Quasiconformal Maps between Bowditch Boundaries of Relatively Hyperbolic Groups
Extracted source text · arXiv 2503.17312
%\documentclass[12pt,twoside]{article}
\documentclass[12pt,reqno]{amsart}
%\documentclass[11pt]{amsart}
%\documentstyle[12pt,bezier,amstex]{article}
%\documentclass[12pt]{puthesis}
\tolerance=1000
\hbadness=10000
\raggedbottom
\usepackage[dvipsnames]{xcolor}
%Packages used
\usepackage{amscd,amsmath,latexsym, amsthm, thmtools, amssymb, mathtools, enumerate,appendix}
%\usepackage{setspace,cite}
\usepackage{lscape,fancyhdr}
%\usepackage{A4}
\usepackage[top=2.5cm, left=3cm, right=3cm]{geometry}
%For commutative diagram
\usepackage[all,cmtip]{xy}
\usepackage{tikz-cd}
\usepackage{tkz-euclide}
%\usepackage[hmarginratio=1:1, vmarginratio =1:1,textheight=23cm,bindingoffset=0.0cm, textwidth=15cm]{geometry}
% for graphics
\usepackage{epsfig,graphicx,tikz,wrapfig,float %(for instant position of images)
}
\usepackage{caption,subcaption}
%\usepackage{hyperref}
%\usepackage[colorlinks=true, urlcolor=blue,bookmarks=true,bookmarksopen=true, citecolor=blue]{hyperref}
%\input{psfig}
%\usegepackege for bibliography
%\usepackage[backend=biber, style=alphabetic, sorting=nty]{biblatex}
%\addbibresource{Bibliography.bib} %Imports bibliography file
\usepackage[alphabetic, msc-links]{amsrefs}
\hypersetup{colorlinks=true, linkcolor=RoyalBlue, urlcolor=magenta, citecolor=RubineRed}
%Margins
\setlength{\unitlength}{0.4in}
\setlength{\textheight}{23cm}
\setlength{\textwidth}{15cm}
\renewcommand{\baselinestretch}{1.2}
\allowbreak
\newtheorem{thmA}{Theorem}
% Set up proclamation styles
\newtheorem{secn}{Definition}[section]
\newtheorem{thm}[secn]{Theorem}
\newtheorem{cor}[secn]{Corollary}
\newtheorem{prop}[secn]{Proposition}
\newtheorem{lem}[secn]{Lemma}
\newtheorem{defn}[secn]{Definition}
\newtheorem{ex}[secn]{Example}
\newtheorem{rem}[secn]{Remark}
\newtheorem{note}[secn]{Note}
\newtheorem{prob}[secn]{Problem}
\newtheorem{algo}[secn]{Algorithm}
\newtheorem{poss}[secn]{Possibility}
\newtheorem{case}[secn]{Case}
\newtheorem{obs}[secn]{Observation}
\numberwithin{equation}{secn}
\numberwithin{figure}{section}
%Set up new commands
%Special Symbol
%\newcommand{\bdry}{\partial}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{document}
%Description of Title Page
\title[Quasiconformal Maps between Bowditch Boundaries]{Quasiconformal Maps between Bowditch Boundaries of Relatively Hyperbolic Groups}
% Address of record for the research reported here
\author{Rana Sardar}
\address{Indian Institute of Technology, Kanpur}
\email{ranasardar20[at]iitk.ac.in}
% Information for second author
%\author{}
%\address{}
%\email{}
% General info
\subjclass[2000]{}
\date{}
%\dedicatory{}
\keywords{Quasi-Conformal Maps, Relatively Hyperbolic Groups, Bowditch Boundaries}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{abstract}
Classifying groups up to quasi-isometry is a fundamental problem in geometric group theory. In the context of hyperbolic and relatively hyperbolic groups, one of the key invariants in this classification is the boundary at infinity. F. Paulin proved that two hyperbolic groups are quasi-isometric if and only if their Gromov boundaries are quasiconformally equivalent. In this article, we extend Paulin's result to relatively hyperbolic groups and their Bowditch boundaries.
A notion of quasiconformal map preserving the shadows of horoballs relative to a point at the Bowditch boundary is defined and we have shown that every coarsely cusp-preserving quasi-isometry between two relatively hyperbolic groups induces a shadow-preserving quasiconformal map between their Bowditch boundaries. Conversely, we have shown that if the Bowditch boundaries of two relatively hyperbolic groups are quasiconformally equivalent and the quasiconformal map coarsely preserves the shadows of horoballs relative to each boundary point, then the quasiconformal map induces a coarsely cusp-preserving quasi-isometry between those groups.
\end{abstract}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\fontfamily{ptm}\selectfont
\maketitle
%\tableofcontents\newpage
\small{\begin{center}{AMS Subject Classification : 20F65, 20F67, 20E08, 51F30}\end{center}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Introduction}
The notion of hyperbolic metric spaces was introduced by M. Gromov \cite{Gro87}. A geodesic metric space is said to be hyperbolic if there exists a constant $\delta\geq 0$ such that for any geodesic triangle $\triangle$, each side of $\triangle$ is contained in a closed $\delta$-neighbourhood of the union of the other two sides.
A hyperbolic group is a finitely generated group whose Cayley graph, equipped with the word metric corresponding to some finite generating set, is a hyperbolic metric space. This definition is independent of the choice of finite generating sets, as hyperbolicity is a quasi-isometric invariant.
The Gromov boundary of a proper hyperbolic metric space provides a compactification of the space by assigning a `boundary' set at infinity.
A relatively hyperbolic group is a generalized notion of a hyperbolic group and has been defined in various ways by different mathematicians. For a survey of equivalent definitions, refer to the article \cite{Hruska} by Hruska. In this article, the definition given by Groves and Manning \cite{GM2008} is followed, which is as follows:
Let $G$ be a finitely generated group with a finite generating set $S$. Equip $G$ with the word metric $d_G,$ induced by $S$. Let $\{H_1, H_2,..., H_n\}$ be a finite collection of infinite-index finitely generated subgroups of $G$. Consider $X$, the Cayley graph of $G$ with respect to $S$, and let $\mathcal{H}_G$ be the collection of all left cosets of $H_1, H_2,..., H_n$.
For each coset $H\in \mathcal{H}_G$, the combinatorial horoball $H^h$ is analogous to a horoball in the hyperbolic space $\mathbb H^n$. The group $G$ is said to be $\delta$-hyperbolic relative to $\mathcal{H}_G$ if the space $X^h$, called the cusped space and obtained by attaching combinatorial horoballs $H^h$ to each $H\in \mathcal{H}_G$, is a $\delta$-hyperbolic metric space. Similar to the hyperbolic case relatively hyperbolic groups are also independent of the choice of finite generating sets. The Gromov boundary $\partial X^h$ of $X^h$ had been studied by Bowditch \cite{Bowditch}, so $\partial X^h$ is termed as the Bowditch boundary of $G$ and is often denoted by $\partial G^h$.
For a relatively hyperbolic group $(G,\mathcal{H}_G)$, the limit set of any $H\in\mathcal{H}_G$ is singleton in $\partial G^h$. A point $a \in \partial G^h$ is called a parabolic endpoint if there exists $H_{a}\in \mathcal{H}_G$ such that the limit set of $H_{a}$ is $\{a\}$.
A quasi-isometry between two hyperbolic groups induces a homeomorphism between their Gromov boundaries, while a coarsely cusp-preserving quasi-isometry between relatively hyperbolic groups induces a homeomorphism between their Bowditch boundaries.
However, the converse does not hold in both cases. There exist hyperbolic groups, due to Pansu, Bourdon and Pajot, with homeomorphic Gromov boundaries that are not quasi-isometric to each other.
For instance, if $G_1$ is the fundamental group of a closed real hyperbolic $4$-manifold and $G_2$ is the fundamental group of a closed complex hyperbolic $2$-manifold. Then $G_1$ and $G_2$ are hyperbolic and their Gromov boundaries $\partial G_1$ and $\partial G_2$ are both homeomorphic to the real $3$-dimensional unit sphere $\mathbb{S}^3$, but $G_1$ and $G_2$ are not quasi-isometric as the conformal dimensions of $\partial G_1$ and $\partial G_2$ are different (see \cite{Mackay-Tyson}*{Theorem 3.2.13, Theorem 3.3.3}).
Given any $k\geq 1$, Kolpakov and Martelli in \cite{Kolpakov-Martelli} constructed infinitely many real hyperbolic $4$-manifolds with $k$ number of cusps. Thus, given a finite-volume complex hyperbolic $2$-manifold $M$, one can construct a real hyperbolic $4$- manifold $N$ such that $M$ and $N$ have the same number of cusps. Then the fundamental groups of $M$ and $N$ are relatively hyperbolic and their Bowditch boundaries are homeomorphic to $\mathbb S^3$. But, similar to the case of closed manifolds, those groups are not quasi-isometric to each other.
In 1996, Paulin \cite{Pau96} proved that if the boundaries of two hyperbolic groups are homeomorphic and quasiconformal equivalent, then they are quasi-isometric to each other, and the converse also holds. There are multiple definitions of quasi-conformal maps; one was introduced by Pansu \cite{Pan1989}. He defined the definition for general spaces and Gromov boundary of a proper hyperbolic space is a particular case. This article only focuses on this particular case and ignores the general definition. Interestingly, this definition is independent of any metric structures of the Gromov boundary.
Let $X$ be a proper hyperbolic metric space with Gromov boundary $\partial X$. Fix a base point $\xi \in \partial X$. The shadow of a subset $Z$ of $X$ relative to $\xi$ is the set of endpoints of geodesics emanating from $\xi$ that intersect $Z$, and is denoted by $\mathcal{O}_\xi(Z)$.
A ring, analogous to an annulus, is a pair $\mathcal{O}_\xi B(x_1,R) \subseteq \mathcal{O}_\xi B(x_2,R)$ of shadows of balls of fixed radius $R > 0$. The collection of all such rings is called a ring structure on $\partial X$, and is denoted by $\mathcal{R}(\xi,R)$.
Let $X$ and $Y$ be two $\delta$-hyperbolic spaces for some $\delta\geq 0$, and $f \colon \partial X^h \to \partial Y^h$ be a homeomorphism.
Then, $f$ is said to be quasiconformal if there exists $R > 2\delta$ and a pair of ring structures $\mathcal{R}(\xi,R)$ and $\mathcal{R}(\xi',R)$ defined on the Gromov boundaries $\partial X^h$ and $\partial Y^h$, respectively, with $\xi' = f(\xi)$ such that $f \colon (\partial X^h, \mathcal{R}(\xi,R)) \to (\partial Y^h, \mathcal{R}(\xi',R))$ sends each ring in $\mathcal{R}(\xi,R))$ to a ring in $\mathcal{R}(\xi',R))$ with a controlled dilation, and $f^{-1}$ does same (see \autoref{defn_qc_maps}).
This article aims to generalize Paulin's result to relatively hyperbolic groups.
Let $(G,\mathcal{H}_G)$ be a relatively hyperbolic group with a Cayley graph $X$, and $\xi$ be a base point on the Bowditch boundary $\partial G^h$. Equip $\partial G^h$ with the ring structure $\mathcal{R}(\xi, R)$.
Then the shadow of each horoball, whose Gromov boundary is not equal to $\{\xi\}$, relative to $\xi$ can be covered by a unique ring (unique in the coarse sense) contained in the ring structure $\mathcal{R}(\xi, R)$ (see \autoref{lem_shadow_of_horospheres}).
Let $(G_1,\mathcal{H}_{G_1})$ and $(G_2,\mathcal{H}_{G_2})$ be two relatively hyperbolic groups. A quasiconformal map $f \colon (\partial G^h_1,\mathcal{R}(\xi,R)) \to (\partial G^h_2,\mathcal{R}(\xi',R))$ that preserves parabolic points and also `roughly' preserves the shadows of each horoball relative to the base point $\xi$, is said to coarsely preserve the shadows of horoballs relative to $\xi$ (refer to \autoref{defn_coarsely_shadows_of_horoballs_preserve}).
The following theorems are the main results of this article:
\begin{thmA} \label{thm_main_theorem_1}
Suppose that the numbers $\delta, C, K\ge 0$ and $\lambda \ge 1$ are given. If $(G_1,\mathcal{H}_{G_1})$ and $(G_2,\mathcal{H}_{G_2})$ are two $\delta$-relatively hyperbolic groups, and if $\varphi \colon G_1 \to G_2$ is a $C$-coarsely cusp-preserving $(\lambda,K)$-quasi-isometry, then the boundary extension map $\partial\varphi \colon \partial G_1^h \to \partial G_2^h$ satisfy the following: for any $R > 2\delta$ and for any pair of points $(\xi,\xi') \in \partial G_1^h \times \partial G_2^h$ with $\xi' = \partial \varphi (\xi)$, there exists a distortion map $\eta \colon [1,\infty) \to [1,\infty)$ and a constant $L \ge 0$ depending on $C, K, R,\lambda$ and $\delta$ such that $\partial\varphi \colon (\partial G_1^h, \mathcal{R}(\xi,R)) \to (\partial G_2^h, \mathcal{R}(\xi',R))$ %on their Bowditch boundaries equipped with the corresponding ring structures
is a $\eta$-quasiconformal homeomorphism, and $L$-coarsely preserves the shadows of horoballs relative to $\xi$ .
Moreover, the map $\eta$ is given by $\eta(t) = e^{A\ln(t)+B}$, where the constants $A \ge 1$ and $B \ge 0$ depend on $C, K, R,\lambda$, and $\delta$.
\end{thmA}
The left multiplication by a group element of a finitely generated group induces a cusp-preserving isometry (in particular, a $(1,0)$-quasi-isometry) on its Cayley graph. In view of \autoref{thm_main_theorem_1}, there exists a distortion map $\eta \colon [1,\infty)\to [1,\infty)$ such that for any element $g$ of a relatively hyperbolic group $G$, the left multiplication $L_g$ induces a $\eta$-quasiconformal map $\partial L_g^h\colon \partial G^h\to\partial G^h$. Thus, the action of a relatively hyperbolic group on itself induces a uniform quasiconformal action on its Bowditch boundary.
\begin{thmA} \label{thm_main_theorem_2}
Let $(G_1,\mathcal{H}_{G_1})$ and $(G_2,\mathcal{H}_{G_2})$ be two $\delta$-relatively hyperbolic groups, and $f\colon \partial G_1^h\to\partial G_2^h$ be a $\eta$-quasiconformal homeomorphism for some $\delta\geq 0$ and some distortion map $\eta \colon [1,\infty) \to [1,\infty)$ such that $f$ and $f^{-1}$ preserves the parabolic endpoints. Then
\begin{enumerate}[$i)$]
\item $f$ induces a quasi-isometry $\Phi f\colon G_1^h \to G_2^h$ between the cusped spaces, where the quasi-isometric constants of $\Phi f$ depend only on $\delta, f$ and $\eta$,
%\item $\Phi f$ is $K$-coarsely cusp-preserving, for some $K \ge 0$ that depends on $\delta$ and $\psi$,
\item the homeomorphism induced by $\Phi f$ between the Bowditch boundaries $\partial G_1^h$ and $\partial G_2^h$ coincides with $f$,
\item there exists a distortion map $\eta' \colon [1,\infty) \to [1,\infty)$ depending on $\delta, f$ and $\eta$ such that for any pair of points $(\zeta,\zeta') \in \partial G_1^h \times \partial G_2^h$ with $\zeta' = f (\zeta)$, the homeomorphism $f \colon (\partial G_1^h, \mathcal{R}(\zeta,R)) \to (\partial G_2^h, \mathcal{R}(\zeta',R))$ is $\eta'$-quasiconformal.
\end{enumerate}
In addition, if there exists $L \ge 0$ such that for each pair of base points $(\zeta,\zeta') \in \partial G_1^h \times \partial G_2^h$ with $\zeta' = \partial \varphi (\zeta)$, the map $f \colon (\partial G_1^h, \mathcal{R}(\zeta,R)) \to (\partial G_2^h, \mathcal{R}(\zeta',R))$ $L$-coarsely preserves the shadows of horoballs relative to $\zeta$. Then
\begin{enumerate}[$i)$]
\item $f$ induces a $C$-coarsely cusp-preserving quasi-isometry $\widetilde{\Phi f} \colon G_1 \to G_2$ between the groups, where the quasi-isometric constants of $\widetilde{\Phi f}$ and $C$ depend only on $\delta, L,f$ and $\eta$,
%\item $\Phi f$ is $K$-coarsely cusp-preserving, for some $K \ge 0$ that depends on $\delta$ and $\psi$,
\item the homeomorphism induced by $\widetilde{\Phi f}$ between the Bowditch boundaries $\partial G_1^h$ and $\partial G_2^h$ coincides with $f$.
\end{enumerate}
\end{thmA}
% another approches of this work
There are two more notions, namely quasisymmetric maps and quasi-M\"obius maps, similar to quasiconformal maps. In several cases, these three notions coincide. A quasisymmetric map between two metric spaces is a generalization of bi-Lipschitz maps and controls the ratios of distances of a pair of points. A quasi-M\"obius map is a homeomorphism that controls the cross-ratio of four distinct points.
