---
title: Reduced Pipe Dreams in Schubert Theory
url: https://www.emergentmind.com/topics/reduced-pipe-dreams
type: topic
---

# Reduced Pipe Dreams in Schubert Theory

Searching arXiv for recent and foundational papers on reduced pipe dreams and closely related developments.
Reduced pipe dreams are combinatorial encodings of permutations, traditionally realized as tilings by crosses and elbows in which the induced strands cross at most once pairwise. They occupy a central position in the combinatorics of Schubert and Grothendieck polynomials, where they function both as generating objects for polynomial expansions and as interfaces to compatible sequences, bumpless pipe dreams, crystal structures, and, more recently, probabilistic models. In the literature summarized here, reduced pipe dreams appear simultaneously as enumerative devices for Schubert calculus, as the reduced sector of broader pipe-dream formalisms, and as the structural backbone for bijections, pattern-avoidance results, and representation-theoretic constructions [2108.11438], [2403.07204], [2303.10463].

## 1. Definition and basic combinatorial model

For \(\pi\in S_\infty\), a pipe dream, also called an RC-graph, is a tiling of the positive quadrant by cross tiles and elbow tiles, with finitely many crosses, such that pipes enter from the north border and exit to the west border, and pipe \(i\) ends at row \(\pi(i)\) [2108.11438]. In a finite staircase model, one works in the staircase Young diagram \((n-1,n-2,\dots,1)\), with each box filled by either a cross or an elbow, and with the associated strands read as a wiring diagram [2009.14120]. The standard reducedness condition is that no two pipes cross twice, or equivalently that each pair of strands crosses at most once [2108.11438], [2009.14120].

The set of reduced pipe dreams of \(\pi\) is denoted \(PD(\pi)\) in [2108.11438], while [2403.07204] writes \(RP(w)\) for the set of reduced pipe dreams for \(w\in S_n\). For a reduced pipe dream \(D\), the set of cross positions is \(cross(D)\), and the weight is
\[
wt(D)=\prod_{(i,j)\in cross(D)} x_i.
\]
This row-weight convention is the standard one used in the Schubert-polynomial expansion [2108.11438].

A compatible-sequence description is built into the model. A compatible sequence of \(\pi\) is a pair
\[
(\mathbf a,\mathbf r)=((a_1,\dots,a_\ell),(r_1,\dots,r_\ell))
\]
with \(\ell=\ell(\pi)\), such that \(\mathbf a\) is a reduced word of \(\pi\), \(r_1\le \cdots \le r_\ell\), \(r_j\le a_j\) for all \(j\), and \(r_j<r_{j+1}\) whenever \(a_j<a_{j+1}\) [2108.11438]. If the crosses of \(D\) are ordered from top to bottom, and within each row from right to left, as
\[
(r_1,j_1)<\cdots<(r_\ell,j_\ell),
\]
then
\[
a_k=r_k+j_k-1.
\]
Equivalently, the cross at \((i,j)\) corresponds to the simple transposition \(s_{i+j-1}\) [2108.11438]. This identifies reduced pipe dreams and compatible sequences combinatorially.

The same row-by-row and right-to-left reading convention also underlies the reduced-pipe-dream formula for Schubert polynomials in [2403.07204]. There, if the crosses in a pipe dream \(D\) are listed row by row from top to bottom, and from right to left within each row, one obtains a reduced word \(a\) by recording \(i+j-1\) for each cross \((i,j)\), together with the compatible sequence \(\beta\) recording the row indices. This is described as the classical rc-graph/pipedream correspondence of Bergeron–Billey and Fomin–Kirillov [2403.07204].

