Recent studies clearly suggest that two components contribute to the Bias Temperature Instability (BTI). The first component is largely recoverable (R) and usually dominates the dynamics within the experimental window, starting from the microsecond regime up to a period of weeks. The second component is largely permanent (P) and may possibly dominate the long-term degradation. The recoverable component is usually attributed to charge trapping inside the oxide, a process typically assumed to follow first-order kinetics. Due to the amorphous nature of the oxide, the capture and emission times of the defects are widely distributed, explaining the contributions over such a large experimental window.
As a consequence of the first-order reaction, the effective time constants of the defects do not depend on frequency and the built-up degradation must be independent of the frequency. However, a number of groups have recently reported a frequency dependence of the degradation. One problem plaguing such studies is the strong impact of the measurement delay and possible synchronization issues between the AC stress signal and the measurement. By exploiting the properties of first-order kinetics it can be shown that, after a recovery time larger than the stress time, the trapped charge after AC stress must equal the trapped charge after DC stress of the same effective duration. Since these recovery times are very large, any issues with measurement delay and synchronization can be avoided. Studying the remaining degradation using this new method clearly shows that the experimental data is inconsistent with a first-order process. This result was expected given our detailed study on individual defects using the Time-Dependent Defect Spectroscopy (TDDS), which showed that charging occurs via an intermediate metastable state. This metastable state fully explains the frequency dependence of the data.