What the Rank of a Matrix Is
The rank of a matrix is the maximum number of linearly independent rows it contains — which, remarkably, is always equal to the maximum number of linearly independent columns. It is written $\rho(A)$ or $\text{rank}(A)$. Linearly independent means no row (or column) can be built as a combination of the others. The rank is the dimension of the space those rows or columns span, so it measures the genuine information content of the matrix.
For a matrix $A$ of order $m \times n$, the rank cannot exceed the smaller of $m$ and $n$:
$$\rho(A) \leq \min(m, n)$$
The only matrix with rank 0 is the zero matrix — every entry is 0, so there is no independent row at all. Every other matrix has rank at least 1. The size of the matrix, its order, sets the ceiling; the dependencies among rows set the actual value.
How Do You Find the Rank of a Matrix?
Two methods dominate. The echelon-form method is the workhorse; the minor method is the definition made concrete.
The echelon-form method
Convert the matrix to row echelon form using elementary row operations (swapping rows, scaling a row, adding a multiple of one row to another). Then count the non-zero rows — that count is the rank.
$$\rho(A) = \text{number of non-zero rows in row echelon form}$$
This is the method to reach for in practice, because row operations are systematic and cheap even for large matrices.
The minor method
A minor is the determinant of a square submatrix. The rank is the order of the largest non-zero minor:
If some $r \times r$ minor is non-zero but every $(r+1) \times (r+1)$ minor is zero, the rank is $r$.
The minor method connects rank directly to the determinant, but for a large matrix it means checking many submatrices, which is why echelon form usually wins.
Variable glossary. $A$ is the matrix; $m$ is its number of rows; $n$ its number of columns; $\rho(A)$ or $\text{rank}(A)$ is the rank; a minor of order $r$ is the determinant of an $r \times r$ submatrix.
[INTERACTIVE: An editable matrix where the reader changes entries and watches the tool row-reduce step by step, highlighting pivots and reporting the rank as the count of non-zero rows live.]
What Are the Properties of Rank?
A handful of properties make rank predictable and useful.
Rank is bounded by size. $\rho(A) \leq \min(m, n)$ for any $m \times n$ matrix. A $3 \times 5$ matrix has rank at most 3.
Row rank equals column rank. The number of independent rows always matches the number of independent columns.
Transpose preserves rank. $\rho(A) = \rho(A^{\mathsf{T}})$, a direct consequence of the previous property.
Full rank means invertible (for square matrices). An $n \times n$ matrix is invertible exactly when its rank is $n$ — that is, when no row is redundant. A square matrix with rank below $n$ is singular.
The zero matrix has rank 0. It is the only matrix with rank 0.
These connect rank to ideas you have already met: a non-zero determinant on a square matrix means full rank, which means an inverse exists.
Examples of Rank of a Matrix
Example 1
Find the rank of $\begin{bmatrix} 1 & 2 \ 3 & 4 \end{bmatrix}$.
The determinant is $(1)(4) - (2)(3) = 4 - 6 = -2 \neq 0$. A non-zero $2 \times 2$ minor exists, so the rank equals the order. Final answer: $\rho(A) = 2$.
Example 2
Find the rank of $\begin{bmatrix} 1 & 2 \ 2 & 4 \end{bmatrix}$.
The determinant is $(1)(4) - (2)(2) = 0$, so there is no non-zero $2 \times 2$ minor. But a $1 \times 1$ minor — any single non-zero entry, such as 1 — is non-zero. The largest non-zero minor has order 1. Final answer: $\rho(A) = 1$. (The second row is just twice the first, so only one row is independent.)
Example 3 (where students lose the mark)
Find the rank of $\begin{bmatrix} 1 & 2 & 3 \ 2 & 4 & 6 \ 3 & 6 & 9 \end{bmatrix}$.
Wrong attempt. A student sees a $3 \times 3$ matrix with no zeros, no obvious repeats, and writes rank 3 — "it is full size, so it must be full rank."
The break. Look at the rows, not the size. Row 2 is $2 \times$ Row 1, and Row 3 is $3 \times$ Row 1. Every row is a multiple of the first, so only one row is genuinely independent. The number of rows is not the rank.
Correct. Row-reduce: subtract $2 \times$ Row 1 from Row 2, and $3 \times$ Row 1 from Row 3:
$$\begin{bmatrix} 1 & 2 & 3 \ 0 & 0 & 0 \ 0 & 0 & 0 \end{bmatrix}$$
One non-zero row remains. Final answer: $\rho(A) = 1$.
Example 4
Find the rank of $\begin{bmatrix} 1 & 2 & 3 \ 2 & 4 & 6 \ 1 & 1 & 1 \end{bmatrix}$ by row reduction.
Subtract $2 \times$ Row 1 from Row 2, and Row 1 from Row 3:
$$\begin{bmatrix} 1 & 2 & 3 \ 0 & 0 & 0 \ 0 & -1 & -2 \end{bmatrix}$$
Swap Row 2 and Row 3 to reach echelon form:
$$\begin{bmatrix} 1 & 2 & 3 \ 0 & -1 & -2 \ 0 & 0 & 0 \end{bmatrix}$$
Two non-zero rows remain. Final answer: $\rho(A) = 2$.
Example 5
Find the rank of the $3 \times 4$ matrix $\begin{bmatrix} 1 & 0 & 2 & 1 \ 0 & 1 & 3 & 2 \ 0 & 0 & 0 & 0 \end{bmatrix}$.
The matrix is already in echelon form: two non-zero rows. Note the ceiling: for a $3 \times 4$ matrix, $\rho(A) \leq \min(3, 4) = 3$, and 2 sits comfortably below it. Final answer: $\rho(A) = 2$.
