Squarefree Powers of Edge Ideals under Graph Joins and Cones
Extracted source text ยท arXiv 2607.04964
\documentclass[12pt]{article}
\usepackage[margin=1in]{geometry}
\usepackage{amssymb,amsfonts,amsmath,amscd,amsthm,amsxtra}
\usepackage[table, dvipsnames]{xcolor}
\usepackage{caption}
\captionsetup{justification=centering}
\usepackage{array}
\usepackage{graphicx,tikz, pgf}
\usepackage{calc}
\usepackage{float}
\usepackage{subcaption}
\usepackage{mathrsfs}
\usepackage{enumerate}
\usepackage{bm}
\usepackage{fancyhdr}
\usepackage{hyperref}
\hypersetup{colorlinks=true,linkcolor=blue,citecolor=blue,urlcolor=blue}
\textwidth=16.00cm
\textheight=23cm
\topmargin=0.00cm
\oddsidemargin=0.0cm
\evensidemargin=0.00cm
\headsep=0.3cm
\setlength{\headheight}{14pt}
\setlength{\parskip}{1.2pt}
\newtheorem{theorem}{Theorem}[section]
\newtheorem{lemma}[theorem]{Lemma}
\newtheorem{proposition}[theorem]{Proposition}
\newtheorem{corollary}[theorem]{Corollary}
\newtheorem{obsr}[theorem]{Observation}
\theoremstyle{definition}
\newtheorem{remark}[theorem]{Remark}
\newtheorem{notation}[theorem]{Notation}
\newtheorem{example}[theorem]{Example}
\newtheorem{definition}[theorem]{Definition}
\newtheorem{conjecture}[theorem]{Conjecture}
\newtheorem{question}[theorem]{Question}
\DeclareMathOperator{\pd}{pd}
\DeclareMathOperator{\reg}{reg}
\DeclareMathOperator{\depth}{depth}
\DeclareMathOperator{\height}{ht}
\DeclareMathOperator{\dimn}{dim}
\def\k{\Bbbk}
\def\c{\mathcal{C}}
\def\s{\subseteq}
\def\K{\mathcal{K}}
\title{A Join-Matching Theorem for Squarefree Powers of Edge Ideals, with Applications to Wheel and Related Graphs}
\author{Bilal Ahmad Wani$^{a}$\footnote{Corresponding author.}\hspace{0.5cm}and Uzair Rafiq Shah$^{b}$}
\date{\footnotesize $^{a,b}$Department of Mathematics, National Institute of Technology Srinagar, 190006, India\\[2pt]
$^{a}$bilalwani@nitsri.ac.in \quad
$^{b}$shahuzair1909@gmail.com}
\begin{document}
\maketitle
\newcommand{\wn}{\mathcal{W}_n}
\newcommand{\cn}{\mathcal{C}_n}
\pagestyle{fancy}
\fancyhf{}
\fancyhf[OHC]{B. A. Wani}
\fancyhf[EHC]{Squarefree powers of edge ideals of wheel graphs}
\fancyfoot[C]{\thepage}
\renewcommand{\headrulewidth}{0pt}
\begin{abstract}
\noindent
For $q\ge 1$, the $q$-th squarefree power $I(G)^{[q]}$ of the edge ideal of a graph $G$ is generated by the squarefree monomials supported on $q$-matchings of $G$; it is the Stanley--Reisner ideal of the complex $\Delta_q(G)=\{F\subseteq V(G):\nu(G[F])<q\}$, where $\nu$ denotes matching number. We prove a general formula for the matching number of an arbitrary graph join,
\[
\nu(G\ast H) = \min\Big(\nu(G)+|V(H)|,\ \ \nu(H)+|V(G)|,\ \ \Big\lfloor\tfrac{|V(G)|+|V(H)|}{2}\Big\rfloor\Big),
\]
via the Tutte--Berge formula, and use it to decompose $\Delta_q(G\ast H)$ for arbitrary graphs $G,H$. Specializing to the wheel graph $\wn=\cn\ast\mathcal K_1$, we determine the Krull dimension and height of $R/I(\wn)^{[q]}$ exactly for all $n\ge 3$, $1\le q\le\lfloor n/2\rfloor$, and -- combining our matching-number computations with a recent Tutte-type Cohen--Macaulayness criterion of Ficarra and Moradi -- prove that at the \emph{top} squarefree power $q=\nu(\wn)=\lceil n/2\rceil$, the ideal $I(\wn)^{[\nu(\wn)]}$ is literally the squarefree Veronese ideal, so that $R/I(\wn)^{[\nu(\wn)]}$ is Cohen--Macaulay with
\[
\dimn = \depth = \reg\big(R/I(\wn)^{[\nu(\wn)]}\big) = 2\Big\lceil\frac n2\Big\rceil-1.
\]
This resolves all four classical invariants at the top power, and confirms there the pattern $\depth(R/I(\wn)^{[q]})=2q-1$ that our computational data (now extended to $n\le13$, every valid $q$) suggests holds throughout. We prove a general depth formula for squarefree powers of cone graphs, via a Betti-splitting exact sequence, that reduces this pattern to two more tractable statements about the underlying cycle alone; both are verified computationally in every case checked but left open in general. We further apply our join-matching and decomposition theorems, by the same extremal method, to fan graphs, multi-hub wheel graphs, complete split graphs, and friendship graphs, obtaining closed-form Krull dimension formulas for three of these families and an explicit partial decomposition for the fourth.
\medskip
\noindent
\textbf{2020 Mathematics Subject Classification:} 13D02, 13F55, 05C70, 05E40\\
\textbf{Key words:} squarefree power, edge ideal, graph join, wheel graph, matching number, Tutte--Berge formula, Cohen--Macaulay, Krull dimension, depth, regularity.
\end{abstract}
\section{Introduction}
Let $R=\Bbbk[x_1,\ldots,x_n]$ be a polynomial ring over a field $\Bbbk$ and let $G$ be a finite simple graph on $\{x_1,\ldots,x_n\}$. The edge ideal $I(G)\subseteq R$, generated by $\{x_ix_j:\{x_i,x_j\}\in E(G)\}$, was introduced by Villarreal \cite{V1}; the graded Betti numbers, projective dimension and regularity of $R/I(G)$ and its ordinary powers have since been studied extensively \cite{herhib,HVT1,jacq,katz,peeva,rova}.
More recently, Erey, Herzog, Hibi and Saeedi Madani \cite{EHHM} introduced the \emph{squarefree powers}
\[
I(G)^{[q]} := \Big(\, \prod_{v\in V(M)} x_v \ \Big| \ M \text{ a } q\text{-matching of } G \,\Big) \subseteq R \qquad (q\ge 1),
\]
which specialize at $q=1$ to the ordinary edge ideal. Since every $q$-matching has exactly $2q$ vertices, $I(G)^{[q]}$ is automatically the Stanley--Reisner ideal of
\begin{equation}\label{eq:deltaq}
\Delta_q(G) := \{ F \subseteq V(G) : \nu(G[F]) < q \}, \qquad \nu(H):=\text{matching number of } H,
\end{equation}
and $\Delta_1(G)$ is the independence complex of $G$. Regularity of squarefree powers has been determined for forests \cite{CFL}, cycles \cite{DRS}, whiskered cycles \cite{DGS}, and Cameron--Walker and several well-covered graph classes \cite{Fak}; the top power $I(G)^{[\nu(G)]}$ has been the subject of particular attention, with Bigdeli, Herzog and Zaare-Nahandi \cite{BHN} proving $\reg(I(G)^{[\nu(G)]})=2\nu(G)$ for \emph{every} graph $G$, and Ficarra and Moradi \cite{FM} recently giving a complete, purely combinatorial (Tutte-type) characterization of when $I(G)^{[\nu(G)]}$ is Cohen--Macaulay. To our knowledge, wheel graphs have not been treated by any of this squarefree-power literature.
The $q=1$ case for wheel graphs is understood via a different route: Mousivand \cite{Mous}, generalizing Emtander, Fr\"oberg, Mohammadi and Moradi \cite{EFMM}, expressed the Betti numbers of $R/I(G\ast H)$ for \emph{any} join of graphs in terms of those of $G,H$, since the independence complex of a join is the disjoint union $\Delta_1(G\ast H)=\Delta_1(G)\sqcup\Delta_1(H)$. In a companion paper \cite{RPW} the author and coauthors compute the graded Betti numbers, projective dimension and regularity of the \emph{ordinary} edge ideal $I(\wn)$ -- that is, exclusively the case $q=1$ -- of the wheel graph; \cite{RPW} does not address squarefree powers $q\ge2$, nor the top-power invariants that are the focus of the present paper. This raises the natural question: does an analogous structural decomposition of $\Delta_q(G\ast H)$ exist for $q\ge2$? We show that it does, and that it follows from a single, purely graph-theoretic fact about matchings in joins.
\section{Preliminaries}
For a graph $X$, $\nu(X)$ is its matching number and $d(X):=|V(X)|-2\nu(X)\ge0$ its (Tutte--Berge) \emph{deficiency}; $X$ has a perfect matching iff $d(X)=0$. The \emph{join} $G\ast H$ of graphs on disjoint vertex sets has $E(G\ast H)=E(G)\cup E(H)\cup\{gh: g\in V(G), h\in V(H)\}$.
\begin{theorem}[Tutte--Berge formula]\label{thm:tb}
For any finite simple graph $X$,
\[
\nu(X) = \frac{1}{2}\Big(|V(X)| - \max_{S\subseteq V(X)}\big(o(X-S)-|S|\big)\Big),
\]
where $o(Y)$ denotes the number of odd-order connected components of $Y$.
