---
title: Lehmer Complexes in Bruhat Intervals
url: https://www.emergentmind.com/topics/lehmer-complexes
type: topic
---

# Lehmer Complexes in Bruhat Intervals

Searching arXiv for the cited Lehmer-complex papers and closely related work.
Lehmer complexes are simplicial complexes attached to lower Bruhat intervals by transporting the interval through a Lehmer code to a multicomplex and then applying the Björner–Frankl–Stanley \(M\)-complex construction. For a finite Coxeter system \((W,S)\) admitting a Lehmer code, each lower interval \([e,w]\) determines an order ideal \(J_w\) in a product of chains, and the associated simplicial complex \(\Delta_w\) has \(f\)-polynomial equal to the Poincaré polynomial of the interval. In the framework developed by Bolognini–Sentinelli, these complexes are vertex-decomposable; in type \(A\) they encode the Poincaré polynomials of smooth Schubert varieties via unimodal permutations, while in type \(F_4\) ordinary Lehmer codes fail to exist but weak Lehmer codes still recover Lehmer complexes interval by interval [2501.03037], [2509.20981].

## 1. Lehmer codes and the Bruhat-order input

Let \((W,S)\) be a finite Coxeter system with exponents \(e_1,\dots,e_n\). A Lehmer code for \((W,S)\) is a bijection
\[
L:W\longrightarrow [e_1+1]_0\times\cdots\times [e_n+1]_0,
\]
where \([m]_0=\{0,1,2,\dots,m\}\), such that the inverse
\[
L^{-1}:\prod_{i=1}^n [e_i+1]_0\longrightarrow (W,<)
\]
is a poset morphism. Equivalently,
\[
L(u)\le L(v)\Longrightarrow u\le v
\]
in Bruhat order. In particular,
\[
\ell(w)=\sum_{i=1}^n L(w)_i.
\]
This condition is the key combinatorial mechanism behind Lehmer complexes: it turns Bruhat intervals into order ideals in a product order [2501.03037].

In type \(A_{n-1}\), that definition recovers the classical inversion-based Lehmer code
\[
\Lambda_n(w)=\bigl(0,\#\{\,i<2:w^{-1}(i)>w^{-1}(2)\},\,\dots,\,
\#\{\,i<n:w^{-1}(i)>w^{-1}(n)\}\bigr),
\]
a bijection onto \(\{0\}\times[1]_0\times\cdots\times[n-1]_0\) whose inverse preserves Bruhat order. Analogous explicit constructions were given for types \(B_n\), \(D_n\), and \(H_3\) by factoring each element into saturated Bruhat chains and reading off their lengths [2501.03037].

The interval of interest is always a lower Bruhat interval
\[
[e,w]=\{v\in W:v\le w\}.
\]
Once a Lehmer code exists, the interval becomes a combinatorial object in \(\mathbb N^n\), and that order-theoretic translation is what the Lehmer complex formalizes.

## 2. From lower Bruhat intervals to multicomplexes

Fix a Lehmer code \(L\). For each \(w\in W\), the lower interval is sent to
\[
J_w=\{\,L(v):v\le w\,\}\subset \prod_{i=1}^n [e_i+1]_0
\cong \mathbb N^n_{\le (e_1+1,\dots,e_n+1)}.
\]
Because the inverse of \(L\) is order-preserving, \(J_w\) is an order ideal. Order ideals in \(\mathbb N^n\) are exactly multicomplexes, so the Bruhat interval is converted into a multicomplex without losing the rank data [2501.03037].

