Differential Equations · Undergraduate

Systems of Linear Differential Equations

Quick answer

A system of linear differential equations, x′ = Ax, describes several quantities whose rates depend on each other. Trying x = e^(λt)v turns it into the eigenvalue problem Av = λv. Each eigenvalue and eigenvector gives a solution moving straight along the eigenvector, growing if λ > 0 and decaying if λ < 0, and every solution is a combination of these. Complex eigenvalues give spirals or circles. Any second-order equation can be rewritten as such a system.

What you'll learn

  • Write a system of linear equations in the form x′ = Ax
  • Solve x′ = Ax with eigenvalues and eigenvectors
  • Fit the constants to an initial condition
  • Classify the origin as a saddle, node or spiral from the eigenvalues

Rates that depend on each other

Two quantities x1x_1 and x2x_2 change at rates that depend on both:

x1′=x1+2x2x2′=2x1+x2x_1' = x_1 + 2x_2 \qquad x_2' = 2x_1 + x_2

In matrix form this is x′=Ax\mathbf{x}' = A\mathbf{x} with A=[1221]A = \begin{bmatrix} 1 & 2\\ 2 & 1 \end{bmatrix}. One equation for xx had the solution eate^{at}, so try x=eλtv\mathbf{x} = e^{\lambda t}\mathbf{v}:

x′=λeλtvAx=eλtAv\mathbf{x}' = \lambda e^{\lambda t}\mathbf{v} \qquad A\mathbf{x} = e^{\lambda t}A\mathbf{v}

These match exactly when Av=λvA\mathbf{v} = \lambda\mathbf{v}: λ\lambda must be an eigenvalue and v\mathbf{v} an eigenvector.

For our AA, the eigenvalues are 33 and −1-1, with eigenvectors (1,1)(1, 1) and (1,−1)(1, -1). Since the equation is linear, combinations of solutions are solutions:

x(t)=c1e3t[11]+c2e−t[1−1]\mathbf{x}(t) = c_1e^{3t}\begin{bmatrix} 1\\ 1 \end{bmatrix} + c_2e^{-t}\begin{bmatrix} 1\\ -1 \end{bmatrix}

Why eigenvectors are the natural directions

Along an eigenvector, AA only stretches: it pushes a point straight out along the same line, or straight back toward the origin. A solution that starts on that line stays on it, growing or shrinking like eλte^{\lambda t}. In coordinates along the eigenvectors, the system splits into separate one-variable equations yi′=λiyiy_i' = \lambda_iy_i, which is diagonalization applied to a differential equation. Every other solution is a blend of these straight-line motions.

The phase portrait of x′ = Ax for A = [[1, 2], [2, 1]] Arrows showing the direction of motion at each point of the plane, with the two eigenvector lines dashed. Along y = x the arrows point outward, away from the origin, because the eigenvalue 3 is positive. Along y = −x they point inward, because the eigenvalue −1 is negative. Elsewhere the arrows curve from the inward line toward the outward one, the pattern of a saddle. −3−2−1123−3−2−1123xy
  • eigenvector lines
The phase portrait of x′ = Ax for A = [[1, 2], [2, 1]]

Worked examples

Common mistakes

Practice problems

  1. Solve x′=[300−1]x\mathbf{x}' = \begin{bmatrix} 3 & 0\\ 0 & -1 \end{bmatrix}\mathbf{x}.

    Answer

    x1=c1e3tx_1 = c_1e^{3t}, x2=c2e−tx_2 = c_2e^{-t}

    Full solution

    The matrix is diagonal, so the equations do not interact.

  2. Solve x′=[4123]x\mathbf{x}' = \begin{bmatrix} 4 & 1\\ 2 & 3 \end{bmatrix}\mathbf{x}.

    Answer

    x=c1e5t(1,1)+c2e2t(1,−2)\mathbf{x} = c_1e^{5t}(1, 1) + c_2e^{2t}(1, -2)

    Full solution

    Trace 77, determinant 1010: λ=5,2\lambda = 5, 2. For 55: (1,1)(1, 1). For 22: (1,−2)(1, -2).

  3. Solve x′=[01−10]x\mathbf{x}' = \begin{bmatrix} 0 & 1\\ -1 & 0 \end{bmatrix}\mathbf{x} with x(0)=(1,0)\mathbf{x}(0) = (1, 0).

    Answer

    x=(cos⁡t, −sin⁡t)\mathbf{x} = (\cos t,\ -\sin t)

    Full solution

    x1′=x2x_1' = x_2 and x2′=−x1x_2' = -x_1, so x1′′=−x1x_1'' = -x_1 and x1=cos⁡tx_1 = \cos t, x2=x1′=−sin⁡tx_2 = x_1' = -\sin t. The point circles clockwise.

  4. Find the eigenvalues and eigenvectors of [1102]\begin{bmatrix} 1 & 1\\ 0 & 2 \end{bmatrix} and write the general solution of x′=Ax\mathbf{x}' = A\mathbf{x}.

    Answer

    x=c1et(1,0)+c2e2t(1,1)\mathbf{x} = c_1e^t(1, 0) + c_2e^{2t}(1, 1)

    Full solution

    The matrix is triangular, so λ=1,2\lambda = 1, 2. For 22: A−2IA - 2I has first row (−1,1)(-1, 1), giving (1,1)(1, 1).

  5. For Example 2, find the solution with x(0)=(1,3)\mathbf{x}(0) = (1, 3).

    Answer

    x=2e−t(1,1)−e−3t(1,−1)\mathbf{x} = 2e^{-t}(1, 1) - e^{-3t}(1, -1)

    Full solution

    c1+c2=1c_1 + c_2 = 1 and c1−c2=3c_1 - c_2 = 3 give c1=2c_1 = 2, c2=−1c_2 = -1.

  6. Write y′′−5y′+6y=0y'' - 5y' + 6y = 0 as a system x′=Ax\mathbf{x}' = A\mathbf{x}.

    Answer

    A=[01−65]A = \begin{bmatrix} 0 & 1\\ -6 & 5 \end{bmatrix}

    Full solution

    x1=yx_1 = y, x2=y′x_2 = y', and y′′=−6y+5y′y'' = -6y + 5y'. Its eigenvalues are 22 and 33, the characteristic roots.

  7. Is the origin stable for a system whose matrix has eigenvalues −1+4i-1 + 4i and −1−4i-1 - 4i?

    Answer

    Yes; solutions spiral in.

    Full solution

    The real part −1-1 is negative, so the size decays like e−te^{-t} while the imaginary part makes the path rotate.

  8. A student writes the solution of the first system as x=c1e3t+c2e−t\mathbf{x} = c_1e^{3t} + c_2e^{-t}. What is missing?

    Answer

    The eigenvectors: x=c1e3t(1,1)+c2e−t(1,−1)\mathbf{x} = c_1e^{3t}(1, 1) + c_2e^{-t}(1, -1).

    Full solution

    x\mathbf{x} is a vector, so each exponential must multiply a direction. Without them, x1x_1 and x2x_2 would be forced to be equal.

Frequently asked questions

How do you solve x′ = Ax?

Find the eigenvalues λ and eigenvectors v of A. Each pair gives a solution e^(λt)v, and for an n × n matrix with n independent eigenvectors, the general solution is c₁e^(λ₁t)v₁ + … + cₙe^(λₙt)vₙ.

Why does e^(λt)v solve the system?

Its derivative is λe^(λt)v, and A times it is e^(λt)Av. These are equal exactly when Av = λv.

What does the phase portrait show?

The paths that solutions trace in the plane of (x₁, x₂). Straight-line paths run along the eigenvectors, and the eigenvalues decide whether paths approach or leave the origin.

When is the origin stable?

When every eigenvalue has negative real part. Then every solution approaches the origin as t grows.

How do you turn a second-order equation into a system?

Set x₁ = y and x₂ = y′. Then x₁′ = x₂, and the equation gives x₂′ in terms of x₁ and x₂.

What to learn next