The complete set of the four macroscopic transport coefficients describing the coupled diffusion of impurity atoms and vacancies in silicon is calculated from the atomistic mechanism by accurately taking into account the effects of the microscopic forces between dopants and vacancies. The aim of these simulations is to come to a decision concerning the validity of models like the pair diffusion model [e.g., M. Yoshida, J. Appl. Phys. 48, 2169 (1977); R. B. Fair and J. C. C. Tsai, J. Electrochem. Soc. 124, 1107 (1977); F. F. Morehead and R. F. Lever, Appl. Phys. Lett. 48, 151 (1986); B. J. Mulvaney and W. B. Richardson, Appl. Phys. Lett. 51, 1439 (1987)] or the “non-Fickian diffusion” model [M. Kurata, Y. Morikawa, K. Nagami, and H. Kuroda, Jpn. J. Appl. Phys. 12, 472 (1973); Y. Morikawa, K. Yamamoto, and K. Nagami, Appl. Phys. Lett. 36, 997 (1980); V. V. Kozlovski, V. N. Lomasov, and L. S. Vlasenko, Radiat. Eff. 106, 37 (1988); O. V. Aleksandrov, V. V. Kozlovski, V. V. Popov, and B. E. Samorukov, Phys. Status Solidi 110, K61 (1988), K. Maser, Exp. Tech. Phys. (Berlin) 34, 213 (1986), K. Maser, Ann. Phys. (Leipzig) 45, 81 (1988), K. Maser, Exp. Tech. Phys. (Berlin) 39, 169 (1991)] that make contradicting predictions for very fundamental properties like the relative direction of the fluxes of dopants and vacancies driven by a vacancy gradient and for the relation α = Td0/Dd0 between two of the four transport coefficients. Simulation results are shown for a variety of assumed interaction potentials that establish a functional dependence between α and measurable quantities, like the factor Dd/Dtracer of enhancement of dopant diffusivity over tracer diffusion, that holds for an arbitrary interaction. The comparison with experimental values for Dd/Dtracer leads to confirmation of the pair diffusion model for boron and phosphorous. For arsenic and antimony, the large scatter of the experimental data prohibits an equally definite conclusion, but at least a qualitative confirmation of pair diffusion theory (i.e., α>0 which means that dopant and vacancy fluxes have the same direction if caused by a vacancy gradient) is possible. © 1998 American Institute of Physics.