## 2. Role in Schubert and Grothendieck polynomial theory

Reduced pipe dreams provide the standard positive combinatorial expansion of Schubert polynomials. In [2108.11438], the Schubert polynomial satisfies
\[
S_\pi=\sum_{D\in PD(\pi)} wt(D).
\]
The equivalent formulation in [2403.07204] is
\[
\mathfrak{S}_w(x_1, \dots, x_n) = \sum_{D \in RP(w)} \mathbf{x}^{wt(D)},
\]
where \(wt(D)\) is the weak composition whose \(i\)-th part is the number of crosses in row \(i\) of \(D\). Thus each reduced pipe dream contributes one monomial, and the Schubert polynomial is the generating function over all such monomials [2403.07204].

This expansion is related to the older reduced-word/compatible-sequence description
\[
\mathfrak{S}_w(x_1, \dots, x_n) = \sum_{a \in R(w)} \sum_{\beta \in C(a)} \mathbf{x}^{wt(\beta)},
\]
and the reduced-pipe-dream model is the geometric-combinatorial realization of that formula [2403.07204].

Reducedness becomes a deformation parameter boundary in Grothendieck-polynomial theory. In [2407.21653], the pipe-dream formula is
\[
\mathfrak{G}_w^\beta(x_1,\dots,x_n) = \sum_{D\in \PD(w)} \beta^{\#\cross(D)-\ell(w)}\wt(D),
\]
where \(\PD(w)\) now denotes all pipe dreams encoding \(w\). The reduced sector is recovered at \(\beta=0\), because only reduced pipe dreams contribute:
\[
\mathfrak{S}_w(x_1,\dots,x_n)=\sum_{D\in \RPD(w)}\wt(D).
\]
This places reduced pipe dreams as the Schubert-theoretic specialization of a larger Grothendieck-theoretic model [2407.21653].

A different extension appears in the theory of padded Schubert polynomials. For a dominant permutation \(\pi\), [2508.11879] fixes a partition \(\lambda\) and defines the associated padded Schubert polynomial by
\[
S_w^\pi(x;y) :=y^\lambda\,S_w\!\left(\frac{x_1}{y_1},\frac{x_2}{y_2},\dots\right),
\]
for \(w\le_L\pi\). The paper then gives a pipe-dream formula
\[
S_w^\pi(x;y)=\sum_{P\in\PD(w)} x^{P()}y^{P()\cap P(\pi)},
\]
where \(P(\pi)\) is a distinguished subset of \(\pi\)-dominated positions in the pipe dream. Although this formulation uses \(\PD(w)\), the starting point is explicitly “the usual Schubert polynomial \(S_w(x)\), defined combinatorially by pipe dreams,” with “reduced” meaning no two pipes cross more than once [2508.11879]. This suggests that reduced pipe dreams remain the organizing model even when additional \(y\)-weights and differential operators are introduced.

## 3. Compatible sequences, pseudo-Yamanouchi conditions, and TSSCPP

A notable recent use of reduced pipe dreams is the reinterpretation of totally symmetric self-complementary plane partitions (TSSCPP) in pipe-dream language [2303.10463]. The paper characterizes TSSCPP as bounded compatible sequences satisfying a Yamanouchi-like condition and thereby puts them in bijection with certain pipe dreams [2303.10463].

The construction starts from the boolean-triangle model for \(\TSSCPP(n)\), encoded by a triangular array \(b=\{b_{i,j}\}\) with entries in \(\{0,1\}\), indexed by
\[
1\le i\le n-1,\qquad n-i\le j\le n-1,
\]
subject to the diagonal partial-sum inequalities
\[
1+\sum_{k=j+1}^{i} b_{k,n-j-1}\;\ge\;\sum_{k=j}^{i} b_{k,n-j}
\qquad\text{for all }1\le j<i\le n-1.
\]
Huang and Striker flip the triangle vertically and left-justify it by defining
\[
y_{i,j}=b_{n-i,i+j-1},\qquad 1\le i\le n-1,\;1\le j\le n-i,
\]
then interpret each \(1\) as a cross tile and each \(0\) as an elbow tile in the triangular region of a pipe dream [2303.10463].

