Where Is the Lunar Mantle and Deep Crust at Crisium? A Perspective From the Luna 20 Samples

Remote sensing observations have been interpreted to indicate that the Crisium basin‐forming event excavated deep crust and upper mantle. Samples from the highlands adjacent to the Crisium basin returned by Luna 20 (L‐20) bring a unique perspective for evaluating this concept. The magmatic lithologies returned from the noritic Hilly and Furrowed Terrain (nHFT) by L‐20 are coarse‐grained feldspar (>300 μm) with inclusions of pyroxene, and finer‐grained norites, troctolites, spinel troctolites, and gabbros (<100 μm). These two suites represent ferroan anorthosites (FANs) and the Mg‐suite, respectively. There is limited evidence for mantle or deep crustal material within the nHFT samples. Ultramafic rocks such as dunites and orthopyroxenites are absent, and Mg‐rich olivine‐ and orthopyroxene‐bearing‐assemblages are derived from magmatic rocks emplaced in the shallow crust. These lithic fragments represent pre‐Crisium episodes of magmatism (Mg‐suite) and lunar magma ocean products (FANs). The lack of deep lithologies at the L‐20 site seems contradictory to excavation models for Crisium. Mineralogical‐chemical differences suggest a higher FAN component in the rim and that this represents FANs excavated from the deep lunar crust. If it exists, the Mg‐rich olivine previously identified within the Crisium rim is most likely related to deep, complementary versions of the Mg‐suite rocks from L‐20. The material associated with the Crisium basin is not derived from the lunar mantle but represents crustal lithologies from the shallow to deep crust, a substantial mantle component may have been incorporated into the Crisium basin impact melt sheet, or that our “Earth‐analog” for the lunar upper mantle is incorrect.

. LRO image of the Mare Crisium and adjacent regions. Smaller impact craters identified. Identified are the Luna 20 and Luna 24 sampling sites (labeled white squares) and numerous craters within and adjacent to Crisium. Olivine-bearing assemblages are indicated by squares (Corley et al., 2018;Powell et al., 2012;Yamamoto et al., 2010). Yellow = olivine with Mg# = 65. Green = olivine with Mg# = 90. White = olivine, no compositional data. Spinel-bearing lithology is indicated by the pink triangle . Base map derived from 100-m WAC global mosaic morphology map. Speyerer et al. (2011). Between 1970 and1976, the Luna sample return program conducted by the USSR collected drill cores along a ∼480 km "traverse" from the Fecunditatis basin (Luna 16) to the lunar highlands between the Fecunditatis and Crisium basins (Luna 20) to the Crisium basin (Luna 24). Locations of the Luna 20 and Luna 24 sampling sites and their geologic context are presented in Figure 1. The total mass of the cores returned from these three missions is approximately 321 g, with Luna 16 returning ∼101 g, Luna 20 returning ∼50 g, and Luna 24 returning ∼170 g.

The Luna 20 Mission
The Luna 20 robotic sample return mission was launched on 14 February 1972. The science goal of the mission was to collect and return soil samples from the lunar highlands. The landing site was in the mountainous lunar highlands area known as the Terra Apollonius. Figure 2 illustrates the surface characteristics of the Luna 20 landing site and its location relative to Mare Fecunditatis and Mare Crisium. The Luna 20 landing site is about 160 km north of the Luna 16 site and about 33 km north of the contact with the "shoreline" of Mare Fecunditatis. "The complexly faulted terrain at this site has probably been involved in many generations of shattering and uplift produced by impacts which formed the large basin ring structures" (Heiken & McEwen, 1972). The core was obtained by using a percussion-rotary drill tool on a stretched-out arm. The drill rate changed automatically depending on the density of the material. Drilling occurred on a 30° angle to local vertical. The tool stopped drilling at a depth of 250 mm due to overheating. When drilling terminated, the sample of lunar soil, which was contained in a plastic tube, was wound into a cylindrical spiral, and placed within the return vehicle. The mission returned to Earth on 25 February 1972. Through exchanges with the NASA Curatorial branch, the United States was allocated 2.69 g (Bell et al., 2022).

Analytical Approaches
Thin sections of sieved material (250-500 μm) collected from the Luna 20 site and allocated to NASA were examined and integrated with data collected during consortium studies in 1973) (Geochimica et Cosmochimica Acta (vol. 37 no. 4). We collected backscattered electron (BSE) images, quantitative EDS analyses and X-ray maps with a TESCAN LYRA3 SEM. After imaging and analysis of all lithic fragments, we focused on crystalline lithologies and conducted further chemical and modal mineralogy analyses. The modal mineralogy of each fragment was determined by point-counting 750-1,200 points on backscattered electron images of the fragments, after identification of the phases by X-ray mapping and EDS and WDS techniques. We calculated normative mineralogy from unpublished broad-beam analyses (Conrad et al., 1973) and converted the wt% values into volume % to be comparable with model mineralogy. The TESCAN LYRA3 scanning electron microscope and a JEOL 8200 electron microprobe were used to collect major element analyses of individual mineral phases. Both instruments are located within the Institute of Meteoritics at the University of New Mexico. Sections were carbon-coated and examined using a TESCAN LYRA3 scanning electron microscope (SEM) operated at 15 kV. Quantitative energy-dispersive analyses (EDS) and elemental X-ray maps were obtained using an IXRF silicon drift energy-dispersive X-ray detector running Iridium Ultra software. Quantitative wavelength-dispersive analyses were obtained with a fully automated JEOL 8200 electron microprobe operated at 15 kV with a focused beam, a beam current of 20 nA, calibrated with pure oxide and natural mineral standards. Count times were 20 or 30 s on peaks, 10 or 20 s (1973) for background measurements, respectively. Most detection limits are between 0.01 and 0.02 elemental wt%. Cation formulas were calculated from oxide totals of individual mineral analyses. These analyses were evaluated for quality using oxide totals and appropriate stoichiometric criteria (e.g., site occupancy, cation totals) based on the crystal chemical rational drawn from several prior works (e.g., Cameron & Papike, 1981;Karner et al., 2006;Papike et al., 1976Papike et al., , 2009. Spinel analyses were plotted on the modified Johnston prism using the Spinel Web software of Antonini et al. (2020). Analyses are presented in an electronic format. Additional unpublished data from Conrad et al. (1973) is available through the Institute of Meteoritics at the University of New Mexico as Special Publication Number 12, UNM Institute of Meteoritics (https://ntrs.nasa.gov/api/citations/19740014355/downloads/19740014355.pdf).

Analysis of Orbital Mineral and Chemical Data
Moon Mineralogy Mapper (M3) data provide the highest spatial-and spectral-resolution mineralogical data for the lunar surface and are therefore ideally suited for characterizing compositional diversity across the Crisium region. For these analyses, we use Planetary Data System-released Level 1 global-mode M 3 radiance images. These data have a nominal spatial resolution of 140-280 m per pixel and spectral resolution of 20-40 nm per channel from 540 to 3,000 nm Green et al., 2011). Photometric, topographic, and thermal corrections were applied following the methods described and validated by Li and Milliken (2016), producing fully calibrated reflectance images comparable to Planetary Data System-released M3 Level 2 reflectance data. However, as noted above, this thermal correction is a significant improvement over the standard M3 thermal correction, enabling a more robust assessment of mineralogy across the lunar surface (Li & Milliken, 2016).
Observing conditions (such as phase angle, solar illumination, detector temperature spacecraft altitude, etc.) varied throughout the mission, affecting spatial resolution as well as detector response . To account for these changing conditions, M3 data are divided into several convenient optical periods (OPs), with small and often systematic artifacts between the OPs (Besse et al., 2013;Boardman et al., 2011). OP2C1 offers the most complete coverage of Crisium, enabling a reliable comparative assessment of the Luna 20 site and Crisium's ring massifs. Calibrated OP2C1 M3 images for the Crisium region were georeferenced to standard lunar coordinates for comparison with Lunar Reconnaissance Orbiter and other remote sensing data.
Lunar surface materials exhibit diagnostic spectral features across the VNIR related to mineralogy, optical maturity, and other factors (Burns & Burns, 1993;Pieters et al., 2000). Because spectral properties across the lunar surface are often dominated by differences in the abundance and composition of pyroxene minerals (relative to plagioclase and olivine), M 3 analytical techniques relevant to such pyroxene-bearing lithologies have been extensively validated and deployed.
For this reason, mineralogical diversity across the Crisium region is well-captured by differences in albedo, band depth, and center wavelength of the 1 and 2 μm spectral absorption bands. Generally speaking, band depths are positively correlated with pyroxene abundance, while band centers are sensitive to pyroxene composition 10.1029/2022JE007409 5 of 24 (Mg-rich orthopyroxenes exhibit short-wavelength absorptions; Ca-, Fe-rich clinopyroxenes exhibit absorptions at longer wavelengths (Klima et al., 2007(Klima et al., , 2011). Parameter maps sensitive to pyroxene abundance and composition were generated using the Parabolas and Linear Continuum (PLC) algorithm, which was validated for use with M 3 data by Moriarty and Pieters (2016). This technique removes a two-part linear continuum based on three optimized tie-points. Band depths and centers are calculated by fitting parabolas to the bottom portions of the 1 and 2 μm absorption bands.
Maps of these parameter values reveal mineralogical diversity in geological context. In these parameter maps, mare basalts exhibit strong, relatively long-wavelength absorption bands and low albedos, noritic materials exhibit shorter-wavelength absorption bands, and feldspathic materials exhibit weak absorption bands and relatively high albedos.
Moon Mineralogy Mapper (M3) online data volumes are available through the PDS Cartography and Imaging Science Node at https://pds-imaging.jpl.nasa.gov/volumes/m3.html.