\end{theorem}
\begin{definition}[\cite{EHHM}]
For $q\ge1$, $I(G)^{[q]} = ( x_{V(M)} : M \text{ a } q\text{-matching of } G )\subseteq R$, where $x_{V(M)}=\prod_{u\in V(M)}x_u$.
\end{definition}
As all $q$-matchings have $2q$ vertices, every monomial $x_{V(M)}$ in this generating set has the same degree $2q$; since a squarefree monomial can properly divide another squarefree monomial only if its degree is strictly smaller, no monomial in the set (of \emph{distinct} such monomials -- two different $q$-matchings may of course yield the same monomial, which simply means that monomial is listed once) can divide another. Hence the generating set is automatically minimal, and $I(G)^{[q]}=I_{\Delta_q(G)}$ for $\Delta_q(G)$ as in \eqref{eq:deltaq}. Hochster's formula \cite{hoch1} gives, for every graph $G$ and $q\ge1$,
\begin{equation}\label{eq:hochq}
\beta_{i,j}\big(R/I(G)^{[q]}\big) = \sum_{W\subseteq V(G),\, |W|=j} \dim_\Bbbk \widetilde H_{j-i-1}\big(\Delta_q(G)[W];\Bbbk\big),
\end{equation}
and $\reg(R/I)=\max\{j-i:\beta_{i,j}\ne0\}$, $\pd(R/I)=\max\{i:\beta_{i,j}\ne0\text{ some }j\}$, $\depth(R/I)=|V(G)|-\pd(R/I)$ (Auslander--Buchsbaum). For a squarefree monomial ideal $I=I_\Delta$, $\dimn(R/I)=\max\{|F|:F\in\Delta\}$ and $\height(I)=|V(G)|-\dimn(R/I)$; $R/I$ is Cohen--Macaulay if $\depth(R/I)=\dimn(R/I)$.
For $1\le k\le n$, the \emph{squarefree Veronese ideal} $\mathfrak m^{[k]}\subseteq \Bbbk[x_1,\ldots,x_n]$ is generated by all squarefree monomials of degree $k$, i.e.\ it is $I_\Delta$ for $\Delta$ the $(k-2)$-skeleton of the full simplex on $n$ vertices.
\begin{theorem}[Herzog--Hibi, {\cite[Theorem 4.2]{HH}}]\label{thm:veronese}
$\Bbbk[x_1,\ldots,x_n]/\mathfrak m^{[k]}$ is Cohen--Macaulay, and $\dimn = \depth = \reg = k-1$.
\end{theorem}
(The regularity statement is the classical fact that squarefree Veronese ideals have linear resolutions; the dimension is immediate since the facets of the $(k-2)$-skeleton of a simplex on $n$ vertices all have size $k-1$.)
\begin{theorem}[Ficarra--Moradi, {\cite[Theorem 1.8]{FM}}]\label{thm:fm}
Let $G$ be a graph on $n$ non-isolated vertices. The following are equivalent:
\begin{enumerate}
\item[(a)] $I(G)^{[\nu(G)]}$ is Cohen--Macaulay;
\item[(b)] $I(G)^{[\nu(G)]}=\mathfrak m^{[2\nu(G)]}$ and $\nu(G)=\lfloor n/2\rfloor$;
\item[(c)] either $G$ has a perfect matching, or $G\setminus\{i\}$ has a perfect matching for every $i\in V(G)$.
\end{enumerate}
\end{theorem}
\begin{definition}
$\wn=\cn\ast\mathcal K_1$: vertex set $\{x_1,\ldots,x_n,v\}$, $\cn$ the rim on $U=\{x_1,\ldots,x_n\}$, $v$ the hub adjacent to all of $U$.
\end{definition}
\section{A general join-matching theorem}
\begin{theorem}\label{thm:joinmatching}
Let $G,H$ be finite simple graphs on disjoint vertex sets, $n_G=|V(G)|$, $n_H=|V(H)|$. Then
\[
\nu(G\ast H) = \min\Big(\nu(G)+n_H,\ \ \nu(H)+n_G,\ \ \big\lfloor\tfrac{n_G+n_H}2\big\rfloor\Big).
\]
\end{theorem}
\begin{proof}
Write $X=G\ast H$. We compute $d(X)=\max_{S\subseteq V(X)}(o(X-S)-|S|)$ from Theorem \ref{thm:tb} by splitting the maximization according to $S=S_G\cup S_H$, $S_G=S\cap V(G)$, $S_H=S\cap V(H)$. These three cases exhaust all $S\subseteq V(X)$: if $S_G\ne V(G)$ and $S_H\ne V(H)$ both fail, then $S_G=V(G)$ or $S_H=V(H)$.
\emph{Case 1: $S_G\ne V(G)$ and $S_H\ne V(H)$.} Then $X-S=(G-S_G)\ast(H-S_H)$ is a join of two non-empty graphs, hence connected, so $o(X-S)\in\{0,1\}$ and $o(X-S)-|S|\le 1-|S|\le1$. At $S=\emptyset$ this equals $\varepsilon:=(n_G+n_H)\bmod 2\in\{0,1\}$; for any $S\ne\emptyset$ in this case, $o(X-S)-|S|\le1-|S|\le0\le\varepsilon$, so the maximum over this case is exactly $\varepsilon$, attained at $S=\emptyset$.
\emph{Case 2: $S_G=V(G)$.} Then $X-S=H-S_H$ (no $G$-vertices remain, so no cross edges survive), so $o(X-S)-|S| = o(H-S_H)-n_G-|S_H|$; maximizing over $S_H$ gives $d(H)-n_G$ by Theorem \ref{thm:tb} applied to $H$.
\emph{Case 3: $S_H=V(H)$.} Symmetrically, gives $d(G)-n_H$.
Hence $d(X)=\max\big(\varepsilon,\ d(H)-n_G,\ d(G)-n_H\big)$, so, using $n_G+n_H-\max(a,b,c)=\min(n_G+n_H-a,\,n_G+n_H-b,\,n_G+n_H-c)$,
\[
\nu(X) = \frac{n_G+n_H-d(X)}2 = \min\Big(\tfrac{n_G+n_H-\varepsilon}2,\ \tfrac{n_G+n_H-d(H)+n_G}2,\ \tfrac{n_G+n_H-d(G)+n_H}2\Big).
\]
The first term equals $\lfloor(n_G+n_H)/2\rfloor$, since subtracting the parity bit $\varepsilon$ before halving always yields the floor; the second is $n_G+\tfrac{n_H-d(H)}2=n_G+\nu(H)$ by Tutte--Berge applied to $H$; the third is $n_H+\nu(G)$ symmetrically. This is the claimed formula.
\end{proof}
\begin{corollary}[Cone lemma, cf.\ \cite{RPW}]\label{cor:cone}
For any graph $H$, $\nu(H\ast\mathcal K_1)=\nu(H)$ if $H$ has a perfect matching, and $\nu(H)+1$ otherwise.
\end{corollary}
\begin{proof}
Take $G=\mathcal K_1$ in Theorem \ref{thm:joinmatching}: $n_G=1,\nu(G)=0$, so $\nu(H\ast\mathcal K_1)=\min(n_H,\ \nu(H)+1,\ \lfloor(n_H+1)/2\rfloor)$. If $d(H)=0$ then $n_H=2\nu(H)$ and $\lfloor(n_H+1)/2\rfloor=\nu(H)<\nu(H)+1\le n_H$ (for $\nu(H)\ge1$; the case $\nu(H)=0,H=\emptyset$ is trivial), so the minimum is $\nu(H)$. If $d(H)\ge1$ then $n_H=2\nu(H)+d(H)\ge2\nu(H)+1$, so $\lfloor(n_H+1)/2\rfloor\ge\nu(H)+1$ and $n_H\ge2\nu(H)+1\ge\nu(H)+1$ (using $\nu(H)\ge0$); since both of the other two terms are $\ge\nu(H)+1$, and $\nu(H)+1$ is itself one of the three terms being minimized, the minimum is exactly $\nu(H)+1$.
\end{proof}
\begin{corollary}\label{cor:km}
For any graph $G$ and $m\ge1$:
\[
\nu(G\ast\mathcal K_m) = \min\Big(\nu(G)+m,\ \big\lfloor\tfrac m2\big\rfloor+n_G,\ \big\lfloor\tfrac{n_G+m}2\big\rfloor\Big), \qquad
\nu(G\ast\overline{\mathcal K_m}) = \min\Big(n_G,\ \nu(G)+m,\ \big\lfloor\tfrac{n_G+m}2\big\rfloor\Big),
\]
where $\overline{\mathcal K_m}$ is the edgeless graph on $m$ vertices.
\end{corollary}
\begin{proof}
Substitute $H=\mathcal K_m$ (so $\nu(H)=\lfloor m/2\rfloor$) or $H=\overline{\mathcal K_m}$ (so $\nu(H)=0$) into Theorem \ref{thm:joinmatching}.
\end{proof}
\begin{corollary}\label{cor:nuwn}
$\nu(\wn) = \lceil n/2\rceil$.
\end{corollary}
\begin{proof}
Take $G=\cn$, $H=\mathcal K_1$ in Theorem \ref{thm:joinmatching}: $\nu(\wn)=\min(\lfloor n/2\rfloor+1,\ n,\ \lceil n/2\rceil)$. Since $\lceil n/2\rceil\le\lfloor n/2\rfloor+1$ always, and $\lceil n/2\rceil\le n$ for $n\ge1$, the minimum is $\lceil n/2\rceil$.
\end{proof}
%Corollary \ref{cor:km} answers, in closed form, the natural question of extending the cone lemma from a single hub to $m$ mutually adjacent or mutually non-adjacent hubs, e.g.\ to the multi-hub wheel graphs $\cn\ast\mathcal K_m$ and $\cn\ast\overline{\mathcal K_m}$, without further work.
