On some Grothendieck expansions
Extracted source text ยท arXiv 2412.18963
\documentclass[10pt]{article}
\usepackage[backend=bibtex,doi=false,isbn=false,url=false,eprint=false]{biblatex}
\addbibresource{ogroth.bib}
\input{preamble.tex}
\definecolor{lightskyblue}{RGB}{176,226,255}
\newcommand{\customvarpi}{\varpi}
\newcommand{\eric}[1]{\todo[inline,size=\tiny,color=green!30]{#1 \\ \hfill --- E.}}
\newcommand{\Eric}[1]{\todo[size=\tiny,color=green!30]{#1 \\ \hfill --- E.}}
\newcommand{\Jiayi}[1]{\todo[size=\tiny,color=blue!30]{#1 \\ \hfill --- J.}}
% \usepackage{fullpage}
\begin{document}
\title{On some Grothendieck expansions}
\author{
Eric MARBERG\\
Department of Mathematics \\
Hong Kong University of Science and Technology \\
{\tt emarberg@ust.hk}
\and
Jiayi WEN\\
Department of Mathematics \\
University of Southern California \\
{\tt jiayiwen@usc.edu}
}
\date{}
\maketitle
\begin{abstract}
The complete flag variety admits a natural action by both the orthogonal group and the symplectic group. Wyser and Yong defined orthogonal Grothendieck polynomials $\mathcal{G}^{\mathsf{O}}_z$ and symplectic Grothendieck polynomials $\mathcal{G}^{\mathsf{Sp}}_z$ as the $K$-theory classes of the corresponding orbit closures. There is an explicit formula to expand $\mathcal{G}^{\mathsf{Sp}}_z$ as a nonnegative sum of Grothendieck polynomials $\mathcal{G}^{(\beta)}_w$, which represent the $K$-theory classes of Schubert varieties. Although the constructions of $\mathcal{G}^{\mathsf{Sp}}_z$ and $\mathcal{G}^{\mathsf{O}}_z$ are similar, finding the $\mathcal{G}^{(\beta)}$-expansion of $\mathcal{G}^{\mathsf{O}}_z$ or even computing $\mathcal{G}^{\mathsf{O}}_z$ is much harder. If $z$ is vexillary then $\mathcal{G}^{\mathsf{O}}_z$ has a nonnegative $\mathcal{G}^{(\beta)}$-expansion, but the associated coefficients are mostly unknown. This paper derives several new formulas for $\mathcal{G}^{\mathsf{O}}_z$ and its $\mathcal{G}^{(\beta)}$-expansion when $z$ is vexillary. Among other applications, we prove that the latter expansion has a nontrivial stability property.
\end{abstract}
\tableofcontents
\section{Introduction}
This article is concerned with combinatorial formulas for expansions of three different families of
\defn{Grothendieck polynomials} related to $K$-theory classes of Schubert varieties.
We first briefly introduce these polynomials in the context of \defn{matrix Schubert varieties}.
We will then discuss the central problem of interest and some of our partial solutions.
\subsection{Grothendieck polynomials}
Let $n$ be a positive integer. Write $S_n$ for the group of permutations of the integers $\ZZ$ with support in $[n]:=\{1,2,\dots,n\}$. Let $\Mat_{n\times n}$ be the affine variety of complex $n\times n$ matrices.
For $w \in S_n$,
the \defn{matrix Schubert variety} $\MX_w$ is the set of matrices $M \in \Mat_{n\times n}$
whose upper $i\times j$ sub-matrices $M_{[i][j]}$ satisfy
\[\rank(M_{[i][j]}) \leq |\{ t \in [i] : w(t) \leq j\}|\quad\text{for all $i,j\in [n]$.}\]
Results in \cite{KnutsonMiller} identify an equivariant $K$-theory class
\[[\MX_w] \in \ZZ[a_1^{\pm1},a_2^{\pm1},\dots,a_n^{\pm1}] \cong K_T(\Mat_{n\times n }). \]
The \defn{Grothendieck polynomial} $\fkG_w$ introduced in \cite{FominKirillov} can be formed from
$[\MX_w]$
by substituting $a_i \mapsto 1+\beta x_i$ for all $i \in [n] $
and dividing by $(-\beta)^{\codim(\MX_w)}$.
\begin{example}
The Grothendieck polynomial of the reverse permutation $w_0 = n\cdots 321 \in S_n$
is the monomial $\fkG_{w_0} = x_1^{n-1}x_2^{n-2}\cdots x_{n-1}$.
As will be explained in Section~\ref{formulas-sect}, all other instances of $\fkG_w$ for $w \in S_n$ can be computed by applying certain
\defn{divided difference operators} to $\fkG_{w_0}$.
When $n=3$ we have
\[
\ba
\fkG_{123} &=1, \\
\fkG_{231} &= x_1,
\ea
\qquad
\ba
\fkG_{132} &=x_1+x_2 + \beta x_1x_2, \\
\fkG_{231} &= x_1x_2,
\ea
\qquad
\ba
\fkG_{312} &=x_1^2, \\
\fkG_{321} &= x_1^2x_2.
\ea
\]
\end{example}
The Grothendieck polynomial $\fkG_w$ always belongs
to
$\NN[\beta][x_1,x_2,\dots]$ and is homogeneous of degree $\codim(\MX_w)$
if we consider $\beta $ to have $\deg \beta=-1$. This degree coincides with the value of the usual Coxeter length function
\be\label{ell-def}
\ell(w) := |\{ (i,j) \in \PP\times \PP : i<j\text{ and }w(i)>w(j)\}|.
\ee
In turn, if $w \in S_\infty := \bigcup_{m\geq 1} S_m$
then the value of $\fkG_w$ does
not
depend on the choice of $n$ with $w \in S_n$.
Finally,
it is known \cite[Cor.~3.3]{MP2021Sp} that the set of polynomials $\{\fkG_w : w \in S_\infty \}$ is a $\ZZ[\beta]$-basis for $ \ZZ[\beta][x_1,x_2,\dots]$.
\subsection{Orthogonal Grothendieck polynomials}
The main subject of this article are the following analogues of Grothendieck polynomials studied in \cite{MP2020,WyserYong}.
Let
\[\textstyle I_n = \{ z \in S_n : z=z^{-1}\} \quand I_\infty = \bigcup_{m\geq 1} I_m = \{ z \in S_\infty : z=z^{-1}\}\]
be the sets of involutions in $S_n$ and $S_\infty$.
For $z \in I_n$ define
\[ \Mat^\O_{n\times n} := \left\{ X \in \Mat_{n\times n} : X^\top = X\right\}
\quand \MX^\O_z = \MX_z \cap \Mat^\O_{n\times n}.\]
Results in \cite{MP2020} identify an equivariant $K$-theory class
\[[\MX^\O_z] \in \ZZ[a_1^{\pm1},a_2^{\pm1},\dots,a_n^{\pm1}] \cong K_T(\Mat^\O_{n\times n }). \]
Similar to $\fkG_w$, the \defn{orthogonal Grothendieck polynomial} $\fkGO_z$ is formed from
$[\MX^\O_z]$
by substituting $a_i \mapsto 1+\beta x_i$ for all $i $
and dividing by $(-\beta)^{\codim(\MX^\O_z)}$.
These polynomials were introduced in \cite{WyserYong} with a slightly different geometric interpretation,
as $K$-theory representatives for the orbit closures of the orthogonal group acting on the complete flag variety.
Each $\fkGO_z$ belongs to $\NN[\beta][x_1,x_2,\dots,x_n]$ and
does not
depend on the choice of $n$ with $z \in I_n$.
If we consider $\beta$ to have $\deg \beta=-1$ then $\fkGO_z$ is homogeneous
of degree $\codim(\MX^\O_z)$. This degree can be computed by
setting
\be\label{ellhat-eq}
\ba
%\ell(z) &:= |\{ (i,j) \in \PP\times \PP : i<j\text{ and }z(i)>z(j)\}|,
%\\
\cyc(z) := |\{ a \in \PP : a<z(a)\}|
\quand
\ellhat(z) := \tfrac{\ell(z) + \cyc(z)}{2} .
\ea
\ee
Then \cite[Lem.~5.4]{Pawlowski-universal} we have
$ \codim(\MX^\O_z)=\ellhat(z) \in \NN$.
\begin{example}
If $w_0 = n\cdots 321 \in I_n$
then there is an explicit product formula
\be\textstyle \fkGO_{w_0} = \prod_{1 \leq i \leq j \leq n-i} (x_i+x_j+\beta x_ix_j).\ee
When $z \in I_n$ is \defn{vexillary} in the sense of being $2143$-avoiding,
then $\fkGO_z$ can be computed by certain applying
divided difference operators to $\fkGO_{w_0}$,
as will be explained in Section~\ref{formulas-sect}.
For $n=3$ such computations give the formulas
\[
\ba
\fkGO_{1} & = 1 = \fkG_{123}, \\[-10pt]\\
\fkGO_{(1,2)} &= 2x_1 + \beta x_1^2 = 2\fkG_{213} + \beta \fkG_{312}, \\[-10pt]\\
\fkGO_{(2,3)} &= 2x_2 + 2x_1 + \beta x_2^2 + 4\beta x_1 x_2 + \beta x_1^2 + 2 \beta^2 x_1 x_2^2 + 2 \beta^2 x_1^2 x_2 + \beta^3 x_1^2 x_2^2 \\ &= 2\fkG_{132} +\beta \fkG_{231}+ \beta \fkG_{1423} + \beta^2 \fkG_{2413}, \\[-10pt]\\
\fkGO_{(1,3)} &= 2x_1 x_2 + 2x_1^2 + 3 \beta x_1^2 x_2 +\beta x_1^3 + \beta^2 x_1^3 x_2
\\&=2\fkG_{231} + 2\fkG_{312} + 3\beta \fkG_{321} + \beta \fkG_{4123} + \beta^2\fkG_{4213}.
\ea
\]
\end{example}
For involutions $z \in I_\infty$ that are not vexillary,
no simple algebraic method is available for computing $\fkGO_z$.
However, at least in the vexillary case, it is known \cite[Prop.~3.29]{MS2023} that
$ \fkGO_z $ is a linear combination of ordinary Grothendieck polynomials $\fkG_w$ with coefficients in $\NN[\beta]$.
%
The nontrivial part of this fact is the non-negativity of the coefficients, which follows from a general geometric result of Brion \cite{Brion2002}.
It is an open problem to describe
the $\fkG$-expansion of $\fkGO_z$ explicitly. This problem is our focus here.
%
%\begin{example}
%The involution $z = (1,4)(2,5)$ is dominant of shape $ (3,3,2)$
%with
%\[\ba
% \fkGO_z&= (x_1\oplus x_1)(x_1\oplus x_2)(x_1\oplus x_3)(x_2\oplus x_2)(x_2\oplus x_3) \\
%&=4 \fkG_{25314} +
%4 \fkG_{35124} +
%4 \fkG_{24513} \\&\quad+
%\beta\(2 \fkG_{45123} +
%2 \fkG_{361245} +
%2 \fkG_{34512} +
%2 \fkG_{263145} +
%6 \fkG_{25413} +
%6 \fkG_{35214} \)\\&\quad+
% \beta^{2} \(\fkG_{461235} +
%2 \fkG_{264135} +
%3 \fkG_{35412} +
%3 \fkG_{362145} +
%3 \fkG_{45213}\)\\&\quad +
% \beta^{3} \(\fkG_{364125} +
% \fkG_{45312} +
%2 \fkG_{462135}\) \\&\quad+
% \beta^{4} \fkG_{463125}.
% \ea
% \]
% \end{example}
\subsection{Symplectic Grothendieck polynomials}
It is instructive to contrast this problem with what is known about the following \defn{symplectic Grothendieck polynomials}.
Let $\Ifpf_n$ be the set of fixed-point-free involutions of $\ZZ$
sending $i \mapsto i-1$ for all even $i \notin [n]$.
This set is in bijection with the fixed-point-free involutions in $S_n$ when $n$ is even.
%Observe that $\Ifpf_n$ is empty if $n$ is odd. % and define $\Ifpf_\infty = \bigcup_{n\geq 0} \Ifpf_n$.
%In examples, we write elements of $\Ifpf_n$ as permutations of $[n]$. % and omit the trailing cycles $(2i-1,2i)$ for $i>n$.
%
For $z \in\Ifpf_n$ define
\[ \Mat^\Sp_{n\times n} := \left\{ X \in \Mat_{n\times n} : X^\top = -X\right\}
\quand \MX^\Sp_z = \MX_z \cap \Mat^\Sp_{n\times n}.\]
As in the previous two cases, there is an equivariant $K$-theory class \cite{MP2020}
\[ [\MX^\Sp_z] \in \ZZ[a_1^{\pm1},a_2^{\pm1},\dots,a_n^{\pm1}] \cong K_T(\Mat^\Sp_{n\times n }). \]
The \defn{symplectic Grothendieck polynomial} $\fkGSp_z$ is formed from
the class $[\MX^\Sp_z]$ % \in K_T(\Mat^\Sp_{n\times n }) = \ZZ[a_1^{\pm1},a_2^{\pm1},\dots,a_n^{\pm1}]\]
by substituting $a_i \mapsto 1+\beta x_i$ for all $i \in [n]$
and dividing by $(-\beta)^{\codim(\MX^\Sp_z)}$.
Like $\fkGO_z$, these polynomials were first considered in \cite{WyserYong} as $K$-theory representatives for a certain set of orbit closures
in the complete flag variety, now for the action of the symplectic group.
Also like $\fkGO_z$,
% the resulting polynomial belongs to
each
$\fkGSp_z$ belongs to $\NN[\beta][x_1,x_2,\dots]$, %if $z \in \Ifpf_\infty:= \bigcup_{n\geq 1} \Ifpf_n$ then $\fkGSp_z$
is homogeneous of degree $\codim(\MX^\Sp_z)$ if we consider $\deg\beta=-1$,
and does not
depend on the choice of $n$ with $z \in \Ifpf_n$.
The polynomials $\fkGSp_z$ are more tractable than $\fkGO_z$ in several respects.
Each $\fkGSp_z$ can be determined inductively by applying divided difference operators to a product formula corresponding to the reverse permutation \cite[\S2.1]{WyserYong}.
There is also a positive $\fkG$-expansion of each $\fkGSp_z$, which can be constructed explicitly in the following way.
Write $w_i=w(i)$ for $w \in S_\infty$. Then let $\approx$ be the transitive closure of the relation
on $S_\infty$ with $v^{-1} \approx w^{-1}$ if some $i \in 2\NN$
and $a<b<c<d$ have
\[ \{v_{i+1}v_{i+2}v_{i+3}v_{i+4},\ w_{i+1}w_{i+2}w_{i+3}w_{i+4}\} \subseteq \{adbc,\ bcad,\ bdac\},\] while $v_j=w_j$ for all $j \notin \{i+1,i+2,i+3,i+4\}$.
For example,
%in one-line notation, we have
\[
(\hs15{2634}\hs)^{-1} \approx (\hs15{3624}\hs)^{-1} \approx (\hs{15{34}}{26}\hs)^{-1} \approx (\hs{3415}26\hs)^{-1} \approx (\hs{3514}26\hs)^{-1}.
\]
%Here, the underline and overline annotations indicate the relevant subwords $ v_{i+1}v_{i+2}v_{i+3}v_{i+4}\approx w_{i+1}w_{i+2}w_{i+3}w_{i+4}$ that manifest each equivalence.
%
Given $z \in \Ifpf_n$ let $a_1<a_2<\dots$ be the integers with $0<a_i<b_i:=z(a_i ) $
and define
$\alpha_\fpf(z)$ to be inverse of $a_1 b_1a_2 b_2 \cdots \in S_\infty.$
% Set $ \cBfpf(z) = \{ w \in S_\infty : w \approx \alpha_\fpf(z)\}$.
Then %, writing $\ell : S_\infty \to \NN$ for the usual Coxeter length function,
we have \cite[Thm.~3.12]{MP2020}
\be\textstyle\label{fkgsp-eq}
\fkGSp_z = \sum_{w\approx \alpha_\fpf(z)} \beta^{\ell(w) - \ellfpf(z)} \fkG_w
\quad\text{where }\ellfpf(z) := \ell(\alpha_\fpf(z)).
\ee
In particular, for each $z \in \Ifpf_n$ there is a map $\GCSp_z : S_\infty \to \{0,1\}$ with
\be\textstyle
\fkGSp_z = \sum_{w \in S_\infty} \GCSp_z(w) \cdot \beta^{\ell(w) - \ellfpf(z)} \cdot \fkG_w.
\ee
The formula \eqref{fkgsp-eq} is a special case of \cite[Thm.~3]{Knutson2009},
which gives a general method of decomposing $K$-theory classes of \defn{multiplicity-free subvarieties}
of generalized flag manifolds. The expansion in \cite{Knutson2009} is somewhat inexplicit as it involves coefficients that may be zero,
and in any case the method
cannot be applied to decompose $\fkGO_z$, which does not represent a multiplicity-free variety.
\subsection{Summary of results}\label{GC-eq-sect}
Both $\fkG_w$ and $\fkGO_z$ are homogeneous if we set $\deg \beta =-1$.
As $ \left\{ \fkG_w : w \in S_\infty \right\}$ is a $\ZZ[\beta]$-basis for $\ZZ[\beta][x_1,x_2,\dots]$,
it follows that
for each $z \in I_\infty$
there is a unique coefficient function
% \be\label{GC-eq}
$\GC_{z} : S_\infty \to \ZZ$
%\ee
with
\be\textstyle
\fkGO_z = \sum_{w \in S_\infty} \GC_z(w) \cdot \beta^{\ell(w) - \ellhat(z)} \cdot \fkG_w.
\ee
%
% \be
% \fkGO_z = \sum_{w \in S_\infty} \GC_z(w) \cdot \beta^{\ell(w) - \ellhat(z)} \cdot \fkG_w.
% \ee
%This holds as the set $\{\fkG_w : w \in S_\infty\}$ is a $\ZZ[\beta]$-basis for $\ZZ[\beta][x_1,x_2,\dots]$
%and since $\fkGO_z$ and $\fkG_w$ are both homogeneous, of respective degrees $\ellhat(z)$ and $\ell(w)$, when $\deg(x_i) = 1$ and $\deg(\beta)=-1$. These facts also imply that
The \defn{support} of this map given by
$
\supp(\GC_z) := \left\{ w \in S_\infty : \GC_z(w) \neq 0\right\}
$
is always a finite set of permutations.
In contrast to $\GCSp_z$,
no simple method is known to compute this set in an efficient way.
When $z \in I_\infty$ is vexillary, as noted earlier, the function $\GC_z$ takes all values in $\NN$.
More precisely,
define the \defn{visible descent set} of $z \in I_\infty$ to be
\be\label{vdes-def} \DesV(z) := \{i \in \PP : z(i) > z(i+1) \leq i\},\ee
and define the \defn{right descent set} of a permutation $w\in S_\infty$ to be
\be\DesR(w) := \{ i \in \PP : w(i)>w(i+1)\}.\ee
Then for any vexillary $z \in I_\infty$ with $\DesV(z) \subseteq [n]$,
one has \cite[Prop.~3.29]{MS2023}
\be\label{vdes-eq}
\fkGO_z \in\NN[\beta]\spanning\left\{\fkG_w : \DesR(w) \subseteq[n]\right\} \subseteq \ZZ[\beta][x_1,x_2,\dots,x_n]
.\ee
Whereas $\GCSp_z$ takes all values in $\{0,1\}$,
the maximum value of $\GC_z$ is unbounded as $z$ varies over all vexillary involutions.
%Outside the vexillary case we have very little information about $\GC_z$.
We mention that if one sets $\beta=0$ then $\fkG_w$, $\fkGO_y$, and $\fkGSp_z$
turn into the \defn{(involution) Schubert polynomials} $\fkS_w$, $\iS_y$, and $\Sfpf_z$ studied in \cite{HMP1,Manivel,WyserYong},
which represent cohomology classes of matrix Schubert varieties rather than $K$-theory classes.
For these polynomials, the relevant expansions are all much simpler.
Both $\iS_y$ and $\Sfpf_z$ are equal to a constant %(which is $2^{\cyc(y)}$ in the first case and $1$ in the second case)
times a multiplicity-free sum of Schubert polynomials $\fkS_w$. Moreover, the terms that appear are predicted by a general formula of Brion \cite{Brion}
and are described combinatorially in \cite{CJW}.
This work contains the first concrete results about the more general $\fkG$-expansion of $\fkGO_z$,
or equivalently the Schubert class expansion of the $K$-theory classes of symmetric matrix Schubert varieties.
Our main theorems can be summarized as follows.
After reviewing some preliminaries, we prove in Section~\ref{prelim-sect}
that the family of polynomials $\fkGO_z $ indexed by vexillary involutions is completely determined by a system of divided difference equations.
Section~\ref{exp1-sect} then derives two exact formulas for $\GC_z $ when $z$ is any \defn{quasi-dominant} involution (see Theorems~\ref{qd-thm} and \ref{dom-thm}).
Several ingredients are needed to prove these theorems,
including the \defn{transition formulas} in \cite{Lenart,LenartSottile} and new identities for certain \defn{involution Grothendieck polynomials} introduced in \cite{MP2021}.
Section~\ref{ex-conj-sect} also presents a general conjecture about the support of $\GC_z$ for any vexillary $z \in I_\infty$,
and a specific prediction of this set for reverse permutations.
In Section~\ref{vex-sect} we explain a new formula for the monomial expansion of $\fkGO_z$ when $z $ is any vexillary involution (see Theorem~\ref{ivex-thm}).
This somewhat complicated result is complementary to an even more complicated Pfaffian formula for the same polynomials
that was derived in \cite{MP2020}.
Our new theorem can be viewed as the first instance of a $K$-theoretic extension of Brion's cohomology formula in \cite{Brion} beyond the multiplicity-free case considered in \cite{Knutson2009}.
%Our new identity follows by an inductive argument using the results in Section~\ref{exp1-sect} as the base case.
This formula may be useful in future work for
determining the \defn{Castelnuovo--Mumford regularity} of symmetric matrix Schubert varieties.
As an application, we show that the
$\fkG$-expansion of $\fkGO_z$ has a nontrivial \defn{shift invariance} property when $z$ is vexillary (see Theorem~\ref{supp-thm}).
This property leads to a new proof of the existence of \defn{stable limits} of orthogonal Grothendieck polynomials,
and affirms \cite[Conj.~3.39]{MS2023} in the vexillary case.
We also derive a number of interesting identities in special cases.
For example, Section~\ref{gr-sect} presents a more explicit formula for $\fkGO_z$ when $z$ is \defn{I-Grassmannian}.
This formula is a non-symmetric generalization of the main theorem in \cite{Chiu.Marberg2023},
which expands Ikeda--Naruse's \defn{$K$-theoretic Schur $Q$-functions} $\GQ_\lambda$ in terms of the \defn{$K$-theoretic Schur $P$-functions} $\GP_\mu$.
\subsection*{Acknowledgments}
This work was partially supported by Hong Kong RGC grants 16306120 and 16304122.
We thank Brendan Pawlowski
for many useful conversations.
\section{Preliminaries}\label{prelim-sect}
This section reviews some facts about divided difference operators and related formulas for $\fkG_w$ and $\fkGO_z$. Then we
develop some properties of a natural \defn{weak order} on
vexillary involutions. Finally, we discuss a variant of $\fkGO_z$ from \cite{MP2021}.
%So far we have not discussed any method of computing $\fkG_w$ or $\fkGO_z$ as polynomials,
%let alone the expansion of the latter in terms of the former.
%This section quickly reviews an algebraic method to compute $\fkG_w$, which can be adapted to $\fkGO_z$ when $z$ is vexillary.
\subsection{Divided difference operators}
Continue to let $\beta$ and $x_i$ be commuting indeterminates.
The group $S_\infty$ acts on $\ZZ[\beta][x_1,x_2,\dots]$ by permuting the $x_i$ variables.
For each $i \in \PP$ the \defn{divided difference operators} $\partial_i$ and $\bpartial_i$ act on $ \ZZ[\beta][x_1,x_2,\dots]$
by the formulas
\be
\partial_i f = \tfrac{f-s_i f}{x_i-x_{i+1}} \quand \bpartial_i f = \partial_i((1+\beta x_{i+1}) f) = -\beta f + (1+\beta x_i) \partial_i f.
\ee
These operators satisfy the Coxeter braid relations for $S_\infty$ along with $\partial_i\partial_i = 0$ and $\bpartial_i \bpartial_i = -\beta \bpartial_i$.
There is a useful product formula
\be\label{fg-eq}
%\partial_i(fg) = s_i f \cdot \partial_i g + \partial_i f \cdot g\quand
\bpartial_i(fg) = (s_i f) \cdot \(\bpartial_i g+\beta g\) + (\bpartial_i f)\cdot g
\ee
which holds for any $f,g\in \ZZ[\beta][x_1,x_2,\dots]$. This implies that if $s_if =f$ then
\be\label{fg-eq2} \bpartial_i(fg) = f \cdot \bpartial_i g.\ee
\subsection{Product formulas and recurrences}\label{formulas-sect}
The \defn{Rothe diagram} of a permutation $w \in S_\infty$ is the set
$ D(w)$ of positive integer pairs $ (i, j)$ satisfying both $ i<w^{-1}(j)$ and $j< w(i)$.
A \defn{partition} is a weakly decreasing sequence of nonnegative integers $\lambda = (\lambda_1\geq \lambda_2\geq \dots \geq 0)$
with a finite sum.
A permutation $w \in S_\infty$ is \defn{dominant}
if
there is a partition $\lambda$ such that
$D(w) =\D_\lambda$ where $\D_\lambda := \{(i,j) : 1 \leq j \leq \lambda_i\}.$
In this case we say that $w$ is dominant of shape $\lambda$.
Every dominant permutation is $132$-avoiding \cite[Ex.~2.2.2]{Manivel} and therefore also vexillary.
% since a permutation is dominant if and only if it is $132$-avoiding \cite[Ex.~2.2.2]{Manivel}.
There is a unique dominant $w \in S_\infty$ of each partition shape $\lambda$,
and results in \cite{FominKirillov} show that the Grothendieck polynomial of this permutation is
\be\label{dom-fk-eq}
\textstyle \fkG_w = \prod_{(i,j) \in \D_\lambda} x_i . \ee
% Let $\DesR(w) := \{ i \in \ZZ : w(i)>w(i+1)\}$ be the \defn{right descent set} of a permutation $w\in S_\infty$.
Then for any $i \in \PP$ and $w \in S_\infty$ we have \cite{FominKirillov}
\be\label{Grothendieck-recursion}
\bpartial_i \fkG_w = \begin{cases} \fkG_{ws_i} &\text{if }i \in\DesR(w) \\ -\beta \fkG_w&\text{if }i \notin \DesR(w).\end{cases}
\ee
Combined with \eqref{dom-fk-eq}, this recursive formula determines $\fkG_w$ for all $w \in S_\infty$.
The dominant permutation $z$ of shape $\lambda$ belongs to $I_\infty$
if and only if $\lambda=\lambda^\top$ is a \defn{symmetric} partition, since $D(z) = D(z^{-1}) = D(z)^\top$.
In this case
%If $z \in I_\infty$ is dominant of shape $\lambda$ then
%$z$ is also vexillary, and
%and one has $i<z(i)$ if and only if $(i,i) \in \D_\lambda$. For the dominant involution $z$ of shape $\lambda$,
\be\label{idom-eq} \fkGO_z = \prod_{\substack{(i,j) \in \D_\lambda \\ i\leq j}} x_i\oplus x_j
\quad\text{where $x\oplus y := x+y+\beta xy$}\ee by \cite[Thm.~3.8]{MP2020}.
% Moreover, if $z \in I_\infty$ is vexillary and $i \in \DesR(z)$ is such that $s_i z s_i\neq z$ is also vexillary, then the formula $\bpartial_i \fkGO_z=\fkGO_{s_izs_i}$ also holds \cite[Prop.~3.23]{MP2020}.
Moreover, if $z \in I_\infty$ is vexillary then
\be\label{orthogonal-recursion}\bpartial_i \fkGO_z = \begin{cases} \fkGO_{s_izs_i} &\text{if }i \in\DesR(z) \\ -\beta \fkGO_z&\text{if }i\notin\DesR(z)\end{cases}
\ee
for all $i \in \PP$ such that $s_i z s_i $ is also vexillary and not equal to $z$ \cite[Prop.~3.23]{MP2020}.
%If these conditions fail, then the formula \eqref{orthogonal-recursion} may no longer be true.
%However,
The recurrence \eqref{orthogonal-recursion} combined with the product formula \eqref{idom-eq}
can be used to calculate $\fkGO_z$ for any vexillary $z \in I_\infty$.
This claim is not obvious, but will be a consequence of Theorem~\ref{vexweak-prop} in the next subsection.
For involutions $z \in I_\infty$ that are not vexillary,
no simple algebraic method using divided difference operators is known for computing $\fkGO_z$.
See the discussion in \cite[\S2.2]{WyserYong}, especially \cite[Examples~2.5 and 2.6]{WyserYong}.
%In particular, these polynomials cannot be expressed using $\bpartial_i$'s.
%We mention that by contrast, the polynomials $\fkGSp_z$ are completely determined by analogues of \eqref{idom-eq} and \eqref{orthogonal-recursion};
% see \cite[Thm.~3.8 and Prop.~3.11]{MP2020}.
\subsection{Vexillary involutions and weak order}
Let $\Ivex$ be the subset of vexillary elements of $I_\infty$,
that is, involutions in $S_\infty$ that are $2143$-avoiding.
This section discusses some concrete ways of identifying the elements of this set. Both \cite{GPP} and \cite[\S4]{Pawlowski-universal} are useful references for this material.
It is sometimes useful to draw involutions $z \in I_\infty$ as \defn{arc diagrams},
which are incomplete matchings on the ordered set $\{1,2,3,\dots,n\}$ in which distinct vertices $i$ and $j$ are connected if and only if $z(i)=j$.
The value of $n$ should be chosen such that $z(i)=i$ for all $i>n$, but we do not always require $n$ to be the minimal integer with this property.
In drawing arc diagrams, we order the vertices from left to right and omit vertex labels. For example,
\[
z = (1,6)(2,4)(5,7)\quad\text{has arc diagram}\quad
\arcstart
{
*{\bullet} \arc{1}{rrrrr} & *{\bullet} \arc{0.5}{rr} & *{\bullet} & *{\bullet} & *{\bullet} \arc{0.5}{rr} & *{\bullet} & *{\bullet}
}
\arcstop.
\]
\begin{remark}\label{arc-vex-rem}
%Recall that a permutation is vexillary if it is $2143$-avoiding.
The $2143$-pattern avoidance condition that characterizes $\Ivex$ is equivalent to the following property of arc diagrams \cite[Fig.~3]{GPP}:
an element of $ I_\infty$ is vexillary if and only if its arc diagram does not contain any of the following as induced vertex-ordered subgraphs:
\[
\arcstart
{
*{\bullet} \arc{0.4}{r} & *{\bullet} & *{\bullet} \arc{0.4}{r} & *{\bullet}
}
\arcstop
\quad
\arcstart
{
*{\bullet} \arc{0.6}{rrr} & *{\bullet} & *{\bullet} \arc{0.6}{rrr} & *{\bullet} & *{\bullet} & *{\bullet}
}
\arcstop
\quad
\arcstart
{
*{\bullet} \arc{0.8}{rrrr} & *{\bullet} & *{\bullet} \arc{0.8}{rrrr} & *{\bullet} \arc{0.5}{rr} & *{\bullet} & *{\bullet} & *{\bullet}
}
\arcstop
\quad
\arcstart
{
*{\bullet} \arc{0.8}{rrrr} & *{\bullet} \arc{0.5}{rr} & *{\bullet} \arc{0.8}{rrrr} & *{\bullet} & *{\bullet} & *{\bullet} & *{\bullet}
}
\arcstop
\quord
\arcstart
{
*{\bullet} \arc{1}{rrrrr} & *{\bullet} \arc{0.6}{rrr} & *{\bullet} \arc{1}{rrrrr} & *{\bullet} \arc{0.6}{rrr} & *{\bullet} & *{\bullet} & *{\bullet} & *{\bullet}
}
\arcstop.
\]
\end{remark}
Define the \defn{vexillary weak order graph} on $\Ivex$ to be the directed graph with labeled edges $v \xrightarrow{i} w$
if $v,w \in \Ivex$ and $i \in \PP$ are such that
\[w=s_i vs_i\quand \ell(v)<\ell(w).\]
One has $\ell(v)<\ell(s_i vs_i)$ precisely when $i$ and $i+1$ are not both fixed points of $v$ and $v(i)<v(i+1)$.
In this case one always has
$\ell(s_ivs_i) = \ell(v)+2$ but it is possible that $s_i v s_i $ is not itself vexillary.
For integers $p,q \in \PP$ with $q \geq 2p$, let
\be\label{dom-def} \dom_{pq} := (1,q)(2,q-1)(3,q-2)\cdots(p,q-p+1) \in \Ivex.\ee
Also define $ \dom_{00} = 1$.
The permutation $\dom_{pq}$ is dominant of shape
\[
\lambda = (q-1,q-2,q-3,\dots,q-p,\underbrace{p,p,p,\dots,p}_{q-2p\text{ terms}},p-1,p-2,p-3,\dots,1).
\]
Finally, given $z \in \Ivex$,
we set $p=q=0$ if $z=1$ and otherwise define
\be\label{pq-eq} p =\cyc(z) = |\{ i \in \PP : i<z(i)\}| \quand
q = \max\{ i \in \PP : z(i)\neq i\}.\ee
Notice that $q \geq 2p$.
The following essential property of $\Ivex$ is possibly known to experts, but we have not been able to find a reference in the literature. %We include a proof below.
\begin{theorem}\label{vexweak-prop}
Let $z \in \Ivex$ with $p$ and $q$ as in \eqref{pq-eq}.
Then the vexillary weak order graph contains a path $z\xrightarrow{i_1} \cdots \xrightarrow{i_k} \dom_{pq}$ with $\{i_1,\dots,i_k\} \subseteq [q-1]$.