Analysis of Contribution of Source Material to Luna 20 Site
We used the approach of Petro and Pieters (2004, 2008, and Petro and Klima (2013) for the region surrounding the Luna 20 site to estimate the contribution of ejecta from the Crisium event and post-Crisium craters to the Luna 20 regolith. Petro and Pieters (2004) presented a model to quantify the effects of basin formation and to estimate the resulting proportion of locally derived material relative to foreign material in the regolith. They utilized the ejecta scaling equations of Pike (1974) and Housen et al. (1983) and the concept of an ejecta mixing ratio from Oberbeck et al. (1975). Although similar in concept to the detailed model of Haskin et al. (2000Haskin et al. ( , 2003, the Petro and Pieters approach readily allows global-scale issues to be addressed. Several permutations of the Oberbeck mixing ratio were evaluated to determine how adjustments to the mixing ratio altered the results of the model. Petro and Pieters (2006) determined that a modified form of the Oberbeck mixing ratio was required. The predicted amount of basin ejecta and the mixing ratio are significant because the product of the two determines the depth of mixing for any basin event. These two parameters allow the depth of the early megaregolith to be estimated across the entire lunar surface. The equations for these important model parameters utilized here are given and described in the appendix of Petro and Pieters (2008).

Introduction
The initial results from studies of Luna 20 materials are summarized in a special issue of Geochemica et Cosmochimica Acta (volume 37 no. 4). Several conclusions were reached from these initial and follow-on studies: (a) The soil sample contains crystalline lithic fragments (20%-36%, with variation between different size fractions) of polymict breccias, impact melt rocks, mare basalts, feldspathic basalts, anorthosites, norites, and troctolites (e.g., Cameron et al., 1973;Crawford & Weigand, 1973;Kridelbaugh & Weill, 1973;Meyer, 1973Meyer, , 2009Prinz et al., 1973;Taylor et al., 1973); (b) Magnesium-rich olivine and pyroxene occur in crystalline rocks and as mineral fragments. Olivine has compositions as magnesian as Fo 94 and the orthopyroxene (Ghose et al., 1973) has compositions as magnesium-rich as Wo 3 En 85 Fs 12 (Cameron et al., 1973). Mineral compositions (e.g., TiO 2 , Cr 2 O 3 , and Mg#) of the highland lithic fragments define two distinct suites (Taylor et al., 1973); (c) the KREEP component was observed to be insignificant in materials derived from this site (Laul & Schmitt, 1973;Nava and Philpotts, 1973;Philpotts et al., 1972;Taylor et al., 1973), although a few KREEPy basalt particles were identified (e.g., Cameron et al., 1973); (d) polymict breccias have a very well-defined Ar retention age of approximately 3.9 Ga (Podosek et al., 1973). Measured Ar/Ar ages of highlands lithic fragments reported in later studies range from 4.42 to 3.84 Ga (e.g., Cohen et al., 2001;Swindle et al., 1991); and Swindle et al. (1991) identified a sample that may represent an impact melt derived from the formation of the Crisium basin, but they concluded that such material was rare in the Luna 20 collection.

Petrography
The fragments observed in this suite of samples include igneous lithologies, breccias, impact melt breccias, impact melts, basalts, and individual minerals. Our focus for this initial examination is the crystalline rocks which appeared to be clast-free and products of crystallization from a melt (magmatic or impact). We did not include clast-bearing impact melt breccias in this analysis (e.g., Liu et al., 2019Liu et al., , 2020. Backscattered electron (BSE) images of examples of different mineralogies and textures are shown in Figure 3. The finer-grained (<100 μm) lithologies include spinel troctolites, troctolites, norites, gabbronorites, and gabbros (Figures 3a-3d). No ultramafic lithologies (e.g., dunites, pyroxenites) were observed. In addition to these lithologies, there are coarser-grained plagioclase fragments (>250 μm) associated with much finer mafic grains (<75 μm), predominantly pyroxene (Figures 3e and 3f). Pyroxenes in all these assemblages do not exhibit exsolution lamellae greater than 2-3 μm in width. An example of pyroxene with fine exsolution lamellae is shown in Figure 4. Earlier X-ray examination of pyroxene from Luna 20 samples indicated sub-micron lamellae of varying degrees of complexity (e.g., Ghose et al., 1973).
We present the modal and normative abundances of phases and their compositions from the Luna 20 sampling site ( Figure 5a). These modal abundances are compared to olivine-bearing lithologies at the rim of the Crisium Basin ( Figure 5b) calculated by Corley et al. (2018) using radiative transfer modeling. Their interpretation of orbital data differentiated between distinct olivine compositions (Fo 90 vs. Fo 65 ) that has implications for the origin of the lithologies (e.g., crust vs. mantle; basalt vs. Mg-suite). Their model mineralogy indicates that they are predominantly anorthosites and noritic and gabbroic anorthosites. At the Luna 20 site, the assemblages containing ferroan mafic phases are dominated by coarse-grained plagioclase (100%-85%). The mafic phases are predominantly low-and high-Ca pyroxene. Olivine is generally a minor phase. The model proportions and mineral chemistries are similar to materials making up the Crisium rim (Corley et al., 2018). The lithologies that have more Mg-rich mafic phases at Luna 20 are finer-grained and have a wide range of plagioclase abundances (38%-90%). They generally plot within the fields of anorthositic norite/gabbro, anorthositic troctolite, norite/gabbro, olivine norite/gabbro, and troctolite ± spinel (Figure 5a). Although unique pure MgAl-spinel + plagioclase assemblages (with minor Mg-rich mafic phases) have been reported to the west of Crisium , the spinel-bearing lithologies from Luna 20 generally contain high Mg# olivine that range up to Fo 96 (Conrad et al., 1973;Simon et al., 2022), but substantially less plagioclase (60%-75%) than suggested for these unique spinel assemblages. Furthermore, the Luna 20 spinel compositions are higher in Fe and Cr than compositions inferred from remote sensing data .
At the rim of the Crisium basin, Corley et al. (2018) identified 60 spectra with signatures that represent olivine-bearing mixtures. These spectra were collected from the rims of small craters, massifs at the rim of Crisium, and from maria both inside and outside the Crisium basin. In Figure 5b, we only plotted those data points from the Crisium inner rim. Only six spectra were estimated by Corley et al. (2018) to represent olivine with Mg# = 90 and all were located on the southern rim of Crisium ( Figure 1). Olivine-bearing spectra from the rim of Crisium are on the Irregular Massif and Platform Massif members that were mapped by Sliz and Spudis (2016). The assemblages with high-Mg olivine (Mg# = 90) are more noritic in composition and do not include the troctolites common at the Luna 20 site. Based on our data and measurements made by Corley et al. (2018), dunites have not been identified along the Crisium rim. Lemelin et al. (2019) did not detect exposures of ultramafic material, though they suggested that such mantle material is possibly present at the sub-pixel scale (<62 m). Ultramafic rocks are absent south of the Crisium rim at the Luna 20 site.  Modal mineralogy of areas in the Crisium basin rim, estimated with radiative transfer modeling for M 3 olivine spectra (Corley et al., 2018). Lithologies are differentiated based on olivine composition (Mg# = 65, Mg# = 90). Fields identified in a and indexed in (b) are 1 = anorthosite, 2 = noritic/gabbro anorthosite, 3 = troctolitic anorthosite, 4 = anorthositic norite/gabbro, 5 = anorthositic troctolite, 6 = norite/gabbro, 7 = olivine norite/gabbro, 8 = troctolite, 9 = pyroxenite. 10 = peridotite, and 11 = dunites. Data are provided in Appendix Most of the more ferroan olivine-bearing assemblages at the Crisium rim are predominantly noritic/gabbro anorthosites and troctolitic anorthosite. Lemelin et al. (2019) concluded that the rim of Crisium was dominated by noritic and gabbroic anorthosite (plagioclase 85% ± 12%, with pyroxene >> olivine (total pyroxene/(total pyroxene + olivine) = 87%), with equal proportions of low and high-Ca pyroxene). This contrasts slightly with the Luna 20 ferroan crystalline rocks that are dominated by pyroxene rather than mixtures of olivine + pyroxene.