\section{Decomposition of $\Delta_q(G\ast H)$}
\begin{lemma}\label{lem:inducedjoin}
For $F_G\subseteq V(G)$, $F_H\subseteq V(H)$, $(G\ast H)[F_G\cup F_H] = G[F_G]\ast H[F_H]$.
\end{lemma}
\begin{proof}
This is immediate from the definitions: the induced subgraph on $F_G\cup F_H$ retains exactly the edges of $G$ within $F_G$, the edges of $H$ within $F_H$, and all cross edges between $F_G$ and $F_H$ -- which is precisely $E(G[F_G])\cup E(H[F_H])\cup\{gh:g\in F_G,h\in F_H\}$, the edge set of $G[F_G]\ast H[F_H]$.
\end{proof}
\begin{theorem}\label{thm:gendecomp}
Let $G,H$ be graphs on disjoint vertex sets and $q\ge1$. For $F=F_G\cup F_H\subseteq V(G)\cup V(H)$,
\[
F\in\Delta_q(G\ast H) \iff \min\Big(\big\lfloor\tfrac{|F_G|+|F_H|}2\big\rfloor,\ |F_G|+\nu(H[F_H]),\ |F_H|+\nu(G[F_G])\Big) < q.
\]
\end{theorem}
\begin{proof}
By Lemma \ref{lem:inducedjoin}, $(G\ast H)[F]=G[F_G]\ast H[F_H]$; apply Theorem \ref{thm:joinmatching} to this join.
\end{proof}
\begin{corollary}[Wheel decomposition, cf.\ \cite{RPW}]\label{cor:wheeldecomp}
For $F=F'\cup\{v\}$ or $F=F'\subseteq U$ ($F'\subseteq U$), $F\in\Delta_q(\wn)$ iff: (a) $v\notin F$ and $\nu(\cn[F'])<q$; or (b) $v\in F$ and $\nu(\cn[F'])<q-1$ if $\cn[F']$ has no perfect matching, or $\nu(\cn[F'])<q$ if it does.
\end{corollary}
\begin{proof}
Take $H=\mathcal K_1$ ($n_H=1,\nu(H)=0$) in Theorem \ref{thm:gendecomp}, with $F_G=F'$.
If $v\notin F$ ($F_H=\emptyset$): the three terms become $\lfloor|F'|/2\rfloor$, $|F'|+\nu(\emptyset)=|F'|$, and $0+\nu(\cn[F'])=\nu(\cn[F'])$. Since $\nu(\cn[F'])\le\lfloor|F'|/2\rfloor\le|F'|$ always (a matching number never exceeds half the vertex count), $\nu(\cn[F'])$ is always the smallest term, giving (a).
If $v\in F$ ($F_H=\{v\}$): the three terms become $\lceil|F'|/2\rceil$, $|F'|+\nu(\mathcal K_1)=|F'|$, and $1+\nu(\cn[F'])$. Since $|F'|\ge\lceil|F'|/2\rceil$ always, the middle term never binds, leaving $\min(\lceil|F'|/2\rceil,\,1+\nu(\cn[F']))$.
If $\cn[F']$ has a perfect matching, $|F'|$ is even and $\nu(\cn[F'])=|F'|/2=\lceil|F'|/2\rceil$, so the two remaining terms are $\nu(\cn[F'])$ and $1+\nu(\cn[F'])$; the minimum is $\nu(\cn[F'])$, giving (b)'s second case.
If $\cn[F']$ has no perfect matching, write $\cn[F']$ as a disjoint union of arcs; a perfect matching fails exactly when at least $t\ge1$ of these arcs have odd length, and each odd-length arc contributes one less to the matching number than half its own vertex count, so $\nu(\cn[F'])=(|F'|-t)/2$. Since $t$ and $|F'|$ have the same parity (each odd arc changes the vertex-count parity), $t\ge1$ and $t\equiv|F'|\pmod2$ together force $t\ge2$ when $|F'|$ is even; a direct check of both parities gives $1+\nu(\cn[F'])=1+(|F'|-t)/2\le\lceil|F'|/2\rceil$ in either case. So $1+\nu(\cn[F'])$ is the binding term, giving (b)'s first case.
\end{proof}
Theorem \ref{thm:gendecomp} applies to \emph{any} join of two graphs at \emph{any} squarefree power $q\ge1$, reducing the study of $\Delta_q(G\ast H)$ entirely to the matching numbers of induced subgraphs of $G$ and of $H$ separately.% we draw on this generality twice more below -- immediately, for a depth formula covering \emph{any} cone graph (Theorem \ref{thm:conedepth}), and later for closed-form dimension formulas for fan, complete split, and friendship graphs (Section \ref{sec:apps}) -- rather than treating the wheel as a one-off computation.
\subsection{Depth formula for cone graphs}\label{ssec:conedepth}
We now extract a genuinely new consequence of Theorem \ref{thm:gendecomp} that goes beyond dimension: a formula computing the \emph{depth} of $R/I(\mathrm{Cone}(G))^{[q]}$ (for $\mathrm{Cone}(G):=G\ast\mathcal K_1$, any graph $G$) in terms of the depths of two auxiliary ideals on $G$'s own, smaller, vertex set. Applied later to $G=\cn$ this will sharpen Conjecture \ref{conj:depth} into two precise, independently verifiable statements about ordinary cycles.
Fix $G$ on $\{x_1,\ldots,x_n\}$, $R=\Bbbk[x_1,\ldots,x_n]$, hub $v$, $S=R[v]$.
\begin{definition}\label{def:hubideal}
For $q\ge1$, let
\[
J_{G,q} := \big(\, x_j\cdot x_{V(M')} \ :\ j\in V(G),\ M' \text{ a } (q-1)\text{-matching of } G\setminus\{x_j\} \,\big) \subseteq R
\]
(for $q=1$ the only $0$-matching is empty, so $J_{G,1}=(x_1,\ldots,x_n)=\mathfrak m$).
\end{definition}
By the same case-splitting argument as in the proof of Corollary \ref{cor:wheeldecomp} (applied to $H=\mathcal K_1$, $G$ arbitrary): writing $F=F'$ or $F=F'\cup\{v\}$ for $F'\subseteq V(G)$, one has $F'\cup\{v\}\in\Delta_q(\mathrm{Cone}(G))$ iff $F'$ avoids every generator support of $J_{G,q}$. Consequently $I(\mathrm{Cone}(G))^{[q]} = I(G)^{[q]} + v\cdot J_{G,q}$: matchings of $\mathrm{Cone}(G)$ avoiding $v$ generate $I(G)^{[q]}$, and matchings using $v$ (an edge $\{v,x_j\}$ together with a $(q-1)$-matching of $G\setminus\{x_j\}$, by the same reasoning as Corollary \ref{cor:cone}'s proof) generate $v\cdot J_{G,q}$.
\begin{lemma}\label{lem:intersecteqI1}
$I(G)^{[q]} \cap J_{G,q} = I(G)^{[q]}$.
\end{lemma}
\begin{proof}
Write $\Delta=\Delta_q(\mathrm{Cone}(G))$. The complex of $I(G)^{[q]}$ is the deletion $\mathrm{del}_\Delta(v)=\Delta[V(G)]=\Delta_q(G)$, and the complex of $J_{G,q}$ is the link $\mathrm{link}_\Delta(v)=\{F'\subseteq V(G):F'\cup\{v\}\in\Delta\}$ (this is exactly the membership condition recalled above). For \emph{any} simplicial complex $\Delta$ and vertex $v$, $\mathrm{link}_\Delta(v)\subseteq\mathrm{del}_\Delta(v)$: if $F\cup\{v\}\in\Delta$ then, as $\Delta$ is downward closed and $F\subseteq F\cup\{v\}$, also $F\in\Delta$, i.e.\ $F\in\Delta[V(G)]$. Since the Stanley--Reisner correspondence sends ideal intersection to complex union, $I(G)^{[q]}\cap J_{G,q}$ is the ideal of $\Delta_q(G)\cup\mathrm{link}_\Delta(v)$, which equals $\Delta_q(G)$ because $\mathrm{link}_\Delta(v)\subseteq\Delta_q(G)$; this ideal is exactly $I(G)^{[q]}$.
\end{proof}
\begin{theorem}[Cone depth theorem]\label{thm:conedepth}
Let $G$ be a graph on $n$ vertices and $q\ge1$. If
\[
d_1 := \depth_R\big(R/I(G)^{[q]}\big) \;>\; d_2 := \depth_R\big(R/J_{G,q}\big),
\]
then $\depth_S\big(S/I(\mathrm{Cone}(G))^{[q]}\big) = d_2+1$.
\end{theorem}
\begin{proof}
Write $I=I(\mathrm{Cone}(G))^{[q]}$, $I_1=I(G)^{[q]}$, $I_2=v\cdot J_{G,q}$, so $I=I_1+I_2$ as recalled above. Since $I_1,J_{G,q}\subseteq R$ don't involve $v$, comparing minimal generators shows $I_1\cap I_2 = I_1\cap(vJ_{G,q}) = v\cdot(I_1\cap J_{G,q})$ (an lcm of a generator of $I_1$ with $v$ times a generator of $J_{G,q}$ is always $v$ times their lcm in $R$). By Lemma \ref{lem:intersecteqI1}, $I_1\cap I_2 = v\cdot I_1$. This gives the standard short exact sequence
\[
0 \to S/(vI_1) \to S/I_1 \oplus S/(vJ_{G,q}) \to S/I \to 0. \tag{$\ast$}
\]
We compute each depth via two elementary facts about $S=R[v]$ and a nonzero ideal $L\subseteq R$:
\begin{itemize}
\item[(i)] $\pd_S(S/L)=\pd_R(R/L)$: $S/L\cong(R/L)\otimes_R S$, and tensoring a minimal free $R$-resolution of $R/L$ with the free (hence flat) $R$-module $S$ yields a minimal free $S$-resolution.
