SYDE 554: Nonlinear Dynamic Systems

Estimated study time: 10 minutes

Table of contents

Sources and References

Strogatz, Nonlinear Dynamics and Chaos (Westview)
Khalil, Nonlinear Systems (Prentice Hall)
Guckenheimer and Holmes, Nonlinear Oscillations, Dynamical Systems, and Bifurcations of Vector Fields (Springer)
Wiggins, Introduction to Applied Nonlinear Dynamical Systems and Chaos (Springer)
Younis, MEMS Linear and Nonlinear Statics and Dynamics (Springer)

Chapter 1: Dynamic Systems Fundamentals

1.1 Linear vs Nonlinear

A linear time-invariant system satisfies superposition and homogeneity. Nonlinear systems do not: doubling the input need not double the output; combined inputs produce effects beyond the sum of individual responses. Consequences include multiple equilibria, limit cycles, chaos, and amplitude-dependent frequency — phenomena absent from linear theory.

1.2 Autonomous and Non-Autonomous

An autonomous system has no explicit time dependence:

\[ \dot{\mathbf{x}} = \mathbf{f}(\mathbf{x}). \]

A non-autonomous system includes time explicitly:

\[ \dot{\mathbf{x}} = \mathbf{f}(\mathbf{x}, t). \]

Forced oscillators and systems with time-varying parameters are non-autonomous, typically analysed by adding time as an extra state variable to produce an equivalent autonomous system in higher dimension.

Chapter 2: Existence, Uniqueness, and Qualitative Theory

2.1 Existence and Uniqueness

Picard-Lindelöf conditions — Lipschitz continuity of \( \mathbf{f} \) — guarantee local existence and uniqueness of solutions to \( \dot{\mathbf{x}} = \mathbf{f}(\mathbf{x}, t) \) through a given initial condition. Without Lipschitz continuity, solutions can branch or fail to exist beyond finite time (finite-time blow-up).

2.2 Equilibrium Points

An equilibrium \( \mathbf{x}^{*} \) satisfies \( \mathbf{f}(\mathbf{x}^{*}) = \mathbf{0} \). Local behaviour near an equilibrium is governed by the Jacobian

\[ J = \left. \frac{\partial \mathbf{f}}{\partial \mathbf{x}} \right|_{\mathbf{x}^{*}}. \]

Eigenvalues of \( J \) classify the equilibrium: stable node or focus (all real parts negative), unstable (any real part positive), saddle (mixed signs), centre (zero real parts — requires nonlinear analysis).

The damped pendulum has equilibria at \( \theta = 0, \pi \) (for simplicity modulo \( 2\pi \)). At 0 the Jacobian has eigenvalues with negative real parts — a stable focus; at \( \pi \) a saddle. Trajectories from the saddle separate basins of attraction.

2.3 Phase Plane Analysis

Two-dimensional systems are fully characterised by trajectories in the phase plane (state variables on the axes). Nullclines \( \dot{x}_1 = 0 \) and \( \dot{x}_2 = 0 \) partition the plane into regions of characteristic flow. Isoclines, trajectories, and separatrices between basins paint a qualitative picture without numerical solution.

Chapter 3: Stability of Equilibria and Lyapunov Theory

3.1 Lyapunov Stability

An equilibrium is Lyapunov stable if trajectories starting nearby remain nearby; asymptotically stable if they also converge to the equilibrium. Lyapunov’s direct method constructs a scalar function \( V(\mathbf{x}) \) that is positive definite and whose time derivative along trajectories,

\[ \dot{V} = \nabla V \cdot \mathbf{f}(\mathbf{x}), \]

is negative definite (asymptotic) or semidefinite (stable).

3.2 LaSalle’s Invariance Principle

When \( \dot{V} \leq 0 \), LaSalle’s principle says trajectories approach the largest invariant set where \( \dot{V} = 0 \). This extends Lyapunov analysis to systems with damping that is zero at isolated points but still drives asymptotic convergence.

3.3 Region of Attraction

Lyapunov functions also bound the region of attraction: the set of initial conditions from which trajectories converge to a given equilibrium. Estimating this region is central to certifying safe operating envelopes for controllers and actuators.

Chapter 4: Limit Cycles and Bifurcations

4.1 Limit Cycles

A limit cycle is an isolated closed orbit in phase space. Van der Pol and other self-excited oscillators exhibit stable limit cycles whose amplitudes are set by nonlinear damping. The Poincaré-Bendixson theorem states that a trajectory trapped in a closed, bounded region of the plane without equilibria must approach a limit cycle — a powerful tool for two-dimensional flows.

