Systems: Real Distinct Eigenvalues — Solving Systems with...

SECTION 1

Historical Context & Motivation

📘 Prerequisites & Audience Note

This lesson covers systems of differential equations solved via eigenvalue methods — a topic typically encountered in a second-year undergraduate course in Ordinary Differential Equations, taken after Calculus II and a first course in Linear Algebra. Prerequisites include familiarity with matrices, determinants, and basic differential equations.

Imagine you have two quantities that are changing over time, but their rates of change depend on each other. For example, the population of rabbits might grow faster when there are fewer foxes, while the fox population depends on how many rabbits are available. Describing these kinds of intertwined relationships requires a system of differential equations — two or more equations that must be solved simultaneously. For centuries, mathematicians searched for systematic ways to untangle these coupled problems.

The key breakthrough came from the study of eigenvalues and eigenvectors, concepts rooted in linear algebra. These tools let us decouple a messy system into independent pieces that we already know how to solve. The development of these ideas spans several centuries and multiple brilliant mathematicians.

1743

Euler & Coupled Oscillations

Leonhard Euler studied systems of coupled oscillators and recognized that certain "principal modes" could simplify the analysis — an early glimpse of eigenvalue thinking.

1826

Cauchy Formalizes Eigenvalues

Augustin-Louis Cauchy introduced the formal theory of eigenvalues for matrices, calling them "characteristic roots" of a matrix equation.

1855

Sylvester Names the Matrix

James Joseph Sylvester coined the term "matrix," giving mathematicians a unified language for the rectangular arrays of numbers used in systems.

1904

Hilbert & Spectral Theory

David Hilbert extended eigenvalue ideas to function spaces of infinite dimension, calling the scaling factors "eigenvalues" (from the German eigen, meaning "own" or "characteristic"). This generalization underpins modern quantum mechanics and functional analysis.

The central question these mathematicians tackled is the one we address in this lesson: given a system of linear differential equations written in matrix form as x′ = Ax, how do we find the general solution? When the matrix A has real, distinct eigenvalues, the answer is beautifully clean: each eigenvalue-eigenvector pair generates one piece of the solution, and we simply add them together.

SECTION 2

Core Principles & Definitions

Before diving into the solution method, let's establish the vocabulary and ideas you'll need. A system of first-order linear differential equations can be written compactly as x′ = Ax, where A is a constant square matrix and x is a vector of unknown functions. This compact notation hides what might be two, three, or even more interrelated equations inside a single expression.

Eigenvalue (λ)

A scalar λ such that Av = λv for some nonzero vector v. Think of it as a "growth rate" that tells how fast a particular mode of the system grows or decays.

Eigenvector (v)

A nonzero vector v satisfying Av = λv. It defines a direction in which the system behaves like a simple exponential — no mixing between variables.

Characteristic Equation

The polynomial equation det(A − λI) = 0 whose roots are the eigenvalues of A. For a 2×2 matrix this is a quadratic, so you can use the quadratic formula.

Real & Distinct

"Real" means the eigenvalues are ordinary real numbers (no imaginary parts). "Distinct" means no eigenvalue is repeated. This is the simplest and most common scenario.

General Solution

The combination x(t) = c₁e^(λ₁t)v₁ + c₂e^(λ₂t)v₂, where c₁ and c₂ are arbitrary constants determined by initial conditions.

✦ KEY TAKEAWAY

Think of eigenvalues and eigenvectors like GPS directions for a system. The eigenvectors point out special straight-line paths the system can follow, and the eigenvalues tell you the speed and direction (toward or away from the origin) along each path. When eigenvalues are real and distinct, you get two independent GPS routes, and every possible trajectory is a blend of those two.

SECTION 3

Visual Explanation — Phase Portrait

The diagram below shows a phase portrait for a 2×2 system with two real, distinct, negative eigenvalues (λ₁ = −3 and λ₂ = −1). Every curve represents a possible trajectory of the system as time moves forward. Notice how every trajectory eventually spirals into the origin — that's because both eigenvalues are negative, so both modes decay.

The dashed cyan line is the eigenvector direction for λ₁ = −3 (fast decay), and the dashed pink line is the eigenvector direction for λ₂ = −1 (slow decay). Trajectories first approach the slow eigenvector direction, then slide along it toward the origin.

Notice an important pattern in the diagram: trajectories that don't start exactly on one of the dashed eigenvector lines are curved. Near the origin, every trajectory becomes nearly parallel to the slow eigenvector (the one with the eigenvalue closer to zero). This happens because the fast component dies out quickly, leaving only the slow component to dominate the long-term behavior.

SECTION 4

Mathematical Framework

Let's build the solution method step by step. We start with a 2×2 system written in matrix form. The same ideas extend to larger systems, but 2×2 is where we develop our intuition.

SYSTEM IN MATRIX FORM

x′(t) = A x(t)

Here x(t) is a column vector of unknown functions [x₁(t), x₂(t)]ᵀ, and A is a constant 2×2 coefficient matrix.

