---
title: Rook Matroids and Their Combinatorial Structures
url: https://www.emergentmind.com/topics/rook-matroids
type: topic
---

# Rook Matroids and Their Combinatorial Structures

Searching arXiv for recent papers on rook matroids and related work.
Rook matroids are matroids defined from non-nesting rook placements on skew Ferrers boards. Given a skew shape $\lambda/\mu$ with $r$ rows and $c$ columns, one considers rook placements that are simultaneously non-attacking and non-nesting, and encodes each placement by the union of its occupied row labels and unoccupied column labels. The resulting family of $c$-subsets satisfies the basis-exchange axiom and therefore defines a matroid $R_{\lambda/\mu}$ on the ground set $[r+c]$ [2410.00127]. The theory places these objects at the intersection of Ferrers-board combinatorics, transversal matroid theory, lattice-path matroids, positroids, and the study of log-concavity phenomena in rook and $P$-Eulerian polynomials. Subsequent work has further developed their positroidal structure, including a characterization in terms of Grassmann necklaces and a description of cyclic flats [2509.21626].

## 1. Definition from non-nesting rook placements

Let $\lambda \supset \mu$ be partitions of length at most $r$, and let $B(\lambda/\mu)\subseteq [r]\times[r+c]$ be the skew Ferrers board with rows labeled $1,\dots,r$ from top to bottom and columns labeled $r+1,\dots,r+c$ from left to right [2410.00127]. A non-attacking rook placement $\rho\subseteq B(\lambda/\mu)$ is a subset containing at most one cell in each row and at most one cell in each column. Two rooks $(i,j)$ and $(i',j')$ are said to nest if $i<i'$ and $j<j'$. A non-nesting rook placement is a non-attacking placement with no nesting pair; the set of such placements is denoted $\operatorname{NN}_{\lambda/\mu}$ [2410.00127].

For $\rho\in \operatorname{NN}_{\lambda/\mu}$, let $R(\rho)\subseteq [r]$ denote the occupied rows and let $C(\rho)\subseteq \{r+1,\dots,r+c\}$ denote the unoccupied columns. Then $|R(\rho)\cup C(\rho)|=c$, and one defines
$$
\operatorname{Bases}(\lambda/\mu)=\{R(\rho)\cup C(\rho): \rho\in \operatorname{NN}_{\lambda/\mu}\}\subset \binom{[r+c]}{c}.
$$
The rook matroid $R_{\lambda/\mu}$ is the matroid on ground set $[r+c]$ whose bases are $\operatorname{Bases}(\lambda/\mu)$ [2410.00127].

The key foundational result is that these sets satisfy the basis-exchange axiom. The proof proceeds by relating placements to transversals of a column-to-row set system
$$
A_j=\{\,\text{rows } i \text{ with } (i,r+j)\in B(\lambda/\mu)\,\}\cup\{r+j\},
$$
and then “straightening” nestings by exchanges within the skew Ferrers board [2410.00127]. In the later formulation of the theory, the same construction is presented as a matroid of rank $c$ on $[r+c]$ with basis set
$$
\mathcal{B}=\{R(\rho)\cup C(\rho): \rho\in \operatorname{NN}_{\lambda/\mu}\},
$$
and the basis-exchange property is cited as established in [AJ–AL, Thm. 3.3], as summarized in [2509.21626].

A point requiring care is that the literature contains two closely related descriptions of the placement data, depending on conventions about board dimensions and full versus partial placements. The invariant definition of the matroid is the basis family built from row labels used by the placement and column labels omitted by it. This suggests that the central combinatorial object is not the rook placement alone, but the induced $c$-subset of the total label set.

