\documentclass[reqno]{amsart}
\usepackage{hyperref}
\AtBeginDocument{{\noindent\small
\emph{Electronic Journal of Differential Equations},
Vol. 2014 (2014), No. 35, pp. 1--13.\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-2014/35\hfil Existence and multiplicity of solutions]
{Existence and multiplicity of solutions for a discrete nonlinear
boundary value problem}
\author[G. A. Afrouzi, A. Hadjian \hfil EJDE-2014/35\hfilneg]
{Ghasem A. Afrouzi, Armin Hadjian} % in alphabetical order
\address{Ghasem A. Afrouzi \newline
Department of Mathematics, Faculty of Mathematical Sciences,
University of Mazandaran, Babolsar, Iran}
\email{afrouzi@umz.ac.ir}
\address{Armin Hadjian \newline
Department of Mathematics, Faculty of Mathematical Sciences,
University of Mazandaran, Babolsar, Iran}
\email{a.hadjian@umz.ac.ir}
\thanks{Submitted November 12, 2013. Published January 29, 2014.}
\subjclass[2000]{39A05, 34B15}
\keywords{Discrete nonlinear boundary value problem; $p$-Laplacian;
\hfill\break\indent multiple solutions; critical point theory}
\begin{abstract}
In this article, we show the existence and multiplicity of positive
solutions for a discrete nonlinear boundary value problem involving the
$p$-Laplacian. Our approach is based on critical point theorems in
finite dimensional Banach spaces.
\end{abstract}
\maketitle
\numberwithin{equation}{section}
\newtheorem{theorem}{Theorem}[section]
\newtheorem{lemma}[theorem]{Lemma}
\newtheorem{corollary}[theorem]{Corollary}
\newtheorem{example}[theorem]{Example}
\newtheorem{remark}[theorem]{Remark}
\allowdisplaybreaks
\section{Introduction}
It is well known that in fields of research, such as
computer science, mechanical engineering, control systems,
artificial or biological neural networks, economics and many others,
the mathematical modelling of important questions leads naturally to
the consideration of nonlinear difference equations. For this
reason, in recent years, many authors have widely developed various
methods and techniques, such as fixed point theorems, upper and
lower solutions, and Brouwer degree, to study discrete problems
(see, e.g.,
\cite{AndRachTis,BerMaw,ChuJi,HendThom,JiChuOrAg,RachTis,TiGe} and
references therein). Recently, also the critical point theory has
aroused the attention of many authors in the study of these problems
(see, e.g.,
\cite{AgPerOr1,AgPerOr2,BiSunZha,BonaCand2,CanMoli,JiZh,KriMihRad}).
Let $N$ be a positive integer, denote with $[1,N]$ the discrete
interval $\{1,\ldots,N\}$ and consider the problem
\begin{equation}
\label{e1.1} % \tag{$P_{\lambda}^{f,q}$}
\begin{gathered}
-\Delta(\phi_p(\Delta u_{k-1}))+q_k \phi_p(u_k)
=\lambda f(k,u_k),\quad k\in[1,N],\\
u_0=u_{N+1}=0,
\end{gathered}
\end{equation}
where, $f:[1,N]\times\mathbb{R}\to\mathbb{R}$ is a
continuous function, $\Delta u_{k-1}:=u_k-u_{k-1}$ is the forward
difference operator, $q_k\geq 0$ for all $k\in[1,N]$,
$\phi_p(s):=|s|^{p-2}s$, $1
2$ and $R>0$ such that,
for all $|\xi|\geq R$, one has
$$
0<\nu G(\xi)\leq\xi g(\xi).
$$
\end{enumerate}
Then, for each
$$
\lambda\in\big]0,\frac{2}{N(N+1)}\sup_{\gamma>0}
\frac{\gamma^2}{\max_{|\xi|\leq\gamma}G(\xi)}\big[,
$$
the problem
\[
\begin{gathered}
-\Delta^2 u_{k-1}+q_ku_k=\lambda g(u_k),\quad k\in[1,N],\\
u_0=u_{N+1}=0,
\end{gathered}
\]
admits at least two non-trivial solutions.
\end{theorem}
Instead, Theorem \ref{the4.3} gives the following theorem.
\begin{theorem}\label{the1.2}
Let $g:\mathbb{R}\to\mathbb{R}$ be a non-negative continuous
function such that
\begin{gather*}
\lim_{\xi\to
0^+}\frac{g(\xi)}{\xi}=+\infty,\quad\lim_{\xi\to +\infty}\frac{g(\xi)}{\xi}=0,\\
\int_0^1 g(x)dx<\frac{1}{2(N+1)}\int_0^2 g(x)dx.
\end{gather*}
Then, for each
$$
\lambda\in\frac{1}{N}\big]\frac{4}{\int_0^2
g(x)dx},\frac{2}{(N+1)\int_0^1 g(x)dx}\big[,
$$
the problem
\begin{gather*}
-\Delta^2 u_{k-1}=\lambda g(u_k),\quad k\in[1,N],\\
u_0=u_{N+1}=0,
\end{gather*}
admits at least three non-negative solutions.
