I TECHNICAL REPORT 66-1

r * '

J. P. LaSALLE AN INVARIANCE PRINCIPLE IN THE

I THEORY OF STABILITY

APRIL, 1968

ff 653 July 66

CENTER FOR DYNAMICAL SYSTEMS

(CODE)
 AN INVARIANCE PRINCIPLE I N TKE THEORY OF STABILITY

J. P. LaSalle
' Center f o r Dynamical Systems
I Brown University

1. In tro d u ctio n .

The purpose of t h i s paper i s t o give a u n i f i e d pr e se nta -

t i o n of Liapunov's th eo ry of s t a b i l i t y t h a t includes t h e c l a s s i c a l

Liapunov theorems on s t a b i l i t y and i n s t a b i l i t y as w e l l as t h e i r

more recen t extensions. The idea being e xploite d h e r e had i t s

beginnings some time ago. It was, however, t h e use made of t h i s

i d e a by Yoshizawa i n [l] i n h i s study of nonautonomous d i f f e r e n t i a l

equations and by Hale i n [2] i n h i s study of autonomous f u n c t i o n a l

d i f f e r e n t i a l equations t h a t caused t h e author t o r e t u r n t o t h i s

s u b j e c t and t o adopt t h e general approach and p o i n t of view of t h i s

paper. This produces some new r e s u l t s f o r dynamical systems defined

by o rd in ary d i f f e r e n t i a l equations which demonstrate t h e e s s e n t i a l

n a t u r e of a Liapunov fu n c tion and which may be u s e f u l i n a p p l i c a t i o n s .

Of g r e a t e r importance, however, i s t h e p o s s i b i l i t y , as a lr e a dy i n -

d i c a t e d by H ale' s r e s u l t s f o r f unc tiona l d i f f e r e n t i a l equations,

*This research was supported i n p a r t by t h e N a t i o n a l p e r o n a u t i c s and

Space Administration under- Grant No. NGR-40-002-015.hd under--Contract
No, Ms8-11264,' i n p a r t by t h e Uniced S t a t e s A i r Force through t h e
A i r Force O ffice of S c i e n t i f i c Research underAGrant N o A F - A F O S R - ~ ~ ~/ ~ ~ ~ ,
and i n p a r t by t h e United S t a t e s A r m y Research OfficejDurham, under
Contract No DA-31-124-ARO-D-270. -,- 1
i
 2

t h a t t h e s e ideas can be extended t o more g e n e r a l c l a s s e s of dynam-

i c a l systems. It i s hoped, f o r insta nc e , t h a t it may be p o s s i b l e

t o do t h i s f o r some s p e c i a l types of dynamicalsystems de f ine d by

p a r t i a l d i f f e r e n t i a l equations.

In s e c t i o n 2 we pr e se nt some b a s i c r e s u l t s f o r or dina r y

d i f f e r e n t i a l equations. Theorem 1 i s a fundamental s t a b i l i t y

theorem f o r nonautonomous systems and i s a modified v e r s i o n of

Yoshizawa's Theorem 6 i n [l]. A simple example shows t h a t t h e

conclusion of t h i s theorem i s t h e b e s t p o s s i b l e . However, when-

ever t h e l i m i t s e t s of s o l u t i o n s a r e known t o have an inva r ia nc e

pr operty then sharper r e s u l t s can be obtained. This "inva r ia nc e

p r i n c i p l e " ex p lain s t h e t i t l e of t h i s paper. It had i t s o r i g i n f o r

autonomous and p e r i o d i c systems i n [3] - [SI, although we pr e se nt

here improved v ersio n s of those results. Miller i n [6] has e sta b-
l i s h e d an invariance p ro p e r ty f o r almost p e r i o d i c systems and ob-

t a i n s thereby a similar s t a b i l i t y theorem for almost p e r i o d i c

systems. Since l i t t l e a t t e n t i o n has been pa id t o theorems which

make p o ssib le estimates of regions of a t t r a c t i o n (regions of asymp-

t o t i c s t a b i l i t y ) f o r nonautonomous systems results of t h i s type a r e

included. S ectio n 3 i s devoted t o a b r i e f disc ussion of some of

Hale's recent r e s u l t s [2] for autonomous f u n c t i o n a l d i f f e r e n t i a l

equations.

2. Ordinary d i f f e r e n t i a l equations.

Consider t h e system
 3

n+1
where x i s an n-vector, f i s a continuous f unc tion on R
n
to R and s a t i s f i e s any one of t h e c onditions guaranteeing unique-

'ness of s o l u t i o n s . For each x in Rn we d e f i n e 1x1 =

(xl
2
+ ... + xn)
23
, and f o r
,
E a c lose d s e t i n Rn we d e f i n e

d(x,E) = Min [ Ix-yl : y i n E). Since w e do not wish t o confine our-

s e l v e s t o bounded s o l u t i o n s , we introduce t h e point a t m and

def in e d(x,m) = 1 x l - l . Thus when w e w r i t e E* = E U[m), we s h a l l

mean d(x,E*) = Min{d(x,E), d(x,m)), If x(t) i s a s o l u t i o n of

(l), we say t h a t x(t) approaches E as t -+m if d(x(t),E) 4 9

as t --$a. If we can f i n d such a s e t E, we have obtained i n -

formation about t h e asymptotic behavior of x(t) as t + 00. The

b e s t t h a t we could hope t o do i s t o f i n d t h e sma lle st c lose d s e t

R that x(t) approaches as t -00. This s e t R i s called the

p o s i t i v e --
l i m i t s e t of x(t) and t h e p o i n t s p in R are called

t h e p o s i t i v e l i m i t p o i n t s of x(t). I n e x a c t l y t h e same way one

defin es x(t) + E as t --f -= , negative l i m i t s e t s , and negative

l i m i t p o in ts. This i s e xa c tly G. D. Bir khof f 's concept of l i m i t

sets. A p o in t p i s a p o s i t i v e l i m i t p o i n t of x(t) i f and only

if t h e r e i s a sequence of times tn approaching 00 as n -+a and

such t h a t x(tn) +p as n 4o . I n t h e above it may be t h a t t h e

maximal i n t e r v a l of d e f i n i t i o n of x(t) is [ 3 , ~ ) . This causes

no d i f f i c u l t y s i n c e i n t h e r e s u l t s t o be pr e se nte d h e r e we need

only with r e s p e c t t o time t r e pla c e mby T. We u s u a l l y ignore

 1
4 I
t h i s p o s s i b i l i t y and speak as though our s o l u t i o n s a r e def'irred on

[o,w) or .
t
a
(-m,m)

i n
Let V (t, x ) be a C fanction O i l [O,m) x R t o R, and

.
let G be m y s e t i n

f u n c t i o n on G
Rn

f o r equation (1) i f
We s h a l l say t h a t

V(t,x) 2
V

0
i s a Liapunov

ana. V(t,x) 5
I
-W(x) 5 0 f o r a l l t > 0 and a l l x in G where W is
I
continuous on Rii to R a d

I
We d efin e (c i s t h e clo sur e of G)
I
-
E = (x, W(x) = 0, x in G).
I
The following r e s u l t i s then a modified but c l o s e l y r e -

l a t e d version o f Yoshizawa's Theorem 6 i n [l]. 1

THEOREM 1. If V i s a Liapunov f u n c t i o n on G for equation (l), I
then each s o l u t i o n x(t) of (1)t h a t remains i n G for a l l

t > to 2 0 approaches E* = E U (a] as t +co , provided one of I

t h e following co n d itio n s i s s a t i s f i e d :

(i) For each p in t h e r e i s a neighborhood 1\s of

I
p such t h a t If(t,x)l i s bounded f o r a l l t >0 and 1
c
ail x in N.

(ti) W is C1 and i s bounded from above o r below

along each s o l u t i o n which remains i n

t > t o 2 0 .
G f o r all
I
I
I
I
 5
If E i s bounded, then each s o l u t i o n of (1) t h a t remains i n G

for t >t 2 0 e i t h e r approaches E or 03 as t+-.

0

Thus t h i s theorem explains p r e c i s e l y t h e na tur e of t h e

information given by a Liapunov function. A Liapunov f unc tion

relative t o a set G d ef ine s a s e t E which under t h e c onditions

of t h e theorem co n tain s ( loc a te s) a l l t h e p o s i t i v e l i m i t s e t s of

solu tio ils which f o r p o s i t i v e time remain i n G. The problem i n

P applying t h e r e s u l t i s t o f i n d "good" Liapunov f unc tions. For

ins tan ce, t h e zero fu n ction V= 0 i s a Liapunov f unc tion f o r tiie

! whole space Rn aiid condition (ii) i s s a t i s f i e d b u t gives no i n -

n
.
1 information s i n c e E = R

c a t i o n s t o note t h a t i f V,
I
It i s t r i v i a l b u t u s e f u l f o r a p p l i -

and V,
L
a r e Liapunov f unc tions on G,

1 then V = V
1
+ V2 i s a l s o a Liapunov f unc tion and E = E l l E2
1
.
If E i s smaller t h m e i t h e r El o r E2 , -then V i s a "better"
I Liapunov f u n c t i o n th an e i t h e r E
1
or E2 and i s always a t l e a s t a s

I "good" as e i t h e r of t h e two.

Condition (5) of Theorem 1 i s e s s e n t i a l l y t h e one used

I by Yoshizawa. We now l o o k a t a simple example where c ondition (ii)

I i s s a t i s f i e d and co n d ition (i)i s not.

t h e conclusion of t h e theorem i s t h e b e s t p o s s i b l e .
The example a l s o shows t h a t

Consider

I *2 + p(t)? + x = 0 where p(t) 2 6 >0 . Define 2 V = x2 + y2 ,

where y = 2 . Then V = -p(t)y
2
6 - 6y
2
and V i s a Liapunov
2 2
f u n c t i o n on R2 Now W = 6y and = 2 6 =~ - 2 S ( x y + p(t)) )

-26xy. Since all s o l u t i o n s a r e e v i d e n t l y bounded f o r all t > 0,

 6
con d itio n (ii) i s s a t i s f i e d . Here E i s t h e x-axis (y = 0)

and f o r each s o l u t i o n x(t), y(t) = 2(t) + 5 as t + . Noting

x + + t
e )S +x =0 has a s o l u t i o n
t h a t t h e equation (2

x ( t ) = 1t e-t , we see t h a t t h i s i s t h e b e s t p o s s i b l e r e s u l t with-

out f u r t h e r r e s t r i c t i o n s on p .
I n order t o use Theorem 1 t h e r e must be some means of

determinLng which SolEtiOiis remain i n G . The following c o r o l l a r y ,

which i s an obvious consequence of Theorem 1, gives one way of

doing t h i s and also provides f o r nonautonomous systems a method f o r

estimating regions of a t t r a c t i o n .

Corollary 1. Assume t h a t t h e r e e x i s t continuous f unc tions u( x)

n
and v(x) on R to R such t h a t U(X) 5 V(t,x) 5 V ( X ) f o r all

t 2 0 . Define
+
Q7 = {x ; U(X) < v} and l e t G
+ be a component
4-
of Qrl Let G denote t h e component of Q = {x ; v ( x ) < 71
11
containing G+ . If V i s a Liapunov f’Jnction on G f o r (1) and

t h e conditions of Theorem 1 a r e s a t i s f i e d , t h e n each s o l u t i o n of

(1) s t a r t i n g i n G+ a t any time t0 2 o remains i n G f o r all

t > to and approaches E* as t +w e If G i s bounded and

Eo = r G C G+ , th en Eo i s an a t t r a c t o r and G
+ i s i n its

r egio n of a t t r a c t i o n .

I n g en eral we know t h a t i f x(t) i s a s o l u t i o n of

(l)--in fact, i f x ( t > i s any continuous f u n c t i o n on R to R”--

the n i t s p o s i t i v e l i m i t s e t i s c lose d and connected. If x(t) is

bounded, then i t s p o s i t i v e l i m i t s e t i s compact. There are, how-

1
I 7

I ever, s p e c i a l c l a s s e s of d i f f e r e n t i a l equations where t h e l i m i t

s e t s of s o l u t i o n s have an a d d i t i o n a l inva r ia nc e pr ope r ty which

I makes p o s s i b l e a refinement ot' Theore= 1. m e f i r s t of t h e s e are

t h e autonomous systems
I a = f(x) (3)

I The l i m i t s e t s of s o l u t i o n s of (3) a r e i n v a r i a n t s e t s . If x(t)

i s defined on [O,m) and i f p i s a p o s i t i v e l i m i t p o i n t of x(t),

t h e n t h e p o i n t s on t h e s o l u t i o n through p on i t s maximal i n t e r -

v a l of d e f i n i t i o n a r e p o s i t i v e l i m i t p o i n t s of x(t). If x(t) is

b o E d e d for t >0 , then it i s de f ine d un [O,mj, i t s positive

limit set R i s compact, noneinpty and s o l u t i o n s through p o i n t s

p of R a r e d efin ed on (-a,..) (ioe., R i s invariant). If

t h e maximal domain of d e f i n i t i o n of x(t) for t >0 i s finite,

then x(t) has no f i n i t e p o s i t i v e l i m i t points: t h a t is, i f the

m a x i r r l a l ijnierval of d e f i n i l i o n o f x(t) for L > 0 is [O,p),

then x(t) + a as t -+p . As we have s a i d before, we w i l l always

speak a s though our s o l u t i o n s are defined on and it should

I
(-m,m)

be remembered t h a t f i n i t e escape time i s always a p o s s i b i l i t y u n l e s s

I t h e r e i s , a s f o r example i n Corollary 2 below, some c ondition t h a t

r u l e s it o u t. I n Corollary 3 below, t h e s o l u t i o n s might w e l l go t o

I infinitjr iii f i n i t e time.

