The Voltage Dependence of NBTI: Resolution of Controversy Surrounding Power Law vs. Exponential

Negative Bias Temperature Instability is typically measured at high stress voltages and then projected to the lower voltage use conditions using a model with either power law or exponential voltage dependence. The choice of the right model is crucial, as it can lead to vastly different estimations of degradation at use conditions. In this work, using a variety of characterization methods, we demonstrate that the NBTI can show an apparent power law voltage dependence during the early stage of degradation, but an exponential voltage dependence model is needed for accurate use-condition projections. We also demonstrate that empirical extrapolation from conventional NBTI measurements, without proper consideration of the recovery effect, may lead to a perceived power law dependence. Finally, using months-long NBTI characterization on various gate stacks and device architectures, we confirm the generality of the behavior that the power law voltage model can significantly underestimate the degradation close to product use conditions.