\documentclass[11pt]{amsart} \usepackage{amsmath,amssymb,indentfirst,hyperref} \newtheorem{theorem}{Theorem} \newtheorem{proposition}[theorem]{Proposition} \title{Schur nonnegativity of double-wiring cluster variables} \author{Qiyue Tang} \date{November 15, 2026} \setlength{\emergencystretch}{2em} \address{Zhili College, Tsinghua University, Beijing 100084, China} \email{tangqy24@mails.tsinghua.edu.cn} \subjclass[2020]{Primary 13F60. Secondary 05E05} \keywords{Double wiring diagrams, cluster variables, Schur nonnegativity, Kazhdan--Lusztig immanants} \begin{document} \begin{abstract} We prove Cuenca's conjecture that cluster variables from double wiring diagrams evaluate to Schur nonnegative symmetric functions on generalized Jacobi--Trudi matrices. The same holds for cluster monomials. \end{abstract} \maketitle \section{Introduction} Fomin and Zelevinsky introduced double wiring diagrams to test total positivity \cite{FZ}. The diagrams select families of matrix minors, one for each chamber. Local moves connect these families, and the determinantal identities for the moves express the product of the two exchanged minors as a sum of two products of common minors. Berenstein, Fomin and Zelevinsky placed these exchanges in the cluster structures of double Bruhat cells \cite{BFZ}. Haiman proved a Schur nonnegativity theorem for Kazhdan--Lusztig immanants \cite[Theorem 1.5]{H}. Rhoades and Skandera formulate the consequence for generalized Jacobi--Trudi matrices \cite[Proposition 2.3]{RS}. Cuenca observed that this result would imply Schur nonnegativity for all double-wiring cluster variables if those variables belonged to the matrix dual canonical basis \cite[Conjecture 1.1]{Cuenca}. Cuenca called a variable Pl\"ucker-adjacent if it is obtained by one mutation from the seed of some double wiring diagram. His Conjecture 1.2 addresses those variables. Using Temperley--Lieb immanants, he proved Schur nonnegativity for a mutation at a chamber of degree at most six \cite[Theorem 1.3 and \S\S 5--6]{Cuenca}. Geiss, Leclerc and Schroer realized quantum matrices as a quantum unipotent coordinate algebra with a cluster structure \cite{GLS}. Grabowski and Launois describe its initial seed explicitly \cite{GL}. Kang, Kashiwara, Kim and Oh proved that quantum unipotent cluster monomials are upper global basis elements up to powers of the quantum parameter \cite{KKKO}. Chepuri and Sherman-Bennett formulate related questions for the big double Bruhat cell in the special linear group, including a conjectural characterization by permutation patterns \cite[\S 4]{CSB}. Here we work with polynomial matrix entries. We identify a double-wiring seed in these entries with the specialized quantum matrix seed, and compare the quantum unipotent and matrix bases in the same coordinates. This proves Cuenca's conjecture and its extension to cluster monomials. Let \(n\geq1\), let \([n]=\{1,\ldots,n\}\), and let \(X=(x_{ij})_{1\leq i,j\leq n}\) be a matrix of commuting indeterminates. For subsets \(I,J\subseteq[n]\) of equal cardinality, write \(\Delta_{I,J}(X)\) for the minor with rows \(I\) and columns \(J\), each taken in increasing order, with \(\Delta_{\varnothing,\varnothing}=1\). Let \(\mathcal A_n\) be the polynomial matrix cluster algebra obtained from a double wiring diagram: its boundary chamber minors are frozen and are not inverted. The chamber with empty label has value \(1\) and is omitted from the seed. An extended cluster consists of its mutable and frozen variables. A double wiring diagram comprises two families of \(n\) wires, red and blue, with each pair of wires of the same color crossing once. The blue wires are numbered increasingly from top to bottom at the left boundary, and the red wires at the right boundary. In each chamber the sets \(I\) of blue labels and \(J\) of red labels below the chamber give its variable \(\Delta_{I,J}(X)\). A chamber with two crossing endpoints is mutable. The others are frozen. The level of a chamber is the number of blue wires below it, which is also the number of red wires below it. Cuenca's quiver is obtained by the following rules \cite[\S 2.2.1]{Cuenca}. Horizontal adjacency across a red crossing gives an arrow to the right, and adjacency across a blue crossing gives one to the left. For two chambers \(c,c'\) in adjacent levels, compare the horizontal intervals of the chambers. If the interval of \(c'\) is contained in that of \(c\) and \(c'\) has endpoints of different colors, the arrow is \(c\to c'\) when the left endpoint of \(c'\) is blue and \(c'\to c\) when it is red. If the intervals interlace with the left endpoint of \(c'\) inside \(c\) and the right endpoint of \(c\) inside \(c'\), and these endpoints have the same color, the arrow is \(c\to c'\) for blue and \(c'\to c\) for red. In the other interlacing order compare the right endpoint of \(c'\) with the left endpoint of \(c\). If their colors agree the arrow is \(c'\to c\) for blue and \(c\to c'\) for red. Omit arrows between frozen vertices. Mutation at a mutable vertex \(k\) replaces its variable \(z_k\) by \(z'_k\) according to \[ z_k z'_k=\prod_{i\to k}z_i+\prod_{k\to j}z_j, \] with one factor per arrow and empty products equal to \(1\). Mutation adds an arrow \(i\to j\) for each two-arrow path \(i\to k\to j\), reverses the arrows incident to \(k\), cancels opposite pairs, and discards arrows joining frozen vertices. The cluster variables are the variables in the seeds reached by finite sequences of these operations. Let \(\Lambda_{\mathbb Z}\) be the ring of symmetric functions over \(\mathbb Z\). For \(t\geq0\), let \(h_t\) be the complete symmetric function of degree \(t\), with \(h_0=1\). Set \(h_t=0\) for \(t<0\). For weakly decreasing sequences \(\mu_1\geq\cdots\geq\mu_n\geq0\) and \(\nu_1\geq\cdots\geq\nu_n\geq0\) of integers, define the generalized Jacobi--Trudi matrix \(J(\mu,\nu)\) by \[ J(\mu,\nu)_{ij}=h_{\mu_i-\nu_j}. \] \begin{theorem} Every cluster variable \(f\) of \(\mathcal A_n\), and more generally every product of nonnegative integer powers of variables from one extended cluster, satisfies \[ f(J(\mu,\nu))\in\bigoplus_{\lambda}\mathbb Z_{\geq0}s_\lambda, \] where \(s_\lambda\) is the Schur function indexed by a partition \(\lambda\). \end{theorem} Zero is included. Specialization to a finite alphabet gives the corresponding assertion for Schur polynomials. \section{External results} \begin{enumerate} \item \textbf{Quantum matrix seed.} Geiss--Leclerc--Schroer \cite[\S 12.4, Corollary 12.10]{GLS} realize quantum \(n\times n\) matrices as a type-A quantum-unipotent cluster algebra. Grabowski--Launois \cite[\S 5, Definition 5.1 and Remark 5.2]{GL} give its initial minors and exchange quiver. \item \textbf{Canonical membership.} Kang--Kashiwara--Kim--Oh \cite[Theorem 11.2.7]{KKKO} place quantum cluster variables in the upper global basis up to powers of \(q^{1/2}\). Their Theorem 11.2.3 and Corollary 7.1.4 give the corresponding statement for cluster monomials, and Corollary 11.2.8 gives the integral cluster structure. The Cartan matrix here is symmetric. \item \textbf{Basis characterizations.} Kimura \cite[Theorem 4.26]{K} gives upper-global-basis triangularity relative to dual PBW vectors. Brundan \cite[\S 6, Theorem 16 and (6.2)]{B} characterizes the intrinsic quantum-matrix basis by its bar and terminal monomials. We compare their normalizations below using \cite[\S\S 6--7]{GLS}. \item \textbf{Immanant realization.} The polynomial matrix dual canonical basis consists of nonzero Kazhdan--Lusztig immanants with repeated rows and columns \cite[Theorem 2.1]{S}, \cite[Theorem 6.4]{SL}. \item \textbf{Schur positivity.} Rhoades--Skandera \cite[Proposition 2.3]{RS}, using Haiman \cite[Theorem 1.5]{H}, prove Schur nonnegativity for every Kazhdan--Lusztig immanant on every generalized Jacobi--Trudi matrix. \end{enumerate} \section{The matrix seed} \begin{proposition}The cluster variables from any double wiring diagram are the specializations at \(q=1\) of the variables from the GLS quantum matrix seed, with the same correspondence of extended clusters.