Acta Mathematica Sinica, New Series

1990, Vol.6, No.3 ,pp . 2 4 3 - - 256

Qualitative Analysis of Van der Pol Type Equa-

tion with Periodic Forcing Term
Zheng Zhiming ( ~ )
Department of Mathematics, Peking University
Received January 14, 1 9 8 8 Revised March 21, 1989

Al~xact. In this paper we analyze the qualitative behaviour of the equation

eJ(+ q(S) J(+~X=bp (t ),

where e is a small parameter. We divide the interval of parameter b into four sets of subintervals, A,
B, C and D. For b ~ A, B or D , we discuss the different structures of the attractors of the equation
and their stabilities. When chaotic phenomena appear, we also estimate the entropy. For b e C, the
set of bifurcation intervals, we analyze the bifurcating type and get a series of consequences from the
results of Newhouse and Palis.

w1. The Vertical Deformation of r with Flow

In this paper we still use the notations and terminologies in [2].
The paper [2] describes the time-dependence of solutions of the system (2.3)
in [2]. We need to know more about their dependence on the initial value. This
dependence is characterized by the partial derivatives ~ ( t ) = ~ ( t , s ) - O x O
( ts, s ) '

q ( t ) =q ( t , s ) = O y ( t , s ) which satisfy the following system

c3s

1 (r/- go ( x ) ~ ) ,
(1.I)
rl=-tr ~(O)=l+O(e), r/(O) =1,

obtained by differentiation of (3.1) by s. We will also need the estimates concern-

ing how sharply the fiber (x, ( s ) , s) is folded after time t; these will by provided

by a ( t , s ) = 32x~ and b ( t , s ) = 02y

Os Os z , which satisfy

h= 1 (b-q~ (x)a)
__ _ 1
__ q~ , (x)~ ~ ,
8
b -sa , (1.2)
a(0)=b(0) =0.

The following propositions describe the solutions of ( 1 . 1 ) , ( 1 . 2 ) . Shortly,

we apply these propositions to analyze the vertical deformation of the "rectangle" r
with the flow. In the propositions e is understood to be sufficiently small.
Proposition 1.1. I f x (t ) behaves as described in i ) o f [2], then for (1.1) we have
244 Zheng Zhiming

r/(-c) < - e c/~, c>0.

Proposition 1.2. I f x (t) crosses only the unstable region and the. cross&g time
, 1 9
t2 - t~<c 2 e l n - - , c 2 < 2 , then for (1.1) we have r/('c)> ~ e-4>10 -2.
Proposition 1.3. Given any cl > 0 . I f x ( t , s) crosses only the unstable region
and r / ( ' c ) t [ - c l , 10-2], then for (1.2) we have
b('c) < - 2 k 2 ( 1 - ( k 0 k )-~)e-2e -2 ,
1
where ko = ~ ( l - ~l(z ) )eP~'-'2 ~and k = l + ab ( l + O) + 2 1-1~ in which 0 satisfies
the following equation
k -I
( - ~ I P + ~ p 2 ) e - e = - 2 ~ p 2 0 + 2p2 bko '

where ~, ~ and p were defined in [2].

Proposition 1.4. Given c~ as above, l f x ( t ) dips and then crosses the unstable
region, then /fr/(z) t [ - c l , 10-2] , we have

