C s;j .q/
j D0
of this nested sequence of cones.

First, we need to show that this intersection is a plane. Suppose that x 2
s;j
C .q/. Then x 2 C s ./, so
ks DF .q/xk D kAss .q/s x C Asu .q/u xk
kAss .q/kks xk C kAsu .q/kku xk . C "/ks xk:
Repeating this sort of estimate, we find that

ks D.F j /.q/xk D ks DF .F j
.q//DF .F j
. C "/j ks xk:
.q// DF .q/xk
On the other hand, if y is also in C s;j .q/ and s x D s y, then repeated applications of the estimates in the lemma on linear invariance of the unstable cone
yield
ku D.F j /.q/x
u D.F j /.q/yk .
"=/j ku x
u yk:
Since D.F j /.q/C s;j .q/ D C s ./, it must, therefore, be the case that
.
126
"=/j ku x u yk
2:
. C "/j ks xk

This implies that
ku x
u yk 2
C "
"=
j
ks xk:
(5.11)
Letting j " 1 in (5.11), we see that for each vs 2 E s there can be no more
than 1 point x in C s;1 .q/ satisfying s x D vs . On the other hand, each
C s;j .q/ contains a plane of dimension dim.E s / (namely, the preimage of E s
under D.F j /.q/), so (since the set of planes of that dimension passing through
the origin is a compact set in the natural topology), C s;1 .q/ contains a plane, as
well. This means that C s;1 .q/ is a plane Pq that is the graph of a linear function
Lq W E s ! E u .
Before we show that Lq D Dh.q/, we make a few remarks.
(a) Because E s C s;j .0/ for every j 2 N, P0 D E s and L0 D 0.
(b) The estimate (5.11) shows that the size of the largest angle between two
vectors in C s;j .q/ having the same projection onto E s goes to zero as
j " 1.
(c) Also, the estimates in the proof of the lemma on linear invariance of the
unstable cone show that the size of the minimal angle between a vector in
C s;1 .F j .q// and a vector outside of C s;0 .F j .q// is bounded away from
zero. Since
C s;j .q/ D D.F j /.q/
C s ./ D D.F j /.q/
C s;0 .F j .q//
and
C s;j C1 .q/ D D.F j C1 /.q/
D D.F j /.q/
D D.F j /.q/
C s ./
DF .F j .q//
C s;1 .F j .q//;
C s ./
this also means that the size of the minimal angle between a vector in
C s;j C1 .q/ and a vector outside of C s;j .q/ is bounded away from zero
(for fixed j ).
(d) Thus, since C s;j C1 .q/ depends continuously on q,
Pq 0 2 C s;j C1 .q 0 / C s;j .q/
for a given j if q 0 is sufficiently close to q. This means that Pq depends
continuously on q.
127
Now, we show that DF .q/ D Lq . Let " > 0 be given. By remark (b) above,
we can choose j 2 N such that
ku v
Lq s vk "ks vk
(5.12)
whenever v 2 C s;j .q/. By remark (c) above, we know that we can choose "0 > 0
such that if w 2 C s;j C1 .q/ and krk "0 kwk, then w C r 2 C s;j .q/. Because
of the differentiability of F j 1 , we can choose > 0 such that
kF
j 1
.F j C1 .q/ C v/ q
D.F
j 1
/.F j C1 .q//vk
"0
kvk
kD.F j C1 /.q/k
(5.13)
whenever kvk . Define the truncated stable cone

C s .; / WD C s ./ \ s 1 E s ./:
From the continuity of F and the -Lipschitz continuity of h, we know that we
can pick > 0 such that
F j C1 .vs C h.vs // 2 fF j C1 .q/g C C s .; /:
(5.14)
whenever kvs s qk < .

Now, suppose that v 2 C s .; /. Then (assuming 1) we know that
kvk , so (5.13) tells us that
F
j 1
.F j C1 .q/ C v/ D q C D.F
j 1
/.F j C1 .q//v C r
D q C D.F j C1 /.q/
for some r satisfying

krk
Let w D D.F j C1 /.q/
1 v.
"0
kD.F j C1 /.q/k
vCr
(5.15)
kvk:
Since v 2 C s ./, w 2 C s;j C1 .q/. Also,
kwk D kD.F j C1 /.q/ 1 vk m.D.F j C1 /.q/

kvk
D
;
kD.F j C1 /.q/k
/kvk
so krk "0 kwk. Thus, by the choice of "0 , w C r 2 C s;j .q/ . Consequently,
(5.15) implies that
F
j 1
.F j C1 .q/ C v/ 2 fqg C C s;j .q/:
Since v was an arbitrary element of C s .; /, we have

F
128
j 1
.fF j C1 .q/g C C s .; // fqg C C s;j .q/:
(5.16)

Set qs WD s q, and suppose that vs 2 E s .r/ satisfies kvs
(5.14),
F j C1 .vs C h.vs // 2 fF j C1 .q/g C C s .; /:
qs k . By
This, the invertibility of F , and (5.16) imply

vs C h.vs / 2 fqg C C s;j .q/;
or, in other words,
vs C h.vs /
qs
h.qs / 2 C s;j .q/:
The estimate (5.12) then tells us that

kh.vs /
h.qs /
Lq .vs
qs /k "kvs
qs k;
which proves that Dh.q/ D Lq (since " was arbitrary).

Remark (d) above implies that Dh.q/ depends continuously on q, so h 2 C 1 .
Remark (a) above implies that Dh.0/ D 0.