\item[(ii)] $\pd_S(S/(vL)) = \pd_R(R/L)$: multiplication by $v$ gives an $S$-module isomorphism $L\otimes_RS\xrightarrow{\cong}vL$ of internal degree $1$, so $\pd_S(vL)=\pd_S(L\otimes_RS)=\pd_R(L)$ by (i) applied to the module $L$; then $\pd_S(S/(vL))=\pd_S(vL)+1=\pd_R(L)+1=\pd_R(R/L)$ (using $\pd_R(L)=\pd_R(R/L)-1$).
\end{itemize}
Applying (i) to $I_1$, (ii) to $J_{G,q}$, and (ii) again to $I_1$ (for the $v I_1$ term), Auslander--Buchsbaum ($S$ has $n+1$ variables, $R$ has $n$) gives
\[
\depth_S(S/I_1)=d_1+1,\qquad \depth_S(S/(vJ_{G,q}))=d_2+1,\qquad \depth_S(S/(vI_1))=d_1+1.
\]
In the notation of $(\ast)$: $A:=S/(vI_1)$ has depth $d_1+1$; $B:=S/I_1\oplus S/(vJ_{G,q})$ has depth $\min(d_1,d_2)+1=d_2+1$ (using $d_1>d_2$); $C:=S/I$. Since $\depth(A)=d_1+1>d_2+1=\depth(B)$, the long exact sequence in local cohomology attached to $(\ast)$ gives, at cohomological degree $i=\depth(B)$: $H^{\depth(B)}_{\mathfrak m}(A)=0$ (as $\depth(B)<\depth(A)$), so the map $H^{\depth(B)}_{\mathfrak m}(B)\to H^{\depth(B)}_{\mathfrak m}(C)$ is injective on the nonzero group $H^{\depth(B)}_{\mathfrak m}(B)$, forcing $H^{\depth(B)}_{\mathfrak m}(C)\ne0$, i.e.\ $\depth(C)\le\depth(B)$. Conversely the general depth-lemma inequality $\depth(C)\ge\min(\depth(B),\depth(A)-1)=\depth(B)$ (using $\depth(A)-1\ge\depth(B)$) gives $\depth(C)\ge\depth(B)$. Hence $\depth(C)=\depth(B)=d_2+1$.
\end{proof}
%We verified Theorem \ref{thm:conedepth}'s hypothesis and conclusion computationally for $G=\cn$, $5\le n\le13$, every $1\le q\le\lfloor n/2\rfloor$ (20 cases in total): in every single case $d_1>d_2$ held, and the predicted depth $d_2+1$ matched the independently computed $\depth(R/I(\wn)^{[q]})$ exactly. Section \ref{sec:data} uses Theorem \ref{thm:conedepth} to sharpen Conjecture \ref{conj:depth}.
\section{Krull dimension and height}
We isolate the combinatorial heart of this section as two short, self-contained lemmas about induced subgraphs of a cycle, each with a complete elementary proof. Theorem \ref{thm:dim} is then a direct assembly of these two lemmas via Corollary \ref{cor:wheeldecomp}.
\begin{lemma}[General extremal arc lemma]\label{lem:arcgen}
Let $n\ge3$ and let $k$ be an integer with $0\le k\le\lfloor n/2\rfloor-1$. Then
\[
\max\big\{\,|F| : F\subsetneq V(\cn),\ \nu(\cn[F])\le k\,\big\} = \Big\lfloor\frac n2\Big\rfloor+k.
\]
\end{lemma}
\begin{proof}
The bound holds trivially for $F=\emptyset$, so assume $F\ne\emptyset$. Since $F\subsetneq V(\cn)$ by hypothesis (so in particular $F\ne V(\cn)$, and the degenerate case ``$\cn[F]$ has no gap at all'' cannot occur), $\cn[F]$ is a disjoint union of $t\ge1$ arcs (maximal paths) of lengths $m_1,\ldots,m_t\ge1$, cyclically separated by $t$ non-empty gaps. Hence $\nu(\cn[F])=\sum_i\lfloor m_i/2\rfloor$ and $|F|+t\le n$.
Write $m_i=2k_i+\varepsilon_i$ with $k_i=\lfloor m_i/2\rfloor\ge0$ and $\varepsilon_i\in\{0,1\}$. Fix, for the moment, both the number of arcs $t$ and the total budget $K:=\sum_ik_i=\nu(\cn[F])\le k$; note that the gap constraint $|F|+t\le n$ (derived above from the existence of $t$ non-empty gaps) depends only on $t$ and on $|F|$ itself, not on how the individual arc lengths $m_i$ are chosen, so it places no further restriction on the parities $\varepsilon_i$. Consequently, for this fixed $(t,K)$, the vertex count $|F|=\sum_im_i=2K+\sum_i\varepsilon_i$ is maximized by taking every $\varepsilon_i=1$ (every arc odd, e.g.\ by letting a single arc absorb the whole budget $K$ and setting all other $k_i=0$), giving $|F|=2K+t$; this is a valid, achievable configuration for the fixed $(t,K)$ in question, not merely an upper bound. Having fixed the arc parities at their optimum for each $(t,K)$, we now optimize over $t$ and then over $K\le k$.
The constraint $|F|+t\le n$ becomes $2K+2t\le n$, i.e.\ $t\le\lfloor(n-2K)/2\rfloor$; since $|F|=2K+t$ is increasing in $t$, the optimal choice is $t=\lfloor(n-2K)/2\rfloor$, giving
\[
|F|(K) = 2K+\Big\lfloor\frac{n-2K}2\Big\rfloor = n-\Big\lceil\frac{n-2K}2\Big\rceil,
\]
which is (weakly) increasing in $K$. So the maximum over $K\le k$ occurs at $K=k$, giving
\[
|F|_{\max} = 2k+\Big\lfloor\frac{n-2k}2\Big\rfloor = 2k+\Big\lfloor\frac n2\Big\rfloor-k = \Big\lfloor\frac n2\Big\rfloor+k,
\]
using $\lfloor(n-2k)/2\rfloor=\lfloor n/2\rfloor-k$ since $k\in\mathbb Z$. This value is achieved: the hypothesis $k\le\lfloor n/2\rfloor-1$ gives $t=\lfloor(n-2k)/2\rfloor\ge1$, so a valid configuration exists (e.g.\ a single odd arc of length $2k+1$ carrying the whole budget $K=k$, together with $t-1$ further isolated single vertices if $t>1$, each contributing $k_i=0$).
\end{proof}
\begin{lemma}[Perfect-matching arc lemma]\label{lem:arcpm}
Let $n\ge3$ and let $k\ge0$ be an integer with $2k\le n-2$. Then
\[
\max\big\{\,|F| : F\subseteq V(\cn),\ \cn[F]\text{ is a disjoint union of even arcs},\ \nu(\cn[F])\le k\,\big\} = 2k.
\]
\end{lemma}
\begin{proof}
If $\cn[F]$ is a disjoint union of even arcs of lengths $m_1,\ldots,m_t$ (possibly $t=0$, i.e.\ $F=\emptyset$), then $\nu(\cn[F])=\sum_i m_i/2$, so $|F|=\sum_im_i=2\nu(\cn[F])\le2k$. Equality is achieved by taking $F$ to be a single even arc of length $2k$: since $2k\le n-2$, the complementary gap has size $n-2k\ge2\ge1$, so this is a valid proper subset of the cycle, and $\nu(\cn[F])=k$ exactly.
\end{proof}
\begin{theorem}\label{thm:dim}
For all $n\ge3$ and $1\le q\le\lfloor n/2\rfloor$,
\[
\dimn\big(R/I(\wn)^{[q]}\big) = \Big\lfloor\frac n2\Big\rfloor+q-1, \qquad
\height\big(I(\wn)^{[q]}\big) = (n+1)-\Big\lfloor\frac n2\Big\rfloor-q+1.
\]
\end{theorem}
\begin{proof}
By Corollary \ref{cor:wheeldecomp}, $\dimn(R/I(\wn)^{[q]})=\max\{|F|:F\in\Delta_q(\wn)\}$, and we split according to whether $v\in F$.
\textbf{Step 1 ($F=F'\subseteq U$).} By Corollary \ref{cor:wheeldecomp}(a), $F'\in\Delta_q(\wn)$ iff $\nu(\cn[F'])\le q-1$. Since $1\le q\le\lfloor n/2\rfloor$, we have $0\le q-1\le\lfloor n/2\rfloor-1$, so Lemma \ref{lem:arcgen} applies with $k=q-1$ (noting $F'=U$ is automatically excluded from the extremal configuration, since $\nu(\cn[U])=\lfloor n/2\rfloor\ge q>q-1$, so $F'=U$ never satisfies the membership condition in this range) and gives
\[
\max\{|F'|\} = \Big\lfloor\frac n2\Big\rfloor+(q-1).
\]
\textbf{Step 2 ($F=\{v\}\cup F'$).} By Corollary \ref{cor:wheeldecomp}(b) there are two mutually exclusive sub-cases.
\begin{itemize}
\item[(i)] $\cn[F']$ has a perfect matching (equivalently, is a disjoint union of even arcs) and $\nu(\cn[F'])\le q-1$. Since $q\le\lfloor n/2\rfloor$ gives $2(q-1)\le 2\lfloor n/2\rfloor-2\le n-2$, Lemma \ref{lem:arcpm} applies with $k=q-1$ and gives $\max\{|F'|\}=2(q-1)$, so $|F|\le 2q-1$.