Moreover, if $i<z(i)$ for all $ i \in [p]$, then this path %$z\xrightarrow{i_1} \cdots \xrightarrow{i_k} \dom_{pq}$
must have $p\notin\{ i_1,\dots,i_k\}.$
\end{theorem}
This result will be proved in two steps. For $i \in \PP$ and $w \in S_\infty$ define
\be D_i(w) := \{ j : (i,j) \in D(w)\}
\quand
\iD(w):=\{(i,j)\in D(w): i\leq j\}.\ee
Then for any finite subset $S \subset\PP\times \PP$ let
\be
\Ess(S):=\{(i,j)\in S: (i+1,j)\notin S\text{ and }(i,j+1)\notin S\}.
\ee
Note that $z\in I_\infty$ is completely determined by $D(z)$, which by symmetry is itself determined by $\iD(z)$.
It is useful to recall that a permutation $w \in S_\infty$ is vexillary if and only if the family of sets $\{ D_i(w): i \in \PP\}$ is totally ordered under inclusion \cite[\S2]{Pawlowski-universal}.
In addition, an involution
$z\in I_\infty$ is vexillary if and only if $\Ess(\iD(z))$ is totally ordered under the partial order that has $(i,j)\preceq (k,l)$ whenever $i\geq k$ and $j\leq l$ \cite[Lem.~4.18]{Pawlowski-universal}.
For $w\in S_\infty$, define $c(w) := (c_1,c_2,c_3,\dots)$ where $c_i := |D_i(w)|$.
It is well-known that $c(w)$ uniquely determines $w$. In particular, $w$ is dominant of shape $\lambda$ if and only if $c(w)=\lambda$.
One has
%The sequence $(c_1,c_2,\cdots)$ uniquely determines $z$ because $z(i) $ is the $(c_i+1)$-th integer in $\NN\setminus\{z(1),z(2),\cdots,z(i-1)\}$. Therefore
$c_i\leq c_{i+1}$ if and only if $w(i)<w(i+1)$. %, and with equality if $\{z(i)+1,\cdots,z(j+1)-1\}\subset\{z(j)\mid 1\leq j<i\}$.
Fix $z \in \Ivex$.
% Assume $z$ is not dominant.
Let $m := \max\{ c_{i+1} : i\in \PP\text{ and } c_i < c_{i+1}\}$ and $j := \min\{ i \in \PP : c_i<c_{i+1} = m\}$.
We call $j$ the \defn{distinguished ascent} of $z$.
Note that $m$ and $j$ are not defined only if $c(z)$ is a partition, i.e. $z$ is dominant.
\begin{lemma}\label{vexweak-path-lem}
If $z \in \Ivex$ is not dominant, then
% its distinguished ascent $j$ is defined and it holds that
$j < \max\{ i \in \PP : z(i) \neq i\}$ and $z \xrightarrow{j} s_j zs_j \in \Ivex$.
\end{lemma}
\begin{proof}
Suppose $z\in \Ivex$ is not dominant and let $q= \max\{i\in \ZZ_{>0}:z(i)\neq i\}$,
so that $z\in S_q$ and $D(z)\subset [q]\times [q]$.
Define $m$ and $j$ as above, and let \[N=\max (D_{j+1}(z)).\]
Because $c_{l+1}=0$ for all $l\geq q$, we must have $j<q$.
% It remains to show that $s_jzs_j\in \Ivex$.
As $c_j<c_{j+1}$, we must also have $z(j)<z(j+1)$ and there exists some $(j+1,l)\in D(z)$ such that $z(j)<l<z(j+1)$. We claim that
$ j\leq z(j).
$
%
To show this, we first check that
\be\label{claim-vex2} \varnothing\neq \{z(j)+1,z(j)+2,\cdots,N-1\}= D_{j+1}(z)\setminus D_{j}(z).\ee
Because $z$ is vexillary, the family of sets $\{D_{i}(z): i\in \PP\}$ is totally ordered. Hence $c_j<c_{j+1}$ implies that $D_j(z)\subset D_{j+1}(z)$, and it suffices to show that
\[\{z(j)+1,z(j)+2,\cdots,N-1\}\subset D_{j+1}(z)\] because $z(j)>\max (D_{j}(z))$. If $i<j$ and $z(j)\in D_i(z)$, then $D_{j+1}(z)\subset D_{i}(z)$ and so $z(i)>N$. However, if $i<j$ and $z(j)\notin D_i(z)$ then $z(i)<z(j)$. Therefore \eqref{claim-vex2} holds.
%$\{z(j)+1,z(j)+2,\cdots,\max D_{j+1}(z)-1\}\subset D_{j+1}(z)$.
Now we prove that $j\leq z(j)$ by
assuming $z(j)<j$ and deriving a contradiction. By the preceding paragraph, we have
\[(j+1,z(j)+1)\in D_{j+1}(z)\quad\text{and also}\quad (z(j)+1,j+1)\in D(z)\] by the symmetry of $D(z)$. Since $z(j)\notin D_{j+1}(z)$ and $z(j)\in D_{z(j)}(z)$, we must have $c_{z(j)+1}>c_{j+1}$. Because $z(j)<j$ is not the distinguished ascent, we have
\[c_{z(j)}\geq c_{z(j)+1}\quad\text{and therefore}\quad D_{z(j)+1}(z)\subset D_{z(j)}(z).\] But this is impossible as $j+1\in D_{z(j)+1}(z)\subset D_{z(j)}(z)$ and $j+1>j=z(z(j))$. Therefore our claim that $j\leq z(j)$ must hold.
Finally, we prove that $s_jzs_j\in \Ivex$ by showing that $\Ess(\iD(s_jzs_j))$ is totally ordered under $\preceq$. We observe that $D(s_jzs_j)$ and $D(z)$ differ only in the rows and columns indexed by $j$ and $j+1$. Because $z(j)\geq j$ and $z(j+1)>j+1$,
the half-diagrams $\iD(z)$ and $\iD(s_jzs_j)$ have the same columns $j$ and $j+1$. Moreover,
\[\iD_{j+1}(s_jzs_j)=\iD_j(z)\quand \iD_j(s_jzs_j) = \iD_{j+1}(z)\cup \{z(j)\}.\]
Because $(j,z(j)+1)\in \iD(s_jzs_j)$, the pair $(j,z(j))$ cannot be in $\Ess(\iD(s_jzs_j))$ and $\Ess(\iD(s_jzs_j))\setminus \Ess(\iD(z))$ has at most an element. In fact:
\bei
\item If $(j+1,N)\notin \Ess(\iD(z))$ then $\Ess(\iD(z)) = \Ess(\iD(s_jzs_j))$.
\item If $(j+1,N)\in \Ess(\iD(z))$ then $(j,N)$ is in $\Ess(\iD(s_jzs_j))$, but this set is still totally ordered as any $(k,l)\succeq (j+1,N)$ must have $k<j$ and $l\geq N$.
\eei
We conclude that $\Ess(\iD(s_jzs_j))$ is totally ordered under $\preceq$ so $s_iz s_i \in \Ivex$.
\end{proof}
Now suppose $z \in \Ivex$ is dominant and has $p>0$ nontrivial cycles.
Define
\be\label{hat-d-eq}
\iD_i(z) := \{ j : (i,j) \in \iD(z)\}
\quand \hat c(z) := (\hat c_1, \hat c_2,\hat c_3,\dots)
\ee
where $\hat c_i := |\iD_i(z)|$.
Because $z$ is dominant, we have
\be
D_i(z) = [i-1+\hat c_i]\quand\iD_i(z) = i-1+ [\hat c_i]\quad\text{for all $i\leq p$,}
\ee
along with $\hat c_i = 0$ and $z(i)\leq i$ for all $i>p$.
%
% Then it is known \cite{?} that
% $D_i(z) = [i-1+\hat c_i]$ and $\iD_i(z) = i-1+ [\hat c_i]$ all $i\leq p$, that
% $\hat c(z)$ is a strict partition
% which uniquely determines $z$,
% and that $\hat c_i = 0$ and $z(i)\leq i$ for all $i>p$.
% %and that $\ell(z) = 2(\hat c_1 + \hat c_2 + \dots + \hat c_p) - p$.
%
% Assume $z$ is dominant.
% Suppose $p>0$ and define $q = \max\{i \in \PP : z(i)\neq i\}$.
Continue to define $q$ as in \eqref{pq-eq}. %\[q = \max\{i \in \PP : z(i)\neq i\}.\]
Then we must have
$D(z) \subseteq \{ (i,j) : i+j\leq q\}$ so $\hat c_i \leq q-2i+1$ for all $i \in [p]$,
%\hat c(z) \subseteq (q-1,q-3,q-5,\dots,q-2p+1)\]
with equality for all $i$ if and only if $z=\dom_{pq}$.
Hence, if $z$ is dominant but not equal to $ \dom_{pq}$ then some minimal $i \in [p]$ has $\hat c_i < q - 2i+1$;
we refer to $j := z(i)$ as the \defn{dominant ascent} of $z$.
Let $\e_i$ be the tuple with $1$ in position $i$ and zeros in all other positions.
\begin{lemma}\label{dominant-path-lem}
Let $z \in \Ivex\setminus\{1\}$ be dominant. Define $p$ and $q$ as in \eqref{pq-eq} and assume $z \neq \dom_{pq}$.
Let $j$ be the dominant ascent of $z$ and set $i = z(j)$.
Then $j\leq q-i $ and $s_j z s_j $ is dominant with $\ell(s_j z s_j) = \ell(z) + 2$ and $\hat c(s_j z s_j) = \hat c(z) + \e_i$.
\end{lemma}
\begin{proof}
Since $\hat c_t = q-2t+1$ for each $t \in [i-1]$,
we must have
$z(t) = q-t+1$ for all $t \in [i-1]$.
Then, in order to have $\hat c_i < q-2i+1$, we must also have $j=z(i) < q-i+1$.
Every $t \in [p]$ must have $t <z(t)$ since $z$ has $p$ nontrivial cycles but $z(t) \leq t$ for all $t >p$.
Therefore $i<z(i)=j$ since $i \in [p]$. As $\ell(z) = |D(z)|$,
it suffices to show that $D(s_j z s_j) = D(z) \sqcup\{(i,j)\}\sqcup\{(j,i)\}$ since the union on the right is
the diagram of a symmetric partition.
The diagrams $D(z)$ and $D(s_jzs_j)$ only differ in the rows and columns indexed by $j$, $j+1$, $z(j)=i$, and $z(j+1)$.
Because $D(z)$ is of partition shape, we must have $z(k)<j$ for any $i<k<z(j+1)$. Hence $z(j+1)\geq j+1$ and $(i,j)\in D(s_jzs_j)$, but $D(s_jzs_j)$ has no pairs of the form $(k,j)$ with $k> i$.
% and $D(z)$ contain no pairs of the form $(k,l)$ with $k=j$ or $k=j+1$ and $k\leq l$.
By the symmetry of the Rothe diagram, $D(s_jzs_j)$ contains $(j,i)$ and no pairs of the form $(j,k)$ with $k>i$.
Because $i<j$ and $j+1\leq z(j+1)$, the diagrams $\iD_j(s_jzs_j)$ and $\iD_{j}(z)$ have the same column $i$ and same column $z(j+1)$ unless $z(j+1)=j+1$.
Thus $D(s_jzs_j)=D(z) \sqcup\{(i,j)\}\sqcup\{(j,i)\}$ as needed.
\end{proof}
We can now give a short proof of Theorem~\ref{vexweak-prop}.
\begin{proof}[Proof of Theorem~\ref{vexweak-prop}]
%The values of $p$ and $q$ do not change if we replace $z$ by $s_j z s_j$ with $j$ as in Lemmas~\ref{vexweak-path-lem} or \ref{dominant-path-lem}, so applying these lemmas in succession will produce
%a directed path $z\xrightarrow{i_1} \cdots \xrightarrow{i_k} \dom_{pq}$ with $\{i_1,\dots,i_k\} \subseteq [q-1]$.
%Suppose $z\in \Ivex$.
If $z$ is not dominant, or dominant but not equal to $\dom_{pq}$,
then we can apply Lemma~\ref{vexweak-path-lem} or \ref{dominant-path-lem} to
find $j \in [q-1]$ with $z \xrightarrow{j} s_j z s_j \in \Ivex$.
As the values of $p$ and $q$ do not change when $z$ is replaced by $s_j z s_j $,
this one-step path can be extended by induction to one of the desired form.
Say that $z$ is \defn{quasi-dominant} if
$i<z(i)$ for all $i\in[p]$.
In this case $z(i) \leq i$ for all $ i>p$, so $\ell(s_p z s_p) \leq \ell(z)$.
Thus, if $z$ is quasi-dominant then we can only have $z \xrightarrow{j}v \in \Ivex$ if $j\neq p$, and then $v$ is quasi-dominant with $\cyc(v) = \cyc(z)=p$, so by induction any path $z\xrightarrow{i_1} \cdots \xrightarrow{i_k} \dom_{pq}$ must have $p\notin\{ i_1,\dots,i_k\}.$
\end{proof}
\subsection{Involution Grothendieck polynomials}\label{iG-sect}
There is another family of polynomials indexed by involutions $z \in I_\infty$ that share several favorable algebraic properties with $\fkGSp_z$.
We shall see that these \defn{involution Grothendieck polynomials}
are closely related to $\fkGO_z$, although they do not currently have a geometric interpretation by themselves.
%Write $\ell : S_\infty \to \NN$ for the usual Coxeter length function counting the number of inversions of a permutation.
Let $\leq$ denote the \defn{(strong) Bruhat order} on $S_\infty$,
which is the transitive closure of the relation with $v<w$ whenever $\ell(v) < \ell(w)$ and $w=vt$ for any transposition $t=(i, j) \in S_\infty$.
For each $u, v \in S_\infty$ there is a unique element $w \in S_\infty$ such that the set-wise product
$
\{ x\in S_\infty : x \leq u\} \{ x \in S_\infty : x \leq v\} $ is equal to $ \{x \in S_\infty : x \leq w\}$.
If we set $u\circ v := w$ then $\circ$ gives an associative operation $ S_\infty \times S_\infty \to S_\infty$, often called the \defn{Demazure product}.
Standard properties of the Bruhat order \cite[\S2.2]{CCG} imply that
$u\circ v = uv$ if and only if $\ell(uv) = \ell(u)+\ell(v)$, that $s_i\circ s_i = s_i$ for all simple transpositions $s_i:=(i,i+1) \in S_\infty$, and that $(u\circ v)^{-1} = v^{-1} \circ u^{-1}$.
It follows
%for $w \in S_\infty$ and $i \in \PP$
that
\be\label{circ-eq}
%s_i \circ s_i = s_i \quand
s_i\circ w =\begin{cases} s_iw &\text{if }i \notin \DesL(w) \\ w &\text{if }i \in \DesL(w)\end{cases}
\quand
w\circ s_i =\begin{cases} ws_i &\text{if }i \notin \DesR(w) \\ w &\text{if }i \in \DesR(w)\end{cases}\ee
for all $w \in S_\infty$ and $i \in \PP$, where $\DesL(w) := \DesR(w^{-1})$.
\begin{remark}\label{bcs-rem}
There is a monomial-positive formula for $\fkG_w$ that can be stated using this notation. In what follows,
we use the term \defn{word} to mean a finite sequence of positive integers $a=a_1a_2\cdots a_k$.
Given such a word, we set $\ell(a) =k$ and $x^a = x_{a_1}x_{a_2}\cdots x_{a_k}$.
A pair of words $(a,i)$ with $\ell(a) = \ell(i)$ is a
\defn{compatible sequence}
if
$ i $
is weakly increasing such that
$i_j < i_{j+1}$ whenever $a_j \leq a_{j+1}$.
Such a compatible sequence is \defn{bounded} if $i_j \leq a_j$ for all $j$.
Given $w \in S_\infty$, let $\cH(w)$ be the set of \defn{Hecke words} $a=a_1a_2\cdots a_k$ with $w =s_{a_1}\circ s_{a_2}\cdots \circ s_{a_k}$. Then \cite[Cor.~5.4]{KnutsonMiller2004} one has
$\textstyle\fkG_w = \sum_{(a,i)} \beta^{\ell(a) - \ell(w)} x^i$
where the sum is over all bounded compatible sequences $(a,i)$ with $a \in \cH(w)$.
\end{remark}
Combining the identities in \eqref{circ-eq} shows that if
$z\in I_\infty$ and $i \in \PP$ then
\be\label{iwi-eq}
s_i\circ z\circ s_i =\begin{cases} zs_i &\text{if }i \notin \DesR(z)\text{ and }zs_i=s_iz \\
s_izs_i &\text{if }i \notin \DesR(z)\text{ and }zs_i\neq s_iz \\
z&\text{if }i \in \DesR(z).\end{cases}
\ee
Note that $zs_i=s_iz$ if and only if $\{i,i+1\} = \{z(i),z(i+1)\}$.
The formula \eqref{iwi-eq} implies that $ w\mapsto w^{-1}\circ w$ is a surjective map $S_\infty\to I_\infty$,
so the set
\be \cB(z) := \{w \in S_\infty : w^{-1} \circ w = z\}\ee is nonempty for $z \in I_\infty$.
%The involution length $ \ellhat(z)$ from \eqref{ellhat-eq} is known to be equal to $\min\{\ell(w) : w \in \cB(z)\}$ \cite[\S3.2]{HMP1}.
\begin{definition}\label{iG-def} The \defn{involution Grothendieck polynomial} of $z \in I_\infty$ is
\[ \textstyle \iG_z := \sum_{w \in \cB(z)} \beta^{\ell(w) - \ellhat(z)} \fkG_w \in \NN[\beta][x_1,x_2,x_3,\dots].\]
%\quad\text{where } \ellhat(z) :=\min\{\ell(w) : w \in \cB(z)\}.\]
\end{definition}
The set $\cB(z)$ was extensively studied in \cite{HMP2}
and the polynomials $\iG_z$ were introduced in \cite[\S4]{MP2021}.
Like $\cBfpf(z)$, the set $\cB(z)$ can be efficiently generated using a certain equivalence relation.
Let $\sim$ be the transitive closure of the relation
on $S_\infty$ that has $v^{-1} \sim w^{-1}$ if there is an index $i \in \PP$
and integers $a<b<c$ such that
\[
\{ v_{i}v_{i+1}v_{i+2} ,\ w_{i}w_{i+1}w_{i+2} \}\subset \{cba,\ cab,\ bca\}\] while $v_j=w_j$ for all $j \notin \{i,i+1,i+2\}$.
This is the transitive relation with
\[ (\cdots cab \cdots)^{-1} \sim (\cdots cba \cdots)^{-1} \sim (\cdots bca \cdots)^{-1}\quad\text{for all }a<b<c,\]
where corresponding ellipses ``$\cdots$'' mask identical subsequences.
For $z \in I_\infty$ let $a_1<a_2<\dots\leq a_q=\max\{ i \in \PP : z(i)\neq i\}$ be the positive integers with $a_i\leq b_i:=z(a_i ) $.
Define
$\alpha_\inv(z)$ to be inverse of the permutation whose one-line notation is formed by removing the repeated letters from
$b_1 a_1b_2 a_2 b_3a_3\cdots.$
Then \cite[\S6.1]{HMP2} we have
\be\label{sim-eq} \cB(z) = \{ w \in S_\infty : w \sim \alpha_\inv(z)\}.
\ee
Recall the definition of $\ellhat(z)$ from \eqref{ellhat-eq}.
It is also known \cite[\S3]{HMP1} that
\be\label{ellhat-eq1}
\ellhat(z) = \ell(\alpha_\inv(z))=\min\{\ell(w) : w \in \cB(z)\} = |\iD(z)|.
\ee
%where $\SubDiag := \{(i,j) \in \PP\times \PP : i\geq j\}$.
From \eqref{ellhat-eq} and \eqref{iwi-eq} we get
\be\label{ellhat-eq2}
\ellhat(s_i\circ z \circ s_i) =\begin{cases} \ellhat(z) &\text{if }i \in \DesR(z) \\
\ellhat(z)+1 &\text{if }i \notin\DesR(z).\end{cases}
\ee
\begin{example}\label{cB-ex2}
If $z = (1,4)(2,5) \in I_\infty$ then $\alpha_\inv(z) = 41523^{-1} = 24513$ and
\[
\ba \cB(z) &= \{24513,\ 25413,\ 25314,\ 35214,\ 35124\} \\&= \left\{41523^{-1},\ 41532^{-1},\ 41352^{-1},\ 43152^{-1},\ 34152^{-1}\right\}
.
\ea
\]
\end{example}
\section{Expansions for quasi-dominant involutions}\label{exp1-sect}
Recall that an element $z \in \Ivex$ is \defn{quasi-dominant} if $i-1<z(i-1)$ whenever $1<i<z(i)$.
Every dominant involution is quasi-dominant.
This section proves two new formulas for $\fkGO_z$ when $z$ is quasi-dominant; see Theorems~\ref{qd-thm} and \ref{dom-thm}.
We then present a general conjecture about the support of $\GC_z$ for any $z \in \Ivex$.
%Our first result (Theorem~\ref{qd-thm}) expresses $\fkGO_z$ as a product of a simple polynomial and $\iG_z$.
%This will serve as the base case for the main theorem in Section~\ref{vex-sect}.
%Our second result (Theorem~\ref{dom-thm}) gives an explicit but not obviously positive $\fkG$-expansion for $\fkGO_z$.
\subsection{Grothendieck expansions in the quasi-dominant case}\label{qd-sect}
For any element $z \in I_\infty$ we define
\be\label{k-eq} k(z) :=
\min \{ i \in \NN: z(j) \leq j\text{ for all } j>i\}.
\ee
Since $D(z)$ is invariant under transpose when $z\in I_\infty$, we also have
\be k(z)=\min\{ i \in \NN : (j,j) \notin D(z) \text{ for all } j>i\}.\ee
\begin{proposition}
Let $z \in I_\infty$. Then it holds that $ k(z) \geq\cyc(z)$,
with equality if and only if $z$ is quasi-dominant.
\end{proposition}
\begin{proof}
As $\{ a \in \PP : a<z(a)\} \subseteq \{1,2,\dots,k(z)\}$
we must have $\cyc(z)\leq k(z)$, with equality only if $z$ is quasi-dominant.
Conversely, if $z$ is quasi-dominant then $z(j) \leq j$ for all $j > \cyc(z)$, so $k(z) \leq \cyc(z)$
whence $k(z) =\cyc(z)$.
\end{proof}
The goal of this section is to prove the following theorem.
%We will explain in the next section how this formula expresses $\fkGO_z$
%as a product of two explicit linear combinations of Grothendieck polynomials.
\begin{theorem}\label{qd-thm}
If $z \in \Ivex$ is quasi-dominant
then $\fkGO_z = \iG_z \prod_{i=1}^{k(z)} (2 + \beta x_i).$
\end{theorem}
The proof of this result appears at the end of this section.
Before giving this argument, we need to derive a few new properties of
the polynomials $\iG_z$.
% from Section~\ref{iG-sect}. To start, we have the following:
\begin{theorem}\label{iG-thm}
Suppose $z \in I_\infty$. Then for any $i \in \PP$ it holds that
\[\bpartial_i \iG_z = \begin{cases} \iG_{zs_i} &\text{if $i \in \DesR(z)$ and $z(i) = i+1$}
\\
\iG_{s_i z s_i} &\text{if $i \in \DesR(z)$ and $z(i) \neq i+1$}
\\
-\beta \iG_z &\text{if $i \notin \DesR(z)$.}\end{cases}
\]
% Moreover, if $z $ is dominant of shape $\lambda$ then $\ds\iG_z = \prod_{\substack{(i,j) \in \D_\lambda \\ i = j}} x_i \prod_{\substack{(i,j) \in \D_\lambda \\ i < j}} x_i\oplus x_j$.
% \be \iG_z = \prod_{\substack{(i,j) \in \D_\lambda \\ i = j}} x_i \prod_{\substack{(i,j) \in \D_\lambda \\ i < j}} x_i\oplus x_j.\ee
\end{theorem}
\begin{proof}
Let $z \in I_\infty$ and $i \in \PP$.
If $i \notin \DesR(z)$
then $z\neq s_i \circ z \circ s_i$ by \eqref{iwi-eq} and every $w \in \cB(z)$ has $w\circ s_i \in \cB(s_i \circ z \circ s_i)$.
In this case,
since $\cB(z)$ and $\cB(s_i\circ z\circ s_i)$ are disjoint sets,
every $w \in \cB(z)$ must have $w\neq w \circ s_i$, which means that $i \notin \DesR(w)$
and $\bpartial_i \fkG_w = -\beta \fkG_w$ so $\bpartial_i \iG_z =-\beta \iG_z$.
Instead suppose $i\in\DesR(z)$.
Define $y \in I_\infty$ to be $ zs_i =s_iz$ if $z(i)=i+1$ or else $s_izs_i$ if $z(i)\neq i+1$.
Then it follows from \eqref{iwi-eq} that
\ben
\item[(a)] one has $\{y,z\} = \{ v \in I_\infty : z=s_i \circ v\circ s_i\}$ and $\ellhat(z) = \ellhat(y)+1$.
\een
Next, choose any $w \in \cB(z)$. If $i \in \DesR(w)$ then $ws_i \circ s_i = w$ so
\[
s_i \circ (ws_i)^{-1} \circ (ws_i) \circ s_i = (ws_i\circ s_i)^{-1} \circ (ws_i \circ s_i) = w^{-1} \circ w = z.
\]
Alternatively, if $i \notin \DesR(w)$ then $ws_i = w\circ s_i$ so
\[
s_i \circ (ws_i)^{-1} \circ (ws_i) \circ s_i= s_i\circ s_i\circ w^{-1} \circ w \circ s_i \circ s_i= s_i \circ z \circ s_i = z.
\]
Given these identites we conclude from property (a) that
\ben
\item[(b)] if $w \in \cB(z)$ then $ws_i \in \cB(y) \sqcup \cB(z)$.
\een
Finally, if $v \in \cB(y)$ then we certainly have $v\circ s_i \in \cB(z)$ as
\[ (v\circ s_i)^{-1} \circ (v\circ s_i) = s_i \circ v^{-1} \circ v \circ s_i = s_i \circ y \circ s_i = z.\]
As $\cB(y)$ and $\cB(z)$ are disjoint, we must also have $v\neq v\circ s_i = vs_i$ so
\ben
\item[(c)] if $v \in \cB(y)$ then $i \notin \DesR(v)$ and $vs_i \in \cB(z)$.
\een
Items (b) and (c) imply
that $\cB(z) = \cC \sqcup \cD \sqcup \cE$
is the union of the disjoint sets
\[
\ba
\cC &:= \{ vs_i : v \in \cB(y)\},
\\
\cD &:= \{ w \in \cB(z) : i \notin \DesR(w)\},
\\
\cE &:= \{ ws_i : w \in \cD \}.
\ea
\]
We can therefore write
\[\ba
\bpartial_i \iG_z
&= \sum_{w \in \cC\sqcup \cD\sqcup \cE} \beta^{\ell(w) - \ellhat(z)} \bpartial_i \fkG_w
\\&= \sum_{v \in \cB(y)} \beta^{\ell(vs_i) - \ellhat(z)} \bpartial_i \fkG_{vs_i} + \sum_{w \in \cD} \beta^{\ell(w) - \ellhat(z)} \(\bpartial_i \fkG_w +\beta \bpartial_i \fkG_{ws_i}\).
%\\&= \sum_{v \in \cB(y)} \beta^{(\ell(v)+1) - (\ellhat(y)+1)} \fkG_v
%+ \sum_{w \in \cD} \beta^{\ell(w) - \ellhat(z)} \(-\beta \fkG_w +\beta\fkG_{w}\)
%\\&= \iG_y.
\ea
\]
The last expression is equal to $\iG_y$, since \eqref{Grothendieck-recursion} implies that if $v \in \cB(y)$ then
\[
\beta^{\ell(vs_i) - \ellhat(z)} \bpartial_i \fkG_{vs_i} = \beta^{(\ell(v)+1) - (\ellhat(y)+1)} \fkG_v = \beta^{\ell(v) -\ellhat(y)} \fkG_v
\]
while if $w \in \cD$ then $\bpartial_i \fkG_w +\beta \bpartial_i \fkG_{ws_i} = -\beta \fkG_w +\beta\fkG_{w}=0$.
\end{proof}
The next result is closely related to \cite[Thm.~4.5]{MP2021}.
The latter theorem
gives a general formula for $\iG_z$ as a sum of polynomial weights
attached to certain \defn{involution Hecke pipe dreams} for $z \in I_\infty$.
It will turn out that this sum has only one term in the case when $z \in I_\infty$ is dominant,
but to explain this we must briefly recall a few definitions.
Recall that $\SubDiag = \{(i,j) \in \PP\times \PP : i\geq j\}$. Write $(i,j) \prec (i',j')$ if $i<i'$ or if $i=i'$ and $j>j'$.
For any subset $D\subset\SubDiag$ let $\word(D)$ be the sequence formed by listing the numbers $i+j-1$ as $(i,j)$ runs over
$D$ in the total order $\prec$. For example,
$ \word(\{(1,2),(1,1),(2,3)\}) = 214.$
Let $D\subset \SubDiag$
and write $\word(D) = w_1w_2\cdots w_{k}$.
Then $D$ is an \defn{involution Hecke pipe dream} for $z \in I_\infty$ if
$ s_{w_1} \circ s_{w_2}\circ \cdots \circ s_{w_k} \in \cB(z)$.
%, or equivalently if \[ s_{w_k}\circ \cdots s_{w_3}\circ s_{w_2} \circ s_{w_1} \circ s_{w_2}\circ s_{w_3}\circ \cdots \circ s_{w_k} = z.\]
In this case we say that $D$ is \defn{reduced} if its size $|D|=k$ is equal to $\ellhat(z)$.
By \cite[Thm.~5.8]{HMP6} every $z\in I_\infty$ has at least one reduced involution Hecke pipe dream.
\begin{lemma}\label{oneplus-lem}
Suppose $D$ is an involution Hecke pipe dream for $z \in I_\infty$ and $(i,j) \in D$.
Then $z$ has a reduced involution Hecke pipe dream containing $(i,j)$.
\end{lemma}
\begin{proof}
Suppose $\word(D) = w_1w_2\cdots w_l$. Define $z_0=1$ and $z_k = s_{w_k}\circ z_{k-1} \circ s_{w_k}$ for $k \in [l]$.
If $D$ is reduced then the desired property is immediate, so assume $D$ is not reduced.
Then it follows from \eqref{ellhat-eq2} that some $q \in [l]$ has $w_q \in \DesR(z_{q-1})$.
If $q$ is chosen to be minimal, then $q>1$ and the ``exchange property'' stated as \cite[Prop.~3.10]{Hultman2007}
implies that there exists an index $p \in [q-1]$ such that
\[ s_{w_1}\circ \cdots \circ s_{w_p} \circ \cdots \circ \widehat{s_{w_q}} \circ \cdots \circ s_{w_l} =s_{w_1}\circ \cdots \circ \widehat{s_{w_p}} \circ \cdots \circ s_{w_q} \circ \cdots \circ s_{w_l} \in \cB(z).\]
The letters $w_p$ and $w_q$ correspond to two different positions in $D$. Only one of these positions can be $(i,j)$, so by removing the other we obtain an involution Hecke pipe dream for $z$ that still contains $(i,j)$ while having strictly fewer elements. From this point, the lemma follows by induction on the size of $D$.
\end{proof}
\begin{lemma}\label{iG-thm2-lem}
Suppose $z \in I_\infty$ is dominant of shape $\lambda$. Then $\D_\lambda \cap \SubDiag$
is the unique involution Hecke pipe dream for $z$.
\end{lemma}
\begin{proof}
\cite[Thm.~3.9 and Prop.~4.15]{HMP6} show that $\D_\lambda \cap \SubDiag$ is the unique reduced involution Hecke pipe dream for $z$. By Lemma~\ref{oneplus-lem},
every involution Hecke pipe dream for $z$ is contained in, and hence equal to, this set.
\end{proof}
The $\beta=0$ case of the following product formula appears in \cite[Thm.~3.26]{HMP1}.
\begin{theorem}\label{iG-thm2}
If $z \in I_\infty$ is dominant of shape $\lambda$ then
\[ \iG_z = \prod_{\substack{(i,j) \in \D_\lambda \\ i = j}} x_i \prod_{\substack{(i,j) \in \D_\lambda \\ i < j}} x_i\oplus x_j
\quad\text{where $x\oplus y := x+y+\beta xy$}.\]
\end{theorem}
\begin{proof}
From Lemma~\ref{iG-thm2-lem}, this result is a special case of \cite[Thm.~4.5]{MP2021}, which states that if $z \in I_\infty$ then $ \textstyle \iG_z = \sum_D \(\prod_{(i,j) \in D, i = j} x_i \prod_{(i,j) \in D, i \neq j} x_i\oplus x_j\)$ where $D$ runs over all involution Hecke pipe dreams for $z$.
\end{proof}
%For a partition $\lambda$ define $k(\lambda) = |\{ (i,j) \in \D_\lambda : i=j\}|$.
\begin{example}
Let $z = (1,4)(2,5) \in I_\infty$. Then $z$ is dominant of shape $(3,3,2)$, so from Example~\ref{cB-ex2} and Theorem~\ref{iG-thm2}
we get the two expressions
\[ \ba \iG_z &=
\fkG_{24513} + \beta\fkG_{25413} + \fkG_{25314} +\beta \fkG_{35214} + \fkG_{35124}
\\
&= x_1x_2(x_1\oplus x_2)(x_1\oplus x_3)(x_2\oplus x_3)
.\ea
\]
\end{example}
We may now deliver the postponed proof of Theorem~\ref{qd-thm}.
\begin{proof}[Proof of Theorem~\ref{qd-thm}]
Let $z\in \Ivex$ be quasi-dominant with $p=\cyc(z) = k(z)$.
Set $q = \max\{i \in \PP : z(i)\neq i\}$.
