The origin of the moon is examined in the context of theories of planetary accretion and of siderophile element abundances inferred for the upper mantles of the earth, moon, and shergottite parent body (SPB=Mars?). The lunar origin hypotheses examined are collisional ejection in a giant imact, and coaccretion from a circumterrestrial disk of metal-depleted material.

If the moon originated in a giant impact, the impactor must have contained free metal and was most probably differentiated with siderophile elements segregated into a core, in order to account for the low absolute and generally nonchondritic relative abundances of moderately and noble siderophile elements in the lunar mantle. If impactor metal was accreted to the earth's upper mantle during the impact and if it perfectly scavenged siderophile elements from the earth's upper mantle into the core, then it is necessary to postulate a late-stage chondritic ''veneer'' in order to account for the high absolute and chondritic relative abundances of noble siderophile elements in the present-day upper mantle of the earth. These elements are less abundant in the mantles of the moon and SPB than in the earth, and they may not be present in chondritic ratios. If they are present in chondritic ratios (the scatter in noble siderophile element data permits this possibility), they are present at approximately 10-3 and 10-4¿CI concentrations in the SPB and moon, respectively. In either case, the noble siderophile elements present a problem for the chondritic ''veneer'' version of the collosional ejection hypothesis because they are not present in any reservoir in the moon at the required abundance levels.

If the impactor metal that was accreted to the earth imperfectly scavenged siderophile elements from the earth's mantle, it is not necessary to postulate a late stage ''veneer.'' However, the giant impact would substantially melt or vaporize the earth's mantle, yet there is no evidence in samples derived from the upper mantle for large-scale vertical compositional zoning that might be expected to result from crystallization in a gravity field, by analogy with the moon and with terrestrial layered intrusions. Mantle nodules from all over the world are homogeneous with respect to compatible siderophile elements such as Ni, Co, and Ir at the hand specimen scale.

Coaccretion hypotheses of lunar origin are inherently less testable using siderophile elements than collisional ejection hypotheses because abundances of these elements in the mantles of the earth and SPB are less useful as constraints. At present, none of the hypotheses of lunar origin satisfactorily accounts for siderophile element concentrations in the mantles of the earth, moon, and SPB. ¿ American Geophysical Union 1987