2015 one-dimensional maps

Upload: hieu-tran-trung

Post on 07-Aug-2018

215 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/21/2019 2015 One-dimensional Maps

    1/26

    One-dimensional maps

    March 23, 2015 1M. Biey - Dip. di Elettronica e Telecomunicazioni - Politecnico di Torino

    “Simple dynamical systems do not necessarily

    lead to simple dynamical behavior” . 

    (R. M. May, 1976)

     __________________________R. M. May, Simple mathematical models with very complicated dynamics, Nature, 261, 459, 1976.

  • 8/21/2019 2015 One-dimensional Maps

    2/26

    Maps

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 2

    The solutions of dynamical models with discrete time,

    are characterized by sequences of points, each of whichis determined from an initial point (the initial condition).

    Such models are called recurrence relationships,

    recurrences, iterations, or simply maps. Their explicit

    form is written:

    X  is a p-dimensional real vector, n denotes the discrete

    time, Λ  is a parameter vector, X 0 is the vector of initial

    conditions.

  • 8/21/2019 2015 One-dimensional Maps

    3/26

    Maps (cont.)

    Maps arise in various ways: As model of natural phenomena

    E.g.: Capital growth: , where I is the annual rate of interest 

     As tools for analyzing differential equationsE.g.: Forward Euler Integration Formula:

     As simple examples of chaos

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 3

    1( ) ( )

    k k k x f x x x hf x  

    1  (1 )

    n n C C    I 

  • 8/21/2019 2015 One-dimensional Maps

    4/26

    Maps (cont.)

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 4

    Simple examples of one-dimensional maps:

    1 0( ), ( 0)

    n n x f x x n x  

    If the vector X  has dimension 1, we have one-

    dimensional  

    maps:

     f is a smooth function from the real line to itself

    and x0 is the initial condition.

    The sequence x0, x1, x2, …. is called the

    orbit   (or trajectory ) starting from x0 .

  • 8/21/2019 2015 One-dimensional Maps

    5/26

    Maps (cont.)

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 5

     A geometric way of thinking:Cobweb construction to

    iterate the map graphically:

  • 8/21/2019 2015 One-dimensional Maps

    6/26

    Fixed points

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 6

    DefinitionSuppose x* satisfies f ( x*) = x*.

    Then, if xn = x*, it follows that

     xn+1 = f ( xn) = f ( x*) = x*

    Hence the orbit remains at x* for all futureiterations.

     x* is called a fixed  (or equilibrium) point .

    Fixed points are the intersections of the mapwith the straight line xn+1 = xn

    What about the stability of a fixed point?

  • 8/21/2019 2015 One-dimensional Maps

    7/26

    Fixed points (cont.)

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 7

    Linear stability of a fixed point x*

     xn = x* + ηn : does ηn  grow or decay as n increases?

     xn+1 = x* + ηn+1 =  f ( x* + ηn) = f ( x*)  + f ′ ( x*) ηn +  (ηn2

     )

    ηn+1  f ′ ( x*) ηn  λ ηn , with  λ = f ′ ( x*)  Then

    η1 = λ η0 , η2 = λ

     η1 = λ2 η0 , ..... , ηn = λ

    n η0 

  • 8/21/2019 2015 One-dimensional Maps

    8/26

    Fixed points (cont.)

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 8

    Linear stabilityIf the multiplier λ satisfies

    ( *) 1 f x 

    then the fixed point x* is (linearly) stable.

    If   the fixed point is unstable.

    Nothing can be said about the marginal case 

    ( *) 1 f x 

    ( *) 1 f x  . In this case, graphical way can help!

  • 8/21/2019 2015 One-dimensional Maps

    9/26

    Fixed points (cont.)

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 9

    Problem:Consider the map . Determine thestability of the fixed point x* = 0.

    1  sin

    n n x x 

     At x* = 0 we have .

    Nothing can be said about stability. We use a

    geometrical approach:

    (0) cos(0) 1 f 

     x* = 0 is a stable equilibrium point!