\end{theorem}
\section{Preliminaries}
Our main tools are Theorems \ref{the2.1} and \ref{the2.2},
consequences of the existence result of a local minimum theorem
\cite[Theorem 3.1]{Bonanno} which is inspired by the Ricceri
Variational Principle \cite{Ricceri}.
For a given non-empty set $X$, and two functionals
$\Phi,\Psi:X\to\mathbb{R}$, we define the following functions
\begin{gather*}
\beta(r_1,r_2):=\inf_{v\in \Phi^{-1}(]r_1,r_2[)}\frac{\sup_{u\in
\Phi^{-1}(]r_1,r_2[)}\Psi(u)-\Psi(v)}{r_2-\Phi(v)},
\\
\rho_2(r_1,r_2):=\sup_{v\in \Phi^{-1}(]r_1,r_2[)}
\frac{\Psi(v)-\sup_{u\in
\Phi^{-1}(]-\infty,r_1])}\Psi(u)}{\Phi(v)-r_1},
\end{gather*}
for all $r_1,r_2\in\mathbb{R}$, with $r_10,
\end{equation}
and for each $\lambda>\frac{1}{\rho(r)}$ the function
$I_{\lambda}:=\Phi-\lambda \Psi$ is coercive.
Then, for each $\lambda>\frac{1}{\rho(r)}$ there is
$u_{0,\lambda}\in\Phi^{-1}(]r,+\infty[)$ such that
$I_{\lambda}(u_{0,\lambda})\leq I_\lambda(u)$ for all $u\in
\Phi^{-1}(]r,+\infty[)$ and $I'_\lambda(u_{0,\lambda})=0$.
\end{theorem}
\begin{remark}\label{rem2.3}\rm
It is worth noticing that whenever $X$ is a finite dimensional
Banach space, a careful reading of the proofs of Theorems
\ref{the2.1} and \ref{the2.2} shows that regarding to the regularity
of the derivative of $\Phi$ and $\Psi$, it is enough to require only
that $\Phi'$ and $\Psi'$ are two continuous functionals on
$X^\ast$.
\end{remark}
Now, consider the $N$-dimensional Banach space
$$
S:=\{u:[0,N+1]\to\mathbb{R} : u_0=u_{N+1}=0\}
$$
endowed with the norm
$$
\|u\|:=\Big(\sum_{k=1}^{N+1}|\Delta u_{k-1}|^p+q_k|u_k|^p\Big)^{1/p}.
$$
In the sequel, we will use the inequality
\begin{equation}\label{e2.3}
\max_{k\in[1,N]}|u_k|\leq\frac{(N+1)^{(p-1)/p}}{2}\|u\|
\end{equation}
for every $u\in S$. It immediately follows, for instance, from
\cite[Lemma 2.2]{JiZh}.
Put
\begin{equation}\label{e2.4}
\Phi(u):=\frac{\|u\|^p}{p},\quad
\Psi(u):=\sum_{k=1}^N F(k,u_k),\quad
I_\lambda(u):=\Phi(u)-\lambda\Psi(u)
\end{equation}
for every $u\in S$, where $ F(k,t):=\int_0^t
f(k,\xi)d\xi$ for every $(k,t)\in[1,N]\times\mathbb{R}$.
Standard arguments show that $I_\lambda\in C^1(S,\mathbb{R})$ as
well as that critical points of $I_\lambda$ are exactly the
solutions of problem \eqref{e1.1}. In fact, one has
\begin{align*}
I_\lambda'(u)(v)
&=\sum_{k=1}^{N+1}[\phi_p(\Delta u_{k-1})\Delta
v_{k-1}+q_k|u_k|^{p-2}u_kv_k-\lambda f(k,u_k)v_k]\\
&=-\sum_{k=1}^N[\Delta(\phi_p(\Delta
u_{k-1}))v_k-q_k|u_k|^{p-2}u_kv_k+\lambda f(k,u_k)v_k]
\end{align*}
for all $u,v\in S$ (see \cite{JiZh} for more details).
Finally, for the reader's convenience we recall
\cite[Theorems 2.2 and 2.3]{BonaCand2} in order to get positive
solutions to problem \eqref{e1.1}, i.e. $u_k>0$ for all $k\in[1,N]$.
\begin{lemma}\label{lem2.4}
Let $u\in S$ and assume that one of the following conditions holds:
\begin{itemize}
\item[(A1)] $-\Delta(\phi_p(\Delta u_{k-1}))+q_k \phi_p(u_k)\geq 0$
for all $k\in[1,N]$;
\item[(A2)] if $u_k\leq 0$, then $-\Delta(\phi_p(\Delta u_{k-1}))
+q_k \phi_p(u_k)=0$.
\end{itemize}
Then, either $u>0$ in $[1,N]$ or $u\equiv0$.
