Picritic glass beads of volcanic origin have been interpreted as representing extremely primitive lunar magmas that approach primary magma compositions. Based on this, they have been used for developing models for lunar magmatism, estimating the mineralogical and chemical characteristics of their mantle sources, constraining the dynamics and structure of the lunar mantle, developing models for lunar core formation, reconstructing the bulk composition of the Moon, and evaluating the genetic relationships between the Earth and the Moon. Yet, we still do not truly understand many aspects of the compositional diversity among individual glass types or even within a single glass type. Using high precision ion microprobe techniques, we have analyzed a total of forty very low-Ti glass beads from the Apollo 15 (A15) site and twenty additional picritic glass beads from other sampling sites for a suite of highly compatible (Ni, Co), weakly compatible to weakly incompatible (Mn, Cr, V), and highly incompatible (Ce, Ti, Zr, Nb) elements. Within the A15 very low-Ti glass population, Ni and Co concentrations are positively correlated, exhibit limited fractionation (of Ni/Co), and are distinctive for individual glass groups as defined by major elements. Manganese is also colinear with both Ni and Co. The highly incompatible elements also appear to be generally colinear with Ni, Co, and Mn, although there are distinguishable compositional ''offsets'' in the glass populations. The Ni/Co ratio does not differ dramatically among the very low-Ti glasses from other sampling sites, although incompatible element concentrations may differ significantly. These trace element data eliminate fractional crystallization and partial melting of a homogeneous source (under extremely low f(o2) conditions) as possible models for the origin of the compositional variations observed in the A15 very low-Ti glass population. These data are consistent with the gen eration of these magmas through partial melting of a spatially restricted, heterogeneous, lunar magma ocean (LMO) cumulate sequence. The compositional variation in these magmas can be produced by at least two different melting processes: (1) Partial melting of a cumulate sequence consisting of relatively early cumulates with a higher proportion of an ''intercumulus melt component'' (Green glass A, D, E) and later cumulates with lower proportions of an ''intercumulus melt component'' (Green glass C). (2) Different degrees of partial melting of a cumulate sequence. Small degrees of partial melting (less than 5%) of a more primitive cumulate mantle source will generate one group of magmas (Green A, D, E), whereas a group of other picritic magmas (Green C) were produced by larger degrees of partial melting (10 to 20%) of a mineralogically similar, but a more evolved cumulate mantle source. In both models, the Green B glass group is formed by either cumulate commingling or magma mixing processes. These magmas cannot be rapidly transported from the source region to the lunar surface in a single eruptive event without extensive fractional crystallization. Yet, the colinear relationship among Ni, Co, and Mn is not compatible with extensive and variable fractional crystallization. Therefore, a polybaric melting model as suggested by Longhi (1992b) appears to be the only alternative to producing relatively unfractionated, Apollo 15 very low-Ti picritic magmas with high-pressure signatures. The trace element data provides additional constrains for this type of model.
