Collision integrals related to binary (dilute gas) diffusion are calculated classically for six species colliding with N 2 . The most detailed calculations make no assumptions regarding the complexity of the potential energy surface, and the resulting classical collision integrals are in excellent agreement with previous semiclassical results for H + N 2 and H 2 + N 2 and with recent experimental results for C n H 2n+2 + N 2 , n = 2-4. The detailed classical results are used to test the accuracy of three simplifying assumptions typically made when calculating collision integrals: (1) approximating the intermolecular potential as isotropic, (2) neglecting the internal structure of the colliders (i.e., neglecting inelasticity), and (3) employing unphysical R −12 repulsive interactions. The effect of anisotropy is found to be negligible for H + N 2 and H 2 + N 2 (in agreement with previous quantum mechanical and semiclassical results for systems involving atomic and diatomic species) but is more significant for larger species at low temperatures. For example, the neglect of anisotropy decreases the diffusion coefficient for butane + N 2 by 15% at 300 K. The neglect of inelasticity, in contrast, introduces only very small errors. Approximating the repulsive wall as an unphysical R −12 interaction is a significant source of error at all temperatures for the weakly interacting systems H + N 2 and H 2 + N 2 , with errors as large as 40%. For the normal alkanes in N 2 , which feature stronger interactions, the 12/6 Lennard-Jones approximation is found to be accurate, particularly at temperatures above ∼700 K where it predicts the full-dimensional result to within 5% (although with somewhat different temperature dependence). Overall, the typical practical approach of assuming isotropic 12/6 Lennard-Jones interactions is confirmed to be suitable for combustion applications except for weakly interacting systems, such as H + N 2 . For these systems, anisotropy and inelasticity can safely be neglected but a more detailed description of the repulsive wall is required for quantitative predictions. A straightforward approach for calculating effective isotropic potentials with realistic repulsive walls is described. An analytic expression for the calculated diffusion coefficient for H + N 2 is presented and is estimated to have a 2-sigma error bar of only 0.7%. © 2014 AIP Publishing LLC. [http://dx.