No abstract uploaded!

We review several recent results concerning two-dimensional systems which exhibit a transport induced by a one-dimensional perturbation of a homogeneous magnetic field. The first concerns the local Iwatsuka model, where a charged particle interacts with a field which is homogeneous outside a finite strip and translationally invariant along it: we present two new sufficient conditions for absolute continuity of the spectrum and show that in most cases the number of open spectral gaps is finite. In the second model the perturbation is a periodic array of point obstacles. In this case the Landau levels remain to be infinitely degenerate eigenvalues, and between them the system has bands of absolutely continuous spectrum.

Graph consists of a set of vertices V={Xj: j I}, a set of finite edges L={Ljn:(Xj, Xn) IL I I} and a set of infinite edges L={Lj: Xj IC} attached to them. We regard it as a configuration space of a quantum system with the Hilbert space

In this paper we investigate the operator H=??(??) in L2 (R2), where> 0 and? is a closed C4 Jordan curve in R2. We obtain the asymptotic form of each eigenvalue of H as tends to in nity. We also get the asymptotic form of the number of negative eigenvalues of H in the strong coupling asymptotic regime. MSC: 35J10, 35P15

Point interactions described by formal shaped potentials were used in quantum mechanics from the early thirties. It lasted several decades, however, until a proper way to handle the corresponding Schrdinger operators was found. Following the observation of Berezin and Faddeev [4], the problem was studied systematically in the eighties. The results are summarized in the monograph [1], references to some recent work are given, eg, in [2, 8]. In its present form the pointinteraction method represents a versatile tool for constructing solvable models whose power has not been yet, to our opinion, appreciated fully in the physical community. Within the standard quantum mechanical formalism point interactions can be constructed as long as the configuration space dimension does not exceed three, and they can be interpreted as limits of suitable families of squeezed potentials. In the onedimensional case the limiting argument admits a straightforward interpretation, because the pointinteraction coupling constant is just the integral of the approximating potential: a slow particle on the line with a wellspread wavefunction sees only the average value of a localized potential. In distinction to that the approximation by scaled potentials in dimension two and three requires existence of a zeroenergy resonance and a particular couplingconstant renormalization; a detailed discussion of the related mathematical problems can be found in [1, 9]. This gives these point interactions a flavour of something exceptional, for instance, one is led to believe that they cannot be used to model welllocalized repulsive potentials. That would have consequences