b(~)> 8k0 e_2 ~-4 In

ab
Proposition 1.5. There exists a c > 0 such that for any s t Y , we have tl(~ )
<l-c.
The proofs of these propositions easily follow from Theorem 3.1 of [ 2] and
we omit them.
We now use these propositions to describe the vertical deformation, of r with
flow. More precisely, for any fiber in r" z~(s)=(x~(s), s), s t Y , let g ( s ) = g~(s)
=y((2fl-1)Tr, z~(s), 0), where z = ( 2 j + 1)Tr is defined as in [2].
PrOposition 1.6. Given any fixed c~> O. The following statements hold.
1
(1) There exists an interval [sl,s2] ~ Y such that g'(s)< - ~ - , f o r s t [ sl,s2].
1
(2) For any st[y~, s~] , if g ' ( s ) t [ - C l , 10-2], then g " ( s ) < - - - .
1
(3) For any s t [ s2,Y2], if g ' ( s ) t [ - c ~ ,10 -2 ] , then g'~(s)>--e "
(4) g'(s)<l-c, for s t Y , and some c>O; and g'(Yl)' g'(Y2 )>10-~" Also,
g(Y2) - g (Y,) = - (Y2- Y, )+ O(e 2).
(5) Finally, if g'( s'), g ' ( s " ) t [ - c ~ , 10-2], then I s ' - s"l < c e 2 , c > 0 .
Proof. Let s~, s~ be those as in Theorem 3.1 of [2]. From i) of Theorem 3.1
1
and Proposition 1.1 we have g ' ( s ) < - - ~ - , s t [ s ~ , s2], which implies (1).By The-
orem 3. 1 we see that ( 2 ) and ( 3 ) follow from Proposition 1 . 3 and i . 4
correspondingly. The first part of (4) follows from Proposition 1.5, and the se-
cond from Proposition 1.2. Finally, we give the proof of (5). Let s ' t[y~, s~],
s U t [ s 2 , Yz] be points for which r/(z, s ' ) , r/(z, s") t [ --CI,10-2]. From Prop-
osition 1.3 and 1.4 we know that z ( t , s ' ), which crosses only the unstable re-
/ / 1
gion, spends a crossing time t 2 - t ~ = 2e I n - -,q+ O ( e ) a n d z (t, s " ) , which dips
Qualitative Analysis of Van der Pol Type Equation with Periodic Forcing Term 245
1
and then crosses the unstable region, spends a dipping time t2u-" t (-c--2e ln-~- + O ( e )
where t l ' =2n+O(e) , t '~=2r~+O(e). It is clear that we can choose a
//
time-interval [ t - , 7]; satisfying [ 7 , T] c [tl', t2'] A [ t 9 1 t~'] and
- 7 = 2e In 1 + O(e) where 7 = 2n + O(e), such that I.x (z, -s) I< 1, for t e [ 7, 7 ]
and s ~ [ s ' , s ' q . Thus we have
$~

J $ t

Since t 'l, tlU=271:'t- O ( 8 ) , we have t, ( s ) =2re + O ( e ) for any s s [ s ' , s "] and
1
- h (s)=2e l n T + O ( e ) . It implies, using an argument in Proposition 1.2, that
~( 7 9 s ) > - jc, s 6 [s ', s ,,] . Substituting this inequality into (l. 3 ), we get the desired
estimate.
To prove the foUowing important proposition, we first give a lemma and omit its
proof. Let
V, = { (x,y)l I x - I I~<2p0 },
where po=e-C/'2 , c>0.
l
Lemma 1.7. All solutions of the system (2.3) of [2] spend a time less than P0Tin Vt.
Remark. Let V_. ={(x,y)l Ix+ II ~< 2po}. Then the conclusion of Lemma 1.7also
holds in V_,.
Let y ~ ( s ) = I ~ z ( ( 2 k + l ) T z , z~(s), 0), where t r y [ - 1 , 1 ] and k = j , j + l , j + 2 ,
the integer j being defined as in ~3 of [ 2]. For yk(s) and g (s) we have the following
proposition.
Proposition 1.8. i)y~(s) depends on tr weakly: there exist c~, c2>0 such that for
any a~, a2 e [ - 1 , 11, I1~, ( s ) - ~ (s+As)I1o, ~<e-q/'2 , .for k=j, j + l , j + 2 and for
any IAsl<e-~'/"' as long as s , s+ As belong to the domains of the corresponding fl.
ii) There exist I c I ~ c Y such that
(1) g ' ( I ) ~ < - 4 . 3 2 -Y , 7 = 2 0 0 ;
(2) 23,-' < ~ g ' ( Y / I , ) < < . I - C , C > O ;
(3) The oscillation of g(s) on each component of I ~ I is O(e2) ; and II, l = O ( e 2 ) .
O (g(s ")+ )> t+e-2
iii) -~- s[ ----T-- a:~, and
0 ,,)+ l + e -2
Ob (g (s s (')> ~ a2= ,
where s ' , s " are the ends of I, s,', s [' are the ends of Ii.
P r o o f of i).
Fix any a, , a = ~ [ - l , 1 ] , and let z~(t,s), i = 1 , 2 , be the solutions of ( 2 . 3 )
with initial values zj(O,s)=(x,~ (s),s). It is enough to show that
ly~(z,s) - y 2 ( ' c , s + As )[co < e -c,/' , (1.4)
and