\item[(ii)] $\cn[F']$ has no perfect matching and $\nu(\cn[F'])\le q-2$ (vacuous if $q=1$). If $q\ge2$, then $0\le q-2\le\lfloor n/2\rfloor-1$, so Lemma \ref{lem:arcgen} applies with $k=q-2$ and gives $\max\{|F'|\}=\lfloor n/2\rfloor+(q-2)$, so $|F|\le\lfloor n/2\rfloor+q-1$.
\end{itemize}
Since $q\le\lfloor n/2\rfloor$, we have $2q-1\le\lfloor n/2\rfloor+q-1$, so sub-case (i) never exceeds sub-case (ii)'s bound, and (ii)'s bound coincides exactly with Step 1's bound.
Combining both steps, $\max\{|F|\}=\lfloor n/2\rfloor+q-1$; the height formula follows from $|V(\wn)|=n+1$.
\end{proof}
\begin{proposition}\label{prop:cycledim}
For all $n\ge3$ and $1\le q\le\lfloor n/2\rfloor$, $\dimn(R/I(\cn)^{[q]})=\lfloor n/2\rfloor+q-1$ as well; the hub vertex does not change the dimension of the squarefree power.
\end{proposition}
\begin{proof}
This is exactly Lemma \ref{lem:arcgen} with $k=q-1$, applied to $F\subseteq V(\cn)$ directly; the hub plays no role in that lemma.
\end{proof}
%We verified Theorem \ref{thm:dim} and Proposition \ref{prop:cycledim} by exhaustive brute-force search for $5\le n\le10$, and separately verified the achievability half by direct construction for $n$ up to $50$.
\begin{proposition}\label{prop:extremal}
Fix $n,q$ with $1\le q\le\lfloor n/2\rfloor$, and set $t=\lfloor(n-2q+2)/2\rfloor$. Choose odd integers $m_1,\ldots,m_t\ge1$ with $\sum_i\lfloor m_i/2\rfloor=q-1$, arrange them cyclically around $\cn$ separated by $t$ gaps each of size $1$ except (if $n-2q$ is odd) exactly one gap of size $2$, and let $F'\subseteq U$ be the resulting set of arc vertices. Then $F'\in\Delta_q(\wn)$ and $|F'|=\lfloor n/2\rfloor+q-1$.
\end{proposition}
\begin{proof}
By construction $\sum m_i+(\text{gaps})=n$ and $\nu(\cn[F'])=\sum\lfloor m_i/2\rfloor=q-1<q$.
\end{proof}
\begin{remark}\label{rmk:extremal_incomplete}
Proposition \ref{prop:extremal} does \emph{not} describe every extremal face. For $n=9,q=2$ (where $\lfloor n/2\rfloor+q-1=5$), direct enumeration finds $45$ extremal faces: $27$ of the all-odd-arc shape $(1,1,3)$ predicted above, but also $9$ of shape $(1,1,1,2)$ (an arc pattern including one even arc, arising because odd $n$ allows a second way to trade arc length against gap length), and $9$ further extremal faces of the form $\{v\}\cup F'$ with $F'$ of shape $(1,1,1,1)$, as in sub-case (i) of Step 2 in the proof of Theorem \ref{thm:dim}. A complete classification of extremal faces is left open.
\end{remark}
\section{Complete resolution of the top squarefree power}\label{sec:top}
The results of Section 5 leave a gap at the top of the range: Theorem \ref{thm:dim} requires $q\le\lfloor n/2\rfloor$, while $\nu(\wn)=\lceil n/2\rceil$ (Corollary \ref{cor:nuwn}) exceeds $\lfloor n/2\rfloor$ whenever $n$ is odd. We now resolve this gap completely, and in fact determine \emph{every} classical homological invariant of $R/I(\wn)^{[q]}$ exactly at $q=\nu(\wn)$, by identifying the ideal itself. We emphasize at the outset that the Cohen--Macaulayness criterion we invoke (Theorem \ref{thm:fm}) is due to Ficarra and Moradi \cite{FM} and is not claimed as new; what is new is the explicit verification, for every wheel graph $\mathcal W_n$ regardless of parity of $n$, that its hypothesis holds, and the closed-form computation of dimension, depth, regularity and projective dimension that this verification unlocks in one stroke.
\begin{theorem}\label{thm:top}
For every $n\ge3$, with $q=\nu(\wn)=\lceil n/2\rceil$:
\[
I(\wn)^{[q]} = \mathfrak m^{[2q]} \qquad (\mathfrak m = (x_1,\ldots,x_n,v)),
\]
so $R/I(\wn)^{[q]}$ is Cohen--Macaulay and
\[
\dimn = \depth = \reg\big(R/I(\wn)^{[q]}\big) = 2q-1 = 2\Big\lceil\frac n2\Big\rceil-1,\]
\[\pd\big(R/I(\wn)^{[q]}\big) = n-2\Big\lceil\frac n2\Big\rceil+2 = \begin{cases}1,& n\text{ odd},\\ 2,& n\text{ even}.\end{cases}
\]
\end{theorem}
\begin{proof}
By Theorem \ref{thm:fm} it suffices to check condition (c): either $\wn$ has a perfect matching, or $\wn\setminus\{i\}$ has a perfect matching for every $i\in V(\wn)$.
\emph{Case $n$ odd.} Then $|V(\wn)|=n+1$ is even, and by Corollary \ref{cor:nuwn}, $\nu(\wn)=\lceil n/2\rceil=(n+1)/2=|V(\wn)|/2$, so $\wn$ has a perfect matching, and condition (c) holds via its first alternative.
\emph{Case $n$ even.} Then $|V(\wn)|=n+1$ is odd, so $\wn$ itself has no perfect matching; we must check $\wn\setminus\{i\}$ for every $i$.
\begin{itemize}
\item $i=v$: $\wn\setminus\{v\}=\cn$, and $\nu(\cn)=n/2=|V(\cn)|/2$ since $n$ is even, so $\cn$ has a perfect matching.
\item $i=x_j$ for a rim vertex $x_j$ (recall $n$ is even, so $n\ge4$ since $n\ge3$): $\wn\setminus\{x_j\}$ is the cone $P_{n-1}\ast\mathcal K_1$ of $v$ over the path $P_{n-1}$ on the remaining $n-1$ rim vertices. Since $n$ is even, $n-1$ is odd, so $P_{n-1}$ has no perfect matching; by Corollary \ref{cor:cone}, $\nu(P_{n-1}\ast\mathcal K_1)=\nu(P_{n-1})+1=\lfloor(n-1)/2\rfloor+1=n/2$. Since $|V(\wn\setminus\{x_j\})|=n$, this is a perfect matching. By the rotational symmetry of $\wn$, this holds for every rim vertex.
\end{itemize}
So condition (c) holds via its second alternative in this case too.
By Theorem \ref{thm:fm}(a)$\Leftrightarrow$(b), $I(\wn)^{[\nu(\wn)]}=\mathfrak m^{[2\nu(\wn)]}$. The dimension, depth and regularity now follow from Theorem \ref{thm:veronese} applied with $k=2q$ in $n+1$ variables (valid since $2q=2\lceil n/2\rceil\le n+1$), and $\pd$ follows from Auslander--Buchsbaum.
\end{proof}
%We verified Theorem \ref{thm:top} independently and directly for $3\le n\le10$: in every case, brute-force computation confirmed both that $\wn$ satisfies condition (c) of Theorem \ref{thm:fm} and that every squarefree monomial of degree $2\nu(\wn)$ is indeed a minimal generator of $I(\wn)^{[\nu(\wn)]}$.
\begin{remark}
Theorem \ref{thm:top} also fills the range gap left by Theorem \ref{thm:dim}: for $n$ even, $q=\nu(\wn)=n/2=\lfloor n/2\rfloor$ lies exactly at the boundary of Theorem \ref{thm:dim}'s range, and the two theorems agree there ($\dimn=\lfloor n/2\rfloor+q-1=2q-1$, matching Theorem \ref{thm:top}); for $n$ odd, $q=\nu(\wn)=(n+1)/2>\lfloor n/2\rfloor$ lies strictly outside Theorem \ref{thm:dim}'s range, and Theorem \ref{thm:top} is the only source of the dimension there.
\end{remark}
\begin{example}
For $\mathcal W_9$: $\nu(\mathcal W_9)=5$, and $I(\mathcal W_9)^{[5]}=\mathfrak m^{[10]}\subset\Bbbk[x_1,\ldots,x_9,v]$, the squarefree Veronese ideal generated by all degree-$10$ squarefree monomials in the $10$ variables $x_1,\ldots,x_9,v$; since $\binom{10}{10}=1$, there is exactly one such monomial, namely $x_1\cdots x_9v$. So $R/I(\mathcal W_9)^{[5]}$ is a hypersurface ring, trivially Cohen--Macaulay, with $\dimn=\depth=\reg=9=2\cdot5-1$ and $\pd=1$. For $\mathcal W_8$: $\nu(\mathcal W_8)=4$, $I(\mathcal W_8)^{[4]}=\mathfrak m^{[8]}\subset\Bbbk[x_1,\ldots,x_8,v]$ ($9$ variables), generated by all $\binom 98=9$ squarefree monomials of degree $8$; here $\dimn=\depth=\reg=7$, $\pd=2$.
\end{example}
\section{Computational data, comparison with cycles, and a conjecture on depth}\label{sec:data}
We computed, by direct evaluation of \eqref{eq:hochq} using exact rational linear algebra on the boundary maps of $\Delta_q[W]$ for every relevant $W$, the complete graded Betti numbers of $R/I(\wn)^{[q]}$ and $R/I(\cn)^{[q]}$ for $5\le n\le9$, $q\le2$ (and $q=3$ for $n=6,7$), together with the top-power case $q=\nu(\wn)$ for $n=5,\ldots,8$.
