Singular solutions for divergence-form elliptic equations involving regular variation theory: existence and classification (Q1679721)

Let \(N\geq 2\) and \(p>1\). This paper deals with the classification of the singularities of positive solutions to the nonlinear equation \[ \mathrm{div}\big(\mathcal{A}(|x|)|\nabla u|^{p-2}\nabla u\big)=b(x)h(u) \eqno(1) \] in the punctured unit ball \(B^*:=B_1\setminus \{0\}\) of \(\mathbb{R}^N\). The function \(\mathcal{A}:(0,1]\rightarrow \mathbb{R}\) is assumed \(C^1\) and positive in \((0,1]\), and satisfying \[ \lim_{t\rightarrow 0^+}tL'_{\mathcal{A}}(t)(L_{\mathcal{A}}(t))^{-1}=0, \] where \(L_{\mathcal{A}}(t):=\mathcal{A}(t)t^{-\vartheta}\), with \(\vartheta>p-N\). The functions \(b:\overline{B_1}\setminus \{0\}\rightarrow \mathbb{R}\) and \(h:\mathbb{R}\rightarrow \mathbb{R}\) are assumed continuous, with \(b\) positive on \(\overline{B_1}\setminus \{0\}\), \(h\) positive in \((0,+\infty)\), \(h(0)=0\), and \(h(t)t^{1-p}\) bounded for \(t>0\) small. Moreover, \(h,b\) satisfy the following further condition: There exist two functions \(L_h\), \(L_b\), slowly varying at \(\infty\) and at \(0\), respectively, such that \[ \lim_{t\rightarrow +\infty}h(t)(t^qL_{h}(t))^{-1}=1, \quad\lim_{|x|\rightarrow 0}b(x)(|x|^\sigma L_{b}(|x|))^{-1}=1, \] where the exponents \(q,\sigma\in \mathbb{R}\) satisfy \(q+1>p>\vartheta-\sigma\). Let \[ \Phi(x)=(N|B_1|)^{-\frac{1}{p-1}}\int_{|x|}^1\bigg(\frac{t^{1-N-\vartheta}}{L_{\mathcal{A}(t)}}\bigg)^{\frac{1}{p-1}}dt \] be the fundamental solution of \[ \mathrm{div}\big(\mathcal{A}(|x|)|\nabla \Phi|^{p-2}\nabla \Phi\big)=\delta_0\quad \text{in}\quad\mathcal{D}'(B_1),\qquad\Phi=0\quad\text{on} \quad\partial B_1, \] where \(\delta_0\) is the Dirac mass at \(0\). The first main result (Theorem 1.1) gives a classification of the singularities at \(0\) of the positive solutions \(u\) of equation \((1)\), by examining the following two possible cases: \(b(x)h(\Phi)\in L^1(B_{1/2})\) and \(b(x)h(\Phi)\notin L^1(B_{1/2})\). This classification is given in terms of weak singularities, strong singularities, and removable singularities which essentially correspond to the situations \[ \lim_{|x|\rightarrow 0}\frac{u(x)}{\Phi(x)}\in (0,+\infty), \quad \lim_{|x|\rightarrow 0}\frac{u(x)}{\Phi(x)}=+\infty, \quad \lim_{|x|\rightarrow 0}\frac{u(x)}{\Phi(x)}=0, \] respectively. The authors prove that \(u\) can have any singularity of these three different kinds when \(b(x)h(\Phi)\in L^1(B_{1/2})\), and that \(u\) can have only a removable singularity when \(b(x)h(\Phi)\notin L^1(B_{1/2})\). The second main result (Theorem 1.2) gives conditions for the existence and uniqueness of positive solutions to equation (1) with prescribed behavior at \(0\), and subjected to the boundary condition \(u=g\) on \(\partial B_1\), where \(g\in C^1(\partial B_1)\) is a nonnegative function. In particular, by this result, one gets the existence of positive solutions having at \(0\) any of the singularities defined above. In the paper some previous results are extended and sharpened. Moreover, in the case of the critical exponent \(q_*=\frac{(N+\sigma)(p-1)}{N+\vartheta-p}\), the explicit behavior of the positive solutions at \(0\) given by Theorem 1.1 is a new result even for the \(p\)-Laplacian equation (that is, for \(\mathcal{A}=1\)).