\end{lemma}
\begin{remark}\label{rem2.5}\rm
If $f:[1,N]\times\mathbb{R}\to\mathbb{R}$ is a non-negative
function, then, owing to Lemma \ref{lem2.4} part (A1), all
solutions of problem ð\eqref{e1.1} are either zero or positive.
Now, let $f:[1,N]\times\mathbb{R}\to\mathbb{R}$ be such that
$f(k,0)=0$ for all $k\in[0,N]$. Put
\[
f^\ast(k,t):=\begin{cases}
f(k,t), & \text{if } t>0,\\
0, & \text{if } t\leq 0.
\end{cases}
\]
Clearly, $f^\ast$ is a continuous function. Owing to Lemma
\ref{lem2.4} part (A2), all solutions of problem
$(P_\lambda^{f^\ast,q})$ are either zero or positive, and hence are
also solutions for \eqref{e1.1}. Hence, we emphasize that when
$(P_\lambda^{f^\ast,q})$ admits non-trivial solutions, then problem
\eqref{e1.1} admits positive solutions, independently of the sign of
$f$.
\end{remark}
\section{Main results}
In this section we establish an existence result of at least one solution,
Theorem \ref{the3.1}, which is based on Theorem \ref{the2.1}, and we point
out some consequences, Theorems \ref{the3.2}, \ref{the3.3} and
Corollary \ref{cor3.6}. Finally, we present an other existence
result of at least one solution, Theorem \ref{the3.7}, which is
based in turn on Theorem \ref{the2.2}.
Put $Q:=\sum_{k=1}^N q_k$. For every two non-negative constants
$\gamma,\delta$, with
$$
(2\gamma)^p\neq(2+Q)(N+1)^{p-1}\delta^p,
$$
we set
$$
a_\gamma(\delta):=\frac{\sum_{k=1}^N\max_{|t|\leq\gamma}F(k,t)-\sum_{k=1}^N
F(k,\delta)}{(2\gamma)^p-(2+Q)(N+1)^{p-1}\delta^p}.
$$
\begin{theorem}\label{the3.1}
Assume that there exist three real constants $\gamma_1,\gamma_2$ and
$\delta$, with
\begin{equation}\label{e3.1}
0\leq\gamma_1<\frac{(2+Q)^{1/p}(N+1)^{(p-1)/p}}{2}\delta<\gamma_2,
\end{equation}
such that
\begin{equation}\label{e3.2}
a_{\gamma_2}(\delta)0$, set
$$
\lambda_\gamma^\star:=\frac{2^p}{p(N+1)^{p-1}}
\frac{\gamma^p}{\sum_{k=1}^N\max_{|t|\leq\gamma} F(k,t)}.
$$
Then, for every $\lambda\in]0,\lambda_\gamma^\star[$, problem
\eqref{e1.1} admits at least one non-trivial solution $\bar{u}\in
S$, such that $|\bar{u}_k|<\gamma$ for all $k\in[1,N]$.
\end{theorem}
\begin{proof}
Fix $\gamma>0$ and $\lambda\in]0,\lambda_\gamma^\star[$. From
\eqref{e3.4} there exists a positive constant $\delta$ with
$$
\delta<\frac{2}{(2+Q)^{1/p}(N+1)^{(p-1)/p}}\gamma,
$$
such that
$$
\frac{(2+Q)\delta^p}{p\sum_{k=1}^N
F(k,\delta)}<\lambda<\frac{(2\gamma)^p}{p(N+1)^{p-1}
\sum_{k=1}^N\max_{|t|\leq\gamma}F(k,t)}.
$$
Hence, owing to Theorem \ref{the3.2}, for every
$\lambda\in]0,\lambda_\gamma^\star[$ problem \eqref{e1.1} admits at
least one non-trivial solution $\bar{u}\in S$, such that
$|\bar{u}_k|<\gamma$ for all $k\in[1,N]$. The proof is complete.
\end{proof}
\begin{remark}\label{rem3.4}\rm
We claim that under the above assumptions, the mapping
$\lambda\mapsto I_{\lambda}(\bar{u})$ is negative and strictly
decreasing in $]0,\lambda^\star_{\gamma}[$. Indeed, the restriction
of the functional $I_{\lambda}$ to $\Phi^{-1}(]0,r_2[)$, where
$r_{2}:=\frac{(2\gamma_2)^p}{p(N+1)^{p-1}}$, admits a global
minimum, which is a critical point (local minimum) of $I_{\lambda}$
in $S$. Moreover, since $w\in\Phi^{-1}(]0,r_2[)$ and
$$
\frac{\Phi(w)}{\Psi(w)}=\frac{(2+Q)\delta^p}{p\sum_{k=1}^N
F(k,\delta)}<\lambda,
$$
we have
$$
I_\lambda(\bar{u})\leq I_\lambda(w)=\Phi(w)-\lambda\Psi(w)<0.
