Numerical results for the Anderson transition .

Answer to cond-mat/0106005, cond-mat/0106006 and additional notes are given concerning my previous comment (cond-mat/0105325). Recently I have discussed the numerical results for the Anderson transition [1]. Objections [2] and [3] have appeared after that. We answer here to the arguments of [2, 3] and give the additional comments. 1. I did not state in [1] that ν = 1 (in fact, my most probable value was 1.25). I tried to show that a problem exists: numerical results contradict to all other information on the critical exponents. It is not my personal problem, and it is not my personal affair to solve it. Indeed, a value ν ≈ 1 is desirable from viewpoint of this other information and I tried to understand, is not it possible to agree it with the raw numerical data by the change of interpretation and by more realistic estimation of errors. I found it possible for the data that I considered as the best. May be I was a little wrong in the latter estimation. Authors of [2] argue that I used the systems of too small size (28 3), " while currently sizes ∼ 50 3 (for ELS) and 18 2 × 10 8 (for TM) are standard " 1. A size 18 2 × 10 8 should kick a man out his feet, while authors of [2] know perfectly well that only the minimal length scale (L = 18) is relevant. In fact, such systems are small. 2 I do not know about " currently " and " standard " but only one result is cited in [2] for systems ∼ 50 3 [6]. With my procedure used, authors of [2] obtained this result with uncertanty ν = 1.25 ÷ 1.75, which is not drastically different from uncertainty 0.8 ÷ 1.7 that I have derived from Zharekeshev and Kramer data [7]. So, it is admitted that the largest scale data give uncertainty 1.25 ÷ 1.75 and not 1% [5]. At the present stage, I am quite satisfied with such progress. 1 Abbreviations of [2]: TM and ELS — transfer matrix and energy level statistics methods, FSS — finite size scaling. 2 I can understand such trick in the text, where authors of [2] are in the course of discussion, but the same thing is present in the figure capture: " the system sizes of TM data are larger than for ELS data " .