A phase portrait is a visual map of how a system changes over time. Instead of plotting a variable against time on a graph, a phase portrait plots the system’s variables against each other, revealing every possible path the system can take from any starting point. It’s one of the most powerful tools in mathematics and science for understanding systems described by differential equations, because a single image captures the full range of a system’s behavior: where it settles down, where it blows up, and where it cycles.
The Core Idea: State Space
To understand a phase portrait, you first need the concept of state space (also called phase space). Every system that changes over time has variables that describe its current condition. A swinging pendulum, for example, has two: its angle and its velocity. Together, these variables define the system’s “state” at any moment. State space is the collection of all possible states the system could be in, and each point in that space represents one specific combination of those variables.
A phase portrait is a sketch of this state space with arrows or curves showing how the system moves through it over time. Each curve, called a trajectory, traces the path the system follows from a particular starting point. You pick a starting condition, and the trajectory shows you the system’s entire future (and past) without needing to solve the equations exactly. A good phase portrait includes a representative selection of these trajectories so you can see the overall pattern.
What Trajectories Tell You
The trajectories in a phase portrait are driven by a vector field. At every point in state space, the differential equations that govern the system assign a direction and speed of change. You can think of this as a field of tiny arrows covering the plane, each one pointing in the direction the system would move if it happened to be at that spot. Trajectories are the paths you get by following those arrows continuously from a given starting point.
Two trajectories can never cross each other (as long as the system is deterministic), which means the phase portrait divides state space into distinct regions of behavior. If you start somewhere inside one region, you stay on a predictable path. This is what makes phase portraits so useful: they give you the global picture of a system at a glance, showing which starting conditions lead to similar outcomes and which lead to dramatically different ones.
Equilibrium Points
The most important features of any phase portrait are its equilibrium points. These are locations in state space where all the rates of change are zero, meaning the system, if placed exactly there, would stay put forever. Nearby trajectories reveal what kind of equilibrium you’re looking at.
- Nodes (sinks and sources): A stable node, or sink, is a point where all nearby trajectories converge inward. The system settles down to rest. An unstable node, or source, is the opposite: trajectories stream outward from it in all directions.
- Saddle points: These are unstable, but in a specific way. Trajectories approach the equilibrium along one direction and flee from it along another. If you picture two highways crossing at an intersection, traffic flows in along one road and out along the other.
- Centers: Trajectories form closed loops around the equilibrium. The system oscillates forever without growing or decaying. A frictionless pendulum produces a center in its phase portrait.
- Spirals: Like centers, but the loops either tighten inward (stable spiral) or expand outward (unstable spiral). A pendulum with friction produces a stable spiral: it oscillates, but each swing gets smaller until it stops.
How Stability Is Determined
For systems described by linear equations, the character of each equilibrium point is determined entirely by quantities called eigenvalues, which emerge from the mathematical structure of the system. You don’t need to know how to calculate them to understand what they mean: eigenvalues are numbers (sometimes involving imaginary parts) that encode the rate and style of motion near an equilibrium.
When both eigenvalues are negative real numbers, trajectories shrink toward the equilibrium, making it a stable node. When both are positive, trajectories expand away, creating an unstable node. When one is positive and one is negative, you get a saddle point. If the eigenvalues have an imaginary component, the system rotates, producing spirals or centers. A negative real part combined with an imaginary part gives a stable spiral. A purely imaginary eigenvalue, with no real part at all, gives a perfect center with no decay.
For nonlinear systems, the same classification applies locally. You zoom in on each equilibrium point, approximate the system as linear in that neighborhood, and classify accordingly. The phase portrait then stitches together these local pictures into a global one.
Limit Cycles and Sustained Oscillations
Not every repeating behavior in a phase portrait is a center. In nonlinear systems, you can find limit cycles: isolated closed loops that nearby trajectories spiral into (or away from). A limit cycle represents sustained oscillations at a fixed amplitude, and what makes it special is that the amplitude doesn’t depend on where the system started. Whether you begin close to the loop or far from it, trajectories converge onto the same repeating path.
This is fundamentally different from the closed orbits around a center, where every starting point produces a different-sized loop. Limit cycles are structurally robust. They show up in models of heartbeats, predator-prey population cycles, and electrical oscillators, anywhere a system naturally locks into a rhythm regardless of initial conditions.
How to Read a Phase Portrait
If you’re constructing or interpreting a phase portrait for the first time, there’s a practical sequence that makes the process manageable.
Start by finding the nullclines. These are curves in state space where one of the rates of change equals zero. Where nullclines for different variables intersect, both rates are zero simultaneously, which means you’ve found an equilibrium point. Mark these clearly.
Next, determine the direction of flow in each region between the nullclines. The nullclines divide the plane into zones where each variable is either increasing or decreasing. You don’t need exact values, just the signs. Draw small arrows in each region indicating the general direction of motion. Then sketch trajectories by following those arrows from various starting points. When a trajectory hits a nullcline, it crosses either horizontally or vertically (because one component of its velocity is zero there), then continues into the next region following the new direction. Most trajectories will flow toward an equilibrium point, spiral around one, or head off toward the edges of the portrait.
The result is a map of all possible long-term behaviors. You can see at a glance which equilibria are attracting, which are repelling, and which regions of initial conditions lead to which outcomes.
Beyond Two Dimensions
Phase portraits are most intuitive in two dimensions, where the state space is a flat plane you can draw on paper. But many real systems have three, ten, or hundreds of variables. Complete phase portraits are generally limited to presenting information in two dimensions, so higher-dimensional systems require workarounds.
The standard approach is to take two-dimensional slices. For a system with three variables, you can create a phase portrait for each pair of variables while holding the third constant. The number of unique pairs you need equals the number of ways to choose 2 variables from your total: a 3-variable system needs 3 portraits, a 4-variable system needs 6, and so on. Interactive tools let you adjust the value of the held-constant variables with sliders, updating the portrait in real time so you can explore how the dynamics shift across different planes of the state space.
Where Phase Portraits Are Used
Phase portraits originated in the study of mechanical systems like pendulums and planetary orbits, but they’ve become essential across the sciences. In biology, phase space analysis has been used to study embryonic development in fruit flies and roundworms, revealing how gene regulatory networks guide cells toward different fates. The same framework applies to models of cell cycle regulation, stem cell differentiation, and microbial metabolism.
In ecology, the classic Lotka-Volterra predator-prey model produces a phase portrait with closed orbits: predator and prey populations cycle endlessly around an equilibrium, and the portrait shows you the relationship between the two populations without needing to track time at all. In neuroscience, phase portraits capture the dynamics of neuron firing, showing how voltage and ion channel states interact to produce action potentials. In engineering, they’re used to analyze control systems, circuit behavior, and robotic motion.
What makes the phase portrait so broadly useful is that it abstracts away the specific details of a system and focuses on structure. Two completely different physical systems, a chemical reaction and an electrical circuit, can produce identical phase portraits if their underlying equations share the same mathematical form. Once you learn to read one, you can read them all.