---
title: Extended Weak Order in Coxeter and Lattice Theory
url: https://www.emergentmind.com/topics/extended-weak-order
type: topic
---

# Extended Weak Order in Coxeter and Lattice Theory

Extended weak order denotes a family of lattice-theoretic enlargements of the classical weak Bruhat order, all designed to retain inversion-set control while enlarging the underlying class of objects. In the Coxeter-theoretic sense introduced by Dyer, the extended weak order is the inclusion poset of biclosed subsets of \(\Phi^+\), the positive roots of a Coxeter system; its finite elements are exactly the inversion sets of group elements, so it contains the usual weak order as an order ideal and coincides with it in finite type [1108.5557]. In parallel, finite Coxeter theory admits the facial weak order on standard parabolic cosets, extending weak order from vertices to all faces of the permutahedron [1602.03158]. Related constructions extend weak-order methods to central hyperplane arrangements [1910.03511], all subsets or antisymmetric closed subsets of a finite root system [1804.06572], reflexive binary relations and integer posets [1701.07995], continuous monotone paths in \(I^d\) [1812.02329], and alternating sign matrices together with their monotone-triangle and bumpless-pipe-dream models [2502.19266, 2606.13518]. The common structural themes are inversion-set characterizations, closure/interior operations, congruence transport, and the search for global lattice structure.

## 1. Classical weak order and the need for extension

For a Coxeter system \((W,S)\) with positive roots \(\Phi^+\), the classical right weak order is characterized by inclusion of inversion sets. In one standard formulation, for \(w\in W\),
\[
P_w=\{\alpha\in\Phi^+\mid \ell(s_\alpha w)<\ell(w)\},
\]
and
\[
x\le y \quad\Longleftrightarrow\quad P_x\subseteq P_y.
\]
Björner’s result gives weak order as a complete meet semilattice for arbitrary Coxeter groups; when \(W\) is finite, it is a lattice [1108.5557].

The fundamental obstruction in infinite type is the failure of joins for arbitrary pairs. Dyer’s extension replaces finite inversion sets by all biclosed subsets of \(\Phi^+\), thereby enlarging the poset while preserving the inversion-set model [1108.5557]. A distinct but complementary extension in finite type replaces group elements by faces of the Coxeter permutahedron; the resulting facial weak order restricts to the ordinary weak order on vertices [1602.03158].

The literature therefore does not use “extended weak order” for a single universal object. Rather, it names several closely related constructions, each preserving weak-order behavior in a larger category. A common misconception is to identify Dyer’s biclosed-set poset with the facial weak order. They are related by shared inversion-set and lattice phenomena, but they are defined on different underlying sets: biclosed root subsets in one case, standard parabolic cosets or faces in the other [1108.5557, 1602.03158].

## 2. Biclosed subsets of roots and Dyer’s extended weak order

Let \(X\subseteq\Phi^+\). In Dyer’s framework, a subset is closed if it is stable under the relevant rank-2 root closure, coclosed if its complement is closed, and biclosed if it is both closed and coclosed. In the 2-closure language, the extended weak order is
\[
\mathcal B(\Phi^+)=\{\,I\subseteq\Phi^+\mid I\text{ is 2-biclosed}\,\},
\]
ordered by inclusion [1108.5557]. In the equivalent root-sum formulation used in later work, a subset \(A\subseteq\Phi^+\) is closed if it is closed under root-sums in the usual sense, coclosed if its complement is closed, and biclosed if both hold; the resulting poset \(B\) of biclosed subsets, ordered by inclusion, is again called the extended weak order [2510.11446].

A foundational fact is that the finite biclosed subsets of \(\Phi^+\) are precisely the inversion sets of elements of \(W\). Thus \(W\) embeds into \(\mathcal B(\Phi^+)\), and in finite type every biclosed set is of the form \(\Delta(w)\) for a unique \(w\in W\), so the extended weak order reduces to the ordinary weak order [1108.5557, 2510.11446].

