Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Let $G$ be your favorite complex semi-simple algebraic group, and let $B\supset T$ be your favorite Borus. For any $w\in W$, we have the Bruhat cell $BwB$, and its closure $\overline{BwB}$.

Now, it's very easy to write down some functions that cut out this variety. Let $V$ be any finite-dimensional representation of highest weight $\lambda$, and let $v$ be highest weight vector, and $\delta$ a non-zero functional killing all but the highest weight space. Then the generalized minors $\omega_{w'\delta,v}(g)=w'\delta(gv)$ $\Delta_{w'\delta,v}(g)=w'\delta(gv)$ for all $w'$ with $w'\lambda > w\lambda$ all vanish on $BwB$ (since $BwBv$ in contained in the sum of weight spaces $\leq w\lambda$), and on no Bruhat cells $Bw'B$ with $w'>w$.

That is, the radical of the ideal generated by these functions is all functions vanishing on $\overline{BwB}$. In fact, it's enough to just consider $\lambda$ fundamental to get an ideal with the correct radical. So, my question is:

Is the ideal generated by these generalized minors already radical?

share|improve this question
9  
+1 for "Borus." –  Moosbrugger Oct 5 '11 at 2:23
    
Thanks. I think. –  Ben Webster Oct 5 '11 at 3:38
add comment

2 Answers

First of all, I think you need to write $w'\lambda< w\lambda$ (look what happens when $w$ is 1).

It seems to me that when $G$ is not $SL(n)$ the answer is no. For example assume that $w=1$. Then you know that the relations are generated by all matrix coefficients $\omega_{\eta,v}$ (your notations) where $\eta$ is a functional which vanishes on the lowest weight vector of $V$ and your generators correspond to $\eta$ being an extremal weight vector. But if the fundamental representations of $G$ are not minuscule I don't see how you get relations with $\eta$ not being an extremal weight vector in $V^*$ (for fundamental $V$) from those with extremal $\eta$ - this seems impossible for degree reasons (if you introduce the multigrading corresponding to $\lambda$).

share|improve this answer
    
Not sure where my brain was when I wrote $\omega$ instead of $\Delta$. Just writing too fast. –  Ben Webster Oct 5 '11 at 14:39
    
I'm also not sure I follow. What's the grading the corresponding to $\lambda$? Shouldn't one pick a coweight rather than a weight to get a grading? –  Ben Webster Oct 5 '11 at 14:48
    
By the way, I think the fact I really wanted was the Theorem from Ramanathan you linked to in your deleted answer. Thanks. –  Ben Webster Oct 5 '11 at 14:54
    
OK, good. That's what seemed to me at the first glance, but now it seems to me that the way you actually formulated the question is different. Ramanathan says that the ideal of a Schubert variety is generated by elements of degree 1 in the homogeneous ring of the flag variety, whereas in your question you are talking about some particular elements of degree 1. –  Alexander Braverman Oct 5 '11 at 17:09
    
Right, but one doesn't always ask the question one actually wanted to know the answer to. –  Ben Webster Oct 6 '11 at 4:42
add comment

Exactly as Alexander said, this will fail for $G/P$ nonminuscule. The smallest example is the closed orbit $SO(5)/P$ of $SO(5)$ acting on ${\mathbb P}({\mathbb C}^4)$. The $T$-weight diagram of this representation is

.1.

111

.1.

The representation arises as the space of sections of ${\mathcal O}(1)$ on $SO(5)/P$. If we take the space of sections over the Schubert point $P/P$, we get just the top $1$. The extremal weight vectors you want to kill correspond to the left, right, and bottom $1$s. But to get the Schubert point on the nose, you have to kill the $1$ in the middle too.

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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