PX391 Nonlinearity, Chaos, Complexity SUMMARY Lecture Notes 7- S C Chapman Limit Cycles Have found that orbits cannot cross, can be attracted to (fixed points), etc. One other possibility is limit cycle. ODE is 'well behaved' ie: all derivatives exist and are continuous – Therefore, all orbits smoothly follow neighbours in phase space. One other possibility only: orbits approach closed curve as t!" limit cycle → NB – complete description of all details is non trivial – here give the basics. Limit cycle – an example F = x + y ! x( x2 + y2 ) Consider G = !( x ! y ) ! y ( x 2 + y 2 ) Fixed point F =G=0 is x = 0, y = 0 y = y + !y x = x + !x Stability analysis F = !x + ! y a! x + b! y = G = !! x + ! y c! x + d ! y p= a+d = 2 - unstable spiral In addition – to look elsewhere in phase plane, rewrite in polars x = r cos ! y = r sin ! x2 + y 2 = r 2 use following identities x 1 dx dy dr +y =r dt dt dt x dy dx d! ! y = r2 dt dt dy q = ad ! bc = 2 p 2 < 4q p>0 PX391 Nonlinearity, Chaos, Complexity SUMMARY Lecture Notes 7- S C Chapman dx = x + y ! x( x2 + y2 ) dt dy = !( x ! y ) ! y ( x 2 + y 2 ) dt then gives r dr = x 2 + xy ! x 2 ( x 2 + y 2 ) ! xy + y 2 ! y 2 ( x 2 + y 2 ) dt = r2 ! r4 r2 ie: " $ $ $ $ $ $ # d! = !x 2 + yx ! xy(x 2 + y 2 ) ! xy ! y 2 + xy(x 2 + y 2 ) dt = !r 2 dr = r (1 ! r 2 ) dt d! = !1 dt Integrate directly – ! = !0 ! t " 2 Ae2t $r = $# 1 + Ae2t % ' '& don't need to integrate r equation to see the limit cycle. dr =0 dt r =1 for any ! (as well as r = 0 the fixed point) trajectory sits on circle r = 1 . For r > 1 r (1 ! r 2 ) < 0 r < 1 r (1 ! r 2 ) < 0 by inspection. Therefore, solution is attracted to r = 1 circle. y either attracted in from r ! " or out from repulsive fixed point at r = 0 ( x = 0, y = 0 ) r =1 x No single cast iron method to find limit cycles – see course texts for some advanced methods. 2 PX391 Nonlinearity, Chaos, Complexity SUMMARY Lecture Notes 7- S C Chapman Example of limit cycle – Van der Pol oscillator Van der Pol, 1926 – Electric circuit with valve (model of heatbeat) Identical to Rayleigh, 1883 – Nonlinear Vibrations 1st experimentally shown limit cycle d 2x dx + ! ( x 2 ! 1) + x = 0 2 dt dt cause of trouble Write as dx =y dt dy = !x ! ! ( x 2 !1) y dt If ! = 0 ! linear pendulum ! = 1 . Symmetries – invariant for ! ! "t; ! ! "! Therefore, solve for ! > 0 - reverse time for ! < 0 ie: ! > 0 growth is ! < 0 damping, etc. Fixed points x = 0, Stability y=0 x = x + !x y = y + !y d!x = !y dt F= dx =y dt G= dy = ! x ! ! ( x 2 ! 1) y dt or work out !F !F =0 =1 !x !y 3 !G !G = "1" 2!x = "! ( x 2 "1) !x !y d! y = !!x + "! y dt PX391 Nonlinearity, Chaos, Complexity SUMMARY Lecture Notes 7- S C Chapman !F =0 !x Evaluate at x , y = 0 !F !G !G =1 = "1 =! !y !x !y then p= !F !G + =! !x !y q= !F !G !F !G " # =1 !x !y !y !x !>0 p>0 q>0 unstable and spiral if p 2 < 4q . Guess there is more …. Since damping term is ! ( x 2 ! 1) this is +ve for large x changes sign as x ! 1 is zero at x = 1 ! (damping) (growth) (neither!) Solve – multiple timescale analysis (Rowlands, appendix) - method of averaging (Drazin, p 193) - handout for result Pendulum by formula We have d! = 0 "! + 1." y # F dt dy n = $% 2 ( $1 ) "! + 0" y # G dt # 0 J =% n %$ !" 2 ( !1 ) dy n = 0 ! y " # 2 ( "1 ) !$ dt d$ = 1.! y + 0 !$ dt # 0 !" 2 !1 n ( ) J =% %$ 1 0 1 & ( 0 (' or ! J = ## a b #" c d $& && &% - same thing since F = a!x + b! y G = c!x + d! y p= a+d =0 q = ad ! bc = " 2 ( !1 ) So, for n even q > 0 centre, 4 n n odd q < 0 saddle (see handout) & ( ('