Example 6
For which value of $k$ does $\begin{bmatrix} 1 & 2 \ 3 & k \end{bmatrix}$ have rank 1?
Rank drops below 2 only when the determinant is zero. Set $\det A = 0$:
$$(1)(k) - (2)(3) = 0$$
$$k - 6 = 0$$
$$k = 6$$
When $k = 6$, the second row is $3 \times$ the first and the rank falls to 1; for any other $k$ the rank is 2. Final answer: $k = 6$.
Why Rank Matters — "how much a matrix really knows"
The idea grew out of solving systems of linear equations. People noticed that some systems with many equations behaved as if they had far fewer — extra equations that added no new constraint. Rank is the precise count of how many of those equations actually pull their weight, and it is what tells you whether a system can be solved at all.
Solving linear systems. Compare the rank of the coefficient matrix with the rank of the augmented matrix: equal ranks mean a solution exists; if that rank also equals the number of unknowns, the solution is unique. This is the Rouché–Capelli theorem, and it governs whether a homogeneous system of linear equations has non-trivial solutions.
Data science and statistics. Rank reveals redundancy in data: a low-rank data matrix means columns are correlated, which underlies dimensionality reduction and techniques like principal component analysis.
Engineering and control. Whether a system is controllable or observable is decided by the rank of a particular matrix.
Image compression. Low-rank approximation throws away the least informative directions of an image matrix to shrink the file while keeping the picture.
A subtlety worth flagging, because it confused me when I first taught it: the rank of a matrix is a structural fact, not something you can read off its size or even its determinant alone (for non-square matrices, there is no determinant). You have to look at the dependencies among the rows. The size only gives you the ceiling.
Where Things Go Sideways With Rank
Mistake 1: Assuming rank equals the number of rows
Where it slips in: Any matrix with no zero entries that looks "full."
Don't do this: Reading the rank off the matrix's size. A $3 \times 3$ matrix does not automatically have rank 3.
The correct way: Reduce to echelon form and count non-zero rows, or check the largest non-zero minor. The first instinct is to equate "big and full of numbers" with "full rank," but a matrix whose rows are multiples of one another can have rank 1 no matter how large it is.
Mistake 2: Stopping row reduction too early
Where it slips in: Counting rows before the matrix is truly in echelon form.
Don't do this: Counting a row as non-zero when more reduction would have zeroed it out.
The correct way: Finish the reduction completely — every pivot to the left of the one below it, all-zero rows pushed to the bottom — before counting. The habit that fixes this is reducing fully and only then counting, rather than counting as you go.
Mistake 3: Hunting minors in the wrong order
Where it slips in: Using the minor method on a larger matrix.
Don't do this: Concluding the rank is low after checking only a few small minors, or testing minors at random.
The correct way: Start from the largest possible minor and work down; the rank is the order of the first non-zero one you find. Testing in a haphazard order wastes effort and invites a wrong conclusion.
Key Takeaways
The rank of a matrix is the number of linearly independent rows, equal to the number of independent columns.
Find it by row-reducing to echelon form and counting non-zero rows, or by locating the largest non-zero minor.
Rank is bounded by $\min(m, n)$, and a square matrix is invertible exactly when it has full rank.
The most common mistake is assuming the rank equals the number of rows — dependent rows lower it.
Rank decides whether a linear system has a solution and how many, via the comparison of coefficient and augmented matrix ranks.
Practice Questions on Rank of a Matrix
Work through these, then check your answers below.
Find the rank of $\begin{bmatrix} 2 & 4 \ 1 & 2 \end{bmatrix}$.
Row-reduce $\begin{bmatrix} 1 & 1 & 1 \ 1 & 2 & 3 \ 1 & 3 & 5 \end{bmatrix}$ and count the non-zero rows.
Find the value of $k$ that makes $\begin{bmatrix} 1 & k \ 4 & 8 \end{bmatrix}$ have rank 1.
State the rank of the $2 \times 4$ matrix $\begin{bmatrix} 1 & 0 & 3 & 0 \ 0 & 0 & 0 & 0 \end{bmatrix}$.
Answer to Question 1: $\det = (2)(2) - (4)(1) = 0$, so no $2 \times 2$ minor is non-zero, but single entries are. The largest non-zero minor has order 1, so $\rho(A) = 1$. (Row 1 is twice Row 2.)
Answer to Question 2: Subtract Row 1 from Rows 2 and 3: $$\begin{bmatrix} 1 & 1 & 1 \ 0 & 1 & 2 \ 0 & 2 & 4 \end{bmatrix}$$ Then subtract $2 \times$ Row 2 from Row 3: $$\begin{bmatrix} 1 & 1 & 1 \ 0 & 1 & 2 \ 0 & 0 & 0 \end{bmatrix}$$ Two non-zero rows remain, so $\rho(A) = 2$.
Answer to Question 3: Rank falls to 1 when $\det A = 0$. So $(1)(8) - (k)(4) = 0$, giving $8 - 4k = 0$, so $k = 2$.
Answer to Question 4: Only the first row is non-zero, so $\rho(A) = 1$. (The ceiling is $\min(2, 4) = 2$, and 1 sits below it.)
A Practical Next Step
If a row reduction goes wrong, come back to Mistake 2 and finish the form before counting. At Bhanzu, our trainers teach rank by reducing first and reading the rank off the result, so students stop guessing it from the matrix's shape.
Want a live Bhanzu trainer to walk through more rank of a matrix problems? Book a free demo class.
Was this article helpful?
Your feedback helps us write better content