Bias temperature instability is one of the most important reliability issues in modern MOSFETs. It is observed when a large bias is applied to the gate of an otherwise grounded transistor and results in a drift of crucial device parameters, most notably the threshold voltage. However, even after four decades of research, bias temperature instability is still a highly puzzling phenomenon that has so far eluded our complete understanding. Particularly intriguing are the complex degradation/recovery patterns, which are observed when the gate bias is modulated. Understanding this behavior is mandatory for any estimation of device degradation in a realistic circuit setting. Quite contrary to the latest experimental observations, most modeling approaches published so far have focused on constant gate bias stress and are remarkably oblivious to any recovery of the degradation which sets in as soon as the stress is removed. In particular, none of these features can be explained by the most commonly used reaction-diffusion model.
Based on our extensive research on the peculiarities observed during recovery, we have recently suggested a new model for bias temperature instability that is able to describe the phenomenon in greater detail. The model assumes the dissociation of silicon-hydrogen bonds, which results in electrically active defects. In equilibrium, that is, after the passivation step used during fabrication and before the application of stress, most hydrogen atoms are bound to silicon dangling bonds, which represents the ground state. Upon application of an electric field and at elevated temperatures the hydrogen atom may overcome the barrier to the neighboring transport state via thermal emission. Capture of a carrier further reduces the binding energy, thereby favoring the complete dissociation of the hydrogen from the silicon dangling bond. Alternatively, instead of complete dissociation when the stress is removed, the particle will return to the ground state, with the time constant determined by the barrier height separating the first transport state from the equilibrium position. Since the gate insulators we are dealing with are amorphous, the barriers separating the ground state from the transport state will be different for each silicon dangling bond and the average behavior over all wells gives the macroscopically observed threshold voltage shift.