\documentclass[reqno]{amsart}
\usepackage{hyperref}

\AtBeginDocument{{\noindent\small
\emph{Electronic Journal of Differential Equations},
Vol. 2017 (2017), No. 35, pp. 1--10.\newline
ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu}
\thanks{\copyright 2017 Texas State University.}
\vspace{8mm}}

\begin{document}
\title[\hfilneg EJDE-2017/35\hfil Existence of solutions]
{Existence of solutions to nonlinear problems with
three-point boundary conditions}

\author[D. P. D. Santos \hfil EJDE-2017/35\hfilneg]
{Dionicio Pastor Dallos Santos}

\address{Dionicio Pastor Dallos Santos \newline
Department of Mathematics, IME-USP,
Cidade Universit\'aria,
 CEP 05508-090, S\~ao Paulo, SP, Brazil}
\email{dionicio@ime.usp.br}

\thanks{Submitted September 21, 2016. Published January 30, 2017.}
\subjclass[2010]{34B15, 34B16, 47H10, 47H11}
\keywords{Boundary value problems; Schauder fixed point theorem;
\hfill\break\indent Leray-Schauder degree; lower and upper solutions}

\begin{abstract}
 Using  Leray-Schauder degree theory and the method of upper and lower solutions,
 we obtain a solution for nonlinear boundary-value problem
 \begin{gather*}
 \big(\varphi(u' )\big)'= f(t,u,u') \\
 l(u,u')=0,
 \end{gather*}
 where $l(u,u')=0$  denotes the  three-point boundary conditions  on
 $[0,T]$, and $\varphi$  is a  homeomorphism such that  $\varphi(0)=0$.
\end{abstract}

\maketitle
\numberwithin{equation}{section}
\newtheorem{theorem}{Theorem}[section]
\newtheorem{lemma}[theorem]{Lemma}
\newtheorem{remark}[theorem]{Remark}
\newtheorem{definition}[theorem]{Definition}
\newtheorem{corollary}[theorem]{Corollary}
\newtheorem{example}[theorem]{Example}
\allowdisplaybreaks

\section{Introduction}

The purpose of this article is to obtain  a solution
for the nonlinear problem
\begin{equation}\label{equ}
\begin{gathered}
\big(\varphi(u' )\big)'   = f(t,u,u')  \\
l(u,u')=0,
\end{gathered}
\end{equation}
where $l(u,u')=0$ denotes the boundary conditions
$u(T) = u'(0)= u'(T)$   or   $u(0) = u(T)= u'(0)$  on the interval
 $[0,T]$, $\varphi$ is a singular or classic homeomorphism such
that $\varphi(0)=0$, and $f: [0, T]\times \mathbb{R} \times
\mathbb{R}\to \mathbb{R}$ is  continuous.

Solvability of  two-point  boundary  value problems can be investigated
by  various methods: fixed point theorems, topological degree arguments,
variational methods, lower and upper functions, etc.,
see  for example, \cite{man4, garcia, grae, huang, maw, oregan}
and the reference therein.
In particular, the author in \cite{oregan} proved the existence of solutions
for the Dirichlet and mixed problems, assuming  $f$ and $\varphi$are  continuous
and that $\varphi$  is strictly  increasing and
satisfies $\varphi(\mathbb{R})=\mathbb{R}$ and
$\varphi^{-1}\in C^{1}(\mathbb{R})$.

Bereanu and  Mawhin \cite{ma1} proved the  existence of solutions
for the periodic bound\-ary-value problem
\begin{gather*}
\big(\varphi(u' )\big)'   = f(t,u,u') \\
u(0)=u(T), \quad  u'(0)=u'(T),
\end{gather*}
assuming that $f : [0, T]\times \mathbb{R} \times \mathbb{R} \to \mathbb{R}$
is a  continuous function and $\varphi: \mathbb{R} \to (-a,a) \ (0<a\leq\infty) $
is an increasing homeomorphism such that $\varphi(0)=0$.
They obtained solutions by using the method of upper and lower solutions
and  the Leray-Schauder degree.
The interest in this class of nonlinear operators $u\mapsto (\varphi(u' ))' $
is mainly due to the fact that they include  the $p$-Laplacian operator
\[
u\mapsto |u'|^{p-2}u',
\]
where $p>1$.

Using the barrier strip argument and topological transversality theorem
the authors in \cite{chinos} obtained the existence of solutions for nonlinear
boundary-value problems
\begin{gather*}
\big(\varphi(u' )\big)'   = f(t,u,u')  \\
u(0)=A, \quad u'(1)=B,
\end{gather*}

where $f: [0, T]\times \mathbb{R} \times \mathbb{R}\to \mathbb{R}$
is  continuous and $\varphi:\mathbb{R} \to \mathbb{R}$ is an increasing
homeomorphism.

Inspired by these results, the main aim of this paper is to study the
existence of solutions for \eqref{equ} using  topological methods based
upon Leray-Schauder degree. The main contribution of this paper is the
extension of some results above cited  to a more general type of boundary
conditions. Such  problems do not seem to have been studied in the literature.

