No abstract

No abstract

No abstract

Shlimak et al. Reply:In the preceding Comment [1] Sarachik and Bogdanovich argue that the new approach for determination of the critical conductivity exponent m of the metal-insulator transition (MIT) at nonzero temperatures T ‫ء‬ , proposed previously in our work [2], is not valid for Si:B, and extrapolation to T 0 is preferable for this case. The main argument in [1] is the fact that the slope m of the straight lines Ds͑T͒ vs DN͞N c becomes progressively smaller as T ‫ء‬ is decreased [ Fig. 1(a) of [1] ]. However, a weak dependence of the slope on T ‫ء‬ does not contradict our method. It has been mentioned in [2] that the data may slightly depend on T ‫ء‬ , but it does not influence significantly the value of m because in the close vicinity of the MIT all curves s͑T͒ are, in fact, almost parallel, while for more distant N, but within the scaling region, the temperature corrections to the conductivity are much smaller than the ones introduced by the increase of impurity concentration. Such behavior is typical for an MIT driven by the impurity concentration ͑N-MIT͒, including Si:B [3]. A different behavior has been observed for an MIT driven by uniaxial stress ͑X-MIT͒ [4]. In this particular case, of course, our method is not valid.It is interesting to note that the data for Si:B shown in the preceding Comment clearly support our result. One can see from Fig. 1(b) of [1] that the slope remains close to 1 for a wide interval of T ‫ء‬ : For 7 samples 0.9 , m , 1.05 while T ‫ء‬ is changed from 0.07 to 1 K, i.e., more than one order of magnitude [5]. This extends the series of Si-based materials (together with Si:P [6] and Si:Sb [7]) for which m ഠ 1.In [8][9][10], impurity concentration of most of the samples is out of the scaling region. We believe that this narrow interval is limited for samples with conductivity almost independent of T . Within the scaling region the above mentioned behavior of the N-MIT is correct (see Figs. 1, 2 in [8,9]). In [10], there is only one sample of Ge:Ga within the scaling region; the neighboring sample is already at the boundary of this region.

We show that one can unify the scaling behavior of the conductivity in Ge:As in the vicinity of the metal-insulator transition driven either by the concentration of impurities N or by the magnetic field B by introducing a new scaling variable Uϭ͓(N/N c )Ϫ(B/B*)Ϫ1͔, where both the critical impurity concentration N c and the characteristic magnetic field B* are constant. ͓S0163-1829͑97͒10004-2͔The metal-insulator transition ͑MIT͒ has been the subject of intensive theoretical and experimental investigation for many years. 1 According to the scaling theory for doped semiconductors, 2 the conductivity at zero temperature (0)ϭ(T→0), when plotted as a function of the impurity concentration N, is equal to zero on the insulating side of the MIT and remains finite on the metallic side, obeying a power law in the vicinity of the transitionwhere N c is the critical-impurity concentration and is the critical-conductivity exponent. The theory 2 predicts ϭ1.We will refer to this transition as N-MIT. For barely metallic samples with NϾN c , the MIT will occur upon the application of a critical magnetic field B c , because a strong magnetic field leads to the shrinkage of the electron wave function. Scaling behavior of the conductivity is also expected in the neighborhood of this magnetic-field-driven metalinsulator transition (B-MIT͒: