284
A NOTE ON SMOOTH GONJUGACY OF NODE MANIFOLDS
Fatma M. Kandil
Mathematics Department ,Faculty of Science, King Abdul Aziz University, Jeddah - Saudi Arabia
ABSTRACT
It was shown that there exists a unique
C
r.smooth invariant manifold for a given dynamical system.Keywords: Smoothes, Manifold, Dynamical system . A.M.S.C: 34A26, 34A25, 34A34, 34C30.
1. INTRODUCTION
Consider a vector field
F R
:
p
R
p, 2
p
which isC
rsmooth(
r
2)
on the ballB
p
{
z
}
with a nodal equilibrium at the origin.F generates the dynamical system
( , )
( , )
x
Lx
P x y
y
My
Q x y
(S)where L and M are diagonalizable matrices of order
n n
andm m
respectively,(0, 0)
(0, 0)
0 , ( , )
n mP
Q
x y
R
R
n
m
p
. The eigenvalues of L and M satisfy
* *
Re
( )
Re
(
)
i i
L
jM
j
for all integer pairs
( , ) {1, 2,... } {1, 2,... }
i j
n
m
. Let the eigenvalues of be indexed in increasing order. In this paper we shall show that there exists a uniqueC
r smooth invariant manifold for (S) even when the large gap conditions
*1/
*nr
(assumed by [4]) does not hold provided the following nonresonance conditions hold. Forj
1, 2,...
n
1 1
, 2
n n
j i i i
i i
k
k
r
(1)where
( ,
k k
1 2,...
k
n)
is an integer vector withk
i
{0,1, 2,... }
r
. If the resonance condition1 n
j i i
i
u
k
oforder
2,..., r
holds for somej
{1, 2,...,
m
}
, and some integer vector( ,
k k
1 2,...
k
n)
then there are examples of n-dimensional manifolds which are less smooth than the order of the resonance condition Our assumptions are:A1)
,
n n m m
L
R
M
R
are diagonalizableA2)
* *
1
* *
1
1
mn
r
A3)
1 1
, 2
1 , j=1,2,...,m
n n
j i i i
i i
k
k
r
A4)
,
(
) ,
2
rP Q
C
B
r
A5)
(0, 0)
0 ,
2,...,
kx
D Q
k
r
under these assumptions then there exists a unique invariant manifold which is the graph of
y
h x
( )
.Remark: Note that (A3) implies that there exists an analytic change of variables which transforms (S) into
( )
S
such that Q is r semiflat with respect to x . Also
*m/
1*r
implies
k
i
*i j,
j
1, 2,...,
m
if1 n
i i
k
r
285
Let X and Y be finite dimensional Banach spaces. All of the major results stated here holds analogously for a map
:
F X
Y
.Let (D) denote the dynamical system obtained by linearizing the vector field near
z
0
. Let( )
S
denote the dynamical system obtained from (S) with Q satisfies (A5). It is known that (S) has many local n dimensional invariant manifolds [see 5] which are graphs ofy
g x
( )
whereg x
( )
may only beC
1 smooth. The partial normal form of one of these manifold can be constructed by solving the following system of partial differential equations over the set of m vectors of polynomials of n variables of order 2 and degree r provided (A3) holds (see 4),( )[
{ , [ ]]
( )
( , [ ]]}
(0)
(0)
0
Dh x
Lx
P x h x
Mh x
Q x h x
h
Dh
(M)However the system
( )
S
with Q satisfying (A5) has a unique solution over the set of rC
function which are little0(
x
r)
. This modivates the following definition.Definition 1: The generalized Poincare’ manifold for (S) satisfying (A4) is the graph of the unique r
C
function( )
p( )
( )
g x
h x
h x
whereh x
p( )
is the vector polynomial determined from (M) andy
h x
( )
is the manifold.2. MAIN RESULTS
In this paper we establish the smoothness of the invariant manfold for (S) which is defined to the graph of
( )
y
h x
forx
B
{
x
R
n:
x
}
for some
0
. The manifold function h isC
ronB
and(0)
0 ,
0,1,...,
kD h
k
r
.A manifold is positively a invariant with respect to (S) if
y t
( )
h x t
[ ( )]
fort
0
, where[ ( ), ( )]
x t
y t
is a positive trajectory of (S) passing through( , )
x y
B
such thaty
h x
( )
. From the variation of constant formula one can show that the manifold is positively invariant if and only if the manifold functionh x
( )
is a fixed point of the integral operator T defined by0
( )( )
Ms{ ( , , ), [ ( , , )]}
T h x
e Q
s x h h Q s x h
ds
(1)[ ( , , )]
( , , ),
{ ( , , ), [ ( , , )]}
d
t x h
L t x h
P
t x h h
t x h
dt
(2)
(0, )
x
0 ,
x
B
,
0
The domain of T is
G
r( )
C
r , where{
(
) :
}
r r
r
C
h
C
B
h
0,1,...
max sup
k( )
1 ,
k(0)
0 ,
0,1, 2,... }
k r x
D h x
D h
k
r
and
0
*
( )
{
(
) :
kr
G
h
C
B
h
* 0
( )
sup
k,
0,1,...
