\documentclass[reqno]{amsart}
\usepackage{hyperref}

\AtBeginDocument{{\noindent\small
\emph{Electronic Journal of Differential Equations},
Vol. 2014 (2014), No. 264, pp. 1--14.\newline
ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu
\newline ftp ejde.math.txstate.edu}
\thanks{\copyright 2014 Texas State University - San Marcos.}
\vspace{9mm}}

\begin{document}
\title[\hfilneg EJDE-2014/264\hfil Solvability of difference equations]
{Solvability of nonlinear difference equations of fourth order}

\author[S. Stevi\'c, J. Dibl\'ik, B. Iri\v{c}anin, Z. \v{S}marda \hfil 
EJDE-2014/264\hfilneg]
{Stevo Stevi\'c, Josef Dibl\'ik, Bratislav Iri\v{c}anin, Zden\v{e}k \v{S}marda}  % in alphabetical order

\address{Stevo Stevi\'c \newline
 Mathematical Institute of the Serbian Academy of Sciences,
Knez Mihailova 36/III, 11000 Beograd, Serbia.\newline
Department of Mathematics, King Abdulaziz University, P.O. Box
80203, Jeddah 21589, Saudi Arabia} 
\email{sstevic@ptt.rs}

\address{Josef Dibl\'{\i}k \newline
Department of Mathematics and Descriptive Geometry, Faculty
of Civil Engineering, 60200, Brno University of Technology, Brno,
Czech Republic}
\email{diblik.j@fce.vutbr.cz, diblik@feec.vutbr.cz}

\address{Bratislav Iri\v{c}anin \newline
Faculty of Electrical Engineering,
Belgrade University, Bulevar Kralja Aleksandra 73,
11000 Beograd, Serbia}
\email{iricanin@etf.rs}

\address{Zden\v{e}k \v{S}marda \newline
Department of Mathematics, Faculty of Electrical Engineering
and Communication, 61600, Brno University of Technology, Brno,
Czech Republic}
\email{smarda@feec.vutbr.cz}

\thanks{Submitted September 21, 2014. Published December 22, 2014.}
\subjclass[2000]{39A10, 39A20}
\keywords{Solution to difference equation; long-term behavior of
solutions; \hfill\break\indent undefinable solutions}

\begin{abstract}
 In this article we show the existence of solutions to the nonlinear
 difference equation
 $$
 x_n=\frac{x_{n-3}x_{n-4}}{x_{n-1}(a_n+b_nx_{n-2}x_{n-3}x_{n-4})},
 \quad n\in\mathbb{N}_0,
 $$
 where the sequences $(a_n)_{n\in\mathbb{N}_0}$ and $(b_n)_{n\in\mathbb{N}_0}$,
 and initial the values $x_{-j}$, $j=\overline{1,4}$, are real numbers.
 Also we find the set of initial values for which solutions are undefinable
 when $a_n\ne 0$ and  $b_n\neq 0$ for every $n\in\mathbb{N}_0$.
 When these  two sequences are constant, we describe the long-term behavior
 of the solutions in detail.
\end{abstract}

\maketitle
\numberwithin{equation}{section}
\newtheorem{theorem}{Theorem}[section]
\newtheorem{remark}[theorem]{Remark}
\newtheorem{definition}[theorem]{Definition}
\allowdisplaybreaks

\section{Introduction}

From the very beginning of the study of difference equations, a
special attention was paid on the solvable ones. Some old results
in the topic can be found, for example, in \cite{lb} and
\cite{ll}. The publication of \cite{amen}, in which Stevi\'c
gave a theoretical explanation for the formula to solutions of the
following difference equation 
\begin{equation}
x_n=\frac{x_{n-2}}{1+x_{n-1}x_{n-2}},\quad
n\in\mathbb{N}_0,\label{b1}
\end{equation}
 presented in \cite{c}, trigged  a
renewed interest in the area (see, e.g., \cite{al0}-\cite{bl},
\cite{amc218-sde,pst2,37264}, \cite{amc218-sys1}-\cite{157943},
\cite{508523}-\cite{tyt2}).
There are also some equations and systems which are recently
studied by using some solvable equations (see, e.g.,
\cite{jdea129,tjm2,jmaa376,ejqtde-maxsde}).

In several papers were later studied some special cases of the
following extension of equation \eqref{b1} 
\begin{equation}
x_n=\frac{x_{n-2}}{a_n+b_nx_{n-1}x_{n-2}},\quad
n\in\mathbb{N}_0,\label{b2}
\end{equation}
 where $(a_n)_{n\in\mathbb{N}_0}$,
$(b_n)_{n\in\mathbb{N}_0}$, and the initial values $x_{-2},x_{-1}$ are
real numbers, as well as some other extensions, by using the main
idea in \cite{amen}
(see, e.g., \cite{al0,al1,bl,pst2,amc218-dcept,amc218-hde,409237}).
Some systems of difference equations which are
extensions of equation \eqref{b1} were studied, in
\cite{amc218-sys1,amc-thos,amc219-ssfo,
amc219-ssdek,amc219-sdeoos,amc219-sstho,508523}.
For related results see
\cite{bb,amc218-sde,37264,amc-218solsys,
amc-mtsde,amc-219sscf,amc219-ssktho,
ejqtde1,157943,541761,108047,jdea205,tyt2}.

Note that, if $(x_n)_{n\ge -2}$ is a solution to equation
\eqref{b2} such that $x_n\ne 0$, $n\ge -2$, then we have that
$$
x_n=\frac{x_{n-1}x_{n-2}}{x_{n-1}(a_n+b_nx_{n-1}x_{n-2})}.
$$
This form of equation \eqref{b2} suggests investigation of the
related equations which in the numerators have more that one
factor, after cancelling the same ones.

Motivated by this idea, here we will study the next difference
equation
\begin{equation}
x_n=\frac{x_{n-3}x_{n-4}}{x_{n-1}(a_n+b_nx_{n-2}x_{n-3}x_{n-4})},\quad
 n\in\mathbb{N}_0,\label{e1}
\end{equation}
where $(a_n)_{n\in\mathbb{N}_0}$, $(b_n)_{n\in\mathbb{N}_0}$ and the initial
values $x_{-j}$, $j\in\{1,2,3,4\}$, are real numbers, which is
naturally imposed for further studies in this direction.

For a solution $(x_n)_{n\ge -s}$ of the difference equation
\begin{equation}
x_n=f(x_{n-1},\ldots,x_{n-s}),\quad n\in\mathbb{N}_0,\label{e}
\end{equation}
where $f:\mathbb{R}^s\to \mathbb{R}$, $s\in \mathbb{N}$, is said that it is periodic
with period $p$, if there is an $n_0\ge -s$ such that
$$
x_{n+p}=x_n,\quad\text{for } n\ge n_0.
$$
If $n_0\ne -s$, sometimes is said that the solution is eventually periodic.
For some results in the area (mostly on classes of equations not
related to differential ones), see, e.g. \cite{jdea128, dcdis4,
amc215-854, k1, kk, kkn, wk, kg, ps4, pss1, dcdis1, 37264, um83-2}
and the references therein.


This article is organized as follows. First, we will show that
equation \eqref{e1} can be solved in closed form. Then, we will
study in detail the long-term behavior of their solutions for the
case when $(a_n)_{n\in\mathbb{N}_0}$ and $(b_n)_{n\in\mathbb{N}_0}$ are constant
sequences. Finally, we will find the domain of undefinable
solutions of the equation for the case when $a_n\ne 0\ne b_n$, for
every $n\in\mathbb{N}_0$.

\section{Closed form solutions for \eqref{e1}}

Let $(x_n)_{n\ge -4}$ be a solution to equation \eqref{e1}. If
$x_{-j}=0$ for some $j\in\{3,4\}$, then clearly $x_0=0$, so that
$x_1$ is not defined. If $x_{-2}=0$, then $x_1=0$, so that $x_2$
is not defined. If $x_{-1}=0$, then clearly $x_0$ is not defined.
So, if $x_{-j}=0$ for some $j\in\{1,2,3,4\}$, then the solution is
not defined.

