Termination of typical wave-function multifractal spectra at the Anderson metal-insulator transition: Field theory description using the functional renormalization group

Abstract

We revisit the problem of wavefunction statistics at the Anderson metal-insulator transition (MIT) of non-interacting electrons in d > 2 spatial dimensions. At the transition, the complex spatial structure of the critical wavefunctions is reflected in the non-linear behavior of the multifractal spectrum of generalized inverse participation ratios (IPRs). Beyond the crossover from narrow to broad IPR statistics, which always occurs for sufficiently large moments of the wavefunction amplitude, the spectrum obtained from a typical wavefunction associated with a particular disorder realization differs markedly from that obtained from the disorder-averaged IPRs. This phenomenon is known as the termination of the multifractal spectrum. We provide a field theoretical derivation for the termination of the typical multifractal spectrum, by combining the non-linear sigma model framework, conventionally used to access the MIT in d = 2 + epsilon dimensions, with a functional renormalization group (FRG) technique. The FRG method deployed here was originally pioneered to study the properties of the two-dimensional (2D) random phase XY model [D. Carpentier and P. Le Doussal, Nucl. Phys. B 588, 565 (2000)]. The same method was used to demonstrate the termination of the multifractal spectrum in the very special problem of 2D Dirac fermions subject to a random Abelian vector potential. Our result shows that the typical multifractal wavefunction spectrum and its termination can be obtained at a generic Anderson localization transition in d > 2, within the standard field theoretical framework of the non-linear sigma model, when combined with the FRG.