Dynamical Systems
Level: 0 1 2 3 4 5 6 7
MSC Classification: 37 (Dynamical systems)


Getting Oriented

Included page "nav:dynamics" does not exist (create it now)

Dynamical systems…

The Basics

Dynamical Systems

Discrete-Time Dynamical System
a pair $(X,f)$ with $f:X\to X$, such that iterates $f^n$ satisfy $f^{t+s}=f^t \circ f^s$.
Continuous Time Dynamical System
a one-parameter family of maps $\{f^t:X\to X\}$ with $f^{t+s}=f^t \circ f^s$ and $f^0=\id$. It is a flow if $t\in\mathbb{R}$ and a semiflow if $t\in\mathbb{R}_0^+$.
the notion of equivalence for dynamical systems. Given systems $(X,f)$ and $(Y,g)$, a semiconjugacy is a surjective map $\pi: Y \to X$ such that $f\circ\pi=\pi\circ g$. A conjugacy is an invertible semiconjugacy.
a point has positive (forward) and negative (backward) semiorbits, whose union is the point's orbit. A periodic point $x$ of period $T$ satisfies $f^T(x)=x$. If $f(x)=x$, $x$ is a fixed point.

A fixed point $x_0$ can be either attracting (when a neighborhood of $x_0$ converges to $x_0$ under $f$), or repelling (when the inverse $f^{-1}$ is attracting at $x_0$). Points can also be eventually periodic.

A subset $A\subset X$ for which $f(A)\subset A$ is $f$-invariant. If $X$ is the only $f$-invariant subset, $(X,f)$ is called minimal. Closely related is the ergodic dynamical system, whose only measurable invariant subsets have either measure zero or full measure.

Canonical Examples

Systems on $S^1$
Consider the circle $S^1$ as the unit interval $I=[0,1]$ with endpoints identified. The rotation map $R_\alpha$ is defined by $R_\alpha(x)=x+\alpha \mod 1$. If $\alpha$ is rational, every orbit is periodic, while if $\alpha$ is irrational, every (positive) semiorbit is dense.

The map defined by $E_m(x)=mx \mod 1, m \in \mathbb{Z}$ is the expanding endomorphism of $S^1$, and is an $|m|$-to-$1$ function.

Sequences and Shifts
Let $\Sigma^+_m=\{0,\ldots,m-1\}^\mathbb{N}$, \ie the set of (one-sided) sequences of $m$ letters. The shift is the $m$-to-$1$ map $\sigma: (x_1,x_2,\ldots) \mapsto (x_2,x_3,\ldots)$. Replacing $\mathbb{N}$ with $\mathbb{Z}$, we obtain the set of two-sided sequences $\Sigma_m$, for which the shift $\sigma$ is invertible.

Given a directed graph with $m$ nodes, we have a subshift (of $\Sigma_m$ or $\Sigma_m^+$) whose elements are sequences corresponding to paths in the graph.

Many systems are conjugate or semiconjugate to shifts. For example, $\phi: \Sigma_m^+ \to S^1$ is a semiconjugacy between $(\Sigma_m^+,\sigma)$ and $(S^1,E_m)$ given by the base $m$ expansion of a sequence in $\Sigma$.

Quadratic maps
the canonical quadratic map is $q_\mu(x)=\mu x (1-x), \mu>0$. The behavior of this system is very different for $\mu\leq 4$ and $\mu>4$. As an example, for $\mu>4$, there are periodic points of any period $T\in\mathbb{N}$.
The Gauss Transformation
the map on $[0,1]$ given by
\begin{align} \phi(x)=\begin{cases}\tfrac{1}{x}-[\tfrac{1}{x}]& x\neq 0,\\ 0& x=0,\end{cases} \end{align}

where $[x]$ is the greatest integer less than or equal to $x$, for $x\in\mathbb{R}$. The map $\phi$ is closely related to continued fractions: if $x=[a_1,a_2,\ldots,a_n]$ is a continued fraction, then $\phi(x)=[a_2,\ldots,a_n]$. Thus, $\phi$ corresponds (via a conjugacy) to shifts on sequences of integers. This property can be used to show that any rational number is uniquely represented by a finite continued fraction.