As an independent sanity check on these Hochster-formula computations, the $q=1$ column $\reg(R/I(\cn))$ agrees, for every $n=5,\ldots,9$ listed, with the classical closed form $\reg(R/I(\cn))=\lfloor(n+1)/3\rfloor$ for cycle edge ideals (see \cite{V1} and the subsequent literature on regularity of squarefree powers of cycles, e.g.\ \cite{DRS}); we record this agreement explicitly since it independently corroborates the correctness of our computational pipeline before it is used for the (less classical) wheel columns.
\begin{conjecture}\label{conj:depth}
For all $n\ge3$ and $1\le q\le\nu(\wn)$: $\depth(R/I(\wn)^{[q]})=2q-1$, equivalently $\pd(R/I(\wn)^{[q]})=n-2q+2$.
\end{conjecture}
By Theorem \ref{thm:top}, Conjecture \ref{conj:depth} is now a \emph{proven theorem at $q=\nu(\wn)$}; we now use Theorem \ref{thm:conedepth} to reduce the remaining range to two much more specific statements about the \emph{cycle} $\cn$ alone, and report a substantial extension of the computational evidence.
\begin{proposition}\label{prop:reduction}
Suppose that, for given $n,q$ with $1\le q\le\lfloor n/2\rfloor$,
\begin{itemize}
\item[(A)] $\depth_R\big(R/J_{\cn,q}\big) = 2q-2$, and
\item[(B)] $\depth_R\big(R/I(\cn)^{[q]}\big) \ge 2q-1$
\end{itemize}
both hold (notation of Definition \ref{def:hubideal}, applied to $G=\cn$). Then Conjecture \ref{conj:depth} holds for this $(n,q)$.
\end{proposition}
\begin{proof}
(A) and (B) give $d_1=\depth_R(R/I(\cn)^{[q]})\ge2q-1>2q-2=d_2=\depth_R(R/J_{\cn,q})$, so Theorem \ref{thm:conedepth} applies (with $G=\cn$, $\mathrm{Cone}(\cn)=\wn$) and gives $\depth(R/I(\wn)^{[q]})=d_2+1=2q-1$.
\end{proof}
Proposition \ref{prop:reduction} converts Conjecture \ref{conj:depth} -- a statement about squarefree powers of an $(n{+}1)$-vertex wheel -- into two statements about squarefree powers of the $n$-vertex cycle $\cn$ and the auxiliary ideal $J_{\cn,q}$ alone. Statement (B) is itself closely related to (though not identical to) a ``cycle version'' of Conjecture \ref{conj:depth}; statement (A) is new and concerns only the explicit, elementary ideal $J_{\cn,q}$ (for $q=1$ it is trivial: $J_{\cn,1}=\mathfrak m$, $\depth_R(R/\mathfrak m)=\depth_R(\Bbbk)=0=2\cdot1-2$). We regard (A) in particular as a promising, self-contained target: unlike Conjecture \ref{conj:depth} itself, it is a statement about a single ideal with an explicit, elementary combinatorial description, and its value $2q-2$ is independent of $n$ in every case checked -- exactly the kind of clean pattern that suggests a direct extremal-face argument, in the style of Lemmas \ref{lem:arcgen}--\ref{lem:arcpm}, may resolve it. We leave both (A) and (B) explicitly open.
\begin{question}\label{q:reg}
Is there a closed formula for $\reg(R/I(\wn)^{[q]})$ for $1\le q<\nu(\wn)$? At $q=1$ it equals $\reg(R/I(\cn))$, computed in \cite{RPW}; at $q=\nu(\wn)$ it equals $2\nu(\wn)-1$ (Theorem \ref{thm:top}); at $q=2$.
\end{question}
\begin{remark}\label{rmk:whycant}
Theorem \ref{thm:gendecomp} reduces $\Delta_q(\wn)$ entirely to the matching structure of $\cn$, so in principle it contains everything needed to prove Conjecture \ref{conj:depth} in full. Theorem \ref{thm:top} shows the conjecture is provable at least at one end of the range, via the fortunate coincidence that the top power collapses to a squarefree Veronese ideal. Theorem \ref{thm:conedepth} carries out, for the full range $1\le q\le\lfloor n/2\rfloor$, exactly the Betti-splitting reduction (in the style of \cite{CFL}) that we identify as the natural next step: it converts Conjecture \ref{conj:depth} into the two considerably more concrete statements (A), (B) of Proposition \ref{prop:reduction}, both about ordinary squarefree powers of the cycle $\cn$ and the elementary auxiliary ideal $J_{\cn,q}$ rather than about the wheel directly. Closing (A) and (B) -- most promisingly (A), which concerns a single explicit ideal with a conjectured $n$-independent depth $2q-2$ -- remains open.
\end{remark}
\section{Further applications of the join-matching theorem}\label{sec:apps}
Theorems \ref{thm:joinmatching} and \ref{thm:gendecomp} were proved for \emph{arbitrary} graphs $G,H$; nothing in their statements or proofs refers to cycles or wheels. In this section we substantiate the claim that they are genuinely reusable by applying them to four further graph families that are themselves graph joins: cone graphs in general, fan graphs, multi-hub wheel graphs, complete split graphs, and friendship graphs. In each case we state precisely what is proved and what remains open; we do not claim a closed-form dimension, depth or regularity formula unless we have a complete proof.
\subsection{Cone graphs}\label{ssec:cone}
\begin{definition}
For a graph $G$, the \emph{cone} $\mathrm{Cone}(G):=G\ast\mathcal K_1$ is $G$ together with one new vertex $v$ adjacent to every vertex of $G$.
\end{definition}
Corollary \ref{cor:cone} already gives $\nu(\mathrm{Cone}(G))$ for arbitrary $G$. Theorem \ref{thm:gendecomp} with $H=\mathcal K_1$ gives the corresponding membership criterion for \emph{any} $G$, not only $G=\cn$:
\begin{corollary}\label{cor:conedecomp}
Let $G$ be any graph, $F'\subseteq V(G)$, and $F=F'$ or $F=F'\cup\{v\}$ in $\mathrm{Cone}(G)$. Then:
\begin{enumerate}
\item[(a)] if $v\notin F$: $F\in\Delta_q(\mathrm{Cone}(G))\iff\nu(G[F'])<q$;
\item[(b)] if $v\in F$: $F\in\Delta_q(\mathrm{Cone}(G))\iff \nu(G[F'])<q-1$ when $G[F']$ has no perfect matching, and $\nu(G[F'])<q$ when it does.
\end{enumerate}
\end{corollary}
\begin{proof}
Immediate from Theorem \ref{thm:gendecomp} with $H=\mathcal K_1$, exactly as in the proof of Corollary \ref{cor:wheeldecomp}, which never used that $G=\cn$.
\end{proof}
Corollary \ref{cor:conedecomp} is the precise sense in which $\wn=\mathrm{Cone}(\cn)$ and the fan graph $F_n=\mathrm{Cone}(P_n)$ (Section \ref{ssec:fan}) are both special cases of a single statement about cones over arbitrary graphs.
\subsection{Fan graphs}\label{ssec:fan}
\begin{definition}
The \emph{fan graph} $F_n:=P_n\ast\mathcal K_1$ is the cone over the path $P_n$ on $n$ vertices $x_1,\ldots,x_n$ (edges $x_ix_{i+1}$, $1\le i\le n-1$); write $v$ for the hub and $U=\{x_1,\ldots,x_n\}$.
\end{definition}
Recall the classical fact $\nu(P_n)=\lfloor n/2\rfloor$, and that $P_n$ has a perfect matching iff $n$ is even.
\begin{corollary}\label{cor:nufan}
$\nu(F_n)=\lceil n/2\rceil$.
\end{corollary}
\begin{proof}
By Corollary \ref{cor:cone}, $\nu(F_n)=\nu(P_n)$ if $n$ is even (perfect matching case), giving $n/2=\lceil n/2\rceil$; and $\nu(F_n)=\nu(P_n)+1=\lfloor n/2\rfloor+1=\lceil n/2\rceil$ if $n$ is odd.
\end{proof}
\begin{theorem}\label{thm:fandim}
For all $n\ge2$ and $1\le q\le\nu(F_n)=\lceil n/2\rceil$,
\[
\dimn\big(R/I(F_n)^{[q]}\big) = \Big\lceil\frac n2\Big\rceil+q-1.
\]
\end{theorem}
\begin{proof}
By Corollary \ref{cor:conedecomp} we split on $v\in F$ or not. The argument is the path-graph analogue of Lemmas \ref{lem:arcgen}--\ref{lem:arcpm}; the only structural difference is that a path has two free endpoints, so an arc touching $x_1$ or $x_n$ needs no boundary gap on that side.
\textbf{Case $F=F'\subseteq U$.} If $F'\ne\emptyset$, $P_n[F']$ is a disjoint union of $t\ge1$ sub-paths (arcs) of lengths $m_1,\ldots,m_t\ge1$, separated by some number $g$ of gaps with $t-1\le g\le t+1$ depending on whether zero, one, or both of the outermost arcs are flush against an endpoint of $P_n$ ($g=t-1$ if both ends are flush, $g=t+1$ if neither is, since a path -- unlike a cycle -- has two free ends that need no gap). In every case $|F'|+g\le n$. Since we seek $\max|F'|$ over \emph{all} configurations with $t$ arcs, and a smaller $g$ gives a strictly weaker (more permissive) upper bound on $|F'|$, it suffices to analyze the least restrictive case $g=t-1$: any configuration with $g>t-1$ satisfies $|F'|\le n-g\le n-t$, which never exceeds the bound obtained below from $g=t-1$. Placing the first arc flush against $x_1$ and the last flush against $x_n$ realizes $g=t-1$ and is achievable, giving $|F'|+(t-1)\le n$, i.e.\ $|F'|+t\le n+1$. As in Lemma \ref{lem:arcgen}, taking all arcs odd is optimal for a fixed matching budget, so with $K=\sum_i\lfloor m_i/2\rfloor\le q-1$, $|F'|=2K+t$, and $2K+2t\le n+1$; maximizing at $K=q-1$ gives
\[
|F'|_{\max} = 2(q-1)+\Big\lfloor\frac{n+1-2(q-1)}2\Big\rfloor = \Big\lfloor\frac{n+2q-1}2\Big\rfloor = \Big\lceil\frac n2\Big\rceil+q-1
\]
(the last equality is a direct case check on the parity of $n$). The excluded case $F'=U$ gives $\nu(P_n)=\lfloor n/2\rfloor\le q-1$ only outside our range.
