Weakly-sparse and strongly flip-flat classes of graphs are uniformly almost-wide
Extracted source text ยท arXiv 2512.04610
%\documentclass[11pt]{article}
%\usepackage[margin=1in]{geometry}
\documentclass[a4paper,UKenglish,autoref, unicode,hidelinks, thm-restate]{lipics-v2021}
\nolinenumbers
\input{packages.tex}
\usepackage{microtype}
\input{macros.tex}
\bibliographystyle{plainurl}% the mandatory bibstyle
%\newcommand{\funding}{
%
% }
\title{Weakly-sparse and strongly flip-flat classes of graphs are uniformly almost-wide} %TODO Please add
\titlerunning{Sparse strongly flip-flats are almost-wide } %TODO optional, please use if title is longer than one line
\author{Fatemeh Ghasemi}{LACL, Universit\'{e} Paris-Est Cr\'{e}teil, France}{fatemeh.ghasemi@lacl.fr}{}{}
\author{Julien Grange}{LACL, Universit\'{e} Paris-Est Cr\'{e}teil, France}{julien.grange@lacl.fr}{https://orcid.org/0009-0005-0470-1781}{}
\author{Mamadou Moustapha Kant\'{e}}{Universit\'{e} Clermont Auvergne, Clermont Auvergne INP, LIMOS, CNRS, Aubi\`{e}re, France}{mamadou.kante@uca.fr}{https://orcid.org/0000-0003-1838-7744}{}
\author{Florent Madelaine}{LACL, Universit\'{e} Paris-Est Cr\'{e}teil, France}{florent.madelaine@u-pec.fr}{https://orcid.org/0000-0002-8528-7105}{}%TODO mandatory, please use full name; only 1 author per \author macro; first two parameters are mandatory, other parameters can be empty. Please provide at least the name of the affiliation and the country. The full address is optional. Use additional curly braces to indicate the correct name splitting when the last name consists of multiple name parts.
\authorrunning{F. Ghasemi, J. Grange, M.M. Kant\'{e}, F. Madelaine} %TODO mandatory. First: Use abbreviated first/middle names. Second (only in severe cases): Use first author plus 'et al.'
\Copyright{Jane Open Access and Joan R. Public} %TODO mandatory, please use full first names. LIPIcs license is "CC-BY"; http://creativecommons.org/licenses/by/3.0/
%\ccsdesc[100]{} %TODO mandatory: Please choose ACM 2012 classifications from https://dl.acm.org/ccs/ccs_flat.cfm
\ccsdesc[500]{Theory of computation~Finite Model Theory}
\ccsdesc[500]{Mathematics of computing~Graph algorithms}
\keywords{Almost-wide, Flip-flatness} %TODO mandatory; please add comma-separated list of keywords
\category{} %optional, e.g. invited paper
\relatedversion{} %optional, e.g. full version hosted on arXiv, HAL, or other respository/website
%\relatedversiondetails[linktext={opt. text shown instead of the URL}, cite=DBLP:books/mk/GrayR93]{Classification (e.g. Full Version, Extended Version, Previous Version}{URL to related version} %linktext and cite are optional
%\supplement{}%optional, e.g. related research data, source code, ... hosted on a repository like zenodo, figshare, GitHub, ...
%\supplementdetails[linktext={opt. text shown instead of the URL}, cite=DBLP:books/mk/GrayR93, subcategory={Description, Subcategory}, swhid={Software Heritage Identifier}]{General Classification (e.g. Software, Dataset, Model, ...)}{URL to related version} %linktext, cite, and subcategory are optional
%\funding{(Optional) general funding statement \dots}%optional, to capture a funding statement, which applies to all authors. Please enter author specific funding statements as fifth argument of the \author macro.
%\acknowledgements{I want to thank \dots}%optional
%\nolinenumbers %uncomment to disable line numbering
%Editor-only macros:: begin (do not touch as author)%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\EventEditors{John Q. Open and Joan R. Access}
\EventNoEds{2}
\EventLongTitle{42nd Conference on Very Important Topics (CVIT 2016)}
\EventShortTitle{CVIT 2016}
\EventAcronym{CVIT}
\EventYear{2016}
\EventDate{December 24--27, 2016}
\EventLocation{Little Whinging, United Kingdom}
\EventLogo{}
\SeriesVolume{42}
\ArticleNo{23}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{document}
\maketitle
\input{chapters/abstract.tex}
\input{chapters/Introduction.tex}
\input{chapters/Preliminaries.tex}
\input{chapters/Main-result2.tex}
%\input{chapters/Main-result.tex}
\input{chapters/Conclusion.tex}
%\input{chapters/Flippergame-boundedround.tex}
%\input{reviewers.tex}
%\input{reviews2.tex}
\bibliography{Bibliography.bib}
\end{document}
% \usepackage[hidelinks]{hyperref}
%\usepackage[dvipsnames]{xcolor}
\usepackage{mathrsfs}
\usepackage{xcolor}
\usepackage{amsmath}
\usepackage{amssymb}
\usepackage{amsthm}
\usepackage{thm-restate}
\usepackage{mdframed}
\usepackage{mathtools}
%\usepackage[hidelinks]{hyperref}
\usepackage[capitalize]{cleveref}
\usepackage{float}
\usepackage{wrapfig}
\usepackage{cancel}
\usepackage{xspace}
\usepackage[textsize=tiny]{todonotes}
\usepackage[all,defaultlines=3]{nowidow}
%\usepackage{space}
\newcommand{\N}{\mathbb{N}}
\mdfdefinestyle{myenvs}{%
hidealllines=true,%
nobreak=true, % comment this to allow breaking
leftmargin=0pt,
rightmargin=0pt,
innerleftmargin=0pt,
innerrightmargin=0pt,
}
%\newmdtheoremenv[style=myenvs]{lemma}{Lemma}[section]
%\newmdtheoremenv[style=myenvs]{corollary}[lemma]{Corollary}
%\newmdtheoremenv[style=myenvs]{theorem}[lemma]{Theorem}
%\newmdtheoremenv[style=myenvs]{fact}[lemma]{Fact}
%\newmdtheoremenv[style=myenvs]{proposition}[lemma]{Proposition}
%\newmdtheoremenv[style=myenvs]{conjecture}[lemma]{Conjecture}
%\newmdtheoremenv[style=myenvs]{definition}[lemma]{Definition}
%\newmdtheoremenv[style=myenvs]{claim}[lemma]{Claim}
%\newmdtheoremenv[style=myenvs]{observation}[lemma]{Observation}
% \newenvironment{claimproof}[1][\proofname]{%
% \renewcommand{\qedsymbol}{$\blacksquare$}%
% \begin{proof}[#1]%
% }{%
% \end{proof}%
% }
% \numberwithin{table}{section}
% \numberwithin{lemma}{section}
% \numberwithin{theorem}{section}
% \numberwithin{claim}{section}
% \numberwithin{corollary}{section}
% \numberwithin{definition}{section}
\usepackage{dirtytalk}
\newcommand{\Cc}[0]{\ensuremath{\mathscr{C}}\xspace}
\newcommand{\Dd}[0]{\ensuremath{\mathscr{D}}\xspace}
\newcommand{\qr}[0]{\mathrm{qr}}
\newcommand{\CC}[0]{\mathrm{\mathscr{C}}}
\newcommand{\DD}[0]{\mathrm{\mathscr{D}}}
\newcommand{\NN}[0]{\mathrm{\mathbb{N}}}
\newcommand{\LL}[0]{\mathrm{\mathcal{L}}}
\newcommand{\PP}[0]{\mathrm{\mathcal{P}}}
\newcommand{\C}{\mathcal{C}}
\newcommand{\Dcal}{\ensuremath{\mathcal{D}}}
\newcommand{\Tcal}{\ensuremath{\mathcal{T}}}
\newcommand{\struc}[1]{\mathbb{#1}}
\newcommand{\flip}[1]{\mathsf{#1}}
\newcommand{\FF}{\mathcal{F}}
\newcommand{\GG}{\mathcal{G}}
\newcommand{\KK}{\mathcal{K}}
\newcommand{\HH}{\mathcal{H}}
\newcommand{\EE}{\mathcal{E}}
\renewcommand{\phi}{\varphi}
\renewcommand{\le}{\leqslant}
\renewcommand{\ge}{\geqslant}
\renewcommand{\leq}{\le}
\renewcommand{\geq}{\ge}
\newcommand{\FO}{\mathrm{FO}}
\newcommand{\Oof}{\mathcal{O}}
\newcommand{\sless}{{\scriptscriptstyle <}}
\newcommand{\sgreater}{{\scriptscriptstyle >}}
\newcommand{\sdc}[2]{\mathrm{sd}_{#1}(K_{#2})}
\newcommand{\trans}[1]{\mathsf{#1}}
\newcommand{\interp}[1]{\mathsf{#1}}
\newcommand{\tp}[1]{\mathrm{tp}(#1)}
\newcommand{\degree}[1]{\mathsf{deg}(#1)}
\newcommand{\ramsey}[2]{\ensuremath{\mathscr{R}(#1,#2)}\xspace}
\newcommand{\fo}{\textsc{FO}\xspace}
\newcommand{\np}{\ensuremath{\mathsf{NP}}\xspace}
\def\ie{\emph{i.e.}\xspace}
\newtheorem*{maintheorem}{Main Theorem}
\begin{abstract}
In this work we take a step towards characterising \emph{strongly flip-flat} classes of graphs.
Strong flip-flatness appears to be the analogue of uniform almost-wideness in the setting of dense classes of graphs.
We prove that strongly flip-flat classes of graphs that are \emph{weakly sparse} are indeed uniformly almost-wide.
\end{abstract}
\section{Introduction}
When Trakhtenbrot showed that the set of \emph{first-order} formulas (\fo formulas for short) that are valid on finite structures is \emph{co-recursively enumerable}, but not \emph{recursively enumerable}, a beam of light was shed on the curious inversion of finite structures and infinite structures; as the same set is recursively enumerable (G\"odel's completeness), but not co-recursively enumerable (Church's theorem) on structures. This gave birth to finite model theory \cite{ebbinghaus2006finite}
%\cite{ebbinghaus2006finite,libkin2004elements}
and boosted the motivation to show its differences with the world of infinites.
Some classical results, such as the \emph{compactness theorem}, almost immediately broke down; some, like the \emph{L\"owenheim-Skolem theorem}, became meaningless; and some remained valid in finite structures, such as \emph{Ehrenfeuchet-Fra\"iss\'e} games. Moreover, many results were discovered to be native of finites and not hold in the infinites such as Fagin showing that on finites structures $\Sigma_1^1= \np$ \cite{baldwin2000finite}.
%\cite{kolaitis1993tutorial,rosen2002some,baldwin2000finite}).
A category of theorems from classical model theory are \emph{preservation} theorems. They describe the relation between syntactic and semantic properties of first-order formulas. \emph{\L{}o\'s-Tarski, Lyndon's positivity} and \emph{Homomorphism preservation theorem} are examples of preservation theorems \cite{hodges1993model}. The \emph{\L{}o\'s-Tarski} theorem, also known as \emph{extension preservation theorem}, which asserts that a first-order formula is preserved under embeddings between all structures if and only if it is equivalent to an existential formula, was shown to fail in the finite from early on \cite{tait1959counterexample}. On the contrary, Rossman showed that the homomorphism preservation theorem asserting that a formula is preserved by homomorphisms if and only if it is existential-positive, holds on \emph{finite structures} \cite{rossman2008homomorphism}. Before Rossman, Atserias et al. attempted to show positive results regarding both extension and homomorphism preservation by further restricting finite structures to tame classes defined by sparsity conditions
%, which allows for methods based on the locality of first-order logic
(with additional closure conditions)~\cite{atserias2008preservation,atserias2006preservation,dawar2024preservation,dawar2010homomorphism}, and the notions such
as \emph{wideness, quasi-wideness}, \emph{almost-wideness} and their uniform versions were introduced for that purposes.
