MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $U$ be the set of unipotent upper triangular matrices and $B$ the upper triangular matrices of $GL_3$. How could I describe $GL_3/U$ ? Using coordinates, in a projective or an affine space.

For example, I already know the identification of $SL_2/U$ with $\mathbb{A}^2 \setminus (0,0)$ and the identification of $SL_2/B$ with $\mathbb{P}^1$ ($B,U$ the standard Borel of $SL_2$ and its unipotent radical).

Thank you.

share|cite|improve this question
GL/B - it is "flag variety". Do you really need its description in coordinates ? It is projective variety, however no one usually thinks about in terms of explicit equations... The open dense cell is lower triangular matrices - so you can easyly get coordiantes on this cell – – Alexander Chervov Jan 31 '12 at 17:38
SL_2/U is not A^2, it is A^2 minus the origin. – Peter McNamara Jan 31 '12 at 17:59
You're both right, I knew it is a projective variety, but as I'm trying to work out a proof in the setting of G= GL_3, I'm willing to explicit the construction with coordinates. – Jan 31 '12 at 18:34
Notice that $B = \mathbb{C}^\times \cdot U$ and so $G/B = (G/U)/\mathbb{C}^\times$. This is one way to see that $SL_2/U = \mathbb{A}^2 - (0,0)$ given that $SL_2/B$ is $\mathbb{P}^1$. I imagine you should also be able to realize $GL_3/U$ as the affine cone over the flag variety $GL_3/B = SL_3/(B\cap SL_3)$. One way to realize the latter as a projective variety is to look at the adjoint action of $SL_3$ on its Lie algebra and look at the orbit of a highest weight. – solbap Jan 31 '12 at 19:04
up vote 3 down vote accepted

Let $V$ be the basic (3-dimensional) representation of $GL(3)$. Then $SL(3)/U$ is the set of all pairs $x\in V, y\in V^*$ where $x$ and $y$ are non-zero and $(x,y)=0$.

The quotient $GL(3)/U$ is non-canonically product of the above by $C^{\times}$. Canonically, you need to choose non-zero $x_i\in \Lambda^i(V)$ (for $i=1,2,3$) such that $x_i\wedge x_j=0$ for all $i$ and $j$ (note that if $x_3$ is fixed then $\Lambda^2(V)$ is canonically the same as $V^*$). This description generalizes immediately to any $GL(n)$ (and in fact to any $G$).

share|cite|improve this answer
I don't understand the first identification of $SL_3/U$, could explain it ? Moreover, do you have some kind of references for the generalization to $GL_n$ or any $G$ ? – Jan 31 '12 at 20:23
Well, you just consider a pair (vector,covector) whose pairing is 0 (both should be assumed to be non-zero, otherwise you get not $SL(3)/U$ but its affine closure). There so many references that I am not even sure where to refer you... The statement is that to specify a point of $G/U$ (for simply connected $G$) you must specify a vector in every fundamental representation of $G$ subject to certain quadratic relations (called Plucker relations). For $GL(n)$ the fundamental representations are the wedge powers of the basic $n$-dimensional representation. – Alexander Braverman Jan 31 '12 at 20:45
The identification $SL_3/U$ with {(vector $\vec v$, 2-tensor $\tau$) : $\vec v \wedge \tau = 0$} is given by $M \mapsto$ (first column of $M$, first two columns wedged together). – Allen Knutson Feb 1 '12 at 1:35

As already mentioned, the quotient $GL_n(\mathbb{C})/B$ is the space $Fl$ of all flags ($0\subset$ a line $\subset$ a 2-plane $\subset\cdots\subset \mathbb{C}^n$). There are line bundles $L_i , i=1,\ldots, n$ on this space where $L_i$ is obtained by pulling back the tautological bundles from Grassmannians of $i$- and $i-1$-planes in $\mathbb{C}^n$ and quotienting one by the other.

Let $L_i^0$ be the total space of $L_i$ minus the zero section and set $L^0$ to be the fibered product $$L_1^0\times_{Fl}L_2^0\times_{Fl}\cdots\times_{Fl}L^0_n.$$ The group $GL_n(\mathbb{C})$ acts transitively on the space $L^0$ and each stabilizer is conjugate to the group of unipotent matrices in $GL_n(\mathbb{C})$. So $GL_n(\mathbb{C})/\mbox{unipotent matrices}\cong L_0$.

share|cite|improve this answer

Let $G$ over a field $k$ be a reductive group and $P$ be parabolic subgroup with Levi decomposition $P =MU$ (over $k$), then $$G / U = \amalg_{w \in N(M) /M} UMwU,$$ where $N_G(M)$ is the normalizer of $M$ in $G$.

Since $MU = UM$ and $Mw = wM$ and $wM \cap M = wU \cap U = \emptyset$ for $w \neq 1$, we encounter that $$ G/U = M \amalg \; \coprod\limits_{w\neq 1} UMw$$ and $$G/P = 1 \amalg \;\coprod\limits_{w\neq 1} Uw.$$

I hope this is clearer now.

share|cite|improve this answer
Actually, I was looking for a more explicit identification, for example, I was able to embed $GL_3/B$ in $\mathbb{P}^2 \times \mathbb{P}^2$, so I was looking for something like that. But I'll see if I can get something from the Bruhat decomposition. Thank you. – Jan 31 '12 at 18:37

Your Answer


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.