In a hyperbolic metric space, corresponding to a geodesic triangle $\triangle (a,b,c)$ with three distinct vertices $a,b,c$ lying either in the space or on its Gromov boundary, there always exists a point $p_{abc}$, called a quasi-center or a quasi-projection, that is uniformly close to each side of $\triangle$. For any four distinct points $a,b,c,d$ on the Gromov boundary, the cross-ratio $[a,b,c,d]$ can be approximated in terms of the distance between the quasi-centers $p_{abc}$ and $p_{acd}$, with an error depending on the hyperbolicity constant of the space.
Mackay and Sisto, in \cite{Mackay-Sisto} proved that a shadow-respecting quasisymmetric homeomorphism between Bowditch boundaries of relatively hyperbolic groups induces a coarsely cusp-preserving quasi-isometry between the groups and vice versa (see \cite{Mackay-Sisto}*{Corollary 1.3}).
In a similar theme, Hruska and Healy \cite{Healy-Hruska}
independently proved that a coarsely cusp preserving quasi-isometry between relatively hyperbolic groups induces a quasi-isometry between their cusped spaces and a quasisymmetric map between their Bowditch boundaries.
Pal and Sardar \cite{PS2024} introduced the notion of a relative quasi-M\"obius homeomorphism, which is as follows: Given three distinct points $a,b,c$ on the Bowditch boundary, where $a$ is a parabolic endpoint of a horoball $H^h$. Then the relative cross-ratio, denoted by $[a,b,c]$, roughly captures the distance of a quasi-center of $(a,b,c)$ from the horosphere-like set $H$. A homeomorphism is called relative quasi-M\"obius if it coarsely preserves the relative cross ratios. They further showed that a relative quasi-M\"obius homeomorphism between Bowditch boundaries of two relatively hyperbolic groups induces a coarsely cusp-preserving quasi-isometry between those groups and vice versa (see \cite{PS2024}*{Theorem 1, Theorem 2}).
In view of \autoref{thm_main_theorem_1}, \autoref{thm_main_theorem_2}, \cite{Mackay-Sisto}*{Corollary 1.3}, and \cite{PS2024}*{Theorem 1, Theorem 2} the following equivalence concludes:
\begin{cor}
Let $(G_1,\mathcal{H}_{G_1})$ and $(G_2,\mathcal{H}_{G_2})$ be two relatively hyperbolic groups, and let $f: \partial G_1^h \to \partial G_2^h$ be a homeomorphism such that $f$ and $f^{-1}$ preserves the parabolic endpoints. Then the following are equivalent:
\begin{enumerate}[$(i)$]
\item $f$ is a quasiconformal map and coarsely preserves the shadows of horoballs relative to every boundary point $\zeta \in \partial G_1^h$ .
\item $f$ is a relative quasi-M\"obius map.
%\item $f$ distorts exit points linearly.
\item $f$ is a shadow-respecting quasisymmetry map.
\end{enumerate}
\end{cor}
%------------------------------------------------------------------%
\subsection{Outline} All the prerequisites related to the hyperbolic and relatively hyperbolic groups are described in \autoref{sec_Preliminaries}. The notion of - shadows of subsets, rings and quasiconformal maps - are defined, and the notion of quasiconformal map that coarsely preserves the shadows of horoballs is also introduced in \autoref{sec_qc_maps}. Finally, the \autoref{thm_main_theorem_1} and \autoref{thm_main_theorem_2} are proved in \autoref{sec_qi_implies_qc} and \autoref{sec_qc_implies_qi}, respectively.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Preliminaries} \label{sec_Preliminaries}
%-----------------------------------------------------------------------%
\subsection{Hyperbolic Spaces}
In a metric space $(X,d_X)$, a path is said to be rectifiable if it has a finite length. The metric $d_X$ is called a length metric if the distance between any two points $x,y \in X$ is equal to the infimum of the lengths of all rectifiable paths joining them. When $d_X$ is a length metric, the space $(X,d_X)$ is called a length space.
A geodesic between two points $x$ and $y$, denoted by $[x,y]$, is a continuous map $\gamma \colon [0,d_X(x,y)] \to X$ that satisfies $\gamma(0)=x$, $\gamma(d_X(x,y))=y$, and $d_X(\gamma(s),\gamma(t))=|s-t| \text{ for all } s,t\in[0,d_X(x,y)]$. The space $(X,d_X)$ is said to be a geodesic metric space if, for every pair of points $x,y \in X$, there exists a geodesic joining them.
A metric space $(X,d_X)$ is said to be proper if every closed ball in $X$ is compact.
\begin{defn}
Let $(X,d_X)$ be a geodesic metric space.
\begin{itemize}
\item Geodesic Triangle: A geodesic triangle in $X$ consists of three points $a,b,c \in X$ (called vertices) and three geodesic segments $[a,b], [b,c], [c,a]$ (called sides) joining them. A geodesic triangle with vertices $a,b,c \in X$ is denoted as $\triangle(a,b,c)$.
\item Slim Triangle: Let $\delta\ge 0$. Given $a,b,c \in X$, a geodesic triangle $\triangle(a,b,c)$ is said to be $\delta$-slim if any side of the triangle is contained within the closed $\delta$-neighbourhood of the union of the other two sides.
\end{itemize}
\end{defn}
% point $p\in X$ is said to be an $r$-center of a triangle $\triangle(x,y,z)$ if the distance from $p$ to each sides of $\triangle(x,y,z)$ is at most $r$. For a $\delta$-thin triangle $\triangle(x,y,z)$, in the above definition, the points $i_a,i_b,i_c$ are $\delta$-centers of $\triangle(x,y,z)$. If $r$ remains fixed for all triangles in a geodesic space then we call the $r$-centers to be quasi-centers.
\begin{defn}[Hyperbolic Metric Spaces]
A geodesic metric space $(X,d_X)$ is said to be $\delta$-hyperbolic, where $\delta \ge 0$, if every geodesic triangle in $X$ is $\delta$-slim. The geodesic metric space is said to be hyperbolic if it is $\delta$-hyperbolic for some $\delta\ge0$.
\end{defn}
Any metric space with finite diameter, the unit disk in the complex plane with respect to the Poincare metric, trees, and Cayley graph of a fundamental group of closed genus $g$ surfaces for $g \ge 2$ are basic examples of hyperbolic spaces. On the other hand, any metric spaces containing a lattice isometric to $\mathbb{Z} \times \mathbb{Z}$, in particular, the Euclidean spaces $\mathbb{R}^n$ for $n \geq 2$ are not hyperbolic.
\begin{defn}[Quasi-isometries and Quasigeodesics]
Let $\lambda \ge 1$ and $K \ge 0$. A (not necessarily continuous) map $f\colon X \to Y$, between two metric spaces $(X,d_X)$ and $(Y,d_Y)$, is said to be $(\lambda, K)$-quasi-isometric embedding if for all $x_1,x_2 \in X$ the following holds
$$\frac{1}{\lambda} d_X(x_1,x_2) - K \le d_Y(f(x_1),f(x_2)) \le \lambda d_X(x_1,x_2) + K.$$
Moreover, if $f$ is coarsely surjective, --- i.e., for every $y \in Y$ there exists an $x\in X$ satisfying $d_Y(f(x),y)\le K$) --- then $f$ is called a $(\lambda, K)$-quasi-isometry. $X$ and $Y$ are said to be quasi-isometric if there exists a quasi-isometry between them.
A quasigeodesic in $X$ is a $(\lambda, K)$-quasi-isometric embedding $\gamma \colon I\to X$, where $I$ is an interval (which may not be closed or bounded) in the real line $\mathbb{R}$. A $(\lambda,\lambda)$-quasigeodesic is called a $\lambda$-quasigeodesic.
\end{defn}
\begin{prop}[Stability of Quasigeodesics, \cite{BH99}] \label{prop_stability_of_quasigeodesics}
For all $\delta , K \ge 0$ and $\lambda \ge 1$ there exists a constant $C_1=C_1(\delta,\lambda,K)$ with the following property:
If $X$ is a $\delta$-hyperbolic space, $\gamma$ is a $(\lambda,K)$-quasigeodesic in $X$ and $[x,y]$ is a geodesic segment joining the endpoints $x$ and $y$ of $\gamma$, then $\gamma$ and $[x,y]$ lie within a $C_1$-neighbourhood of each other.
\end{prop}
\begin{thm}[Quasi-isometries Preserve Hyperbolicity]
Let $X$ and $Y$ be two proper geodesic metric spaces and $f\colon X \to Y$ be a quasi-isometry. If one of these spaces is hyperbolic, then the other must also be hyperbolic.
\end{thm}
%------------------------------------------------------%
\subsection{Hyperbolic Groups}
A group $G$ is said to be \textit{finitely generated} if there exists a finite subset $S \subseteq G$ such that every elements in $G$ can be expressed as a finite product of elements of $S$ and their inverses. A finitely generated group may have multiple finite generating sets. Each finite generating set $S$ of $G$ defines a Cayley graph, denoted by $Cay(G,S)$, which is also a metric space with respect to the word metric. If $S_1$ and $S_2$ are two finite generating sets of $G$, then the Cayley graphs $Cay(G,S_1)$ and $Cay(G,S_2)$ are quasi-isometric. Therefore, the following definition is independent of the choice of generating sets.
\begin{defn}[Hyperbolic Groups, \cites{Gro87,BH99}]
A finitely generated group $G$ is said to be hyperbolic if its Cayley graph with respect to some (and hence every) finite generating set is a hyperbolic metric space.
\end{defn}
\begin{prop}[\cite{BH99}*{Chapter III.$\Gamma$, Proposition 2.2}] Hyperbolic groups are finitely presented.
\end{prop}
%------------------------------------------------------------------%
%\bigskip
\subsection{Gromov Boundaries}
\begin{defn}[Gromov Boundaries]
Two geodesic rays $\gamma_1,\gamma_2\colon [0,\infty) \to X$ in a hyperbolic metric space $(X,d_X)$ are said to be asymptotic if $sup\{d_X(\gamma_1(t)\gamma_2(t)) \mid t \ge 0\}$ is finite. Being asymptotic is an equivalence relation on geodesic rays in $X$. For each geodesic ray $\gamma$ the equivalence class is denoted by $[\gamma]$.
Let $w\in X$ be a base point. The Gromov boundary of $X$ based at $w$ is defined by
$$\partial_w X \coloneqq \{[\gamma] \mid \gamma\colon [0,\infty)\to X\ \text{is a geodesic ray with}\ \gamma(0)=w\}.$$
The Gromov boundary of a hyperbolic group $G$ is the Gromov boundary of its Cayley graph.
\end{defn}
\begin{defn}
For any $a \in \partial_w X$ and $r\ge 0$ the set is defined as
\begin{align*}
V(a,r) \coloneqq \{ b\in\partial_wX \mid \liminf_{t\to\infty} (\gamma_1(t),\gamma_2(t) )_w \ge r, \text{ for some $\gamma_1,\gamma_2$ with $[\gamma_1]=a, [\gamma_2]=b$} \}
\end{align*}
\end{defn}
\begin{lem}
The family $\{V(a,r) \mid r \ge 0\}$ forms a basis of neighbourhoods at $a\in \partial_w X$.
\end{lem}
\begin{lem}[Visibility of $\partial X$, \cite{BH99}]
If $X$ is proper $\delta$-hyperbolic metric space, then for each pair of distinct points $\xi_1,\xi_2\in \partial X$ there exist a geodesic line $c\colon \mathbb{R} \to X$ with $c(-\infty)=\xi_1$ and $c(+\infty)=\xi_2$.
\end{lem}
\begin{defn}[Gromov Products]
Let $X$ be a $\delta$-hyperbolic space and $w \in X$ be a base point, where $\delta \ge 0$. The Gromov product for a pair of points $a,b$ in $X$ is denoted by $(a,b)_w$ and is defined as $$(a,b)_w = \frac{1}{2}(d_X(a,w) + d_X(b,w) - d_X(a,b)).$$
Whenever, $a,b \in X \cup \partial X$, the gromov product is defined by
$$(a,b)_w = \sup \liminf_{i,j \to \infty}(a_i,b_j),$$
where the supremum is taken over all the sequences $\{a_i\}$ and $\{b_j\}$ such that $a_i \to a$ and $b_j \to b$.
\end{defn}
\begin{lem}
There exists $C_2=C_2(\delta) \ge 0$ such that for any $a,b \in X \cup \partial X$ with $a \ne b$, $(a,b)_w \approx_{C_2} d_X(w,[a,b])$.
\end{lem}
%-------------------------------------------------------------%
%Required results
\begin{thm}[Quasi-isometry Invariants, \cite{BH99}]\label{thm_qi_implies_homeomorphism}
Let $X$ and $Y$ be two proper hyperbolic geodesic spaces. If $f\colon X\to Y$ is a quasi-isometric embedding, then $\partial f\colon \partial X\to \partial Y$ defined by $c(\infty)\mapsto (f\circ c)(\infty)$ is an embedding. Moreover, if $f\colon X \to Y$ is a quasi-isometry, then $\partial f$ is a homeomorphism.
\end{thm}
\begin{lem}\label{lem_bounded_diameter_of_points_on_a_geodesic}
For each $r, \delta \ge 0$, there exists $C_3=C_3(r,\delta) \ge 0$ such that if $X$ is a proper $\delta$-hyperbolic space, then for any geodesic triangle $\triangle(a, b, c)$ with vertices $a,b,c$ in $X \cup \partial X$, the diameter of the set of points on the geodesic $[a, c]$ which are within $r$ distance from both sides $[a, b]$ and $[c, b]$ of $\triangle(a, b, c)$, is at most $C_3$.
\end{lem}
\begin{proof}
Let $p, q$ be two points on the geodesic $[a, c]$ at most $r$ distance from both the sides $[a, b]$ and $[c, b]$. Then there are points $p_1$ and $q_1$ on $[a, b]$, and $p_2$ and $q_2$ on $[c, b]$, such that $d_Z(p, p_i)\le r$ and $d_Z(q, q_i)\le r$ for $i=1,2$. Without loss of generality assume that $p$ is closer to $a$ than $q$.
\begin{figure}[H]
\centering
\begin{tikzpicture}[scale=.5]
\draw (0,0)--(8,0)--(4,6)--(0,0);
\draw (4,6)--(3,0);
\draw[gray] (4,6)--(5,0);
\draw[dashed] (3,0)--(1,1.4);
\draw[gray,dashed] (5,0)--(1.5,{9/4});
\draw[dashed] (3,0)--(7,{3/2});
\draw[gray,dashed] (5,0)--(6.5,{9/4});
\node at (-.5,0) {$a$};
\node at (4,6.5) {$b$};
\node at (8.5,0) {$c$};
%\fill[blue] (3,2) circle (1.5pt);
%\fill[magenta] (4.3,2) circle (1.5pt);
\node at (3,-.5) {$p$};
\node at (5,-.5) {$q$};
\node at (.4,1.5) {$p_1$};
\node at (1,2.5) {$q_1$};
\node at (7.7,1.5) {$p_2$};
\node at (7,2.5) {$q_2$};
\end{tikzpicture}
\caption{}
\end{figure}
Case-$1$: Let $a,b,c$ be in $Z$. Consider the triangle $\triangle(a,q,b)$, then $$d_X(a,q) + d_X(q,b) \le d_X(a,b) + 2r \le d_X(a,p) + d_X(p,b) + 2r.$$
This implies that
$$d_X(p,q)+d_X(q,b)\le d_X(p,b)+2r.$$
\noindent Similarly, from the triangle $\triangle(b,p,c)$,
$$d_X(p,q)+d_X(p,b)\le d_X(q,b)+2r.$$
\noindent Hence, $$d_X(p,q)\le 2r.$$
\noindent Case-$2$: If at least one of $a,b,c$ is in $\partial Z$, consider the sequences $\{a_n\}$, $\{b_n\}$, $\{c_n\}$ in $Z$ such that $a_n\to a$, $b_n\to b$, $c_n\to c$, and $a_n,c_n\in[a,c]$. Then there is a positive integer $n$ such that $p,q\in[a,_n,c_n]$, $p_1,q_1$ are at most $3 \delta$ distance from $[a_n,b_n]$ and $p_2,q_2$ are at most $3 \delta$ distance from $[b_n,c_n]$. Hence $p,q$ are at most $r + 3 \delta$ distance from both the sides $[a_n,b_n]$ and $[b_n,c_n]$ of the triangle $\triangle(a_n,b_n,c_n)$. Then by case-$1$, $$d_X(p,q) \le 2 (r + 3 \delta).$$
\end{proof}
%--------------------------------------------------------------------%
\subsection{Relatively Hyperbolic Groups}
\begin{defn}[Combinatorial Horoballs, \cite{GM2008}]
Let $(H,d_H)$ be a metric space. The combinatorial horoball based on $H$, denoted by $\mathcal{H}(H,d_H)$ is the graph defined as follows:
\begin{enumerate}
\item $\mathcal{H}^{(0)}= H \times \mathbb{N}\cup\{0\}$.
\item $\mathcal{H}^{(1)}$ contains the following two types of edges
\begin{enumerate}
\item For each $n \ge 0$ and $x \in H$, there is a vertical edge connecting $(x,n)$ and $(x,n+1)$.