Olivine and Plagioclase
On a traditional Mg# (mafic silicates)-An# (plagioclase) diagram for distinguishing highland crustal lithologies, the compositions of coexisting olivine and plagioclase in Luna 20 highland crystalline assemblages plot within the Mg-suite and FAN-suite fields ( Figure 6). Unlike the range of FANs observed in the Apollo sample suite, the Luna 20 samples with FAN affinities only plot in the most Mg-rich portion of the FAN field. Those samples with an Mg-suite affinity show a somewhat limited relationship between lithology and distribution in the Mg-suite field, although in a very general sense the spinel troctolites tend to plot at higher Mg# and An#, while the norites, gabbronorites, and gabbros plot at lower values of Mg# and An#. Troctolites are in intermediate positions in this plot.

Pyroxene
Compositions are plotted within pyroxene quadrilaterals in Figure 7. As expected, lithologies that plot within the FAN field in Figure 6 have pyroxenes that are overall more iron-rich ( Figure 7) than those that plot in the Mg-suite field ( Figure 7). FANs have mineral assemblages with both low-and high-Ca pyroxene, whereas the Mg-suite members have mostly low-Ca pyroxene. Low-Ca pyroxenes in the FANs plot predominantly between Wo 10-5 , whereas the high-Ca pyroxenes plot between Wo 45-30. There are pyroxenes that plot along mixing lines between the high-and low-Ca pyroxenes, indicative of fine exsolution lamellae. Based on the superimposed isotherms, most of the pyroxenes experienced low temperature reequilibration. Low-Ca pyroxenes in the Mg-suite samples plot predominantly between Wo 15-2 , and the high-Ca pyroxenes plot between Wo 48-40 . As in the FANs, some Mg-suite pyroxenes plot along mixing lines between the high-and low-Ca pyroxenes. Most appear to have equilibrated at temperatures near 800°C.  Warren (1993), Norman et al., 1995, Shervais andMcGee (1999), and Shearer et al. (2015). Data from this study (see Figure 3), companion paper by Simon et al. (2022), unpublished analyses of Conrad et al. (1973), and published analyses in special issue of Geochimica et Cosmochimica Acta 37 (1973). Data points do not include lithic fragments that are unambiguously impact melts such as shown in Figure 3h or lithologies that exhibit substantial mineral composition heterogeneity. Plotted data does include potential granulites such as shown in Figure 3i and ambiguous lithologies such as in Figures 3f and 3g.

Spinel
Of 166 particles examined in our study of Luna 20 lithic fragments, 31 contain spinel; of these, 12 are impact melt rocks and 13 are crystalline igneous rocks that we are considering in this manuscript. All are feldspar-rich but not modally spinel-rich, which is typical of spinel-bearing lithologies among Apollo samples and unlike the one found in lunar meteorite ALHA 81005 (Gross & Treiman, 2011;Gross et al., 2014) or suggested from orbital data (e.g., Pieters et al., 2011). Spinel was also found in fused soil particles and devitrified glass. Spinel grain sizes are mostly <20 μm. Most occurrences are anhedral. Compositions range from nearly pure MgAl 2 O 4 to chromian ulvöspinel. Fe-Mg and Cr-Al relationships are illustrated in Figure 8. Contents of Fe are strongly anticorrelated with Mg, and Cr is strongly anticorrelated with Al. Cr-rich grains are richer in V than Cr-poor grains. Spinel in spinel troctolites is relatively Cr-poor and Al-rich, as previously reported (Brett et al., 1973;Haggerty, 1973). Cr-rich ulvöspinel is more common in plutonic than in impact melt rocks, and its composition is similar to previously reported Luna 20 Cr-ulvöspinel compositions (Brett et al., 1973;Haggerty, 1973). One melt rock has Cr-spinel enclosed in Mg-Al spinel ( Figure 4 of Simon et al., 2022). Simon et al. (2021Simon et al. ( , 2022 discussed compositional variation, spinel crystal chemistry, and petrologic interpretation in much greater detail.

Bulk Chemical Composition
Individual Luna 20 samples were not available for this study, so the bulk compositions we consider are the major element data produced by modal recombination of our modal data and mineral chemistries, unpublished data previously generated by broad beam EPMA analysis of lithic fragments at the Institute of Meteoritics (Conrad et al., 1973), and geochemical data from studies conducted soon after the return of Luna 20 (Laul & Schmitt, 1973;Laul et al., 1982;Philpotts et al., 1972;Prinz et al., 1973;Taylor et al., 1973).
Major element analyses illustrate overlaps and differences among anorthosites (FANs), noritic and troctolitic anorthosites (FANs), anorthositic norites and troctolites (Mg-suite), and spinel troctolites (Mg-suite). In Figure 9, bulk Al 2 O 3 contents are plotted against bulk FeO contents ( Figure 9a) and against Mg# (Figure 9b). The four categories of highland lithologies define a trajectory of decreasing Al 2 O 3 (%) and increasing FeO (%) from anorthosites ⇒ noritic and troctolitic anorthosites ⇒ Mg-suite norites and troctolites (Figure 9a). There is considerable overlap among the Mg-suite lithologies. The anorthosites sampled at Luna 20 are similar to those collected at the Apollo sites. For example, Apollo 16 anorthosite 60025 lies on the trend of the Luna 20 anorthosites plotted in Figure 9a. Although generally finer-grained, the anorthositic norites and troctolites overlap with Apollo mission-collected norites and troctolites (e.g., 78235,76535,78255,76335). The ultramafic rocks (pyroxenites, dunites) and mare basalts are significantly distinct from Luna 20 lithologies (Figure 9a). There are similar relationships with regards to these lithologies and Mg/(Mg + Fe) among the Luna 20 lithologies and between the Luna 20 and Apollo highland lithologies (Figure 9b).
Representative minor (e.g., TiO 2 ) and trace element abundances are presented in Figure 10. The plot of TiO 2 versus FeO (Figure 10a) shows relationships among the different crystalline lithologies similar to those illustrated in Figure 9. In the sequence from anorthosites ⇒ noritic and troctolitic anorthosites ⇒ Mg-suite norites and  (Simon et al., 1981). Pyroxene from the Mg suite lithologies are shown in the lower quadrilateral. Other data from this study, companion paper by Simon et al. (2022), unpublished analyses of Conrad et al. (1973), and published analyses in special issue of Geochimica et Cosmochimica Acta 37 (1973). troctolites, there are increases in both FeO and TiO 2 contents. Compared to the samples collected from the Apollo sites, the Luna 20 anorthositic norites and troctolites (±spinel) overlap with many of the Mg-suite lithologies, and the anorthosites overlap with the FANs. It is quite revealing that the anorthositic norites and troctolites (±spinel) do not overlap with the compositions of the Crisium melt rocks (dark green ellipses in Figure 10a) identified by Spudis and Sliz (2016) through "impact-derived windows" through the mare basalts within the Crisium basin. The impact melts within the Crisium basin are more FeO-rich than the Luna 20 crystalline rocks examined in this study (Figure 10a). Based on major element correlations, the impact melt rocks in Crisium presumably have lower Al 2 O 3 abundances (15% Al 2 O 3 ), higher FeO (10%), and perhaps lower Mg/(Mg + Fe) than the crystalline rocks examined in this study.
There are very few Luna 20 lithic fragments that have been both fully described petrographically and analyzed for trace elements (Laul & Schmitt, 1973;Vinogradov, 1973). Data for these samples and Luna 20 soils are plotted in Figures 10b-10d using potential indicators of KREEP (K 2 O, Th, La N /Lu N ) to better understand the potential role of the KREEP component in magmatic rocks associated with Crisium. The samples are generally incompatible element-poor compared to KREEP-rich norites and breccias from Apollo sites (e.g., clasts in 15455). However, they are similar to other troctolites and norites from Apollo landing sites. Therefore, based on only these few Luna 20 samples, it is difficult to ascertain the role of KREEP in their origin. This general lack of incompatible element-rich lithologies (e.g., KREEP-rich norites) reflects reduced abundance of a KREEP component, which may be attributed to the limited addition of Imbrium ejecta to the site and/or the limited contribution of KREEP to the magmatic lithologies making up the crust.