1 and
r1}
x
h x
k
r
h
x
From now on assume 1, , / 2 1
2
. Let
K
k denote a bound on,
k k
P
Q
computed overB
for1,...
k
r
andM
kdenotes a constant depending onK K
1,
2,...
K
k. Theorem 1: (i) The solution
( , , )
t x h
of (2) satisfies* 1 1 ( 2 )
( , , )
t x h
x e
K t,
x
B
,
t
0
286 (ii) The solution
i( , ,
t x h
i) ,
i
1, 2
satisfies* 1 1 ( 2 )
2 1
sup |
2[
2]
1[ ]
1 1,
0
K tt R
h
h
K e
t
1
,
2 rh h
C
.Proof: The results follows directly by using the variation of constants formula for the solutions of (2) and Grnwall inequality.
Theorem 2: T is a well defined map from
intoG
r( )
.Proof: (i) For
0
1
select0
,
/ 2
and choose
0
small enough so thatsup ( )
x
h x
. Sincesup ( , , )
t Bx
t x h
fort
0
(because L is a diagonal matrix) thecomposition
Q
{ ( , , ), [ ( , , )]}
t x h h
t x h
is well defined andC
r for( , )
t x
R
B
.(ii)
0 0
{ , [ ]}
( )
M r rp
Q
h
Th x
e
ds
* *
1 2 1 02
m r r K s r
r
e
x
K ds
* *
11 2 1 2
r m r r K x Kr
.
Since
*m/
*1r
, the integral is well defined for if
0
is chosen small enough so that* * 1 1
2
mr
K
r
.(iii)
* *
11 1
( ) k r k2 2
r m
Th x x K r r K
if 1/ r k for
* *
1 1
0,1,... ,
2
mr
k
r
K
r
and *1 *1
2 m r
r
K rK . Thus the range of T is a subset of
( )
rG
.Theorem 3: T is a contraction on
G
r( )
in the norm* r
. Proof: Select
h h
1,
2
G
r( )
, than2 1 0 2 2 2 1 1 1
0
{
,
[
]}
{ , [ ]
Ms p
Th
Th
e
Q
h
Q
h
ds
*
2 2 2 1 2 1
0
{
,
[
]}
{ ,
[ ]}
m u s
e
Q
h
Q
h
1 2 1 1 1 1
{ ,
[ ]}
{ , [ ]}
Q
h
Q
h
ds
*
2 2 2 1 2 1
2 1
0 2 1
{ , [ ]} { , [ ]}
{
ms r
r r
Q h Q h
e
2 1 1 1
1 1 1 [ ] [ ] } r r h h
K ds
2 1 2 2 2 1 1 1
0 2 1
{ , [ ]} { , [ ]} 2 m r r s r r
Q h Q h
e
2 1 2 2 2 1 1 1
2 1
{ , [ ]} { , [ ]}
2
r r
r r
Q h Q h
*
1 1 2 1 1 1
( 2 )
1
1
[
]
[
]
r r r K s
r
h
h
K e
x
ds
287
* 1 1
2 1 2 1
1 0
m
r
r k k
s
r k
e
K
*
1 1 1 2 1 1 1
( 2 )
2 1
1
[
]
[
]
2
r r r K sr r
K h
h
e
x
K
ds
* *
1 1 1 2 1 1 1
( 2 )
2 2 1 1
0 1
[
]
[
]
3
[
]
[
]
m r r K s r
r r
K h
h
e
x
K h
h
ds
* *
11
2
1(3
1)
2 1 *r
m r
r
r K
K
K
h
h
Thus T is contraction on
G
r( )
in the norm* r
if
* 1 1
2
mr
K
r
and3
K
r
(2
r
1)
K
1
r
*1 m.From now on replacing h by
h T
0,
byT
0, denoteD
xk
by
k,
D h x
k( )
byh x
k( )
and( )
kD Th x
byT
k . LetE
k denote the set of functionh x
k( )
which are continuous andB
and K linearoperators on
1 k
n
i
x R
for each fixedx
B
. The setk r k
E
G
is a Banach space with the generalized infinity norm0
0
( )
( )r k sup k p
k r k
k
h x h x
x
.
The equations of variations of order K are given below, Let id denote the identity map on
R
n,
id x
( )
x
,1
L
P
xP h
y 1
1
(3)1
(0, , )( )
1,
n n
x h h
I
R
11 ,
1
{ , ( )} ...
2
(0, , )[ ,... ](0) 0
q
k
k
i i q
k k i
q q k
n
k k
L Dp D P h D D
k
x h h h R
(4)
Similar we can prove he following theorem .
Theorem 4:
T
k is a contraction onE
k
G
r k for fixedh h
0, ,...