Dyer’s central conjecture is that the extended weak order is always a complete lattice. In affine type this is now established: for an untwisted affine Coxeter group, \(\bigl(\mathcal B(\Phi),\subseteq\bigr)\) is a complete lattice, with
\[
\bigvee_{i\in I} B_i=\overline{\bigcup_{i\in I} B_i},
\qquad
\bigwedge_{i\in I} B_i=\Bigl(\bigcap_{i\in I} B_i\Bigr)^\circ,
\]
where \(\overline{(-)}\) is biclosed closure and \((-)^\circ\) biclosed interior [2311.05737]. Barkley’s affine-symmetric-group model strengthens the structural picture: for \(W=\widetilde S_n\), the extended weak order \(Dyer(W)\) is a quotient of the lattice \(L_n\) of translation-invariant total orderings of \(\mathbb Z\); \(L_n\) is profinite, algebraic, and completely semidistributive, and \(Dyer(\widetilde S_n)\) inherits these properties [2502.05875].

The conjectural join formula is more refined. Dyer’s second conjecture describes the join through an algebraic-geometric construction on reflections. In finite type this becomes a statement about the left-reflection set \(T_L(u\vee v)\). It is proved in types \(I_2(m)\) and \(A_{n-1}\), and computationally verified in types \(H_3\) and \(F_4\) using Sage [2510.11446].

## 3. Facial weak order on permutahedra and arrangements

For a finite Coxeter system \((W,S)\), the faces of the permutahedron \(\Perm(W)\) are indexed by standard parabolic cosets
\[
P_W=\{\,xW_I\mid x\in W^I,\ I\subseteq S\}.
\]
Palacios–Ronco’s facial weak order is generated by two cover types:
\[
xW_I\lessdot xW_{I\cup\{s\}}
\quad\text{if }s\notin I\text{ and }x\in W^{I\cup\{s\}},
\]
and
\[
xW_I\lessdot xsW_{I\setminus\{s\}}
\quad\text{if }s\in I.
\]
Under \(w\mapsto wW_\varnothing\), ordinary weak order appears as the induced suborder on the vertices [1602.03158].

Two equivalent characterizations are central. The geometric one uses the root inversion set
\[
\mathbf R(xW_I)=x(\Phi^-\cup\Phi_I^+)\subseteq\Phi,
\]
with
\[
xW_I\le yW_J
\Longleftrightarrow
\mathbf R(xW_I)\cap\Phi^+\subseteq \mathbf R(yW_J)\cap\Phi^+
\text{ and }
\mathbf R(xW_I)\cap\Phi^-\supseteq \mathbf R(yW_J)\cap\Phi^-.
\]
The combinatorial one uses minimal and maximal coset representatives:
\[
xW_I\le yW_J
\Longleftrightarrow
x^I\le y^J
\text{ and }
x_I\le y_J
\]
in ordinary weak order on \(W\) [1602.03158].

These descriptions imply that the facial weak order is a lattice. For \(X=xW_I\) and \(Y=yW_J\),
\[
X\wedge Y=z_\wedge W_{K_\wedge},
\qquad
X\vee Y=z_\vee W_{K_\vee},
\]
where
\[
z_\wedge=x\wedge y,\quad K_\wedge=D_L\bigl(z_\wedge^{-1}(x_Iy_J)\bigr),
\]
and
\[
z_\vee=x_I\vee y_J,\quad K_\vee=D_L\bigl(z_\vee^{-1}(xy)\bigr).
\]
Moreover, every lattice congruence of the ordinary weak order induces a lattice congruence of the facial weak order; geometrically, the induced classes are the cones of the fan obtained by gluing chambers of the Coxeter fan that belong to the same weak-order congruence class. The descent congruence yields the facial boolean lattice on the faces of the cube, and Reading’s \(c\)-Cambrian congruence yields the facial \(c\)-Cambrian lattice on the faces of the generalized associahedron [1602.03158].

