A procedure that expresses a linear endomorphism of a finite-dimensional vector space (over a suitable field) as a direct sum of Jordan blocks, or equivalently decomposes an endomorphism into the commuting sum of a semisimple (diagonalisable) part and a nilpotent part that share the same generalized eigenspaces.

Principle

Separate spectral (semisimple) behavior from nilpotent behavior so that an operator is represented as s+n with s semisimple, n nilpotent, and [s,n]=0; when the base field is algebraically closed this yields a canonical block decomposition up to ordering.

Demonstration

Over the complex numbers, a 3×3 matrix with characteristic polynomial (λ−2)^3 may be conjugated to a Jordan matrix with one 3×3 Jordan block (nilpotent part nonzero) or to a diagonal matrix when all Jordan blocks are 1×1; equivalently the matrix equals its diagonalizable part (s) plus its nilpotent part (n) with s and n commuting.

Misapplication

Applying Jordan decomposition without checking hypotheses: attempting to write an operator over a non–algebraically closed field in Jordan normal form, or forcing a decomposition for an infinite-dimensional operator without establishing a compatible generalized eigenspace decomposition.

Consequence

When applicable, the decomposition gives a canonical local description of the operator, simplifies computation of functions of the operator (e.g. exponential, polynomial evaluation) and clarifies invariant subspace structure and similarity classification.

Reversal

The inversion is thinking of an operator solely as 'diagonalizable' (semisimple) or solely as 'nilpotent'; reversing yields the extremes: purely semisimple operators (nilpotent part zero) versus purely nilpotent operators (semisimple part scalar).

Boundary

Applies primarily to linear endomorphisms of finite-dimensional vector spaces where the minimal polynomial splits; it excludes general nonlinear maps, operators without a splitting field, and many infinite-dimensional operators unless additional spectral hypotheses hold.

Semantic Tension

Tension arises between Jordan normal form (a matrix-level canonical form over an algebraically closed field) and rational canonical form (which works over arbitrary fields); both decompose structure but differ in uniqueness and field-dependence.

Synthesis

Jordan decomposition organizes a linear map into commuting semisimple and nilpotent components (or explicit Jordan blocks) so one can treat spectral eigenvalue multiplicity and generalized eigenspace nilpotence separately for classification and computation.