This is the spin-off of the question I previously asked.

First, let me remind you some notation from that question:

$G_0$ - compact, simply connected Lie group giving rise (by complexification) to a semi-simple Lie Group $G$

$\mathfrak{g}$ - Lie algebra of $G$

$\delta=\frac{1}{2}\sum_{\alpha>0}\alpha$ (summation over all possitive roots of $\mathfrak{g}$)

$V_{\lambda}$ - finite dimensional complex vector space on which $\mathfrak{g}$ is irreducibly represented (with the highest weight $\lambda$ and highest weight vector $v_\lambda$).

$\mathbb{P}V_\lambda$ - complex projective space of $V_\lambda$

$\pi: V_\lambda\rightarrow \mathbb{P}V_\lambda$ - mapping onto the projective space

$O_{v_{\lambda}}^0 $ - orbit of $G_0$ through $v_\lambda$

$\Omega:V_\lambda\otimes V_\lambda \rightarrow V_\lambda \otimes V_\lambda$ - second order Casimir represented on $V_\lambda\otimes V_\lambda$

Now, my the true question starts:

In the article by Lichteinstein ( check here )it is proven that the necessary and sufficient condition for $\pi(v)$ to be in $\pi(O_{v_{\lambda}}^0)$ is the following:

$\Omega (v\otimes v) =\langle 2\lambda+2\delta,2\lambda\rangle (v\otimes v)$

In the above $\langle\cdot ,\cdot \rangle $ is a standard inner product on Cartan algebra dual $\mathfrak{h}^*$ .

I came up with the way of reasoning that might simplify arguments for this statement - they rely on a rather heavy (from my perspective) machinery. $\Omega$ decomposes into a direct sum of multiples of identity acting on irreducible representations $V_\alpha$ of $\mathfrak{g}$ on which $V_\lambda\otimes V_\lambda$ decomposes:


$\beta(\alpha)$ is some function of the weight $\alpha$- see above, $\mathcal{A}$ denotes some collection of weights - there may be some overlappings (if that is the case corresponding spaces $V_\alpha$ and $V'_\alpha$ are different). .

Now, it is easy to see that in the above sum there is only one $\alpha_0=2\lambda$ (it is the weight of the irreducible representation generated on $V_\lambda\otimes V_\lambda$ by $v_\lambda\otimes v_\lambda$ ). As $\Omega$ commutes with all unitary operators in the representation of $G_0$ (see notation above) in $V_\lambda \otimes V_\lambda$, clearly: $\Omega(v\otimes v)=\beta(2 \lambda) v\otimes v$ for $v\in O_{v_\lambda}^0$.

In order to have the complete characterization of $\pi(O_{v_\lambda}^0)$ I only have to know that $v \otimes v \in V_{2 \lambda}$ implies that $v\otimes v$ is proportional to some $\Phi(g) v_\lambda \otimes \Phi(g) v_\lambda$, where $\Phi$ - representation of $G_0$ in $V_\lambda$, $g\in G_0$.

One possibility would be to allow $G_0$ to act transitively on $\pi(V_{2\lambda})$ where $V_{2 \lambda}$ treated as a subspace of $V_\lambda\otimes V_\lambda$. Yet, this condition is not fulfilled for large dimensions of $V_\lambda$. Let me state the final question.

When $v \otimes v \in V_{2 \lambda}$ ( $V_{2 \lambda}$ is treated as as subspace of $V_{\lambda}\otimes V_\lambda$) implies that $v\otimes v$ is proportional to some $\Phi(g) v_\lambda \otimes \Phi(g) v_\lambda$?


1 Answer 1


This always holds by a theorem of Kostant. See Chapter 10, section 6.6 of Procesi's (wonderful) book.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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