On the other hand, if there is an $n\in\mathbb{N}_0$, say $n=n_0$, such
that $x_{n_0}=0$ and $x_n\ne 0$ for $0\le n\le n_0-1$. Then
$x_{n_0-3}=0$ or $x_{n_0-4}=0$, so that it must be $n_0\le 3$. If
$n_0\in\{0,1,2\}$, then clearly $x_{-j}=0$ for some
$j\in\{1,2,3,4\}$. If $n_0=3$, then $x_0=0$ (the case already
treated) or $x_{-1}=0$. Hence, in all the cases there is a
$j\in\{1,2,3,4\}$ such that $x_{-j}=0$, so that according to the
first part of the consideration such solutions are not defined.

Therefore, for every well-defined solution of equation \eqref{e1}
\begin{equation}
 x_{-j}\ne 0,\quad 1\le j\le 4,\label{c7}
\end{equation}
is equivalent to $x_n\ne 0$, $n\ge -4$.

Hence, for solutions satisfying \eqref{c7}, the change of
variables 
\begin{equation}
y_n=\frac1{x_nx_{n-1}x_{n-2}},\quad n\ge
-2,\label{a3}
\end{equation}
is possible and the sequence $(y_n)_{n\ge -2}$
satisfies the equation
\begin{equation}
y_n=a_ny_{n-2}+b_n,\quad n\in\mathbb{N}_0,\label{l1}
\end{equation}
which means that
\begin{equation}
y_{2m+i}=a_{2m+i}y_{2(m-1)+i}+b_{2m+i},\label{l2}
\end{equation}
for every
$m\in\mathbb{N}_0$ and $i\in\{0,1\}$, that is, $(y_{2m+i})_{m\ge -1}$,
$i\in\{0,1\}$, are solutions to the difference equations
 \begin{equation}
z_m=a_{2m+i}z_{m-1}+b_{2m+i},\quad m\in\mathbb{N}_0,\label{l3}
\end{equation}
$i\in\{0,1\}$.

By a known formula, it follows that 
\begin{equation}
y_{2m+i}=y_{i-2}\prod_{j=0}^ma_{2j+i}+\sum_{l=0}^mb_{2l+i}\prod_{j=l+1}^ma_{2j+i},
\quad m\in\mathbb{N}_0,\label{l5}
\end{equation}
$i\in\{0,1\}$, are general solutions to
the equations in \eqref{l3}.
From \eqref{a3} it follows that
$$
x_{3m+i}=\frac1{y_{3m+i}x_{3m+i-1}x_{3m+i-2}}
=\frac{y_{3m+i-1}}{y_{3m+i}}x_{3(m-1)+i},
$$
$i\in\{0,1,2\}$, and consequently
$$
x_{3m+i}=\frac{y_{3m+i-1}}{y_{3m+i}}\frac{y_{3m+i-4}}{y_{3m+i-3}}x_{3(m-2)+i},
$$
$i\in\{0,1,2\}$, so by using the change $m\to 2m+j$, $m\in\mathbb{N}_0$,
 $j\in\{0,1\}$, is obtained
$$
x_{6m+3j+i}=\frac{y_{6m+3j+i-1}}{y_{6m+3j+i}}
\frac{y_{6m+3j+i-4}}{y_{6m+3j+i-3}}x_{6(m-1)+3j+i},
$$
$i\in\{0,1,2\}$, $j\in\{0,1\}$, which can be written in the  form
\begin{equation}
x_{6m+j}=\frac{y_{6m+j-1}}{y_{6m+j}}
\frac{y_{6m+j-4}}{y_{6m+j-3}}x_{6(m-1)+j},\quad m\in\mathbb{N}_0,\label{l6}
\end{equation}
$j\in\overline{0,5}$, as far as $6m+j\ge 2$.
From \eqref{l6} it follows that
\begin{equation}
x_{6m+l}=x_{l-6}\prod_{s=0}^m\frac{y_{6s+l-1}}{y_{6s+l}}\frac{y_{6s+l-4}}{y_{6s+l-3}},\quad m\ge
-1,\label{l7}
\end{equation}
for $l=\overline{2,7}$.

Employing the formulas in \eqref{l5}, in equalities \eqref{l7} for
$l$ even and odd separately, we have 
\begin{align*}
x_{6m+2i}&= x_{2i-6}\prod_{s=0}^m\frac{y_{6s+2i-1}}{y_{6s+2i}}
 \frac{y_{6s+2i-4}}{y_{6s+2i-3}}\\
&= x_{2i-6}\prod_{s=0}^m\frac{y_{-1}\prod_{j=0}^{3s+i-1}a_{2j+1}+\sum_{l=0}^{3s+i-1}b_{2l+1}\prod_{j=l+1}^{3s+i-1}a_{2j+1}}
{y_{-2}\prod_{j=0}^{3s+i}a_{2j}+\sum_{l=0}^{3s+i}b_{2l}\prod_{j=l+1}^{3s+i}a_{2j}}\\
&\quad \times\frac{y_{-2}\prod_{j=0}^{3s+i-2}a_{2j}+\sum_{l=0}^{3s+i-2}b_{2l}
 \prod_{j=l+1}^{3s+i-2}a_{2j}} {y_{-1}\prod_{j=0}^{3s+i-2}a_{2j+1}
 +\sum_{l=0}^{3s+i-2}b_{2l+1} \prod_{j=l+1}^{3s+i-2}a_{2j+1}}\\
&= x_{2i-6}\prod_{s=0}^m\frac{(x_{-1}x_{-2}x_{-3})^{-1}
 \prod_{j=0}^{3s+i-1}a_{2j+1}+\sum_{l=0}^{3s+i-1}b_{2l+1}
 \prod_{j=l+1}^{3s+i-1}a_{2j+1}}{(x_{-2}x_{-3}x_{-4})^{-1}
 \prod_{j=0}^{3s+i}a_{2j}+\sum_{l=0}^{3s+i}b_{2l}\prod_{j=l+1}^{3s+i}a_{2j}}\\
&\quad\times\frac{(x_{-2}x_{-3}x_{-4})^{-1}\prod_{j=0}^{3s+i-2}a_{2j}
 +\sum_{l=0}^{3s+i-2}b_{2l}\prod_{j=l+1}^{3s+i-2}a_{2j}}
 {(x_{-1}x_{-2}x_{-3})^{-1}\prod_{j=0}^{3s+i-2}a_{2j+1}
 +\sum_{l=0}^{3s+i-2}b_{2l+1}\prod_{j=l+1}^{3s+i-2}a_{2j+1}}\\
&= x_{2i-6}\prod_{s=0}^m\frac{\prod_{j=0}^{3s+i-1}a_{2j+1}
 +x_{-1}x_{-2}x_{-3}\sum_{l=0}^{3s+i-1}b_{2l+1}\prod_{j=l+1}^{3s+i-1}a_{2j+1}}
 {\prod_{j=0}^{3s+i}a_{2j}+x_{-2}x_{-3}x_{-4}\sum_{l=0}^{3s+i}b_{2l}
 \prod_{j=l+1}^{3s+i}a_{2j}}\\
&\quad\times\frac{\prod_{j=0}^{3s+i-2}a_{2j}+x_{-2}x_{-3}x_{-4}
 \sum_{l=0}^{3s+i-2}b_{2l}\prod_{j=l+1}^{3s+i-2}a_{2j}}
 {\prod_{j=0}^{3s+i-2}a_{2j+1}+x_{-1}x_{-2}x_{-3}
 \sum_{l=0}^{3s+i-2}b_{2l+1}\prod_{j=l+1}^{3s+i-2}a_{2j+1}},
\end{align*}
for $m\ge -1$, $i\in\{1,2,3\}$,  and
\begin{align*}
x_{6m+2i+1}
&= x_{2i-5}\prod_{s=0}^m\frac{y_{6s+2i}}{y_{6s+2i+1}}
 \frac{y_{6s+2i-3}}{y_{6s+2i-2}}\\
&= x_{2i-5}\prod_{s=0}^m\frac{y_{-2}\prod_{j=0}^{3s+i}a_{2j}
 +\sum_{l=0}^{3s+i}b_{2l}\prod_{j=l+1}^{3s+i}a_{2j}}
{y_{-1}\prod_{j=0}^{3s+i}a_{2j+1}+\sum_{l=0}^{3s+i}b_{2l+1}
 \prod_{j=l+1}^{3s+i}a_{2j+1}}\\
&\quad\times\frac{y_{-1}\prod_{j=0}^{3s+i-2}a_{2j+1}
 +\sum_{l=0}^{3s+i-2}b_{2l+1}\prod_{j=l+1}^{3s+i-2}a_{2j+1}}
{y_{-2}\prod_{j=0}^{3s+i-1}a_{2j}+\sum_{l=0}^{3s+i-1}b_{2l}
 \prod_{j=l+1}^{3s+i-1}a_{2j}}\\
&= x_{2i-5}\prod_{s=0}^m\frac{(x_{-2}x_{-3}x_{-4})^{-1}
 \prod_{j=0}^{3s+i}a_{2j}+\sum_{l=0}^{3s+i}b_{2l}\prod_{j=l+1}^{3s+i}a_{2j}}
 {(x_{-1}x_{-2}x_{-3})^{-1}\prod_{j=0}^{3s+i}a_{2j+1}+\sum_{l=0}^{3s+i}b_{2l+1}\prod_{j=l+1}^{3s+i}a_{2j+1}}\\
&\quad\times\frac{(x_{-1}x_{-2}x_{-3})^{-1}\prod_{j=0}^{3s+i-2}a_{2j+1}
 +\sum_{l=0}^{3s+i-2}b_{2l+1}\prod_{j=l+1}^{3s+i-2}a_{2j+1}}
{(x_{-2}x_{-3}x_{-4})^{-1}\prod_{j=0}^{3s+i-1}a_{2j}
 +\sum_{l=0}^{3s+i-1}b_{2l}\prod_{j=l+1}^{3s+i-1}a_{2j}}\\
&= x_{2i-5}\prod_{s=0}^m\frac{\prod_{j=0}^{3s+i}a_{2j}
 +x_{-2}x_{-3}x_{-4}\sum_{l=0}^{3s+i}b_{2l}\prod_{j=l+1}^{3s+i}a_{2j}}
 {\prod_{j=0}^{3s+i}a_{2j+1}+x_{-1}x_{-2}x_{-3}\sum_{l=0}^{3s+i}b_{2l+1}
 \prod_{j=l+1}^{3s+i}a_{2j+1}}\\
&\quad\times\frac{\prod_{j=0}^{3s+i-2}a_{2j+1}
 +x_{-1}x_{-2}x_{-3}\sum_{l=0}^{3s+i-2}b_{2l+1}\prod_{j=l+1}^{3s+i-2}a_{2j+1}}
{\prod_{j=0}^{3s+i-1}a_{2j}+x_{-2}x_{-3}x_{-4}
 \sum_{l=0}^{3s+i-1}b_{2l}\prod_{j=l+1}^{3s+i-1}a_{2j}},
\end{align*}
for $m\ge -1$, $i\in\{1,2,3\}$.