% \todo[inline]{why uniformly is missing here? It is also missing in all places where almost wide is used}
These notions proved to be useful in other contexts. Informally, a class of graphs $\mathscr{C}$ is said quasi-wide if, for every integer $r$ and in any large enough graph $G$
from $\mathscr{C}$, we can find a small set of vertices $S$ and a large enough set $B\subseteq V(G)\setminus S$ such that any two vertices in $B$ are at
distance at least $r+1$ in $G\setminus S$, \ie, the induced graph on $V(G)\setminus S$. Being \emph{uniformly quasi-wide}, means that the above statement is true not only for
any large enough graph, but also for their large enough subgraphs. The well-known notion of \emph{nowhere dense} classes of graphs are characterised to be
exactly the uniformly quasi-wide classes of graphs \cite{nevsetvril2012sparsity}, and
this characterisation lies at the heart of another combinatorial characterisation of nowhere dense classes of graphs through the notion of \emph{splitter
game}, which helped proving that \emph{$\fo$ model checking} on nowhere dense classes of graphs is
\emph{fixed parameter tractable} \cite{grohe2017deciding}.
It is therefore natural to ask for which other graph classes such characterisations or algorithmic results can be obtained, and there is a large effort from
the finite model theory and algorithmic/structural graph theory communities to push them to hereditary dense structures (see for instance
\cite{dreier2024flip,gajarsky2020first,dreier2024first,torunczyk2023flip} to cite a few). With this aim, the well-studied model-theoretic notion of \emph{monadic stability} beautifully manifested itself.
The concept of monadic stability was introduced by Baldwin and Shelah \cite{baldwin1985second}, and was mainly studied within \emph{model theory} and in connection with Shelah's \emph{classification theory} \cite{shelah1990classification}. Their work provides an understanding of the structure of monadically stable classes from the model-theoretic perspective.
However, describing their combinatorial properties has started only recently. Immediately after, the two combinatorial characterisations of monadically stable graph classes through the notions of \emph{flip-flatness} \cite{dreier2023indiscernibles} and the \emph{flipper game} \cite{flipper-game}, came handy in proving that $\fo$-model checking is fixed parameter tractable on monadically stable classes of graphs \cite{ssmc, dreier2024first}. Informally, flip-flatness means that in a big enough graph we can perform a certain number of \emph{flips}, that is flipping the adjacency between a pair of subsets of vertices, such that in the final structure, there exists a large enough set of vertices that are pairwise far away. When the number of flips is independent from the distance between the far-away subset of vertices and only depends on the class, we call the class \emph{strongly flip-flat}.
Recently, Eleftheriadis extended the result on extension preservation theorem to a restriction of the strongly flip-flat classes of graphs \cite{eleftheriadis2024extension}.
We recall Eleftheriadis' theorem after explaining the notation. If $G$ is a graph and $F=(A,B)$ is a pair of subsets of vertices of $G$, the \emph{flip of $G$ on $F$}, denoted by $G\oplus F$, is the graph obtained from $G$ by
flipping the
edges between vertices in $A$ and vertices in $B$, where $A$ and $B$ can have non-empty intersection. Flipping edges between $A$ and $B$ means replacing every
edge by a non-edge and every non-edge by an edge. One can check that $G\oplus F\oplus F' = G\oplus F'\oplus F$. A family $\mathcal{F}$ of $k$ pairs of subsets of vertices is called a \emph{$k$-flip in $G$}, and for
a \emph{$k$-flip} $\mathcal{F} := \{F_1,\ldots,F_k\}$ in $G$, we denote by $G \oplus \mathcal{F}$ the graph $G\oplus F_1 \oplus F_2\oplus \cdots \oplus
F_k$, and by $G\star_{\mathcal{F}} G$ the graph $G'\oplus \mathcal{F'}$ where $G'$ is the disjoint union of two copies $G_1$
and $G_2$ of $G$ and $\mathcal{F'}=\{(A_1,B_2)\mid (A,B)\in \mathcal{F}, A_1\ \textrm{(resp. $B_2$)}$ the copy of $A\ \textrm{(resp. $B$)}$ in $G_1\ \textrm{(resp. $G_2$)}\}$.
% Even though the desired characterisation of dense structures allowing
% nice structural and algorithmic properties such as \emph{\fo model-checking} being \emph{fixed-parameter tractable}, a large amount of tools developped for \emph{nowhere dense}
% classes of graphs has been extended to cope with hereditary dense structures. Building upon such tools, Eleftheriadis extended in
% \cite{eleftheriadis2024extension} the result on \emph{extension preservation theorem} on \emph{nowhere dense} classes of graphs to some dense classes of
% graphs. If $G$ is a graph and $F=\{A,B\}$ is an unordered pair of subsets of vertices of $G$, the \emph{flip of $G$ on $F$}, denoted by $G\oplus F$, is the graph obtained from $G$ by
% flipping\footnote{Flipping edges means replace every edge by a non-edge and every non-edge by an edge.} the
% edges between vertices in $A$ and vertices in $B$. One can check that $G\oplus F\oplus F' = G\oplus F'\oplus F$. A family $\mathcal{F}$ of $k$ pairs of subsets of vertices is called a \emph{$k$-flip in $G$}, and for
% a \emph{$k$-flip} $\mathcal{F} := \{F_1,\ldots,F_k\}$ in $G$, we denote by $G \oplus \mathcal{F}$ the graph $G\oplus F_1 \oplus F_2\oplus \cdots \oplus
% F_k$, and by $G\star_{\mathcal{F}} G$ the graph $G'\oplus \mathcal{F'}$ where $G'$ is the disjoint union of two copies of $G_1$
% and $G_2$ of $G$ and $\mathcal{F'}=\{\{A_1,B_2\}\mid \{A,B\}\in \mathcal{F}, A_1\ \textrm{(resp. $B_2$)}$ the copy of $A \textrm{(resp. $B$)}$ in $G_1 \textrm{(resp. $G_2$)}\}$.
%
%\say{
\begin{theorem}[\cite{eleftheriadis2024extension}] Let $\C$ be a hereditary class of graphs. Suppose that there is some $k \in \N$ such that, for all $r \in
\N$, there is a function $f_r:\N \to \N$ satisfying that, for every $m \in \N$ and every $G \in \C$ of size at least $f_r(m)$, there is a $k$-partition $P$ of
$V(G)$, some $k$-flip $\mathcal{F} \subseteq P\times P$, and $A \subseteq V(G)$ such that
\begin{enumerate}
\item $|A|\geq m$;
\item $A$ is $r$-independent in $G \oplus \mathcal{F}$;
\item\label{ass:3} $G\star_{\mathcal{F}}G \in \C$.
\end{enumerate}
Then extension preservation holds over $\C$.
\end{theorem}
% (with additional closure conditions)\cite{atserias2008preservation,atserias2006preservation,dawar2024preservation,dawar2010homomorphism}.
%
% It was in these works that notions such as \emph{wideness, quasi-wideness} and \emph{almost-wideness} were introduced.
% These notions proved to be useful in other contexts. \emph{Quasi-wideness}, informally, means that in a large enough structure $G$ from our class we can find a small set of elements $S$ such that in $G\setminus S$, there is a large enough set $B\subseteq G$ where its elements are pairwise far-away. Being uniform, or as it is called \emph{uniformly quasi-wide}, means that the above statement is true not only for large enough structures but also for their large enough substructures. \emph{Nowhere dense} classes of graphs are characterised by \emph{uniformly quasi-wideness} \cite{nevsetvril2012sparsity}.
% For example, \emph{quasi-wideness}, which informally says that in any large enough structure we can find a small set of vertices such that after the removal of these vertices, it remains a large set of pairwise far-away elements, characterises \emph{hereditary nowhere dense} classes of graphs \cite{nevsetvril2010first,nevsetvril2011nowhere}.
% This characterisation is in the core of another combinatorial characterisation for \emph{nowhere dense} class of graphs, namely \emph{splitter game}, which helped proving that \emph{$\mathsf{FO}$-model checking} is \emph{fixed parameter tractable} on \emph{nowhere dense classes of graphs} \cite{grohe2017deciding}.
% We quote his theorem here.
%
%
% \textbf{Theorem.} Fix a hereditary class of graphs $\C$. Suppose that there is some $k \in \N$ such that for all $r \in \N$ there is a function $f_r:\N \to \N$ satisfying that for every $m \in \N$ and every $G \in \C$ of size at least $f(m)$ there is a $k$-partition $P$ of $V(G)$, some $k$-flip $F $, and $A \subseteq V(G)$ such that
%
% \begin{enumerate}
% \item $|A|\geq m$;
% \item $A$ is $r$-independent in $G \oplus \mathcal{F}$;
% \item\label{ass:3} $G\star_{\mathcal{F}}G \in \C$.
% \end{enumerate}
% Then extension preservation holds over $\C$.
% \end{theorem}
%% \todo{rewrite the theorem with our notation.}
% }
The first two conditions are imposing strong flip-flatness and the third condition is a closure condition which does not naturally hold.
%and is restricting the tameness notion which is imposed by the first two conditions. This tameness notion is called \emph{strongly flip-flatness}.
% Informally, \emph{strongly flip-flatness} means that in a big enough graph we can perform a certain number of flips, that is flipping the adjacency between a pair of subsets of vertices, such that in the final structure there exists a large enough set of vertices that are pairwise far-away. Note that the number of flips is fixed and only depends on the class of graphs; when the number of flips can depend on the distance between the far-away subset of vertices we call the notion \emph{flip-flatness}\cite{dreier2023indiscernibles}.
% \emph{Flip-flatness} is another important concept that helped pushing the $\mathsf{FO}$-model checking results \cite{dreier2024first}. To explain the significance of this concept, it suffices to say that its \emph{structural \cite{dreier2024flip}, combinatorial \cite{flipper-game}}, and \emph{logic based \cite{dreier2023indiscernibles}} characterisation are studied.
% Despite the well roundedness of \emph{flip-flatness}, the precise position of \emph{strongly flip-flatness} within this framework remains unclear.
While it is evident that strongly flip-flat classes of graphs are a subset of flip-flat graph classes, a clear characterisation is still needed.
% Transduction of \emph{hereditary quasi-wide} classes of graphs are included in \emph{flip-flat} classes of graph. A \emph{transduction}, which is denoted by $T_{\delta,\varphi}$, where $\delta$ and $\varpghi$ are a unary and binary $\mathsf{FO}$-formulas, is when you are allowed to colour the graphs of a class $\CC$ with a fixed number of colours, then change the vertex set and the adjacency of each particular graph as $\delta$ and $\varphi$ dictate, then take any arbitrary subgraph.
We initiate a characterisation of strongly flip-flat graph classes by looking at their intersection with \emph{weakly sparse} classes of graphs, \ie,
classes of graphs that exclude some biclique. The reason we believe this is a good starting point is that, for many classes of dense graphs, their intersection
with weakly sparse classes of graphs has been observed to be well-behaved classes of graphs, and as such examples we can cite:
\begin{itemize}
\item Bounded shrubdepth $\cap$ Weakly sparse $=$ Bounded treedepth \cite{de2025transducing} (see also the discussion of \cite{gajarsky2022stable}),
\item Bounded clique-width $\cap$ Weakly sparse $=$ Bounded tree-width \cite[Theorem 5.9(4)]{courcelle2000upper},
\item Bounded linear clique-width $\cap$ Weakly sparse $=$ Bounded path-width \cite{gurski2000tree}. In \cite{gurski2000tree} Gurski and Wanke only
state the result for graphs of bounded clique-width, which is an improvement on the bound given in \cite[Theorem 5.9(4)]{courcelle2000upper}. However,
they indeed prove that one can construct a tree decomposition from a given clique-width expression (equivalently NLC-width expression), while keeping the shape of the expression. Hence, when the expression has the form of a path, \ie, linear clique-width is bounded, the resulting tree will look like a path, \ie, path-width is bounded.
\item Monadically stable $\cap$ Weakly sparse $=$ Nowhere dense \cite{dvovrak2018induced,podewski1978stable, adler2014interpreting}
\item Monadically dependent $\cap$ Weakly sparse $=$ Nowhere dense \cite{dvovrak2018induced} (see also \cite{nevsetvril2021classes}), and
\item Bounded flip-width $\cap$ Weakly sparse $=$ Bounded expansion \cite{torunczyk2023flip}.
\end{itemize}
Hence, following this line of research, we prove the following.
\begin{maintheorem}
A class of graphs $\Cc$ is a weakly sparse and strongly flip-flat if and only if $\CC$ is uniformly almost-wide.