Higher Differentiability
Lemma. (Higher Differentiability) If F is C k , then h is C k .
Proof. Weve already seen that this holds for k D 1. We show that it is true for
all k by induction. Let k 2, and assume that the lemma works for k 1. Define
a new map H W Rn Rn ! Rn Rn by the formula

p
F .p/
H
WD
:
v
DF .p/v
Since F is C k , H is C k 1 . Note that

p
F .F .p//
F 2 .p/
2
H
D
D
;
v
DF .F .p//DF .p/v
D.F 2 /.p/v

p
F .F 2 .p//
F 3 .p/
H
D
D
;
v
DF .F 2 .p//D.F 2 /.p/v
D.F 3 /.p/v
3
and, in general,

p
F j .p/
H
D
:
v
D.F j /.p/v
j
129
Also,
so
3
2

DF .p/
0
p
5;
DH
D4
v
D 2 F .p/v DF .p/
2
3

DF
.0/
0
0
5;
DH
D4
0
0
DF .0/
which is hyperbolic and invertible, since DF .0/ is. Applying the induction hypothesis, we can conclude that the fixed point of H at the origin (in Rn Rn ) has
a local stable manifold W that is C k 1 .
Fix q 2 Wrs , and note that F j .q/ ! 0 as j " 1 and
n
o
Pq D v 2 Rn lim D.F j /.q/v D 0 :

j "1
This means that
Since W has a C k

)
q
Pq D v 2 R
2W :
v
(
dependence on q, so does Pq . Hence, h is C k .
Flows
Now we discuss how the Stable Manifold Theorem for maps implies the Stable
Manifold Theorem for flows. Given f W ! Rn satisfying f .0/ D 0, let
F D '.1; /, where ' is the flow generated by the differential equation
xP D f .x/:
(5.17)
If f is C k , so is F . Clearly, F is invertible and F .0/ D 0. Our earlier discussion

on differentiation with respect to initial conditions tells us that
d
Dx '.t; x/ D Df .'.t; x//Dx '.t; x/
dt
and Dx '.0; x/ D I , where Dx represents differentiation with respect to x. Setting
g.t/ D Dx '.t; x/jxD0 ;
this implies, in particular, that
130
d
g.t/ D Df .0/g.t/
dt

and g.0/ D I , so
g.t/ D e tDf .0/:
Setting t D 1, we see that

e Df .0/ D g.1/ D Dx '.1; x/jxD0 D Dx F .x/jxD0 D DF .0/:
Thus, DF .0/ is invertible, and if (5.17) has a hyperbolic equilibrium at the origin
then DF .0/ is hyperbolic.
Since F satisfies the hypotheses of the Stable Manifold Theorem for maps,
we know that F has a local stable manifold Wrs on some box B.r/. Assume that
< 1 and that r is small enough that the vector field of (5.17) points into B.r/
on C s ./ \ @B.r/. (See the estimates in Section 3.4.) The requirements for a
point to be in Wrs are no more restrictive then the requirements to be in the local
stable manifold Wrs of the origin with respect to the flow, so Wrs Wrs .
We claim that, in fact, these two sets are equal. Suppose they are not. Then
there is a point q 2 Wrs n Wrs . Let x.t/ be the solution of (5.17) satisfying
x.0/ D q. Since limj "1 F j .q/ D 0 and, in a neighborhood of the origin, there
is a bound on the factor by which x.t/ can grow in 1 unit of time, we know that
x.t/ ! 0 as t " 1. Among other things, this implies that
(a) x.t/ Wrs for some t > 0, and
(b) x.t/ 2 Wrs for all t sufficiently large.
Since Wrs is a closed set and x is continuous, (a) and (b) say that we can pick t0
to be the earliest time such that x.t/ 2 Wrs for every t t0 .
Now, consider the location of x.t/ for t in the interval t0 1; t0 /. Since
x.0/ 2 Wrs , we know that x.j / 2 Wrs for every j 2 N. In particular, we can
choose t1 2 t0 1; t0 / such that x.t1 / 2 Wrs . By definition of t0 , we can choose
t2 2 .t1 ; t0 / such that x.t2 / Wrs . By the continuity of x and the closedness of
Wrs , we can pick t3 to the be the last time before t2 such that x.t3 / 2 Wrs . By
definition of Wrs , if t 2 t0 1; t0 / and x.t/ Wrs , then x.t/ B.r/; hence,
x.t/ must leave B.r/ at time t D t3 . But this contradicts the fact that the vector
field points into B.r/ at x.t3 /, since x.t3 / 2 C s ./ \ @B.r/. This contradiction
implies that no point q 2 Wrs n Wrs exists; i.e., Wrs D Wrs .
The exponential decay of solutions of the flow on the local stable manifold is
a consequence of the similar decay estimate for the map, along with the observation that, near 0, there is a bound to the factor by which a solution can grown in
1 unit of time.
131
5.7 Center Manifolds

Definition
Recall that for the linear differential equation
xP D Ax
(5.18)
the corresponding invariant subspaces E u , E s , and E c had the characterizations
o
[n
Eu D
t# 1
c>0
o
[n
n
x 2 R lim je ct e tA xj D 0 ;
E D
s
t "1
c>0
and
Ec D
\n
c>0
x 2 Rn lim je ct e tA xj D lim je
t# 1
t "1
o
e xj D 0 :
ct tA
The Stable Manifold Theorem tells us that for the nonlinear differential equation
xP D f .x/;
(5.19)
with f .0/ D 0, the stable manifold W s .0/ and the unstable manifold W u .0/
have characterizations similar to E s and E u , respectively:
o
[n
s
n
x 2 R lim je ct '.t; x/j D 0 ;
W .0/ D
c>0
and
W u .0/ D
t "1
[n
x 2 Rn lim je
c>0
t# 1
ct
o
'.t; x/j D 0 ;
where ' is the flow generated by (5.19). (This was only verified when the equilibrium point at the origin was hyperbolic, but a similar result holds in general.)
Is there a useful way to modify the characterization of E c similarly to get
a characterization of a center manifold W c .0/? Not really. The main problem
is that the characterizations of E s and E u only depend on the local behavior of
solutions when they are near the origin, but the characterization of E c depends
on the behavior of solutions that are, possibly, far from 0.
Still, the idea of a center manifold as some sort of nonlinear analogue of
E c .0/ is useful. Heres one widely-used definition:
Definition. Let A D Df .0/. A center manifold W c .0/ of the equilbrium point 0
of (5.19) is an invariant manifold whose dimension equals the dimension of the
invariant subspace E c of (5.18) and which is tangent to E c at the origin.
132
Center Manifolds
Nonuniqueness
While the fact that stable and unstable manifolds are really manifolds is a theorem (namely, the Stable Manifold Theorem), a center manifold is a manifold by
definition. Also, note that we refer to the stable manifold and the unstable manifold, but we refer to a center manifold. This is because center manifolds are not
necessarily unique. An extremely simple example of nonuniqueness (commonly
credited to Kelley) is the planar system
(
xP D x 2
yP D y:
Clearly, E c is the x-axis, and solving the system explicitly reveals that for any
constant c 2 R the curve

