A refined subsolution estimate of weak subsolutions to second order linear elliptic equations with a singular vector field (Q498639): Difference between revisions

In this paper, the author considers second order linear elliptic equations of divergence type: \[ \begin{aligned} -\operatorname{div}(A(x)\triangledown u)& +b(x)\triangledown u \\ =-\sum_{i,j=1}^n\partial_{j}(a_{ij}(x)\partial_{i}u) & +\sum_{i=1}^n{}b_{i}(x)\partial_{i}u=0 \in \Omega,\end{aligned} \tag{DE} \] where \(\Omega\) is a domain in \(\mathbb R^n\), \(n\geq 3\), \(A(x)=(a_{ij})_{1\leq i,j\leq n}\) is measurable and satisfies the uniform ellipticity condition, i.e. there exist positive constants \(0<\nu\leq L<\infty\) such that \[ |a_{ij}(x)|\leq L,\sum_{i,j=1}^n a_{ij}(x)\xi _{i}\xi _{j}\geq \nu|\xi |^2,\forall\xi \in \mathbb R^n,\forall x\in \Omega. \tag{A} \] The author assumes that the vector field \(b(x)=(b_{i}(x))_{1\leq i\leq n}\) belongs to \(L_{\mathrm{loc}}^2(\Omega)\). Moreover the following conditions for the vector field b are assumed: Condition (B). A vector field \(b\in L_{\mathrm{loc}}^2(\Omega)\) can be represented as \(b=b^{(1)}+b^{(2)}+b^{(3)}+b^{(4)}\) and each \(b^{(i)}\in L_{\mathrm{loc}}^2(\Omega),i=1,4\), satisfies the following conditions: {\parindent=6mm \begin{itemize} \item[1.] \(b^{(1)}\) belong to some Lorentz space \(L^{n,q}(\Omega)\) with \(n\leq q<\infty\); \item [2.] \(b^{(2)}\) is small relative to the lower bound \(\nu\) of (A) in the following sense: there exists a constant \(B_2=B_2(\Omega)<\nu\) such that \[ \int_{\Omega}|b^{(2)} |{}^2\phi^2 dx\leq (B_2)^2\int_{\Omega}|\triangledown\phi |{}^2 dx,\forall\phi \in C_{c}^{\infty}(\Omega). \] \item [3.] \(b^{(3)}\) satisfies form boundedness condition and \(\operatorname{div} b^{(3)}\leq 0\) in the distribution sense, i.e. there exists a constant \(B_3=B_3(\Omega)<\infty\) such that \[ \int_{\Omega}|b^{(3)}|^2\phi^2 dx\leq (B_3)^2\int_{\Omega}|\triangledown\phi |{}^2 dx,\forall\phi \in C_{c}^{\infty}(\Omega), \] \[ \int_{\Omega}b^{(3)}\triangledown\phi dx\geq 0,\forall\phi \in C_{c}^{\infty}(\Omega),\phi \geq 0. \] \item [4.] \(b^{(4)}=(b_{i}^{(4)})_{1\leq i\leq n}\) can be written in the form \(b_{i}^{(4)}=\sum_{j=1}^n\partial_{j}V_{ij}\) in the distribution sense, where \(V=(V_{ij})\) satisfies \(V_{ij}=-V_{ji}\) and \(V_{ij}\in \text{BMO}(\Omega)\), where \(\text{BMO}(\Omega)\) consists of all locally integrable functions \(f\) satisfying \(\| f\|{}_{\text{BMO}(\Omega)}<\infty\). Define \(\| V\|{}_{\text{BMO}(\Omega)}=\sum_{i,j}\| V_{ij}\|{}_{\text{BMO}(\Omega)}\). \end{itemize}} The purpose of this paper is to establish a refined subsolution estimate and a weak Harnack inequality for weak subsolutions and weak supersolutions without using the approximating procedure on the vector field b. The paper is organized as follows: First, by using the properties of the form \(\int b\cdot \triangledown uvdx\) (Lemma 7 and Lemma 8) the author proves the Cacciopoli type inequality when \(b^{(1)}\) is small enough (Lemma 1). Also, he gets the subsolution estimate when \(b^{(1)}\) is sufficient small (Lemma 9) using this. Next, using the weak maximum principle and Lemma 2, the author proves the subsolution estimate without smallness of \(b^{(1)}\) (Theorem 1). Finally, using Lemma 2 and the subsolution estimate, the author obtains the weak Harnack inequality (Theorem 2).