All of the models mentioned so far have in common that the concentration of the diffusion vehicles (point defects) do not deviate from their equilibrium values. Unfortunately this cannot generally be assumed. Oxidation, oxynitridation and nitridation are well known to perturb the point defect concentrations. Oxidation generates excess silicon self-interstitials, tending to enhance the diffusivities of those atoms which diffuse with a significant interstitialcy component (for instance boron, arsenic and phosphorus) and tending to retard diffusivities of atoms which diffuse predominantly via a vacancy mechanism (like antimony) [Hu83], [Ant82].
In spite of the evidence of dopant fluxes driven by the gradient of the
point defect concentration, substitutional dopants are assumed to adhere to
the dual diffusion model. Impurities 
diffuse with preference 
via
interstitialcy and with complementary preference 
via vacancy mechanism. The diffusivity is then given by (3.2-28)
where 
and 
are the silicon self-interstitial and vacancy
concentrations, respectively, and the superscripts 
indicate quantities
at equilibrium point defect conditions.
The two implemented models for oxidation enhanced/retarded diffusion
(OED/ORD) differ just in the calculation of the point defect concentration.
The static OED/ORD model OEDS calculates the interstitial and
vacancy concentrations from analytical approximations, whereas the model
OED determines the local point defect concentration by solving
diffusion equations. Both models require the growth of the oxide to be known
in advance (as shown in Program 3.6-2), more precisely the
models need the velocity of each point at the 
-interface, and
the shape of the silicon segment at any point of time.
Here, we will
describe the two models OEDS and OED in
Sections 3.2.3 and 3.2.3, respectively.