I' (y~(z,s)-y2(z,s+As))
/C0
< e -q/': (1.5)
246 Zheng Zhiming

We now show the first inequality . Since

[y~(z,s)-y2('c,s+ As)[= [y (z, x,t (s ), s ) - y ( z , xq ( s + A s ) , s + As)l
0 0
<~ I--~-~ Y(z, x,t (s )+ O, A x , s ) I I A x I + I -~-s y ( z , s + O2As )llAs l,while
Oy Oy
I--~-s I, I --~-xI do not exceed g/' and the initial values for zl, z2 differ by less
than e-C/"2 , these facts imply (1.4).
Remark. The estimate ( 1 . 4 ) holds for the horizontal components x; of zi as
well.
Using L e m m a 1.7, we can get inequality (1.5) and omit the simple proof.
P roof of ii ).
We first define I and It. Fix c~=4.32y. F r o m the facts that g'(y~)> 10 -5,
g'(si)<-e"~,i=l,2, it is obvious that there exist s~(yt, st), s"~ (s2, Y2) such
that g ' ( s ' ) = - c ~ , g'(s ~' ) = - c ~ . Using the reduction to absurdity, we can easi-
ly show the uniqueness of s', s ~'. Define I = [ s', s"]. [z is defined in the same
way: take s ( e ( y t , st), s~te(s2, Y2) such that g'( s( ) = g'( s ' ( ) = 1 0 -2, and let
It = [s (. s ~'] . Existence and uniqueness of s (, s "t is proven identically to that of
s ', s u. We also note that I=It and I I ~ l < c ~ ( P r o p o s i t i o n 1.6 ( 5 ) ) . We next
prove (1), (2) of ii).
F r o m proposition 1.1 we have g~(s)<-e~/"<-c 91 s~[st, sz], if e is small
enough. Therefore, to prove ( I ) of i i ) , we need only to show g ' ( s ) <~-c~,
sE.[ s', s,)1,) (s2 , s # ] . A s s u m e the contrary: there exists an s*~[s',s~) U (s2,s"]
such that g ' ( s * ) > - c~. Without loss of generality we suppose that s * E [s ',st ),
g ' (s*)~.<10 -2 and g'(s) 6 [ - q , 10-2], for all s s [ s ',s*]. These inequalities, by
Proposition 1.3, imply that g U ( s ) < 0 , s6[s ~,s*]. But since g'(s*) > g ' ( s ' ) ,
there exists an ~" s (s /, s*) such that g " > 0 , which gives a contradiction. In the
same way, we can show that g'(s) > 1 0 -z, for all s ~ Y \ I t .
The proof of (3) of ii) is obvious. In fact, for s~It~I, g~(s)e[ - c t , 10-2],
and It = O(~ ) (Proposition 1.6 ( 5 ) ) ; therefore the oscillation of g on either com-
ponent of I t \ I is less than c ~ . This proves ii).
P r o o f of iii).
To finish the proof of (iii), it suffices to show
l + e -2
(1) - ~ - y ( z , s b ) > - - 0~2~ ,
3

0 y(z,s';b) > l + e -2
0~ 2 ~, "~
Ob 3
d d d (= d //
(2) ~ s ' = o (1) , db S =o(1), -d-ffs o ( 1 ) and ~--b- s l= o (1) .
Indeed, if (1), (2) hold, then

0 (g(~, s ', b ) + s O = Og(s' ) ds' + Og + > _l + e_ -2

Ob Os db Ob 4 ~27~ "

Same for the right ends of I, It, i. e. iii) holds.

Now we prove (1). Let q-- ~
0
x ( t , s, b), r = - ~b
- y (t , s , b). Then (q,r)
Qualitative Analysis of Van der Pol Type Equation with Periodic Forcing Term 247

satisfies the following system

= 1 (r- r
(1.6)
[" = - s q + p ( t ) , q(O)=O(e) , r(0)=0.
Tracing the evolution of ( q , r ) , it follows from the definition of z and
1 1 ( 1 + e_ 2 ) a27~.
t ~ - t ~ = 2 e t n - - + O ( e ) that r(l:) > -~-
P r o o f of (2).
We outline now the proof of (2); we do it only for s ', estimations for other
derivatives will not have any more essential difficulties. By definition of s ',

Os y ( ~' s ",b ) = - c i, differentiation w . r . t b gives ds ' _ g s'1b ( s ' , b ) Since

" ab g"s (s ' , b ) "
c (Proposition 1.3), ( 2 ) will follow if we show
at s = s ' w e have [ g ,'~,(s,b ) [ > --~
that

g sb ~

' ( t " S ', b ), v = y ~b

The proof runs as follows. Let u = x ~b '~ ( t , s ', b). Then (u ' v)
satisfy the system with initial data:

t~ = -
1 (v-tp(x)u)-
- -
1
-

(1.7)
;~= - e u , u(O)=O(e-C/~2), v(O )--0.

