Definition

Given a θ\theta of a g\mathfrak g, its Cartan decomposition is the vector-space direct sum

g=kp,k=ker(θ1),p=ker(θ+1).\mathfrak g=\mathfrak k\oplus\mathfrak p,\qquad \mathfrak k=\ker(\theta-1),\quad \mathfrak p=\ker(\theta+1).

Because θ\theta is a , the brackets satisfy

[k,k]k,[k,p]p,[p,p]k.[\mathfrak k,\mathfrak k]\subseteq\mathfrak k,\qquad [\mathfrak k,\mathfrak p]\subseteq\mathfrak p,\qquad [\mathfrak p,\mathfrak p]\subseteq\mathfrak k.

Thus k\mathfrak k is a , whereas p\mathfrak p is generally not. The decomposition is orthogonal for the positive associated with θ\theta.

Compact and noncompact directions

The subalgebra k\mathfrak k is maximally compactly embedded and integrates, under the usual group hypotheses, to a KK. The space p\mathfrak p models the at the base point of the G/KG/K. The bracket inclusions are the infinitesimal algebraic signature of its symmetric-space geometry Helgason, Chapter V.

Example

For g=gln(R)\mathfrak g=\mathfrak{gl}_n(\mathbb R) and θ(X)=XT\theta(X)=-X^{\mathsf T},

k=so(n),p={X:XT=X}.\mathfrak k=\mathfrak{so}(n),\qquad \mathfrak p=\{X:X^{\mathsf T}=X\}.

Exponentiating p\mathfrak p produces the positive-definite symmetric matrices. The resulting group-level statement is the familiar polar decomposition and is a model for the .

Terminological distinction

This kp\mathfrak k\oplus\mathfrak p decomposition should not be confused with a root-space decomposition relative to a . A maximal abelian subspace ap\mathfrak a\subseteq\mathfrak p instead gives the , which records how the noncompact directions decompose under ada\operatorname{ad}\mathfrak a.

References
  1. A. W. Knapp, Lie Groups Beyond an Introduction, 2nd ed., Birkhäuser, 2002. Publisher record. Relevant: Chapter VI, §2.
  2. S. Helgason, Differential Geometry, Lie Groups, and Symmetric Spaces, American Mathematical Society, 2001. DOI record. Relevant: Chapter V on symmetric decompositions.