## 2. Basic examples and first structural phenomena

Several examples illustrate the range of behavior of rook matroids. For the $3\times 2$ rectangle $\lambda=(2,2,2)$, the number of non-nesting placements of size $k$ is $\binom{3}{k}\binom{2}{k}$, and the resulting rook matroid is isomorphic to a uniform matroid $U_{k,5}$ [2410.00127]. More generally, in the positroidal treatment, for a rectangular board $\lambda/\mu=(n-k)^k$ with $r=k$ and $c=n-k$, one has $R_{\lambda/\mu}\cong U_{k,n}$ [2509.21626]. Its Grassmann necklace is explicitly
$$
I_i=[i,i+1,\dots,i+k-1],\qquad i=1,\dots,n
$$
cyclically, and there are no nontrivial inner- or outer-corner cyclic flats [2509.21626].

A more distinctive skew example arises from $\lambda=(5,4,4,2,1)$ and $\mu=(3,1,0,0,0)$, where $r=5$ and $c=5$. In this case there are exactly $\binom{5}{2}=10$ non-nesting placements of $5$ rooks, so $R_{\lambda/\mu}$ has $10$ bases of size $5$ [2509.21626]. The associated Grassmann necklace is listed explicitly as
$$
I_1=\{1,2,3,4,5\},\;I_2=\{2,3,4,5,6\},\;\dots,\;I_{10}=\{6,7,8,9,10\},
$$
and the inner corners at cells $(2,5)$ and $(4,7)$ yield cyclic flats that reflect the nontrivial geometry of the matroid [2509.21626].

One example is especially important for separating rook matroids from lattice-path matroids. For $\lambda/\mu=332/1$, one has $r=4$ and $c=3$, and the rank-$3$ bases are
$$
\{456,146,256,246,245,346,126,234\}.
$$
This family fails basis exchange and therefore cannot be the full basis set used in the defining construction [2410.00127]. At the same time, the same skew shape yields a rook matroid isomorphic to $Q_6$, the rank-$3$ quaternary excluded minor for lattice-path matroids [2410.00127]. The example therefore serves as a structural obstruction rather than a failure of the rook-matroid definition itself.

This dual role of $332/1$ is central: it identifies a specific skew subshape at which rook matroids diverge from lattice-path matroids, while still remaining inside the broader classes of transversal matroids and positroids.

## 3. Closure properties and relations to other matroid classes

Rook matroids satisfy several natural closure properties. If $(\lambda/\mu)'$ is the conjugate skew shape, then
$$
R_{\lambda/\mu}^*\cong R_{\lambda'/\mu'}
$$
so the class is closed under duality [2410.00127]. It is also closed under direct sums:
$$
R_{\lambda_1/\mu_1}\oplus R_{\lambda_2/\mu_2}\cong R_{(\lambda_1/\mu_1)\oplus(\lambda_2/\mu_2)},
$$
where the Ferrers diagrams are appended northeast of one another [2410.00127].

By contrast, rook matroids are not minor-closed. Deleting a row or column corresponds to removing that row or column from the skew shape, and contraction has an analogous interpretation, but the class fails closure under taking minors: for example, $R_{5543/321}\setminus 8$ is not a rook matroid [2410.00127]. This non-minor-closed behavior distinguishes the class from many familiar matroid families defined by forbidden-minor characterizations.

Rook matroids lie inside several broader matroid classes. First, they are transversal matroids. Using the set system
$$
A_j=\{\text{rows } i \text{ with } (i,r+j)\in B(\lambda/\mu)\}\cup\{r+j\},
$$
one obtains a transversal presentation, and $R_{\lambda/\mu}$ is the transversal matroid of $(A_1,\dots,A_c)$ [2410.00127]. The later positroid-focused account phrases this as an embedding into the transversal matroid of the bipartite graph $\Gamma$ determined by the board, with the non-nesting restriction selecting a subfamily of transversals while preserving matroidality [2509.21626].