Hence  the following theorem holds.

\begin{theorem} \label{thm1} 
 If $(x_n)_{n\ge -4}$ is a well-defined solution of equation \eqref{e1}, 
then it can be represented in the form
\begin{equation}
\begin{aligned}
x_{6m+2i}
&= x_{2i-6}\prod_{s=0}^m\frac{\prod_{j=0}^{3s+i-1}a_{2j+1}+x_{-1}x_{-2}x_{-3}
 \sum_{l=0}^{3s+i-1}b_{2l+1}\prod_{j=l+1}^{3s+i-1}a_{2j+1}}
{\prod_{j=0}^{3s+i}a_{2j}+x_{-2}x_{-3}x_{-4}\sum_{l=0}^{3s+i}b_{2l}
 \prod_{j=l+1}^{3s+i}a_{2j}}\\
&\quad\times\frac{\prod_{j=0}^{3s+i-2}a_{2j}+x_{-2}x_{-3}x_{-4}
 \sum_{l=0}^{3s+i-2}b_{2l}\prod_{j=l+1}^{3s+i-2}a_{2j}}
{\prod_{j=0}^{3s+i-2}a_{2j+1}+x_{-1}x_{-2}x_{-3}\sum_{l=0}^{3s+i-2}b_{2l+1}
 \prod_{j=l+1}^{3s+i-2}a_{2j+1}},
\end{aligned}\label{l9}
\end{equation}
and
\begin{equation}
\begin{aligned}
x_{6m+2i+1}
&= x_{2i-5}\prod_{s=0}^m\frac{\prod_{j=0}^{3s+i}a_{2j}+x_{-2}x_{-3}x_{-4}
 \sum_{l=0}^{3s+i}b_{2l}\prod_{j=l+1}^{3s+i}a_{2j}}
{\prod_{j=0}^{3s+i}a_{2j+1}+x_{-1}x_{-2}x_{-3}\sum_{l=0}^{3s+i}b_{2l+1}
 \prod_{j=l+1}^{3s+i}a_{2j+1}}\\
&\quad \times\frac{\prod_{j=0}^{3s+i-2}a_{2j+1}+x_{-1}x_{-2}x_{-3}
 \sum_{l=0}^{3s+i-2}b_{2l+1}\prod_{j=l+1}^{3s+i-2}a_{2j+1}}
{\prod_{j=0}^{3s+i-1}a_{2j}+x_{-2}x_{-3}x_{-4}\sum_{l=0}^{3s+i-1}b_{2l}
 \prod_{j=l+1}^{3s+i-1}a_{2j}},
\end{aligned}\label{l10}
\end{equation}
for $m\ge -1$, $i\in\{1,2,3\}$.
\end{theorem}

\begin{remark} \label{rmk1} \rm
The formulas in \eqref{l9} and \eqref{l10}  can be regarded as an 
integral formula for general solution of equation \eqref{e1}. 
In fact, they include  non-defined solutions, which will be described in
detail in the last section of this article.
\end{remark}

\section{Constant coefficients case}

In this section we study equation \eqref{e1}  when
$$
a_n=a,\quad b_n=b,\quad n\in \mathbb{N}_0,
$$
where $a$ and $b$ are some real constants.
In this case, equation \eqref{e1} becomes
\begin{equation}
x_n=\frac{x_{n-3}x_{n-4}}{x_{n-1}(a+bx_{n-2}x_{n-3}x_{n-4})},
\quad n\in\mathbb{N}_0.\label{e1c}
\end{equation}
If $x_{-j}\ne 0$, $j=\overline{1,4}$, from \eqref{l9} and \eqref{l10} we
have
\begin{equation}
\begin{aligned}
x_{6m+2i}
&=x_{2i-6}\prod_{s=0}^m\frac{(x_{-1}x_{-2}x_{-3})^{-1}
 \prod_{j=0}^{3s+i-1}a+\sum_{l=0}^{3s+i-1}b\prod_{j=l+1}^{3s+i-1}a}
{(x_{-2}x_{-3}x_{-4})^{-1}\prod_{j=0}^{3s+i}a+\sum_{l=0}^{3s+i}b
 \prod_{j=l+1}^{3s+i}a}\\
&\quad \times\frac{(x_{-2}x_{-3}x_{-4})^{-1}\prod_{j=0}^{3s+i-2}a
 +\sum_{l=0}^{3s+i-2}b\prod_{j=l+1}^{3s+i-2}a}
 {(x_{-1}x_{-2}x_{-3})^{-1}\prod_{j=0}^{3s+i-2}a
 +\sum_{l=0}^{3s+i-2}b\prod_{j=l+1}^{3s+i-2}a}\\
&=x_{2i-6}\prod_{s=0}^m\frac{(x_{-1}x_{-2}x_{-3})^{-1}a^{3s+i}
 +b\sum_{l=0}^{3s+i-1}a^{3s+i-1-l}}
{(x_{-2}x_{-3}x_{-4})^{-1}a^{3s+i+1}+b\sum_{l=0}^{3s+i}a^{3s+i-l}}\\
&\quad\times\frac{(x_{-2}x_{-3}x_{-4})^{-1}a^{3s+i-1}
 +b\sum_{l=0}^{3s+i-2}a^{3s+i-2-l}}
{(x_{-1}x_{-2}x_{-3})^{-1}a^{3s+i-1}+b\sum_{l=0}^{3s+i-2}a^{3s+i-2-l}},
\end{aligned}\label{l11}
\end{equation}
$m\ge -1$, $i\in\{1,2,3\}$, and
\begin{equation}
\begin{aligned}
x_{6m+2i+1}&= x_{2i-5}\prod_{s=0}^m\frac{(x_{-2}x_{-3}x_{-4})^{-1}
 \prod_{j=0}^{3s+i}a+\sum_{l=0}^{3s+i}b\prod_{j=l+1}^{3s+i}a}
{(x_{-1}x_{-2}x_{-3})^{-1}\prod_{j=0}^{3s+i}a+\sum_{l=0}^{3s+i}b
 \prod_{j=l+1}^{3s+i}a}\\
&\quad\times\frac{(x_{-1}x_{-2}x_{-3})^{-1}\prod_{j=0}^{3s+i-2}a
 +\sum_{l=0}^{3s+i-2}b\prod_{j=l+1}^{3s+i-2}a}
{(x_{-2}x_{-3}x_{-4})^{-1}\prod_{j=0}^{3s+i-1}a
 +\sum_{l=0}^{3s+i-1}b\prod_{j=l+1}^{3s+i-1}a}\\
&= x_{2i-5}\prod_{s=0}^m\frac{(x_{-2}x_{-3}x_{-4})^{-1}a^{3s+i+1}
 +b\sum_{l=0}^{3s+i}a^{3s+i-l}}
{(x_{-1}x_{-2}x_{-3})^{-1}a^{3s+i+1}+b\sum_{l=0}^{3s+i}a^{3s+i-l}}\\
&\quad \times\frac{(x_{-1}x_{-2}x_{-3})^{-1}a^{3s+i-1}
 +b\sum_{l=0}^{3s+i-2}a^{3s+i-2-l}}
{(x_{-2}x_{-3}x_{-4})^{-1}a^{3s+i}+b\sum_{l=0}^{3s+i-1}a^{3s+i-1-l}},
\end{aligned}\label{l12}
\end{equation}
for $m\ge -1$, $i\in\{1,2,3\}$.


