Let $G$ be a connected Lie Group and $K<G$ a maximal compact subgroup.

Denote by $\Omega^q(G/K)^G$ the $G$-invariant real-valued $q$-forms on the manifold $G/K$, i.e. those forms $\omega$ s.t. $g^*\omega=\omega$, where $g$ denotes the left translation mapping $hK$ to $(gh)K$.

Evaluation at the identity $eK$ yields an isomorphism $$\Omega^q(G/K)^G\cong\hom_{\mathfrak k}(\Lambda^q\mathfrak{g/k},\mathbb{R})$$

where $\mathfrak g$ and $\mathfrak k$ denote the Lie Algebras of $G$ and $K$.

I understand that a $G$-invariant form is determined by its values at the identity and so the isomorphism is given by mapping $\omega$ to $\omega_{eK}$.

But what does $\hom_{\mathfrak k}$ mean? Maybe $\mathfrak k$-equivariant? But what is the $\mathfrak k$-action and and why isn't it just $\mathbb R$-linear homomorphisms?


It should be $\operatorname{Hom}(\Lambda^q(\mathfrak{g}/\mathfrak{k}),\mathbb{R})^K$, invariant under $K$. It isn't just the $\mathbb{R}$-linear stuff. Imagine $G$ is the rotation group of the sphere, $K$ the subgroup fixing the north pole. Then to be invariant under $G$, you need to invariant under all transformations of $G$, in particular under all transformations in $K$. But these turn the north pole around, so you need to be invariant under the action of $K$ on the tangent plane at the north pole. If $K$ is connected, then $K$-invariance is the same as being null for the $\mathfrak{k}$-action, so you can write it as $\operatorname{Hom}(\Lambda^q(\mathfrak{g}/\mathfrak{k}),\mathbb{R})^{\mathfrak{k}}$.

| cite | improve this answer | |
  • $\begingroup$ Very helpful, thanks. Could you maybe state the $\mathfrak k$-action explicitly? $\endgroup$ – Peter Aug 28 '13 at 18:13
  • 1
    $\begingroup$ Get $G$ to act on its Lie algebra $\mathfrak{g}$ by the adjoint action. Restrict this from $G$ to $K$ to get an action of $K$ on $\mathfrak{g}$, which preserves $\mathfrak{k}$, and so induces an action on the quotient $\mathfrak{g}/\mathfrak{k}$. Call this action $\rho$. Every action of $K$ induces an action of $\mathfrak{k}$ by differentiation: define $\rho(A)$ for $A \in \mathfrak{k}$ by $e^{\rho(A)}=\rho(e^A)$. $\endgroup$ – Ben McKay Aug 29 '13 at 6:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.