Under this translation, the diagonal-sum inequalities become a local condition equivalent to a pseudo-Yamanouchi counting rule on the associated bounded compatible sequence. If a pipe dream corresponds to
\[
(\mathbf a,\mathbf r)=((a_1,\dots,a_\ell),(r_1,\dots,r_\ell)),
\]
then it must satisfy
\[
r_1\ge r_2\ge \cdots \ge r_\ell,\qquad a_i\ge r_i,\qquad r_i>r_{i+1}\ \text{if }a_i\ge a_{i+1},
\]
and is pseudo-Yamanouchi if for all \(1\le k\le \ell\) and all \(1\le j\le n-2\),
\[
1+\cnt(k,j)(\mathbf a)\ge \cnt(k,j+1)(\mathbf a),
\]
where \(\cnt(k,j)(\mathbf a)\) is the number of entries equal to \(j\) among \(a_1,\dots,a_k\) [2303.10463]. The key characterization is
\[
\TSSCPP(n)\quad\longleftrightarrow\quad\{\text{pseudo-Yamanouchi pipe dreams in }\PD(n)\},
\]
and this bijection preserves weight [2303.10463].

Reduced pipe dreams enter decisively when the paper imposes pattern avoidance. A theorem of Gao states that if \(\pi\) avoids \(1432\), then any two reduced pipe dreams of \(\pi\) are connected by simple slides. Huang and Striker prove that pseudo-Yamanouchi-ness is preserved under simple slides and show that the bottom pipe dream is pseudo-Yamanouchi. Hence for \(1432\)-avoiding permutations, every reduced pipe dream of \(\pi\) is pseudo-Yamanouchi, giving
\[
\TSSCPP^{\mathrm{red}(\pi)}\;\cong\;\PD^{\mathrm{red}(\pi)}
\qquad(\pi\text{ avoids }1432).
\]
This is the reduced, \(1432\)-avoiding TSSCPP model [2303.10463].

A plausible implication is that reducedness here is not merely technical: it isolates the sector in which the local slide dynamics is sufficiently rigid to identify all reduced realizations with the TSSCPP-compatible condition.

## 4. Canonical bijection with reduced bumpless pipe dreams and ASM correspondences

Reduced pipe dreams are linked canonically to reduced bumpless pipe dreams in [2108.11438]. A reduced bumpless pipe dream is a tiling by six tile types encoding pipes that travel from \((\infty,i)\) to \((\pi(i),\infty)\) in a northeast direction, with reducedness again meaning that no two pipes cross twice [2108.11438]. The set is denoted \(BPD(\pi)\), and its weight is
\[
wt(D)=\prod_{(i,j)\in blank(D)} x_i.
\]
Lam–Lee–Shimozono’s theorem gives
\[
S_\pi=\sum_{D\in BPD(\pi)} wt(D),
\]
so reduced bumpless pipe dreams and reduced pipe dreams both enumerate the same Schubert polynomial [2108.11438].

The central result of [2108.11438] is a direct, explicit, weight-preserving bijection
\[
\varphi:BPD(\pi)\to PD(\pi).
\]
The map is constructed by iterating a “pop” operation on bumpless pipe dreams. Given \(D\), one repeatedly finds the smallest row index \(r\) containing a blank tile, marks the rightmost blank tile in that row, performs a sequence of local moves, and obtains \(\nabla D\in BPD(s_a\pi)\) while recording \(pop(D)=(a,r)\). Iterating this yields
\[
\varphi(D)=\big(\mathbf a=(a_1,\ldots,a_\ell),\mathbf r=(r_1,\ldots,r_\ell)\big),
\]
where
\[
pop(\nabla^{i-1}D)=(a_i,r_i)\qquad\text{for }i=1,\dots,\ell.
\]
The output is exactly a compatible sequence, hence a reduced pipe dream [2108.11438].

The paper further proves canonicity by showing that \(\varphi\) preserves Monk’s rule. Equivalently,
\[
\varphi\circ x_\alpha{\rightsquigarrow}=x_\alpha{\rightsquigarrow}\circ\varphi,
\qquad
\varphi\circ m_{s,\beta}=m_{s,\beta}\circ\varphi.
\]
This shows that the bijection is not just weight-preserving but structurally compatible with Schubert-calculus recursions [2108.11438].