Tracing the evolution o f ( u , v), w e h a v e v(z)=O(+). The proof of Proposi-

tion 1.8 is completed.
w 2. Basic Properties of F
In this section we will prove that the map F satisfies the condition H. We
first introduce the condition H , which consists of two parts. The first part
describes the map F in term of circle map o f f ; t h e second part of the condition
H imposes some conditions on d F which allow us to reconstruct the information
on F from that on the circle map f . The details are as follows. Let

f~(u)=I-IuF(z,(u) ), (2.1)
here u ~ S 1, a ~ [ - 1,1], z, ( u ) = (u, o') and 1-I, being vertical projection on the
u - axis. Fix o', and let z = z , ( s ) = ( x , ( s ) , s ) belong to the corresponding fiber
in r. we recall that N+_ (z) = z ( ( 2 k + l ) n , z, 0 ) = ( x ~ , y ~ ) , k being the
smallest integer such that (x k, y k ) ~ - r . Since z(5rc, s) ~ - S , it is easy to prove
k = j , j + 1, or j + 2 (for the definition o f j see [2] ), and we omit this simple
proof. Applying the linear map T to (x~, y~), we can rewrite (2.1) a s
1
f~(u) = -d-- ( ( - Y ~ + 1 - 2rcp)cos 0 + (x~ + 1 )sin0), (2.2)
248 Zheng Zhiming

where u is related to s via T , d is the distance between the straight lines on

which the upper and lower sides of r lie correspondingly, 0 is the angle between
x -axis and the straight line on which the lower side of r lies.
Condition H. An annulus map F: R--~ R, F = F ( 9 , b ) w i t h the parameter
ranging in b~<~b<~b2, satisfies the following properties (H~, H2):
H~. Properties o f f , (u)=f~ (u, b).
( 1 ) f ~ ( u ) depends on tr weakly:
I f , - f ~ , Ic0= O(e-C/e), o-, r .
( 2 ) There exist 7 = 2 0 0 , k~>32, 0 < c < l , and two intervals A cA~ c S z ,
whose ends depend on b (not r ), and IAtl = 0 ( 8 ) such that
fd(u) > / q , , for u e A ,
-l+c< f d ( u ) < - y -~, for u e S ~ N A~ .
( 3 ) Oscillation o f f ( u ) on each of two components of A~",. A is less than

(4) The ends o f f ~ ( A ( b ) , b) rotate clockwise with respect to the ends of

Al(b) as b increases:
d
db (f~(x~(b), b ) - x / ( b ) ) < - o g , i = 1 , 2,
/__\

where x~(b), x'~(b), i = 1 , 2 are the ends of A, At correspondingly, and c o = O ( + ;

k /
is independent of b. f~(u) is illustrated in the following figure.

lifo(u) To state the second part of H, we first intro-

duce some notations. Denote
Q= {(x, y)
Q, = {(x,y)eR, x eA, }.
},

We also define a horizontal sector in R 2 as

,7 ), 1,71 < # 141}.
Similarly for Pa we call the set
JL--U
A 1
v,2= {(4. , ) . 141<u21,1 }
Fig. 1 a vertical sector.
H2. Properties of dF.
(1) dF(z): H,~-" H,, V z e Q U Q~.
(2) dF-'(z) : V~n ~; V~/2 , V z e F ( Q U Q~).
(3) dF-l(z), zeF(QU QJ), is expanding in Vw2: [q-,[>kl [q[, where (4, r/)
V~/2 ,, (4-i, ~l-,)=dF-l(z)(~,q) and k, > e -l.
(4) d F ( z ) , zeQ~, is contracting : IdF(z)[ < c < l , V z e Q ~ .
The following lemmas describe the connection between x k and yk in (2.2) and
how these coordinates depend on k. They provide the desired connection between
f~ and g , which reduces the analysis off~ to that of g.
Qualitative Analysis of Van der Pol Type Equation with Periodic Forcing Term 249

Lemma 2.1. There exist c~=c~ (b, e) and c2 such that for i= k, k - 1, k - 2,

(1) Y-cox~=c,+O(e-~/'=), ds (y'-c00d)= O(e_~/~2 ), ( 3 ) s - c o x , ( s ) + c

(2)--~- ~

=O(e-~, /'= ), (4) ~ 3 (y~-c0od)= O(e_~/~ ), where co= --}-

1 ( l + ( 1 - 4eE )T1 ).

Lemma 2 . 2 . For i=k, k - I , k-2, we have (1) yi-y~-~=d+O(e2),

d i ~ d dd = O ( e ) ,
(2)--d-~sy=Z--~y'-'+O(e-~/e),(3)-~-- (4)--~-
a (yi_yi-i ) = o ( 1 ) , as
d
I--~s yq<~c, where 2 = e - 2 ~ ~ 1+ O(~).
The proofs of the above lemmas are the same as those of corresponding con-
clusions in [3] and [1] and we omit them.
Now we prove the main theorem in this section.
Theorem 2.3. The map F, induced by the system (2.3) of[2] , satisfies the re-
quirements stated in H.
Proof. We first prove that f~ satisfies H t .
Let z, ( s ) = T-t(u, tr). From the definitions of T and r, we have G , ( s + k s )
= T - l ( u , tr'), here IAsi=O(e-~/~).We first suppose k = k ( G ( s ) ) = k ( G , ( s + A s ) ) .
Then

f, ( u ) - f , , (u) = - 71 ( - ( y , - y , , ) c o s O + ( ~ - ~ , ) s i n 0 ) .
By Proposition 1.8 and Lemma 2.1 (1), we have

I = O(e -c'e)
and
Ig-~, I = c01 lyff-y~, l + O(e -c'e ) = O (e-C/'=).

