Lithium was chosen as the simplest known metal for the first application of quantum Monte Carlo methods in order to evaluate the accuracy of conventional one-electron band theories.' Lithium has been extensively studied using such techniques. The KKR method [1], the linear muffin tin orbital method (LMTO) [2], the augmented spherical wave method (ASW) [3] and a linear combinations of gaussian type orbitals (LCGTO) method [4] agree in their predictions of the equa tion of state. Thesj results

more »

... also consistent with experimental data available at low compressions [5j [6] and agree with quantum-statistica"-models [7] [8] at high pressures. Band theory calculations have certain limitations in general and soecifieally in their application to lithium. Results depend on such factors as charge shape approximations (muffin tins), pseudopotentials (a special problem for lithium where the lack of p core states requires a strong pseudopotential), and the form and parameters chosen for the exchange potential. The calculations are all oneelectron methods in which the correlation effects are included in an ad hoc manner. This approximation may be particularly poor in the high compression regime, where the core states become delocalized. Furthermore, band theory pro vides only self-consistent results rather than strict limits on th' e< energies. The quantum Monte Carlo method is a totally different technique usin^ a many-body rather than a mean field approach which yields an upper bound on the energies. QUANTUM MANY-BODY ALGORITHM The SehrSdinger equation was solved for a system of M fixed lithium atoms and W=3Af electrons using the quantum Monte Carlo algorithm previously developed for the electron gas [9] [10]. This technique does not approximate the 3N dimensional problem by reducing it to a set of equations of lower dimen sionality, but solves it exactly within statistical error bars. The algorithm f permanent address: Commissariat l'energie atomique, Centre d'etudes de Limeil-