This article is organized as follows.
 In Section 2, we introduce some notations and  preliminaries, which
will be crucial in the proofs of our results. Section 3  is devoted
to the study of existence of solutions for  boundary-value problems
\begin{gather*}
\big(\varphi(u' )\big)'   = f(t,u,u')  \\
u(T)=u'(0)=u'(T),
\end{gather*}
where $\varphi:(-a,a)\to \mathbb{R}$ (we call it singular).
We call \emph{solution } of this problem any function
$u: [0, T]\to \mathbb{R}$ of class $C^{1}$ such that $ \max_{ [0, T]}|u'(t)|<a$,
satisfying  the boundary conditions and  the function
 $\varphi(u')$ is continuously differentiable and
$(\varphi(u'(t) ))'  = f(t,u(t),u'(t))$ for all $t\in [0,T]$.
Combining the method of upper and lower solutions and the fixed point
theorem of Schauder, we prove in Section 4 an  existence result
(Theorem \ref{gato}) for  boundary-value problems of the form
\begin{gather*}
\big(\varphi(u' )\big)' = f(t,u,u') \\
u(0)=u(T)=u'(0),
\end{gather*}
where $\varphi: \mathbb{R} \to \mathbb{R}$  is an increasing homeomorphism.
A \emph{solution } of this problem is any function  $u: [0, T]\to \mathbb{R}$
of class $C^{1}$ satisfying the boundary conditions and satisfying that
$\varphi(u')$ is continuously differentiable and
$(\varphi(u'(t) ))'  = f(t,u(t),u'(t))$ for all $t\in [0,T]$.

\section{Preliminaries}\label{S2}

We first introduce some notation.
Let $C=C( [0,T], \mathbb R)$ denote the Banach space of 
continuous functions from $ [0, T]$ to $\mathbb R$ endowed with the
 uniform norm $\| \cdot \|_{\infty}$, and  $C^{1}=C^{1}( [0,T], \mathbb R)$
the Banach space of continuously differentiable functions  from $ [0, T]$
 to $\mathbb R$ equipped with the usual norm
$\|u\|_1=\|u\|_{\infty} +  \|u' \|_{\infty}$.

We introduce the following operators:
the  \textit{Nemytskii operator} $N_f:C^{1} \to C $,
\begin{center}
$N_f (u)(t)=f(t,u(t),u'(t))$,
\end{center}
the  \textit{integration operator}   $H:C \to C^{1}$,
\begin{center}
$ H(u)(t)=\int_0^t u(s)ds$,
\end{center}
also the following continuous linear mappings:
\begin{gather*}
K:C \to C^{1}, \quad  K(u)(t)=-\int_t^T u(s)ds , \\
Q:C \to C, \quad  Q(u)(t)=\frac{1}{T}\int_0^T u(s)ds, \\
S:C \to C, \quad  S(u)(t)=u(T), \\
P:C \to C, \quad  P(u)(t)=u(0).
\end{gather*}
The following  technical result proved by Bereanu and Mawhin  is needed 
for the construction of the equivalent fixed point problem (see \cite{mabe}).

\begin{lemma}\label{lebema1}
 For  each $h\in C$,  there exists a unique  
$Q_{\varphi}=Q_{\varphi}(h)\in \operatorname{Im}(h)$
(where $\operatorname{Im}(h)$ denotes the range of  $h$) such that
\begin{center}
$\int_0^T \varphi^{-1}(h(t)-Q_{\varphi}(h))dt = 0$.
\end{center}
Moreover, the function  $Q_{\varphi}:C \to \mathbb{R}$ is continuous  
and sends bounded sets into bounded sets.
\end{lemma}

\section{Boundary value problems with  singular $\varphi$-Laplacian}
\label{S3}

 In this section we are interested in boundary-value problems of the type
\begin{equation}\label{equa1}
\begin{gathered}
\big(\varphi(u' )\big)'   = f(t,u,u')  \\
u(T) = u'(0) = u'(T),
\end{gathered}
\end{equation}
where  $\varphi: (-a,a) \to  \mathbb{R}$ is  a homeomorphism such 
that $\varphi(0)=0$ and $f: [0, T]\times \mathbb{R} \times \mathbb{R}\to \mathbb{R} $ 
is a continuous function. We remember that an  \emph{solution } of this problem  
is any function $u: [0, T]\to \mathbb{R}$ of class $C^{1}$ such that 
$ \max_{ [0, T]}|u'(t)|<a$, satisfying  the boundary conditions and  the 
function  $\varphi(u')$ is continuously differentiable and 
$(\varphi(u'(t) ))'  = f(t,u(t),u'(t))$ for all $t\in [0,T]$.

Let us consider the operator $M_1:C^{1}\to C^{1}$,
\[
M_1(u)= S(u)+ Q(N_{f}(u)) + K(\varphi^{-1} [H( N_f (u)
-Q(N_{f}(u)))+\varphi(S(u))]).
\]
Here  $\varphi^{-1}$ is understood as the operator  
$\varphi^{-1}: C \to B_{a}(0)\subset C$ defined for 
$\varphi^{-1}(v)(t)=\varphi^{-1}(v(t))$. The symbol $B_{a}(0)$ denoting the 
open ball of center $0$ and radius $a$ in $C$. It is clear that 
$\varphi^{-1}$ is continuous and sends bounded sets into bounded sets.
When $\varphi: \mathbb{R} \to  \mathbb{R}$, such an operator has been 
considered  in \cite{dallos}.

