Multiplicity results for nonlinear periodic fourth order difference equations with parameter dependence and singularities (Q990834)

First I must introduce some notations of the authors. Let \(T\in {\mathbb N}, I=\{0,\dots,T-1\}, J=\{0,\dots,T+3\}.\) If \(P\) is a bounded subset of \({\mathbb Z}\) and \(\xi:P\to{\mathbb R}\) then \[ \xi_*=\min_{j\in P}\{\xi(j)\},\quad \xi^*=\max_{j\in P}\{\xi(j)\},\quad \|\xi\|=(| \xi|)^*. \] The authors use also the maps \(g:I\to{\mathbb R},\,c:I\to{\mathbb R}\) and a number \(\lambda>0.\) They denote by \(f\) a continuous map \(f:(0+\infty)\to(0+\infty)\) or a continuous map \(f:[0+\infty)\to[0+\infty).\) They consider also \(M\in{\mathbb R},\) such that the problem \[ u(k+4)+Mu(k)=0,\;k\in{\mathbb N};\quad u(T+i)=u(i),\,i=0,\dots,3\tag{1.2} \] is non-resonant (i.e. it has only trivial solution) and the related Green's function has constant sign. The authors consider the following problem: \[ \begin{cases} u(k+4)+Mu(k)=h(k),\quad k\in I \\u(T+i)=u(i),\quad i=0,\dots,3,\end{cases}\tag{1.1} \] where \(h(k)=\lambda g(k)f(u(k))+c(k),\) with \(k\in I.\) In section 2 they consider the problem (1.1) with arbitrary \(h(k),\) and study the Green function \(G_M:J\times I\to{\mathbb R}\) such that \[ u(k)=\sum_{j=0}^{T-1} G_M(k,j)h(j),\, k\in J. \tag{2.2} \] The authors collect results of their study of the function \(G_M\) in the following (we cite) ``\textbf{Proposition 2.1.} The related Green's function \(G_M:J\times I\to{\mathbb R}\) of problem (1.2) satisfies the following properties: 1. \(G_M\) is strictly negative on \(J\times I\) if and only if \(T\) is odd and \(M<-1,\) 2. \(G_M\) is non-positive on \(J\times I,\) not identically zero on \(J\times I\) and vanishes on some points \(J\times I\) if and only if \(T\) is even and if \(M<-1,\) 3. \(G_M\) is strictly positive on \(J\times I\) if and only one of the two following assertions holds: (a) \(T\) is odd and \(M\in(-1,0),\) (b) \(T=1\) and \(M\geq0,\) 4. \(G_M\) is non-negative on \(J\times I,\) not identically zero on \(J\times I\) and vanishes on some points \(J\times I\) if and only if one of the three following assertions holds: (a) \(T\) is even and \(M\in(-1,0)\) (b) \(T\geq2\) \(M=0\) (c) \(T=2,\,4\) and \(M>0.\)'' In section 3 the authors define for each \(\lambda>0\) the operator \[ {\mathcal T}_\lambda:{\mathcal D}({\mathcal T}_\lambda)\equiv \{u(k)>0\,\,\text{for all}\,\,k\in J\}\to{\mathbb R}^{T+4} \] given by \[ {\mathcal T}_\lambda u(j)=\left(\sum_{j=0}^{T-1}G_M(k,j)g(j)f(u(j))\right)+\left(\sum_{j=0}^{T-1}G_M(k,j)c(j)\right) \] and remark that \(u>0\) is a solution of the problem (1.1) if and only if \(u=T_\lambda u.\) They introduce the following notations: \[ \gamma(k)=\sum_{j=0}^{T-1}G_M(k,j)c(j),\,k\in J, \] \[ \alpha=\min_{(k,j)\in J\times I} G_M(k,j),\quad \beta=\max_{(k,j)\in J\times I}G_M(k,j), \] \[ f_0=\lim_{u\to 0^+}f(u),\quad f_\infty=\lim{u\to +\infty}f(u) \] (if they exist). Further they make the following assumptions: \[ \gamma_*\geq0.\tag{\(H0\)} \] \[ \text{Either \(T\) is odd and \(M\in(-1,0)\) or \(T=1\) and \(M\geq0\).