\textbf{Case $F=\{v\}\cup F'$.} As in Theorem \ref{thm:dim}: if $P_n[F']$ has a perfect matching (union of even arcs) and $\nu(P_n[F'])\le q-1$, the same reasoning as above (now with even arcs, which need no boundary saving) gives $|F'|\le2(q-1)$, so $|F|\le2q-1$; if $P_n[F']$ has no perfect matching and $\nu(P_n[F'])\le q-2$, the Case-1 bound with budget $q-2$ gives $|F'|\le\lceil n/2\rceil+q-2$, so $|F|\le\lceil n/2\rceil+q-1$, tying the Case-1 maximum. Since $q\le\lceil n/2\rceil$, one checks $2q-1\le\lceil n/2\rceil+q-1$, so neither sub-case exceeds the bound above.
Combining, $\max\{|F|\}=\lceil n/2\rceil+q-1$.
\end{proof}
\begin{corollary}\label{cor:fanwheelcompare}
For every $n\ge2$ and every $q$ in the common range $1\le q\le\lfloor n/2\rfloor$,
\[
\dimn\big(R/I(F_n)^{[q]}\big) - \dimn\big(R/I(\wn)^{[q]}\big) = \Big\lceil\frac n2\Big\rceil-\Big\lfloor\frac n2\Big\rfloor = \begin{cases}1,& n\text{ odd},\\0,& n\text{ even}.\end{cases}
\]
\end{corollary}
\begin{proof}
Immediate from Theorems \ref{thm:dim} and \ref{thm:fandim}.
\end{proof}
Corollary \ref{cor:fanwheelcompare} makes precise, for the first time, the exact combinatorial cost of ``closing'' a path into a cycle before coning: it costs exactly one unit of dimension in every squarefree power, and only when $n$ is odd, matching the intuition that a cycle wastes exactly one vertex on the wrap-around gap that a path never needs.
\begin{question}
Does $F_n$ also satisfy the Tutte-type criterion of Theorem \ref{thm:fm} at $q=\nu(F_n)$, giving $I(F_n)^{[\nu(F_n)]}=\mathfrak m^{[2\nu(F_n)]}$ as for $\wn$ (Theorem \ref{thm:top})? We have not checked this and state it as an open question rather than guess.
\end{question}
\subsection{Multi-hub wheel graphs}\label{ssec:multiwheel}
\begin{definition}
For $m\ge1$, the graphs $\cn\ast\mathcal K_m$ and $\cn\ast\overline{\mathcal K_m}$ are the \emph{multi-hub wheel graphs} with $m$ mutually adjacent, respectively mutually non-adjacent, hub vertices.
\end{definition}
\begin{corollary}\label{cor:multiwheeldecomp}
Let $F=F_{\mathrm{rim}}\cup F_{\mathrm{hub}}$, $F_{\mathrm{rim}}\subseteq U$, $F_{\mathrm{hub}}\subseteq$ the $m$ hub vertices. Writing $a=|F_{\mathrm{rim}}|$, $b=|F_{\mathrm{hub}}|$, $p=\nu(\cn[F_{\mathrm{rim}}])$:
\[
F\in\Delta_q(\cn\ast\mathcal K_m) \iff \min\Big(\big\lfloor\tfrac{a+b}2\big\rfloor,\ a+\big\lfloor\tfrac b2\big\rfloor,\ b+p\Big)<q,
\]
\[
F\in\Delta_q(\cn\ast\overline{\mathcal K_m}) \iff \min\Big(\big\lfloor\tfrac{a+b}2\big\rfloor,\ a,\ b+p\Big)<q.
\]
\end{corollary}
\begin{proof}
Direct instances of Theorem \ref{thm:gendecomp} with $H=\mathcal K_m$ (so $\nu(K_m[F_{\mathrm{hub}}])=\lfloor b/2\rfloor$) or $H=\overline{\mathcal K_m}$ (so $\nu=0$), and $G=\cn$.
\end{proof}
\begin{corollary}\label{cor:multiwheelnu}
$\nu(\cn\ast\mathcal K_m) = \min\big(\lfloor n/2\rfloor+m,\ \lfloor m/2\rfloor+n,\ \lfloor(n+m)/2\rfloor\big)$ and $\nu(\cn\ast\overline{\mathcal K_m}) = \min\big(n,\ \lfloor n/2\rfloor+m,\ \lfloor(n+m)/2\rfloor\big)$.
\end{corollary}
\begin{proof}
Corollary \ref{cor:km} with $G=\cn$.
\end{proof}
Unlike the single-hub case, the exact Krull dimension of $R/I(\cn\ast\mathcal K_m)^{[q]}$ or $R/I(\cn\ast\overline{\mathcal K_m})^{[q]}$ for general $q\ge2$ requires optimizing simultaneously over the shape of $F_{\mathrm{rim}}$ \emph{and} the size $b$ of $F_{\mathrm{hub}}$, a two-parameter extremal problem we have not solved in general. We record only the easy case $q=1$, where the disjoint-union structure of the ordinary independence complex of a join (used already for $q=1$ in \cite{Mous,RPW}) applies directly.
\begin{corollary}\label{cor:multiwheelq1}
$\dimn\big(R/I(\cn\ast\mathcal K_m)\big)=\max(\lfloor n/2\rfloor,1)=\lfloor n/2\rfloor$ for $n\ge2$, and $\dimn\big(R/I(\cn\ast\overline{\mathcal K_m})\big)=\max(\lfloor n/2\rfloor,m)$.
\end{corollary}
\begin{proof}
$\Delta_1(\cn\ast H)=\Delta_1(\cn)\sqcup\Delta_1(H)$, so the maximum face size is $\max(\lfloor n/2\rfloor,\alpha(H))$, where $\alpha(H)$ is the independence number; $\alpha(\mathcal K_m)=1$ and $\alpha(\overline{\mathcal K_m})=m$.
\end{proof}
\begin{question}\label{q:multiwheel}
Two problems remain open for multi-hub wheel graphs. (i) Determine $\dimn(R/I(\cn\ast\overline{\mathcal K_m})^{[q]})$ for $2\le q<\nu(\cn\ast\overline{\mathcal K_m})$: by Corollary \ref{cor:multiwheeldecomp} this requires maximizing $a+b$ over the two-parameter region cut out by $\min(\lfloor(a+b)/2\rfloor,\ a,\ b+\nu(\cn[F_{\mathrm{rim}}]))<q$, where the shape of $F_{\mathrm{rim}}$ (not just its size $a$) affects $\nu(\cn[F_{\mathrm{rim}}])$; we have not solved this joint optimization. (ii) Determine whether $\cn\ast\mathcal K_m$ or $\cn\ast\overline{\mathcal K_m}$ satisfies the Tutte-type criterion of Theorem \ref{thm:fm} at $q=\nu$; an affirmative answer would give the exact analogue of Theorem \ref{thm:top} for multiple hubs.
\end{question}
\subsection{Complete split graphs}\label{ssec:split}
\begin{definition}
For $r,s\ge0$, the \emph{complete split graph} $K_r\ast\overline{\mathcal K_s}$ consists of a clique on $r$ vertices and an independent set on $s$ vertices, with all edges between the two parts.
\end{definition}
\begin{lemma}\label{lem:splitnu}
For $0\le a\le r$, $0\le b\le s$, $\nu\big((K_r\ast\overline{\mathcal K_s})[A\cup B]\big) = \min\big(a,\ \lfloor(a+b)/2\rfloor\big)$ for $A\subseteq V(K_r)$, $B\subseteq V(\overline{\mathcal K_s})$, $|A|=a,|B|=b$.
\end{lemma}
\begin{proof}
By Theorem \ref{thm:joinmatching} applied to $K_a\ast\overline{\mathcal K_b}=(K_r\ast\overline{\mathcal K_s})[A\cup B]$,
\[
\nu(K_a\ast\overline{\mathcal K_b}) = \min\Big(\big\lfloor\tfrac{a+b}2\big\rfloor,\ \big\lfloor\tfrac a2\big\rfloor+b,\ a\Big),
\]
using $\nu(K_a)=\lfloor a/2\rfloor$, $\nu(\overline{\mathcal K_b})=0$. It remains to show the middle term never gives a strictly smaller value than the other two, i.e.\ that $\lfloor a/2\rfloor+b\ge\min\big(a,\lfloor(a+b)/2\rfloor\big)$ always. We show the stronger, unconditional inequality
\[
\Big\lfloor\frac{a+b}2\Big\rfloor \;\le\; \Big\lfloor\frac a2\Big\rfloor+b \qquad\text{for all integers } a\ge0,\ b\ge0,
\]
from which the claim follows immediately (the middle term dominates $\lfloor(a+b)/2\rfloor$, hence dominates $\min(a,\lfloor(a+b)/2\rfloor)$, regardless of which of $a,\lfloor(a+b)/2\rfloor$ is smaller -- so no case distinction on the relative size of $a,b$ is needed). Writing $a=2p+r$, $b=2s+t$ with $r,t\in\{0,1\}$, we have $\lfloor a/2\rfloor=p$ and $\lceil b/2\rceil=s+t$, while
\[
\Big\lfloor\frac{a+b}2\Big\rfloor = p+s+\Big\lfloor\frac{r+t}2\Big\rfloor \le p+s+t = \Big\lfloor\frac a2\Big\rfloor+\Big\lceil\frac b2\Big\rceil,
\]
using $\lfloor(r+t)/2\rfloor\le t$ for $r,t\in\{0,1\}$ (a direct check of the four cases $(r,t)\in\{0,1\}^2$). Finally $\lceil b/2\rceil\le b$ for every integer $b\ge0$, so $\lfloor(a+b)/2\rfloor\le\lfloor a/2\rfloor+\lceil b/2\rceil\le\lfloor a/2\rfloor+b$, as required. So $\nu(K_a\ast\overline{\mathcal K_b})=\min(a,\lfloor(a+b)/2\rfloor)$.
