An equilibrium is a state where the autonomous vector field vanishes.
Linear stability studies the leading evolution of a small perturbation around a specified solution. Near an equilibrium of , write . Dropping terms quadratic and higher in the perturbation gives , where is the Jacobian matrix .
An eigenvalue with positive real part gives a growing mode. If every eigenvalue has negative real part, sufficiently small perturbations decay and the smooth nonlinear equilibrium is locally asymptotically stable. Eigenvalues on the imaginary axis require additional analysis: nonlinear terms can change the outcome, and repeated neutral eigenvalues can have algebraic growth even in a linear system. The test also describes local behavior; a nearby stable equilibrium can coexist with a different remote attractor. State, parameter values and perturbation class are therefore part of a stability claim.
For a discrete dynamical system, eigenvalues of the linearization strictly inside the unit circle imply local asymptotic stability; an eigenvalue outside implies instability. Eigenvalues on the unit circle require more information. A nontrivial Jordan block there can already make the linearized iterates unbounded.
Even a bounded linearization leaves nonlinear stability undecided. The two displayed maps of the complex plane have the same derivative, a rotation matrix. Their radial updates are and . For small positive radius, the first decreases to zero and the second moves out of every sufficiently small fixed neighborhood. Thus identical unit-modulus linear multipliers can accompany attraction or instability.
For a scalar ordinary differential equation and an equilibrium point with , the linearization gives . Thus small perturbations decay like with . This is a local exponential recovery time. In several dimensions, the slowest decaying stable eigenvalue typically controls asymptotic recovery, subject to additional polynomial factors for nontrivial Jordan blocks.
For and , with , the positive equilibrium is , . Its Jacobian has trace and determinant . The unit-disc root conditions reduce to . At the roots form a nonreal conjugate pair and cross the unit circle; their argument satisfies . The unique threshold lies between one and because increases strictly and at threshold.
For near an equilibrium point, the stability matrix is the Jacobian matrix of the vector field evaluated at that point. Perturbations obey to first order. Its eigenvalues determine linear growth and decay: strictly negative real parts imply local asymptotic stability, while a positive real part gives instability. Zero real parts require further analysis. The stability matrix of a renormalization-group fixed point applies this construction to beta-function flows.
A conservative equilibrium with one real pair of linear eigenvalues and one imaginary pair has saddle and centre directions. Generic perturbations have an exponentially growing component, even though finely chosen initial conditions can remain in the centre-stable subspace. The local L2 Lagrange point dynamics is an example.
A sideband instability is growth of disturbances with wavenumbers slightly displaced from a carrier pattern's wavenumber. A real physical perturbation uses the displaced mode together with its complex conjugate; a complex amplitude equation can couple those two sidebands through its cubic term. The Eckhaus instability is a longitudinal phase-modulation example.
For a spatially periodic nonlinear pattern, a linear stability disturbance is written with periodic on the pattern cell. The conjugated operator acts on that fixed cell and determines the sideband eigenvalues. Translation invariance supplies zero eigenfunctions at , namely spatial derivatives of the base pattern. Their small- eigenvalues determine long-wavelength phase stability; other parts of the spectrum can give independent instabilities.
Suppose a Bloch stability operator has expansion , with a simple neutral eigenfunction and an adjoint neutral function normalized by . Expand and . The Fredholm alternative gives and the cell problem , . Applying the adjoint solvability condition at second order gives the displayed phase-diffusion coefficient. Multiple translation phases or a conserved mean require a coupled matrix problem rather than this scalar formula.
A spatial sideband is a Fourier perturbation with wavenumber close to a pattern's carrier wavenumber, such as or . A real physical field also includes the conjugate Fourier component. Linearizing a nonlinear complex amplitude equation can couple the two displaced components, so testing only a single uniform amplitude misses their joint linear stability problem.
Overstability is oscillatory instability: a perturbation grows while continuing to change sign or direction periodically. A real perturbation can contain with and , corresponding to a complex-conjugate pair of linear growth rates with positive real part.
The oscillation period is approximately , while sets the envelope-growth time. Successive peak heights can therefore reveal growth even when a short segment of the trace resembles an ordinary oscillation. This differs from a stationary instability with a purely real positive growth rate and from a damped oscillation with a negative envelope growth rate. The distinction is useful when classifying how an equilibrium first becomes unstable as a control parameter changes, including an oscillatory onset associated with a Hopf bifurcation.
For a real Jacobian matrix , both eigenvalues have strictly negative real parts if and only ifThe characteristic polynomial is . If its roots are real, positive product makes their signs agree and negative sum makes both negative. If they are a complex conjugate pair, each real part is half the trace. Conversely, two eigenvalues with negative real parts have negative real sum and positive product.
For a continuous-time nonlinear system, the linearization stability theorem turns this strict test into local asymptotic stability. If , the real eigenvalues have opposite signs and the equilibrium is a saddle. Equality in either condition needs separate analysis. A discrete map instead requires eigenvalues inside the unit disk; the trace-determinant inequalities for negative real parts are not its stability test.
Articles by others on the same topic
There are currently no matching articles.