}\tag{\(H1\)} \] \[ g(j)\geq0\text{ for all }j\in I \text{ and }\sum_{j=0}^{T-1}g(j)>0. \tag{\(H2\)} \] \[ f:(0+\infty)\to(0+\infty)\text{ is continuous.}\tag{\(H3\)} \] It follows from Proposition 2.1 and \((H1)\) that \(\alpha>0.\) The authors define the cone \[ {\mathcal K}:=\bigg\{u\geq0,\,\text{on}\,\,J,\, \min_{k\in J}\geq\sigma\| u\|\bigg\}, \] where \[ \sigma=\min\left(\frac\alpha\beta,\, \frac{\gamma_*}{\gamma^*}\right), \] if \(\|\gamma\|>0,\) and \(\sigma=\frac\alpha\beta,\) if \(\|\gamma\|=0.\) After that, for \(0<r<R\), they define \[ {\mathcal K}_{r,R}=\{u\in{\mathcal K}: r\leq\|\gamma\|\leq R\}, \] and, as sufficient condition for the solvability of problem (1.1), declare the following (we cite) ``\textbf{Theorem 3.1.} Assume that condition \((H0), (H1), (H2)\) and \((H3)\) are fulfilled. Then for each \(\lambda>0\) and \(0<r<R\) the operator \({\mathcal T}_\lambda\) is well defined and completely continuous. Moreover, if either (i) \(R\geq\|{\mathcal T}_\lambda u\|\) for all \(u\in{\mathcal K}\) with \(\| u\|=R,\) and \(r\leq \|{\mathcal T}_\lambda u\|\) for all \(u\in{\mathcal K}\) with \(\| u\|=r,\) or (ii) \(R\leq\|{\mathcal T}_\lambda u \|\) and for all \(u\in{\mathcal K}\) with \(\| u\|=R,\) \(r\geq \|{\mathcal T}_\lambda u\|\) for all \(u\in{\mathcal K}\) with \(\| u\|=r,\) then there exists \(u\) in \({\mathcal K}_{r,R}\) such that \({\mathcal T}_\lambda (u)=u.\)'' Unfortunately, the authors do not give a detailed proof of this Theorem, and make only the following statement (we cite): ``The proof is analogous to the one given in [6, Teorem 2.1] and follows from classical expansion/contraction Krasnoselskii's fixed Teorem.'' Reviewer's remark: [6] is the paper by \textit{A. Cabada} and \textit{J. A. Cid}, ``On the sign of the Green's function associated to Hill's equation with an indefinite potential'', Appl. Math. Comput. 205, No. 1, 303--308 (2008; Zbl 1161.34014). Assertions, which have been deduced by the authors on the base of this Theorem 3.1 are collected in the following Theorem 3.2 (Parts 1 -- 3) and Theorem 3.3. (Parts 1 -- 6). \textbf{Theorem 3.2.} Assume that conditions \((H1)\), \((H2)\) and \((H3)\) are fulfilled. If moreover \(\gamma_*>0\) the following results hold: Part 1. There exists \(\lambda>0\) such that problem (1.1) has a positive solution for all \(\lambda\in(0,\lambda_0)),\) Part 2. If \(f_\infty=0\) then problem (1.1) has a positive solution for all \(\lambda>0,\) Part 3. If \(f_\infty=+\infty\) then there exists \(\lambda>0\) such that problem (1.1) has two positive solutions for all \(\lambda\in(0,\lambda_0),\) Part (4). If \(f_0>0\) and \(f_\infty>0\) then there exists \(\lambda>0\) such that problem (1.1) has no positive solution for all \(\lambda>\lambda_0.\) \textbf{Theorem 3.3.