---
title: Weak Lehmer Codes in F4 and Beyond
url: https://www.emergentmind.com/topics/weak-lehmer-codes
type: topic
---

# Weak Lehmer Codes in F4 and Beyond

Searching arXiv for the cited papers to ground the article in current records.
Weak Lehmer codes are finite families of injective, rank-preserving encodings that replace the single global order-preserving product-of-chains parametrization required by a classical Lehmer code. In the classical setting of permutations, the Lehmer code records inversion counts and identifies $S_n$ with a rectangular product of chains. In more general Coxeter-theoretic settings, especially for lower Bruhat intervals, analogous factorizations of rank-generating functions often suggest such a structure, but that suggestion can fail at the level of order embedding. The paper "A weak Lehmer code for type $F_4$" establishes precisely this phenomenon for $F_4$: although the full Poincaré polynomial factors as a product of $q$-analogues, no global Lehmer code exists, yet a weaker local theory still yields multicomplexes and Lehmer complexes for every lower Bruhat interval [2509.20981].

## 1. Classical origin and the meaning of “weak”

For a permutation $\sigma \in S_n$, the classical Lehmer code is
$$
c(\sigma)=(c_1(\sigma),\dots,c_n(\sigma)), \qquad
c_i(\sigma)=|\{j>i:\sigma_j<\sigma_i\}|.
$$
It encodes the inversion set of $\sigma$ and gives a bijection
$$
S_n \to \prod_{i=1}^n [n-i]_0,
$$
where $[m]_0=\{0,1,\dots,m\}$. Under componentwise order on the target, the inverse map is order-preserving, so the image is a product-of-chains subposet whose rank-generating function matches that of $S_n$ [2509.20981].

This classical picture separates into two logically distinct features. One is numerical: the rank-generating function factors as a product of $q$-analogues. The other is order-theoretic: there is an order-preserving inverse from a rectangular box of code vectors into the ambient poset. The weak Lehmer code formalism retains the first kind of control locally on intervals while relaxing the second globally. In type $F_4$, this relaxation is necessary because the Bruhat order does not admit a product of chains as a subposet, even though the Poincaré polynomial factors [2509.20981].

A useful contrast appears in type $A$. In the symmetric group, several Lehmer-code-derived constructions model weak-order phenomena without giving a global code for the full weak order. Denoncourt showed that the Lehmer codes of permutations in a weak-order interval form a distributive lattice under the product order, and that its rank-generating function matches that of the interval [1102.2689]. In a different direction, the consecutive Lehmer code for parabolic quotients of $S_n$ is not injective on the full parabolic quotient, but on $(\alpha,231)$-avoiding elements it becomes bijective and componentwise order realizes the parabolic Tamari lattice [2009.05342]. This suggests that “weak” may mean either intervalwise distributive refinement or quotientwise order realization, depending on context. In type $F_4$, the weak Lehmer code of [2509.20981] is a further generalization: a finite family of codes suffices to realize every lower Bruhat interval as a multicomplex, possibly after precomposition by an automorphism.

## 2. The $F_4$ obstruction to a strong Lehmer code

Let $W(F_4)$ be the finite Weyl/Coxeter group of rank $4$ with simple reflections $S=\{s_1,s_2,s_3,s_4\}$ and Coxeter graph
$$
s_1 - s_2 -4- s_3 - s_4,
$$
where “$4$” denotes $m_{2,3}=4$. Bruhat order $\le$ on $W(F_4)$ is graded by the length function $\ell$, and for $w\in W(F_4)$ the lower interval $[e,w]$ is $\{v\in W:v\le w\}$ [2509.20981].

The exponents of $F_4$ are $\{1,5,7,11\}$. Writing $[k]_q:=\sum_{j=0}^{k-1} q^j$, the full group’s rank-generating function factors as
$$
P_{F_4}(q)=\sum_{w\in W(F_4)} q^{\ell(w)}
=[2]_q[6]_q[8]_q[12]_q.
$$
Such a factorization guarantees a rank-preserving bijection to a product of chains, but it does not by itself provide an order-preserving inverse. The negative result of [2509.20981] shows that this distinction is essential in type $F_4$.