1h
k1 in the normr k
E
k
Gr
k
r k . Theorem 5: There exists a unique manifold (constructed as the graph ofy
h x
( )
) which is invariant with respect to (S). The map h is r times continuously differentiable onB
for some
0
and(0)
0 ,
0,1, 2,...
kD h
k
r
.Proof: By Theorem 4, and continuity of
0 1 2
( )[ ,
,...
]
kk
T
h
h h
h
in( ,
h h
0 1,...
h
k)
fork
0,1, 2,...
r
the map(
T T
0,
1...
T
r)
is a fiber contraction from0
1 1
( ) ( )
r r
r r k r k r k r k
k k
G C x E G G C x E G
By induction one can show that if
h
0
C
r
and
h h
1,
2,...
h
k
E
1
...
E
k then(
T T
0,
1,...
T
r)
is a map fromr
C
toC
k in the sense thatT h
(
0)
C
k andD t h
k(
0)
T
k(
h
0)[ ,
h h
1 2,...
h
k]
. By uniform convergence, Taylor’s theorem, and the attractivity property of the filter contraction theorem [2, 6, 8] it follows thath
* is inr
C
for some
0
. One can extend the differentialabity ofh
* toB
if one employs the continuous extensions of( , [ ])
k r k
x
D Q x h x
fork
1, 2,...,
r
.
288
( )[
{ , ( )}]
( )
{ , ( )}
(0)
0 ,
2,3,...,
kDh x
Lx
P x h x
Mh x
Q x h x
D h
k
r
(M)Systems with a form similar to (M) have solutions which provide linearizing and partially linearizing transformations of dynamical systems. Such transformations can be used to constract local horizontal and vertical invariant foliations of
B
under milder conditions than those required for the existence of linearizing transformation.The first step towards developing such a theory is to generalize concept of an invariant manifold by dropping the spectral separation requirement while retaining the explicit construction as the graph of
y
h x
( )
. Thus we can now make an arbitrary assignment of dependent and independent variables in order to consider cases of spectral overlap. If the eigenvalues
i(
j)
are associated with the independent (dependent)variables and are indexed in order of increasing real part then the following cases are possible(i)
0
1* *m *1 *n(ii)
0
1* *1u
*m
*n(iii)
0
*1 1*u
m*
*n(iv)
0
1* *1 *n *m(v)
0
*1 1* *n *m(vi)
0
*1 *n *1 *mTheorem 6: If
DF
(0)
isC
rand diagonalizable such that the (A1– A4) hold then there exists a unique invariant manifold for (S) which is the graph ofy
h x
( )
.Proof: The prove is formally identical to that of Theorem 5. The condition (A1) guarantees that one may consider a system with nonlinearity Q in r semiflat form with respect to x . The bound on the spectral gap guarantees that the maps
T
k,
k
0,1,...
r
, are contractions in the normal r k forh h
0, ...
1h
k1 fixed. Together with continuity ofk
T
is( , ...
h h
0 1h
k)
one concludes that(
T T T
0, ...
1 r)
is a fiber contraction. By induction it can be shown that Tis a map from
C
r toC
kin the sense previously discussed. Then Taylor’s theorem, uniform convergence, and the attractivity property of the fiber contraction theorem implyC
rsmoothness of the manifold.Remarks: (1) The classical linearizing conjugacy theorem of Sternberg is a corollary of Theorem 6, because case (iii) can be modified to include
* * * *
1 1
...
n m
withn
m
(2) All of the sprevious results can be extended to get
C
manifolds forC
vector fields because the limit of the contraction conditions of Theorems 5 asr
can be met by manifold function h defined onB
, for some 0.(iii) A result similar to Theorem 5 for the case when
DF
(0)
is not diagonaliable should hold but the estimates for constractions and the partial normalization are considerably more intricate. There is also a difficulty with local invariance due to polynomial factors which prevent the compositionsQ
0(
id h o t x
, )
( , )
from being well defined on finite time intervals[0, ]
T
if L is not diagonalizable.REFERENCES
1- R. Abraham and J. Robbin: Transversal Mapping and Flow. Benjamin, Boston 1967.
2- C. Bota, About the invariant and minimal sets of a dynamical system on a topological space, The Fifth International Worksop on Analele Universitat. Ii din Timi, Soara September 18-22, 2001, Romania Fascicula-Mathematica
3- J. Hale: Integral nanifolds of perturbed differential system. Ann. Math. 73 (1961), 496-531.
4- M. Hirsch, C. Pugh and M. Shub; Invariant Manifold Lect. Note in Math. No 583, Springer Velag 1976.
5- R. Dela Liave: Some new invariant manifold Theorems and applications to smooth conjugacies of dynamical system, mansicript, 1992. 6- A. Osinga: Two-dimensional invariant manifolds in four-dimensional dynamical Systems, Birstol BS8 1TR, UK December 22, 2003. 7- G. Sell: Smooth Linearization near fixed point. Amer. J. Math. Vol.107, (1989), 1035-1091.