\begin{lemma}\label{mate2}
 A map $u \in C^{1}$ is  a solution  of \eqref{equa1} if and only if  $u$ 
is a fixed point of the operator $M_1$.
\end{lemma}

\begin{proof}
For $ u \in C^{1}$, we have the following equivalences:
\begin{align*}
&(\varphi(u'))'  = N_f (u), \;   u'(T)= u'(0), \; u'(0)=u(T) \\
&\Leftrightarrow(\varphi(u'))' = N_f(u)- Q(N_{f}(u)),\; 
  Q(N_{f}(u))=0, \;  u'(0)=u(T) \\
&\Leftrightarrow \varphi(u')= H( N_f (u)-Q(N_{f}(u)))+\varphi(u'(0)),\;
 Q(N_{f}(u))=0, \;  u'(0)=u(T) \\
&\Leftrightarrow u'= \varphi^{-1} [H( N_f (u)-Q(N_{f}(u)))+\varphi(u'(0))],\;
 Q(N_{f}(u))=0,  \;  u'(0)=u(T) \\
&\begin{aligned}
\Leftrightarrow u
 &= u(T)+ K(\varphi^{-1} [H( N_f (u)-Q(N_{f}(u))) + \varphi(u'(0))]),\\
 &\quad Q(N_{f}(u))=0, \;  u'(0)=u(T)
\end{aligned} \\
&\Leftrightarrow  u= u(T)+ Q(N_{f}(u)) + K(\varphi^{-1}
  [H( N_f (u)-Q(N_{f}(u)))+\varphi(u(T))]) \\
&\Leftrightarrow  u= S(u)+ Q(N_{f}(u)) + K(\varphi^{-1} [H( N_f (u)-Q(N_{f}(u)))
 +\varphi(S(u))]).
\end{align*}
\end{proof}

\begin{remark} \label{rmk3.2} \rm
 Note that  $u'(T)= u'(0) \Leftrightarrow  Q(N_{f}(u))=0$.
\end{remark}

\begin{lemma}\label{mate3}
  The operator $M_1:C^{1}\to C^{1}$ is completely continuous.
\end{lemma}

\begin{proof}
Let $ \Lambda \subset C^{1}$ be a bounded set. 
Then, if $u \in \Lambda$, there exists a constant  $\rho>0$ such  that
\begin{equation}\label{dp4}
\| u\|_1\leq \rho.
\end{equation}
Next, we show that  $\overline{M_1(\Lambda)}\subset  C^{1}$ is a compact set. 
Let $(v_n)_n $  be a sequence in   $M_1(\Lambda)$,  and
 let $(u_n)_n$  be a sequence  in  $\Lambda$ such that  $v_n=M_1(u_n)$. 
Using \eqref{dp4}, we have that there exists a constant  $L_1>0$ such that,  
for all $n\in \mathbb{N}$,
\[
 \| N_{f}(u_n)\|_{\infty}\leq L_1,
\]
which implies 
\[
\| H(N_{f}(u_n)-Q(N_f(u_n))) \|_{\infty}\leq 2L_1T.
\]
Hence the sequence  $ (H(N_{f}(u_n)-Q(N_f(u_n))))_n$ is bounded in  $C$.
Moreover, for  $t, t_1\in [0, T]$ and for all $n\in \mathbb{N}$, we have
\begin{align*}
&| H(N_{f}(u_n)-Q(N_f(u_n)))(t) - H(N_{f}(u_n)-Q(N_f(u_n)))(t_1)  |\\
&\leq \big|\int_{t_1}^{t}N_f(u_n)(s) ds\big| 
 + \big| \int_{t_1}^{t}Q(N_f(u_n))(s) ds\big|\\
&\leq L_1|t-t_1| +  |t-t_1|\| Q(N_f(u_n))\|_{\infty}\\
&\leq 2L_1|t-t_1|,
\end{align*}
which  implies that  $ \big(H(N_{f}(u_n)-Q(N_f(u_n)))\big)_n$ is equicontinuous. 
Thus, by the Arzel\`a-Ascoli theorem there is a subsequence of  
$(H(N_{f}(u_n)-Q(N_f(u_n))))_n$, which we call
$(H(N_{f}(u_{n_{j}})-Q(N_f(u_{n_{j}}))))_{j}$, which is  convergent in $C$.
 Then, passing to a subsequence if  necessary, we obtain that the sequence 
$$ 
(H(N_{f}(u_{n_{j}})-Q(N_f(u_{n_{j}}))) +  \varphi(S(u_{n_{j}})))_{j}
$$
is convergent in  $C$. Using the fact that   
$\varphi^{-1}: C \to B_{a}(0)\subset C$ is continuous it follows from
\[
M_1(u_{n_{j}})'=\varphi^{-1} 
[(H(N_{f}(u_{n_{j}})-Q(N_f(u_{n_{j}})))   + \varphi(S(u_{n_{j}}))) ]
\]
that the sequence  $(M_1(u_{n_{j}})')_{j}$ is convergent in  $C$. 
Therefore, passing if necessary to a subsequence, we have that  
$(v_{n_{j}})_{j}=( M_1(u_{n_{j}}))_{j}$  is convergent in $C^{1}$. 
Finally, let  $(v_n)_n$ be a sequence in  $\overline{M_1(\Lambda)}$.
 Let  $(z_n)_n\subseteq M_1(\Lambda)$  be such that
\[
\lim_{n \to \infty}\| z_n-v_n\|_1=0.
\]
Let $(z_{n_{j}})_{j}$ be  a subsequence of  $(z_n)_n$  such that  converge to $z$. 
It follows that  $z\in \overline{M_1(\Lambda)}$ and  $(v_{n_{j}})_{j}$ 
converge to $z$. This completes the proof.
\end{proof}


