Matrix Calculator

What is a matrix calculator?

A matrix calculator is an online tool that performs linear algebra operations on rectangular arrays of numbers. Common operations include addition, subtraction, multiplication, scalar multiplication, transpose, determinant, inverse, and row reduction to reduced row echelon form (RREF). These operations are central to linear algebra, which is foundational to machine learning, computer graphics, structural engineering, economics, and most quantitative science.

This free matrix calculator supports matrices up to 4×4 and performs eight distinct operations: A+B, A−B, A×B, k·A, Aᵀ, det(A), A⁻¹, and RREF(A). For determinant and inverse calculations, step-by-step working is displayed — showing cofactor expansion for the determinant and the adjugate method for the inverse. RREF shows each row operation (swap, scale, eliminate) in sequence, making it useful for checking homework and understanding Gaussian elimination.

How to use the matrix calculator

Set the dimensions of Matrix A and Matrix B using the row and column dropdowns (1 to 4 for each dimension). Click any cell in the grid to enter a value — Tab moves to the next cell horizontally, making data entry fast on keyboard. Decimal values and negative numbers are fully supported; leave cells at 0 if not needed. For operations requiring only one matrix (transpose, determinant, inverse, RREF), only Matrix A needs values. For binary operations (A+B, A−B, A×B), both matrices must have compatible dimensions.

Click the operation button to compute and view the result grid below. To solve a system of linear equations, create an augmented matrix [A|b] by setting Matrix A to have one extra column for the constants, then run RREF(A). The solution reads directly from the rightmost column of the result.

Matrix operations, determinants, and inverses explained

Addition and subtraction are elementwise operations requiring identical dimensions. Multiplication A×B requires A's column count to equal B's row count; the result has A's row count and B's column count. The (i,j) element of the product is the dot product of row i from A and column j from B. Scalar multiplication kA multiplies every element by k. Transpose Aᵀ mirrors the matrix over its main diagonal, turning an m×n matrix into n×m.

The determinant encodes whether a matrix is invertible (det ≠ 0) and by what factor it scales areas or volumes. The inverse A⁻¹ 'undoes' A: multiplying A by A⁻¹ returns the identity matrix. For practical equation-solving, RREF applied to the augmented matrix [A|b] is numerically more stable than computing A⁻¹ and multiplying — it simultaneously reveals consistency, uniqueness, and the solution values. The rank of A (the number of non-zero rows in RREF) determines which of the three solution types (unique, infinite, none) applies.

Why step-by-step working matters

Most online matrix calculators show the answer but not how they got there. This one shows every intermediate step for the three operations students most need to understand: cofactor expansion for determinants, the adjugate method for matrix inverses, and individual row operations (scale, swap, eliminate) for RREF. That step-by-step output is what makes the tool useful for homework checking rather than just answer-grabbing.

The RREF implementation uses partial pivoting — swapping to the largest absolute-value element in each column before dividing — which improves numerical stability and avoids the floating-point division errors that appear in naive implementations. Values below 1e-10 are set to exact zero, so results stay clean even for ill-conditioned inputs where a simpler algorithm would show 0.0000000001 instead of 0.

How it compares to Wolfram Alpha, Symbolab, and Mathway

Wolfram Alpha is the most powerful alternative but caps free matrix queries and hides step-by-step working behind a $7/month Pro subscription. Symbolab and Mathway require an account sign-up and limit daily free step-by-step calculations to roughly 5–10 per day. Desmos handles graphing well but does not support matrix operations. This calculator is different: no account, no daily limit, and no paywall. Every operation — including full working for determinants, inverses, and RREF — is free on every use.

It also runs entirely in your browser. Your matrix entries never leave your device. That matters when you're working on homework problems or proprietary engineering data that shouldn't be sent to a third-party server. Close the tab and the data is gone.

Matrix multiplication explained: the dot product rule and non-commutativity

Matrix multiplication is not elementwise — you cannot simply multiply matching positions as you do with addition. The product C = A×B is defined only when A has dimensions m×n and B has dimensions n×p; the shared dimension n must match. The result C is m×p. Each entry C(i,j) is computed as the dot product of row i of A with column j of B: C(i,j) = A(i,1)·B(1,j) + A(i,2)·B(2,j) + … + A(i,n)·B(n,j). For a concrete 2×2 example with A = [[1,2],[3,4]] and B = [[5,6],[7,8]]: C(1,1) = 1·5 + 2·7 = 19, C(1,2) = 1·6 + 2·8 = 22, C(2,1) = 3·5 + 4·7 = 43, C(2,2) = 3·6 + 4·8 = 50, giving C = [[19,22],[43,50]].

A critical property of matrix multiplication is non-commutativity: AB ≠ BA in general. Even when both products are defined (i.e., A and B are both square), reversing the order almost always changes the result. Geometrically, if A and B represent linear transformations such as rotations and scalings, AB applies B first then A, while BA applies A first then B — different sequences produce different outcomes, just as rotating then scaling a shape is not the same as scaling then rotating it.

The computational cost of matrix multiplication also deserves attention. The naive algorithm requires O(n³) arithmetic operations to multiply two n×n matrices — doubling the size multiplies the work eightfold. For large matrices (n in the thousands, as encountered in deep learning and scientific computing), this is prohibitive. Strassen's algorithm reduces the complexity to approximately O(n^2.807) by cleverly reducing the number of scalar multiplications from 8 to 7 per 2×2 block. More recent algorithms push the exponent further toward O(n^2.37), and practical libraries like BLAS and LAPACK implement highly optimized blocked variants that exploit CPU cache architecture, making large matrix products feasible in neural network training.