\item For each $n \ge 1$ and $x,y \in H$, if $0<d_H(x,y)\le 2^n$ then there is a horizontal edge connecting $(x,n)$ and $(y,n)$.
\end{enumerate}
\end{enumerate}
\end{defn}
Endowed $\mathcal{H}(H,d_H)$ with the induced length metric, say $d_\mathcal{H}$, and consider all the edges have unit length. Then $(\mathcal{H}(H,d_H),d_{\mathcal{H}})$, is a hyperbolic metric space (see \cite{GM2008}). For convenience, hereafter this hyperbolic space is denoted by $(H^h,d_{H^h})$ and is termed `horoball' instead of combinatorial horoball. The topological boundary of $H^h$ is $H$, and is called a horosphere-like.
\begin{defn}[Relatively Hyperbolic Groups] \label{defn_rel_hyp}
Let $G$ be a finitely generated group, and $\{H_1,\dots,H_n\}$ be a finite collection of infinite index finitely generated subgroups of $G$.
Let $X$ be a Cayley graph of $G$ with respect to some finite generating set, and $\mathcal{H}_G$ be the collection of all left cosets of the subgroups $H_1, H_2,..., H_n$. The group $G$ is said to be $\delta$-hyperbolic relative to $\{H_1,H_2,...,H_n\}$ or simply $(G,\mathcal{H}_G)$ is a $\delta$-relatively hyperbolic group if the cusped space (or sometimes say coned off Cayley graph) $X^h$, obtained by attaching combinatorial horoball $H^h$ to each $H\in \mathcal{H}_G$, is a $\delta$-hyperbolic metric space. The group $G$ is said to be hyperbolic relative to $\{H_1,H_2,...,H_n\}$ if there exists $\delta\geq 0$ such that $(G,\mathcal{H}_G)$ is a $\delta$-relatively hyperbolic group.
The members of $\mathcal{H}_G$ are called horosphere-like sets, or simply horospheres, and conjugates of $H_1,\dots,H_n$ are called peripheral subgroups of $G$.
\end{defn}
Drutu \cite{Drutu} proved that the relative hyperbolicity is a quasi-isometric invariant. Thus, the above \autoref{defn_rel_hyp} of a relatively hyperbolic group is independent of the (finite) generating sets.
\begin{defn}[Coarsely Cusp-preserving Maps] \label{defn_cusp_preserving}
Let $(G_1,\mathcal{H}_{G_1})$ and $(G_2,\mathcal{H}_{G_2})$ be two relatively hyperbolic groups with Cayley graphs $X$ and $Y$, respectively. For a given constant $C \ge 0$, a bijection $\varphi\colon X^h \to Y^h$ is said to be $C$-coarsely cusp-preserving if
\begin{enumerate}[$i)$]
\item for any $H \in \mathcal{H}_{G_1}$ there exists $H' \in \mathcal{H}_{G_2}$ such that the image $\varphi(H)$ contained in the $C$-neighbourhood of $H'$, and similarly
\item for any $H' \in \mathcal{H}_{G_2}$ there exists $H \in \mathcal{H}_{G_1}$ such that the inverse image $\varphi^{-1}(H')$ is contained in the $C$-neighbourhood of $H$.
\end{enumerate}
A bijection $\varphi \colon X^h \to Y^h$ is called coarsely cusp-preserving if it is $C$-coarsely cusp-preserving for some $C \ge 0$.
\end{defn}
It is known that a coarsely cusp-preserving quasi-isometry between two relatively hyperbolic groups induces a quasi-isometry between their cusped spaces (see \cite{Pal}*{Lemma 1.2.31}, \cite{Groff}*{Corollary 6.5}).
\begin{thm} \label{thm_cusp_presv_qi_implies_homeomorphism}
Let $\delta,C,K \ge 0$ and $\lambda \ge 1$. If $(G_1,\mathcal{H}_{G_1})$ and $(G_2,\mathcal{H}_{G_2})$ are two relatively hyperbolic groups with Cayley graphs $X$ and $Y$, respectively, and if $\varphi \colon G_1 \to G_2$ is a $C$-coarsely cusp-preserving $(\lambda, K)$-quasi-isometry, then $\varphi$ induces a $(\lambda',K')$-quasi-isometry $\varphi^h \colon X^h \to Y^h$, where the constants $\lambda' \ge 1, K' \ge 0$ are depend on $\delta,C,K$ and $\lambda$.
\end{thm}
Hence, by \autoref{thm_qi_implies_homeomorphism} the boundary extension map $\partial \varphi^h \colon \partial X^h \to \partial Y^h$ is a homeomorphism, and hence the Gromov boundaries of cusped spaces are homeomorphic. Thus, the Gromov boundary of a cusped space is a suitable candidate to be the boundary of a relatively hyperbolic group, and this boundary is known in the literature as the Bowditch boundary.
%\subsection{Bowditch Boundaries}
\begin{defn}[Bowditch Boundaries]
Let $(G,\mathcal{H}_G)$ be a relatively hyperbolic group with a Cayley graph $X$. Then the Gromov boundary $\partial G^h$ or $\partial X^h$ of the cusped space $X^h$ is called the Bowditch boundary of $G$.
A point $a \in \partial G^h$ is said to be parabolic if it is a Gromov boundary of a horoball (or combinatorial horoball) $H^h$ for some $H \in \mathcal{H}_G$.
\end{defn}
The following properties of uniform boundedness satisfied by a horoball in a relatively hyperbolic group will be used later (see \cite{Pal}*{Lemmas 1.2.38, Lemma 1.2.39}, \cite{Farb}*{Lemma 4.4}).
\begin{thm}[Visually Bounded] \label{thm_visually_bounded}
Let $(G,\mathcal{H}_G)$ be a $\delta$-relatively hyperbolic group for some $\delta \ge 0$. Then there exists a constant $C_4 = C_4(\delta) \ge 0$ such that
\begin{enumerate}[$i)$]
\item horoballs are visually bounded; --- if $H^h $ is a horoball and $x\notin H^h$, then the diameter of the set of first entry points of geodesics emanating from $x$ to $H^h$ is at most $C_4$, and
\item the diameter of the image of a geodesic $\gamma$ under nearest point projection onto a horoball that is disjoint from $\gamma$ is at most $C_4$.
\end{enumerate}
\end{thm}
\begin{lem}[Uniformly Properly Embedded Lemma \cite{Pal}*{Lemmas 1.2.12, 1.2.19})] \label{lem_Uniformly_Properly_embedded_Lemma}
Let $G$ be a relatively hyperbolic group with Cayley graph $X$, then $G$ and the family $\{H_n = H \times \{n\} \colon H \in \mathcal{H}_G, \ n \ge 0\}$ are uniformly properly embedded in $X^h$. That is, for any $r \ge 0$, there exists $C_8 = C_8(r,\delta)$ such that
\begin{enumerate}[$i)$]
\item if $x,y \in X$ with $d_{X^h}(x,y) \le r$, then $d_G(x,y) \le C_8$, and
\item if $x,y \in H_n$ for $n \ge 1$ with $d_{X^h}(x,y) \le r$, then $d_{H_n}(x,y) \le C_8$.
\end{enumerate}
\end{lem}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\newpage
\section{Quasiconformal Maps on Boundaries at Infinity} \label{sec_qc_maps}
\subsection{Ring Structures on Boundaries at Infinity}
In the complex plane, an annulus is ideally defined as the set of points lying between two concentric circles centered at the same point. However, from a different perspective, an annulus can also be interpreted as a pair of balls with the same center such that one is contained in the other. A ring is analogous to an annulus was defined by Pansu \cite{Pan1989} based on the concept of shadows of balls and using this defined the quasi-conformal maps. His definitions are defined for general spaces and the Gromov boundary of a proper hyperbolic space is a particular case. This article only focuses on this particular case and ignores the general definition. Interestingly, this definition is independent of any metric structures on the Gromov boundaries.
Note that if the Gromov boundary of a hyperbolic metric space contains only one point, then every structure becomes trivial. Therefore, hereafter it is assumed that the boundary at infinity contains at least two points.
\begin{defn}
Let $(X,d_X)$ be a proper hyperbolic metric space, and let $\xi$ be a base point on the Gromov boundary $\partial X$. Then the shadow of a subset $Z \subseteq X$ relative to $\xi$, denoted by $\mathcal{O}_\xi(Z)$, is the set of endpoints (not $\xi$) of geodesics emanating from $\xi$ that intersect $Z$. Formally,
$$\mathcal{O}_\xi(Z) = \{a \in \partial X \setminus \{\xi\} \mid \text{there exists a geodesic } [\xi,a], \text{ that intersects } Z\}.$$
\begin{figure}[H]
\centering
\begin{tikzpicture}
\filldraw[gray!70] ({2.5*cos 330},{2.5*sin 330}) arc (60:81.8:13);
\filldraw[gray!70] ({2.5*cos 330},{2.5*sin 330}) -- ({2.5*cos 171.5},{2.5*sin 171.5}) -- ({2.5*cos 198.3},{2.5*sin 198.3});
\filldraw[gray!70] ({2.5*cos 171.5},{2.5*sin 171.5}) arc (171.5:198.3:2.5);
\filldraw[gray!30] ({2.5*cos 330},{2.5*sin 330}) arc (60:75:13);
\filldraw[gray!30] ({2.5*cos 330},{2.5*sin 330}) -- (-1,0.07) -- (-1.1,-.53);
%\filldraw[gray!40] ({2.5*cos 171.5},{2.5*sin 171.5}) arc (171.5:198.3:2.5);
\filldraw[white] ({2.5*cos 330},{2.5*sin 330}) arc (60:109:5.5);
\draw[white,thick] ({2.5*cos 330},{2.5*sin 330})--({2.5*cos 199},{-.1+2.5*sin 199});
\draw[line width =.5 mm] ({2.5*cos 171.5},{2.5*sin 171.5}) arc (171.5:199:2.5);
%\filldraw[gray] (-.37,-.353) circle (.3cm);
\filldraw[gray!140] (-1.1,-.23) circle (.3cm);
%\draw (-.37,-.353) circle (.3cm);
\draw (-1.1,-.23) circle (.3cm);
%\filldraw (-.37,-.32) circle (1pt);
%\filldraw (-1.1,-.21) circle (1pt);
\draw[dashed] ({2.5*cos 330},{2.5*sin 330}) arc (60:81.8:13);
\draw[dashed] ({2.5*cos 330},{2.5*sin 330}) arc (60:88.1:10);
\draw[dashed] ({2.5*cos 330},{2.5*sin 330}) arc (60:94.8:8);
\draw[dashed] ({2.5*cos 330},{2.5*sin 330}) arc (60:102:6.5);
\draw[dashed] ({2.5*cos 330},{2.5*sin 330}) arc (60:109:5.5);
%\draw ({2.5*cos 330},{2.5*sin 330}) arc (60:118:4.5);
\draw (0,0) circle (2.5 cm);
\node at ({.25+2.5*cos 330},{-.2+2.5*sin 330}) {$\xi$};
\node[white] at (-1.1,-.21) {$Z$};
\node at ({-.75+2.5*cos 180},{-.2+2.5*sin 180}) {$\mathcal{O}_{\xi}(Z)$};
%to do centre the image
\node[white] at ({.75-2.5*cos 180},{.2-2.5*sin 180}) {$\mathcal{O}_{\xi}(Z)$};
%\draw[-stealth] (-.37,-.32) --(.5,.6);
%\draw[-stealth] (-1.1,-.23) --(-.23,.6);
%\node at (-.03,.8) {$x_1$};
%\node at (.7,.8) {$x_2$};
\end{tikzpicture}
\caption{Shadow of a ball}
\end{figure}
%The point $\xi$ can be thought as a source of lights.
\end{defn}
\begin{lem} \label{lem_Existence_of_rings}
Let $(X,d_X)$ be a proper $\delta$-hyperbolic space. Suppose $R > 2\delta$ is a positive real number, and $a \in \partial X \setminus \{ \xi \}$. Then for any $x,x_1$ and $x_2$ on the geodesic $[\xi,a]$,
\begin{enumerate}[$i)$]
\item if $x_1 \in [x,a]$ and $d_X(x_1,x) \ge R+\delta$, then $\mathcal{O}_\xi B(x_1,R) \subseteq \mathcal{O}_\xi B(x,R).$
\item if $x_2 \in [x, \xi]$ and $d_X(x,x_2) \ge R+\delta$, then $\mathcal{O}_\xi B(x,R) \subseteq \mathcal{O}_\xi B(x_2,R).$
\end{enumerate}
\end{lem}
\begin{proof}
\begin{figure}[H]
\centering
\begin{tikzpicture}
\draw (-5,0)--(7,0);
\filldraw (0,0) circle (1pt);
\filldraw (-4,0) circle (1pt);
\filldraw (4,0) circle (1pt);
\draw (0,0) circle (.8cm);
\draw (-4,0) circle (.8cm);
\draw (4,0) circle (.8cm);
\draw[bend right=3] (-5,.8) edge (7,0);
\draw[bend left=3] (-3,-1.2) edge (7,0);
\node at (0,-.3) {$x$};
\node at (-4,-.3) {$x_1$};
\node at (4,.3) {$x_2$};
\node at (7.3,0) {$\xi$};
\end{tikzpicture}
\caption{}
\end{figure}
Let $b \in \mathcal{O}_\xi B(x_1,R)$ and suppose that the geodesic $[\xi, b]$ intersects the ball $B(x_1,R)$ at a point $z$. Consider the geodesic triangle $\triangle(x_1,z,\xi)$. By the $\delta$-slimness of geodesic triangles, the point $x$ is $2\delta$-close to the geodesic ray $[z,\xi]$. This implies that $[\xi,b]$ intersects the ball $B(x,2\delta)$, which is contained within the ball $B(x,R)$. Hence, $b \in \mathcal{O}_\xi B(x,R)$, and the assertion $(i)$ is proved.
The proof of $(ii)$ follows similarly to the proof of $(i)$, by replacing $x_1$ and $x$ with $x$ and $x_2$, respectively.
\end{proof}
%\begin{defn}
%The set of shadows of every balls of $X$ from the base point $\xi$ is denoted by $Conf(\xi,R)$ and is said to be a quasiconformal structure on $\partial X \setminus \{\xi\}$.
%\end{defn}
\begin{defn}[Rings] \label{defn_annulus_by_shadows_from_boundary}
Given $\delta \ge 0$ and $R > 2\delta$. Let $(X,d_X)$ be a proper $\delta$-hyperbolic metric space, and let $\xi \in \partial X$ be a base point. A ring is a pair of shadows $\mathcal{O}_\xi B(x_1,R) \subseteq \mathcal{O}_\xi B(x_2,R)$, where $B(x_i,R)$ denotes the ball of radius $R$, centered at $x_i \in X$, and $x_1, x_2$ and $\xi$ are aligned (i.e., lie in a single geodesic) with $x_2$ being closer to $\xi$ than $x_1$. (Refer to the \autoref{fig_ring})
\begin{figure}[H]
\centering
\begin{tikzpicture}
\filldraw[gray!40] ({2.5*cos 330},{2.5*sin 330}) arc (60:81.8:13);
\filldraw[gray!40] ({2.5*cos 330},{2.5*sin 330}) -- ({2.5*cos 171.5},{2.5*sin 171.5}) -- ({2.5*cos 199},{2.5*sin 199});
\filldraw[gray!40] ({2.5*cos 171.5},{2.5*sin 171.5}) arc (171.5:198.3:2.5);
\filldraw[white] ({2.5*cos 330},{2.5*sin 330}) arc (60:109:5.5);
\draw[white,thick] ({2.5*cos 330},{2.5*sin 330})--({2.5*cos 199},{-.1+2.5*sin 199});
\draw[line width =.5 mm] ({2.5*cos 171.8},{2.5*sin 171.8}) arc (171.8:164.8:2.5);
\draw[line width =.5 mm] ({2.5*cos 198.8},{2.5*sin 198.8}) arc (198.8:208.3:2.5);
\filldraw[gray] (-.37,-.353) circle (.3cm);
\filldraw[gray] (-1.1,-.23) circle (.3cm);
\draw (-.37,-.353) circle (.3cm);
\draw (-1.1,-.23) circle (.3cm);
\filldraw (-.37,-.32) circle (1pt);
\filldraw (-1.1,-.21) circle (1pt);
\draw ({2.5*cos 330},{2.5*sin 330}) arc (60:75:19);
\draw[dashed] ({2.5*cos 330},{2.5*sin 330}) arc (60:81.8:13);
\draw ({2.5*cos 330},{2.5*sin 330}) arc (60:94.8:8);
\draw[dashed] ({2.5*cos 330},{2.5*sin 330}) arc (60:109:5.5);
\draw ({2.5*cos 330},{2.5*sin 330}) arc (60:118:4.5);
\draw (0,0) circle (2.5 cm);
\node at ({.25+2.5*cos 330},{-.2+2.5*sin 330}) {$\xi$};
\node at ({-.35+2.5*cos 180},{-.2+2.5*sin 180}) {$a$};
\draw[-stealth] (-.37,-.32) --(.5,.6);
\draw[-stealth] (-1.1,-.23) --(-.23,.6);
\node at (-.03,.8) {$x_1$};
\node at (.7,.8) {$x_2$};
\end{tikzpicture}
\caption{Ring}
\label{fig_ring}
\end{figure}
Moreover, if $d_X(x_1,x_2) = \log(t)$ for some $t \ge 1$, then the ring $\mathcal{O}_\xi B(x_1,R) \subseteq \mathcal{O}_\xi B(x_2,R)$ is called a $t$-ring, and sometimes it is denoted either by $\mathcal{O}_\xi B(x_1,R) \subseteq t\mathcal{O}_\xi B(x_1,R)$ or by $\frac{1}{t}\mathcal{O}_\xi B(x_2,R) \subseteq \mathcal{O}_\xi B(x_2,R)$. If $x_1$ and $x_2$ lie on the geodesic $[\xi,a]$ for some $a \in \partial X \setminus \{ \xi \}$, then the ring is said to be centered at $a$.
