The process and property by which a linear operator or matrix is represented, relative to some basis, by a diagonal matrix; equivalently, the ambient vector space admits a basis made of eigenvectors for the operator.

Principle

A linear map is diagonalizable precisely when there exists a full set of linearly independent eigenvectors, which happens algebraically when the minimal polynomial factors into distinct linear factors over the base field (so algebraic multiplicity equals geometric multiplicity for each eigenvalue).

Demonstration

A symmetric 3×3 real matrix with three distinct real eigenvalues is diagonalizable by an orthogonal change of basis; explicitly A = Q Λ Q^T where Λ is diagonal with the eigenvalues on the diagonal and Q has the orthonormal eigenvectors as columns.

Misapplication

Asserting that every matrix is diagonalizable over its base field; attempting to diagonalize a non-diagonalizable matrix (for example a single Jordan block of size >1) will fail because it lacks a complete eigenvector basis.

Consequence

When an operator is diagonalizable, computing polynomials, powers, and functions of the operator reduces to applying the scalar function to the diagonal entries; spectral computations and stability analysis become straightforward.

Reversal

The converse simplification is the Jordan (or rational) canonical forms: when diagonalization is impossible one instead expresses the operator in Jordan blocks or companion blocks which retain nilpotent structure.

Boundary

Requires existence of enough eigenvectors; diagonalization depends on the base field (an operator diagonalizable over an extension may not be diagonalizable over the original field). It excludes operators with nontrivial Jordan nilpotent parts.

Semantic Tension

Tension exists between diagonalizability and normality: normal operators over C are diagonalizable via a unitary basis, but diagonalizability alone does not imply normality; also diagonalization (similarity to a diagonal) differs from congruence or singular-value decompositions.

Synthesis

Diagonalization isolates pure scaling actions of an operator by choosing an eigenvector basis; it is the strongest linear simplification, attainable exactly when the spectrum produces a full eigenbasis, and it converts operator problems into scalar ones.