If $a\ne 1$, then from \eqref{l11} and
\eqref{l12} we have 
\begin{equation}
\begin{aligned}
x_{6m+2i}
&=x_{2i-6}\prod_{s=0}^m\frac{(x_{-1}x_{-2}x_{-3})^{-1}(1-a)a^{3s+i}+b(1-a^{3s+i})}
{(x_{-2}x_{-3}x_{-4})^{-1}(1-a)a^{3s+i+1}+b(1-a^{3s+i+1})}\\
&\quad \times\frac{(x_{-2}x_{-3}x_{-4})^{-1}(1-a)a^{3s+i-1}+b(1-a^{3s+i-1})}
{(x_{-1}x_{-2}x_{-3})^{-1}(1-a)a^{3s+i-1}+b(1-a^{3s+i-1})}\\
&=x_{2i-6}\prod_{s=0}^m\frac{((x_{-1}x_{-2}x_{-3})^{-1}(1-a)-b)a^{3s+i}+b}
{((x_{-2}x_{-3}x_{-4})^{-1}(1-a)-b)a^{3s+i+1}+b}\\
&\quad \times\frac{((x_{-2}x_{-3}x_{-4})^{-1}(1-a)-b)a^{3s+i-1}+b}
{((x_{-1}x_{-2}x_{-3})^{-1}(1-a)-b)a^{3s+i-1}+b},
\end{aligned}\label{l14}
\end{equation}
for $m\ge -1$, $i\in\{1,2,3\}$, and
\begin{equation}
\begin{aligned}
x_{6m+2i+1}
&=x_{2i-5}\prod_{s=0}^m\frac{(x_{-2}x_{-3}x_{-4})^{-1}(1-a)a^{3s+i+1}+b(1-a^{3s+i+1})}
{(x_{-1}x_{-2}x_{-3})^{-1}(1-a)a^{3s+i+1}+b(1-a^{3s+i+1})}\\
&\quad \times\frac{(x_{-1}x_{-2}x_{-3})^{-1}(1-a)a^{3s+i-1}+b(1-a^{3s+i-1})}
{(x_{-2}x_{-3}x_{-4})^{-1}(1-a)a^{3s+i}+b(1-a^{3s+i})}\\
&=x_{2i-5}\prod_{s=0}^m\frac{((x_{-2}x_{-3}x_{-4})^{-1}(1-a)-b)a^{3s+i+1}+b}
{((x_{-1}x_{-2}x_{-3})^{-1}(1-a)-b)a^{3s+i+1}+b}\\
&\quad \times\frac{((x_{-1}x_{-2}x_{-3})^{-1}(1-a)-b)a^{3s+i-1}+b}
{((x_{-2}x_{-3}x_{-4})^{-1}(1-a)-b)a^{3s+i}+b},
\end{aligned}\label{l15}
\end{equation}
for $m\ge -1$, $i\in\{1,2,3\}$.


\subsection*{Case $a=1$} From \eqref{l11} and
\eqref{l12} we have
\begin{equation}
\begin{aligned}
&x_{6m+2i}\\
&=x_{2i-6}\prod_{s=0}^m\frac{(x_{-1}x_{-2}x_{-3})^{-1}+b(3s+i)}
{(x_{-2}x_{-3}x_{-4})^{-1}+b(3s+i+1)}\frac{(x_{-2}x_{-3}x_{-4})^{-1}+b(3s+i-1)}
{(x_{-1}x_{-2}x_{-3})^{-1}+b(3s+i-1)},
\end{aligned} \label{l16}
\end{equation}
 for $m\ge -1$, $i\in\{1,2,3\}$, and
\begin{equation}
\begin{aligned}
&x_{6m+2i+1}\\
&=x_{2i-5}\prod_{s=0}^m\frac{(x_{-2}x_{-3}x_{-4})^{-1}+b(3s+i+1)}
{(x_{-1}x_{-2}x_{-3})^{-1}+b(3s+i+1)}\frac{(x_{-1}x_{-2}x_{-3})^{-1}+b(3s+i-1)}
{(x_{-2}x_{-3}x_{-4})^{-1}+b(3s+i)},
\end{aligned}\label{l17}
\end{equation}
 for $m\ge -1$, $i\in\{1,2,3\}$.


\section{Long-term behavior of solutions to \eqref{e1c}}

Before we formulate and prove the main results
in this section, we want to introduce the following
notation
$$
y_{-1}=(x_{-1}x_{-2}x_{-3})^{-1},\quad 
y_{-2}=(x_{-2}x_{-3}x_{-4})^{-1},
$$
which are consistent with the considerations and notation in the
previous section (see the change of variables \eqref{a3}).
Set 
\begin{equation}
p_m^{2i}=\frac{((y_{-1}(1-a)-b)a^{3m+i}+b)((y_{-2}(1-a)-b)a^{3m+i-1}+b)}
{((y_{-2}(1-a)-b)a^{3m+i+1}+b)((y_{-1}(1-a)-b)a^{3m+i-1}+b)}\label{ps}
\end{equation}
and
\begin{equation}
p_m^{2i+1}=\frac{((y_{-2}(1-a)-b)a^{3m+i+1}+b)((y_{-1}(1-a)-b)a^{3m+i-1}+b)}
{((y_{-1}(1-a)-b)a^{3m+i+1}+b)((y_{-2}(1-a)-b)a^{3m+i}+b)},\label{ps1}
\end{equation}
for $m\ge -1$ and $i\in\{1,2,3\}$.

\subsection*{Case $a\ne -1$, $b\ne 0$} 
First we describe the long-term behavior of well-defined solution of
equation \eqref{e1c} for the case $a\ne -1$, $b\ne 0$.


\begin{theorem} \label{thm2}
Assume that $a\ne -1$, $b\ne 0$ and
$(x_n)_{n\ge -4}$ is a well-defined solution of equation
\eqref{e1c}. Then the following statements are true.
\begin{itemize}
\item[(a)]  If $|a|>1$, $y_{-1}\ne b/(1-a)\ne y_{-2}$, then $x_n\to 0$ as
$n\to+\infty$.

\item[(b)] If $|a|>1$, $y_{-1}=b/(1-a)\ne y_{-2}$, then $x_{6m+2i}\to
0$, $i\in\{1,2,3\}$ as $m\to+\infty$.

\item[(c)] If $|a|>1$, $y_{-1}=b/(1-a)\ne y_{-2}$, then 
$|x_{6m+2i+1}|\to \infty$, $i\in\{1,2,3\}$  as $m\to+\infty$.

\item[(d)] If $|a|>1$, $y_{-1}\ne b/(1-a)=y_{-2}$, then 
$|x_{6m+2i}|\to \infty$, $i\in\{1,2,3\}$ as $m\to+\infty$.

\item[(e)] If $|a|>1$, $y_{-1}\ne b/(1-a)=y_{-2}$, then 
$x_{6m+2i+1}\to 0$, $i\in\{1,2,3\}$ as $m\to+\infty$.