} Assume that conditions \((H1), (H2)\) and \((H3)\) are fulfilled. If moreover \(c\equiv0\) on \(I\) the following results hold: Part 1. If \(f_0=+\infty\) or \(f_\infty=+\infty\) then there exists \(\lambda>0\) such that problem (1.1) has a positive solution for all \(\lambda\in(0,\lambda_0),\) Part 2. If \(f_\infty=0\) then there exists \(\lambda>0\) such that problem (1.1) has a positive solution for all \(\lambda>\lambda_0,\) Part 3. If \(f_0=+\infty\) and \(f_\infty=0\) then problem (1.1) has a positive solution for all \(\lambda>0,\) Part 4. If \(f_0=+\infty\) and \(f_\infty=+\infty\) then there exists \(\lambda>0\) such that problem (1.1) has two positive solutions for all \(\lambda\in(0,\lambda_0),\) Part 5. If \(f_0=0\) and \(f_\infty=+\infty\) then there exists \(\lambda>0\) such that then problem (1.1) has a positive solution for all \(\lambda>0,\) Part 6. If \(f_0=0\) and \(f_\infty=0\) then there exists \(\lambda>0\) such that problem (1.1) has two positive solutions for all \(\lambda>\lambda_0),\) Part 7. If \(f_0>0\) and \(f_\infty>0\) then there exists \(\lambda>0\) such that problem (1.1) has no positive solution for all \(\lambda>\lambda_0).\) As illustration authors consider the problem (1.1) with \(f(x)=x^{-\delta}\) (the authors denote this specification as \textbf{Problem (3.5)}) and make the following assertions: 1. If \(\delta>-1\) then problem (3.5) has a positive solution for all \(\lambda>0,\) 2. If \(\delta=-1\) then there exists \(\lambda>0\) such that problem (3.5) has no positive solution for all \(\lambda>\lambda_0).\) 3. If \(\delta<-1\) and \(\gamma_*>0\) then there exists \(\lambda>0\) such that problem (3.5) has two positive solutions for all \(\lambda\in(0,\lambda_0),\) 4. If \(\delta<-1\) and \(c\equiv0\) on \(I\) then problem (3.5) has a positive solution for every \(\lambda>0.\) In section 4, the authors consider the case when the Green's function is non-negative on \(J\times I.\) Instead of condition \(H(1)\) they assume \[ \text{Either \(T\) is even and \(M\in(-1,0)\) or \(T\geq2\) and \(M=0\) or \(T=2,\,4\) and \(M>0\).}\tag{\(\widetilde{H1}\)} \] In this case they introduce the following cone: \[ \widetilde{{\mathcal K}}:=\bigg\{u\geq0,\,\text{on}\,\,I,\, \sum_{k=0}^{T-1}u(k)\geq\widetilde{\sigma}\| u\|\bigg\}, \] where \[ \widetilde{\sigma}=\min\left(\frac1{\beta(1+M)} \frac1{\|\gamma\|}\sum_{k=0}^{T-1}\gamma_k\right), \] if \(\|\gamma\|>0,\) and \(\widetilde{\sigma}=\frac1{\beta(1+M)},\) if \(\|\gamma\|=0.\) The authors suppose in this section that \[ f:[0,+\infty)\to[0,+\infty)\text{ is continuous and }f(u)>0\text{ for all }u.\tag{\(\widetilde{H3}\)} \] Further the authors for \(0<r<R\) introduce \[ \widetilde{\mathcal K}_{r,R}= \left\{u\in\widetilde{\mathcal K}\colon r\leq\| u\|\leq R\right\} \] and formulate the following (we cite) ``\textbf{Theorem 4.1.} Assume that condition \((H0),(\widetilde{H1}),(H2)\) and \((\widetilde{H3})\) are fulfilled. Then for each \(\lambda>0\) and \(0<r<R\) the operator \({\mathcal T}_\lambda: \widetilde{\mathcal K}_{r,R}\to\widetilde{\mathcal K}\) given by (3.1) is well defined and completely continuous. Moreover, if either (i) \(R\geq\|{\mathcal T}_\lambda u\|\) for all \(u\in\widetilde{\mathcal K}\) with \(\| u\|=R,\) and \(r\leq \|{\mathcal T}_\lambda u\|\) for all \(u\in\widetilde{{\mathcal K}}\) with \(\| u\|=r,\) or (ii) \(R\leq\|{\mathcal T}_\lambda u \|\) and for all \(u\in\widetilde{\mathcal K}\) with \(\| u\|=R,\) \(r\geq \|\widetilde{{\mathcal T}}_\lambda u\|\) for all \(u\in\widetilde{{\mathcal K}}\) with \(\| u\|=r,\) then there exists \(u\) in \(\widetilde{{\mathcal K}}_{r,R}\) such that \({\mathcal T}_\lambda (u)=u.