These estimates imply If,(u) - f , , (u)l= O(e -J3 ).

For k(z,(s) ) v~ k(z,,(s+ As ) ), we suppose k = k (z,) < k(z,, ). By Proposition
1.8, l y k , ( s ) - y } , ( s + A s ) l = O ( e - ~ / ' 2 ) . This implies that there exists a point z*=
(x*, y * ) 6 - l - i such that l y * - ~ , ( s + As)I=O(e-C/~'). Thus IDly*-y~, %') (s+As)[
= O(e-C/'2). Since we identify z with Dz in r, we have [yk,%)(s)--yk,~"r As)I
= O ( e -J~ )and [xkC~,')(S)--xk! ~') (S + As)[ = O (e-~/e). These estimates imply
HI (1).
To continue our discussion, we prove the following estimates.
1 (s+ 1 -2r~p+E~), where E~= O(e2), ~d & = O ( e 2) and
(a) u = -~- O = O(e=);
cob

(b) If ~ ' ( u ) l < c , then I f ~ ' ( u ) - f , ' ( u ) l = O(~);

C
(c) If [ f j ( u ) I > c , then [ f / ( u ) l > -~- 9

Since
250 Zheng Zhiming

1
u= --ff ((s+ 1 - 2 n p ) c o s O - ( x , - 1 )sin0)

= 1 (s+ 1 - 2 r c p + E t )
d
d
where El = (s+ 1 - 2 z r p ) ( c o s 0 - 1 ) - ( x ~ - 1 )sin0. Then ~ x~(s) = 1+ O(~) and
0 x,(s)=O(e) imply that (a) holds.
Ob
Proof of (b). If k= k( G )= k(G. ), we have from Proposition 1.8 and Lemma 2.1
that
d
If" (u) -f,',(u)l = I --~- ( ~ ( s ) - ~ , ( s + As) )l(1 + O(e) ) = O(e -~/~ ).

If k ( z , ) < k ( z , . ) , then Lemma 2.1 and 2.2 imply that

Ifj(u)-f~,(u)l <~ d ' (s) - y*,!'.' (s+ as) )1 (1 + o ( ~ ) )

+l__~_s
, dy.k~--.)(s+As)ifA~(l+O(e ~ ) ) - 1 [

where i = 0 , 1 , or 2. The boundedness o f f ' , ( u ) implies that ~d y. ~z)

, (s+ As) is
bounded, and thus we obtain the desired estimate (b).
Proof of (c). If,'(u)l= If~', ( u ) - f " (u)+fd(u)l~>14 y*,5"" )(s + As)l(1 + o(~))
> - ~C- . The last inequality is got from the definition o f f , ( u ) and the condition
G,(u)l>c.
We now prove HI ( 2 ) . Fix a=ao and let A c [ 0 , 1] correspond to I c Y :
A=I-I,,T(z,o (I) ). Analogously we define A~= I-[~T(Go(It)). Obviously,
[A~l=O(e) . It follows from Lemma 2. I and 2.2 that for all u~A.
f~o (u) = -g~o(S)2' (1 + O(e) ) , s 6 I.
Using Proposition 1.8 and estimate (c) , we thus obtain f ' ( u ) > 3 2 y , for all
u~A, a ~ [ - 1,1].
We can also look at the second estimation of ( 2 ) in the same way. In fact,
we have from Proposition 1.8 and estimate (b) that
-l+c<f,'(u)<-y -1, u 6 S ' ~ A l , a ~ [ - 1 , 1 ] .
Proof of (3) is simple. S i n c e ~ ( u ) can be rewritten in the form
1
f , ( u ) = ~ ( - g , . ( s ) + 1-2r~p+Ez ) - i ,
where i = 0 , 1, or 2, and E2= O ( ~ ) . Note that R is an annulus, we have from
Proposition 1.8 that
_ _

n,,2= ~ (n,,~+ o ( d ) ) = o(~),

Qualitative Analysis of Van der Pol Type Equation with Periodic Forcing Term 251

where ~"~1,2 ' ~"~1,2 are the oscillations o f f , g on each of the two components of
A ~ A , I t \ I respectively .
Finally, we prove HI (4). We only prove the first inequality of (4) , because
the second can be proven in the same way.
We have from Lemma 2.1 and 2.2 that
1
f,(x, (b))= ~ ( - g , ( s " ) + 1-2rcp+ E3), (2.3)

