Jordan decomposition

Jordan decomposition
Name	Jordan decomposition
Field	Linear algebra, Lie algebra, Algebraic group theory
Introduced	1870s
Introduced by	Camille Jordan
Notable for	structure theorem for linear operators, decomposition into semisimple and nilpotent parts

Jordan decomposition provides a canonical splitting of linear transformations and group elements into simpler components that reflect eigenstructure and unipotent behavior. In finite-dimensional Linear algebra and in the theory of Lie algebras and Algebraic groups, the decomposition separates an operator or element into commuting parts with distinct algebraic character: semisimple (diagonalizable) and nilpotent (or unipotent) components. This construction underlies classification results such as the Jordan normal form and the structure theory of representations for GL(n), SL(n), O(n), and Sp(2n).

Introduction

The Jordan decomposition appears across several branches of mathematics as a tool to analyze linear maps, matrices, and group elements by reducing them to canonical pieces. In the matrix setting it is tightly connected to the Jordan normal form and the spectral decomposition used in study of Weyl characters, while in the context of Algebraic groups and Lie algebras it takes the name Jordan–Chevalley decomposition, linking work of Camille Jordan with later developments by Claude Chevalley and others. It plays a central role in representation theory for groups such as GL(n, C), in algebraic geometry for actions of Galois groups, and in differential equations studied by H. Poincaré and Sophus Lie.

Jordan decomposition in linear algebra

For a linear operator T on a finite-dimensional vector space over an algebraically closed field (commonly C), one can express T uniquely as the sum T = S + N where S is semisimple (diagonalizable) and N is nilpotent, with S and N commuting. This additive decomposition corresponds to the partition of the minimal polynomial of T into pairwise coprime factors, and it is manifested concretely by the Jordan normal form classification of matrices used in work on Galois theory and invariant subspaces studied by Issai Schur. For matrices over non-algebraically closed fields such as R or finite fields like F_p, the decomposition exists over suitable extensions and is compatible with field restrictions appearing in the theory of Frobenius endomorphisms.

Jordan–Chevalley decomposition (algebraic groups and Lie algebras)

In the setting of linear algebraic groups and Lie algebras, the Jordan–Chevalley decomposition generalizes the matrix case: an element g in a linear algebraic group G can be written uniquely as g = g_s g_u = g_u g_s with g_s semisimple and g_u unipotent, commuting in G. For x in a Lie algebra 𝔤 of G there is an analogous decomposition x = x_s + x_n with x_s semisimple and x_n nilpotent and [x_s,x_n]=0. This structure is crucial in the classification of conjugacy classes in groups like GL(n), SL(n, C), E8, SO(n), and in the study of representation-theoretic objects appearing in the work of Harish-Chandra, George Lusztig, and Jacobson.

Properties and uniqueness

Uniqueness: In all standard contexts the semisimple and nilpotent (or semisimple and unipotent) parts are unique and commute. Functoriality: the decomposition is preserved under homomorphisms of linear algebraic groups and under representations, a property exploited in the theories of Chevalley groups and Tannakian formalism. Spectral characterization: the semisimple part has spectrum equal to the set of eigenvalues of the original operator with the same multiplicities modulo nilpotent contribution; the nilpotent part has spectrum {0}. Behavior under conjugacy: conjugacy classes decompose compatibly, enabling classification of conjugacy in groups such as GL(n, F_q) and in exceptional groups like G2 and F4. Compatibility with exponentials and logarithms: over C the exponential map intertwines additive and multiplicative decompositions linking Lie algebra elements with group elements as used in work by Élie Cartan.

Computation and algorithms

Algorithmic extraction of the semisimple and nilpotent parts can be done using polynomial functional calculus: given a matrix A with minimal polynomial m(t), one constructs idempotent polynomials p(t), q(t) with p+q=1 so that S=p(A) and N=q(A)A. Computational techniques include rational canonical form algorithms from Frobenius theory, algorithms using primary decomposition from Artin-style methods, and iterative schemes implemented in computer algebra systems used by researchers at institutions such as IHÉS and MIT. Over finite fields, algorithms exploit efficient factorization routines for polynomials over F_q and use ideas from Moll and computational group theory packages developed in projects like GAP and Magma.

Applications and examples

Applications span linear differential equations studied by Paul Painlevé and S. Lie, classification of algebraic group actions on varieties in Alexander Grothendieck’s program, and modular representation theory of finite groups of Lie type used by Robert Steinberg. Explicit examples include decomposing Jordan blocks for a nilpotent matrix, separating rotation and scaling parts for matrices in SO(2), and distinguishing semisimple conjugacy classes in SL(2, C) relevant to monodromy of linear ODEs studied by Riemann and Fuchs. In number theory the decomposition appears in analysis of Frobenius elements in Galois group representations and in the study of étale cohomology in work following Alexander Grothendieck.

Historical context and attribution

The ideas trace to Camille Jordan’s 19th-century work on linear substitutions and the development of the Jordan normal form. The extension to algebraic groups is associated with Claude Chevalley’s mid-20th-century contributions, and subsequent formalizations were advanced by Nathan Jacobson, Élie Cartan, and others in the development of Lie theory. Later computational and representation-theoretic refinements involve contributors such as George Lusztig, Harish-Chandra, and authors of modern computational algebra systems at institutions including University of Sydney and University of Cambridge.

Category:Linear algebra Category:Lie algebras