\)'' Again the authors do not give a detailed proof of this Theorem, and make only the following statement (we cite): ``We only verify that operator \({\mathcal T}_\lambda\) maps \(\widetilde{{\mathcal K}}_{r,R}\) into \(\widetilde{{\mathcal K}}.\) The rest of the proof follows similar arguments to the ones in [6, Teorem 3.1]\dots'' Let me offer the following counterexample to this ``Theorem 4.1'' now. \textbf{Counterexample 1.} Let \(T=2, \mu=-M\in[0,1).\) Then \(u(k+4)-\mu u(k)=h(k)\) for \(k=0,\,1\) , \(u(i)=u(t+i)\) for \(i=0,\,\dots,\,3.\) Therefore \(u(4)=u(2)=u(0)=\frac1{1-\mu}h(0),\,u(5)=u(3)=u(1)=\frac1{1-\mu}h(0)\). In this case \(I=\{0,\,1\}\), \(J=\{0,\,\dots,\,5\},\) \[ G_M=\left(\begin{matrix} G_M(0,0)&G_M(0,1)\\G_M(1,0)&G_M(1,1)\\G_M(2,0)&G_M(2,1)\\ G_M(3,0)&G_M(3,1)\\G_M(4,0)&G_M(4,1) \\G_M(5,0)&G_M(5,1) \end{matrix}\right)=\frac1{1-\mu}\left(\begin{matrix} 1&0\\0&1\\1&0\\0&1\\1&0 \end{matrix}\right), \] \(\alpha=0\) and \(\beta=1.\) Further we have \[ {\mathcal T}_\lambda u(k)=\lambda g(k)u(k)+\gamma(k), \] where \[ u(2k+i)=u(i)\,\,\text{for}\,\,i=0,\,1,\,k=0,1,2 \] \[ g(2k+i)=g(i)\,\,\text{for}\,\,i=0,\,1,\,k=0,1,2 \] We give now a counterexample to ``Theorem 4.1'' in the case, when \(T=2,M\in(-1,0),\gamma^*=0,\) and the sufficient condition (ii) of this ``Theorem'' is fulfilled. We take \(\mu\in(0,1/2),\eta=\mu/(1-\mu)\in(0,1).\) Then \(1+\eta=1/(\beta(1-\mu))=\tilde{\sigma}.\) If \(u\in\tilde{{\mathcal K}}\), then \[ u(j)+u(1-j)\geq\tilde{\sigma}| u|=(1+\eta)| u|\geq (1+\eta)u(j) \] and \[ u(1-j)\geq\eta u(j) \] for \(j=0,1.\) So, the inequality \(u(0)+u(1)\geq\tilde{\sigma}| u|\) is equivalent to condition \(u(j)/u(k)\geq\eta,\,\{j,k\}\subset\{0,1\}\) We take further \[ b_0>0,r>b_0,\,R>r/\eta,\,B_0\in(0,R), g_0=\lambda=1,\, g_1>\frac r{b_0\eta}. \] We construct an auxiliary continuous piecewise linear function \[ f(x):[0,+\infty)\to[0,+\infty), \] as follows: \[ f(x)=\frac {(1-\mu)b_0}rx\,\,\text{for}\,\,x\in[0,r], \] \[ f(x)=(1-\mu)b_0\frac{R\eta-x}{R\eta-r}+\frac{1-\mu}{g(1)}B_0\frac{x-r}{R\eta-r} \] for \(x\in[r,R\eta],\) \[ f(x)=\frac{1-\mu}{g(1)}B_0\,\,\text{for}\,\,x\in[R\eta,+\infty). \] If \(\| u\|=r\) then \(u(i)\in[r\eta,r]\) for \(i=0,1.\) Therefore \[ {\mathcal T}_1u(0)=g(0)f(u(0))/(1-\mu)\leq b_0< r, \] \[ (1-\mu)b_0\geq f(u(1))=(1-\mu)u(1)b_0/r\geq(1-\mu)b_0\eta, \] \[ {\mathcal T}_1u(1)=g(1)f(u(1))/(1-\mu)\geq g(1)b_0\eta>r, \] and, hence, \[ \|{\mathcal T}_1u\|>r. \] If \(\| u\|=R\) then \[ u(i)\in[R\eta,R],\,f(u(i))=\frac{1-\mu}{g(1)}B_0 \] for \(i=0,1.\) Therefore \[ {\mathcal T}_1u(0)=f(u(0))/(1-\mu)=B_0/g(1)<R, \] \[ {\mathcal T}_1u(1)=g(1)f(u(1))=B_0<R, \] and hence \[ \| {\mathcal T}_1u\|<R. \] Clearly, \({\mathcal T}_1u(0)<u(0),\) if \(u(0)>0.