Due to the amorphous structure of the interface, the binding energy of the
-bonds at the interface is not constant but varies from site to site.
Electron-spin-resonance (ESR) studies revealed the binding energies as
distributed Gaussian with a variance of 
[70], i.e. weaker bonds
break first, while stronger bonds remain passivated. Longer stress times or a
larger applied electric field are required to break those stronger bonds
[71].
Charge pumping (CP) as the measurement of choice for the assessment of the
amount of interface states (cf. Chapter 2.4) revealed some interesting facts. The
observed amount of recovering interface states accessed via CP after NBTI stress
was too small to be able to explain the overall recovery of 
. Therefore a
part of the community [62, 72, 73, 74] considered the generated interface
states as permanent once created. This assumption will be discussed in
Chapter 8.
Quite in contrast, Mahapatra et al. stated that CP measurements in the range of seconds are too slow to detect the recovery of interface states because of inherent delay of the measurement setup. Another possibility to explain the missing recovery involves the CP technique itself, as it pulses into accumulation which in turn causes unwanted additional relaxation [75].
In order to clarify the issue of how interface states contribute to recovery, Li
et al. developed the on-the-fly fast interface trap CP method (OFIT) [51, 25],
described in Chapter 2.5. Based on the results of this OFIT method [51, 24, 76],
which showed recovery faster than a second, but also revealed long-term recovery,
Grasser et al. derived a BTI-model based on interface states only in [77]. Therein
they describe two distinct components of the recovery as two facets of a single
degradation mechanism proceeding as a series of steps. By assuming dissociation
of 
bonds (dispersive bond breaking) the so-called double-well
(and subsequently refined triple-well) model is able to describe quite
complex stress-relaxation-patterns. Though the mathematics in this model
describe the NBTI phenomenon correctly, its microscopic assumptions
are likely unjustified [78], an issue that will be examined in detail in
Chapter 5.