A square matrix is invertible if and only if its determinant is not zero. That single test is the fastest way to prove invertibility for most matrices you’ll encounter in a course or on an exam. But the determinant is just one entry point into a web of equivalent conditions, and depending on what information you already have about a matrix, a different method may be easier or more illuminating.
The Determinant Test
For any n × n matrix A, compute det(A). If the result is any nonzero number, the matrix is invertible. If det(A) = 0, the matrix is singular and has no inverse. This works for square matrices of any size.
For a 2×2 matrix, the determinant is ad − bc. For a 3×3 matrix, you can expand along a row or column using cofactors. Beyond 3×3, row reduction to an upper triangular form and then multiplying the diagonal entries is usually faster than cofactor expansion, since cofactor expansion grows in complexity very quickly.
One practical warning: if you’re checking invertibility on a computer rather than by hand, the determinant can be misleading. Floating-point arithmetic may return a very small number instead of exactly zero, or it may produce a wildly large or small determinant that obscures the true situation. For computational work, rank-based methods or singular value decomposition are more reliable. For homework and proofs, the determinant is perfectly fine.
Row Reduce and Count Pivots
Row reduce the matrix to echelon form. If every row has a leading 1 (a pivot), meaning you get n pivots for an n × n matrix, the matrix is invertible. If any row becomes all zeros during elimination, the matrix is singular.
This method is especially useful for larger matrices where computing the determinant by cofactor expansion would be tedious. You’re doing the same kind of arithmetic either way, but row reduction gives you a clear visual signal: a full staircase of pivots means invertible, a missing step means not. As a bonus, if you augment the matrix with the identity and keep reducing all the way to reduced row echelon form, you get the actual inverse on the right side.
The Invertible Matrix Theorem
Linear algebra has a powerful result that ties together many properties of a square matrix. For an n × n matrix A, all of the following statements are equivalent. Proving any one of them proves all the others, including invertibility:
- A has n pivots (full row reduction yields no zero rows)
- The columns of A are linearly independent (no column can be written as a combination of the others)
- The columns of A span Rⁿ (they cover the entire space)
- Ax = b has a unique solution for every b
- Ax = 0 has only the trivial solution x = 0
- rank(A) = n
- nullity(A) = 0 (the null space contains only the zero vector)
- det(A) ≠ 0
- 0 is not an eigenvalue of A
This theorem is the reason your professor might accept very different-looking proofs for the same invertibility question. If you can show the columns are linearly independent, you’ve proven invertibility just as rigorously as if you computed the determinant. Pick whichever condition is easiest to verify given what you know about the matrix.
Check the Rank
The rank of an n × n matrix is the number of linearly independent rows (or equivalently, columns). If rank(A) = n, the matrix is called “full rank” and is invertible. If the rank falls short of n, the matrix is singular.
You find the rank by row reducing and counting pivots, so in practice this method overlaps with the pivot-counting approach above. But “rank” is the language you’ll often see in proofs and theoretical arguments. If a problem tells you the rank of a matrix directly, or if you can determine it from context (for instance, knowing the matrix represents a transformation that maps n-dimensional space onto all of n-dimensional space), you can skip the computation entirely and cite full rank as your proof.
Linear Independence of Columns
If you can show that the columns of A are linearly independent, the matrix is invertible. Two ways to do this: set up the equation c₁v₁ + c₂v₂ + … + cₙvₙ = 0 (where v₁ through vₙ are the columns) and show the only solution is all coefficients equal to zero. Alternatively, row reduce and confirm there are no free variables.
This approach is particularly natural when a problem gives you the column vectors explicitly and asks you to prove invertibility. It also connects to the geometric intuition: the columns of an invertible matrix point in truly different directions in n-dimensional space, so no information is lost when you multiply by the matrix.
The Eigenvalue Test
A matrix is invertible if and only if zero is not one of its eigenvalues. If you’ve already computed the eigenvalues for another reason, this gives you invertibility for free. If any eigenvalue equals zero, the matrix is singular.
The logic is straightforward. An eigenvalue of zero means there’s a nonzero vector x such that Ax = 0x = 0. That means the equation Ax = 0 has a nontrivial solution, which directly contradicts invertibility. Conversely, if all eigenvalues are nonzero, the matrix maps every nonzero vector to a nonzero result, and the transformation can be reversed.
Singular Value Decomposition
In computational and applied settings, singular value decomposition (SVD) offers the most robust test. SVD breaks any matrix into three components, and the key information lives in the “singular values,” a set of non-negative numbers associated with the matrix. If all singular values are nonzero, the matrix is invertible. If any singular value is zero, it’s not.
The intuition here is that each singular value represents how much the matrix stretches space along a particular direction. A zero singular value means the matrix crushes an entire direction down to nothing. Once that information is destroyed, there’s no way to recover the original input, so no inverse can exist. The rank of the matrix equals the number of nonzero singular values, which ties this method back to the rank condition.
SVD is rarely something you’d compute by hand for a homework problem, but in programming environments like MATLAB, Python (NumPy), or R, it’s often the preferred method because it handles numerical issues gracefully.
The Adjugate Formula
For a concrete construction of the inverse, the adjugate (or classical adjoint) formula states that A⁻¹ = (1/det(A)) × adj(A), where adj(A) is the transpose of the cofactor matrix. This formula simultaneously proves invertibility and produces the inverse, since it only works when det(A) ≠ 0.
For a 2×2 matrix, this is quick: swap the diagonal entries, negate the off-diagonal entries, and divide by the determinant. For 3×3, you compute nine cofactors, transpose them, and divide by the determinant. Beyond 3×3, this method becomes impractical compared to row reduction, but it remains important in theoretical proofs because it gives an explicit closed-form expression for the inverse.
Choosing the Right Method
If you’re working a homework problem and have the actual matrix in front of you, computing the determinant (for 2×2 or 3×3) or row reducing (for anything larger) is usually the fastest path. If the problem gives you properties of the matrix rather than specific entries, look at the invertible matrix theorem and identify which condition you can verify most directly.
For proofs in a more abstract setting, where you’re working with a matrix described by its properties rather than its numbers, the eigenvalue condition, rank condition, or linear independence of columns often provide cleaner arguments than trying to compute a determinant symbolically. And if you’re writing code, skip the determinant entirely and use rank, SVD, or a built-in library function designed for numerical stability.