The main theorem states that the Bruhat order of $W(F_4)$ does not admit a product of chains as a subposet; equivalently, $F_4$ does not admit a Lehmer code in the sense of an order-preserving bijection
$$
W(F_4)\to [1]_0\times[5]_0\times[7]_0\times[11]_0
$$
with rank compatibility [2509.20981]. This answers negatively, in type $F_4$, the question of Billey, Fan and Losonczy whether a rank-symmetric lower Bruhat interval admitting a product factorization of its rank-generating function also admits such a Lehmer code.

The obstruction is formulated through truncated rank layers. Let
$$
C=[1]_0\times[5]_0\times[7]_0\times[11]_0,\qquad
\rho(x)=x_1+x_2+x_3+x_4.
$$
For $k\in\mathbb N$, set
$$
C(k)=\{x\in C:\rho(x)\le k\},\qquad
F_4(k)=\{w\in W(F_4):\ell(w)\le k\}.
$$
Then
$$
\sum_{x\in C(k)} q^{\rho(x)}=\sum_{w\in F_4(k)} q^{\ell(w)}.
$$
The authors compute all injective poset morphisms $L^{-1}:C(6)\to F_4(6)$ and reduce them up to the automorphism group
$$
\mathrm{Aut}(F_4,\le)\cong \mathbb Z_2\times\mathbb Z_2,
$$
generated by inversion $\iota(w)=w^{-1}$ and the diagram automorphism $\psi(s_i)=s_{4-i+1}$ [2509.20981]. They then attempt to extend these immersions to rank $7$. For each coatom-set configuration arising from an element of rank $7$ in $C$, one must find a corresponding upper bound in $F_4(7)$; for every candidate immersion, some required configuration has no such upper bound. Therefore no order-preserving bijection onto $C$ exists.

A plausible implication is that factorizations of Poincaré polynomials in exceptional types should not be interpreted as evidence for global rectangular order models without additional structural input. In type $F_4$, the factorization survives, but the order-embedding fails.

## 3. Formal definition of weak Lehmer codes

Let $(W,S)$ be a Coxeter system of finite rank $k:=|S|$. A finite set
$$
L^W=\{L_1,\dots,L_h\}
$$
is a weak Lehmer code if the following conditions hold [2509.20981]:

1. Each $L_i:W\to \mathbb N^k$ is injective and rank-preserving, meaning that $\ell(w)$ is the sum of the coordinates of $L_i(w)$.

2. For any $w\in W$ and any $i$, if
$$
\left|\max\{L_i(v):v\le w\}\right|=1
$$
in the componentwise order, then
$$
\{L_i(v):v\le w\}
$$
is a multicomplex, that is, an order ideal of $\mathbb N^k$.

3. For every $w\in W$, there exist $i$ and an automorphism $\phi\in \mathrm{Aut}(W,\le)$ such that
$$
\{(L_i\circ\phi)(v):v\le w\}
$$
is a multicomplex.

This definition relaxes the single global product-of-chains embedding in two ways. First, it allows a finite family rather than one code. Second, it requires the multicomplex property only intervalwise, possibly after precomposition by an automorphism [2509.20981]. What is retained is sufficient for the rank-generating function of every lower Bruhat interval to be realized as the Hilbert series of a multicomplex and therefore to feed the canonical Lehmer-complex construction of [BS, Section 2], as quoted in [2509.20981].

The $F_4$ theory also introduces two derived classes of elements relative to a fixed $L\in\{L_1,L_2\}$. An element $w\in W$ is $L$-principal if
$$
\left|\max\{L(v):v\le w\}\right|=1.
$$
In that case,
$$
\{L(v):v\le w\}=\{x\in \mathbb N^4:x\le L(w)\},
$$
hence
$$
h_w(q)=\prod_{i=1}^4 [L(w)_i+1]_q.
$$
The set $\mathrm{Pr}(L)$ of $L$-principal elements forms a meet-semilattice isomorphic to its coordinate image under $L$ [2509.20981].