.x; y/ 2 R2 x < 0 and y D ce 1=x [ .x; 0/ 2 R2 x 0
is a center manifold.
Existence
There is a Center Manifold Theorem just like there was a Stable Manifold Theorem. However, the goal of the Center Manifold Theorem is not to characterize
a center manifold; that is done by the definition. The Center Manifold Theorem
asserts the existence of a center manifold.
We will not state this theorem precisely nor prove it, but we can give some
indication how the proof of existence of a center manifold might go. Suppose
that none of the eigenvalues of Df .0/ have real part equal to , where is a
given real number. Then we can split the eigenvalues up into two sets: Those
with real part less than and those with real part greater than . Let E be the
vector space spanned by the generalized eigenvectors corresponding to the first
set of eigenvalues, and let E C be the vector space spanned by the generalized
eigenvectors corresponding to the second set of eigenvalues. If we cut off f so
that it is stays nearly linear throughout Rn , then an analysis very much like that in
the proof of the Stable Manifold Theorem can be done to conclude that there are
invariant manifolds called the pseudo-stable manifold and the pseudo-unstable
manifold that are tangent, respectively, to E and E C at the origin. Solutions
x.t/ in the first manifold satisfy e t x.t/ ! 0 as t " 1, and solutions in the
second manifold satisfy e t x.t/ ! 0 as t # 1.
Now, suppose that is chosen to be negative but larger than the real part
of the eigenvalues with negative real part. The corresponding pseudo-unstable
manifold is called a center-unstable manifold and is written W cu .0/. If, on the
133
other hand, we choose to be between zero and all the positive real parts of
eigenvalues, then the resulting pseudo-stable manifold is called a center-stable
manifold and is written W cs .0/. It turns out that
W c .0/ WD W cs .0/ \ W cu .0/
is a center manifold.
Center Manifold as a Graph

Since a center manifold W c .0/ is tangent to E c at the origin it can, at least locally,
be represented as the graph of a function h W E c ! E s E u . Suppose, for
simplicity, that (5.19) can be rewritten in the form
(
xP D Ax C F .x; y/
(5.20)
yP D By C G.x; y/;
where x 2 E c , y 2 E s E u , the eigenvalues of A all have zero real part, all of
the eigenvalues of B have nonzero real part, and F and G are higher order terms.
Then, for points x C y lying on W c .0/, y D h.x/. Inserting that into (5.20) and
using the chain rule, we get
Dh.x/Ax C F .x; h.x// D Dh.x/xP D yP D Bh.x/ C G.x; h.x//:
Thus, if we define an operator M by the formula
.M/.x/ WD D.x/Ax C F .x; .x//
B.x/
G.x; .x//;
the function h whose graph is the center manifold is a solution of the equation
Mh D 0.
5.8 Computing and Using Center Manifolds

Approximation
Recall that we projected our equation onto E c and onto E s E u to get the system
(
xP D Ax C F .x; y/
(5.21)
yP D By C G.x; y/;
and that we were looking for a function h W E c ! E s E u satisfying .Mh/ 0,
where
134
.M/.x/ WD D.x/Ax C F .x; .x//
B.x/
G.x; .x//:
Computing and Using Center Manifolds

Except in the simplest of cases we have no hope of trying to get an explicit
formula for h, but because of the following theorem of Carr we can approximate
h to arbitrarily high orders.
Theorem. (Carr) Let be a C 1 mapping of a neighborhood of the origin in Rn
into Rn that satisfies .0/ D 0 and D.0/ D 0. Suppose that
.M/.x/ D O.jxjq /
as x ! 0 for some constant q > 1. Then
jh.x/
.x/j D O.jxjq /
as x ! 0.
Stability
If we put y D h.x/ in the first equation in (5.20), we get the reduced equation
xP D Ax C F .x; h.x//;
(5.22)
which describes the evolution of the E c coordinate of solutions on the center