Bolognini–Sentinelli then apply the Björner–Frankl–Stanley \(M\)-complex construction. For \(\mathbf d=(d_1,\dots,d_k)\), the complex \(M_{\mathbf d}\) is a pure, vertex-colored, shellable simplicial complex of dimension \(\sum_i d_i-1\). If \(J\subset [d_1]\times\cdots\times[d_k]\) is an order ideal, then \(M_{\mathbf d}(J)\) is its \(M\)-complex. In the Coxeter-theoretic setting,
\[
\mathbf d=(e_1+1,\dots,e_n+1),\qquad
\Delta_w=M_{\mathbf d}(J_w).
\]

Each face \(F\subset F_x\) of \(\Delta_w\) inherits a multidegree \(\deg(F)\) by counting how many vertices of each color occur in \(F\). The resulting \(f\)-polynomial is
\[
f_{\Delta_w}(t)=\sum_{F\in\Delta_w} t^{\deg(F)}
=\sum_{x\in J_w} t^{\sum_i x_i}
=\sum_{v\le w} t^{\ell(v)}
=P_{[e,w]}(t).
\]
Thus the Lehmer complex is not merely attached to the interval: its face enumeration reproduces the interval’s Poincaré, or rank-generating, polynomial exactly [2501.03037].

## 3. Polynomial identities and topological properties

A classical change of variables relates the \(f\)- and \(h\)-polynomials of a \(d\)-dimensional simplicial complex:
\[
h_\Delta(t)=(1-t)^d\,f_\Delta\!\bigl(\tfrac{t}{1-t}\bigr).
\]
Within the Lehmer-complex construction, one obtains the following statement: for every lower Bruhat interval \([e,w]\) in a finite Coxeter group admitting a Lehmer code,
\[
h_{\Delta_w}(t)=f_{\Delta_w}(t)=\sum_{v\le w} t^{\ell(v)},
\]
and moreover \(\Delta_w\) is vertex-decomposable, hence shellable [2501.03037].

The vertex-decomposability argument proceeds recursively. One peels off a highest-colored vertex whose deletion and link both remain \(M\)-complexes of smaller multicomplexes, and continues until the complex reduces to the void or a point. This gives an explicit inductive structure rather than a purely existential shelling argument [2501.03037].

The principal significance of this construction is that Poincaré polynomials of lower Bruhat intervals are realized as \(h\)-polynomials of vertex-decomposable simplicial complexes. In the terminology of the paper, the Lehmer complex is therefore a bridge between Bruhat-order combinatorics and a class of simplicial complexes with strong recursive and shellability properties.

## 4. Type \(A\), unimodal permutations, and smooth Schubert varieties

In type \(A\), Lehmer complexes interact directly with Schubert geometry. For \(S_n\), the Schubert variety \(X_w\subset GL_n/B\) is smooth if and only if \(w\) avoids the patterns \(3412\) and \(4231\). Its Poincaré polynomial factors as
\[
P_{[e,w]}(q)=\prod_{i=1}^r [\lambda_i(w)+1]_q,
\]
where \(\lambda(w)=(\lambda_1>\lambda_2>\cdots>\lambda_r)\) is the exponent partition of \(w\) [2501.03037].

A permutation \(u\in S_n\) is unimodal if there exists \(k\) such that
\[
u(1)<u(2)<\dots<u(k)
\quad\text{and}\quad
u(k)>u(k+1)>\dots>u(n).
\]
Equivalently, \(u\) avoids \(312\) and \(213\). Bolognini–Sentinelli prove that the set of Poincaré polynomials of all smooth Schubert varieties in \(GL_n/B\) coincides with
\[
\{\,P_{[e,u]}(q):u\text{ unimodal in }S_n\,\}.
\]
In particular there are \(2^{n-1}\) distinct such polynomials, in bijection with partitions \(\lambda_1>\cdots>\lambda_r\) of size \(\ell(u)\) having strictly decreasing parts [2501.03037].

The small-rank examples make the correspondence concrete. In type \(A_3\), the Lehmer complex of the full group \(S_4\) is \(M_{(1,2,3)}\), whose \(h\)-polynomial is
\[
[2]_q[3]_q[4]_q=(1+q)(1+q+q^2)(1+q+q^2+q^3).
\]
Among the unimodal permutations in \(S_4\), the element \(u=2143\) has Lehmer code \((0,1,1)\), exponent partition \((2,1)\), and
\[
P_{[e,2143]}(q)=[3]_q[2]_q=(1+q+q^2)(1+q).
\]
These examples illustrate that the Lehmer complex packages the same rank data that appears in Schubert-theoretic factorization formulas [2501.03037].