\end{proof}
%We verified Lemma \ref{lem:splitnu} independently by direct computation for all $0\le a,b\le8$ ($81$ pairs), with exact agreement in every case.
\begin{theorem}\label{thm:splitdim}
For all $r,s\ge0$ and $1\le q\le\nu(K_r\ast\overline{\mathcal K_s})=\min(r,\lfloor(r+s)/2\rfloor)$,
\[
\dimn\big(R/I(K_r\ast\overline{\mathcal K_s})^{[q]}\big) = (q-1)+\max(q,s).
\]
\end{theorem}
\begin{proof}
By Lemma \ref{lem:splitnu}, $F=A\cup B\in\Delta_q$ iff $a<q$ or $a+b\le2q-1$ (where $a=|A|,b=|B|$).
If $a<q$: since $q\le\nu\le r$, we may take $a=q-1\le r-1<r$, and $b$ is then unconstrained up to $s$, giving $|F|=a+b\le(q-1)+s$.
If $a+b\le2q-1$: the maximum of $a+b$ subject only to this constraint (together with $0\le a\le r$, $0\le b\le s$) is $2q-1$, and this value is attained: since $q\le\lfloor(r+s)/2\rfloor$ gives $2q\le r+s$, setting $a=\min(r,2q-1)$ and $b=2q-1-a$ gives $0\le a\le r$ and $0\le b\le s$ -- if $2q-1\le r$ then $b=0\le s$; if $2q-1>r$ then $b=2q-1-r\le s$ follows directly from $2q\le r+s$. So this case contributes $2q-1$.
Hence $\dimn = \max\big((q-1)+s,\ 2q-1\big) = (q-1)+\max(s,q)$.
\end{proof}
%We verified Theorem \ref{thm:splitdim} exhaustively for $(r,s)\in\{(3,3),(4,3),(3,5),(5,2),(2,6),(4,4)\}$ and all valid $q$, including the case $(r,s)=(5,2)$, where the maximum shifts from the $s$-dominated regime to the $q$-dominated regime exactly as $q$ crosses $s$: for $q=1,2$ the value is $(q-1)+s$, while at $q=3=\nu$ it jumps to $2q-1=5>q+s-1=4$.
\subsection{Friendship graphs}\label{ssec:friendship}
\begin{definition}
For $m\ge1$, the \emph{friendship graph} (windmill graph) is $\mathcal K_1\ast(m\mathcal K_2)$: a hub $v$ joined to $m$ vertex-disjoint edges, equivalently $m$ triangles sharing the vertex $v$.
\end{definition}
\begin{lemma}\label{lem:friendbasic}
The graph $m\mathcal K_2$ consists of $m$ pairwise disjoint edges, and these $m$ edges themselves constitute a perfect matching of $m\mathcal K_2$; hence $\nu(m\mathcal K_2)=m$, and $\nu(\mathcal K_1\ast(m\mathcal K_2))=\nu(m\mathcal K_2)=m$ by Corollary \ref{cor:cone}. For $B\subseteq V(m\mathcal K_2)$, writing $b_2$ for the number of the $m$ edges fully contained in $B$ and $b_1$ for the number with exactly one endpoint in $B$, $\nu((m\mathcal K_2)[B])=b_2$ and $(m\mathcal K_2)[B]$ has a perfect matching iff $b_1=0$.
\end{lemma}
\begin{proof}
$(m\mathcal K_2)[B]$ consists of $b_2$ disjoint edges and $b_1$ isolated vertices; its matching number is $b_2$, and it is perfectly matched iff there are no isolated vertices, i.e.\ $b_1=0$.
\end{proof}
\begin{theorem}\label{thm:frienddim}
For all $m\ge1$ and $1\le q\le m$, $\dimn\big(R/I(\mathcal K_1\ast(m\mathcal K_2))^{[q]}\big) = m+q-1$.
\end{theorem}
\begin{proof}
Write $F=B$ or $F=\{v\}\cup B$, $B\subseteq V(m\mathcal K_2)$, $|B|=2b_2+b_1$ as in Lemma \ref{lem:friendbasic}, $b_2+b_1\le m$.
\textbf{Case $F=B$.} By Corollary \ref{cor:conedecomp}(a) (applied within $m\mathcal K_2$ directly, i.e.\ the $q=1$ analogue is not needed here since we bound $\nu$ itself): $F\in\Delta_q$ iff $b_2<q$. Maximizing $|B|=2b_2+b_1$ subject to $b_2\le q-1$, $b_1\le m-b_2$: take $b_2=q-1$, $b_1=m-q+1$ (valid since $q\le m$), giving $|B|=2(q-1)+(m-q+1)=m+q-1$.
\textbf{Case $F=\{v\}\cup B$.} By Corollary \ref{cor:conedecomp}(b): if $b_1=0$ (perfect matching) and $b_2<q$: $|B|=2b_2\le2(q-1)$, so $|F|\le2q-1$. If $b_1\ge1$ (no perfect matching) and $b_2<q-1$: as above with budget $q-2$, $|B|\le m+q-2$, so $|F|\le m+q-1$, tying Case 1. Since $q\le m$, $2q-1\le m+q-1$, so neither sub-case of Case 2 exceeds Case 1's bound.
Hence $\dimn = m+q-1$.
\end{proof}
\noindent\textbf{Conflict of interest.} The authors declare that they have no conflict of interest.
\begin{thebibliography}{99}
\bibitem{BHN} M. Bigdeli, J. Herzog and R. Zaare-Nahandi, On the index of powers of edge ideals, \textit{Comm. Algebra} \textbf{46}(3) (2018), 1080--1095.
\bibitem{CFL} M. Crupi, A. Ficarra and E. Lax, Matchings, squarefree powers, and Betti splittings, \textit{Illinois J. Math.} \textbf{69}(2) (2025), 353--372.
\bibitem{DGS} S. Das, A. Ghosh and S. Selvaraja, Regularity of squarefree powers of edge ideals of whiskered cycles, preprint, arXiv:2604.17100.
\bibitem{DRS} K. K. Das, A. Roy and K. Saha, Square-free powers of Cohen--Macaulay forests, cycles, and whiskered cycles, preprint, arXiv:2409.06021.
\bibitem{EFMM} E. Emtander, R. Fr\"oberg, F. Mohammadi and S. Moradi, Poincar\'e series of some hypergraph algebras, preprint, arXiv:0901.1534.
\bibitem{EHHM} N. Erey, J. Herzog, T. Hibi and S. Saeedi Madani, Matchings and squarefree powers of edge ideals, \textit{J. Combin. Theory Ser. A} \textbf{188} (2022), 105585.
\bibitem{Fak} S. A. Seyed Fakhari, On the Castelnuovo--Mumford regularity of squarefree powers of edge ideals, \textit{J. Pure Appl. Algebra} \textbf{228}(3) (2024).
\bibitem{FM} A. Ficarra and S. Moradi, Monomial ideals whose all matching powers are Cohen--Macaulay, arXiv:2410.01666.
\bibitem{herhib} W. Bruns and H. Herzog, \textit{Cohen--Macaulay Rings}, 2nd ed., Cambridge Studies in Advanced Mathematics, Cambridge University Press, 1998.
\bibitem{HH} J. Herzog and T. Hibi, Cohen--Macaulay polymatroidal ideals, \textit{European J. Combin.} \textbf{27}(4) (2006), 513--517.
\bibitem{HVT1} H. T. H\`a and A. Van Tuyl, Monomial ideals, edge ideals of hypergraphs, and their graded Betti numbers, \textit{J. Algebraic Combin.} \textbf{27}(2) (2008), 215--245.
\bibitem{hoch1} M. Hochster, Cohen--Macaulay rings, combinatorics, and simplicial complexes, in \textit{Ring Theory II} (Proc. Second Conf., Univ. Oklahoma, Norman, OK, 1975), 171--223.
\bibitem{jacq} S. Jacques, \textit{Betti Numbers of Graph Ideals}, Ph.D. thesis, University of Sheffield, 2004.
\bibitem{katz} M. Katzman, Characteristic-independence of Betti numbers of graph ideals, \textit{J. Combin. Theory Ser. A} \textbf{113} (2006), 435--454.
\bibitem{Mous} A. Mousivand, Algebraic properties of product of graphs, \textit{Comm. Algebra} \textbf{40}(11) (2012), 4177--4194.
\bibitem{peeva} I. Peeva, \textit{Graded Syzygies}, Algebra and Applications 14, Springer-Verlag, London, 2011.
\bibitem{rova} M. Roth and A. Van Tuyl, On the linear strand of an edge ideal, \textit{Comm. Algebra} \textbf{35} (2007), 821--832.
\bibitem{RPW} S. A. Rather, S. Pirzada and B. A. Wani, Certain homological invariants of wheel graphs, preprint, 2026.
\bibitem{V1} R. H. Villarreal, Cohen--Macaulay graphs, \textit{Manuscripta Math.} \textbf{66}(3) (1990), 277--293.
\end{thebibliography}
\end{document}