$$
Next, we observe that
$$
I_{\lambda}(u)=\lambda\Big(\frac{\Phi(u)}{\lambda}-\Psi(u)\Big),
$$
for every $u\in S$ and fix
$0<\lambda_1<\lambda_2<\lambda^\star_{\gamma}$. Set
\begin{gather*}
m_{\lambda_1}:=\Big(\frac{\Phi(\bar{u}_1)}{\lambda_1}-\Psi(\bar{u}_1)\Big)
=\inf_{u\in \Phi^{-1}(]0,r_2[)}\Big(\frac{\Phi(u)}{\lambda_1}-\Psi(u)\Big),
\\
m_{\lambda_2}:=\Big(\frac{\Phi(\bar{u}_2)}{\lambda_2}-\Psi(\bar{u}_2)\Big)
=\inf_{u\in \Phi^{-1}(]0,r_2[)}\Big(\frac{\Phi(u)}{\lambda_2}-\Psi(u)\Big).
\end{gather*}
Clearly, as claimed before, $m_{\lambda_i}<0$ (for $i=1,2$), and
$m_{\lambda_2}\leq m_{\lambda_1}$ thanks to $\lambda_1<\lambda_2$.
Then the mapping $\lambda\mapsto I_{\lambda}(\bar{u})$ is strictly
decreasing in $]0,\lambda^\star_{\gamma}[$ owing to
$$
I_{\lambda_2}(\bar{u}_2)=\lambda_2m_{\lambda_2}\leq
\lambda_2m_{\lambda_1}<\lambda_1m_{\lambda_1}=I_{\lambda_1}(\bar{u}_1).
$$ This concludes the proof of our claim.
\end{remark}
\begin{remark}\label{rem3.5}\rm
In other words, Theorem \ref{the3.3} ensures that if the asymptotic
condition at zero \eqref{e3.4} is verified then, for every parameter
$\lambda$ belonging to the real interval $]0,\lambda^\star[$, where
$$
\lambda^\star:=\frac{2^p}{p(N+1)^{p-1}}\sup_{\gamma>0}
\frac{\gamma^p}{\sum_{k=1}^N\max_{|t|\leq\gamma}F(k,t)},
$$
problem \eqref{e1.1} admits at least one non-trivial solution
$\bar{u}\in S$.
\end{remark}
\begin{corollary}\label{cor3.6}
Let $\alpha:[1,N]\to\mathbb{R}$ be a non-negative and
non-zero function and let $g:[0,+\infty)\to\mathbb{R}$ be a
continuous function such that $g(0)=0$. Assume that there exist two
positive constants $\gamma,\delta$, with
$$
\delta<\frac{2}{(2+Q)^{1/p}(N+1)^{(p-1)/p}}\gamma,
$$
for which
\begin{equation}\label{e3.5}
\frac{\max_{0\leq t\leq\gamma}G(t)}{\gamma^p}
<\Big(\frac{2^p}{(2+Q)(N+1)^{p-1}}\Big)\frac{G(\delta)}{\delta^p},
\end{equation}
where $ G(t):=\int_0^t g(\xi)d\xi$ for all
$t\in\mathbb{R}$. Then, for each
$$
\lambda\in\frac{1}{p\sum_{k=1}^N\alpha_k}\big]\frac{(2+Q)\delta^p}{G(\delta)},
\frac{(2\gamma)^p}{(N+1)^{p-1}\max_{0\leq t\leq\gamma}G(t)}\big[,
$$
the problem
\begin{equation} \label{e3.6} %\tag{$P_{\lambda}^{\alpha g,q}$}
\begin{gathered}
-\Delta(\phi_p(\Delta u_{k-1}))+q_k \phi_p(u_k)
=\lambda \alpha_k g(u_k),\quad k\in[1,N],\\
u_0=u_{N+1}=0,
\end{gathered}
\end{equation}
admits at least one positive solution $\bar{u}\in S$, such that
$\bar{u}_k<\gamma$ for all $k\in[1,N]$.
\end{corollary}
\begin{proof}
Put
\[
f(k,t):=\begin{cases}
\alpha_k g(t), & \text{if } t\geq 0,\\
0, & \text{if } t<0,
\end{cases}
\]
for every $k\in[1,N]$ and $t\in\mathbb{R}$. The conclusion follows
from Theorem \ref{the3.2} owing to \eqref{e3.5} and taking into
account Lemma \ref{lem2.4} part (A2).
\end{proof}
Finally, we present an application of Theorem \ref{the2.2} which we
will use in next section to obtain multiple solutions.
\begin{theorem}\label{the3.7}
Assume that there exist two real constants
$\bar{\gamma},\bar{\delta}$, with
$$
0<\bar{\gamma}<\frac{(2+Q)^{1/p}(N+1)^{(p-1)/p}}{2}\bar{\delta},
$$
such that
\begin{equation}\label{e3.7}
\sum_{k=1}^N\max_{|t|\leq\bar{\gamma}}F(k,t)<\sum_{k=1}^N
F(k,\bar{\delta}),
\end{equation}
and
\begin{equation}\label{e3.8}
\limsup_{|\xi|\to +\infty}\frac{F(k,\xi)}{|\xi|^p}\leq
0\quad \text{uniformly in } k.