\end{maintheorem}
$\fo$-transductions are a way of constructing new classes of graphs from a given class $\CC$ by the means of $\fo$ formulas. Basically, we have the power to colour
the vertices of the graphs arbitrarily, but with a constant number of colours, and then change the vertex and edge set as the given $\fo$ formulas
dictates. These formulas can use the colours. When a class $\CC$ has property $\mathcal{P}$, for any $\fo$-transduction of $\CC$, we say that it is \emph{structurally}
$\mathcal{P}$. For example, it has been shown that structurally nowhere dense classes of graphs are monadically stable and hence flip-flat \cite{podewski1978stable,
adler2014interpreting, dreier2023indiscernibles}. On the other hand, it is shown that, for some weakly sparse classes of graphs, their $\fo$-transductions are exactly some well-known, yet
dense classes of graphs. As a prominent example, it is proven that structurally bounded tree-depth classes of graphs are exactly bounded shrubdepth classes of
graphs \cite{ganian2012trees}. In addition to this, Eleftheriadis also showed that in the setting of hereditary classes of graphs, transductions of uniformly almost-wide classes of graphs are strongly flip-flat
\cite{eleftheriadis2024extension}. Therefore, we conjecture the following, supported by the \emph{sparsification conjecture} stating that monadically stable graph classes are exactly the structurally nowhere dense classes of
graphs.
\begin{conjecture}\label{conjecture}
Every strongly flip-flat graph class is structurally uniform almost-wide.
\end{conjecture}
We believe our main theorem is a step towards proving the conjecture, and allow us to confirm it on structurally bounded expansion graph classes.
Ne{\v{s}}et{\v{r}}il and Ossona De Mendez showed that for a hereditary class of graphs $\CC$, if there are $s\in \NN$ and a function
$t:\NN\rightarrow\NN$ such that, for every $r\in \NN$, the graphs in the depth-$r$ shallow minor of $\CC$ exclude $K_{s,t(r)}$, then $\CC$ is uniformly almost-wide
\cite{nevsetvril2012sparsity}. As second point supporting the conjecture, we also argue that there exist infinitely many $s\in\NN$ such that, for every $r\in\NN$, the
$r$-subdivisions of $K_{s,N}$, for arbitrary large $N$, cannot be strongly flip-flat and we leave open the question of using them to characterise
strongly flip-flat graph classes using \emph{shallow vertex-minors} \cite{buffiere2024shallow}. % as the one given in \cite[Theorem 8.4]{nevsetvril2012sparsity}.
%After completing our work, we became aware of independent results by M\"ahlmann which establish a statement closely related to ours \cite{maehlmann2024}(Lemma 13.10). In that work, M\"ahlmann proves that \emph{weakly sparse} and \emph{flip-flat} classes of graphs are \emph{uniformly quasi-wide}. As \emph{flip-flatness} characterises \emph{monadic stability} and \emph{uniform quasi-wideness} characterises \emph{nowhere denseness}, that result mirrors a corollary of \cite{dvovrak2018induced,nevsetvril2020linear} showing that \emph{monadically stable} claases of graph that are \emph{weakly sparse} are \emph{nowhere dense}. Although, our main statements are different, ours studies \emph{strong flip-flatness} and M\"ahlmann studies \emph{flip-flatness}, the proof are close in nature. However, our approach has the novelty of applying an induction on the set of flips instead of the radius. Therefore, we believe our work brings value to the ongoing research on this topic.
After completing our work, we became aware of independent results by M\"ahlmann \cite[Lemma 13.10]{maehlmann2024}, which establish a statement closely related
to ours. In that work, M\"ahlmann proves that weakly sparse and flip-flat classes of graphs are uniformly quasi-wide. Since
flip-flatness characterizes monadic stability, and uniform quasi-wideness characterizes nowhere denseness, this result mirrors a
corollary of \cite{dvovrak2018induced,nevsetvril2020linear}, which shows that monadically stable and weakly sparse classes of graphs are
nowhere dense.
While our main results differ, ours concern strong flip-flatness, whereas M\"ahlmann focuses on flip-flatness, the proofs are similar in spirit. Our approach, however, introduces a novel point of view: an induction on the set of flips instead of the usual induction on the independence radius. We believe this perspective brings new insights and contributes meaningfully to the ongoing research in this area.
%
% \textcolor{red}{
% \begin{itemize}
% \item Motivation behind the preservation theorems.
% \item The preservation result on dense graphs.
% \item The result holds on a certain classes of graphs that are strongly flip-flat; hence the motivation to characterise these classes.
% \item The result showing that structurally almost-wide classes are strongly flip-flat.
% \item Citing all the results looking at the intersection of dense structures with weakly sparse classes. (why it made sense for us to look at this intersection as well.)
% \item Conjecture.
% \item Examples of classes that are not almost-wide and examples of classes that are not strongly flip-flat.
% \end{itemize}}
\section{Preliminaries}
For two sets $A$ and $B$, we denote by $A\setminus B$ the set $\{x\in A\mid x\not \in B\}$ and by $A\Delta B$ the set $(A\setminus B)\cup (B\setminus A)$. The
set of positive integers is denoted by $\mathbb{N}$.
We assume familiarity with standard graph theoretical notions, and refer to \cite{bondy2008graph,diestel2024graph}, and deal only with undirected
graphs\footnote{One can extend the results to directed graphs by using a coding of directed graphs, through an invertible \fo-transduction, into bipartite
undirected graphs.}. The vertex set of a graph $G$ is denoted by
$V(G)$ and its edge set by $E(G)$. An edge between $u$ and $v$ in a graph $G$ is denoted by $uv$ (equivalently $vu$).
We recall that, in a graph $G$, a path between two vertices $u$ and $v$ in $V(G)$ is a sequence of vertices $u=w_1,\cdots,w_l=v$ such that $w_i\neq w_j$, for
all $1\leq i< j\leq l$, and, for all $1\leq i< l$, we have that $w_iw_{i+1}\in E(G)$; we say that the length of this path is $l-1$; we often denote such a path
by $v-u$ when we are not interested in the internal vertices of the path. The distance between a pair of vertices is the length of the shortest path between them. For a subset $S$ of $V(G)$, we denote by $G[S]$ the subgraph of $G$ induced by $S$,
\ie, $V(G[S])=S$ and $E(G[S])=(S\times S)\cap E(G)$; by $G\setminus S$ we mean the subgraph of $G$ induced by $V(G)\setminus S$. We say that a set $B\subseteq
V(G)$ is \emph{$r$-independent in $G$} if, for all pairs $u,v\in B$, the distance between $u$ and $v$ in $G$ is strictly greater than $r$. By $K_t$ and $K_{t,t}$, we denote the complete graph on $t$ vertices and the complete bipartite graph with $t$ vertices in each part, respectively. A \emph{half-graph} of order $n$ is a bipartite graph with sides $\{a_1,\cdots,a_n\}$ and $\{b_1,\cdots,b_n\}$ such that, for all $1\leq i,j\leq n$, $a_i$ and $b_j$ are adjacent if and only if $i\leq j$.
\emph{Half-graphs} are graph theoretical representations of linear orders.
% By the $s$-subdivision of $G$, we mean the graph where each edge of $G$ is replaced by a path of length $s+1$, i.e. we put $s$ extra vertices on each edge of $G$. We denote the $s$-subdivision of $G$ by $G^s$.
%
For an integer $r$, the $r$-subdivision of a graph $G$ is the graph obtained from $G$
by replacing each edge $uv$ of $G$ by a path $u-v$ of length at most $r+1$ that intersects $V(G)$ only on its endpoints.
\paragraph*{Sparse graph classes.} We adopt the notion of graph sparsity as developed in \cite{nevsetvril2012sparsity}.
\begin{definition}
Let $\CC$ be a class of graphs. We say that $\CC$ is \textbf{weakly sparse} if there exists a constant $t\in \NN$ such that every graph $G\in \CC$ excludes $K_{t,t}$ as a subgraph.
\end{definition}
A graph $H$ is a \emph{depth-$r$ shallow minor} of $G$ if there are disjoint connected subgraphs $C_1,\ldots, C_{|V(H)|}$ of
$G$, each of radius at most $r$, and a bijective mapping $\rho:V(H)\to \{C_1,\ldots,C_{|V(H)|}\}$ such that $uv\in E(H)$ only if there is an edge between a vertex in $\rho(u)$ and a
vertex in $\rho(v)$.
A class of graphs $\CC$ is nowhere dense if and only, for every integer $r$, there exists an integer $t_r$ such that all the graphs of $\CC$ exclude $K_{t_r}$ as depth-$r$ shallow minor.
% \footnote{A graph $H$ is a minor of a graph $G$ if there are disjoint connected subgraphs $C_1,\ldots, C_{|V(H)|}$ of
% $G$ and a bijective mapping $\rho:V(H)\to \{C_1,\ldots,C_{|V(H)|}\}$ such that $uv\in E(H)$ only if there is an edge between a vertex in $\rho(u)$ and a
% vertex in $\rho(v)$.}.
% Since we obtain $K_t$ by contracting each pair of vertices composed of two vertices from the opposite sides of
% $K_{t,t}$,
By taking $t$ pairs of vertices from opposite sides of $K_{t,t}$ as disjoint connected subgraphs, one constructs $K_t$ as a depth-$1$ shallow minor.
Therefore, we can conclude that nowhere dense classes of graphs are weakly sparse.
Let us now introduce the core notions of quasi-wideness.
\begin{definition}\label{definition: uniformly quasi-wide and almost-wide classes}
Let $\CC$ be a class of graphs. We say that $\CC$ is \textbf{uniformly quasi-wide} if, for every $r\in \NN$, there exists a function $N_{r}:\NN\rightarrow\NN$
and an integer $s_r$ such that, for all $m\in \NN$ and $G\in \CC$, and for all $A\subseteq V(G)$ of size at least $N_{r}(m)$, we can find sets $S\subseteq V(G)$ and $B\subseteq A$ where $|S|\leq s_r$ and $|B|> m$, such that $B$ is $r$-independent in $G\setminus S$. A graph class $\CC$ is said to be \textbf{uniformly almost-wide} if $s_r:= s$ is independent of $r$. (See~\autoref{fig:definition of almost wide}.)
We say that $(A,S)$ is an \emph{$(r,m)$-widenable instance of $G$ with respect to $s$ and $N_r$} if $A\subseteq V(G)$ with $|A|\geq N_r(m)$ and $S\subseteq
V(G)$ with $|S|\leq s$, and we can find a set $B\subseteq A$ such that $|B|\geq m$ and $B$ is $r$-independent in $G\setminus S$. When $s$ and $N_r$ are clear
from the context we only say that $(A,S)$ is an $(r,m)$-widenable instance of $G$.
\end{definition}
\begin{figure}[h]
\label{fig: fig1}
\center
\begin{tikzpicture}
\node[draw, minimum width=10cm, minimum height=6cm] (rect) at (0,0) {};
\node[draw, minimum width=8cm, minimum height=4cm] (rect2) at (0.1,0.3) {};
\node[above right=2pt and 2pt of rect] {G};
\node[above right=2pt and 2pt of rect2] {A};
\node at (-2,1.5) {\textcolor{red}{\Large$\times$}};
\node at (1,2) {\textcolor{red}{\Large$\times$}};
\node at (0,-2) {\textcolor{red}{\Large$\times$}};
\node at (-3,0) {\textcolor{red}{\Large$\times$}};
\node at (3,-1.5) {\textcolor{red}{\Large$\times$}};
\node at (0.5,-1) {\textcolor{red}{\Large$\times$}};
\draw[thick, blue]
(-3,-1.2)
.. controls (-2,-1.5) and (-1.5,-1.6) ..
(-0.5,-1.4)
.. controls (0.3,0) and (1.2,-1.2) ..
(2,-1)
.. controls (2.5,-0.5) and (2.4,0.8) ..
(2,1.2)
.. controls (1.5,1.8) and (0.2,1.7) ..
(-1,1.5)
.. controls (-2,1.2) and (-2.5,0.5) ..
(-3,-0.5)
.. controls (-3.2,-0.8) and (-3.1,-1.1) ..