To apply  the Leray-Schauder degree to the equivalent fixed point operator $M_1$, 
for $\lambda \in  [0,1]$, we introduce the family of boundary-value problems
\begin{equation}\label{misto3}
\begin{gathered}
(\varphi(u' ))'   = \lambda N_{f}(u)+(1-\lambda)Q(N_{f}(u))  \\
u(T)=u'(0)=u'(T).
\end{gathered}
\end{equation}
Note that \eqref{misto3} coincide with \eqref{equa1} for $\lambda =1$. 
So, for  each $ \lambda \in [0,1]$, the  nonlinear operator associated  
with \eqref{misto3} by Lemma \ref{mate2} is the  operator $M(\lambda,\cdot)$, 
where $M$  is defined on $[0,1]\times C^{1}$  by
\begin{equation}\label{eme}
 M(\lambda,u)=S(u)+Q(N_{f}(u)) + K(\varphi^{-1} [\lambda H( N_f (u)
-Q(N_{f}(u)))+\varphi(S(u))]).
\end{equation}
Using the same arguments as in the proof  of Lemma \ref{mate3} 
 we show that the operator  $M$ is completely continuous. 
Moreover, using the same reasoning as in Lemma \ref{mate2}, 
 system \eqref{misto3} is equivalent to the problem
\[
u=M(\lambda, u).
\]
The following lemma gives  a priori bounds for the possible fixed points of $M$.

\begin{lemma}\label{cota}
Let $f: [0,T]\times \mathbb{R}\times \mathbb{R}\to \mathbb{R}$ be continuous. 
If $(\lambda,u)\in  [0,1]\times C^{1} $ is such that  $u=M(\lambda, u)$, then
\begin{center}
$\|u\|_1\leq a(2+T)$.
\end{center}
\end{lemma}

\begin{proof}
Let $ [0,T]\times C^{1}$ be  such that $u=M(\lambda, u)$. Then
\[
u =M(\lambda, u)= S(u)+Q(N_{f}(u)) + K(\varphi^{-1} 
[\lambda H( N_f (u)-Q(N_{f}(u)))+\varphi(S(u))]).
\]
Differentiating, we obtain
\[
u' = [M(\lambda, u)]' = \varphi^{-1} 
[\lambda H( N_f (u)-Q(N_{f}(u)))+\varphi(S(u))],
\]
so that $\|u'\|_{\infty}\leq a$.
 Because $u \in C^{1}$ is such that $u(T)=u'(0)$, we have 
\[
|u(t)|\leq |u(T)|+ \int_ 0 ^T  |u'(s)| ds \leq a +aT,\quad t\in[0,T],
\]
 and hence  $\|u\|_1 =\|u\|_{\infty}+ \|u'\|_{\infty}\leq a+aT +a=a(2+T)$.
 This completes the proof
\end{proof}

\subsection{Existence result} %\label{S:0}

We can now prove an existence theorem for \eqref{equa1}. 
We denote by $\deg_{B}$ the Brouwer degree and  for $\deg_{LS}$ the
Leray-Schauder degree, and define the mapping 
$G:\mathbb{R}^2\to \mathbb{R}^2$ by
\[
G (x,y)= (xT+yT^2-yT-\frac{1}{T}\int_0^T f(t,x+yt,y)dt, y-x-yT).
\]

\begin{theorem}\label{prin1}
Let $f: [0,T]\times \mathbb{R}\times \mathbb{R}\to \mathbb{R}$ be continuous. 
Then for all $\rho>a(2+T)$
\[
\deg_{LS}(I-M(1,\cdot), B_{\rho}(0),0) 
= \deg_{B}(G, B_{\rho}(0) \cap  \mathbb{R}^2,0).
\]
If furthermore
\[
\deg_{B}(G,B_{\rho}(0) \cap  \mathbb{R}^2,0)\neq 0,
\]
then problem \eqref{equa1} has at least one solution.
\end{theorem}

\begin{proof}
Let $M$  be the operator given by  \eqref{eme} and let $\rho>a(2+T)$.
 Using Lemma \ref{cota}, we have that, for each $\lambda \in  [0,T]$, 
the Leray-Schauder degree $\deg_{LS}(I-M(\lambda,\cdot), B_{\rho}(0),0)$ 
is well defined, and  by the homotopy invariance, one has
\[
\deg_{LS}(I-M(0,\cdot),B_{\rho}(0),0)= \deg_{LS}(I-M(1,\cdot),B_{\rho}(0),0).
\]
On the other hand,
\[
\deg_{LS}(I-M(0,\cdot),B_{\rho}(0),0) 
= \deg_{LS}(I-(S+QN_{f}+KS) ,B_{\rho}(0),0).
\]
But the range of the mapping
\[
u \mapsto S(u)+Q(N_{f}(u))+K(S(u))
\]
is contained in the subspace  of related functions, isomorphic to 
$\mathbb{R}^2$. Using  homotopy invariance and reduction properties
of  Leray-Schauder degree \cite{man7}, we obtain
\begin{align*}
& \deg_{LS}(I-(S+QN_{f}+KS) ,B_{\rho}(0),0)\\
& = \deg_{B}\big(I-(S+QN_{f}+KS)\big|_{\overline{B_{\rho}(0) 
\cap  \mathbb{R}^2}} ,B_{\rho}(0)\cap \mathbb{R}^2, 0\big)\\
&=\deg_{B}(G,B_{\rho}(0)\cap  \mathbb{R}^2,0)\neq 0.
\end{align*}
Then, $\deg_{LS}(I-M(1,\cdot),B_{\rho}(0),0)\neq 0$. Hence, there exists  
$u\in B_{\rho}(0)$ such that  $M_1(u)=u$, which is a solution  for \eqref{equa1}.
\end{proof}