Dermenjian, Hohlweg, McConville, and Pilaud extended the facial weak order from finite Coxeter groups to essential central hyperplane arrangements. For a chosen base region \(B\), the facial weak order \(\FW(A,B)\) on all faces has four equivalent descriptions: via intervals in the poset of regions, via cover relations, via oriented-matroid covectors, and via root-inversion-type sets \(R(F)\) related to the corresponding zonotope. When the poset of regions is a lattice, \(\FW(A,B)\) is a lattice; for simplicial arrangements it is semidistributive, with join-irreducibles characterized by
\[
F\in\JIrr(\FW)
\Longleftrightarrow
M_F\in\JIrr(\Reg(A))
\text{ and }
\codim(F)\in\{0,1\}.
\]
Topologically, every open interval is either contractible or homotopy-equivalent to a sphere, and the Möbius function is correspondingly either \(0\) or a sign [1910.03511].

## 4. Ambient lattices on root subsets, relations, and continuous paths

One important line of work constructs weak-order-like ambient lattices large enough to contain classical objects as sublattices. For a finite crystallographic root system \(\Phi\), Gay and Pilaud define an extended weak order on all subsets \(R(\Phi)=2^\Phi\) by
\[
R\preccurlyeq S
\quad\Longleftrightarrow\quad
R^+\supseteq S^+
\text{ and }
R^-\subseteq S^-.
\]
This is a lattice, with
\[
R\wedge S=(R^+\cup S^+)\sqcup(R^-\cap S^-),
\qquad
R\vee S=(R^+\cap S^+)\sqcup(R^-\cup S^-).
\]
For crystallographic root systems, the induced subposet on antisymmetric closed subsets of roots, the \(\Phi\)-posets, is again a lattice. This ambient lattice contains sublattices corresponding to elements, intervals, and faces of the permutahedron, as well as Cambrian and Boolean analogues associated with generalized associahedra and cubes [1804.06572].

Chatel, Pilaud, and Pons establish an analogous extension on all reflexive binary relations on \([n]\). If \(R^{Inc}\) and \(R^{Dec}\) denote the increasing and decreasing parts of a relation \(R\), then
\[
R\le S
\quad\Longleftrightarrow\quad
R^{Inc}\supseteq S^{Inc}
\text{ and }
R^{Dec}\subseteq S^{Dec},
\]
with meet and join given by the corresponding Boolean formulas. The restriction to integer posets is again a lattice after appropriate transitive deletion and closure operations. This construction recovers classical weak order on permutations and induced sublattices for intervals and faces of permutahedra and associahedra [1701.07995].

Santocanale’s continuous weak order extends multinomial lattices from discrete words to images of continuous monotone paths in the cube \(I^d\). If \(C_f\) and \(C_g\) are the images of coordinate-wise monotone continuous maps \(f,g:I\to I^d\) with common endpoints, then
\[
C_f\le C_g
\quad\Longleftrightarrow\quad
f(t)\le g(t)\text{ coordinate-wise for all }t\in I.
\]
The resulting lattice \(L(I^d)\) is complete and self-dual, fails distributivity when \(d\ge 3\), and has join-irreducibles parameterized by one-step maps \(j_{x,y}\) and points \(p\in(0,1)^d\). Every element is an, possibly infinite, join of these join-irreducibles, but there are no completely join-irreducible or compact elements. The lattice is the Dedekind–MacNeille completion of the directed colimit of finite multinomial lattices arising from subdivisions of \(I\) [1812.02329].

## 5. Alternating sign matrices, monotone triangles, and weak-order operators

Hamaker–Reiner’s extension of weak order to alternating sign matrices is encoded by operators \(\pi_i:\ASM(n)\to\ASM(n)\) that minimally modify row \(i\) while preserving the other corner sums. A cover in ASM weak order is
\[
B\prec A
\quad\Longleftrightarrow\quad
\pi_i(A)=B\neq A
\]
for some \(i\). Escobar, Klein, and Weigandt place this order on algebro-geometric footing. If \(X_A\) is the ASM variety indexed by \(A\), then
\[
\codim X_A
=
\min\{\,k:\text{ there exists a saturated weak chain of length }k\text{ from }e\text{ to }A\,\}.
\]
They also prove compatibility with \(K\)-theoretic divided difference operators: for the double ASM Grothendieck polynomial \(G_A(x,y)\),
\[
\pi_i(G_A)=G_{\pi_i(A)}.
\]
A derivative formula involving \(\nabla\), \(E\), and column descents extends the usual Grothendieck-polynomial recurrence, and the whole formalism generalizes to arbitrary unions of matrix Schubert varieties indexed by antichains in strong order [2502.19266].