%In this case, the shadow $\mathcal{O}_\xi B(x_2,R)$ is denoted by $t \mathcal{O}_\xi B(x_1,R)$ and the shadow $\mathcal{O}_\xi B(x_1,R)$ is denoted by $\frac{1}{t} \mathcal{O}_\xi B(x_2,R)$. That is, each of the following represents a $t$-ring
%\begin{itemize}
%\item $\mathcal{O}_\xi B(x_1,R) \subseteq \mathcal{O}_\xi B(x_2,R)$ with $d_X(x_1,x_2) = \log(t)$, and $x_1,x_2$ and $\xi$ are aligned,
%\item $\mathcal{O}_\xi B(x_1,R) \subseteq t\mathcal{O}_\xi B(x_1,R)$,
%\item $\frac{1}{t}\mathcal{O}_\xi B(x_2,R) \subseteq \mathcal{O}_\xi B(x_2,R)$.
%\end{itemize}
\end{defn}
\begin{defn}[Ring Structures]
The set of all possible rings on $\partial X \setminus \{\xi\}$ based on the shadows of balls relative to the base point $\xi$ is denoted by $\mathcal{R}(\xi,R)$ and is said to be a ring structure on $\partial X $. %; that is $$\mathcal{R}(\xi,R) = \{B \subseteq t B \mid B = \mathcal{O}_\xi B(x,R) , x \in X, t \ge 1\}.$$
\end{defn}
%Therefore, for different base points and different radii, there are different ring structures on the Gromov boundaries.
%---------------------------------------------------------------%
%\subsection{Quasiconformal Maps}
\begin{defn}[Quasiconformal Maps] \label{defn_qc_maps}
Let $(X,d_X)$ and $(Y,d_Y)$ be two proper $\delta$-hyperbolic metric spaces, and $\mathcal{R}(\xi,R)$ and $\mathcal{R}(\xi',R)$ be two ring structures on their Gromov boundaries $\partial X$ and $\partial Y$, respectively, where $\delta
\ge 0$, $(\xi,\xi') \in \partial X \times \partial Y$ and $R > 2\delta$. Let $\eta \colon [1,\infty) \to [1,\infty)$ be a continuous map, called a distortion map. A homeomorphism $f \colon ( \partial X, \mathcal{R}(\xi,R)) \to (\partial Y, \mathcal{R}(\xi',R))$ with $f(\xi) = \xi'$ is said to be $\eta$-quasiconformal if
for any $t$-ring in $B \subseteq tB$ in $\mathcal{R}(\xi,R)$ there exists an $\eta(t)$-ring $B' \subseteq \eta(t)B'$ in $\mathcal{R}(\xi',R)$ such that
$$B' \subseteq f(B) \subseteq f(tB) \subseteq \eta(t)B',$$
and for any $t$-ring $B' \subseteq tB'$ in $\mathcal{R}(\xi',R)$ there exists an $\eta(t)$-ring $B \subseteq \eta(t)B$ in $\mathcal{R}(\xi,R)$ such that
$$B \subseteq f^{-1}(B') \subseteq f^{-1}(tB') \subseteq \eta(t)B.$$
A homeomorphism $f\colon \partial X \to \partial Y$ is said to be quasiconformal if it is quasiconformal with respect to some ring structures defined on $\partial X$ and $\partial Y$; that is, there exists a constant $R > 2\delta$ and a pair of base points $\xi.\xi'$ with $\xi'=f(\xi)$ and a distortion map $\eta \colon [1,\infty) \to [1,\infty)$ such that $f \colon ( \partial X, \mathcal{R}(\xi,R)) \to (\partial Y, \mathcal{R}(\xi',R))$ with $f(\xi) = \xi'$ is said to be $\eta$-quasiconformal.
\end{defn}
\begin{defn}[Equivalent Ring Structures]
Two ring structures $\mathcal{R}_1$ and $\mathcal{R}_2$ defined on the Gromov boundary $\partial X$ of a $\delta$-hyperbolic metric space $X$ are said to be $\eta$-equivalent if the identity map $Id_{\partial X} \colon (\partial X, \mathcal{R}_1) \to (\partial X, \mathcal{R}_2)$ is $\eta$-quasiconformal, for some distortion map $\eta \colon [1,\infty) \to [1,\infty)$.
\end{defn}
\begin{prop} \label{prop_ring_structures_are_independent_on_R}
Any two ring structures $\mathcal{R}(\xi,R)$ and $\mathcal{R}(\xi,R')$ defined on the Gromov boundary $\partial X$ for different radii $R,R' > 2\delta$ are equivalent. Moreover, there exists a constant $C_5=C_5(\delta,R,R') \ge 1$ such that the corresponding distortion map $\eta \colon [1,\infty) \to [1,\infty)$ is defined by $\eta(t) = C_5t$, for all $t \in [1,\infty)$.
\end{prop}
\begin{proof}
Let $\mathcal{O}_\xi B(x_1,R) \subseteq \mathcal{O}_\xi B(x_2,R)$ be a $t$-ring in $\mathcal{R}(\xi,R)$ centered at $a \in \partial X \setminus \{\xi\}$, where $x_1,x_2 \in X$ and $t \ge 1$. Then $a,x_1,x_2$, and $\xi$ are aligned, and $d_X(x_1,x_2) = \ln(t)$.
Let $x_1' \in [x_1,a]$ and $x_2 \in [x_2,\xi]$ such that $d_X(x_1',x_1) = R' + \delta$ and $d_X(x_2',x_2) = R + \delta$. Then by \autoref{lem_Existence_of_rings}, each geodesic that emanates from $\xi$ and intersects the ball $B(x_1',R')$ (or the ball $B(x_2,R)$) also intersects the ball $B(x_1,R)$ (respectively, the ball $B(x_2',R')$). Hence,
$$\mathcal{O}_\xi B(x_1',R')\subseteq \mathcal{O}_\xi B(x_1,R) \subseteq \mathcal{O}_\xi B(x_2,R) \subseteq \mathcal{O}_\xi B(x_2',R'),$$
and $$d_X(x_1',x_2') = \ln(t) + R+R'+2\delta = \ln(te^{R+R'+2\delta}).$$
Let $C_5 = e^{R+R'+2\delta} \ge 1$. Then $\mathcal{O}_\xi B(x_1',R') \subseteq \mathcal{O}_\xi B(x_2',R')$ is a $C_5t$-ring in $\mathcal{R}(\xi,R')$.
Similarly, for any $t$-ring $\mathcal{O}_\xi B(x_1',R') \subseteq \mathcal{O}_\xi B(x_2',R')$ in $\mathcal{R}(\xi,R')$ there exists a $C_5t$-ring $\mathcal{O}_\xi B(x_1,R) \subseteq \mathcal{O}_\xi B(x_2,R)$ in $\mathcal{R}(\xi,R)$ such that
$$\mathcal{O}_\xi B(x_1,R)\subseteq \mathcal{O}_\xi B(x_1',R') \subseteq \mathcal{O}_\xi B(x_2',R') \subseteq \mathcal{O}_\xi B(x_2,R).$$
Hence the result follows.
\end{proof}
%\begin{prop}
%Let $(X,d_X)$ and $(Y,d_Y)$ be two proper $\delta$-hyperbolic metric spaces, and let $f \colon \partial X \to \partial Y$ be a homeomorphism. Then $f \colon (\partial X, \mathcal{R}_\omega(\partial X)) \to (\partial Y, \mathcal{R}_{\omega'}(\partial Y))$ is quasiconformal if and only if
%for every pair $(\xi, \xi'=f(\xi)) \in \partial X \times \partial Y$ and every $R > 2\delta$ the map $f \colon (\partial X \setminus \{\xi\}, \mathcal{R}_{\{\xi,R\}}(\partial X \setminus \{\xi\})) \to (\partial Y \setminus \{\xi\}, \mathcal{R}_{\{\xi',R\}}(\partial Y \setminus \{\xi\}))$ is quasiconformal.
%\end{prop}
%\begin{proof}
%Assume that $f \colon (\partial X, \mathcal{R}_\omega(\partial X)) \to (\partial Y, \mathcal{R}_{\omega'}(\partial Y))$ is a $\eta$-quasiconformal homeomorphism for some distortion map $\eta\colon [1,\infty) \to [1,\infty)$. Let $B \subseteq tB$ be a $t$-ring in $\mathcal{R}_{\{\xi,R\}}(\partial X \setminus \{\xi\})$ for $t \ge 1$.
%\end{proof}
%----------------------------------------------------%
\subsection{Shadows of Horoballs}
In the context of relatively hyperbolic groups, similar to the Gromov boundary, the notion of shadows, rings and quasi-conformal maps can be defined on the Bowditch boundaries. Furthermore, the shadows of horoballs play an important role in this situation. Let $(G,\mathcal{H}_G)$ be a $\delta$-relatively hyperbolic group with a Cayley graph $X$, and $\mathcal{R}(\xi,R)$ be a ring structure on the Bowditch boundary $\partial G^h$, where $\delta \ge 0$, $\xi \in \partial G^h$ and $R > 2\delta$.
Then the shadow of each horoball, whose limit set is not equal to $\{\xi\}$, relative to the point $\xi$ can be covered by a unique (unique in the coarse sense) ring contained in $\mathcal{R}(\xi, R)$ in the following sense:
%The shadow of each horoball, whose Gromov boundary is not equal to $\{\xi\}$, from the point $\xi$ is the set of other endpoints of geodesics that emanate from $\xi$ and intersect the horoball. This shadow can be covered coarsely by a unique ring in the ring structure $\mathcal{R}(\xi, R)$ defined on the Bowditch boundary $\partial X^h$ in the following sense.
\begin{lem} \label{lem_shadow_of_horospheres}
There exists a constant $T_1 = T_1(\delta,R) \ge 0$ with the following property: if $H \in \mathcal{H}_G$ is a horosphere-like set centered at a parabolic point $a_H \in \partial G^h \setminus \{\xi\}$, then there exists a $T_1$-ring $\mathcal{O}_\xi B(p_{\{\xi,H\}}, R) \subseteq T_1 \mathcal{O}_\xi B(p_{\{\xi,H\}}, R)$ in $\mathcal{R}(\xi,R))$ centered at $a_H$ such that
$$\mathcal{O}_\xi B(p_{\{\xi,H\}}, R) \subseteq \mathcal{O}_\xi (H^h) \subseteq T_1 \mathcal{O}_\xi B(p_{\{\xi,H\}}, R).$$
Moreover, the point $p_{\{\xi,H\}}$ can be chosen as a point at most $(R+\delta)$-distance from $H$, and the collection of such points has a uniformly bounded diameter that depends on $R$ and $\delta$.
\end{lem}
\begin{proof}
Let $e_H$ be the first entry point of geodesic $[\xi, a_H]$ to $H^h$. Then by \autoref{thm_visually_bounded} there exists a constant $C_4=C_4(\delta) \ge 0$ such that the set of first entry points of geodesics joining $\xi$ and $a_H$ to $H^h$ has a diameter at most $C_4$. Therefore, any geodesic emanating from $\xi$ that intersects $H^h$ also intersects the ball $B(e_H, C_4)$ and hence, $\mathcal{O}_\xi (H^h) \subseteq \mathcal{O}_\xi B(e_H, C_4)$.
\begin{figure}[H]
\centering
\begin{tikzpicture}
\draw (0,0) circle (2.5 cm);
\filldraw[gray!20] (0,1.7) circle (.8 cm);
\draw (0,1.7) circle (.8 cm);
\filldraw[gray!30] (0,0) circle (.35 cm);
\draw (0,0) circle (.35 cm);
\filldraw[gray!50] (0,1.3) circle (.3 cm);
\draw (0,1.3) circle (.3 cm);
\draw (0,-2.5)--(0,2.5);
\draw[bend left = 10] (0,-2.5) edge (1.08,2.25);
\draw[bend right = 10] (0,-2.5) edge (-1.08,2.25);
\draw[-stealth] (0,.9) -- (1.2,.9);
\draw[-stealth] (0,0) -- (1.2,0);
\draw[-stealth] (0,1.3) -- (1.2,1.3);
\filldraw (0,0) circle (1pt);
\filldraw (0,0.9) circle (1pt);
\filldraw (0,1.3) circle (1pt);
\node at (0,2.8) {$a_H$};
\node at (0,-2.8) {$\xi$};
\node at (1.8,0) {$p'_{\{\xi,H\}}$};
\node at (1.5,.9) {$e_H$};
\filldraw[white] (2,1.2) circle (.35cm);
\node at (1.8,1.3) {$p_{\{\xi,H\}}$};
\end{tikzpicture}
\caption{}
\end{figure}
Now consider the points $p_{\{\xi,H\}} \in [e_H, a_H]$ and $p'_{\{\xi,H\}} \in [e_H, \xi]$ with $d_{X^h}(e_H,p_{\{\xi,H\}}) = R + \delta$ and $d_{X^h}(e_H,p'_{\{\xi,H\}}) = C_4 + \delta$. Then $d_{X^h}(p_{\{\xi,H\}},p'_{\{\xi,H\}}) \le C_4 + R + 2\delta$. Let $T_1 = e^{(C_4 + R + 2\delta)}$, then $T_1$ depends only on $\delta$ (since $R$ is fixed) and
$$\mathcal{O}_\xi B(p_{\{\xi,H\}}, R) \subseteq \mathcal{O}_\xi (H^h) \subseteq M_1 \mathcal{O}_\xi B(p_{\{\xi,H\}}, R).$$
\end{proof}
It is concluded from \autoref{lem_shadow_of_horospheres} that there always exists an inclusion from the collection $\mathcal{H}_G$ of horosphere-like sets into the ring structure $\mathcal{R}(\xi, R)$ on $\partial G^h$ defined by
$$H \longmapsto (\mathcal{O}_\xi B(p_{\{\xi,H\}}, R) \subseteq T_1 \mathcal{O}_\xi B(p_{\{\xi,H\}}, R)),$$
and any two such inclusions are also uniformly bounded.
Let $(G_1,\mathcal{H}_{G_1})$ and $(G_2,\mathcal{H}_{G_2})$ be two $\delta$-relatively hyperbolic groups and $f\colon (\partial G_1^h, \mathcal{R}(\xi,R)) \to (\partial G_2^h, \mathcal{R}(\xi',R))$ be a $\eta$-quasiconformal map, for some $\delta \ge 0$ and for some distortion map $\eta \colon [0,\infty) \to [1,\infty)$.
Then $f$ must sends the $T_1$-ring $\mathcal{O}_\xi B(p_{\{\xi,H\}}, R) \subseteq T_1 \mathcal{O}_\xi B(p_{\{\xi,H\}}, R)$ to a $\eta(T_1)$-ring in $\mathcal{R}(\xi',R)$. The following definition gives a notion of a quasiconformal map that coarsely preserves the shadow of horoballs.
%The following definition of a relative quasiconformal map imposes some condition on a quasiconformal map so that the shadow of each horoball in $\partial X^h$ coarsely maps to a shadow of some horoball $\partial Y^h$.
\begin{defn}[Coarsely-Preserves the Shadows of Horoballs] \label{defn_coarsely_shadows_of_horoballs_preserve}
Let $(G_1,\mathcal{H}_{G_1})$ and $(G_2,\mathcal{H}_{G_2})$ be two $\delta$-relatively hyperbolic groups and $f \colon \partial G_1^h \to \partial G_2^h$ be a $\eta$-quasiconformal map, for some $\delta \ge 0$ and for some distortion map $\eta \colon [0,\infty) \to [1,\infty)$. Let $L \ge 1$ be a constant. Then $f$ is said to $L$-coarsely preserves the shadows of horoballs relative to $\xi \in \partial G_1^h$ if
\begin{enumerate}[$i)$]
\item $f \colon (\partial G_1^h, \mathcal{R}(\xi,R)) \to (\partial G_2^h, \mathcal{R}(\xi',R))$ is a $\eta$-quasiconformal map,
\item $f$ and $f^{-1}$ preserve the parabolic points,
\item if $H \in \mathcal{H}_{G_1}$ is a horosphere-like set centered at $a \in \partial G_1^h \setminus \{\xi\}$ and $H' \in \mathcal{H}_Y$ is the horosphere-like set centered at $f(a)=a' \in \partial G_2^h \setminus \{\xi'\}$ with
$$\mathcal{O}_\xi B(p_{\{\xi,H\}}, R) \subseteq \mathcal{O}_\xi (H^h) \subseteq T_1 \mathcal{O}_\xi B(p_{\{\xi,H\}}, R), \text{ and}$$
$$\mathcal{O}_{\xi'} B(q_{\{\xi',H'\}}, R) \subseteq \mathcal{O}_{\xi'} (H'^h) \subseteq T_1 \mathcal{O}_{\xi'} B(q_{\{\xi',H'\}}, R).$$
Then
\begin{align*}
\frac{1}{L}\mathcal{O}_{\xi'} B(q_{\{\xi',H'\}}, R) \subseteq f(\mathcal{O}_\xi B(p_{\{\xi,H\}}, R)) \subseteq f(\mathcal{O}_\xi (H^h)) \\
\subseteq f(T_1 \mathcal{O}_\xi B(p_{\{\xi,H\}}, R)) \subseteq L \mathcal{O}_{\xi'} B(q_{\{\xi',H'\}}, R).
\end{align*}
\end{enumerate}
\end{defn}
%\item if $H \in \mathcal{H}_{G_1}$ is a horosphere-like set centered at $a \in \partial G_1^h \setminus \{\xi\}$ and $H' \in \mathcal{H}_Y$ is the horosphere-like set centered at $f(a)=a' \in \partial Y^h \setminus \{\xi'\}$ with $$\mathcal{O}_\xi B(p_{\{\xi,H\}}, R) \subseteq \mathcal{O}_\xi (H^h) \subseteq T_1 \mathcal{O}_\xi B(p_{\{\xi,H\}}, R),$$ and $$\mathcal{O}_{\xi'} B(q_{\{\xi',H'\}}, R) \subseteq \mathcal{O}_{\xi'} (H'^h) \subseteq T_1 \mathcal{O}_{\xi'} B(q_{\{\xi',H'\}}, R).$$ Then
%\begin{align*}
%\frac{1}{L}\mathcal{O}_{\xi'} B(q_{\{\xi',H'\}}, R) \subseteq f(\mathcal{O}_\xi B(p_{\{\xi,H\}}, R)) \subseteq f(\mathcal{O}_\xi (H^h)) \\
%\subseteq f(T_1 \mathcal{O}_\xi B(p_{\{\xi,H\}}, R)) \subseteq L \mathcal{O}_{\xi'} B(q_{\{\xi',H'\}}, R).
%\end{align*}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Quasi-isometry induces Quasi-conformal Mapping} \label{sec_qi_implies_qc}
Let $\delta, C, K \ge 0$ and $\lambda \ge 1$. Let $(G_1,\mathcal{H}_{G_1})$ and $(G_2,\mathcal{H}_{G_2})$ be two proper $\delta$-relatively hyperbolic groups with Cayley graph $x$ and $Y$, respectively, and $\varphi \colon G_1 \to G_2$ be a $C$-coarsely cusp preserving $(\lambda, K)$-quasi-isometry. Then by \autoref{thm_cusp_presv_qi_implies_homeomorphism} $\varphi$ induces a quasi-isometry between their cusped spaces $X^h$ to $Y^h$ and the resulting quasi-isometric constants depend on $\delta,\lambda,C$ and $K$. Without loss of generality, assume that $\varphi$ is a $(\lambda,K)$-quasi-isometry from $X^h$ to $Y^h$, and let $\psi\colon Y^h \to X^h$ be an inverse quasi-isometry of $\varphi$, which is also a $(\lambda, K)$-quasi-isometry. Then $d_{X^h}(\psi\varphi(x),x) \le K$ for all $x \in X^h$ and $d_{Y^h}(\varphi\psi(y),y) \le K$ for all $y \in Y^h$.
The boundary extension maps on Gromov boundaries induced by $\varphi$ and $\psi$ are denoted by $\partial \varphi\colon \partial G_1^h \to \partial G_2^h$ and $\partial \psi\colon \partial G_2^h \to \partial G_1^h$, respectively. Note that $\partial \varphi$ and $\partial \psi$ are homeomorphisms and inverses of each other.
\begin{thm} \label{thm_qi_implies_qc}
There exists a continuous map $\eta\colon [1,\infty) \to [1,\infty)$, depending on $\lambda, K$, and $\delta$ such that for any $R > 2\delta$ and for any pair of points $(\xi,\xi') \in \partial G_1^h \times \partial G_2^h$ with $\xi' = \partial \varphi (\xi)$, the boundary extension map $\partial\varphi \colon (\partial G_1^h, \mathcal{R}(\xi,R)) \to (\partial G_2^h, \mathcal{R}(\xi',R))$ on their Bowditch boundaries equipped with the corresponding ring structures is a $\eta$-quasiconformal map.
Moreover, the map $\eta$ is given by $\eta(t) = e^{A\ln(t)+B}$, where the constants $A \ge 1$ and $B \ge 0$ depend on $C, K, R, \lambda$, and $\delta$.
\end{thm}
\begin{proof}
Let $(\xi,\xi') \in \partial G_1^h \times \partial G_2^h$ be a pair of points with $\xi' = \partial \varphi (\xi)$. Since for any two positive real numbers $R,R' > 2\delta$ the corresponding ring structures in $\partial G_1^h$ (similarly, in $\partial G_2^h$) are equivalent (see \autoref{prop_ring_structures_are_independent_on_R}), it is enough to show that the given statement is true for a particular positive real number $R_0 >2\delta$.
To find a suitable $R_0 > 2\delta$, recall the Stability Theorem for quasigeodesics (\autoref{prop_stability_of_quasigeodesics}), which implies that: there exists a constant $C_1=C_1(\delta,\lambda,K) \ge 0$ such that each $(\lambda,K)$-quasigeodesic is contained in the $C_1$-neighbourhood of some geodesic with the same endpoints as the quasigeodesic.
Now, choose a fixed real number $R_0 > (\lambda C_1 + C_1 + 2K + 2\lambda\delta)$ and consider the ring structures $\mathcal{R}(\xi,R_0)$ and $\mathcal{R}(\xi,R_0)$ on the Bowditch boundaries $\partial G_1^h$ and $\partial G_2^h$, respectively, and let $\mathcal{O}_\xi B(x_1,R_0) \subseteq \mathcal{O}_\xi B(x_2,R_0)$ be a $t$-ring in $\mathcal{R}(\xi,R_0)$, centered at some point $a \in \partial G_1^h \setminus \{\xi\}$, where $t \ge 1$. Then $a,x_1, x_2$ and $\xi$ are aligned (that is, $x_1,x_2 \in [\xi,a]$) with $x_2$ is being closer to $\xi$ than $x_1$, and $d_{X^h}(x_1,x_2) = \ln(t)$.
Then by the Stability Theorem there exist a constant $C_1=C_1(\delta,\lambda,K) \ge 0$ and two points $y_1$ and $y_2$ on the geodesic $[\xi', a']$, where $a' = \partial \varphi (a) \in \partial G_2^h$ such that
$$d_Y(\varphi (x_i) , y_i) \le C_1, \text{ for } i=1,2.$$
Now, let $R_1=\frac{1}{\lambda} \left( R_0 - (\lambda C_1 + C_1 +2K) \right)$ and $R_2 = (2C_1 + \lambda R_0 + K)$. Then $R_2 \ge R_0 \ge R_1 > 2\delta$, and claim that,
$$\mathcal{O}_{\xi'} B(y_1,R_1) \subseteq f(\mathcal{O}_{\xi} B(x_1,R_0)) \subseteq f(\mathcal{O}_{\xi} B(x_2,R_0)) \subseteq \mathcal{O}_{\xi'} B(y_2,R_2).$$
Let $b'$ be a point in $\mathcal{O}_{\xi'} B(y_1,R_1)$, and let the geodesic $[\xi', b']$ intersect the ball $B(y_1,R_1)$ at $q$. Since $\psi\colon Y^h \to X^h$ is an inverse quasi-isometry of $\varphi$ as considered in the beginning of this section, then by the stability theorem there exists a point $p \in [\xi,b]$, where $b=\partial\psi(b')=\partial\varphi^{-1}(b')$ such that $d_X(p,\psi(q)) \le C_1$. Therefore,
\begin{align*}
d_{X^h}(p,x_1) & \le d_{X^h}(p,\psi(q)) + d_{X^h}(\psi(q), \psi \varphi (x_1)) + d_{X^h}(\psi \varphi (x_1), x_1)\\
& \le C_1 + \lambda (R_1+C_1) + 2K \\
& < R_0.
\end{align*}
This implies that the geodesic $[\xi,b]$ intersects the ball $B(x_1,R)$ and so $b \in \mathcal{O}_{\xi} B(x_1,R)$. Hence,
$$\mathcal{O}_{\xi'} B(y_1,R_1) \subseteq f(\mathcal{O}_{\xi} B(x_1,R_0)).$$
Now for the other inequality, let $b$ be a point in $\mathcal{O}_{\xi} B(x_2,R)$, and let the geodesic $[\xi,b]$ intersect the ball $B(x_2,R_0)$ at $p$. Then there exists a point $q \in [\xi',b']$, where $b' = \partial\varphi (b)$ such that $d_Y(\varphi(p),q) \le C_1$. Therefore,
\begin{align*}
d_{Y^h}(y_2,q) & \le d_{Y^h}(y_2, \varphi(x_2)) + d_{Y^h}(\varphi(x_2), \varphi(p)) + d_{Y^h}(\varphi(p), q) \\
& \le C_1 + \lambda R_0 + K + C_1 \\
& = R_2.
\end{align*}
This implies that the geodesic $[\xi',b']$ intersects the ball $B(y_2,R_2)$, and so $b' \in \mathcal{O}_{\xi'} B(y_2,R_2)$. Hence,
$$f(\mathcal{O}_{\xi} B(x_2,R_0)) \subseteq \mathcal{O}_{\xi'} B(y_2,R_2).$$
Thus, the claim is proved.
Now consider the points $z_1 \in [a',y_1]$, and $z_2 \in [y_2, \xi']$ such that $d_Y(z_1,y_1) = R_0 + \delta$ and $d_Y(y_2,z_2) = R_2 + \delta$. (Refer to \autoref{fig_thm_1})
\begin{figure}[H]
\centering
\begin{tikzpicture}
\draw (-1,0) circle (.7cm);
\draw (5,0) circle (.9cm);
%\draw (-4,0) circle (.9cm);
\draw (-5,0)--(8,0);
\draw (-4,0)--(-4,.9);
\draw (2,0)--(2,1);
%\draw[dashed] (-1,0)--(-1,1);
%\draw[dashed] (5,0)--(5,2);
\draw[bend right=4] (-4,.9) edge (8,0);
%\draw[dashed] (-2,1) -- (0,1);
\draw[bend right=5] (2,1) edge (8,0);
%\draw[dashed] (5,1) -- (6,1);
\filldraw (-4,0) circle (1pt);
\filldraw (-1,0) circle (1pt);
\filldraw (2,0) circle (1pt);
\filldraw (5,0) circle (1pt);
\node at (-4,-.3) {$z_1$};
\node at (-1,-.3) {$y_1$};
\node at (2,-.3) {$y_2$};
\node at (5,-.3) {$z_2$};
\node at (-3,-.3) {$R_0 + \delta$};
\node at (3.5,-.3) {$R_2 + \delta$};
\node at (-3.5,.5) {$\le R$};
%\node at (-1.5,.8) {$\le R_1$};
\node at (2.5,.5) {$\le R_2$};
%\node at (5.5,.8) {$\le R$};
\node at (8.5,0) {$\xi'$};
\node at (-5.5,0) {$a$};
\end{tikzpicture}
\caption{}
\label{fig_thm_1}
\end{figure}
Then by \autoref{lem_Existence_of_rings},
\begin{itemize}
\item $\mathcal{O}_{\xi'} B(z_1,R_0) \subseteq \mathcal{O}_{\xi'} B(y_1, 2\delta) \subseteq \mathcal{O}_{\xi'} B(y_1, R_1)$ (since $R_1 > 2 \delta$), and
\item $\mathcal{O}_{\xi'} B(y_2,R_2) \subseteq \mathcal{O}_{\xi'} B(z_2, 2\delta) \subseteq \mathcal{O}_{\xi'} B(z_2, R_0)$ (since $R_0 > 2 \delta$).
\end{itemize}
Therefore, $$\mathcal{O}_{\xi'} B(z_1,R_0) \subseteq f (\mathcal{O}_\xi B(x_1,R_0)) \subseteq f (t\mathcal{O}_\xi B(x_2,R_0)) \subseteq \mathcal{O}_{\xi'} B(z_2,R_0).$$
Finally, the distance between $z_1$ and $z_2$ is given by
\begin{align*}
d_{Y^h}(z_1,z_2) & \le d_{Y^h}(y_1,y_2) + R_0 + R_2 + 2\delta \\
& \le d_{Y^h}(\varphi(x_1), \varphi(x_2)) + 2C_1 + R_0 + R_2 + 2\delta \\
& \le \lambda \ln(t) + K + 2C_0 + R_0 + R_2 + 2\delta.
\end{align*}
Let $A_0 = \lambda$ and $B_0 = K + 2C_0 + R + R_2 + 2\delta$, and define the map $\eta_0\colon [1,\infty) \to [1,\infty)$ by
$$\eta_0(t) = e^{A_0 \ln(t) + B_0}.$$
Hence, $\partial\varphi\colon (\partial G_1^h, \mathcal{R}(\xi,R_0)) \to (\partial G_2^h, \mathcal{R}(\xi',R_0))$ is a $\eta_0$-quasiconformal map.
Now by \autoref{prop_ring_structures_are_independent_on_R}, for any $R > 2\delta$ there exists a constant $C_5 = C_5(\delta,R,R_0) \ge 1$ such that the identity map $Id_X \colon (\partial G_1^h, \mathcal{R}(\xi,R)) \to (\partial G_1^h, \mathcal{R}(\xi,R_0))$ and the identity map $Id_Y \colon (\partial G_2^h, \mathcal{R}(\xi',R_0) \to (\partial G_2^h, \mathcal{R}(\xi',R))$ are $\eta_{\{R,R_0\}}$-quasiconformal maps, where the distortion map $\eta_{\{R,R_0\}}\colon [1,\infty) \to [1,\infty)$ is given by by $\eta_{\{R,R_0\}}(t)=C_5t$.
Therefore, the map $\partial\varphi\colon (\partial G_1^h, \mathcal{R}(\xi,R)) \to (\partial G_2^h, \mathcal{R}(\xi',R))$ is a $(\eta_{\{R,R_0\}} \circ \eta_0 \circ \eta_{\{R,R_0\}})$-quasiconformal map and $(\eta_{\{R,R_0\}} \circ \eta_0 \circ \eta_{\{R,R_0\}})(t) = C_5 e^{A_0\ln(C_5t)+B_0} = e^{A\ln(t)+B}$ (say). Hence the result follows.
\end{proof}
\begin{cor}[Hyperbolic Cases]
Let $\delta, K \ge 0$ and $\lambda \ge 1$. If $(X,d_X)$ and $(Y,d_Y)$ are two proper $\delta$-hyperbolic metric spaces, and if $\varphi\colon X \to Y$ is a $(\lambda, K)$-quasi-isometry, then there exists a continuous map $\eta\colon [1,\infty) \to [1,\infty)$, depending on $\lambda, K$ and $\delta$ such that for any $R > 2\delta$ and for any pair of points $(\xi,\xi') \in \partial X \times \partial Y$ with $\xi' = \partial \varphi (\xi)$, the boundary extension map $\partial\varphi\colon (\partial X, \mathcal{R}(\xi,R)) \to (\partial Y, \mathcal{R}(\xi',R))$ of $\varphi$ on their Gromov boundaries equipped with ring structures is a $\eta$-quasiconformal map.
Moreover, the distortion map $\eta$ is given by $\eta(t)=e^{A\ln(t)+B}$, where the constants $A \ge 1$ and $B \ge 0$ depend on $\lambda, K$ and $\delta$.
\end{cor}
\begin{thm}
There exists a constant $L=L(C,K,R,\lambda,\delta) \ge 1$ such that for any pair of base points $(\xi,\xi') \in \partial G_1^h \times \partial G_2^h$ with $\partial\varphi(\xi)=\xi'$ the quasiconformal map $\partial\varphi\colon (\partial G_1^h, \mathcal{R}(\xi,R)) \to (\partial G_2^h, \mathcal{R}(\xi',R))$ $L$-coarsely preserves the shadows of horoballs.
\end{thm}
\begin{proof}
Let $H \in \mathcal{H}_X$ be a horosphere-like set in $X^h$ centered at $a \in \partial G_1^h \setminus \{\xi\}$, and let $H' \in \mathcal{H}_Y$ be the horosphere-like set in $Y^h$ centered at $a' = \partial\varphi(a) \in \partial G_2^h$.
