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

We define the affine Grassmannian to be the quotient $Gr = GL_n(\mathbb{C}((t)))/GL_n(\mathbb{C}[[t]])$ where $\mathbb{C}((t))$ is the field of formal Laurent series and $\mathbb{C}[[t]]$ is the ring of formal power series. (The affine Grassmannian can be defined more generally, but here we restrict to a special case.) If we let $B$ be the Borel subgroup of upper triangular matrices in $GL_n(\mathbb{C})$, $T$ a maximal torus, then the Weyl group $W=N(T)/T$ is just $S_n$. Let $\widetilde{W} = \mathbb{Z}^{n-1} \rtimes W$ denote the affine Weyl group. Then for $i = 1, 2,...,n-1$ the affine permutations in $\widetilde{W}$ correspond to the usual permutation matrices in $GL_n(\mathbb{C})$, namely the identity matrix with columns $i$ and $i+1$ interchanged. The matrix for the affine permutation $s_0$ has ones along the diagonal in rows $2, 3, ..., n-1$, has $t$ in the right hand corner, and $t^{-1}$ in the bottom left corner. Let $I$ denote the Iwahori subgroup, that is, the inverse image of $B$ under the reduction map $GL_n(\mathbb{C}((t))) \rightarrow GL_n(\mathbb{C})$. Then $I$ is the set of upper triangular matrices mod $t$. I read somewhere that $GL_n(\mathbb{C}((t)))$ has a decomposition

$GL_n(\mathbb{C}((t))) = \cup IwGL_n(\mathbb{C}[[t]])$

where $w$ varies across the affine permutation matrices. This decomposition is supposed to induce the Bruhat decomposition of the affine Grassmannian into Schubert cells. Now something here is wrong. Since $I \subset GL_n(\mathbb{C}[[t]])$, and the determinant of any affine permutation matrix is 1 or -1, we have that for any $w \in \widetilde{W}$ the determinant of any matrix in $IwGL_n(\mathbb{C}[[t]])$ has power series determinant, but the matrix

t^{-1} 0

0 t^{-1}

is in $GL_n(\mathbb{C}((t)))$ with determinant $t^{-2}$ and inverse

t 0

0 t

So, my question is, what is wrong here? What is the correct decomposition and indexing set?

share|cite|improve this question
up vote 3 down vote accepted

There are two discrepancies: first, the "strict" affine Bruhat decomposition applies to SL(n), not GL(n); second, the decomposition you're wanting, in the strict case, would be $G=\bigcup_w IwI$, that is, with Iwahori on both sides. (This would make it a disjoint union.)

To adapt to GL(n), the affine Weyl group needs to be enlarged to include the obvious extra diagonal elements.

This doesn't depend so much on $\mathbb C((t))$ and $\mathbb C[[t]]$, as it applies to the field of fractions of a (maybe complete) discrete valuation ring.

share|cite|improve this answer
OK, yes, I was thinking that the double cosets would be IwI, rather than IwK, but I was confused because I thought we were working over GL(n). So, if I want the indexing set to be affine permutation matrices, I need to work with SL(n); that gives Gr = U IwK, where K = SL(n, C[[t]]) and the indexing set is diagonal matrices in C((t)) with entries whose exponents are integers that sum to 0. I guess now the question is: what is the indexing set for the decomposition of GL(n)? – Mehta Jan 5 '12 at 0:05
I didn't see where you defined affine permutation matrix, but anyway, it should be a monomial matrix with powers of t. If you want the $IwK$ decomposition (which I would call $IwP$, but never mind) to be disjoint, let $w$ only vary over $diag(t^{d_1}, t^{d_2}, \ldots)$ with $(d_1 \geq d_2 \geq \ldots d_n) \in \integers^n$. – Allen Knutson Jan 16 '12 at 23:25

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.