Second, rook matroids are closely related to lattice-path matroids. If $U$ and $L$ are the upper and lower boundary paths of $\lambda/\mu$, then the lattice-path matroid $P_{\lambda/\mu}$ is another transversal matroid on the same ground set, constructed from admissible east-step indices in each column [2410.00127]. There is a bijection between non-crossing or non-nesting placements and lattice paths obtained by sending rooks to valleys of a path, and this yields
$$
|\operatorname{Bases}(R_{\lambda/\mu})|=|\operatorname{Bases}(P_{\lambda/\mu})|.
$$
Moreover,
$$
R_{\lambda/\mu}\cong P_{\lambda/\mu}
\quad\text{iff}\quad
\lambda/\mu \text{ avoids } 332/1,
$$
equivalently iff $Q_6$ is not a minor [2410.00127]. The same source states that the isomorphism can be made explicit via a “spine path” and a “path permutation.” The Tutte polynomials also coincide:
$$
T(P_{\lambda/\mu};x,y)=T(R_{\lambda/\mu};x,y)
$$
[2410.00127].

Third, rook matroids are positroids. In the initial treatment, this is shown using Oh’s criterion: for the natural labeling $1,\dots,r$ on rows and $r+1,\dots,r+c$ on columns, the Grassmann necklace term $I_i$ is the lexicographically minimal basis in the $i$-order, and one checks that it is the $i$-extremal rook placement [2410.00127]. The later work gives a new proof through sort-closedness: if $I,J$ are bases arising from non-nesting rook placements, then the sorted pairs $\operatorname{sort}_1(I,J)$ and $\operatorname{sort}_2(I,J)$ also arise from non-nesting rook placements, so the basis family is sort-closed, and hence the matroid is a positroid by the Lam–Postnikov criterion [2509.21626].

## 4. Grassmann necklaces, positroidal characterization, and cyclic flats

A major refinement of the theory is the characterization of rook matroids among positroids by means of Grassmann necklaces [2509.21626]. For a loop- and coloop-free positroid $M$ of rank $k$ on $[n]$, with Grassmann necklace
$$
\mathcal{I}=(I_1,\dots,I_n),\qquad I_i\in \binom{[n]}{k},
$$
the paper introduces row and column data extracted from each $I_i$:
$$
R_i=I_i\cap[1,n-k],\qquad r_i=\min R_i \text{ or } 0,
$$
$$
C_i=I_i\cap[n-k+1,n],
$$
together with ordered complements
$$
[n-k+1,n]\setminus C_i=\{s_\ell<\cdots<s_1\}.
$$
The main theorem states that $M\cong R_{\lambda/\mu}$ for some skew shape $\lambda/\mu\subset[r]\times[n-k+1,n]$ if and only if five explicit conditions hold, including the normalization $I_{n-k+1}=[n-k+1,n]$, inequalities governing the row minima and column maxima, and compatibility rules that produce inner-corner and outer-corner data from gaps in the necklace [2509.21626]. In that case the skew board $\lambda/\mu$ is uniquely recovered from the inner- and outer-corner sets $(IC(\mathcal{I}),OC(\mathcal{I}))$, and the necklace of $R_{\lambda/\mu}$ is exactly $\mathcal{I}$ [2509.21626].

This result answers a question of Thomas Lam and gives an intrinsic recognition theorem for rook matroids inside the class of positroids [2509.21626]. Conceptually, it replaces the external Ferrers-board construction by internal necklace data. A plausible implication is that rook matroids can be studied through the combinatorial infrastructure of positroids without always referring back to rook placements.

The same work identifies an important subclass of cyclic flats. In a rook matroid $R_{\lambda/\mu}$, the essential cyclic flats are the cyclic intervals associated to inner and outer corners of the skew shape [2509.21626]. If $(i,j)$ is an inner corner, then
$$
F_{i,j}=[j+1,i]\subset [1,\dots,r+c]
$$
is a cyclic flat of rank
$$
\operatorname{rk}(F_{i,j})=|I_{j+1}\cap [j+1,i]|=r+c-j.
$$
If $(i,j)$ is an outer corner, then
$$
G_{i,j}=[i,j-1]
$$
is a cyclic flat of rank
$$
\operatorname{rk}(G_{i,j})=j-1-r
$$
[2509.21626].

