Next, the phenomenological impact of the stress and recovery field on the threshold voltage shift is shown. Again, the results of large-area devices and the results obtained from their nanoscale counterparts are compared.
Considering the threshold voltage shift of large-area transistors recorded during stress, as visible in Figure 6.3 (left), it can be clearly seen that the absolute threshold voltage shift increases at larger stress biases. This observation is intuitively clear, because a higher stress bias leads to higher oxide field and as a consequence more defects can become charged. Also visible is the continuous characteristics of the threshold voltage shift which can be attributed to the contribution of many defects with very small step heights to the total threshold voltage shift.
In contrast, for nanoscale devices the impact of the stress bias can be illustrated best in terms of single defects which become charged during the stress cycle, see Figure 6.3 (right). Thereby each defect has its individual charge capture time which is strongly bias and temperature dependent. As a consequence of the decreasing capture time at higher stress biases, the probability of a single defect to become charged increases when the stress bias is increased.
Analogously to the stress bias dependence of the threshold voltage shift, the threshold voltage shift does also depend on the gate bias during recovery. As can be seen in Figure 6.4 (left), the larger the recovery bias gets, the slower the device recovery proceeds.
In line with the accumulated threshold voltage shift during stress from Figure 6.3 (right), the recovery of nanoscale devices proceeds in a discrete manner too, see Figure 6.4 (right). The emission time of such charge emission events can be very sensitive to the recovery bias. However, in contrast to the charge capture events, charge emission of single defects can also be bias-independent. Thus the emission time of a defect does not change with the gate bias. This behavior can be seen for defect #2 which remains unaffected by an change of the gate bias, whereas the two other defects emit their charge at shorter emission times for lower recovery bias. In general, the bias independent emission time is associated with fixed oxide traps whereas the bias dependent emission times are due to switching traps. As one can easily see, providing an accurate model of the bias dependence of BTI is pretty challenging as for instance the field dependence of individual defects is observed to be on one hand negligible and on the other hand very strong.« PreviousUpNext »Contents