For $w\in \mathrm{Pr}(L)$, let
$$
O_w=\{v\in \mathrm{Pr}(L):h_v=h_w\}.
$$
Then $w$ is $L$-unimodal if $L(w)$ is the lexicographically minimal element in $\{L(v):v\in O_w\}$. The induced Bruhat order on $U(L)$ is isomorphic to componentwise order on $\{L(w):w\in U(L)\}$ [2509.20981]. These notions isolate strata on which the weak code behaves particularly rigidly.

## 4. Explicit construction in type $F_4$

The construction in [2509.20981] is based on a parabolic decomposition relative to
$$
J=\{s_1,s_2,s_3\}.
$$
Five saturated Bruhat chains are fixed:

- $X_1: e \triangleleft s_2$.
- $X_2: e \triangleleft s_3 \triangleleft s_3s_2 \triangleleft s_3s_2s_3$.
- $X_3: e \triangleleft s_1 \triangleleft s_2s_1 \triangleleft s_3s_2s_1 \triangleleft s_2s_3s_2s_1 \triangleleft s_1s_2s_3s_2s_1$.
- $Y_1$, a saturated chain of length $15$ beginning at $e$ and ending at $s_4 s_3 s_2 s_3 s_1 s_2 s_3 s_4 s_3 s_2 s_3 s_1 s_2 s_3 s_4$.
- $Y_2$, a saturated chain beginning at $s_4s_3s_2s_3$ and ending at $s_4 s_3 s_2 s_3 s_1 s_2 s_3 s_4 s_3 s_2 s_1$.

These chains realize the parabolic structure:
$$
W(F_4)_J = X_3X_1X_2,
$$
the parabolic subgroup of type $B_3$ generated by $J$, while the minimal left coset representatives for $W(F_4)/W(F_4)_J$ decompose into $Y_1\sqcup Y_2$. Consequently,
$$
W(F_4)=X_3X_1X_2Y_1\;\sqcup\;X_3X_1X_2Y_2,
$$
and every element has a unique factorization
$$
w=x_3x_1x_2u
$$
with $x_i\in X_i$, $u\in Y_1\cup Y_2$, and
$$
\ell(w)=\ell(x_3)+\ell(x_1)+\ell(x_2)+\ell(u)
$$
[2509.20981].

Using this factorization, two maps $L_1,L_2:W(F_4)\to \mathbb N^4$ are defined. For $w=x_3x_1x_2u$,
$$
L_1(x_3x_1x_2u)=
\begin{cases}
(\ell(x_1),\ell(x_3),\ell(x_2),\ell(u)), & u\in Y_1,\\
(\ell(x_1),\ell(x_3),\ell(u),\ell(x_2)), & u\in Y_2,
\end{cases}
$$
and
$$
L_2(x_3x_1x_2u)=
\begin{cases}
(\ell(x_1),\ell(x_3),\ell(x_2),\ell(u)), & u\in Y_1,\\
(\ell(x_1),\ell(x_3),\ell(x_2)+4,\ell(u)-4), & u\in Y_2.
\end{cases}
$$
Both maps are injective and rank-preserving because each chain is linear and has a unique element of each length [2509.20981].

Their images are described explicitly as unions of boxes in $\mathbb N^4$:

| Map | Image description |
|---|---|
| $L_1$ | $[1]_0\times[5]_0\times[3]_0\times[15]_0 \;\cup\; [1]_0\times[5]_0\times[11]_0\times[3]_0$ |
| $L_2$ | $[1]_0\times[5]_0\times[3]_0\times[15]_0 \;\cup\; [1]_0\times[5]_0\times[7]_0\times[7]_0$ |

These sit inside the ambient product $[1]_0\times[5]_0\times[11]_0\times[15]_0$ under componentwise order [2509.20981].

Theoremally, the set $\{L_1,L_2\}$ is a weak Lehmer code for type $F_4$: condition (2) holds whenever the image on an interval has a unique maximal element, and condition (3) holds globally after allowing automorphisms generated by inversion and the diagram automorphism [2509.20981]. This gives an explicit encoding and decoding mechanism for every element of $W(F_4)$, without asserting the existence of a single rectangular order embedding.