Let us give now an application of Theorem \ref{prin1}.

\begin{example} \label{examp3.6}\rm
We consider the boundary-value problem
\begin{equation}\label{exemplo}
\begin{gathered}
( \varphi(u'))' = e^{u'} +e \\
 u(T)=u'(0)=u'(T),
\end{gathered}
\end{equation}
where  $\varphi(s)=s/\sqrt{1-s^2}$.

It is not difficult to verify that $\varphi:(-1,1) \to \mathbb{R}$  is a  
homeomorphism and $f(t,x,y)=e^{y}+e$  is a continuous function. 
If we choose  $\rho> 2+T$, then the equation
\begin{align*}
G(x,y)
&=\Big(xT+yT^2-yT-\frac{1}{T}\int_{0}^{T} f(t,x+yt,y)dt, y-x-yT\Big)=(0,0)\\
&= \Big(xT+yT^2-yT-\frac{1}{T}\int_{0}^{T}(e^{y}-e)dt, y-x-yT\Big)=(0,0)\\
&= \Big(xT+yT^2-yT-e^{y}+e, y-x-yT\Big)=(0,0)
\end{align*}
does not have solutions on  $\partial B_{\rho}(0)\cap\mathbb{R}^2$. 
Then we have that the Brouwer degree 
$\deg_{B}(G,B_{\rho}(0) \cap \mathbb{R}^2,(0,0))$ is well defined and, 
by the properties of that degree, we have that
\[
 \deg_{B}(G, B_{\rho}(0)\cap\mathbb{R}^2, (0,0))
= \sum_{(x,y) \in G^{-1}(0,0)}\operatorname{sgn}J_{G}(x,y) 
= \operatorname{sgn}(-e)=-1,
\]
where $(0,0)$ is a regular value of $G$ and $J_{G}(x,y)=$det$ G'(x,y)$ 
is the Jacobian of $G$ at $(x,y)$. So, using Theorem \ref{prin1}, 
we obtain that  the boundary-value problem \eqref{exemplo} 
has at least one solution.
\end{example}

\begin{remark}\label{obs} \rm
Using the family of boundary-value problems
\begin{equation}\label{equa9}
\begin{gathered}
(\varphi(u' ))'   = \lambda N_{f}(u)+(1-\lambda)Q(N_{f}(u)) \\
u(0) = u'(0) = u'(T)
\end{gathered}
\end{equation}
which gives the completely continuous homotopy $\widetilde{M}$  defined on 
$[0,1]\times C^{1}$  by
\[
\widetilde{M}(\lambda,u)=  P(u)+Q(N_{f}(u)) 
+ H(\varphi^{-1} [\lambda H( N_f (u)-Q(N_{f}(u)))+\varphi(P(u))]),
\]
and similar a priori bounds as in the Lemma \ref{cota}, it is not difficult 
to see  that \eqref{equa9} has a solution  for $\lambda=1$.
\end{remark}

 \section{Boundary value problems with $\varphi$-Laplacian}
\label{S4}

 In this section we study the existence of solutions for the boundary-value problem
 \begin{equation}\label{equa2}
\begin{gathered}
(\varphi(u' ))'   = f(t,u,u')  \\
u(0) = u(T) = u'(0),
\end{gathered}
\end{equation}
where $\varphi: \mathbb{R} \to \mathbb{R}$ is  an increasing homeomorphism,  
$\varphi(0)=0$ and $f: [0, T]\times \mathbb{R} \times \mathbb{R}\to \mathbb{R}$ 
is continuous.
 
Let us consider the operator $M_1:C^{1} \to C^{1}$,
\[
M_1(u) \varphi^{-1}(-Q_{\varphi}(H(N_{f}(u)))) 
+ H(\varphi^{-1}  [H(N_{f}(u))-Q_{\varphi}(H(N_{f}(u)))]).
\]
As in the previous section, here $\varphi^{-1}$  with an abuse of notation 
is understood as the operator  $\varphi^{-1}: C \to C$ defined for 
$\varphi^{-1}(v)(t)=\varphi^{-1}(v(t))$. It is clear that $\varphi^{-1}$ 
is continuous and sends bounded sets into bounded sets.

