Orbit of a Dynamical System

Overview

An orbit records which states a deterministic system visits from a specified initial state. The same rule may produce a fixed state, a repeating cycle, or an infinite collection of distinct states. Orbit language separates this elementary iteration from later questions about convergence, recurrence, density and statistical behavior.

An orbit is a set of visited states. The indexed trajectory retains their visiting times as well, so it carries information that the set alone suppresses.

Prerequisites

Start with a set $X$ and a self-map $T:X\to X$. Define $T^0$ to be the identity and $T^{n+1}=T\circ T^n$. The exponent denotes repeated composition: $T^2(x)=T(T(x))$, not the square of $T(x)$. Continuity and a metric are needed only when discussing closures and limits below.

Definition and conventions

The forward orbit of $x\in X$ is

Here $\mathbb N$ includes zero, so the initial state belongs to its orbit. If $T$ is bijective, define negative iterates using its inverse and the full orbit by $\mathcal O(x)=\{T^n(x):n\in\mathbb Z\}$. For a noninvertible map, a backward trajectory may have several choices or no continuation; a single inverse map is unavailable.

For a globally defined flow $\varphi^t$, replace integer time by real time: the full orbit uses $t\in\mathbb R$ and the forward orbit uses $t\ge0$. A solution of an equation defined only on a limited time interval requires its own domain convention.

Intuition

Starting from $T(x)$ discards the first step of the trajectory. Consequently, $T(\mathcal O^+(x))\subseteq\mathcal O^+(x)$, but equality can fail. This is forward invariance, rather than a guarantee that every visited state has a predecessor within the forward orbit.

Full orbits of an invertible map behave differently: moving the starting point along its orbit changes only the time origin. They partition $X$ because the relation $y=T^n(x)$ for some integer $n$ is reflexive, symmetric and transitive.

Examples and non-examples

For $T(x)=x/2$ on $\mathbb R$, the forward orbit of $1$ is $\{1,1/2,1/4,\ldots\}$. Its closure also contains zero, while its omega-limit set consists only of zero: these are the limits approached along times tending to infinity. Zero is never visited from $1$. The full orbit is $\{2^k:k\in\mathbb Z\}$ because this map is invertible on $\mathbb R$.

The phase space matters. Restrict the same formula to $[0,1]$ and it is no longer surjective, so it has no inverse self-map of that interval.

For a finite example, take $X=\{a,b,c,d\}$ with $T(a)=b$, $T(b)=c$, $T(c)=b$, and $T(d)=c$. The forward orbit of $a$ is $\{a,b,c\}$. It is finite, but $a$ is not periodic: it never returns. The cycle $b,c$ has least period two, and $a$ has a transient step before reaching it. This refutes the implication “finite forward orbit means periodic initial point” for arbitrary self-maps.

Important properties

A point is periodic if $T^p(x)=x$ for some positive integer $p$. Its least such $p$ is its least period, and its forward orbit contains exactly $p$ points.

More generally, a forward orbit is finite exactly when the trajectory is eventually periodic. Proof sketch: finiteness forces $T^i(x)=T^j(x)$ for some $i<j$. Applying further iterates gives repetition with period $j-i$ after time $i$. Conversely, a finite transient segment followed by a repeating segment visits finitely many states. If $T$ is invertible, applying $T^{-i}$ cancels the transient segment and proves periodicity of $x$ itself.

For continuous $T$ on a metric space, the closure of a forward orbit is forward invariant. Indeed, if $y$ is a limit of visited states, continuity makes $T(y)$ a limit of their successors. This argument explains why an orbit closure is often a useful state space even when the orbit is not closed.

Connections and further reading

Discrete forward orbits are countable, but continuous-time orbits need not be. A dense orbit is one whose closure fills the phase space; topological transitivity investigates closely related open-set connections. Poincaré recurrence instead concerns returns to measurable regions, using an invariant measure. Neither an orbit's mere infinitude nor its countability settles those questions.

Read Bullett, Chapter 2, pp.1–2 for periodicity conventions and Sarig, §1.1 for the wider orbit viewpoint.