4.2 Bifurcations

Bifurcations are qualitative changes in dynamics as a parameter varies. Saddle-node bifurcations create or destroy equilibrium pairs; transcritical exchanges stability between two equilibria; pitchfork produces symmetric equilibrium branching; Hopf bifurcations convert stable foci into limit cycles. Normal forms capture the essential local dynamics near each bifurcation type.

A Hopf bifurcation occurs when a pair of complex-conjugate eigenvalues of the Jacobian cross the imaginary axis as a parameter varies; beyond the crossing, a limit cycle emerges whose amplitude grows as the square root of the parameter distance.

4.3 Nonlinear Phenomena

Multistability — multiple coexisting attractors — arises when equilibria, limit cycles, or chaotic sets share state space. Hysteresis follows: the attractor a trajectory settles on depends on its history. Jump phenomena occur when a continuously varied parameter crosses a saddle-node; the system snaps to a distant attractor, a behaviour exploited in switches and avoided in sensitive controls.

Chapter 5: Advanced Nonlinear Topics

5.1 Poincaré Sections

A Poincaré section is a lower-dimensional cross-section of phase space intersected transversely by trajectories. The flow induces a Poincaré map on the section that reduces continuous-time dynamics to discrete-time iteration — a tool for analysing limit cycles (fixed points of the map), period-doubling, and chaos (complex map dynamics).

5.2 Multi-Frequency Oscillations

Coupled oscillators with different natural frequencies can synchronise, lock into rational frequency ratios, or exhibit quasi-periodic motion on tori. Devil’s staircases and Arnold tongues organise the parameter space of mode-locking; their structure recurs in circadian rhythms, cardiac dynamics, and laser arrays.

5.3 Chaos

Chaotic systems are deterministic yet unpredictable due to sensitive dependence on initial conditions (positive Lyapunov exponent). Lorenz, Rössler, and duffing systems exemplify low-dimensional chaos. Strange attractors with fractal geometry are the asymptotic sets. Chaos complicates prediction but can be beneficial — efficient mixing in microfluidics, signal encryption, and exploration in machine learning — or harmful — unpredictable vibrations, irregular heartbeats.

Chaos is not noise. It is fully deterministic and, in principle, predictable from exact initial conditions; but real-world measurement precision bounds practical prediction to a finite horizon that shrinks with Lyapunov exponent.

Chapter 6: Applications and MEMS/NEMS

6.1 Mechanical and Electromechanical Systems

Mechanical systems exhibit nonlinearity through large-amplitude geometric terms, contact and friction, nonlinear springs, and coupling between degrees of freedom. MEMS and NEMS devices — microcantilevers, resonators, capacitive actuators — routinely operate near pull-in, mode-coupling, and parametric resonances. Duffing’s equation

\[ \ddot{x} + 2\zeta\omega_0 \dot{x} + \omega_0^{2} x + \alpha x^{3} = F\cos(\Omega t) \]

captures amplitude-dependent frequency and bistability characteristic of these devices.

6.2 Electrical and Chemical Systems

Nonlinear electrical systems include tunnel diodes, Van der Pol oscillators realized in RLC circuits, Chua’s circuit (chaotic), and saturating magnetic circuits. Chemical systems — Belousov-Zhabotinsky reaction, glycolytic oscillations, enzyme kinetics — host limit cycles, chaos, and wave patterns. Engineering applications range from oscillator design to biochemical sensing.

6.3 Ecological and Biological Systems

Population dynamics, epidemic spread, neural networks, and cardiac dynamics are all nonlinear. The Hodgkin-Huxley equations for nerve impulses exhibit action-potential limit cycles; FitzHugh-Nagumo and Morris-Lecar reduce these to two-dimensional caricatures amenable to phase-plane analysis. Predator-prey, SIR-model, and reaction-diffusion equations yield rich structures including travelling waves, Turing patterns, and chaos.

6.4 Design and Control Implications

Engineers design with nonlinearity rather than around it. Nonlinear control — gain scheduling, feedback linearisation, sliding mode, backstepping — handles systems that defy linear approximation. Nonlinear observers, model predictive control, and data-driven methods extend the toolkit. Understanding limit cycles, bifurcations, and chaos guides when to stabilise, when to exploit, and when to avoid.

A MEMS resonator near its Duffing bistability can amplify small signals by biasing just below the jump; beyond it, the amplitude switches abruptly — the basis of highly sensitive mass-detection sensors and mechanical switches.

Nonlinear dynamics provides the rigorous language for behaviours linear analysis cannot represent — equilibria, limit cycles, bifurcations, chaos — and the analytical tools (Lyapunov, phase plane, Poincaré, bifurcation theory) to engineer with these behaviours rather than pretend them away.

Students emerge equipped to analyse and design systems in which nonlinearity is central: sensors, actuators, circuits, biological interfaces, and any domain where richness of dynamics exceeds the reach of linear approximation.