Variations on a theme of Jost and Pais (Q2466515): Difference between revisions

This paper gives an interesting extension of a one-dimensional result of \textit{R.\,Jost} and \textit{A.\,Pais} [Phys.\ Rev.~(2) 82, 840--851 (1951; Zbl 0042.45206)] to the higher-dimensional case. Jost and Pais considered the Schrödinger equation \([-\frac{d^2}{dx^2}+V(x)]\psi(x) =z^{1/2}\psi(x)\) on the half-line \((0,\infty)\) with integrable potential functions \(V(x)\) to prove a formula which reduces the (Fredholm) determinant associated with the Birman--Schwinger kernel of the corresponding one-dimensional Schrödinger operator to a simple Wronski determinant of distributional solutions. And this Wronski determinant also equals the value of the so-called Jost function at the boundary \(x=0\). A little more detail is the following: Let \(H_0^D,H_0^N\) be, respectively, the Dirichlet, Neumann Laplacians in \(L^2((0,\infty);dx)\), and \(H^D :=H_0^D +V\), \(H^N :=H_0^N +V\). Set \(u := \exp(i\arg V)|V|^{1/2}\), \(v :=|V|^{1/2}\) so that \(V=uv\). Then both the operators \(\overline{u(H_0^D-zI)^{-1}v}\) and \(\overline{u(H_0^N-zI)^{-1}v}\) on \(L^2((0,\infty);dx)\) are of trace class, and \[ \begin{aligned} &\det(I+\overline{u(H_0^D-zI)^{-1}v})=W(f(z,\cdot),\phi^D(z,\cdot)) =f(z,0),\\ &\det(I+\overline{u(H_0^N-zI)^{-1}v}) = -\frac{W(f(z,\cdot),\theta^D(z,\cdot))}{iz^{1/2}} =\frac{f'(z,0)}{iz^{1/2}}. \end{aligned} \] Here, the three functions \(\phi^D(z,\cdot)\), \(\theta^D(z,\cdot)\) and \(f(z,\cdot)\), having \(z\) as a parameter being a nonzero complex number with \(\text{Im}(z^{1/2}) > 0\), are, respectively, a fundamental system of two solutions and the Jost solution, which are to be defined via Volterra integral equations, while \(W(f,g)\) denotes the Wronskian of \(f\) and \(g\). In the paper under review, the authors first illustrate what is behind the title, and how they want to extend this result to the multi-dimensional case. As may be seen, taking the ratios of these two determinants above yields a modern formulation of the formula: \[ \begin{multlined}\frac{\det(I+\overline{u(H_0^N-zI)^{-1}v})}{\det(I+\overline{u(H_0^D-zI)^{-1}v})} =1-(\overline{\gamma_N(H^D-zI)^{-1}V[\gamma_D(H_0^N-\bar{z}I)^{-1}]^*})\\ =\frac{W(f(z,\cdot),\phi^N(z,\cdot))} {iz^{1/2}W(f(z,\cdot),\phi^D(z,\cdot))} =\frac{f'(z,0)}{iz^{1/2}f(z,0)} =\frac{m^D(z)}{m_0^D(z)} =\frac{m_0^N(z)}{m^N(z)},\end{multlined} \] where \(\gamma_D,\gamma_N\) are the Dirichlet, Neumann (boundary) traces, and \(m_0^D,m^D,m_0^N\) and \(m^N\) are the so-called Weyl--Titchmarsh \(m\)-functions, satisfying \(m_0^N=-1/m_0^D\), \(m^N=-1/m^D\). Roughly speaking, it is just a formula of this form that the authors have obtained as the main result, by in fact replacing the half-line by an open set \(\Omega \subset{\mathbb{R}}^n\), \(n\geq 2\), which has a compact, nonempty boundary \(\partial\Omega\) satisfying certain regularity conditions, to consider the Schrödinger operator on \(L^2(\Omega;d^nx)\). Nevertheless, since the operators concerned on \(L^2(\Omega;d^nx)\) become no more of trace class, the associated determinants need to be modified to the regularized determinants, and their ratio is reduced to the regularized determinant -- with some additional exponential factor -- associated with an operator on \(L^2(\partial\Omega;d^{n-1}\sigma)\). Thus the result reveals an analogy with the above one-dimensional case, where \(\Omega\) corresponds to the half-line \((0,\infty)\) and its boundary \(\partial\Omega\) to the one-point set \(\{0\}\). It will be of new significance that the Dirichlet-to-Neumann map \(M_{\Omega}^D(z)\) turns out to emerge as the operator-valued generalization of the Weyl--Titchmarsh \(m\)-function \(m^D(z)\).