The phase after switching off a device after a certain stress time or between switching on and off is called recovery or relaxation. During this recovery phase the degraded parameters start to revert to their initial values, but within times that depend on the prior stressing conditions. Now the question arises how the interplay between BTI stress and recovery during the operation of a MOSFET does affect the reliability of the device. This is important because only when charactising and modeling both stress and recovery a realistic lifetime extrapolation of the device is possible.
Within the RD theory the recovery is explained via back diffusion of hydrogen. Once the stress is removed, the quasi-equilibrium at the interface causes all interfacial hydrogen to immediately passivate the dangling bonds. By the time all interfacial hydrogen is bound and no more hydrogen is available at the interface, hydrogen located deeper in the oxide is assumed to diffuse back and control the entire recovery as the limiting process. During recovery the back diffusion can be approximated by the empirical expression
|
| (3.2) |
This equation is related to the universal recovery [61] which will be discussed in
Chapter 4.1 more thoroughly. Interestingly, the diffusion-limited recovery
in (3.2) yields about 
recovery within 4 decades of time, whereas
experimental data still show recovery over at least 10 decades of time [61].
Moreover, since the RD recovery only depends on the ratio of relaxation to
stress time, the model as such is not capable of explaining any other
experimentally observed recovery behavior, e.g. dependence on temperature or
stress voltage. The RD model is also not able to explain the dynamic
behavior1
of NBTI when applying alternating stress and relaxation sequences with a
varying duty factor (DF) or duty cycle (DC) [20, 6, 62, 30].