## 5. Type \(F_4\): failure of ordinary Lehmer codes and the weak replacement

Type \(F_4\) provides the main obstruction to a uniform ordinary theory. Let \((W,S)\) be the Coxeter system of type \(F_4\), with exponents \(1,5,7,11\). If \(w_0\) is the longest element, then
\[
h_{w_0}(q)=\sum_{v\le w_0} q^{\ell(v)}
=[2]_q[6]_q[8]_q[12]_q.
\]
Despite this product factorization, the Bruhat order of \(F_4\) does not admit a product of chains as subposet. Equivalently, there is no bijection
\[
L:W\longrightarrow [0,1]\times[0,5]\times[0,7]\times[0,11]
\]
whose inverse preserves Bruhat order. This answers negatively, in type \(F_4\), the question asked by Billey, Fan and Losonczy [2509.20981].

The proof strategy is combinatorial and obstruction-based. One checks that any order-preserving immersion would have to send the rank-\(6\) elements bijectively onto the \(264\) tuples in
\[
C(6)=\{(i,j,k,\ell)\in \{0,1\}\times\{0,\dots,5\}\times\{0,\dots,7\}\times\{0,\dots,11\}: i+j+k+\ell=6\}.
\]
Up to the automorphism group of \(F_4\), generated by inversion \(w\mapsto w^{-1}\) and the Dynkin-diagram symmetry \(s_i\mapsto s_{5-i}\), there are \(264\) such immersions. For each candidate, some rank-\(7\) tuple has a set of lower covers whose images do not admit a common upper cover in Bruhat order, so no consistent extension to level \(7\) exists [2509.20981].

To recover the interval-by-interval multicomplex construction, the paper introduces a weak Lehmer code. For a finite-rank Coxeter system with \(|S|=k\), a finite collection
\[
L^W=\{L_1,\dots,L_h\}
\]
of rank-preserving injections \(L_i:W\to\mathbb N^k\) is a weak Lehmer code if two conditions hold. First, whenever the set of maximal elements
\[
\max\{L_i(v):v\le w\}
\]
is a singleton, then \(\{L_i(v):v\le w\}\) is an order ideal in \(\mathbb N^k\). Second, for every \(w\in W\), some \(L_i\), up to automorphism of \((W,S)\), makes \(\{L_i(v):v\le w\}\) an order ideal. A classical Lehmer code is the special case \(h=1\), with bijectivity onto the full product of chains [2509.20981].

In type \(F_4\), exactly two maps \(L_1,L_2:F_4\to\mathbb N^4\) suffice. The construction uses a parabolic decomposition into saturated Bruhat chains
\[
X_1:\ e\vartriangleleft s_2,
\]
\[
X_2:\ e\vartriangleleft s_3\vartriangleleft s_3s_2\vartriangleleft s_3s_2s_3,
\]
\[
X_3:\ e\vartriangleleft s_1\vartriangleleft s_2s_1\vartriangleleft\cdots\vartriangleleft s_1s_2s_3s_2s_1,
\]
together with chains \(Y_1,Y_2\subset {}^{\{s_1,s_2,s_3\}}F_4\), each of length \(15\) or \(7\), satisfying
\[
F_4=X_3X_1X_2Y_1\;\bigsqcup\;X_3X_1X_2Y_2.
\]
For a factorization \(x=x_3x_1x_2u\) with \(x_i\in X_i\) and \(u\in Y_1\cup Y_2\),
\[
L_1(x)=
\begin{cases}
(\ell(x_1),\ell(x_3),\ell(x_2),\ell(u)), & u\in Y_1,\\
(\ell(x_1),\ell(x_3),\ell(u),\ell(x_2)), & u\in Y_2,
\end{cases}
\]
and
\[
L_2(x)=
\begin{cases}
(\ell(x_1),\ell(x_3),\ell(x_2),\ell(u)), & u\in Y_1,\\
(\ell(x_1),\ell(x_3),\ell(x_2)+4,\ell(u)-4), & u\in Y_2.
\end{cases}
\]
These maps are injective and rank-preserving, and for every lower interval one of them, up to automorphism, yields the required order ideal. The pair \(\{L_1,L_2\}\) is therefore a weak Lehmer code for \((F_4,S)\). Once
\[
J=\{L_i(v):v\le w\}\subseteq \mathbb N^4
\]
is an order ideal with \(f\)-polynomial \(h_w(q)\), the general machinery produces a balanced vertex-decomposable simplicial complex, called the Lehmer complex of \([e,w]\), whose \(h\)-polynomial is again \(h_w(q)\) [2509.20981].

