Linear Algebra › Section B · Matrix algebra › Topic B.2
The inverse of a matrix
To solve 5x = 15 you multiply both sides by 1/5. Some square matrices have a partner that undoes them in the same way, and finding that partner turns out to be ordinary row reduction in disguise.
Undoing a matrix
With numbers, every non-zero \(a\) has a reciprocal \(a^{-1}\) with \(a^{-1}a = 1\), and that is how \(ax = b\) is solved. For matrices the role of 1 is played by the identity \(I\), and the question becomes: is there a matrix that undoes \(A\), so that applying \(A\) and then it returns every vector unchanged? Because matrix multiplication is not commutative, the partner has to work on both sides. And, as with the number 0, many non-zero matrices have no such partner at all.
Definition and uniqueness
An \(n \times n\) matrix \(A\) is invertible if there is an \(n \times n\) matrix \(C\) with
Such a \(C\) is unique: if \(B\) also works, then \(B = BI = B(AC) = (BA)C = IC = C\). So we may call it the inverse and write \(A^{-1}\). A matrix with no inverse is called singular, and an invertible one is also called nonsingular. Only square matrices are invertible in this sense, and a square matrix with a row or column of zeros is always singular.
The 2 × 2 formula
For a \(2 \times 2\) matrix there is a shortcut:
Swap the diagonal entries, change the signs of the other two, and divide by \(ad - bc\). That number is the determinant, \(\det A\), and the formula works exactly when \(ad - bc \ne 0\). If \(ad - bc = 0\), \(A\) is singular. Determinants of larger matrices come later, in "Determinants and cofactor expansion".
Solving a system with the inverse
If \(A\) is invertible, then for every \(\mathbf{b}\) the equation \(A\mathbf{x} = \mathbf{b}\) has exactly one solution, \(\mathbf{x} = A^{-1}\mathbf{b}\). It is a solution, since \(A(A^{-1}\mathbf{b}) = \mathbf{b}\). It is the only one, since any solution \(\mathbf{u}\) gives \(\mathbf{u} = A^{-1}A\mathbf{u} = A^{-1}\mathbf{b}\). This is neat for theory and for small systems. For large systems, computing \(A^{-1}\) is slower and less accurate than row reducing \([A \ \mathbf{b}]\), so software rarely forms the inverse.
Worked example: a 2 × 2 system
Solve \(4x_1 + 3x_2 = 5\) and \(2x_1 + 2x_2 = 2\). 1. The coefficient matrix has \(a = 4\), \(b = 3\), \(c = 2\), \(d = 2\), so \(ad - bc = 8 - 6 = 2\), which is non-zero. 2. Apply the formula:
3. Multiply: \(x_1 = 1(5) - 1.5(2) = 2\) and \(x_2 = -1(5) + 2(2) = -1\). 4. Check in the original equations: \(8 - 3 = 5\) and \(4 - 2 = 2\). Both hold, so \(\mathbf{x} = (2, -1)\).
Rules for inverses
If \(A\) and \(B\) are invertible \(n \times n\) matrices, then: \((A^{-1})^{-1} = A\), since \(A\) undoes \(A^{-1}\); \((AB)^{-1} = B^{-1}A^{-1}\), with the order reversed; and \((A^T)^{-1} = (A^{-1})^T\). For a non-zero scalar, \((cA)^{-1} = \frac{1}{c}A^{-1}\). The product rule is easy to verify:
and the same works on the other side. Think of socks and shoes: you put socks on first, but take shoes off first. A product of any number of invertible matrices is invertible, with the inverses multiplied in reverse order.
Elementary matrices
An elementary matrix comes from doing one row operation to an identity matrix. Multiplying \(A\) on the left by it performs that same operation on \(A\). For example,
replaces row 2 of any \(3 \times 3\) matrix by row 2 minus twice row 1. Every row operation can be reversed, so every elementary matrix is invertible, and its inverse is the elementary matrix of the reverse operation: here, the same matrix with \(+2\) in place of \(-2\). A swap undoes itself, and scaling a row by \(k \ne 0\) is undone by scaling by \(1/k\).
Why row reduction finds the inverse
Suppose row operations take \(A\) to \(I\). In matrix form, \(E_k \cdots E_2E_1A = I\), so the product \(E_k \cdots E_1\) is \(A^{-1}\). Apply those same operations to \(I\) and you build exactly that product. Hence the algorithm: row reduce the \(n \times 2n\) matrix \([A \ I]\). If the left half becomes \(I\), the right half is \(A^{-1}\). If the left half cannot be reduced to \(I\), \(A\) is singular. The theorem behind this says \(A\) is invertible exactly when it is row equivalent to \(I_n\). Another view: column \(j\) of \(A^{-1}\) is the solution of \(A\mathbf{x} = \mathbf{e}_j\), the \(j\)th column of \(I\).
Worked example: a 3 × 3 inverse
Invert the matrix with rows (1, 0, 1), (2, 1, 3) and (1, 1, 3). Write each row of \([A \ I]\) as left part | right part. 1. Start: (1, 0, 1 | 1, 0, 0), (2, 1, 3 | 0, 1, 0), (1, 1, 3 | 0, 0, 1). 2. Row 2 minus 2 times row 1, and row 3 minus row 1: (0, 1, 1 | −2, 1, 0) and (0, 1, 2 | −1, 0, 1). 3. Row 3 minus row 2: (0, 0, 1 | 1, −1, 1). 4. Clear column 3 above it: row 2 minus row 3 gives (0, 1, 0 | −3, 2, −1), and row 1 minus row 3 gives (1, 0, 0 | 0, 1, −1). The left half is now \(I\), so
5. Check one entry of \(AA^{-1}\): row 2 of \(A\) with column 1 gives \(0 - 3 + 3 = 0\), as \(I\) requires.
Common mistakes to avoid
First, writing \((AB)^{-1} = A^{-1}B^{-1}\); the order must reverse. Second, forgetting to divide by \(ad - bc\) in the \(2 \times 2\) formula, or using it when \(ad - bc = 0\). Third, writing a matrix fraction such as \(B/A\): it is ambiguous, because \(A^{-1}B\) and \(BA^{-1}\) usually differ. Fourth, row reducing only part of \([A \ I]\); every operation must act on the whole row. Fifth, talking about the inverse of a non-square matrix.
Key takeaways
A square matrix A is invertible when some A⁻¹ gives A⁻¹A = I and AA⁻¹ = I; the inverse is unique, and a matrix without one is singular. For 2 × 2 matrices, swap the diagonal, negate the off-diagonal and divide by ad − bc, which must be non-zero. Invertible A gives the unique solution x = A⁻¹b of Ax = b. (AB)⁻¹ = B⁻¹A⁻¹ and (Aᵀ)⁻¹ = (A⁻¹)ᵀ. Each row operation is multiplication by an invertible elementary matrix, which is why row reducing [A I] to [I A⁻¹] finds the inverse.
Flashcards
Flashcard
A square matrix A with some A⁻¹ such that A⁻¹A = I and AA⁻¹ = I
Test your understanding
Question 1 of 5
What is the inverse of the matrix with rows (5, 2) and (7, 3)?
Question 2 of 5
A and B are invertible n × n matrices. Which expression equals (AB)⁻¹?
Question 3 of 5
For which value of k is the matrix with rows (2, k) and (4, 6) singular?
Question 4 of 5
E is the 3 × 3 identity with −4 placed in row 3, column 1. What does multiplying A on the left by E⁻¹ do?
Question 5 of 5
While row reducing [A I] for a 3 × 3 matrix A, the left half produces a row of zeros. What can you conclude?
