The maximum Markovian self-adjoint extensions of generalized Schrödinger operators (Q1187569)

Let \(G\) be an open subset of \(\mathbb{R}^ d\), let \(\lambda^ d\) be \(d\)- dimensional Lebesgue measure on \(G\) and let \(\rho\) be a Borel measurable function defined on \(G\) which is \(\lambda^ d\)-almost everywhere strictly positive, which is locally square integrable and which possesses the property that the partial derivatives \(\nabla_ j\rho\), \(1\leq j\leq d\), also are locally square integrable. Let \(m\) be the Radon measure \(m= \rho^ 2 \lambda^ d\) and let \(S\) be a symmetric linear operator in \(L^ 2(G, m)\); denote by \({\mathcal A}_{\mathcal M}(S)\) the family of all selfadjoint extensions of \(S\) that generate Markovian semigroups and call an element of \({\mathcal A}_{\mathcal M}(S)\) a Markovian extension of \(S\). The family \({\mathcal A}_{\mathcal M}(S)\) will be equipped with the following partial order: \(A_ 1< A_ 2\) if \(D(A_ 1)\subseteq D(A_ 2)\) and if \((\sqrt{- A_ 1} u,\sqrt{-A_ 1} u)_ m\geq (\sqrt{- A_ 2} u,\sqrt{- A_ 2} u)_ m\) for all \(u\in D(\sqrt{- A_ 1})\). Then the Friedrichs extension is the minimal operator in \({\mathcal A}_{\mathcal M}(S)\). The author is especially interested in the case \(S:= L_ p\), a so-called generalized Schrödinger operator, defined by \(L_ \rho\varphi= \Delta\varphi+ 2\sum^ d_{j= 1} \rho^{-1} \nabla_ j \rho\nabla_ j \varphi\), \(\varphi\in C^ \infty_ 0(G)\), which is symmetric in \(L^ 2(G, m)\). Define the ``Sobolev-like'' subspace \({\mathcal F}^ +\) as the space of all functions \(u\in L^ 2(G, m)\) with the property that, for all \(1\leq j\leq d\), the distribution \(D_ j u\) on \(G\) given by \(D_ j u= \nabla_ j u- 2(\rho^{-1} \nabla_ j \rho)u\) belongs to \(L^ 2(G, m)\). If \(\rho\equiv 1\), then \({\mathcal F}^ +\) is the Sobolev space \(W^{1,2}(G)\). Define the quadratic form \({\mathcal E}^ +(u, v)\) as follows: \({\mathcal E}^ +(u, v)= \sum^ d_{j= 1} \int_ G D_ j u D_ j v\rho^ 2 d\lambda^ d\), where \(u\) and \(v\) belong to \({\mathcal F}^ +\). The author proves that the selfadjoint operator corresponding to \({\mathcal E}^ +\) is the maximal operator in \({\mathcal A}_{\mathcal M}(L_ p)\). As a consequence the restriction of \(L_ p\) to \(C^ \infty_ 0(G)\) possesses a unique Markovian extension if and only if \(({\mathcal E}^ 0, {\mathcal F}^ 0)\), being the smallest closed extension of \({\mathcal E}^ +\) restricted to \(C^ \infty_ 0(G)\), is the same as \(({\mathcal E}^ +, {\mathcal F}^ +)\). This means that the ``Sobolev'' spaces \({\mathcal F}^ +\) and \({\mathcal F}^ 0\) coincide. Upon taking \(\rho\) identically 1, it then easily follows that, in case \(\mathbb{R}^ d\backslash G\) is (1,2)-polar, the operator \(L_ 1\) confined to \(C^ \infty_ 0(G)\) possesses a unique Markovian extension. It also follows that the operators \(L_ \rho\) may have a unique Markovian extension without being essentially selfadjoint. The author gives several examples of this situation. He also indicates that his techniques can be adapted to the infinite-dimensional setting.