5.4 OFIT versus CP

As in conventional CP measurements, care has to be taken that parasitic tunneling currents and geometry effects do not pollute the measured charge pumping current . The first problem is even more severe in the OFIT technique since there the low level gate voltage equals the stress voltage, resulting in excessive tunneling in thin oxides [51]. In order to avoid these problems, also large-area devices with thick oxides () are used. As shown in Fig. 5.12, the measured during stress and recovery are qualitatively identical for three completely different technologies ( thick , thin , and ).

Figure 5.12: Comparison of OFIT and CP results. For OFIT the offset between end of stress and beginning of relaxation is comparable for different technologies and geometries (not shown). Previously this was explained by fast recovery. This fast recovery is absent in CP (indicated for ). Furthermore, continuous gate pulsing affects both stress and relaxation, causing a faster recovery of interface states with increasing number of measurements (black circles vs. black squares, both of OFIT).

Quite remarkably, continuous application of OFIT pulses (as well as CP measurements) has a dramatic impact on both the stress and the recovery characteristics. In particular, with 3 measurements per decade of relaxation time, is quasi constant during recovery, while up to 100 measurements per decade in time result in approximately 80% recovery of .

Another fact is that the first OFIT measurement point during stress is already responsible for at least 30% of the total degradation. Likewise, the first measurement taken during recovery at already shows recovery while the rest of the recovery depends basically on the number of measurements per decade. Though not shown here, the same behavior is obtained for CP. To be able to understand how recovery works here, a deeper analysis of the charge pumping technique is needed.