To transform problem \eqref{equa2} to a fixed point problem  we use 
 Lemma \ref{lebema1}.

\begin{lemma}\label{dallos11}
  A map $u \in C^{1}$ is a solution of  \eqref{equa2} if and only if  $u$ 
is a fixed point of the operator $M_1$.
\end{lemma}

\begin{proof}
 If  $u\in C^{1}$ is solution of \eqref{equa2}, then
\[
(\varphi(u'(t)))' = N_f (u)(t)=f(t,u(t),u'(t)), \quad  
u(0)=u(T),\quad  u(0)=u'(0)
\]
for all $t\in [0, T]$. Applying  $H$ to both  members  and using the fact 
that  $u(0)=u'(0)$, we deduce that
\[
 \varphi(u'(t)) = \varphi(u(0)) + H(N_{f}(u))(t).
\]
Composing with the function $\varphi^{-1}$, we obtain
\[
u'(t) =\varphi^{-1} [ H(N_{f}(u))(t)+c],
\]
where $c=\varphi (u(0))$. Integrating from 0 to  $t\in [0, T]$, we  have
\[
  u(t)=u(0)+ H( \varphi^{-1}  [H(N_{f}(u))+c ])(t).
\]
Because  $u(0)=u(T)$, we have
\[
\int_0^T \varphi^{-1} [H(N_{f}(u))(t)+c]dt=0.
\]
Using Lemma  \ref{lebema1}, it follows that $c=-Q_{\varphi}(H(N_{f}(u)))$. 
Hence,
\[
u=\varphi^{-1}(-Q_{\varphi}(H(N_{f}(u))))
+ H ( \varphi^{-1} [H(N_{f}(u))-Q_{\varphi}(H(N_{f}(u)))] ) .
\]
Now suppose that $u\in C^{1}$ be such that $ u=M_1(u)$. It follows that
\begin{equation}\label{pof}
u(t)=\varphi^{-1}(-Q_{\varphi}(H(N_{f}(u))))
 + H( \varphi^{-1} [H(N_{f}(u))-Q_{\varphi}(H(N_{f}(u)))])(t)
\end{equation} 
for all $t\in [0, T]$. Since 
\[ 
\int_0^T \varphi^{-1} [H(N_{f}(u))(t)-Q_{\varphi}(H(N_{f}(u)))]dt=0,
\]
  we have  $u(0)=u(T)$. Differentiating  \eqref{pof}, we obtain 
\begin{align*}
u'(t)
&=\varphi^{-1} [H(N_{f}(u))-Q_{\varphi}(H(N_{f}(u)))](t)\\
&= \varphi^{-1} [H(N_{f}(u))(t)-Q_{\varphi}(H(N_{f}(u)))].%\\
\end{align*}
Applying   $\varphi$  to both of its members, and differentiating we have
\[
(\varphi(u'(t)))' = N_f (u)(t), \quad  u(0)=u(T), \quad u(0)=u'(0)
\]
for all $t\in [0, T]$. This completes the proof.
\end{proof}

Using an argument similar to the one introduced in the proof of
\cite[Lemma 4.2]{dallos}, it is not difficult to see that 
$M_1:C^{1} \to C^{1} $ is well defined and completely continuous.

 \subsection{Upper and lower solutions}
\label{S:0}

 The functions considered as lower and upper solutions for the initial 
problem  \eqref{equa2} are defined as follows.

\begin{definition} \label{def4.2} \rm
A lower solution $\alpha$ (resp. upper solution $\beta$) of \eqref{equa2} 
is a function $\alpha \in C^{1}$ such that
 $\varphi(\alpha')\in C^{1}, \  \alpha'(0)\geq \alpha(0)=\alpha(T)$ (resp. $\beta \in C^{1}, \  \varphi(\beta')\in C^{1}, \  \beta'(0)\leq \beta(0)=\beta(T))$ and
\begin{equation}\label{pas1}
(\varphi(\alpha'(t)))' \geq f(t,\alpha(t),\alpha'(t)) \quad
  (\text{resp. }  (\varphi(\beta'(t)))'
\leq f(t,\beta(t),\beta'(t))
\end{equation}
for all $t\in  [0,T]$.
\end{definition}
We can now prove  some existence results for \eqref{equa2}. 
These results are inspired on works by Bereanu and  Mawhin \cite{ma1} 
and Carrasco and Minh\'os  \cite{minhos}.

\begin{theorem}\label{gato}
Suppose that \eqref{equa2} has a lower solution $\alpha$ and  an upper solution 
$\beta$  such that $\alpha(t)\leq \beta(t)$ for all $t\in  [0,T]$. 
If there exists a continuous function $g(t,x)$ on $ [0,T]\times \mathbb{R}$ 
such that
\begin{equation}\label{pas2}
|f(t,x,y)|\leq |g(t,x)|,  \quad\text{for all } (t, x,y)\in   
[0,T]\times \mathbb{R}\times \mathbb{R},
\end{equation}
then \eqref{equa2} has a solution $u$ such that 
$\alpha(t)\leq u(t)\leq \beta(t)$ for all $t\in  [0,T]$.
\end{theorem}

\begin{proof}\label{gato1}
Let $\alpha$, $\beta$ be, respectively, lower and upper solutions  of  
\eqref{equa2}. Let $\gamma: [0,T]\times \mathbb{R}\to \mathbb{R}$ 
be the continuous function defined by
$$
\gamma(t,x) = \begin{cases}
 \beta(t), & x>\beta(t)\\
 x, & \alpha(t)\leq x\leq \beta(t) \\
 \alpha(t), & x <\alpha(t),
\end{cases}
$$
and define $F:  [0,T]\times \mathbb{R}\times \mathbb{R} \to \mathbb{R}$ by 
$F(t,x,y)= f(t,\gamma (t,x),y) +\frac{x-\gamma(t,x)}{1+|x-\gamma(t,x)|}$. 
We consider the modified problem
\begin{equation}\label{equa3}
\begin{gathered}
(\varphi(u' ))'   = F(t,u,u')  \\
u(0)= u(T)=u'(0).
\end{gathered}
\end{equation}
For clearness, the proof will follow several steps.
\smallskip