(-3.1,-1.1)
-- cycle;
\node[blue] at (2.4,1) {$B$};
\foreach \x/\y in {-2.5/-0.8, -1.5/-1.1, -0.5/-0.6, 0.7/0, 1.5/-0.8,
2/0, 1.2/0.8, 0/0.7, -0.8/0.6} {
\filldraw[draw=green!60!black, fill=white, thick] (\x,\y) circle (0.4);
\fill[green!60!black] (\x,\y) circle (1.2pt);
}
% Label r on one disk
\draw[->, thick] (0.7,0) -- ++(30:0.4) node[midway, below ] {\small $r$};
\end{tikzpicture}
\caption{ Assuming $|A|> N_r(m)$ is a subset of vertices of $G$, a graph in a uniformly quasi-wide class of graphs $\mathscr{C}$, then there are sets of vertices $S$ (the red crosses), with $|S|\leq s_r$,
and $B\subseteq A$, with $|B|\geq m$, such that $B$ is $r$-independent in $G\setminus S$. The function $N_r$ and the constant $s_r$ both depend on $r$ and the class
$\mathscr{C}$. When $s_r$ is independent from $r$ and only depends on the
class, we call that class uniformly almost-wide.}
\label{fig:definition of almost wide}
% In a graph $G$, for any large enough subset of vertices $A$, one can find a set $S$ of $s$ vertices (the red crosses) such that after removing
% those vertices, there will be a large subset of $A$, namely $B$, with $r$-independent vertices. When $s$ is independent from $r$ and only depends on the
% class, we call that class uniformly almost-wide, otherwise it is called uniformly quasi-wide.}\todo[inline]{The caption is just repeating the definition
% above. A possible one : Assuming $|A|>N_r(m)$ is a subset of vertices of $G$, a graph in a uniformly quasi-wide class of graphs $\mathscr{C}$, then there are $S\leq s_r$
% and $|B|\geq m$ such that $B$ is $r$-independent in $G\setminus S$. The function $N_r$ and the constant $s_r$ both depend on $r$ and the class
% $\mathscr{C}$. When $s_r$ is independent from $r$ and only depends on the
% class, we call that class uniformly almost-wide.
\end{figure}
It has been shown that when it comes to hereditary classes of graphs, \ie, closed under induced subgraphs, uniform quasi-wideness characterises
nowhere denseness, and uniform almost-wideness characterises classes for which there exists a function $t:\NN\rightarrow \NN$ and an integer $s$
such that for every $r\in \N$, it excludes $K_{s,t(r)}$ as depth-$r$ shallow minor \cite{nevsetvril2012sparsity}.
\paragraph*{Beyond sparse graph classes.} While nowhere dense classes of graphs are exactly the graph classes closed under subgraph and for which \fo
model-checking is fixed parameter tractable, there are other important graph classes for which \fo model-checking is still fixed parameter tractable, such as
unit-interval graphs or map graphs. It is conjectured in \cite{ALS2016} that hereditary graph classes for which \fo
model-checking is fixed parameter tractable correspond exactly to monadically dependent graph classes, a model-theoretic notion which will be defined later. The notions introduced in this quest are the ones behind the notion of
strongly flip-flat graph classes.
\begin{definition}[Flips]
Let $G$ be a graph. A \emph{flip} $\flip{F}$ is an operation that is specified by a pair of sets $\flip{F}= (A,B)$, where $A,B\subseteq V(G)$($A$ and $B$ can intersect and even be equal). We denote this operation by $G\oplus\flip{F}$ and it operates on $G$ by complementing the adjacency between $A$ and $B$, more precisely
\begin{align*}
G\oplus \flip{F} := (V(G),E(G)\Delta ((A\times B)\cup (B\times A)) ).%, \{uv\mid uv\in E(G) ~xor~ (u,v)\in (A\times B)\cup (B\times A)\}).
\end{align*}
By $G\oplus\FF$, where $\FF =\{\flip{F_1},\cdots,\flip{F_k}\}$ is a set of flips, we mean $G\oplus \flip{F_1}\oplus \cdots\oplus\flip{F_k}$. For a flip $\flip{F}=(A,B)$ and a set of vertices $P\subseteq V(G)$, by $\flip{F}\setminus{P}$ we mean $(A\setminus P,B\setminus P)$. To simplify the notation, instead of $(G\setminus P)\oplus \flip{F}\setminus{P}$ we write $(G\setminus P)\oplus \flip{F}$.
%We say that a flip $\flip{F}=(A,B)$ is minimal with respect to $C\subseteq V(G)$ and $r\in\NN$, if $C$ is $r$-independent in $G\oplus \flip{F}$, and
%there is no flip $\flip{F'}=(A'\subseteq A,B'\subseteq B)$ such that $C$ is $r$-independent in $G\oplus \flip{F}'$.
%for every vertex $v\in A\cup B$ there are at least two vertices $u_1,u_2\in C$ such that $v$ is on a path of length less than $2r$ between $u_1$ and $u_2$.
%
%We say that $\mathcal{F}$ scatters $B$ with respect to $r$ and $G$ if $B$ is $r$-independent in $G\oplus\mathcal{F}$. When $r$ and $G$ are clear from the context, we say that $\mathcal{F}$ scatters $B$.
\end{definition}
\begin{observation}\label{Observation: repetition and order of flips}
For any graph $G$ and \emph{flips} $\flip{F},\flip{F'}$ we have $G\oplus \flip{F}\oplus\flip{F'} = G\oplus\flip{F'}\oplus \flip{F}$ and $G\oplus\flip{F}\oplus\flip{F}=G$.
\end{observation}
\begin{lemma}\label{lemma: flips do not intersect}
Let $G$ be a graph and $\FF=\{\flip{F_1},\cdots,\flip{F_k}\}$ be a set of flips. There exists a set of flips $\FF'$ where $G\oplus\FF = G\oplus \FF'$ and $|\FF'|\leq \binom{2^{2k}+1}{2}$ such that $\FF'$ does not flip the adjacency of any fixed pair of vertices more than once.
\end{lemma}
%${\cap_{i \in I} P_i \setminus \cup_{j \notin I} P_j | for I subset of [2k]} \cup {V(G) \setminus \cup_{i \in [2k]} P_i}
%$
\begin{proof}
Let $\flip{F_1},\cdots,\flip{F_k}$ be equal to $(P_1,P_2),\cdots,(P_{2k-1},P_{2k})$, respectively.
Let $\mathsf{Rest}= \{V(G) \setminus \cup_{i \in [2k]} P_i\} $ and
$$\mathsf{Partition}=\{\cap_{i \in I} P_i \setminus \cup_{j \notin I} P_j \mid I\subseteq [2k],I\neq \emptyset\} \cup \{\mathsf{Rest}\}.$$
%
% We construct $\PP_i$ recursively for $0\leq i\leq 2k$ as follows:
% \begin{itemize}
% \item $\PP_0 =\{\bigcap_{\substack{i\in [2k]}}P_i\}$,
% \item $\PP_{i} = \{\bigcap_{t\in I}P_t \setminus \bigcup_{0\leq j<i}\bigcup_{P\in\PP_{j}}P\mid I\subseteq [2k],|I| = 2k-i\}$ for $1\leq i\leq 2k-1$, and
% \item $\PP_{2k}=\{V(G)\setminus \bigcup_{1\leq i\leq 2k}P_i\}$.
% \end{itemize}
% The sets in $\PP_i$ for $i< 2k$ are intersections of any $2k-i$ sets of $\PP$ minus any vertex that appears in more than $2k-i$ sets of $\PP$, i.e., the vertices in the sets of $\PP_i$ are vertices that are in the intersection of exactly $2k-i$ sets of $\PP$. The only set in $\PP_{2k}$ is the set of vertices that do not appear in any element of $\PP$. Let $\mathsf{Partition} = \bigcup_{0\leq i\leq 2k}\PP_i$, then $|\mathsf{Partition}|\leq 1+\binom{2k}{2k-1}+\cdots+\binom{2k}{1}+1$, that is $|\mathsf{Partition}|\leq 2^{2k}$.
Indeed $\mathsf{Partition}$ is a partition of the vertices of $G$, and for a part $Q\in \mathsf{Partition}$, if $P_i\cap Q\neq\emptyset$, then $Q\subseteq P_i$. Notice that $|\mathsf{Partition}|\leq 2^{2k}+1$.
% Suppose that $Q\in \PP_t$, then there are $P_{i_1},\cdots,P_{i_t}$ sets in $\PP$ such that they all include $Q$. If $P_i$ is not one of $P_{i_1},\cdots,P_{i_t}$, then we have found $t+1$ sets such that they all include $P_i\cap Q$, this means that we should have removed all the elements of $P_i\cap Q$ from $Q$ before adding it to $\PP_t$. Hence, $P_i$ is one of $P_{i_1},\cdots,P_{i_t}$, and so it includes $Q$.
This implies that each $P_i$ can be written as a union of some elements of the $\mathsf{Partition}$. Now, assume that for each $P_i$ there is a set of indices $I_i$ such that $P_i=\bigcup_{j\in I_i}Q_j$, and $Q_j$, $j\in I_i$, being elements of $\mathsf{Partition}$, then a flip $F_t=(P_{2t-1}, P_{2t})$ is equivalent to $\FF_t=\{(Q_j,Q_s)\mid j\in I_{2t-1}, s\in I_{2t}\}$.
We construct $\FF'$ to be a refinement of $\FF$ by
taking flips from $\FF_t$, for $1\leq t\leq k$, that appear in an odd number of $\FF_t$.
As an even repetition of a flip does not alter the graph (Observation \ref{Observation: repetition and order of flips}), the other flips of $\FF_t$, $1\leq t\leq k$, i.e. those that appear an even number of times, can be avoided. Thus, $\FF'$ has at most $\binom{2^{2k}+1}{2}$ many elements.
Now, as the sets in the flips of $\FF'$ are parts of a partition, a fixed pair of vertices occurs in a unique flip. Thus, $\FF'$ does not change the adjacency of a pair of vertices more than once.
\end{proof}
\begin{corollary}
A set of flips $\FF=\{\flip{F_1},\cdots,\flip{F_k}\}$ on a graph $G$, can be seen as a binary relation on a partition of the vertex set of $G$. This binary relation specifies which pairs of parts undergo a flip and which do not.
\end{corollary}
%Let $\mathsf{P}$ be a partition of the vertex set of a graph $G$ and $F$ be a binary relation on this partition. By $G\bigtriangleup_{F}\mathsf{P}$ we denote the set of flips corresponding to $F$ on $\mathsf{P}$.
We proceed to the formal definition of flip-flatness and strong flip-flatness, these notions are a generalisation of uniformly
quasi-wideness and uniformly almost-wideness, where \emph{vertex deletion} is replaced by \emph{performing flips}.
\begin{definition}
Let $\CC$ be a class of graphs. We say that $\CC$ is \textbf{flip-flat} if, for every $r\in \NN$, there exists a function $N_r:\NN\rightarrow \NN$ and an integer $s_r$ such that for all $m\in \NN$ and $G\in \CC$, and for all $A\subseteq V(G)$ of size at least $N_r(m)$, we can find a set of flips $\FF$ and $B\subseteq A$ where $|\FF|\leq s_r$ and $|B|\geq m$, such that $B$ is $r$-independent in $G\oplus \FF$. A class $\CC$ is said to be \textbf{strongly flip-flat} if $s_r:= s$ is independent of $r$. (See~\autoref{fig: fig2}.)
We say that $(A,\FF)$ is an \emph{$(r,m)$-flippable instance of $G$} with respect to $N_r$ if $A\subseteq V(G)$ with $|A|\geq N_r(m)$ and $\FF$ is a set of
flips and we can find a set $B\subseteq A$ with $|B|\geq m$ such that $B$ is $r$-independent in $G\oplus \FF$. When $N_r$ is clear from the context, we only say
that $(A,\FF)$ is an $(r,m)$-flippable instance of $G$.
\end{definition}
\begin{figure}[h]
\center
\begin{tikzpicture}
\node[draw, minimum width=10cm, minimum height=6cm] (rect) at (0,0) {};
\node[draw, minimum width=8cm, minimum height=4cm] (rect2) at (0.2,0.5) {};
\node[above right=2pt and 2pt of rect] {G};
\node[above right=2pt and 2pt of rect2] {A};
% \draw[thick, blue]
% (-3,-1.2)
% .. controls (-2,-1.5) and (-1.5,-1.6) ..
% (-0.5,-1.4)
% .. controls (0.3,-1) and (1.2,-1.2) ..
% (2,-1)
% .. controls (2.5,-0.5) and (2.4,0.8) ..
% (2,1.2)
% .. controls (1.5,1.8) and (0.2,1.7) ..
% (-1,1.5)
% .. controls (-2,1.2) and (-2.5,0.5) ..
% (-3,-0.5)
% .. controls (-3.2,-0.8) and (-3.1,-1.1) ..
% (-3.1,-1.1)
% -- cycle;
% \node[blue] at (2.4,1) {$B$};
\draw[thick, blue]
(-3,-1.2)
.. controls (-2,-1.5) and (-1.5,-1.6) ..
(-0.5,-1.4)
.. controls (0.3,0) and (1.2,-1.2) ..