\label{prop:seed}\end{proposition} \begin{proof} For \(n=1\), the seed consists of the frozen variable \(x_{11}\), so the seed identification is immediate. Assume \(n\geq2\). For \(1\leq r,s\leq n\), define the integer intervals \[ R(r,s)=[n-r+1,n-r+s]\cap[n],\qquad C(r,s)=[n-s+1,n-s+r]\cap[n]. \] These sets have equal cardinality. The seed of \cite[Definition 5.1]{GL} has variables \(\Delta_{R(r,s),C(r,s)}\), frozen precisely when \(r=n\) or \(s=n\). Its arrows are \[ (r,s)\longrightarrow(r,s+1),\qquad (r,s)\longrightarrow(r+1,s),\qquad (r,s)\longrightarrow(r-1,s-1), \] when the endpoints exist, with arrows between frozen vertices omitted. Number adjacent crossings from top to bottom by \(s_1,\ldots,s_{n-1}\), and use the reduced sorting word \[ (s_{n-1}\cdots s_1)(s_{n-1}\cdots s_2)\cdots s_{n-1} \] for each colour, performing the crossings in this word from left to right. Place all red crossings before all blue crossings. Number blue wires increasingly from top to bottom at the left boundary and red wires increasingly from top to bottom at the right boundary, following Cuenca's convention. At chamber level \(k\), meaning that \(k\) wires of each colour lie below the chamber, the red segment gives \[ \Delta_{[n-k+1,n],[j,j+k-1]},\qquad 1\leq j\leq n-k+1, \] and the blue segment gives \[ \Delta_{[i,i+k-1],[n-k+1,n]},\qquad n-k+1\geq i\geq1. \] The central southeast minor is counted once. After the first \(a\) descending blocks the blue labels are \(n,n-1,\ldots,n-a+1,1,\ldots,n-a\) from top to bottom, since each block moves the largest remaining label to the front of the remaining segment. The red label order is obtained by replacing each label \(u\) in this list by \(n+1-u\), because the initial red order is reversed. At level \(k\), only the crossing \(s_{n-k}\) in each of the first \(n-k\) blocks changes the bottom \(k\) labels. Reading those sets gives the two lists above. The nonunit chambers therefore number \[ \sum_{k=1}^{n}\bigl(2(n-k)+1\bigr)=n^2. \] They are exactly the minors indexed by \((r,s)\) above, and their unbounded chambers give the stated frozen variables. To compare arrows, set \(N=n(n-1)/2\). Within either sorting word, the \(a\)-th crossing at level \(k\) occurs at position \[ T(a,k)=(a-1)n-\frac{a(a-1)}2+k, \qquad 1\leq a\leq n-k. \] For \(1\leq k1 \] whenever the two successive crossings occur in these comparisons. Together with the same-level arrows, the directions are east, south and northwest. Since \[ (r,s)=(n-i+1,n-j+1), \] this is the global opposite of the quiver in \cite{GL}. Reversing all arrows interchanges the two numerator terms in each commutative exchange relation, and the rule for quiver mutation commutes with global reversal. Thus every mutation sequence produces the same commutative variables in the two seeds. Let \(w_0\) be the reversing permutation in the symmetric group on \([n]\). Double reduced words for \((w_0,w_0)\) are connected by braid, commutation and colour-interchange moves. The corresponding chamber seeds are mutation equivalent \cite[proof of Theorem 17 and Lemma 18]{FZ}, \cite[Definition 2.4.1 and Proposition 2.4.3]{FWZ}. In the generic matrix exchange identities of \cite{FZ}, the full determinant is retained as a common chamber minor. Thus the mutation equivalence identifies the polynomial matrix variables for every double-wiring diagram with the determinant kept as a frozen variable. In the integral quantum cluster algebra of \cite[Corollary 11.2.8]{KKKO}, specialize by sending \(q^{1/2}\) to \(1\). To check each exchange, assign the value \(1\) to all variables of the algebraically independent initial commutative cluster. The ordinary exchange rule then gives a positive value at every step, so each ordinary variable is a nonzero rational function. Inductively the specialized quantum identity determines the new variable uniquely, since the variable being replaced has a nonzero specialization. This identity is the ordinary exchange identity because its parameter powers specialize to \(1\). \end{proof} \section{The matrix dual canonical basis} \begin{proposition}In the GLS quantum matrix realization, the upper global basis and the intrinsic matrix dual canonical basis have the same specialization at \(q=1\) in the matrix coordinate integral form.\end{proposition} \begin{proof} Let \(q\) be an indeterminate and work first over \(\mathbb Q(q)\). The quantum matrix algebra has generators \(x_{ij}\) and relations \[ \begin{aligned} x_{ij}x_{ik}&=q x_{ik}x_{ij} &&(j\ell),\\ x_{ij}x_{k\ell}-x_{k\ell}x_{ij} &=(q-q^{-1})x_{i\ell}x_{kj} &&(i