Critical indices of Anderson transition: something is wrong with numerical results

Numerical results for Anderson transition are critically discussed. A simple procedure to deal with corrections to scaling is suggested. With real uncertainties taken into account, the raw data are in agreement with a value ν = 1 for the critical index of the correlation length in three dimensions. Critical indices s and ν for the conductivity σ and the localization radius ξ of wave functions near the Anderson transition [1] are determined by relations σ ∼ τ s , ξ ∼ τ , (1) where τ is the dimensionless distance to a critical point. On the modern level, the first estimates of s and ν were given by a scaling theory of localization [2], s = 1 and ν = 1/ǫ for the space dimensionality d = 2 + ǫ. They suggest that s ≈ ν ≈ 1 for d = 3. A little later, a self-consistent theory of Vollhardt and Wölfle [3] was developed: it gave for arbitrary d [3, 4] s = 1 , d > 2 ; ν = 