DIFFERENTIAL EQUATIONS • SYSTEMS OF DIFFERENTIAL EQUATIONS

Stability in Systems — Stability and Long-Term Behavior in Systems

Discover how equilibrium points and the behavior of solutions predict whether a system settles down, grows without bound, or oscillates forever.

Historical Context & Motivation

Long before engineers designed autopilot systems or biologists modeled predator-prey populations, mathematicians grappled with a deceptively simple question: if you nudge a system slightly away from its resting point, will it return to rest or fly off to extremes? This question of stability sits at the heart of differential equations and has shaped centuries of scientific progress. Understanding stability allows us to predict whether a bridge will withstand gusts of wind, whether a chemical reaction will reach a safe equilibrium, or whether a population of animals will thrive or collapse.

The development of stability theory followed a path from celestial mechanics to modern control engineering. Each breakthrough built on the idea that the long-term behavior of a system—whether it settles, grows, or oscillates—can be determined by analyzing the system's equations near special points called equilibrium points.

1687
Newton's Principia
Isaac Newton published his laws of motion and universal gravitation, creating the first differential equations to describe planetary orbits and laying the groundwork for asking whether those orbits are stable over long periods of time.
1788
Lagrange's Stability Criteria
Joseph-Louis Lagrange studied small oscillations around equilibrium positions, introducing the key idea that we can understand a complicated system's behavior near a rest point by studying a simpler, approximate version of it.
1892
Lyapunov's Stability Theory
Aleksandr Lyapunov published his doctoral thesis, rigorously defining stable, unstable, and asymptotically stable equilibria. His methods—which use energy-like measuring functions to track whether solutions grow or shrink—remain foundational in engineering and applied mathematics today.
1960s
Modern Control Theory
Engineers applied stability analysis to design feedback controllers for aircraft, spacecraft (including the Apollo missions), and industrial processes, turning abstract mathematics into life-saving technology.

The central question this lesson addresses is: given a system of differential equations, how can we determine whether solutions near an equilibrium point will move toward it, move away, or keep circling around it? By the end, you will be able to read a direction field, identify equilibrium points, and use algebraic tests to predict long-term behavior—the same core ideas used by scientists and engineers worldwide.

Core Principles & Definitions

Before diving into calculations, let's establish the key ideas. A system of differential equations describes how two or more quantities change simultaneously over time. For instance, the temperature inside a building and the energy used by its heater form a coupled system—each affects the other. We study such systems by focusing on their equilibrium points and asking what happens to nearby solutions as time increases.

1

Equilibrium Point

A point where all rates of change equal zero. The system is perfectly at rest here. Also called a fixed point or critical point.
2

Stable Equilibrium

If solutions that start near the equilibrium stay near it (and possibly converge to it), the equilibrium is stable. Think of a marble at the bottom of a bowl—push it, and it rolls back.
3

Unstable Equilibrium

If even a tiny push sends solutions away from the equilibrium, it is unstable. Picture a marble balanced on top of an upside-down bowl—any nudge sends it rolling away.
4

Asymptotic Stability

A stronger form of stability: solutions not only stay close but actually converge to the equilibrium as time goes on. The system genuinely returns to rest, like a pendulum slowing down due to friction.
5

Phase Portrait

A diagram showing possible trajectories of a system by plotting one variable against the other (instead of against time). Arrows show the direction of change, revealing whether solutions flow toward, away from, or around equilibria.
KEY TAKEAWAY
Think of stability like a ball on different surfaces. A ball at the bottom of a valley is stable—disturb it and it rolls back. A ball on a hilltop is unstable—any push sends it away. A ball on a perfectly flat table is neutrally stable—it stays wherever you push it but doesn't return. Stability analysis tells us which surface our system's equilibrium sits on.

Visual Explanation — Phase Portraits

A phase portrait is the most powerful visual tool for understanding stability. Instead of plotting each variable against time separately, we plot the variables against each other on the same graph. Every point in the plane represents a possible state of the system, and arrows show how that state evolves over time. The diagram below shows three fundamental types of equilibrium behavior.