\end{equation}
Then, for each $\lambda>\tilde{\lambda}$, where
$$
\tilde{\lambda}:=\frac{(2+Q)(N+1)^{p-1}\bar{\delta}^p-(2\bar{\gamma})^p}
{p(N+1)^{p-1}\big(\sum_{k=1}^N
F(k,\bar{\delta})-\sum_{k=1}^N\max_{|t|\leq\bar{\gamma}}F(k,t)\big)},
$$
problem \eqref{e1.1} admits at least one non-trivial solution
$\tilde{u}\in S$, such that
$\|\tilde{u}\|>\frac{2\bar{\gamma}}{(N+1)^{(p-1)/p}}$.
\end{theorem}
\begin{proof}
Take the real Banach space $S$ as defined in Section 2, and put
$\Phi,\Psi,I_\lambda$ as in \eqref{e2.4}. Our aim is to apply
Theorem \ref{the2.2} to function $I_\lambda$. The functionals $\Phi$
and $\Psi$ satisfy all regularity assumptions requested in Theorem
\ref{the2.2}; see Remark \ref{rem2.3}. Moreover, by standard
computations, the assumption \eqref{e3.8} implies that
$I_\lambda,\,\lambda>0$, is coercive. So, our aim is to verify
condition \eqref{e2.2} of Theorem \ref{the2.2}. To this end, we put
$$
r:=\frac{(2\bar{\gamma})^p}{p(N+1)^{p-1}},
$$
and pick $w\in S$, defined as
\[
w_k:=\begin{cases}
\bar{\delta}, & \text{if } k\in[1,N],\\
0, & \text{if } k=0,\,k=N+1.
\end{cases}
\]
Arguing as in the proof of Theorem \ref{the3.1} we obtain that
$$
\rho(r)\geq p(N+1)^{p-1}\frac{\sum_{k=1}^N
F(k,\bar{\delta})-\sum_{k=1}^N\max_{|t|\leq
\bar{\gamma}}F(k,t)}{(2+Q)(N+1)^{p-1}\bar{\delta}^p-(2\bar{\gamma})^p}.
$$
So, from our assumption it follows that $\rho(r)>0$.
Hence, from Theorem \ref{the2.2} for each $\lambda>\tilde{\lambda}$,
the functional $I_\lambda$ admits at least one local minimum
$\tilde{u}$ such that
$\|\tilde{u}\|>2\bar{\gamma}/\big((N+1)^{(p-1)/p}\big)$ and the
conclusion is achieved.
\end{proof}
\section{Multiplicity results}
The main aim of this section is to present multiplicity results.
First, as consequence of Theorem \ref{the3.1}, taking into account
the classical theorem of Ambrosetti and Rabinowitz, we have the
following multiplicity result.
\begin{theorem}\label{the4.1}
Let the assumptions of Theorem \ref{the3.1} be satisfied. Assume
also that $f(k,0)\neq 0$ for some $k\in [1,N]$. Moreover, let
\begin{itemize}
\item[(AR)] there exist constants $\nu>p$ and $R>0$ such
that, for all $|\xi|\geq R$ and for all $k\in[1,N]$, one has
\begin{equation}\label{e4.1}
0<\nu F(k,\xi)\leq \xi f(k,\xi).
\end{equation}
\end{itemize}
Then, for each
$\lambda\in\frac{1}{p(N+1)^{p-1}}]\frac{1}{a_{\gamma_1}(\delta)},
\frac{1}{a_{\gamma_2}(\delta)}[$,
problem \eqref{e1.1} admits at least two non-trivial solutions
$\bar{u}_1,\bar{u}_2$, such that
\begin{equation}\label{e4.2}
\frac{2\gamma_1}{(N+1)^{(p-1)/p}}
<\|\bar{u}_1\|
<\frac{2\gamma_2}{(N+1)^{(p-1)/p}}.
\end{equation}
\end{theorem}
\begin{proof}
Fix $\lambda$ as in the conclusion. So, Theorem \ref{the3.1} ensures
that problem \eqref{e1.1} admits at least one non-trivial solution
$\bar{u}_1$ satisfying the condition \eqref{e4.2} which is a local
minimum of the functional $I_\lambda$.
Now, we prove the existence of the second local minimum distinct
from the first one. To this end, we must show that the functional
$I_\lambda$ satisfies the hypotheses of the mountain pass theorem.
Clearly, the functional $I_\lambda$ is of class $C^1$ and
$I_\lambda(0)=0$.
We can assume that $\bar{u}_1$ is a strict local minimum for
$I_\lambda$ in $S$. Therefore, there is $\rho>0$ such that
$\inf_{\|u-\bar{u}_1\|=\rho}I_\lambda(u)>I_\lambda(\bar{u}_1)$, so
condition \cite[$(I_1)$, Theorem 2.2]{Rab} is verified.