Then by \autoref{lem_shadow_of_horospheres} there exists a constant $T_1=T_1(\delta,R)$ and two points $p_{\{\zeta,H\}} \in [\xi,a]$ and $q_{\{\zeta',H'\}} \in [\xi',a']$ at a distance at most $(R + \delta)$ from $H$ and $H'$, respectively, such that
$$\mathcal{O}_\zeta B(p_{\{\zeta,H\}},R) \subseteq \mathcal{O}_\zeta(H^h) \subseteq T_1 \mathcal{O}_\zeta B(p_{\{\zeta,H\}},R),$$
and
$$\mathcal{O}_{\zeta'} B(q_{\{\zeta',H'\}},R) \subseteq \mathcal{O}_{\zeta'}(H'^h) \subseteq T_1 \mathcal{O}_{\zeta'} B(q_{\{\zeta',H'\}},R).$$
Since $\varphi$ is a $(\lambda,K)$-quasi-isometry, there exists a point $y_1,y_2 \in [\xi',a']$ such that $d_{Y^h}(\varphi(p_{\{\zeta,H\}}), y_1) \le C_1$ and $d_{Y^h}(\varphi(p'_{\{\zeta,H\}}), y_2) \le C_1$.
As in the proof of \autoref{thm_qi_implies_qc} there exists $z_1 \in [y_1,a']$, and $z_2 \in [y_2,\xi']$ with $d_{Y^h}(\varphi(z_1,z_2) = \ln(\eta(T_1))$ such that
$$\mathcal{O}_{\xi'} B(z_1,R) \subseteq f(\mathcal{O}_\xi B(p_{\{\zeta,H\}},R)) \subseteq f(\mathcal{O}_\xi(H^h)) \subseteq f(T_1 \mathcal{O}_\xi B(p_{\{\zeta,H\}},R)) \subseteq \mathcal{O}_{\xi'} B(z_2,R).$$
Hence $d_{Y^h}(\varphi(z_1,y_1) \le \ln(\eta(T_1))$ and $d_{Y^h}(\varphi(z_2,y_2) \le \ln(\eta(T_1))$.
Also, since $\varphi$ is $C$-coarsely cusp preserving and $d_{X^h}(p_{\{\zeta,H\}}, H) \le R + \delta$, then $d_{Y^h}(\varphi(p_{\{\zeta,H\}}), H') \le C + \lambda(R \delta) + K = C' (say)$. Therefore,
\begin{align*}
d_{Y^h}(z_1, H') &\le d_{Y^h}(z_1, y_1) + d_{Y^h}(y_1, \varphi(p_{\{\zeta,H\}})) + d_{Y^h}(\varphi(p_{\{\zeta,H\}}), H') \\
& \le \ln(\eta(T_1)) + C_1 + C',
\end{align*}
and so, $d_{Y^h}(z_2, H') \le 2\ln(\eta(T_1)) + C_1 + C'$.
Let $L = e^{(2\ln(\eta(T_1)) + C_1 + C')+(R+\delta)} \ge 1$, and consider the points $q_1 \in [q_{\{\zeta',H'\}},a']$ and $q_2 \in [q_{\{\zeta',H'\}},xi']$ with $d_{Y^h}(q_1, q_{\{\zeta',H'\}}) = \ln(L) = d_{Y^h}(q'_{\{\zeta',H'\}},q_2)$. Then
$$\mathcal{O}_{\xi'} B(q_1,R) \subseteq \mathcal{O}_{\xi'} B(z_1,R) \subseteq f(\mathcal{O}_\xi(H^h)) \subseteq \mathcal{O}_{\xi'} B(z_2,R) \subseteq \mathcal{O}_{\xi'} B(q_2,R).$$
Then $L$ depends on $C,K,R,\lambda$, and $\delta$. Also, $\mathcal{O}_{\xi'} B(q_2,R) = L \mathcal{O}_{\xi'} B(q'_{\{\zeta',H'\}}),R)$, $\mathcal{O}_{\xi'} B(q_{\{\zeta',H'\}}),R) = L \mathcal{O}_{\xi'} B(q_1,R)$, and so $\mathcal{O}_{\xi'} B(q_1,R) = \frac{1}{L} \mathcal{O}_{\xi'} B(q_{\{\zeta',H'\}}),R)$. Therefore,
$$\frac{1}{L} \mathcal{O}_{\xi'} B(q_{\{\zeta',H'\}}),R) \subseteq f(\mathcal{O}_\xi(H^h)) \subseteq L \mathcal{O}_{\xi'} B(q'_{\{\zeta',H'\}}),R).$$
Also, for $\partial \varphi^{-1} = \partial \psi$ the result holds. Hence, the result follows.
\end{proof}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%\newpage
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Quasi-Isometry induced by a Quasi-Conformal Mapping} \label{sec_qc_implies_qi}
Let $(G_1,\mathcal{H}_{G_2})$ and $(G_2,\mathcal{H}_{G_2})$ be two $\delta$-relatively hyperbolic groups with Cayley graphs $X$ and $Y$, respectively, and let $f\colon \partial G_1^h \to \partial G_2^h$ be a $\eta$-quasiconformal map between their Bowditch boundaries for some $\delta \ge 0$ and some distortion map $\eta \colon [1,\infty) \to [1,\infty)$ such that $f$ and $f^{-1}$ preserve the parabolic points. Then, by the definition of quasiconformal maps, there exists a positive real number $R > 2\delta$ and a pair of base points $(\xi,\xi') \in \partial G_1^h \times \partial G_2^h$ depending on $f$ and $\delta$ with $f(\xi) = \xi'$ such that $f \colon (\partial G_1^h, \mathcal{R}(\xi,R)) \to (\partial G_2^h, \mathcal{R}(\xi',R))$ is $\eta$-quasiconformal. By \autoref{prop_ring_structures_are_independent_on_R} without loss of generality assume that $R > 2\delta +1$.
If $x \in X$, $t \ge 1$, and $B \subseteq tB$ (if exists) is a $t$-ring in $\mathcal{R}(\xi,R)$ satisfying $B \subseteq \mathcal{O}_{\xi} B(x,R) \subseteq tB$, then $f$ sends this $t$-ring to a $\eta(t)$-ring in $\mathcal{R}(\xi',R)$. Consider
$$E_{\{\xi,\xi'\}}(x,t,B) = \{ y \in Y \mid \mathcal{O}_{\xi'} B(y,R) \subseteq f(B)\subseteq f(\mathcal{O}_{\xi} B(x,R)) \subseteq f(tB) \subseteq \eta(t) \mathcal{O}_{\xi'} B(y,R) \},$$
and
$$E_{\{\xi,\xi'\}}(x,t) = \bigcup_{B \subseteq tB} \{ E_{\{\xi,\xi'\}}(x,k,B) \mid B \subseteq \mathcal{O}_\xi B(x,R) \subseteq tB \}.$$
\begin{prop} \label{prop_E(x,t)_is_bounded}
There exists a constant $D=D(\delta, \eta, t, R) \ge 0$ such that for each $x\in X^h$ and $t \ge R + \delta$, the set $E_{\{\xi,\xi'\}}(x,t)$ is a non-empty, bounded subset of $Y^h$ and has a diameter at most $D$.
\end{prop}
\begin{proof}
Let $x \in X^h$ and $t \ge R + \delta$. Then by \autoref{lem_Existence_of_rings} there always exists a $t$-ring $B \subseteq tB$ in $\mathcal{R}(\xi,R)$ satisfying $B \subseteq \mathcal{O}_\xi B(x,R) \subseteq tB$. Hence, by the definition of quasiconformal maps, the set $E_{\{\xi,\xi'\}}(x,t,B)$ is non-empty,
%for each $t$-ring $B \subseteq tB$ in $\mathcal{R}(\xi,R)$ satisfying $B \subseteq \mathcal{O}_\xi B(x,R) \subseteq tB$,
and hence $E_{\{\xi,\xi'\}}(x,t)$ is non-empty. To prove the boundedness, let $y_1$ and $y_2$ be two points in $E_{\{\xi,\xi'\}}(x,k)$. Then for each $i=1,2$, there exists a $t$-ring $B_i \subseteq tB_i$ in $\mathcal{R}(\xi,R)$ with $B_i\subseteq \mathcal{O}_\xi B(x,R) \subseteq tB_i$ such that
\begin{gather*}
\mathcal{O}_{\xi'}B(y_i,R)\subseteq f(B_i)\subseteq f(\mathcal{O}_{\xi} B(x,R))\subseteq f(tB_i)\subseteq \eta(t)\mathcal{O}_{\xi'}B(y_i,R).
\end{gather*}
For $i=1,2$, let $\mathcal{O}_{\xi'}B(y_i',R) = \eta(t)\mathcal{O}_{\xi'}B(y_i,R)$, where $y_i' \in Y^h$, $y_i,y_i'$ and $\xi'$ are aligned and $d_Y(y_i,y_i')=\ln(\eta(t))$. Then, both $\mathcal{O}_{\xi'}B(y_1,R)$ and $\mathcal{O}_{\xi'}B(y_2,R)$ are contained in the intersection $\mathcal{O}_{\xi'}B(y_1',R) \cap \mathcal{O}_{\xi'}B(y_2',R)$. This implies that for each $i=1,2$, the geodesic $\gamma_i$, emanating from $\xi'$ (i.e., $\gamma_i(-\infty)= \xi'$) and passing through $y_i$ and $y_i'$, intersects the closed $2\delta$-neighbourhood of both $B(y_1',R)$ and $B(y_2',R)$. There are three possible cases:
\noindent\underline{Case-$1$:} Either the ball $B(y_1',R + 2\delta)$ intersects on the segment $[y_2,y_2']$ or the ball $B(y_2',R + 2\delta)$ intersects on the segment $[y_1,y_1']$ (refer to \autoref{fig_E(x,t)_is_bounded_1}).
\begin{figure}[H]
\centering
%\includegraphics[width=0.5\linewidth]{}
\begin{tikzpicture}
\draw[bend left=10] (0,1) edge (9,0);
\filldraw (1.5,1.05) circle (1pt);
\filldraw (4.7,.94) circle (1pt);
\draw (1.5,1.05) circle (.5 cm);
\draw (4.7,.94) circle (.6 cm);
%\filldraw (0,0) circle (1pt);
\draw[bend left=20] (0,-1) edge (9,0);
\filldraw (1.5,-.34) circle (1pt);
\filldraw (6.2,.45) circle (1pt);
\draw (1.5,-.34) circle (.5 cm);
\draw (6.2,.45) circle (.6 cm);
\node at (1.5,1.25) {$y_1$};
\node at (4.7,1.2) {$y_1'$};
\node at (1.5,-.6) {$y_2$};
\node at (6.2,0.2) {$y_2'$};
\node at (9.3,0) {$\xi'$};
\node at (7.6,1.1) {$B(y_2',R + 2\delta)$};
\node at (0.2,-.2) {$B(y_2,R)$};
\node at (-.5,1) {$\gamma_1$};
\node at (-.5,-.9) {$\gamma_2$};
\end{tikzpicture}
\caption{}
\label{fig_E(x,t)_is_bounded_1}
\end{figure}
Without loss of generality, assume that the ball $B(y_1',R + 2\delta)$ intersects on the segment $[y_2,y_2']$ at $q$. Therefore,
\begin{align*}
d_{Y^h}(y_1,y_2) & \le d_{Y^h}(y_1,y_1') + d_{Y^h}(y_1',q) + d_Y(q,y_2)\\
& \le \ln(\eta(t)) + R + 2\delta + \ln(\eta(t)) \text{ (uniformly bounded.)}
\end{align*}
Hence $E_{\{\xi,\xi'\}}(x,t)$ is a bounded subset of $Y^h$, and the bound depends on $t,\eta,\delta$, and $R$.
\noindent\underline{Case-$2$:} The ball $B(y'_1,R)$ intersects $\gamma_2$ on $[y_2,\gamma_2(\infty)]$ at $q$ (refer to \autoref{fig_E(x,t)_is_bounded_2}).
\begin{figure}[H]
\centering
%\includegraphics[width=0.5\linewidth]{}
\begin{tikzpicture}
\draw[bend left=10] (-2,.5) edge (9,0);
\filldraw (-1.3,.58) circle (1pt);
\filldraw (1.3,.8) circle (1pt);
\draw (-1.3,.58) circle (.5 cm);
\draw (1.3,.8) circle (.6 cm);
%\filldraw (0,0) circle (1pt);
\draw[bend left=10] (-2.5,-.3) edge (9,0);
\filldraw (4,.45) circle (1pt);
\filldraw (6.5,.32) circle (1pt);
\draw (4,.45) circle (.5 cm);
\draw (6.5,.32) circle (.6 cm);
\node at (-1.3,.8) {$y_1$};
\node at (1.3,1.15) {$y_1'$};
\node at (4,.2) {$y_2$};
\node at (6.5,0) {$y_2'$};
\node at (9.3,0) {$\xi'$};
\node at (3,1.5) {$B(y_1',R + 2\delta)$};
%\node at (2.8,-.2) {$B(y_2,R)$};
\node at (-2.5,.5) {$\gamma_1$};
\node at (-3,-.3) {$\gamma_2$};
\filldraw (1.3,0.32) circle (1pt);
\node at (1.3,-.1) {$q$};
\end{tikzpicture}
\caption{}
\label{fig_E(x,t)_is_bounded_2}
\end{figure}
Claim that $d_{Y^h}(q,y_2) \le 2R + 5\delta$. If $d_{Y^h}(q,y_2) > 2R + 5\delta$, then
$$\mathcal{O}_{\xi'} B(y_1',R) \subseteq \mathcal{O}_{\xi'} B(q,2R+4\delta) \subseteq \mathcal{O}_{\xi'} B(y_2,2\delta).$$
Therefore, any geodesic emanating from $\xi'$ intersecting the ball $B(y_1',R)$ must intersect the ball $B(y_2,2\delta)$. Since $R > 2\delta+1$, $B(y_2,R) \setminus B(y_2,2\delta) \ne \phi$ (non-empty), and contains at least one vertex of $X^h$. Therefore, any geodesic passing through $B(y_2,R) \setminus B(y_2,2\delta)$ does not intersects the ball $B(y_1',R)$, which contradict the fact that $\mathcal{O}_{\xi'} B(y_2,R) \subseteq \mathcal{O}_{\xi'} B(y_1',R)$. Hence the claim follows, and the distance $d_{Y^h}(y_,y_2)$ is uniformly bounded and the bound depends on $\delta, \eta, t$ and $R$.
\noindent\underline{Case-$3$:} The ball $B(y'_1,R)$ intersects $\gamma_2$ on $[y_2,\xi]$. Then similar to case-$2$, by interchanging the role of $y_1'$ and $y_2'$ in case-$2$ it can be shown that the distance between $y_1$ and $y_2$ is uniformly bounded and the bound depends on $\delta, \eta, t$ and $R$.
\end{proof}
%\begin{lem}\label{lem: Diameter map is an increasing map under t}
%The diameter of the set $E_{\{\xi,\xi'\}}(x,t)$ induces an increasing map with respect to the parameter $t$.
%\end{lem}
\begin{thm} \label{thm_qc_implies_qi_for_cusped_spaces}
The map $f$ induces a quasi isometry $\Phi f \colon X^h \to Y^h$ between the cusped spaces and the quasi-isometric constants of $\Phi f$ depend only on $\delta,\eta$ and $R$.
\end{thm}
\begin{proof}
Let $x_1$ and $x_2$ be two points in $X^h$ with $d_{X^H}(x_1,x_2)\le R$. Also, let $m \in X^h$ be a point such that $d_{X^h}(m,x_1) = d_{X^h}(m,x_2) =\frac{d_{X^h}(x_1,x_2)}{2}$. Then
\begin{gather*}
B(m,R/2)\subseteq B(x_1,R)\subseteq B(m,2R), \text{ and } \\
B(m,R/2)\subseteq B(x_2,R)\subseteq B(m,2R).
\end{gather*}
\begin{figure}[H]
\centering
%\includegraphics[width=0.5\linewidth]{}
\begin{tikzpicture}
\filldraw[gray!10] (0,.4) circle (.95 cm);
\filldraw[gray!30] (0,-.4) circle (.95 cm);
\filldraw[gray] (0,0) circle (.5 cm);
\draw (0,.4) circle (.95 cm);
\draw (0,-.4) circle (.95 cm);
\draw (-5.5,0) -- (7,0);
%circles at mid point
\filldraw (0,0) circle (1pt);
\draw (0,0) circle (.5 cm);
\draw (0,0) circle (1.5 cm);
%required circles
\draw (-4,0) circle (.95 cm);
\draw (3.5,0) circle (.95 cm);
%bend lines
\draw[bend right=5] (-.4,1.7) edge (7,0);
\draw[bend right=5] (-4.75,1.1) edge (7,0);
\filldraw (-4,0) circle (1pt);
\filldraw (0,0.4) circle (1pt);
\filldraw (0,-.4) circle (1pt);
\filldraw (3.5,0) circle (1pt);
%node
%\node at (-5.8,0) {$a$};
\node at (-4,-.3) {$m_1$};
\node at (3.5,-.3) {$m_2$};
\node at (7.3,0) {$\xi$};
\draw[-stealth] (.2,-.2) --(1.7,-1.2);
\node at (2.2,-1.5) {$B(m,\frac{R}{2})$};
\draw[-stealth] (-.5,-1) --(-1.7,-1.2);
\node at (-2.5,-1.3) {$B(x_2,R)$};
\draw[-stealth] (-.5,1) --(-1.6,1.4);
\node at (-2.3,1.5) {$B(x_1,R)$};
\draw[-stealth] (0,-1.5) --(-1,-1.7);
\node at (-1.8,-1.8) {$B(m,2R)$};
\end{tikzpicture}
\caption{}
\end{figure}
Then by \autoref{lem_Existence_of_rings} there exist two points $m_1$ and $m_2$ on the geodesic through $\xi$ and $m$ such that
\begin{itemize}
\item $d_{X^h}(m_1,m) = R + \delta$,
\item $d_{X^h}(m,m_2) = 2R + \delta$, and
\item $\mathcal{O}_\xi B(m_1,R)\subseteq \mathcal{O}_\xi B(m,R/2)\subseteq \mathcal{O}_\xi B(m,2R)\subseteq \mathcal{O}_\xi B(m_2,R)$.
\end{itemize}
Therefore, the pair $\mathcal{O}_\xi B(m_1,R) \subseteq \mathcal{O}_\xi B(m_2,R)$ is a $(3R + 2\delta)$-ring in $\mathcal{R}(\xi,R)$.
%such that for each $i=1,2$, $$\mathcal{O}_\xi B(m_1,R)\subseteq \mathcal{O}_\xi B(x_i,R)\subseteq t_0 \mathcal{O}_\xi B(m_1,R).$$
Consider $t_0 = 3R + 2\delta$, %3R + 2\delta + e^{R +\delta}\eta(e^{R + \delta})
and let $\Phi f \colon X^h \to Y^h$ be a map defined by sending each point $x \in X^h$ to a point $\Phi f (x)$ in $E_{\{\xi,\xi'\}}(x,t_0)$.
Then for each $i=1,2$,
$$\mathcal{O}_\xi B(m_1,R)\subseteq \mathcal{O}_\xi B(x_i,R)\subseteq t_0 \mathcal{O}_\xi B(m_1,R).$$
Since $f$ is $\eta$-quasiconformal, there exists a point $y$ in $Y^h$ such that
$$\mathcal{O}_{\xi'}B(y,R) \subseteq f(\mathcal{O}_\xi B(m_1,R)) \subseteq f(\mathcal{O}_\xi B(x_i,R)) \subseteq f(t_0\mathcal{O}_\xi B(m_1,R)) \subseteq \eta(t_0)\mathcal{O}_{\xi'}B(y,R),$$
for each $i=1,2$. This implies that the point $y$ is contained in both $E_{\{\xi,\xi'\}}(x_1,t_0)$ and $E_{\{\xi,\xi'\}}(x_2,t_0)$. Then by \autoref{prop_E(x,t)_is_bounded} there exists a constant $D=D(\delta, \eta, t_0, R) \ge 0$ such that $d_{Y^h}(\Phi f(x_i), y) \le D$, for each $i=1,2$. Therefore, $$d_{Y^h}(\Phi f(x_1), \Phi f(x_2)) \le 2D.$$
Now, let $x_1,x_2 \in X^h$ be any two points and $\gamma \colon [0,d_{X^h}(x_1,x_2)] \to X^h$ be a geodesic joining $x_1$ and $x_2$. Consider the partition $0=t_0 < t_1 < \dots < t_n = d_{X^h}(x_1,x_2)$ of the interval $[0,d_{X^h}(x_1,x_2)]$ such that $d_{X^h}(\gamma(t_{i-1}),\gamma(t_i)) = R$ for $1 \le i < n$, and $d_{X^h}(\gamma(t_{n-1}),\gamma(t_n)) \le R$. Hence, $d_{Y^h}(\Phi f(\gamma(t_{i-1}), \Phi f(\gamma(t_{i})) \le 2D$, for all $1 \le i \le n$. Therefore,
\begin{align*}
d_{Y^h}(\Phi f(x_1), \Phi f(x_2)) & \le \sum_{i=1}^n d_{Y^h}(\Phi f(\gamma(t_{i-1}), \Phi f(\gamma(t_{i})) \\
& \le 2nD \\
& \le \frac{2D}{R} d_{X^h}(x_1,x_2) + 2D
\end{align*}
Hence, there exist two constants $\lambda_1 = \lambda_1(\delta,\eta,R) \ge 1$ and $K_1 = K_1(\delta,\eta,R) \ge 0$ such that for all $y_1,y_2 \in Y^h$
$$d_{Y^h}(\Phi f(x_1), \Phi f(x_2)) \le \lambda_1 d_{X^h}(x_1,x_2)+ K_1.$$
Similarly, since the map $f^{-1}$ is also $\eta$-quasiconformal, there exist two constants $\lambda_2 = \lambda_2(\delta,\eta,R) \ge 1$, $K_2 = K_1(\delta,\eta,R) \ge 0$ and a map $\Psi f^{-1} \colon Y^h \to X^h$ (constructed similar to $\Phi f$) such that for all $y_1,y_2 \in Y^h$
$$d_{X^h}(\Psi f^{-1}(y_1), \Psi f^{-1}(y_2)) \le \lambda_2 d_{Y^h}(y_1,y_2)+ K_2.$$
Claim that there exists a constant $N \ge 0$ depending on $\delta,\eta$ and $R$ such that $d_{X^h}(\Psi f^{-1} \circ \Phi f (x),x) \le N$ for all $x \in X^h$ and $d_{Y^h}(\Phi f \circ \Psi f^{-1}(y),y) \le N$ for all $Y \in Y^h$.
Let $x \in X^h$ be a point, and let $y = \Phi f (x)$, $x'= \Psi f^{-1} \circ \Phi f (x)$. Then,
$$\mathcal{O}_{\xi'} B(y,R) \subseteq f(\mathcal{O}_{\xi} B(x,R)) \subseteq \eta(t_0) \mathcal{O}_{\xi'} B(y,R),$$
and
$$\mathcal{O}_{\xi} B(x',R) \subseteq f^{-1}(\mathcal{O}_{\xi'} B(y,R)) \subseteq \mathcal{O}_{\xi} B(x,R) \subseteq f^{-1}(\eta(t_0) \mathcal{O}_{\xi'} B(y,R)) \subseteq \eta^2(t_0) \mathcal{O}_{\xi'} B(x',R).$$
Let $\mathcal{O}_{\xi'} B(x'',R) = \eta^2(t_0) \mathcal{O}_{\xi'} B(x',R)$ and $\gamma$ be a geodesic passing through $\xi, x'$ and $x''$ with $\gamma(-\infty)=\xi$. Then the ball $B(x,R)$ intersects the geodesic $\gamma$ either on $[x',\gamma(\infty)]$ or on $[x',x'']$ or on $[x'',\xi]$. Similar to the proof of \autoref{prop_E(x,t)_is_bounded}, the distance between $x$ and $x'$ is uniformly bounded and the bound depend on $\delta,\eta$ and $R$. Hence the claim follows.
%Therefore, for any $x_1,x_2 \in X^h$
%\begin{align*}
%d_{X^h}(x_1,x_2) & \le d_{X^h}(x_1,\Psi f^{-1} \circ \Phi f (x_1)) + d_{X^h}(\Psi f^{-1} \circ \Phi f (x_1),\Psi f^{-1} \circ \Phi f (x_2)) \\
%& \qquad + d_{X^h}(\Psi f^{-1} \circ \Phi f (x_2),x_2) \\
%& \le N + (\lambda_2 d_{Y^h}(\Phi f (x_1),\Phi f (x_2)) + K_2) + N \\
%& = \lambda_2 d_{Y^h}(\Phi f (x_1),\Phi f (x_2)) + K_2 +2N.
%\end{align*}
Hence, $\Phi f$ is a quasi-isometry with an inverse quasi-isometry $\Psi$ and the quasi-isometric constants depend only on $\delta, \eta$ and $R$.
\end{proof}
\begin{thm}
The boundary extension map $\partial \Phi f\colon \partial G_1^h \to \partial G_2^h$ of $\Phi f$ coincides with $f$.
\end{thm}
\begin{proof}
Let $a$ be a point on the Bowditch boundary $\partial G_1^h$, and $[\xi,a]$ be a geodesic joining $\xi$ to $a$. This is enough to show that the image of the geodesic $[\xi,a]$ under the quasi-isometry $\Phi f$ is contained in a bounded neighbourhood of a geodesic $[\xi',f(a)]$ joining $\xi'$ and $f(a)$.
Let $x$ be a point on the geodesic $[\xi,a]$. Then $\Phi f (x) \in E_{\{\xi,\xi'\}}(x,t_0)$ and
$$\mathcal{O}_{\xi'} B(\Phi f (x),R) \subseteq f(\mathcal{O}_\xi B(a,r)) \subseteq \eta(t_0) \mathcal{O}_{\xi'} B(\Phi f (x),R),$$
where $t_0$ is taken as in the proof of \autoref{thm_qc_implies_qi_for_cusped_spaces}. If for some $y \in Y^h$, $\mathcal{O}_{\xi'} B(y,R) = \eta(t_0) \mathcal{O}_{\xi'} B(\Phi f (x),R)$, then the geodesic $[\xi',f(a)]$ intersects the ball $B(y,R)$. Therefore, the point $\Phi f (x)$ is at most at a distance $\eta(t_0) + R$ from the geodesic $[\xi',f(a)]$. Hence the result follows.
%Consider the $t_0$-annulus $\mathcal{O}_\xi B(x,R) \subseteq t_0 \mathcal{O}_\xi B(x,R)$ centered at $a$, where $t_0 = 3R + \delta$. Then by definition of quasiconformal map there exists a $\eta(t_0)$-annulus $\mathcal{O}_{\xi'} B(y,R) \subseteq \eta(t_0) \mathcal{O}_{\xi'} B(y,R)$ centered at $f(a)$ such that $$\mathcal{O}_{\xi'} B(y,R) \subseteq f(\mathcal{O}_\xi B(x,R)) \subseteq f(t_0 \mathcal{O}_\xi B(x,R)) \subseteq \eta(t_0) \mathcal{O}_{\xi'} B(y,R).$$
%This implies that $y$ is in both $E_{\{\xi,\xi'\}}(x,t_0)$ and $[\xi,f(a)]$, and hence by \autoref{prop_E(x,t)_is_bounded} $\Phi f(x)$ is contained in the closed $(2\eta(t_0)+R+\delta)$-neighbourhood of $[\xi,f(a)]$. This follows the theorem.
\end{proof}
\begin{cor}[Hyperbolic Cases]
Let $G_1$ and $G_2$ be two $\delta$-hyperbolic groups, and let $f \colon \partial G_1 \to \partial G_2$ be a $\eta$-quasiconformal map, for some $\delta \ge 0$ and some distortion map $\eta \colon [1,\infty) \to [1,\infty)$. Then
\begin{enumerate}[$i)$]
\item $f$ induces a quasi-isometry $\Phi f \colon G_1 \to G_2$.
\item the homeomorphism induced by $\Phi f$ from $\partial G_1$ to $\partial G_2$ is coincides with $f$.
\end{enumerate}
\end{cor}
Using \autoref{thm_qi_implies_qc}, it is concluded in the following statement that if a homeomorphism $f \colon \partial G_1^h \to \partial G_2^h$ being quasiconformal implies that there exists a distortion map $\eta'$ such that for every pair of base points $(\zeta,\zeta') \in \partial G_1^h \times \partial G_2^h$, the map $f \colon (\partial G_1^h, \mathbf{R}(\zeta,R)) \to (\partial G_2^h, \mathcal{R}(\zeta',R')$ is $\eta'$-quasiconformal.
\begin{cor} \label{cor_qc_is_independet_of_base_point}
Let $(G_1,\mathcal{H}_{G_1})$ and $(G_2,\mathcal{H}_{G_2})$ be two proper $\delta$-relatively hyperbolic groups, and let $f \colon \partial G_1^h \to \partial G_2^h$ be a $\eta$-quasiconformal map, for some $\delta \ge 0$ and some distortion map $\eta \colon [1,\infty) \to [1,\infty)$. Then there exists a distortion map $\eta' \colon [1,\infty) \to [1,\infty)$ depending on $\delta, \eta$ and $R$ such that for any pair of points $(\zeta,\zeta') \in \partial G_1^h \times \partial G_2^h$ with $\zeta'= f(\zeta)$, the homeomorphism $f \colon (\partial G_1^h, \mathcal{R}(\zeta,R)) \to (\partial G_2^h, \mathcal{R}(\zeta',R))$ is $\eta'$-quasiconformal.
Moreover, $\eta'$ is given by $\eta'(t) = e^{A \in(t) + B}$ for all $t \ge 1$, where the constants depend on $\delta,\eta$ and $R$.
\end{cor}
Suppose that $(\zeta_1,\zeta_1')$ and $(\zeta_2,\zeta_2')$ are two pairs of base points in $\partial G_1^h \times \partial G_2^h$ with $\zeta_i' = f(\zeta_i)$ for $i=1,2$. Therefore, by the above construction in \autoref{thm_qc_implies_qi_for_cusped_spaces}, $f$ induces two quasi-isometries namely $\Phi f_{(\zeta_1,\zeta_1')}$ and $\Phi f_{(\zeta_2,\zeta_2')}$ from $X^h$ to $Y^h$.
One can ask, is there any suitable relation between those quasi-isometries? The answer is `Yes'. They are bounded in the following sense;
%if a homeomorphism $f$ is a quasiconformal with respect to some pair of base points $(\xi,\xi') \in \partial X^h \times \partial Y^h$ with $\xi' = f(\xi)$ then so is for any pair of base points $(\xi,\xi') \in \partial X^h \times \partial Y^h$ with $\zeta' = f(\zeta)$.
\begin{lem} \label{lem_two_qi_with_same_bdry_are_bdd}
Given $\delta, K_1, K_2 \ge 0$ and $\lambda_1, \lambda_2 \ge 1$. Let $(G_1,\mathcal{H}_{G_1})$ and $(G_2,\mathcal{H}_{G_2})$ be two $\delta$-relatively hyperbolic groups with Cayley graphs $X$ and $Y$, respectively. If $\phi_1\colon X^h \to Y^h$ is a $(\lambda_1,K_1)$-quasi-isometry and $\phi_2\colon X^h \to Y^h$ is a $(\lambda_2,K_2)$-quasi-isometry such that their boundary extension maps on the Bowditch boundary coincide; that is, $\partial \phi_1 = \partial \phi_2$, then there exists a constant $D_1=D_1(\delta,\lambda_1,\lambda_2,K_1,K_2) \ge 0$ such that the difference between $\phi_1$ and $\phi_2$ is at most $D_1$; that is, $d_{X^h}(\phi_1(x),\phi_2(x)) \le D$ for all $x \in X^h$.
\end{lem}
To prove \autoref{lem_two_qi_with_same_bdry_are_bdd}, the notion of quasi-projection and some related results have been used. Here are some previews of them.
In a $\delta$-hyperbolic metric space $X$, where $\delta \ge 0$, corresponding to a triple $(a,b,c)$ of three distinct points in $X \cup \partial X$ there always exists a point $p_{abc}$, called a quasi-projection, in $X$ such that the distance of $p_{abc}$ from each side of each triangle $\triangle (a,b,c)$ with vertices $a,b,c$ is uniformly bounded, and the bound depends only on the hyperbolic constant $\delta$.
For each point $x \in X^h$ there exist three distinct points $a,b,c \in \partial G_1^h$ such that $x$ is uniformly close to quasi-projections of $(a,b,c)$ and the bound depends only on the hyperbolicity constant $\delta$ (\cite{PS2024}*{Proposition 3.21, Corollary 3.22, Remark 3.24}). Also, a quasi-isometry between two hyperbolic metric spaces coarsely preserves the quasi-projections (\cite{PS2024}*{Lemma 3.7}). (For further details about quasi-projection see \cites{Pau96,PS2024})
\begin{proof}[Proof of \autoref{lem_two_qi_with_same_bdry_are_bdd}]
Let $x$ be a point in $X^h$. Then there exist three distinct points $a,b,c \in \partial G_1^h$ such that $x$ is uniformly close to quasi-projections of $(a,b,c)$ and the bound depends only on the hyperbolicity constant $\delta$. Let $f = \partial \phi_1 = \partial \phi_2$, and let $q \in Y^h$ be a quasi-projection for the triple $(f(a),f(b),f(c))$. Then $\phi_1(x)$ is uniformly close to $q$ and the bound depends on $\delta,\lambda_1$ and $K_1$. Similarly, $\phi_2(x)$ is uniformly close to $q$, the bound depends on $\delta,\lambda_2$ and $K_2$. Hence, the result follows.
\end{proof}
\begin{cor} \label{cor_two_induced_qi_are_bdd}
There exists a constant $M_1 = M_1(\delta,\eta,R) \ge 0$ such that the induced quasi-isometries $\Phi f_{(\zeta_1,\zeta_1')}\colon X^h \to Y^h$ and $\Phi f_{(\zeta_2,\zeta_2')} \colon X^h \to Y^h$ can differ by at most $M_1$.
\end{cor}
\begin{thm}
Let $f$ and $\Phi f$ be defined as \autoref{thm_qc_implies_qi_for_cusped_spaces}. In addition, if for every pair of base points $(\zeta,\zeta') \partial G_1^h \times \partial G_2^h$ with $\zeta' = f(\zeta)$, the quasiconformal map $f\colon (\partial G_1^h,\mathcal{R}(\zeta,R)) \to (\partial G_2^h,\mathcal{R}(\zeta',R))$ $L$-coarsely preserves the shadows of horoballs relative to $\zeta$, then there exists a constant $C = C(\delta,\eta,R,L) \ge 0$ such that the quasi-isometry $\Phi f\colon X^h \to Y^h$ is a $C$-coarsely cusp-preserving map.