Three types of equilibrium in a 2D system. Left: a stable node where all trajectories converge to the origin. Center: an unstable node where all trajectories diverge. Right: a saddle point where trajectories approach along one direction and retreat along another.

In the stable node (left), every trajectory curves inward toward the equilibrium at the origin. No matter where you start nearby, the system eventually reaches rest. In the unstable node (center), the reverse happens: solutions flee the origin in every direction. The saddle point (right) is the trickiest—solutions approach along one direction but are repelled along the other, making the equilibrium unstable overall because most starting positions lead to trajectories that eventually escape.

Mathematical Framework — Analyzing Stability Algebraically

For a two-equation linear system, we can write both equations together in a compact form. The numbers that appear in those equations—how strongly each variable affects each rate of change—are arranged in a rectangular grid of numbers called a coefficient matrix. Two simple summary numbers computed from this grid, called the trace and the determinant, turn out to be all we need to classify stability in most cases.

TWO-VARIABLE LINEAR SYSTEM
dx/dt = ax + by, dy/dt = cx + dy
Here x and y are the two quantities changing over time, and a, b, c, d are constants describing how each quantity affects the other's rate of change. This is the standard form of a two-variable linear system.
TRACE AND DETERMINANT
trace = a + d, det = ad − bc
The trace is the sum of the two main-diagonal entries: trace = a + d. The determinant is computed as: det = ad − bc. Both can be found by simple arithmetic once you know the four coefficients a, b, c, d.

These two numbers contain all the information we need to determine stability. The key insight is that the trace tells us whether solutions grow or shrink on average, while the determinant tells us whether the equilibrium is a node or saddle. A negative trace means the system is losing energy overall; a positive trace means it is gaining energy.

STABILITY SHORTCUT
Asymptotically stable ⟺ trace < 0 AND det > 0
If the trace is negative (the sum a + d is negative) and the determinant is positive (ad is larger than bc), then all nearby solutions decay to the equilibrium—it is asymptotically stable. If the determinant is negative, the equilibrium is a saddle point and is always unstable.

To go further and fully classify the equilibrium type (for example, to distinguish a node from a spiral), we solve the characteristic equation—a quadratic equation built from the trace and determinant. Its two solutions are called the characteristic values (also known as eigenvalues in more advanced courses). You already know how to solve quadratic equations using the quadratic formula, so this step uses familiar algebra.

CHARACTERISTIC EQUATION
λ² − (trace)λ + det = 0
Solving this quadratic using the quadratic formula gives two characteristic values λ₁ and λ₂. The discriminant (trace)² − 4·det tells us whether those values are two distinct real numbers (discriminant > 0), a repeated real number (discriminant = 0), or a pair of complex numbers involving a square root of a negative number (discriminant < 0).

When the discriminant is negative, the characteristic values involve the square root of a negative number. In that case, we write them as λ = α ± βi, where α is the real part and β is the imaginary part (using the symbol i to denote √(−1), a concept introduced in algebra courses). The real part α determines growth or decay, and the imaginary part β produces oscillation. You do not need to go deeply into complex number arithmetic here—the main point is: if the discriminant is negative, the system spirals, and the sign of the trace still tells you whether the spiral moves inward (stable) or outward (unstable).

Classifying Equilibrium Points

The characteristic values of the coefficient matrix determine the classification of the equilibrium. The following table organizes the possibilities based on the trace and determinant—numbers you can always compute quickly from the four coefficients in your system. Notice that the sign of the trace decides stability, while the sign of the discriminant decides whether trajectories are straight or spiral.

Classification of equilibrium points by trace, determinant, and discriminant
Conditions on Trace & DetCharacteristic ValuesClassificationStability
det > 0, discriminant > 0, trace < 0Both characteristic values real and negativeStable nodeAsymptotically stable
det > 0, discriminant > 0, trace > 0Both characteristic values real and positiveUnstable nodeUnstable
det < 0Characteristic values real, opposite signsSaddle pointUnstable
det > 0, discriminant < 0, trace < 0Complex characteristic values, negative real partStable spiralAsymptotically stable
det > 0, discriminant < 0, trace > 0Complex characteristic values, positive real partUnstable spiralUnstable
det > 0, discriminant < 0, trace = 0Complex characteristic values, zero real partCenterStable (not asymptotically)
The trace-determinant plane maps stability. The region with negative trace and positive determinant corresponds to stable behavior (nodes or spirals). The region with positive trace and positive determinant is unstable. A negative determinant always means a saddle point (unstable). Points with zero trace and positive determinant produce centers with pure oscillation.

