Electron localization in a metal, ultimately leadingto a metalinsulator (MI) transition, can occur because of disorder (Anderson transition) or electron-electron interactions (Mott-Hubbard transition). Both effects play a role in heavily doped semiconductors which have become prototype systems for the study of MI transitions. In this review we focus on phosphorus-doped Si. The statistical distribution of donor atoms on an atomic scale as the origin of random disorder can be checked by scanning tunneling microscopy. Long-range Coulomb interactions lead to AltshulerAronov corrections to the density of states N(EF) at the Fermi level and the electrical conductivity a(T) on the metallic side of the MI transition, and to a soft Coulomb gap at EF and Efros-Shklovskii variable-range hopping on the insulating side. On-site Coulomb interactions, on the other hand, lead to the formation of localized magnetic moments and the Kondo effect on the metallic side, and to a Hubbard splitting of thedonor band on the insulating side. The MI transition in Si:P can be tuned by varying the P concentration or -for barely insulating samples -by application of uniaxial stress S. The continuous stress tuning allows the observation of dynamic scaling of a(T,S) and hence a reliable determination of the critical exponent # of the extrapolated zero-temperature conductivity a(0)~l S -S¢ Iv, i.e. # --1, and of the dynamical exponent z --3.
IntroductionThe problem of the metal-insulator (MI) transition in heavily doped semiconductors has come of age. Although these materials have become prototype systems for the study of a MI transition driven by the combined effects of disorder (Anderson transition) and electron-electron interactions (Mott-Hubbard transition), these effects are only beginning to be disentangled experimentally. In a strict sense, the MI transition viewed as a transition from extended to localized states at the Fermi level, occurs only at T = 0 because at finite temperature T thermally activated charge transport may occur. Hence a metal is defined by a finite dc conductivity a(T) for T --~0, while for an insulator a(0) = 0. Zero-point quantum fluctuations become important at a quantum phase transition between these two groundstates [1]. The determination of the critical exponent # describing how a(0) vanishes at T = 0 as a function of a control parameter 5 at a critical value 5c, i.e. a(0)~1 5 -5c I", has been a subject of considerable controversy, as will be discussed below. However, using both carrier concentration N and uniaxial stress S to