(2,-1)
.. controls (2.5,-0.5) and (2.4,0.8) ..
(2,1.2)
.. controls (1.5,1.8) and (0.2,1.7) ..
(-1,1.5)
.. controls (-2,1.2) and (-2.5,0.5) ..
(-3,-0.5)
.. controls (-3.2,-0.8) and (-3.1,-1.1) ..
(-3.1,-1.1)
-- cycle;
\node[blue] at (2.4,1) {$B$};
%\foreach \x/\y in {-2.5/-0.8, -1.5/-1.1, -0.5/-0.9, 0.7/0, 1.5/-0.8,
% 2/0, 1.2/0.8, 0/0.7, -1/0.6} {
% \filldraw[draw=green!60!black, fill=white, thick] (\x,\y) circle (0.4);
% \fill[green!60!black] (\x,\y) circle (1.2pt);
% }
%
% % Label r on one disk
% \draw[->, thick] (0.7,0) -- ++(30:0.4) node[midway, above right] {\small $r$};
%
%\draw[red, thick, rotate=30] (-1.5,2) ellipse (0.3 and 0.6);
%\draw[red, thick, rotate=150] (2,0) ellipse (0.3 and 0.6);
%
%\draw[red, thick] (1,2) ellipse (0.3 and 0.6);
%\draw[red, thick] (0.3,2) ellipse (0.3 and 0.6);
%
%\draw[red, thick] (-3,0) ellipse (0.3 and 0.6);
%\draw[red, thick] (-3.6,0) ellipse (0.3 and 0.6);
%
%\draw[red, thick, rotate=45] (-0.7,-2) ellipse (0.3 and 0.6);
%\draw[red, thick] (0.3,-2) ellipse (0.3 and 0.6);
%
%\draw[red, thick] (3,-1.5) ellipse (0.3 and 0.6);
%\draw[red, thick] (3.3,-1.5) ellipse (0.3 and 0.6);
%
\foreach \x/\y in {-2.5/-0.8, -1.5/-1.1, -0.5/-0.6, 0.7/0, 1.5/-0.8,
2/0, 1.2/0.8, 0/0.7, -0.8/0.6} {
\filldraw[draw=green!60!black, fill=white, thick] (\x,\y) circle (0.4);
\fill[green!60!black] (\x,\y) circle (1.2pt);
}
% Label r on one disk
\draw[->, thick] (0.7,0) -- ++(30:0.4) node[midway, above right] {\small $r$};
\draw[red, thick, rotate=30] (-1.5,2) ellipse (0.3 and 0.6);
\draw[red, thick, rotate=150] (2,0) ellipse (0.3 and 0.6);
\draw[red, thick] (1,2) ellipse (0.3 and 0.6);
\draw[red, thick] (0.3,2) ellipse (0.3 and 0.6);
\draw[red, thick] (-3,0) ellipse (0.3 and 0.6);
\draw[red, thick] (-3.6,0) ellipse (0.3 and 0.6);
\draw[red, thick, rotate=45] (-0.5,-1.5) ellipse (0.3 and 0.6);
\draw[red, thick] (0.3,-1.5) ellipse (0.3 and 0.6);
\draw[red, thick] (3,-1.5) ellipse (0.3 and 0.6);
\draw[red, thick] (3.3,-1.5) ellipse (0.3 and 0.6);
\end{tikzpicture}
\caption{
Assuming $|A|> N_r(m)$ is a subset of vertices of $G$, a graph in a flip-flat class of graphs $\mathscr{C}$, then there is a set of flips $\mathcal{F}$ (the red ellipses), with $|\mathcal{F}|\leq s_r$,
and a set $B\subseteq A$, with $|B|\geq m$, such that $B$ is $r$-independent in $G\oplus \mathcal{F}$. The function $N_r$ and the constant $s_r$ both depend on $r$ and the class
$\mathscr{C}$. When $s_r$ is independent from $r$ and only depends on the
class, we call that class strongly flip-flat.
%In a graph $G$, for any large enough subset of vertices $A$, one can find a set $\mathcal{F}$ of $s$ flips (the red pairs of ellipses) such that after
% flipping those flips, there will be a large subset of $A$, namely $B$, with $r$-independent vertices. When $s$ is independent from $r$ and only depends on the
% class, we call that class strongly flip-flat, otherwise it is called flip-flat.
}
\label{fig: fig2}
% \todo[inline]{similar to Figure 1}
\end{figure}
Note that one can mimic the removal of a vertex $v$ by the flip $(\{v\}, N(v))$.
However, flip-flatness gives us a purely combinatorial characterisation of monadic stability as well, that we recall now.
One can think of graphs as finite relational structures with the signature $\tau_{E}=\{E\}$, where $E$ is the relation symbol indicating the edge relation. We
refer to \cite{Libkin2004elements} for the definition of \fo formulas. We recall here the notion of a transduction. Let $c \in \N$ and $\delta(x)$ and $\phi(x,y)$ be
first-order formulas over the extended signature $\tau_E^c:= \tau_E \cup \{C_1,\dots,C_c\}$, where $C_1,\cdots, C_c$ are unary predicates. We think of these
predicates as colours marking the vertices. The \emph{transduction $T_{\delta,\phi}$} is the operation that maps a graph $G$ to the class $T_{\delta,\phi}(G)$
containing the graphs $H$ such that there exists a $\tau^c_E$-expansion $G^+$ of $G$, i.e., a colouring of $G$ with colours $\{C_1,\dots,C_c\}$, satisfying $V(H)=\{v \in V(G): G^+ \models \delta(v)\}$ and
\[E(H):= \left\{(u,v)\in V(H)^2 \colon u \neq v \land G^+ \models (\phi(u,v) \lor \phi(v,u))\right\}.\]
We write $T_{\delta,\phi}(\C):=\bigcup_{G \in \C} T_{\delta,\phi}(G)$ and say that \emph{$\Dcal$ is a transduction of a class $\C$}, or \emph{$\C$ transduces $\Dcal$}, if there exists a transduction $\Tcal_{\delta,\phi}$ such that $\Dcal \subseteq \Tcal_{\delta,\phi}(\C)$. A graph class $\C$ is monadically dependent (resp. monadically stable) if $\C$ does not transduce the class of all graphs (resp. half-graphs). We remark that these are not the original definitions of monadic stability and monadic dependence, but it follows from the work of Baldwin and Shelah that these notions can be defined by transductions \cite{baldwin1985second}.
We say that a class $\CC$ is structurally $\mathcal{P}$, if there exists a transduction $T_{\delta, \varphi}$ and a class $\DD$ with property $\mathcal{P}$ such that $\CC\subseteq T_{\delta, \varphi}(\DD)$.
%When a class $\CC$ has property $\mathcal{P}$, for any transduction $T_{\delta, \varphi}$, we say that $T_{\delta,\varphi}(\CC)$ is \emph{structurally} $\mathcal{P}$.
\begin{theorem}[\cite{dreier2023indiscernibles}]
A class of graphs is monadically stable if and only if it is flip-flat.
\end{theorem}
Since an \fo-transduction preserves monadic stability, one can first ask whether \fo-transductions preserve strong flip-flatness, and indeed yes as shown in
\cite{eleftheriadis2024extension}.
\begin{proposition}[\cite{eleftheriadis2024extension}]\label{prop:transduce-sflipflat} If $\mathscr{C}$ is strongly flip-flat, then $T_{\delta, \varphi}(\mathscr{C})$ is also strongly flip-flat,
for any \fo-transduction $T_{\delta, \varphi}$.
\end{proposition}
It has been proven that classes of graphs that are monadically stable and weakly sparse are nowhere dense
\cite{dvovrak2018induced,podewski1978stable, adler2014interpreting}. Thanks to Proposition \ref{prop:transduce-sflipflat}, every \fo-transduction of a uniformly almost-wide graph class is indeed
strongly flip-flat. Hence our motivation to show that classes of graphs that are strongly flip-flat and weakly sparse are uniformly
almost-wide.
\section{Main result}
We are ready to state and prove the main result.
\begin{theorem}\label{theorem: main result}
A class of graphs $\Cc$ is a weakly sparse and strongly flip-flat if and only if $\CC$ is uniformly almost-wide.
\end{theorem}
We start with the following lemma.
\begin{lemma}\label{case of k=1}
Let $G$ be a graph, $t_0,r$ and $m$ be integers, and $B\subseteq V(G)$ be a set with $|B|\geq t^2_0+t_0+m-2$, such that all the elements of $B$ are pairwise at distance less than or equal to $r$. Assume there exists a flip $\flip{F}=(P_1,P_2)$ such that:
\begin{itemize}
\item $\flip{F}$ does not contain a $K_{t_0,t_0}$ with one side in $P_1$ and the other side in $P_2$,
\item and $B$ is $r$-independent in $G\oplus \flip{F}$.
\end{itemize}
Then we can find a set $S\subseteq V(G)$, with $|S|\leq t^2_0+t_0-2$ and a set $B'\subseteq B$ with $|B'|\geq m$ such that $B'$ is $r$-independent in
$G\setminus S$.
\end{lemma}
\begin{proof}
It is safe to assume that $t^2_0+t_0-2\geq 8t_0$, since if we exclude $K_{t_0,t_0}$ then we exclude $K_{t,t}$ for all $t\geq t_0$ as well.
As all the elements of $B$ are pairwise at distance less than or equal to $r$, we infer that $\flip{F}$ is responsible for breaking the short paths between all pairs
of vertices from $B$.
For a path $u=w_1,\cdots,w_l=v$ of length less than or equal to $r$ between two vertices $u$ and $v$, that is broken by $\flip{F}$, we define
\emph{elementary segments} to be the parts $w_1,w_2,\cdots,w_i$ and $w_j,w_{j+1},\cdots,w_l$ such that $\{w_1,w_2,\cdots,w_i\}\cap(P_1 \cup P_2)= \{w_i\} $ and $\{w_j,w_{j+1},\cdots,w_l\}\cap (P_1 \cup P_2)=\{w_j\}$, i.e., the parts consisting of $v$ or $u$ and only one element of $P_1 \cup P_2$.
Hence, each such path has
exactly $2$ elementary segments. As the length of these paths is less than or equal to $r$, and at least one of their edges is between $P_1$ and $P_2$,
we conclude that at least one of their elementary segments has length strictly less than $\lceil\frac{r}{2}\rceil$; we call this a \emph{short
elementary segment}. Note that two short elementary segments with different endpoints in $B$ cannot share an endpoint in $P_1\cup P_2$ as this would
be a path of length less than or equal to $r$ between the elements of $B$ that is not affected by the flip.
Let us fix $8t_0$ vertices of $B$, say $\{v_1,v_2,\cdots,v_{4t_0}, u_1,\cdots,u_{4t_0}\}$, and focus on the paths in $G$ of length less than or equal to $r$ between pairs
$(v_1,u_1),(v_2,u_2),\cdots,(v_{4t_0},u_{4t_0})$. These are $4t_0$ paths of length less than or equal to $r$ with different endpoints, having at least $4t_0$ elementary segments that
have length strictly less than $\lceil\frac{r}{2}\rceil$. At least $2t_0$ of these short elementary segments have an endpoint in the same side of $\flip{F}$. Say, without loss of generality, that $v_1 - a_1, v_2 - a_2, \cdots, v_{2t_0} - a_{2t_0}$ are the short elementary segments, where $a_1, a_2, \cdots, a_{2t_0} \in P_1$, and possibly $v_i = a_i$ for some $i$. Now, we divide the argument based on the parity of $r$.
\textbf{The even case: $r=2r'$.} Pick $t_0+1$ of these short elementary segments. Say, without loss of generality $v_1-a_1, v_2-a_2, \cdots,
v_{t_0+1}-a_{t_0+1}$. Observe that every element in $P_2$ is adjacent to at least $t_0$ of $a_1,a_2,\cdots,a_{t_0+1}$. Indeed, if $u\in P_2$ is not adjacent
to $a_i,a_j$ where $a_i\neq a_j$, then after the flip we will have that $v_i-a_iua_j-v_j$ is a path of length less than or equal to $r$ between $v_i$ and $v_j$;
recall that $v_i$ and $v_j$ belong to $B$ and after performing $\flip{F}$ their distance in $G\oplus\flip{F}$ is strictly greater than $r$. For a vertex $a$ among $a_1,a_2,\cdots,a_{t_0+1}$, there are at most $t_0-1$ vertices in $P_2$ that are not adjacent to $a$; as otherwise, these $t_0$ vertices of $P_2$ and $\{a_1,a_2,\cdots,a_{t_0+1}\}\setminus\{a\}$ form a $K_{t_0,t_0}$. Hence, besides these $(t_0-1)(t_0+1)$ vertices of $P_2$, all the remaining form a biclique with $a_1,a_2,\cdots,a_{t_0+1}$. Thus, $|P_2|\leq (t_0-1)(t_0+1)+(t_0-1)$ (Figure \ref{Figure: figure}). We conclude that at least one of $P_1$ or $P_2$ has at most $t_0^2+t_0-2$ many vertices. Without loss of generality, let $P_2$ be the one such that $|P_2|\leq t_0^2+t_0-2$; now, $B\setminus P_2$ is $r$-independent in $G\setminus P_2$, and we have that $|B\setminus P_2|\geq m$.