Integrating condition \eqref{e4.1} shows that there exist constants
$a_1,a_2>0$ such that
$$
F(k,t)\geq a_1|t|^\nu-a_2
$$
for all $k\in[1,N]$ and $t\in\mathbb{R}$. Now, choosing any
$u\in S\setminus\{0\}$, one has
\begin{align*}
I_\lambda(tu)
&= (\Phi-\lambda\Psi)(tu)\\
&= \frac{1}{p}\|tu\|^p-\lambda\sum_{k=1}^N F(k,tu_k)\\
&\leq \frac{t^p}{p}\|u\|^p-\lambda t^\nu
a_1\sum_{k=1}^N|u_k|^\nu+\lambda a_2 N\to -\infty
\end{align*}
as $t\to +\infty$, so condition \cite[$(I_2)$, Theorem 2.2]{Rab} is satisfied.
So, the functional $I_\lambda$ satisfies the
geometry of mountain pass. Moreover, by standard computations,
$I_\lambda$ satisfies the Palais-Smale condition. Hence, the
classical theorem of Ambrosetti and Rabinowitz ensures a critical
point $\bar{u}_2$ of $I_\lambda$ such that
$I_\lambda(\bar{u}_2)>I_\lambda(\bar{u}_1)$. So, $\bar{u}_1$ and
$\bar{u}_2$ are two distinct solutions of \eqref{e1.1} and the proof
is complete.
\end{proof}
Next, as a consequence of Theorems \ref{the3.7} and \ref{the3.2},
the following theorem of the existence of three solutions is
obtained and its consequence for the nonlinearity with separable
variables is presented.
\begin{theorem}\label{the4.2}
Assume that \eqref{e3.8} holds. Moreover, assume that there exist
four positive constants $\gamma,\delta,\bar{\gamma},\bar{\delta}$,
with
$$
\frac{(2+Q)^{1/p}(N+1)^{(p-1)/p}}{2}\delta
<\gamma\leq\bar{\gamma}
<\frac{(2+Q)^{1/p}(N+1)^{(p-1)/p}}{2}\bar{\delta},
$$
such that \eqref{e3.3}, \eqref{e3.7} and
\begin{equation}\label{e4.3}
\frac{\sum_{k=1}^N\max_{|t|\leq\gamma}F(k,t)}{(2\gamma)^p}<\frac{\sum_{k=1}^N
F(k,\bar{\delta})-\sum_{k=1}^N\max_{|t|
\leq\bar{\gamma}}F(k,t)}{(2+Q)(N+1)^{p-1}\bar{\delta}^p-(2\bar{\gamma})^p}.
\end{equation}
are satisfied. Then, for each
$$
\lambda\in\Lambda=\big]\max\big\{\tilde{\lambda},
\frac{(2+Q)\delta^p}{p\sum_{k=1}^N
F(k,\delta)}\big\},\frac{(2\gamma)^p}{p(N+1)^{p-1}
\sum_{k=1}^N\max_{|t|\leq\gamma}F(k,t)}\big[,
$$
problem \eqref{e1.1} admits at least three solutions.
\end{theorem}
\begin{proof}
First, we observe that $\Lambda\neq\emptyset$ owing to \eqref{e4.3}.
Next, fix $\lambda\in\Lambda$. Theorem \ref{the3.2} ensures a
non-trivial solution $\bar{u}$ such that
$\|\bar{u}\|<\frac{2\gamma}{(N+1)^{(p-1)/p}}$ which is a local
minimum for the associated functional $I_\lambda$, as well as
Theorem \ref{the3.7} guarantees a non-trivial solution $\tilde{u}$
such that $\|\tilde{u}\|>\frac{2\bar{\gamma}}{(N+1)^{(p-1)/p}}$
which is a local minimum for $I_\lambda$. Hence, the mountain pass
theorem as given by Pucci and Serrin (see \cite{PuSe1}) ensures the
conclusion.
\end{proof}
\begin{theorem}\label{the4.3}
Assume that $g:\mathbb{R}\to\mathbb{R}$ is a non-negative
continuous function such that
\begin{gather}
\limsup_{\xi\to 0^+}\frac{G(\xi)}{\xi^p}=+\infty, \label{e4.4}\\
\limsup_{\xi\to +\infty}\frac{G(\xi)}{\xi^p}=0. \label{e4.5}
\end{gather}
Further, assume that there exist two positive constants
$\bar{\gamma},\bar{\delta}$, with
$$
\bar{\gamma}<\frac{(2+Q)^{1/p}(N+1)^{(p-1)/p}}{2}\bar{\delta},
$$
such that
\begin{equation}\label{e4.6}
\frac{G(\bar{\gamma})}{\bar{\gamma}^p}
<\Big(\frac{2^p}{(2+Q)(N+1)^{p-1}}\Big)\frac{G(\bar{\delta})}{\bar{\delta}^p}.