This bijection becomes a bridge to alternating sign matrices (ASM) in [2303.10463]. Gao–Huang give a direct bijection
\[
\varphi:\BPD^{\mathrm{red}(\pi)}\to \PD^{\mathrm{red}(\pi)},
\]
preserving weight, and under the usual ASM–BPD correspondence, blank tiles correspond exactly to the NW-zero positions of the ASM. Consequently,
\[
\wt(A)=\prod_{(i,j)\in \nw(A)} x_i
\]
agrees with the BPD weight [2303.10463]. Combining the TSSCPP-to-pipe-dream translation with the reduced BPD-to-pipe-dream bijection yields, for every permutation \(\pi\), an explicit weight-preserving injection
\[
\varphi:\TSSCPP^{\mathrm{red}(\pi)}\hookrightarrow \ASM^{\mathrm{red}(\pi)}.
\]
If \(\pi\) avoids \(1432\), the map is a bijection:
\[
\TSSCPP^{\mathrm{red}(\pi)}\;\cong\;\ASM^{\mathrm{red}(\pi)}.
\]
For permutations avoiding both \(1432\) and \(2143\), the paper constructs a different bijection preserving natural poset structures [2303.10463].

The same paper defines \(\slide(\pi)\), the poset of reduced pipe dreams of \(\pi\), ordered by simple slides starting from the bottom pipe dream. In the \(1432\)- and \(2143\)-avoiding case, the bumpless-pipe-dream droop poset decomposes as a product of slide posets and dual slide posets:
\[
\droop(\pi)\cong \slide(\sigma_1)\times\cdots\times \slide(\sigma_k)\times \slide(\tau_1)^*\times\cdots\times \slide(\tau_\ell)^*.
\]
This yields
\[
\ASM(\pi)\cong \TSSCPP^{\mathrm{red}(\pi)}
\]
with the corresponding product-poset structure preserved [2303.10463].

## 5. Crystal operators and Demazure-crystal structure

Reduced pipe dreams support a Demazure crystal structure when Bergeron–Billey chute moves are restricted appropriately [2403.07204]. The paper begins from the reduced-pipe-dream expansion of the Schubert polynomial and then defines crystal operators by a pairing process between adjacent rows.

Fix \(1\le i<n\). Starting from the rightmost cross in row \(i\), one looks for an unpaired cross in row \(i+1\) lying weakly to the right, choosing the leftmost such cross if there are several; if found, the two crosses are paired. Moving leftward through row \(i\) leaves some crosses unpaired [2403.07204]. The lowering operator \(f_i\) moves the leftmost unpaired cross in row \(i\) down to row \(i+1\), jumping over a maximal rectangle of adjacent crosses. If \((i,j)\) is the leftmost unpaired cross in row \(i\), then one finds \(m\) such that
\[
(i,j-m), (i+1,j-m) \notin D_+,\qquad (i,j-k), (i+1,j-k) \in D_+ \text{ for } 1 \le k < m,
\]
and defines
\[
f_i(D)_+ = \{D_+ \setminus (i,j)\} \cup \{(i+1,j-m)\}.
\]
If there is no unpaired cross, or if all tiles to the left are already crosses, then \(f_i(D)=0\). The raising operator \(e_i\) is the inverse move [2403.07204].

These operators are well-defined on reduced pipe dreams and satisfy
\[
wt(f_i(D)) = wt(D)-\alpha_i,\qquad \alpha_i=e_i-e_{i+1},
\]
and
\[
wt(e_i(D)) = wt(D)+\alpha_i,
\]
with \(e_i\) and \(f_i\) mutually inverse on nonzero images [2403.07204]. The main theorem states that for any \(w\in S_n\), the operators \(e_i\) and \(f_i\) define a type \(A_{n-1}\) Demazure crystal structure on \(RP(w)\). In particular,
\[
RP(w) \cong \bigcup_{\substack{D \in RP(w) \\ e_i(D)=0,\ \forall\, 1\le i<n}} B_{\pi_D}(wt(D)),
\]
where \(\pi_D\) is the truncating permutation attached to the highest weight pipe dream \(D\) [2403.07204].

