5.8 Results

Based on the previous results we are now able to better understand the charge pumping current I
 cp   measured with the OFIT sequence. The presence of additional charges contributes to the signal when the pulse amplitude ΔVG   is increased. A large spread of time constants larger than that of the interface states is necessary to explain the results. By assuming oxide traps with a distributed thermally activated barrier one is able to explain the measurement results with good accuracy. Whereas interface states seem to not respond to an increasing electric field and due to their small time constants account for Icp   at low (10kHz  ) and high frequencies (1MHz  ), the oxide traps are by far slower due to the assumed barrier ΔE
   B   they have to surmount. That is why oxide traps only affect Icp   at lower frequencies, i.e. 10kHz  .

The particularly troublesome part is the application of the OFIT technique during the stress phase, where both oxide traps ΔNot   and additionally created interface states ΔNit   add to Icp   . These contributions are absent during the initial reference measurements and during the OFIT recovery measurements both taken at V     = V
 G,low    rel   . This has fundamental consequences on OFIT measurements: Initially, a reference Icp   is recorded. Following this reference measurement, the gate voltage low-level VG,low   is switched to Vstr   . Due to the much larger ΔVG   now a significant contribution of   ox
Icp   is obtained. Furthermore, with the large pulse amplitude, additional interface states are created, which is the intended effect of this OFIT measurement. However, without this the additional increase in Icp   due to oxide traps must not be attributed to interface states created by degradation. Consequently, I ot
 cp   needs to be corrected in the measurement data. Using the mentioned extrapolation method of            2
ΔNot  = AE ox   reveals that the 30% initial increase in Icp   is entirely due to oxide traps. The corrected last stress value in Fig. 5.19 is identical to the first value at the recovery, leading to the conclusion that no fast interface state recovery occurs.


PIC


Figure 5.19: Oxide traps lead to a spurious increase in the charge pumping signal during stress. Using the scheme developed for Fig. 2.10, a corrected Icp   is obtained. The smooth transition between the corrected Icp   during stress and the Icp   during recovery suggests that no fast recovery takes place.