Following Garrett's book, we have the usual description of affine buildings of SL(n) in terms of homothety classes of lattices. So let $F$ be a local field, $\mathcal{O}$ be its ring of integers and $\mathfrak{m} = <\pi>$ be its unique maximal ideal. Let $V = F^{4}$ and $G = SL(V)$. Let $e_{1}, \ldots, e_{4}$ be the standard basis vectors. Then one of the chambers is given by the following lattices: $$ L_{0} = \mathcal{O}e_{1} + \mathcal{O}e_{2} + \mathcal{O}e_{3} + \mathcal{O}e_{4},$$ $$ L_{1} = \mathcal{O}\pi^{-1}e_{1} + \mathcal{O}e_{2} + \mathcal{O}e_{3} + \mathcal{O}e_{4},$$ $$ L_{2} = \mathcal{O}\pi^{-1}e_{1} + \mathcal{O}\pi^{-1}e_{2} + \mathcal{O}e_{3} + \mathcal{O}e_{4},$$ $$ L_{3} = \mathcal{O}\pi^{-1}e_{1} + \mathcal{O}\pi^{-1}e_{2} + \mathcal{O}\pi^{-1}e_{3} + \mathcal{O}e_{4}. $$

The type of lattice $L_{i}$ is $i$ according to typing given in Garrett on page 329. Hence they are all of different type (which they have to be since they form a chamber).

Now $G$ acts in a type preserving manner (claimed in Garrett on Page $329-330$) on the building. However, I do not see it. Also, I seem to run into a problem with the following example.

Let $g \in G$ be given by $$ g = \pmatrix{1 & 0 & \pi^{-1} & 0 \\ 0 & 1 & 0 & 0 \\ 0 & 0 & 1 & 0 \\ 0 & 0 & 0 & 1 \\} $$

Then $gL_{0} = L_{1}$ which have different types. Where am I making the mistake?

EDIT: To see the last statement, observe that $gL_{0}$ is the lattice

$$ gL_{0} = \mathcal{O}(1+\pi^{-1})e_{1} + \mathcal{O}e_{2} + \mathcal{O}e_{3} + \mathcal{O}e_{4}.$$

Clearly, $gL_{0} \subseteq L_{1}$. To see the other inequality, we observe that $(1 + \pi)$ is a unit in the ring (otherwise $1$ belongs to the maximal ideal)and $(1+\pi)^{-1}(1+\pi^{-1}) = \pi^{-1}$.

  • 6
    $\begingroup$ I see that $gL_0$ is contained in $L_1$, but i don't see equality. $\endgroup$ – user1688 Aug 11 '16 at 10:03
  • 3
    $\begingroup$ Since $\det g=1$, you can't have $g L_0 = L_1$. $\endgroup$ – David E Speyer Aug 11 '16 at 12:15
  • 2
    $\begingroup$ It is also not equal up to homothety. Further, any base-change matrix from $L_0$ to $L_1$ must have determinant of absolute value $|\pi^{-1}|$. $\endgroup$ – user1688 Aug 12 '16 at 8:01
  • 3
    $\begingroup$ We have $gL_0 = \mathcal{O}e_1 + \mathcal{O}e_2+\mathcal{O}(e_3+\pi^{-1}e_1)+\mathcal{O}e_4$. The expression for $gL_0$ that you wrote is incorrect. $\endgroup$ – Aurel Aug 18 '16 at 10:22

The groups $G={\rm SL}(V)$ and $G'={\rm GL}(V) \supset G$ have the same affine building. If the action of $G$ is type preserving, the action of $G'$ is not! Your particular element $g$ lies in $G'$ but not in $G$.

If an element of $G'$ stabilizes a chamber globally then it stabilizes it pointwise. This is not true for $G$: given a chamber $C$, on can find an element $\Pi$ of $G'$ stablizing $C$ and inducing a cyclic permutation of the vertices of $C$.

More generally the action of a simply connected reductive group (like ${\rm SL}(N)$, ${\rm Sp}(2N)$, ...) on its building is type preserving. But this is not true in general.


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.