Oxide defects play an important role in both the operation as well as the reliability of semiconductor devices. In Metal-Oxide-Semiconductor (MOS) transistors, they have been linked to the occurrence of Random Telegraph Noise (RTN), visible as random switching of, for instance, the drain current between two discrete levels under constant bias conditions. RTN has traditionally been investigated for nMOS transistors and explained by random capture and emission of channel electrons into and from oxide traps. While conventional RTN appears to be the prevalent form, various anomalies have been detected, such as defects switching between more than a single level and defects that temporarily stop switching altogether.
Recent evidence strongly suggests that the defects conventionally linked to RTN are also responsible for slow charging and discharging transients in MOS transistors. Such oxide charging phenomena can be, for instance, observed under negative bias temperature stress in a phenomenon known as the Negative Bias Temperature Instability (NBTI). Characterization of these oxide defects has so far to rely on macroscopic observations on an ensemble of defects, making the identification of microscopic defect properties difficult. Our recently developed measurement technique, the Time Dependent Defect Spectroscopy (TDDS), allows for extraction of capture and emission time constants of individual defects over a wide range of temperature and bias conditions. Application of the TDDS to pMOS transistors revealed similar behavior as known from nMOS transistors. In addition, a number of similar anomalies have been observed. The most dominant feature is the change of the discrete step height associated with the defect upon charge capture of another defect located in the same current percolation path. This purely electrostatic interaction can also cause a significant change of the emission time constant. Furthermore, defects can temporarily stop contributing capture and emission events for random amounts of time, ranging from a few hours to many months. Finally, the capture and emission times show interesting bias dependencies that indicate the existence of metastable defect states.
Based on the dataset acquired by the TDDS, a refined model for oxide traps was developed. Each defect is assumed to exist in two charge states, either neutral or positive. Each charge state then has a metastable minimum, which is required in order to explain the peculiar bias dependencies of the capture and emission time constants. Transitions between the states are calculated using non-radiative multiphonon emission theory. A detailed comparison with the experimental data revealed that the previously considered anomalous behavior occurs much more often than not and should thus be considered to be the regular case.