More complicated analytic quadratic map is
xn+1 = fc( xn ) = xn2 + c.
On complex plane it generates the famous Mandelbrot and Julia fractal sets. In spite of apparent simplicity it has very rich dynamics. For this map regions of regular and chaotic dynamics are entangled in an intricate manner and scenarios of transition to chaos are common for many other dynamical systems.

Iteration diagram

Dynamics of 1D real maps is useful to trace on iteration diagram shown below. The blue curve is f oN(x) = f(f(...f(x))) the N-th iteration of f(x). Diagonal y = x is the green line. -2 ≤ x,y ≤ 2. As since f(0) = C then for N = 1 the C value coincides with y(0). Dependence xn on n is plotted in the right window.
Controls: Drag the blue curve to change C and starting point xo values. Press <Enter> to set new parameters from the text fields.
To plot the first iteration we draw vertical red line from the starting point xo toward the blue curve y = f(x) = x2 + c, where yo = f(xo ). To get the second iteration we draw red horizontal line to the green diagonal y = x, where x1 = yo . Then draw again vertical line to the blue curve to get y1 = f(x1 ) and so on.
Points fc: xo → x1 → x2 → ... for some c and xo values make orbit of the point xo (it is plotted to the right).

Critical points

For an analytic map points where f '(xc ) = 0 are called critical points. Every stable cycle attracts at least one critical point. Quadratic map has the only critical point xc = 0. Therefore it can have only one attracting cycle and xc is used as the starting point to find the cycle.

Fixed points

For C = -1/2 iterations go quickly to attracting fixed point   x = f(x) of the map. Fixed points correspond to intersections of y = x and y = f(x) (green and blue) curves. There are always two fixed points (may be complex) for a quadratic map because of two roots of quadratic equation
f(x) - x = x2 + c - x = 0,     x1,2 = 1/2 ∓ (1/4 - c)½.
The first derivative of a map at a fixed point
m = f '(x) = 2x
is called multiplier (or the eigenvalue) of the point. For small enough δx
f(x + δx) = f(x) + mδx + O(δx2) ≈ x + mδx.

So a fixed point is stable (attracting), superstable, repelling, indifferent (neutral) according as its multiplier satisfies |m| < 1, |m| = 0, |m| > 1 or |m| = 1.

The second fixed point (the right intersection) is always repelling. For |x| > x2 iterations go to infinity. For |x| < x2 they go to the attracting fixed point x1. This interval is the basin of attraction of the point.

Attracting cycles

For C = -1 the map has attractin period-2 cycle (the left picture above). The second iteration of the map f o2 has two attracting fixed points z3 and z4 .

Lyapunov exponent

For a continuous map xn+1 = f(xn ) a small deviation δxo of coordinate xo leads to a small change in x1
δx1 = f '(xo) δxo.
For n iterations
δxn = δxo i=0,n-1 f '(xi ).
Then the Lyapunov exponent is determined as
Λ = limn → ∞ Ln ,
Ln = 1/n log|δxn /δxo| = 1/n ∑i=0,n-1 ln |f '(xi )|
.
For a chaotic orbit |δxn| grows with increasing of n so Λ > 0.
You see below chaotic quadratic map for c = -2 with positive Lyapunov exponent L calculated for shown finite orbit segment.

For attracting cycle below L is negative

Contents   Previous: Strange Cantor repeller   Next: Bifurcation diagram
updated 3 July 2007