Rubey's Poset: Lattice in Pipe Dream Combinatorics

• Rubey’s poset is a lattice of reduced pipe dreams for a fixed permutation, characterized by generalized chute moves that preserve reducedness and connectivity.
• It employs coordinate-indexed move operations to recursively construct joins and meets, ensuring both termination and confluence through antidiagonal indices.
• A global tableau model links Rubey’s poset with Lehmer tableaux, offering a unified framework to analyze Schubert polynomial combinatorics and local interval geometry.

Rubey’s poset is the partial order on the reduced pipe dreams, or rc-graphs, for a fixed permutation $w$, with cover relations given by generalized chute moves that slide a cross down-right while preserving reducedness and the pipe connectivity type $w$. The conjecture that this poset is always a lattice has now been proved. Recent accounts give two complementary descriptions of the same structure: a direct move-theoretic proof based on coordinate-indexed operations $\mathcal{M}_{ij}$ and recursive constructions of joins and meets, and a global model identifying the poset with a componentwise ordered family of Lehmer tableaux. Together these results show that Rubey’s poset is a lattice, and more specifically a semidistributive polygonal lattice whose polygons are all diamonds or pentagons (Billey et al., 25 Jul 2025, Axelrod-Freed et al., 17 Jul 2025).

1. Ambient combinatorics: permutations, Rothe diagrams, and reduced pipe dreams

Rubey’s poset is defined for a fixed permutation $w \in S_n$, written in one-line notation as

$w=[w(1),w(2),\dots,w(n)].$

Its inversion set is

$$\mathrm{Inv}(w):=\{(i,j):1\le i<j\le n \ \text{and}\ w(i)>w(j)\},$$

and its Coxeter length is $\ell(w)=|\mathrm{Inv}(w)|$.

A standard geometric encoding of $w$ is the Rothe diagram

$$D(w):=\{(i,j)\in [n]\times [n] : j<w(i)\ \text{and}\ i<w^{-1}(j)\}.$$

Equivalently, $D(w)$ is the set of boxes strictly to the left of a $w$0 in the same row and strictly above a $w$1 in the same column in the permutation matrix of $w$2. One has $w$3.

A pipe dream is a tiling of the infinite first-quadrant grid by cross tiles and elbow tiles, with pipes entering from the left in rows $w$4 and exiting at the top in columns $w$5. A pipe dream $w$6 represents $w$7 if the pipe entering at row $w$8 exits at column $w$9. Writing $\mathcal{M}_{ij}$0 for the set of cross positions, $\mathcal{M}_{ij}$1 is reduced if any two pipes cross at most once and

$\mathcal{M}_{ij}$2

Standard constructions imply that all crosses of a reduced pipe dream lie in a bounded staircase-like region determined by $\mathcal{M}_{ij}$3. The recent literature uses both $\mathcal{M}_{ij}$4 and $\mathcal{M}_{ij}$5 for the set of reduced pipe dreams for $\mathcal{M}_{ij}$6.

2. Generalized chute and ladder moves

Rubey’s generalized moves extend the classical chute and ladder moves on rc-graphs. In the ladder formulation, for a box $\mathcal{M}_{ij}$7 one considers an NE–SW ladder segment

$\mathcal{M}_{ij}$8

for maximal $\mathcal{M}_{ij}$9 such that the boxes remain in the relevant bounding region and the local configuration is admissible. Informally, a generalized chute move is permitted when the tiles along the ladder are elbows except for a single cross at one endpoint; the move toggles that unique cross to the other endpoint. The inverse operation is a generalized ladder move.

A complementary staircase-diagram formulation uses a combinatorial rectangle $w \in S_n$0 in $w \in S_n$1. If $w \in S_n$2 and $w \in S_n$3 contain bump tiles, $w \in S_n$4 contains either a bump or elbow tile, and all other boxes in $w \in S_n$5 are cross tiles, then one may perform a chute move changing the bump at $w \in S_n$6 to a cross tile and the cross at $w \in S_n$7 to a bump tile. The result is again a reduced pipe dream for the same permutation.

Fixing $w \in S_n$8, Rubey’s covering relation is generated by a single down-right generalized chute move: $w \in S_n$9 The partial order is the reflexive-transitive closure,

$w=[w(1),w(2),\dots,w(n)].$0

Generalized ladder moves are therefore the cover relations in the opposite direction. Rubey also proved that $w=[w(1),w(2),\dots,w(n)].$1 has minimum and maximum elements (Axelrod-Freed et al., 17 Jul 2025).

3. The lattice theorem via the move operations $w=[w(1),w(2),\dots,w(n)].$2

The direct proof of the lattice property introduces a family of coordinate-indexed move operations $w=[w(1),w(2),\dots,w(n)].$3. For a reduced pipe dream $w=[w(1),w(2),\dots,w(n)].$4, $w=[w(1),w(2),\dots,w(n)].$5 is defined by taking the maximal admissible ladder through $w=[w(1),w(2),\dots,w(n)].$6, performing the associated generalized chute move, and then continuing along consecutive admissible ladder segments in a canonical way until the cross reaches the southeast endpoint or the process is blocked by reducedness constraints. Its inverse $w=[w(1),w(2),\dots,w(n)].$7 performs the corresponding sequence of generalized ladder moves. Conceptually, $w=[w(1),w(2),\dots,w(n)].$8 is a canonical “push southeast along the ladder through $w=[w(1),w(2),\dots,w(n)].$9” operation (Billey et al., 25 Jul 2025).