\noindent\textbf{Step  1} We  show that if  $u$ is a solution of \eqref{equa3}, 
then $ \alpha(t)\leq u(t)\leq \beta(t)$  for all $t\in  [0,T]$ and hence $u$ 
is a solution of \eqref{equa2}. 
Let $u$ be a solution of the modified problem \eqref{equa3} and suppose 
by contradiction that there is some $t_{0}\in  [0,T]$ such that
\begin{equation}\label{equa4}
 \max_{[0,T]}(\alpha(t)-u(t)) = \alpha(t_{0})-u(t_0)>0.
\end{equation}
If $t_{0}\in (0,T)$,  there are  sequences $(t_{k})$ in 
$ [t_{0}-\epsilon, t_{0})$ and $(t'_{k})$ in
 $(t_{0}, t_{0}+\epsilon ]$ converging to $t_{0}$ such that 
$ \alpha'(t_{k})-u'(t_{k})\geq 0$  and $\alpha'(t'_{k})-u'(t'_{k})\leq 0$. 
Therefore  $\alpha'(t_{0})=u'(t_{0})$. Using the fact that $\varphi$ is an 
increasing homeomorphism, this implies 
$(\varphi(\alpha'(t_{0}) ))'\leq  (\varphi(u'(t_{0}) ))'$. 
By  \eqref{pas1} we get the contradiction
\begin{align*}
(\varphi(\alpha'(t_{0}) ))' 
&\leq  (\varphi(u'(t_{0})))'  
= F(t_{0},u(t_{0}),u'(t_{0})) \\
&\leq f(t_{0},\alpha(t_{0}), \alpha'(t_{0}))) +\frac{u(t_{0})
 -\alpha(t_{0})}{1+|u(t_{0})-\alpha(t_{0}) |}  \\
&< f(t_{0},\alpha(t_{0}), \alpha'(t_{0}))) \leq (\varphi(\alpha'(t_{0}) ))'.
\end{align*}
So $\alpha(t)\leq u(t)$ for all $t\in (0,T)$.
If the maximum is attained at $t_{0}=0$ then
\[
  \max_{[0,T]}(\alpha(t)-u(t)) = \alpha(0)-u(0)>0.
\]
Using the fact  that $u$  is solution of \eqref{equa3} and $\alpha'(0)\leq u'(0)$, 
we obtain the following contradiction
\[
\alpha(0)\leq \alpha'(0)\leq u'(0)=u(0)<\alpha(0).
\]
If
\[
  \max_{[0,T]}(\alpha(t)-u(t)) = \alpha(T)-u(T)>0,
\]
 then, since  $\alpha(0)=\alpha(T)$  and   $u(0)=u(T)$ we obtain again a 
contradiction. In consequence we have that $\alpha(t)\leq u(t)$ for all 
$t\in  [0,T]$. In a similar way we can prove that $u(t)\leq \beta(t)$ for all 
$t\in  [0,T]$.
\smallskip

\noindent\textbf{Step  2}
 We  show that   problem \eqref{equa3}  has at least one solution.
 Let $u\in C^{1}$ and define the operator $M_{F}: C^{1} \to C^{1}$ by
\[
M_{F}(u)=\varphi^{-1}(-Q_{\varphi}(H(N_{F}(u)))) 
 + H(\varphi^{-1}  [H(N_{F}(u))-Q_{\varphi}(H(N_{F}(u)))]),
\]
with $F(t,u,u')=f(t,\gamma (t,u),u') +\frac{u-\gamma(t,u)}{1+|u-\gamma(t,u)|}$,
 by  Lemma \ref{dallos11} it is enough to
prove that $M_{F}$ has a fixed point. Under the hypothesis of the theorem, 
the operator $M_{F}$ is bounded. Indeed, if $v=M_{F}(u)$ then
\begin{equation}\label{pepa15}
\varphi(v') =  [H(N_F(u))- Q_{\varphi}(H(N_F(u)))],
\end{equation}
where
\begin{align*}
|H(N_F(u))(t)| 
&\leq  \int_0^T |f(s,\gamma(s,u(s)),u'(s)) 
 +\frac{u(s)-\gamma(s,u(s))}{1+|u(s)-\gamma(s,u(s))|}|ds  \\
&\leq \int_0^T |f(s,\gamma(s,u(s)),u'(s))|ds + T  \\
& \leq \int_0^T |g(s,\gamma(s,u(s)))|ds  + T\\
& \leq \sigma T +T,
\end{align*}
with $\sigma:=   \sup_{s\in[0,T]}|g(s,\gamma(s,u(s)))|$.
 Using  \eqref{pepa15}, we have that
\begin{equation}\label{pepa}
|\varphi(v'(t))|\leq 2(\sigma T +T):= \delta  \quad t\in  [0,T],
\end{equation}
and hence
\begin{equation}\label{pepa1}
\|v'\|_{\infty}\leq \omega,
\end{equation}
where $\omega= \max \{|\varphi^{-1}(\delta)|, \  |\varphi^{-1}(-\delta)|\}$. 
Because $v \in C^{1}$ is
such that $v(0) = v'(0)$, we have
\[
|v(t)|\leq  |v(0)| + \int_0^T |v'(s)|ds \leq  \omega+ T \omega   \quad 
t\in  [0,T],
\]
and hence
\[
\|v\|_1=\|v\|_{\infty}+ \|v'\|_{\infty}\leq  \omega +  T\omega + \omega 
=\omega(2+T).
\]
As the operator $M_{F}$ is completely continuous and bounded, we can use 
Schauder's fixed point theorem to deduce the existence of at least one 
fixed point in $\overline{B_{\omega(2+T)}(0)}$. The proof is complete.
\end{proof}