These connected cyclic flats generate the facet-defining inequalities of the base polytope:
$$
x(E)=c,
$$
$$
x([j+1,i])\le r+c-j \qquad \text{for all inner corners},
$$
$$
x([i,j-1])\le j-1-r \qquad \text{for all outer corners}
$$
[2509.21626]. This description makes the positroidal geometry of rook matroids concrete and links the Ferrers-board combinatorics directly to polyhedral data.

## 5. Enumerative theory and ultra-log-concavity

For a skew shape $\lambda/\mu$ with $c$ columns, the non-nesting rook polynomial is defined by
$$
M_{\lambda/\mu}(t)=\sum_{k=0}^c r_k(\lambda/\mu)t^k,
$$
where $r_k$ is the number of non-nesting placements of size $k$ [2410.00127]. A multivariate refinement is the basis polynomial
$$
r_{\lambda/\mu}(x_1,\dots,x_r;y_{r+1},\dots,y_{r+c})
=\sum_{\rho\in \operatorname{NN}}
\prod_{i\in R(\rho)}x_i\prod_{j\in C(\rho)}y_j.
$$
This polynomial records the row/column encoding that defines the matroid itself [2410.00127].

The central enumerative result is ultra-log-concavity. The argument invokes the Stanley–Yan theorem: for any matroid $M$ on ground set $E$ and subset $T\subseteq E$, if
$$
f_i=\left|\{B\in\operatorname{Bases}: |B\cap T|=i\}\right|,
$$
then the sequence $(f_i)$ is ultra-log-concave with no internal zeros [2410.00127]. Applying this to $M=R_{\lambda/\mu}$ and $T$ equal to the set of row labels gives the inequality
$$
\left(\frac{r_k}{\binom{c}{k}}\right)^2
\ge
\left(\frac{r_{k-1}}{\binom{c}{k-1}}\right)
\left(\frac{r_{k+1}}{\binom{c}{k+1}}\right),
$$
so the coefficient sequence of $M_{\lambda/\mu}(t)$ is ultra-log-concave [2410.00127].

This contrasts sharply with the classical theory of unrestricted rook placements. The non-nesting rook polynomial need not be real-rooted in general; an explicit counterexample is given by the skew shape $\lambda/\mu=888888765/76654321$ [2410.00127]. Thus the principal regularity property is ultra-log-concavity rather than real-rootedness. This distinction is structurally important: it shows that the matroidal mechanism supplies strong coefficient inequalities even when zero distributions of the generating polynomial do not exhibit the strongest expected behavior.

Additional symmetry phenomena occur in special cases. The polynomial $M_{\lambda/\mu}$ is palindromic if and only if $\lambda/\mu$ decomposes into “squarecases,” described as the condition that outer corners lie on the northwest diagonal [2410.00127]. In that situation $M_{\lambda/\mu}$ is $\gamma$-positive and satisfies
$$
(-1)^{\lfloor n/2\rfloor}M_{\lambda/\mu}(-1)\ge 0
$$
[2410.00127]. These results place non-nesting rook polynomials in a broader family of enumerative objects where strong coefficient constraints survive in the absence of general real-rootedness.

## 6. Width-two posets and $P$-Eulerian polynomials

One of the most consequential applications of rook matroids is to the Eulerian theory of width-two posets [2410.00127]. Starting from the spine path of a skew shape, the rows and columns are labeled by the steps of that path: north steps are labeled bottom-to-top and east steps left-to-right, producing two chains $C_1$ and $C_2$ of lengths $r$ and $c$ [2410.00127]. Cover relations are then imposed as follows: for each outer corner one declares $i\to j$ with $i\in C_1$ and $j\in C_2$, while for each inner corner one declares $j\to i$ [2410.00127]. The resulting poset is a naturally labeled width-two poset $(P,\omega)$.