Introduction
The remotely sensed data from Lunar Prospector, Chandrayaan-1, Kaguya, and LRO enable the mineralogy and geochemistry of the Luna 20 samples to be placed within a regional context and compared to the Crisium rim ( Figures 11 and 12). Further, the Luna 20 samples provide ground truth for the petrology of the nHFT Terrain south of the Crisium basin (e.g., Moriarty et al., 2022).  Brett et al. (1973) combined with analyses from this study and companion manuscript by Simon et al. (2022). Data were plotted with the algorithm and software of Antonini et al. (2020).

Mineralogy
To first order, spectral diversity across the Crisium region is dominated by variations in the composition of pyroxenes and the relative abundances of plagioclase and mafic minerals ( Figure 11). As noted above, pyroxenes in Luna 20 samples preserve evidence of a complicated crystallization history, including complex and fine exsolution lamellae and partial inversion of pigeonite to orthopyroxene (Ghose et al., 1973;Meyer, 1973Meyer, , 2009Prinz et al., 1973). As noted by Moriarty and Pieters (2016), this complexity precludes the derivation of precise, quantitative mineral compositions from orbital spectra. Nevertheless, remote spectral analyses are critical for determining compositional heterogeneity on regional scales and in geologic context.
The Crisium interior is dominated by Fe-and Ca-rich mare basalts. However, several impact craters that penetrated or pre-date the mare (e.g., Peirce, Picard, Yerkes) exhibit relatively Mg-rich pyroxenes (e.g., Runyon et al., 2019). It is likely that these craters expose Crisium impact melt rocks. Since the depth of melting exceeds the depth of excavation for basin-scale impacts (e.g., Cintala & Grieve, 1998), and lunar impacts excavate materials from  depths up to ∼10% of their diameter (e.g., Melosh, 1989), Crisium's impact melt pool should include abundant materials from beneath the ∼40 km thick lunar crust (Wieczorek et al., 2013). Future numerical modeling of the Crisium-forming impact would greatly refine our understanding of the relevant depths of melting and excavation.
Crisium's rim and adjacent highlands exhibit notable spectral diversity (Figure 11). Apparent variations in pyroxene abundance and composition may be related to mixing between Crisium ejecta (including upper crust and lower crust/upper mantle components), adjacent highlands material, and ejecta from later impact events, convolved with differences in optical maturity. Weak 1-micron integrated band depth across both the rim and nHFT indicate the lack of abundant widespread mafic materials (probably significantly less than 10% ). Although band center measurements can be unreliable for weak absorption bands given the presence of noise and observation artifacts in M 3 data, both the rim and nHFT exhibit relatively short-wavelength 2-micron band centers compared to local mare basalts, indicating the presence of pyroxenes lower in Ca, Fe. These may be pyroxene compositions similar to possible Crisium impact melt exposed by the Peirce, Picard, and Yerkes craters (Figure 11a) (Runyon et al., 2019), although band centers across the rim and nHFT are slightly longer in wavelength. This possible link between noritic Crisium ejecta and noritic Crisium impact melt implies that some noritic lithologies observed in Luna 20 samples may be related to deep-seated materials excavated from the lower crust/upper mantle by the Crisium-forming impact. Yet, in the samples from Luna 20, there is very limited mineralogical and textural indication (e.g., pyroxene exsolution) of a deep crustal origin.
The 1-micron integrated band depth map in Figure 11b demonstrates that both the Crisium rim and nHFT are low in mafic minerals, although there is evidence for heterogeneity on local scales. Overall, band depths are slightly elevated in the nHFT relative to the rim. This subtle difference may reflect differences between relatively pristine uplifted target material in the rim and more well-developed soils in the nHFT.
Clear spectral differences between the rim and nHFT are observed in the Pure Anorthosite Ratio map (Figure 11c), using similar methods as discussed by Pieters et al. (2014). This parameter (defined as the ratio of the sum of reflectance values at 1 and 1.5 micron to the reflectance value a 1.25 micron) targets the presence of pure, crystalline plagioclase, which exhibits an absorption band centered at 1.25 micron. Outcrops of crystalline plagioclase have been observed around the Crisium rim (Donaldson , supporting the presence of widespread crystalline plagioclase indicated by this parameter map. In conjunction with the low mafic abundances indicated by the 1-micron integrated band depth (Figure 11b), this suggests that the rim of Crisium has exposed primarily anorthositic crust, whereas the surrounding nHFT does not exhibit pure crystalline plagioclase, consistent with a more well-developed regolith.
A potential caveat is that olivine's primary absorption feature extends to wavelengths overlapping the plagioclase absorption and could potentially cause false positive identifications in the Pure Anorthosite Ratio Moriarty et al., 2022). Olivine spectral signatures have been identified in localized outcrops around the Crisium rim (e.g., Corley et al., 2018), so it is plausible that some areas exhibiting high Pure Anorthosite Ratio values are dominated by olivine rather than pure anorthosite spectral signatures.  Prettyman et al., 2002) for the Crisium region. (a) FeO, (b) Th. (c) TiO 2 abundance from LROC (Sato et al., 2017). These maps are stretched to highlight regional variations.
Apart from localized outcrops, is the widespread elevation in Pure Anorthosite Ratio across the Crisium rim due to anorthosite or olivine? Spectra of plagioclase-olivine mixtures provide insight into this question. Cheek and Pieters (2014) demonstrated that olivine dominates the spectra of plagioclase-olivine mixtures at abundances as low as 2%-5%, significantly increasing band depths. Because band depths are generally very low across the Crisium rim (Figure 11b), this suggests that the widespread Pure Anorthosite Ratio elevation across the rim is probably due to crystalline plagioclase, although minor (<∼5%) amounts of olivine cant be completely ruled out. Further, Warren and Korotev (2022) argued that the evolution of anorthositic regolith may preferentially lead to an overestimation of the "pure-anorthosite ratio" values due to the importance of the 10-20 micron size fraction in bulk reflectance measurements and the importance of impact/agglutinitic glass in this size fraction. This process may provide artificial enrichment at the rim of Crisium, but this cannot rule out the difference between the rim and the nHFT.
The final conclusions based on the orbital mineralogy observations are (a) the nHFT exhibits some local heterogeneity but is regionally consistent, indicating that samples collected from the Luna 20 site are probably representative; (b) pyroxenes across the rim and nHFT are relatively low in Ca, Fe, possibly similar to Crisium impact melt sheet material; and (c) the nHFT has higher mafic phase abundance and lower crystalline plagioclase (and/ or olivine) abundance than the Crisium rim materials.

