Key Concepts of Phase Plane Analysis

Why This Matters

Phase plane analysis is where linear algebra and differential equations come together in a visual framework. It lets you connect eigenvalue analysis, matrix operations, and solution behavior to understand how dynamical systems evolve over time. The phase plane transforms abstract equations into geometric intuition, showing you why solutions spiral, converge, or diverge rather than just that they do.

This topic bridges nearly everything you've learned: eigenvalues determine stability, eigenvectors set trajectory directions, and linearization lets you apply matrix techniques to nonlinear problems. When you see a phase portrait, you should immediately think about the underlying Jacobian, its eigenvalues, and what those values tell you about long-term behavior. Don't just memorize classifications; know what mathematical features produce each type of behavior.

---

Foundations: Setting Up the Phase Plane

Before analyzing behavior, you need to understand how we represent dynamical systems geometrically. The phase plane converts a system of two first-order ODEs into a two-dimensional picture where every point represents a complete state of the system.

Definition of a Phase Plane

• Two-dimensional state space where each axis represents one dependent variable (like $x$ and $y$, or position and velocity)
• Every point is a complete system state, meaning if you know where you are in the phase plane, you know everything about the system at that instant
• Time is implicit: trajectories show the sequence of states but don't explicitly mark when each state occurs

Direction Fields and Trajectories

At each point $(x, y)$, the system $\dot{x} = f(x,y),\; \dot{y} = g(x,y)$ defines a vector $\begin{pmatrix} f(x,y) \\ g(x,y) \end{pmatrix}$. Plotting these vectors across the plane gives you a direction field, which shows the instantaneous direction of motion everywhere.

Trajectories are solution curves that follow these vectors, representing the actual path a system takes from a given initial condition. For autonomous systems (where $f$ and $g$ don't depend explicitly on $t$), the uniqueness theorem guarantees that trajectories never cross. This constraint is fundamental to all phase portrait analysis: if two trajectories crossed, a single initial condition would lead to two different futures, violating uniqueness.

---

Finding Critical Points: Equilibria and Nullclines

Equilibrium points are where the system is stationary, and understanding the flow around them is the first step in characterizing any system.

Equilibrium Points and Their Classification

An equilibrium point (also called a critical point or fixed point) occurs where $\dot{x} = 0$ and $\dot{y} = 0$ simultaneously. At such a point, the system has no tendency to move.

These points are classified by the behavior of nearby trajectories:

• Nodes (stable or unstable): trajectories approach or recede along definite directions
• Saddle points: trajectories approach along one direction and recede along another
• Spirals (stable or unstable): trajectories wind inward or outward
• Centers: trajectories form closed loops around the equilibrium

The long-term behavior of most initial conditions is determined by which equilibrium they approach (or are repelled from).

Nullclines and Their Significance

Nullclines are curves that help you map out the flow without solving anything.

• The $x$-nullcline is the set of points where $\dot{x} = 0$. On this curve, the velocity vector is purely vertical (only $\dot{y}$ is nonzero).
• The $y$-nullcline is the set of points where $\dot{y} = 0$. On this curve, the velocity vector is purely horizontal.
• Intersections of the two nullclines are equilibrium points, since both derivatives vanish there.

Between the nullclines, you can determine the sign of $\dot{x}$ and $\dot{y}$ in each region, which reveals the general flow direction. This makes nullclines one of the fastest ways to sketch a rough phase portrait by hand.

---

Stability Analysis: The Eigenvalue Connection

This is where linear algebra becomes essential. The eigenvalues of the Jacobian matrix evaluated at an equilibrium point completely determine local stability for hyperbolic equilibria (those whose eigenvalues have nonzero real parts).

Linearization of Nonlinear Systems

For a nonlinear system $\dot{x} = f(x,y),\; \dot{y} = g(x,y)$, the idea is to zoom in near an equilibrium point $(x_0, y_0)$ and approximate the system by its linear part. The tool for this is the Jacobian matrix:

$J = \begin{pmatrix} \frac{\partial f}{\partial x} & \frac{\partial f}{\partial y} \\ \frac{\partial g}{\partial x} & \frac{\partial g}{\partial y} \end{pmatrix}$

Here's the process:

1. Find the equilibrium points by solving $f(x,y) = 0$ and $g(x,y) = 0$ simultaneously.
2. Compute the four partial derivatives in $J$.
3. Evaluate $J$ at the equilibrium point $(x_0, y_0)$ to get a constant matrix.
4. Analyze the eigenvalues of that constant matrix to determine local behavior.

This linearization is valid when the eigenvalues have nonzero real parts (the hyperbolic case). In that situation, the Hartman-Grobman theorem guarantees the nonlinear system behaves qualitatively like its linearization near the equilibrium.

