Conductance noise in an out-of-equilibrium two-dimensional electron system

TL;DR

This study investigates conductance noise in a 2D electron system in silicon at low temperatures, revealing non-Gaussian, history-dependent noise linked to out-of-equilibrium states and Coulomb interactions.

Contribution

It demonstrates how density changes induce strong non-Gaussian noise and reveals the system's aging and approach to equilibrium through detailed statistical analysis.

Findings

Non-Gaussian noise appears after cooling from high temperature.

Density changes significantly affect the free energy landscape.

Non-Gaussian PDFs evolve towards Gaussian as the system ages.

Abstract

A study of the conductance noise in a two-dimensional electron system (2DES) in Si at low temperatures (T) reveals the onset of large, non-Gaussian noise after cooling from an equilibrium state at a high T with a fixed carrier density n_s. This behavior, which signifies the falling out of equilibrium of the 2DES as T->0, is observed for n_s<n_g (n_g - glass transition density). A protocol where density is changed by a small value \Delta n_s at low T produces the same results for the noise power spectra. However, a detailed analysis of the non-Gaussian probability density functions (PDFs) of the fluctuations reveals that \Delta n_s has a qualitatively different and more dramatic effect than \Delta T, suggesting that \Delta n_s induces strong changes in the free energy landscape of the system as a result of Coulomb interactions. The results from a third, waiting-time (t_w) protocol, where n_s is changed temporarily during t_w by a large amount, demonstrate that non-Gaussian PDFs exhibit history dependence and an evolution towards a Gaussian distribution as the system ages and slowly approaches equilibrium. By calculating the power spectra and higher-order statistics for the noise measured over a wide range of the applied voltage bias, it is established that the non-Gaussian noise is observed in the regime of Ohmic or linear response, i.e. that it is not caused by the applied bias.