Geochemistry
Elemental abundance maps from the Lunar Prospector gamma ray spectrometer ( Figure 12) provide further insight into the compositional context of Luna 20 samples and the chemical variation observed in the Crisium region. Ground truth of the orbital data is provided by the samples collected and analyzed from the Luna 20 and 24 sites. For example, regolith (<1 mm size fraction) from Luna 20 has the following chemistry: FeO = 7.27 wt%, TiO 2 = 0.48 wt%, K = 564 ppm, and Th = 1.32 ppm (Papike et al., , 1998. Near the other end of the spectrum, the regolith from the Luna 24 site is dominated by a mare basalt component and has the following chemistry: FeO = 20.2 wt%, TiO 2 = 1.0 wt%, K = 224 ppm, and Th = 0.40 ppm (Papike et al., , 1998. Compared to Luna 20 site, the adjacent nHFT and Crisium rim is lower in FeO (∼5.0 wt %) and Th. As shown in Figure 12c, TiO 2 shows limited variation in the nHFT (∼0.50 wt%). Zhu et al. (2013) illustrated limited variation of K within the Crisium rim (400-700 ppm), nHFT (400-800 ppm), and the Luna 20 site (800 ppm). The Luna 20 site appears to be slightly higher in FeO (7.0 wt%) and Th (1.4 ppm) than much of the nHFT (5-7 wt%). It is unlikely that these slight enrichments at the Luna 20 site are products of additions of mare basalts and other non-local ejecta, as significant volumes of these materials have not been identified in the Luna 20 samples. The FeO, Th, TiO 2 , and K abundances at the Luna 20 site and in the nHFT indicate that the returned samples represent a well-mixed regolith incorporating multiple lithologies from the region. The orbital data suggest subtle differences between the Crisium rim and the surrounding nHFT. Further, the PKT has an average higher K (1768 ± 530, 2004 ± 708) (Zhu et al., 2013) and Th (4.8 ppm) (e.g., Jolliff et al., 2000) than the region immediately surrounding the Crisium basin.
Relative to the nHFT, the rim of Crisium exhibits lower Fe, Th, and K abundances. This is not surprising as it was shown above (Figure 11c) to have a higher anorthosite component than the nHFT. Surface materials across Crisium's rim exhibit elemental abundances consistent with impact-driven mixing between crustal materials, Crisium ejecta, mare basalts, and non-local ejecta. Specific evidence for this is observed at two young craters (Proclus and Condorcet A; see Figure 1) that appear to have excavated relatively pure highlands crustal materials (very low Fe, Th, Ti) from beneath mixed, intermediate surface materials.
The rim of Crisium, the nHFT and presumably the pure highland crustal material (e.g., as viewed through Proclus and Condorcet A) has lower Th abundances than the PKT to the west. This low Th abundance is indicative of a lower regional KREEP component. This appears to be consistent with the small population of the potentially KREEPy Luna 20 sample analyses shown in Figure 10 and discussed in Section 4.4.

Modeling Contribution of Material to the Luna 20 Site
Using the approach of Petro and Pieters (2006), for the region surrounding the Luna 20 site, we estimated the contribution of post-Crisium craters to the sample site to better understand the provenance of regolith components. The closest large crater is Ameghino, a 9 km-diameter, Imbrian-aged crater (Figure 1). Based on crater scaling models (Petro and Pieters, 2006), we estimate ∼80 cm of ejecta could have been introduced to the Luna 20 sampling site. The Ameghino impact occurred in Crisium ejecta, which itself contains a non-negligible amount of Crisium impact melt and crustal materials.

Extending the Luna 20 Samples Beyond the Collection Site
The Luna 20 core penetrated approximately 25 cm into the regolith. The core exhibited no visible layering and based on the proportions of fused soil, it is likely sub-mature in character (Simon et al., 1981). Estimates of maturity based on I S /FeO were not determined (Morris, 1978). The material sampled by the Luna 20 mission was predominantly derived locally with a large contribution from Ameghino impact crater and to a lesser degree from other post-Crisium local craters (see Section 5.4). The orbital data from recent missions such as Lunar Prospector, Kaguya, Chandrayaan-1, and LRO (Figures 11 and 12) indicate that the noritic Hilly and Furrowed Terrain  is fairly homogenous and therefore mixing at the Luna 20 site predominantly involved material from this terrain. Returned samples indicate only minor amounts of exotic materials. On a larger scale, a large proportion of nHFT represents materials excavated by the Crisium event. Three-dimensional numerical impact simulations of the significantly larger South Pole-Aitkin (SPA) basin-forming event (estimates of the SPA transient cavity are 800-1,400 km; those for the Crisium transient cavity are 487-560 km) by Melosh et al. (2017) indicate that the volume of deepest material should decrease outward from the inner crater rim, but mantle material (50-75 km) should still be ejected past the final crater rim. Further, the deeper-seated ejecta should lie in the upper portion of the ejecta blanket with shallower crustal material lying in the deeper portion of the ejecta stratigraphy, assuming an overturned "flap" of material. Therefore, deeper material excavated by the Crisium event should be in the nHFT sampled by Luna 20. However, this estimated stratigraphy may be modified by several complex processes both during and after impact (summarized by ).

Relationship Between Luna 20 Samples and the Lithologies Associated With the Inner Ring of Mare Crisium
Based upon two-and three-dimensional numerical impact simulations of lunar basin formation (e.g., Melosh et al., 2017;Miljkovic et al., 2015), there are potential differences between the inner ring of the Crisium basin and the nHFT sampled by Luna 20. Potentially, there is a greater volume of deeper lithologies in the inner ring that were excavated by the Crisium event than in the nHFT. This may be reflected in the presence of shallow mantle or deep crust mineral assemblages, as indicated by differences in mineralogy (e.g., olivine, pyroxene, crystalline plagioclase) and chemistry (e.g., Al 2 O 3 , FeO, MgO). In a comparison between samples from Luna 20 and the modal mineralogy of the Crisium basin rim, estimated with radiative transfer modeling for M 3 olivine spectra (Corley et al., 2018), the anorthosites at Luna 20 appear to be dominated by pyroxene (i.e., pyroxene/(total pyroxene + olivine) ∼ 100%), whereas the Crisium basin rim has more olivine though with pyroxene >> olivine (total pyroxene/(total pyroxene + olivine) = 87%). Further, there appears to be more diversity of more mafic rocks at the Luna 20 site compared to the Crisium inner rim ( Figure 5). Further, as implied by Figures 5 and 11, the anorthosite (and/or olivine) component (e.g., plagioclase) decreases with distance from the Crisium rim to the Luna 20 sampling site. These mineralogical differences are reflected by increases in incompatible element and FeO contents from the rim to the Luna 20 site in the nHFT. Therefore, there are distinct mineralogical differences between the rim of Crisium and the Luna 20 site. Does this also reflect distinct crustal and mantle lithologies and depths of origin for these regions surrounding Crisium?