manifold. Another theorem of Carrs states that if all the eigenvalues of Df .0/
are in the closed left half-plane, then the stability type of the origin as an equilibrium solution of (5.21) (Lyapunov stable, asymptotically stable, or unstable)
matches the stability type of the origin as an equilibrium solution of (5.22).
These results of Carr are sometimes useful in computing the stability type of
the origin. Consider, for example, the following system:
(
xP D xy C ax 3 C by 2 x
yP D y C cx 2 C dx 2 y;
where x and y are real variables and a, b, c, and d are real parameters. We know
that there is a center manifold, tangent to the x-axis at the origin, that is (locally)
of the form y D h.x/. The reduced equation on the center manifold is
xP D xh.x/ C ax 3 C bh.x/2 x:
(5.23)
To determine the stability of the origin in (5.23) (and, therefore, in the original system) we need to approximate h. Therefore, we consider the operator M
defined by
.M/.x/ D 0 .x/x.x/ C ax 3 C b..x//2 x C .x/
cx 2
dx 2 .x/;
135
and seek polynomial (satisfying .0/ D 0 .0/ D 0) for which the quantity
.M/.x/ is of high order in x. By inspection, if .x/ D cx 2 then .M/.x/ D
O.x 4 /, so h.x/ D cx 2 C O.x 4 /, and (5.23) becomes
xP D .a C c/x 3 C O.x 5 /:
Hence, the origin is asymptotically stable if aCc < 0 and is unstable if aCc > 0.
What about the borderline case when a C c D 0? Suppose that a C c D 0 and
lets go back and try a different , namely, one of the form .x/ D cx 2 C kx 4 .
Plugging this in, we find that .M/.x/ D .k cd /x 4 C O.x 6 /, so if we choose
k D cd then .M/.x/ D O.x 6 /; thus, h.x/ D cx 2 C cdx 4 C O.x 6 /. Inserting
this in (5.23), we get
xP D .cd C bc 2 /x 5 C O.x 7 /;
so the origin is asymptotically stable if cd C bc 2 < 0 (and a C c D 0) and is
unstable if cd C bc 2 > 0 (and a C c D 0).
What if aCc D 0 and cd Cbc 2 D 0? Suppose that these two conditions hold,
and consider of the form .x/ D cx 2 C cdx 4 C kx 6 for some k 2 R yet to be
determined. Calculating, we discover that .M/.x/ D .k b 2 c 3 /x 6 C O.x 8 /,
so by choosing k D b 2 c 3 , we see that h.x/ D cx 2 C cdx 4 C b 2 c 3 x 6 C O.x 8 /.
Inserting this in (5.23), we see that (if a C c D 0 and cd C bc 2 D 0)
xP D
b 2 c 3 x 7 C O.x 9 /:
Hence, if a C c D cd C bc 2 D 0 and b 2 c 3 > 0 then the origin is asymptotically

stable, and if a C c D cd C bc 2 D 0 and b 2 c 3 < 0 then the origin is unstable.
It can be checked that in the remaining borderline case a C c D cd C bc 2 D
2
3
b c D 0, h.x/ cx 2 and the reduced equation is simply xP D 0. Hence, in this
case, the origin is Lyapunov stable, but not asymptotically stable.
Bifurcation Theory
Bifurcation theory studies fundamental changes in the structure of the solutions
of a differential equation or a dynamical system in response to change in a parameter. Consider the parametrized equation
xP D F .x; "/;
(5.24)
where x 2 Rn is a variable and " 2 Rp is a parameter. Suppose that F .0; "/ D 0

for every ", that the equilibrium solution at x D 0 is stable when " D 0, and
that we are interested in the possibility of persistent structures (e.g., equilibria or
periodic orbits) bifurcating out of the origin as " is made nonzero. This means
136
Computing and Using Center Manifolds

that all the eigenvalues of Dx F .0; 0/ have nonpositive real part, so we can project
(5.24) onto complementary subspaces of Rn and get the equivalent system
(
uP D Au C f .u; v; "/
vP D Bv C g.u; v; "/;
with the eigenvalues of A lying on the imaginary axis and the eigenvalues of B
lying in the open right half-plane. Since the parameter " does not depend on time,
we can append the equation "P D 0 to get the expanded system
8
<uP D Au C f .u; v; "/

(5.25)
vP D Bv C g.u; v; "/
:
"P D 0:
The Center Manifold Theorem asserts the existence of a center manifold for the
origin that is locally given by points .u; v; "/ satisfying an equation of the form
v D h.u; "/:
Furthermore, a theorem of Carr says that every solution .u.t/; v.t/; "/ of (5.25)
for which .u.0/; v.0/; "/ is sufficiently close to zero converges exponentially
quickly to a solution on the center manifold as t " 1. In particular, no persistent structure near the origin lies off the center manifold of this expanded system. Hence, it suffices to consider persistent structures for the lower-dimensional
equation
uP D Au C f .u; h.u; "/; "/:
137
6
Periodic Orbits
6.1 Poincare-Bendixson Theorem
Definition. A periodic orbit of a continuous dynamical system ' is a set of the
form

'.t; p/ t 2 0; T
for some time T and point p satisfying '.T; p/ D p. If this set is a singleton,
we say that the periodic orbit is degenerate.
Theorem. (Poincare-Bendixson) Every nonempty, compact !-limit set of a C 1

