Lecture 5

Outline:

Limit Cycles

Definition and examples How to rule out limit cycles

Gradient systems Liapunov functions Dulacs criterion

Poincare-Bendixson t eorem Hopf !ifurcations Poincare maps

Limit Cycles

D"#: $ limit cycle is an isolated closed tra%ectory&

"xamples:

Heart !eat' pacema(er neurons' daily r yt ms in !ody temperature' c emical reactions' &&&

Limit Cycles

$re in erently a non-linear p enomenon

) at a!out closed or!its in linear systems*

=A x x

purely complex ei0envalues

+ ey can ave solutions t at are closed or!its, Linearity implies: if x-t. is a solution so is c x-t. $mplitude is determined !y /Cs' any small pertur!ation persists forever

/n contrast: in limit cycles in nonlinear systems t e structure of t e system determines t e amplitude and s ape of t e limit cycle&

"xamples -1.

2ot ard to construct examples of limit cycles in polar coordinates

x-t.' y-t. -3 sine or cosine' -3 periodic functions in x $nd y directions

4an der Pol Oscillator

dx8dt

"5uation arose in connection wit nonlinear electrical circuits used in t e first radios Dampin0 term

Causes lar0e aplitude osci6s to decay' !ut pumps t em !ac( if t ey !ecome too small

4dP osci as a uni5ue sta!le LC for eac 37

) at to do*

9uppose we want to analy:e a 0iven ;d flow

<ule out existence of limit cycles

/ndex t eory -last lecture. )e (now t e system !elon0s to a class of systems t at cannot ave LCs --3 e&0& Gradient systems. Construct Liapunov function Dulac6s criterion &&& Poincare-Bendixson t eorem

Prove existence of a LC

Gradient 9ystems

D"#: $ gradient system is a system t at can !e written in t e form x = V & 4 is t en called a potential function& <emar(s:

/n components:

x = x V ( x , y ) y = y V ( x , y )

"very 1d system is a 0radient system How to c ec( if a ;d system is a 0radient system* = f ( x , y ) is a 0s' t en: f / y = g / x /f x y = g(x , y)

"xample Gradient 9ystems

y + 2xy 2 2 x+ x y

-1.

/s -1. a 0radient system* /f so' find a potential& C ec( condition f / y = g / x

f / y =1 + 2x

g / x =1 + 2x
2

#ind potential* Partial inte0ration&

V x = y + 2xy V = xy x y + C ( y ) 2 2 2 V y = x + x C y ( y )= y = x+ x y 2 3 dC / dy = y C ( y )= 1 / 3 y 2 3 V ( x , y )= xy x y + 1 / 3 y

Gradient 9ystems

+H"O<"=: Closed or!its are impossi!le in 0radient systems& Proof:

9uppose t ere is a closed or!it& Consider c an0e in 4 after one circuit& On t e one and: 4>7 -4 is a scalar. On t e ot er and:
T T dV ) dt V =0 dt =0 ( V x dt T 2 V =0 x dt < 0 -contradiction.

Liapunov #unctions

Occasionally possi!le to find ?ener0y-li(e@ functions t at decrease alon0 tra%ectories D"#: Consider a system dx8dt>f-x. wit a fixed point at xA& $ Liapunov function is a cont& diff& real-valued function 4-x.' suc t at:

4-x.37 for all xBxA and 4-xA.>7 -4 is pos& definite. d48dtC7 for all xBxA -all tra%ectories are ?down
ill towards xA.

+ en xA is 0lo!ally asymptotically sta!le&

/n particular t e system as no limit cycles&

Liapunov #unctions

?Proof@: ) y does t e existence of a Liapunov function rule out closed or!its*

Consider c an0e of 4 alon0 a closed or!it& 4 is real-valued' after one circuit one s ould ave 4>7 On t e ot er and:

V =0

dV dt < 0 dt

"xistence of a L# implies t at tra%ectories move monotonically down t e 0rap of 4 toward xA

Liapunov #unctions

Pro!lem: How to find Liapunov functions* 2o 0eneral met od' re5uires divine inspiration :-. Often sums of s5uares wor( &&&
x y

= x + 4y 3 = x y
2

-;.
2

$nsat::

V ( x , y )= x + a y

3 V =2 x x + 2a y y = 2 x (x + 4y )+ 2ay ( x y ) 2 4 =2x + 8xy 2ay 2axy

C oose a>D

=2x 8y

Liapunov #unctions

/s V ( x , y )= x + 4y -;.*

a Liapunov function for

4 3 7 for all -x'y.,>-xA'yA.>-7'7. 4-7'7.>7 d48dt C 7 for -x'y.,>-7'7. !y construction&

+ eorem a!ove t en implies t at -7'7. is 0lo!ally asymptotically sta!le&

Dulac6s Criterion

Let dx8dt>f-x. !e a cont& diff& 4# in a simply connected su!set < of t e plane& /f t ere exists a cont& diff& real-valued function 0-x. suc t at ( g x ) as one si0n t rou0 out < t en t ere are no closed or!its lyin0 entirely in <&

) dA = g x x n dl A ( g C
B7 >7 since

x n

Dulac in Practice &&&

<e5uires a real-valued function 0 wit t e a!ove properties 9ame pro!lem as wit Liapunov functions

2o 0eneral met od to find 0' divine inspiration &&&

9o far: ne0ative results -rulin0 out limit cycles.& How to prove t at limit cycles must exist*