\end{equation}
Then, for each
$$
\lambda\in\frac{1}{p\sum_{k=1}^N\alpha_k}
\big]\frac{(2+Q)\bar{\delta}^p}{G(\bar{\delta})},
\frac{(2\bar{\gamma})^p}{(N+1)^{p-1}G(\bar{\gamma})}\big[,
$$
problem \eqref{e3.6} admits at least three non-negative solutions.
\end{theorem}
\begin{proof}
Clearly, \eqref{e4.5} implies \eqref{e3.8}. Moreover, by choosing
$\delta$ small enough and $\gamma=\bar{\gamma}$, simple computations
show that \eqref{e4.4} implies \eqref{e3.3}. Finally, from
\eqref{e4.6} we get \eqref{e3.7} and also \eqref{e4.3}. Hence,
Theorem \ref{the4.2} ensures the conclusion.
\end{proof}
Finally, we present two applications of our results.
\begin{example}\label{ex4.4}\rm
Consider the problem
\begin{equation}\label{e4.7}
\begin{gathered}
-\Delta^2 u_{k-1}+q_ku_k=\lambda\Big(\frac{1}{6}+|u_k|^2u_k\Big),\quad
k\in[1,N],\\
u_0=u_{N+1}=0.
\end{gathered}
\end{equation}
Let $g(t)=\frac{1}{6}+|t|^2t$ for all $t\in\mathbb{R}$. Obviously,
$g(0)\neq 0$. Since
$$
\lim_{\xi\to 0^{+}}\frac{g(\xi)}{\xi}
=\lim_{\xi\to 0^{+}}\big(\frac{1}{6\xi}+|\xi|^2\big)=+\infty,
$$
condition \eqref{e1.2} holds true. Choose $\nu=3$ and $R=1$, we have
$$
0<3G(\xi)\leq \xi g(\xi),
$$
for all $|\xi|\geq 1$. Moreover, one has
$$
\frac{2}{N(N+1)}\sup_{\gamma>0}\frac{\gamma^2}{\max_{|\xi|\leq\gamma}G(\xi)}=
\frac{2}{N(N+1)}\sup_{\gamma>0}\frac{12\gamma}{2+3\gamma^3}=
\frac{6}{N(N+1)}.
$$
Then, owing to Theorem \ref{the1.1}, for each
$\lambda\in]0,\frac{6}{N(N+1)}[$, problem \eqref{e4.7}
admits at least two non-trivial solutions.
\end{example}
\begin{example}\label{ex4.5}\rm
Consider the problem
\begin{gather*}
-\Delta^2 u_{k-1}=\frac{1}{10}\Big(\frac{u_k^8}{e^{u_k}}+1\Big),\quad k\in[1,3],\\
u_0=u_4=0.
\end{gather*}
Then, owing to Theorem \ref{the1.2}, it admits three positive
solutions. In fact, one has
\begin{gather*}
\lim_{\xi\to 0^+}\frac{g(\xi)}{\xi}
=\lim_{\xi\to0^+}\frac{\frac{\xi^8}{e^\xi}+1}{\xi}=+\infty,
\\
\lim_{\xi\to +\infty}\frac{g(\xi)}{\xi}
=\lim_{\xi\to +\infty}\frac{\frac{\xi^8}{e^\xi}+1}{\xi}=0.
\end{gather*}
Moreover, taking into account that
$$
G(t)=t-\frac{\sum_{i=0}^8\frac{8!}{i!}t^i}{e^t}+8!,\quad\forall t\in\mathbb{R},
$$
one has $G(1)<\frac{1}{8}G(2)$ and
$\frac{4}{3G(2)}<\frac{1}{10}<\frac{1}{6G(1)}$.
\end{example}
\subsection*{Acknowledgments}
This research work was supported by a research grant from the
University of Mazandaran.
\begin{thebibliography}{99}
\bibitem{AfHaHe} G. A. Afrouzi, A. Hadjian, S. Heidarkhani;
\emph{Non-trivial solutions for a two-point boundary value problem},
Ann. Polon. Math., \textbf{108} (2013), 75--84.
\bibitem{AgPerOr1} R. P. Agarwal, K. Perera, D. O'Regan;
\emph{Multiple positive solutions of singular and nonsingular
discrete problems via variational methods},
Nonlinear Anal., \textbf{58} (2004), 69--73.
\bibitem{AgPerOr2} R. P. Agarwal, K. Perera, D. O'Regan;
\emph{Multiple positive solutions of singular discrete $p$-Laplacian
problems via variational methods},
Adv. Difference Equ., Vol. \textbf{2005} (2005), 93--99.
\bibitem{AmbRab} A. Ambrosetti, P. H. Rabinowitz;
\emph{Dual variational methods in critical point theory and
applications},
J. Funct. Anal., \textbf{14} (1973), 349--381.
\bibitem{AndRachTis} D. R. Anderson, I. Rachunkov\'{a}, C. C. Tisdell;
\emph{Solvability of discrete Neumann boundary value problems},
Adv. Differential Equations, \textbf{2} (2007), 93--99.