planar flow that does not contain an equilibrium point is a nondegenerate periodic orbit.
We will prove this theorem by means of 4 lemmas. Throughout our discussion, we will be referring to a C 1 planar flow ' and the corresponding vector
field f .
Definition. If S is a line segment in R2 and p1 ; p2 ; : : : is a (possibly finite) sequence of points lying on S, then we say that this sequence is monotone on S if
.pj pj 1 / .p2 p1 / 0 for every j 2.
Definition. A (possibly finite) sequence p1 ; p2 ; : : : of points on a trajectory (i.e.,
an orbit) T of ' is said to be monotone on T if we can choose a point p and
times t1 t2 such that '.tj ; p/ D pj for each j .
Definition. A transversal of ' is a line segment S such that f is not tangent to
S at any point of S.
Lemma. If a (possibly finite) sequence of points p1 ; p2 ; : : : lies on the intersection of a transversal S and a trajectory T , and the sequence is monotone on T ,
139
6. P ERIODIC O RBITS
then it is monotone on S.
Proof. Let p be a point on T . Since S is closed and f is nowhere tangent
to S, the times t at which '.t; p/ 2 S form an increasing sequence (possibly
biinfinite). Consequently, if the lemma fails then there are times t1 < t2 < t3 and
distinct points pi D '.ti ; p/ 2 S, i 2 f1; 2; 3g, such that
fp1 ; p2 ; p3 g D '.t1 ; t3 ; p/ \ S
and p3 is between p1 and p2 . Note that the union of the line segment p1 p2 from
p1 to p2 with the curve '.t1 ; t2 ; p/ is a simple closed curve in the plane, so by
the Jordan Curve Theorem it has an inside I and an outside O. Assuming,
without loss of generality, that f points into I all along the interior of p1 p2 ,
we get a picture something like:
p1 b
O
bp2
Note that
I [ p1 p2 [ '.t1 ; t2 ; p/
is a positively invariant set, so, in particular, it contains '.t2 ; t3 ; p/. But the fact
that p3 is between p1 and p2 implies that f .p3 / points into I, so '.t3 "; p/ 2 O
for " small and positive. This contradiction implies that the lemma holds.
The proof of the next lemma uses something called a flow box. A flow box
is a (topological) box such that f points into the box along one side, points out
of the box along the opposite side, and is tangent to the other two sides, and the
140
Poincare-Bendixson Theorem
restriction of ' to the box is conjugate to unidirectional, constant-velocity flow.
The existence of a flow box around any regular point of ' is a consequence of
the C r -rectification Theorem.
Lemma. No !-limit set intersects a transversal in more than one point.
Proof. Suppose that for some point x and some transversal S, !.x/ intersects S
at two distinct points p1 and p2 . Since p1 and p2 are on a transversal, they are
regular points, so we can choose disjoint subintervals S1 and S2 of S containing,
respectively, p1 and p2 , and, for some " > 0, define flow boxes B1 and B2 by

Bi WD '.t; x/ t 2 "; "; x 2 Si :
Now, the fact that p1 ; p2 2 !.x/ means that we can pick an increasing sequence of times t1 ; t2 ; : : : such that '.tj ; x/ 2 B1 if j is odd and '.tj ; x/ 2 B2
if j is even. In fact, because of the nature of the flow in B1 and B2 , we can assume that '.tj ; x/ 2 S for each j . Although the sequence '.t1 ; x/; '.t2 ; x/; : : :
is monotone on the trajectory T WD .x/, it is not monotone on S, contradicting
the previous lemma.
Definition. An !-limit point of a point p is an element of !.p/.
Lemma. Every !-limit point of an !-limit point lies on a periodic orbit.
Proof. Suppose that p 2 !.q/ and q 2 !.r/. If p is a singular point, then it
obviously lies on a (degenerate) periodic orbit, so suppose that p is a regular
point. Pick S to be a transversal containing p in its interior. By putting a
suitable flow box around p, we see that, since p 2 !.q/, the solution beginning
at q must repeatedly cross S. But q 2 !.r/ and !-limit sets are invariant, so the
solution beginning at q remains confined within !.r/. Since !.r/ \ S contains at
most one point, the solution beginning at q must repeatedly cross S at the same
point; i.e., q lies on a periodic orbit. Since p 2 !.q/, p must lie on this same
periodic orbit.
Lemma. If an !-limit set !.x/ contains a nondegenerate periodic orbit P, then
!.x/ D P.
Proof. Fix q 2 P. Pick T > 0 such that '.T; q/ D q. Let " > 0 be given.
By continuous dependence, we can pick > 0 such that j'.t; y/ '.t; q/j < "
whenever t 2 0; 3T =2 and jy qj < . Pick a transversal S of length less than
141
with q in its interior, and create a flow box

B WD '.t; u/ u 2 S; t 2 ;
for some 2 .0; T =2. By continuity of '.T; /, we know that we can pick a
subinterval S 0 of S that contains q and that satisfies '.T; S 0 / B. Let tj be the
j th smallest element of

t 0 '.t; x/ 2 S 0 :
Because S 0 is a transversal and q 2 !.x/, the tj are well-defined and increase

to infinity as j " 1. Also, by the lemma on monotonicity, j'.tj ; x/ qj is a
decreasing function of j .
Note that for each j 2 N, '.T; '.tj ; x// 2 B, so, by construction of S and B,
'.t; '.T; '.tj ; x/// 2 S for some t 2 T =2; T =2. Pick such a t. The lemma
on monotonicity implies that
'.t; '.T; '.tj ; x/// 2 S 0 :
This, in turn, implies that t C T C tj 2 ft1 ; t2 ; : : :g, so
tj C1
tj t C T 3T =2:
(6.1)
Now, let t t1 be given. Then t 2 tj ; tj C1 / for some j 1. For this j ,

j'.t; x/
'.t
tj ; q/j D j'.t
tj ; '.tj ; x//
'.t
tj ; q/j < ";
since, by (6.1), jt tj j < jtj C1 tj j < 3T =2 and since, because '.tj ; x/ 2 S 0

S, jq '.tj ; x/j < .
Since " was arbitrary, we have shown that
lim d.'.t; x/; P/ D 0:
t "1
Thus, P D !.x/, as was claimed.
142
Now, we get to the proof of the Poincare-Bendixson Theorem itself. Suppose

!.x/ is compact and nonempty. Pick p 2 !.x/. Since C .p/ is contained in
the compact set !.x/, we know !.p/ is nonempty, so we can pick q 2 !.p/.
Note that q is an !-limit point of an !-limit point, so, by the third lemma, q lies
on a periodic orbit P. Since !.p/ is invariant, P !.p/ !.x/. If !.x/
contains no equilibrium point, then P is nondegenerate, so, by the fourth lemma,
!.x/ D P.
Lienards Equation
6.2 Lienards Equation

Suppose we have a simple electrical circuit with a resistor, an inductor, and a
capacitor as shown.
iC
b
C
L
iL
iR
b
Kirchhoffs current law tells us that

iL D iR D
iC ;
(6.2)
and Kirchhoffs voltage law tells us that the corresponding voltage drops satisfy
VC D VL C VR :
(6.3)
By definition of the capacitance C ,

