A matrix (or linear endomorphism) that is not invertible; over a field this is equivalent to having determinant zero, nonmaximal rank, or a nontrivial kernel.

Principle

Singularity signals loss of full rank: columns (or rows) become linearly dependent, the linear map fails to be bijective, and algebraic invariants such as determinant vanish, affecting solvability of linear systems.

Demonstration

The 2×2 matrix [[1,2],[2,4]] is singular because its rows are linearly dependent and det = 1·4 − 2·2 = 0; the system Ax = b with such A either has no solution or infinitely many if b lies outside or inside the column space respectively.

Misapplication

Equating numerical ill-conditioning (a nearly singular matrix with tiny determinant in floating-point arithmetic) with algebraic singularity leads to confusion; also assuming singularity implies nilpotence or diagonalizability failures is incorrect.

Consequence

A singular matrix cannot be inverted, so linear systems may lack unique solutions; spectral consequences include zero as an eigenvalue, and algebraic constructions that require inverses (matrix factorizations, change-of-basis) must be modified or replaced by pseudoinverses.

Reversal

The opposite is a nonsingular (invertible) matrix: full rank, nonzero determinant over a field, and a bijective linear map. One can remove singularity by restricting to invariant subspaces or by perturbation in analytic contexts.

Boundary

Over fields the equivalences (determinant zero ↔ noninvertible ↔ rank drop) hold; over general commutative rings with zero divisors the determinant can vanish without noninvertibility behaving as over fields, and one must consider left/right inverses in noncommutative rings.

Semantic Tension

Singular competes with 'degenerate form' and with the analytical notion of 'ill-conditioned' — algebraic singularity is an exact statement about invertibility while ill-conditioning is numeric and context-dependent; clarify which is meant.

Synthesis

A singular matrix is an algebraic failure of invertibility manifested by zero determinant, rank deficiency, and a nontrivial kernel; recognizing singularity determines solvability, spectral properties, and the need for alternative tools like pseudoinverses or restriction to images.