Statistical properties of the Green function in finite size for Anderson localization models with multifractal eigenvectors

For Anderson localization models with multifractal eigenvectors on disordered samples containing N sites, we analyze in a unified framework the consequences for the statistical properties of the Green function. We focus in particular on the imaginary part of the Green function at coinciding points G(xx)(I)( E - i eta) I and study the scaling with the size N of the moments of arbitrary indices q when the broadening follows the scaling eta = c/N-delta For thestandard scaling regime delta = 1, we find in the two limits c e 1 and c c 1 that the moments are governed by the anomalous exponents Delta(q) of individual eigenfunctions, without the assumption of strong correlations between the weights of consecutive eigenstates at the same point. For the non- standard scaling regimes 0< delta < 1, we obtain that the imaginary Green function follows some Frechet distribution in the typical region, while rare events are important to obtain the scaling of the moments. We describe the application to the case of Gaussian multifractality and to the case of linear multifractality.