Other meanings of Lyapunov exponent
Dynamical systems
The Lyapunov exponent (or Lyapunov characteristic exponent) is a quantity that characterizes the average exponential rate of separation of infinitesimally close trajectories in a dynamical system. A positive largest Lyapunov exponent indicates chaos, as nearby orbits diverge exponentially; a negative exponent signals convergence to a fixed point or limit cycle; a zero exponent often corresponds to periodic or quasiperiodic motion along a neutral direction.1
The Lyapunov exponent is defined for a continuous dynamical system ẋ = f(x) as the limit λ = lim_{t→∞} (1/t) ln |δx(t)| / |δx(0)|, where δx(t) is the evolution of an initial perturbation δx(0). For discrete maps, the definition uses the product of Jacobian matrices along an orbit. The multiplicative ergodic theorem of Oseledets (1968) guarantees the existence of the Lyapunov exponents for almost every initial condition with respect to an invariant measure, and the full spectrum (ordered λ₁ ≥ λ₂ ≥ … ≥ λₙ) describes the expansion/contraction rates in each orthogonal direction.1 The sum of all exponents equals the average divergence of the flow (the trace of the Jacobian).
Numerical estimation of the largest Lyapunov exponent is often performed using the algorithm of Benettin, Galgani, Giorgilli, and Strelcyn (1980), which evolves a set of orthonormal tangent vectors and periodically reorthogonalizes them via Gram–Schmidt.2 The Wolf algorithm (1985) is another common method for experimental time series. Lyapunov exponents are central to characterizing chaos in fluid dynamics, population biology, celestial mechanics, and quantum systems. For example, the Lorenz system (1963) has a positive largest exponent ~0.9, confirming its chaotic attractor. In cardiology, the largest Lyapunov exponent of heart rate variability has been used as a marker of pathological states.
Early conceptual work on what later became Lyapunov exponents was done by Nikolay S. Krylov (1930s) in the context of the Boltzmann–Gibbs approach, but the rigorous definition is due to Aleksandr Lyapunov (1892) in his study of stability of motion. An overlooked nuance: Lyapunov exponents are defined in the limit of infinite time; for finite times, one obtains “finite-time Lyapunov exponents” (FTLEs) which are used in Lagrangian coherent structures to identify mixing barriers in fluid flows. Another edge case: in systems with coexisting attractors, the Lyapunov exponents depend on the basin of attraction. The Lyapunov exponent can be negative even in a chaotic system if the spectrum as a whole is considered (e.g., the sum of all exponents is negative for dissipative systems). The concept also applies to stochastic systems via the “random Lyapunov exponent.”
The Lyapunov exponent is intimately related to the Kolmogorov–Sinai entropy, which for certain systems equals the sum of positive Lyapunov exponents (Pesin’s formula).3 The Oseledets theorem extends the concept to infinite-dimensional systems (e.g., delay-differential equations). In the context of Hamiltonian systems, Lyapunov exponents are symmetric and sum to zero, reflecting phase space volume preservation. Lyapunov functions (not to be confused with exponents) are used to prove stability. The largest Lyapunov exponent is a key input in the calculation of the Kaplan–Yorke dimension, an estimate of the fractal dimension of a strange attractor.4 The field continues to evolve, with applications in machine learning (e.g., detecting chaos in reservoir computing).
This article treats the Lyapunov exponent as a measure of average exponential separation in dynamical systems; for the related concept in control theory, see Lyapunov function.
Help improve the encyclopedia. Reports go straight to the site manager.