Exponential Stability¶

An equilibrium's trajectory deviations obey a common exponentially decaying bound relative to their initial size over a specified domain.

Version: v1 · 2026-10-03 · History
Domain-specific #: 13213
Domain group: Formal Sciences
Origin domain: Mathematics
Subdomains: Dynamical Systems, Stability Theory → Mathematics
Aliases: Exponential stability of an equilibrium

Core Idea¶

Exponential stability is a quantitative stability property of an equilibrium: deviations of all trajectories from a specified initial-state region are bounded by a common exponentially shrinking multiple of their starting deviations. For an autonomous continuous-time system, this has the form $\|x(t)-x_*\|\leq K e^{-\lambda t}\|x_0-x_*\|$ for all $t\geq0$, with $K\geq1$ and $\lambda>0$ independent of the initial state in that region. A discrete-time version uses $Kq^k$ with $0<q<1$. The property is local if the initial-state condition is confined to a neighborhood and global if the same kind of bound holds on the full stated domain. This is stronger than mere eventual convergence in general.[^{ref-9a0e0ebc596b][}ref-c9ca18fcba8e]

Scope of Application¶

The definition applies to normed state trajectories near a declared equilibrium, not automatically to every decaying output. In finite-dimensional autonomous continuous LTI systems, all matrix eigenvalues in the open left half-plane yield the property; in discrete LTI systems, they must lie inside the unit circle. These are conditional certificates, not universal definitions for nonlinear or time-varying systems. A Lyapunov proof can also establish a nonlinear rate if the Lyapunov function is comparable to a positive power of state norm and decreases at a matching proportional rate; mere decrease is insufficient.[^{ref-0fddc3d08d1d][}ref-6927e73d7e40][^ref-9a0e0ebc596b]

Clarity¶

The term makes the rate and quantifiers explicit: which equilibrium, which norm, which initial states, and which common $K$ and positive rate? Asymptotic stability may lack this envelope even though a trajectory approaches the equilibrium. A global claim cannot be inferred from a neighborhood-only proof, and an LTI matrix test should not be imported into an arbitrary nonlinear model.[^{ref-9a0e0ebc596b][}ref-c9ca18fcba8e]

Manages Complexity¶

The envelope compresses many transient trajectories into a defensible bound. $K$ permits some initial overshoot and the positive rate fixes a ceiling on persistent error, but the inequality is not an exact trajectory formula or a quantified perturbation-robustness guarantee. Jordan polynomial factors in a stable linear model can still fit a slightly slower exponential bound; forcing the exact spectral-edge rate may fail.[^{ref-0fddc3d08d1d][}ref-6927e73d7e40]

Abstract Reasoning¶

Specify the equilibrium and an initial-state region; translate to zero if convenient. Establish one norm bound proportional to the initial error and exponentially decaying for every admitted start and every future time. Then select a proof method appropriate to the model—spectral criteria for finite-dimensional autonomous LTI dynamics, or sufficiently strong Lyapunov inequalities for nonlinear dynamics. The resulting property is a rate-qualified subtype of the broad live Stability prime, not a synonym for a Hurwitz-stable matrix.[^{ref-9a0e0ebc596b][}ref-c9ca18fcba8e]

Knowledge Transfer¶

A constructed continuous nonlinear example $\dot x=-x-x^3$ has $V=x^2/2$ and $\dot V\leq-2V$, giving $|x(t)|\leq e^{-t}|x_0|$ globally. A constructed discrete Jordan update with matrix $\begin{pmatrix}1/2&1\\0&1/2\end{pmatrix}$ has spectral radius $1/2$ and a polynomial transient, but $\|A^k x_0\|\leq K_q q^k\|x_0\|$ for any fixed $q$ between $1/2$ and $1$. The shared pattern is the common initial-error-relative envelope; the certificates and the appropriate time notation differ.[^{ref-9a0e0ebc596b][}ref-c9ca18fcba8e][^ref-6927e73d7e40]

[^ref-9a0e0ebc596b]: Murat Arcak, EE C222 Lecture 8 Notes, University of California, Berkeley, 2025, PDF pp.2–3 and p.5. [^ref-c9ca18fcba8e]: Berkeley IEOR 265, Lecture 9: Stability, PDF pp.1–2. [^ref-6927e73d7e40]: Stephen Boyd, EE263: Eigenvectors and Diagonalization, Stanford University, 2008, PDF pp.33–34. [^ref-0fddc3d08d1d]: Stephen Boyd, EE263: Solution via Laplace Transform and Matrix Exponential, Stanford University, 2007–08, PDF pp.24–27.

Relationships to Other Abstractions¶

Current abstraction Exponential Stability Domain-specific

Parents (1) — more general patterns this builds on

Exponential Stability is a kind of Stability Prime

A quantitative exponential-rate subtype of equilibrium stability.

Hierarchy path (1) — routes to 1 parentless root

Exponential Stability → Stability

Neighborhood in Abstraction Space¶

Exponential Stability sits in a sparse region of the domain-specific corpus (64^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Dynamical Systems & Differential Structures (37 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-08