Depth of Excavation
There are three approaches that can be used to better understand the depth of origin of the crystalline lithologies that are documented from Luna 20: (a) interpretation of bulk rock compositions; (b) constraints from exsolution lamellae in pyroxenes; and (c) implications of the presence of spinel in some lithologies.
Although composition cannot be directly related to the depth of origin, it can provide insights into crustal or mantle lithologies excavated. Lithic fragments making up the magmatic suite of Luna samples consist of ferroan anorthosites with pyroxene as the dominant mafic silicate, and a suite of norites and troctolites (±spinel) with mafic silicates with higher Mg# than FANs (Figures 5a and 6). There are also individual mineral fragments that were potentially derived from members of these two suites. Reid et al. (1973) and Papike et al. (1982) concluded that the Luna 20 soils and lithic fragments are dominated by anorthositic norites and troctolites (±spinel). Concepts for the composition of the upper mantle developed by Wieczorek et al. (2006Wieczorek et al. ( , 2013, Elkins-Tanton et al. (2011), Melosh et al. (2017, Elardo et al. (2020), , and  involve LMO cumulates perhaps modified by cumulate overturn or mixing. Under these different scenarios, mineralogies for the upper mantle generally consist of Mg-rich dunites or orthopyroxenites or hybridized mantle consisting of mixtures of early-and late-stage LMO cumulates. There were no mineral assemblages (e.g., orthopyroxene, dunite, evolved LMO cumulates) that clearly have a lunar mantle origin based on these contemporary concepts. This suggests that Luna 20 magmatic rocks were excavated from the crust. Compared to the igneous lithic fragments returned by Luna 20, the dominant rock types identified by orbital data making up the rim of Crisium are plagioclase-rich with predominantly ferroan mafic phases (Mg# = 65), so they are most likely dominated by FAN-suite lithologies such as anorthosites, noritic anorthosites, and troctolitic anorthosites (Corley et al., 2018;Figure 5b).
Similar to the Luna 20 site, Corley et al. (2018) and Lemelin et al. (2019) failed to detect any ultramafic lithologies in the rim of Crisium, despite the spatial resolution of ∼62 m per pixel of the Multiband Imager data set and the results of hydrocode models (Miljković et al., 2015) that predict the presence of minor mantle material at Crisium as well as other basins. To rationalize these observations, Lemelin et al. (2019) suggested that the impact process is extremely efficient at mixing plagioclase-rich crustal material with mantle material into the ring material on a scale of several tens of meters. This efficient mixing may obscure or dilute evidence for a mantle component, as observed in the South Pole-Aitken Basin . However, the samples, such those collected at the Luna 20 site have a resolution on the order of 10s of microns, and no hidden excavated mantle component was identified at that scale.
The ferroan anorthosites and the Mg-suite rocks are texturally distinct from each other. Among the Luna 20 samples, ferroan anorthosites are commonly represented by coarse (>400 μm), single grains of calcic plagioclase with finer inclusions, commonly <50 μm, of ferroan mafic silicates (Figures 3e and 3f). Larger grains of plagioclase were not identified in the present study as the samples examined were of a limited size range (250-500 μm). These plutonic grain sizes and textures are typical of many large FAN samples collected during the Apollo program (e.g., Marks et al., 2019;McCallum & O'Brien, 1996;McCallum & Schwartz, 2001;Papike et al., 1998). On the other hand, the magnesian anorthositic norites and troctolites have much finer-grained textures (Figures 3a-3d), commonly on the scale of 100 μm or less. This contrasts with some of the Mg-suite rocks that crystallized in the deep crust. For example, plutonic troctolite 76535 has a grain size of 2-3 mm and has a calculated a depth of origin of approximately 47 km (e.g., Dymek et al., 1975;Gooley et al., 1974;. Although not quantitative, the typical grain sizes of the ferroan anorthosites suggest derivation from mid-to shallow crustal levels, whereas the grain sizes of the Mg-suite rocks may reflect emplacement in shallower crustal regimes. A more quantitative tool for estimating the depth of origin and cooling rate are the widths of exsolution lamellae observed in pyroxene. In a single pyroxene grain with a composition similar to that expected in ferroan anorthosites and associated with calcic plagioclase (An 97 ), Ghose et al. (1973) identified complex and fine exsolution lamellae (from 10Å to 1 μm) and partial inversion of pigeonite to orthopyroxene. In a comparison to finer lamellae observed in mare basalts (near surface) and coarser lamellae observed in terrestrial pyroxenes from the Skaergaard intrusion (2 km), they concluded that the exsolution represented crystallization in a lunar crustal environment at a depth of less than 2 km. Pyroxene exsolution features in other ferroan anorthosites from Luna 20 are varied, with some of the thicker lamellae reaching a width of 2-3 μm (Figure 4). How do these wider pyroxene lamellae compare with others found in ferroan anorthosites? Marks et al. (2019) reconstructed the thermal history of ferroan anorthosites 60016 and 60025 by integrating exsolution lamellae in pyroxene with multiple chronometers with distinct blocking temperatures (Ar-Ar, Rb-Sr, Sm-Nd). McCallum and O'Brien (1996) estimated that 60025 cooled at a rate of 18°C/Myr from 1,100°C to 800°C and from this they calculated a depth of emplacement of 21 km. Cooling at a moderate rate resulted in partial homogenization of the pyroxene and olivine grains and exsolution of clinopyroxene from low-Ca pyroxene, producing thick exsolution lamellae that are ∼40 μm in width. Marks et al. (2019) concluded that the pyroxene reached a temperature of approximately 850°C (Sm-Nd blocking temperature) by 4,367 Ma. The final stage of cooling likely occurred in the upper crust, where a secondary set of fine exsolution lamellae was produced (Hodges & Kushiro, 1973), and where the sample was heavily brecciated (McCallum & O'Brien, 1996). Marks et al. (2019) determined that this excavation event occurred at 4,302 Ma while this lithology was at a temperature below 800°C.
Ferroan anorthosite represented by a clast in sample 60016 had a different thermal history. The amoeboid-shaped pyroxene and olivine suggest a slower cooling rate (<<18°C/Myr), keeping the lithology above the pyroxene solvus for an extended period of time. This prevented exsolution in the pyroxene but enabled subsolidus growth of olivine and pyroxene. At 4,302 Myr, the sample was excavated from a crustal environment with a temperature above 850°C and quickly cooled in the upper crust, producing fine exsolution lamellae (1-3 μm in width). Based on a comparison with these two Apollo 16 samples, the ferroan anorthosites from Luna 20 have a somewhat different thermal history. The exsolution lamellae observed in pyroxene (Figure 4) in some of the ferroan anorthosites suggest rapid cooling in a shallow crustal environment (2-4 km). Based solely on the exsolution, this could represent an emplacement depth or a depth in an ejecta blanket following excavation. The former is much more likely. Unlike 60016, there is no textural evidence for extensive subsolidus growth of pyroxene. Further, the large anorthite grains in the Luna 20 core have Ar-Ar ages that predate the Crisium event by at least 300 Myr (Cohen et al., 2001;Swindle et al., 1991). If the ferroan anorthosites were excavated by the Crisium event, the formation of the exsolution lamellae may be more closely related to the preserved crystallization age rather than the younger excavation event.
In the more magnesian pyroxenes, presumably from Mg-suite lithologies, Ghose et al. (1973) identified fine augite exsolution lamellae (from 10Å to 1 μm) in orthopyroxene. They further concluded that the orthopyroxene was primary, and not derived from the inversion of pigeonite. In our examination of pyroxene from these Mg-suite lithologies, the scale of the lamellae was at the lower limits of our SEM and EPMA resolution capabilities. Clearly, any exsolution products in these pyroxenes are on the sub-micron scale. Based on the nature of the exsolution lamellae in the pyroxenes from the Mg-suite, these lithologies cooled quickly in fairly shallow crustal environments at depths of approximately 2 km (Ghose et al., 1973;Marks et al., 2019;McCallum & O'Brien, 1996;McCallum & Schwartz, 2001;Shearer and Papike, 2005).
In contrast to the conditions of crystallization implied by composition, textures, and exsolution lamellae in pyroxenes, the occurrence of Mg-Al spinel in the troctolites may imply a deeper origin for the Mg-suite rocks at the Luna 20 site. Mg-Al spinel is relatively rare in lunar rocks and occurs predominantly in troctolites, troctolitic cataclasites, and unique lunar meteorites (e.g., ALHA 81005; Gross & Treiman, 2011). Although a number of previous studies have indorsed a deep crust-shallow mantle origin for spinel-bearing lithologies including samples from Luna 20 (e.g., Anderson, 1973;Baker & Herzberg, 1980;Bence et al., 1974;Dymek et al., 1976;Marvin et al., 1989;Prinz et al., 1973;Snyder et al., 1999;Takeda et al., 2006), Gross and Treiman (2011) outlined three distinct hypotheses for the endogenic origin of lunar spinel-bearing lithologies on the Moon. First, lunar spinel-bearing rocks could have formed as cumulates or restites from basaltic systems at high pressures (e.g., Delano, 1976;Marvin et al., 1989;Prinz et al., 1973), similar to spinel-peridotites formed in Earth's mantle at pressures between 1 and 2-3 GPa. Basaltic magmas at such high pressures can produce phenocrysts of spinel, which are rarely found at Earth's surface. Second, lunar spinel-bearing rocks could have formed from troctolitic melts at relatively low pressures (Marvin & Walker, 1985;Walker et al., 1973). Third, spinel-rich rocks could form by the assimilation of anorthosite into olivine-saturated melts to form troctolitic melt compositions (Finnila et al., 1994;Gross & Treiman, 2011;Morgan et al., 2006;Prissel et al., 2014;Treiman et al., 2019). Both scenarios 2 and 3 are consistent with the shallow origin implied by the exsolution lamellae and the finer-grained textures observed in the Luna 20 Mg-suite rocks.
Although unrelated to the Mg-suite rocks at the Luna 20 site, Simon et al. (2021Simon et al. ( , 2022 illustrated that the crystallization of Mg-Al spinel occurs at lower pressures in fairly rapidly cooled impact-produced melt breccia. In Luna 20 sample 22003, 6 (fragment 3), subhedral Cr-Fe spinel is enclosed in subhedral Mg-Al spinel, and relatively Fe-rich olivine is enclosed in more forsteritic olivine. Plagioclase laths appear to have nucleated on the olivine and radiate away from it. There are fine intergrowths of pyroxene and plagioclase in the matrix. Although this melt breccia is a product of impact, many of the characteristics are similar to those generated by plagioclase assimilation models proposed for the formation of Mg-Al-spinel-bearing magmatic assemblages. Perhaps this melt breccia assemblage is a product of an analogous process involving the mixing and melting of appropriate crustal target rocks. The final conclusion that can be reached by the low-pressure Mg-Al models proposed above and the observations of Simon et al. (2021Simon et al. ( , 2022 on impact melt breccias is that the occurrence of Mg-Al spinel in the spinel troctolites at the Luna 20 site does not constrain the depth of origin. Many of the petrologic observations made from the magmatic lithologies at Luna 20 provide little evidence for mantle or deep crustal assemblages being excavated. However, orbital observations and excavation modeling based on coupling parameters and equations of Miljkovic et al. (2017) predict that deep crustal (28 km) and shallow mantle (73 km) components in near-equal proportions make up the rim of the Crisium basin (Lemelin et al., 2019). There are several interpretations for this apparent change in depth of origin for those samples examined from Luna 20 (shallow crust) and material from the rim of Crisium (deep crust) and impact melt sheet (mantle + crust melting) revealed by younger craters penetrating the Crisium basin-filling basalts (Runyon et al., 2019). First, if most impact models are correct, there must be a transition from deeper lithologies preserved in the Crisium rim (and incorporated into the impact melt sheet) to more shallow rocks preserved in the outer portions of the nHFT. Second, the lack of deep lithologies at the Luna 20 site seems contradictory to models such as Melosh et al. (2017), but perhaps the nHFT is mixed such that the Luna 20 mission did not sample this material. Third, as illustrated in Sections 6.1 and 6.2, the Crisium rim is much more plagioclase-rich and mafic-poor and has lower incompatible element contents than the material making up the nHFT. This difference suggests a higher FAN component in the rim. This perhaps represents FANs excavated from the deep lunar crust. The interpretation of the orbital data and the Luna 20 observations further suggests that the crust at Crisium consists of FANs and Mg-suite rocks, and therefore the Mg-rich olivine within the rim is mostly related to deep, complementary versions of the Mg-suite rocks from Luna 20. Finally, we must conclude that either the material associated with the Crisium basin as sampled at the Luna 20 landing site is not derived from the lunar mantle, that our "Earth-analog" assumptions concerning the upper mantle of the Moon are incorrect, or that a significant proportion of the excavated mantle was incorporated as a component within the impact melt sheet. The new M 3 observations presented here (specifically the weak 1-micron absorption bands across the rim demonstrated in Figure 11b) are unequivocally inconsistent with the presence of significant volumes of ultramafic mantle materials as previously claimed (Lemelin et al., 2019).

