The viscosity of simple dense fluids ; a memory-function approach

A gaussian memory function is used to compute the viscosity of dense fluid argon over a large region of states from the Lennard-Jones potential and the radial distribution function. The superposition approximation is used to evaluate the triplet correlation function and estimates of the errors introduced by this approximation are investigated. The agreement with experimental viscosities is quite satisfactory in the liquid region, but less so at lower densities and higher temperatures. The computed stress autocorrelation function exhibits those characteristics which are known from molecular dynamics studies. The stress autocorrelation remains monotonic even close to the triple point, where the same memory function leads to the well-known backscattering behaviour in the velocity autocorrelation function. This, too, is in agreement with the results of molecular dynamics.