01 / 21

Stability in
Differential Equation Systems

Understand how equilibrium points govern long-term behavior, and learn to classify them using eigenvalues and phase portraits.

Differential Equations

What Is an Equilibrium Point?

An equilibrium point (also called a fixed point or critical point) of a system \( \mathbf{x}' = \mathbf{f}(\mathbf{x}) \) is a point \( \mathbf{x}^* \) where the system does not change:

Think of a ball on a landscape: at the bottom of a valley or at the top of a hill, the ball is in equilibrium. The question is: what happens if we nudge it slightly?

Stable vs Unstable: The Big Idea

Stable Equilibrium
Nearby solutions are attracted toward \( \mathbf{x}^* \). Like a ball at the bottom of a bowl: nudge it and it rolls back.
Unstable Equilibrium
Nearby solutions are repelled from \( \mathbf{x}^* \). Like a ball balanced on top of a hill: any nudge sends it away.

Visualizing Stability: Phase Portraits

Stable Node Trajectories converge Unstable Node Trajectories diverge

A phase portrait shows how trajectories behave near equilibria. Stable nodes attract all nearby paths; unstable nodes repel them.

The Eigenvalue Classification

For a linear system \( \mathbf{x}' = A\mathbf{x} \), the stability of the origin is determined entirely by the eigenvalues \( \lambda_1, \lambda_2 \) of the matrix \( A \).

This works because solutions contain terms like \( e^{\lambda t} \). When \( \operatorname{Re}(\lambda) < 0 \), those terms decay. When \( \operatorname{Re}(\lambda) > 0 \), they grow.

Eigenvalues in the Complex Plane

Re Im STABLE \( \operatorname{Re}(\lambda) < 0 \) UNSTABLE \( \operatorname{Re}(\lambda) > 0 \) \( \lambda = -2 \) \( -1 + 3i \) \( -1 - 3i \) \( \lambda = +2 \) \( \operatorname{Re}(\lambda) = 0 \): marginal / center

Full Classification of Equilibria

Eigenvalues Type Stability
\( \lambda_1, \lambda_2 < 0 \) (real) Stable node Asymp. stable
\( \lambda_1, \lambda_2 > 0 \) (real) Unstable node Unstable
\( \lambda_1 < 0 < \lambda_2 \) (real) Saddle point Unstable
\( \alpha \pm \beta i,\ \alpha < 0 \) Stable spiral Asymp. stable
\( \alpha \pm \beta i,\ \alpha > 0 \) Unstable spiral Unstable
\( \pm \beta i \) (purely imaginary) Center Stable (not asymp.)

Asymptotically Stable vs Stable

Asymptotically Stable
Solutions near \( \mathbf{x}^* \) not only stay close, they actually converge to \( \mathbf{x}^* \) as \( t \to \infty \). Requires \( \operatorname{Re}(\lambda_i) < 0 \) for all \( i \).
Stable (Lyapunov)
Solutions that start near \( \mathbf{x}^* \) stay near it, but may never converge. Centers (purely imaginary eigenvalues) exhibit this: orbits circle forever.

The Saddle Point: Mixed Eigenvalues

Stable manifold Unstable manifold Saddle: \( \lambda_1 < 0 < \lambda_2 \)

When eigenvalues have opposite signs, the equilibrium is a saddle point. Solutions approach along the stable manifold but are repelled along the unstable manifold. Overall: unstable.

How to Classify an Equilibrium

1
Find equilibria. Solve \( \mathbf{f}(\mathbf{x}^*) = \mathbf{0} \) to locate all equilibrium points.
2
Linearize (if nonlinear). Compute the Jacobian \( J = \frac{\partial \mathbf{f}}{\partial \mathbf{x}} \) evaluated at \( \mathbf{x}^* \).
3
Find eigenvalues. Solve \( \det(J - \lambda I) = 0 \) to get \( \lambda_1, \lambda_2 \).
4
Check real parts. All \( \operatorname{Re}(\lambda_i) < 0 \)? Stable. Any \( \operatorname{Re}(\lambda_i) > 0 \)? Unstable.
5
Classify the type. Real vs complex, same sign vs opposite: node, spiral, saddle, or center.