By Theorem~\ref{vexweak-prop}, there is a directed path $z\xrightarrow{i_1} \cdots \xrightarrow{i_k} \dom_{pq}$ in the vexillary weak order graph with $p\notin\{ i_1,\dots,i_k\}$,
so
\[\textstyle \fkGO_z = \bpartial_{i_1} \cdots \bpartial_{i_k} \fkGO_{\dom_{pq}}= \bpartial_{i_1} \cdots \bpartial_{i_k} \( \iG_{\dom_{pq}} \prod_{i=1}^{p}(2+\beta x_i)\)\]
by \eqref{idom-eq}, \eqref{orthogonal-recursion}, and Theorem~\ref{iG-thm2}.
As $\prod_{i=1}^{p}(2+\beta x_i)$ is symmetric in $x_i$ and $x_{i+1}$ when $i \neq p$,
we can use \eqref{fg-eq2} and Theorem~\ref{iG-thm} to rewrite the last expression as
the desired product
$\textstyle ( \bpartial_{i_1} \cdots \bpartial_{i_k} \iG_{\dom_{pq}} )
\prod_{i=1}^{p}(2+\beta x_i)
= \iG_z \prod_{i=1}^{p}(2+\beta x_i)$. %=\iG_z \prod_{i=1}^{k(z)}(2+\beta x_i).$
\end{proof}
\subsection{Transition formulas and applications}
Results in \cite{Lenart,LenartSottile}
give explicit \defn{transition formulas} for certain products of Grothendieck polynomials. We review these here and then use them to derive an expression for $\GC_z$ when $z$ is quasi-dominant.
Following \cite{LenartSottile}, we write
\[ v\xrightarrow{(a,b)}w\qquad\text{for $v,w \in S_\infty$ and positive integers $a<b$}
\] to indicate that $w = v(a,b)$ and $\ell(w) = \ell(v)+1$, meaning that $w$ covers $v$ in the Bruhat order on $S_\infty$.
The length condition holds precisely when $v(a) < v(b)$ and
no $i$ with $a<i<b$ has $v(a) < v(i) <v(b)$.
\begin{theorem}[{Lenart \cite[Thm.~3.1]{Lenart}}]
\label{lenart-transition} If $k \in \PP$ and $v \in S_\infty$ then
\[\textstyle
(1+\beta x_k) \fkG_v = \sum_\gamma
%\varepsilon_k(\gamma)\cdot \beta^{\ell(\gamma)}
(-1)^{p} \cdot \beta^{p+q}
\cdot \fkG_{\last(\gamma)}
\]
where the sum is over all saturated chains $\gamma$ in Bruhat order on $S_\infty$ of the form
{\be\label{lenart-eq}
v = v_0 \xrightarrow{(a_1,k)}%v_1 \xrightarrow{(a_2,k)}
\cdots \xrightarrow{(a_p,k)} v_p \xrightarrow{(k,b_1)} %v_{p+1} \xrightarrow{(k,b_2)}
\cdots
\xrightarrow{(k,b_q)}v_{p+q} = \last(\gamma)
\ee}%
with $p,q\in \NN$ and
$ a_p < a_{p-1} < \dots < a_1 < k < b_q< b_{q-1} <\dots <b_1.$
% where one sets $\varepsilon_k(\gamma) := (-1)^{p}$ and $\ell(\gamma) := p+q$.
\end{theorem}
A permutation $w \in S_\infty$ is \defn{$n$-Grassmannian}
if $w(i) < w(i+1)$ for all positive integers $i\neq n$.
We say that an element of $S_\infty$ is \defn{Grassmannian} if it is $n$-Grassmannian for some $n$.
The identity permutation is the unique element of $S_\infty$ that is $n$-Grassmannian for all $n \in \NN$.
\begin{remark}\label{grass-sym-rem}
%One has $\fkG_w \in \ZZ[\beta][x_1,x_2,\dots,x_n]$ if and only if $\DesR(w) \subseteq [n]$ (see \cite[Prop.~3.5]{MS2023}).
%Also, a polynomial
% $f \in \ZZ[\beta][x_1,x_2,\dots,x_n]$ is $S_n$-invariant if and only if $\bpartial_i f = -\beta f$ for all $i\in[n-1]$.
The polynomials $\fkG_w$ indexed by $n$-Grassmannian permutations are a $\ZZ[\beta]$-basis
for the subring of $S_n$-invariants in $\ZZ[\beta][x_1,x_2,\dots,x_n]$ by \cite[Thm.~2.2]{lenart2000}.
In particular, $\fkG_w$ is a symmetric element of $\ZZ[\beta][x_1,x_2,\dots,x_n]$ if and only if $w$ is $n$-Grassmannian.
\end{remark}
Given a partition $\lambda$ with at most $n$ nonzero parts,
there is a unique $n$-Grassmannian permutation $w \in S_\infty$
with
$(w(1) -1, w(2) -2,\dots, w(n) - n) = (\lambda_n,\dots,\lambda_2, \lambda_1).$
We denote this permutation using the notation $[\lambda \mid n]$.
%Every $n$-Grassmannian permutation can be expressed in this way.
\begin{example}\label{Gr-ex}
In one-line notation we have
$ [1^n \mid n] = 234\cdots (n+1)1$ and
\[ [ n \mid n] = 1,2,3,\cdots, n-1,2n,n,n+1,n+2,\dots,2n-1,\]
along with $[\emptyset\mid n] = 1$ for all $n$.
For $0<j<n$ we have
\[[1^{n-j} \mid n] = 1,2,3,\cdots,j,j+2,j+3,j+4,\dots,n+1,j+1.\]
\end{example}
In the following lemma, we interpret $[1^0\mid n]$ as the permutation $[\emptyset\mid n] = 1$.
\begin{lemma}\label{1gr-lem}
If $0\leq j\leq n$ then
\[ (1+\beta x_{n+1}) \fkG_{[1^{j} \mid n]} = \sum_{i=j}^n (-\beta)^{i-j} \fkG_{[1^i\mid n]} - \sum_{i=j+1}^{n+1} (-\beta)^{i-j} \fkG_{[1^{i}\mid n+1]}.\]
\end{lemma}
\begin{proof}
By inspecting Example~\ref{Gr-ex}, we see that if $v_0= [1^{j} \mid n]$ and $k=n+1$, then
the only way to construct a saturated chain in Bruhat order
\[
v_0 \xrightarrow{(a_1,k)}v_1 \xrightarrow{(a_2,k)}\cdots \xrightarrow{(a_p,k)} v_p \xrightarrow{(k,b_1)}v_{p+1} \xrightarrow{(k,b_2)}
\cdots
\xrightarrow{(k,b_q)} v_{p+q}
\]
with $a_p<a_{p-1}<\dots<a_1<k<b_q<b_{q-1}<\dots<b_1$
is to have
\[ a_i = n-j+1-i\text{ for all $i \in [p]$}\quand q\in\{0,1\}\quand b_1= k+1=n+2.\]
Under these conditions, the end of the chain is $v_{p+q}=[1^{j+p+q}\mid n+q]$.
Thus,
Theorem~\ref{lenart-transition} gives
$
(1+\beta x_{n+1}) \fkG_{[1^{j} \mid n]} = \sum_{p=0}^{n-j} \sum_{q=0}^1 (-1)^p \beta^{p+q} \fkG_{[1^{j+p+q}\mid n+q]}$
which is equivalent to the desired formula.
\end{proof}
Fix a positive integer $k$. The \defn{$k$-Bruhat order} on $S_\infty$ is
transitive closure of the relation with $v<_k w$ whenever $v \xrightarrow{(a,b)}w$ and $a\leq k <b$.
%For integer pairs $(a,b)$ and $(c,d)$ we write
%\be
%(a,b) \prec (c,d)\quad\text{to mean that either}\quad (b>d) \text{ or }(b=d\text{ and }a<c).
%\ee
\begin{definition}\label{kpieri-def}
An \defn{unmarked $k$-Pieri chain} from $v\in S_\infty$ to $w \in S_\infty$ is a saturated chain in $k$-Bruhat order of the form
\be\label{kpieri} v=v_0 \xrightarrow{(a_1,b_1)} v_1 \xrightarrow{(a_2,b_2)} \cdots \xrightarrow{(a_q,b_q)} v_q = w\ee
satisfying $1\leq a_i \leq k < b_i$ for all $i \in [q]$ along with the following two conditions:
\ben
%\item[(P0)] if $a_j =a_i$ for some $1 \leq j <i < q$ then $(a_i,b_i) \prec (a_{i+1},b_{i+1})$; and
\item[(P0)] if $a_j =a_i > a_{i+1}$ for some $1 \leq j <i < q$ then $b_i>b_{i+1}$; and
\item[(P1)] $b_1\geq b_2\geq \dots \geq b_q$.
\een
\end{definition}
An essential and non-obvious property of this definition is
that for any permutations $v,w \in S_\infty$ at most one unmarked $k$-Pieri chain exists from $v$ to $w$
\cite[Thm.~2.2]{LenartSottile}.
Following \cite{LenartSottile}, we write $v \xrightarrow{c(k)} w$ if such a chain exists. The reference \cite{LenartSottile}
gives an explicit algorithm to find the unique chain $v \xrightarrow{c(k)} w$, but we will not review this here.
Suppose $v,w \in S_\infty$ are such that $v \xrightarrow{c(k)} w$,
and assume that the unique unmarked $k$-Pieri chain from $v$ to $w$ has the form \eqref{kpieri}.
Define $\forced_k(v,w)$ to be the number of indices $i \in [q]$ such that either
\bei
\item $b_1=b_2=\dots=b_i$ and $a_1>a_2>\dots>a_i$, or
\item $i<q$ and $b_i=b_{i+1}$ and $a_i>a_{i+1}$.
% NOT ($b_i>b_{i+1}$, or $b_i=b_{i+1}$ and $a_{i}<a_{i+1}$)
%$(a_i,b_i) \not \prec (a_{i+1},b_{i+1})$.
\eei
Then define $\prohib_k(v,w)$ to be the number of indices $i \in [q]$ such that
\bei
\item $a_j=a_i$ for some $1\leq j <i$.
\eei
Notice that we have $\forced_k(v,w)\geq1$ if $v\neq w$ and that
\be\label{fp-obs}
\forced_k(v,w) + \prohib_k(v,w) \leq \ell(w) - \ell(v)
\quand
k \leq \ell(w) - \ell(v) - \prohib_k(v,w).
\ee
\begin{theorem}[{Lenart--Sottile \cite[Cor.~1.16]{LenartSottile}}]
\label{lensot-thm1}
If $0<p\leq k$ and $v \in S_\infty$ then
\[
\fkG_{[1^p\mid k]} \fkG_{v} = \sum_{v\xrightarrow{c(k)}w}\tbinom{\ell(w) - \ell(v) - \forced_k(v,w) - \prohib_k(v,w)}{p - \forced_k(v,w)}\cdot \beta^{\ell(w)-\ell(v) - p} \cdot \fkG_w
\]
where the sum is over all permutations $w \in S_\infty$ such that $v\xrightarrow{c(k)} w$.
\end{theorem}
We note
one more formula, which is a special case of the previous result:
\begin{theorem}[{Lenart \cite[Thm.~3.1]{lenart2000}}]\label{lenart-pieri}
Suppose $0<p\leq k$ and $\lambda$ is a partition with at most $k$ parts. Then
\[
\fkG_{[1^p\mid k]} \fkG_{[\lambda \mid k]} = \sum_\mu \tbinom{\cols(\mu/\lambda)-1}{|\mu/\lambda|-p}
\cdot \beta^{|\mu/\lambda|-p}
\cdot \fkG_{[\mu\mid k]}
\]
where the summation ranges over all partitions $\mu$ with $\ell(\mu)\leq k$ obtained by adding a vertical strip to $\lambda $ of weight at least $p$, and where $|\mu/\lambda|$ and $\cols(\mu/\lambda)$ respectively denote the numbers of cells and columns in $\D_\mu \setminus \D_\lambda$.
\end{theorem}
We can now turn Theorem~\ref{qd-thm} into an exact, though not manifestly positive, formula for the coefficient function $\GC_z : S_\infty \to\NN$ from Section~\ref{GC-eq-sect}.
% with $\fkGO_z = \sum_{w \in S_\infty} \GC_z(w) \cdot \beta^{\ell(w) - \ellhat(z)} \cdot \fkG_w$.
\begin{lemma}\label{prod-lem}
For all positive integers $k$ it holds that
\[ \textstyle
\prod_{i=1}^k (2+\beta x_i) = 2^k - \sum_{j=1}^k2^{k-j} \cdot (-\beta)^j \cdot \fkG_{[1^j \mid k]}.
\]
\end{lemma}
\begin{proof}
Our argument is by induction on $k$. Since $\fkG_{[1^1\mid 1]} = x_1$, the base case when $k=1$ is immediate. When $k>1$, we may assume by induction that
\[
\ba
\textstyle \prod_{i=1}^k (2+\beta x_i)
&
= \textstyle(2+\beta x_k) \(2^{k-1} - \sum_{j=1}^{k-1} 2^{k-1-j} (-\beta)^j \fkG_{[1^j \mid k - 1]} \)\\
&
=\textstyle (2+\beta x_k) \(2^{k} - \sum_{j=0}^{k-1} 2^{k-1-j}(-\beta)^j \fkG_{[1^j \mid k - 1]} \)
\ea
\]
where we interpret $[1^0\mid k-1]=1$. By Lemma~\ref{1gr-lem}
we have
\[
\ba
\textstyle (2+&\beta x_k)\textstyle\sum_{j=0}^{k-1} 2^{k-1-j}(-\beta)^j \fkG_{[1^j \mid k - 1]} =
\\ & \sum_{j=0}^{k-1} 2^{k-1-j} \( (-\beta)^j\fkG_{[1^j \mid k - 1]} + \sum_{i=j}^{k-1} (-\beta)^i \fkG_{[1^i \mid k - 1]} - \sum_{i=j+1}^{k} (-\beta)^i \fkG_{[1^{i}\mid k]}\).
\ea
\]
The coefficient of $(-\beta)^j \fkG_{[1^j\mid k-1]}$ in the last expression is
\[ 2^{k-1-j} + (2^{k-1} + 2^{k-2} + \dots + 2^{k-1-j}) = 2^k\]
for each $j=0,1,\dots,k-1$, while the coefficient of $-(-\beta)^j \fkG_{[1^j\mid k]}$ is
\[ 2^{k-1} + 2^{k-2} + \dots + 2^{k-1-(j-1)} = 2^k - 2^{k-j}\]
for each $j=1,2,\dots,k$. Thus $(2+\beta x_k)\sum_{j=0}^{k-1} 2^{k-1-j} (-\beta)^j \fkG_{[1^j \mid k - 1]}$ is equal to
\[ \textstyle 2^k + 2^k \sum_{j=1}^{k-1} (-\beta)^j \fkG_{[1^j\mid k-1]} - \sum_{j=1}^k (2^k-2^{k-j}) (-\beta)^j\fkG_{[1^j\mid k]}.
\]
Lemma~\ref{1gr-lem} also
shows that $(2+\beta x_k) 2^{k}=(2+\beta x_k) 2^{k} \fkG_{[1^0 \mid k - 1]}$ is equal to
\[ \textstyle
2^{k+1} + 2^k \sum_{j=1}^{k-1} (-\beta)^j \fkG_{[1^j \mid k - 1]} - 2^k \sum_{j=1}^k (-\beta)^j \fkG_{[1^j \mid k]}.
\]
These expressions have difference $ \ds \prod_{i=1}^k (2+\beta x_i) = 2^k - \sum_{j=1}^k2^{k-j} (-\beta)^j \fkG_{[1^j \mid k]}.$
\end{proof}
For $v,w \in S_\infty$, recall the definitions of $\forced_k(v,w)$ and $\prohib_k(v,w)$ from Theorem~\ref{lensot-thm1}.
Set $\epsilon_k(v,v) = 1$ and $\rho_k(v,v)=2^k$, and for $v \xrightarrow{c(k)} w\neq v$ define
\be
\epsilon_k(v,w) := (-1)^{1+\forced_k(v,w)}
\quand
\rho_k(v,w) := 2^{k + \ell(v) - \ell(w) + \prohib_k(v,w)}.
\ee
Set $\epsilon_k(v,w)= \rho_k(v,w)=0$ when we do not have $v \xrightarrow{c(k)} w$.
\begin{theorem}\label{dom-thm}
Suppose $z \in \Ivex$ is quasi-dominant and $k=k(z)$. Then
\[\textstyle \GC_z(w) = \sum_{v \in \cB(z)}
\epsilon_k(v,w) \rho_k(v,w) \geq 0\quad\text{for all $w \in S_\infty$}.\]
\end{theorem}
\begin{proof}
We start with the formula $ \fkGO_z = \iG_z \prod_{i=1}^{k} (2 + \beta x_i)$ from Theorem~\ref{qd-thm}. Substituting in Lemma~\ref{prod-lem} and applying Theorem~\ref{lensot-thm1} turns this to
\[
\fkGO_z = \sum_{v \in \cB(z)}
\(2^{k}\cdot \beta^{\ell(v)-\ellhat(z)} \cdot \fkG_v
-
\sum_{\substack{w \in S_\infty \\ v\xrightarrow{c(k)}w}} \Sigma_{v,w}^k \cdot \beta^{\ell(w)-\ellhat(z)} \cdot \fkG_w\)
\]
where for $v \xrightarrow{c(k) }w$ we define
\[\textstyle\Sigma_{v,w}^k :=
\sum_{p=1}^k
(-1)^p 2^{k-p} \tbinom{\ell(w) - \ell(v) - \forced_k(v,w) - \prohib_k(v,w)}{p - \forced_k(v,w)}.
\]
If $v=w$ then $\ell(w) - \ell(v) =\forced_k(v,w) = \prohib_k(v,w) = 0$ so $\epsilon_k(v,w) \rho_k(v,w) = 2^k$ and $\Sigma_{v,w}^k =0$. If $v\xrightarrow{c(k)}w$ but $v\neq w$ then we must have $\forced_k(v,w)\geq 1$, so in view of \eqref{fp-obs} we can rewrite $\Sigma_{v,w}^k$ as
\[\textstyle\Sigma_{v,w}^k
= (-1)^{\forced_k(v,w)} 2^{k-\forced_k(v,w)} \sum_{m \in \NN} (-1/2)^m \tbinom{\ell(w) - \ell(v) - \forced_k(v,w) - \prohib_k(v,w)}{m}
\]
which transforms via the binomial theorem to
$ \Sigma_{v,w}^k = - \epsilon_k(v,w) \rho_k(v,w)$.
Using these identities, our expression above for $\fkGO_z$ becomes the desired formula
$
\fkGO_z = \sum_{v \in \cB(z)} \sum_{w \in S_\infty} \epsilon_k(v,w) \rho_k(v,w) \cdot \beta^{\ell(w)-\ellhat(z)} \cdot \fkG_w.
$
% which is equivalent to the desired identity.
\end{proof}
\subsection{Examples and conjectures}\label{ex-conj-sect}
Fix $1\neq z \in I_\infty$, define $k=k(z)$ as in \eqref{k-eq}, and
%\be k=k(z)=\min\{ i \in \NN : (j,j) \notin D(z) \text{ for all } j>i\} \ee
%as above, and
let
\be
j=j(z) := \max\{ j \in \PP : z(i) = i\text{ for all }1\leq i\leq j\}.
\ee
For the identity permutation
we set $j(1)=1$ and $k(1)=0$.
\begin{definition}\label{zpieri-def}
Fix $z \in I_\infty$ and set $j=j(z)$ and $k=k(z)$.
For permutations $v,w \in S_\infty$ we write $v \xrightarrow{[z]}w$ if there exists an unmarked $k$-Pieri chain
\[ v=v_0 \xrightarrow{(a_1,b_1)} v_1 \xrightarrow{(a_2,b_2)} \cdots \xrightarrow{(a_q,b_q)} v_q = w\]
as in Definition~\ref{kpieri-def}, such that for each $i\in[q]$ it additionally holds that
\[ j \leq a_i \leq k < b_i \qquand \text{$a_i<z(a_i)$ or $z(b_i)<b_i$} .\]
Finally, let $\cB^+(z) := \left\{ w \in S_\infty : v\xrightarrow{[z]} w \text{ for some }v \in \cB(z)\right\}$.
\end{definition}
\begin{example}
If $z=1 \in I_\infty$ then $j(z) = 1>0=k(z)$ so only empty chains satisfy the conditions in Definition~\ref{zpieri-def}.
Therefore $\cB^+(1) = \cB(1) = \{1\}$.
\end{example}
%For $v,w \in S_\infty$ we write $v \xrightarrow{[z]}w$ if there exists an unmarked $k$-Pieri chain
%%$ v=v_0 \xrightarrow{(a_1,b_1)} v_1 \xrightarrow{(a_2,b_2)} \cdots \xrightarrow{(a_q,b_q)} v_q = w$
%as in Definition~\ref{kpieri-def} that has
%$ j \leq a_i \leq k < b_i $ and also either $a_i<z(a_i)$ or $z(b_i)<b_i$ for each $i\in[q]$.
%Finally,
%let
%\be\cB^+(z) := \left\{ w \in S_\infty : v\xrightarrow{[z]} w \text{ for some }v \in \cB(z)\right\}.\ee
%Also define $\cB^+(1) = \cB(1) = \{1\}$.
Based on computations and the preceding theorem, the set $\cB^+(z)$ appears to give a good approximation for $\supp(\GC_z)$. In particular:
%\begin{example}
%If $z=1 \in I_\infty$ then $j(z) = 1>0=k(z)$ so only empty chains satisfy the conditions in Definition~\ref{zpieri-def},
%and thus $\cB^+(1) = \cB(1) = \{1\}$.
%\end{example}
%There is a useful case when $\cB^+(z)$ is equal to the set in \eqref{dom-supp-eq}.
\begin{corollary}\label{qd-supp-cor}
If $z \in \Ivex$ is quasi-dominant and $k=k(z)$ then
\[
\supp(\GC_z) \subseteq \cB^+(z)
=\left\{ w \in S_\infty : v\xrightarrow{c(k)} w \text{ for some }v \in \cB(z)\right\}.
\]
\end{corollary}
\begin{proof}
In this case $j(z)=1$ and $k(z)=k$, so $a<z(a)$ for all $j\leq a \leq k$.
Thus all unmarked $k$-Pieri chains from $v$ to $w$ satisfy the conditions in Definition~\ref{zpieri-def},
\end{proof}
If $a$ and $b$ are integers then let $[a,b] = \{ i \in \ZZ : a\leq i \leq b\}$.
For $w \in S_\infty$ let
\be\supp(w) := \{ i \in \PP : w(i) \neq i\} .\ee
\begin{proposition} \label{supp-prop}
If $z \in I_\infty$ has $\supp(z) \subseteq [a,b]$ for integers $a\leq b$,
then \[\supp(v) \subseteq [a,b]\quand \supp(w) \subseteq [a-1,b+1]\]
for all $v \in \cB(z)$ and $w \in \cB^+(z)$.
\end{proposition}
\begin{proof}
If $z=1$ then $\supp(z) = \supp(v) = \supp(w) = \varnothing$
for all $v \in \cB(z)$ and $w \in \cB^+(z)$
so the result holds trivially. Assume $z\neq 1$.
Let $j=j(z)$ and $k = k(z)$ and suppose $v \in \cB(z)$.
Since $\alpha_\inv(z)$ fixes all integers $i \notin[a,b]$,
it follows from \eqref{sim-eq} that we have $\supp(v) \subseteq [a,b]$.
Therefore, if $c \leq k$ and $ b+1< d$ then we must have $\ell(v(c,d)) \neq \ell(v)+1$,
since if $v(c) < v(d)$ then it also holds that $v(c) < b+1 = v(b+1) < d = v(d)$.
Hence,
if there exists any unmarked $k$-Pieri chain of the form \eqref{kpieri} from $v$ to $w \in S_\infty$, then
this chain must have $b+1 \geq b_1 \geq b_2 \geq \dots \geq b_q>k$. Since to have $v \xrightarrow{[z]} w$ we also require that $j \leq a_i$ for all $i$,
we conclude that $v \xrightarrow{[z]} w$ implies $\supp(w) \subseteq[j,b+1]$.
The result follows since $a-1 \leq j$ as $\supp(z) \subseteq [a,b]$.
\end{proof}
We have used a computer to verify the following for all vexillary $z \in I_{11}$:
\begin{conjecture}\label{b+conj}
If $z \in \Ivex$ then $\cB(z)\subseteq \supp(\GC_z) \subseteq \cB^+(z)$.
\end{conjecture}
Both containments in this conjecture can be strict,
but in some notable cases we have $\supp(\GC_z) = \cB^+(z)$.
Among the dominant (respectively, vexillary) involutions in $S_8$,
the equality $\supp(\GC_z) = \cB^+(z)$ holds in 67 of 70 cases (respectively, 179 of 323 cases).
% (67, 70) dom s8
%(179, 323) vex s8
%First consider the transposition
%$
%t_{ij} := (i, j) \in I_\infty$
%for $i<j$.
\begin{example}\label{t-ex}
Suppose $z=t_n:=(1, n)\in I_n$.
Then $z$ is dominant of shape $\lambda=(n-1,1^{n-2})$ and $ \cB(z)$ consists of the permutations in $S_n$
whose {inverses} in one-line notation are the shuffles of $n1$ and $234\cdots(n-1)$
with at most one letter between $n$ and $1$.
The larger set $\cB^+(z)$ consists of the permutations in $S_{n+1}$
whose {inverses} in one-line notation are the shuffles of the words $n1$ and $234\cdots(n-1)(n+1)$
with at most two letters between $n$ and $1$, excluding the inverse of $234\cdots(n-1)(n+1)n1$.
We shall see below that $\supp(\GC_z) = \cB^+(z)$.
\end{example}
We can generalize this example.
Let $\Sh(n)$ denote the set of words obtained by shuffling
the 2-letter word $n2$ with then $(n-1)$-letter word
\[1 \hs\widehat 2\hs 3\cdots (n-1)\hs\widehat n \hs(n+1)= 1345\cdots (n-3)(n-2)(n-1) (n+1).\]
Define $\cX(n)$ to be the subset of words in $\Sh(n)$ for which
\begin{itemize}
\item the first letter is $1$ while the last letter is $n+1$; and
\item at most one letter appears between $n$ and $2$.
\end{itemize}
Define $\cY(n)\subset\Sh(n)$ be the set of words with exactly one or exactly two letters between $n$ and $2$.
Note that $\cX(n)$ and $\cY(n)$ have nonempty intersection.
%
\begin{example}
We have
$\cX(3) = \left\{ 1\overline{3}\underline{2}4 \right\}
%\left\{{\footnotesize\barr{ccccccc} 1\overline{3}\underline{2}4 \\ \ \\ \ \earr}\right\}
$, $ \cY(3) =
\left\{
\overline{3}1\underline{2}4, \ \ 1\overline{3}4\underline{2}, \ \ \overline{3}14\underline{2}\right\}
%\left\{{\footnotesize\barr{ccccccccc} \ \\
%\overline{3}1\underline{2}4, && 1\overline{3}4\underline{2}, \\
%&\overline{3}14\underline{2}
%\earr}\right\}
$, and
\[\Sh(3)=\left\{{\barr{ccccccccc} \overline{3}\underline{2}14, &&1\overline{3}\underline{2}4, && 14\overline{3}\underline{2}\\
&\overline{3}1\underline{2}4, && 1\overline{3}4\underline{2}, \\
&&\overline{3}14\underline{2}
\earr}\right\}.\]
%along with
%We have annotated the letters $n=3$ and $2$.
%Similarly, we have
%\[
%\ba
%\cX(5) &= \left\{{\footnotesize\barr{ccccccc} 1\overline{5}\underline{2}346,&& 13\overline{5}\underline{2}46,&& 134\overline{5}\underline{2}6, \\
%& 1\overline{5}3\underline{2}46, && 13\overline{5}4\underline{2}6,
%\earr}\right\}
%\ea\]
%and
%\[\ba
%\cY(5) &= \left\{{\footnotesize\barr{ccccccccc}
%\overline{5}1\underline{2}346, && 1\overline{5}3\underline{2}46, && 13\overline{5}4\underline{2}6, &&134\overline{5}6\underline{2}, \\
%&\overline{5}13\underline{2}46,&& 1\overline{5}34\underline{2}6,&& 13\overline{5}46\underline{2}
%\earr}\right\}.
%\ea
%\]
\end{example}
Define
%$c_4=(1,2,5,4) = 25314 \in S_5$ and for $n\geq 5$ let \[ c_n=(1,2,5,6,7,\dots,n) = 253467\cdots n1 \in S_n.\]
$c_4=(1,4,5,2) = 41352 \in S_5$ and for $n\geq 5$ let \[ c_n=(1,n,n-1,n-2,\dots,7,6,5,2) = n1342567\cdots (n-2)(n-1) \in S_n.\]
Finally set $\cZ(n) = \varnothing$ for $n\leq 3$ and $\cZ(n) = \{c_n\}$ for $n\geq 4$.
\begin{proposition}\label{t-prop}
Let $n>2$ and $z=(2,n) \in \Ivex$. Then the following holds:
\ben
\item[(a)] $\cB(z)= \left\{w^{-1} : w \in \cX(n)\right\}$.
\item[(b)] $\cB^+(z)=\left\{w^{-1} : w \in \cX(n)\cup \cY(n)\sqcup \cZ(n)\right\}$.
\item[(c)] The orthogonal Grothendieck polynomial of $z$ is
\[\fkGO_{z} = \sum_{w^{-1} \in \cX(n)} 2 \cdot \beta^{w(2)-w(n)-1}\cdot \fkG_w + \sum_{w^{-1} \in \cY(n)} \beta^{w(2)-w(n)-1}\cdot \fkG_w
\]
and consequently
$\supp(\GC_z) =\left\{w^{-1} : w \in \cX(n)\cup \cY(n)\right\}$.
% $\GC_z(w) \in \{0,1,2,3\}$ for all $w \in S_\infty$.
%\[\GC_z(w) = \begin{cases} 2&\text{if }w^{-1} \in \cX(n)\setminus \cY(n) \\
% 3 &\text{if }w^{-1} \in \cX(n) \cap \cY(n) \\
% 1 &\text{if }w^{-1} \in \cY(n) \setminus \cX(n) \\
% 0&\text{otherwise}.
% \end{cases}\]
\een
\end{proposition}
By the results in Section~\ref{shift-sect}, this proposition
determines $\cB(z)$, $\cB^+(z)$, and $\fkGO_z$ for all transpositions $z=(i,j) \in I_\infty$.
For these elements,
Conjecture~\ref{b+conj} holds but $ \cB^+(z)\setminus \supp(\GC_z)$ contains one extra element if $1<i<i+1<j$.
\begin{proof}
Part (a) is straightforward to derive from \eqref{sim-eq}, and part (b) then follows as an exercise from the definitions. We omit the details.
Let $\Sh_1(n)$ denote the set of shuffles of
$n1$ and $ 2345\cdots (n+1)$, viewed as permutations in one-line notation. Then
define $\cX_1(n)$ to be the subset of words in $\Sh_1(n)$ for which
the last letter is $n+1$ and at most one letter appears between $n$ and $1$.
Finally let $\cY_1(n)$ be the subset of words in $\Sh_1(n)$ for which exactly one or exactly two letters appear between $n$ and $1$.
Now set $y = (1,n) \in I_\infty$.
We shall first use Theorem~\ref{dom-thm} to prove that
\be\label{t1n-eq}
\fkGO_{y}
= \sum_{w^{-1}\in \cX_1(n)}2\cdot \beta^{w(1)-w(n)-1}\cdot \fkG_w + \sum_{w^{-1}\in \cY_1(n)}\beta^{w(1)-w(n)-1}\cdot \fkG_w.
\ee
It follows from \eqref{sim-eq} that $\cB(y) = \{ w^{-1} : w \in \cX_1(n)\}$.
Therefore, if $v \in \cB(y)$ and $1<i_1<i_2<i_3<\dots$ are the indices $i$ with $v(1) < v(i)$, then we must have $v(1)<v(i_1)<v(i_2)<v(i_3)<\cdots$
and also $v(i_1) = v(1)+1$.
Set $j := v(1)$
and $k:=k(y)=1$. Given the previous paragraph, we see that if $v \in \cB(y)$ then a chain $v\xrightarrow{c(k)}w $
is either trivial with $v=w$ or a single Bruhat cover of the form $v\xrightarrow{(1,b)}w$ where $w=v(1,b) = s_j v$.
For chains of the first type we have $\varepsilon_k(v,w) = 1$ and $\rho_k(v,w) = 2$,
and $w^{-1}=v^{-1}$ gives an arbitrary element of $\cX_1(n)$.
For chains of the second type we have $\varepsilon_k(v,w)=1$ and $\rho_k(v,w)=1$,
and the permutations $w$ range over exactly the set of inverses of the elements of $\cY_1(n)$.
The only thing left to show to deduce \eqref{t1n-eq} from Theorem~\ref{dom-thm}
is that if $w \in S_\infty$ has $w^{-1} \in \cX_1(n)\cup \cY_1(n)$ then $\ell(w) - \ellhat(y) = w(1)-w(n)-1$.
This holds since
in such a permutation $w$
it must hold that $w(n)<w(n+1)$, so $w^{-1}$ has $w(1)-1$ inversions involving $1$, along with $n-1 - (w(n)-1)$ inversions involving $n$,
and a single inversion involving both $1$ and $n$, which gives
\[
\ell(w) = \ell(w^{-1}) = w(1) -1 + n-1 - (w(n)-1) - 1 = w(1) - w(n) + n-2.
\]
As we have $\ellhat(y) = n-1$ by \eqref{ellhat-eq}, the identity \eqref{t1n-eq} follows.
Let $z = (2,n) \in I_\infty$.
To prove part (c) we apply $\bpartial_1$ to both sides of \eqref{t1n-eq}.
On the left side we get $\bpartial_1\fkGO_y = \fkGO_z$ by \eqref{orthogonal-recursion}.