Petrogenesis of the Crystalline Rocks at the Luna 20 Site
It was recognized in earliest studies that two suites of crystalline highland rocks occurred at the Luna 20 site (e.g., Taylor et al., 1973). The data collected in the present study and integrated with previous data indicate that these two suites represent the ferroan anorthosites and the Mg-suite. The former are lunar magma ocean flotation cumulates and the latter are cumulates that crystallized from shallowly emplaced magmas. The Ar-Ar ages for these two suites of rocks indicate that they represent highlands crust formed prior to the Crisium event by at least 300 Myr (Cohen et al., 2001;Swindle et al., 1991). This would be consistent with many crystallization ages derived for similar lithologies collected by the Apollo missions (e.g., Borg et al., 2015Borg et al., , 2020Shearer & Papike, 2005). The orthopyroxene and olivine observed in these two suites are products of crustal magmatism and are not contributions from the mantle. Numerous models have been proposed for the origin of Mg-suite magmas (e.g., see reviews in Shearer and Papike (2005)). Some of the observations observed in rocks from the Luna 20 site offer constraints on this origin.
To produce Mg-suite magmas in a planetary body with a sub-chondritic (Al 2 O 3 /MgO) mantle, Elardo et al. (2020) suggested melting of a hybrid cumulate package consisting of deep mantle dunite, crustal anorthosite, and urKREEP at the base of the crust under the Procellarum KREEP Terrane (PKT). Estimates of the Mg# of the various LMO cumulate component range from 92 in the earliest dunite cumulate to 59 in urKREEP (e.g., Elardo et al., 2011Elardo et al., , 2020Elkins-Tanton et al., 2011;Rapp and Draper, 2018;Warren, 1989). In this model, these three LMO cumulate components are brought into close proximity by overturn resulting from the unstable density and thermal structure of the cumulate pile. Early and deep dunite LMO cumulates with an Mg# of ∼90 rose to the base of the ferroan anorthosite lunar crust due to their buoyancy relative to colder, later, and denser Fe-and Ti-rich cumulates. This produces a potential high-Al 2 O 3 /MgO source. As a result of the increased temperature at the base of the lunar crust generated by the hot, early cumulates and the heat-producing elements associated with KREEP, this hybridized source rock melted to form Mg-suite magmas saturated in Mg-rich olivine and calcic plagioclase, with a substantial KREEP component. How does this model fit observations made for the Mg-suite lithologies associated with Crisium?
Within this dunite-anorthosite (±KREEP) hybrid source near the crust-mantle interface, melting would have occurred at the olivine-pyroxene-spinel peritectic or the plagioclase-pyroxene-spinel peritectic within the Olivine-Anorthite-Silica ternary. The peritectic temperature at which melting would occur is dependent upon the proportions of dunite and anorthosite. Still, these melts would have super-chondritic, relatively high Al 2 O 3 /MgO. Once these melts were transported to the upper lunar crust environment, they would no longer be in the spinel primary crystallization field. At low pressures, these melts would crystallize olivine + plagioclase. Although the crystallization products of these melt compositions would be analogous to most Mg-suite rocks, spinel would not be in the assemblage, and this does not fit the textural observations of the spinel troctolites. Further, olivine in the spinel troctolites from Luna 20 has resorption features  which cannot be accounted for by crystallization in the olivine field, along the plagioclase-olivine cotectic, or at the Olivine + Low-Ca pyroxene + Plagioclase peritectic. The observed olivine texture was most likely produced by the reaction Olivine + Melt ⇒ Low-Ca pyroxene + Plagioclase. Although these observations argue against the origin of Mg-suite magmas capable of crystallizing spinel, they do not eliminate a hybrid source model for the production of all other Mg-suite lithologies.
A Mg-suite hybridization model that requires an upper mantle dominated by olivine is not consistent with remote sensing observations. Remotely collected data are more consistent with an upper mantle that is spectrally dominated by pyroxene. Specifically, mantle materials ejected and melted during the South Pole-Aitken basin-forming event appear to be dominated by pyroxene + ilmenite-bearing cumulates + urKREEP LMO dregs Moriarty & Pieters, 2013Nakamura et al., 2009).
Although this upper mantle model may be relevant to the SPA region of the Moon, is it relevant to Crisium and other portions of the Moon? The impact melt associated with the Crisium basin appears to be dominated by low-Ca pyroxene (Runyon et al., 2020). If mantle materials were a component in the production of the Crisium impact melt sheet (as suggested by impact models published by Miljković et al. (2015)), this would suggest that the mantle excavated by the Crisium event was dominated by pyroxene.
A final scenario for generating the Mg-suite and associated spinel troctolites is the assimilation of ferroan anorthosite crust by Mg-rich, mantle-derived magmas (Finnila et al., 1994;Gross & Treiman, 2011;Morgan et al., 2006;Prissel et al., 2014;Shearer & Papike, 2005;Treiman et al., 2019). This has been demonstrated to be a favorable process to account for observations of magmatic rocks with high spinel contents (e.g., pink-spinel anorthosites, ALHA 81005), but is it appropriate for these shallow troctolites (±spinel) and norites? Treiman (2011), Prissel et al. (2014), and Treiman et al. (2019) demonstrated that in the Forsterite-Anorthosite-SiO 2 ternary and Forsterite-Anorthite binary systems, the assimilation of anorthositic wall rock by an olivine-saturated melt could produce melts in the spinel stability field. These melts would have a crystallization sequence of spinel ⇒ spinel + plagioclase ⇒ spinel + melt react to plagioclase + olivine ⇒ plagioclase + olivine ⇒ olivine + melt reacts to plagioclase + low-Ca pyroxene. This appears to be the approximate fractional crystallization sequence recorded in the spinel troctolites. Gross and Treiman (2011) suggested that such a process would displace previously crystallized olivine off the liquidus, resulting in the resorption features observed in the olivine in the troctolite. The addition of a ferroan component could account for the relatively Fe-rich quench olivines found in the impact melt rock . The assimilation of FAN material by sub-chondritic (Al 2 O 3 /MgO) olivine-saturated magmas also provides a process for producing Mg-suite parent magmas. This petrogenetic hypothesis is the best fit thus far for the origin of the spinel troctolites. However, both the assimilation and source hybridization models are appropriate for the origin of most of the Mg-suite.