0
where ~ E3= o(1). Moreover, Proposition 1.7, 1.8 imply that
0 0
Ob g , ( s ' ) = ~ go(S')+O(e). (2.4)

The estimates (2.3) and (2.4) imply that we can rewritef,(x,(b))in the form
1
f ' ( x ' ( b ) ) = - d ( - g ~ o ( S ' ) + 1 - 2 n p + E4), (2.5)

where ~ 0 E4= o (1). On the other hand, we have from (a) that
1
x,'(b)= ~ (sl'+ 1 - 2 r c p + El), (2.6)

O
where - ~ - E l = O ( e ) . (2.4), (2.5)and (2.6) imply that
d" (f, (xl(b) ) - x,'(b ) ) = - (g~ (s ') + s,'+ E5 ), (2.7)

where - ~0- E5 = o (1). Noting that d, obOd = O ( e ) , we have from Proposition 1.8
that
0 c
Ob ( f , ( x t ( b ) ) - x ~ ( b ) ) < - - - ' e c>0,

by differentiating ( 2 . 7 ) w . r. t. b. This completes the proof of H I.

We will show that H~ holds, it follows from the definition of the map T "
r--,- R that for R, H, , and V,t2 correspond to the sectors

H= (~,r/): 1-cte<--;" < l + c 2 e ,c,, c1>0,

and

W e a,so not t at for to map

" x ' l l
(~,r/)0 ~-,- (~ ,r/),, which is given by the linearized system ( 1 . 1 ) along the solution
z(t,s) of (2.3) in [2] with the initial value z(O,s)=z,(s) which belongs to a fiber
in r. To prove H 2 , it suffices to show that
(a) L,V~ V, Lj- I c H , for all s ~ I U ( Y \ I , ) , and
(b) L ; " is expanding in Vfor s e I ( . . ) ( Y \ I,) by a large factor M(e) , more
252 Zheng Zhiming

precisely, if G0 =(G0, r / 0 ) = L ~ t ( ~ , ql), (~1' /']1)I~g, then Ir >M(e)I~,[,where

M(e)~ -~ oo, e ~ O, and Ls is contracting in R 2, for s t Y \ I ~ .
P r o o f of ( a ) .
Let (t, (2r be two points in R 2 on the different boundaries of the sector
V, such that the segment joining them belongs to V. It is easy to show that L,
(t, L, (2 belong to the different connected .components of the complementary sector
R2~ V. Thus we have LsIV~V.
We prove the second part o f ( a ) . Let ( ( 0 ) = ( ~ ( 0 ) , r / ( 0 ) ) ~ H . Then

r/(0) _ l + O ( e ) . T o prove L ~ H c H , we need only to show r/(z)

~(.z) = 1+ O(e:).This is
4(0) -
done in the following veay. It is easy to prove

r/(10u) =l+O(e), for all s ~ I ( J I ~ ,

as stated in Proposition 1.1, 1.2. Let u ( t ) = ~t/(t)

(t) " Then for t > 10u, we have
1 I
/J=--g-- --U2+ - - U,
g 8
u( 10re ) = 1 + O(e). (2.8)
For the time-interval [10rc,z ], this differential equation has a stable stationary
l+ /1-4e
point uo= 2 Let ut= u-Uo. Then we can rewrite (2.8) as

~1 -4e 2 1
u~ = u t - - - u~ , u~(lOzc)=O(e). (2.9)
g

Solving the equation, we get Ul(Z)=O(e-C/"2). It implies u ( z ) = 1+ O(e 2) , and

this finishes the proof of (a).
P r o o f of ( b ) .
We first prove that L~-tis expanding in V by a large factor. Let s ~ I , for in-
s t a n c e , and show that for ( 0 = L ~ t ( t with (,~ V ( t h u s (o~ V ) , we
have 140 [ > M I~l, with M = M ( e ) - " oo , e --~ 0, where I(il-- max(lr l, 117,1) = I~il,
i = 1 , 2 . Without loss of generality we choose I~1--1, then we, want to show
I~0 I > M . We only consider the case where z ( t , s ) c r o s s e s the unstable region only;
the case where it also dips can be treated analogously. In our notation
0 < h < t 2 < 5 7 z are the times when the solution z ( t , s ) crosses the lines x = 1,
x = - 1 respectively.
Let ( ( t ) = ( r q ( t ) ) be a solution of the system (1.1) with ( ( 0 ) = ( 0 . Trac-
ing ( ( t ) backward in time , we have I((0)l>ec/"l((z)l. Thus M ( e ) = g / " , c > 0 .
Finally, we show Ls is contracting in R 2 for s.s Y \ It. Since for s ~ Y \ I1,
IILJI= sup [L,(01, it is sufficient to show ILs(0 l< c < 1, for any [Col= 1. Let (1= (1,1),
I(01~t
(2=(1, -1) and Lt be the segment joining ( t and (2. we prove only
Qualitative Analysis of Van der Pol Type Equation with Periodic Forcing Term 253