CHARACTERISTIC EQUATION

det(A − λI) = 0

Expanding this determinant for a 2×2 matrix A = [[a, b], [c, d]] gives the quadratic λ² − (a + d)λ + (ad − bc) = 0. The quantity (a + d) is the trace of A, and (ad − bc) is its determinant.

EIGENVECTOR EQUATION

(A − λI) v = 0

For each eigenvalue λ found above, solve this equation for the vector v. You'll get a free parameter, so just pick the simplest nonzero vector.

GENERAL SOLUTION

x(t) = c₁ e^(λ₁t) v₁ + c₂ e^(λ₂t) v₂

Each eigenvalue-eigenvector pair (λᵢ, vᵢ) produces one fundamental solution e^(λᵢt) vᵢ. The constants c₁ and c₂ are determined by initial conditions x(0).

💡 Why does this work?

If you substitute x(t) = e^(λt) v into x′ = Ax, the left side gives λe^(λt) v and the right side gives e^(λt)(Av). Canceling e^(λt) (which is never zero) leaves Av = λv — exactly the eigenvalue equation. So each eigenvalue-eigenvector pair automatically produces a solution!

SECTION 5

Classifying the Behavior

When both eigenvalues are real and distinct, the behavior of the system depends on the signs of the eigenvalues. The sign determines whether solutions grow, decay, or do a mix of both. The table below summarizes the three main cases.

Behavior classification for real distinct eigenvalues

Case	Eigenvalue Signs	Phase Portrait Type	Long-Term Behavior
1	λ₁ < 0 and λ₂ < 0	Stable node	All solutions decay to the origin
2	λ₁ > 0 and λ₂ > 0	Unstable node	All solutions grow away from the origin
3	λ₁ < 0 and λ₂ > 0 (or vice versa)	Saddle point	Solutions approach along one eigenvector, flee along the other

From left to right: a stable node (both eigenvalues negative), an unstable node (both positive), and a saddle point (one negative, one positive). The saddle point is unstable since most initial conditions eventually escape to infinity.

Understanding these three cases lets you predict the system's behavior before you even compute the full solution. If you know the signs of the eigenvalues, you already know whether solutions decay, grow, or exhibit the "approach then flee" pattern of a saddle. This qualitative insight is just as important as the quantitative formula.

SECTION 6

Worked Example

Let's walk through a complete example. We'll solve the system x′ = Ax where A = [[1, 2], [3, 2]], and then apply the initial condition x(0) = [6, 4]ᵀ.

Solve x′ = Ax with A = [[1, 2], [3, 2]] and x(0) = [6, 4]ᵀ

Step 1 — Write the Characteristic EquationCompute det(A − λI) = 0. We get det([[1−λ, 2], [3, 2−λ]]) = (1 − λ)(2 − λ) − (2)(3) = λ² − 3λ + 2 − 6 = λ² − 3λ − 4 = 0.

λ² − 3λ − 4 = 0

Step 2 — Find the EigenvaluesFactor the quadratic: λ² − 3λ − 4 = (λ − 4)(λ + 1) = 0. This gives two real, distinct eigenvalues.

λ₁ = 4 and λ₂ = −1

Step 3 — Find Eigenvector v₁ for λ₁ = 4Solve (A − 4I)v = 0. We get [[1−4, 2], [3, 2−4]]v = [[-3, 2], [3, -2]]v = 0. The first row gives −3v₁ + 2v₂ = 0, so v₂ = (3/2)v₁. Choosing v₁ = 2 gives v₂ = 3.

v₁ = [2, 3]ᵀ

Step 4 — Find Eigenvector v₂ for λ₂ = −1Solve (A − (−1)I)v = 0. We get [[2, 2], [3, 3]]v = 0. The first row gives 2v₁ + 2v₂ = 0, so v₂ = −v₁. Choosing v₁ = 1 gives v₂ = −1.

v₂ = [1, −1]ᵀ

Step 5 — Write the General SolutionCombine the eigenvalue-eigenvector pairs: x(t) = c₁ e^(4t) [2, 3]ᵀ + c₂ e^(−t) [1, −1]ᵀ.

x(t) = c₁ e^(4t) [2, 3]ᵀ + c₂ e^(−t) [1, −1]ᵀ

Step 6 — Apply Initial ConditionsSet t = 0: x(0) = c₁[2, 3]ᵀ + c₂[1, −1]ᵀ = [6, 4]ᵀ. This gives the system 2c₁ + c₂ = 6 and 3c₁ − c₂ = 4. Adding these equations: 5c₁ = 10, so c₁ = 2. Then c₂ = 6 − 2(2) = 2.

c₁ = 2, c₂ = 2

Step 7 — Write the Particular SolutionSubstitute c₁ = 2 and c₂ = 2 back into the general solution to get the final answer.

x(t) = 2e^(4t) [2, 3]ᵀ + 2e^(−t) [1, −1]ᵀ = [4e^(4t) + 2e^(−t), 6e^(4t) − 2e^(−t)]ᵀ

✓ Check Your Work

You can verify by differentiating x(t) and confirming it equals Ax(t). Also check that x(0) = [4(1) + 2(1), 6(1) − 2(1)]ᵀ = [6, 4]ᵀ, which matches the initial condition. Always verify — it only takes a minute and catches algebraic mistakes.

