Analysis and optimization of stress conditions for gate oxide wearout using Monte Carlo simulation

The stress conditions for gate oxide stressing and the sample size determine the duration and the accuracy of the lifetime projection. Often the sample size is compromised to reduce stress time. In this presentation we study the statistical variations related to sample size and stress conditions by means of Monte-Carlo simulation. The simulation allows us to determine the deviation of each parameter from the entered value that was used for the simulation. The analysis focuses on intrinsic distributions in the 5-6 nm oxide thickness range. From the simulated tbd-distributions t63.2%, the Weibull slope and the acceleration parameter gamma are determined and used for the lifetime projection. Distributions of the parameters resulting from many trials showed that t63.2% and gamma are normally distributed while the slope and the projected lifetimes show a log-normal distribution. We investigated the impact of several options to reduce stress duration and their impact on accuracy. For all studied options we observed that with reduced stress duration accuracy is also reduced. A new stress matrix that includes different areas is introduced for improved Weibull slope determination. This matrix results in more precise lifetime projections despite a time saving. However, any desired level of accuracy requires the investment of an appropriate sample size