Highest weight pipe dreams are exactly those annihilated by all \(e_i\). The paper proves that if \(D\) is highest weight, then \(wt(D)\) is a partition. Each connected component under the crystal moves is a Demazure crystal generated from a highest weight pipe dream, and the monomials of \(\mathfrak{S}_w\) are partitioned into crystal components whose characters are key polynomials. Consequently,
\[
\mathfrak{S}_w(x_1,\dots,x_n)
= \sum_{\substack{D \in RP(w) \\ e_i(D)=0,\ \forall\,1\le i<n}}
\kappa_{a_D}(x_1,\dots,x_n),
\]
where the composition \(a_D\) is uniquely determined by the highest weight pipe dream \(D\) [2403.07204].

A weight-preserving bijection
\[
\phi: RP(w) \longrightarrow RFC(w^{-1}), \qquad (i,j)\in D_+ \longmapsto i+j-1 \in (r^i),
\]
to reduced factorizations with cutoff transports the Assaf–Schilling crystal structure back to reduced pipe dreams. The paper proves that \(\phi\) preserves the pairing process and intertwines the lowering and raising operators:
\[
\phi(f_i(D)) = f_i(\phi(D)),\qquad
\phi(e_i(D)) = e_i(\phi(D)).
\]
This identifies the restricted chute moves as crystal operators rather than merely local combinatorial moves [2403.07204].

This suggests that reduced pipe dreams constitute a particularly rigid realization of Schubert combinatorics: the same objects support polynomial enumeration, reduced-word encoding, and a full Demazure-crystal structure.

## 6. Extensions, probabilistic models, and generalized settings

Reduced pipe dreams are the type \(A\) prototype for a broader family of pipe-dream models. In [2009.14120], the model is extended uniformly to the classical Weyl groups \(B_n\), \(C_n\), and \(D_n\) by introducing signed pipe dreams with \(\Gamma\)-blocks and faucets. The central formulas are
\[
\mathfrak{C}_w^{[k]}(z;x)   = \sum_{D\in \mathrm{PD}_C^k(w)} z^{\beta(D)}x^{\alpha(D)},
\]
\[
\mathfrak{B}_w^{[k]}(z;x)   = \sum_{D\in \mathrm{PD}_B^k(w)} z^{\beta(D)}x^{\alpha(D)},
\]
\[
\mathfrak{D}_w^{[k]}(z;x)   = \sum_{D\in \mathrm{PD}_D^k(w)} z^{\beta(D)}x^{\alpha(D)}.
\]
These generalize the type \(A\) identity
\[
\mathfrak{S}_w(z)=\sum_{D\in \mathrm{PD}_A(w)} z^{\beta(D)}.
\]
The paper also proves the existence and uniqueness of a bottom pipe dream in each classical type [2009.14120]. Reduced pipe dreams in type \(A\) are therefore the base case of a larger Coxeter-theoretic framework rather than an isolated construction.

A different generalization appears in the study of random permutations arising from pipe dreams [2407.21653]. Fix \(n\) and the staircase
\[
\delta_n=\{(i,j): i+j\le n\}.
\]
Each box is independently filled by a cross with probability \(p\) and an elbow with probability \(1-p\), producing one of the \(2^{\binom n2}\) pipe dreams of order \(n\). A reduction procedure replaces any crossing of two pipes that have already crossed earlier by an elbow, producing a unique reduced pipe dream \(D'\). The associated permutation \(w(D)\) is defined from this reduction [2407.21653].

