Abstract
A study is made of a series-expansion procedure which gives the leading terms of the n-particle distribution function p(n)(1,2,,n) as explicit functionals in the radial distribution function g(r). The development of the series is based on the cluster-expansion formalism applied to the Abe form for p(n) expressed as a product of the generalized Kirkwood superposition approximation pK(n) and a correction factor exp[A(n)(1,2,,n)]. An ordering parameter μ is introduced to determine A(n) and p(n) in the form of infinite power series in μ, and the postulate of minimal complexity is employed to eliminate an infinite number of possible classes of solutions in a sequential relation connecting A(n1) and A(n). Derivation of the series for p(n) and many other algebraic manipulations involving a large number of cluster integrals are greatly simplified by the use of a scheme which groups together all cluster terms having, in a certain way, the same source term. In particular, the scheme is useful in demonstrating that the nature of the series structure of p(3) is such that its three-point Fourier transform S(3)(k1,k2,k3) has as a factor the product of the three liquid-structure functions S(k1)S(k2)S(k3). The results obtained to order μ4 for A(3),p(3), and S(3) agree with those derived earlier in a more straightforward but tedious approach. The result for p(4) shows that the convolution approximation pc(4), which contains μ3 terms, must be supplemented by a correction of O(μ3) in order to be accurate through third order. The μ-expansion approach is also examined for the cluster expansion of the correlation function in the Bijl-Dingle-Jastrow description of a manyboson system, and then compared with the number-density expansion formula by using the Gaussian model for g(r)1 to evaluate cluster integrals. A testing procedure based on the requirement p(3)(1,2,2)=0 is developed to study accuracy of the μ-ordered approximations for p(3). Numerical results obtained to orders μ2,μ3, and μ4 with the Gaussian model indicate substantial improvements with each increase in the order of truncation in the power series of p(3). A brief discussion is presented concerning the asymptotic behavior of g(r) in the context of equilibrium statistical mechanics.