## 5. Multicomplexes, Lehmer complexes, and interval invariants

A multicomplex is an order ideal $J\subset \mathbb N^k$ under componentwise order. Its $f$-polynomial is
$$
\sum_{x\in J} q^{\rho(x)},\qquad \rho(x)=x_1+\cdots+x_k.
$$
For type $F_4$, given any $w\in W(F_4)$, one chooses $i$ and $\phi$ as guaranteed by the weak-code definition and sets
$$
J(w)=\{(L_i\circ \phi)(v):v\le w\}\subset \mathbb N^4.
$$
Then $J(w)$ is a multicomplex and
$$
\sum_{x\in J(w)} q^{\rho(x)}
=
\sum_{v\le w} q^{\ell(v)}
=
h_w(q),
$$
where $h_w(q)$ is the rank-generating function of the interval $[e,w]$ [2509.20981].

This is the principal structural payoff of weak Lehmer codes. The global order-embedding is unavailable, but every lower Bruhat interval still acquires an explicit multicomplex model. By the canonical construction described in [BS, Section 2], and quoted in [2509.20981], such a multicomplex produces a balanced vertex-decomposable simplicial complex whose $h$-polynomial is the interval rank-generating function. In the terminology of [2509.20981], this yields the Lehmer complex of $[e,w]$ in type $F_4$.

The behavior of principal and unimodal elements further refines this picture. If $w$ is $L$-principal, then the interval image is the full box below $L(w)$, and the rank-generating function factors as a product of $q$-analogues:
$$
h_w(q)=\prod_{i=1}^4 [L(w)_i+1]_q.
$$
For $L_1$, the set of palindromic Poincaré polynomials is realized by $L_1$-unimodal elements together with the longest element:
$$
\mathrm{Pal}(F_4)=\{h_w:w\in U(L_1)\}\cup\{[2]_q[6]_q[8]_q[12]_q\}.
$$
Moreover, both $\mathrm{Pr}(L_1)\cup\{w_0\}$ and $U(L_1)\cup\{w_0\}$ are lattices under Bruhat order, isomorphic to their coordinate images under $L_1$. By contrast, $U(L_2)$ is strictly smaller than $\mathrm{Pal}(F_4)\setminus\{w_0\}$ and does not form a lattice [2509.20981].

This suggests that weak Lehmer codes do more than recover Hilbert series: they stratify the Bruhat order into regions where factorization, palindromicity, and lattice structure are simultaneously visible in coordinates.

## 6. Examples and comparison with earlier Lehmer-code frameworks

Several explicit examples in [2509.20981] illustrate how weak Lehmer codes behave beyond the principal-box case. Let
$$
w=s_4s_3s_1s_2s_3s_4s_2s_1.
$$
This element is fixed by $\mathrm{Aut}(F_4,\le)$, and
$$
L_1(w)=(1,1,6,0).
$$
The multicomplex
$$
J(w)=\{L_1(v):v\le w\}
$$
has maximal elements
$$
(0,5,1,1),\ (1,1,0,5),\ (1,1,1,4),\ (1,1,3,1),\ (1,1,6,0),\ (1,2,1,2),\ (1,3,1,1).
$$
The associated Lehmer complex of $[e,w]$ has $|[e,w]|=100$ facets, and its generating function agrees with $h_w(q)$ [2509.20981]. This is a genuinely nontrivial interval: the code image is a multicomplex with several maximal elements rather than a single principal box.

A second example concerns principal and unimodal behavior. Let $c=s_1s_2s_3s_4$ and $w=cs_1$. Then
$$
L_1(w)=(1,2,1,1),
$$
so
$$
h_w(q)=[2]_q^3[3]_q,
$$
and $w\in \mathrm{Pr}(L_1)$. Its orbit by palindromic factorization is
$$
O_w=\{cs_1,cs_2,cs_3\},
$$
with images
$$
L_1(cs_1)=(1,2,1,1),\quad
L_1(cs_2)=(1,1,2,1),\quad
L_1(cs_3)=(1,1,1,2).
$$
The lexicographically minimal one is $L_1(cs_3)$, so $cs_3\in U(L_1)$ [2509.20981].