%
To evaluate $\bpartial_1$ applied to the right side of \eqref{t1n-eq}, fix
$w \in S_\infty$ with $w^{-1} \in \cX_1(n)\cup \cY_1(n)$. Then observe that
$w(1) < w(2)$ if and only if $w^{-1} = n1234\cdots(n-1) \in \cX_1(n)$,
and in this case it holds that $(ws_1)^{-1} = n2134\cdots (n-1) \in \cX_1(n)$ and
\[
\bpartial_1 \( \beta^{ws_1(1) - w(n)-1} \fkG_{ws_1} + \beta^{w(1) - w(n)-1} \fkG_w\)
=
\bpartial_1\( \beta \fkG_{ws_1} + \fkG_w\) = 0.
\]
The map $w \mapsto s_1 w$ defines a bijection $\cY_1(n)\to \cY(n)$ as well as a bijection
\[ \cX_1(n) \setminus \{n1234\cdots(n-1), \ n2134\cdots (n-1)\} \to \cX(n),\]
so it follows from \eqref{Grothendieck-recursion} that $\bpartial_1$ transforms \eqref{t1n-eq}
to the formula in part (c).
\end{proof}
It can be useful to represent $\cB^+(z)$ as the following direct graph.
For $v,w \in S_\infty$ we write $v \lessdot_L w$ when
$\ell(w) = \ell(v)+1$ and $ w=s_i v$ for some $ i\in \PP$.
We turn $\cB^+(z)$ into a directed graph by adding edges $v \to w$
whenever $v \lessdot_L w$.
Figures~\ref{t1n-fig}, \ref{gn-fig}, and \ref{w0-fig} show some instances of this graph.
\begin{remark}
In all examples that we have computed, the directed graph structure on $\cB^+(z)$ just described is connected. This is not the case if the left weak order relation $\lessdot_L$
is replaced by the right weak order. It would be interesting to know if this connectedness property is a general phenomenon.
\end{remark}
% \begin{figure}[h]
% \centerline{
% \resizebox{.5\width}{!}{%
% \begin{tikzpicture}[xscale=1.6, yscale=1,>=latex,baseline=(z.base)]
% \node at (0,0.0) (z) {};
% \node[draw] at (0.5,0) (11) {\tiny $4231^{-1}:1$};
% \node[draw] at (1.5,0) (12) {\tiny $24351^{-1}:1$};
% \node[draw,fill=lightskyblue] at (0,1) (21) {\tiny $4213^{-1}:3$};
% \node[draw,fill=lightskyblue] at (1,1) (22) {\tiny $2431^{-1}:3$};
% \node[draw] at (2,1) (23) {\tiny $23451^{-1}:1$};
% \node[draw,fill=lightskyblue] at (0,2) (31) {\tiny $4123^{-1}:2$};
% \node[draw,fill=lightskyblue] at (1,2) (32) {\tiny $2413^{-1}:2$};
% \node[draw,fill=lightskyblue] at (2,2) (33) {\tiny $2341^{-1}:2$};
% %
% \draw[->,thick] (31) -- (21);
% \draw[->,thick] (32) -- (21);
% \draw[->,thick] (32) -- (22);
% \draw[->,thick] (33) -- (22);
% \draw[->,thick] (33) -- (23);
% \draw[->,thick] (21) -- (11);
% \draw[->,thick] (22) -- (11);
% \draw[->,thick] (22) -- (12);
% \draw[->,thick] (23) -- (12);
% \end{tikzpicture}}
% \quad
% \resizebox{.5\width}{!}{%
% \begin{tikzpicture}[xscale=2, yscale=1,>=latex,baseline=(z.base)]
% \node at (0,0.0) (z) {};
% \node[draw] at (1,0) (00) {\tiny $4516273^{-1}:1$};
% \node[draw] at (0,1) (21) {\tiny $4156273^{-1}:2$};
% \node[draw] at (1,1) (22) {\tiny $451623^{-1}:2$};
% \node[draw] at (2,1) (23) {\tiny $4512673^{-1}:2$};
% \node[draw] at (0,2) (31) {\tiny $415623^{-1}:4$};
% \node[draw] at (1,2) (32) {\tiny $4152673^{-1}:4$};
% \node[draw] at (2,2) (33) {\tiny $451263^{-1}:4$};
% \node[draw, fill=lightskyblue] at (1,3) (41) {\tiny $415263^{-1}:8$};
% %
% \draw[->,thick] (21) -- (00);
% \draw[->,thick] (22) -- (00);
% \draw[->,thick] (23) -- (00);
% \draw[->,thick] (31) -- (21);
% \draw[->,thick] (31) -- (22);
% \draw[->,thick] (32) -- (21);
% \draw[->,thick] (32) -- (23);
% \draw[->,thick] (33) -- (22);
% \draw[->,thick] (33) -- (23);
% \draw[->,thick] (41) -- (31);
% \draw[->,thick] (41) -- (32);
% \draw[->,thick] (41) -- (33);
% \end{tikzpicture}}
% \quad
% \resizebox{.5\width}{!}{%
% \begin{tikzpicture}[xscale=1.8, yscale=1,>=latex,baseline=(z.base)]
% \node at (0,0.0) (z) {};
% \node[draw] at (2,0) (00) {\tiny $43521^{-1}:1$};
% \node[draw] at (1,1) (11) {\tiny $43251^{-1}:3$};
% \node[draw] at (2,1) (12) {\tiny $34521^{-1}:1$};
% \node[draw] at (3,1) (13) {\tiny $43512^{-1}:1$};
% \node[draw] at (0,2) (21) {\tiny $42351^{-1}:2$};
% \node[draw] at (1,2) (22) {\tiny $34251^{-1}:3$};
% \node[draw, fill=lightskyblue] at (2,2) (23) {\tiny $4321^{-1}:6$};
% \node[draw] at (3,2) (24) {\tiny $34512^{-1}:1$};
% \node[draw] at (4,2) (25) {\tiny $43152^{-1}:3$};
% \node[draw,fill=lightskyblue] at (0,3) (31) {\tiny $4231^{-1}:4$};
% \node[draw] at (1,3) (32) {\tiny $32451^{-1}:2$};
% \node[draw,fill=lightskyblue] at (2,3) (33) {\tiny $3421^{-1}:6$};
% \node[draw] at (3,3) (34) {\tiny $34152^{-1}:2$};
% \node[draw,fill=lightskyblue] at (4,3) (35) {\tiny $4312^{-1}:6$};
% \node[draw] at (5,3) (36) {\tiny $41352^{-1}:2$};
% \node[draw, fill=lightskyblue] at (1.5,4) (41) {\tiny $3241^{-1}:4$};
% \node[draw, fill=lightskyblue] at (3,4) (42) {\tiny $3412^{-1}:4$};
% \node[draw, fill=lightskyblue] at (4.5,4) (43) {\tiny $4132^{-1}:4$};
% %
% \draw[->,thick] (11) -- (00);
% \draw[->,thick] (12) -- (00);
% \draw[->,thick] (13) -- (00);
% \draw[->,thick] (21) -- (11);
% \draw[->,thick] (22) -- (11);
% \draw[->,thick] (22) -- (12);
% \draw[->,thick] (23) -- (11);
% \draw[->,thick] (24) -- (12);
% \draw[->,thick] (24) -- (13);
% \draw[->,thick] (25) -- (13);
% \draw[->,thick] (31) -- (21);
% \draw[->,thick] (31) -- (23);
% \draw[->,thick] (32) -- (22);
% \draw[->,thick] (33) -- (22);
% \draw[->,thick] (33) -- (23);
% \draw[->,thick] (34) -- (24);
% \draw[->,thick] (34) -- (25);
% \draw[->,thick] (35) -- (23);
% \draw[->,thick] (35) -- (25);
% \draw[->,thick] (36) -- (25);
% \draw[->,thick] (41) -- (32);
% \draw[->,thick] (41) -- (33);
% \draw[->,thick] (42) -- (33);
% \draw[->,thick] (42) -- (34);
% \draw[->,thick] (42) -- (35);
% \draw[->,thick] (43) -- (35);
% \draw[->,thick] (43) -- (36);
% \end{tikzpicture}}
% }
% \caption{The digraphs $ \cB^+(z)$ when $z$ is $t_4=(1,4)$ (left), $g_{3}=(1,4)(2,5)(3,6)$ (middle),
% or $w_0=4321$ (right).
% The data in each box is $\boxed{w:\GC_z(w)}$, with $w$ given in inverse one-line notation.
% The blue vertices are the elements of $\cB(z)$.
% As $\GC_z(w)$ is always nonzero, one has $\supp(\GC_z) = \cB^+(z)$ in these cases.
% }\label{t1n-fig}
% \end{figure}
\begin{figure}[h]
\centerline{
\begin{tikzpicture}[xscale=2, yscale=1,>=latex,baseline=(z.base)]
\node at (0,0.0) (z) {};
\node[draw] at (0.5,0) (11) {\tiny $4231^{-1}:1$};
\node[draw] at (1.5,0) (12) {\tiny $24351^{-1}:1$};
\node[draw,fill=lightskyblue] at (0,1) (21) {\tiny $4213^{-1}:3$};
\node[draw,fill=lightskyblue] at (1,1) (22) {\tiny $2431^{-1}:3$};
\node[draw] at (2,1) (23) {\tiny $23451^{-1}:1$};
\node[draw,fill=lightskyblue] at (0,2) (31) {\tiny $4123^{-1}:2$};
\node[draw,fill=lightskyblue] at (1,2) (32) {\tiny $2413^{-1}:2$};
\node[draw,fill=lightskyblue] at (2,2) (33) {\tiny $2341^{-1}:2$};
%
\draw[->,thick] (31) -- (21);
\draw[->,thick] (32) -- (21);
\draw[->,thick] (32) -- (22);
\draw[->,thick] (33) -- (22);
\draw[->,thick] (33) -- (23);
\draw[->,thick] (21) -- (11);
\draw[->,thick] (22) -- (11);
\draw[->,thick] (22) -- (12);
\draw[->,thick] (23) -- (12);
\end{tikzpicture}
}
\caption{The directed graph $ \cB^+(z)$ for $z=t_{4}=(1,4)$.
The data in each box is $\boxed{w:\GC_z(w)}$ with $w$ in inverse one-line notation.
The blue vertices are the elements of $\cB(z)$.
One has $\supp(\GC_z) = \cB^+(z)$ in this case.
}\label{t1n-fig}
\end{figure}
%The permutation
%$g_{1,n} := [n^n; n]%= (n+1)(n+2)\cdots 2n, 1,2\cdots n
%= (1, n+1)(2, n+2)\cdots(n, 2n) \in I_{2n}$
%is both $n$-Grassmannian and dominant of shape $\lambda=(n^n)$.
\begin{example}\label{g-ex}
Suppose $z=g_n:= (1, n+1)(2, n+2)\cdots(n, 2n) \in I_{2n}$. Then $z$ is dominant of shape $\lambda=(n^n)$ and
the set $ \cB(z)$ consists of the single element whose {inverse} is
$
(n+1)1(n+2)2\cdots (2n)n\in S_{2n}.$
One can show that the larger set
$\cB^+(z)$ consists of the $2^n$ permutations whose {inverses} have the form
$ (n+1)a_1b_1a_2b_2\cdots a_nb_n \in S_{2n+1}$
where $\{a_i,b_i\} = \{ i, n+1+i\}$ for $i \in [n]$.
% See Figure~\ref{gn-fig}.
\end{example}
\begin{figure}[h]
\centerline{
\begin{tikzpicture}[xscale=2.5, yscale=1.2,>=latex,baseline=(z.base)]
\node at (0,0.0) (z) {};
\node[draw] at (1,0) (00) {\tiny $4516273^{-1}:1$};
\node[draw] at (0,1) (21) {\tiny $4156273^{-1}:2$};
\node[draw] at (1,1) (22) {\tiny $451623^{-1}:2$};
\node[draw] at (2,1) (23) {\tiny $4512673^{-1}:2$};
\node[draw] at (0,2) (31) {\tiny $415623^{-1}:4$};
\node[draw] at (1,2) (32) {\tiny $4152673^{-1}:4$};
\node[draw] at (2,2) (33) {\tiny $451263^{-1}:4$};
\node[draw, fill=lightskyblue] at (1,3) (41) {\tiny $415263^{-1}:8$};
%
\draw[->,thick] (21) -- (00);
\draw[->,thick] (22) -- (00);
\draw[->,thick] (23) -- (00);
\draw[->,thick] (31) -- (21);
\draw[->,thick] (31) -- (22);
\draw[->,thick] (32) -- (21);
\draw[->,thick] (32) -- (23);
\draw[->,thick] (33) -- (22);
\draw[->,thick] (33) -- (23);
\draw[->,thick] (41) -- (31);
\draw[->,thick] (41) -- (32);
\draw[->,thick] (41) -- (33);
\end{tikzpicture}}
\caption{The directed graph $ \cB^+(g_3)$, presented as in Figure~\ref{t1n-fig}.
}\label{gn-fig}
\end{figure}
The previous example has an extension analogous to Proposition~\ref{t-prop},
which shows that $\supp(\GC_z) = \cB^+(z)$ whenever $z=g_n$; see Proposition~\ref{g-prop}.
% \begin{figure}[h]
%\centerline{\includegraphics[scale=0.3]{g12.png}\qquad \includegraphics[scale=0.3]{g13.png}}
%\caption{The digraphs $\cB(z) \subseteq \cB^+(z)$ when $z \in I_\infty$ is $g_{2}=(1,3)(2,4)$ and $g_{3}=(1,4)(2,5)(3,6)$, presented using the conventions as in Figure~\ref{t1n-fig}.}\label{gn-fig}
%\end{figure}
\begin{example}
Finally let $w_0 = n\cdots 321\in I_n$ be the longest element of $S_n$. Then $w_0$
is dominant of shape $\lambda=(n-1,\dots,3,2,1)$.
The set $\cB(w_0)$ does not seem to have any description simpler than \eqref{sim-eq}.
One can check that the larger set
$ \cB^+(w_0)$ contains the permutations in $S_{n+1}$ whose {inverses} in one-line notation have the form
$
u_1\cdots u_i(n+1) u_{i+1}\cdots u_n
$
where $u=u_1u_2\dots u_n $ is the {inverse} of an element of $ \cB(w_0)$ and $i \in [n]$ has $n \geq 2 u_j$ for each $i<j\leq n$. %See Figure~\ref{w0-fig}.
%We conjecture, but do not know how to prove, that $\supp(\GC_{w_0}) = \cB^+(w_0)$ for all $n$.
%This has been checked by computer for $n\leq 11$.
\end{example}
\begin{figure}[h]
\centerline{
\begin{tikzpicture}[xscale=1.8, yscale=1,>=latex,baseline=(z.base)]
\node at (0,0.0) (z) {};
\node[draw] at (2,0) (00) {\tiny $43521^{-1}:1$};
\node[draw] at (1,1) (11) {\tiny $43251^{-1}:3$};
\node[draw] at (2,1) (12) {\tiny $34521^{-1}:1$};
\node[draw] at (3,1) (13) {\tiny $43512^{-1}:1$};
\node[draw] at (0,2) (21) {\tiny $42351^{-1}:2$};
\node[draw] at (1,2) (22) {\tiny $34251^{-1}:3$};
\node[draw, fill=lightskyblue] at (2,2) (23) {\tiny $4321^{-1}:6$};
\node[draw] at (3,2) (24) {\tiny $34512^{-1}:1$};
\node[draw] at (4,2) (25) {\tiny $43152^{-1}:3$};
\node[draw,fill=lightskyblue] at (0,3) (31) {\tiny $4231^{-1}:4$};
\node[draw] at (1,3) (32) {\tiny $32451^{-1}:2$};
\node[draw,fill=lightskyblue] at (2,3) (33) {\tiny $3421^{-1}:6$};
\node[draw] at (3,3) (34) {\tiny $34152^{-1}:2$};
\node[draw,fill=lightskyblue] at (4,3) (35) {\tiny $4312^{-1}:6$};
\node[draw] at (5,3) (36) {\tiny $41352^{-1}:2$};
\node[draw, fill=lightskyblue] at (1.5,4) (41) {\tiny $3241^{-1}:4$};
\node[draw, fill=lightskyblue] at (3,4) (42) {\tiny $3412^{-1}:4$};
\node[draw, fill=lightskyblue] at (4.5,4) (43) {\tiny $4132^{-1}:4$};
%
\draw[->,thick] (11) -- (00);
\draw[->,thick] (12) -- (00);
\draw[->,thick] (13) -- (00);
\draw[->,thick] (21) -- (11);
\draw[->,thick] (22) -- (11);
\draw[->,thick] (22) -- (12);
\draw[->,thick] (23) -- (11);
\draw[->,thick] (24) -- (12);
\draw[->,thick] (24) -- (13);
\draw[->,thick] (25) -- (13);
\draw[->,thick] (31) -- (21);
\draw[->,thick] (31) -- (23);
\draw[->,thick] (32) -- (22);
\draw[->,thick] (33) -- (22);
\draw[->,thick] (33) -- (23);
\draw[->,thick] (34) -- (24);
\draw[->,thick] (34) -- (25);
\draw[->,thick] (35) -- (23);
\draw[->,thick] (35) -- (25);
\draw[->,thick] (36) -- (25);
\draw[->,thick] (41) -- (32);
\draw[->,thick] (41) -- (33);
\draw[->,thick] (42) -- (33);
\draw[->,thick] (42) -- (34);
\draw[->,thick] (42) -- (35);
\draw[->,thick] (43) -- (35);
\draw[->,thick] (43) -- (36);
\end{tikzpicture}}
\caption{The directed graph $ \cB^+(4321)$, presented as in Figure~\ref{t1n-fig}.
}\label{w0-fig}
\end{figure}
Extending this example,
let
\be w_{ij} := (i, j)(i+1,j-1)(i+2,j-2)\cdots (i+k,j-k) \in \Ivex\ee for any integers $1\leq i<j$ where $k=\lfloor\tfrac{j-i-1}{2}\rfloor$. Computer calculations support the following conjecture.
The case when $i=1$ has been checked for all
$j\leq 11$.
\begin{conjecture}
One has
$\supp(\GC_{w_{ij}}) = \cB^+(w_{ij})$ if $i=1$ or $j-i$ is odd.
\end{conjecture}
When $i>1$ and $j-i$ is even, the set $\cB^+(w_{ij})$ can be much larger than $\supp(\GC_{w_{ij}})$.
We do not have any conjecture for $\supp(\GC_{w_{ij}})$ in these cases,
or for the values of $\GC_{w_{ij}}$ in general.
This set of coefficients can be quite complicated,
although its maximum seems to be a central binomial coefficient; see Figure~\ref{values-fig}.
\begin{figure}[h]
\centerline{\begin{tabular}{| l | l|}
\hline&\\[-8pt]
$n$ & nonzero values of $\GC_{w_0}$ for $w_0=n\cdots 321\in I_n$ \\[-8pt]&\\ \hline & \\[-8pt]
1 & 1 \\[-8pt]&\\
2 & 1, 2 \\[-8pt]&\\
3 & 1, 2, 3 \\[-8pt]&\\
4 & 1, 2, 3, 4, 6 \\[-8pt]&\\
5 & 1, 2, 3, 4, 6, 9, 10 \\[-8pt]&\\
6 & 1, 2, 3, 4, 6, 8, 9, 10, 12, 18, 20 \\[-8pt]&\\
7 & 1, 2, 3, 4, 5, 6, 8, 9, 10, 12, 13, 14, 15, 18, 20, 27, 30, 33, 34, 35 \\[-8pt]&\\
8 & 1, 2, 3, 4, 5, 6, 8, 9, 10, 12, 13, 14, 15, 16, 18, 20, 24, 27, 30, 33, \\ & 34, 35, 36, 40, 54, 60, 66, 68, 70
\\[-8pt]&\\ \hline
\end{tabular}}
\caption{Nonzero values of $\GC_{w_0}$ for $w_0=n\cdots 321\in I_n$.}\label{values-fig}
\end{figure}
% \begin{figure}[h]
% \centerline{
% %\includegraphics[scale=0.3]{t12.png}\qquad
% \includegraphics[scale=0.3]{t14.png}
% \quad
% \includegraphics[scale=0.3]{g13.png}
% \quad
% \includegraphics[scale=0.25]{w14.png}}
% \caption{The digraphs $ \cB^+(z)$ when $z$ is $t_4=(1,4)$ (left), $g_{3}=(1,4)(2,5)(3,6)$ (middle),
% or $w_0=4321$ (right).
% The data in each box is $\boxed{w:\GC_z(w)}$, with $w$ given in inverse one-line notation.
% The blue vertices are the elements of $\cB(z)$.
% As $\GC_z(w)$ is always nonzero, one has $\supp(\GC_z) = \cB^+(z)$ in these cases.
% }\label{t1n-fig}
% \end{figure}
\section{Expansions for vexillary involutions}\label{vex-sect}
The first main result of this section is a generalization of Theorem~\ref{qd-thm}
that holds any vexillary involution; see Theorem~\ref{ivex-thm}. This formula is more complicated than in the quasi-dominant case,
and we use Section~\ref{setting-sect} to explain the relevant notation.
The rest of the section derives some interesting consequences.
\subsection{Setting up the main theorem}\label{setting-sect}
A generic vexillary involution $z \in \Ivex$
has cycle notation \[z = (a_1, b_1)(a_2, b_2)\cdots (a_q, b_q)\]
where $1\leq a_1 <a_2 < \dots <a_q < \min\{b_1,b_2,\dots,b_q\}$.
We refer to the numbers $a_i$ as \defn{left endpoints} and to the numbers $b_i$ as \defn{right endpoints}.
The \defn{left segments} of $z$ are the maximal subsets of consecutive left endpoints,
that is, the equivalence classes in $\{a_1,a_2,\dots,a_q\}$ under the transitive closure of the relation with $a_i \sim a_j$ if $|a_j-a_i|\leq 1$.
There is at most one left segment containing $1$, which we refer to as the \defn{immobile segment}.
All other left segments are \defn{mobile}.
Suppose $L$ is a mobile left segment of $z$ and define $c_0: = \min(L) - 1$. Notice that we must have $c_0 = z(c_0) \in \PP$.
Now, for any subset $ S \subseteq \{a_1,a_2,\dots, a_q\}$ define $\sigma_{S,L} \in S_\infty$ to be the cyclic permutation
\be \sigma_{S,L}: = (c_0,c_1,c_2,\dots,c_k) \quad\text{where }S\cap L = \{c_1<c_2<\dots<c_k\}.\ee
This is the identity element when $S\cap L$ is empty.
The conjugated permutation
\[ (\sigma_{S,L})^{-1} \cdot z \cdot \sigma_{S,L} \]
is not always vexillary, but
its arc diagram is obtained from $z$ by simply replacing each edge $\{c_i < z(c_i)\}$ for $i \in [k]$ with $\{ c_{i-1} < z(c_i)\}$.
Also define \be \textstyle \sigma_{S}: = \prod_L \sigma_{S,L}\ee where the product is over all mobile left segments of $z$.
This product can be taken in any order since as $L$ varies the permutations $\sigma_{S,L}$ have disjoint supports.
\begin{example} \label{sigma-ex}
Suppose our vexillary involution is
\[z = (2, 7)(3, 8)(4, 6)(5, 9)
=
\arcstart
{
*{\bullet} & *{\bullet} \arc{1.2}{rrrrr} & *{\bullet} \arc{1.2}{rrrrr} & *{\bullet} \arc{0.6}{rr} & *{\bullet} \arc{1}{rrrr} & *{\bullet} & *{\bullet} & *{\bullet} & *{\bullet} \\
1 & 2 & 3 & 4 & 5 & 6 & 7 & 8 & 9
}
\arcstop.
\]
This permutation has a unique left segment $L=\{2,3,4,5\}$, which is mobile.
For the subset $S = \{2,4,5\}$ we have $\sigma_{S,L} = (1,2,4,5)$ and
\[ (\sigma_{S,L})^{-1} \cdot z\cdot \sigma_{S,L} =
(1,7)(2,6)(3,8)(4,9)
= \arcstart
{
*{\color{red}\bullet} \arc{1.4}{rrrrrr} & *{\color{red}\bullet} \arc{1}{rrrr} & *{\bullet} \arc{1.2}{rrrrr} & *{\color{red}\bullet} \arc{1.2}{rrrrr} & *{\color{red}\bullet} & *{\bullet} & *{\bullet} & *{\bullet} & *{\bullet} \\
1 & 2 & 3 & 4 & 5 & 6 & 7 & 8 & 9
}
\arcstop.
\]
\end{example}
Suppose $a_i$ and $a_j$ are left endpoints of $z$ in the same left segment with $i<j$.
We say that $a_j$ is a \defn{crossing bound} of $a_i$ if $\{i\} = \{ t : i\leq t <j\text{ and }b_t<b_j\}$.
Now, given a subset $ S \subseteq \{a_1,a_2,\dots, a_q\}$ we define
\be
\label{varpi-eq}
\textstyle\customvarpi_{z,S} := \prod_{i \in [q]}\customvarpi_{z,S}^{(a_i)}
\ee where
\be
\customvarpi_{z,S}^{(a)} := \begin{cases}
-1&\text{if $a \in S$ and $S$ contains any crossing bound of $a$} \\
2+\beta x_{a} &\text{if }a \notin S \\
1 + \beta x_{a} &\text{otherwise}.
\end{cases}
\ee
We also let $\crb_z(a_i)$ denote the set of crossing bounds of $a_i$.
\begin{example} \label{sigma-ex2}
If $z=(2, 7)(3, 8)(4, 6)(5, 9)$ is as in Example~\ref{sigma-ex} then
\[\ba \customvarpi_{z,\varnothing} &= (2+\beta x_2)(2+\beta x_3)(2+\beta x_4)(2+\beta x_5),\\
\customvarpi_{z,\{2,4,5\} } &= -(1+\beta x_2)(2+\beta x_3)(1+\beta x_5).\ea\]
\end{example}
\begin{definition}
A set $ S \subseteq \{a_1,a_2,\dots, a_q\}$ is \defn{shiftable} relative to $z$ if both
\bei
\item[(1)] no element of $S$ is in the left segment of $z$ containing $1$, and
\item[(2)] if some $a_i \notin S$ then $S$ does not contain any crossing bound of $a_i$.
\eei
Let $\shiftable(z)$ be the set of all such subsets, and for each $S \in \shiftable(z)$
define
\[\iG_{z,S} := \iG_v\quad\text{where }v = (\sigma_S)^{-1} \cdot z\cdot \sigma_S.\]
\end{definition}
The empty set is always shiftable, as is the set of all left endpoints of $z$ outside the unique immobile segment.
A singleton set $S = \{b\}$ is in $\shiftable(z)$
if and only if $b$ belongs to a mobile left segment $L$ of $z$ with the property that every $a \in L$ with $a<b$ has $z(a)>z(b)$.
\begin{example}
When $z=(2, 7)(3, 8)(4, 6)(5, 9)$ there are 9 shiftable subsets of left endpoints,
given by $\emptyset$, $\{2\}$, $\{4\}$, $\{2,3\}$, $\{2,4\}$, $\{4,5\}$,
$\{2,3,4\}$, $\{2,4,5\}$, and $\{2,3,4,5\}$, and it turns out that $\fkGO_z $ can be expressed as a sum of 9 terms
\[
\ba \fkGO_z &=
(2 + \beta x_2)(2 + \beta x_3)(2 + \beta x_4)(2 + \beta x_5) \cdot \iG_{(2, 7)(3, 8)(4, 6)(5, 9)} \\&\quad +
\beta\cdot (1 + \beta x_2)(2 + \beta x_3)(2 + \beta x_4)(2 + \beta x_5)\cdot \iG_{(1, 7)(3, 8)(4, 6)(5, 9)} \\&\quad +
\beta\cdot(2 + \beta x_2)(2 + \beta x_3)(1 + \beta x_4)(2 + \beta x_5)\cdot \iG_{(1, 6)(2, 7)(3, 8)(5, 9)} \\&\quad -
\beta^2\cdot(1 + \beta x_3)(2 + \beta x_4)(2 + \beta x_5)\cdot \iG_{(1, 7)(2, 8)(4, 6)(5, 9)} \\&\quad +
\beta^2\cdot(1 + \beta x_2)(2 + \beta x_3)(1 + \beta x_4)(2 + \beta x_5)\cdot \iG_{(1, 7)(2, 6)(3, 8)(5, 9)} \\&\quad -
\beta^2\cdot(2 + \beta x_2)(2 + \beta x_3)(1 + \beta x_5)\cdot \iG_{(1, 6)(2, 7)(3, 8)(4, 9)} \\&\quad -
\beta^3\cdot(1 + \beta x_3)(1 + \beta x_4)(2 + \beta x_5)\cdot \iG_{(1, 7)(2, 8)(3, 6)(5, 9)} \\&\quad -
\beta^3\cdot(1 + \beta x_2)(2 + \beta x_3)(1 + \beta x_5)\cdot \iG_{(1, 7)(2, 6)(3, 8)(4, 9)} \\&\quad +
\beta^4\cdot(1 + \beta x_3)(1 + \beta x_5)\cdot\iG_{(1, 7)(2, 8)(3, 6)(4, 9)}.
\ea
\]
\end{example}
These terms correspond to the right side of the following formula.
\begin{theorem}\label{ivex-thm}
If $z \in \Ivex$ then $\fkGO_z = \sum_{S\in\shiftable(z)} \customvarpi_{z,S} \cdot \beta^{|S|} \cdot \iG_{z , S}$.
%where the sum is over all shiftable subsets of left endpoints of $z$.
\end{theorem}
We give the proof of this theorem in the next section.
This result should be compared with \cite[Thm.~3.26]{MP2020},
which presents a different formula for $\fkGO_z$ in the vexillary case in terms of certain Pfaffians.
Our new formula in Theorem~\ref{ivex-thm} is more explicit and easier to compute.
\begin{remark}\label{locally-noncrossing-remark}
For a general $z \in \Ivex$, there are many cancellations in the formula in Theorem~\ref{ivex-thm},
although we know that $\fkGO_z\in \NN[\beta][x_1,x_2,\dots]$.
One can show that $\customvarpi_{z,S} \in \NN[\beta][x_1,x_2,\dots]$
for all $S \in \shiftable(z)$ if and only if
$z$ is \defn{locally noncrossing} in the sense that
any two numbers $a<b$ in the same mobile left segment of $z$ have $z(a)>z(b)$.
In this case, since every $\iG_y\in \NN[\beta][x_1,x_2,\dots]$,
Theorem~\ref{ivex-thm} gives a monomial-positive formula for $\fkGO_z$.
The numbers $1$, $2$, $4$, $9$, $20$, $47$, $109$, $261$, $621$, $1511$, $3664$, $9052$, $\dots$ of locally noncrossing vexillary involutions in $S_n$
do not appear to given a known integer sequence.
The slightly more natural sequence of locally noncrossing vexillary involutions $z \in S_{n+1}$ with $z(1)=1$
(whose counts are $1$, $2$, $4$, $8$, $17$, $36$, $77$, $166$, $359$, $780$, $1700$, $3715$, $\dots$)
also appears to be unknown.
\end{remark}
\subsection{Proof of Theorem~\ref{ivex-thm}}\label{proof-sect}
We continue the setup of the previous section, so $z \in \Ivex$ is a vexillary involution
with left endpoints $a_1<a_2<\dots<a_q$ and right endpoints $b_1,b_2,\dots,b_q$.
We begin with three technical lemmas about shiftable subsets.
\begin{lemma}\label{shiftable-lem}
Let $S \in \shiftable(z)$. Choose a left segment $L$ of $z$ with $i \in S\cap L$.
Write $j $ for the smallest element $j \in S$ with $i<j$, when this exists.
\ben
\item[(a)] If $\crb_z(i)\cap S = \varnothing$
and $i \neq \max (S\cap L)$ then $z(i) > z(j)$.
\item[(b)] If $\crb_z(i)\cap S \neq \varnothing$
then $i \neq \max (S\cap L)$ and $z(i) < z(j)$.
\een
\end{lemma}
\begin{proof}
First suppose $\crb_z(i)\cap S = \varnothing$
and $i \neq \max (S\cap L)$ but
the smallest $j \in S$ with $i<j$ has $z(i) < z(j)$. Then since $j$ cannot be a crossing bound of $i$,
the maximal $t$ with $ i \leq t < j $ and $z(t)<z(j)$ must be greater than $i$.
But then $j \in \crb_z(t)$ so we must have $t \in S$ since $j \in S$ and $S$ is shiftable.
This contradicts the minimality of $j$ and proves part (a).
For part (b), suppose $\crb_z(i)\cap S \neq \varnothing$. Then some $k \in S\cap L$
has $i<k$ and $k \in \crb_z(i)$, so $i\neq\max(S\cap L)$.
If $j=k$ then $z(i) < z(j)$ by the definition of a crossing bound.
If $j<k$ then since $\{i\} = \{ t: i\leq t < k \text{ and }z(t)<z(k)\}$,
we must have $z(i) < z(k) < z(j)$.
\end{proof}
\begin{lemma}\label{shiftable-lem2}
Suppose $L$ is a left segment of $z$ and $S \in \shiftable(z)$. Assume $i<j$ are left endpoints with
$i \in S\cap L$ and $j \in L\setminus S$.
If every integer $t$ with $i\leq t <j$ has $z(t)>z(j)$,
then $S \sqcup\{j\} \in \shiftable(z)$.
\end{lemma}
\begin{proof}
We argue by induction on $i$.
The union $S \sqcup\{j\}$ can only fail to be shiftable if some left endpoint $h\in L$ has $h \notin S$ and $j \in \crb_z(h)$. Such an element must have $z(h)<z(j)$ so $h<i$, and then every $t$ with $h<t<j$ must have $z(t)>z(j)$. This is impossible if $i = \min (L)$, so the lemma holds in that base case.
Assume $i> \min(L)$ and such as $h$ exists.
As we cannot have $ i \in \crb_z(h)$ since $h\notin S$ and $i \in S$, there is some maximal $t$ with $h<t <i$ and $z(j)<z(t)<z(i)$.
Then by construction $i \in \crb_z(t)$ so $t \in S\cap L$ since $S$ is shiftable.
But this means the hypotheses of the lemma are satisfied
if we replace $i$ by the smaller left endpoint $t$, so we conclude by induction that $S \sqcup\{j\} \in \shiftable(z)$.