ILs(0l<c< 1, for all (0610; other cases can be treated analogously. To prove our
statment, it suffices to show ILs~il<c<l. i = 1 , 2 . In fact, for any (0E/t, we can
write G0 as (o=t~ t + ( 1 - t ) ~ z , t ~ [ 0 , 1 ] . Thus IL~0l < ~ t l L , ~ t l + ( 1 - t ) l L ~ l < c , and
we finish our proof.
From the lineared system (1.1) with initial value ~, we have ff ( t l ) = (~ (t~),
r / ( t L ) ) = ( l + O(e), 1+ O(e)). Since sell, by Proposition 1.5 we have that
10-~(1+O(~))~< r/(z) ~ < ( 1 - c s ) ( l + O ( e ) ) , 0 < c s < l . (2.10)
It follows from (2.10) and Lemma 2.1 that I Z ~ [ < c < l . Analogously, we also
have IZ, d21 < c < 1. This finishes the proof of (b) and the theorem is established.
F ( R ) is illustrated in the following figure.

,V'Q

Figure 2

w3. Main Results

We now describe the main results of the map F. Since the annulus map F=
F(',b, e) determined by system (2.3)of [2] satisfies the condition H, then as we
said in [2], F has the following behaviour (see [1] Theorem 2.3).
Theorem 3. 1. For e small enough, the interval [bl ,bz] consists of two
alternating types of subintervals Ak , Bk separated by the bifurcating gaps Cj :
t b,, =A, Uc, Us, Uc UA U---A,Uc ,-IUB,,
such that throughout each subinterval Ak or Bk , the qualitative behaviour of the
map F is preserved, and
I ) For b ~ Ak, the map F has exactly one pair of fixed points, a sink and a
saddle. Moreover, F is an Morse-Smale diffeornorphism (see Figure 3 ( a ) ) .
II ) For b ~ Bk , the following holds.
(1) The map F possesses two sinks and an invariant hyperbolic Cantor set C.
Any point not on the stable manifoM of C tends to one of the two sinks for the for-
ward iterations. Similarly, any point not on the unstable manifold of C leaves R u n ~
der the backward iterations.
(2) Let
~B={o'=(...o'_~o'0o-~-..)ta;=0,1,2,3; o',.o-j+~00, 10, 21, 22, 23, 30}.

n is equipped with usual topology. Then, the map F Ic is described in terms of the
left shift 0 acting on ~ B " More precisely, there is a homeomarphism z : ~',n ~ C
254 Zheng Zhiming

such that Flc = zOz-', where O" ~ n ~ ~ B is defined by [Off]i = (7i+1'

(3) The measure of the stable and the unstable manifolds of C is zero.
(4) The map F is structurally stable (see Figure 3 ( b ) ) .

(a) (b)
Figure 3

III ) In every gap Cj there exists a bjeCj for which F has a nondegenerate
homoclinic tangency. Applying the results of Newhouse and Palis ([ 4 ] - [6] ) we ob-
tain uncoantably many values of b ~ Cs , for each of which F has infinitely many
sinks. Furthermore, F passes through infinitely many open intervals with different
structurally stable topological types. More precisely, for any given v > 0 there exist a
5 > 0 and an open subset B~ of 5- neighborhood U~ of bs such that for b ~ U 6 ~ B 6,
F is structurally stable with the bifurcations confined to B 6 . The measure o f B6 is
less than yrS. U6~Ba consists of infinitely many components to which there corre-
spond infinitely many structurally stable types of F.
2/-1
IV)

