Source: Abstract Linear Algebra, UIUC
Tags: matrix rank, row operations, column operations, elementary operations, matrix inverse, augmented matrix, row echelon form, invertible matrix, full rank, linear algebra
Difficulty: Intermediate Prerequisites: Basic matrix arithmetic (addition, scalar multiplication), familiarity with systems of linear equations and row echelon form.
Matrix rank and invertibility are two of the most important structural ideas in linear algebra. Rank tells you how much "independent information" a matrix carries, and it controls whether systems of equations have solutions, whether transformations are injective or surjective, and whether a matrix can be inverted. If you are comfortable with these two concepts, a large portion of linear algebra falls into place. You should already be familiar with basic matrix notation and how to set up a system of linear equations as a matrix equation.
The rank of a matrix is the number of linearly independent rows (or columns). A square matrix is invertible precisely when its rank equals its size. You find the inverse by row-reducing the augmented matrix [A | I] until the left side becomes the identity.
Rank of a matrix
The number of linearly independent rows (equivalently, columns) in a matrix. In simple terms, it tells you how many rows are "doing something new" rather than being copies or combinations of the others.
Elementary row operations
Three permitted moves: swap two rows, multiply a row by a nonzero scalar, or add a scalar multiple of one row to another. Think of these as legal rewrites that change the look of the matrix without changing its rank or the solution set of the associated system.
Elementary column operations
The column analogues of row operations: swap columns, scale a column, or add a multiple of one column to another. These change the column labelling but preserve the column rank.
Row echelon form (REF)
A matrix shape where each leading entry sits to the right of the leading entry in the row above, and all entries below each leading entry are zero. Think of it as a staircase of pivots descending from the top-left.
Invertible matrix (nonsingular matrix)
A square matrix A for which there exists a matrix A⁻¹ such that AA⁻¹ = A⁻¹A = Iₙ. In simple terms, it is a matrix whose transformation can be completely undone.
Augmented matrix
A matrix formed by appending extra columns to the right of the original, typically written [A | B]. Used as a bookkeeping device so that row operations act on both sides simultaneously.
The goal is to simplify a matrix using elementary row and column operations until you can count the independent rows or columns by inspection.
Row-reduce the matrix toward row echelon form; the number of nonzero rows in that form equals the rank.
For a 2 × 2 matrix: check whether one row is a scalar multiple of the other. If it is, the rank is 1; if not, the rank is 2.
For a 3 × 3 matrix: apply row operations to introduce zeros below each pivot. Count the resulting nonzero rows to read off the rank.
Rank counts the maximum number of linearly independent rows. The row rank and column rank of any matrix are always equal.
If rank(A) equals the number of rows, every row is independent. If it equals the number of columns, every column is independent.
An n × n matrix is invertible if and only if its rank equals n (full rank).
Equivalently, A is invertible when its row echelon form has a pivot in every column, or when det(A) ≠ 0.
Construct the augmented matrix [A | Iₙ], where Iₙ is the n × n identity matrix.
Apply elementary row operations to transform the left block into the identity.
When the left block reaches Iₙ, the right block has become A⁻¹.
If at any stage you get a row of all zeros on the left, A is not invertible and you can stop.
For a 2 × 2 matrix A = [[a, b], [c, d]]:
A⁻¹ = (1 / (ad – bc)) × [[d, –b], [–c, a]]
provided ad – bc ≠ 0.
Inverting matrices is how engineers solve systems of equations in circuit analysis, structural mechanics, and control theory. When you calibrate a sensor array or solve a least-squares regression, you are relying on rank to confirm the system has a unique solution.
Students sometimes believe row rank and column rank can differ. They cannot; for any matrix, the two are always equal.
A common error is attempting to invert a non-square matrix. Only square matrices can be invertible in the standard sense.
Some students think that if a matrix has no zero rows it must be full rank. A row can be a linear combination of other rows without being all zeros.
Forgetting to check for invertibility before computing: if you row-reduce and get a zero row on the left side of [A | I], the inverse does not exist.
⚠️ "Find the rank" is one of the most common exam instructions. Be comfortable row-reducing quickly and reading off pivot positions.
⚠️ The augmented-matrix method for inverses is a standard exam problem. Practise the full procedure on 3 × 3 matrices until it is automatic.
⚠️ The equivalence "A is invertible ⟺ rank(A) = n ⟺ det(A) ≠ 0" is tested constantly, often as a true/false or "which of the following are equivalent" question.
True or false: A 4 × 3 matrix can have rank 4.
Fill in the blank: A square matrix is invertible if and only if its rank equals ______.
True or false: Swapping two rows of a matrix changes its rank.
Fill in the blank: To compute A⁻¹, you row-reduce the augmented matrix [A | ______].
True or false: If the determinant of a matrix is zero, the matrix is invertible.
Answers: 1. False (rank ≤ min(rows, columns) = 3). 2. n (its number of rows/columns). 3. False (rank is unchanged by elementary row operations). 4. Iₙ (the identity matrix). 5. False (det = 0 means not invertible).
Q: Describe the procedure for finding the inverse of a 3 × 3 matrix using row reduction.
A: Form the augmented matrix [A | I₃]. Apply elementary row operations to the entire augmented matrix until the left block is reduced to the identity I₃. The right block then contains A⁻¹. If a zero row appears on the left, A has no inverse.
Q: A 3 × 3 matrix has two identical rows. What is the maximum possible rank?
A: The maximum rank is 2, because the two identical rows contribute only one independent row.
Q: Why does rank(A) = n guarantee that A is invertible?
A: Rank n means every column is a pivot column, so the system Ax = b has a unique solution for every b. This is precisely what it means for A to have an inverse.
Q: Can a matrix have different row rank and column rank?
A: No. The row rank and column rank of any matrix are always equal. This common value is called the rank of the matrix.
This material connects directly to determinants (covered in the next set of notes), because det(A) ≠ 0 is another test for invertibility. It also links to eigenvalues: a matrix with rank less than n has 0 as an eigenvalue. Understanding rank is essential when you reach the dimension theorem (rank–nullity theorem) later in the course.
matrix rank, row rank, column rank, elementary row operations, elementary column operations, row echelon form, reduced row echelon form, RREF, invertible matrix, nonsingular matrix, singular matrix, augmented matrix, Gauss–Jordan elimination, pivot, full rank, rank–nullity theorem, matrix inverse computation, linear independence of rows