Magnetoresistance and dephasing in a two-dimensional electron gas at intermediate conductances

We study, both theoretically and experimentally, the negative magnetoresistance (MR) of a two-dimensional (2D) electron gas in a weak transverse magnetic field B. The analysis is carried out in a wide range of zero-B conductances g (measured in units of e(2)/h), including the range of intermediate conductances gsimilar to1. Interpretation of the experimental results obtained for a 2D electron gas in GaAs/InxGa1-xAs/GaAs single quantum well structures is based on a theory that takes into account terms of higher orders in 1/g. We show that the standard weak localization (WL) theory is adequate for ggreater than or similar to5. Calculating the corrections of second order in 1/g to the MR, stemming from both the interference contribution and the mutual effect of WL and Coulomb interaction, we expand the range of a quantitative agreement between the theory and experiment down to significantly lower conductances gsimilar to1. We demonstrate that at intermediate conductances the negative MR is described by the standard WL digamma-functions expression, but with a reduced prefactor alpha. We also show that at not very high g the second-loop corrections dominate over the contribution of the interaction in the Cooper channel, and therefore appear to be the main source of the lowering of the prefactor alphasimilar or equal to1-2/pig. The fitting of the MR allows us to measure the true value of the phase breaking time within a wide conductance range ggreater than or similar to1. We further analyze the regime of a weak insulator, when the zero-B conductance is low g(B=0)<1 due to the localization at low temperature, whereas the Drude conductance is high g(0)>1, so that a weak magnetic field delocalizes electronic states. In this regime, while the MR still can be fitted by the digamma-functions formula, the experimentally obtained value of the dephasing rate has nothing to do with the true one. The corresponding fitting parameter in the low-T limit is determined by the localization length and may therefore saturate at T-->0, even though the true dephasing rate vanishes.