\item[(f)] If $|a|<1$, then the sequences $(x_{6m+j})_{m\in\mathbb{N}_0}$
converge for every $j=\overline{0,5}$.

\item[(g)] If $y_{-1}=b/(1-a)=y_{-2}$ or $a=0$, then
$x_{6m+j}=x_{j-6}$, $m\in\mathbb{N}_0$, $j=\overline{2,7}$.

\item[(h)] If $a=1$, then $x_n\to 0$ as $n\to+\infty$.
\end{itemize}
\end{theorem}


\begin{proof} (a): From \eqref{ps} and \eqref{ps1}, we have
$$
p_m^{2i}=\frac{((y_{-1}(1-a)-b)+(b/a^{3m+i}))((y_{-2}(1-a)-b)+(b/a^{3m+i-1}))}
{((y_{-2}(1-a)-b)a+(b/a^{3m+i}))((y_{-1}(1-a)-b)+(b/a^{3m+i-1}))}\to \frac1a
$$
and
$$
p_m^{2i+1}=\frac{((y_{-2}(1-a)-b)+(b/a^{3m+i+1}))((y_{-1}(1-a)-b)+(b/a^{3m+i-1}))}
{((y_{-1}(1-a)-b)+(b/a^{3m+i+1}))((y_{-2}(1-a)-b)a+(b/a^{3m+i-1}))}\to \frac1a,
$$
as $m\to+\infty$, for every $i\in\{1,2,3\}$, which means that
\begin{equation}
\lim_{m\to+\infty}p_m^j=\frac1a,\label{f7}
\end{equation}
for every $j=\overline{2,7}$.
From \eqref{l14}, \eqref{l15}, \eqref{f7} and the assumption $|a|>1$,
statement (a) follows easily.
\smallskip

(b) and  (c): In this case we have 
\begin{gather}
p_m^{2i}=\frac{(y_{-2}(1-a)-b)a^{3m+i-1}+b}
{(y_{-2}(1-a)-b)a^{3m+i+1}+b}\to \frac1{a^2},\label{ps2}
\\
p_m^{2i+1}=\frac{(y_{-2}(1-a)-b)a^{3m+i+1}+b}
{(y_{-2}(1-a)-b)a^{3m+i}+b}\to a,\label{ps3}
\end{gather}
as $m\to+\infty$, for every $i\in\{1,2,3\}$,
From \eqref{l14}, \eqref{l15}, \eqref{ps2}, \eqref{ps3} 
and the assumption $|a|>1$, these two statements follow easily.
\smallskip

(d) and (e): In this case we have 
\begin{gather}
p_m^{2i}=\frac{(y_{-1}(1-a)-b)a^{3m+i}+b}
{(y_{-1}(1-a)-b)a^{3m+i-1}+b}\to a,\label{ps5}
\\
p_m^{2i+1}=\frac{(y_{-1}(1-a)-b)a^{3m+i-1}+b}
{(y_{-1}(1-a)-b)a^{3m+i+1}+b}\to \frac1{a^2},\label{ps6}
\end{gather}
as $m\to+\infty$, for every $i\in\{1,2,3\}$,
From \eqref{l14}, \eqref{l15}, \eqref{ps5}, \eqref{ps6} and the 
assumption $|a|>1$, these two statements
follow easily.
\smallskip

 (f):  Using the asymptotic relation 
\begin{equation}
(1+x)^{-1}=1-x+O(x^2),\label{ar}
\end{equation}
when $x$ is in a neighborhood of zero, we have
\begin{equation}
\begin{aligned}
p_m^{2i}
&=\frac{(1+(y_{-1}(1-a)-b)a^{3m+i}/b)(1+(y_{-2}(1-a)-b)a^{3m+i-1}/b)}
{(1+(y_{-2}(1-a)-b)a^{3m+i+1}/b)(1+(y_{-1}(1-a)-b)a^{3m+i-1}/b)}\\
&=1+\frac1b\Big((y_{-1}(1-a)-b)\Big(1-\frac1a\Big)+(y_{-2}(1-a)-b)
\Big(\frac1a-a\Big)\Big)a^{3m+i}\\
&\quad +o(a^{3m})
\end{aligned}\label{ps7}
\end{equation}
and
\begin{equation}
\begin{aligned}
p_m^{2i+1}
&=\frac{(1+(y_{-2}(1-a)-b)a^{3m+i+1}/b)(1+(y_{-1}(1-a)-b)a^{3m+i-1}/b)}
{(1+(y_{-1}(1-a)-b)a^{3m+i+1}/b)(1+(y_{-2}(1-a)-b)a^{3m+i}/b)}\\
&=1+\frac1b\Big((y_{-2}(1-a)-b)(a-1)+(y_{-1}(1-a)-b)
\Big(\frac1a-a\Big)\Big)a^{3m+i}\\
&\quad +o(a^{3m}),
\end{aligned}\label{ps8}
\end{equation}
for every  $i\in\{1,2,3\}$ and sufficiently large $m$.
From \eqref{ps7}, \eqref{ps8}, the assumption $|a|<1$, and by a
known result on the convergence of products the result follows easily.
\smallskip

 (g): The result follows from direct calculations and formulas
\eqref{l14} and \eqref{l15}.

 (h): Let
\begin{gather*}
r_m^{2i}=\frac{y_{-1}+bi+3bm}{y_{-2}+b(i+1)+3bm}
\frac{y_{-2}+b(i-1)+3bm}{y_{-1}+b(i-1)+3bm},\\
r_m^{2i+1}=\frac{y_{-2}+b(i+1)+3bm}
{y_{-1}+b(i+1)+3bm}\frac{y_{-1}+b(i-1)+3bm}
{y_{-2}+bi+3bm},
\end{gather*} 
for $i\in\{1,2,3\}$.
Then we have
\begin{equation}
r_m^{2i}
=\frac{\Big(1+\frac{y_{-1}+bi}{3bm}\Big)}{\Big(1+\frac{y_{-2}+b(i+1)}{3bm}\Big)}
\frac{\Big(1+\frac{y_{-2}+b(i-1)}{3bm}\Big)}{\Big(1+\frac{y_{-1}+b(i-1)}{3bm}\Big)}
=1-\frac1{3m}+O\Big(\frac1{m^2}\Big) \label{f22}
\end{equation}
and
\begin{equation}
 r_m^{2i+1}
=\frac{\Big(1+\frac{y_{-2}+b(i+1)}{3bm}\Big)}
{\Big(1+\frac{y_{-1}+b(i+1)}{3bm}\Big)}\frac{\Big(1+\frac{y_{-1}+b(i-1)}{3bm}\Big)}
{\Big(1+\frac{y_{-2}+bi}{3bm}\Big)}
=1-\frac1{3m}+O\Big(\frac1{m^2}\Big).\label{f22a}
\end{equation}

From \eqref{f22} and \eqref{f22a}, we have that the products in
\eqref{l16}, \eqref{l17} are equiconvergent with the product
$$
\prod_{j=1}^n\Big(1-\frac1{3j}+O\Big(\frac1{j^2}\Big)\Big),
$$
that is, with the  sequence
\begin{equation}
\exp\Big(\sum_{j=1}^n\ln\Big(1-\frac1{3j}+O\Big(\frac1{j^2}\Big)\Big)\Big)=
\exp\Big(-\frac13\sum_{j=1}^n\Big(\frac1{j}+O\Big(\frac1{j^2}\Big)\Big)\Big).
\label{ps10}
\end{equation}
From \eqref{ps10}, and the fact that
$\lim_{n\to\infty}\sum_{j=1}^n\frac1j=+\infty$,
the statement  follows.

\subsection*{Case $a=-1$, $b\ne 0$}
 Here we describe long-term behavior of well-defined solutions of \eqref{e1c} 
for the case $a=-1$, $b\ne 0$, by using the next two formulas 
$$
x_{6m+2i}=x_{2i-6}\prod_{s=0}^m\frac{(2y_{-1}-b)(-1)^{3s+i}+b}
{(2y_{-2}-b)(-1)^{3s+i+1}+b}\cdot\frac{(2y_{-2}-b)(-1)^{3s+i-1}+b}
{(2y_{-1}-b)(-1)^{3s+i-1}+b}
$$ 
and 
$$
x_{6m+2i+1}=x_{2i-5}\prod_{s=0}^m\frac{(2y_{-2}-b)(-1)^{3s+i+1}+b}
{(2y_{-1}-b)(-1)^{3s+i+1}+b}\cdot\frac{(2y_{-1}-b)(-1)^{3s+i-1}+b}
{(2y_{-2}-b)(-1)^{3s+i}+b},
$$ 
for $m\ge -1$ and $i\in\{1,2,3\}$,
which are obtained from \eqref{l14} and \eqref{l15} with $a=-1$.