The invariance property of t h e l i m i t s e t s of s o l u t i o n s

1 of autonomous systems ( 3 ) now ena5les us t o r e f i n e Theorem 1.

n
.
I Let V be a C1 fu n ctio n on R to R If G i s any a r b i t r a r y

I
1
 1
8 1
set in Rn , we say t h a t V i s a Liapunov f unc tion -
on G f o r
1
equation (3) if fl= (grad V). f does not change s i g n on G .
Define E = ( x ; i(x) = 3 , x ir, G 1 , where
-
G i s ihe I
. .
closu re of G

w i l l be a closed s e t .
Let M be t h e l a r g e s t i n v a r i a n t s e t i n

The fundamental s t a b i l i t y theorem f o r

E M
I
autonomous systems i s then t h e following: I
THEORFM 2 .

each s o l u t i o n
If V i s a Liapunov f unc tion on G f o r ( 3 ) , the n
x(t) of (3) t h a t remains i n G f o r a l l t > 0
1
(t < 0 ) approaches M* = M U (m) as t -+ m (t -m). If M is I
bounded, then e i t h e r x(t) -+M or x(t) as t (t .
I
+ m -+m -m)

This one theorem contains a l l of t h e u s u a l Liapunov l i k e

theorems on s t a b i l i t y and i n s t a b i l i t y of autonomous systems.

however, t h e r e a r e no conditions of d e f i n i t e n e s s f o r V or V
Here

,
I
and it i s o f t e n p o s s i b l e t o obta in s t a b i l i t j r information about a 1
system w i t i i t h e s e more g ene r a l types of Liapunov f unc tions. The

f i r s t c o r o l l a r y below i s a s t a b i l i t y r e s u l t which f o r applications I

has been q u i t e u s e f u l a i d t h e second i l l u s t r a t e s how one obta ins

i n f o m a t t o n on i n s t a b i l i t y .
v
Cetaev's i n s t a b i l i t y theorem i s
I
s i m i l a r l y an immediate consequence of Theorem 2 ( se e s e c t i o n 3).
1
COROLLARY 2. Let G be a component of Q = ( x ; V(x) < 7 ) .
Ass-me tliai G i s bon-ided, V 6 0 on G
7
, and M
0 -
= M n G cG . I
Then M
0
i s an a t t r a c t o r and G i s i n i t s region of a t t r a c t i o n . 1
I f , i n addition, V i s c onsta nt on t h e boundary of Mo , the n
1
I
I
 9

Mo i s a s t a b l e a t t r a c t o r .
Note t h a t if Mo c o n s i s t s of a s i n g l e p o i n t p ,
then p i s asymptotically s t a b l e and G provides an estimate of

i t s region of asymptotic s t a b i l i t y .

COROLLARY 3. Assume t h a t r e l a t i v e t o (3) t h a t V > 0 on G

and on t h e boundary of G that V = 0 . Then each s o l u t i o n of

(3) s t a r t i n g i n G approaches m as t -+ m ( o r possibly i n

f i n i t e time).

There a r e a l s o some s p e c i a l c l a s s e s of nonautonomous

systems where t h e l i m i t s e t s o f s o l u t i o n s have an invariance

property. The simplest of these a r e p e r i o d i c systems (see [ 3 ] ) .

2 = f(t,x) , f ( t + T,x) = f(t) for a l l t and x . (4)

Eere i n order t o avoid introducing t h e concept of a p e r i o d i c

approach of a s o l u t i o n of (4) t o a s e t and t h e concept of a

p e r i o d i c l i m i t p o i n t l e t us confine ourselves t o s o l u t i o n s x(t)

of (4) which a r e bounded f o r t >0 . Let R be t h e p o s i t i v e

l i m i t s e t of such a s o l u t i o n x ( t ) , and l e t p be a p o i n t i n R .
Then t h e r e i s a s o l u t i o n o f (4) s t a r t i n g a t p which remains i n

