We have now demonstrated that the ratio of the activation enthalpies for nearest neighbor hopping of the two types of vacancy in InP are in the 4 to 1 ratio of the masses of the atoms that hop and are in good agreement with the values predicted by the ballistic model (BM), which identifies the activation enthalpy as a kinetic energy. A consequence is that relatively light atoms should hop to second neighbor vacancy sites as easily as heavy atoms hop to nearest neighbor vacancies, and do so without the complication of forming or annihilating antisites. For example, in GaAs B, C, N, Al, Si, and P are predicted to hop to second neighbor sites as easily as As hops to a first neighbor site. Of course, they hop to nearest neighbor sites even more readily. The import of this on the array of point defects predicted to occur as a heterojunction anneals is rather complex. Experiment supports our contention of complexity. For example, in As rich,n‐type AlGaAs Meietal. reportD(Al)∝n1 exp(−2.9 eV/kT) if the doping is with Te butD(Al) ∝n3 exp(−4 eV/kT) if the doping is with Si, wherenis the free electron density. We have studied the Al diffusion in AlGaAs with the aid of a microcomputer simulation that works out the complex consequences of one’s assumptions re thermochemical parameters at the atomic level. Our previously published BM assumptions lead to the conclusion that Si promotes the conversion of Ga vacancies to trivacancies while Te does not; they also identify 2.9 eV as the sum of the Ga second‐neighbor hopping enthalpy and the enthalpy of formation of single negative Ga vacancy inn‐type GaAs and identify 4 eV as the sum of the enthalpy of formation of the trivacancy plus its nearest neighbor hopping energy in GaAs. The simulation also finds that Al by itself converts As vacancies to trivacancies plus metal antisite. It shows the trivacancies to diffuse very rapidly and to entrain Al.We propose that the cation sublattice dominance of Si induced host interdiffusion follows from this entrainment. We also suggest these complexes are related to DX and account for the anomalous variation of the parameters of that defect with Al concentration. Except for the assumption that the hopping kinetic energy be proportional to the mass of the hopping atom, the conclusions stated above are insensitive to the precise values of atomic level parameters assumed. Quantities such as activation energies for macroscopic diffusion are sensitive to the precise values and some of these are either poorly determined or controversial. More simulation than we have accomplished to date is required to establish the relation between precise atomic level values and observable macroscopic quantities.