Topological entropy

Topological entropy is a non-negative real number that measures the complexity of a dynamical system by quantifying the exponential growth rate of the number of distinguishable orbit segments as time increases. Introduced by Adler, Konheim, and McAndrew in 1965 and later reformulated by Dinaburg and Bowen using spanning and separating sets, topological entropy captures the intrinsic information-production rate of a system — the rate at which new distinctions emerge as the dynamics unfolds.

For a map f on a compact metric space, the topological entropy h_top(f) is defined as the limit superior of (1/n) log N(n, ε), where N(n, ε) is the maximum number of orbit segments of length n that can be distinguished at precision ε. Two segments are distinguishable if their distance at some time in [0, n-1] exceeds ε. As ε → 0, the number of distinguishable segments grows exponentially for chaotic systems, and the exponential growth rate is the entropy.

On shift spaces, topological entropy has an exact combinatorial formula: for a subshift of finite type with transition matrix M, h_top = log λ, where λ is the largest eigenvalue of M. This formula connects the continuous notion of entropy to the discrete machinery of linear algebra. For smooth systems, Margulis and Katok developed formulas relating topological entropy to volume growth of unstable manifolds, linking entropy to the geometry of the phase space.

Topological entropy is related to measure-theoretic entropy through the variational principle: h_top(f) = sup_μ h_μ(f), where the supremum is taken over all invariant measures. The measures that achieve the supremum are called measures of maximal entropy, and for hyperbolic systems they are unique and coincide with the SRB measure.

Topological entropy is the system's answer to the question: how fast are you making me think? A system with zero entropy is predictable; a system with positive entropy is chaotic; a system with infinite entropy is pathological. The remarkable fact is that this single number — a limit of logarithms of counts — captures the difference between order and chaos with complete precision.