\textbf{The odd case: $r=2r'+1$.} Suppose first that at least $t_0$ many of these short elementary segments have length exactly $r'$. Say,
without loss of generality $v_1-a_1, v_2-a_2, \cdots, v_{t_0}-a_{t_0}$ have length exactly $r'$. Let $b_1-u_1,b_2-u_2,\cdots, b_{t_0}-u_{t_0}$ be the other elementary
segments of these paths, i.e. $b_1,b_2,\cdots,b_{t_0}\in P_2$ and for all $i=1,\cdots,t_0$ we have that $v_i-a_i$ and $b_i-u_i$ are the elementary
segments of $v_i-u_i$. As for all $i\in [t_0]$ the length of $v_i-a_i$ is exactly $r'$ we conclude that the length of $b_i-u_i$ is less than or equal to $r'$ as
well. Therefore, the sets $\{a_1,\cdots,a_{t_0}\}$ and $\{b_1,\cdots,b_{t_0}\}$ form a biclique in $G$. Indeed, if $a_i$ and $b_j$, when $i\neq j$, are not
adjacent in $G$, then after the flip $v_i-a_ib_j-u_j$ will be a path of length less than or equal to $2r'+1 = r$, which is contradictory and a biclique on $\{a_1,\cdots,a_{t_0}\}$ and $\{b_1,\cdots,b_{t_0}\}$ is also a contradiction to weakly sparseness. Hence, at least $t_0+1$ of the short elementary segments have length strictly less than $r'$. Say, without loss of generality $v_1-a_1, v_2-a_2, \cdots, v_{t_0+1}-a_{t_0+1}$.
Now, as in the even case, all the vertices $u\in P_2$ are adjacent to at least $t_0$ of $a_1,a_2,\cdots,a_{t_0+1}$ and we get the same bound on the size of $P_2$. Thus, in this case, $B\setminus P_2$ is also $r$-independent in $G\setminus P_2$, and we have $|B\setminus P_2|\geq m$; hence the statement follows.
\begin{figure}
\center
\input{chapters/lemma.tex}
\caption{The paths from $v_i$ to $a_i$ are short elementary segments. For each $a_i$ there are at most ${t_0-1}$ many vertices in $P_2$ that are not adjacent to $a_i$. To not overcomplicate the figure we avoided putting edges. The dotted lines mean \emph{not adjacent}, where there is no dotted line, there is an edge.}\label{Figure: figure}
\end{figure}
\end{proof}
By $\ramsey{m}{n}$ we denote the \emph{Ramsey number} for $m$ and $n$, that is,
the smallest integer $N$ such that every graph of size~$N$ contains either
a clique of size~$m$ or an independent set of size~$n$.
We are ready to prove the Theorem \ref{theorem: main result}. Let
\begin{displaymath}
\mathscr{R}^k(m,n) := \underbrace{\ramsey{m}{\ramsey{m}{\cdots,\ramsey{m}{n}}}}_\text{k times}.
\end{displaymath}
% \mathscr{R}^k(m,n) := \underbrace{\ramsey{m}{\ramsey{m}{\cdots,\ramsey{m}{n}}}}_\text{k times}
%$$
% where $\ramsey{m}{n}$ is the \emph{Ramsey} number of $m$ and $n$.
\begin{proof}
% Let $s$ be the size of the set of flips given by the strongly flip-flatness of $\CC$. We prove this by induction on $s$. For $s=0$, the statement follows. As the induction hypothesis, we assume that if a class $\CC$ is weakly sparse, with parameter $t_0$, and for every $r\in \NN$ there exists a function $N_r:\NN\rightarrow \NN$ such that for all $G\in \CC$ and $m\in \NN$ and for all $A\subseteq V(G)$ with $|A|\geq N_r(m)$, we can find $B\subseteq A$ with $|B|\geq m$ and $\mathcal{F}=\{\flip{F_1},\cdots,\flip{F_{k-1}}\}$ such that $B$ is $r$-independent in $G\oplus\mathcal{F}$, then $\CC$ is uniformly almost-wide and $s'=(k-1)(t_0^2+t_0-2)$ is the constant witnessing the wideness.
Let us first prove that uniformly almost-wide classes are strongly flip-flat.
Recall that removing a vertex $v$ can be done by performing a flip between $v$ and the neighbourhood of $v$, i.e., the flip $(\{v\}, N(v))$; note that after this flip the vertex $v$ is completely independent from the rest of the graph and is not adjacent to any other vertex. For a
uniformly almost-wide class of graphs $\CC$ there exists an integer $k$ such that for any $r\in \NN$ there exists a function $N_r\colon \NN\rightarrow\NN$, so that for any graph $G\in\CC$ and any set of vertices $A\subseteq V(G)$, with $|A|\geq N_r(m)$, we can find $k$ vertices $v_1,\cdots,v_k$ such that there exists a set of vertices $B\subseteq V(G)\setminus\{v_1,\cdots,v_k\}$, where $|B|\geq m$ and $B$ is $r$-independent in $G\setminus \{v_1,\cdots,v_k\}$. Now, instead of removing $\{v_1,\cdots,v_{k}\}$ we iteratively perform the flips $(\{v_1\}, N(v_1)),\cdots ,(\{v_k\},N(v_k))$, notice that the iterative application of the flips means that $N(v_i)$ is the neighbourhood of $v$ after the first $i-1$ flips. With these flips, we completely isolate $v_1,\cdots,v_k$ from the rest of the graph and therefore $B$ will be $r$-independent in $G\oplus \{(\{v_1\}, N(v_1)),\cdots ,(\{v_k\},N(v_k))\}$. Hence, for any $r\in \NN$ the same function $N_r$ and the same fixed integer $k$ witness strong flip-flatness of $\CC$. In addition, uniformly almost-wide classes of graphs are weakly sparse, as for any integer $k$, we cannot find any set of $2$-independent vertices after removing $k$ vertices in $K_{k+1,k+1}$.
Now, for the other direction, let $k$ be the size of the set of flips given by the strong flip-flatness of $\CC$. We prove this by induction on $k$. For $k=0$, the statement follows.
As the induction hypothesis, we assume that if a class $\CC$ is weakly sparse, with parameter $t_0$, and for every $r\in \NN$ there exists a function $N_r:\NN\rightarrow \NN$ such that for all $G\in \CC$ and $m\in \NN$ and for all $A\subseteq V(G)$ with $|A|\geq N_r(m)$, we can find a $(r,m)$-flippable instance $(A,\FF)$ with $|\FF|<k$, then $\CC$ is uniformly almost-wide and $s'=(k-1)(t_0^2+t_0-2)$ is the constant witnessing the wideness.
% Now, suppose that $\CC$ is weakly sparse, with parameter $t_0$ such that $t_0^2+t_0-2\geq 4t_0+2$, and strongly flip-flat with $s=k$. We define $N_r'(m) = N_r(\mathscr{R}^k(m, t_0^2+t_0+m-2))$ to be the function in the definition of \emph{uniform almost-wideness}. For an arbitrary set $A\subseteq V(G)$ with $|A|\geq N_r'(m)$ we can find $\mathcal{F}=\{\flip{F_1},\cdots,\flip{F_k}\}$ and $B\subseteq A$ with $|B|\geq \mathscr{R}^k(m, t_0^2+4t_0+m+6)$ such that $B$ is $r$-independent in $G\oplus\mathcal{F}$.
Now, suppose that $\CC$ is weakly sparse, with parameter $t_0$ such that $t_0^2+t_0-2\geq 4t_0+2$, and strongly flip-flat with parameter $k$. We define $N_r'(m) = N_r(\mathscr{R}^k(m, t_0^2+t_0+m-2))$ to be the function in the definition of uniform almost-wideness. For any graph $G\in\CC$ and an arbitrary set $A\subseteq V(G)$ with $|A|\geq N_r'(m)$ we can find a $(r,m)$-flippable instance $(A,\FF)$.
% If we manage to find a subset $B'\subseteq A$ with $|B'|\geq m$ and a set of flip $\mathcal{F}'$ with $|\mathcal{F}'|\leq k-1$ such that $B'$ is $r$-independent in $G\oplus \mathcal{F}'$, then we use the induction hypothesis for this particular set $A$ to find a set $S$ of size at most $s'$ such that in $G\setminus S$, $A$ has a subset of size at least $m$ which is $r$-independent.
If $|\FF|\leq k-1$ then we use the induction hypothesis for this particular set $A$ to find a set $S$ such that $(A,S)$ is a $(r,m)$-widenable instance of $G$.
Hence, we can focus on the case where no subset of $A$ of size $m$ is $r$-independent after any $k-1$ flips; however there exists a set $B\subseteq A$ with $|B|\geq \mathscr{R}^k(m, t_0^2+t_0+m-2)$ such that $B$ is $r$-independent in $G\oplus\FF$.
Let us examine the iterative effect of $\FF:=\{F_1,\ldots,F_k\}$ on $B$. As $|B|\geq \mathscr{R}^k(m, t_0^2+t_0+m-2)$ at the beginning we face two
cases, either $B$ contains a set of vertices of size $m$ where all the elements are pairwise at distance strictly greater than $r$, or it contains a set of vertices of size $\mathscr{R}^{k-1}(m, t_0^2+t_0+m-2)$ where all the elements are pairwise at distance less than or equal to $r$. Because of our assumption on the necessity of at least $k$ flips, the former can not happen and so there exists $B_0\subseteq B$ with $|B_0|\geq \mathscr{R}^{k-1}(m, t_0^2+t_0+m-2)$ such that all its elements are pairwise at distance less than or equal to $r$ in $G$. We use the same argument on $B_0$ but in $G\oplus \flip{F_1}$, this gives us a set $B_1\subseteq B_0$ with $|B_1|\geq \mathscr{R}^{k-2}(m, t_0^2+t_0+m-2 )$ such that all its elements are pairwise at distance less than or equal to $r$ in $G\oplus\flip{F_1}$. We repeat similarly and get $B_{k-1}\subseteq B$ with $|B_{k-1}|\geq t_0^2+t_0+m-2$ such that all its elements are pairwise at distance less than or equal to $r$ in $G\oplus\flip{F_1}\oplus\cdots\oplus\flip{F_{k-1}}$.
As $B$ is $r$-independent in $G\oplus \FF$,
% As $\FF$ is responsible for scattering $B$,
we conclude that after the application of $\flip{F_k}$ on $G\oplus\flip{F_1}\oplus\cdots\oplus\flip{F_{k-1}}$, all the elements of $B_{k-1}$ should be
at distance strictly greater than $r$.
%
% With this description of $\flip{F_k}$, we can claim that at least one side of $\flip{F_k}$ has size less than $t^2_0+t_0-2$. The argue goes as in the base case (theorem \ref{case of k=1}). Let $\flip{F_k}=(P_1^k,P_2^k)$ and $|P_2^k|\leq t_0^2+t_0-2$.
