Inverse Matrix: Formula, Conditions, and Examples

An inverse matrix undoes a linear transformation represented by an invertible square matrix. For a square matrix $A$, an inverse $A^{-1}$ satisfies both $A{A}^{-1}=I$ and $A^{-1}A=I$. The superscript negative one does not mean taking the reciprocal of each entry. It describes an inverse under matrix multiplication, where order and dimensions matter.

A square matrix is invertible exactly when its determinant is nonzero

A square matrix is invertible exactly when its determinant is nonzero. A zero determinant means the transformation collapses some nonzero direction, so different inputs can produce the same output and no two-sided inverse can recover them. Such a matrix is called singular.

For a two-by-two matrix, the determinant is $ad-bc$. Swap the diagonal entries $a$ and $d$, change the signs of $b$ and $c$, and divide every resulting entry by the determinant. This formula is valid only when that denominator is not zero.

Explore a two-by-two matrix and its inverse

Change the entries to see the transformed square, determinant and inverse.

Entry aEntry bEntry cEntry d

Worked example: calculate and verify the inverse

Take $A$ with rows $(2,1)$ and $(5,3)$. Its determinant is $2\times 3-1\times 5=1$, so the matrix is invertible. Apply the two-by-two rule to obtain rows $(3,-1)$ and $(-5,2)$. Multiplying in either order gives the identity.

Because the determinant is 1, every entry in this example remains an integer. If the determinant were another nonzero number, fractions would appear and should be kept exact until a decimal approximation is genuinely needed.

One inverse can handle many systems with the same coefficient matrix

For $Ax=b$, multiply on the left by $A^{-1}$. Associativity gives $\left(A^{-1}A\right)x=A^{-1}b$, so $x=A^{-1}b$. With the matrix above and $b=(7,19)$, the product gives $x=(2,3)$. Substitution checks that $2\times 2+1\times 3=7$ and $5\times 2+3\times 3=19$.

Computing a full inverse is not always the best numerical method for one large system. Direct factorization can require less computation and behave better under rounding. The inverse remains valuable when many right-hand sides share the same matrix or when the symbolic structure itself is the goal.

Gauss-Jordan elimination works beyond two by two

Place $A$ beside the identity matrix in an augmented array. Apply valid elementary row operations to the whole array. If the left side becomes the identity, the right side is the inverse. If a required pivot cannot be created and the left side retains a zero row, the matrix is singular.

Row swaps, nonzero row scaling, and adding a multiple of one row to another are reversible. Applying the same operations to the identity block builds the inverse step by step. The method also verifies invertibility without first expanding a large determinant.

The inverse of a product reverses the order

If $A$ and $B$ are invertible and their product is defined, then ${\left(AB\right)}^{-1}=B^{-1}{A}^{-1}$. The order reverses because $B^{-1}$ must undo $B$ first. Also, ${\left(A^T\right)}^{-1}={\left(A^{-1}\right)}^T$ and ${\left(A^k\right)}^{-1}={\left(A^{-1}\right)}^k$ for integer powers where the expressions are defined.

A nonzero determinant proves exact invertibility, but it does not by itself show that inversion is numerically stable. The determinant depends on the matrix's scale, so its absolute size is not a reliable standalone measure of sensitivity. Numerical work instead examines conditioning, often through a condition number: a large condition number means that small input or rounding errors can produce much larger changes in the solution or computed inverse.