This diagram is your roadmap. Once you compute the trace and determinant and check the discriminant, you immediately know the system's long-term behavior. When the discriminant (trace² − 4·det) is positive, the characteristic values are real numbers and trajectories approach straight-line directions—giving nodes or saddle points. When the discriminant is negative, the characteristic values involve imaginary numbers and trajectories curve in spirals or closed loops around the equilibrium.

Worked Example — Classifying an Equilibrium

Consider the system dx/dt = −3x + y and dy/dt = −2y. We want to determine the equilibrium point, compute the trace and determinant, classify the equilibrium, and describe the long-term behavior.

Classifying an Equilibrium for a 2-Variable Linear System
1
Step 1 — Identify the coefficients and find the equilibriumThe system has the form dx/dt = ax + by and dy/dt = cx + dy with a = −3, b = 1, c = 0, d = −2. The equilibrium point occurs where both rates of change equal zero simultaneously: −3x + y = 0 and −2y = 0. Solving these gives y = 0 and then x = 0, so the equilibrium is at (x, y) = (0, 0).
a = −3, b = 1, c = 0, d = −2; equilibrium at the origin (0, 0)
2
Step 2 — Compute the trace and determinantTrace = a + d = (−3) + (−2) = −5. Determinant = ad − bc = (−3)(−2) − (1)(0) = 6 − 0 = 6. Since the trace is negative and the determinant is positive, we can already predict the system is asymptotically stable using the stability shortcut.
trace = −5, det = 6
3
Step 3 — Find the characteristic valuesCompute the discriminant: (trace)² − 4·det = (−5)² − 4(6) = 25 − 24 = 1 > 0. Because the discriminant is positive, the characteristic values are two distinct real numbers. Solve the characteristic equation λ² − (trace)λ + det = 0, which becomes λ² + 5λ + 6 = 0. Factoring: (λ + 3)(λ + 2) = 0, so λ₁ = −3 and λ₂ = −2.
discriminant = 1 > 0; λ₁ = −3, λ₂ = −2 (both real)
4
Step 4 — Classify the equilibriumBoth characteristic values are real and negative, and the discriminant is positive. According to our classification table, this makes the origin a stable node (also called a sink). All trajectories approach the origin as t → ∞ along straight-line directions.
Stable node — asymptotically stable
5
Step 5 — Describe long-term behaviorBecause |λ₁| = 3 > |λ₂| = 2, the x-component decays faster than the y-component. Trajectories initially move toward the origin quickly in the x-direction, then slowly approach along the y-direction. In the long run, every solution decays to (0, 0).
All solutions → (0, 0) as t → ∞

Strengths & Limitations of Linear Stability Analysis

The trace-determinant method we've learned is powerful, but it has boundaries. Understanding when it works—and when it doesn't—is essential for applying it correctly.

Strengths versus limitations of trace-determinant stability analysis
StrengthsLimitations
Gives exact stability classification for any 2-variable constant-coefficient linear systemOnly directly applies to linear systems; nonlinear systems (where terms like x² or xy appear) require additional approximation steps near each equilibrium
The trace-determinant shortcut allows fast checking with only basic arithmeticWhen the trace equals zero exactly (centers), the method cannot determine stability of related nonlinear systems without additional analysis
Extends naturally to larger systems using the same underlying principleDoes not reveal how far from the equilibrium the stability prediction holds—it only describes local behavior close to the equilibrium
Connects to phase portraits, giving geometric intuition alongside algebraCannot analyze systems where the coefficients themselves change over time
KEY TAKEAWAY
Think of linear stability analysis like using a magnifying glass near an equilibrium point. It gives you a perfectly clear view of local behavior—what happens right around that point. But it can't tell you what's happening far away from the equilibrium, just as a magnifying glass can't show you the entire landscape. For the bigger picture, mathematicians and engineers use additional tools such as energy-based analysis or computer simulation.