For the longest element $w_0$,
$$
h_{w_0}(q)=[2]_q[6]_q[8]_q[12]_q.
$$
Although this factors as a product of $q$-analogues, there is no global strong Lehmer code embedding onto $[1]_0\times[5]_0\times[7]_0\times[11]_0$ with order-preserving inverse. Nevertheless, the weak-code framework still produces a multicomplex recovering $h_{w_0}(q)$ [2509.20981]. This sharply distinguishes factorization of generating functions from existence of a strong code.

Relative to earlier Lehmer-code constructions, the $F_4$ framework occupies a distinct position. Denoncourt’s interval theory in type $A$ shows that the Lehmer codes of permutations in a weak-order interval form a distributive lattice, with a canonical base poset of join-irreducibles [1102.2689]. Tomie’s analysis of Denoncourt’s base posets studies when these interval-derived posets are $B_2$-free, characterizing that condition by avoidance of the patterns $3412$ and $3421$ [1111.3094]. In the parabolic type-$A$ setting, the consecutive Lehmer code realizes the parabolic Tamari lattice on $(\alpha,231)$-avoiding elements via componentwise order [2009.05342]. The weak Lehmer code for $F_4$ differs from all of these in that it is designed for Bruhat order in an exceptional Weyl group, it requires a finite family of maps rather than one, and its target is not a single product-of-chains image but a collection of intervalwise multicomplexes [2509.20981].

## 7. Broader significance, limitations, and open problems

The strong/weak distinction is explicit in [2509.20981]. A strong Lehmer code requires a single bijection
$$
L:W\to \prod_{i=1}^k [e_i]_0
$$
whose inverse is order-preserving, so the image is a product-of-chains subposet. A weak Lehmer code relaxes this to a finite family $\{L_i\}$, each injective and rank-preserving, such that for every $w$ there is at least one $L_i$, possibly composed with an automorphism, for which the image of $[e,w]$ is a multicomplex [2509.20981].

What is retained under this relaxation is substantial. Weak codes still realize rank-generating functions as multicomplex Hilbert series, provide explicit combinatorial encoding and decoding via parabolic chains, and yield poset isomorphisms on principal and unimodal strata [2509.20981]. What is lost is equally clear: there is no single global order-embedding onto a rectangular box, and there is no uniform surjectivity onto one product $\prod_i [e_i]_0$.

The paper places the $F_4$ result in a wider landscape. In classical types $A_n$, $B_n$, and $D_n$, strong Lehmer codes exist, and thus multicomplexes for all lower intervals are already available [2509.20981]. By contrast, a Coxeter system of type $\widetilde{C}_2$ does not admit any weak Lehmer code, and universal Coxeter systems with $|S|\ge 3$ do not admit weak codes because some lower intervals have rank sequences that are not $M$-sequences, for example $1,3,6,12$ [2509.20981]. The $F_4$ construction therefore sits between positive classical results and stronger negative results where even weak local control fails.

The exceptional types $E_6$, $E_7$, and $E_8$ remain open. The construction for $F_4$ suggests a possible route through carefully chosen parabolic factorizations and canonical chains partitioning $W$ into a small number of blocks, but obstacles remain in ensuring conditions (2) and (3) uniformly, especially uniqueness of maxima and the multicomplex ideal property [2509.20981]. The same source also identifies further problems: classifying $L$-principal and $L$-unimodal elements in exceptional types, optimizing the number $h$ of maps in a weak code family, and relating weak-code images to pattern-avoidance and rational smoothness through palindromicity [2509.20981].

This suggests a general interpretation of weak Lehmer codes as a compromise between algebraic factorization and order-theoretic rigidity. They do not recover a global rectangular model of Bruhat order, but they preserve enough structure to encode every lower interval by multicomplex data, thereby extending the Lehmer-code philosophy into settings where the classical product-of-chains paradigm provably breaks down.

Source: https://www.emergentmind.com/topics/weak-lehmer-codes