Inverse and Determinant

Reversing a Transformation

In the previous lesson we saw that every matrix represents a linear transformation. A natural question arises: can that transformation be reversed? If A maps vector x to Ax, is there a matrix that maps Ax back to x?

This reverse operation — if it exists — is the matrix inverse, written A⁻¹. It satisfies A⁻¹A = AA⁻¹ = I, where I is the identity matrix (the transformation that maps every vector to itself). Applying A and then A⁻¹ (or vice versa) returns you to where you started.

Definition of the Inverse

A^{-1}A = AA^{-1} = I

A is invertible (also called non-singular) if and only if there exists a matrix A⁻¹ such that both products equal the identity. Not every square matrix has an inverse — only those that don't "collapse" space.

When Does the Inverse Exist?

The inverse exists precisely when the matrix represents a reversible transformation — one that maps distinct inputs to distinct outputs, losing no information. Three equivalent conditions must all hold:

Square: The matrix must be n × n. Only square matrices can be invertible.
Full rank: The rank of A must equal n (all rows and columns are linearly independent).
Non-zero determinant: det(A) ≠ 0. The determinant is zero if and only if the transformation collapses space into a lower-dimensional subspace.

These three conditions are equivalent: any one of them implies the other two. A matrix that fails these conditions is called singular — it has no inverse.

Geometric Intuition

Think of a linear transformation as a machine. If you put in box A and the machine squashes it flat, you cannot recover the original box from the flat sheet. A singular matrix is exactly this kind of squashing machine — it maps multiple inputs to the same output, so you can't tell which input produced which output. An invertible matrix is a machine that simply rearranges, rotates, and stretches — reversible operations.

Computing the 2×2 Inverse

For a 2×2 matrix, the inverse formula is compact and memorable. Given matrix A with det(A) ≠ 0:

2×2 Inverse Formula

A^{-1}=\frac{1}{ad-bc}\begin{pmatrix}d&-b\\-c&a\end{pmatrix}

The recipe: swap the diagonal entries (a and d), negate the off-diagonal entries (b and c), then divide everything by the determinant ad − bc. This formula breaks down when ad − bc = 0, confirming that singular matrices have no inverse.

For example, if A = [[3, 1], [2, 1]], then det(A) = 3·1 − 1·2 = 1, so A⁻¹ = [[1, −1], [−2, 3]]. You can verify: multiply A by A⁻¹ and you get the 2×2 identity matrix.

Computing the Inverse via Row Reduction

For larger matrices, the row reduction method is systematic. Form the augmented matrix [A | I] by placing the identity matrix I alongside A. Then apply Gaussian elimination to transform A into I. Whatever operations you perform on A, you simultaneously perform on I — and I transforms into A⁻¹.

The steps:

Write the augmented matrix [A | I_n]
Use elementary row operations to reduce the left half to reduced row echelon form
If the left half becomes I, the right half is now A⁻¹
If the left half gets a row of zeros, A is singular (no inverse exists)

Properties of the Inverse

The matrix inverse obeys several important algebraic rules:

Product Rule

(AB)⁻¹ = B⁻¹A⁻¹

Order reverses — undo B first, then A (like putting on shoes then socks: take off socks first, then shoes).

Transpose Rule

(Aᵀ)⁻¹ = (A⁻¹)ᵀ

Inverse and transpose commute. Transposing an invertible matrix gives another invertible matrix.

Double Inverse

(A⁻¹)⁻¹ = A

Inverting twice recovers the original matrix. The inverse of the inverse is the matrix itself.

Scalar Rule

(cA)⁻¹ = (1/c)A⁻¹

Scaling a matrix by c scales its inverse by 1/c (provided c ≠ 0).

The Determinant: Geometric Meaning

The determinant is a scalar computed from a square matrix that encodes how the corresponding linear transformation scales volumes. In 2D, it measures how areas change; in 3D, how volumes change.

Specifically: if you take a unit square (or unit cube), apply the transformation, and measure the resulting parallelogram (or parallelepiped), the area (or volume) equals |det(A)|. The sign of det(A) tells you whether orientation is preserved (positive) or flipped (negative).