Employing these formulas we obtain
\begin{align}
&\begin{aligned}
x_{12m+2i}
&=x_{2i-6}\prod_{s=0}^{2m}\frac{(2y_{-1}-b)(-1)^{3s+i}+b}
{(2y_{-1}-b)(-1)^{3s+i-1}+b}\\
&=x_{2i-6}\frac{(2y_{-1}-b)(-1)^i+b}
{(2y_{-1}-b)(-1)^{i-1}+b}\prod_{s=0}^{m-1}\frac{b^2-(2y_{-1}-b)^2}
{b^2-(2y_{-1}-b)^2}\\
&=x_{2i-6}\frac{(2y_{-1}-b)(-1)^i+b}
{(2y_{-1}-b)(-1)^{i-1}+b},
\end{aligned}\label{c1}\\
&\begin{aligned}
x_{12m+6+2i}&=x_{2i-6}\prod_{s=0}^{2m+1}\frac{(2y_{-1}-b)(-1)^{3s+i}+b}
{(2y_{-1}-b)(-1)^{3s+i-1}+b}\\
&=x_{2i-6}\prod_{s=0}^m\frac{b^2-(2y_{-1}-b)^2}
{b^2-(2y_{-1}-b)^2}\\
&=x_{2i-6},
\end{aligned}\label{c2}\\
&\begin{aligned}
x_{12m+2i+1}&=x_{2i-5}\prod_{s=0}^{2m}\frac{(2y_{-2}-b)(-1)^{3s+i+1}+b}
{(2y_{-2}-b)(-1)^{3s+i}+b}\\
&=x_{2i-5}\frac{(2y_{-2}-b)(-1)^{i+1}+b}{(2y_{-2}-b)(-1)^{i}+b}\prod_{s=0}^{m-1}\frac{b^2-(2y_{-2}-b)^2}
{b^2-(2y_{-2}-b)^2}\\
&=x_{2i-5}\frac{(2y_{-2}-b)(-1)^{i+1}+b}{(2y_{-2}-b)(-1)^{i}+b},
\end{aligned}\label{c3}\\
&\begin{aligned}
x_{12m+6+2i+1}&=x_{2i-5}\prod_{s=0}^{2m+1}\frac{(2y_{-2}-b)(-1)^{3s+i+1}+b}
{(2y_{-2}-b)(-1)^{3s+i}+b}\\
&=x_{2i-5}\prod_{s=0}^m\frac{b^2-(2y_{-2}-b)^2}
{b^2-(2y_{-2}-b)^2}\\
&=x_{2i-5}
\end{aligned}\label{c5}
\end{align} 
for $m\ge -1$ and $i\in\{1,2,3\}$.

From \eqref{c1}-\eqref{c5} the following theorem follows.
\end{proof}


\begin{theorem} \label{thm3}
 Assume that $a=-1$, $b\ne 0$. Then
every well-defined solution $(x_n)_{n\ge -4}$ of equation
\eqref{e1c} is twelve-periodic and is given by formulas
\eqref{c1}-\eqref{c5}.
\end{theorem}

The twelve-periodicity of every well-defined solution 
$(x_n)_{n\ge -4}$ of equation \eqref{e1c} in the case $a=-1$, $b\ne 0$, can be
proved also without calculations in the following way. First note
that the sequence
$$
y_n=\frac1{x_nx_{n-1}x_{n-2}},\quad n\ge -2,
$$
satisfies the  recurrence relation
$$
y_n=b-y_{n-2},\quad n\in\mathbb{N}_0,
$$
from which it follows that
$$
y_n=y_{n-4},\quad n\ge 2;
$$
that is, sequence $(y_n)_{n\ge-2}$ is four-periodic, and
consequently the sequence $u_n=1/y_n$, $n\ge-2$, is also four-periodic.
Further, we have 
\begin{equation}
x_n=\frac{u_n}{x_{n-1}x_{n-2}}=\frac{u_n}{u_{n-1}}x_{n-3},\quad
n\ge -1.\label{c6}
\end{equation}
By using relation \eqref{c6} four times,
we obtain
$$
x_n=\frac{u_n}{u_{n-1}}\frac{u_{n-3}}{u_{n-4}}\frac{u_{n-6}}{u_{n-7}}\frac{u_{n-9}}{u_{n-10}}x_{n-12},\quad
n\ge 8.
$$
This along with four-periodicity of $(u_n)_{n\ge-2}$
implies twelve-periodicity of $(x_n)_{n\ge -4}$.

\subsection*{Case $a\ne 0$, $b=0$}  
If $a\ne 0$ and $b=0$ then equation \eqref{e1c}
becomes 
$$
x_n=\frac{x_{n-3}x_{n-4}}{x_{n-1}a},\quad n\in\mathbb{N}_0,
$$
and formulas \eqref{l14}-\eqref{l17} also hold, from which we obtain
$$
x_{6m+2i}=\frac{x_{2i-6}}{a^{m+1}},
$$ 
for $m\ge -1$, $i\in\{1,2,3\}$, and
$$
x_{6m+2i+1}=\frac{x_{2i-5}}{a^{m+1}},
$$ 
for $m\ge -1$, $i\in\{1,2,3\}$, which means that
\begin{equation}
x_{6m+j}=\frac{x_{j-6}}{a^{m+1}},\label{g3}
\end{equation}
for every $m\ge -1$ and $j=\overline{2,7}$.
Using \eqref{g3} we obtain the following theorem.

\begin{theorem} \label{thm4} 
 Assume that $a\ne 0$, $b=0$, and
$(x_n)_{n\ge -4}$ is a well-defined solution of equation
\eqref{e1c}.  Then the following statements are true.
\begin{itemize}

\item[(a)] If $|a|>1$, then $x_n\to 0$ as $n\to+\infty$.

\item[(b)] If $|a|<1$, then $|x_n|\to \infty$ as $n\to+\infty$.

\item[(c)] If $a=1$, then the sequence $(x_n)_{n\ge -4}$ is
six-periodic.

\item[(d)] If $a=-1$, then the sequence $(x_n)_{n\ge -4}$ is
twelve-periodic.
\end{itemize}
\end{theorem}


\section{Domain of undefinable solutions for \eqref{e1}}

We have already shown that solutions of equation \eqref{e1} are
not defined if $x_{-j}=0$ for some $j\in\{1,2,3,4\}$. A natural
problem is to describe the set of all initial values for which
solutions to equation \eqref{e1} are not defined.

\begin{definition}[\cite{amc219-dussde}] \label{def1}\rm
Consider the difference equation 
\begin{equation}
x_n=f(x_{n-1},\ldots,x_{n-s},n),\quad
n\in\mathbb{N}_0,\label{b1a}
\end{equation}
where $s\in\mathbb{N}$, and $x_{-i}\in\mathbb{R}$,
$i=\overline{1,s}$. The string of numbers
$x_{-s},\ldots,x_{-1},x_0,\ldots,x_{n_0}$ where $n_0\ge -1$, is
called an {\it undefined solution} of equation \eqref{b1a} if
$$
x_j=f(x_{j-1},\ldots,x_{j-s},j)
$$
for $0\le j<n_0+1$, and $x_{n_0+1}$ is not a defined number; that
is, the quantity $f(x_{n_0},\ldots,x_{n_0-s+1},n_0+1)$ is not
defined.
\end{definition}

The set of all initial values $x_{-s},\ldots,x_{-1}$ which
generate undefined solutions of equation \eqref{b1a} is called
{\it domain of undefinable solutions} of the equation.
This domain is characterized  in the next theorem for the case
 $a_n\ne 0$, $b_n\ne 0$, $n\in\mathbb{N}_0$. 


