Abstract
For low-pressure simulations of liquid-like states of pure substances, the statistical and systematic errors contained in the results for the compressibility factor are investigated. Calculations with particle numbers between 108 and 2048 are reported, giving a clear picture of the magnitude of the possible errors. Two state points have been considered, for which the total compressibility factor is much smaller than unity. The detected total error is found to amount to 15–30 per cent for convenient simulations with 500 particle and a cutoff sphere of 2·30σ. Furthermore the applicability of methods like the Baxter technique for continuation of the pair correlation function to large r for the long range corrections neglected in the direct evaluation of the virial has been tested. For cutoff radii smaller than or equal to 2·30σ such a correction should be preferred to the usual correction term calculated on the basis of g(r) = 1 (for r larger than the cutoff sphere). When the potential is truncated at 3·3σ or at larger separations, the normal long-range correction suffices to avoid systematic errors of that type.