Convergence of spectra of mesoscopic systems collapsing onto a graph (Q5940340): Difference between revisions

In the last decade research was started on theory of differential operators restricted to a graph. Recent progress in manufacturing allows us to have circuits composed of extremely thin strips of semiconductors. Thus the domain of a differential operator describing electronic phenomena shrinks to a graph. The resulting one-dimensional mathematical models are known as ``quantum wires''. Other labels have also been used. Similar but not identical problems arose in the study of thin photonic crystals. Thus arose the theory of differential or pseudodifferential operators on graphs. Along the edge of the graph the spectral problems for these operators shrink to a simple problem like \(-d^2u/dx^2= \lambda^2u\) with appropriate boundary conditions at each vertex. But proving convergence of the spectrum, as the domain shrinks to a graph, is not easy. The difficulty arises in formulating corresponding consistent boundary conditions at each vertex. Convergence of the solutions as the domain shrinks to a smooth curve was proved for the Dirichlet and Neumann Laplace operators. More recently some convergence results for the heat equation with Neumann Laplacian were proved by \textit{Y. Saito} [The limiting equation of the Neumann Laplacian on shrinking domains, preprint (1999)] on a ``fattened graph'', meaning an \(\varepsilon\)-neighborhood of the graph. The authors acknowledge in their final remarks that presently reviewed theory follows initial series of articles of Rubinstein and Schatzman published between 1996 and 1999. See for example [\textit{M. Schatzman}, On the eigenvalue of the Laplace operator on a thin set with Neumann boundary conditions, Appl. Anal. 61, 293-306 (1996; Zbl 0865.35098)]. In this article the authors also enclose the graph in tubes. These tubes do not end in the vortices, but instead at a distance \(\gamma\) of the order of \(\varepsilon\) away from the vertex point of the graph. They consider the Neumann-Laplace operator \(-\Delta_\varepsilon\) in \(L^2(M^\varepsilon_j)\) defined in the \(\varepsilon\)-neighborhood of the tube \(M_j\). The Laplacian is reduced to the ordinary differential operator \(A_p\) with domain \(M_j: A_pf(x_j)= p^{-1}d/dx_j(p df/dx_j)\). \(f\) is continuous through the vertex, but \(p\) generally has jump discontinuities in the ``very short'' \(\gamma\)-transition sleeve between the tube and the spherical neighborhood of the vertex. The authors prove that in the limit (as \(\varepsilon\to 0\)) \(\lim\lambda_n(- \Delta_\varepsilon)= \lambda_n(A_p)\). As the authors claimed in their introduction this proof is not easy. In the next section they prove that their results apply operators which include electromagnetic field effects.