  • 8/21/2019 2015 One-dimensional Maps

    10/26

    Logistic Map

    Logistic (or quadratic) map[*]

    :

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 10

    1  (1 )

    n n n x rx x  (1)

     ________________________________

    [*] Discrete-time analog of the logistic equation for population growth 

  • 8/21/2019 2015 One-dimensional Maps

    11/26

    Logistic Map (cont.)

    Fixed points:

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 11

    * 0, (0 4)

    1* 1 , ( 1)

    x r 

    x r r 

    REMARKS:For 0 < r  < 1 there is just an equilibrium point at the origin ( x* can notbe negative)

     A second equilibrium point appears for r  > 1

    It is stable for 1< r  < 3

    For r  > 3 a period-2 cycle is born

  • 8/21/2019 2015 One-dimensional Maps

    12/26

    Bifurcations

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 12

     A bifurcation represents the sudden 

    appearance of a qualitatively different solutionfor a nonlinear system as some parameter is

    varied.

    r

      x 

    Logistic map

    Bifurcations occur in four basic varieties:flip, fold, pitchfork, and transcritical (Rasband 1990)

  • 8/21/2019 2015 One-dimensional Maps

    13/26

    Bifurcations (cont.)

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 13

    In the logistic map, at r = 1 the equilibrium x = 0

    becomes unstable and a new equilibrium point appears:that is called a transcritical bifurcation.

     At r  = 3, x* becomes unstable and a 2-cycle appears.

    That is called a flip (or period-doubling) bifurcation

    FLIP

    TRANSCRITICAL

  • 8/21/2019 2015 One-dimensional Maps

    14/26

    Bifurcations (cont.)

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 14

    In one-dimensional maps a 2-cycle exists if and

    only if there are two points p and q such that

     f ( p) = q and f (q) = p

    Equivalently,  p must satisfy

     f ( f ( p)) = p

    Hence,  p is a fixed point of the second-iterate

    map f 2( x) =  f ( f ( p)) . The same for q.

    In conclusion:

    p  and q  are fixed points of the second-iterate

    map.

  • 8/21/2019 2015 One-dimensional Maps

    15/26

    Bifurcations (cont.)

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 15

  • 8/21/2019 2015 One-dimensional Maps

    16/26

    Bifurcations (cont.)

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 16

     p and q are evaluated by solving the equation f 2( x) –  x = 0:

    2 (1 ) 1 (1 ) 0r x x rx x x  

    The fixed points of the 1-st iterate are solution of theabove equation (in fact, f ( x*) = x*, so also f 2( x*) = x*).

     After factoring out  x and [ x  – (1 – 1/r )], we get a

    quadratic equation:

    2   1 1 11 1 0x x r r r 

  • 8/21/2019 2015 One-dimensional Maps

    17/26

    Bifurcations (cont.)

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 17

     p and q are given by

    ( 1) ( 3)( 1),

    2

    r r r p q 

     A 2-cycle exists for all r  > 3. For r  < 3 the roots arecomplex, which means that a 2-cycle does not exist.

    What about its stability?

  • 8/21/2019 2015 One-dimensional Maps

    18/26

    Bifurcations (cont.)

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 18

    For evaluating the stability of the 2-cycle, it is sufficient toevaluate the stability of a fixed point, since p and q are fixed

    points of the 2nd-iterate map.

    Let us compute the multiplier, say, at p (*):

    ( ( ( ))) ( ( )) ( ) ( ) ( )x p

    d  f f x f f p f p f q f p

    dx 

     _____________________________________

    (*) Remember that:

  • 8/21/2019 2015 One-dimensional Maps

    19/26

    Bifurcations (cont.)

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 19

    Evaluate the multiplier at p:

  • 8/21/2019 2015 One-dimensional Maps

    20/26

    Bifurcations (cont.)

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 20

    The first values where bifurcations take place:

    Source: S. H. Strogatz, Nonlinear dynamics and chaos, Perseus Books, 1994.

