A continuous map on a metric space has Devaney chaos when it has topological transitivity, a dense set of periodic points, and sensitive dependence on initial conditions. Transitivity allows trajectories to pass between prescribed open regions; dense periodic points place exactly repeating trajectories arbitrarily close to every state; sensitivity requires a fixed positive separation scale for some nearby initial conditions under iteration.
The two-symbol full shift illustrates all three conditions. A finite coordinate block specifies a basic open set. Concatenating two prescribed blocks with suitable spacing gives a sequence whose shift moves between the two open sets, while repeating a block gives a periodic sequence in any such neighborhood. Changing a sufficiently distant coordinate produces an initially close sequence that becomes visibly different when that coordinate shifts to the center.
These are global properties of the chosen invariant state space. A positive local stretching rate or an unstable fixed point alone does not prove all three conditions.
A map is topologically transitive when, for every pair of nonempty open sets , some iterate satisfies .
A continuous map is topologically mixing if for every pair of nonempty open sets , there is an integer such that for every . By comparison, topological transitivity asks only for some hitting time for each pair.
For the full tent map, every open interval contains a sufficiently small branch interval of an iterate. That iterate maps the branch across the whole unit interval. Every later image still contains the whole interval, so it meets every nonempty open . This proves mixing and illustrates why a uniform tail of hitting times is stronger than a single possible visit.
Mixing here describes how open sets overlap under a deterministic map. It is a topological property; a statement about decay of correlations for a specified invariant probability measure is a separate measure-theoretic assertion.
Periodic points are dense when every nonempty open set contains a point fixed by some positive iterate.
The butterfly effect is the intuitive idea that an arbitrarily small change in an initial condition can eventually change the observed future substantially. In dynamical systems, sensitive dependence makes this precise for a map on a metric space : there is one separation scale such that
The same works at every point, whereas and the separation time may depend on the point and the requested closeness. For a flow map, replace the iteration of a map by with some . The state space matters: restricting to a single fixed point removes the alternative nearby states needed for sensitivity.
Sensitivity is compatible with deterministic evolution. Given an exact initial condition, there is still one orbit; uncertainty concerns which exact state a measurement represents. It also allows accurate short-term prediction. For a continuous function , each fixed finite iterated function is continuous, so sufficiently close initial states remain close at any specified finite step. What fails is a single initial accuracy that controls every future step.
The logistic map illustrates both sides. At , nearby orbits in the basin of attraction of its attracting fixed point approach one another. At , arbitrarily close states can eventually separate by a macroscopic fraction of . The definition asserts the existence of separating neighbours, not that every nearby pair must separate: and have exactly the same next state under the logistic map and hence the same subsequent orbit. Nor must separation increase at every step.
Figure 1.
The same deterministic logistic rule can shrink or amplify initial uncertainty
. Each pair starts at and and uses the same logistic map parameter. At , the orbits approach the same attracting fixed point. At , this pair eventually gives visibly different states. The logarithmic panels expose changes hidden in the upper panels; dashed slopes illustrate the eventual contraction factor and a representative exponential growth factor . They are reference slopes, not exact formulas for these finite separations. Distances fluctuate and remain bounded by one. These computations illustrate the phenomenon; two numerical orbits alone do not establish the quantified definition. Details.
A positive Lyapunov exponent describes average exponential growth of an infinitesimal displacement along an orbit, when the relevant limit exists. It is related to sensitivity but is not a substitute for its uniform definition. For a differentiable interval map, the chain rule gives
At , the logistic map has for Lebesgue measure-almost every initial condition, while exceptional orbits can have different exponents. Finite-separation growth is only approximately exponential while the linear approximation remains useful. In a bounded state space, a separation cannot grow exponentially forever.
Sensitivity alone is also weaker than Devaney chaos, which additionally requires topological transitivity and dense periodic points. For example, on the real line has sensitive dependence and a positive Lyapunov exponent, but its only periodic point is zero, so it is not Devaney-chaotic. This distinguishes growth of uncertainty from the richer organization of recurrent orbits.
Edward N. Lorenz's "Deterministic Nonperiodic Flow" (1963) studied a thermal convection model with three ordinary differential equations. Its numerical orbits exhibited persistent nonperiodic motion and instability to small changes of the initial condition. Lorenz discussed the consequences for very-long-range weather forecasting: a deterministic law need not make a detailed distant future reliably recoverable from approximate observations. The historically relevant lesson of the Lorenz system is this growth of uncertainty under deterministic evolution.[ref]
A forecast horizon is the duration over which a prediction meets a specified error tolerance. If an initial uncertainty grows approximately as with positive Lyapunov exponent , then solving for when it reaches a tolerance gives the displayed estimate. Here is the natural logarithm and is measured per step of the discrete dynamical system; for a flow map, use the exponent per unit time instead. The estimate assumes the growth approximation remains valid up to the chosen tolerance and excludes additional model errors.
Reducing by a factor ten buys approximately more steps, rather than multiplying the forecast duration by ten. For typical logistic map orbits at , , so one extra decimal digit buys roughly steps in this approximation. Better measurements help, but their benefit is logarithmic. A forecast of a probability distribution or a stable long-term average asks a different question from predicting each individual state.
The doubling map is the discrete dynamical system
on the unit circle represented by . In a binary expansion, it shifts the digits one place to the left. This makes a small uncertainty in later digits become a large uncertainty after sufficiently many iterates.
Its Devaney chaos can be checked directly. The periodic points of the doubling map are the grids ; their mesh tends to zero, giving dense periodic points. Every nonempty open arc contains an interval for large enough . Its image under covers the unit circle, proving topological mixing and hence topological transitivity. Finally, for any initial point and arbitrarily large , choose modulo one. Their initial circle metric distance tends to zero, while the distance between and is . This proves sensitive dependence on initial conditions with a fixed separation threshold smaller than .
The doubling map preserves normalized Lebesgue measure on the unit circle. Each point has exactly two preimages, and modulo one, and each inverse branch scales lengths by .
For the doubling map on , a binary expansion gives
Multiplication by two removes the first digit after reduction modulo one. Thus binary cylinder sets describe exactly which branches the first few iterates visit.
Concatenating two prescribed finite blocks with enough intervening digits constructs a point whose itinerary visits the associated cylinders in order. Repeating a finite block gives a periodic point arbitrarily close to any prescribed finite prefix. These observations prove topological mixing and density of periodic points.
At dyadic rationals, terminating and eventually-all-one expansions describe the same number; one consistent convention is to choose the expansion ending in zeros. Therefore the numerical map is not globally bijectively conjugate to the unrestricted one-sided binary sequence space without accounting for endpoint identifications. Binary coding remains a useful representation with those identifications made explicit.
The fixed points of are
Thus has fixed points.
Let with an integer . For the doubling map, a point fixed by lies on a periodic orbit whose least return time divides . Every proper divisor of divides , so all smaller-period points fixed by are exactly the points fixed by . Using the periodic points of the doubling map, subtraction gives the number of points of exact period :
For example, gives two points of exact period two, and , forming one periodic orbit. To count distinct periodic orbits of length , divide the displayed point count by . The separate case has one fixed point, ; the subtraction formula requires .

Articles by others on the same topic (0)

There are currently no matching articles.