\end{proof}
\begin{lemma}\label{shiftable-lem3}
Suppose $L$ is a left segment of $z$ that intersects $S \in \shiftable(z)$.
Let $m:= \min(S\cap L)$ and assume $i\in L\setminus S$ has $i< m$.
Then $z(m) < z(i)$.
\end{lemma}
\begin{proof}
Suppose instead that $z(i)<z(m)$. Since $S$ is shiftable but does not contain $i$, the left endpoint
$m$ cannot be a crossing bound for $i$.
Hence the maximal left endpoint $t\in L$ with $i\leq t<m$ and $z(t) <z(m)$ is greater than $i$.
But then $m $ is a crossing bound for $t$, so we must have $t \in S$
since $S$ is shiftable and contains $m$.
This contradicts the minimality of $m=\min(S\cap L)$.
\end{proof}
We now begin the main proof of this section.
\begin{proof}[Proof of Theorem~\ref{ivex-thm}]
The basic strategy in this long proof is fairly straightforward.
We will argue by reverse induction on the length of $z$, the base case being when $z$ is dominant. Then,
$z$ only has one left segment, which is immobile, so only the empty set is shiftable,
and the desired formula for $\fkGO_z$ reduces to the one given in Theorem~\ref{qd-thm}.
Now assume that $z \in \Ivex$ is such that $\fkGO_z = \sum_{S\in \shiftable(z)} \customvarpi_{z,S} \cdot \beta^{|S|}\cdot \iG_{z , S}$.
Suppose $i \in \PP$ has $z(i)>z(i+1)$ and $z \neq s_i zs_i \in \Ivex$. Set \[y := s_i zs_i \in \Ivex.\]
Then $\fkGO_y = \bpartial_i \fkGO_z$ by \eqref{orthogonal-recursion},
so by Theorem~\ref{vexweak-prop} it suffices to show that
\be\label{goal-eq}
\sum_{S\in \shiftable(y)} \customvarpi_{y,S} \cdot \beta^{|S|}\cdot \iG_{y , S}
=
\sum_{S\in \shiftable(z)} \beta^{|S|}\cdot \bpartial_i(\customvarpi_{z,S} \cdot \iG_{z , S}).
\ee
The argument to check this splits into several cases according to how $i$ and $i+1$ are
distributed among the left endpoints, right endpoints, and fixed points of $z$.
Since $i$ is a descent of $z$,
it cannot occur that $i$ and $i+1$ are both fixed points of $z$, or
that $i$ is a fixed point and $i+1$ is a left endpoint,
or
that $i$ is a right endpoint and $i+1$ is a fixed point
Also, since $y$ and $z$ are distinct and both vexillary, it cannot occur that
$i$ is a right endpoint and $i+1$ is a left endpoint,
or that $i$ is a left endpoint and $i+1$ is a right endpoint.
These considerations exclude 5 of 9 cases, leaving 4 possibilities:
\ben
\item[(a)] The number $i$ is a fixed point of $z$ while $i+1$ is a right endpoint.
\item[(b)] Both $i$ and $i+1$ are right endpoints of $z$.
\item[(c)] Both $i$ and $i+1$ are left endpoints of $z$.
\item[(d)] The number $i$ is a left endpoint of $z$ while $i+1$ is a fixed point.
\een
We explain why \eqref{goal-eq} holds in each of these cases below.
\ben
\item[(a)] Suppose $i$ is a fixed point of $z$ and $i+1$ is a right endpoint.
\een
In this case $y$ and $z$ have the same left segments, crossing bounds,
and shiftable subsets.
Suppose $S \in \shiftable(y)= \shiftable(z)$.
As $i$ must be greater than the largest left endpoint of $z\in \Ivex$,
it follows that $\bpartial_i$ commutes with $\customvarpi_{y,S}=\customvarpi_{z,S}$ and $s_i$ commutes with $\sigma_S$. Moreover, $i$ is still a fixed point and $i+1$ is still a right endpoint of $(\sigma_S)^{-1} \cdot z \cdot \sigma_S$, so we have $\bpartial_i \iG_{z,S} = \iG_{y,S}$ and
\be\label{case-a-eq}
\bpartial_i(\customvarpi_{z,S} \cdot \iG_{z , S}) %= \customvarpi_{z,S} \cdot \bpartial_i \iG_{z,S}
= \customvarpi_{y,S} \cdot \iG_{y,S}
.\ee
Substituting this formula turns the left side of \eqref{goal-eq} into the right.
\ben
\item[(b)] Suppose $i$ and $i+1$ are both right endpoints of $z$.
\een
In this case $i$ is again greater than the largest left endpoint of $z$,
so $\bpartial_i$ commutes with $\customvarpi_{z,S}$ for all $S \in \shiftable(z)$.
Since $i$ is a descent of $z$, for some indices $1\leq j <k \leq q$
it holds that $z(i+1) = a_j$ and $z(i) = a_k$. If $a_j$ and $a_k$ are not in the same left segment of $z$,
or are both in the unique immobile left segment (when this exists),
then the argument to deduce \eqref{goal-eq} is the same as in case (a),
since then $y$ and $z$ have the same left segments, crossing bounds,
and shiftable subsets, so \eqref{case-a-eq} again holds.
Instead assume that $a_j$ and $a_k$ belong to the same mobile left segment of $z$.
Then $y=s_izs_i$ has the same left segments as $z$,
and almost the same nontrivial cycles:
the only difference is that $(a_j,i+1)$ and $(a_k,i)$ have been replaced by $(a_j,i)$ and $(a_k,i+1)$.
This means that the crossing bounds
for the left endpoints of $y$ are the same as for $z$, with one exception:
if $i+1<b_t$ for all $j<t<k$
then $a_k$ will be a crossing bound for $a_j$ in $y$ but not in $z$.
If this exception does not occur then \eqref{case-a-eq} again holds for all
$S \in \shiftable(y)=\shiftable(z)$, which suffices to prove \eqref{goal-eq}.
For the exceptional case, we make a claim:
\begin{claim} Assume we are in the case when $i+1<b_t$ for all $j<t<k$. Let
\[
\ba
\cE &:=\{ S \in \shiftable(z) : a_j \notin S \text{ and }a_k \in S\}, \\
\cF &:= \{ S \in \shiftable(z) : a_j,a_k \in S \text{ and }a_t \notin S\text{ for all }j<t<k\} .
\ea
\]
Then
$ \shiftable(y) = \shiftable(z) \setminus \cE$
and $S \mapsto S\sqcup\{a_j\}$ is a bijection $\cE \to \cF$.
\end{claim}
\begin{proof}[Proof of the claim]
We can always remove $a_j$ from $T \in \cF$ to obtain a shiftable set $T\setminus\{a_j\}\in\cE$,
since such a set $T$ contains no crossing bounds for $a_j$, as these would have to be of the form $a_t$ for some $j<t<k$.
On the other hand,
if $S \in \cE$ then any left endpoint $a_t$ that has $a_j$ as a crossing bound in $z$ must also have $a_k$
as a crossing bound, and therefore is already in $S$, so we can add $a_j$ to get a shiftable subset
$S\sqcup\{a_j\} \in \shiftable(z)$.
Now we check that $a_t \notin S$ for all $j<t<k$ if $a_j \notin S \in \shiftable(z)$.
This follows by induction on $t$ as each $a_t$ is a crossing bound for one of $a_j,a_{j+1},\dots,a_{t-1}$ as
$b_j =i+1< b_t$.
Hence
if $a_j,a_{j+1},\dots,a_{t-1} \notin S$ then also $a_t \notin S$.
We conclude that if $S \in \cE$ then the union $S\sqcup\{a_j\} $ is shiftable and contained in $\cF$.
\end{proof}
Continue to assume $i+1<b_t$ for all $j<t<k$, so that the claim applies.
If $S \in \shiftable(y) \setminus \cF$ then $S$ contains neither $a_j$ nor $a_k$ so $\customvarpi_{y,S} = \customvarpi_{z,S}$.
If instead $S \in \cF$ then
we obtain $\customvarpi_{y,S}$ from $\customvarpi_{z,S} =\prod_{t\in[q]} \customvarpi_{z,S}^{(a_t)}$ by replacing the single
factor $\customvarpi_{z,S}^{(a_j)} = 1+\beta x_{a_j}$ by $\customvarpi_{y,S}^{(a_j)} =-1$. This means that if $S \in \cF$
then
\be
\label{case-b-eq1}
\customvarpi_{y,S} = \customvarpi_{z,S} - \customvarpi_{z,T}\quad\text{for $T := S\setminus \{a_j\} \in \cE$.}
\ee
Finally, we observe that if $S \in \shiftable(y)=\shiftable(z)\setminus\cE$ then
\be
\label{case-b-eq2}
\beta^{|S|} \cdot \bpartial_i(\customvarpi_{z,S} \cdot \iG_{z , S}) = \customvarpi_{z,S} \cdot \beta^{|S|} \cdot \bpartial_i \iG_{z,S} =\customvarpi_{z,S} \cdot \beta^{|S|} \cdot \iG_{y,S}
\ee
while if $T \in \cE$ and $S := T \sqcup\{a_j\} \in \cF$ then, as $a_t \notin T$ for $j<t<k$, we have
\be
\label{case-b-eq3}
\beta^{|T|}\cdot \bpartial_i(\customvarpi_{z,T} \cdot \iG_{z ,T}) =
\customvarpi_{z,T} \cdot \beta^{|T|}\cdot \bpartial_i \iG_{z,T} =
- \customvarpi_{z,T} \cdot \beta^{|S|} \cdot \iG_{y,S}.
\ee
Since $T\mapsto T\sqcup\{a_j\}$ is a bijection $\cE\to\cF$,
substituting \eqref{case-b-eq2} and \eqref{case-b-eq3} into the right side of \eqref{goal-eq}
and then using \eqref{case-b-eq1} gives the left side \eqref{goal-eq}, as needed.
\ben
\item[(c)] Suppose $i$ and $i+1$ are both left endpoints of $z$.
\een
In this case $y$ and $z$ have the same left endpoints and left segments,
% and we can write $i=a_j$ and $i+1=a_{j+1}$ for some $1\leq j< q$ with $b_j>b_{j+1}$.
and the numbers $i$ and $i+1$ belong to the same left segment.
Notice that the set $\crb_z(i) $ of crossing bounds for $i$ in $z$ is empty since whenever $i <t $ and $z(i) < z(t)$ we also have $z(i+1)<z(t)$.
Our proof of \eqref{goal-eq} in this case reduces to two somewhat technical claims,
which we state and prove below.
\begin{claim}
\label{c-claim1}
Continue to assume $i$ and $i+1$ are both left endpoints of $z$. Then
\[
\sum_{\substack{S \in \shiftable(z) \\ i \in S}} \beta^{|S|}\cdot \bpartial_i(\customvarpi_{z,S} \cdot \iG_{z , S})
=
0.
\]
\end{claim}
\begin{proof}
If $i$ is part of a left segment of $z$ containing $1$, then the equality holds trivially because the sum is empty.
We may therefore
assume that $i$ and $i+1$ belong to a left segment that does not contain $1$. (This assumption is not actually used in any part of the following argument, however.)
%For any left endpoint $a$ of $z$, recall that $\crb_z(a)$ denotes its set of crossing bounds.
To prove the claim, it suffices to verify the following three statements:
\ben
\item[(1)] If $ S \in \shiftable(z)$ has $i,i+1 \in S$ and $S \cap \crb_z(i+1)=\varnothing$ then
\[
\beta^{|S|}\cdot \bpartial_i(\customvarpi_{z,S} \cdot \iG_{z , S}) = \customvarpi_{z,S} \cdot \beta^{|S|}\cdot\iG_{z,T}
\]
for the set $T:=S\setminus\{i+1\} $, which belongs to $ \shiftable(z)$ and satisfies
\[
\beta^{|T|}\cdot \bpartial_i(\customvarpi_{z,T} \cdot \iG_{z , T}) = - \customvarpi_{z,S} \cdot \beta^{|S|}\cdot \iG_{z , T}.
\]
\item[(2)] If $T \in \shiftable(z)$ has $i \in T$ and $i+1\notin T$ then $ S:= T\sqcup\{i+1\} $ satisfies
\[
i,i+1\in S \in \shiftable(z)
\quand
S \cap \crb_z(i+1)=\varnothing.\]
\item[(3)] If $i,i+1 \in S\in \shiftable(z)$ and $S \cap \crb_z(i+1)\neq \varnothing$ then
$
\bpartial_i(\customvarpi_{z,S} \iG_{z , S}) = 0.
$
\een
To show (1), choose $ S \in \shiftable(z)$ with $i,i+1 \in S$ and $S \cap \crb_z(i+1)=\varnothing$ and define $T:=S\setminus\{i+1\} $. Since $S$ does not contain any crossing bounds for $i+1$,
it is evident that $T \in \shiftable(z)$.
Since $\crb_z(i)=\varnothing$, we have
\[\customvarpi_{z,S}^{(i)}\customvarpi_{z,S}^{(i+1)} = (1+\beta x_i)(1+\beta x_{i+1})\] so $\bpartial_i$ commutes with $\customvarpi_{z,S}$.
It follows from Lemma~\ref{shiftable-lem}(a) that
\[\bpartial_i( \iG_{z , S}) = \iG_{z,T}
\quand
\bpartial_i( \iG_{z , T}) =-\beta \iG_{z,T}.
\]
Therefore, we conclude from \eqref{fg-eq} that
\[
\beta^{|S|} \cdot \bpartial_i(\customvarpi_{z,S} \cdot \iG_{z , S})
%= \beta^{|S|} \cdot \customvarpi_{z,S} \cdot \bpartial_i \iG_{z , S}
=
\customvarpi_{z,S} \cdot \beta^{|S|} \cdot\iG_{z,T}
\]
and
\[
\beta^{|T|} \cdot \bpartial_i(\customvarpi_{z,T} \cdot \iG_{z , T}) = (\bpartial_i \customvarpi_{z,T})\cdot \beta^{|S|-1} \cdot \iG_{z , T}.
\]
It remains to check that $\bpartial_i \customvarpi_{z,T} = - \customvarpi_{z,S} \cdot \beta$. As we have \[
\customvarpi_{z,T}^{(i)}\customvarpi_{z,T}^{(i+1)} = (1+\beta x_i)(2+\beta x_{i+1})\]
the desired identity follows by checking that
\[
\bpartial_i\( (1+\beta x_i)(2+\beta x_{i+1})\) = -\beta(1+\beta x_i)(1+\beta x_{i+1}).
\]
This proves item (1).
We proceed to item (2). Assume $T \in \shiftable(z)$ has $i \in T$ and $i+1\notin T$. Define $ S:= T\sqcup\{i+1\} $. Since $T$ is shiftable, we must have $T\cap \crb_z(i+1) = \varnothing$, so it also holds that $S \cap \crb_z(i+1) = \varnothing$. Finally, Lemma~\ref{shiftable-lem2} shows that $S \in \shiftable(z)$
as needed.
For item (3), assume $S\in \shiftable(z)$ has $i,i+1 \in S$ and $S \cap \crb_z(i+1)\neq \varnothing$. Then
$
\customvarpi_{z,S}^{(i)}\customvarpi_{z,S}^{(i+1)} = -1-\beta x_i
$ so $\bpartial_i \customvarpi_{z,S}=0$.
It suffices in view of \eqref{fg-eq} to check that $\bpartial_i \iG_{z , S} = -\beta \iG_{z,S}$,
and this follows by Lemma~\ref{shiftable-lem}(b).
\end{proof}
Our second claim involves the following definitions.
Given $S\in\shiftable(z)$, let $m_S$ be the smallest element of $ S$ with $i+1<m_S$, if this exists. Then let
\[
\ba
\cA &:= \{ S \in \shiftable(z) : i,i+1 \notin S\}, \\
\mathcal{B} &:= \{ S \in \shiftable(z) : i\notin S,\ i+1\in S,\ S \cap \crb_z(i+1)= \varnothing \}, \\
\cC &:= \{ S \in \shiftable(z) : i\notin S,\ i+1\in S,\ S \cap \crb_z(i+1)\neq \varnothing,\ z(m_S)<z(i) \}, \\
\cD &:= \{ S \in \shiftable(z) : i\notin S,\ i+1\in S,\ S \cap \crb_z(i+1)\neq \varnothing,\ z(i)<z(m_S) \}.
\ea
\]
Lemma~\ref{shiftable-lem} ensures that $m_S$ is defined in the formulas for $\cC$ and $\cD$.
If $X$ and $Y$ are two sets then we let $X\vartriangle Y := (X \setminus Y) \sqcup (Y\setminus X)$.
To establish \eqref{goal-eq}, it suffices to combine our first claim above with the following claim below:
\begin{claim}\label{c-claim2}
Continue to assume $i$ and $i+1$ are both left endpoints of $z$. Then
\[ \shiftable(y) = \cA \sqcup \{ S \vartriangle\{i,i+1\} : S \in \mathcal{B} \sqcup \cC\}
\sqcup \{ S \sqcup \{i\} : S \in \mathcal{B} \sqcup \cC \sqcup \cD\}.\]
Moreover, if $S \in \cA\sqcup\mathcal{B}\sqcup\cC\sqcup\cD$ and
$T := S\vartriangle \{i,i+1\}$ and $U := S\sqcup\{i\}$, then
\[
\beta^{|S|}\cdot \bpartial_i(\customvarpi_{z,S} \cdot \iG_{z , S})=
\begin{cases}
\customvarpi_{y,S} \cdot \beta^{|S|}\cdot \iG_{y,S}&\text{if }S \in \cA \\
\customvarpi_{y,T}\cdot \beta^{|T|}\cdot \iG_{y , T} +\customvarpi_{y,U} \cdot \beta^{|U|}
\cdot \iG_{y , U} &\text{if }S \in \mathcal{B}\sqcup\cC \\
\customvarpi_{y,U} \cdot \beta^{|U|} \cdot \iG_{y , U} &\text{if }S \in \cD.
\end{cases}
\]
\end{claim}
\begin{proof}
We can derive the formula for $\beta^{|S|}\cdot \bpartial_i(\customvarpi_{z,S} \cdot \iG_{z , S})$
directly from \eqref{fg-eq},
since the setup determines the value of $\customvarpi_{z,S}^{(i)} \customvarpi_{z,S}^{(i+1)}$ in each case
and Lemma~\ref{shiftable-lem} tells us how to evaluate $\bpartial_i \iG_{z,S}$ via Theorem~\ref{iG-thm}.
The argument is similar to the proof of Claim~\ref{c-claim1} (and also to the proofs of Claims~\ref{d-claim2} and \ref{d-claim3} below) so we omit the details.
The nontrivial statement left to prove is the desired expression for $\shiftable(y)$.
To show this, we start by observing that while
$\crb_z(i) = \varnothing$, it holds that
\[\ba
\crb_y(i) &= \{i+1\}\sqcup \{ t \in \crb_z(i+1) : z(i+1) < z(t)< z(i)\},
\\
\crb_y(i+1) &= \{ t \in \crb_z(i+1) : z(i) < z(t)\}.\ea
\]
Meanwhile, for any other left endpoint $t$, the set of crossing bounds $\crb_y(t)$ is either equal to
$\crb_z(t)$, or formed from $\crb_z(t)$ by removing $i$ or $i+1$, and the second case only occurs when $\crb_z(t)$ already contains at least $i$. Specifically, for any left endpoint $t \notin\{i,i+1\}$, we are in one of the following cases:
\ben
\item[(i)] when $t<i$ and $z(t) < z(i+1)$, it can happen that
\[ i,i+1\in \crb_z(t)
\quand \crb_y(t) = \crb_z(t) \setminus\{i+1\}.\]
\item[(ii)] when $t<i$ and $z(i+1)<z(t)<z(i)$, it can happen that
\[i \in \crb_z(t),\ i+1\notin\crb_z(t)
\quand
\crb_y(t) = \crb_z(t) \setminus\{i\}.\]
\item[(iii)] if cases (i) or (ii) do not apply, then $i,i+1 \notin \crb_y(t)=\crb_z(t)$.
\een
It follows that if $S$ is a set containing neither $i$ nor $i+1$, then $S \in \shiftable(z)$ if and only if $S \in \shiftable(y)$.
Alternatively, if $S \in \shiftable(z)$ and $i \notin S$,
then the only way $S\sqcup\{i\}$ can fail to be in $\shiftable(y)$ is if some left endpoint $t \notin S$
has $i \in \crb_y(t)$. In view of (i), (ii), and (iii), this can only happen when $i+1 \in \crb_z(t)$, which would mean that $i+1 \notin S$.
Thus, if $S \in \shiftable(z)$ has $i \notin S$ and $i+1 \in S$, so that
$S \in \mathcal{B}\sqcup\cC\sqcup\cD$,
then we have $S\sqcup\{i\}\in \shiftable(z)$ and this union contains both $i$ and $i+1$.
Conversely, if $U \in \shiftable(y)$ has $i,i+1 \in U$ then since $\crb_z(i)=\varnothing$,
the only way $U \setminus \{i\}$ can fail to be in $\shiftable(z)$ is if some left endpoint $t \notin U$
has $i +1\notin \crb_y(t)$ but $i+1 \in \crb_z(t)$. Inspecting (i) and (ii) shows that this situation is impossible,
so if $U \in \shiftable(y)$ has $i,i+1 \in U$ then $U \setminus\{i\} \in \shiftable(z)$.
Finally suppose $S$ is a set containing $i+1$ but not $i$. Form $T := S\vartriangle\{i,i+1\}$ by removing $i+1$ and adding $i$.
It is not possible for a left endpoint $t$ to have $i \in \crb_y(t)$ and $i+1\notin \crb_z(t)$,
so if $S \in \shiftable(z)$ then we can only fail to have $T \in \shiftable(y)$
if $S \cap \crb_y(i+1) \neq \varnothing$.
This only occurs if $S$ contains some $t \in \crb_z(i+1)$ with $z(i)<z(t)$,
but then we must have $t=m_S$ or $z(i)<z(t) < z(m_S)$.
We conclude that if $S \in \mathcal{B}\sqcup\cC$ then $T \in \shiftable(y)$.
Conversely, since $\crb_z(i)=\varnothing$, if $T \in \shiftable(y)$
then we can only fail to have $S \in \shiftable(z)$
is there is some left endpoint $t \notin T$ with $i\notin \crb_y(t)$ and $i+1 \in \crb_z(t)$.
Inspecting (i) and (ii) shows that this situation is impossible,
so if $T \in \shiftable(y)$ then $S \in \shiftable(z)$.
In fact,
if $T \in \shiftable(y)$ then $S \in \mathcal{B} \sqcup \cC$,
since $i+1 \in S$ and if $m_S$ is defined
then $y(m_S)< y(i+1)$ by Lemma~\ref{shiftable-lem3}
so
$z(m_S) =y(m_S)< y(i+1)=z(i)$.
This argument shows that the given union is contained in $\shiftable(y)$,
and that each element of $\shiftable(y)$---except those containing $i+1$ but not $i$---is contained in this union.
As no set in $\shiftable(y)$ contains $i+1$ but not $i$ since $i+1 \in \crb_y(i)$,
this completes the proof of our claim.
\end{proof}
The identity \eqref{goal-eq} in case (c) is immediate from Claims~\ref{c-claim1} and \ref{c-claim2}.
This leaves just one final case to consider:
\ben
\item[(d)] Suppose $i$ is a left endpoint of $z$ and $i+1$ is a fixed point.
\een
In this case $i$ is the last element of its left segment in $z$, so $\crb_z(i) = \varnothing$.
However, the left endpoints for $y$ are no longer the same as for $z$. Compared to $z$,
the set of left endpoints for $y$ loses $i$ and gains $i+1$.
This means that the left segments of $y$ also differ from $z$.
Let $L$ be the left segment of $z$ containing $i$.
If $i+2$ is not a left endpoint of $z$, then the left segments of $y$ are formed from those of $z$
by replacing $L$ by $L\setminus\{i\}$ and adding a new singleton segment $\{i\}$.
If $i+2$ is a left endpoint of $z$, then it begins a new left segment $M$ of $z$.
In this event, the left segments of $y$ are formed from those of $z$
by replacing $L$ by $L\setminus \{i\}$ and $M$ by $\{i+1\}\sqcup M$.
These sub-cases require slightly different arguments,
which we handle as usual through a series of claims. To start, we have this close relative of Claim~\ref{c-claim1}:
\begin{claim}\label{d-claim1} Continue to assume that $i<z(i)$ and $i+1=z(i+1)$. Then
\[
\sum_{\substack{S \in \shiftable(z) \\ i \in S}} \beta^{|S|}\cdot \bpartial_i(\customvarpi_{z,S} \cdot \iG_{z , S})
=
0.
\]
\end{claim}
\begin{proof}
This holds for a much simpler reason than Claim~\ref{c-claim1}.
Assume $S\in \shiftable(z)$ has $i \in S$. Since $i$ is the last element of its left segment, we have $\crb_z(i)=\varnothing$, so
$
\customvarpi_{z,S}^{(i)} = 1+\beta x_i$
and $\bpartial_i \customvarpi_{z,S}=0$.
At the same time, it is also clear that $\bpartial_i \iG_{z , S} = -\beta \iG_{z,S}$
so in fact $\bpartial_i(\customvarpi_{z,S} \cdot \iG_{z , S})=0$ by \eqref{fg-eq}.
\end{proof}
We now have two variants of Claim~\ref{c-claim2}.
Let $\cX = \{S \in \shiftable(z) : i \notin S\}$.
\begin{claim}\label{d-claim2}
If $i$ but not $i+2$ is a left endpoint of $z$, and $i+1=z(i+1)$,
then \[\shiftable(y) = \cX \sqcup \{ S\sqcup\{i+1\} : S \in \cX\}.\]
Moreover, in this case, if $S \in \cX$ and $T := S\sqcup\{i+1\}$ then
\[
\beta^{|S|}\cdot \bpartial_i(\customvarpi_{z,S} \cdot \iG_{z , S})
=
\customvarpi_{y,S}\cdot \beta^{|S|}\cdot \iG_{y , S} +\customvarpi_{y,T} \cdot \beta^{|T|}
\cdot \iG_{y , T}.
\]
\end{claim}
\begin{proof}
In this case $\crb_y(t) = \crb_z(t) \setminus \{i\}$ for all left endpoints $t\neq i$,
although the set difference may often satisfy $\crb_z(t) \setminus \{i\} = \crb_z(t)$.
This property makes it clear that $\cX \subset \shiftable(y)$,
Then, since $\{i\}$ is a left segment in $y$,
it also follows that $ \{ S\sqcup\{i+1\} : S \in \cX\} \subset \shiftable(y)$.
Conversely, if $S \in \shiftable(y)$ then $i\notin S$ as $i$ is not a left endpoint of $y$,
and the same observation about crossing bounds implies that $S\setminus\{i+1\} \in \cX$.
This proves our formula for $\shiftable(y)$.
Now let $S \in \cX$ and $T := S\sqcup\{i+1\}$.
Let $\cQ = \prod_{a \neq i,i+1} \customvarpi_{z,S}^{(a)} = \prod_{a \neq i, i+1} \customvarpi_{y,S}^{(a)}$
where the products are over the left endpoints $a$ shared by $y$ and $z$. Then
\[
\customvarpi_{z,S} = (2+\beta x_i)\cQ,
\quad
\customvarpi_{y,S} = (2+\beta x_{i+1})\cQ,
\quand
\customvarpi_{y,T} = (1+\beta x_{i+1})\cQ.
\]
As $\cQ$ does not involve the variables $x_i$ or $x_{i+1}$,
we have
\[ s_i(\customvarpi_{z,S}) = \customvarpi_{y,S}
\quand
\bpartial_i(\customvarpi_{z,S}) = -\beta \cQ,\] so we can compute that
\[ s_i(\customvarpi_{z,S})\cdot \beta + \bpartial_i(\customvarpi_{z,S}) = \beta \cdot ((2+\beta x_{i+1}) - 1) \cdot \cQ =
\beta\cdot \customvarpi_{y,T}.\]
Since it is clear that
$\iG_{z,S} = \iG_{y,T}$ and $\bpartial_i \iG_{z,S} = \iG_{y,S}$,
the desired expression for $ \beta^{|S|}\cdot \bpartial_i(\customvarpi_{z,S} \cdot \iG_{z , S})$ follows from \eqref{fg-eq} using the preceding identities.
\end{proof}
Now suppose that $i$ and $i+2 $ are both left endpoints of $z$ while $i+1=z(i+1)$.
As above let $M$ be the left segment of $z$ containing $i+2$.
Then define
\[
\ba
%\cU &= \{ S \in \cX: S\cap L =\varnothing\}, \\
\cV &= \{ S \in \cX: S\cap M =\varnothing\text{ or } z(i)>z(\min(S\cap M)) \},\\
\cW &=\{ S \in \cX : S\cap M\neq\varnothing\text{ and }z(i)<z(\min(S\cap M)) \}.
\ea
\]
Here is our second variant of Claim~\ref{c-claim2}:
\begin{claim}\label{d-claim3}
Assume $i$ and $i+2$ are left endpoints of $z$ while $i+1=z(i+1)$. Then
\[ \shiftable(y) = \cV
\sqcup \{ S \sqcup \{i+1\} : S \in \cV \sqcup \cW=\cX\}.\]
Moreover, if $S \in\cV\sqcup\cW$ and $T := S\sqcup\{i+1\}$ then
\[
\beta^{|S|}\cdot \bpartial_i(\customvarpi_{z,S} \cdot \iG_{z , S})=
\begin{cases}
\customvarpi_{y,S}\cdot \beta^{|S|}\cdot\iG_{y , S} +\customvarpi_{y,T} \cdot\beta^{|T|} \cdot \iG_{y , T} &\text{if }S \in \cV \\
\customvarpi_{y,T} \cdot \beta^{|T|}\cdot \iG_{y , T} &\text{if }S \in \cW.
\end{cases}
\]
\end{claim}
\begin{proof}
The proof of this statement is similar to Claim~\ref{d-claim2}, and only slightly more complicated.
As in the earlier proof, we have $\crb_y(t) = \crb_z(t) \setminus \{i\}$ for all left endpoints $t\neq i$,
though it may often hold that $ \crb_z(t) \setminus \{i\} =\crb_z(t)$.
Whereas $\crb_z(i)=\varnothing$, the set of crossing bound $\crb_y(i+1)$ may be a nonempty subset of the left segment $M$.
Suppose $S \in \cX$.
Then it is clear from the previous paragraph and the definition of a shiftable set that $S\sqcup\{i+1\} \in \shiftable(y)$.
Moreover, the only way we can fail to have $S \in \shiftable(y)$ is if $S\cap \crb_y(i+1)\neq \varnothing$.
If $S\cap M$ is empty, then this is impossible so
$S \in \shiftable(y)$.
Assume instead that $S\cap M \neq \varnothing$.
Let $m =\min(S\cap M)$ and suppose
$z(m) <z(i)$. Then $y(m) = z(m) < z(i) = y(i+1)$, which means that $m\notin \crb_y(i+1)$,
and
also that no $t \in \crb_y(i+1)$ is greater than $m$, since if $m<t$ and $y(i+1) < y(t)$ then $y(m) <y(i+1)< y(t)$.
In other words, it follows that $S\cap \crb_y(i+1)=\varnothing$,
so $S \in \shiftable(y)$.
We conclude that if $S \in \cV$ then $S\in \shiftable(y)$.
This shows that $\shiftable(y)$ contains the given union.
Conversely, suppose $T \in \shiftable(y)$ and let $S := T\setminus \{i+1\}$.
Then $i \notin S$ since $i$ is not a left endpoint of $y$, and it follows
from our earlier observations about the crossing bounds of $y$ versus $z$ that
$S \in \cX$.
Assume $i+1 \notin T$ so that $S=T$. It remains only to check that $S \in \cV$.
For this, observe that if $S\cap M \neq \varnothing$
and $m=\min(S\cap M) = \min(T\cap M)$,
then we have $y(m) < y(i+1)$ by Lemma~\ref{shiftable-lem3}
and therefore $z(m) = y(m) < y(i+1) = z(i)$.
This completes our proof of the formula for $\shiftable(y)$.
We now compute $\beta^{|S|}\cdot \bpartial_i(\customvarpi_{z,S} \cdot \iG_{z , S})$
using \eqref{fg-eq}.
The argument is almost the same as in the proof of Claim~\ref{d-claim2}.
Choose $S \in \cX$ and set $T := S\sqcup\{i+1\}$.
Again let $\cQ = \prod_{a \neq i,i+1} \customvarpi_{z,S}^{(a)} = \prod_{a \neq i,i+1} \customvarpi_{y,S}^{(a)}$
where the products are over the left endpoints $a$ shared by $y$ and $z$.
Then we have
\[ \begin{cases}
\customvarpi_{z,S} = (2+\beta x_i)\cQ
\\
\customvarpi_{y,S} = (2+\beta x_{i+1})\cQ
\end{cases}
\quand
\customvarpi_{y,T} =\begin{cases} (1+\beta x_{i+1})\cQ &\text{if }S \in \cV \\
-\cQ &\text{if } S\in \cW,
\end{cases}
\]
using Lemma~\ref{shiftable-lem} for the second identity.
As $\cQ$ does not involve $x_i$ or $x_{i+1}$,
\[ s_i(\customvarpi_{z,S}) = \customvarpi_{y,S}
\quand
\bpartial_i(\customvarpi_{z,S}) = -\beta \cQ .\]
Observe that $\iG_{z,S} = \iG_{y,T}$ for all $S \in \cX$ along with $\bpartial_i \iG_{z,S} = \iG_{y,S}$ for $S \in \cV$
and
$\bpartial_i \iG_{z,S} = -\beta\iG_{z,S}$ for $S \in \cW$.