Example 1: Linear System

1
Eigenvalues: The matrix is diagonal, so \( \lambda_1 = -3 \) and \( \lambda_2 = -1 \).
2
Real parts: Both \( \operatorname{Re}(\lambda_1) = -3 < 0 \) and \( \operatorname{Re}(\lambda_2) = -1 < 0 \).
3
Classification: Both real and negative: stable node. The origin is asymptotically stable.

Example 2: Complex Eigenvalues

1
Characteristic equation: \( \lambda^2 + 2\lambda + 5 = 0 \).
2
Eigenvalues: \( \lambda = \frac{-2 \pm \sqrt{4 - 20}}{2} = -1 \pm 2i \).
3
Real part: \( \alpha = -1 < 0 \) with \( \beta = 2 \neq 0 \). This is a stable spiral: trajectories spiral inward.
Stable spiral: \( \lambda = -1 \pm 2i \)

Example 3: Nonlinear System

1
Find equilibria: Set \( x(1 - y) = 0 \) and \( y(-1 + x) = 0 \). Two equilibria: \( (0, 0) \) and \( (1, 1) \).
2
Jacobian: \( J = \begin{pmatrix} 1 - y & -x \\ y & -1 + x \end{pmatrix} \).
3
At \( (0,0) \): \( J = \begin{pmatrix} 1 & 0 \\ 0 & -1 \end{pmatrix} \), eigenvalues \( \lambda = 1, -1 \). Saddle point: unstable.
4
At \( (1,1) \): \( J = \begin{pmatrix} 0 & -1 \\ 1 & 0 \end{pmatrix} \), eigenvalues \( \lambda = \pm i \). Center: stable but not asymptotically stable.

Shortcut: The Trace-Determinant Method

For a \( 2 \times 2 \) matrix \( A \), you can classify the equilibrium using \( \tau = \operatorname{tr}(A) \) and \( \Delta = \det(A) \) without computing eigenvalues directly:

\( \tau = \operatorname{tr}(A) \)
Sum of diagonal entries: \( a_{11} + a_{22} \). Controls decay vs growth.
\( \Delta = \det(A) \)
Product minus cross: \( a_{11}a_{22} - a_{12}a_{21} \). Controls node vs saddle.

Common Mistakes to Avoid

✕ Forgetting to Linearize
Applying the eigenvalue test directly to a nonlinear system without computing the Jacobian first. The eigenvalue method applies to the linearized system at each equilibrium.
✕ Confusing Stable with Asymptotically Stable
A center is stable (bounded orbits) but not asymptotically stable (solutions don't converge). These are different properties.
✕ One Negative Eigenvalue = Stable
If \( \lambda_1 < 0 \) but \( \lambda_2 > 0 \), it is a saddle, which is unstable. All eigenvalues must have negative real parts for stability.

Summary of Stability Criteria

Criterion Trace-Det Eigenvalue
Stable node \( \tau < 0,\;\Delta > 0,\;\tau^2 > 4\Delta \) \( \lambda_1, \lambda_2 < 0 \) (real)
Stable spiral \( \tau < 0,\;\Delta > 0,\;\tau^2 < 4\Delta \) \( \alpha < 0,\;\beta \neq 0 \)
Center \( \tau = 0,\;\Delta > 0 \) \( \lambda = \pm \beta i \)
Unstable node \( \tau > 0,\;\Delta > 0,\;\tau^2 > 4\Delta \) \( \lambda_1, \lambda_2 > 0 \) (real)
Unstable spiral \( \tau > 0,\;\Delta > 0,\;\tau^2 < 4\Delta \) \( \alpha > 0,\;\beta \neq 0 \)
Saddle \( \Delta < 0 \) \( \lambda_1 < 0 < \lambda_2 \)

Question 1 of 5

The eigenvalues of matrix \( A \) are \( \lambda = -3 \pm 4i \). What type of equilibrium does the origin have?

Question 2 of 5

If \( A = \begin{pmatrix} 2 & 0 \\ 0 & 5 \end{pmatrix} \), the origin is:

Question 3 of 5

A system has \( \operatorname{tr}(A) = 0 \) and \( \det(A) = 9 \). The equilibrium is:

Question 4 of 5

Which condition on the Jacobian's eigenvalues guarantees that a nonlinear system's equilibrium is asymptotically stable?

Question 5 of 5

Given eigenvalues \( \lambda_1 = -2 \) and \( \lambda_2 = 3 \), the equilibrium is a: