Coherent Backscattering near the Two-Dimensional Metal-Insulator Transition

We have studied corrections to conductivity due to the coherent backscattering in low-disordered two-dimensional electron systems in silicon for a range of electron densities including the vicinity of the metal-insulator transition, where the dramatic increase of the spin susceptibility has been observed earlier. We show that the corrections, which exist deeper in the metallic phase, weaken upon approaching the transition and practically vanish at the critical density, thus suggesting that the localization is suppressed near and at the transition even in zero field.

Figure 1

Shubnikov–de Haas (SdH) oscillations in sample NH50.3064.9N at $n_s=7.3\times {10}^{11}{\mathrm{c}\mathrm{m}}^{-2}$ (a) and $n_s=1.17\times {10}^{11}{\mathrm{c}\mathrm{m}}^{-2}$ (b). The vertical arrows in (b) show the positions of the SdH minima corresponding to Landau filling factors $\nu =14$, 10, and 6 as labeled. Temperatures are indicated in the figures. Measuring current is 2 nA (a) and 200 pA (b). The experimentally measured amplitudes of the oscillations are shown in the insets by solid circles together with the calculated dependences (solid lines); the parameters used in calculations are indicated in the insets. The lower inset in (b) shows a close-up view of two SdH oscillations corresponding to $\nu =14$ and 10 measured at three temperatures.

Figure 2

Longitudinal magnetoconductivity in a weak perpendicular magnetic field at $T=42\mathrm{m}\mathrm{K}$ for a range of electron densities which are indicated near each curve in units of ${10}^{11}{\mathrm{c}\mathrm{m}}^{-2}$ in samples NH50.3064.9N (a) and NH50.4064.7N (b).

Figure 3

Magnetoconductivity in sample NH50.3064.9N at $T=42\mathrm{m}\mathrm{K}$ and at different electron densities indicated near the curves in units of ${10}^{11}{\mathrm{c}\mathrm{m}}^{-2}$. (a) The curves are vertically shifted for clarity, and two lower curves are multiplied by 6 (the middle one) and 50 (the lower one), which makes them quantitatively similar to the upper (unmodified) curve. The thick dashed line is the fit by Eq. (1) with $\alpha =0.6$ and ${\tau }_{\varphi }=30\mathrm{p}\mathrm{s}$. (b) Unsuccessful attempt to fit the experimental data for $n_s=0.85\times {10}^{11}{\mathrm{c}\mathrm{m}}^{-2}$ by Eq. (1) with the same prefactor $\alpha =0.6$ as above but reduced ${\tau }_{\varphi }=6\mathrm{p}\mathrm{s}$ (the lower dashed curve, which is shifted for clarity). However, the data can be fitted reasonably well by choosing ${\tau }_{\varphi }=30\mathrm{p}\mathrm{s}$ as in (a), but with the prefactor 10 times smaller ($\alpha =0.06$; the upper dashed curve).