Differential Equations › Section E · Systems of differential equations › Topic E.1
Review: matrices, vectors, and eigenvalues
Every method in this section turns a system of differential equations into a question about a matrix. This review gathers the linear algebra you will lean on: solving linear systems, matrix arithmetic, determinants, inverses, independence and, above all, eigenvalues and eigenvectors.
Why matrices, and why now
A single equation such as \(y' = ky\) is solved by one exponential. The equations in this section track several quantities at once, each changing at a rate that depends on all of them. Written with vectors and matrices, such a system becomes \(\mathbf{x}' = A\mathbf{x}\), and solving it comes down to understanding the matrix \(A\). This topic reviews the tools in the order you will use them. The linear algebra course covers each in depth, in topics such as "Matrix operations", "Eigenvectors and eigenvalues" and "The characteristic equation".
Solving linear systems by row reduction
A linear system is stored as an augmented matrix: one row per equation, holding its coefficients and then its right-hand side. Three row operations never change the solutions: swap two rows, multiply a row by a non-zero number, and add a multiple of one row to another. Use them to reach triangular form, with zeros below the main diagonal, then back substitute from the bottom row up. Every linear system ends in one of three ways: exactly one solution, no solution (some row reads 0 = a non-zero number), or infinitely many solutions (a row of zeros leaves a free variable).
Worked example: a three-by-three system
Solve \(x_1 - 2x_2 + x_3 = 7\), \(2x_1 + x_2 - x_3 = 0\) and \(-x_1 + 3x_2 + 2x_3 = 1\). 1. The augmented rows are (1, -2, 1 | 7), (2, 1, -1 | 0) and (-1, 3, 2 | 1). 2. Subtract 2 times row 1 from row 2 and add row 1 to row 3, giving (0, 5, -3 | -14) and (0, 1, 3 | 8). 3. Swap those two rows, then subtract 5 times the new row 2 from row 3: (0, 0, -18 | -54). 4. Back substitute: \(x_3 = 3\), then \(x_2 = 8 - 3x_3 = -1\), then \(x_1 = 7 + 2x_2 - x_3 = 2\). 5. Check in the third equation: \(-2 - 3 + 6 = 1\).
Matrix arithmetic
Matrices of the same size add entry by entry, and a scalar multiplies every entry. The product \(AB\) needs \(A\) to have as many columns as \(B\) has rows; its entry in row \(i\), column \(j\) pairs row \(i\) of \(A\) with column \(j\) of \(B\), multiplying and adding. Usually \(AB \ne BA\). The identity matrix \(I\) has 1s on the main diagonal and 0s elsewhere, so \(AI = IA = A\). A matrix times a column vector gives a column vector:
This is exactly the operation on the right side of \(\mathbf{x}' = A\mathbf{x}\).
Determinants, inverses and independence
For a 2 × 2 matrix with rows (a, b) and (c, d), \(\det A = ad - bc\); larger determinants come from cofactor expansion, as in "Determinants and cofactor expansion". A matrix with determinant 0 is singular. A nonsingular matrix has an inverse, which in the 2 × 2 case is
and then \(A\mathbf{x} = \mathbf{b}\) has exactly one solution, \(A^{-1}\mathbf{b}\). A singular \(A\) gives none or infinitely many; in particular, \(A\mathbf{x} = \mathbf{0}\) has non-zero solutions exactly when \(\det A = 0\). Likewise, \(n\) vectors in \(\mathbb{R}^n\) are linearly independent exactly when the matrix with them as columns has a non-zero determinant: (1, 2) and (3, 6) give \(6 - 6 = 0\), so they are dependent.
Eigenvalues and eigenvectors
An eigenvector of a square matrix \(A\) is a non-zero vector \(\mathbf{v}\) with \(A\mathbf{v} = \lambda\mathbf{v}\); the number \(\lambda\) is its eigenvalue. Rewrite this as \((A - \lambda I)\mathbf{v} = \mathbf{0}\). By the previous card, a non-zero solution exists exactly when
which is the characteristic equation. For a 2 × 2 matrix it reads \(\lambda^2 - (\operatorname{tr} A)\lambda + \det A = 0\), where the trace \(\operatorname{tr} A\) is the sum of the diagonal entries. Then solve \((A - \lambda I)\mathbf{v} = \mathbf{0}\) for each \(\lambda\). Any non-zero multiple of an eigenvector is again one, so choose a convenient one.
Worked example: two real eigenvalues
Find the eigenvalues and eigenvectors of
1. The trace is 5 and the determinant is \(4 + 2 = 6\), so \(\lambda^2 - 5\lambda + 6 = 0\), giving \(\lambda = 2\) and \(\lambda = 3\). 2. For \(\lambda = 2\), \(A - 2I\) has rows (2, -2) and (1, -1), so \(v_1 = v_2\) and \(\mathbf{v} = (1, 1)\). 3. For \(\lambda = 3\), \(A - 3I\) has rows (1, -2) and (1, -2), so \(v_1 = 2v_2\) and \(\mathbf{v} = (2, 1)\). 4. Check: \(A(2, 1) = (6, 3) = 3(2, 1)\). If the rows of \(A - \lambda I\) are not proportional, the eigenvalue is wrong.
Multiplicity and complex eigenvalues
An eigenvalue that appears once as a root is simple; one that appears \(k\) times has multiplicity \(k\). Eigenvectors for different eigenvalues are always linearly independent, but an eigenvalue of multiplicity \(k\) has anywhere from 1 to \(k\) independent eigenvectors. The matrix with rows (3, 1) and (-1, 1) has characteristic equation \((\lambda - 2)^2 = 0\), yet \(A - 2I\) has rows (1, 1) and (-1, -1), so every eigenvector is a multiple of (1, -1). When \(A\) is real, complex eigenvalues come in conjugate pairs \(\alpha \pm \beta i\), and conjugating an eigenvector for one gives an eigenvector for the other.
Worked example: complex eigenvalues
Find the eigenvalues and eigenvectors of
1. The trace is 4 and the determinant is \(3 + 2 = 5\), so \(\lambda^2 - 4\lambda + 5 = 0\). 2. The quadratic formula gives \(\lambda = \frac{4 \pm \sqrt{-4}}{2} = 2 \pm i\). 3. For \(\lambda = 2 + i\), the first row of \(B - \lambda I\) is \((-1 - i, -2)\). Taking \(v_1 = 2\) forces \(v_2 = -1 - i\), so \(\mathbf{v} = (2, -1 - i)\). 4. Check with the second row, (1, 1 - i): \(2 + (1 - i)(-1 - i) = 2 - 2 = 0\). 5. Conjugating, \(\lambda = 2 - i\) has eigenvector \((2, -1 + i)\).
Common mistakes to avoid
First, subtracting \(\lambda\) from every entry instead of only the diagonal entries when forming \(A - \lambda I\). Second, getting the sign of the middle term wrong: the 2 × 2 equation is \(\lambda^2 - (\operatorname{tr} A)\lambda + \det A = 0\). Third, accepting the zero vector as an eigenvector; it satisfies \(A\mathbf{0} = \lambda\mathbf{0}\) for every \(\lambda\), so the definition excludes it. Fourth, assuming a repeated eigenvalue always brings two independent eigenvectors. Fifth, reading eigenvalues off a row-reduced form of \(A\): row operations change them.
Key takeaways
Row reduction solves linear systems, which have one, none or infinitely many solutions. A square matrix is invertible exactly when its determinant is non-zero; when it is zero, \(A\mathbf{x} = \mathbf{0}\) has non-zero solutions. Eigenvalues solve \(\det(A - \lambda I) = 0\), which for 2 × 2 matrices is \(\lambda^2 - (\operatorname{tr} A)\lambda + \det A = 0\), and eigenvectors are the non-zero solutions of \((A - \lambda I)\mathbf{v} = \mathbf{0}\). Distinct eigenvalues give independent eigenvectors; repeated ones may not; real matrices have complex eigenvalues in conjugate pairs.
Flashcards
Flashcard
The coefficients and right-hand sides of a linear system in one array
Test your understanding
Question 1 of 5
Row reduction of a linear system produces the row (0, 0, 0 | 5). What does this tell you?
Question 2 of 5
What is the determinant of the matrix with rows (3, -2) and (4, 1)?
Question 3 of 5
For the matrix with rows (2, 1) and (4, -1), which equation gives the eigenvalues?
Question 4 of 5
Which vector is an eigenvector of the matrix with rows (5, -4) and (2, -1) for the eigenvalue 3?
Question 5 of 5
A real 2 × 2 matrix has eigenvalue 3 + 2i with eigenvector (1, 1 − i). What else must be true?