2×2 Determinant

\det\begin{pmatrix}a&b\\c&d\end{pmatrix}=ad-bc

The 2×2 determinant ad − bc equals the signed area of the parallelogram formed by the two column vectors of A. If det > 0, orientation is preserved. If det < 0, the transformation includes a reflection. If det = 0, the parallelogram degenerates — the two columns are parallel (linearly dependent).

Computing the 3×3 Determinant

The 3×3 determinant is computed by cofactor expansion along any row or column. Expanding along the first row:

3×3 Cofactor Expansion

\det(A)=a_{11}(a_{22}a_{33}-a_{23}a_{32})-a_{12}(a_{21}a_{33}-a_{23}a_{31})+a_{13}(a_{21}a_{32}-a_{22}a_{31})

Each term is an entry from the first row multiplied by the 2×2 determinant of the submatrix formed by deleting that entry's row and column (the minor). Signs alternate: +, −, + for the first row. The signs follow a checkerboard pattern across the entire matrix.

Key Determinant Properties

The determinant has several powerful algebraic properties that make it much more than just a formula:

det(AB) = det(A) · det(B): The determinant of a product equals the product of determinants. Geometric interpretation: if A scales areas by det(A) and B scales areas by det(B), their composition scales areas by det(A) · det(B).
det(Aᵀ) = det(A): Transposing a matrix does not change its determinant.
det(A⁻¹) = 1/det(A): Since A·A⁻¹ = I and det(I) = 1, we get det(A)·det(A⁻¹) = 1.
Row swap negates det: Swapping any two rows multiplies the determinant by −1.
Row scale scales det: Multiplying a row by scalar c multiplies the determinant by c.
Adding row multiple: Adding a multiple of one row to another does not change the determinant.

Multiplicativity

\det(AB)=\det(A)\cdot\det(B)

This is the most important determinant property. It tells us that the "volume-scaling factor" of a composition of transformations is the product of the individual factors. It also implies det(A) ≠ 0 ↔ A is invertible, since det(A)·det(A⁻¹) = 1 requires both to be non-zero.

Singularity: When the Inverse Fails

A matrix is singular (non-invertible) if and only if det(A) = 0. This corresponds to a linear transformation that collapses space into a lower-dimensional subspace — all of ℝ² gets mapped onto a line, or all of ℝ³ gets flattened onto a plane.

Singular matrices arise naturally in many contexts:

One column is a multiple of another (linearly dependent columns)
One row is a linear combination of other rows
A projection matrix (e.g., [[1,0],[0,0]]) — it kills the y-component entirely
A matrix representing a physical system with no unique solution

Applications

The matrix inverse and determinant appear throughout engineering and science:

Solving Ax = b: If A is invertible, the unique solution is x = A⁻¹b. This is the conceptual basis for solving linear systems, though in practice one uses factorization methods rather than explicitly computing A⁻¹.
Cramer's Rule: Expresses each component of the solution as a ratio of determinants — elegant but computationally expensive for large systems.
Change of basis: Coordinate transformations between different bases use the inverse of the change-of-basis matrix.
Computer graphics: Camera transformations and projection matrices must often be inverted to "unproject" screen coordinates back to 3D space.
Control systems: Stability analysis uses eigenvalues (roots of the characteristic polynomial det(A − λI) = 0).
Statistics: The covariance matrix Σ and its inverse Σ⁻¹ appear in multivariate distributions and the Mahalanobis distance.

Key Takeaways

A matrix A has an inverse A⁻¹ if and only if it is square, full rank, and has non-zero determinant. For 2×2 matrices, the inverse is A⁻¹ = (1/det)[d, −b; −c, a]. For larger matrices, use row reduction on [A|I]. The determinant measures how the transformation scales areas/volumes: det = 0 means singular (collapses space), |det| > 1 means expansion, |det| < 1 means contraction. The sign of det indicates whether orientation is preserved. Key property: det(AB) = det(A)·det(B), so volume-scaling factors compose multiplicatively.