Mg-Suite Magmatism Outside the PKT
An important characteristic of Mg-suite rocks collected during the Apollo program is that they have a trace element signature that is consistent with a KREEPy component (Papike et al., 1994(Papike et al., , 1996. Snyder et al. (1995) suggested that magmas similar to KREEP basalts were parental magmas to the Mg-suite. Elardo et al. (2020) demonstrated that the KREEP component in the upper lunar mantle may enhance Mg-suite melt production through both melting point depression and radioactive heating. These observations imply that there is a fundamental petrogenetic relationship between KREEP and the Mg-suite. Do the heat-producing elements in KREEP drive Mg-suite magmatism, and thereby restrict this style of magmatism to the PKT region of the Moon?
The distribution of the Mg-suite may only be regional and closely associated with the PKT, or it may be Moonwide (with differing proportions of a KREEP component). Distinguishing between these differences in distribution is critical to understanding the origin of the Mg-suite magmas, and more broadly planetary differentiation and crustal building processes on an asymmetric planetary body. Using orbital data, Shearer et al. (2015) demonstrated (see their Figure 15) that Mg-suite magmatism may extend outside the PKT. The Crisium region is outside the PKT, and the Th levels in the region (Figures 10 and 12) suggest that the KREEP component is relatively low in abundance in the crust. Yet, the occurrence of Mg-suite lithologies at the Luna 20 site indicates that this style of magmatism is a significant component in the lunar crust.

Conclusions
It has been proposed that lunar mantle material was excavated by the Crisium event and perhaps other basin-forming events (e.g., Corley et al., 2018;Lemelin et al., 2019;Miljkovic et al., 2015;Wieczorek et al., 2013;Yamamoto et al., 2010). The rarity of dunites associated with these large basins has been attributed to its limited abundance in the upper lunar mantle (e.g., Arnold et al., 2016;Lucey et al., 2014;Melosh et al., 2017;Moriarty & Pieters, 2018;Sun & Lucey, 2023). There is limited evidence for this from the Luna 20 samples. Ultramafic rocks such as dunites and orthopyroxenites are rare, and Mg-rich olivine-and orthopyroxene-bearing-assemblages appear to be derived from magmatic rocks emplaced in the shallow crust. This is reflected by numerous lithologies with fine pyroxene exsolution lamellae, suggestive of rapid cooling in shallow crustal environments. Spinel-bearing lithologies at Crisium have been interpreted as representing deep crustal lithologies (30-40 km) (e.g., Cohen et al., 2001;Snyder et al., 1999). However, numerous studies (Gross & Treiman, 2011;Simon et al., 2022;Treiman et al., 2019;Walker et al., 1973Walker et al., , 1975 and our observations suggest an alternative interpretation in which the spinel represents crystallization in a shallow crustal environment. Do these lithologies that seemingly crystallized in the shallow crust represent Mg-suite and FAN plutonic suites or crystallization products of an impact melt sheet associated with the Crisium impact event? Modeling the crystallization sequence of large impact melt sheets on the Moon (e.g., Hurwitz & Kring, 2014;Vaughan & Head, 2014) produces abundant norites and ultramafic cumulates that appear not to match the lithologies represented by the present samples. Perhaps this difference is a product of the proportions and composition of crustal and mantle lithologies that were melted. Vaughan and Head (2014) and Hurwitz and Kring (2014), however, used a variety of starting compositions in their modeling. Along these lines of evidence, as illustrated in Figure 10a, the Luna 20 crystalline rocks appear to be different in composition from Crisium melt rocks. Another problem with the interpretation that these lithologies represent cumulates from a Crisium melt sheet is the chronology established for these rocks by Ar-Ar ages (Cohen et al., 2001 and references within). These data suggest crystallization ages that pre-date the Crisium event and fall within the range of ages for the Mg-suite. Although multiple chronometers would better define the crystallization history, we are left with the conclusion that these lithic fragments represent pre-Crisium episodes of shallowly emplaced Mg-suite magmas and FAN magmatism related to primordial differentiation.
From orbital data and empirical calculations of excavation during the Crisium basin forming event, it was previously proposed that the rim of Crisium consisted of lithologies derived from the upper mantle and deep crust. Our comparison between the Crisium ring and the Luna 20 site indicates that the inner ring has a higher ferroan anorthosite component. The mafic component of the ferroan anorthosite at the Luna 20 site is dominated by pyroxene. The Luna 20 mineral assemblages do not reflect derivation from the shallow mantle or deep crust. This may mean that deep lithologies are only in the inner ring and incorporated as components in the Crisium impact melt and that the outer portions of the Crisium ejecta are derived from the intermediate (FANs) and shallow crust (Mg-suite). These observations suggest that the inner ring of Crisium is dominated by deep crustal ferroan anorthosites and Mg-suite rocks. The Mg-suite rocks are reflected in the Mg-rich olivines observed adjacent to the Crisium basin and may be deeper crustal lithologies with a petrologic link to the Mg-suite rocks at Luna 20. Either the occurrence of mantle rocks is rare or the upper mantle is not dominated by ultramafic rocks such as dunites and orthopyroxenites.
Finally, Mg-suite magmas were derived from either the melting of a hybrid mantle source (e.g., Elardo et al., 2020) or through plagioclase assimilation by a mantle-derived, olivine-saturated basaltic magma. However, the spinel-bearing troctolites may only be produced through the assimilation of crustal plagioclase. The Mg-suite lithologies associated with the Crisium basin indicate that this style of early lunar crust building is Moon-wide and not associated only with the PKT region of the Moon.

Data Availability Statement
Data for this manuscript may be found in (a) companion manuscript Simon et al. (2022) in JGR Planets which has been published, (b) https://doi.org/10.5281/zenodo.7796375, Data tables 1, 2, and 3 which include modal mineralogy, mineral chemistries, and examples of broadbeam analyses and CIPW normative mineralogy by Conrad et al. (1973), (c) complete unpublished data by Conrad et al. (1973) available through Special Publication Number 12, UNM Institute of Meteoritics (https://ntrs.nasa.gov/api/citations/19740014355/downloads/19740014355.pdf), (d) published data presented in Luna 20 special issue of Geochimica et Cosmochimica Acta, 37(1973). data and support for Noah Petro and Dan Moriarty. We are indebted to Jim Papike for access to unpublished data and information from his initial work on these samples in the mid-1970s and for providing inspiration for this project. We are also appreciative of the University of New Mexico archiving services for providing unpublished data produced in the Institute of Meteoritics in the early 1970s (e.g., Conrad et al., 1973). The author's appreciate the review provided by K. Joy. The manuscript was significantly improved by her comments and insights.