The relaxation behavior of stressed pMOSFETs depends on the oxide electric field and stress time of the performed stress. Especially when dealing with PBTI, the harsher the conditions of stress the later the device starts to relax significantly. By using the limited observation period for NBTI and PBTI as part of the full recovery shape the experimental findings can be explained. The full features of the recovery curve can only be identified after moderate BTI stress, where the relaxation after a certain time accelerates, and slows down again later. Furthermore, deeper analysis of the relaxation characteristics provides information on the distribution of capture and emission times of the defects assumed responsible for BTI. For the case of the PBTI measurements presented in this chapter, especially the distribution of emission time constants depends on the applied oxide electric field during stress. A higher oxide electric field shifts and broadens this distribution. This change in the distribution shows that with a deeper understanding of single capture and emission times it might be possible to reveal the actual origin of the BTI phenomenon.
A method for the detection of the change of a real single defect state, e.g. an electron emission, was already reported by Karwath et al. more than 20 years ago. They used the deep level transient spectroscopy (DLTS)1 to observe the emission times of single isolated defects in small-area MOSFETs by step-like current transients [118]. Such a step-like behavior at the emission time of a defect is also obtained by the time dependent defect spectroscopy (TDDS) [115, 113]. Here small devices are repeatedly stressed (100 times or more) and the averaged relaxation curve is then monitored showing the discharging behavior of the single defects. In order to be able to determine the different capture times of the defects, different stress pulses have to be applied [12]. When a large number of stress and relaxation sequences are collected on a map, the emission times can be obtained. These maps are similar to those presented in this chapter [116, 111, 114]. The major difference lies in the size of the investigated samples. In larger MOSFETs the averaging of the relaxation curves is neither necessary nor reasonable, because already inherent due the large number of defects present there [115]. Although the superposition of many defects is not yet fully understood, it was shown in [116] that the discrete step-like recovery observed in small (narrow) devices is indeed comparable with the nearly continous recovery behavior obtained for large (wide) devices. The averaging of many small devices also yields a log-like behavior, giving a very strong hint that the underlying mechanism is the same.