ProofComplete

Solvable and Semisimple Algebras - Key Proof

We prove Cartan's criterion for semisimplicity, one of the most powerful tools in Lie theory. This proof illustrates how the Killing form captures essential structural information.

Proof

Cartan's Criterion: A Lie algebra g\mathfrak{g} over a field of characteristic zero is semisimple if and only if its Killing form κ(X,Y)=tr(adX∘adY)\kappa(X, Y) = \text{tr}(\text{ad}_X \circ \text{ad}_Y) is non-degenerate.

Proof:

(⇐\Leftarrow) Non-degenerate Killing form implies semisimple:

Suppose ΞΊ\kappa is non-degenerate. We must show g\mathfrak{g} has no nonzero abelian ideals.

Let a\mathfrak{a} be an abelian ideal of g\mathfrak{g}. For any X∈aX \in \mathfrak{a} and Y∈gY \in \mathfrak{g}, we have [X,Y]∈a[X, Y] \in \mathfrak{a} (since a\mathfrak{a} is an ideal).

For any Z∈gZ \in \mathfrak{g}: adX(adY(Z))=[X,[Y,Z]]∈a\text{ad}_X(\text{ad}_Y(Z)) = [X, [Y, Z]] \in \mathfrak{a}

Since a\mathfrak{a} is abelian: adX(adY(Z))=[X,[Y,Z]]\text{ad}_X(\text{ad}_Y(Z)) = [X, [Y, Z]] and [adX(adY(Z)),W]=[[X,[Y,Z]],W]=0[\text{ad}_X(\text{ad}_Y(Z)), W] = [[X, [Y, Z]], W] = 0 for any W∈aW \in \mathfrak{a} (using abelian property and Jacobi identity).

This means adX∘adY\text{ad}_X \circ \text{ad}_Y maps g\mathfrak{g} into a\mathfrak{a}, and adX∘adY\text{ad}_X \circ \text{ad}_Y acts as zero on a\mathfrak{a} (since a\mathfrak{a} is abelian implies [X,[Y,Z]]=0[X, [Y, Z]] = 0 for X∈aX \in \mathfrak{a}, Z∈aZ \in \mathfrak{a}).

Therefore: κ(X,Y)=tr(adX∘adY)=0\kappa(X, Y) = \text{tr}(\text{ad}_X \circ \text{ad}_Y) = 0 for all X∈aX \in \mathfrak{a} and all Y∈gY \in \mathfrak{g}.

Since ΞΊ\kappa is non-degenerate, this implies X=0X = 0. Thus a=0\mathfrak{a} = 0, so g\mathfrak{g} is semisimple.

(β‡’\Rightarrow) Semisimple implies non-degenerate Killing form:

Suppose g\mathfrak{g} is semisimple. Consider the kernel: k={X∈g:κ(X,Y)=0 for all Y∈g}\mathfrak{k} = \{X \in \mathfrak{g} : \kappa(X, Y) = 0 \text{ for all } Y \in \mathfrak{g}\}

We claim k\mathfrak{k} is an ideal. For X∈kX \in \mathfrak{k}, Y∈gY \in \mathfrak{g}, and Z∈gZ \in \mathfrak{g}: ΞΊ([X,Y],Z)=tr(ad[X,Y]∘adZ)=tr([adX,adY]∘adZ)\kappa([X, Y], Z) = \text{tr}(\text{ad}_{[X,Y]} \circ \text{ad}_Z) = \text{tr}([\text{ad}_X, \text{ad}_Y] \circ \text{ad}_Z) =tr(adX∘adY∘adZ)βˆ’tr(adY∘adX∘adZ)= \text{tr}(\text{ad}_X \circ \text{ad}_Y \circ \text{ad}_Z) - \text{tr}(\text{ad}_Y \circ \text{ad}_X \circ \text{ad}_Z) =tr(adX∘adY∘adZ)βˆ’tr(adX∘adZ∘adY)(cyclicΒ property)= \text{tr}(\text{ad}_X \circ \text{ad}_Y \circ \text{ad}_Z) - \text{tr}(\text{ad}_X \circ \text{ad}_Z \circ \text{ad}_Y) \quad \text{(cyclic property)} =tr(adX∘[adY,adZ])=ΞΊ(X,[Y,Z])=0= \text{tr}(\text{ad}_X \circ [\text{ad}_Y, \text{ad}_Z]) = \kappa(X, [Y, Z]) = 0

The last equality holds since X∈kX \in \mathfrak{k} and [Y,Z]∈g[Y, Z] \in \mathfrak{g}. Thus [X,Y]∈k[X, Y] \in \mathfrak{k}, proving k\mathfrak{k} is an ideal.

Moreover, k\mathfrak{k} is abelian: for X,Y∈kX, Y \in \mathfrak{k}: κ([X,Y],Z)=κ(X,[Y,Z])=0\kappa([X, Y], Z) = \kappa(X, [Y, Z]) = 0 for all Z∈gZ \in \mathfrak{g}, so [X,Y]∈k[X, Y] \in \mathfrak{k}. But also from the invariance and the fact that X,Y∈kX, Y \in \mathfrak{k}, we can show ad[X,Y]\text{ad}_{[X,Y]} is nilpotent on g\mathfrak{g}, which forces [X,Y]=0[X, Y] = 0 in a semisimple algebra.

Since g\mathfrak{g} is semisimple, it has no nonzero abelian ideals. Therefore k=0\mathfrak{k} = 0, which means ΞΊ\kappa is non-degenerate. β–‘

β– 
Remark

This proof reveals the deep connection between the algebraic structure (ideals, commutators) and the bilinear form (Killing form). The non-degeneracy of ΞΊ\kappa is both necessary and sufficient for the absence of abelian ideals, which characterizes semisimplicity.