C
dVC
D iC ;
dt
(6.4)
and by Faradays Law

d iL
D VL ;
(6.5)
dt
where L is the inductance of the inductor. We assume that the resistor behaves
nonlinearly and satisfies the generalized form of Ohms Law:
L
VR D F .iR /:
(6.6)
143
Let x D iL and f .u/ WD F 0 .u/. By (6.5),
xP D
1
VL ;
L
so by (6.3), (6.4), (6.6), and (6.2)

1 dVL
1
1
xR D
D .VPC VPR / D
L dt
L
L

1 1
D
. x/ f .x/xP
L C
1
iC
C
d iR
F .iR /
dt
0
Hence,
1
1
f .x/xP C
x D 0:
L
LC
By rescaling f and t (or, equivalently, by choosing units judiciously), we get
Lienards Equation:
xR C f .x/xP C x D 0:
xR C
We will study Lienards Equation under the following assumptions on F and

f:
(i) F .0/ D 0;
(ii) f is Lipschitz continuous;
(iii) F is odd;
(iv) F .x/ ! 1 as x " 1;
(v) for some > 0, F ./ D 0 and F is positive and increasing on .; 1/;
(vi) for some > 0, F ./ D 0 and F is negative on .0; /.
Assumption (vi) corresponds to the existence of a region of negative resistance. Apparently, there are semiconductors called tunnel diodes that behave
this way.
By setting y D xP CF .x/, we can rewrite Lienards Equation as the first-order
system
(
xP D y F .x/
(6.7)
yP D x:
144
Definition. A limit cycle for a flow is a nondegenerate periodic orbit P that is the
!-limit set or the -limit set of some point q P.
Lienards Equation
Theorem. (Lienards Theorem) The flow generated by (6.7) has at least one
limit cycle. If D then this limit cycle is the only nondegenerate periodic
orbit, and it is the !-limit set of all points other than the origin.
The significance of Lienards Theorem can be seen by comparing Lienards
Equation with the linear equation that would have resulted if we had assumed
a linear resistor. Such linear RCL circuits can have oscillations with arbitrary
amplitude. Lienards Theorem says that, under suitable hypotheses, a nonlinear
resistor selects oscillations of one particular amplitude.
We will prove the first half of Lienards Theorem by finding a compact, positively invariant region that does not contain an equilibrium point and then using
the Poincare-Bendixson Theorem. Note that the origin is the only equilibrium
point of (6.7). Since
d 2
.x C y 2 / D 2.x xP C y y/
P D
dt
2xF .x/;
assumption (vi) implies that for " small, R2 n B.0; "/ is positively invariant.
The nullclines x D 0 and y D F .x/ of (6.7) (i.e. curves along which the
flow is either vertical or horizontal) separate the plane into four regions A, B, C,
and D, and the general direction of flow in those regions is as shown below. Note
that away from the origin, the speed of trajectories is bounded below, so every
solution of (6.7) except .x; y/ D .0; 0/ passes through A, B, C, and D in succession an infinite number of times as it circles around the origin in a clockwise
direction.
145
We claim that if a solution starts at a point .0; y0 / that is high enough up on
the positive y-axis, then the first point .0; yQ0 / it hits on the negative y-axis is
closer to the origin then .0; y0 / was. Assume, for the moment, that this claim is
true. Let S1 be the orbit segment connecting .0; y0 / to .0; yQ0 /. Because of the
symmetry in (6.7) implied by Assumption (iii), the set

S2 WD .x; y/ 2 R2 . x; y/ 2 S1
is also an orbit segment. Let
S3 WD .0; y/ 2 R2
and let
S4 WD .0; y/ 2 R2

yQ0 < y < y0 ;

y0 < y < yQ0 ;

S5 WD .x; y/ 2 R2 x 2 C y 2 D "2 ;
for some small ". Then it is not hard to see that [5iD1 Si is the boundary of a
compact, positively invariant region that does not contain an equilibrium point.
y0
yQ0
yQ0
y0
To verify the claim, we will use the function R.x; y/ WD .x 2 C y 2 /=2, and
show that if y0 is large enough (and yQ0 is as defined above) then
146
R.0; y0 / > R.0; yQ0 /:
Lienards Theorem
6.3 Lienards Theorem

Recall, that were going to estimate the change of R.x; y/ WD .x 2 C y 2 /=2 along
the orbit segment connecting .0; y0 / to .0; yQ0 /. Notice that if the point .a; b/ and
the point .c; d / lie on the same trajectory then
Z .c;d /
R.c; d / R.a; b/ D
dR:
.a;b/
(The integral is a line integral.) Since RP D xF .x/, if y is a function of x along

the orbit segment connecting .a; b/ to .c; d /, then
Z c P
Z c
R
xF .x/
R.c; d / R.a; b/ D
dx D
dx:
(6.8)
x
P
y.x/
F .x/
a
a
If, on the other hand, x is a function of y along the orbit segment connecting
.a; b/ to .c; d /, then
Z d P
Z d
R
x.y/F .x.y//
R.c; d / R.a; b/ D
dy D
dy
x.y/
b yP
b
Z d
D
F .x.y// dy:
(6.9)
b
We will chop the orbit segment connecting .0; y0 / to .0; yQ0 / up into pieces and
use (6.8) and (6.9) to estimate the change R in R along each piece and, therefore, along the whole orbit segment.
Let D C 1, and let
B D sup jF .x/j:
0x
Consider the region
In R,

