• Home
  • Textbooks
  • Chaotic Dynamics: An Introduction
  • The characterization of chaotic attractors

Chaotic Dynamics: An Introduction

Gregory L. Baker, Jerry P. Gollub

Chapter 5

The characterization of chaotic attractors - all with Video Answers

Educators


Chapter Questions

02:02

Problem 1

The length of the Cantor set may be determined by subtracting the length of each segment taken out of the set during each step in its formation. For example, a length $\frac{1}{3}$ is taken out in the first step. $2\left(\frac{1}{5}\right)$ in the second step, $4\left(\frac{1}{27}\right)$ in the third step, and so on. Form the infinite geometric series which this process describes and show that the sum approaches 1. Therefore the Cantor set has no length.

Nick Johnson
Nick Johnson
Numerade Educator
View

Problem 2

Construct a fractal that is similar to the Cantor set, but instead remove the middle $\frac{1}{2}$ from each previous section. Show that its dimension is $\frac{1}{2}$.

Victor Salazar
Victor Salazar
Numerade Educator
View

Problem 3

Construct a 'two-dimensional' Cantor set in the following way. Draw two Cantor sets at right angles to each other so they just touch at one corner. Then fill in those two-dimensional regions where each 'one-dimensional' set intersects the other. In other words form the set which is the Cartesian product of the two original sets, as in Figure (a) below. Show that the capacity dimension is 2 ( $\log 2 / \log 3$ ). (This set resembles the invariant set of the horseshoe transformation; that is, the points which remain from the original set after many iterations.)

Victor Salazar
Victor Salazar
Numerade Educator
01:13

Problem 4

Following a procedure analogous to that used in Problem 5.1, calculate the area of the two-dimensional Cantor set defined in
FIGURE CAN'T COPY.

Karly Williams
Karly Williams
Numerade Educator
09:20

Problem 5

Construct a Cantor-like set by taking squares of relative area $\frac{1}{9}$ out of the center of larger squares. The process is illustrated in Figure (b). Show that the dimension of the structure is 3 ( $\log 2 / \log 3)$.

Kevin Shryock
Kevin Shryock
Numerade Educator

Problem 6

The Cantor set may be used to study some properties of information dimension. Recall that $l=-\sum_{i=1}^s p_i \log p_i$. Calculate $l$ for each state in the development of the set (as shown in Figure 5.1), assuming that each line segment is equally probable at each iteration. Then use the defining expression for information dimension to show that $d_1=d_c=\log 2 / \log 3$, in this case.

Check back soon!

Problem 7

Repeat the calculation of Problem 5.6 but do not assume equal probabilities. At every iteration of the set let the right segment have twice the probability of the left segment. For example, when there 7. Develop the expression for entropy when there are $2^n$ segments. Show that, in the limit as $n$ tends to infinity, $d_1=1-\left(\frac{3}{3}\right)(\log 2 / \log 31$.

Check back soon!
01:04

Problem 8

By finding the maximum of the function $I=-\sum_{i=1}^N p_i \log p_i$, show that $d_1$ is maximized when $d_1=d_{\mathrm{c}}$ (See Problem 4.5.)

Victor Salazar
Victor Salazar
Numerade Educator
01:55

Problem 9

Show that the generalized dimensions are equivalent to $d_{\mathrm{c}}$ and $d_1$ for $q=0$ and $q=1$, respectively.
FIGURE CAN'T COPY.

Surendra Kumar
Surendra Kumar
Numerade Educator

Problem 10

For a pendulum with $q=5$ and $y=1.5$ the Lyapunov exponents are $\lambda_1=0.06, \lambda_2=0 . \lambda_3=-0.26$. Verify the relation between Lyapunov exponents and the damping factor. Calculate the Lyapunov dimension for the attractor in $(\theta, \omega, \psi)$ space.

Check back soon!
03:08

Problem 11

Estimate the time for predictability of the pendulum of Problem 5.10, assuming its state is initially known to within $1 \%$ of the range of the phase variables. How would you expect this time to change if the damping were increased?

Zulfiqar Ali
Zulfiqar Ali
Numerade Educator

Problem 12

Moon and $\mathrm{Li}$ (1985) suggested a way to estimate Lyapunov exponents. The method is based on the fact that at a saddle point the stable and unstable trajectories meet and deline directions of convergence and divergence of trajectories, respectively. The rates of these behaviors approximate the respective Lyapunov exponents. To study the behavior of the trajectories in the saddle point region the equation of motion is linearized at the saddle point and then solved. Furthermore it is only necessary to solve the homogeneous part of the differential equation because the saddle point is a fixed point in any Poincaré section taken at frequency $\omega_{\mathrm{D}}$. An examin- ation of Poincaré sections taken at various phases shows that $\theta= \pm \pi$ appears to be the $\theta$ saddle point coordinate and therefore the required homogeneous, linearized equation is
$$
\mathrm{d}^2 \theta / \mathrm{dt}^2+(\mathrm{d} \theta / \mathrm{d} t) / q-\theta=0 .
$$

Substitute the solution $\theta=e^{m t}$ into the differential equation to obtain the relation
$$
m=-\frac{1}{2 q} \pm\left(\frac{1}{4 q^2}+1\right)
$$
which has both positive and negative values. These values are the estimates of $\lambda_{+}$and $\lambda_{\text {_ }}$. Check that they sum to $\nabla \cdot F$. Substitute these values into the Kaplan-Yorke relation for $d_1$, thereby obtaining a relation for $d_{\mathrm{L}}$ in terms of $q$. Find $d_{\mathrm{L}}$ in the limiting cases where $q=0$ (infinite damping) and $q=\infty$ (no damping). Calculate $d_1$ in the cases where $q=2$ and $q=5$. Do your results match those given in the text and developed in a previous problem? If not suggest a reason. (Moon and $\mathrm{Li}$ change the value of the damping parameter to an effective value giving numerical results that better match the dimension. If you were to follow a similar procedure. what value of $q$ would you choose for essential agreement with the numerically derived $d_{\mathrm{L}}$ ?)

Check back soon!

Problem 13

Use the program PENDLYAP in Appendix B to calculate Lyapunov exponents in various cases. Examine both chaotic and nonchaotic states. Use the bifurcation diagrams to choose the appropriate values of $g$.

Check back soon!