Poincare Bendixson + eorem

9uppose

< is a closed !ounded su!set of t e plane and dx8dt>f-x. is a cont& diff& 4# on an open set containin0 < < does not contain any fixed points + ere exists a tra%& t at is ?confined@ in <

+ en' eit er C is a closed or!it or C spirals toward one& < contains a closed or!it&

P-B-+

Limits types of !e aviour in t e p ase plane:

#ixed points' oscillations 9elf-sustainin0 oscillations' !ut no chaos,

PB+ in applications:

#irst two conditions are easy to s ow Often ard to construct a ?trappin0 re0ion@ to prove existence of a confined tra%ectory 9ituation often simpler' if system as simple representation in polar coordinates

P-B-+ in $ction

Consider:

r = r ( 1 r )+ r cos = 1

Does t e closed or!it at r>1 survive for small *

r max ( 1 r

2 max

)+ r max cos < 0

e.g. r max > 1 + 2 r min ( 1 r min )+ r min cos > 0 e.g. r min < 1
+ ere must !e a limit cycle in

0.999 1 < r < 1.001 1 +

Hopf Bifurcations

How can a #P lose sta!ility*

9ta!le node +ranscritical' 9addle node' Pitc for( !ifurcations

9ta!le spiral Hopf !ifurcations

9upercritical Hopf Bifurcation

Below !if& point: small oscillations 0radually a!atin0 $!ove !if& point: small oscillations 0radually !uilt up to small amplitude 9ta!le spiral c an0es into an unsta!le spiral surrounded !y a small elliptical limit cycle

9upercritical Hopf E "xample -7.

controls sta!ility of #P controls fre5uency of rotations B controls dependence of fre5uency on amplitude for lar0e amplitudes

C7: ori0in is sta!le spiral >7: ori0in still is ?wea(@ sta!le spiral 37: ori0in unsta!le spiral F limit cycle at r =

9upercritical Hopf E "xample -1.

) at appens to ei0envalues*

/n cartesian coordinates &&&

sin x =r cos r =( r r 3 ) cos r (+ br 2 ) sin 2 2 2 2 =([ x + y ]) x (+ b [ x + y ]) y = x y + cubic terms ... y = x + y + cubic terms J=

x = r cos y =r sin

= i

9upercritical Hopf Bifurcation

"xample illustrates two 0eneral rules:

9i:e of LC 0rows continuously from :ero c

#re5uency of limit cycle at !irt is 0iven !y /m- . at

However:

"i0envalues 0enerally don6t cross x>7 in strai0 t lines $t !irt limit cycle 0enerally is an ellipse -not circle as in example.

9u!critical Hopf Bifurcations

Cu!ic term now desta!ili:in0,

Poincare maps

Gseful for studyin0 swirlin0 flows dx8dt>f-x. n dim& flow' 9 n-1 dim& surface of section transverse to flow Poincare map is a mappin0 of 9 to itself o!tained !y followin0 tra%ectories from one intersection wit 9 to t e next' i&e& if x( is (t intersection' t en: x(F1>P-x(.&

#P of P correspond to closed or!its

Poincare maps -1.

Converts pro!lems a!out closed or!its -tou0 . into pro!lems a!out fixed points of a map -easy . "&0& sta!ility of limit cycle -3 sta!ility of fixed point of Poincare map However: typically impossi!le to find a formula for P

"xample

= 1 Poincare map for r = r ( 1 r )

Let 9 !e positive x axis' time of fli0 t for one return is t>;&

r1
0

2! dr =0 dt =2 ! 2 r ( 1 r ) 4 !

r 1=( 1 + e

( r 1 ))
2

2 0

1/ 2

P ( r )=( 1+ e

4 !

( r 1))

1 /2

* * r = P ( r ) #ixed points*

r * =1

Co!we! Construction

rA is sta!le and uni5ue -3 system as sta!le limit cycle 0iven !y t e circle r>1&

Linear 9ta!ility of Closed Or!its

Given a system dx8dt>f-x. wit a closed or!it& /s it sta!le* -3 /s correspondin0 fixed point of Poincare map sta!le* "xplore fate of an infinitesimal pertur!ation v 7&
x + v 1= P ( x + v 0) * * 2 = P ( x )+[ DP ( x )] v 0 + O (v 0 ) v 1=[ DP ( x *)] v 0
* *

DP-xA. is lineari:ed Poincare map at xA

Linear 9ta!ility of Closed Or!its

+ e closed or!it is linearly sta!le if and only if for all ei0envalues of DP-xA.' i'i>1'&&&'n-1 it olds HiHC1&
n 1 j=1 n 1 j=1 n 1 * j =1 n 1 j =1

v 0= " v j , 0 e j

v 1= DP ( x ) " v j , 0 e j=" v j , 0 j e j

v k =... =" v j , 0 ( j )k e j

+ e ei0envalues i are also called t e c aracteristic or Floquet multipliers of t e closed or!it& /n 0eneral' re5uires numerical inte0ration &&&

9ummary

Iou s ould remem!er:

) at a limit cycle is and w y it is only found in nonlinear systems Conditions under w ic t ere cannot !e a limit cycle

Gradient systems and Liapunov functions

+ e Poincare Bendixson t eorem and w y it is important Hopf !ifurcations Poincare maps and ow to ar0ue a!out sta!ility of closed or!its