Considering the relation between F and Poiricare' map D, we obtain the

global behaviour of D from Theorem 3.1.
Theorem 3.2. For any e small enough, the following results hold.
D= D (', b, e) has exactly one totally unstable fixed point zo (b ) 9 Moreover,
I) For b e A k , D has exactly one pair o f periodic points o f period ( 2 n - 1 )
1
with an integer n = n ( e , k ) ~ - - ~ constant throughout each Ak, and n(e, k+ 1 ) =
n ( e , k ) - 1. One of the two points is a sink and the other is a saddle. Any point
which does not lie on the stable manifold of the saddle (except for zo(b ) ) tends to
the sink. We also note that D is structurally stable.
II ) For b ~ B k , the invariant set of D consists of (besides the totally unstable
point zo(b ) ) two sink-saddle pairs of periods 2n+ 1 and 2 n - 1 and o f an invariant
2n+3
hyperbolic Cantor set C*, to which the saddles belong. The set C*= ~ D iC.
i=0
The attractor I'D of D is simply the Cantor set C* together with its unstable
manifold and the two sinks. Because the measure of C* and itsstable(and unstable)
Qualitative Analysis of Van der Pol Type Equation with Periodic Forcing Term 255

manifolds is zero, almost all points tend to one of the two sinks for positive iterates
and almost all points tend to zo(b ) or oo for negative iterates.
The set of rotation numbers of 1) is exactly a closed interval 2n+ 1 '

2re -] D also is structurally stable, just like That in .the case I).
2 n - 1 .J "
III) In every gap Cj , there are uncountably many values of b ~ Cj, for each
of which the map D has iniinitely many sinks. In addition, as b travels through Cj,
the map D passes through infinitely many bifurcations separated by infinitely many
open b- subintervals, to which there correspond infinitely many different (structurally
stable) topological types of D.
2
IV ) F o r b> - ~ , the map D has a unique totally stable fixed point (See [16] ).
Finally, we give an estimate of the topological entropy o f F for b ~Bk. Let
h ( F ) denote the topological entropy of F. Then we have
Theorem 3.3. For b ~ B k , h(F)>~log ( 1 + ~ ' 2 - ) .
P r o o f . By the definition of topological entropy , h ( F ) >~h ( FIc ) 9
Since FIc=~Oz -~, where ~:~s~-,- C is a homeomorphism, we have h(FIc) = h ( 0 ) ,
where 0, acting on ~ s , is a left shift. To prove the theorem, we now need only
to show h ( 0 ) = l o g ( l + ~ v / ~" ). Define a matrix B :

1 1 1t
l
0
0 1 1 1
B=
1 0 0 0
0 1 1 1

Then y ' n can be rewritten as

~'n= {a=(...a_taoa,...)la,=O, 1, 2, 3 ; b,.,,+~ 4=0},

where bij, i, j = 0, 1, 2, 3, are the elements of B. It is easy to verify that B is an

irreducible 4 x 4 matrix and the largest positive eigenvalue of B is 1 + V~- 9 Thus
we have from [14] that h ( 0 ) = l o g ( l + x / ~ - ), thus completing the proof of the the-
orem.
Acknowledgement.
The author would like to thank most sincerely my supervisor, Prof. Zhang
Zhifen, for her guidance and help.

References
[ 1] Levi, M., Qualitative analysis of the periodically forced relaxation oscillation, Ph. D. thesis, New York
University, 1978.
[ 2] Zheng Zhiming, The time dependence of the solutions of a Van der Pol type equation with periodic
forcing term (to appear).
[ 3] Chaotic phenomena in a Lienard equation with periodic forced oscillation, Aa~ances
in Math., 14 (1985).
[ 4] Nowhouse, S., Ditfeomorphisms with infinitely many sinks, Topology, 13 (i974), 9 - 1 8 .
256 Zheng Zhiming

[5] Zheng Zhiming , The abundance of wild hyperbolic sets and non-smooth stable sets for
diffeomorphisms, IHES, 1977, 101-151.
[ 6] &Palis, J . , Cycles and bifurcation theory, Asterisque, 31 (1976), 43-141.
[ 7] Golubitsky, M. & Guillemin, V., Stable mappings 'and their singularities.
[ 8] Palis, J., On Morse-Smale dynamical systems, Topology, 3 (1969), 385-404.
[ 9] Robbin, J., A structural stability theorem, Ann. Math., 94 (1971), 447-493.
[10] Robinson, C., Stability theorems and hyperbolicity in dynamical systems, Rocky Mount. J. Math., 7
(1977).
[11] Moser, J., Stable and random motions in dynamical systems, princeton university press, 1973.
[12] Smale, S., Differentiable dynamical systems, Amer. Math. Soc., 73 (1967), 747-817.
[ 13] " , The mathematics of time; essays on dynamical systems, economic process and related
topics, 1980.
[ 14] Peter Waiters, An Introduction to Ergodic Theory, 1982.
[ 15] Guckenheimer, J. & Holmes, P., Nonlinear oscillations, dynamical systems and bifurcations of vector
fields.
[ 16] Lloyd , N. G., On the non- autonomous Van der Pol equation ~ t h large parameter, Proc. Camb.
Phil. Soc. (1972), 213-227.