Hyperbolic Total Automorphisms
consider the torus $T^2$ as the unit square with opposite sides identified. Then, a $2 \times 2$ integer matrix $A$ induces an automorphism of the torus with $A: \binom{x_1}{x_2} \mapsto A\binom{x_1}{x_2} \mod 1.$ If $A$ is invertible ($|\det A|=1$) and the eigenvalues of $A$ do not lie on the unit circle, then $A: T^n \to T^n$ is called a hyperbolic toral automorphism because it expands in contracts in complementary directions.
The Horseshoe
a system on the unit square with map $f$ defined by ‘bending’ the square into a horseshoe and placing it atop the square. Thus, $f$ maps the square to two horizontal strips, and $f^{-1}$ maps the square to two vertical strips.

The horseshoe set $H$ is the $f$-invariant subset of the square, and is the product of two Cantor sets. The system $(H,f)$ is conjugate with the shift system $\Sigma_2$. It is hyperbolic because it expands and contracts in different directions, but its behavior survives under small perturbations.

The Solenoid
the subset $S$ of the solid torus $T=S^1\times D^2$ forward invariant under
\begin{align} F(\phi,x,y)=(2\phi,\lambda x+\tfrac{1}{2}\cos 2\pi\phi, \lambda y + \tfrac{1}{2}\sin 2\pi\phi), \quad \lambda\in(0,\tfrac{1}{2}) \text{ fixed}. \end{align}

$S$ is a Cantor set, and $(S,F)$ is a hyperbolic system. It is conjugate to the symbolic system $\Phi$ consisting of sequences $(\phi_1,\phi_2,\ldots)$ in $(S^1)^\mathbb{N}$ for which $\phi_i=2\phi_{i+1} \mod 1$, with map $\alpha:(\phi_1,\phi_2,\ldots)\mapsto(2\phi_1,\phi_1,\phi_2,\ldots)$.

Basic Concepts

Flows, Suspension, and Cross-Section
Given an autonomous differential equation $\dot x=F(x)$, we can consider the flow $f^t(x)$. Given any flow and a subset $A\subset X$ for which any point $x\in A$ implies $f^t(x)\in A$ a ‘discrete’ time $t$ later, one can define a map $g: A \to A$ by $g(x)=f^{\tau(x)}(x)$, where $\tau(x)$ is the ‘return time’ of $x$, i.e., the first intersection of the orbit of $x$ and $A$. This gives the cross-section dynamical system $(A,g)$.

Conversely, given a map $f: X \to X$ we have a quotient space $X^*=X\times\mathbb{R}^+/~$, where $(f(x),0)~(x,1)$. Then the suspension of $f$ is the semiflow $f^t: X^* \to X^*$ given by flowing from $(x,0)$ to $(x,1)$, then from $(f(x),0)$ to $(f(x),1)$, and so on.

Chaos and Lyapunov Exponents
Chaos is essentially sensitive dependence on initial conditions. A simple example is the circle endomorphism $(S^1,E_m)$, because the orbits of nearby points diverge. The sensitivity of initial conditions strongly correlates with a positive Lyapunov Exponent, which is defined by
\begin{align} \mathbb{C}hi(x,v)=\limsup_{n\to\infty}\frac{1}{n}\log\|df^n(x)v\|. \end{align}

This measures the exponential growth rate of tangent vectors along orbits.

Given a map $f:X \to X$, a compact subset $C\subset X$ is an attractor if there is an open set $U \supset C$ with $f(\overline{U}) \subset U$ and $C=\cap_{n\geq 0} f^n(U)$. Thus, nearby points converge to $C$. The set of points which converge to $C$ is called the basin of attraction $BA(C)$. An open set $U$ with compact closure and $f(\overline{U})\subset U$ is called a trapping region because $\cap_{n\geq 0} f^n(U)$ is an attractor. The existence of an attractor is often proven by finding a trapping region. Attractors can come in many forms: fixed points, periodic orbits, hyperbolic attractors, and chaotic attractors. Examples of chaotic attractors include the Lorenz attractor and the H\'enon Attractor.

Topological Dynamics

Going Further

The Road Ahead

Unless otherwise stated, the content of this page is licensed under Creative Commons Attribution-ShareAlike 3.0 License