\) But \(u\not\in\widetilde{\mathcal K}_{r,R},\) if \(u(0)=0.\) \(\blacksquare\) In the rest of this review I bound myself by transferring further assertions of the authors. \textbf{Theorem 4.2.} Assume that conditions \((H0)\), \((\widetilde{H1})\), \((H2)\) and \((\widetilde{H3})\) are fulfilled. If moreover \[ \sum_{j=0}^{T-1} g(j)>0, \] (reviewer's remark: this inequality contains in \((H2)\)) then the following results hold: 1. There exists \(\lambda>0\) such that problem (1.1) has a non-negative solution for all \(\lambda\in(0,\lambda_0),\) 2. If \(f_\infty=0\) then problem (1.1) has a non-negative solution for all \(\lambda>0.\) Further the authors, instead of condition \((H2)\), impose a stronger condition on the function \(g\): \[ g_*>0\tag{\(\widetilde{H2}\)} \] \textbf{Theorem 4.3.} Assume that conditions \((H0)\), \((\widetilde{H1}),\, (\widetilde{H2})\) and \((\widetilde{H3})\) are fulfilled. Then following assertions are satisfied: 1. If \(f_\infty=\infty\) then there exists \(\lambda>0\) such that problem (1.1) has a non-negative solution for all \(\lambda\in(0,\lambda_0),\) 2. If \(f_\infty=\infty\) and \(f_\infty=0\) then problem (1.1) has a non-negative solution for all \(\lambda>0.\) 3. If \(f_0>0\) and \(f_\infty>0\) then there exists \(\lambda_0>0\) such that problem (1.1) has no non-negative solution for all \(\lambda>\lambda_0).\) \textbf{Theorem 4.4.} Assume that \(\delta>0\) and that \((H0)\), \((\widetilde{H1})\), \((\widetilde{H2})\) are fulfilled. Let \[ \Lambda(k)=\sum_{j=0}^{T-1}G_M(k,j)g(j)\,\,\text{for all}\,\,k\in I, \tag{4.1} \] and suppose that there exists some \(r>0\) such that the following inequalities hold \[ \frac{\lambda\Lambda_*} {\left(\frac{\lambda\Lambda^*}{r^\delta}+\gamma^*\right)^\delta}+\gamma_*\geq r \tag{4.2} \] and \[ \max(\lambda\Lambda^*/r^\delta+\gamma_*) ,\,\lambda\Lambda_*/r^\delta+\gamma^*)\geq r.\tag{4.3} \] Then problem (3.5) has a positive solution \(u\) such that \(u(k)\geq r\) for all \(k\in J.\) \textbf{Corollary 4.1.} Suppose that conditions \((H0)\), \((\widetilde{H1})\) and \((\widetilde{H2})\) are fulfilled. If \(\delta>0,\,\gamma_*>0\) then problem (3.5) has a positive solution for all \(\lambda>0.\) Moreover, such solution is such that \(x(t)\geq\gamma_*\) for all \(t\in[0,T].\) \textbf{Corollary 4.2.} Assume that \(\delta>0,\,\gamma_*=0\) and that conditions \((H0)\), \((\widetilde{H1}), (\widetilde{H2})\) are fulfilled. Then problem (3.5) has a positive solution for all \(\lambda>0.\) In the end of section 4, the authors make two following remarks: \textbf{Remark 4.1.} Notice that if \(\gamma_*=0,\) conditions (4.2) and (4.3) are satisfied if and only if \(c\equiv0\) on \(I\) and function \(\Lambda\) is a constant. Therefore in the case \(\gamma_*=0\) Theorem 4.4 has not non-trivial applications for strong singularities (\(\delta\geq1\)). \textbf{Remark 4.2.} The results given in this section can be applied for problem (3.5) when \(\delta\leq0.