piatra.institute

PIATRA . INSTITUTE

Morphologies of stability

canonical patterns by which dynamical systems hold form through disturbance
based on 2015, Strogatz, Nonlinear Dynamics and Chaos · 1940, Kramers, Brownian motion in a field of force
active pattern
Point attractor
The simplest stabilizer: deviations generate a restoring drift back toward a set point. Stability here means convergence to rest.
dx/dt = -k(x - x*) + ξ(t)
stability index
0.800
Lyapunov λ
-1.200
energy
1.1760
error |x - x*|
1.400
claude opus 4.8April 2026·initial implementation with four canonical stabilizing patterns
State line, restoring flow toward target
x*x
History, error decay over time
sensitivity analysis · stability index
noise
Δ0.676
restoring force k
Δ0.509
baseline: 0.800 · each parameter swept min→max while others held constant

What counts as stable?

The four patterns in this playground represent four fundamentally different answers to the question “what does it mean for a system to hold its form?” A point attractor holds form by returning to rest. A bistable switch holds form by occupying one of two valleys. A limit cycle holds form by sustaining a rhythm. A consensus network holds form by coordinating many units into agreement. These are not the same kind of stability. They are different morphologies of stability.

Point attractor

The simplest case. A linear restoring force drives the state back toward a target:

dxdt=−k(x−x∗)+ξ(t)\frac{dx}{dt} = -k(x - x^*) + \xi(t)

The Lyapunov function V=12k(x−x∗)2V = \tfrac{1}{2}k(x - x^*)^2 decreases monotonically along trajectories. The decay is exponential with time constant τ=1/k\tau = 1/k. This is stability as convergence to rest. The system has a single equilibrium and every initial condition flows toward it.

Bistable switch

A particle in a double-well potential with damping and noise:

V(x)=a(x2−1)2+tilt⋅xV(x) = a(x^2 - 1)^2 + \text{tilt} \cdot x
d2xdt2=−γdxdt−dVdx+ξ(t)\frac{d^2x}{dt^2} = -\gamma \frac{dx}{dt} - \frac{dV}{dx} + \xi(t)

Kramers escape theory (1940) gives the transition rate between wells: rate∝exp⁡(−2ΔV/σ2)\text{rate} \propto \exp(-2\Delta V / \sigma^2), where ΔV\Delta V is the barrier height and σ2\sigma^2 the noise intensity. This is stability as basin selection. The system has multiple stable states, and identity depends on which valley you occupy.

Limit cycle

The Hopf normal form in complex notation:

dzdt=(μ−∣z∣2) z+iωz+ξ(t)\frac{dz}{dt} = (\mu - |z|^2)\,z + i\omega z + \xi(t)

When μ>0\mu > 0, the origin is unstable and trajectories converge to a stable orbit of radius r∗=μr^* = \sqrt{\mu}. The Floquet exponent for radial perturbations is λr=−2μ\lambda_r = -2\mu, governing how fast the system returns to the orbit after a kick. This is stability as rhythm. The system never rests, but its pattern of motion is self-restoring.

Consensus network

DeGroot dynamics with an external anchor:

dxidt=c(xˉ−xi)+s(a−xi)+ξi(t)\frac{dx_i}{dt} = c(\bar{x} - x_i) + s(a - x_i) + \xi_i(t)

Each agent ii is pulled toward the group mean xˉ\bar{x} with coupling cc and toward an anchor aa with stubbornness ss. The effective convergence rate is λ=c(1−1/N)+s\lambda = c(1 - 1/N) + s. This is stability as coordination. The stable object is a collective configuration, not any single unit.

The philosophical bridge

“Becoming” becomes legible when a process can repeatedly recover its own form. That is the bridge from dynamics to entityhood. A circle drawn on paper is a static shape. A droplet becoming round under surface tension is a self-stabilizing process. The distinction matters: the droplet maintains its form through active correction, not through inertness.

The four patterns here represent four ways a system can achieve this: by relaxing to a point, by selecting a basin, by sustaining a rhythm, or by coordinating across units. Each is a morphology of stability, a distinct way that “holding form” can be realized in a dynamical system.

Open questions

This playground covers four canonical patterns, but stability in real systems involves additional morphologies: metastability (long-lived transients that aren’t true equilibria), self-organized criticality (systems that tune themselves to the edge of instability), excitable systems (stable rest with threshold-triggered excursions), and autopoiesis (systems that actively reconstruct their own boundary). Each deserves its own exploration.

Calibration

Each case drives the noise-free core of one morphology and checks it against a textbook result: a fixed point stays put, a displaced state relaxes to its target, the Hopf orbit settles at radiusμ\sqrt{\mu}, the subcritical regime collapses to the origin, and the consensus network reaches agreement.

Model changelog

v1April 2026
  • Four dynamical systems: point attractor, bistable switch, Hopf limit cycle, DeGroot consensus
  • Real-time simulation with requestAnimationFrame and configurable noise injection
  • Custom SVG phase portraits: state line with flow arrows, double-well potential landscape, 2D orbit trail, network graph
  • Five presets encoding distinct stability morphologies including near-bifurcation regime
  • Lyapunov exponent estimation and stability index for each pattern
  • Kramers escape rate approximation for bistable barrier crossings
  • Parameter sweep and sensitivity analysis across pattern-relevant parameters
  • Perturbation, randomization, and state reset controls
read the research companion →