Simple bifurcation and global curves of solutions of \(p\)-Laplacian problems with radial symmetry (Q2358281): Difference between revisions

The author considers the boundary value problem \[ -(r^{N-1 }\phi_p(u'(r)))'= \lambda r^{N-1} f(r,u(r)), \] where \(r \in (0,1)\), \(N \geq 1\) is an integer, \(p \in {\mathbb R}\) satisfies \(2 \neq p > 1\), \(\phi_p(s)=|s|^{p-1} \mathrm{sign} s\), \(s \in {\mathbb R}\), \(\lambda \geq 0\). The boundary conditions are expressed as follows: if \(N=1\) \(BC_N(u)=(u(0),u(1))=(0,0)\) and if \(N > 1\) \(BC_N(u)=(u'(0),u(1))=(0,0)\). It is also assumed that the function \(f\) has the form \(f(r,\xi)=g(r,\xi) \phi_p(\xi)\), where \(g \in C^1\) and \(g_0(r)=g(r,0)>0\). The main novelty of the present paper consists in the fact that the author is able to treat any \(2 \neq p >1\), while in many papers it is assumed that \(p>2\). For this reason, a crucial difficulty in the proofs is the differentiability of the involved operators. In particular, it is first proved that \(\phi_p\) and \(S_p\), the unique solution of the problem \(-\Delta_p(u)=r^{N-1} h\), \(h \in L^1(0,1)\) are differentiable. Then, the author uses the fact that the given equation is equivalent to \[ -\Delta_p(u)=\lambda r^{N-1} g(u) \phi_p(u), (\lambda,u) \in {\mathbb R}\times {\mathcal D}_p, \] where \({\mathcal D}_p=\{u \in C^1: u \;\text{satisfies \;the \;boundary \;condition \;and} \;r^{N-1 }\phi_p(u')\in W^{1,1}(0,1) \}\); moreover, the latter equation can be written as \(F(\lambda,u)=0\), \((\lambda,u) \in {\mathbb R}\), with \(F(\lambda, u) = u - \lambda ^{p^*} S_p(g(u) \phi_p(u))\), \(p^*=1/p-1\) and \(Y_p=C^1[0,1]\) if \(1<p<2\) and \(Y_p=W^{1,1}(0,1)\) if \(p>2\). Then, the differentiability of \(F\) at solutions of \(F(\lambda,u)=0\) is proved. The main result on simple bifurcation reads as follows: Let \(v_k\) be the \(k\)th eigenvalue of the problem \(-\Delta_p(u)=\lambda_k r^{N-1 } g_0 \phi_p(v_k)\) and let \(Z_p=\{z \in Y_p: \int_0^1 r^{N-1}g_0 \phi_p(v_k) z =0\}\). Then, there exist \(\epsilon >0\), a neighbourhood \(U_p\) of \((\lambda_k,0)\) and a \(C^1\) map \(s \mapsto (\lambda(s),z(s))\) from \((-\epsilon,\epsilon) \to {\mathbb R}\times Z_p\) such that \((\lambda(0),z(0))=(\lambda_k,0)\) and \(\{(\lambda,u) \in U_p : u \neq 0, -\Delta_p(u)=\lambda r^{N-1 } g(u) \phi_p(u) \}\) = \(\{(\lambda (s), s(v_k+z(s)) : 0 \neq s\in (-\epsilon, \epsilon) \}\). As a corollary, a result on the nodal properties of the solutions obtained in the main result is proved. Finally, smooth curves of positive solutions are obtained as well. The proof is performed using the continuation approach, which relies on the implicit function theorem.