## 6. Examples, further structure, and terminological scope

The \(F_4\) examples show the construction in explicit form. For
\[
w=s_4s_3s_1s_2s_3s_4s_2s_1,
\]
one has \(\ell(w)=8\), \(w\) is fixed by every Dynkin automorphism, and
\[
L_1(w)=(1,1,6,0).
\]
The lower interval \([e,w]\) has \(100\) elements. Its corresponding multicomplex \(J=\{L_1(v):v\le w\}\subset \mathbb N^4\) has maximal elements
\[
\{(0,5,1,1),(1,1,0,5),(1,1,1,4),(1,1,3,1),(1,1,6,0),(1,2,1,2),(1,3,1,1)\},
\]
and the associated Lehmer complex is a pure simplicial complex on seven vertices, one for each coatom, with \(100\) facets. For the shorter element
\[
w'=s_1s_2s_3s_4s_1,\qquad \ell(w')=5,
\]
one finds
\[
L_1(w')=(1,2,1,1),\qquad h_{w'}(q)=[2]_q^3[3]_q.
\]
Here \(J=\{L_1(v):v\le w'\}\) is an order ideal generated by the single maximal chain \((1,2,1,1)\), so \(J\) is isomorphic to \([0,1]\times[0,2]\times[0,1]\times[0,1]\), and the Lehmer complex is a direct product of four chains of lengths \(2,3,2,2\) [2509.20981].

The \(F_4\) theory also supplies additional order-theoretic structure. Every lower interval \([e,w]\subseteq F_4\) has rank sequence an \(M\)-sequence, realized by the weak code as the \(f\)-vector of an explicit multicomplex in \(\mathbb N^4\). The Lehmer complex of each interval is balanced and vertex-decomposable, hence Cohen–Macaulay, and its \(h\)-polynomial is exactly the Bruhat–Poincaré polynomial \(h_w(q)\). If \(w\) is \(L_i\)-principal, meaning that \(\max\{L_i(v):v\le w\}\) is a singleton, then the set \(\Pr(L_i)\) of \(L_i\)-principal elements is a meet-semilattice under Bruhat order, identified with the componentwise order on \(\{L_i(w):w\in \Pr(L_i)\}\subset \mathbb N^4\). Likewise, the \(L_i\)-unimodal elements form a lattice \(\mathcal U(L_i)\) isomorphic to a sublattice of \(\mathbb N^4\); in type \(F_4\), \(\mathcal U(L_1)\cup\{w_0\}\) is precisely in bijection with the set \(\Pal(F_4)\) of palindromic Poincaré polynomials of Schubert varieties, ordered by reverse inclusion of inversion sets [2509.20981].

A common terminological confusion arises from the word “complex.” In the combinatorial setting considered here, a Lehmer complex is a simplicial complex attached to a Bruhat interval. By contrast, the neural-network paper “Efficient and Interpretable Neural Networks Using Complex Lehmer Transform” studies a weighted Lehmer transform with complex-valued parameter \(s=a+ib\) and introduces complex Lehmer activation units; in that context, “complex” refers to the complex domain and to phase-sensitive behavior, not to a simplicial complex [2501.15223]. This distinction is purely terminological, but it separates two mathematically unrelated uses of the word “Lehmer” in current arXiv literature.

Source: https://www.emergentmind.com/topics/lehmer-complexes