Thus, if $S \in \cV$ then \eqref{fg-eq} can be written as
\[
\ba
\beta^{|S|}\cdot \bpartial_i(\customvarpi_{z,S} \cdot \iG_{z , S})
%&= \beta^{|S|}\cdot s_i (\customvarpi_{z,S}) \cdot \(\bpartial_i \iG_{z , S}+\beta \iG_{z , S}\) + \beta^{|S|}\cdot\bpartial_i (\customvarpi_{z,S})\cdot \iG_{z , S}
&=
\beta^{|S|}\cdot \cP_{y,S} \cdot \( \iG_{y , S}+\beta \iG_{y , T}\) - \beta^{|S|}\cdot \beta \cQ \cdot \iG_{y , T}
\\&=
\cP_{y,S} \cdot \beta^{|S|}\cdot \iG_{y , S}+((2+\beta x_{i+1}) - 1)\cdot \cQ \cdot \beta^{|T|}\cdot \iG_{y , T}
\\&
= \cP_{y,S} \cdot \beta^{|S|}\cdot \iG_{y , S}+\cP_{y,T}\cdot \beta^{|T|}\cdot \iG_{y,T}
\ea
\]
whereas if $S \in \cW$ then
\eqref{fg-eq} gives
\[
\beta^{|S|}\cdot \bpartial_i(\customvarpi_{z,S} \cdot \iG_{z , S})
=
\beta^{|S|}\cdot (-\beta \cQ) \cdot \iG_{z , S}
=
\customvarpi_{y,T}\cdot \beta^{|T|} \cdot \iG_{y,T}
\]
as needed.
\end{proof}
The identity \eqref{goal-eq} in case (d) is immediate from Claims~\ref{d-claim1}, \ref{d-claim2}, \ref{d-claim3}:
the first claim simplifies the sum on the right, and then whichever of the other two claims applies
(according to whether $i+2$ is a left endpoint) transforms the remaining terms to the sum on the left.
This completes our proof by induction.
\end{proof}
\subsection{Shift invariance and stable limits}\label{shift-sect}
In this section we consider how $\fkG_w$, $\iG_y$, and $\fkG_z$
vary when we apply natural shifting operations to the indexing permutations.
This will allow us to state some additional results about the \defn{stable limits} of these polynomials.
Define $\downshift : \ZZ[\beta][x_1,x_2,\dots]\to \ZZ[\beta][x_1,x_2,\dots]$ to be the $\ZZ[\beta]$-algebra homomorphism sending
$x_1 \mapsto 0$ and $x_{i+1} \mapsto x_i$.
%Notice that $w \in S_\infty$ can be expressed as $w =1 \times v$ for some $v \in S_\infty$ if and only if $w(1)=1$.
Recall that we view permutations $w \in S_\infty$ as maps $w : \ZZ \to \ZZ$ with $w(i)=i$ for $i\leq 0$.
Given $w \in S_\infty$ and an integer $n\geq0$, define $w\downarrow n$ to be the permutation of $\ZZ$
that sends $i \mapsto w(i+n)-n$.
Notice for $w \in S_\infty$ that $w\downarrow n\in S_\infty$ if and only if $w$ fixes each $i \in [n]$,
and in this case $\ell(w\downarrow n) = \ell(w)$.
\begin{proposition}\label{downshift-prop}
If $w \in S_\infty$ then $\downshift(\fkG_w) =
\begin{cases}
\fkG_{w\downarrow 1} &\text{if $w(1)=1$} \\
0&\text{if }w(1) \neq 1.
\end{cases}$
\end{proposition}
\begin{proof}
This identity is possibly well-known, and in any case follows from Remark~\ref{bcs-rem},
upon noting two observations.
First, if $w(1) = 1$ then all letters of all Hecke words for $w$ are at least $2$, and the operation on words
that decrements every letter by one is a bijection $\cH(w) \to \cH(w\downarrow 1)$.
Second, if $w(1) \neq 1$ then every Hecke word for $w$ must have some letter equal to $1$, so in any bounded compatible sequence $(a,i)$ for $w$ it must hold that $i_1 = a_1 = 1$.
%From here, the rest of the proof is a straightforward exercise.
%
%Fix $w \in S_\infty$ and
%recall the formula for $\fkG_w$ explained in Remark~\ref{bcs-rem}.
%If $w(1) \neq 1$ then every Hecke word for $w$ must have some letter equal to $1$, so in any bounded compatible sequence $(a,i)$ for $w$ it must hold that $i_1 = a_1 = 1$. Therefore if $w(1) \neq 1$ then $\fkG_w$ is divisible by $x_1$ and $\downshift(\fkG_w) = 0$.
%
%Assume $w(1) = 1$. Then all letters of all Hecke words for $w$ are at least $2$, and the operation on words $\decr : a \mapsto (a_1-1)(a_2-1)(a_3-1)\cdots $ that decrements every letter by one is a bijection $\cH(w) \to \cH(w\downarrow 1)$.
%It follows that
%\[\ba
%\downshift(\fkG_w) &= \sum_{\substack{(a,i) \in \BCS(w) \\ i_1>1} }\beta^{\ell(a)-\ell(w)} x^{\decr( i)}
%=
%\sum_{(a,i) \in \BCS(w\downarrow 1)} \beta^{\ell(a)-\ell(w)} x^{ i}
%=\fkG_{w\downarrow 1}\ea\]
%where $\BCS(w)$ denotes the set of bounded compatible sequences for $w$.
\end{proof}
If $z \in \Ivex$ fixes each $i \in [n]$ then the permutation $z\downarrow n$ is also in $\Ivex$.
\begin{proposition}\label{yz-down-prop}
Suppose $y \in I_\infty$ and $z \in \Ivex$. Then
\[
\downshift(\iG_y) =
\begin{cases}
\iG_{y\downarrow 1} &\text{if $y(1)=1$} \\
0&\text{if }y(1) \neq 1
\end{cases}
\quand
\downshift(\fkGO_z) =
\begin{cases}
\fkGO_{z\downarrow 1} &\text{if $z(1)=1$} \\
0&\text{if }z(1) \neq 1.
\end{cases}
\]
\end{proposition}
\begin{proof}
One can deduce from \eqref{sim-eq} that if $y(1) =1$ then
every $w \in \cB(y)$ has $w(1)=1$ and $w\mapsto w\downarrow 1$ is a length-preserving bijection $\cB(y) \to \cB(y\downarrow 1)$.
Likewise, if $y(1)\neq 1$ then
every $w \in \cB(y)$ also has $w(1)\neq 1$. The
identity for $\downshift(\iG_y) $ is immediate from Proposition~\ref{downshift-prop} given these observations.
For $S \in \shiftable(z)$ let $S\downarrow 1 = \{ s - 1: s \in S\}$.
As $\downshift$ is an algebra homomorphism,
Theorem~\ref{ivex-thm} implies that \[\textstyle
\downshift(\fkGO_z)=
\sum_{S\in\shiftable(z)} \downshift(\customvarpi_{z,S}) \cdot \beta^{|S|} \cdot \downshift(\iG_{z , S})
.\]
Suppose $z(1) = 1$. Let $L$ be the left segment of $z$ containing $2$ or set $L =\varnothing$ if $z(2)=2$.
Then our first identity implies that
$\downshift(\iG_{z , S})$ is zero when $S\cap L\neq \varnothing$
and equal to $\iG_{z\downarrow 1, S\downarrow 1}$ when $S\cap L= \varnothing$, in which case we also have $\downshift(\customvarpi_{z,S})=\customvarpi_{z\downarrow 1,S\downarrow 1}$.
As $S\mapsto S\downarrow 1$ is a bijection \[\{S \in \shiftable(z) : S\cap L= \varnothing\} \to \shiftable(z\downarrow 1),\] we conclude that
$\downshift(\fkGO_{z}) = \fkGO_{z\downarrow 1}$.
Alternatively, if $z(1) \neq 1$ then $y := (\sigma_S)^{-1}\cdot z \cdot \sigma_S$ has $y(1) \neq 1$ for all $S \in \shiftable(z)$, so
$\downshift(\iG_{z , S})$ is always zero and $\downshift(\fkGO_z)=0$.
\end{proof}
Given $w \in S_\infty$, define
$1\times w \in S_\infty$
to be the permutation that sends $1\mapsto 1$ and $i+1 \mapsto w(i)+1$ for all $i\in\PP$.
For positive integers $n$
we inductively define $1^n \times w \in S_\infty$ to be $1\times w$ if $n=1$ or $1\times (1^{n-1}\times w)$ if $n>1$.
Note that
\[
(1^n \times w) \downarrow n = w\quad\text{for all }w\in S_\infty.\]
Now define
$\upshift : \ZZ[\beta][x_1,x_2,\dots]\to \ZZ[\beta][x_1,x_2,\dots]$ to be the $\ZZ[\beta]$-linear map
sending $\fkG_w \mapsto \fkG_{1\times w}$ for $w \in S_\infty$.
By Proposition~\ref{downshift-prop}, we have
\be\label{shift-eq} \downshift \circ \upshift = 1
\quand
\upshift \circ \downshift(\fkG_w) = \begin{cases}
\fkG_w &\text{if }w(1) = 1\\
0&\text{if }w(1)\neq 1.
\end{cases}
\ee
Note, however, that $\upshift$ is not an algebra homomorphism, and thus differs from the more familiar right inverse of $\downshift$
sending $x_i \mapsto x_{i+1}$ for all $i \in \PP$.
\begin{proposition}
It holds that $\upshift(\iG_z) = \iG_{1\times z}$ for all $z \in I_\infty$.
\end{proposition}
\begin{proof}
The operation $w\mapsto 1\times w$ preserves $\ell$ and $\ellhat$
and given a bijection $\cB(z) \to \cB(1\times z)$,
so the result is clear from Definition~\ref{iG-def}.
\end{proof}
The following theorem gives a similar formula for $\upshift(\fkGO_z)$.
This is the one of the main results of this section, and is much less easy to prove.
\begin{theorem}\label{supp-thm}
Let $z \in \Ivex$.
Then $\upshift(\fkGO_z) = \fkGO_{1\times z}$ if and only if $z(1)=1$.
\end{theorem}
Before proving this theorem we consider an example and state one lemma.
\begin{example}
%The hypothesis $z(1)= 1$ in the last part of Theorem~\ref{supp-thm} is necessary. For example,
We have
$\fkGO_{(1, 2)} = 2 \fkG_{21} + \beta \fkG_{312}$
but
\[
\fkGO_{1\times (1, 2)}= \fkGO_{(2, 3)} = 2 \fkG_{132} + \beta \fkG_{1423} + \beta \fkG_{231} + \beta^2 \fkG_{2413}
\neq \upshift(\fkGO_{(1,2)}).
\]
However, it does hold that \[
\fkGO_{1\times (2, 3)}= \fkGO_{(3, 4)} = 2 \fkG_{1243} + \beta \fkG_{12534} + \beta \fkG_{1342} + \beta^2 \fkG_{13524}
= \upshift(\fkGO_{(2,3)}).\]
\end{example}
Given $v=v_1v_2v_3\cdots \in S_\infty$ with $v_j=1$, define
$ v^{\natural}=v_1^{\natural}v_2^{\natural}v_3^{\natural}\cdots \in S_\infty$ to be the permutation with
$v^{\natural}_1=2$, $v^{\natural}_{j+1}= 1$, and $v^{\natural}_{i+1} = 1+v_i$ for all $i\neq j$.
%Also define $1\times v:= 1(1+v_1)(1+v_2)\cdots \in S_\infty$
%to be the permutation that fixes $1$ and sends $i+1 \mapsto 1+v_i$ for all $i\in\PP$.
If
\[v=3467125\quad\text{then}
\quad v^{\natural} = 24578136\quand 1\times v = 14578236.\]
Notice that $ v^{\natural} = (1,2)\cdot (1\times v)= s_1\cdot (1\times v)$.
\begin{lemma}\label{supp-lem}
Fix integers $1 < i_1 < i_2 <\dots <i_l$ and
let
\[\textstyle \Pi = \prod_{j\in[l]}(1+\beta x_{i_j})
\quand
\Pi^+ = \prod_{j\in[l]}(1+\beta x_{1+i_j}) .
\]
Suppose $w \in S_\infty$ has $w(1)=1$. Let $a_{v},b_{v} \in \ZZ[\beta]$ for $v \in S_\infty$ be such that
\[\textstyle \Pi\cdot \fkG_w
= \sum_{v \in S_\infty} a_{v} \cdot \fkG_v
\quand
\Pi^+\cdot \fkG_{1\times w} = \sum_{v \in S_\infty} b_{v} \cdot \fkG_v.
\]
Then for all $v \in S_\infty$ it holds that
$b_{v^{\natural}} =\begin{cases} - a_{v} &\text{if }v(1) \neq 1 \\ 0 &\text{if }v(1)=1.\end{cases}$
\end{lemma}
We point out that
as $\fkG_w$ is homogeneous if we view $\beta$ as having degree $-1$,
the coefficients $a_v$ and $b_v $ in this lemma are always integers times $ \beta^{\ell(v)-\ell(w)}$.
\begin{proof}
Since $1\times w$ fixes $1$ and $2$ and since $\Pi^+$ does not involve
$x_1$ or $x_2$, it follows from \eqref{fg-eq} and
Theorem~\ref{iG-thm} that $\bpartial_i (\Pi^+\cdot \fkG_{1\times w}) = -\beta\cdot \Pi^+ \cdot \fkG_{1\times w}$.
In view of \eqref{Grothendieck-recursion}, this is only possible if $b_u =0$ whenever $u(1)>u(2)$.
The latter condition holds whenever $u=v^\natural$ for some $v\in S_\infty$ with $v(1)=1$.
Next, let $ i = (i_1,i_2,\dots,i_l)$ and $i^+ = (1+i_1,1+i_2,\dots,1+i_l)$.
Define an \defn{$i$-chain} to be an ascending chain in Bruhat order formed by
starting with a chain of the form \eqref{lenart-eq} with $k=i_l$,
then appending chains of the same form with $k=i_{l-1}, i_{l-2},\dots,i_1$ in succession.
We say that an $i$-chain is \defn{odd} (respectively, \defn{even}) if the sum of the values of $p$ from \eqref{lenart-eq}
corresponding to the $l$ segments of the chain is odd (respectively, even).
Finally, fix $v \in S_\infty$ with $v(1)\neq 1$.
If we use Theorem~\ref{lenart-transition} to multiply $\fkG_{1\times w}$ first by $(1+\beta x_{i_l})$,
then by $(1+\beta x_{i_{l-1}}) $, then by $(1+\beta x_{i_{l-2}})$, and so, to ultimately obtain the $\fkG$-expansion of $\Pi^+\cdot \fkG_{1\times w}$, then we see that
the coefficient $b_{v^\natural}$ is exactly the difference between the numbers of even and odd $i^+$-chains from $1\times w$ to $v^\natural$.
Similarly, the coefficient $a_v$ in the $\fkG$-expansion of $\Pi\cdot \fG_w$ is
the difference between the numbers of even and odd $i$-chains from $w$ to $v$.
The key idea of this proof rests in the following observations.
Because $1\times w$ fixes $1$ and $2$ while $v^\natural(1)=2$, in any $i^+$-chain $\gamma$ from $1\times w$ to $v^\natural$,
there must be an ascent of the form $\xrightarrow{(2,i_j)}$ (as otherwise we will have $\last(\gamma)(1)=1$) and the first such ascent must be followed
immediately an ascent of the form $\xrightarrow{(1,i_j)}$ (as otherwise $2$ will never return to the first position in the one-line representation of $\last(\gamma)$),
and then all other ascents in the chain $\xrightarrow{(a,b)}$ must have $a>1$ (as any such ascent with $a=1$ must come later in the chain, but if there is such a later ascent then we will have $\last(\gamma)(1)>2$).
In view of these properties, we deduce that removing $\xrightarrow{(1,i_j)}$
and then replacing all other ascents $\xrightarrow{(a,b)}$ by $\xrightarrow{(a-1,b-1)}$
transforms $\gamma$ to an $i$-chain from $w$ to $v$. Conversely, by considering the natural inverse operation we see that every $i$-chain from $w$ to $v$ arises in this way for a unique choice of $\gamma$.
As the bijection just given between $i^+$-chains from $1\times w$ to $v^\natural$ and $i$-chains from $w$ to $v$
reverses parity, we conclude that $b_{v^\natural} = -a_v$ as needed.
\end{proof}
We are now ready to prove Theorem~\ref{supp-thm}.
\begin{proof}[Proof of Theorem~\ref{supp-thm}]
Assume $z(1) \neq 1$ and let $L$ be the left segment of $1\times z$ containing $2$,
so that $\{i -1 : i \in L\}$ is the unique immobile left segment of $z$.
We first check that $\upshift(\fkGO_z)\neq \fkGO_{1\times z} $.
Sending $S \mapsto 1 +S$ gives a bijection from $\shiftable(z)$ to the family of shiftable sets $T \in \shiftable(1\times z)$ with $T\cap L = \varnothing$.
Since $ \customvarpi_{ z,S} = \customvarpi_{1\times z,1+S}$ and $\upshift(\iG_{z , S}) = \iG_{1\times z,1+S}$,
it follows from Theorem~\ref{ivex-thm} that
\[ \fkGO_{1\times z} - \upshift(\fkGO_z) = \sum_{\substack{S\in\shiftable(1\times z) \\ S\cap L \neq \varnothing}} \customvarpi_{1\times z,S} \cdot \beta^{|S|} \cdot \iG_{1\times z , S}.
\]
This difference is divisible by $\beta$. Moreover, first dividing by $\beta$ and then setting $\beta=0$
kills all terms in the sum indexed by sets $S$ with $|S| \neq 1$.
Consulting \eqref{varpi-eq}, we see that this operation
transforms $\fkGO_{1\times z} - \upshift(\fkGO_z)$ to the sum
\[\textstyle
\sum_{n \in L} \customvarpi_{1\times z,\{n\}}\big|_{\beta =0} \cdot \iG_{1\times z , \{n\}}\big|_{\beta =0}
=
2^{\cyc(z)-1} \sum_{n \in L} \iS_{t_n (1\times z)t_n}
\]
where $t_n = (1,n) \in S_\infty$ and $\iS_y := \iG_y\big|_{\beta=0}$ is the \defn{involution Schubert polynomial} considered in \cite{HMP1}.
This sum cannot be zero as involution Schubert polynomials are nonzero elements of $\NN[x_1,x_2,\dots]$,
so $\fkGO_{1\times z} - \upshift(\fkGO_z) \neq 0$.
Now assume instead that $z(1)=1$. Then the formula for $\fkGO_{1\times z}$ in Theorem~\ref{ivex-thm}
expands as a sum of terms of the form $ \Pi\cdot \fkG_w$ as in Lemma~\ref{supp-lem},
and the corresponding formula for $\fkGO_{1^2\times z}$
is obtained by replacing each term $ \Pi\cdot \fkG_w$ by $\Pi^+\cdot \fkG_{1\times w}$.
It therefore follows from Lemma~\ref{supp-lem} that if $v \in S_\infty$ has $v(1)\neq 1$
then $\GC_{1^2\times z}(v^\natural) = -\GC_{1\times z}(v)$.
As $\GC_{1\times z}$ and $\GC_{1^2\times z}$ both only take nonnegative values \cite[Prop.~3.29]{MS2023},
we must have $\GC_{1\times z}(v) = 0$ if $v(1)\neq 1$.
We conclude from Proposition~\ref{yz-down-prop} and \eqref{shift-eq} that
\[\fkGO_{1\times z} = \upshift\circ\downshift(\fkGO_{1\times z}) = \upshift(\fkGO_z)\]
as claimed.
\end{proof}
Using Theorem~\ref{supp-thm}, we can now prove another nontrivial result.
Comparing this with Proposition~\ref{supp-prop}
gives some heuristic support for Conjecture~\ref{b+conj}.
\begin{proposition}
Suppose $z \in \Ivex$ has $\supp(z) \subseteq [a,b]$ for some integers $a\leq b$. Then $\supp(w) \subseteq [a-1,b+1]$
for all $w \in \supp(\GC_z)$.
\end{proposition}
\begin{proof}
%To prove part (b),
%we drop the assumption that $z(1)=1$ but suppose $\supp(z) \subseteq [a,b]$
%for integers $a\leq b$.
Define $p$ and $q$ as in \eqref{pq-eq}, so that $q \leq b$, and let $\delta = \dom_{pq}$ be the dominant involution defined in \eqref{dom-def}.
It follows from Corollary~\ref{qd-supp-cor} and Proposition~\ref{supp-prop}
that $\supp(w) \subseteq [1,q+1]$ for all $w \in \supp(\GC_{\delta})$.
Since $\fkGO_z$ can be obtained from $\fkGO_{\delta}$ by applying a sequence of divided difference
operators $\bpartial_i$ with $i <q$,
we also have $\supp(w) \subseteq[1,q+1]\subseteq [1,b+1]$ for all $w \in \supp(\GC_z)$.
It remains to check that $\supp(w)\subseteq [a-1,\infty)$ for all $w \in \supp(\GC_z)$.
This is trivial unless $a\geq 3$,
and then the desired containment follows from
the fact that
$\fkGO_z = \upshift^{a-2}(\fkGO_y)$
for $y := z\downarrow (a-2) \in \Ivex$
by Theorem~\ref{supp-thm}.
\end{proof}
For $z \in I_\infty$ we
extend the domain of the Grothendieck coefficient function $\GC_z$ by setting $
\GC_z(w) = 0$ if $w \notin S_\infty$.
The following identities
are immediate from
Proposition~\ref{yz-down-prop} and
Theorem~\ref{supp-thm}.
\begin{corollary}\label{shift-cor}
Fix $n \in \NN$ and $z \in \Ivex$.
\ben
\item[(a)] If $z \downarrow n \in S_\infty$ then $\GC_{z\downarrow n }(w) = \GC_z(1^n \times w)$ for all $w \in S_\infty$.
\item[(b)] If $z(1)=1$ then $\GC_{1^n\times z }(w) = \GC_z(w \downarrow n)$ for all $w \in S_\infty$.
\een
\end{corollary}
%\begin{proof}
%Part (a) follows from Theorem~\ref{supp-thm}
%while part (b) follows from Proposition~\ref{yz-down-prop}.
%\end{proof}
The properties in this corollary are consistent with Conjecture~\ref{b+conj} since
it is easy to show that
if $n \in \NN$ and $z \in \Ivex$ has $z(1)=1$ then
\be
\label{BB-eq}
\ba
\cB(1^n\times z ) &= \left\{ w \in S_\infty : w \downarrow n \in \cB(z)\right\},
\\
\cB^+(1^n \times z ) &= \left\{ w \in S_\infty : w \downarrow n \in \cB^+(z)\right\}.
\ea
\ee
In fact, the identity for $\cB(1^n\times z ) $ holds even if $z(1) \neq 1$.
Theorem~\ref{supp-thm} has an application concerning the \defn{stable limit} of $\fkGO_z$.
The \defn{symmetric Grothendieck function} of $w \in S_\infty$
is
\be\textstyle G_w := \lim_{n\to \infty} \fkG_{1^n\times w}\ee where the limit is taken in the sense of formal power series. This means the limit exists precisely when the coefficients of any fixed monomial
in $\fkG_{1^n\times w}$ is eventually a constant sequence.
It is known \cite{Buch2002,BuchEtAl,FominKirillov} that $G_w$ always exists and is a formal power series that is symmetric in the $x_i$ variables. Moreover, every symmetric power series over $\ZZ[\beta]$ is given by a unique, possibly infinite $\ZZ[\beta]$-linear combination
of $G_w$'s \cite[Thm.~6.12]{Buch2002}.
Consulting Remark~\ref{bcs-rem}, one obtains the explicit generating function
$G_w = \sum_{(a,i)} \beta^{\ell(a)-\ell(w)} x^i$
where the sum is over all (not necessarily bounded) compatible sequences $(a,i)$ with $a \in \cH(w)$.
Following \cite{MP2020}, for $z \in \Ivex$ we define an analogous power series limit
\be\label{gq-eq}\textstyle \GQ_z := \lim_{n\to \infty} \fkGO_{1^n\times z}.\ee
We can immediately see that this limit converges to a symmetric function:
\begin{corollary}\label{supp-cor} If $z \in \Ivex$ has $z(1)=1$ then
\[\textstyle \GQ_z = \sum_{w \in S_\infty} \GC_{z}(w) \cdot \beta^{\ell(w) - \ellhat(z)} \cdot G_w.\]
Moreover, it holds for any $z \in \Ivex$ that $\GQ_z = \GQ_{1\times z}$.
\end{corollary}
%This formula shows that $\GQ_z$ exists for any $z \in \Ivex$
%since
% $\GQ_z = \GQ_{1\times z}$.
\begin{proof}
If $z(1)=1$ then
$\fkGO_{1^n\times z} = \sum_{w \in S_\infty} \GC_{z}(w) \cdot \beta^{\ell(w) - \ellhat(z)} \cdot \fkG_{1^n\times w}$ holds by Theorem~\ref{supp-thm}
and the corollary follows by taking the limit as $n\to \infty$.
\end{proof}
\begin{remark}
Comparing Corollary~\ref{supp-cor} with \cite[Lem.~3.7 and Prop.~3.29]{MS2023} shows that the conjecture \cite[Conj.~3.39]{MS2023} holds when $z$ is vexillary.
\end{remark}
A stronger result about $\GQ_z$ is known: if $z \in \Ivex$ then $\GQ_z$ is equal to the \defn{$K$-theoretic Schur $Q$-function} $\GQ_\mu$ defined in \cite{Ikeda.Naruse2013} when $\mu=\sh(z)$ is the \defn{involution shape} of $z$ \cite[Thm.~4.11]{MP2020}.
The involution shape $\sh(z)$ is the transpose of the partition that sorts the sequence of numbers
$\hat c(z)$ %(\hat c_1, \hat c_2, \hat c_3,\dots)$
%with $\hat c_i = |\{ j : (i,j) \in D(z)\text{ if }i\geq j\}|$ as in
from \eqref{hat-d-eq}.
One can view Corollary~\ref{supp-cor} as a new $G$-positive expansion of $\GQ_\mu$.
% The identity
% $\GQ_z = \GQ_{\sh(z)}$ was proved in \cite{MP2020}
% using results in \cite{Anderson,NakagawaNaruse}.
% Theorem~\ref{supp-thm} leads to a simpler derivation of the fact that $\GQ_z$ is a symmetric formal power series. We also get a
% new formula relating the $G$-expansion of $\GQ_z$ to the $\fkG$-expansion of $\fkGO_z$.
%
It is an open problem to show that
\eqref{gq-eq} converges when $z \in I_\infty$ is not vexillary.
By contrast, it is clear from \eqref{BB-eq} that for any $z \in I_\infty$ the limit
\be\label{gp-eq1}
\textstyle \GP_z := \lim_{n\to \infty} \iG_{1^n\times z }
\ee
converges in the sense of formal power series to the symmetric function
\be\label{gp-eq2}
\textstyle \GP_z =\sum_{w \in \cB(z)} \beta^{\ell(w)-\ellhat(z)} G_w.
\ee
In fact, it is known \cite[Thm.~1.9]{Mar2019} that the power series $\GP_z$ are all $\NN[\beta]$-linear combinations of the
\defn{$K$-theoretic Schur $P$-functions} $\GP_\mu$ defined in \cite{Ikeda.Naruse2013}.
We can derive one other formula for $\GQ_z$ using this notation and Theorem~\ref{ivex-thm}.
Given $z \in \Ivex$
and $S \in \shiftable(z)$,
define $\GP_{z , S} := \GP_{y}$ where $y:=(\sigma_S)^{-1}\cdot z \cdot \sigma_S$
and form $ \Theta_{z,S} \in \ZZ$ from $ \customvarpi_{z,S}$ by setting $x_1=x_2=\dots=0$.
More explicitly, if $L= \{ a\in \PP : a<z(a)\}$ then
\[\Theta_{z,S} = (-1)^{|\{ a \in S \hs : \hs \crb_z(a) \cap S \neq \varnothing\}|} 2^{|L|-| S|} = \pm 2^{\cyc(z)-| S|}. \]
One can show that $\Theta_{z,S}>0$ for all $S \in \shiftable(z)$ if and only if $z\in \Ivex$
is locally noncrossing in the sense defined in Remark~\ref{locally-noncrossing-remark}
\begin{corollary}\label{ivex-cor}
If $z \in \Ivex$ has $z(1)=1$ then
\[\textstyle\GQ_z = \sum_{S\in\shiftable(z)} \Theta_{z,S} \cdot \beta^{|S|} \cdot \GP_{z , S}.\]
\end{corollary}
\begin{proof}
Form $\customvarpi_{z,S}^{\uparrow n}$ from $\customvarpi_{z,S}$ by applying the $\ZZ[\beta]$-algebra homomorphism that sends $x_i \mapsto x_{i+n}$ for all $i \in \PP$.
Theorem~\ref{ivex-thm} implies that if $z(1)=1$ then
$\fkGO_{1^n\times z} = \sum_{S\in\shiftable(z)} \customvarpi_{z,S}^{\uparrow n} \cdot \beta^{|S|} \cdot \iG_{1^n\times (\sigma_S)^{-1}\cdot z \cdot \sigma_S}$.
The result follows by taking the limit $n\to \infty$, which turns $ \customvarpi_{z,S}^{\uparrow n}$ into $ \Theta_{z,S}$ and $\iG_{1^n\times (\sigma_S)^{-1}\cdot z \cdot \sigma_S}$ into $\GP_{z , S}$.
\end{proof}
\subsection{Grassmannian identities}\label{gr-sect}
There is an interesting special case when the formulas for $\fkGO_z$ and $\GQ_z$ in the previous sections become considerably more explicit. We explain this below after defining some additional notation.
Recall the definition of the visible descent set $\DesV(z)$ from \eqref{vdes-def}.
Following \cite{HMP3}, we say $z \in I_\infty$ is \defn{$n$-I-Grassmannian}
%if
%its \defn{visible descent set}
%\be\DesV(z) := \{i \in \PP : z(i) > z(i+1) \leq i\}\ee satisfies
$\DesV(z)\subseteq \{n\}$.
An element of $I_\infty$ is \defn{I-Grassmannian} if it is $n$-I-Grassmannian for some $n \in \NN$.
For a \defn{strict} partition $\mu=(\mu_1>\mu_2>\dots>\mu_r>0)$ with $r>0$ nonzero parts and largest part $\mu_1\leq n$, define $\langle \mu \mid n\rangle \in I_\infty$
to be the permutation %given in cycle notation as
\be (n+1-\mu_1, n+1)(n+1-\mu_2,n+2)\cdots(n+1-\mu_r,n+r) \in I_\infty.\ee
When $\mu=\emptyset$ is empty let $\langle \mu \mid n\rangle =1$ for all $n \in \PP$.
This notation deliberately mimics our convention for writing $n$-Grassmannian permutations $[\lambda\mid n]\in S_\infty$, for reasons that shall soon be apparent.
For each strict partition $\mu$ with largest part $\mu_1\leq n$,
the involution $\langle \mu \mid n \rangle$ turns out \cite[\S4.1]{HMP3} to be the unique
$n$-I-Grassmannian element of $I_\infty$ with involution shape $\sh(\langle \mu \mid n \rangle)=\mu$.
From this observation and Remark~\ref{arc-vex-rem},
we deduce that every $I$-Grassmannian involution is vexillary.
Thus, we can apply
Theorem~\ref{ivex-thm} when $z$ is $I$-Grassmannian.
%and in this case our general formula for $\fkGO_z$ becomes considerably more explicit. We explain this below after defining some additional notation.
Let $\SD_\mu := \{ (i,i+j-1) : (i,j) \in \D_\mu\}$ denote the \defn{shifted Young diagram}
of a strict partition $\mu$. Write $\mu\subseteq\lambda$ if $\mu$ and $\lambda$ are strict partitions
with $\SD_\mu\subseteq \SD_\lambda$.
When this containment holds we define
\[
\SD_{\lambda/\mu} := \SD_\lambda\setminus \SD_\mu
\quand |\lambda/\mu| := |\lambda|-|\mu|
\quad\text{where $|\mu| := |\SD_\mu| $.}\]
The shifted skew shape $\SD_{\lambda/\mu}$ is a \defn{vertical strip} if it contains at most one position in each row,
or equivalently if $\lambda_i-\mu_i \in \{0,1\}$ for all $i \in \PP$.
Assume this holds and let $r$ be the number of nonzero parts of $\lambda$.
Define $\topRows(\lambda/\mu)$ to be the set of indices $i \in [r]$
such that row $i$ of $\SD_{\lambda/\mu}$ is empty or
such that $(i+1,j) \notin \SD_{\lambda/\mu}$ for the unique column $j$ with $(i,j) \in \SD_{\lambda/\mu}$.
Then let
\be
\textstyle
\customvarpi_{\lambda/\mu}^{(n)} :=
\prod_{i \in\topRows(\lambda/\mu) } (2 +\mu_i- \lambda_i +\beta x_{n+1-\mu_i}).
\ee
Also write $\cols(\lambda/\mu) = |\{ j : (i,j) \in \SD_{\lambda/\mu}\text{ for some }i \in \PP\}|$
for the number of distinct columns occupied by the positions in $\SD_{\lambda/\mu}$.
\begin{example}
If $\mu =(8,7,4,3,1)$ and $\lambda = (9,7,5,4,2)$ then
\[
\ytableausetup{boxsize=0.3cm,aligntableaux=center}
\SD_{\lambda/\mu} = \begin{ytableau}
\none[\cdot] & \none[\cdot] & \none[\cdot] & \none[\cdot] & \none[\cdot] & \none[\cdot] & \none[\cdot] & \none[\cdot] & \ \\
\none & \none[\cdot] & \none[\cdot] & \none[\cdot] & \none[\cdot] & \none[\cdot] & \none[\cdot] & \none[\cdot] \\
\none & \none & \none[\cdot] & \none[\cdot] & \none[\cdot] & \none[\cdot] & \ \\
\none & \none & \none & \none[\cdot] & \none[\cdot] & \none[\cdot] & \ \\
\none & \none & \none & \none & \none[\cdot] & \
\end{ytableau}
\]
so $|\lambda/\mu| = 4$, $\cols(\lambda/\mu) =3$, $\topRows(\lambda/\mu) = \{1,2,4,5\}$, and
\[
\customvarpi_{\lambda/\mu}^{(9)} = (1+ \beta x_{2})(2+\beta x_{3}) (1+\beta x_{7})(1+\beta x_9).