\begin{corollary}\label{pepa3}
Let $f(t,x,y)=f(t,x)$  be  a continuous function. If \eqref{equa2} has 
a lower solution $\alpha$ and a upper solution $\beta$ such that 
$\alpha(t)\leq \beta(t)$ for all $t\in  [0,T]$, then problem  
\eqref{equa2} has a solution such that  $\alpha(t)\leq u(t)\leq \beta(t)$ 
for all $t\in  [0,T]$.
\end{corollary}

\begin{example} \rm
We consider the boundary-value problem
\begin{equation}\label{pepa4}
\begin{gathered}
\big(|u'|^{p-2}u'\big)'=  \frac{\sin(u+1)-1 +4ue^{u^2t}}{1+t^2}, \\
u(0)=u(T)=u'(0),
\end{gathered}
\end{equation}
where $p\in (1,\infty)$. As $f(t,x)$ is a continuous function, 
and the functions $\alpha(t)=-1$ and $\beta(t)=1$ are lower and upper 
solutions of \eqref{pepa4}, respectively, then, by  Corollary \eqref{pepa3}, 
we obtain  that \eqref{pepa4} has at least one solution $u$ such that
$-1\leq u(t)\leq 1$  for all $t\in  [0,T]$.
\end{example}

\subsection*{Acknowledgements}
This research was supported by  CAPES and CNPq/Brazil.

\begin{thebibliography}{99}

\bibitem{man4} C. Bereanu, J. Mawhin; 
\emph{Boundary value problems for some nonlinear systems with singular
 $\varphi$-laplacian}, J. Fixed Point Theory Appl. 4 (2008), 57-75.

\bibitem{ma1} C. Bereanu, J. Mawhin; 
\emph{Periodic solutions of nonlinear perturbations of  $\varphi$-laplacians
 with possibly bounded $\varphi$}, Nonlinear  Anal. 68 (2008), 1668-1681.

\bibitem{mabe} C. Bereanu, J. Mawhin; 
\emph{Existence and multiplicity results for some nonlinear problems with 
singular $\varphi$-laplacian}, J. Differential Equations. 243 (2007), 536-557.

\bibitem{minhos} H. Carrasco, F. Minh\'os; 
\emph{Unbounded solutions for functional problems on the half-line}, 
Abstract and Applied Analysis. doi:10.1155/2016/8987374.

\bibitem{dallos} D. P. D. Santos; 
\emph{Existence of solutions for some nonlinear problems  with boundary 
value conditions},  Abstract and Applied  Analysis. doi:10.1155/2016/5283263.

\bibitem{man7} K. Deimling;
\emph{Nonlinear Functional Analysis}, Springer, Berlin, 1985.

\bibitem{garcia} M. Garc\'ia-Huidobro, R. Man\'asevich, F. Zanolin; 
\emph{A Fredholm-like result for strongly nonlinear second order ODE's}, 
J. Differential Equations. 114 (1994), 132-167.

\bibitem{grae} J. Graef, L. Kong, F. Minhos, J. Fialho; 
\emph{On the lower and upper solutions method for higher order functional
 boundary-value problems}, Applicable
Analysis and Discrete Mathematic. 5 (2011), 133-146.

\bibitem{huang} Y. X. Huang, G. Metzen; 
\emph{The existence of solutions to a class of semilinear equations},
 Differential Integral Equations. 8 (1995), 429-452.

\bibitem{chinos} R. Ma, L. Zhang, R. Liu; 
\emph{Existence results for nonlinear problems with $\varphi$-Laplacian},
 Electronic Journal of Qualitative Therory of Differential Equations. 22 (2015), 1-7.


\bibitem{man} R. Man\'asevich,  J. Mawhin; 
\emph{Periodic solutions for nonlinear systems with $p$-laplacian-like operators}, 
Differential Equations. 145 (1997), 367-393.

\bibitem{maw} J. Mawhin; 
\emph{Some boundary value problems for Hartman-type perturbations of the ordinary 
vector $p$-Laplacian}, Nonlinear Anal. 40 (2000), 497-503.

\bibitem{oregan} D. O'Regan; 
\emph{Some general principles and results for $(\varphi(y'))'=qf(t,y,y')$, 
 $0<t<1$}, SIAM J. Math. Anal. 24 (1993), 648-668.

\end{thebibliography}

\end{document}
