Phys. Rev. A 4, 806 (1971).' Two of the sources of error in Monte Carlo calculations are the statistical error, associated with the fact that chains of finite length are used, and the statisticalmechanical error, associated with the fact that only a finite number of particles are used. A third source of error in our use of Monte Carlo hard-sphere data results from the empirical formulas employed to fit the data. Verlet and Weis (Ref. 6) estimated the statistical error of their hard-sphere calculation to be of the order of 0.01 in g&(r). Their empirical fit, Eq. (Cl), differs from their Monte Carlo results by at most 0. 03. An estimate of the statistical-mechanical error can be obtained by comparing Monte Carlo hard-sphere calculations of two groups of investigators using different numbers of particles. When we compare the Verlet and Weis empirical gd(r) (calculated for 864 particles) with the results of Barker and Henderson ( 100 particles) [Mol. Phys. (to be published)], we find that at high densities the Barker-Henderson peak heights are lower (by as much as 0.08 at pd3=0. 85) and that near r= l. 5d the Barker-Henderson values for gz(r) are higher by around 0. 03.~G . E.Highly accurate x-ray diffraction measurements are presented for the static structure factor a(q) for liquid Na (at 100 and 200'C) and liquid K (at 65 and 135'C). A detailed error analysis is presented showing that the over-all root-mean-square error in a(q) never exceeds 2. 5% for any value of the momentum transfer q and the relative root-mean-square error in a(q) between different temperatures is always less than 1.5%. We discuss and demonstrate the reliability of the tabulated values for the atomic form factor and the Compton-scattering correction. A brief discussion is included of the relative merits of x-ray vs neutron diffraction for obtaining the static structure factor.

No abstract

No abstract

No abstract

No abstract