The multivariate $P$-Eulerian polynomial is defined by
$$
\widetilde{W}_{P}(x,y)
=
\sum_{\sigma\in \operatorname{LinExt}(P)}
\prod_{i\in \operatorname{DesBot}(\sigma)}x_i
\prod_{j\in \operatorname{ColAsc}(\sigma)}y_j,
$$
where
$$
\operatorname{DesBot}(\sigma)=\{\sigma_k:\sigma_{k-1}>\sigma_k\}
$$
and
$$
\operatorname{ColAsc}(\sigma)=\{\sigma_k:\sigma_k<\sigma_{k+1}\text{ and }\sigma_k\in C_2\}
$$
[2410.00127]. The key theorem states that under the bijection between the skew shape and the width-two poset,
$$
\widetilde{W}_P(x,y)=r_{\lambda/\mu}(x,y),
$$
and in particular the specialization $x_i\mapsto t$, $y_j\mapsto 1$ recovers the non-nesting rook polynomial $M_{\lambda/\mu}(t)$ [2410.00127].

As a consequence, for any naturally labeled width-two poset $P$, the univariate $P$-Eulerian polynomial $W_P(t)$ is ultra-log-concave [2410.00127]. The same result states that the stronger multivariate polynomial is Lorentzian [2410.00127]. In the language of conjectural Eulerian positivity and log-concavity, this establishes that the log-concavity part of Brenti’s conjecture holds for width $2$, and it “completes the story of the Neggers–Stanley conjecture for naturally labeled width two posets” [2410.00127].

The significance of this correspondence is methodological as well as enumerative. It transfers a matroid-theoretic coefficient theorem to a poset-linear-extension problem by way of a geometric-combinatorial encoding through skew Ferrers boards. This suggests that rook matroids function as an intermediary object connecting matroid basis enumeration and descent-type statistics on linear extensions.

## 7. Position within current research

The initial development of rook matroids establishes a self-contained framework including definitions, structural theorems, connections to transversal matroids, lattice-path matroids, and positroids, together with enumerative consequences for non-nesting rook polynomials and $P$-Eulerian polynomials [2410.00127]. A central message is that non-nesting restrictions, unlike classical rook-placement conditions, produce a matroidal class with unusually rich interaction between geometry, enumeration, and order theory.

Subsequent work shifts emphasis to the positroidal viewpoint [2509.21626]. The Grassmann-necklace characterization gives necessary and sufficient conditions for a positroid to arise as a rook matroid, thereby answering a question of Lam. The determination of essential cyclic flats and corresponding facet inequalities further situates rook matroids within the combinatorics of positroid polytopes [2509.21626]. This indicates that the class is not merely a special family of transversal matroids defined by Ferrers-board data, but also a recognizable subclass of positroids with explicit necklace and polyhedral signatures.

Several structural boundaries are now clear. Rook matroids are closed under duals and direct sums but not minors [2410.00127]. They coincide with lattice-path matroids exactly in the $332/1$-avoiding regime, and the appearance of $Q_6$ marks the obstruction [2410.00127]. Their enumerative generating polynomials are ultra-log-concave, but not generally real-rooted [2410.00127]. These contrasts are not incidental; they delineate a class that is simultaneously robust enough to support matroidal and positroidal machinery, yet narrow enough to display distinctive behavior not shared by better-known families.

Within algebraic and geometric combinatorics, rook matroids therefore occupy a specific niche: they encode non-nesting configurations on skew Ferrers boards as bases of matroids, inherit transversal and positroidal structure, interface sharply with lattice-path matroids through the excluded configuration $332/1$, and furnish a matroid-based route to ultra-log-concavity results for width-two $P$-Eulerian polynomials [2410.00127; 2509.21626].

Source: https://www.emergentmind.com/topics/rook-matroids