The probability of a given permutation is
\[
(\mathbf w=w) = \sum_{D\in \PD(w)} p^{\#\cross(D)}(1-p)^{\#\elbow(D)},
\]
which becomes
\[
(\mathbf w=w) = (1-p)^{\binom n2}\, \mathfrak{G}_w^{\beta=1}\!\left(\frac{p}{1-p},\dots,\frac{p}{1-p}\right).
\]
Thus the law of the random permutation is governed by \(\beta=1\) Grothendieck polynomials under principal specialization [2407.21653].

The paper proves that this model is equivalent to a colored stochastic six-vertex model and, after forgetting colors, to a discrete-time TASEP with parallel update and geometric jumps. For the permutation height function
\[
H(x,y) = \#\Bigl(\{w^{-1}(x),w^{-1}(x+1),\dots,w^{-1}(n)\}\cap\{y,y+1,\dots,n\}\Bigr),
\]
one obtains a deterministic limit shape, a limiting permuton, and \(n^{1/3}\)-scale Tracy–Widom GUE fluctuations in the curved region [2407.21653]. Although the random model begins with arbitrary pipe dreams, the reduction procedure shows that reduced pipe dreams are again the canonical representatives of permutation data.

The same paper also studies the \(q=1\) non-reduced model, where one reads the permutation from the raw random tiling without resolving double crossings. There the inversion number is of order \(n^{3/2}\), and the permuton converges to the identity permuton [2407.21653]. This sharp contrast clarifies the structural role of reducedness: passing to reduced pipe dreams changes both the algebraic interpretation and the asymptotic permutation statistics.

## 7. Structural themes and common misconceptions

Several recurring themes organize the modern theory of reduced pipe dreams.

First, reduced pipe dreams are not merely one model among many for Schubert polynomials. They are linked explicitly to compatible sequences [2108.11438], to reduced factorizations with cutoff [2403.07204], to reduced bumpless pipe dreams via a canonical bijection [2108.11438], and to TSSCPP and ASM in pattern-avoiding regimes [2303.10463]. The same objects support polynomial, bijective, and crystal-theoretic formalisms.

Second, reducedness is stronger than a convenience condition. In the standard definition, it means that no two pipes cross twice [2108.11438], [2403.07204]. In Grothendieck-theoretic settings, non-reduced pipe dreams contribute only when the deformation parameter allows them; at \(\beta=0\), the Schubert specialization forces the reduced sector [2407.21653]. In TSSCPP-related work, reducedness is the regime in which slide connectivity and pattern avoidance yield clean bijections [2303.10463].

Third, reduced pipe dreams should not be conflated with bumpless pipe dreams. The latter are distinct tiling objects with six tile types and blank-tile weights [2108.11438]. Their equivalence with reduced pipe dreams is highly nontrivial and is established by an explicit bijection \(\varphi\), not by identification of definitions [2108.11438].

Fourth, local moves on pipe dreams come in different forms with different purposes. Bergeron–Billey chute moves, simple slides, and Monk moves are not interchangeable notions. In [2403.07204], restricted chute moves define crystal operators; in [2303.10463], simple slides organize reduced pipe dreams into posets and preserve pseudo-Yamanouchi-ness under pattern avoidance; in [2108.11438], Monk moves are the operators relevant for canonicity of the BPD–PD bijection. A plausible implication is that much of the current literature distinguishes reduced-pipe-dream structures not by the underlying tilings but by the local move systems imposed on them.

Finally, reduced pipe dreams now function as a nexus connecting algebraic combinatorics, Schubert calculus, representation theory, and integrable probability. In the available literature, they generate Schubert polynomials [2108.11438], decompose into Demazure crystal components [2403.07204], mediate bijections with TSSCPP, ASM, and bumpless pipe dreams [2303.10463], [2108.11438], generalize to classical Weyl groups [2009.14120], and arise as the reduced images of random pipe-dream models with KPZ asymptotics [2407.21653]. This combination of exact combinatorial control and extensibility explains their continuing prominence in current research.

Source: https://www.emergentmind.com/topics/reduced-pipe-dreams