R for a l l t in (-w,oo) ; t h a t i s , i f one s t a r t s a t p at the

proper time t h e s o l u t i o n remains i n R f o r a l l time. Tnis i s t h e

sense now i n which R i s an invariant s e t . Let V(t,x) be C1

on R x Rn and p erio d ic i n t of pe r iod T . For an a r b i t r a r y

set G of Rn we say t h a t V i s a Liapunov f unc tion -

on G for
 10

f o r t h e g erio d ic system (4) if V does not change s i g n f o r a l l

t and a l l x in G . Define E = { (t,x); V( t,x) = 0, x in 3 }

and l e t M b e t h e union of a l l s o l u t i o n s x(t) of (4) with t h e
pr o p erty t h a t (t,x(t)) is in E for a l l t . M could be c a l l e d

"the largest invariant s e t r e l a t i v e t o E". One the n obta ins t h e

following v ersio n of Theorem 2 f o r p e r i o d i c systems:

THEOFEM 3. If V i s a Liapunov f u n c t i o n on G f o r the periodic

system (4), then each s o l u t i o n of (4) t h a t i s bounded and remains

In G fo r a l l t > 0 (t < 0) approaches M as t -+ 00 (t + - 0 ~ ) .

I n [6] M i l l e r showed t h a t t h e l i m i t s e t s of s o l u t i o n s

of almost p e r i o d i c systems have a s i m i l a r inva r ia nc e pr ope r ty and

from t h i s he o b tain s a r e s u l t q u i t e l i k e Theorem 3 f o r almost

p e r i o d i c systems. This then y i e l d s f o r pe r iodic and almost p e r i o d i c

systems a whole chain of theorems on s t a b i l i t y and i n s t a b i l i t y

q u i t e similar t o t h a t f o r autonomous systems. For example, one has

COKOLLARY 4. Let
+
Qrl = { x; V(t,x) < 7, all t i n [O,T] } , and
let G
+ be a component of Q
+ . Let G be t h e compofient of
11
Q = { x ; V(t,x) < 11 f o r some t i n [O,T]
7
} c onta ining G+ . If G

i s bounded, V 5 0 f o r a l l t and a l l x in G , and if M

0
=

M fl G C G+, then Mo i s an a t t r a c t o r and G

+ i s i n i t s region of

attractioc. If V(t,x) = q ( t ) f o r all t and all x on t h e

boundary of Mo , th en Mo i s a s t a b l e a t t r a c t o r .

Our l a s t example of an inva r ia nc e p r i n c i p l e f o r or dina r y

 11

differential equations is that due to Yoshizawa in [l] for "asymp-

totically autonomous" systems. It is a consequence of Theorem 1

and results by Markus and Opial (see [l] for references) on the

limit sets of such systems. A system of the form

is said to be asymptotically autonomous if (i) g(t,x) -+9 as

t -+w uniformly for x in an arbitrary compact set of Rn ,
W

(ii) J Ih(t,q(t))l dt < w Yor all cp bounded and continuous

0
on [O,w) to Rn . The combined results of Markus and Opial then

state that the positive limit sets of solutions of (3) are in-

variant sets of k = F(x) . Using this, Yoshizawa then improved

Theorem 1 for asymptotically autonomous systems.

It turns out to be useful, as we shall illustrate in a

moment on the simplest possible example, in studying systems (1)

which are not necessarily asymptotically autonomous to state the

theorem in the following manner:

THEOREM 4. If, in addition to the conditions of Theorem 1, it is

known that a solution x(t) of (1) remains in G for t >0
and is also a solution of an asymptotically autonomous system (?),

then x ( t ) approaches M* = M U {OD) as t + w , where M is the

largest invariant set of 2 = F(x) in E .

It can happen that the system (1) is itself asymptotically

autonomous in which case the above theorem can be applied. However,

 ..-
Aid

as t h e following example i l l u s t r a t e s , t h e o r i g i n a l system may not

i t s e l f be asymptotically autonomous but it s t i l l may be p o s s i b l e

t o co n stru ct f o r each s o l u t i o n of (1)an a symptotic a lly autonomous

system (7) which it a l s o s a t i s f i e s .