By Lemma \ref{lemma: flips do not intersect}, it is safe to assume that the flips $\flip{F_1},\cdots,\flip{F_k}$ do not alter an edge twice. Hence, $\flip{F_k}$ is acting on the original edges of $G$, which excludes $K_{t_0,t_0}$ as subgraph, and not the edges added by $\flip{F_1},\cdots,\flip{F_{k-1}}$, i.e., there is no $K_{t_0,t_0}$ with its bipartition included in the side of the flip. Additionally, $B_{k-1}$ has size larger than $t^2_0+t_0+m-2$ and is
$r$-independent in $(G\oplus\flip{F_1}\oplus\cdots\oplus\flip{F_{k-1}})\oplus \flip{F_k}$;
% scattered by $\flip{F_k}$ with respect to $r$ in $G\oplus\flip{F_1}\oplus\cdots\oplus\flip{F_{k-1}}$;
thus, by Lemma \ref{case of k=1}, we can find sets $S$ with $|S|\leq t_0^2+t_0-2$ and $B'$ with $|B'|\geq m$ as promised. As $(G\oplus\flip{F_1}\oplus\cdots\oplus\flip{F_{k-1}})\setminus S$ and $(G\setminus S)\oplus\flip{F_1}\oplus\cdots\oplus\flip{F_{k-1}}$ are equal, we are now able to use the induction hypothesis on $(G\setminus S)\oplus\flip{F_1}\oplus\cdots\oplus\flip{F_{k-1}}$. Thus, we have a set $B''$ with $|B''|\geq m$ and a set $S'$ with $|S'|\leq (k-1)(t_0^2+t_0-2)$ such that $B''$ is $r$-independent in $(G\setminus S)\setminus S'$, that is $B''$ is $r$-independent in $G\setminus (S\cup S')$ where $|S\cup S'|\leq k(t_0^2+t_0-2)$.
Therefore, for any $(r,m)$-flippable instance $(A,\FF)$ of $G$ with respect to $N_r'$ and $|\FF|\leq k$, we can find a $(r,m)$-widenable instance $(A,S)$ of $G$ with respect to $s'$ and $N_r'$ where $s'\leq k(t_0^2+t_0-2)$. As $s'$ is independent from $r$, we conclude that $\CC$ is uniformly almost-wide.
% It is clear that $B\setminus P_2^k$ is $r$-independent in $(G\setminus P_2^k)\oplus\flip{F_1}\oplus\cdots\oplus\flip{F_{k-1}}$. Thus, we managed to find a set of size greater than $m$ and a set of flips of size $k-1$ such that the set of vertices is $r$-independent in the graph after the application of the flips and as the removal of vertices keep the class weakly sparse, we can use the induction hypothesis to find a set $B''\subseteq A$ with $|B''|\geq m$ and a set $S$ with $|S|\leq s'$ such that $B''$ is $r$-independent in $G\setminus(P_2^k\cup S)$. We have that $|P_2^k\cup S|\leq k(t_0^2+t_0-2)$ and we are done.
%
% that all classes of graphs that are weakly sparse and strongly flip-flat with $s=k-1$, are uniformly almost-wide and $s':= (k-1)(t_0^2+t_0-2)$ is the number of vertices witnessing the wideness.
%
%
% Let $\CC$ be a weakly sparse class of graphs with $t_0$ being the size of the largest biclique such that for all $r\in \NN$ we can find a function $N_r:\NN\rightarrow\NN$ so that for all $G\in \CC$ and $m\in \NN$, and for all sets $A\subseteq V(G)$ with $|A|\geq N_r(m)$, we can find a set of flips $\mathcal{F}=\{\flip{F_1},\cdots,\flip{F_k}\}$ and a set $B\subseteq A$ with $|B|\geq m$ such that $B$ is $r$-independent in $G\oplus\mathcal{F}$.
%
% We define $N_r'(m)= N_r(\mathscr{R}^k(m,t_0^2+4t_0+m+6))$ and $s'':= k(t_0^2+t_0-2)$ to be the function and the constant witnessing the uniformly almost-wideness.
%
%
%Now, for a fixed $G\in \CC$ and $r\in\NN$ and for any arbitrary set $A\subseteq V(G)$ with $|A|\geq N_r'(m)$ we get a set of flips $\mathcal{F}$ that scatters $A$. If $|\mathcal{F}|<k$, then we use the induction hypothesis to find a set of size at most $(k-1)(t_0^2+t_0-2)< s''$ vertices to remove and get the set $B\subseteq A$ with $|B|\geq m$ which is $r$-independent in $G$ after the removal of vertices. Thus, let $\mathcal{F}=\{\flip{F_1},\cdots,\flip{F_k}\}$ and assume that we can not find a set $B\subseteq A$ with $|B|\geq m$ which is $r$-independent in $G\oplus\mathcal{F}'$ for any $\mathcal{F'}\subset \mathcal{F}$.
%
%
% We examine the iterative effect of $\mathcal{F}$ on $b$. As $|B|\geq \mathscr{R}^k(m,t_0^2+4t_0+m+6)$, it either contains a subset of vertices of size $m$ where all the elements are pairwise at distance greater than $2r$, which is contradictory as we need at least $k$ flips, or it contains a subset $A_0\subseteq A$ with $|A_0|\geq \mathscr{R}^{k-1}(m,t_0^2+4t_0+m+6)$, where all the elements are pairwise at distance less than $2r$. Now, $A_0$ in $G\oplus \flip{F_1}$ either has a subset of size $m$ where all the elements are pairwise at distance greater than $2r$, which again is contradictory, or we can find $A_1\subseteq A_0$ with $|A_1|\geq \mathscr{R}^{k-2}(m,t_0^2+4t_0+m+6)$, where all the elements are pairwise at distance less than $2r$. We proceed similarly and find $A_{k-1}\subseteq A$ with $|A_{k-1}|\geq \ramsey{m}{t_0^2+4t_0+m+6}$, and $A_{k-1}$ should contain a subset of size at least $t_0^2+4t_0+m+6$ where all its elements are pairwise at distance less than $2r$ in $G\oplus\flip{F_1}\oplus \cdots\oplus\flip{F_{k-1}}$. Therefore,
%
% Let us fix $r$ and $m$, then for an arbitrarily graph $G\in\CC$ and for any set $A\subseteq V(G)$ with $|A|\geq N_r'(m)$, we can find a set of flips $\mathcal{F}$ and a set $B\subseteq A$ with $$|B|\geq \underbrace{\ramsey{m}{\ramsey{m}{\cdots,\ramsey{m}{t_0^2+4t_0+m+6}}}}_\text{k times},$$ such that $B$ is $r$-independent in $G\oplus\mathcal{F}$. If $|\mathcal{F}|<k$, then by the induction hypothesis we are done. Thus, it is safe to assume that we need $k$ flips. As $|A|\geq \underbrace{\ramsey{m}{\ramsey{m}{\cdots,\ramsey{m}{t_0^2+4t_0+m+6}}}}_\text{k times}$, and we need exactly $k$ flips to scatter a subset of $A$, we conclude that the last flip is responsible for breaking paths of length less than $2r$ between all pairs of at least $t_0^2+4t_0+m+6$ many vertices.
%
% Let $\flip{F_i}=(P_1^i,P_2^i)$ for $1\leq i\leq k$, and w.l.o.g. assume that $|P_2^k|\leq |P_1^k|$. Let $G'=(V(G), E(G)\setminus (P_1^k\times P_2^k)\cup(P_2^k\times P_1^k))$. It is clear that $B$ is $r$-independent in $G'\oplus \{\flip{F_1},\cdots,\flip{F_{k-1}}\}$, and $G'$ excludes $K_{t_0,t_0}$. Subsequently, we can use the induction hypothesis. Thus, for an arbitrary $A\subseteq V(G')$ with $|A|\geq N_r'(m)$ the same $B\subseteq A$ is $r$-independent in $G'\setminus S$ where $|S|\leq (k-1)(t_0^2+t_0-2)$. This implies that $B\setminus P_2^k$ is $r$-independent in $G\setminus (P_2^k\cup S)$. As $\flip{F_k}$ is scattering at least $t_0^2+4t_0+m+6$ vertices that are pairwise at distance less than $2r$, we can conclude that at least one side has at most $t_0^2+t_0-2$ many elements and so $|P_2^k\cup S|\leq k(t_0^2+t_0-2)$ and $|B\setminus P_2^k|\geq m$.
\end{proof}
While our argument extends to show that flip-flat graph classes that are weakly sparse are uniformly quasi-wide, this result is equivalent to the known result that monadically stable and weakly sparse classes are nowhere dense \cite{dvovrak2018induced,podewski1978stable, adler2014interpreting}, so we do not elaborate further.
As a corollary, we obtain the following characterization of strongly flip-flat classes that have structurally bounded expansion. We say that a class $\CC$ of graphs has \emph{bounded expansion} if there exists a function $f \colon \NN \to \NN$ such that for every $r \in \NN$ and every depth-$r$ shallow minor $H$ of a graph $G \in \CC$, we have $\frac{\|H\|}{|H|} \leq f(r)$, where $\|H\|$ denotes the number of edges and $|H|$ the number of vertices of $H$.
Note that for bounded expansion classes, we require the edge density of depth-$r$ shallow minors to be bounded, whereas for nowhere dense classes, we ask that large cliques be excluded as depth-$r$ shallow minors. Accordingly, every class of bounded expansion is nowhere dense.
Let us recall that a structurally bounded expansion class $\CC$ of graphs is a class of graphs for which we can find a bounded expansion class of graphs $\DD$ and a transduction $T_{\delta,\varphi}$ such that $\CC\subseteq T_{\delta,\varphi}(\DD)$.
\begin{theorem}\label{thm:struct-bd-expansion} A structurally bounded expansion graph class $\mathscr{C}$ is strongly flip-flat if and only if there is a uniformly
almost-wide graph class $\mathscr{D}$ of bounded expansion and a transduction $\tau$ such that $\mathscr{C}\subseteq \tau(\mathscr{D})$.
\end{theorem}
\begin{proof} If $\mathscr{C}$ is structurally bounded expansion, then there are transductions $\tau_1$ and $\tau_2$ such that
$\tau_1(\mathscr{C})$ has bounded expansion and $\mathscr{C}\subseteq \tau_2(\tau_1(\mathscr{C}))$ \cite{gajarsky2020first}. Because transductions preserve
strong flip-flatness (Proposition \ref{prop:transduce-sflipflat}), we
have that $\tau_1(\mathscr{C})$ is strongly flip-flat and by Theorem \ref{theorem: main result}, we can conclude that $\tau_1(\mathscr{C})$ is uniformly almost-wide.
\end{proof}
Theorem \ref{thm:struct-bd-expansion} confirms Conjecture~\ref{conjecture} for graph classes of structurally bounded expansion.
% For $r\geq 1$ it is clear that the graphs of this class are weakly sparse and as it has been shown in \cite{nevsetvril2010first}(Theorem $3.21$) they are not almost-wide; thus, by theorem \ref{theorem: main result} this class is not strongly flip-flat. However, we prove this without the use of theorem \ref{theorem: main result}. In this approach we only rely on the fact that a set of flips can be seen as a partition of vertex set.
%\begin{proof}
%We prove by contradiction.
%
%Let $k$ be the number of flips in the definition of strongly flip-flatness. Then by lemma \ref{lemma: flips do not intersect} we can think of a particular set of flip $\mathcal{F}$ of size at most $k$ as a partition of vertex set into $2^{2k}$ parts and a binary relation on this partition dictating flips between pairs of parts.
%
%Suppose that we are asked to find a large set of $r'$-independent vertices. We pick $r'-1$-subdivided bipartite graph $G = (S\cup L, E(G))$ where $|S|> (2^{2k})^{(2r'+1)}$ and $|L|$ is arbitrary large. The distance between any two vertices $u,v\in L$ is exactly $2r'$. We take a subset $A$ with $|A|\geq N_{r'}(m)$ from $L$. By our assumption there exists a set of flips that scatters a set $B\subseteq A$ of size $m$. Let us think of the equivalent partition of this set of flips as a colouring $c:V(G)\rightarrow [2^{2k}]$ of vertices. Before the flips, all the elements of $B$ are pairwise at distance less than $2r'$. Hence the flips should break such paths for all pairs of vertices from $B$. Note that these paths do not share edges. There are $2r'+1$ many vertices on each such path; thus there are $(2^{2k})^{2r'+1}$ colour patterns for paths.
%We have $\binom{m}{2}\times |S|$ many paths and as $|S|> (2^{2k})^{2r'+1}$, we can find at least two paths with the same colour pattern such that they do not share their intersection with $S$.
%
%
% Let $u_1-u_{r'+1}-u_{2r'+1}$ and $v_1-v_{r'+1}-v_{2r'+1}$, where $u_1,u_{2r'+1},v_1,v_{2r'+1}\in B$ and $u_{r'+1},v_{r'+1}\in S$, be those two paths with the same colour pattern, i.e. $c(u_i)=c(v_i)$ for $1\leq i\leq 2r'+1$. They do not share their intersection with $S$, hence $u_{r'+1}\neq v_{r'+1}$.