R WD .x; y/ 2 R2 x 2 0; ; y 2 B C ; 1/ :

dy
x

D
D 1I
dx
y F .x/

hence, if y0 > B C 2 , then the corresponding trajectory must exit R through
its right boundary, say, at the point .; y /. Similarly, if yQ0 < B 2 , then
the trajectory it lies on must have last previously hit the line x D at a point
.; yQ /. Now, assume that as y0 ! 1, yQ0 ! 1. (If not, then the claim clearly
holds.) Based on this assumption we know that we can pick a value for y0 and
a corresponding value for yQ0 that are both larger than B C 2 in absolute value,
and conclude that the orbit segment connecting them looks qualitatively like:
147
.0; y0 /
.; y /
.; yQ /
.0; yQ0 /
We will estimate R on the entire orbit segment from .0; y0 / to .0; yQ0 / by
considering separately, the orbit segment from .0; y0 / to .; y /, the segment
from .; y / to .; yQ /, and the segment from .; yQ / to .0; yQ0 /.
First, consider the first segment. On this segment, the y-coordinate is a function y.x/ of the x-coordinate. Thus,
Z
xF .x/
jR.; y / R.0; y0 /j D
dx
F .x/
0 y.x/
Z
Z
B
xF .x/
dx

y.x/ F .x/ dx
B
0
0 y0
2B
D
!0
y0 B
as y0 ! 1. A similar estimate shows that jR.0; yQ0 / R.; yQ /j ! 0 as
y0 ! 1.
On the middle segment, we know that the x-coordinate is a function x.y/ of
the y-coordinate y. Hence,
R.; yQ /
R.; y / D
yQ
y
F .x.y// dy
jy
as y0 ! 1.
Putting these three estimates together, we see that
148
R.0; yQ0 /
R.0; y0 / !
yQ jF . / !
Lienards Theorem
as y0 ! 1, so jyQ0 j < jy0 j if y0 is sufficiently large. This shows that the orbit
connecting these two points forms part of the boundary of a compact, positively
invariant set that surrounds (but omits) the origin. By the Poincare-Bendixson
Theorem, there must be a limit cycle in this set.
Now for the second half of Lienards Theorem. We need to show that if
D (i.e., if F has a unique positive zero) then the limit cycle whose existence
weve deduced is the only nondegenerate periodic orbit and it attracts all points
other than the origin. If we can show the uniqueness of the limit cycle, then the
fact that we can make our compact, positively invariant set as large as we want
and make the hole cut out of its center as small as we want will imply that it
attracts all points other than the origin. Note also, that our observations on the
general direction of the flow imply that any nondegenerate periodic orbit must
circle the origin in the clockwise direction.
So, suppose that D and consider, as before, orbit segments that start on
the positive y-axis at a point .0; y0 / and end on the negative y-axis at a point
.0; yQ0 /. Such orbit segments are nested and fill up the open right half-plane.
We need to show that only one of them satisfies yQ0 D y0 . In other words, we
claim that there is only one segment that gives
R.0; yQ0 /
R.0; y0 / D 0:
Now, if such a segment hits the x-axis on 0; , then x all along that
segment, and F .x/ 0 with equality only if .x; y/ D .; 0/. Let x.y/ be the
x-coordinate as a function of y and observe that
R.0; yQ0 /
R.0; y0 / D
yQ0
F .x.y// dy > 0:
(6.10)
y0
We claim that for values of y0 generating orbits intersecting the x-axis in .; 1/,
R.0; yQ0 / R.0; y0 / is a strictly decreasing function of y0 . In combination with
(6.10) (and the fact that R.0; yQ0 / R.0; y0 / < 0 if y0 is sufficiently large), this
will finish the proof.
Consider 2 orbits (whose coordinates we denote .x; y/ and .X; Y /) that intersect the x-axis in .; 1/ and contain selected points as shown in the following
diagram.
149
.0; Y0 /
b
b
.; Y /
.0; y0 /
.; y /
.; y /
.; yQ /
.; yQ /
.0; yQ0 /
b
b
.0; YQ0 /
.; YQ /
Note that
R.0; YQ0 /
R.0; Y0 / D R.0; YQ0 / R.; YQ /

C R.; YQ / R.; yQ /
C R.; yQ /
C R.; y /
C R.; Y /
R.; y /
(6.11)
R.; Y /
R.0; Y0 /
DW 1 C 2 C 3 C 4 C 5 :
Let X.Y / and x.y/ give, respectively, the first coordinate of a point on the
outer and inner orbit segments as a function of the second coordinate. Similarly,
let Y.X / and y.x/ give the second coordinates as functions of the first coordinates (on the segments where thats possible). Estimating, we find that
1 D
XF .X /
dX <
Y.X / F .X /
2 D
3 D
yQ
F .X.Y // d Y <
xF .x/
dx D R.0; yQ0 /
y.x/ F .x/
R.; yQ /;
(6.12)
YQ
F .X.Y // d Y < 0;
(6.13)
yQ
yQ
F .x.y// dy D R.; yQ /
4 D
150
R.; y /; (6.14)
F .X.Y // d Y < 0;
(6.15)
Lienards Theorem
and
5 D
XF .X /
dX <
Y.X / F .X /
xF .x/
dx D R.; y /
y.x/ F .x/
R.0; y0 /:
(6.16)
By plugging, (6.12), (6.13), (6.14), (6.15), and (6.16) into (6.11), we see that
R.0; YQ0 /
R.0; Y0 /
< R.0; yQ0 /
C R.; yQ /
R.; yQ / C 0
C R.; y /
R.; y / C 0
R.0; y0 /
D R.0; yQ0 /
R.0; y0 /:
This gives the claimed monotonicity and completes the proof.
151

Theory of Odes 2008

Enviado por

Dados do documento

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Theory of Odes 2008

Enviado por

Direitos autorais:

Formatos disponíveis

Theory of Ordinary Differential Equations

Theory of Ordinary Differential Equations

LaSalles Invariance Principle . . . . . . . . . . . . . . . . . .