To formulate joins and meets, let the antidiagonal index of a box be $$\mathrm{Inv}(w):=\{(i,j):1\le i<j\le n \ \text{and}\ w(i)>w(j)\},$$0. Generalized chute moves push crosses to larger $$\mathrm{Inv}(w):=\{(i,j):1\le i<j\le n \ \text{and}\ w(i)>w(j)\},$$1, while generalized ladder moves pull them to smaller $$\mathrm{Inv}(w):=\{(i,j):1\le i<j\le n \ \text{and}\ w(i)>w(j)\},$$2. For $$\mathrm{Inv}(w):=\{(i,j):1\le i<j\le n \ \text{and}\ w(i)>w(j)\},$$3, define

$$\mathrm{Inv}(w):=\{(i,j):1\le i<j\le n \ \text{and}\ w(i)>w(j)\},$$4

A join is obtained by starting from $$\mathrm{Inv}(w):=\{(i,j):1\le i<j\le n \ \text{and}\ w(i)>w(j)\},$$5 and recursively applying

$$\mathrm{Inv}(w):=\{(i,j):1\le i<j\le n \ \text{and}\ w(i)>w(j)\},$$6

where $$\mathrm{Inv}(w):=\{(i,j):1\le i<j\le n \ \text{and}\ w(i)>w(j)\},$$7 is minimal in a fixed total order refining $$\mathrm{Inv}(w):=\{(i,j):1\le i<j\le n \ \text{and}\ w(i)>w(j)\},$$8 and then lexicographic order, chosen so that $$\mathrm{Inv}(w):=\{(i,j):1\le i<j\le n \ \text{and}\ w(i)>w(j)\},$$9 decreases by $\ell(w)=|\mathrm{Inv}(w)|$0. Termination yields

$\ell(w)=|\mathrm{Inv}(w)|$1

Dually, starting from $\ell(w)=|\mathrm{Inv}(w)|$2 and applying inverse moves gives

$\ell(w)=|\mathrm{Inv}(w)|$3

The proof that these formulas give the join and meet uses termination and confluence. Termination is controlled by the monotonic increase of a rank-like statistic, such as the multiset of antidiagonal indices of crosses, under a well-order. Confluence is obtained from local diamond lemmas for overlapping ladders. The canonical normal form reached by reconciling $\ell(w)=|\mathrm{Inv}(w)|$4 upward toward $\ell(w)=|\mathrm{Inv}(w)|$5 is then the least element above both, and the canonical normal form reached by reconciling $\ell(w)=|\mathrm{Inv}(w)|$6 downward toward $\ell(w)=|\mathrm{Inv}(w)|$7 is the greatest element below both.

The same framework yields an explicit comparability criterion. Define the antidiagonal crossing profile

$\ell(w)=|\mathrm{Inv}(w)|$8

A necessary condition for $\ell(w)=|\mathrm{Inv}(w)|$9 is

$w$0

More precisely,

$w$1

where $w$2 records the index of the unique cross along $w$3. Operationally, comparability can therefore be tested by scanning ladders and comparing the positions of their unique crosses. The same paper observes that each move changes the antidiagonal index of exactly one cross and strictly decreases the cardinality of the relevant mismatch set, so the number of steps in the recursive join or meet computation is at most $w$4; local admissibility checks are constant-time per move if adjacency and ladder-membership data structures are maintained (Billey et al., 25 Jul 2025).

4. The global tableau model: inversions tableaux and Lehmer tableaux

A second description of Rubey’s poset replaces local moves by a global combinatorial model. For

$w$5

viewed inside the lower-right staircase, a column-injective tableau for $w$6 is a map

$w$7

such that $w$8 iff $w$9, and the nonzero entries in each column are distinct. Balancedness is imposed on every $$D(w):=\{(i,j)\in [n]\times [n] : j<w(i)\ \text{and}\ i<w^{-1}(j)\}.$$0-shape $$D(w):=\{(i,j)\in [n]\times [n] : j<w(i)\ \text{and}\ i<w^{-1}(j)\}.$$1 by

$$D(w):=\{(i,j)\in [n]\times [n] : j<w(i)\ \text{and}\ i<w^{-1}(j)\}.$$2

equivalently, $$D(w):=\{(i,j)\in [n]\times [n] : j<w(i)\ \text{and}\ i<w^{-1}(j)\}.$$3 is the median of the entries on $$D(w):=\{(i,j)\in [n]\times [n] : j<w(i)\ \text{and}\ i<w^{-1}(j)\}.$$4. An inversions tableau is a balanced column-injective tableau in which every entry in row $$D(w):=\{(i,j)\in [n]\times [n] : j<w(i)\ \text{and}\ i<w^{-1}(j)\}.$$5 is at most $$D(w):=\{(i,j)\in [n]\times [n] : j<w(i)\ \text{and}\ i<w^{-1}(j)\}.$$6. The set of such tableaux is denoted $$D(w):=\{(i,j)\in [n]\times [n] : j<w(i)\ \text{and}\ i<w^{-1}(j)\}.$$7.