\]
\end{example}
\begin{corollary}\label{igrass-cor}
If $\mu$ is a strict partition with largest part at most $n$
then
\[
\fkGO_{\langle \mu \mid n \rangle} =
\sum_\lambda (-1)^{\cols(\lambda/\mu)} \cdot \customvarpi_{\lambda/\mu}^{(n)} \cdot (-\beta)^{|\lambda/\mu|} \cdot \iG_{\langle \lambda \mid n\rangle}
\]
where the sum is over strict partitions $\lambda$ with the same number of parts as $\mu$
and with largest part at most $n$, such that $\lambda \supseteq\mu$ and $\SD_{\lambda/\mu}$ is a vertical strip.
\end{corollary}
\begin{proof}
Let $z = \langle\mu \mid n \rangle$. Then
as $S$ ranges over the elements of $\shiftable(z)$,
the permutations $ (\sigma_S)^{-1}\cdot z\cdot \sigma_S$
range over exactly the involutions $\langle \lambda\mid n\rangle$ for $\lambda$
as described in the corollary.
Additionally, when going across the correspondence $S\leftrightarrow \langle \lambda\mid n\rangle$,
one has $|S| = |\lambda/\mu|$,
while the number of factors $\customvarpi_{z,S}^{(a)}$ equal to $-1$ is $|\lambda/\mu| - \cols(\lambda/\mu)$
and the product of the other factors $\customvarpi_{z,S}^{(a)}$ is exactly $\customvarpi_{\lambda/\mu}^{(n)}$.
Verifying these claims is a straightforward exercise, and then the desired formula
for $\fkGO_{\langle \mu \mid n\rangle}$
is immediate from Theorem~\ref{ivex-thm}.
\end{proof}
\begin{example}
If we fix $n=4$ and write $\langle \mu_1\mu_2\cdots\mu_r \rangle$ instead of $\langle \mu\mid 4\rangle $ then
\[
\fkGO_{\langle 32 \rangle} =
(2+\beta x_{2})(2+\beta x_{3})\iG_{\langle 32 \rangle}
+
(1+\beta x_{2})(2+\beta x_{3}) \beta \iG_{\langle 42 \rangle}
-
(1+\beta x_{3}) \beta^2 \iG_{\langle 43 \rangle}.\]
\end{example}
Corollary~\ref{igrass-cor} is a non-symmetric
generalization of \cite[Thm.~1.1]{Chiu.Marberg2023},
which one recovers by taking stable limits in the following way.
Let $\mu$ be a strict partition with $\mu_1\leq n$.
Observe that $1^m\times \langle \mu \mid n \rangle = \langle \mu \mid m+n\rangle$ for all $m \in \NN$.
Therefore
\be
\fkGO_{1^m\times \langle \mu \mid n \rangle} =
\sum_\lambda (-1)^{\cols(\lambda/\mu)} \cdot \customvarpi_{\lambda/\mu}^{(m+n)} \cdot (-\beta)^{|\lambda/\mu|} \cdot \iG_{1^m\times\langle \lambda \mid n\rangle }
\ee
where the sum is over strict partitions $\lambda$ as in Corollary~\ref{igrass-cor}.
We have $\lim_{m\to \infty} \customvarpi_{\lambda/\mu}^{(m+n)} = 2^{|\{i \in [r] : \lambda_i=\mu_i\}|} = 2^{\ell(\mu) - |\lambda/\mu|}$ with convergence in the sense of formal power series,
where $\ell(\mu)$ denotes the number of nonzero parts of $\mu$.
Hence, taking the limit as $m\to \infty$ of both sides above gives
\be\label{almost-eq}
\GQ_{\langle \mu \mid n \rangle} =
\sum_\lambda (-1)^{\cols(\lambda/\mu)} \cdot 2^{\ell(\mu)-|\lambda/\mu|}\cdot (-\beta)^{|\lambda/\mu|} \cdot \GP_{\langle \lambda \mid n\rangle}
\ee
where the sum is over strict partitions $\lambda$ as in Corollary~\ref{igrass-cor}.
This identity is nearly the same as \cite[Thm.~1.1]{Chiu.Marberg2023},
which asserts the same equation but with ``$\GQ_{\langle \mu \mid n \rangle}$''
and ``$\GP_{\langle \lambda \mid n \rangle}$''
replaced by
``$\GQ_{\mu}$''
and ``$\GP_{ \lambda}$''.
The subtle difference between these statements, which is obscured by our notation,
is that $\GQ_{\langle \mu \mid n \rangle}$
and $\GP_{\langle \lambda \mid n \rangle}$
are instances of
the stable limits \eqref{gq-eq} and \eqref{gp-eq1},
while $\GQ_\lambda$ and $\GP_\lambda$ are the symmetric functions defined in a quite different way in \cite{Ikeda.Naruse2013}.
However, as mentioned earlier, it is actually already known \cite[Thm.~4.11]{MP2020} that $\GQ_{\langle \mu \mid n \rangle} = \GQ_{\sh(\langle \mu \mid n \rangle)} = \GQ_\mu$.
To complete the circle, it would suffice to have the analogous identity $
\GP_{\langle \mu \mid n \rangle} = \GP_\mu
$. This does not seem to have appeared in the literature,
but we can prove it here:
\begin{proposition}
If $\mu$ is a strict partition and $n\geq \mu_1$
then
$
\GP_{\langle \mu \mid n \rangle} = \GP_\mu
$.
\end{proposition}
\begin{proof}
Our argument is very similar to the proofs of \cite[Prop.~4.14]{MP2021Sp}
and
\cite[Prop.~3.37]{MS2023}.
Suppose $\mu$ has $r$ nonzero parts, so that
$\langle \mu \mid n \rangle $ is the involution $ (\phi_1 , n+1)(\phi_2 , n+2) \cdots (\phi_r , n+r)$ where $\phi_i = n+1-\mu_i$.
Then
\[
\textstyle
x^{\mu} \prod_{i=1}^{r} \prod_{j=i+1}^n \tfrac{x_i\oplus x_j}{x_i} = x_1^{1-\phi_1}x_2^{2-\phi_2}\cdots x_r^{r-\phi_r} \iG_w
\]
for the dominant permutation $w := (1 , n+1)(2 , n+2)\cdots (r , n+r)$,
since by Theorem~\ref{iG-thm2} it holds that $\iG_w = \prod_{i=1}^r x_i\cdot \prod_{i=1}^r \prod_{j=i+1}^n x_i\oplus x_j$.
Define $\bpi_i f := \bpartial(x_i f)$ for $i \in \PP$; see \cite[\S4.2]{MP2021Sp} for the basic properties of these
\defn{isobaric divided difference operators}.
For integers $0 <a \leq b$ let \be\bpi_{b \searrow a} := \bpi_{b-1}\bpi_{b-2}\cdots \bpi_a\ee
so that $\bpi_{a\searrow a} = 1$, and define $\bpartial_{b\searrow a}$ analogously.
Exactly as in the proofs of \cite[Prop.~4.14]{MP2021Sp}
or
\cite[Prop.~3.37]{MS2023}, one can check by an easy calculation that
\[ \bpi_{\phi_j \searrow j} \bpi_{\phi_{j+1}\searrow j+1} \cdots \bpi_{\phi_r \searrow r} \( x_1^{1-\phi_1}x_2^{2-\phi_2}\cdots x_r^{r-\phi_r} \iG_w\)\] is fixed by $\bpi_i$ for all $j -1 < i < \phi_{j-1} < \phi_j$,
and so is equal to
\[
\bpartial_{\phi_1 \searrow 1} \bpartial_{\phi_{2}\searrow 2} \cdots \bpartial_{\phi_r \searrow r} \iG_w = \iG_{\langle\mu\mid n\rangle}
\]
by \cite[Lem.~4.16]{MP2021Sp} and Theorem~\ref{iG-thm}.
Let
$\mathsf{stab}_n = \bpi_{2\searrow 1} \bpi_{3\searrow 1} \bpi_{4\searrow 1}\cdots \bpi_{n\searrow 1}.$
Since every $w \in \cB(z)$ for $z \in I_\infty$ has $\DesR(w)\subseteq \DesR(z)$ by \eqref{iwi-eq},
it follows from \cite[Lem.~3.7]{MS2023}
that
\[ G_w(x_1,x_2,\dots,x_n) =\mathsf{stab}_n( \fkG_w)\quad\text{for all }w \in \cB(\langle \mu \mid n\rangle).\]
Next, it is known \cite[Eq.~(3.17)]{MS2023} that
\[\GP_\mu(x_1,x_2,\dots,x_n) = \mathsf{stab}_n \( \textstyle x^{\mu} \prod_{i=1}^{r} \prod_{j=i+1}^n \tfrac{x_i\oplus x_j}{x_i} \).\]
Finally, one has $\mathsf{stab}_n \pi_i = \mathsf{stab}_n$ if $i \in [n-1]$ since the $\pi_i$ operators are idempotent and
obey the braid relations for $S_\infty$ \cite[\S4.2]{MP2021Sp}.
Combining these observations with \eqref{gp-eq2} shows that
\[\GP_\mu(x_1,x_2,\dots,x_n) = \mathsf{stab}_n (\iG_{\langle \mu\mid n\rangle})=\GP_{\langle \mu\mid n\rangle}(x_1,x_2,\dots,x_n).\]
If $N\geq n$ then $\GP_{\langle \mu \mid n \rangle} =\GP_{1^{N-n}\times \langle \mu \mid n \rangle} = \GP_{\langle \mu \mid N \rangle} $ so \[\GP_\mu(x_1,x_2,\dots,x_N) = \GP_{\langle \mu \mid n\rangle}(x_1,x_2,\dots,x_N ).\]
On taking the limit $N \to\infty$ we conclude that $\GP_\mu = \GP_{\langle \mu \mid n\rangle}$.
\end{proof}
There is one case where we can turn Corollary~\ref{igrass-cor} into an explicit positive formula.
Define $ g_{ij} = (i, j+1)(i+1,j+2)(i+3,j+4)\cdots(j,2j-i+1)$
for any integers $0<i<j$. This coincides with the $I$-Grassmannian involution
$\langle \mu \mid j\rangle$ for $\mu = (m,m-1,m-2,\dots,1)$ and $m=j-i+1$.
By \cite[Thm.~3.35 and Cor.~3.36]{HMP1},
the involutions $g_{ij}$ are the only elements of $\Ivex$ that are $321$-avoiding,
that is, both vexillary and \defn{fully commutative}.
Notice that $1^n \times g_{ij} = g_{i+n,j+n}$.
Thus, to compute $\cB(g_{ij})$, $\cB^+(g_{ij})$, and $\fkGO_{g_{ij}}$ for arbitrary integers $0<i<j$, it suffices by the results in Section~\ref{shift-sect}
to consider just the case when $i=2$.
For this purpose, given any $w\in S_\infty$ recall that $\DesL(w) = \DesR(w^{-1})$ and let $\ODes_L(w) := \{ i \in \DesL(w) : i\text{ is odd}\}$.
\begin{proposition}\label{g-prop}
Assume $n>2$ and define $z \in \Ivex$ to be the involution
\[z= g_{2n} =\langle(n-1,\dots,3,2,1)\mid n\rangle= (2,n+1)(3,n+2)\cdots(n,2n-1).\] Then the following properties hold:
\ben
\item[(a)] $\cB(z)$ contains only the inverse of
$1(n+1)2(n+2)3\cdots (2n-1)n(2n).$
\item[(b)] $\cB^+(z)$ consists of the $2^{n}$ permutations whose inverses in one-line notation
have the form $a_1b_1a_2b_2\cdots a_{n} b_{n} $
where $\{a_i,b_i\}=\{i,n+i\}$ for each $i\in [n]$.
\item[(c)] If $u_{\max} \in \cB^+(z)$ is the inverse of $(n+1)1(n+2)2\cdots (2n)n$
then
\[\textstyle\fkGO_{z} = \sum_{w \in \cB^+(z)} 2^{n-1-|\ODes_L(w)|} \beta^{|\ODes_L(w)|} \fkG_w +
\tfrac{1}{2} (-\beta)^n \fkG_{u_{\max}}
\]
and consequently
\[\GC_z(w) = \begin{cases} 2^{n-1-|\ODes_L(w)|}&\text{if }u_{\max} \neq w \in \cB^+(z) \\
1 &\text{if $u_{\max}=w$ and $n$ is even} \\
0 &\text{if $u_{\max}=w$ and $n$ is odd, or $w \notin \cB^+(z)$}.
\end{cases}\]
\een
\end{proposition}
See Figure~\ref{g2n-fig} for an illustration of this proposition.
\begin{proof}
Part (a) is immediate from \eqref{sim-eq} and part (b) follows as an easy exercise from the definitions. We omit the details.
We may prove part (c) in the following way.
First, it is straightforward using Remark~\ref{arc-vex-rem} to check that $s_izs_i$ and $z$ are vexillary for $i\in [n-1]$.
%
% We claim that $s_izs_i$ and $z$ are vexillary for $i\in [n-1]$.
% Observe that the arc diagram of $z=g_{2,n}$ satisfies the following properties:
% \begin{enumerate}[label=(\arabic{*})]
% \item there is no isolated vertex under arcs,
% \item no left endpoint is greater than a right endpoint, and
% \item any pair of arcs intersects.
% \end{enumerate}
% The arc diagram contains at least one of the involution patterns in Remark~\ref{arc-vex-rem} only if condition (2) fails or conditions (1) and (3) fail at the same time.
% Because the arc diagram of $s_izs_i$ is obtained form the arc diagram of $z$ by swapping the left endpoint of arcs involving with vertices $i$ and $i+1$,
% (2) never fails if $i\in[n-1]$. Moreover, (1) fails only if $i=1$ and (3) fails only if $1<i\leq n-1 $. Thus $s_izs_i$ is vexillary for any $i\in [i-1]$.
%
Because $\Des_R(z) = \Des_V(z) =\{n\}$, we have $\fkGO_z\in \ZZ[\beta][x_1,x_2,\cdots,x_n]$
by \eqref{vdes-eq}
%by \cite[Prop.~3.29]{MS2023}
and $\bpartial_i(\fkGO_z) =-\beta \fkGO_z$ for all $i\in [n-1]$ by \eqref{orthogonal-recursion}.
The latter identity implies that $\fkGO_z$ is symmetric in $x_1,x_2,\cdots,x_n$.
By Remark~\ref{grass-sym-rem},
we therefore have
\[\textstyle
\fkGO_z= \sum_{\mu} \GC_z([\mu\mid n]) \cdot \beta^{|\mu| - \ellhat(z)} \cdot \fkG_{[\mu\mid n]}
\]
where the sum is over partitions $\mu$ with at most $n$ nonzero parts.
On the other hand,
as $z\in \Ivex$ has $z(1)=1$, Corollary~\ref{supp-cor} implies that
\[\textstyle\GQ_z=\sum_{\mu} \GC_z([\mu\mid n]) \cdot \beta^{|\mu| - \ellhat(z)} \cdot G_{[\mu\mid n]}.\]
Since
the polynomial
$G_{[\mu\mid n]}(x_1,x_2,\cdots,x_n)$ obtained from $G_{[\mu\mid n]}$
by setting the variables $x_{n+1}=x_{n+2}=\dots=0$ is equal to
$ \fkG_{[\mu| n]}$ for all partitions $\mu$ with at most $n$ parts (compare
\cite[Thm.~6.12]{Buch2002} and \cite[Thm.~2.2]{lenart2000}), we have
\be\label{GQ-suff-eq}
\textstyle
\GQ_z(x_1,x_2,\cdots,x_n) = \sum_{\mu} \GC_z([\mu\mid n]) \cdot \beta^{|\mu| - \ellhat(z)} \cdot \fkG_{[\mu\mid n]}.
\ee Hence,
it suffices to compute
the $\fkG$-expansion of $\GQ_z(x_1,x_2,\cdots,x_n)$.
We can accomplish this by assembling some miscellaneous identities from \cite{Chiu.Marberg2023,Ikeda.Naruse2013,LM2019,MP2020}.
First, \cite[Thm.~4.11]{MP2020} tells us that $\GQ_z$ is equal to the $K$-theoretic Schur $Q$-function indexed by $\rho_{n-1}=(n-1,n-2,\dots,3,2,1)$, which can be expanded in terms of $K$-theoretic Schur $P$-functions via \cite[Thm.~1.1]{Chiu.Marberg2023} as
\[
\textstyle
\GQ_z= 2^{n-1}\GP_{\rho_{n-1}}-2^{n-1}\sum_{j=1}^{n-1}(-\beta/2)^j \GP_{\rho_{n-1}+1^j}.
\]
Next, it follows from \cite[Prop.~2.3]{Ikeda.Naruse2013} that
\[
\GP_{\rho_{n-1}+1^j}(x_1,x_2,\cdots,x_n) = \GP_{\rho_{n-1}}(x_1,x_2,\cdots,x_n)G_{1^j}(x_1,x_2,\cdots,x_n).
\]
Since by \cite[Thm.~6.8]{LM2019} we have $\GP_{\rho_{n-1}} = G_{\rho_{n-1}}$,
we can rewrite this as \[
\GP_{\rho_{n-1}+1^j}(x_1,x_2,\cdots,x_n)
= \fkG_{[\rho_{n-1}| n]}\fkG_{[1^j\mid n]}.
\]
%In \cite[Prop.~2.3]{Ikeda.Naruse2013}, the formula uses factorial grothendieck polynomials $\GP_{\rho_{n-1}+(1)^j}(x_1,\cdots,x_n) = \GP_{\rho_{n-1}}(x_1,\cdots,x_n)\fkG_{[(1)^j\mid n]}(x_1,\cdots,x_n|b)$ and $\fkG_{[(1)^j\mid n]}(x_1,\cdots,x_n|b) =\fkG_{[(1)^j\mid n]}(x_1,\cdots,x_n)$ if we specialize $b=0$.
If $I$ is a subset of $[n]$ then we let $\mu_{I}= \rho_{n-1}+\sum_{i\in I} \e_{i}$.
Theorem~\ref{lenart-pieri} says that
\[\textstyle
\fkG_{[\rho_{n-1}| n]}\fkG_{[1^j\mid n]}=
\sum_{p=j}^n \sum_{|I|=p} \tbinom{p-1}{p-j}
\cdot \beta^{p-j}
\cdot \fkG_{[\mu_{I}\mid n]}.
\]
%Combining these facts, we conclude from \eqref{GQ-suff-eq} that
%$\supp(\GC_z)$ contains only the permutations $[\mu_{I}\mid n]$ for $I\subseteq [n]$.
%
%Now,
Thus,
if
$|I|=p< n$ then
the coefficient of $\fkG_{[\mu_{I}\mid n]}$ in $ \GQ_z(x_1,x_2,\dots,x_n)$
is
\[
- 2^{n-1}\sum_{j=1}^p (-\tfrac{\beta}{2})^{j} \beta^{p-j} \tbinom{p-1}{p-j}=2^{n-2}\beta^p\underbrace{\textstyle\sum_{j=0}^{p-1} (-\tfrac{1}{2})^{-j} \tbinom{p-1}{p-1-j}}_{=(-\frac{1}{2} + 1)^{p-1} = 2^{1-p}} = 2^{n-1-|I|}\beta^{|I|}.
\]
If $I=[n]$ then $\mu_{I} = \rho_n$ and %$\GC_z([\rho_n\mid n])$
the coefficient of $\fkG_{[\rho_n\mid n]}$ in $ \GQ_z(x_1,x_2,\dots,x_n)$
is
\[
- 2^{n-1}\sum_{j=1}^{n-1} (-\tfrac{\beta}{2})^{j} \beta^{n-j} \tbinom{n-1}{n-j}=2^{n-2} \beta^n\big(
2^{1-n} -(-2)^{1-n}
\big)= \begin{cases}
\beta^n &\text{ $n$ is even}\\
0 &\text{ $n$ is odd.}\\
\end{cases}
\]
The desired formula for $\GC_z$ follows from these computations via \eqref{GQ-suff-eq}
since the elements of $\cB^+(z)$
described in part (b) are precisely the Grassmannian permutations $[\mu_{I}\mid n]$ for $I \subseteq[n]$, and it holds that
$u_{\max}=[\rho_n\mid n]$ along with
\[\ell([\mu_I \mid n]) - \ellhat(z) = |\mu_I| - |\rho_{n-1}| = |I| = \ODes([\mu_{I}\mid n]).\]
This completes our proof of the proposition.
\end{proof}
Using Corollary~\ref{shift-cor}, we see
from this proposition that
Conjecture~\ref{b+conj} holds if $z=g_{ij}$ for any integers $0<i<j$.
In fact, we have $\supp(\GC_{g_{ij}})= \cB^+(g_{ij})$
unless $i>1$ and $j-i$ is odd, in which case $ \cB^+(g_{ij})$ contains one extra element.
%\begin{corollary}
%If $z=g_{ij}$ then $\cB(z)\subseteq \supp(\GC_z)\subseteq \cB^+(z)$ with
%\[|\cB^+(z)\setminus \supp(\GC_z)| = \begin{cases}
%1&\text{if $i>1$ and $j-i$ is odd} \\
%0&\text{otherwise}.
%\end{cases}\]
%\end{corollary}
\begin{figure}[h]
% \centerline{
% %\includegraphics[scale=0.35]{t25.png}\qquad
% \includegraphics[scale=0.3]{g23.png}
% \quad
% \includegraphics[scale=0.3]{g24.png}
% }
\centerline{
\resizebox{\textwidth}{!}{%
\begin{tikzpicture}[xscale=2, yscale=1.875,>=latex,baseline=(z.base)]
\node at (0,1.5) (z) {};
\node[draw,fill=lightgray] at (1,0) (11) {\tiny $415263^{-1}:0$};
\node[draw] at (0,1) (21) {\tiny $145263^{-1}:1$};
\node[draw] at (1,1) (22) {\tiny $412563^{-1}:1$};
\node[draw] at (2,1) (23) {\tiny $41523^{-1}:1$};
\node[draw] at (0,2) (31) {\tiny $142563^{-1}:2$};
\node[draw] at (1,2) (32) {\tiny $14523^{-1}:2$};
\node[draw] at (2,2) (33) {\tiny $41253^{-1}:2$};
\node[draw,fill=lightskyblue] at (1,3) (41) {\tiny $14253^{-1}:4$};
%
\draw[->,thick] (41) -- (31);
\draw[->,thick] (41) -- (32);
\draw[->,thick] (41) -- (33);
\draw[->,thick] (31) -- (21);
\draw[->,thick] (31) -- (22);
\draw[->,thick] (32) -- (21);
\draw[->,thick] (32) -- (23);
\draw[->,thick] (33) -- (22);
\draw[->,thick] (33) -- (23);
\draw[->,thick] (21) -- (11);
\draw[->,thick] (22) -- (11);
\draw[->,thick] (23) -- (11);
\end{tikzpicture}
\quad
\begin{tikzpicture}[xscale=2.1, yscale=1.5,>=latex,baseline=(z.base)]
\node at (0,2) (z) {};
\node[draw] at (2.5,0) (11) {\tiny $51627384^{-1}:1$};
\node[draw] at (1,1) (21) {\tiny $51623784^{-1}:1$};
\node[draw] at (2,1) (22) {\tiny $51267384^{-1}:1$};
\node[draw] at (3,1) (23) {\tiny $15627384^{-1}:1$};
\node[draw] at (4,1) (24) {\tiny $5162734^{-1}:1$};
\node[draw] at (0,2) (31) {\tiny $51263784^{-1}:2$};
\node[draw] at (1,2) (32) {\tiny $15623784^{-1}:2$};
\node[draw] at (2,2) (33) {\tiny $5162374^{-1}:2$};
\node[draw] at (3,2) (34) {\tiny $15267384^{-1}:2$};
\node[draw] at (4,2) (35) {\tiny $5126734^{-1}:2$};
\node[draw] at (5,2) (36) {\tiny $1562734^{-1}:2$};
\node[draw] at (1,3) (41) {\tiny $15263784^{-1}:4$};
\node[draw] at (2,3) (42) {\tiny $5126374^{-1}:4$};
\node[draw] at (3,3) (43) {\tiny $1562374^{-1}:4$};
\node[draw] at (4,3) (44) {\tiny $1526734^{-1}:4$};
\node[draw,fill=lightskyblue] at (2.5,4) (51) {\tiny $1526374^{-1}:8$};
%
\draw[->,thick] (51) -- (41);
\draw[->,thick] (51) -- (42);
\draw[->,thick] (51) -- (43);
\draw[->,thick] (51) -- (44);
\draw[->,thick] (41) -- (31);
\draw[->,thick] (41) -- (32);
\draw[->,thick] (41) -- (34);
\draw[->,thick] (42) -- (31);
\draw[->,thick] (42) -- (33);
\draw[->,thick] (42) -- (35);
\draw[->,thick] (43) -- (32);
\draw[->,thick] (43) -- (33);
\draw[->,thick] (43) -- (36);
\draw[->,thick] (44) -- (34);
\draw[->,thick] (44) -- (35);
\draw[->,thick] (44) -- (36);
\draw[->,thick] (31) -- (21);
\draw[->,thick] (31) -- (22);
\draw[->,thick] (32) -- (21);
\draw[->,thick] (32) -- (23);
\draw[->,thick] (33) -- (21);
\draw[->,thick] (33) -- (24);
\draw[->,thick] (34) -- (22);
\draw[->,thick] (34) -- (23);
\draw[->,thick] (35) -- (22);
\draw[->,thick] (35) -- (24);
\draw[->,thick] (36) -- (23);
\draw[->,thick] (36) -- (24);
\draw[->,thick] (21) -- (11);
\draw[->,thick] (22) -- (11);
\draw[->,thick] (23) -- (11);
\draw[->,thick] (24) -- (11);
\end{tikzpicture}}}
\caption{The directed graphs $ \cB^+(z)$ when $z$ is %$t^+_5=(2,5)$ (left) and
$g_{23}=(2,4)(3,5)$ (left) and $g_{24}=(2,5)(3,6)(4,7)$ (right), presented using the conventions in Figure~\ref{t1n-fig}.