Then an nth order ordinary differential equation is an equation of the form

F .t; x.t/; x.t/;

If I  R is an interval, then x W I ! E is said to be a solution of (1.2) on I if

and let y W I ! E n be defined by y D .y1 ; : : : ; yn /. For i D 1; 2; : : : ; n

and, given F as in (1.1), define Gn W dom.Gn /  R E n E n ! Rj by

Letting G W dom.Gn /  R E n E n ! E n 1 Rj be defined by

ODEs and Dynamical Systems

Equations Resolved with Respect to the Derivative

where F W dom.F /  R Rn Rn ! Rn , and x0 ; p0 are given elements of Rn

is autonomous if f doesnt really depend on t, i.e., if dom.f / D R  for some

where f D .f1 ; : : : ; fn /T and

Exercise 1 Suppose that:

ODEs and Dynamical Systems

 ' W R  !  is a continuous dynamical system;

Show that y solves the IVP

In these notes we will also discuss discrete dynamical systems. Given  

1. '.0; x/ D x for every x 2 ;

1.2 Existence of Solutions

where f W dom.f /  R Rn ! Rn is continuous, and .t0 ; a/ 2 dom.f / is

f .s; x.s// ds:

f .s; xk .s// ds:

f .s; xk .s// dx; if t  t0 C 1=k

Theorem. (Arzela-Ascoli) Every uniformly bounded, uniformly equicontinuous

Theorem. (Uniform Convergence) If a sequence of continuous functions hk W

We are now in a position to state and prove the Cauchy-Peano Existence

M=k < D jxk .t1 /

jf .s; xk .s//j dx  jaj C .b

Step 3: .xk / is uniformly equicontinuous.

jxk .t1 / xk .t2 /j D

1=k/M < jaj C :

Since xk is constant on t0 ; t0 C 1=k and continuous at t0 C 1=k, the estimate

f .s; xk` .s// ds

f .s; xk` .s// ds:

f .s; x.s// ds;

and by the boundedness of f the second integral converges to 0. Thus, taking

Exercise 2 How many solutions of the IVP

1.3 Uniqueness of Solutions

Exercise 3 Assume that the conditions of the Cauchy-Peano Theorem

for every t 2 t0 ; t0 C b, and that V .t0 / D ".

Definition. An inner product space is a vector space together with a function

Definition. A sequence .xn / in a metric space X converges to x if for every

Hence, .xk / is a Cauchy sequence. Since X is complete, .xk / converges to some

1.4 Picard-Lindelof Theorem

is equivalent to the standard supremum norm k k1 on C.t0 b; t0 C b/, so this

f .s; x.s// ds:

Step 1: If x 2 X then T .x/ makes sense.

so T .x/.t/ 2 B.a; /. Hence, T .x/ 2 X .

Taking the supremum over all t 2 t0 b; t0 C b, we find that T is a contraction

1.5 Intervals of Existence

2.t; a/; t C 2.t; a/ B.a; 2.t; a//  D:

Note that the collection of sets

Q WD min .ti ; xi / iD1 :

Since K0 is a compact subset of D, there is a constant M > 0 such that f is

for every .t0 ; a/ 2 K, so f is bounded by M on each such product set. According

where d.x; Rn n / is the distance from x to the complement of . It is not

Exercise 4 Suppose that  is a subset of Rn , that f W  ! Rn and

1.6 Dependence on Parameters

where  2 Rk . We are interested in studying how the solution of (1.17) depends

for every t 2 t0 ; T . Then

is continuous. Suppose, furthermore, that f .t; ; / is Lipschitz continuous with

This theorem shows continuous dependence on parameters if, in addition to

If I R is an interval, then x W I ! E is said to be a solution of (1.2) on I if

and, given F as in (1.1), define Gn W dom.Gn / R E n E n ! Rj by

Letting G W dom.Gn / R E n E n ! E n 1 Rj be defined by

where F W dom.F / R Rn Rn ! Rn , and x0 ; p0 are given elements of Rn

is autonomous if f doesnt really depend on t, i.e., if dom.f / D R for some

' W R ! is a continuous dynamical system;

In these notes we will also discuss discrete dynamical systems. Given

1. '.0; x/ D x for every x 2 ;

where f W dom.f / R Rn ! Rn is continuous, and .t0 ; a/ 2 dom.f / is

f .s; xk .s// dx; if t t0 C 1=k

jf .s; xk .s//j dx jaj C .b

2.t; a/; t C 2.t; a/ B.a; 2.t; a// D:

where d.x; Rn n / is the distance from x to the complement of . It is not

Exercise 4 Suppose that is a subset of Rn , that f W ! Rn and

where 2 Rk . We are interested in studying how the solution of (1.17) depends

is continuous. Suppose, furthermore, that f .t; ; / is Lipschitz continuous with

jx1 .t/ x2 .t/j D f .s; x1 .s/; 1 / f .s; x2 .s/; 2 ds

fx .t; x.t; /; /y.t; /

f .t; x.t; /; /;

.t; ; / jf .t; x.t; /; C 2 / f .t; x.t; /; /j

C jfx .t; x.t; / C 1 x; C /j jz.t; ; /j

fx .t; x.t; /; C /j jy.t; /j

fx .t; x.t; /; /j jy.t; /j

p.t; ; / C .jfx .t; x.t; /; /j C p.t; ; //jz.t; ; /j;

jfx .t; x.t; /; /j K

which gives X.t/ "e .KC"/.t

Given t t0 , pick K so large that (1.22) holds. As ! 0, we can take "

can be made arbitrarily small by taking m ` N for N sufficiently large.

If B2 does map everything to zero, then kB2 k D kB1 B2 k D 0, so kB1 B2 k

Given B 2 L.Rn ; Rn /, recall that that 2 C is an eigenvalue of B if Bx D x

Proof. Note first that .T k I /N.T

j I / N.T j I /. We claim that, in

has k as its only

If D a C bi 2 C n R is an eigenvalue of an operator T 2 L.Rn ; Rn /, and