For a reduced pipe dream $$D(w):=\{(i,j)\in [n]\times [n] : j<w(i)\ \text{and}\ i<w^{-1}(j)\}.$$8, the pipes $$D(w):=\{(i,j)\in [n]\times [n] : j<w(i)\ \text{and}\ i<w^{-1}(j)\}.$$9 and $D(w)$0 cross exactly once iff $D(w)$1. Let $D(w)$2 be the row where they cross, and define

$D(w)$3

The fundamental theorem is that

$D(w)$4

is a bijection (Axelrod-Freed, 15 Jul 2025).

The Lehmer form of a column-injective tableau is then defined boxwise by

$D(w)$5

For $D(w)$6, $D(w)$7 is a Lehmer tableau, and $D(w)$8 is partially ordered componentwise. The composite

$D(w)$9

is a bijection. Its key local feature is that an increment at a box $w$00 in tableau form adds exactly $w$01 to the entry at $w$02 in Lehmer form; increments commute, and generalized chute moves become controlled monotone updates of Lehmer coordinates. The central theorem is therefore a global one: $w$03 This identifies Rubey’s poset with a subposet of a product of chains and gives a global description of the chute-move order (Axelrod-Freed et al., 17 Jul 2025).

5. Local interval geometry, semidistributivity, and auxiliary symmetries

The global tableau model permits a precise local analysis of intervals. When $w$04 and $w$05 arise from chute rectangles $w$06 and $w$07, the geometry of the interval $w$08 is controlled by how the southwest corners of the two rectangles interact. If $w$09 and $w$10 are not corners of the other rectangle, then

$w$11

componentwise, and the interval is a diamond. If $w$12, the same componentwise-maximum formula holds and the interval is a pentagon. If $w$13, the interval is still a pentagon, although a componentwise-maximum formula is not asserted in $w$14 itself. These local cases underlie the proof that every pair of covers of a common lower element has a join.

The same paper proves the stronger structural theorem that, for every $w$15, $w$16 is a semidistributive polygonal lattice whose polygons are all diamonds or pentagons. Here semidistributivity means that the lattice satisfies

$w$17

Polygonality means that whenever two elements cover their meet, or are covered by their join, the interval between those endpoints is a polygon, i.e. a union of two saturated chains sharing only endpoints.

Two auxiliary transformations play an important role. The transpose operation $w$18 is a bijection $w$19 and a poset anti-isomorphism. The triforce embedding sends $w$20 into $w$21, where

$w$22

and realizes $w$23 as an interval in the larger poset. It is used to reduce the $w$24 pentagon case to the $w$25 case (Axelrod-Freed et al., 17 Jul 2025).

6. Context within Schubert combinatorics and current directions

Rubey’s poset sits inside the combinatorics of Schubert polynomials. Reduced pipe dreams index the monomials of the Schubert polynomial $w$26, and generalized chute and ladder moves are local transformations preserving reducedness and the target permutation. In this sense, the poset organizes the network of rc-graphs for a fixed $w$27. The order also respects antidiagonal height data: generalized chute moves push crosses southeast, so the poset admits both a local geometric interpretation in the grid and a global monotone interpretation through Lehmer coordinates. Its relation to weak order or Bruhat order is indirect; the rc-graph order is finer and depends on pipe-dream configurations rather than on the permutation order itself.

Historically, Rubey defined the chute-move poset and conjectured that it is always a lattice, building on Bergeron–Billey and Serrano–Stump. Ceballos–Padrol–Sarmiento proved the conjecture in the special case of permutations beginning with $w$28 and avoiding the pattern $w$29, where one obtains $w$30-Tamari lattices. The full conjecture was then resolved in 2025 by two complementary approaches: a move-theoretic proof via $w$31, joins, meets, and a ladderwise comparability criterion, and a tableau-theoretic proof via the isomorphism $w$32. Axelrod-Freed–Defant–Mularczyk–Nguyen–Tung also note that Billey–McCausland–Minnerath independently and simultaneously proved the conjecture by a different approach (Billey et al., 25 Jul 2025, Axelrod-Freed et al., 17 Jul 2025).

The current scope and limitations are sharply defined. Rubey’s poset is a poset of reduced pipe dreams for a fixed permutation $w$33; allowing nonreduced configurations breaks move admissibility and the order. For permutations of small length the poset may be a chain, and the generalized moves degenerate to the classical chute and ladder moves. The papers also isolate several open directions: whether distributivity occurs for special families such as $w$34-avoiding or vexillary permutations; the diameter of the Hasse diagram under generalized moves and the complexity of random walk or uniform generation of rc-graphs; and stronger connections to Edelman–Greene insertion, bumpless pipe dreams, and the geometry of brick polytopes. These questions indicate that Rubey’s poset is now understood as a lattice-theoretic object, but not yet exhausted as a geometric or algorithmic one.