Eigenvalues and Eigenvectors in Phase Plane Analysis

Once you have the Jacobian evaluated at an equilibrium, its eigenvalues $\lambda_1, \lambda_2$ tell you almost everything:

• Real parts determine stability: both negative means stable, both positive means unstable, opposite signs means saddle.
• Imaginary parts indicate oscillation: pure imaginary eigenvalues give centers, complex eigenvalues with nonzero real part give spirals.
• Eigenvectors determine trajectory geometry. For real eigenvalues, trajectories approach or depart along the eigenvector directions. The eigenvector associated with the eigenvalue of larger magnitude determines the direction trajectories follow when far from the equilibrium; the one with smaller magnitude dominates close in.

Stability Classification via Trace and Determinant

Computing eigenvalues directly can be tedious. The trace-determinant method offers a shortcut. For the Jacobian $J$:

• $\tau = \text{tr}(J) = \lambda_1 + \lambda_2$
• $\Delta = \det(J) = \lambda_1 \cdot \lambda_2$

The classification rules:

• Asymptotically stable: $\tau < 0$ and $\Delta > 0$ (both eigenvalues have negative real parts)
• Unstable node or spiral: $\tau > 0$ and $\Delta > 0$
• Saddle point: $\Delta < 0$ (eigenvalues have opposite signs)
• Node vs. spiral: determined by the discriminant $\tau^2 - 4\Delta$. If positive, eigenvalues are real (node). If negative, eigenvalues are complex (spiral).
• Borderline cases: $\tau = 0$ with $\Delta > 0$ gives pure imaginary eigenvalues (center in the linear system), and $\Delta = 0$ means at least one zero eigenvalue (non-isolated or degenerate equilibrium).

---

Beyond Equilibria: Global Behavior

Local analysis tells you what happens near equilibrium points, but phase plane analysis also reveals global structures. Limit cycles and bifurcations describe behaviors that linearization alone cannot capture.

Limit Cycles and Periodic Solutions

A limit cycle is an isolated closed trajectory in the phase plane. "Isolated" means there's no continuous family of closed orbits next to it; nearby trajectories either spiral toward it (stable limit cycle) or away from it (unstable limit cycle).

Limit cycles represent self-sustained oscillations: the system settles into a periodic orbit regardless of where it starts (as long as it starts in the basin of attraction). This is fundamentally different from the centers of linear systems, where closed orbits exist in continuous families and any perturbation destroys the periodicity.

The Poincaré-Bendixson theorem provides a way to prove limit cycles exist: if a trajectory in a planar autonomous system remains in a bounded region that contains no equilibrium points, then the trajectory must approach a limit cycle. This theorem only applies in two dimensions.

Bifurcations and Their Types

A bifurcation occurs when a small change in a parameter causes a qualitative change in the system's behavior, such as equilibria appearing, disappearing, or changing stability.

Two particularly important types:

• Saddle-node bifurcation: two equilibria (one stable, one unstable) collide and annihilate each other as a parameter crosses a critical value. Before the bifurcation, two equilibria exist; after, none do.
• Hopf bifurcation: a stable equilibrium loses stability as its complex eigenvalues cross the imaginary axis, and a limit cycle is born. The equilibrium transitions from a stable spiral to an unstable spiral surrounded by a stable limit cycle.

Bifurcation diagrams plot equilibrium locations and their stability against a parameter, revealing the critical thresholds where behavior changes.

---

Phase Portraits: Putting It All Together

Phase portraits synthesize everything: equilibria, stability, nullclines, and global flow. Different system types produce characteristic portrait patterns that you should be able to recognize and construct.

Phase Portraits for Different System Types

• Linear systems produce nodes, saddles, spirals, or centers depending entirely on eigenvalues. Limit cycles cannot occur in linear systems.
• Nonlinear systems can exhibit multiple equilibria with different stability types, limit cycles, and complex basins of attraction.
• Conservative systems (like undamped oscillators) have a conserved quantity, producing centers and families of closed orbits. Dissipative systems (with energy loss or gain) show attracting equilibria or limit cycles instead.

---

Quick Reference Table

---

Self-Check Questions

1. If a system has eigenvalues $\lambda = -2 \pm 3i$ at an equilibrium point, what type of equilibrium is it, and is it stable? What would change if the real part were positive?

2. Compare and contrast nullclines and eigenvectors. Both give directional information, but how do they differ in what they tell you about the system?

3. A nonlinear system has a bounded region with no equilibrium points inside. What does the Poincaré-Bendixson theorem tell you must exist, and how does this differ from what's possible in linear systems?

4. Given a Jacobian with $\text{tr}(J) = 0$ and $\det(J) > 0$, classify the equilibrium. Why does linearization fail to fully characterize stability in this case?

5. Explain how a Hopf bifurcation differs from a saddle-node bifurcation in terms of what happens to equilibria and what new structures might appear.