    What happens if r  > r ∞ ? 

    1 6

  • 8/21/2019 2015 One-dimensional Maps

    21/26

    Chaos

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 21

    Qualitative definition(1):

    Chaos is a bounded  aperiodic long-term

    behavior   in a deterministic  system that exhibits

    sensitive dependence on the initial conditions,

    thereby rendering long-term predictionimpossible.

    1. Bounded: trajectories remain confined in a finite set

    2.  Aperiodic long-term behavior: there are trajectories which do not settle

    down to fixed points, periodic orbits or quasi periodic orbits as t → ∞. 

    3. Deterministic: the system has no random or noisy inputs or parameters.4. Sensitive dependence on the initial conditions: nearby trajectories separate

    exponentially fast.

     _____________________________

    (1) S.H. Strogatz, Nonlinear dynamics and chaos, p. 323, Perseus Books Pu.,

    1994.

  • 8/21/2019 2015 One-dimensional Maps

    22/26

    Lyapunov Exponent in One Dimension

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 22

    In a chaotic system neighboring orbits should separate

    exponentially fast, on average.

    Let us compute the separation between two orbitsstarting from two nearby initial conditions: x0 and x0 + ε 

    To what extent two solutions starting from nearly equal

    initial conditions are repelling each others?

    1st orbit:2

    0 0 0 0, ( ), ( ), , ( )n x f x f x f x  

    2nd orbit:2

    0 0 0 0, ( ), ( ), , ( )n x f x f x f x  

  • 8/21/2019 2015 One-dimensional Maps

    23/26

    Lyapunov Exponent in One Dimension

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 23

    The absolute value of separation after n iterates of the two

    maps is

    01 2 0

    '' ' '( ) ( ) ( )n 

    n n n x x f f x f x f x  

    0 0( ) ( )n n 

    n x f x f x  

     To a 1st order approximation we have

    Then, taking the log

    1

    0

    ln ln '( ) lnn 

    n i i 

    x f x 

  • 8/21/2019 2015 One-dimensional Maps

    24/26

    Lyapunov Exponent in One Dimension

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 24

    The average value of the sum is

    0( )

    0 0( ) ( )

      n x n n 

    n x f x f x e  

    0( )x   is called the Lyapunov exponent (LE) of the orbit.

    It represents the mean exponential separation of two orbits that

    have near initial conditions. A positive LE is a signature of chaos 

    1

    00

    '1( ) ln ( )n 

    i i 

    x f x n 

     So we have

    From which we get

    0ln ( ) ln

    n x n x 

  • 8/21/2019 2015 One-dimensional Maps

    25/26

    Lyapunov Exponent in One Dimension

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 25

    Recall the last equation:

    dividing by |ε|, taking the limit for ε → 0, andremembering that we get

    0

    0

    ( )

    0

    ( )( )

    n n x    n n 

    x    df x e f x 

    dx 

    1

    00

    0( )  1 1 1lim ln lim ln ( ) lim ln ( )

    n n n 

    i n n n i 

    x x 

     f x f x n n n 

     

    Taking the limit as n → ∞, we get the definition of LE:

    0( )

    0 0( ) ( )  n x n n 

    n x f x f x e  

    0( )n x 

    n x e 

  • 8/21/2019 2015 One-dimensional Maps

    26/26

    Lyapunov Exponent in One Dimension

    March 23, 2015 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 26

    Two fundamental facts [*]

     :The LE does not depend on the trajectory. More

    precisely, it is the same for all the x0 in the basin of

    attraction of a given attractor

    The limit for n going to infinity exists

    In general, one needs to use a computer to

    calculate Lyapunov exponents

    [*] V.I. Oseledets, A multiplicative ergodic theorem. Lyapunov characteristic numbers

    for dynamical systems, Trans.Moscow Math. Soc. 19(1968), 197-231.

    Moscov.Mat.Obsch.19(1968),179-210.