Consider again t h e example

? = y (6)
y = - x - P(t)Y , o < s ~ p ( t )m ~
f o r all t > 0

Now we have t h e a d d i t i o n a l assumption t h a t p(t) i s bounded from

above. Let (x(t), -y ( t ) ) be any s o l u t i o n of (6). A s was argued

pr evio u sly below Theorem 1, a l l s o l u t i o n s a r e bounded and y(t) +0

as t + Q) . Now (E(t), y(t)) satisfies k = y ,y =

-x - p ( t ) y ( t ) , and t h i s system i s asymptotically autonomous t o

(") 2 = y , $ = -x . With t h e same Liapunov f unc tion a s before,

E i s t h e x - a x i s and t h e l a r g e s t i n v a r i a n t s e t of (*) i n E i s the

origin. Thus f o r (6) t h e o r i g i n i s a symptotic a lly s t a b l e i n t h e

large

3. Autonomous f u n c t i o n a l d i f f e r e n t i a l equation.

Difference d i f f e r e n t i a l equations of t h e form

2(t) = f(t,x(t),x(t-r)) 9 r > O (7)

have been stu d ied almost a s long as or dina r y d i f f e r e n t i a l equations

and t h e s e a s w e l l a s o th e r types of systems a r e of t h e ge ne r a l form

 13
is iri
where x Rn and x
t i s t h e f unc tion defined on [-r,O]

by x (2) = x (t+ . r), -r S 2 6 0. Thus xt i s t h e f unc tion t h a t

t
describ es t h e p a s t h i s t o r y of the system on t h e i n t e r v a l [t-r,t]
and i n order t o consider it as an element i n t h e space C of

contizuous f m c - t i o n s a l l defined on t h e same i n t e r v a l [-r,O], x,

b

i s taken t o be t h e fu n ction whose graph i s t h e t r a n s l a t i o n of t h e

graph of x on t h e i n t e r v a l [t-r,t] t o the interval [-r,O] .

Since such equations have had a long h i s t o r y it seems s u r p r i s i n g

that; it i s only w ith in t h e l a s t 10 ye a r s o r s o t h a t t h e geometric

theo ry of o rd in ary d i f f e r e n t i a l equations has been s u c c e s s f u l l y

c a r r i e d over t o f u n c t i o n a l d i f f e r e n t i a l equations. Krasovskii [8]

has demonstrated t h e effec tive ne ss of a geometric approach i n ex-

tending t h e c l a s s i c a l Liapunov theory, inc luding t h e converse

theorems, t o f u n c t i o n a l d i f f e r e n t i a l equations. An account of o t h e r

asp ects of t h e i r theory which have yie lde d t o t h i s geometric approach

can be found i n t h e paper [ g ] by Hale. What we wish t o do he r e i s

t o p resen t Hale's extension i n [2] of t h e r e s u l t s of Se c tion 2 of

t h i s paper t o autonomous f unc tiona l d i f f e r e n t i a l equations

K = f ( xt ) . (9)

It i s t h i s extension t h a t has had s o f a r t h e g r e a t e s t success i n

studying s t a b i l i t y p r o p e r t i e s of t h e s o l u t i o n s of systems (9), and

it i s p o s s i b l e t h a t t h i s may lead t o a s i m i l a r theory f o r s p e c i a l

c l a s s e s of systems defined by p a r t i a l d i f f e r e n t i a l equations.

With r 4 3 t h e space C i s t h e space of continuous

 14
f unctio n s cp on [-r,O] to Rn with l{cpll =

max (Icp(~)l; -r 5 T s 01. Convergence i n c i s uniform conver-

gence on [-r,O]. A fu n c tion x defined on [-r,m) to Rn to

-
s a i d t o be a s o l u t i o n of (9) s a t i s f y i n g t h e i n i t i a l c ondition cp

a t time t = 0 i f t h e r e i s an a > 0 such t h a t % ( t=) f ("t )

for a l l t in [O,a) and x 0 = cp . Remember x 0 = cp means

X(T) = cp(z), -r 5 T 5 0. At t = 0, k i s t h e r i g h t hand deriv-

ative. The ex isten ce uniqueness theorems a r e q u i t e s i m i l a r t o

those f o r ordinary d i f f e r e n t i a l equations. If f i s locally

L ip sch itzian on C, then f o r each cp in C t h e r e i s one and only

one s o l u t i o n of (9) and t h e s o l u t i o n depends continuously on cp .

The s o l u t i o n can a l s o be extended i n C for t >0 as long a s it

remains bounded. A s i n Section 2, we w i l l always speak as though

s o l u t i o n s a r e d efin ed on [-r,m). The space C i s now t h e s t a t e

space of (9) and through each p o i n t cp of C t h e r e i s t h e motion

o r flow x
t
starting a t cp defined by t h e s o l u t i o n x(t) of (9)
s a t i s f y i n g a t time t = 0 t h e i n i t i a l c ondition cp; xt' 0 5 t <m,

i s a curve i n C which s t a r t s a t time t = 0 at cp. I n analogy

t o Section 2 with C repla c ing Rn, xt r e pla c ing x(t), and

IIx
t 11 rep lacin g Ix(t)l, we de f ine t h e d i s t a n c e d(xt,E) of xt

from a closed s e t E of C t o be d ( xt ,E) = min (llxt-$l! ; $ E E).

