6
$\begingroup$

Let $\left(W\text{, }S\right)$ be a Coxeter system. For every $w\in W$ let us write $\left|w\right|$ for the length of $w$. Set $\lambda\left(e\right)=1$ where $e\in W$ denotes the neutral element of the group and define coefficients $\lambda\left(w\right)\in\mathbb{Z}$ recursively via \begin{eqnarray} \sum_{v\in W \text{: }\left|v^{-1}w\right|=\left|w\right|-\left|v\right|}\lambda\left(v\right)=0 \end{eqnarray} for any $w\in W$. For example this gives us $\lambda\left(e\right)=1$, $\lambda\left(s\right)=\left(-1\right)$ for all $s\in S$, $\lambda\left(st\right)=1$ if $m_{st}=2$ and $\lambda\left(st\right)=0$ if $m_{st}\neq2$ (for the notation see Wikipedia) for all $s\text{, }t\in S$ with $s\neq t$, and so on.

My question is: Does there always exist some $l\in\mathbb{N}$ such that $\lambda\left(w\right)=0$ for all $w\in W$ with $\left|w\right|\geq l$?

In the case that $\left(W\text{, }S\right)$ is right-angled (i.e. $m_{st}\in\left\{ 2\text{, }\infty\right\}$ for all $s\text{, }t\in S$ with $s\neq t$) this is true and we can take $l=\left|S\right|$. I'm wondering if in the general case this property also holds. If no, is it possible to characterize the property in an instructive way?

$\endgroup$
8
$\begingroup$

The answer is yes. I claim there are at most $2^n$ elements of $W$ for which $\lambda(w) \neq 0$. Specifically, let $\vee$ be the join in weak order, which is a semi-lattice. If $\{ s_1, s_2, \ldots, s_j \} \subseteq S$ and $s_1 \vee s_2 \vee \cdots \vee s_j$ is defined, then I claim that $\lambda(s_1 \vee s_2 \vee \cdots \vee s_j)=(-1)^j$, and otherwise I claim that $\lambda(w)=0$. Thus $N$ can be taken to be the greatest length of $s_1 \vee s_2 \vee \cdots \vee s_j$, restricting ourselves to cases where this join is defined.

The condition $|v^{-1} w| = |w| - |v|$ says that $v \leq w$ in weak order. Therefore, this is the recursion defining the Mobius function of weak order. The Mobius function of a finite lattice $L$ can be computed by Rota's crosscut theorem. (The first online reference I could find was Theorem 1.3 here.) Namely, let $A$ be the set of minimal elements of $L$ -- in this case, this is the set $S$. Then $$\mu(x) = \sum_{B \subseteq A,\ \bigvee B = x} (-1)^{|B|}.$$

So $\mu(w)=0$ if $w$ is not a join of elements of $S$. If $w$ is a join of elements of $S$, then it is so in only one way (this is a way that weak order is simpler than a general lattice) so we get the description from the first paragraph.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Thanks for your reply. Great answer! $\endgroup$ – worldreporter14 Jun 24 '19 at 9:34

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.