\bibitem{BerMaw} C. Bereanu, J. Mawhin;
\emph{Boundary value problems for second-order nonlinear difference
equations with discrete $\phi$-Laplacian and singular $\phi$},
J. Difference Equ. Appl., \textbf{14} (2008), 1099--1118.
\bibitem{BiSunZha} L.-H. Bian, H.-R. Sun, Q.-G. Zhang;
\emph{Solutions for discrete $p$-Laplacian periodic boundary value
problems via critical point theory},
J. Difference Equ. Appl., \textbf{18} (2012), 345--355.
\bibitem{Bonanno} G. Bonanno;
\emph{A critical point theorem via the Ekeland variational
principle},
Nonlinear Anal., \textbf{75} (2012), 2992--3007.
\bibitem{BonaCand2} G. Bonanno, P. Candito;
\emph{Infinitely many solutions for a class of discrete non-linear
boundary value problems},
Appl. Anal., \textbf{88} (2009), 605--616.
\bibitem{BonaMoliRad1} G. Bonanno, G. Molica Bisci, V. R\u{a}dulescu;
\emph{Nonlinear elliptic problems on Riemannian manifolds and
applications to Emden-Fowler type equations},
Manuscripta Math., \textbf{142} (2013), 157--185.
\bibitem{BonaMoliRad2} G. Bonanno, G. Molica Bisci, V. R\u{a}dulescu;
\emph{Weak solutions and energy estimates for a class of nonlinear
elliptic Neumann problems},
Adv. Nonlinear Stud., \textbf{13} (2013), 373--389.
\bibitem{BonaSci2} G. Bonanno, A. Sciammetta;
\emph{Existence and multiplicity results to Neumann problems for
elliptic equations involving the $p$-Laplacian},
J. Math. Anal. Appl., \textbf{390} (2012), 59--67.
\bibitem{CanMoli} P. Candito, G. Molica Bisci;
\emph{Existence of two solutions for a second-order discrete
boundary value problem},
Adv. Nonlinear Stud., \textbf{11} (2011), 443--453.
\bibitem{ChuJi} J. Chu, D. Jiang;
\emph{Eigenvalues and discrete boundary value problems for the
one-dimensional $p$-Laplacian},
J. Math. Anal. Appl., \textbf{305} (2005), 452--465.
\bibitem{Dag2} G. D'Agu\`{i};
\emph{Multiplicity results for nonlinear mixed boundary value
problem},
Bound. Value Probl., Vol. \textbf{2012} (2012), No. 134, 1--12.
\bibitem{HendThom} J. Henderson, H. B. Thompson;
\emph{Existence of multiple solutions for second order discrete
boundary value problems},
Comput. Math. Appl., \textbf{43} (2002), 1239--1248.
\bibitem{JiChuOrAg} D. Jiang, J. Chu, D. O'Regan, R. P. Agarwal;
\emph{Positive solutions for continuous and discrete boundary value
problems to the one-dimensional $p$-Laplacian},
Math. Inequal. Appl., \textbf{7} (2004), 523--534.
\bibitem{JiZh} L. Jiang, Z. Zhou;
\emph{Three solutions to Dirichlet boundary value problems for
$p$-Laplacian difference equations},
Adv. Difference Equ., Vol. \textbf{2008} (2008), Article ID 345916,
1--10.
\bibitem{KriMihRad} A. Krist\'aly, M. Mihailescu, V. R\u{a}dulescu;
\emph{Discrete boundary value problems involving oscillatory
nonlinearities: small and large solutions},
J. Difference Equ. Appl., \textbf{17} (2011), 1431--1440.
\bibitem{PuSe1} P. Pucci, J. Serrin;
\emph{A mountain pass theorem},
J. Differential Equations, \textbf{63} (1985), 142--149.
\bibitem{RachTis} I. Rachunkov\'{a}, C. C. Tisdell;
\emph{Existence of non-spurious solutions to discrete Dirichlet
problems with lower and upper solutions},
Nonlinear Anal., \textbf{67} (2007), 1236--1245.
\bibitem{Rab} P. H. Rabinowitz;
\emph{Minimax Methods in Critical Point Theory with Applications to
Differential Equations},
CBMS Reg. Conf. Ser. Math., Vol. 65, Amer. Math. Soc., Providence,
RI, 1986.
\bibitem{Ricceri} B. Ricceri;
\emph{A general variational principle and some of its applications},
J. Comput. Appl. Math., \textbf{113} (2000), 401--410.
\bibitem{TiGe} Y. Tian, W. Ge;
\emph{Multiple positive solutions of boundary value problems for
second-order discrete equations on the half-line},
J. Difference Equ. Appl., \textbf{12} (2006), 191--208.
\end{thebibliography}
\end{document}