\begin{theorem} \label{thm5} 
 Assume that $a_n\ne 0$, $b_n\ne 0$,
$n\in\mathbb{N}_0$. Then the domain of undefinable solutions of equation
\eqref{e1} is the  set 
\begin{equation}
\begin{aligned}
\mathcal{U}
&= \cup_{m\in\mathbb{N}_0}\cup_{i=0}^1
\Big\{(x_{-4},\ldots,x_{-1})\in\mathbb{R}^4:x_{i-2}x_{i-3}x_{i-4}=
\frac1{c_m},\text{ where } \\
&\quad c_m:=-\sum_{j=0}^m\frac{b_{2j+i}}{a_{2j+i}}\prod_{l=0}^{j-1}
\frac1{a_{2l+i}}\ne 0\Big\}
\cup\cup_{j=1}^4\Big\{(x_{-4},\ldots,x_{-1})\in\mathbb{R}^4:
x_{-j}=0\Big\}.
\end{aligned}\label{c8}
\end{equation}
\end{theorem}

\begin{proof}
The considerations at the beginning of Section 2 show
that the domain of undefinable solutions of equation \eqref{e1}
contains the set
$$
\cup_{j=1}^4\big\{(x_{-4},\ldots,x_{-1})\in\mathbb{R}^4:
x_{-j}=0\big\}.
$$
Now assume $x_{-j}\ne 0$, $j=\overline{1,4}$ (i.e. $x_n\ne 0$ for every
$n\ge -4$). If a solution $(x_n)_{n\ge -4}$ with such initial
values is not defined then it must be 
\begin{equation}
x_{n-2}x_{n-3}x_{n-4}=-\frac{a_n}{b_n}\label{nds}
\end{equation}
for some
$n\in\mathbb{N}_0$ (here we use the condition $b_n\ne 0$, $n\in\mathbb{N}_0$).

Now recall that the change of variables \eqref{a3} implies that
equation \eqref{e1} is equivalent to the equations in \eqref{l2}.
Hence, this along with \eqref{nds} implies that solution
$(x_n)_{n\ge -4}$ is not defined if
$$
y_{2(m-1)+i}=-\frac{b_{2m+i}}{a_{2m+i}}
$$ 
for some $m\in\mathbb{N}_0$ and $i\in\{0,1\}$.
Set
$$
f_{2m+i}(t):=a_{2m+i}t+b_{2m+i},\quad m\in\mathbb{N}_0,\;i\in\{0,1\}.
$$
Then $f_{2m+i}^{-1}(t)=(t-b_{2m+i})/a_{2m+i}$, $m\in\mathbb{N}_0$,
$i\in\{0,1\}$, and specially  
\begin{equation}
f_{2m+i}^{-1}(0)=-\frac{b_{2m+i}}{a_{2m+i}},\quad
m\in\mathbb{N}_0,\;i\in\{0,1\}.\label{b3a}
\end{equation}
Now write equations in
\eqref{l2} as
$$
y_{2m+i}=f_{2m+i}(y_{2(m-1)+i}),\quad m\in\mathbb{N}_0,
$$
for $i\in\{0,1\}$.
Then, we have
\begin{equation}
 y_{2m+i}=f_{2m+i}\circ f_{2(m-1)+i}\circ
\cdots \circ f_i(y_{i-2}),\quad
m\in\mathbb{N}_0,\;i\in\{0,1\}.\label{b5a}
\end{equation}
Equalities \eqref{b3a} and \eqref{b5a} imply that
$$
y_{2(m-1)+i}=-\frac{b_{2m+i}}{a_{2m+i}}
$$
for some $m\in\mathbb{N}_0$, $i\in\{0,1\}$,
if and only if
\begin{equation}
y_{i-2}= f^{-1}_i\circ\cdots\circ f^{-1}_{2m+i}(0).\label{b6a}
\end{equation}
From \eqref{b6a} we obtain
$$
y_{i-2}=-\sum_{j=0}^m\frac{b_{2j+i}}{a_{2j+i}}\prod_{l=0}^{j-1}\frac1{a_{2l+i}},
$$
for some $m\in\mathbb{N}_0$ and $i\in\{0,1\}$, which along with the
relations
$$
y_{i-2}=\frac1{x_{i-2}x_{i-3}x_{i-4}},\quad i\in\{0,1\},
$$
 implies that the first union
in \eqref{c8} belongs to the domain of undefinable solutions and
consequently the result.
\end{proof}


\subsection*{Acknowledgements} 
The second author was supported by the Operational Programme Research and
Development for Innovations, No. CZ.1.05/2.1.00/03.0097 (AdMaS).
The fourth author was supported by Project no. FEKT-S-14-2200 of
Faculty of Electrical Engineering and Communication, Brno
University of Technology, Czech Republic. This paper is also
supported by the Serbian Ministry of Science projects III 41025,
III 44006 and OI 171007.


\begin{thebibliography}{00}

\bibitem{al0} M.~Aloqeili;
 Dynamics of a $k$th order rational difference
equation, {\it Appl. Math. Comput.} \textbf{181} (2006), 1328-1335.

\bibitem{al1} M.~Aloqeili;
 Dynamics of a rational difference equation, {\it Appl. Math. Comput.} 
\textbf{176} (2006), 768-774.

\bibitem{am2} A.~Andruch-Sobilo, M.~Migda;
 On the rational recursive sequence
$x_{n+1}=ax_{n-1}/(b+cx_nx_{n-1})$, {\it Tatra Mt. Math. Publ.}
\textbf{43} (2009), 1-9.

\bibitem{bl} I.~Bajo, E.~Liz;
 Global behaviour of a second-order nonlinear
difference equation, {\it J. Differ. Equations Appl.} \textbf{17}
(10) (2011), 1471-1486.

\bibitem{jdea129} K.~Berenhaut, S.~Stevi\' c;
 The behaviour of the positive solutions of the difference equation 
$x_n=A+(x_{n-2}/x_{n-1})^p$,
{\it J. Difference Equ. Appl.} \textbf{12} (9) (2006), 909-918.

\bibitem{bb} L.~Berezansky, E.~Braverman;
On impulsive Beverton-Holt difference equations and their applications, 
{\it J. Differ. Equations Appl.} \textbf{10} (9) (2004), 851-868.

\bibitem{jdea128} L.~Berg, S.~Stevi\' c;
 Periodicity of some classes of holomorphic difference equations,
{\it J. Difference Equ. Appl.} \textbf{12} (8) (2006), 827-835.

\bibitem{amc218-sde} L.~Berg, S.~Stevi\' c;
 On some systems of difference equations, {\it Appl. Math. Comput.} 
\textbf{218} (2011), 1713-1718.

\bibitem{lb} L.~Brand;
 A sequence defined by a difference equation, 
{\it Amer. Math. Monthly} \textbf{62} (7) (1955), 489-492.

\bibitem{c} C.~Cinar;
 On the positive solutions of difference equation, 
{\it Appl. Math. Comput.} \textbf{150} (1) (2004), 21-24.

\bibitem{dcdis4} B.~Iri\v canin, S.~Stevi\' c;
 Some systems of nonlinear difference equations of
higher order with periodic solutions, {\it Dyn. Contin. Discrete
Impuls. Syst. Ser. A Math. Anal.} \textbf{13 a} (3-4) (2006),
499-506.

\bibitem{amc215-854} B.~Iri\v canin, S.~Stevi\' c;
Eventually constant solutions of a rational difference
equation, {\it Appl. Math. Comput.}  \textbf{215} (2009), 854-856.

\bibitem {k1} G.~L.~Karakostas;
 Asymptotic 2-periodic difference equations with diagonally
self-invertible responces, {\it J. Differ. Equations Appl.} 
\textbf{6} (2000), 329-335.

\bibitem{kk} C.~M.~Kent, W.~Kosmala;
 On the nature of solutions of the difference equation 
$x_{n+1}=x_nx_{n-3}-1$, {\it Int. J. Nonlinear
Anal. Appl.} \textbf{2} (2) (2011), 24-43

\bibitem{kkn} C.~M.~Kent, M.~Kustesky M, A.~Q.~Nguyen, B.~V.~Nguyen;
 Eventually periodic solutions of
$x_{n+1}=\max\{A_n/x_n, B_n/x_{n-1}\}$ when the parameters are two
cycles, {\it Dyn. Contin. Discrete Impuls. Syst. Ser. A Math.
Anal.} \textbf{10} (1-3) (2003), 33-49.

\bibitem{wk} W.~Kosmala;
 A period 5 difference equation, {\it Int. J. Nonlinear Anal. Appl.} 
\textbf{2} (1) (2011), 82-84.