SECTION 7

Strengths & Limitations

The eigenvalue method is powerful but not universal. Understanding when it works well — and when you need a different approach — is key to becoming fluent with systems of differential equations.

Strengths and limitations of the eigenvalue method for systems

Strengths	Limitations
Gives an exact, closed-form solution — no approximations or numerical methods needed	Only works directly when A is a constant matrix (coefficients don't change over time)
Reveals the qualitative behavior (stable, unstable, saddle) from eigenvalue signs alone	Requires eigenvalues to be real and distinct — complex or repeated eigenvalues need modified techniques
Scales naturally to 3×3, 4×4, and larger systems using the same recipe	For large systems, finding eigenvalues by hand becomes impractical (use technology for n ≥ 4)
Each eigenvalue-eigenvector pair is independent, so errors in one don't cascade to others	Doesn't handle nonlinear systems directly — though eigenvalues of the linearization can still describe local behavior near an equilibrium point

✦ KEY TAKEAWAY

The real distinct eigenvalue method is like a master key that opens a specific type of lock perfectly. When you encounter complex eigenvalues (oscillatory systems) or repeated eigenvalues (degenerate systems), you'll need a slightly different key — but the underlying philosophy of finding special directions remains the same.

SECTION 8

Connection to Advanced Theory

The real distinct eigenvalue case is your launching point for more advanced topics. Once you're comfortable here, you'll extend the same core ideas — find eigenvalues, find eigenvectors, build the solution — to situations that require a few extra tools.

How this lesson connects to more advanced system types

Feature	Real Distinct Eigenvalues (This Lesson)	Advanced Extensions
Eigenvalue type	Two real, unequal numbers	Complex conjugate pairs (a ± bi) or repeated roots
Solution form	Pure exponentials: c₁e^(λ₁t)v₁ + c₂e^(λ₂t)v₂	Complex case: exponentials × sines and cosines. Repeated case: add te^(λt) terms
Phase portrait	Nodes (stable/unstable) or saddle points	Spirals and centers (complex), star nodes and improper nodes (repeated)
Matrix requirement	Constant coefficients, homogeneous	Nonhomogeneous systems add a particular solution found via undetermined coefficients or other standard methods covered in a differential equations course

In many science and engineering applications, a nonlinear system can be approximated near a steady state (called an equilibrium point) by a linear system of the form x′ = Ax, where A captures the local rates of change. The eigenvalues of that matrix A then tell you whether small disturbances grow or decay — exactly the analysis done in population ecology, circuit design, and control theory. Mastering the real distinct case gives you the conceptual foundation for all of these applications.

SECTION 9

Practice Problems

PROBLEM 1 — CONCEPTUAL

A 2×2 system x′ = Ax has eigenvalues λ₁ = −5 and λ₂ = −2. Without solving, describe the long-term behavior of every solution. What type of phase portrait does this system have?

PROBLEM 2 — BASIC CALCULATION

Find the eigenvalues of the matrix A = [[5, 1], [0, 3]]. Verify that they are real and distinct.

PROBLEM 3 — INTERMEDIATE

Solve the system x′ = Ax where A = [[3, −1], [1, 1]]. Find the eigenvalues, eigenvectors, and general solution.

PROBLEM 4 — APPLIED

Two interconnected tanks contain salt solutions. Tank 1 has x₁(t) kg of salt and Tank 2 has x₂(t) kg. The flow rates yield the system x₁′ = −3x₁ + x₂ and x₂′ = 2x₁ − 2x₂. Find the general solution using eigenvalues and eigenvectors. If initially x₁(0) = 10 kg and x₂(0) = 0 kg, find the particular solution.

PROBLEM 5 — CRITICAL THINKING

Consider a general 2×2 matrix A = [[a, b], [c, d]]. Derive a condition on a, b, c, and d that guarantees the eigenvalues are real and distinct. Express your answer in terms of the trace T = a + d and determinant D = ad − bc. Then explain geometrically what happens as the eigenvalues approach each other (i.e., as the "distinct" condition starts to fail).

SUMMARY

Lesson Summary

A system of linear differential equations x′ = Ax with constant coefficients can be solved by finding the eigenvalues and eigenvectors of the matrix A. You form the characteristic equation det(A − λI) = 0 to find eigenvalues, then solve (A − λI)v = 0 for each eigenvector. When the eigenvalues are real and distinct, the general solution is x(t) = c₁e^(λ₁t)v₁ + c₂e^(λ₂t)v₂, with constants c₁ and c₂ determined by initial conditions.

The signs of the eigenvalues reveal the system's qualitative behavior: two negative eigenvalues give a stable node, two positive give an unstable node, and mixed signs produce a saddle point. The eigenvectors define straight-line trajectories in the phase plane, and all other trajectories are combinations of these fundamental directions. This method extends naturally to complex and repeated eigenvalue cases, making it the cornerstone technique for solving systems of differential equations.