Tsetseris and co-workers investigated the subject of BTI at atomic level using the ab-initio simulation code VASP . They propose a proton based dissociation model  to describe the BTI induced breaking of Si-H bonds at the silicon-dielectric interface, which is described in the following.
The breaking of passivated
bonds at the
interface and the
generation of dangling silicon bonds,
, has been suggested as direct
mechanism in the reaction-diffusion model
Through ab-initio calculations  a dissociation energy barrier of 0.95eV is obtained, when the Fermi-level is at the valence band edge, which is the situation of an n-type substrate in inversion. Due to the lower energy barrier, this process can be activated by BTI temperatures.
As source for the required protons the authors suggested hydrogen bound to substrate dopants. The activation energy to dissociate a P-H complex has calculated to be 1.3eV, which is in good agreement with an experimental value of 1.18eV . These energy values are valid in n-type silicon at flat band conditions where hydrogen exists as .
In the depletion region the stability of the P-H complex changes dramatically. In this region the preferred charge state of hydrogen changes from negative to neutral, , resulting in a much lower dissociation activation energy for a P-H complex of only -eV. For a certain period of time the hydrogen atom located in the inversion layer stays neutral and then transfers into a positively charged proton, , by picking up a hole.
At negative bias stress conditions the free protons drift to the interface. As the energy barrier to cross the interface is very high (1eV) the protons migrate rapidly along the interface. As both, dangling bonds and the protons are positively charged it is very unlikely that a proton passivates an interface trap. They will preferably be located close to Si-H bonds which they can break by forming (6.30). Figure 6.15 gives a schematic overview of the processes involved.
An additional mechanism can be the injection of protons into the dielectric leading to positive oxide charges . These positive charges prevent further protons to be injected. Only when the oxide charges diffuse away from the interface new protons can be injected leading to further degradation .
This model is capable of explaining the different susceptibility of n- and p-channel MOSFETs to positive and negative bias stress (Section 6.3.7)
When assuming that the atomic hydrogen in the bulk diffuses faster than the molecular hydrogen in the dielectric, Tsetseris' model results in a time exponent of . This is not in agreement with recent measurement data (Section 6.3.6) and therefore the model in the present form might be incomplete.