\bibitem{kg} R.~Kurshan, B.~Gopinath;
Recursively generated periodic sequences, {\it Canad. J. Math.} 
\textbf{24} (6) (1974), 1356-1371.

\bibitem{ll} H.~Levy, F.~Lessman;
 {\it Finite Difference Equations}, Dover Publications, Inc., New York, 1992.

\bibitem{ps4} G.~Papaschinopoulos, C.~J.~Schinas;
 Invariants and oscillation for systems of two nonlinear difference
equations, {\it Nonlinear Anal. TMA} \textbf{46} (7) (2001), 967-978.

\bibitem{pss1} G. Papaschinopoulos, C.~J.~Schinas, G. Stefanidou;
On a $k$-order system of Lyness-type difference
equations, {\it Adv. Differ. Equat.} Volume 2007, Article ID
31272, (2007), 13 pages.

\bibitem{pst2} G. Papaschinopoulos, G. Stefanidou;
Asymptotic behavior of the solutions of a class of rational difference equations, 
{\it Inter. J. Difference Equations} \textbf{5} (2) (2010), 233-249.

\bibitem{dcdis1} S.~Stevi\'c;
 On the recursive sequence  $x_{n+1}=\alpha_n+(x_{n-1}/x_n)$ II, 
{\it Dyn. Contin. Discrete Impuls. Syst.} \textbf{10a} (6) (2003), 911-916.

\bibitem{tjm2} S.~Stevi\'c;
On the recursive sequence 
$x_{n+1}=A/\prod_{i=0}^k x_{n-i}+1/\prod_{j=k+2}^{2(k+1)}x_{n-j}$,
{\it Taiwanese J. Math.} \textbf{7} (2) (2003), 249-259.

\bibitem{amen} S.~Stevi\'c;
 More on a rational recurrence relation, {\it Appl. Math. E-Notes}
\textbf{4} (2004), 80-85.

\bibitem{37264} S.~Stevi\'c;
 A short proof of the Cushing-Henson conjecture, 
{\it Discrete Dyn. Nat. Soc.} Vol. 2006, Article ID
37264,  (2006), 5 pages.

\bibitem{um83-2} S.~Stevi\'c;
 Periodicity of max difference equations,
 {\it Util. Math.} \textbf{83} (2010), 69-71.

\bibitem{jmaa376} S.~Stevi\'c;
 On a nonlinear generalized max-type difference equation,
 {\it J. Math. Anal. Appl.} \textbf{376} (2011), 317-328.

\bibitem{amc218-sys1} S.~Stevi\'c;
 On a system of difference equations, {\it Appl. Math. Comput.} \textbf{218} (2011),
3372-3378.

\bibitem{amc218-dcept} S.~Stevi\'c;
 On the difference equation $x_n=x_{n-2}/(b_n+c_nx_{n-1}x_{n-2})$, 
{\it Appl. Math. Comput.} \textbf{218} (2011), 4507-4513.

\bibitem{amc-thos} S.~Stevi\'c;
 On a third-order system of difference equations, 
{\it Appl. Math. Comput.} \textbf{218} (2012), 7649-7654.

\bibitem{amc-218solsys} S.~Stevi\'c;
 On some solvable systems of difference equations, {\it Appl. Math. Comput.} 
\textbf{218} (2012), 5010-5018.

\bibitem{amc218-hde} S.~Stevi\'c;
 On the difference equation $x_n=x_{n-k}/(b+cx_{n-1}\cdots x_{n-k})$,
{\it Appl. Math. Comput.} \textbf{218} (2012), 6291-6296.

\bibitem{amc-mtsde} S.~Stevi\'c;
 Solutions of a max-type system of difference
equations, {\it Appl. Math. Comput.} \textbf{218} (2012), 9825-9830.


\bibitem{amc219-dussde} S.~Stevi\'c;
 Domains of undefinable solutions
of some equations and systems of difference equations, {\it Appl.
Math. Comput.} \textbf{219} (2013), 11206-11213.

\bibitem{amc219-ssfo} S.~Stevi\'c;
 On a solvable system of difference equations of fourth order, 
{\it Appl. Math. Comput.}
\textbf{219} (2013), 5706-5716.

\bibitem{amc219-ssdek} S.~Stevi\'c;
 On a solvable system of difference equations of $k$th
order, {\it Appl. Math. Comput.} \textbf{219} (2013), 7765-7771.

\bibitem{amc219-sdeoos} S.~Stevi\'c;
 On a system of difference equations of odd order solvable in closed form, {\it
Appl. Math. Comput.} \textbf{219} (2013) 8222-8230.


\bibitem{amc-219sscf} S.~Stevi\'c;
 On a system of difference equations which can be solved in closed form, 
{\it Appl. Math. Comput.} \textbf{219} (2013), 9223-9228.


\bibitem{amc219-sstho} S.~Stevi\'c;
 On the system of difference equations 
$x_n=c_ny_{n-3}/(a_n+b_ny_{n-1}x_{n-2}y_{n-3})$,
$y_n=\gamma_n x_{n-3}/(\alpha_n+\beta_n x_{n-1}y_{n-2}x_{n-3})$, {\it Appl.
Math. Comput.} \textbf{219} (2013), 4755-4764.

\bibitem{amc219-ssktho} S.~Stevi\'c;
 On the system $x_{n+1}=y_nx_{n-k}/(y_{n-k+1}(a_n+b_ny_nx_{n-k}))$,
$y_{n+1}=x_ny_{n-k}/(x_{n-k+1}(c_n+d_nx_ny_{n-k}))$, {\it Appl.
Math. Comput.} \textbf{219} (2013), 4526-4534.

\bibitem{ejqtde1} S.~Stevi\'c, M.~A.~Alghamdi, A.~Alotaibi, N.~Shahzad;
 On a higher-order system of difference equations, 
{\it Electron. J. Qual. Theory Differ. Equ.} Article
No. 47, (2013), pages 1-18.

\bibitem{157943} S.~Stevi\'c, M.~A.~Alghamdi, D.~A.~Maturi, N.~Shahzad;
 On a class of solvable difference equations, {\it
Abstr. Appl. Anal.} Vol. 2013, Article ID 157943, (2013), 7 pages.

\bibitem{ejqtde-maxsde} S.~Stevi\'c M.~A.~Alghamdi, A.~Alotaibi, N.~Shahzad;
Boundedness character of a max-type system of difference equations of second 
order, {\it Electron. J. Qual. Theory Differ. Equ.} Article No.
45, (2014), pages 1-12.

\bibitem{508523} S.~Stevi\'c, J.~Dibl\'{\i}k, B.~Iri\v{c}anin, Z.~\v Smarda;
 On a third-order system of difference equations with variable
coefficients, {\it Abstr. Appl. Anal.} vol. 2012, Article ID
508523, (2012), 22 pages.

\bibitem{541761} S.~Stevi\'c, J.~Dibl\'{\i}k, B.~Iri\v{c}anin, Z.~\v Smarda;
 On some solvable difference equations and systems of difference equations, {\it
Abstr. Appl. Anal.} Vol. 2012, Article ID 541761, (2012), 11
pages.

\bibitem{409237} S.~Stevi\'c, J.~Dibl\'{\i}k, B.~Iri\v{c}anin, Z.~\v Smarda;
 On the difference equation $x_n=a_nx_{n-k}/(b_n+c_nx_{n-1}\cdots x_{n-k})$,
{\it Abstr. Appl. Anal.} Vol. 2012, Article ID 409237, (2012), 19
pages.

\bibitem{108047} S.~Stevi\'c, J.~Dibl\'{\i}k, B.~Iri\v{c}anin, Z.~\v{S}marda;
 On the difference equation
$x_{n+1}=x_nx_{n-k}/(x_{n-k+1}(a+bx_nx_{n-k}))$, {\it Abstr. Appl.
Anal.} Vol. 2012, Article ID 108047, (2012), 9 pages.

\bibitem{jdea205} S.~Stevi\'c, J.~Dibl\'{\i}k, B.~Iri\v{c}anin, Z.~\v{S}marda;
 On a solvable system of rational difference
equations, {\it J. Difference Equ. Appl.} \textbf{20} (5-6) (2014),
811-825.

\bibitem{tyt2} D.~T.~Tollu, Y.~Yazlik, N.~Taskara;
 On fourteen solvable systems of difference equations, 
{\it Appl. Math. Comput.} \textbf{233} (2014), 310-319.

\end{thebibliography}

\end{document}