\end{thm}
\begin{proof}
Let $H \in \mathcal{H}_{G_1}$ be a horosphere-like set in $X^h$ centered at the parabolic point $a_H$, and $x \in H$ be a point. There exists $\zeta \in \partial G_1^h \setminus \{a_H\}$ such that $x$ lies on the geodesic $[\zeta,a_H]$.
Then by \autoref{lem_shadow_of_horospheres} there exists a constant $T_1=T_1(\delta,R) \ge 1$ and a $T_1$-ring $\mathcal{O}_\zeta B(p_{\{\zeta,H\}},R) \subseteq T_1 \mathcal{O}_\zeta B(p_{\{\zeta,H\}},R)$ in $\mathcal{R}(\zeta,R)$ such that
$$\mathcal{O}_\zeta B(p_{\{\zeta,H\}},R) \subseteq \mathcal{O}_\zeta(H^h) \subseteq T_1 \mathcal{O}_\zeta B(p_{\{\zeta,H\}},R),$$
Moreover, $d_{X^h}(p_{\{\zeta,H\}},x) = d_{X^h}(p_{\{\zeta,H\}},H) \le R+\delta$.
Let $H' \in \mathcal{H}_{G_2}$ be the horosphere-like set in $Y^h$ centered at the parabolic point $a_{H'}=f(a_H)$, and let $\zeta' = f(\zeta)$. Then $\zeta' \in \partial G_2^h \setminus \{a_{H'}\}$.
Then by \autoref{lem_shadow_of_horospheres} there exists a constant $T_1=T_1(\delta,R) \ge 1$ and a $T_1$-ring $\mathcal{O}_{\zeta'} B(q_{\{\zeta',H'\}},R) \subseteq T_1 \mathcal{O}_{\zeta'} B(q_{\{\zeta',H'\}},R)$ in $\mathcal{R}(\zeta',R)$ such that
$$\mathcal{O}_{\zeta'} B(q_{\{\zeta',H'\}},R) \subseteq \mathcal{O}_{\zeta'}(H'^h) \subseteq T_1 \mathcal{O}_{\zeta'} B(q_{\{\zeta',H'\}},R).$$
Since $f$ $L$-coarsely preserves the shadows of horoballs relative to $\zeta$, then
\begin{align*}
\frac{1}{L} \mathcal{O}_{\zeta'} B(q_{\{\zeta',H'\}},R) \subseteq f(\mathcal{O}_{\zeta} B(p_{\{\zeta,H\}},R)) &\subseteq f(\mathcal{O}_{\zeta}(H^h)) \\
&\subseteq f(T_1 \mathcal{O}_{\zeta} B(p_{\{\zeta,H\}},R)) \subseteq L \mathcal{O}_{\zeta'}B(q_{\{\zeta',H'\}},R).
\end{align*}
In particular,
$$\frac{1}{L} \mathcal{O}_{\zeta'} B(q_{\{\zeta',H'\}},R) \subseteq f(\mathcal{O}_{\zeta} B(p_{\{\zeta,H\}},R)) \subseteq L \mathcal{O}_{\zeta'}B(q_{\{\zeta',H'\}},R).$$
Also, by \autoref{cor_qc_is_independet_of_base_point}, the map $f \colon (\partial G_1^h, \mathcal{R}(\zeta,R)) \to (\partial G_2^h, \mathcal{R}(\zeta',R))$ is a $\eta'$-quasiconformal, for some distortion map $\eta' \colon [1,\infty) \to [1,\infty)$ depends on $\delta,\eta$ and $R$. Let $\Phi f_{\{\zeta,\zeta'\}}\colon X^h \to Y^h$ be the induced quasi-isometry and $y = \Phi f_{\{\zeta,\zeta'\}}(p_{\{\zeta,H\}})$. Then, $y \in E_{\{\zeta,\zeta'\}}(x,t_0)$, and
$$\mathcal{O}_{\zeta'} B(y,R) \subseteq f(\mathcal{O}_{\zeta} B(p_{\{\zeta,H\}},R)) \subseteq \eta'(t_0) \mathcal{O}_{\zeta'}B(y,R).$$
Therefore, both the shadows $\frac{1}{L} \mathcal{O}_{\zeta'} B(q_{\{\zeta',H'\}},R)$ and $\mathcal{O}_{\zeta'} B(y,R)$ are contained in the intersection $L \mathcal{O}_{\zeta'}B(q_{\{\zeta',H'\}},R) \cap \eta'(T_1) \mathcal{O}_{\zeta'}B(\Phi f(x),R)$. Therefore, similar to the proof of \autoref{prop_E(x,t)_is_bounded}, the distance between $y = \Phi f_{\{\zeta,\zeta'\}}(p_{\{\zeta,H\}})$ and $q_{\{\zeta',H'\}}$ is uniformly bounded and the bound depends on $\delta,\eta,R,L$. Hence,
\begin{align*}
d_{Y^h}(\Phi f_{\{\zeta,\zeta'\}} (x), H') \le d_{Y^h}(\Phi f_{\{\zeta,\zeta'\}} (x), \Phi f_{\{\zeta,\zeta'\}}(p_{\{\zeta,H\}})) + d_{Y^h}(\Phi f_{\{\zeta,\zeta'\}}(p_{\{\zeta,H\}}), q_{\{\zeta',H'\}}),
\end{align*}
which is uniformly bounded and the bound depends on $\delta,\eta,R,L$. Again, by \autoref{cor_two_induced_qi_are_bdd}, $\Phi f(x)$ and $\Phi f_{\{\zeta,\zeta'\}} (x)$ is uniformly bounded and hence, $\Phi f(x)$ is uniformly close to $H'$, and the bound depends only on $\delta,\eta, R$, and $L$. Hence the result follows.
\end{proof}
Let $\varphi \coloneqq Pr_{G_2} \circ \Phi f$ and $\psi \coloneqq Pr_{G_1} \circ \Psi f^{-1}$, where $Pr_{G_1} \colon X^h \to {G_1}$ and $Pr_{G_2} \colon Y^h \to {G_2}$ are two nearest point projection maps on ${G_2}$ and ${G_2}$, respectively.
\begin{prop}
If $\Phi f\colon \partial G_1^h \to \partial G_2^h$ is a coarsely cusp-preserving quasi-isometry, so is $\varphi\colon {G_1} \to {G_2}$.
\end{prop}
\begin{proof}
Let $C,K \ge 0$ , $\lambda \ge 1$, and let $\Phi f \colon X^h \to Y^h$ be a $C$-coarsely cusp-preserving $(\lambda,K)$-quasi-isometry. Suppose that $x,z \in G_1$ with $d_{G_1}(x,z) = 1$. Then there exists $H_1, H_2 \in \mathcal{H}_{G_1}$ such that $x \in H_1$ and $z \in H_2$. Then $d_{X^h}(x,z) \le 1$., and hence, $d_{Y^h}(\Phi f (x), \Phi f (z)) \le \lambda + K$.
Since $\Phi f$ is a $C$-coarsely cusp-preserving, there exist $H_1',H_2' \in \mathcal{H}_{G_2}$ such that $d_{Y^h}(\Phi f (x), H_1') \le C$ and $d_{Y^h}(\Phi f (z), H_2') \le C$. Therefore, $d_{Y^h}(\Phi f (x), \varphi x) \le C$ and $d_{Y^h}(\Phi f (z), \varphi z) \le C$. Hence,
$$d_{Y^h}(\varphi x, \varphi z) \le \lambda + K + 2C.$$
Also, since $G_2$ is properly embedded in $Y^h$, then by \autoref{lem_Uniformly_Properly_embedded_Lemma} there exists $C_8 = C_8 (\delta, \lambda, K,C)$ such that $d_{G_2}(\varphi x, \varphi z) \le C_8$.
Now, let $x,z \in G_1$ be arbitrary, and consider the partition $x=x_0,x_1,\dots,x_{n}=z$ of the geodesic $[x,z]$ with $d_{G_1}(x_{i-1},x_i) = 1$, for all $i$ and $x_i \in [x_{i-1},x_{i+1}]$, for $1 \le i \le n-1$. Then,
\begin{align*}
d_{G_2}(\varphi x, \varphi z) & \le \sum_{i=1}^{n} d_{G_2}(\varphi x_{i-1}, \varphi x_i) \\
& \le n C_8 \\
& = C_8 d_{G_1}(x,z)
\end{align*}
Similarly, for all $y,w \in {G_2}$, $$d_{G_1}(\psi y, \psi w) \le C_8 d_{G_2}(y,w).$$
Now for any $x \in G_1$,
\begin{align*}
& d_{Y^h}(\Phi f (x), \varphi (x)) \le K \\
\implies & d_{X^h}(\Psi f \circ \Phi f (x), \Psi f \circ \varphi (x)) \le \lambda K + \epsilon \\
\implies & d_{X^h}(x, \Psi f \circ \varphi (x)) \le (\lambda K + \epsilon) + \epsilon \\
\implies & d_{X^h}(x, \psi \circ \varphi (x)) \le (\lambda K + 2\epsilon) + K
\end{align*}
Since, $G_1$ is properly embedded in $X^h$, there exists $C_8' = C_8'(\delta, \lambda, \epsilon)$ such that for any $x \in G_1$ $d_{G_1}(x, \psi \circ \varphi (x)) \le C_8'$. Similarly, it can be shown that for any $y \in G_2$, $d_{G_2}(y, \varphi \circ \psi (y)) \le C_8'$.
Hence, $\varphi$ is a quasi-isometry and $\psi$ is an inverse quasi-isometry of it.
Since for any $x \in G_1$, $d_{X^h}(\varphi (x), H') \le 2C$, which is uniformly bounded and hence, $\varphi$ is also coarsely cusp-preserving quasi-isometry.
\end{proof}
Let $\varphi^h \colon X^h \to Y^h$ be an extension of $\varphi$ to the cusped spaces defined by
$$\varphi^h (x) =
\begin{cases}
\varphi(x), \text{ if } x \in G_1 \\
\Phi f (x), \text{ if } x \notin G_1.
\end{cases}$$
Then for all $x \in X^h$, the distances between $\varphi^h (x)$ and $\Phi f (x)$ are uniformly bounded, and therefore their boundary extension maps from $\partial G_1^h$ to $\partial G_2^h$ are the same and equal to $f$. Therefore, $\varphi \colon G_1 \to G_2$ is the required quasi-isometry, and hence \autoref{thm_main_theorem_2} is proved.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\bigskip \textit{\textbf{Acknowledgment.}}
The author would like to thank Prof. Abhijit Pal, Associate Professor in the Department of Mathematics and Statistics at Indian Institute of Technology Kanpur, India, for several fruitful discussions.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%\section{Appendix}
%\appendixname{}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%\newpage
%\bibliography{Bibliography.bib}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{thebibliography}{wide entry}
\bibitem[BH99]{BH99}
Bridson, Martin R. and Haefliger, Andr\'{e}.
\newblock{\it Metric spaces of non-positive curvature,}
\newblock{Grundlehren der mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences]. volume 319, Springer-Verlag, Berlin, 1999.}
\newblock{\href{https://doi.org/10.1007/978-3-662-12494-9}{MR1744486}.}
\bibitem[Bow12]{Bowditch}
Bowditch, B. H..
\newblock{\it Relatively Hyperbolic Groups.}
\newblock{International Journal of Algebra and Computation. volume 22, 2012.} \newblock{\href{https://mathscinet.ams.org/mathscinet/relay-station?mr=2922380}{MR2922380}.}
\bibitem[Dru09]{Drutu}
Dru\c tu, Cornelia.
\newblock{\it Relatively hyperbolic groups: geometry and quasi-isometric invariance.}
\newblock{Commentarii Mathematici Helvetici. A Journal of the Swiss Mathematical Society. volume 84, pages 503-546, 2009.}
\newblock{\href{https://doi.org/10.4171/CMH/171}{MR2507252}.}
\bibitem[Far98]{Farb}
Farb, B.
\newblock{\it Relatively hyperbolic groups.}
\newblock{Geometric and Functional Analysis. volume 8, pages 810-840, 1998.} \newblock{\href{https://doi.org/10.1007/s000390050075}{MR1650094}.}
\bibitem[GM08]{GM2008}
Groves, Daniel and Manning, Jason Fox.
\newblock{\it Dehn filling in relatively hyperbolic groups.}
\newblock{Israel J. Math. volume 168, pages 317-429, 2008.} \newblock{\href{https://doi.org/10.1007/s11856-008-1070-6}{MR2448064}.}
\bibitem[Gro13]{Groff}
Groff, Bradley W.
\newblock{\it Quasi-isometries, boundaries and {JSJ}-decompositions of relatively hyperbolic groups.}
\newblock{Journal of Topology and Analysis. volume 5, pages 451-475, 2013.} \newblock{\href{https://doi.org/10.1142/S1793525313500192}{MR3152211}.}
%-------------------------------------------%
\bibitem[Gro87]{Gro87}
Gromov, M.
\newblock{\it Hyperbolic groups. In Essays in group theory, of }
\newblock{Math. Sci. Res Inst. Publ. volume 8, pages 75–263, Springer, New York, 1987.} \newblock{\href{https://doi.org/10.1007/978-1-4613-9586-7_3}{MR919829}.}
%-------------------------------------------%
\bibitem[HH21]{Healy-Hruska}
Brendan Burns Healy and G. Christopher Hruska.
\newblock{\it Cusped spaces and quasi-isometries of relatively hyperbolic groups.}
\newblock{arXiv. https://arxiv.org/abs/2010.09876, 2021.}
\newblock{\href{https://arxiv.org/abs/2010.09876}{arXiv}.}
%-------------------------------------------%
\bibitem[Hru10]{Hruska}
Hruska, G. Christopher.
\newblock{\it Relative hyperbolicity and relative quasiconvexity for countable groups.}
\newblock{Algebraic \& Geometric Topology. volume 10, pages 1807-1856, 2010.} \newblock{\href{https://doi.org/10.2140/agt.2010.10.1807}{MR2684983}.}
%-------------------------------------------%
\bibitem[KM13]{Kolpakov-Martelli}
Kolpakov, Alexander and Martelli, Bruno.
\newblock{\it Hyperbolic four-manifolds with one cusp.}
\newblock{Geometric and Functional Analysis. volume 23, pages 1903-1933, 2013.} \newblock{\href{https://doi.org/10.1007/s00039-013-0247-2}{MR3132905}.}
%-------------------------------------------%
\bibitem[MS24]{Mackay-Sisto}
Mackay, John M. and Sisto, Alessandro.
\newblock{\it Maps between relatively hyperbolic spaces and between their boundaries.}
\newblock{Trans. Amer. Math. Soc. volume 377, pages 1409-1454, 2008.}
\newblock{\href{https://doi.org/10.1090/tran/9063}{MR4688555}.}
%-------------------------------------------%
\bibitem[MT10]{Mackay-Tyson}
Mackay, John M. and Tyson, Jeremy T.
\newblock{\it Conformal dimension.}
\newblock{University Lecture Series, American Mathematical Society, Providence, RI. volume 54, 2010.}
\newblock{\href{https://doi.org/10.1090/ulect/054}{MR2662522}.}
%-------------------------------------------%
\bibitem[Pal10]{Pal}
Pal, Abhijit.
\newblock{\it Cannon-Thurston Maps and Relative Hyperbolicity.}
\newblock{Indian Statistical Institute , Kolkata
Doctoral Theses. 28, 2010.} \newblock{\href{https://digitalcommons.isical.ac.in/doctoral-theses/28}{https://digitalcommons.isical.ac.in/doctoral-theses/28}.}
%-------------------------------------------%
\bibitem[Pan89]{Pan1989}
Pansu, Pierre.
\newblock{\it Dimension conforme et sph\`ere \`a l'infini des
vari\'{e}t\'{e}s \`a courbure n\'{e}gative.}
\newblock{Ann. Acad. Sci. Fenn. Ser. A I Math. Annales Academiae Scientiarum Fennicae. Series A I. Mathematica. volume 14, pages 177--212, 1989.} \newblock{\href{https://doi.org/10.5186/aasfm.1989.1424}{MR1024425}.}
%-------------------------------------------%
\bibitem[Pau96]{Pau96}
Paulin, Fr\'{e}d\'{e}ric.
\newblock{\it Un groupe hyperbolique est d\'{e}termin\'{e} par son bord.}
\newblock{J. London Math. Soc. (2). volume 54, pages 50-70, 1996.} \newblock{\href{https://doi.org/10.1112/jlms/54.1.50}{MR1395067}.}
%-------------------------------------------%
\bibitem[PS24]{PS2024}
Pal, Abhijit and Sardar, Rana.
\newblock{\it Maps between Boundaries of Relatively Hyperbolic Groups.} \newblock{arXiv:2401.14863v2,2024.\href{https://arxiv.org/abs/2401.14863v2}{arXiv}}
\end{thebibliography}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\end{document}