Seemingly pure agglutinitic glass is riddled with many submicrometer-sized vesicles and clasts, as shown by scanning electron microscopy of ion-etched polished grain mounts. Electron mircoprobe analyses of agglutinitic glasses that have to be identified with relatively low power optical microscopes are difficult and tend to be biased by very fine grained contaminants. Energy dispersive x ray analysis (EDXA) of small spots (1 μm¿1 μm) of pure agglutinitic glass, carefully selected to be free of clasts or vesicles as determined by secondary electron imagery, can overcome some of the analytical problems. We have performed over 600 EDXA of pure glass of 32 large agglutinates from five different Apollo landing sites. Analyses adjacent to or away from clasts do not show any difference in chemical composition and suggest that secondary assimilation of clasts by the primary impact melt has not been a major process affecting the final composition. The average glass composition of several agglutinates in any single soil is different from the bulk composition of the soil; it is biased toward that of the finest fraction of the soil. This is consistent with agglutination models advocating the fusion and/or incorporation of the finest soil fraction. Most of the large multigenerational agglutinates contain small glassy areas or domains that are remarkably homogeneous in chemical composition. They may represent the melt from single impact events, and their homogeneity indicates that total melting of a microtarget in lunar soils upon micrometeoritic bombardment might be the origin of agglutinitic glass. The compositions of these domains are different in different agglutinates, and even within single multigenerational agglutinates, and bear no systematic relationship to the composition of either the bulk soils or their finest fractions. It seems likely that random small targets in lunar soils (grains ~100 μm or less) can only produce very small volumes of melt that are not representative of the soil itself.

The total melting of microtargets is not contradictory to models invoking incorporation or fusion of the finest fraction because such models consider average compositions of several agglutinates with respect to the parent soils; they therefore refer to larger scale effects of agglutination. A large soil grain, if hit directly by a micrometeorite smaller than the grain itself, is likely to behave more likely to behave more like a solid target; a collection of very fine soil grains, if hit by a micrometeorite larger than the collection of the grains, will behave as a porous target. Because porous targets produce relatively more melt than solid targets, the average melt composition from a large number of random micrometeoritic impacts is expected to be biased to that of the finest fraction of lunar soils.