A nonnegative integer (or cardinal) describing the maximal number of linearly independent rows or columns of a matrix, equivalently the dimension of the image (column space) of a linear map; it quantifies the effective linear degrees of freedom of the operator or matrix.

Principle

Rank equals the dimension of the image of a linear transformation and is invariant under change of basis; in finite dimensions it satisfies the rank–nullity theorem: dim(domain) = rank + nullity, and full rank characterizes invertibility for square matrices.

Demonstration

A 3×3 identity matrix has rank 3; a 3×3 matrix with two independent rows has rank 2 and maps R^3 onto a 2-dimensional subspace; the linear map T: R^4 → R^3 given by a 3×4 matrix of rank 3 has image dimension 3 and nullity 1.

Misapplication

Confusing matrix rank with tensor rank or multilinear rank, computing rank over the wrong field (for example using real arithmetic when the problem is over a finite field), or assuming numerical rank computed approximately equals exact algebraic rank without stability checks.

Consequence

Rank determines solvability of linear systems (a system has solutions constrained by rank conditions), invertibility of square matrices (full rank), and the dimension of images and cokernels; it also controls dimension counts in linear algebraic geometry and determines the number of independent constraints.

Reversal

Nullity (the dimension of the kernel) measures the deficiency of rank; one may instead focus on cokernel dimension or corank (codimension of the image) which quantifies constraints unmet by the map's image.

Boundary

Over general rings rank can be ambiguous (several non-equivalent notions such as Smith rank or free rank exist); for infinite-dimensional operators one may speak of finite rank, infinite rank, or Fredholm index instead of a simple integer rank.

Semantic Tension

Rank is easily conflated with related invariants (tensor rank, analytic rank, numerical rank, or free rank over rings); specifying the underlying field or ring and whether one means matrix/operator/tensor rank prevents misinterpretation.

Synthesis

Rank measures the size of the image of a linear map — the number of independent output directions — and together with nullity provides the fundamental partition of domain dimension that governs solvability, invertibility, and constraint counts in linear settings.