Connection to Nonlinear Systems & Advanced Theory

Most real-world systems are nonlinear—meaning the equations contain terms like x², xy, or sin(x) rather than just constants times x and y. Fortunately, a powerful idea called linearization lets us apply everything we've learned to nonlinear systems—at least close to an equilibrium. The idea is that any smooth curve looks nearly straight when you zoom in close enough. Similarly, a nonlinear system behaves approximately like a simpler linear system right near an equilibrium point. We find that linear approximation by computing rates of change (using calculus) at the equilibrium, then apply our trace-determinant method to the resulting linear system.

Linear vs. nonlinear stability analysis
FeatureLinear Systems (this lesson)Nonlinear Systems (advanced)
Number of equilibriaUsually one (the origin)Can have many; must find and classify each separately
Stability methodTrace and determinant of the coefficient grid directlyTrace and determinant of the linearized (zoomed-in) version at each equilibrium
Global behaviorFully described by trace and determinantMay include closed repeating loops (limit cycles), chaotic behavior, and sudden changes in stability as parameters vary
Borderline cases (center)Conclusive: pure oscillationInconclusive: the nonlinear terms can tip the balance to stable spiraling in, unstable spiraling out, or true oscillation

In more advanced courses, you'll encounter Lyapunov's direct method, which uses energy-like functions to prove stability without solving the equations at all—similar to showing a ball will stay in a valley by showing its energy can only decrease. You'll also study how stability changes when a parameter in the system is gradually varied, such as slowly increasing the gain in an audio amplifier until feedback causes it to squeal. These powerful ideas all grow from the algebraic foundations laid in this lesson.

Practice Problems

PROBLEM 1CONCEPTUAL
In your own words, explain the difference between stable and asymptotically stable. Give a physical example of each.
PROBLEM 2BASIC CALCULATION
Consider the system dx/dt = −x and dy/dt = −4y. Compute the trace and determinant of the coefficient grid, use the stability shortcut to determine stability, and classify the type of equilibrium at the origin.
PROBLEM 3INTERMEDIATE
For the system dx/dt = 2x − y and dy/dt = x + 4y, compute the trace and determinant, use the stability shortcut to determine whether the equilibrium is stable or unstable, and identify the equilibrium type.
PROBLEM 4APPLIED
A simplified model of two competing species has populations x and y governed by dx/dt = −2x + y and dy/dt = x − 2y. Use the trace-determinant method to classify the equilibrium at the origin and explain what the model predicts about both populations over time.
PROBLEM 5CRITICAL THINKING
Suppose a system has equations dx/dt = −βy and dy/dt = βx, where β is a positive constant. Compute the trace and determinant, classify the equilibrium, and explain why this case is special. What happens to the classification if we change the system to dx/dt = −εx − βy and dy/dt = βx − εy for a small positive constant ε?

Lesson Summary

A system of differential equations models how coupled quantities change over time. Its equilibrium points are the states where everything is at rest (all derivatives equal zero). The long-term behavior of solutions near an equilibrium is determined by two simple numbers computed from the coefficients: the trace (sum of the main-diagonal coefficients) and the determinant (ad − bc). If the trace is negative and the determinant is positive, the equilibrium is asymptotically stable (a stable node or stable spiral). If the trace is positive or the determinant is negative, the equilibrium is unstable. A zero trace with positive determinant produces a center with perpetual oscillation.

The discriminant of the characteristic equation (trace² − 4·det) tells us whether trajectories approach in straight lines (discriminant ≥ 0, giving nodes) or in spirals (discriminant < 0). Phase portraits give geometric insight into how trajectories flow near equilibria—whether they converge straight in (nodes), spiral in (spirals), orbit endlessly (centers), or escape along special directions (saddle points). This framework extends to nonlinear systems through linearization—approximating a nonlinear system by a simpler linear one near each equilibrium—making these ideas among the most widely used tools in applied mathematics and engineering.

Varsity Tutors • Differential Equations • Stability in Systems — Stability and Long-Term Behavior in Systems