\) In this case we have a regular function, which trivially satisfies condition \((\widetilde{H3}).\) Assuming that conditions \((H0), (H1)\) and \((H2)\) are fulfilled, as a consequence of results shown in this section, we arrive at the following assertions: 1. If \(\delta\leq0\) and \(\sum_{j=0}^{T-1}g(j)>0\) then there exists \(\lambda_0>0\) such that problem (3.5) has a non-negative solution for all \(\lambda\in(0,\lambda_0)),\) 2. If \(-1<\delta\leq0\) then problem (1.1) has a non-negative solution for all \(\lambda>0,\) 3. If \(\delta=-1\) then there exists \(\lambda_1>0\) such that problem (1.1) has no non-negative solution for all \(\lambda>\lambda_1).\) In section 5, the authors consider the case when the Green's function for the problem (1.2) is non-positive on \(J\times I\). In this case, instead of problem (1.1) authors study the following problem: \[ \begin{cases} y(k+4)-Ny(k)=-h(k),\quad k\in I \\ u(T+i)=u(i),\quad i=0,\dots,3\end{cases},\tag{5.1} \] with the same \(h(k)\) as in problem (1.1), and \[ N=-M\tag{*)(1} \] with \(M\) considered in problem (1,1). The authors denote by \(G^-_N\) the Green's function to the problem \[ -y(k+4)+Ny(k)=0,\,k\in I, u(T+i)=u(i),\,i=0,\,\dots,\,3\tag{5.2} \] Reviewer's remark: Clearly, \[ G^-_N=-G_M.\tag{*)(2} \] After that, making use of (*)(1) and (*)(2), authors transform Proposition 2.1. into the following Proposition 5.1: \textbf{Proposition 5.1.} The related Green's function \(G^-_N:J\times I\to{\mathbb R}\) of problem (5.2) satisfies the following properties: 1. \(G^-_N\) is strictly positive on \(J\times I\) if and only if \(T\) is odd and \(N>1,\) 2. \(G^-_N\) is non-negative on \(J\times I,\) not identically zero on \(J\times I\) and vanishes on some points \(J\times I\) if and only if \(T\) is even and if \(N>1,\) 3. \(G^-_N\) is strictly negative on \(J\times I\) if and only one of the two following assertions holds: (a) \(T\) is odd and \(N\in(0,1),\) (b) \(T=1\) and \(N\leq0,\) 4. \(G^-_N\) is non-positive on \(J\times I,\) not identically zero on \(J\times I\) and vanishes on some points \(J\times I\) if and only if one of the three following assertions holds: (a) \(T\) is even and \(N\in(0,1)\) (b) \(T\geq2\) \(N=0\) (c) \(T=2,\,4\) and \(N<0.\) Finally, the authors make the following assertions: (we cite) ``So replacing \((H1)\) by the following one: \[ T\,\,\text{is odd and}\,\, N>1\tag{\(\overline{H1}\)} \] we conclude that all assertions of Theorem 3.2 and 3.3 remain valid for problem (5.1). The same comment is valid if we refer to theorems 4.2 and 4.3 concerning problem (5.1) and replacing condition \((\widetilde{H1})\) by \[ T\,\,\text{is even and}\,\,N>1.\tag{\(\widetilde{H1}^*\)} \] If we consider problem \[ \begin{cases} -y(k+4)+Ny(k)=\lambda g(k)u^{-\delta}(k),\quad k\in I \\u(T+i)=u(i),\quad i=0,\dots ,3\end{cases} \] we deduce similar results to Example 3.1, Theorem 4.4 and Corollaries 4.1 and 4.2 under this new assumptions.'' Conclusion. In my opinion, the paper contains an essential defect. Nevertheless it is interesting.