The determinant: geometric meaning and the singular matrix

The determinant is a single scalar that encodes fundamental structural information about a square matrix. For a 2×2 matrix A = [[a,b],[c,d]], det(A) = ad − bc. For a 3×3 matrix, cofactor expansion along the first row gives det(A) = a(ei − fh) − b(di − fg) + c(dh − eg), where the letters label the nine entries in reading order. The alternating signs follow the checkerboard pattern + − + / − + − / + − + for 3×3, and the same principle extends recursively to any size.

Geometrically, the absolute value of the determinant measures how much the linear transformation represented by the matrix scales areas (in 2D) or volumes (in 3D). A 2×2 matrix with |det| = 3 stretches parallelograms by a factor of 3. The sign indicates orientation: a positive determinant preserves the handedness of coordinate axes (no reflection), while a negative determinant includes a reflection. A determinant of exactly 1 means the transformation preserves area or volume perfectly — orthogonal matrices (pure rotations and reflections) always satisfy |det| = 1.

When det(A) = 0, the matrix is called singular. Geometrically, a singular transformation collapses space: a 2D transformation with zero determinant maps the entire plane onto a line (or a point), destroying information irreversibly. Algebraically, this means the rows (or columns) of the matrix are linearly dependent — at least one row can be expressed as a combination of the others. A singular matrix has no inverse, and any system Ax = b with a singular A either has no solution or infinitely many, never a unique one. Checking det(A) before attempting inversion is therefore always the first diagnostic step.

Matrix inverse and solving linear systems

The inverse of a square matrix A, written A⁻¹, is the unique matrix satisfying A·A⁻¹ = A⁻¹·A = I (the identity). It exists if and only if det(A) ≠ 0. For a 2×2 matrix A = [[a,b],[c,d]], the closed-form inverse is A⁻¹ = (1/(ad−bc)) × [[d,−b],[−c,a]] — swap the main diagonal, negate the off-diagonal, divide by the determinant. For larger matrices, the general formula is A⁻¹ = (1/det(A)) × adj(A), where adj(A) (the adjugate) is the transpose of the cofactor matrix. Each cofactor C(i,j) = (−1)^(i+j) × det(M(i,j)), where M(i,j) is the submatrix formed by deleting row i and column j from A.

The most important application of the inverse is solving systems of linear equations. A system of n equations in n unknowns can be written as Ax = b, where A is the coefficient matrix, x is the column vector of unknowns, and b is the column vector of constants. If A is invertible, the unique solution is x = A⁻¹b — left-multiply both sides by A⁻¹. This is conceptually clean but computationally wasteful for large systems. Gaussian elimination (equivalently, RREF on the augmented matrix [A|b]) finds the same solution in roughly a third of the arithmetic operations and is numerically more stable.

Numerical stability is a serious concern for matrix inversion. An ill-conditioned matrix — one whose determinant is close to zero but not exactly zero — amplifies rounding errors enormously during inversion. The condition number of a matrix (the ratio of its largest to smallest singular value) quantifies this sensitivity: a condition number of 10⁶ means that six decimal digits of precision in the input can be completely lost in the output. Professional numerical libraries therefore avoid explicit inversion and instead prefer LU decomposition, which factors A into a lower triangular matrix L and an upper triangular matrix U (A = LU), then solves Ly = b and Ux = y with two fast triangular solves. This approach is both faster and far more numerically robust than computing A⁻¹ directly.

Eigenvalues, eigenvectors, and their applications

An eigenvector of a square matrix A is a non-zero vector v that satisfies Av = λv, where λ is the corresponding eigenvalue. In plain terms: when A acts on v as a linear transformation, it only scales v by the factor λ — it does not rotate v out of its original direction. Every direction that a matrix leaves unchanged (up to scaling) corresponds to an eigenvector, and the amount of scaling in that direction is the eigenvalue. This makes eigenvectors the 'natural axes' of a linear transformation.

To find eigenvalues, rearrange Av = λv as (A − λI)v = 0. This homogeneous system has a non-zero solution only when A − λI is singular, i.e., when det(A − λI) = 0. Expanding that determinant yields the characteristic polynomial — a degree-n polynomial in λ whose roots are the eigenvalues. For a 2×2 matrix [[a,b],[c,d]] the characteristic polynomial is λ² − (a+d)λ + (ad−bc) = 0, revealing that the sum of eigenvalues equals the trace (a+d) and their product equals the determinant (ad−bc). Once each eigenvalue is known, substituting back into (A − λI)v = 0 and solving for v gives the corresponding eigenvectors.

Eigenvalues and eigenvectors are not merely theoretical — they power some of the most impactful algorithms in computing. Google's PageRank algorithm ranks web pages by finding the dominant eigenvector of the link transition matrix: the eigenvector associated with eigenvalue 1 gives the long-run probability of landing on each page during a random walk through the web. In machine learning, Principal Component Analysis (PCA) finds the directions of maximum variance in a dataset by computing the eigenvectors of the data's covariance matrix; the eigenvalues measure how much variance each direction captures, allowing high-dimensional data to be compressed by retaining only the top eigenvectors. In structural engineering, eigenvalues of the stiffness matrix correspond to the natural vibration frequencies of a structure, which must be computed to ensure a bridge or building won't resonate catastrophically under external loading.