Here, the grey box indicates the unique element of $\cB^+(z)$ not in $\supp(\GC_z)$.}\label{g2n-fig}
\end{figure}
\newpage
\printbibliography
\end{document}
%=======================================================
% usepackage
\usepackage{latexsym}
\usepackage{amssymb}
\usepackage{amsthm}
\usepackage{amscd}
\usepackage{amsmath}
\usepackage{mathrsfs}
\usepackage{graphicx}
\usepackage{hyperref}
\usepackage{shuffle}
\usepackage{stmaryrd}
\usepackage{enumitem}
\usepackage[noabbrev]{cleveref}
\usepackage{ytableau}
\newcommand\crefrangeconjunction{\textendash}
\usepackage[all]{xy}
\input xy \xyoption{frame}
\xyoption{dvips}
\usepackage[colorinlistoftodos]{todonotes}
\usetikzlibrary{chains,scopes,decorations.markings}
\usepackage{tikz-cd}
%=========================================================================
% theorem declarations
\theoremstyle{definition}
\newtheorem* {theorem*}{Theorem}
\newtheorem* {conjecture*}{Conjecture}
\newtheorem{theorem}{Theorem}[section]
\newtheorem{thmdef}[theorem]{Theorem-Definition}
\newtheorem{propdef}[theorem]{Proposition-Definition}
\newtheorem{question}[theorem]{Question}
\theoremstyle{definition}
\newtheorem{observation}[theorem]{Observation}
\newtheorem {claim}{Claim}
\newtheorem* {example*}{Example}
\newtheorem* {fact}{Fact}
\newtheorem{lemma}[theorem]{Lemma}
\theoremstyle{definition}
\newtheorem{definition}[theorem]{Definition}
\theoremstyle{definition}
\newtheorem* {notation}{Notation}
\newtheorem{conjecture}[theorem]{Conjecture}
\newtheorem{proposition}[theorem]{Proposition}
\newtheorem{corollary}[theorem]{Corollary}
\newtheorem{algorithm}[theorem]{Algorithm}
\newtheorem* {remark*}{Remark}
\newtheorem{remark}[theorem]{Remark}
\theoremstyle{definition}
\newtheorem {example}[theorem]{Example}
\theoremstyle{definition}
\newtheorem* {examples}{Examples}
\theoremstyle{definition}
\newtheorem* {exercise}{Exercise}
\theoremstyle{definition}
\newtheorem{problem}[theorem]{Problem}
\theoremstyle{definition}
\newtheorem* {remarks}{Remarks}
\newcommand{\Fl}{\operatorname{Fl}}
\newcommand{\std}{\operatorname{std}}
\newcommand{\Proj}{\operatorname{Proj}}
\newcommand{\dist}{\operatorname{dist}}
\newcommand{\Des}{\operatorname{Des}}
\newcommand{\DesR}{\operatorname{Des}_R}
\newcommand{\DesL}{\operatorname{Des}_L}
\newcommand{\im}{\operatorname{im}}
\newcommand{\supp}{\operatorname{supp}}
\newcommand{\Inv}{\operatorname{Inv}}
\newcommand{\Ess}{\operatorname{Ess}}
\newcommand{\rank}{\operatorname{rank}}
\renewcommand{\dim}{\operatorname{dim}}
\newcommand{\DesV}{\operatorname{Des}_V}
\newcommand{\Asc}{\operatorname{Asc}}
\newcommand{\asc}{\operatorname{asc}}
\newcommand{\des}{\operatorname{des}}
\newcommand{\last}{\operatorname{end}}
\newcommand{\shift}{\operatorname{shift}}
\newcommand{\upshift}{\mathsf{shift}_{[\uparrow]}}
\newcommand{\downshift}{\mathsf{shift}_{[\downarrow]}}
\def\topRows{\mathsf{topRows}}
\def\cols{\mathsf{cols}}
\def\rW{]}
\def\lW{[}
\def\modu{\ (\mathrm{mod}\ }
\def\wh{\widehat}
\def\({\left(}
\def\){\right)}
\newcommand{\sP}{\mathscr{P}}
\newcommand{\FF}{\mathbb{F}}
\newcommand{\CC}{\mathbb{C}}
\newcommand{\QQ}{\mathbb{Q}}
\newcommand{\cP}{\mathcal{P}}
\newcommand{\cQ}{\mathcal{Q}}
\newcommand{\cK}{\mathcal{K}}
\newcommand{\cR}{\mathcal{R}}
\newcommand{\V}{\mathcal{V}}
\newcommand{\cS}{\mathcal{S}}
\newcommand{\cI}{\mathcal{I}}
\newcommand{\cJ}{\mathcal{J}}
\newcommand{\cC}{\mathcal{C}}
\newcommand{\cD}{\mathcal{D}}
\newcommand{\cZ}{\mathcal{Z}}
\def\cS{\mathcal{S}}
\def\cX{\mathcal{X}}
\def\cW{\mathcal{W}}
\def\cY{\mathcal{Y}}
\def\tr{\mathrm{tr}}
\def\Hom{\mathrm{Hom}}
\def\End{\mathrm{End}}
\def\CC{\mathbb{C}}
\def\RR{\mathbb{R}}
\def\ZZ{\mathbb{Z}}
\def\Aut{\mathrm{Aut}}
\def\sl{\mathfrak{sl}}
\def\Ind{\mathrm{Ind}}
\def\GL{\mathrm{GL}}
\def\ind{\mathrm{ind}}
\def\Res{\mathrm{Res}}
\def\Inf{\mathrm{Inf}}
\def\ch{\mathrm{ch}}
\def\spanning{\textnormal{-span}}
\def\Irr{\mathrm{Irr}}
\def\wt{\widetilde}
\def\cyc{\textsf{cyc}}
\newcommand{\decr}{\mathsf{decr}}
\newcommand{\BCS}{\mathsf{BCS}}
\def\fG{\mathcal{G}^{(\beta)}}
\def\sh{\mathrm{sh}}
\def\diag{\mathrm{diag}}
\def\ts{\textstyle}
\newcommand{\fkn}{\mathfrak{n}}
\newcommand{\cM}{\mathcal{M}}
\newcommand{\cN}{\mathcal{N}}
\newcommand{\cL}{\mathcal{L}}
\newcommand{\g}{\mathfrak{g}}
\newcommand{\One}{{1\hspace{-.15.5cm} 1}}
\newcommand{\one}{{1\hspace{-.11cm} 1}}
\newcommand{\sgn}{\mathrm{sgn}}
\newcommand{\fka}{\mathfrak{a}}
\newcommand{\fkh}{\mathfrak{h}}
\def\fk{\mathfrak}
\def\fkm{\mathfrak{m}}
\def\barr{\begin{array}}
\def\earr{\end{array}}
\def\ba{\begin{aligned}}
\def\ea{\end{aligned}}
\def\be{\begin{equation}}
\def\ee{\end{equation}}
\def\Cyc{\mathrm{Cyc}}
\def\Fix{\mathrm{Fix}}
\def\Pair{\mathrm{Pair}}
\def\pair{\mathrm{pair}}
\def\qquand{\qquad\text{and}\qquad}
\def\quand{\quad\text{and}\quad}
\def\qquord{\qquad\text{or}\qquad}
\def\quord{\quad\text{or}\quad}
\newcommand{\gl}{\mathfrak{gl}}
\def\inv{{\mathsf{inv}}}
\def\I{\mathcal{I}}
\def\M{\mathcal{M}}
\def\cH{\mathcal H}
\def\cM{\mathcal M}
\def\omdef{\overset{\mathrm{def}}}
\def\uS{\underline S}
\def\hs{\hspace{0.5mm}}
\def\uW{\underline W}
\def\uw{\underline w}
\def\RIGHT{\{\to\}}
\def\LEFT{\{\leftarrow\}}
\def\ds{\displaystyle}
\def\proj{\mathrm{proj}}
\def\id{\mathrm{id}}
\def\PP{\mathbb{Z}_{>0}}
\def\NN{\mathbb{Z}_{\geq 0}}
\def\dash{\cdots\cdot\cdot\hs}
\def\fkS{\fk S}
\def\tphi{\tilde \phi}
\def\tpsi{\tilde \psi}
\def\ben{\begin{enumerate}}
\def\een{\end{enumerate}}
\def\bei{\begin{itemize}}
\def\eei{\end{itemize}}
\def\cT{\mathscr{T}}
\def\sU{\mathscr{U}}
\def\cE{\mathcal E}
\def\invol{\textbf{invol}}
\newcommand{\Mon}[1]{\mathscr{F}(#1)}
\def\uL{ \underline{L}}
\def\uR{R}
\def\hs{\hspace{0.5mm}}
\def\simstar{{\hs\hs\sim_{\I_*}\hs\hs}}
\def\fpf{{\mathsf {FPF}}}
\def\fp{\mathrm{fp}}
\def\D{\mathsf{D}}
\newcommand{\wfpf}{\Theta}
\def\DesF{\mathrm{Des}_\fpf}
\def\DesI{\mathrm{Des}_I}
\def\DesIF{\mathrm{Des}_{F}}
\def\Dfpf{\hat D_{\fpf}}
\def\cfpf{\hat c_{\fpf}}
\def\ilambda{\mu}
\def\flambda{\nu}
\def\ellhat{\hat\ell}
\def\Ffpf{\hat F^\fpf}
\def\Sfpf{\hat {\fk S}^\fpf}
\def\ifpf{\iota}
\def\Ifpf{I^\fpf}
\def\x{\textbf{x}}
\def\y{\textbf{y}}
\def\z{\textbf{z}}
\def\a{\textbf{a}}
\def\b{\textbf{b}}
\def\c{\textbf{c}}
\def\d{\textbf{d}}
\def\e{\textbf{e}}
\newcommand{\xRightarrow}[2][]{\ext@arrow 0359\Rightarrowfill@{#1}{#2}}
\renewcommand{\O}{\mathsf{O}}
\newcommand{\Sp}{\mathsf{Sp}}
\renewcommand{\r}[1]{\textcolor{red}{#1}}
\def\dbldots{: \hspace{.55mm} : \hspace{.55mm} :}
\def\L{\underline L}
\def\R{R}
\newcommand{\cA}{\mathcal{A}}
\newcommand{\cB}{\mathcal{B}}
\newcommand{\iB}{\hat\cB}
\newcommand{\fB}{\hat\cB^\fpf}
\def\cAfpf{\cA_\fpf}
\def\cRfpf{\hat\cR_\fpf}
\def\iR{\hat\cR}
\def\iHR{\cH\hat\cR}
\def\act{\rtimes}
\newcommand\dact{\hat\circ}
\def\dem{\circ}
\def\invdemspace{\hspace{.85mm}}
\newcommand{\geninvdem}[3]{#2 \invdemspace \dact#1\invdemspace #3}
\newcommand{\idem}[2]{\geninvdem{_*}{#1}{#2}}
\newcommand{\sidem}[2]{\geninvdem{}{#1}{#2}}
\def\F{\mathscr{F}}
\newcommand{\arc}[2]{ \ar @/^#1pc/ @{-} [#2] }
\def\arcstop{\endxy\ }
\def\arcstart{\ \xy<0cm,-.06cm>\xymatrix@R=.1cm@C=.10cm }
\newcommand{\arcstartc}[1]{\ \xy<0cm,-.15cm>\xymatrix@R=.1cm@C=#1cm}
\def\ellhat{\hat\ell}
\def\Ffpf{\hat F^\fpf}
\def\Sfpf{\hat {\fk S}^\fpf}
\def\iS{\hat \fkS}
\newcommand{\push}{\mathord{\downarrow}}
\newcommand{\pull}{\mathord{\uparrow}}
\newcommand{\ipush}{\mathord{\Downarrow}}
\newcommand{\ipull}{\mathord{\Uparrow}}
\def\sM{\mathscr{M}}
\def\ttS{{\tt S}}
\def\s{\mathbf{s}}
\def\t{\mathbf{t}}
\def\r{\mathbf{r}}
\def\customstar{\hs\hat \bullet_{\mathrm{des}}\hs}
\def\red{\emph{reduce}}
\def\sA{\mathscr{A}}
\def\sB{\mathscr{B}}
\def\sR{\mathscr{R}}
\def\sC{\mathscr{C}}
\def\sE{\mathscr{E}}
\def\ihB{\underline{\isB}}
\def\sI{\mathscr{I}}
\def\tS{\tilde S}
\def\tX{G}
\def\tI{\tilde I}
\def\cG{\mathcal{G}}
\newcommand{\sAq}[1]{\sA_{\bullet,\geq #1}}
\newcommand{\sAp}[1]{\sA_{\leq #1,\bullet}}
\def\sX{\mathscr{X}}
\def\sY{\mathscr{Y}}
\def\cWmin{\cM}
\def\cV{\mathcal{V}}
\def\NCSM{\textsf{NCSP}}
\def\XA{\cW_{{\tt embed}}}
\def\HA{\cA_{{\tt hecke}}}
\def\bar{\overline}
\def\scov{<\hspace{-1.6mm}\vartriangleleft_B}
\def\bcov{<\hspace{-1.6mm}\blacktriangleleft_B}
\def\lleq{\leq \hspace{-1.6mm}\leq}
\def\X{\mathcal{X}}
\def\Y{\mathcal{Y}}
\def\Z{\mathcal{Z}}
\def\wB{\bar1\hs\bar 2\hs\bar 3\cdots \bar n}
\def\Id{Id}
\def\cRwin{\cR^{\mathrm{win}}}
\def\iF{\hat F}
\def\SL{\mathrm{SL}}
\def\O{\mathrm{O}}
\def\W{W^{\mathsf{BC}}}
\def\DGroup{W^{\mathsf{D}}}
\def\ellD{\ell^{\mathsf{D}}}
\def\wD{w_0^{\mathsf{D}_n}}
\def\HA{\cA_{\mathsf{Hecke}}}
\def\cAB{\cA_{\mathsf{Brion}}}
\def\RSphi{\phi_{\mathrm{RS}}}
\def\bRSphi{\bar\phi_{\mathrm{RS}}}
\def\cAfpfD{\cAfpf^{\mathsf{D}}}
\def\cAD{\cA^{\mathsf{D}}}
\def\Aphi{\iota^{\mathsf{A}}}
\def\BCDphi{\iota^{\mathsf{BC}}}
\def\RSphiB{\phi_{\mathrm{B}}}
\def\bRSphiB{\bar\phi_{\mathrm{B}}}
\def\SGamma{\Gamma^{\mathsf{sym}}}
\def\SSGamma{\Gamma^{\mathsf{skew}}}
\def\Clans{\mathsf{Clans}}
\def\fpf{\mathsf{fpf}}
\def\BC{\mathsf{BC}}
\def\cAk{\cA_k}
\def\cAkfpf{\cA_k^{\fpf}}
\def\yfpf{y^{\fpf}}
\def\gfpf{g^{\fpf}}
\def\shfpf{\sh^{\fpf}}
\def\signs{\mathsf{signs}}
\definecolor{darkred}{rgb}{0.7,0,0} % darkred color
\newcommand{\defn}[1]{{\color{darkred}\emph{#1}}} % emphasis of a definition
\def\AO{\mathsf{AO}}
\def\MAO{\fkm\mathsf{AO}}
\newcommand{\Sym}{\mathsf{Sym}}
\newcommand{\QSym}{\mathsf{QSym}}
\def\wzeta{\bar\zeta}
\def\zetaq{\zeta_{\mathsf{Q}}}
\def\wzetaq{\wzeta_{\mathsf{Q}}}
\def\op{\mathsf{op}}
\def\oX{\overset{\to}{X}}
\def\oGamma{\overset{\to}{\Gamma}}
\def\bDelta{\Delta^{(\beta)}}
\def\mQSym{\fkm\QSym}
\def\mSym{\fkm\Sym}
\newcommand{\cO}{\mathcal{D}}
\def\wO{\bar\cO}
\def\wG{\bar\cG}
\def\wP{\bar\cP}
\def\htimes{\mathbin{\bar\otimes}}
\def\bX{\bar X}
\def\Set{\mathsf{Set}}
\def\bare{\bar e}
\def\bGamma{\bar \Gamma}
\def\mL{\cM}
\def\bL{\bar L}
\def\antipode{\mathsf{S}}
\def\wGwt{\mathfrak{m}\mathsf{WGraphs}}
\def\wzetaw{\wzeta_{\mathsf{WGraphs}}}
\def\kam{\overline{\widetilde m}}
\def\align{\mathsf{align}}
\def\SetSSYT{\mathsf{SetSSYT}}
\def\source{\mathsf{source}}
\def\oDelta{\overset{\to}\Delta}
\def\ozetaG{\overset{\to}{\zeta_\cG}}
\def\ozetaP{\overset{\to}{\zeta_\cP}}
\def\ocG{\overset{\to}\cG}
\def\ocP{\overset{\to}\cP}
\def\tm{\widetilde m}
\def\btm{\overline{\widetilde m}}
\def\oT{\mathscr{G}}
\def\inc{\mathsf{inc}}
\usepackage{pbox}
\tikzset{every loop/.style={min distance=10 mm, in=60, out=120, looseness=10}}
\tikzset{
dot/.style = {circle, fill, inner sep=2.2,outer sep=0},
% dot/.default = 6pt % size of the circle diameter
}
\numberwithin{equation}{section}
\allowdisplaybreaks[1]
\UseCrayolaColors
\def\KK{\mathbb{K}}
\def\merge{\mathsf{merge}}
\newcommand{\arxiv}[1]{\href{https://arxiv.org/abs/#1}{\texttt{arXiv:#1}}}
\def\op{\mathrm{decr}}
\def\ozeta{\zeta_{\oG}}
\def\bpartial{\partial^{(\beta)}}
\def\bpi{\pi^{(\beta)}}
\def\fkG{\mathcal{G}^{(\beta)}}
\def\fkGO{\mathcal{G}^{\mathsf{O}}}
\def\fkGSp{\mathcal{G}^{\mathsf{Sp}}}
\def\MX{M\hspace{-0.5mm}X}
\def\Mat{\mathsf{Mat}}
\def\codim{\mathrm{codim}}
\def\cF{\mathcal{F}}
\def\cU{\mathcal{U}}
\def\crb{\mathsf{crb}}
\def\shiftable{\mathsf{shiftable}}
\def\dom{\mathsf{dom}}
\def\iG{\widehat{\cG}}
\def\ellfpf{\ell_{\mathsf{fpf}}}
\def\ellhat{\ell_{\mathsf{inv}}}
\def\ODes{\mathrm{ODes}}
% \def\varpi{\Pi}
\def\SD{\mathsf{SD}}
\def\forced{\mathsf{F}}
\def\prohib{\mathsf{P}}
\def\simplus{\xrightarrow{[n]}}
\def\GC{\operatorname{GC}^{\mathsf{O}}}
\def\GCSp{\operatorname{GC}^{\mathsf{Sp}}}
\def\GQ{GQ}
\def\GP{GP}
\def\Sh{\mathsf{Sh}}
\def\cBfpf{\mathcal{B}_\fpf}
\def\cB{\mathcal{B}_\inv}
\def\SubDiag{\mathsf{SubDiag}}
\def\word{\mathsf{word}}
\newcommand{\Ivex}{I^{\mathsf{vex}}_\infty}
\def\iD{\hat D}
% $ biblatex auxiliary file $
% $ biblatex bbl format version 3.2 $
% Do not modify the above lines!
%
% This is an auxiliary file used by the 'biblatex' package.
% This file may safely be deleted. It will be recreated as
% required.
%
\begingroup
\makeatletter
\@ifundefined{ver@biblatex.sty}
{\@latex@error
{Missing 'biblatex' package}
{The bibliography requires the 'biblatex' package.}
\aftergroup\endinput}
{}
\endgroup
\datalist[entry]{nty/global//global/global}
\entry{CCG}{book}{}
\name{author}{2}{}{%
{{hash=BA}{%
family={Bj\"orner},
familyi={B\bibinitperiod},
given={A.},
giveni={A\bibinitperiod},
}}%
{{hash=BF}{%
family={B.},
familyi={B\bibinitperiod},
given={Francesco},
giveni={F\bibinitperiod},
}}%
}
\list{publisher}{1}{%
{Springer, New York}%
}
\strng{namehash}{BABF1}
\strng{fullhash}{BABF1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{B}
\field{sortinithash}{B}
\field{pages}{xiv+363}
\field{series}{Graduate Texts in Mathematics}
\field{title}{Combinatorics of {C}oxeter groups}
\field{volume}{231}
\field{year}{2005}
\endentry
\entry{Brion2002}{article}{}
\name{author}{1}{}{%
{{hash=BM}{%
family={Brion},
familyi={B\bibinitperiod},
given={M.},
giveni={M\bibinitperiod},
}}%
}
\strng{namehash}{BM1}
\strng{fullhash}{BM1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{B}
\field{sortinithash}{B}
\field{note}{Special issue in celebration of Claudio Procesi's 60th birthday}
\field{number}{1}
\field{pages}{137\bibrangedash 159}
\field{title}{Positivity in the {G}rothendieck group of complex flag varieties}
\field{volume}{258}
\field{journaltitle}{J. Algebra}
\field{year}{2002}
\endentry
\entry{Brion}{article}{}
\name{author}{1}{}{%
{{hash=BM}{%
family={Brion},
familyi={B\bibinitperiod},
given={M.},
giveni={M\bibinitperiod},
}}%
}
\strng{namehash}{BM1}
\strng{fullhash}{BM1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{B}
\field{sortinithash}{B}
\field{number}{1}
\field{pages}{137\bibrangedash 174}
\field{title}{The behaviour at infinity of the {B}ruhat decomposition}
\field{volume}{73}
\field{journaltitle}{Comment. Math. Helv.}
\field{year}{1998}
\endentry
\entry{Buch2002}{article}{}
\name{author}{1}{}{%
{{hash=BAS}{%
family={Buch},
familyi={B\bibinitperiod},
given={A.\bibnamedelima S.},
giveni={A\bibinitperiod\bibinitdelim S\bibinitperiod},
}}%
}
\strng{namehash}{BAS1}
\strng{fullhash}{BAS1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{B}
\field{sortinithash}{B}
\field{pages}{37\bibrangedash 78}
\field{title}{A Littlewood-Richardson rule for the $K$-theory of Grassmannians}
\field{volume}{189}
\field{journaltitle}{Acta Math.}
\field{year}{2002}
\endentry
\entry{BuchEtAl}{article}{}
\name{author}{5}{}{%
{{hash=BAS}{%
family={Buch},
familyi={B\bibinitperiod},
given={A.\bibnamedelima S.},
giveni={A\bibinitperiod\bibinitdelim S\bibinitperiod},
}}%
{{hash=KA}{%
family={Kresch},
familyi={K\bibinitperiod},
given={A.},
giveni={A\bibinitperiod},
}}%
{{hash=SM}{%
family={Shimozono},
familyi={S\bibinitperiod},
given={M.},
giveni={M\bibinitperiod},
}}%
{{hash=TH}{%
family={Tamvakis},
familyi={T\bibinitperiod},
given={H.},
giveni={H\bibinitperiod},
}}%
{{hash=YA}{%
family={Yong},
familyi={Y\bibinitperiod},
given={A.},
giveni={A\bibinitperiod},
}}%
}
\list{publisher}{1}{%
{Springer}%
}
\strng{namehash}{BAS+1}
\strng{fullhash}{BASKASMTHYA1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{B}
\field{sortinithash}{B}
\field{pages}{359\bibrangedash 382}
\field{title}{Stable Grothendieck polynomials and K-theoretic factor sequences}
\field{volume}{340}
\field{journaltitle}{Math. Ann.}
\field{year}{2008}
\endentry
\entry{CJW}{article}{}
\name{author}{3}{}{%
{{hash=CMB}{%
family={Can},
familyi={C\bibinitperiod},
given={M.\bibnamedelima B.},
giveni={M\bibinitperiod\bibinitdelim B\bibinitperiod},
}}%
{{hash=JM}{%
family={Joyce},
familyi={J\bibinitperiod},
given={M.},
giveni={M\bibinitperiod},
}}%
{{hash=WB}{%
family={Wyser},
familyi={W\bibinitperiod},
given={B.},
giveni={B\bibinitperiod},
}}%
}
\strng{namehash}{CMBJMWB1}
\strng{fullhash}{CMBJMWB1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{C}
\field{sortinithash}{C}
\field{pages}{207\bibrangedash 225}
\field{title}{Chains in weak order posets associated to involutions}
\field{volume}{137}
\field{journaltitle}{J. Combin. Theory Ser. A}
\field{year}{2016}
\endentry
\entry{Chiu.Marberg2023}{article}{}
\name{author}{2}{}{%
{{hash=CYC}{%
family={Chiu},
familyi={C\bibinitperiod},
given={Y.\bibnamedelima C.},
giveni={Y\bibinitperiod\bibinitdelim C\bibinitperiod},
}}%
{{hash=ME}{%
family={Marberg},
familyi={M\bibinitperiod},
given={E.},
giveni={E\bibinitperiod},
}}%
}
\strng{namehash}{CYCME1}
\strng{fullhash}{CYCME1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{C}
\field{sortinithash}{C}
\field{number}{6}
\field{pages}{1419\bibrangedash 1445}
\field{title}{Expanding K-theoretic Schur Q-functions}
\field{volume}{6}
\field{journaltitle}{Algebr. Comb.}
\field{year}{2023}
\endentry
\entry{FominKirillov}{article}{}
\name{author}{2}{}{%
{{hash=FS}{%
family={Fomin},
familyi={F\bibinitperiod},
given={S.},
giveni={S\bibinitperiod},
}}%
{{hash=KA}{%
family={Kirillov},
familyi={K\bibinitperiod},
given={A.},
giveni={A\bibinitperiod},
}}%
}
\strng{namehash}{FSKA1}
\strng{fullhash}{FSKA1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{F}
\field{sortinithash}{F}
\field{pages}{183\bibrangedash 190}
\field{title}{Grothendieck polynomials and the Yang-Baxter equation}
\field{journaltitle}{Proceedings of the Sixth Conference in Formal Power Series and Algebraic Combinatorics, DIMACS}
\field{year}{1994}
\endentry
\entry{GPP}{article}{}
\name{author}{3}{}{%
{{hash=GO}{%
family={Guibert},
familyi={G\bibinitperiod},
given={O.},
giveni={O\bibinitperiod},
}}%
{{hash=PE}{%
family={Pergola},
familyi={P\bibinitperiod},
given={E.},
giveni={E\bibinitperiod},
}}%
{{hash=PR}{%
family={Pinzani},
familyi={P\bibinitperiod},
given={R.},
giveni={R\bibinitperiod},
}}%
}
\strng{namehash}{GOPEPR1}
\strng{fullhash}{GOPEPR1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{G}
\field{sortinithash}{G}
\field{number}{2}
\field{pages}{153\bibrangedash 174}
\field{title}{Vexillary involutions are enumerated by {M}otzkin numbers}
\field{volume}{5}
\field{journaltitle}{Ann. Comb.}
\field{year}{2001}
\endentry
\entry{HMP6}{article}{}
\name{author}{3}{}{%
{{hash=HZ}{%
family={Hamaker},
familyi={H\bibinitperiod},
given={Z.},
giveni={Z\bibinitperiod},
}}%
{{hash=ME}{%
family={Marberg},
familyi={M\bibinitperiod},
given={E.},
giveni={E\bibinitperiod},
}}%
{{hash=PB}{%
family={Pawlowski},
familyi={P\bibinitperiod},
given={B.},
giveni={B\bibinitperiod},
}}%
}
\strng{namehash}{HZMEPB1}
\strng{fullhash}{HZMEPB1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{H}
\field{sortinithash}{H}
\field{number}{5}
\field{pages}{1310\bibrangedash 1346}
\field{title}{Involution pipe dreams}
\field{volume}{74}
\field{journaltitle}{Canad. J. Math.}
\field{year}{2022}
\endentry
\entry{HMP1}{article}{}
\name{author}{3}{}{%
{{hash=HZ}{%
family={Hamaker},
familyi={H\bibinitperiod},
given={Z.},
giveni={Z\bibinitperiod},
}}%
{{hash=ME}{%
family={Marberg},
familyi={M\bibinitperiod},
given={E.},
giveni={E\bibinitperiod},
}}%
{{hash=PB}{%
family={Pawlowski},
familyi={P\bibinitperiod},
given={B.},
giveni={B\bibinitperiod},
}}%
}
\strng{namehash}{HZMEPB1}
\strng{fullhash}{HZMEPB1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{H}
\field{sortinithash}{H}
\field{pages}{217\bibrangedash 260}
\field{title}{Involution words: counting problems and connections to {S}chubert calculus for symmetric orbit closures}
\field{volume}{160}
\field{journaltitle}{J. Combin. Theory Ser. A}
\field{year}{2018}
\endentry
\entry{HMP2}{article}{}
\name{author}{3}{}{%
{{hash=HZ}{%
family={Hamaker},
familyi={H\bibinitperiod},
given={Z.},
giveni={Z\bibinitperiod},
}}%
{{hash=ME}{%
family={Marberg},
familyi={M\bibinitperiod},
given={E.},
giveni={E\bibinitperiod},
}}%
{{hash=PB}{%
family={Pawlowski},
familyi={P\bibinitperiod},
given={B.},
giveni={B\bibinitperiod},
}}%
}
\strng{namehash}{HZMEPB1}
\strng{fullhash}{HZMEPB1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{H}
\field{sortinithash}{H}
\field{number}{3}
\field{pages}{701\bibrangedash 743}
\field{title}{Involution words {II}: braid relations and atomic structures}
\field{volume}{45}
\field{journaltitle}{J. Algebraic Combin.}
\field{year}{2017}
\endentry
\entry{HMP3}{article}{}
\name{author}{3}{}{%
{{hash=HZ}{%
family={Hamaker},
familyi={H\bibinitperiod},
given={Z.},
giveni={Z\bibinitperiod},
}}%
{{hash=ME}{%
family={Marberg},
familyi={M\bibinitperiod},
given={E.},
giveni={E\bibinitperiod},
}}%
{{hash=PB}{%
family={Pawlowski},
familyi={P\bibinitperiod},
given={B.},
giveni={B\bibinitperiod},
}}%
}
\strng{namehash}{HZMEPB1}
\strng{fullhash}{HZMEPB1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{H}
\field{sortinithash}{H}
\field{number}{17}
\field{pages}{5389\bibrangedash 5440}
\field{title}{Schur P-positivity and involution Stanley symmetric functions}
\field{journaltitle}{IMRN}
\field{year}{2019}
\endentry
\entry{Hultman2007}{article}{}
\name{author}{1}{}{%
{{hash=HA}{%
family={Hultman},
familyi={H\bibinitperiod},
given={A.},
giveni={A\bibinitperiod},
}}%
}
\strng{namehash}{HA1}
\strng{fullhash}{HA1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{H}
\field{sortinithash}{H}
\field{number}{6}
\field{pages}{2787\bibrangedash 2798}
\field{title}{The combinatorics of twisted involutions in {C}oxeter groups}
\field{volume}{359}
\field{journaltitle}{Trans. Amer. Math. Soc.}
\field{year}{2007}
\endentry
\entry{Ikeda.Naruse2013}{article}{}
\name{author}{2}{}{%
{{hash=IT}{%
family={Ikeda},
familyi={I\bibinitperiod},
given={T.},
giveni={T\bibinitperiod},
}}%
{{hash=NH}{%
family={Naruse},
familyi={N\bibinitperiod},
given={H.},
giveni={H\bibinitperiod},
}}%
}
\strng{namehash}{ITNH1}
\strng{fullhash}{ITNH1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{I}
\field{sortinithash}{I}
\field{pages}{22\bibrangedash 66}
\field{title}{K-theoretic analogues of factorial Schur P- and Q- functions}
\field{volume}{243}
\field{journaltitle}{Adv. Math.}
\field{year}{2013}
\endentry
\entry{Knutson2009}{misc}{}
\name{author}{1}{}{%
{{hash=KA}{%
family={Knutson},
familyi={K\bibinitperiod},
given={A.},
giveni={A\bibinitperiod},
}}%
}
\strng{namehash}{KA1}
\strng{fullhash}{KA1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{K}
\field{sortinithash}{K}
\verb{eprint}
\verb 0902.1930
\endverb
\field{title}{Frobenius splitting and {M}\"obius inversion}
\field{eprinttype}{arxiv}
\field{year}{Preprint (2009), {\tt arXiv:0902.1930}}
\endentry
\entry{KnutsonMiller}{article}{}
\name{author}{2}{}{%
{{hash=KA}{%
family={Knutson},
familyi={K\bibinitperiod},
given={A.},
giveni={A\bibinitperiod},
}}%
{{hash=ME}{%
family={Miller},
familyi={M\bibinitperiod},
given={E.},
giveni={E\bibinitperiod},
}}%
}
\strng{namehash}{KAME1}
\strng{fullhash}{KAME1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{K}
\field{sortinithash}{K}
\field{number}{3}
\field{pages}{1245\bibrangedash 1318}
\field{title}{Gr\"obner geometry of {S}chubert polynomials}
\field{volume}{161}
\field{journaltitle}{Ann. of Math.}
\field{year}{2005}
\endentry
\entry{KnutsonMiller2004}{article}{}
\name{author}{2}{}{%
{{hash=KA}{%
family={Knutson},
familyi={K\bibinitperiod},
given={A.},
giveni={A\bibinitperiod},
}}%
{{hash=ME}{%
family={Miller},
familyi={M\bibinitperiod},
given={E.},
giveni={E\bibinitperiod},
}}%
}
\strng{namehash}{KAME1}
\strng{fullhash}{KAME1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{K}
\field{sortinithash}{K}
\field{pages}{161\bibrangedash 176}
\field{title}{Subword complexes in Coxeter groups}
\field{volume}{184}
\field{journaltitle}{Adv. Math.}
\field{year}{2004}
\endentry
\entry{Lenart}{article}{}
\name{author}{1}{}{%
{{hash=LC}{%
family={Lenart},
familyi={L\bibinitperiod},
given={C.},
giveni={C\bibinitperiod},
}}%
}
\strng{namehash}{LC1}
\strng{fullhash}{LC1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{L}
\field{sortinithash}{L}
\field{number}{1-2}
\field{pages}{137\bibrangedash 158}
\field{title}{A {$K$}-theory version of {M}onk's formula and some related multiplication formulas}
\field{volume}{179}
\field{journaltitle}{J. Pure Appl. Algebra}
\field{year}{2003}
\endentry
\entry{lenart2000}{article}{}
\name{author}{1}{}{%
{{hash=LC}{%
family={Lenart},
familyi={L\bibinitperiod},
given={C.},
giveni={C\bibinitperiod},
}}%
}
\list{publisher}{1}{%
{Springer}%
}
\strng{namehash}{LC1}
\strng{fullhash}{LC1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{L}
\field{sortinithash}{L}
\field{pages}{67\bibrangedash 82}
\field{title}{Combinatorial aspects of the {{K}}-theory of {{G}}rassmannians}
\field{volume}{4}
\field{journaltitle}{Ann. Comb.}
\field{year}{2000}
\endentry
\entry{LenartSottile}{article}{}
\name{author}{2}{}{%
{{hash=LC}{%
family={Lenart},
familyi={L\bibinitperiod},
given={C.},
giveni={C\bibinitperiod},
}}%
{{hash=SF}{%
family={Sottile},
familyi={S\bibinitperiod},
given={F.},
giveni={F\bibinitperiod},
}}%
}
\strng{namehash}{LCSF1}
\strng{fullhash}{LCSF1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{L}
\field{sortinithash}{L}
\field{number}{5}
\field{pages}{2317\bibrangedash 2342}
\field{title}{A {P}ieri-type formula for the {$K$}-theory of a flag manifold}
\field{volume}{359}
\field{journaltitle}{Trans. Amer. Math. Soc.}
\field{year}{2007}
\endentry
\entry{LM2019}{article}{}
\name{author}{2}{}{%
{{hash=LJB}{%
family={Lewis},
familyi={L\bibinitperiod},
given={J.\bibnamedelima B.},
giveni={J\bibinitperiod\bibinitdelim B\bibinitperiod},
}}%
{{hash=ME}{%
family={Marberg},
familyi={M\bibinitperiod},
given={E.},
giveni={E\bibinitperiod},
}}%
}
\strng{namehash}{LJBME1}
\strng{fullhash}{LJBME1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{L}
\field{sortinithash}{L}
\field{pages}{1929\bibrangedash 1972}
\field{title}{Enriched set-valued $P$-partitions and shifted stable Grothendieck polynomials}
\field{volume}{299}
\field{journaltitle}{Math. Z.}
\field{year}{2021}
\endentry
\entry{Manivel}{book}{}
\name{author}{1}{}{%
{{hash=ML}{%
family={Manivel},
familyi={M\bibinitperiod},
given={L.},
giveni={L\bibinitperiod},
}}%
}
\list{publisher}{2}{%
{American Mathematical Society, Providence, RI}%
{Soci\'{e}t\'{e} Math\'{e}matique de France, Paris}%
}
\strng{namehash}{ML1}
\strng{fullhash}{ML1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{M}
\field{sortinithash}{M}
\field{note}{Translated from the 1998 French original by John R. Swallow, Cours Sp\'{e}cialis\'{e}s, 3}
\field{pages}{viii+167}
\field{series}{SMF/AMS Texts and Monographs}
\field{title}{Symmetric functions, {S}chubert polynomials and degeneracy loci}
\field{volume}{6}
\field{year}{2001}
\endentry
\entry{Mar2019}{article}{}
\name{author}{1}{}{%
{{hash=ME}{%
family={Marberg},
familyi={M\bibinitperiod},
given={E.},
giveni={E\bibinitperiod},
}}%
}
\strng{namehash}{ME1}
\strng{fullhash}{ME1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{M}
\field{sortinithash}{M}
\field{pages}{Paper No. 105216}
\field{title}{A symplectic refinement of shifted Hecke insertion}
\field{volume}{173}
\field{journaltitle}{J. Combin. Theory Ser. A}
\field{year}{2020}
\endentry
\entry{MP2020}{article}{}
\name{author}{2}{}{%
{{hash=ME}{%
family={Marberg},
familyi={M\bibinitperiod},
given={E.},
giveni={E\bibinitperiod},
}}%
{{hash=PB}{%
family={Pawlowski},
familyi={P\bibinitperiod},
given={B.},
giveni={B\bibinitperiod},
}}%
}
\strng{namehash}{MEPB1}
\strng{fullhash}{MEPB1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{M}
\field{sortinithash}{M}
\field{pages}{Paper No. 107299}
\field{title}{{$K$}-theory formulas for orthogonal and symplectic orbit closures}
\field{volume}{372}
\field{journaltitle}{Adv. Math.}
\field{year}{2020}
\endentry
\entry{MP2021Sp}{article}{}
\name{author}{2}{}{%
{{hash=ME}{%
family={Marberg},
familyi={M\bibinitperiod},
given={E.},
giveni={E\bibinitperiod},
}}%
{{hash=PB}{%
family={Pawlowski},
familyi={P\bibinitperiod},
given={B.},
giveni={B\bibinitperiod},
}}%
}
\strng{namehash}{MEPB1}
\strng{fullhash}{MEPB1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{M}
\field{sortinithash}{M}
\field{number}{1}
\field{pages}{Paper No. 106463}
\field{title}{On some properties of symplectic {G}rothendieck polynomials}
\field{volume}{225}
\field{journaltitle}{J. Pure Appl. Algebra}
\field{year}{2021}
\endentry
\entry{MP2021}{article}{}
\name{author}{2}{}{%
{{hash=ME}{%
family={Marberg},
familyi={M\bibinitperiod},
given={E.},
giveni={E\bibinitperiod},
}}%
{{hash=PB}{%
family={Pawlowski},
familyi={P\bibinitperiod},
given={B.},
giveni={B\bibinitperiod},
}}%
}
\strng{namehash}{MEPB1}
\strng{fullhash}{MEPB1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{M}
\field{sortinithash}{M}
\field{number}{2}
\field{pages}{273\bibrangedash 287}
\field{title}{Principal specializations of {S}chubert polynomials in classical types}
\field{volume}{4}
\field{journaltitle}{Algebr. Comb.}
\field{year}{2021}
\endentry
\entry{MS2023}{misc}{}
\name{author}{2}{}{%
{{hash=ME}{%
family={Marberg},
familyi={M\bibinitperiod},
given={E.},
giveni={E\bibinitperiod},
}}%
{{hash=ST}{%
family={Scrimshaw},
familyi={S\bibinitperiod},
given={T.},
giveni={T\bibinitperiod},
}}%
}
\strng{namehash}{MEST1}
\strng{fullhash}{MEST1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{M}
\field{sortinithash}{M}
\verb{eprint}
\verb 2302.04226
\endverb
\field{title}{Key and {L}ascoux polynomials for symmetric orbit closures}
\field{eprinttype}{arxiv}
\field{year}{Preprint (2023), {\tt arXiv:2302.04226}}
\endentry
\entry{Pawlowski-universal}{article}{}
\name{author}{1}{}{%
{{hash=PB}{%
family={Pawlowski},
familyi={P\bibinitperiod},
given={B.},
giveni={B\bibinitperiod},
}}%
}
\strng{namehash}{PB1}
\strng{fullhash}{PB1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{P}
\field{sortinithash}{P}
\field{pages}{1417\bibrangedash 1461}
\field{title}{Universal graph Schubert varieties}
\field{volume}{26}
\field{journaltitle}{Transform. Groups}
\field{year}{2021}
\endentry
\entry{WyserYong}{article}{}
\name{author}{2}{}{%
{{hash=WB}{%
family={Wyser},
familyi={W\bibinitperiod},
given={B.},
giveni={B\bibinitperiod},
}}%
{{hash=YA}{%
family={Yong},
familyi={Y\bibinitperiod},
given={A.},
giveni={A\bibinitperiod},
}}%
}
\strng{namehash}{WBYA1}
\strng{fullhash}{WBYA1}
\field{labelnamesource}{author}
\field{labeltitlesource}{title}
\field{sortinit}{W}
\field{sortinithash}{W}
\field{number}{1}
\field{pages}{267\bibrangedash 290}
\field{title}{Polynomials for symmetric orbit closures in the flag variety}
\field{volume}{22}
\field{journaltitle}{Transform. Groups}
\field{year}{2017}
\endentry
\enddatalist
\endinput