Comparatively heavy isotopic compositions of moderately volatile elements (MVE) in lunar rocks have been advocated to reflect the loss of light isotopes during devolatilization processes from the Moon. In this study we present new gallium (Ga) isotope data for lunar highland rocks, with a focus on the Ferroan Anorthosite Suite (FAS). These are commonly thought to be direct crystallization products from the late lunar magma ocean (LMO) and should contain the majority of the Ga inventory of the Moon. As such, FAS rocks are crucial for identifying the processes that drove Ga isotope fractionation as well as for inferring the Ga isotopic composition of the bulk Moon. Our data reveal that FAS samples have a range in 871Ga from -0.27 to 0.22%o and are generally isotopically light in Ga compared to other lunar lithologies, but straddle values typical of terrestrial rocks. Although Ga is defined as an MVE, these Ga isotope variations do not correspond with concentrations of more volatile elements, indicating that Ga isotope variations in the FAS are not primarily controlled by devolatilization processes. Instead, the Ga isotopic compositions of bulk FAS rocks broadly correlate with the composition of plagioclase, with the calcium content of plagioclase decreasing as Ga becomes isotopically heavier. This suggests that fractionation of Ga isotopes in FAS rocks was caused by the preferential incorporation of isotopically light Ga into plagioclase during the later solidification stages of the LMO. The progressive crystallization and extraction of plagioclase forces the residual melt towards increasingly heavier Ga isotope ratios, corroborating similar conclusions derived from correlations between 871Ga and Eu* in the mare basalt suite. Using Ga isotope partitioning calculations, we demonstrate that an isotope fractionation coefficient between plagioclase and coexisting melt of -0.3 to -0.4%o could explain the observed range of 871Ga values in FAS, mare basalt suite rocks, and KREEP. These calculations allow for a first order estimate of the Ga isotopic composition of the bulk silicate Moon prior to plagioclase fractionation and suggest it was close to the composition of the bulk silicate Earth. This would imply that the Moon did not lose a substantial fraction of its Ga inventory during accretion, consistent with new constraints from RbSr isotope systematics that indicate the Moon's volatile deficit was primarily inherited from Theia. In conjunction with the overlap in non-mass-dependent isotope ratios, these collective observations could be reconciled if Theia and the proto-Earth formed in similar regions of the inner Solar System that were already volatile-depleted.

Due to the lack of rock samples directly from the deep part of the Moon, experiments and numerical simulation are effective methods to understand the early evolution of the Moon. Since the 1970s, the Lunar Magma Ocean (LMO) evolution model has been verified and modified by a large number of experimental petrology and geochemical work. However, the original composition of the Moon and the depth of its magma ocean, which are the two most critical parameters of LMO models remain controversial. The different lunar crust thickness estimated from lunar seismic data compared to that estimated from gravity data, the volatile content of lunar samples, and the widespread of Mg and Al-rich spinet (Cr (#) <5) discovered from interpreting the new remote sensing data affect our assessment on the starting composition and the depth of LMO, and the fractional crystallization process thereafter. In this paper, we review a series of high temperature and high pressure experimental petrology and experimental geochemistry results on the Moon's early evolution by focusing on: (1) The influence of refractory elements and volatile content of LMO's composition and its depth on the thickness of lunar crust and the Moon's mineral constitution formed through early differentiation. (2) The rationality of stability of high pressure mineral garnet deep inside lunar mantle and it effect on the distribution of trace elements during the evolution of lunar. (3) The petrogenesis of the Moon's special components, including volcanic glasses and Mg-suite, and their indication on the composition of the Moon's deep interior. (4) The constraint of lunar core composition on the Moon's material source, especially the abundance of trace elements. Based on the latest observation and the new analysis results of lunar samples, we evaluate the existing LMO evolution models and propose a LMO model with garnet as an important constituent mineral inside the Moon. We also discuss the necessary work need to be done to improve the new LMO model.