%
%% It is safe to assume that $u_{s+1}\neq v_{s+1}$, as there are definitely two different vertices in $S$ with the same colour and for each pair of elements in $B$ there are paths going through both of those two vertices.
%
%We use a pebble to travers from $u_1$ to one of $u_{2r'+1}$ or $v_{2r'+1}$ in $2r'$ steps after the flips; this shows that even after the flips there is path of length $2r'$ between elements of $B$. This gives us a contradiction.
%
%After the application of the flips, put the pebble on $u_1$, if there is an edge between $u_1$ to $u_2$, we move the pebble to $u_2$; however if there is no edge between $u_1$ and $u_2$ we infer that the adjacency has been flipped. In addition to that we know that $c(u_2)=c(v_2)$ and there had not been an edge between vertices of $u_1-u_{r'+1}-u_{2r'+1}$ and $v_1-v_{r'+1}-v_{2r'+1}$ prior to the flips. Thus, the flip had added an edge between $v_1$ and $u_2$. In this case we move the pebble to $u_2$. So, w.l.o.g. assume that at step $i$ the pebble is on $u_i$, if there is an edge between $u_{i}$ and $u_{i+1}$ we move the pebble to $u_{i+1}$ if not, there should be an edge between $u_{i}$ and $v_{i+1}$; hence we move the pebble to $v_{i+1}$. As these paths did not share an edge and vertices before the flips, we can continue this procedure for $2r'$ steps and end up on $u_{2r'+1}$ or $v_{2r'+1}$.
% Suppose that $k$ is the universal number of flips we need to scatter arbitrary large sets of vertices in each $G \in \mathcal{C}$. Let $s:= k+1$, $r:= s$ and $m:=\textit{ very large}$. We show that there is no constant number of flips that can make a large number of vertices $r$-independent. We can think of a $k$-flip as a partition of vertices, let us color each vertex with the part it belongs to. Let $A$ be the part of $K^{s}_{s,N}$ that has $s$ vertices and $B$ be the part that has $N$ vertices. Now, for a large enough $m$, we can find two vertices $v_i$ and $v_j$ in $B$ and two vertices $x$ and $y$ in $A$ such that the paths $v_i,\cdots,x,\cdots,v_j$ and $v_i,\cdots,y,\cdots,v_j$ have the same color pattern. Now, as the length of these paths is $2s$ and we want to make $v_i$ and $v_j$ $s$-independent, we should perform flips such that it breaks these paths. The similarity of these two paths implies that each flip just changes the graph up to a labelling and keep these path the same isomorphically. Thus, for this large $m$, we will not be able to make all of them $s$-independent.
%
%\end{proof}
\tikzset{every picture/.style={line width=0.75pt}} %set default line width to 0.75pt
\begin{tikzpicture}[x=0.75pt,y=0.70pt,yscale=-1,xscale=1]
%uncomment if require: \path (0,495); %set diagram left start at 0, and has height of 495
%Rounded Rect [id:dp1840579081474819]
\draw (402,107.5) .. controls (406.42,107.5) and (410,111.08) .. (410,115.5) -- (410,413.5) .. controls (410,417.92) and (406.42,421.5) .. (402,421.5) -- (378,421.5) .. controls (373.58,421.5) and (370,417.92) .. (370,413.5) -- (370,115.5) .. controls (370,111.08) and (373.58,107.5) .. (378,107.5) -- cycle ;
%Rounded Rect [id:dp43659636205968144]
\draw (302,152) .. controls (306.42,152) and (310,155.58) .. (310,160) -- (310,319) .. controls (310,323.42) and (306.42,327) .. (302,327) -- (278,327) .. controls (273.58,327) and (270,323.42) .. (270,319) -- (270,160) .. controls (270,155.58) and (273.58,152) .. (278,152) -- cycle ;
%Curve Lines [id:da8821375138958147]
\draw (206,164.5) .. controls (246,134.5) and (240,213.5) .. (280,183.5) ;
%Curve Lines [id:da3789139931412354]
\draw (208,226.5) .. controls (248,196.5) and (239,235.5) .. (279,205.5) ;
%Curve Lines [id:da7276516603722737]
\draw (205,305.5) .. controls (245,275.5) and (238,334.5) .. (278,304.5) ;
%Straight Lines [id:da2641105234060074]
\draw [dash pattern={on 0.84pt off 2.51pt}] (297,172.5) -- (381,131.5) ;
%Straight Lines [id:da8209598278868032]
\draw [dash pattern={on 0.84pt off 2.51pt}] (301,178.5) -- (375,176.5) ;
%Straight Lines [id:da010236799365302418]
\draw [dash pattern={on 0.84pt off 2.51pt}] (301,200.5) -- (377,198.5) ;
%Straight Lines [id:da32013017368350716]
\draw [dash pattern={on 0.84pt off 2.51pt}] (303,209.5) -- (376,242.5) ;
%Straight Lines [id:da18951529515421806]
\draw [dash pattern={on 0.84pt off 2.51pt}] (308,296.5) -- (373,298.5) ;
%Straight Lines [id:da2296003933165892]
\draw [dash pattern={on 0.84pt off 2.51pt}] (297,307.5) -- (371,343.5) ;
%Straight Lines [id:da8809991771839264]
\draw (346,346.5) -- (375,369) ;
%Straight Lines [id:da6724569148240183]
\draw (352,371.5) -- (374,402) ;
% Text Node
\draw (284,165.4) node [anchor=north west][inner sep=0.75pt] {$a_{1}$};
% Text Node
\draw (281,192.4) node [anchor=north west][inner sep=0.75pt] {$a_{2}$};
% Text Node
\draw (274,285.4) node [anchor=north west][inner sep=0.75pt] {$a_{t_{0} +1}$};
% Text Node
\draw (193,157.4) node [anchor=north west][inner sep=0.75pt] {$v_{1}$};
% Text Node
\draw (192,220.4) node [anchor=north west][inner sep=0.75pt] {$v_{2}$};
% Text Node
\draw (184,300.4) node [anchor=north west][inner sep=0.75pt] {$v_{t_{0} +1}$};
% Text Node
\draw (282,126.4) node [anchor=north west][inner sep=0.75pt] {$P_{1}$};
% Text Node
\draw (381,82.4) node [anchor=north west][inner sep=0.75pt] {$P_{2}$};
% Text Node
\draw (382,118.9) node [anchor=north west][inner sep=0.75pt] {$b_{1}^{1}$};
% Text Node
\draw (377,232.4) node [anchor=north west][inner sep=0.75pt] {$b_{t_{0} -1}^{2}$};
% Text Node
\draw (380,188.4) node [anchor=north west][inner sep=0.75pt] {$b_{1}^{2}$};
% Text Node
\draw (378,163.4) node [anchor=north west][inner sep=0.75pt] {$b_{t_{0} -1}^{1}$};
% Text Node
\draw (383,134.9) node [anchor=north west][inner sep=0.75pt] {$\vdots $};
% Text Node
\draw (379,201.9) node [anchor=north west][inner sep=0.75pt] {$\vdots $};
% Text Node
\draw (376,285.4) node [anchor=north west][inner sep=0.75pt] {$b_{1}^{t_{0} +1}$};
% Text Node
\draw (375,331.4) node [anchor=north west][inner sep=0.75pt] {$b_{t_{0} -1}^{t_{0} +1}$};
% Text Node
\draw (382,307.4) node [anchor=north west][inner sep=0.75pt] {$\vdots $};
% Text Node
\draw (380,252.4) node [anchor=north west][inner sep=0.75pt] {$\vdots $};
% Text Node
\draw (282,239.4) node [anchor=north west][inner sep=0.75pt] {$\vdots $};
% Text Node
\draw (380,359.4) node [anchor=north west][inner sep=0.75pt] {$c_{1}$};
% Text Node
\draw (377,396.4) node [anchor=north west][inner sep=0.75pt] {$c_{t_{0} -1}$};
% Text Node
\draw (383,375.4) node [anchor=north west][inner sep=0.75pt] {$\vdots $};
\end{tikzpicture}
\section{Conclusion}
A corollary of our main theorem is the confirmation of Conjecture \ref{conjecture} on strongly flip-flat graph classes that have structurally bounded expansion, partly
answering the characterisation question raised in \cite{eleftheriadis2024extension}. It seems however that there is
still a long way to completely characterise strongly flip-flat classes of graphs. A confirmation of our conjecture \ref{conjecture} will give a
characterisation of strongly flip-flat graph classes as its reverse has been proved (Proposition \ref{prop:transduce-sflipflat}).
In order to prove Conjecture \ref{conjecture}, we must find a nowhere dense preimage for any class $\CC$ of strongly flip-flat graph classes and then
argue that this preimage itself is strongly flip-flat, \ie, equivalent of saying that it is almost-wide. This might be as hard as the
\emph{sparsification conjecture} \cite{ossonademendez:LIPIcs.STACS.2021.2}, stated below.
\begin{conjecture}
Every monadically stable graph class is structurally nowhere dense.
\end{conjecture}
A first step towards Conjecture \ref{conjecture} is probably assuming that the sparsification conjecture is true, and then prove the following, that we
state as a conjecture.
\begin{conjecture}[Weak version of Conjecture \ref{conjecture}]
Let $\CC$ be a strongly flip-flat class of graphs and $\DD$ be a nowhere dense class of graphs. If there exists a transduction $T_{\delta,\varphi}$ such that
$\CC\subseteq T_{\delta,\varphi}(\DD)$, then there exists a class $\DD'$ of uniformly almost-wide graphs and a transduction $T'_{\delta,\varphi}$ such that $\CC\subseteq
T'_{\delta,\varphi}(\DD')$.
\end{conjecture}
Our conjecture may not be the easiest way to characterise strongly flip-flat graph classes. One can, for instance, look into forbidden induced subgraphs. In the following we give an example of a nowhere dense class of graphs that is not strongly flip-flat. We denote by $G^s$ the $s$-subdivision of
a graph $G$.
% be the class of $s$-subdivided complete bipartite graphs where one side has $s$ vertices and the other has arbitrary many vertices
\begin{theorem}\label{Example: not SFF} Let $\CC:=\{K^s_{s,N}\mid s\in \NN,~~N\in\NN\}$. Then $\CC$ is not strongly flip-flat.
\end{theorem}
\begin{proof}
Every graph $G$ in $\CC$ excludes $K_{2,2}$ as subgraph; hence $\CC$ is weakly sparse. Note that $K^{2s}_s$ is the largest $2s$-subdivided clique, where the main vertices of the clique are the $s$ vertices of the small side of $K^{s}_{s,N}$. Therefore, $\CC$ is nowhere dense as well.
Let $r:= s+1$. We pick $K^s_{s,N}$ with an arbitrary large $N$ and an arbitrary large set
of vertices $A$ from the large side of $K_{s,N}^s$. To make any subset of $A$ $r$-independent, we need to remove all the $s$ vertices of the other
side. Therefore $\CC$ is not almost-wide (one can infer that $\CC$ is not almost-wide by \cite[Theorem 8.4]{nevsetvril2012sparsity}). By Theorem \ref{theorem: main result}, $\CC$ is not strongly flip-flat.
\end{proof}
%we already have one example of a graph class that is \emph{flip-flat}, but not \emph{strongly flip-flat} (Theorem \ref{Example: not SFF}).
As an alternative, one could look for a characterisation via forbidden induced subgraphs occurring in flips, as in the case of monadic stability provided in \cite{dreier2024flip} or bounded shrubdepth provided in \cite{DBLP:conf/icalp/Mahlmann25}, or in terms of shallow vertex-minors, as in \cite{buffiere2024shallow}.
Characterisations can be combinatorial and in terms of games, such as the flipper game for flip-flat classes of graphs
\cite{torunczyk2023flip}.
%For instance, can \emph{flipper} win, on \emph{strongly flip-flat} classes of graphs, in a constant number of rounds independent from the given radius?
We remark that the flipper cannot, in general, win in a constant number of rounds---independent of the radius---on strongly flip-flat classes of graphs. This game, in which the flipper has a strategy requiring a constant number of rounds independent of the radius, corresponds to the radius infinity game; as we can consider radius $n$ when playing on an $n$-vertex graph. This characterises instead the class of graphs with
bounded shrubdepth.
%\textcolor{red}{
%\begin{itemize}
% \item Problems we can look into, such as if a strongly flip-flat classes is known to be transduced from a nowhere dense class, can we find another preimage which is strongly flip-flat as well?
% \item Other approaches such as forbidden obsrtuctions, constant round flipper game, and/or finding the preimage.
%\end{itemize}}