Extracting statistical distributions of RTN originating from both acceptor-like and donor-like traps

Abstract

The impact of Random Telegraph Noise (RTN) on devices increases, as the device sizes are downscaled. Against a reference level, it is commonly observed that RTN can fluctuate both below and above this level. The modelling of RTN, however, was typically carried out only in the direction where drain current reduces. In reality, this current reduction can be compensated by simultaneous current increases. This calls the accuracy of the onedirectional RTN modelling into questions. Separating the fluctuation in one direction from the other is difficult experimentally. In this paper, we review the recently proposed integral methodology for achieving this separation. In contrast with early works, the integral methodology does not require selecting devices with fluctuation only in one direction. The RTN in all devices are measured and grouped together to form one dataset. It is then statically analyzed by assuming the presence of fluctuation in both directions. In this way, the separation is carried out numerically, rather than experimentally. Based on the maximum likelihood estimation, the popular statistical distributions are tested against experimental data. It is found that the General Extreme Value (GEV) distribution agrees best with the experimental threshold voltage shift, when compared with the Exponential and Lognormal distributions.