The p o s i t i v e l i m i t s e t of xt i s then de f ine d i n a manner completely

analogous t o Section 2. Because t h e r e a r e some important d i f f e r e n c e s

we s h a l l be s a t i s f i e d h er e with r e s t r i c t i n g ourselves t o motions

x
t
hoimded f o r t > 0. One of t h e d i f f e r e n c e s he r e i s t h a t i n

C c lo sed and bounded s e t s a r e not always compact. Another i s t h a t

although we have uniqueness of s o l u t i o n s i n t h e f u t u r e two motions

s t a r t i n g from d i f f e r e n t i n i t i a l c onditions can come toge the r i n

f i n i t e time t
0
> 0; a f t e r t h i s the y coincide f o r t 2 t
0
. (The

motions d efin e semi-groups and not n e c e s s a r i l y groups. )

Hale i n [2] has, however, shown t h a t t h e p o s i t i v e l i m i t

sets R of bounded motions x a r e nonempty, compact, connected,

t
invariant sets i n C . Invariance he r e i s i n t h e sense t h a t , i f

x i s a motion s t a r t i n g a t a point of R, the n t h e r e i s an exten-

t
s i o n onto (-w,-r] such t h a t x(t) i s a s o l u t i o n of (9) f o r a l l

t in (-m,w) and x
t
remains i n R for a l l t . With t h i s

r e s u l t he i s th en ab le t o obtain a r e s u l t which i s s i m i l a r t o

Cor o llary 1 of S ectio n 2.

For cp E C let x (cp)

t
denote t h e motion defined by (9)
starting at cp . For V a continuous f unc tion on C to R define

V and QR by
- 1
+(cp) = lim 7 rv(xT(cp))-v(cp)l.
T-+*

and

THEOREM 3. If V i s a Liapunov f unc tion on G f o r (9) and xt is

a t r a j e c t o r y of (9) which remains i n G and i s bounded f o r t > 0,

then x
t +M as t + 00 .
 16
I
Hale has a l s o given t h e following more u s e f u l ve r sion 1
of t h i s r e s u l t .

COROLLARY 5. Define Q = ( ~ p ; V(cp) < v] and l e t G be Q or

I
v v
a component of Q
v
. Assume t h a t V i s a Liapunov f unc tion on
1
G for (9) and t h a t e i t h e r (i) G i s bounded o r (iii) [ c p ( O ) [ is

bounded for cp in G . Then each t r a j e c t o r y s t a r t i n g i n G 1

.
approaches M as t + 00

The following i s an extension of Fe ta e vls i n s t a b i l i t y

I
theorem. This i s a somewhat s i m p l i f i e d ve r sion of H a le 's Theorem

i n [ 2 ] , which should have s t a t e d "V(cp) > 0 on U when cp # 0

4
I
and V ( 0 ) = 0'' and a t t h e end 'I... i n t e r s e c t t h e boundary of I
This i s c l e a r from h i s proof and i s necessary s i n c e he
C
Y
...'I.

wanted t o g en eralize t h e u s u a l statment of Cetaev's theorem t o i n -

Y I
clude t h e p o s s i b i l i t y t h a t t h e equilibrium p o i n t be i n s i d e

w e l l as on i t s boundary.
U as
I
COROLLARY 6. Let p e C be an equilibrium p o i n t of (9) contained 1
i n t h e closure of an open s e t

p . Assume t h a t (i) V
U and l e t

i s a Liapunov f u n c t i o n on
N be a neighborhood of

G = U f' N,
1
(ii) M n G i s e i t h e r t h e empty s e t o r p, (5%:) V(cp) < 11 on G 1
when cp # p9 and (iv) V(p) = 7 and V(cp) = 7 on t h a t p a r t of

t h e boundary o f G inside N. Then p i s unsta ble . I n f a c t , if I

No i s a 5ourided neighborhood of p pr ope r ly contained i n N the n

each t r a j e c t o r y s t a r t i n g a t a p o i n t of Go = G fl No othe r than p

lea v es No i n f i n i t e time.
 17
-
Proof. By t h e co n d itio n s of the c o r o l l a r y and Theorem 6 each

trajectory starting inside a t a p o i n t othe r than p must

Go
e i t h e r leav e
GO,
approach i t s boundary o r approach p .
Conditions (i)and ( i v ) imply t h a t it cannot reach o r approach t h a t

p a r t of t h e boundary of Go inside No nor can it approach p

as t +m . Now (ii) s t a t e s t h a t t h e r e a r e no p o i n t s of M on

t h a t p a r t of t h e boundary of No i n s i d e G . Hence each such

t r a j e c t o r y must leav e No i n f i n i t e time. Since p i s either i n

t h e i n t e r i o r o r on t h e boundary of G, each neighborhood of p

con tain s such t r a j e c t o r i e s , and p i s t h e r e f o r e unsta ble .

I n [2] it was shown t h a t t h e equilibrium p o i n t cp = 0 of

R ( t ) = a x3 (t)+ bx3 (t-r)

was u n stab le if a > 0 and Ibl < I a1 . Using t h e same Liapunov

f u n c t i o n and Theorem 6 we c a n show a b i t more. With

4a -r

4 t
V(X,) = --
(t) + $ .I xb(e)ae
4a t-r

and

which i s nonpositive when I bl < 1 a1 (negative d e f i n i t e with r e -

spect t o cp(0) and cp(-r)); t h a t i s , V is 8 Liapimov f u n c t i o n

c,n C and E = ( c p ; cp(0) = cp(-r) = O} . Therefore M i s simply

t h e n u l l function cp = 0 . If' a > 0, t h e region G = { c p ; V(cp) < 0)

 I
18
I
i s nonempty, and no t r a j e c t o r y s t a r t i n g i n G can have cp = 0 as 1
a p o s i t i v e l i m i t p o i n t nor can it le a ve G . Hence by Theorem 5
each t r a j e c t o r y s t a r t i n g i n G must be unbounded. Since cp = 3
1
i s a boundary p o i n t of

[2] t h a t if a < 0 and

G, it is unsk;i'!)l.?.

lbl < 1 ai, the n

It i s a l s o e a s i l y seen

cp = 0 i s a symptotic a lly
1
stable i n the large. I
I n [ 2 ] Hale has a l s o extended t h i s the or y f o r systems

with i n f i n i t e l a g (r = m), and i n t h a t same paper give s a number

I
of s i g n i f i c a n t examples of t h e a p p l i c a t i o n s of t h i s theory.
I
I
1
1
1
1
I
I
1
1
1
I
I
I
I Ret'erences

I [l] Yoshizawa, T., Asymptotic behavior of s o l u t i o n s of a system

of d i f f e r e n t i a l equations, Contrib. t o D i f f . Eq., 1(1963),
371-387
I
[2] Hale, J., S u f f i c i e n t conditions f o r s t a b i l i t y and i n s t a b i l i t y
I of autonomous f u n c t i o n a l d i f f e r e n t i a l equations, J. of D i f f .
Eq., 1(1963), 432-482.
I [3] LaSalle, J., Some extensions of Liapunov's second method, IRE

I Trans. on C i r c u i t Theory, CT-7 (1$0), 320-527.

LaSalle, J., Asymptotic s t a b i l i t y c r i t e r i a , Proc. of Symposia

I [4]
i n Applied Mathematics, V o l . 13, Hydrodynamic I n s t a b i l i t y ,
Amer. Math. SOC., Providence, R . I . , 1962, 299-307.
I [3] LaSslle, J., and Lefschetz, So, S t a b i l i t y by Liapunov's D i r e c t
I Method with Applications, Academic Press, New York, 1961.

I [6] Miller, R , , On almost periodic d i f f e r e n t i a l equations, B u l .

Amer. Math. SOC o , 70 (l964), 792-793

I [7] Miller, R., Asymptotic behavior of nonlinear d e l a y - d i f f e r e n t i a l

equatLms, J, of D i f f . Eq., 3(1965), 293-305.
1
[8] Krasovskii, N.N., S t a b i l i t y of Motion, S t a n Y m i University Press,
1 1963( t r a n s l a t i o n o f 1939 Russian E d i t i o n ) .

I [9] Hale, J., "Geometric theory of f u n c t i o n a l d i f f e r e n t i a l equations",

Proc. of An I n t e r n a t i o n a l Symposium on D i f f e r e n t i a l Equations
and Dynamical Systems, University of Puerto Rico,Mayaguez,
I
P. R.,
Dec. 1963, Academic Press, New York ( t o appear).

1
I