The later compatibility theorem identifies this ASM weak order with the Hamaker–Reiner weak order on monotone triangles under the standard bijection \(A\mapsto m_A\):
\[
\pi_i^{\MT}(m_A)=m_{\pi_i^A(A)}.
\]
The same cover relations admit three explicit realizations. On ASMs, a cover occurs exactly when there is an essential cell in row \(i\), and it is produced by toggling a \(2\times 2\) submatrix. On monotone triangles, it is obtained by lowering one essential entry by \(1\). On bumpless pipe dreams, it is realized by a simultaneous undroop of all liftable intervals in row \(i\). For fixed \(i\), each fiber
\[
\Fib_i(A)=\{\,C\in\ASM(n):\pi_i(C)=\pi_i(A)\,\}
\]
is a sublattice of strong Bruhat order on \(\ASM(n)\) [2606.13518].

These results show that weak-order operators can govern both combinatorial covering relations and algebro-geometric invariants. This suggests a broader principle: extensions of weak order are often most effective when the order, the operator calculus, and the geometry of the corresponding varieties are compatible.

## 6. Structural themes, current status, and open problems

Several structural properties recur across the various extensions. First, inversion-set or covector models remain decisive: biclosed subsets of \(\Phi^+\) for Dyer’s order, root inversion sets \(\mathbf R(xW_I)\) for facial weak order, sign-vectors and zonotope rays for arrangements, corner sums for ASMs, and join-continuous coordinate functions for continuous paths [1108.5557, 1602.03158, 1910.03511, 2502.19266, 1812.02329]. Second, closure/interior operations typically encode join and meet, whether as 2-closure of unions, biclosed closure and interior, or analogous deletion/closure procedures in induced sublattices [1108.5557, 2311.05737, 1701.07995]. Third, congruence transport is pervasive: weak-order congruences extend to facial weak order, while quotient constructions in other settings preserve lattice-theoretic information [1602.03158].

The present state of Dyer’s conjecture is sharply stratified. Finite type is classical. Affine type is established via clean arrangements and suitable total orders on \(\Phi^+\) [2311.05737]. For the affine symmetric group, the quotient model through translation-invariant total orders gives additional profinite and semidistributive structure [2502.05875]. The rank-3 universal Coxeter group provides the first non-spherical, non-affine case where the conjecture is proved: every biclosed set is weakly separable by a real hyperplane, implying that the extended weak order is a lattice [2509.00871].

The open problems are correspondingly precise. Dyer’s general lattice-property conjecture remains open for arbitrary Coxeter systems [2510.11446]. The join-formula conjecture is proved in types \(I_2(m)\) and \(A_{n-1}\), verified computationally in \(H_3\) and \(F_4\), and open in further types including \(D_n\), \(E_6\), \(E_7\), and \(E_8\) [2510.11446]. A plausible implication is that methods based on separability, clean arrangements, or profinite semidistributive models may continue to propagate to broader classes, but that extrapolation is not yet a theorem in the cited sources.

A second misconception is that once a weak order is extended to a lattice, it should also be distributive. The continuous weak order explicitly shows otherwise: \(L(I^d)\) is self-dual and complete, but when \(d\ge 3\) it is not distributive [1812.02329]. The shared conclusion across the literature is therefore not distributivity, but the more flexible package of lattice existence, semidistributivity in special cases, explicit irreducibles, and geometric realization.

In this broader sense, extended weak order has become a unifying framework for relating Coxeter combinatorics, hyperplane arrangements, root-system closure operators, generalized permutahedral geometry, and algebro-geometric models such as ASM varieties. The term now denotes not merely one poset, but an active program of extending weak-order methods to larger combinatorial and geometric universes while preserving the inversion-theoretic backbone that makes weak order tractable.

Source: https://www.emergentmind.com/topics/extended-weak-order